* [PATCH 02/40] drm/i915: Rearrange i915_scheduler.c
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 03/40] drm/i915: Pass i915_sched_node around internally Chris Wilson
` (44 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
To avoid pulling in a forward declaration in the next patch, move the
i915_sched_node handling to after the main dfs of the scheduler.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/i915_scheduler.c | 210 +++++++++++++-------------
1 file changed, 105 insertions(+), 105 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c
index ec22c3fe7360..b7488c31e3e9 100644
--- a/drivers/gpu/drm/i915/i915_scheduler.c
+++ b/drivers/gpu/drm/i915/i915_scheduler.c
@@ -35,109 +35,6 @@ static inline bool node_signaled(const struct i915_sched_node *node)
return i915_request_completed(node_to_request(node));
}
-void i915_sched_node_init(struct i915_sched_node *node)
-{
- INIT_LIST_HEAD(&node->signalers_list);
- INIT_LIST_HEAD(&node->waiters_list);
- INIT_LIST_HEAD(&node->link);
- node->attr.priority = I915_PRIORITY_INVALID;
- node->semaphores = 0;
- node->flags = 0;
-}
-
-static struct i915_dependency *
-i915_dependency_alloc(void)
-{
- return kmem_cache_alloc(global.slab_dependencies, GFP_KERNEL);
-}
-
-static void
-i915_dependency_free(struct i915_dependency *dep)
-{
- kmem_cache_free(global.slab_dependencies, dep);
-}
-
-bool __i915_sched_node_add_dependency(struct i915_sched_node *node,
- struct i915_sched_node *signal,
- struct i915_dependency *dep,
- unsigned long flags)
-{
- bool ret = false;
-
- spin_lock_irq(&schedule_lock);
-
- if (!node_signaled(signal)) {
- INIT_LIST_HEAD(&dep->dfs_link);
- list_add(&dep->wait_link, &signal->waiters_list);
- list_add(&dep->signal_link, &node->signalers_list);
- dep->signaler = signal;
- dep->flags = flags;
-
- /* Keep track of whether anyone on this chain has a semaphore */
- if (signal->flags & I915_SCHED_HAS_SEMAPHORE_CHAIN &&
- !node_started(signal))
- node->flags |= I915_SCHED_HAS_SEMAPHORE_CHAIN;
-
- ret = true;
- }
-
- spin_unlock_irq(&schedule_lock);
-
- return ret;
-}
-
-int i915_sched_node_add_dependency(struct i915_sched_node *node,
- struct i915_sched_node *signal)
-{
- struct i915_dependency *dep;
-
- dep = i915_dependency_alloc();
- if (!dep)
- return -ENOMEM;
-
- if (!__i915_sched_node_add_dependency(node, signal, dep,
- I915_DEPENDENCY_ALLOC))
- i915_dependency_free(dep);
-
- return 0;
-}
-
-void i915_sched_node_fini(struct i915_sched_node *node)
-{
- struct i915_dependency *dep, *tmp;
-
- GEM_BUG_ON(!list_empty(&node->link));
-
- spin_lock_irq(&schedule_lock);
-
- /*
- * Everyone we depended upon (the fences we wait to be signaled)
- * should retire before us and remove themselves from our list.
- * However, retirement is run independently on each timeline and
- * so we may be called out-of-order.
- */
- list_for_each_entry_safe(dep, tmp, &node->signalers_list, signal_link) {
- GEM_BUG_ON(!node_signaled(dep->signaler));
- GEM_BUG_ON(!list_empty(&dep->dfs_link));
-
- list_del(&dep->wait_link);
- if (dep->flags & I915_DEPENDENCY_ALLOC)
- i915_dependency_free(dep);
- }
-
- /* Remove ourselves from everyone who depends upon us */
- list_for_each_entry_safe(dep, tmp, &node->waiters_list, wait_link) {
- GEM_BUG_ON(dep->signaler != node);
- GEM_BUG_ON(!list_empty(&dep->dfs_link));
-
- list_del(&dep->signal_link);
- if (dep->flags & I915_DEPENDENCY_ALLOC)
- i915_dependency_free(dep);
- }
-
- spin_unlock_irq(&schedule_lock);
-}
-
static inline struct i915_priolist *to_priolist(struct rb_node *rb)
{
return rb_entry(rb, struct i915_priolist, node);
@@ -239,6 +136,11 @@ i915_sched_lookup_priolist(struct intel_engine_cs *engine, int prio)
return &p->requests[idx];
}
+void __i915_priolist_free(struct i915_priolist *p)
+{
+ kmem_cache_free(global.slab_priorities, p);
+}
+
struct sched_cache {
struct list_head *priolist;
};
@@ -440,9 +342,107 @@ void i915_schedule_bump_priority(struct i915_request *rq, unsigned int bump)
spin_unlock_irqrestore(&schedule_lock, flags);
}
-void __i915_priolist_free(struct i915_priolist *p)
+void i915_sched_node_init(struct i915_sched_node *node)
{
- kmem_cache_free(global.slab_priorities, p);
+ INIT_LIST_HEAD(&node->signalers_list);
+ INIT_LIST_HEAD(&node->waiters_list);
+ INIT_LIST_HEAD(&node->link);
+ node->attr.priority = I915_PRIORITY_INVALID;
+ node->semaphores = 0;
+ node->flags = 0;
+}
+
+static struct i915_dependency *
+i915_dependency_alloc(void)
+{
+ return kmem_cache_alloc(global.slab_dependencies, GFP_KERNEL);
+}
+
+static void
+i915_dependency_free(struct i915_dependency *dep)
+{
+ kmem_cache_free(global.slab_dependencies, dep);
+}
+
+bool __i915_sched_node_add_dependency(struct i915_sched_node *node,
+ struct i915_sched_node *signal,
+ struct i915_dependency *dep,
+ unsigned long flags)
+{
+ bool ret = false;
+
+ spin_lock_irq(&schedule_lock);
+
+ if (!node_signaled(signal)) {
+ INIT_LIST_HEAD(&dep->dfs_link);
+ list_add(&dep->wait_link, &signal->waiters_list);
+ list_add(&dep->signal_link, &node->signalers_list);
+ dep->signaler = signal;
+ dep->flags = flags;
+
+ /* Keep track of whether anyone on this chain has a semaphore */
+ if (signal->flags & I915_SCHED_HAS_SEMAPHORE_CHAIN &&
+ !node_started(signal))
+ node->flags |= I915_SCHED_HAS_SEMAPHORE_CHAIN;
+
+ ret = true;
+ }
+
+ spin_unlock_irq(&schedule_lock);
+
+ return ret;
+}
+
+int i915_sched_node_add_dependency(struct i915_sched_node *node,
+ struct i915_sched_node *signal)
+{
+ struct i915_dependency *dep;
+
+ dep = i915_dependency_alloc();
+ if (!dep)
+ return -ENOMEM;
+
+ if (!__i915_sched_node_add_dependency(node, signal, dep,
+ I915_DEPENDENCY_ALLOC))
+ i915_dependency_free(dep);
+
+ return 0;
+}
+
+void i915_sched_node_fini(struct i915_sched_node *node)
+{
+ struct i915_dependency *dep, *tmp;
+
+ GEM_BUG_ON(!list_empty(&node->link));
+
+ spin_lock_irq(&schedule_lock);
+
+ /*
+ * Everyone we depended upon (the fences we wait to be signaled)
+ * should retire before us and remove themselves from our list.
+ * However, retirement is run independently on each timeline and
+ * so we may be called out-of-order.
+ */
+ list_for_each_entry_safe(dep, tmp, &node->signalers_list, signal_link) {
+ GEM_BUG_ON(!node_signaled(dep->signaler));
+ GEM_BUG_ON(!list_empty(&dep->dfs_link));
+
+ list_del(&dep->wait_link);
+ if (dep->flags & I915_DEPENDENCY_ALLOC)
+ i915_dependency_free(dep);
+ }
+
+ /* Remove ourselves from everyone who depends upon us */
+ list_for_each_entry_safe(dep, tmp, &node->waiters_list, wait_link) {
+ GEM_BUG_ON(dep->signaler != node);
+ GEM_BUG_ON(!list_empty(&dep->dfs_link));
+
+ list_del(&dep->signal_link);
+ if (dep->flags & I915_DEPENDENCY_ALLOC)
+ i915_dependency_free(dep);
+ }
+
+ spin_unlock_irq(&schedule_lock);
}
static void i915_global_scheduler_shrink(void)
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 03/40] drm/i915: Pass i915_sched_node around internally
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
2019-05-08 8:06 ` [PATCH 02/40] drm/i915: Rearrange i915_scheduler.c Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 10:15 ` Tvrtko Ursulin
2019-05-08 8:06 ` [PATCH 04/40] drm/i915: Check for no-op priority changes first Chris Wilson
` (43 subsequent siblings)
45 siblings, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
To simplify the next patch, update bump_priority and schedule to accept
the internal i915_sched_ndoe directly and not expect a request pointer.
add/remove: 0/0 grow/shrink: 2/1 up/down: 8/-15 (-7)
Function old new delta
i915_schedule_bump_priority 109 113 +4
i915_schedule 50 54 +4
__i915_schedule 922 907 -15
v2: Adopt node for the old rq local, since it no longer is a request but
the origin node.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_scheduler.c | 36 ++++++++++++++-------------
1 file changed, 19 insertions(+), 17 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c
index b7488c31e3e9..f32d0ee6d58c 100644
--- a/drivers/gpu/drm/i915/i915_scheduler.c
+++ b/drivers/gpu/drm/i915/i915_scheduler.c
@@ -186,7 +186,7 @@ static void kick_submission(struct intel_engine_cs *engine, int prio)
tasklet_hi_schedule(&engine->execlists.tasklet);
}
-static void __i915_schedule(struct i915_request *rq,
+static void __i915_schedule(struct i915_sched_node *node,
const struct i915_sched_attr *attr)
{
struct intel_engine_cs *engine;
@@ -200,13 +200,13 @@ static void __i915_schedule(struct i915_request *rq,
lockdep_assert_held(&schedule_lock);
GEM_BUG_ON(prio == I915_PRIORITY_INVALID);
- if (i915_request_completed(rq))
+ if (node_signaled(node))
return;
- if (prio <= READ_ONCE(rq->sched.attr.priority))
+ if (prio <= READ_ONCE(node->attr.priority))
return;
- stack.signaler = &rq->sched;
+ stack.signaler = node;
list_add(&stack.dfs_link, &dfs);
/*
@@ -257,9 +257,9 @@ static void __i915_schedule(struct i915_request *rq,
* execlists_submit_request()), we can set our own priority and skip
* acquiring the engine locks.
*/
- if (rq->sched.attr.priority == I915_PRIORITY_INVALID) {
- GEM_BUG_ON(!list_empty(&rq->sched.link));
- rq->sched.attr = *attr;
+ if (node->attr.priority == I915_PRIORITY_INVALID) {
+ GEM_BUG_ON(!list_empty(&node->link));
+ node->attr = *attr;
if (stack.dfs_link.next == stack.dfs_link.prev)
return;
@@ -268,15 +268,14 @@ static void __i915_schedule(struct i915_request *rq,
}
memset(&cache, 0, sizeof(cache));
- engine = rq->engine;
+ engine = node_to_request(node)->engine;
spin_lock(&engine->timeline.lock);
/* Fifo and depth-first replacement ensure our deps execute before us */
list_for_each_entry_safe_reverse(dep, p, &dfs, dfs_link) {
- struct i915_sched_node *node = dep->signaler;
-
INIT_LIST_HEAD(&dep->dfs_link);
+ node = dep->signaler;
engine = sched_lock_engine(node, engine, &cache);
lockdep_assert_held(&engine->timeline.lock);
@@ -319,13 +318,20 @@ static void __i915_schedule(struct i915_request *rq,
void i915_schedule(struct i915_request *rq, const struct i915_sched_attr *attr)
{
spin_lock_irq(&schedule_lock);
- __i915_schedule(rq, attr);
+ __i915_schedule(&rq->sched, attr);
spin_unlock_irq(&schedule_lock);
}
+static void __bump_priority(struct i915_sched_node *node, unsigned int bump)
+{
+ struct i915_sched_attr attr = node->attr;
+
+ attr.priority |= bump;
+ __i915_schedule(node, &attr);
+}
+
void i915_schedule_bump_priority(struct i915_request *rq, unsigned int bump)
{
- struct i915_sched_attr attr;
unsigned long flags;
GEM_BUG_ON(bump & ~I915_PRIORITY_MASK);
@@ -334,11 +340,7 @@ void i915_schedule_bump_priority(struct i915_request *rq, unsigned int bump)
return;
spin_lock_irqsave(&schedule_lock, flags);
-
- attr = rq->sched.attr;
- attr.priority |= bump;
- __i915_schedule(rq, &attr);
-
+ __bump_priority(&rq->sched, bump);
spin_unlock_irqrestore(&schedule_lock, flags);
}
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* Re: [PATCH 03/40] drm/i915: Pass i915_sched_node around internally
2019-05-08 8:06 ` [PATCH 03/40] drm/i915: Pass i915_sched_node around internally Chris Wilson
@ 2019-05-08 10:15 ` Tvrtko Ursulin
0 siblings, 0 replies; 71+ messages in thread
From: Tvrtko Ursulin @ 2019-05-08 10:15 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 08/05/2019 09:06, Chris Wilson wrote:
> To simplify the next patch, update bump_priority and schedule to accept
> the internal i915_sched_ndoe directly and not expect a request pointer.
>
> add/remove: 0/0 grow/shrink: 2/1 up/down: 8/-15 (-7)
> Function old new delta
> i915_schedule_bump_priority 109 113 +4
> i915_schedule 50 54 +4
> __i915_schedule 922 907 -15
>
> v2: Adopt node for the old rq local, since it no longer is a request but
> the origin node.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/i915_scheduler.c | 36 ++++++++++++++-------------
> 1 file changed, 19 insertions(+), 17 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c
> index b7488c31e3e9..f32d0ee6d58c 100644
> --- a/drivers/gpu/drm/i915/i915_scheduler.c
> +++ b/drivers/gpu/drm/i915/i915_scheduler.c
> @@ -186,7 +186,7 @@ static void kick_submission(struct intel_engine_cs *engine, int prio)
> tasklet_hi_schedule(&engine->execlists.tasklet);
> }
>
> -static void __i915_schedule(struct i915_request *rq,
> +static void __i915_schedule(struct i915_sched_node *node,
> const struct i915_sched_attr *attr)
> {
> struct intel_engine_cs *engine;
> @@ -200,13 +200,13 @@ static void __i915_schedule(struct i915_request *rq,
> lockdep_assert_held(&schedule_lock);
> GEM_BUG_ON(prio == I915_PRIORITY_INVALID);
>
> - if (i915_request_completed(rq))
> + if (node_signaled(node))
> return;
>
> - if (prio <= READ_ONCE(rq->sched.attr.priority))
> + if (prio <= READ_ONCE(node->attr.priority))
> return;
>
> - stack.signaler = &rq->sched;
> + stack.signaler = node;
> list_add(&stack.dfs_link, &dfs);
>
> /*
> @@ -257,9 +257,9 @@ static void __i915_schedule(struct i915_request *rq,
> * execlists_submit_request()), we can set our own priority and skip
> * acquiring the engine locks.
> */
> - if (rq->sched.attr.priority == I915_PRIORITY_INVALID) {
> - GEM_BUG_ON(!list_empty(&rq->sched.link));
> - rq->sched.attr = *attr;
> + if (node->attr.priority == I915_PRIORITY_INVALID) {
> + GEM_BUG_ON(!list_empty(&node->link));
> + node->attr = *attr;
>
> if (stack.dfs_link.next == stack.dfs_link.prev)
> return;
> @@ -268,15 +268,14 @@ static void __i915_schedule(struct i915_request *rq,
> }
>
> memset(&cache, 0, sizeof(cache));
> - engine = rq->engine;
> + engine = node_to_request(node)->engine;
> spin_lock(&engine->timeline.lock);
>
> /* Fifo and depth-first replacement ensure our deps execute before us */
> list_for_each_entry_safe_reverse(dep, p, &dfs, dfs_link) {
> - struct i915_sched_node *node = dep->signaler;
> -
> INIT_LIST_HEAD(&dep->dfs_link);
>
> + node = dep->signaler;
> engine = sched_lock_engine(node, engine, &cache);
> lockdep_assert_held(&engine->timeline.lock);
>
> @@ -319,13 +318,20 @@ static void __i915_schedule(struct i915_request *rq,
> void i915_schedule(struct i915_request *rq, const struct i915_sched_attr *attr)
> {
> spin_lock_irq(&schedule_lock);
> - __i915_schedule(rq, attr);
> + __i915_schedule(&rq->sched, attr);
> spin_unlock_irq(&schedule_lock);
> }
>
> +static void __bump_priority(struct i915_sched_node *node, unsigned int bump)
> +{
> + struct i915_sched_attr attr = node->attr;
> +
> + attr.priority |= bump;
> + __i915_schedule(node, &attr);
> +}
> +
> void i915_schedule_bump_priority(struct i915_request *rq, unsigned int bump)
> {
> - struct i915_sched_attr attr;
> unsigned long flags;
>
> GEM_BUG_ON(bump & ~I915_PRIORITY_MASK);
> @@ -334,11 +340,7 @@ void i915_schedule_bump_priority(struct i915_request *rq, unsigned int bump)
> return;
>
> spin_lock_irqsave(&schedule_lock, flags);
> -
> - attr = rq->sched.attr;
> - attr.priority |= bump;
> - __i915_schedule(rq, &attr);
> -
> + __bump_priority(&rq->sched, bump);
> spin_unlock_irqrestore(&schedule_lock, flags);
> }
>
>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread
* [PATCH 04/40] drm/i915: Check for no-op priority changes first
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
2019-05-08 8:06 ` [PATCH 02/40] drm/i915: Rearrange i915_scheduler.c Chris Wilson
2019-05-08 8:06 ` [PATCH 03/40] drm/i915: Pass i915_sched_node around internally Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 10:16 ` Tvrtko Ursulin
2019-05-08 8:06 ` [PATCH 05/40] drm/i915: Bump signaler priority on adding a waiter Chris Wilson
` (42 subsequent siblings)
45 siblings, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
In all likelihood, the priority and node are already in the CPU cache
and by checking them first, we can avoid having to chase the
*request->hwsp for the current breadcrumb.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/i915_scheduler.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c
index f32d0ee6d58c..5581c5004ff0 100644
--- a/drivers/gpu/drm/i915/i915_scheduler.c
+++ b/drivers/gpu/drm/i915/i915_scheduler.c
@@ -200,10 +200,10 @@ static void __i915_schedule(struct i915_sched_node *node,
lockdep_assert_held(&schedule_lock);
GEM_BUG_ON(prio == I915_PRIORITY_INVALID);
- if (node_signaled(node))
+ if (prio <= READ_ONCE(node->attr.priority))
return;
- if (prio <= READ_ONCE(node->attr.priority))
+ if (node_signaled(node))
return;
stack.signaler = node;
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* Re: [PATCH 04/40] drm/i915: Check for no-op priority changes first
2019-05-08 8:06 ` [PATCH 04/40] drm/i915: Check for no-op priority changes first Chris Wilson
@ 2019-05-08 10:16 ` Tvrtko Ursulin
0 siblings, 0 replies; 71+ messages in thread
From: Tvrtko Ursulin @ 2019-05-08 10:16 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 08/05/2019 09:06, Chris Wilson wrote:
> In all likelihood, the priority and node are already in the CPU cache
> and by checking them first, we can avoid having to chase the
> *request->hwsp for the current breadcrumb.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> ---
> drivers/gpu/drm/i915/i915_scheduler.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c
> index f32d0ee6d58c..5581c5004ff0 100644
> --- a/drivers/gpu/drm/i915/i915_scheduler.c
> +++ b/drivers/gpu/drm/i915/i915_scheduler.c
> @@ -200,10 +200,10 @@ static void __i915_schedule(struct i915_sched_node *node,
> lockdep_assert_held(&schedule_lock);
> GEM_BUG_ON(prio == I915_PRIORITY_INVALID);
>
> - if (node_signaled(node))
> + if (prio <= READ_ONCE(node->attr.priority))
> return;
>
> - if (prio <= READ_ONCE(node->attr.priority))
> + if (node_signaled(node))
> return;
>
> stack.signaler = node;
>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread
* [PATCH 05/40] drm/i915: Bump signaler priority on adding a waiter
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (2 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 04/40] drm/i915: Check for no-op priority changes first Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 06/40] drm/i915: Convert inconsistent static engine tables into an init error Chris Wilson
` (41 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
The handling of the no-preemption priority level imposes the restriction
that we need to maintain the implied ordering even though preemption is
disabled. Otherwise we may end up with an AB-BA deadlock across multiple
engine due to a real preemption event reordering the no-preemption
WAITs. To resolve this issue we currently promote all requests to WAIT
on unsubmission, however this interferes with the timeslicing
requirement that we do not apply any implicit promotion that will defeat
the round-robin timeslice list. (If we automatically promote the active
request it will go back to the head of the queue and not the tail!)
So we need implicit promotion to prevent reordering around semaphores
where we are not allowed to preempt, and we must avoid implicit
promotion on unsubmission. So instead of at unsubmit, if we apply that
implicit promotion on adding the dependency, we avoid the semaphore
deadlock and we also reduce the gains made by the promotion for user
space waiting. Furthermore, by keeping the earlier dependencies at a
higher level, we reduce the search space for timeslicing without
altering runtime scheduling too badly (no dependencies at all will be
assigned a higher priority for rrul).
v2: Limit the bump to external edges (as originally intended) i.e.
between contexts and out to the user.
Testcase: igt/gem_concurrent_blit
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/gt/selftest_lrc.c | 12 ++++++++----
drivers/gpu/drm/i915/i915_request.c | 9 ---------
drivers/gpu/drm/i915/i915_scheduler.c | 11 +++++++++++
drivers/gpu/drm/i915/i915_scheduler_types.h | 3 ++-
4 files changed, 21 insertions(+), 14 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index 4b042893dc0e..5b3d8e33f1cf 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -98,12 +98,14 @@ static int live_busywait_preempt(void *arg)
ctx_hi = kernel_context(i915);
if (!ctx_hi)
goto err_unlock;
- ctx_hi->sched.priority = INT_MAX;
+ ctx_hi->sched.priority =
+ I915_USER_PRIORITY(I915_CONTEXT_MAX_USER_PRIORITY);
ctx_lo = kernel_context(i915);
if (!ctx_lo)
goto err_ctx_hi;
- ctx_lo->sched.priority = INT_MIN;
+ ctx_lo->sched.priority =
+ I915_USER_PRIORITY(I915_CONTEXT_MIN_USER_PRIORITY);
obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
if (IS_ERR(obj)) {
@@ -958,12 +960,14 @@ static int live_preempt_hang(void *arg)
ctx_hi = kernel_context(i915);
if (!ctx_hi)
goto err_spin_lo;
- ctx_hi->sched.priority = I915_CONTEXT_MAX_USER_PRIORITY;
+ ctx_hi->sched.priority =
+ I915_USER_PRIORITY(I915_CONTEXT_MAX_USER_PRIORITY);
ctx_lo = kernel_context(i915);
if (!ctx_lo)
goto err_ctx_hi;
- ctx_lo->sched.priority = I915_CONTEXT_MIN_USER_PRIORITY;
+ ctx_lo->sched.priority =
+ I915_USER_PRIORITY(I915_CONTEXT_MIN_USER_PRIORITY);
for_each_engine(engine, i915, id) {
struct i915_request *rq;
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index fa955b7b6def..9510db566a58 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -488,15 +488,6 @@ void __i915_request_unsubmit(struct i915_request *request)
/* We may be recursing from the signal callback of another i915 fence */
spin_lock_nested(&request->lock, SINGLE_DEPTH_NESTING);
- /*
- * As we do not allow WAIT to preempt inflight requests,
- * once we have executed a request, along with triggering
- * any execution callbacks, we must preserve its ordering
- * within the non-preemptible FIFO.
- */
- BUILD_BUG_ON(__NO_PREEMPTION & ~I915_PRIORITY_MASK); /* only internal */
- request->sched.attr.priority |= __NO_PREEMPTION;
-
if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &request->fence.flags))
i915_request_cancel_breadcrumb(request);
diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c
index 5581c5004ff0..d215dcdf0b1e 100644
--- a/drivers/gpu/drm/i915/i915_scheduler.c
+++ b/drivers/gpu/drm/i915/i915_scheduler.c
@@ -387,6 +387,16 @@ bool __i915_sched_node_add_dependency(struct i915_sched_node *node,
!node_started(signal))
node->flags |= I915_SCHED_HAS_SEMAPHORE_CHAIN;
+ /*
+ * As we do not allow WAIT to preempt inflight requests,
+ * once we have executed a request, along with triggering
+ * any execution callbacks, we must preserve its ordering
+ * within the non-preemptible FIFO.
+ */
+ BUILD_BUG_ON(__NO_PREEMPTION & ~I915_PRIORITY_MASK);
+ if (flags & I915_DEPENDENCY_EXTERNAL)
+ __bump_priority(signal, __NO_PREEMPTION);
+
ret = true;
}
@@ -405,6 +415,7 @@ int i915_sched_node_add_dependency(struct i915_sched_node *node,
return -ENOMEM;
if (!__i915_sched_node_add_dependency(node, signal, dep,
+ I915_DEPENDENCY_EXTERNAL |
I915_DEPENDENCY_ALLOC))
i915_dependency_free(dep);
diff --git a/drivers/gpu/drm/i915/i915_scheduler_types.h b/drivers/gpu/drm/i915/i915_scheduler_types.h
index 166a457884b2..3e309631bd0b 100644
--- a/drivers/gpu/drm/i915/i915_scheduler_types.h
+++ b/drivers/gpu/drm/i915/i915_scheduler_types.h
@@ -66,7 +66,8 @@ struct i915_dependency {
struct list_head wait_link;
struct list_head dfs_link;
unsigned long flags;
-#define I915_DEPENDENCY_ALLOC BIT(0)
+#define I915_DEPENDENCY_ALLOC BIT(0)
+#define I915_DEPENDENCY_EXTERNAL BIT(1)
};
#endif /* _I915_SCHEDULER_TYPES_H_ */
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 06/40] drm/i915: Convert inconsistent static engine tables into an init error
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (3 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 05/40] drm/i915: Bump signaler priority on adding a waiter Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 07/40] drm/i915: Seal races between async GPU cancellation, retirement and signaling Chris Wilson
` (40 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Remove the modification of the "constant" device info by promoting the
inconsistent intel_engine static table into an initialisation error.
Now, if we add a new engine into the device_info, we must first add that
engine information into the intel_engines.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/gt/intel_engine_cs.c | 28 ++++++++---------------
1 file changed, 9 insertions(+), 19 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
index 4c3753c1b573..6434170ea4b6 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
@@ -372,15 +372,14 @@ void intel_engines_cleanup(struct drm_i915_private *i915)
*/
int intel_engines_init_mmio(struct drm_i915_private *i915)
{
- struct intel_device_info *device_info = mkwrite_device_info(i915);
- const unsigned int engine_mask = INTEL_INFO(i915)->engine_mask;
- unsigned int mask = 0;
unsigned int i;
int err;
- WARN_ON(engine_mask == 0);
- WARN_ON(engine_mask &
- GENMASK(BITS_PER_TYPE(mask) - 1, I915_NUM_ENGINES));
+ /* We always presume we have at least RCS available for later probing */
+ if (GEM_WARN_ON(!HAS_ENGINE(i915, RCS0))) {
+ err = -ENODEV;
+ goto cleanup;
+ }
if (i915_inject_load_failure())
return -ENODEV;
@@ -392,25 +391,16 @@ int intel_engines_init_mmio(struct drm_i915_private *i915)
err = intel_engine_setup(i915, i);
if (err)
goto cleanup;
-
- mask |= BIT(i);
}
- /*
- * Catch failures to update intel_engines table when the new engines
- * are added to the driver by a warning and disabling the forgotten
- * engines.
- */
- if (WARN_ON(mask != engine_mask))
- device_info->engine_mask = mask;
-
- /* We always presume we have at least RCS available for later probing */
- if (WARN_ON(!HAS_ENGINE(i915, RCS0))) {
+ /* Catch failures to update intel_engines table for new engines. */
+ if (GEM_WARN_ON(INTEL_INFO(i915)->engine_mask >> i)) {
err = -ENODEV;
goto cleanup;
}
- RUNTIME_INFO(i915)->num_engines = hweight32(mask);
+ RUNTIME_INFO(i915)->num_engines =
+ hweight32(INTEL_INFO(i915)->engine_mask);
i915_check_and_clear_faults(i915);
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 07/40] drm/i915: Seal races between async GPU cancellation, retirement and signaling
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (4 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 06/40] drm/i915: Convert inconsistent static engine tables into an init error Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 10:21 ` Tvrtko Ursulin
2019-05-08 11:24 ` [PATCH] " Chris Wilson
2019-05-08 8:06 ` [PATCH 08/40] dma-fence: Refactor signaling for manual invocation Chris Wilson
` (39 subsequent siblings)
45 siblings, 2 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Currently there is an underlying assumption that i915_request_unsubmit()
is synchronous wrt the GPU -- that is the request is no longer in flight
as we remove it. In the near future that may change, and this may upset
our signaling as we can process an interrupt for that request while it
is no longer in flight.
CPU0 CPU1
intel_engine_breadcrumbs_irq
(queue request completion)
i915_request_cancel_signaling
... ...
i915_request_enable_signaling
dma_fence_signal
Hence in the time it took us to drop the lock to signal the request, a
preemption event may have occurred and re-queued the request. In the
process, that request would have seen I915_FENCE_FLAG_SIGNAL clear and
so reused the rq->signal_link that was in use on CPU0, leading to bad
pointer chasing in intel_engine_breadcrumbs_irq.
A related issue was that if someone started listening for a signal on a
completed but no longer in-flight request, we missed the opportunity to
immediately signal that request.
Furthermore, as intel_contexts may be immediately released during
request retirement, in order to be entirely sure that
intel_engine_breadcrumbs_irq may no longer dereference the intel_context
(ce->signals and ce->signal_link), we must wait for irq spinlock.
In order to prevent the race, we use a bit in the fence.flags to signal
the transfer onto the signal list inside intel_engine_breadcrumbs_irq.
For simplicity, we use the DMA_FENCE_FLAG_SIGNALED_BIT as it then
quickly signals to any outside observer that the fence is indeed signaled.
v2: Sketch out potential dma-fence API for manual signaling
Fixes: 52c0fdb25c7c ("drm/i915: Replace global breadcrumbs with per-context interrupt tracking")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/dma-buf/dma-fence.c | 1 +
drivers/gpu/drm/i915/gt/intel_breadcrumbs.c | 70 ++++++++++++++-------
drivers/gpu/drm/i915/i915_request.c | 1 +
drivers/gpu/drm/i915/intel_guc_submission.c | 1 -
4 files changed, 51 insertions(+), 22 deletions(-)
diff --git a/drivers/dma-buf/dma-fence.c b/drivers/dma-buf/dma-fence.c
index 3aa8733f832a..9bf06042619a 100644
--- a/drivers/dma-buf/dma-fence.c
+++ b/drivers/dma-buf/dma-fence.c
@@ -29,6 +29,7 @@
EXPORT_TRACEPOINT_SYMBOL(dma_fence_emit);
EXPORT_TRACEPOINT_SYMBOL(dma_fence_enable_signal);
+EXPORT_TRACEPOINT_SYMBOL(dma_fence_signaled);
static DEFINE_SPINLOCK(dma_fence_stub_lock);
static struct dma_fence dma_fence_stub;
diff --git a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
index fe455f01aa65..7053a90e5cb5 100644
--- a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
+++ b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
@@ -23,6 +23,7 @@
*/
#include <linux/kthread.h>
+#include <trace/events/dma_fence.h>
#include <uapi/linux/sched/types.h>
#include "i915_drv.h"
@@ -96,9 +97,30 @@ check_signal_order(struct intel_context *ce, struct i915_request *rq)
return true;
}
+static void
+__dma_fence_signal_timestamp(struct dma_fence *fence, ktime_t timestamp)
+{
+ fence->timestamp = timestamp;
+ set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
+ trace_dma_fence_signaled(fence);
+}
+
+static void
+__dma_fence_signal_notify(struct dma_fence *fence)
+{
+ struct dma_fence_cb *cur, *tmp;
+
+ list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
+ INIT_LIST_HEAD(&cur->node);
+ cur->func(fence, cur);
+ }
+ INIT_LIST_HEAD(&fence->cb_list);
+}
+
void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
{
struct intel_breadcrumbs *b = &engine->breadcrumbs;
+ const ktime_t timestamp = ktime_get();
struct intel_context *ce, *cn;
struct list_head *pos, *next;
LIST_HEAD(signal);
@@ -122,6 +144,11 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
GEM_BUG_ON(!test_bit(I915_FENCE_FLAG_SIGNAL,
&rq->fence.flags));
+ clear_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags);
+
+ if (test_and_set_bit(DMA_FENCE_FLAG_SIGNALED_BIT,
+ &rq->fence.flags))
+ continue;
/*
* Queue for execution after dropping the signaling
@@ -129,14 +156,6 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
* more signalers to the same context or engine.
*/
i915_request_get(rq);
-
- /*
- * We may race with direct invocation of
- * dma_fence_signal(), e.g. i915_request_retire(),
- * so we need to acquire our reference to the request
- * before we cancel the breadcrumb.
- */
- clear_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags);
list_add_tail(&rq->signal_link, &signal);
}
@@ -159,7 +178,12 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
struct i915_request *rq =
list_entry(pos, typeof(*rq), signal_link);
- dma_fence_signal(&rq->fence);
+ __dma_fence_signal_timestamp(&rq->fence, timestamp);
+
+ spin_lock(&rq->lock);
+ __dma_fence_signal_notify(&rq->fence);
+ spin_unlock(&rq->lock);
+
i915_request_put(rq);
}
}
@@ -261,19 +285,17 @@ void intel_engine_fini_breadcrumbs(struct intel_engine_cs *engine)
bool i915_request_enable_breadcrumb(struct i915_request *rq)
{
- struct intel_breadcrumbs *b = &rq->engine->breadcrumbs;
-
- GEM_BUG_ON(test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags));
-
- if (!test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags))
- return true;
+ lockdep_assert_held(&rq->lock);
+ lockdep_assert_irqs_disabled();
- spin_lock(&b->irq_lock);
- if (test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags) &&
- !__request_completed(rq)) {
+ if (test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags)) {
+ struct intel_breadcrumbs *b = &rq->engine->breadcrumbs;
struct intel_context *ce = rq->hw_context;
struct list_head *pos;
+ spin_lock(&b->irq_lock);
+ GEM_BUG_ON(test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags));
+
__intel_breadcrumbs_arm_irq(b);
/*
@@ -303,8 +325,8 @@ bool i915_request_enable_breadcrumb(struct i915_request *rq)
GEM_BUG_ON(!check_signal_order(ce, rq));
set_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags);
+ spin_unlock(&b->irq_lock);
}
- spin_unlock(&b->irq_lock);
return !__request_completed(rq);
}
@@ -313,9 +335,15 @@ void i915_request_cancel_breadcrumb(struct i915_request *rq)
{
struct intel_breadcrumbs *b = &rq->engine->breadcrumbs;
- if (!test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags))
- return;
+ lockdep_assert_held(&rq->lock);
+ lockdep_assert_irqs_disabled();
+ /*
+ * We must wait for b->irq_lock so that we know the interrupt handler
+ * has released its reference to the intel_context and has completed
+ * the DMA_FENCE_FLAG_SIGNALED_BIT/I915_FENCE_FLAG_SIGNAL dance (if
+ * required).
+ */
spin_lock(&b->irq_lock);
if (test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags)) {
struct intel_context *ce = rq->hw_context;
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 9510db566a58..cbe4098cd6ec 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -437,6 +437,7 @@ void __i915_request_submit(struct i915_request *request)
set_bit(I915_FENCE_FLAG_ACTIVE, &request->fence.flags);
if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &request->fence.flags) &&
+ !test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &request->fence.flags) &&
!i915_request_enable_breadcrumb(request))
intel_engine_queue_breadcrumbs(engine);
diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c
index 380d83a2bfb6..ea0e3734d37c 100644
--- a/drivers/gpu/drm/i915/intel_guc_submission.c
+++ b/drivers/gpu/drm/i915/intel_guc_submission.c
@@ -23,7 +23,6 @@
*/
#include <linux/circ_buf.h>
-#include <trace/events/dma_fence.h>
#include "gt/intel_engine_pm.h"
#include "gt/intel_lrc_reg.h"
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* Re: [PATCH 07/40] drm/i915: Seal races between async GPU cancellation, retirement and signaling
2019-05-08 8:06 ` [PATCH 07/40] drm/i915: Seal races between async GPU cancellation, retirement and signaling Chris Wilson
@ 2019-05-08 10:21 ` Tvrtko Ursulin
2019-05-08 11:24 ` [PATCH] " Chris Wilson
1 sibling, 0 replies; 71+ messages in thread
From: Tvrtko Ursulin @ 2019-05-08 10:21 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 08/05/2019 09:06, Chris Wilson wrote:
> Currently there is an underlying assumption that i915_request_unsubmit()
> is synchronous wrt the GPU -- that is the request is no longer in flight
> as we remove it. In the near future that may change, and this may upset
> our signaling as we can process an interrupt for that request while it
> is no longer in flight.
>
> CPU0 CPU1
> intel_engine_breadcrumbs_irq
> (queue request completion)
> i915_request_cancel_signaling
> ... ...
> i915_request_enable_signaling
> dma_fence_signal
>
> Hence in the time it took us to drop the lock to signal the request, a
> preemption event may have occurred and re-queued the request. In the
> process, that request would have seen I915_FENCE_FLAG_SIGNAL clear and
> so reused the rq->signal_link that was in use on CPU0, leading to bad
> pointer chasing in intel_engine_breadcrumbs_irq.
>
> A related issue was that if someone started listening for a signal on a
> completed but no longer in-flight request, we missed the opportunity to
> immediately signal that request.
>
> Furthermore, as intel_contexts may be immediately released during
> request retirement, in order to be entirely sure that
> intel_engine_breadcrumbs_irq may no longer dereference the intel_context
> (ce->signals and ce->signal_link), we must wait for irq spinlock.
>
> In order to prevent the race, we use a bit in the fence.flags to signal
> the transfer onto the signal list inside intel_engine_breadcrumbs_irq.
> For simplicity, we use the DMA_FENCE_FLAG_SIGNALED_BIT as it then
> quickly signals to any outside observer that the fence is indeed signaled.
>
> v2: Sketch out potential dma-fence API for manual signaling
>
> Fixes: 52c0fdb25c7c ("drm/i915: Replace global breadcrumbs with per-context interrupt tracking")
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> ---
> drivers/dma-buf/dma-fence.c | 1 +
> drivers/gpu/drm/i915/gt/intel_breadcrumbs.c | 70 ++++++++++++++-------
> drivers/gpu/drm/i915/i915_request.c | 1 +
> drivers/gpu/drm/i915/intel_guc_submission.c | 1 -
> 4 files changed, 51 insertions(+), 22 deletions(-)
>
> diff --git a/drivers/dma-buf/dma-fence.c b/drivers/dma-buf/dma-fence.c
> index 3aa8733f832a..9bf06042619a 100644
> --- a/drivers/dma-buf/dma-fence.c
> +++ b/drivers/dma-buf/dma-fence.c
> @@ -29,6 +29,7 @@
>
> EXPORT_TRACEPOINT_SYMBOL(dma_fence_emit);
> EXPORT_TRACEPOINT_SYMBOL(dma_fence_enable_signal);
> +EXPORT_TRACEPOINT_SYMBOL(dma_fence_signaled);
>
> static DEFINE_SPINLOCK(dma_fence_stub_lock);
> static struct dma_fence dma_fence_stub;
> diff --git a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
> index fe455f01aa65..7053a90e5cb5 100644
> --- a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
> +++ b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
> @@ -23,6 +23,7 @@
> */
>
> #include <linux/kthread.h>
> +#include <trace/events/dma_fence.h>
> #include <uapi/linux/sched/types.h>
>
> #include "i915_drv.h"
> @@ -96,9 +97,30 @@ check_signal_order(struct intel_context *ce, struct i915_request *rq)
> return true;
> }
>
> +static void
> +__dma_fence_signal_timestamp(struct dma_fence *fence, ktime_t timestamp)
> +{
> + fence->timestamp = timestamp;
> + set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
> + trace_dma_fence_signaled(fence);
> +}
> +
> +static void
> +__dma_fence_signal_notify(struct dma_fence *fence)
> +{
> + struct dma_fence_cb *cur, *tmp;
> +
> + list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
> + INIT_LIST_HEAD(&cur->node);
> + cur->func(fence, cur);
> + }
> + INIT_LIST_HEAD(&fence->cb_list);
> +}
> +
> void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
> {
> struct intel_breadcrumbs *b = &engine->breadcrumbs;
> + const ktime_t timestamp = ktime_get();
> struct intel_context *ce, *cn;
> struct list_head *pos, *next;
> LIST_HEAD(signal);
> @@ -122,6 +144,11 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
>
> GEM_BUG_ON(!test_bit(I915_FENCE_FLAG_SIGNAL,
> &rq->fence.flags));
> + clear_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags);
> +
> + if (test_and_set_bit(DMA_FENCE_FLAG_SIGNALED_BIT,
> + &rq->fence.flags))
I wanted this direct/open-coded manipulation to also be clearly called
out for.
To follow the convention in other helpers here, if you could wrap it in
something like __dma_fence_signal_return I could live with it.
Regards,
Tvrtko
> + continue;
>
> /*
> * Queue for execution after dropping the signaling
> @@ -129,14 +156,6 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
> * more signalers to the same context or engine.
> */
> i915_request_get(rq);
> -
> - /*
> - * We may race with direct invocation of
> - * dma_fence_signal(), e.g. i915_request_retire(),
> - * so we need to acquire our reference to the request
> - * before we cancel the breadcrumb.
> - */
> - clear_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags);
> list_add_tail(&rq->signal_link, &signal);
> }
>
> @@ -159,7 +178,12 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
> struct i915_request *rq =
> list_entry(pos, typeof(*rq), signal_link);
>
> - dma_fence_signal(&rq->fence);
> + __dma_fence_signal_timestamp(&rq->fence, timestamp);
> +
> + spin_lock(&rq->lock);
> + __dma_fence_signal_notify(&rq->fence);
> + spin_unlock(&rq->lock);
> +
> i915_request_put(rq);
> }
> }
> @@ -261,19 +285,17 @@ void intel_engine_fini_breadcrumbs(struct intel_engine_cs *engine)
>
> bool i915_request_enable_breadcrumb(struct i915_request *rq)
> {
> - struct intel_breadcrumbs *b = &rq->engine->breadcrumbs;
> -
> - GEM_BUG_ON(test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags));
> -
> - if (!test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags))
> - return true;
> + lockdep_assert_held(&rq->lock);
> + lockdep_assert_irqs_disabled();
>
> - spin_lock(&b->irq_lock);
> - if (test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags) &&
> - !__request_completed(rq)) {
> + if (test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags)) {
> + struct intel_breadcrumbs *b = &rq->engine->breadcrumbs;
> struct intel_context *ce = rq->hw_context;
> struct list_head *pos;
>
> + spin_lock(&b->irq_lock);
> + GEM_BUG_ON(test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags));
> +
> __intel_breadcrumbs_arm_irq(b);
>
> /*
> @@ -303,8 +325,8 @@ bool i915_request_enable_breadcrumb(struct i915_request *rq)
> GEM_BUG_ON(!check_signal_order(ce, rq));
>
> set_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags);
> + spin_unlock(&b->irq_lock);
> }
> - spin_unlock(&b->irq_lock);
>
> return !__request_completed(rq);
> }
> @@ -313,9 +335,15 @@ void i915_request_cancel_breadcrumb(struct i915_request *rq)
> {
> struct intel_breadcrumbs *b = &rq->engine->breadcrumbs;
>
> - if (!test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags))
> - return;
> + lockdep_assert_held(&rq->lock);
> + lockdep_assert_irqs_disabled();
>
> + /*
> + * We must wait for b->irq_lock so that we know the interrupt handler
> + * has released its reference to the intel_context and has completed
> + * the DMA_FENCE_FLAG_SIGNALED_BIT/I915_FENCE_FLAG_SIGNAL dance (if
> + * required).
> + */
> spin_lock(&b->irq_lock);
> if (test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags)) {
> struct intel_context *ce = rq->hw_context;
> diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
> index 9510db566a58..cbe4098cd6ec 100644
> --- a/drivers/gpu/drm/i915/i915_request.c
> +++ b/drivers/gpu/drm/i915/i915_request.c
> @@ -437,6 +437,7 @@ void __i915_request_submit(struct i915_request *request)
> set_bit(I915_FENCE_FLAG_ACTIVE, &request->fence.flags);
>
> if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &request->fence.flags) &&
> + !test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &request->fence.flags) &&
> !i915_request_enable_breadcrumb(request))
> intel_engine_queue_breadcrumbs(engine);
>
> diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c
> index 380d83a2bfb6..ea0e3734d37c 100644
> --- a/drivers/gpu/drm/i915/intel_guc_submission.c
> +++ b/drivers/gpu/drm/i915/intel_guc_submission.c
> @@ -23,7 +23,6 @@
> */
>
> #include <linux/circ_buf.h>
> -#include <trace/events/dma_fence.h>
>
> #include "gt/intel_engine_pm.h"
> #include "gt/intel_lrc_reg.h"
>
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread* [PATCH] drm/i915: Seal races between async GPU cancellation, retirement and signaling
2019-05-08 8:06 ` [PATCH 07/40] drm/i915: Seal races between async GPU cancellation, retirement and signaling Chris Wilson
2019-05-08 10:21 ` Tvrtko Ursulin
@ 2019-05-08 11:24 ` Chris Wilson
2019-05-08 11:50 ` Tvrtko Ursulin
1 sibling, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 11:24 UTC (permalink / raw)
To: intel-gfx
Currently there is an underlying assumption that i915_request_unsubmit()
is synchronous wrt the GPU -- that is the request is no longer in flight
as we remove it. In the near future that may change, and this may upset
our signaling as we can process an interrupt for that request while it
is no longer in flight.
CPU0 CPU1
intel_engine_breadcrumbs_irq
(queue request completion)
i915_request_cancel_signaling
... ...
i915_request_enable_signaling
dma_fence_signal
Hence in the time it took us to drop the lock to signal the request, a
preemption event may have occurred and re-queued the request. In the
process, that request would have seen I915_FENCE_FLAG_SIGNAL clear and
so reused the rq->signal_link that was in use on CPU0, leading to bad
pointer chasing in intel_engine_breadcrumbs_irq.
A related issue was that if someone started listening for a signal on a
completed but no longer in-flight request, we missed the opportunity to
immediately signal that request.
Furthermore, as intel_contexts may be immediately released during
request retirement, in order to be entirely sure that
intel_engine_breadcrumbs_irq may no longer dereference the intel_context
(ce->signals and ce->signal_link), we must wait for irq spinlock.
In order to prevent the race, we use a bit in the fence.flags to signal
the transfer onto the signal list inside intel_engine_breadcrumbs_irq.
For simplicity, we use the DMA_FENCE_FLAG_SIGNALED_BIT as it then
quickly signals to any outside observer that the fence is indeed signaled.
v2: Sketch out potential dma-fence API for manual signaling
v3: And the test_and_set_bit()
Fixes: 52c0fdb25c7c ("drm/i915: Replace global breadcrumbs with per-context interrupt tracking")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/dma-buf/dma-fence.c | 1 +
drivers/gpu/drm/i915/gt/intel_breadcrumbs.c | 78 +++++++++++++++------
drivers/gpu/drm/i915/i915_request.c | 1 +
drivers/gpu/drm/i915/intel_guc_submission.c | 1 -
4 files changed, 59 insertions(+), 22 deletions(-)
diff --git a/drivers/dma-buf/dma-fence.c b/drivers/dma-buf/dma-fence.c
index 3aa8733f832a..9bf06042619a 100644
--- a/drivers/dma-buf/dma-fence.c
+++ b/drivers/dma-buf/dma-fence.c
@@ -29,6 +29,7 @@
EXPORT_TRACEPOINT_SYMBOL(dma_fence_emit);
EXPORT_TRACEPOINT_SYMBOL(dma_fence_enable_signal);
+EXPORT_TRACEPOINT_SYMBOL(dma_fence_signaled);
static DEFINE_SPINLOCK(dma_fence_stub_lock);
static struct dma_fence dma_fence_stub;
diff --git a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
index fe455f01aa65..c092bdf5f0bf 100644
--- a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
+++ b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
@@ -23,6 +23,7 @@
*/
#include <linux/kthread.h>
+#include <trace/events/dma_fence.h>
#include <uapi/linux/sched/types.h>
#include "i915_drv.h"
@@ -96,9 +97,39 @@ check_signal_order(struct intel_context *ce, struct i915_request *rq)
return true;
}
+static bool
+__dma_fence_signal(struct dma_fence *fence)
+{
+ return !test_and_set_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags);
+}
+
+static void
+__dma_fence_signal__timestamp(struct dma_fence *fence, ktime_t timestamp)
+{
+ fence->timestamp = timestamp;
+ set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
+ trace_dma_fence_signaled(fence);
+}
+
+static void
+__dma_fence_signal__notify(struct dma_fence *fence)
+{
+ struct dma_fence_cb *cur, *tmp;
+
+ lockdep_assert_held(fence->lock);
+ lockdep_assert_irqs_disabled();
+
+ list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
+ INIT_LIST_HEAD(&cur->node);
+ cur->func(fence, cur);
+ }
+ INIT_LIST_HEAD(&fence->cb_list);
+}
+
void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
{
struct intel_breadcrumbs *b = &engine->breadcrumbs;
+ const ktime_t timestamp = ktime_get();
struct intel_context *ce, *cn;
struct list_head *pos, *next;
LIST_HEAD(signal);
@@ -122,6 +153,10 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
GEM_BUG_ON(!test_bit(I915_FENCE_FLAG_SIGNAL,
&rq->fence.flags));
+ clear_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags);
+
+ if (!__dma_fence_signal(&rq->fence))
+ continue;
/*
* Queue for execution after dropping the signaling
@@ -129,14 +164,6 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
* more signalers to the same context or engine.
*/
i915_request_get(rq);
-
- /*
- * We may race with direct invocation of
- * dma_fence_signal(), e.g. i915_request_retire(),
- * so we need to acquire our reference to the request
- * before we cancel the breadcrumb.
- */
- clear_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags);
list_add_tail(&rq->signal_link, &signal);
}
@@ -159,7 +186,12 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
struct i915_request *rq =
list_entry(pos, typeof(*rq), signal_link);
- dma_fence_signal(&rq->fence);
+ __dma_fence_signal__timestamp(&rq->fence, timestamp);
+
+ spin_lock(&rq->lock);
+ __dma_fence_signal__notify(&rq->fence);
+ spin_unlock(&rq->lock);
+
i915_request_put(rq);
}
}
@@ -261,19 +293,17 @@ void intel_engine_fini_breadcrumbs(struct intel_engine_cs *engine)
bool i915_request_enable_breadcrumb(struct i915_request *rq)
{
- struct intel_breadcrumbs *b = &rq->engine->breadcrumbs;
-
- GEM_BUG_ON(test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags));
+ lockdep_assert_held(&rq->lock);
+ lockdep_assert_irqs_disabled();
- if (!test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags))
- return true;
-
- spin_lock(&b->irq_lock);
- if (test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags) &&
- !__request_completed(rq)) {
+ if (test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags)) {
+ struct intel_breadcrumbs *b = &rq->engine->breadcrumbs;
struct intel_context *ce = rq->hw_context;
struct list_head *pos;
+ spin_lock(&b->irq_lock);
+ GEM_BUG_ON(test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags));
+
__intel_breadcrumbs_arm_irq(b);
/*
@@ -303,8 +333,8 @@ bool i915_request_enable_breadcrumb(struct i915_request *rq)
GEM_BUG_ON(!check_signal_order(ce, rq));
set_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags);
+ spin_unlock(&b->irq_lock);
}
- spin_unlock(&b->irq_lock);
return !__request_completed(rq);
}
@@ -313,9 +343,15 @@ void i915_request_cancel_breadcrumb(struct i915_request *rq)
{
struct intel_breadcrumbs *b = &rq->engine->breadcrumbs;
- if (!test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags))
- return;
+ lockdep_assert_held(&rq->lock);
+ lockdep_assert_irqs_disabled();
+ /*
+ * We must wait for b->irq_lock so that we know the interrupt handler
+ * has released its reference to the intel_context and has completed
+ * the DMA_FENCE_FLAG_SIGNALED_BIT/I915_FENCE_FLAG_SIGNAL dance (if
+ * required).
+ */
spin_lock(&b->irq_lock);
if (test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags)) {
struct intel_context *ce = rq->hw_context;
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 9510db566a58..cbe4098cd6ec 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -437,6 +437,7 @@ void __i915_request_submit(struct i915_request *request)
set_bit(I915_FENCE_FLAG_ACTIVE, &request->fence.flags);
if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &request->fence.flags) &&
+ !test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &request->fence.flags) &&
!i915_request_enable_breadcrumb(request))
intel_engine_queue_breadcrumbs(engine);
diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c
index 380d83a2bfb6..ea0e3734d37c 100644
--- a/drivers/gpu/drm/i915/intel_guc_submission.c
+++ b/drivers/gpu/drm/i915/intel_guc_submission.c
@@ -23,7 +23,6 @@
*/
#include <linux/circ_buf.h>
-#include <trace/events/dma_fence.h>
#include "gt/intel_engine_pm.h"
#include "gt/intel_lrc_reg.h"
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* Re: [PATCH] drm/i915: Seal races between async GPU cancellation, retirement and signaling
2019-05-08 11:24 ` [PATCH] " Chris Wilson
@ 2019-05-08 11:50 ` Tvrtko Ursulin
0 siblings, 0 replies; 71+ messages in thread
From: Tvrtko Ursulin @ 2019-05-08 11:50 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 08/05/2019 12:24, Chris Wilson wrote:
> Currently there is an underlying assumption that i915_request_unsubmit()
> is synchronous wrt the GPU -- that is the request is no longer in flight
> as we remove it. In the near future that may change, and this may upset
> our signaling as we can process an interrupt for that request while it
> is no longer in flight.
>
> CPU0 CPU1
> intel_engine_breadcrumbs_irq
> (queue request completion)
> i915_request_cancel_signaling
> ... ...
> i915_request_enable_signaling
> dma_fence_signal
>
> Hence in the time it took us to drop the lock to signal the request, a
> preemption event may have occurred and re-queued the request. In the
> process, that request would have seen I915_FENCE_FLAG_SIGNAL clear and
> so reused the rq->signal_link that was in use on CPU0, leading to bad
> pointer chasing in intel_engine_breadcrumbs_irq.
>
> A related issue was that if someone started listening for a signal on a
> completed but no longer in-flight request, we missed the opportunity to
> immediately signal that request.
>
> Furthermore, as intel_contexts may be immediately released during
> request retirement, in order to be entirely sure that
> intel_engine_breadcrumbs_irq may no longer dereference the intel_context
> (ce->signals and ce->signal_link), we must wait for irq spinlock.
>
> In order to prevent the race, we use a bit in the fence.flags to signal
> the transfer onto the signal list inside intel_engine_breadcrumbs_irq.
> For simplicity, we use the DMA_FENCE_FLAG_SIGNALED_BIT as it then
> quickly signals to any outside observer that the fence is indeed signaled.
>
> v2: Sketch out potential dma-fence API for manual signaling
> v3: And the test_and_set_bit()
>
> Fixes: 52c0fdb25c7c ("drm/i915: Replace global breadcrumbs with per-context interrupt tracking")
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> ---
> drivers/dma-buf/dma-fence.c | 1 +
> drivers/gpu/drm/i915/gt/intel_breadcrumbs.c | 78 +++++++++++++++------
> drivers/gpu/drm/i915/i915_request.c | 1 +
> drivers/gpu/drm/i915/intel_guc_submission.c | 1 -
> 4 files changed, 59 insertions(+), 22 deletions(-)
>
> diff --git a/drivers/dma-buf/dma-fence.c b/drivers/dma-buf/dma-fence.c
> index 3aa8733f832a..9bf06042619a 100644
> --- a/drivers/dma-buf/dma-fence.c
> +++ b/drivers/dma-buf/dma-fence.c
> @@ -29,6 +29,7 @@
>
> EXPORT_TRACEPOINT_SYMBOL(dma_fence_emit);
> EXPORT_TRACEPOINT_SYMBOL(dma_fence_enable_signal);
> +EXPORT_TRACEPOINT_SYMBOL(dma_fence_signaled);
>
> static DEFINE_SPINLOCK(dma_fence_stub_lock);
> static struct dma_fence dma_fence_stub;
> diff --git a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
> index fe455f01aa65..c092bdf5f0bf 100644
> --- a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
> +++ b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
> @@ -23,6 +23,7 @@
> */
>
> #include <linux/kthread.h>
> +#include <trace/events/dma_fence.h>
> #include <uapi/linux/sched/types.h>
>
> #include "i915_drv.h"
> @@ -96,9 +97,39 @@ check_signal_order(struct intel_context *ce, struct i915_request *rq)
> return true;
> }
>
> +static bool
> +__dma_fence_signal(struct dma_fence *fence)
> +{
> + return !test_and_set_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags);
> +}
> +
> +static void
> +__dma_fence_signal__timestamp(struct dma_fence *fence, ktime_t timestamp)
> +{
> + fence->timestamp = timestamp;
> + set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
> + trace_dma_fence_signaled(fence);
> +}
> +
> +static void
> +__dma_fence_signal__notify(struct dma_fence *fence)
> +{
> + struct dma_fence_cb *cur, *tmp;
> +
> + lockdep_assert_held(fence->lock);
> + lockdep_assert_irqs_disabled();
> +
> + list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
> + INIT_LIST_HEAD(&cur->node);
> + cur->func(fence, cur);
> + }
> + INIT_LIST_HEAD(&fence->cb_list);
> +}
> +
> void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
> {
> struct intel_breadcrumbs *b = &engine->breadcrumbs;
> + const ktime_t timestamp = ktime_get();
> struct intel_context *ce, *cn;
> struct list_head *pos, *next;
> LIST_HEAD(signal);
> @@ -122,6 +153,10 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
>
> GEM_BUG_ON(!test_bit(I915_FENCE_FLAG_SIGNAL,
> &rq->fence.flags));
> + clear_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags);
> +
> + if (!__dma_fence_signal(&rq->fence))
> + continue;
>
> /*
> * Queue for execution after dropping the signaling
> @@ -129,14 +164,6 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
> * more signalers to the same context or engine.
> */
> i915_request_get(rq);
> -
> - /*
> - * We may race with direct invocation of
> - * dma_fence_signal(), e.g. i915_request_retire(),
> - * so we need to acquire our reference to the request
> - * before we cancel the breadcrumb.
> - */
> - clear_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags);
> list_add_tail(&rq->signal_link, &signal);
> }
>
> @@ -159,7 +186,12 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
> struct i915_request *rq =
> list_entry(pos, typeof(*rq), signal_link);
>
> - dma_fence_signal(&rq->fence);
> + __dma_fence_signal__timestamp(&rq->fence, timestamp);
> +
> + spin_lock(&rq->lock);
> + __dma_fence_signal__notify(&rq->fence);
> + spin_unlock(&rq->lock);
> +
> i915_request_put(rq);
> }
> }
> @@ -261,19 +293,17 @@ void intel_engine_fini_breadcrumbs(struct intel_engine_cs *engine)
>
> bool i915_request_enable_breadcrumb(struct i915_request *rq)
> {
> - struct intel_breadcrumbs *b = &rq->engine->breadcrumbs;
> -
> - GEM_BUG_ON(test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags));
> + lockdep_assert_held(&rq->lock);
> + lockdep_assert_irqs_disabled();
>
> - if (!test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags))
> - return true;
> -
> - spin_lock(&b->irq_lock);
> - if (test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags) &&
> - !__request_completed(rq)) {
> + if (test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags)) {
> + struct intel_breadcrumbs *b = &rq->engine->breadcrumbs;
> struct intel_context *ce = rq->hw_context;
> struct list_head *pos;
>
> + spin_lock(&b->irq_lock);
> + GEM_BUG_ON(test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags));
> +
> __intel_breadcrumbs_arm_irq(b);
>
> /*
> @@ -303,8 +333,8 @@ bool i915_request_enable_breadcrumb(struct i915_request *rq)
> GEM_BUG_ON(!check_signal_order(ce, rq));
>
> set_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags);
> + spin_unlock(&b->irq_lock);
> }
> - spin_unlock(&b->irq_lock);
>
> return !__request_completed(rq);
> }
> @@ -313,9 +343,15 @@ void i915_request_cancel_breadcrumb(struct i915_request *rq)
> {
> struct intel_breadcrumbs *b = &rq->engine->breadcrumbs;
>
> - if (!test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags))
> - return;
> + lockdep_assert_held(&rq->lock);
> + lockdep_assert_irqs_disabled();
>
> + /*
> + * We must wait for b->irq_lock so that we know the interrupt handler
> + * has released its reference to the intel_context and has completed
> + * the DMA_FENCE_FLAG_SIGNALED_BIT/I915_FENCE_FLAG_SIGNAL dance (if
> + * required).
> + */
> spin_lock(&b->irq_lock);
> if (test_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags)) {
> struct intel_context *ce = rq->hw_context;
> diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
> index 9510db566a58..cbe4098cd6ec 100644
> --- a/drivers/gpu/drm/i915/i915_request.c
> +++ b/drivers/gpu/drm/i915/i915_request.c
> @@ -437,6 +437,7 @@ void __i915_request_submit(struct i915_request *request)
> set_bit(I915_FENCE_FLAG_ACTIVE, &request->fence.flags);
>
> if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &request->fence.flags) &&
> + !test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &request->fence.flags) &&
> !i915_request_enable_breadcrumb(request))
> intel_engine_queue_breadcrumbs(engine);
>
> diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c
> index 380d83a2bfb6..ea0e3734d37c 100644
> --- a/drivers/gpu/drm/i915/intel_guc_submission.c
> +++ b/drivers/gpu/drm/i915/intel_guc_submission.c
> @@ -23,7 +23,6 @@
> */
>
> #include <linux/circ_buf.h>
> -#include <trace/events/dma_fence.h>
>
> #include "gt/intel_engine_pm.h"
> #include "gt/intel_lrc_reg.h"
>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread
* [PATCH 08/40] dma-fence: Refactor signaling for manual invocation
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (5 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 07/40] drm/i915: Seal races between async GPU cancellation, retirement and signaling Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 11:25 ` [PATCH] " Chris Wilson
2019-05-08 8:06 ` [PATCH 09/40] drm/i915: Restore control over ppgtt for context creation ABI Chris Wilson
` (38 subsequent siblings)
45 siblings, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Move the duplicated code within dma-fence.c into the header for wider
reuse.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/dma-buf/Makefile | 10 +-
drivers/dma-buf/dma-fence-trace.c | 28 +++
drivers/dma-buf/dma-fence.c | 28 +--
drivers/gpu/drm/i915/gt/intel_breadcrumbs.c | 21 --
include/linux/dma-fence-types.h | 248 ++++++++++++++++++++
include/linux/dma-fence.h | 244 ++-----------------
6 files changed, 312 insertions(+), 267 deletions(-)
create mode 100644 drivers/dma-buf/dma-fence-trace.c
create mode 100644 include/linux/dma-fence-types.h
diff --git a/drivers/dma-buf/Makefile b/drivers/dma-buf/Makefile
index 1f006e083eb9..56e579878f26 100644
--- a/drivers/dma-buf/Makefile
+++ b/drivers/dma-buf/Makefile
@@ -1,5 +1,11 @@
-obj-y := dma-buf.o dma-fence.o dma-fence-array.o dma-fence-chain.o \
- reservation.o seqno-fence.o
+obj-y := \
+ dma-buf.o \
+ dma-fence.o \
+ dma-fence-array.o \
+ dma-fence-chain.o \
+ dma-fence-trace.o \
+ reservation.o \
+ seqno-fence.o
obj-$(CONFIG_SYNC_FILE) += sync_file.o
obj-$(CONFIG_SW_SYNC) += sw_sync.o sync_debug.o
obj-$(CONFIG_UDMABUF) += udmabuf.o
diff --git a/drivers/dma-buf/dma-fence-trace.c b/drivers/dma-buf/dma-fence-trace.c
new file mode 100644
index 000000000000..eb6f282be4c0
--- /dev/null
+++ b/drivers/dma-buf/dma-fence-trace.c
@@ -0,0 +1,28 @@
+/*
+ * Fence mechanism for dma-buf and to allow for asynchronous dma access
+ *
+ * Copyright (C) 2012 Canonical Ltd
+ * Copyright (C) 2012 Texas Instruments
+ *
+ * Authors:
+ * Rob Clark <robdclark@gmail.com>
+ * Maarten Lankhorst <maarten.lankhorst@canonical.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ */
+
+#include <linux/dma-fence-types.h>
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/dma_fence.h>
+
+EXPORT_TRACEPOINT_SYMBOL(dma_fence_emit);
+EXPORT_TRACEPOINT_SYMBOL(dma_fence_enable_signal);
+EXPORT_TRACEPOINT_SYMBOL(dma_fence_signaled);
diff --git a/drivers/dma-buf/dma-fence.c b/drivers/dma-buf/dma-fence.c
index 9bf06042619a..64582c4e0b6f 100644
--- a/drivers/dma-buf/dma-fence.c
+++ b/drivers/dma-buf/dma-fence.c
@@ -24,13 +24,6 @@
#include <linux/dma-fence.h>
#include <linux/sched/signal.h>
-#define CREATE_TRACE_POINTS
-#include <trace/events/dma_fence.h>
-
-EXPORT_TRACEPOINT_SYMBOL(dma_fence_emit);
-EXPORT_TRACEPOINT_SYMBOL(dma_fence_enable_signal);
-EXPORT_TRACEPOINT_SYMBOL(dma_fence_signaled);
-
static DEFINE_SPINLOCK(dma_fence_stub_lock);
static struct dma_fence dma_fence_stub;
@@ -136,7 +129,6 @@ EXPORT_SYMBOL(dma_fence_context_alloc);
*/
int dma_fence_signal_locked(struct dma_fence *fence)
{
- struct dma_fence_cb *cur, *tmp;
int ret = 0;
lockdep_assert_held(fence->lock);
@@ -152,15 +144,10 @@ int dma_fence_signal_locked(struct dma_fence *fence)
* still run through all callbacks
*/
} else {
- fence->timestamp = ktime_get();
- set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
- trace_dma_fence_signaled(fence);
+ __dma_fence_signal_timestamp(fence, ktime_get());
}
- list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
- list_del_init(&cur->node);
- cur->func(fence, cur);
- }
+ __dma_fence_signal_notify(fence);
return ret;
}
EXPORT_SYMBOL(dma_fence_signal_locked);
@@ -188,18 +175,11 @@ int dma_fence_signal(struct dma_fence *fence)
if (test_and_set_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
return -EINVAL;
- fence->timestamp = ktime_get();
- set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
- trace_dma_fence_signaled(fence);
+ __dma_fence_signal_timestamp(fence, ktime_get());
if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &fence->flags)) {
- struct dma_fence_cb *cur, *tmp;
-
spin_lock_irqsave(fence->lock, flags);
- list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
- list_del_init(&cur->node);
- cur->func(fence, cur);
- }
+ __dma_fence_signal_notify(fence);
spin_unlock_irqrestore(fence->lock, flags);
}
return 0;
diff --git a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
index 7053a90e5cb5..7dd154f75086 100644
--- a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
+++ b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
@@ -23,7 +23,6 @@
*/
#include <linux/kthread.h>
-#include <trace/events/dma_fence.h>
#include <uapi/linux/sched/types.h>
#include "i915_drv.h"
@@ -97,26 +96,6 @@ check_signal_order(struct intel_context *ce, struct i915_request *rq)
return true;
}
-static void
-__dma_fence_signal_timestamp(struct dma_fence *fence, ktime_t timestamp)
-{
- fence->timestamp = timestamp;
- set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
- trace_dma_fence_signaled(fence);
-}
-
-static void
-__dma_fence_signal_notify(struct dma_fence *fence)
-{
- struct dma_fence_cb *cur, *tmp;
-
- list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
- INIT_LIST_HEAD(&cur->node);
- cur->func(fence, cur);
- }
- INIT_LIST_HEAD(&fence->cb_list);
-}
-
void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
{
struct intel_breadcrumbs *b = &engine->breadcrumbs;
diff --git a/include/linux/dma-fence-types.h b/include/linux/dma-fence-types.h
new file mode 100644
index 000000000000..18e7511c0eed
--- /dev/null
+++ b/include/linux/dma-fence-types.h
@@ -0,0 +1,248 @@
+/*
+ * Fence mechanism for dma-buf to allow for asynchronous dma access
+ *
+ * Copyright (C) 2012 Canonical Ltd
+ * Copyright (C) 2012 Texas Instruments
+ *
+ * Authors:
+ * Rob Clark <robdclark@gmail.com>
+ * Maarten Lankhorst <maarten.lankhorst@canonical.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ */
+
+#ifndef __LINUX_DMA_FENCE_TYPES_H
+#define __LINUX_DMA_FENCE_TYPES_H
+
+#include <linux/kref.h>
+#include <linux/ktime.h>
+
+struct dma_fence;
+struct dma_fence_ops;
+struct dma_fence_cb;
+
+/**
+ * struct dma_fence - software synchronization primitive
+ * @refcount: refcount for this fence
+ * @ops: dma_fence_ops associated with this fence
+ * @rcu: used for releasing fence with kfree_rcu
+ * @cb_list: list of all callbacks to call
+ * @lock: spin_lock_irqsave used for locking
+ * @context: execution context this fence belongs to, returned by
+ * dma_fence_context_alloc()
+ * @seqno: the sequence number of this fence inside the execution context,
+ * can be compared to decide which fence would be signaled later.
+ * @flags: A mask of DMA_FENCE_FLAG_* defined below
+ * @timestamp: Timestamp when the fence was signaled.
+ * @error: Optional, only valid if < 0, must be set before calling
+ * dma_fence_signal, indicates that the fence has completed with an error.
+ *
+ * the flags member must be manipulated and read using the appropriate
+ * atomic ops (bit_*), so taking the spinlock will not be needed most
+ * of the time.
+ *
+ * DMA_FENCE_FLAG_SIGNALED_BIT - fence is already signaled
+ * DMA_FENCE_FLAG_TIMESTAMP_BIT - timestamp recorded for fence signaling
+ * DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT - enable_signaling might have been called
+ * DMA_FENCE_FLAG_USER_BITS - start of the unused bits, can be used by the
+ * implementer of the fence for its own purposes. Can be used in different
+ * ways by different fence implementers, so do not rely on this.
+ *
+ * Since atomic bitops are used, this is not guaranteed to be the case.
+ * Particularly, if the bit was set, but dma_fence_signal was called right
+ * before this bit was set, it would have been able to set the
+ * DMA_FENCE_FLAG_SIGNALED_BIT, before enable_signaling was called.
+ * Adding a check for DMA_FENCE_FLAG_SIGNALED_BIT after setting
+ * DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT closes this race, and makes sure that
+ * after dma_fence_signal was called, any enable_signaling call will have either
+ * been completed, or never called at all.
+ */
+struct dma_fence {
+ struct kref refcount;
+ const struct dma_fence_ops *ops;
+ struct rcu_head rcu;
+ struct list_head cb_list;
+ spinlock_t *lock;
+ u64 context;
+ u64 seqno;
+ unsigned long flags;
+ ktime_t timestamp;
+ int error;
+};
+
+enum dma_fence_flag_bits {
+ DMA_FENCE_FLAG_SIGNALED_BIT,
+ DMA_FENCE_FLAG_TIMESTAMP_BIT,
+ DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT,
+ DMA_FENCE_FLAG_USER_BITS, /* must always be last member */
+};
+
+typedef void (*dma_fence_func_t)(struct dma_fence *fence,
+ struct dma_fence_cb *cb);
+
+/**
+ * struct dma_fence_cb - callback for dma_fence_add_callback()
+ * @node: used by dma_fence_add_callback() to append this struct to fence::cb_list
+ * @func: dma_fence_func_t to call
+ *
+ * This struct will be initialized by dma_fence_add_callback(), additional
+ * data can be passed along by embedding dma_fence_cb in another struct.
+ */
+struct dma_fence_cb {
+ struct list_head node;
+ dma_fence_func_t func;
+};
+
+/**
+ * struct dma_fence_ops - operations implemented for fence
+ *
+ */
+struct dma_fence_ops {
+ /**
+ * @use_64bit_seqno:
+ *
+ * True if this dma_fence implementation uses 64bit seqno, false
+ * otherwise.
+ */
+ bool use_64bit_seqno;
+
+ /**
+ * @get_driver_name:
+ *
+ * Returns the driver name. This is a callback to allow drivers to
+ * compute the name at runtime, without having it to store permanently
+ * for each fence, or build a cache of some sort.
+ *
+ * This callback is mandatory.
+ */
+ const char * (*get_driver_name)(struct dma_fence *fence);
+
+ /**
+ * @get_timeline_name:
+ *
+ * Return the name of the context this fence belongs to. This is a
+ * callback to allow drivers to compute the name at runtime, without
+ * having it to store permanently for each fence, or build a cache of
+ * some sort.
+ *
+ * This callback is mandatory.
+ */
+ const char * (*get_timeline_name)(struct dma_fence *fence);
+
+ /**
+ * @enable_signaling:
+ *
+ * Enable software signaling of fence.
+ *
+ * For fence implementations that have the capability for hw->hw
+ * signaling, they can implement this op to enable the necessary
+ * interrupts, or insert commands into cmdstream, etc, to avoid these
+ * costly operations for the common case where only hw->hw
+ * synchronization is required. This is called in the first
+ * dma_fence_wait() or dma_fence_add_callback() path to let the fence
+ * implementation know that there is another driver waiting on the
+ * signal (ie. hw->sw case).
+ *
+ * This function can be called from atomic context, but not
+ * from irq context, so normal spinlocks can be used.
+ *
+ * A return value of false indicates the fence already passed,
+ * or some failure occurred that made it impossible to enable
+ * signaling. True indicates successful enabling.
+ *
+ * &dma_fence.error may be set in enable_signaling, but only when false
+ * is returned.
+ *
+ * Since many implementations can call dma_fence_signal() even when before
+ * @enable_signaling has been called there's a race window, where the
+ * dma_fence_signal() might result in the final fence reference being
+ * released and its memory freed. To avoid this, implementations of this
+ * callback should grab their own reference using dma_fence_get(), to be
+ * released when the fence is signalled (through e.g. the interrupt
+ * handler).
+ *
+ * This callback is optional. If this callback is not present, then the
+ * driver must always have signaling enabled.
+ */
+ bool (*enable_signaling)(struct dma_fence *fence);
+
+ /**
+ * @signaled:
+ *
+ * Peek whether the fence is signaled, as a fastpath optimization for
+ * e.g. dma_fence_wait() or dma_fence_add_callback(). Note that this
+ * callback does not need to make any guarantees beyond that a fence
+ * once indicates as signalled must always return true from this
+ * callback. This callback may return false even if the fence has
+ * completed already, in this case information hasn't propogated throug
+ * the system yet. See also dma_fence_is_signaled().
+ *
+ * May set &dma_fence.error if returning true.
+ *
+ * This callback is optional.
+ */
+ bool (*signaled)(struct dma_fence *fence);
+
+ /**
+ * @wait:
+ *
+ * Custom wait implementation, defaults to dma_fence_default_wait() if
+ * not set.
+ *
+ * The dma_fence_default_wait implementation should work for any fence, as long
+ * as @enable_signaling works correctly. This hook allows drivers to
+ * have an optimized version for the case where a process context is
+ * already available, e.g. if @enable_signaling for the general case
+ * needs to set up a worker thread.
+ *
+ * Must return -ERESTARTSYS if the wait is intr = true and the wait was
+ * interrupted, and remaining jiffies if fence has signaled, or 0 if wait
+ * timed out. Can also return other error values on custom implementations,
+ * which should be treated as if the fence is signaled. For example a hardware
+ * lockup could be reported like that.
+ *
+ * This callback is optional.
+ */
+ signed long (*wait)(struct dma_fence *fence,
+ bool intr, signed long timeout);
+
+ /**
+ * @release:
+ *
+ * Called on destruction of fence to release additional resources.
+ * Can be called from irq context. This callback is optional. If it is
+ * NULL, then dma_fence_free() is instead called as the default
+ * implementation.
+ */
+ void (*release)(struct dma_fence *fence);
+
+ /**
+ * @fence_value_str:
+ *
+ * Callback to fill in free-form debug info specific to this fence, like
+ * the sequence number.
+ *
+ * This callback is optional.
+ */
+ void (*fence_value_str)(struct dma_fence *fence, char *str, int size);
+
+ /**
+ * @timeline_value_str:
+ *
+ * Fills in the current value of the timeline as a string, like the
+ * sequence number. Note that the specific fence passed to this function
+ * should not matter, drivers should only use it to look up the
+ * corresponding timeline structures.
+ */
+ void (*timeline_value_str)(struct dma_fence *fence,
+ char *str, int size);
+};
+
+#endif /* __LINUX_DMA_FENCE_TYPES_H */
diff --git a/include/linux/dma-fence.h b/include/linux/dma-fence.h
index 974717d6ac0c..770525eeb593 100644
--- a/include/linux/dma-fence.h
+++ b/include/linux/dma-fence.h
@@ -21,6 +21,7 @@
#ifndef __LINUX_DMA_FENCE_H
#define __LINUX_DMA_FENCE_H
+#include <linux/dma-fence-types.h>
#include <linux/err.h>
#include <linux/wait.h>
#include <linux/list.h>
@@ -30,226 +31,7 @@
#include <linux/printk.h>
#include <linux/rcupdate.h>
-struct dma_fence;
-struct dma_fence_ops;
-struct dma_fence_cb;
-
-/**
- * struct dma_fence - software synchronization primitive
- * @refcount: refcount for this fence
- * @ops: dma_fence_ops associated with this fence
- * @rcu: used for releasing fence with kfree_rcu
- * @cb_list: list of all callbacks to call
- * @lock: spin_lock_irqsave used for locking
- * @context: execution context this fence belongs to, returned by
- * dma_fence_context_alloc()
- * @seqno: the sequence number of this fence inside the execution context,
- * can be compared to decide which fence would be signaled later.
- * @flags: A mask of DMA_FENCE_FLAG_* defined below
- * @timestamp: Timestamp when the fence was signaled.
- * @error: Optional, only valid if < 0, must be set before calling
- * dma_fence_signal, indicates that the fence has completed with an error.
- *
- * the flags member must be manipulated and read using the appropriate
- * atomic ops (bit_*), so taking the spinlock will not be needed most
- * of the time.
- *
- * DMA_FENCE_FLAG_SIGNALED_BIT - fence is already signaled
- * DMA_FENCE_FLAG_TIMESTAMP_BIT - timestamp recorded for fence signaling
- * DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT - enable_signaling might have been called
- * DMA_FENCE_FLAG_USER_BITS - start of the unused bits, can be used by the
- * implementer of the fence for its own purposes. Can be used in different
- * ways by different fence implementers, so do not rely on this.
- *
- * Since atomic bitops are used, this is not guaranteed to be the case.
- * Particularly, if the bit was set, but dma_fence_signal was called right
- * before this bit was set, it would have been able to set the
- * DMA_FENCE_FLAG_SIGNALED_BIT, before enable_signaling was called.
- * Adding a check for DMA_FENCE_FLAG_SIGNALED_BIT after setting
- * DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT closes this race, and makes sure that
- * after dma_fence_signal was called, any enable_signaling call will have either
- * been completed, or never called at all.
- */
-struct dma_fence {
- struct kref refcount;
- const struct dma_fence_ops *ops;
- struct rcu_head rcu;
- struct list_head cb_list;
- spinlock_t *lock;
- u64 context;
- u64 seqno;
- unsigned long flags;
- ktime_t timestamp;
- int error;
-};
-
-enum dma_fence_flag_bits {
- DMA_FENCE_FLAG_SIGNALED_BIT,
- DMA_FENCE_FLAG_TIMESTAMP_BIT,
- DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT,
- DMA_FENCE_FLAG_USER_BITS, /* must always be last member */
-};
-
-typedef void (*dma_fence_func_t)(struct dma_fence *fence,
- struct dma_fence_cb *cb);
-
-/**
- * struct dma_fence_cb - callback for dma_fence_add_callback()
- * @node: used by dma_fence_add_callback() to append this struct to fence::cb_list
- * @func: dma_fence_func_t to call
- *
- * This struct will be initialized by dma_fence_add_callback(), additional
- * data can be passed along by embedding dma_fence_cb in another struct.
- */
-struct dma_fence_cb {
- struct list_head node;
- dma_fence_func_t func;
-};
-
-/**
- * struct dma_fence_ops - operations implemented for fence
- *
- */
-struct dma_fence_ops {
- /**
- * @use_64bit_seqno:
- *
- * True if this dma_fence implementation uses 64bit seqno, false
- * otherwise.
- */
- bool use_64bit_seqno;
-
- /**
- * @get_driver_name:
- *
- * Returns the driver name. This is a callback to allow drivers to
- * compute the name at runtime, without having it to store permanently
- * for each fence, or build a cache of some sort.
- *
- * This callback is mandatory.
- */
- const char * (*get_driver_name)(struct dma_fence *fence);
-
- /**
- * @get_timeline_name:
- *
- * Return the name of the context this fence belongs to. This is a
- * callback to allow drivers to compute the name at runtime, without
- * having it to store permanently for each fence, or build a cache of
- * some sort.
- *
- * This callback is mandatory.
- */
- const char * (*get_timeline_name)(struct dma_fence *fence);
-
- /**
- * @enable_signaling:
- *
- * Enable software signaling of fence.
- *
- * For fence implementations that have the capability for hw->hw
- * signaling, they can implement this op to enable the necessary
- * interrupts, or insert commands into cmdstream, etc, to avoid these
- * costly operations for the common case where only hw->hw
- * synchronization is required. This is called in the first
- * dma_fence_wait() or dma_fence_add_callback() path to let the fence
- * implementation know that there is another driver waiting on the
- * signal (ie. hw->sw case).
- *
- * This function can be called from atomic context, but not
- * from irq context, so normal spinlocks can be used.
- *
- * A return value of false indicates the fence already passed,
- * or some failure occurred that made it impossible to enable
- * signaling. True indicates successful enabling.
- *
- * &dma_fence.error may be set in enable_signaling, but only when false
- * is returned.
- *
- * Since many implementations can call dma_fence_signal() even when before
- * @enable_signaling has been called there's a race window, where the
- * dma_fence_signal() might result in the final fence reference being
- * released and its memory freed. To avoid this, implementations of this
- * callback should grab their own reference using dma_fence_get(), to be
- * released when the fence is signalled (through e.g. the interrupt
- * handler).
- *
- * This callback is optional. If this callback is not present, then the
- * driver must always have signaling enabled.
- */
- bool (*enable_signaling)(struct dma_fence *fence);
-
- /**
- * @signaled:
- *
- * Peek whether the fence is signaled, as a fastpath optimization for
- * e.g. dma_fence_wait() or dma_fence_add_callback(). Note that this
- * callback does not need to make any guarantees beyond that a fence
- * once indicates as signalled must always return true from this
- * callback. This callback may return false even if the fence has
- * completed already, in this case information hasn't propogated throug
- * the system yet. See also dma_fence_is_signaled().
- *
- * May set &dma_fence.error if returning true.
- *
- * This callback is optional.
- */
- bool (*signaled)(struct dma_fence *fence);
-
- /**
- * @wait:
- *
- * Custom wait implementation, defaults to dma_fence_default_wait() if
- * not set.
- *
- * The dma_fence_default_wait implementation should work for any fence, as long
- * as @enable_signaling works correctly. This hook allows drivers to
- * have an optimized version for the case where a process context is
- * already available, e.g. if @enable_signaling for the general case
- * needs to set up a worker thread.
- *
- * Must return -ERESTARTSYS if the wait is intr = true and the wait was
- * interrupted, and remaining jiffies if fence has signaled, or 0 if wait
- * timed out. Can also return other error values on custom implementations,
- * which should be treated as if the fence is signaled. For example a hardware
- * lockup could be reported like that.
- *
- * This callback is optional.
- */
- signed long (*wait)(struct dma_fence *fence,
- bool intr, signed long timeout);
-
- /**
- * @release:
- *
- * Called on destruction of fence to release additional resources.
- * Can be called from irq context. This callback is optional. If it is
- * NULL, then dma_fence_free() is instead called as the default
- * implementation.
- */
- void (*release)(struct dma_fence *fence);
-
- /**
- * @fence_value_str:
- *
- * Callback to fill in free-form debug info specific to this fence, like
- * the sequence number.
- *
- * This callback is optional.
- */
- void (*fence_value_str)(struct dma_fence *fence, char *str, int size);
-
- /**
- * @timeline_value_str:
- *
- * Fills in the current value of the timeline as a string, like the
- * sequence number. Note that the specific fence passed to this function
- * should not matter, drivers should only use it to look up the
- * corresponding timeline structures.
- */
- void (*timeline_value_str)(struct dma_fence *fence,
- char *str, int size);
-};
+#include <trace/events/dma_fence.h>
void dma_fence_init(struct dma_fence *fence, const struct dma_fence_ops *ops,
spinlock_t *lock, u64 context, u64 seqno);
@@ -561,6 +343,28 @@ static inline signed long dma_fence_wait(struct dma_fence *fence, bool intr)
struct dma_fence *dma_fence_get_stub(void);
u64 dma_fence_context_alloc(unsigned num);
+static inline void
+__dma_fence_signal_timestamp(struct dma_fence *fence, ktime_t timestamp)
+{
+ fence->timestamp = timestamp;
+ set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
+ trace_dma_fence_signaled(fence);
+}
+
+static inline void
+__dma_fence_signal_notify(struct dma_fence *fence)
+{
+ struct dma_fence_cb *cur, *tmp;
+
+ lockdep_assert_held(fence->lock);
+
+ list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
+ INIT_LIST_HEAD(&cur->node);
+ cur->func(fence, cur);
+ }
+ INIT_LIST_HEAD(&fence->cb_list);
+}
+
#define DMA_FENCE_TRACE(f, fmt, args...) \
do { \
struct dma_fence *__ff = (f); \
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH] dma-fence: Refactor signaling for manual invocation
2019-05-08 8:06 ` [PATCH 08/40] dma-fence: Refactor signaling for manual invocation Chris Wilson
@ 2019-05-08 11:25 ` Chris Wilson
2019-05-08 11:52 ` Tvrtko Ursulin
0 siblings, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 11:25 UTC (permalink / raw)
To: intel-gfx
Move the duplicated code within dma-fence.c into the header for wider
reuse.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/dma-buf/Makefile | 10 +-
drivers/dma-buf/dma-fence-trace.c | 28 +++
drivers/dma-buf/dma-fence.c | 32 +--
drivers/gpu/drm/i915/gt/intel_breadcrumbs.c | 30 ---
include/linux/dma-fence-types.h | 248 +++++++++++++++++++
include/linux/dma-fence.h | 251 +++-----------------
6 files changed, 321 insertions(+), 278 deletions(-)
create mode 100644 drivers/dma-buf/dma-fence-trace.c
create mode 100644 include/linux/dma-fence-types.h
diff --git a/drivers/dma-buf/Makefile b/drivers/dma-buf/Makefile
index 1f006e083eb9..56e579878f26 100644
--- a/drivers/dma-buf/Makefile
+++ b/drivers/dma-buf/Makefile
@@ -1,5 +1,11 @@
-obj-y := dma-buf.o dma-fence.o dma-fence-array.o dma-fence-chain.o \
- reservation.o seqno-fence.o
+obj-y := \
+ dma-buf.o \
+ dma-fence.o \
+ dma-fence-array.o \
+ dma-fence-chain.o \
+ dma-fence-trace.o \
+ reservation.o \
+ seqno-fence.o
obj-$(CONFIG_SYNC_FILE) += sync_file.o
obj-$(CONFIG_SW_SYNC) += sw_sync.o sync_debug.o
obj-$(CONFIG_UDMABUF) += udmabuf.o
diff --git a/drivers/dma-buf/dma-fence-trace.c b/drivers/dma-buf/dma-fence-trace.c
new file mode 100644
index 000000000000..eb6f282be4c0
--- /dev/null
+++ b/drivers/dma-buf/dma-fence-trace.c
@@ -0,0 +1,28 @@
+/*
+ * Fence mechanism for dma-buf and to allow for asynchronous dma access
+ *
+ * Copyright (C) 2012 Canonical Ltd
+ * Copyright (C) 2012 Texas Instruments
+ *
+ * Authors:
+ * Rob Clark <robdclark@gmail.com>
+ * Maarten Lankhorst <maarten.lankhorst@canonical.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ */
+
+#include <linux/dma-fence-types.h>
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/dma_fence.h>
+
+EXPORT_TRACEPOINT_SYMBOL(dma_fence_emit);
+EXPORT_TRACEPOINT_SYMBOL(dma_fence_enable_signal);
+EXPORT_TRACEPOINT_SYMBOL(dma_fence_signaled);
diff --git a/drivers/dma-buf/dma-fence.c b/drivers/dma-buf/dma-fence.c
index 9bf06042619a..8196a179fdc2 100644
--- a/drivers/dma-buf/dma-fence.c
+++ b/drivers/dma-buf/dma-fence.c
@@ -24,13 +24,6 @@
#include <linux/dma-fence.h>
#include <linux/sched/signal.h>
-#define CREATE_TRACE_POINTS
-#include <trace/events/dma_fence.h>
-
-EXPORT_TRACEPOINT_SYMBOL(dma_fence_emit);
-EXPORT_TRACEPOINT_SYMBOL(dma_fence_enable_signal);
-EXPORT_TRACEPOINT_SYMBOL(dma_fence_signaled);
-
static DEFINE_SPINLOCK(dma_fence_stub_lock);
static struct dma_fence dma_fence_stub;
@@ -136,7 +129,6 @@ EXPORT_SYMBOL(dma_fence_context_alloc);
*/
int dma_fence_signal_locked(struct dma_fence *fence)
{
- struct dma_fence_cb *cur, *tmp;
int ret = 0;
lockdep_assert_held(fence->lock);
@@ -144,7 +136,7 @@ int dma_fence_signal_locked(struct dma_fence *fence)
if (WARN_ON(!fence))
return -EINVAL;
- if (test_and_set_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) {
+ if (!__dma_fence_signal(fence)) {
ret = -EINVAL;
/*
@@ -152,15 +144,10 @@ int dma_fence_signal_locked(struct dma_fence *fence)
* still run through all callbacks
*/
} else {
- fence->timestamp = ktime_get();
- set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
- trace_dma_fence_signaled(fence);
+ __dma_fence_signal__timestamp(fence, ktime_get());
}
- list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
- list_del_init(&cur->node);
- cur->func(fence, cur);
- }
+ __dma_fence_signal__notify(fence);
return ret;
}
EXPORT_SYMBOL(dma_fence_signal_locked);
@@ -185,21 +172,14 @@ int dma_fence_signal(struct dma_fence *fence)
if (!fence)
return -EINVAL;
- if (test_and_set_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
+ if (!__dma_fence_signal(fence))
return -EINVAL;
- fence->timestamp = ktime_get();
- set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
- trace_dma_fence_signaled(fence);
+ __dma_fence_signal__timestamp(fence, ktime_get());
if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &fence->flags)) {
- struct dma_fence_cb *cur, *tmp;
-
spin_lock_irqsave(fence->lock, flags);
- list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
- list_del_init(&cur->node);
- cur->func(fence, cur);
- }
+ __dma_fence_signal__notify(fence);
spin_unlock_irqrestore(fence->lock, flags);
}
return 0;
diff --git a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
index c092bdf5f0bf..d1f8572100c3 100644
--- a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
+++ b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
@@ -23,7 +23,6 @@
*/
#include <linux/kthread.h>
-#include <trace/events/dma_fence.h>
#include <uapi/linux/sched/types.h>
#include "i915_drv.h"
@@ -97,35 +96,6 @@ check_signal_order(struct intel_context *ce, struct i915_request *rq)
return true;
}
-static bool
-__dma_fence_signal(struct dma_fence *fence)
-{
- return !test_and_set_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags);
-}
-
-static void
-__dma_fence_signal__timestamp(struct dma_fence *fence, ktime_t timestamp)
-{
- fence->timestamp = timestamp;
- set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
- trace_dma_fence_signaled(fence);
-}
-
-static void
-__dma_fence_signal__notify(struct dma_fence *fence)
-{
- struct dma_fence_cb *cur, *tmp;
-
- lockdep_assert_held(fence->lock);
- lockdep_assert_irqs_disabled();
-
- list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
- INIT_LIST_HEAD(&cur->node);
- cur->func(fence, cur);
- }
- INIT_LIST_HEAD(&fence->cb_list);
-}
-
void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
{
struct intel_breadcrumbs *b = &engine->breadcrumbs;
diff --git a/include/linux/dma-fence-types.h b/include/linux/dma-fence-types.h
new file mode 100644
index 000000000000..18e7511c0eed
--- /dev/null
+++ b/include/linux/dma-fence-types.h
@@ -0,0 +1,248 @@
+/*
+ * Fence mechanism for dma-buf to allow for asynchronous dma access
+ *
+ * Copyright (C) 2012 Canonical Ltd
+ * Copyright (C) 2012 Texas Instruments
+ *
+ * Authors:
+ * Rob Clark <robdclark@gmail.com>
+ * Maarten Lankhorst <maarten.lankhorst@canonical.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ */
+
+#ifndef __LINUX_DMA_FENCE_TYPES_H
+#define __LINUX_DMA_FENCE_TYPES_H
+
+#include <linux/kref.h>
+#include <linux/ktime.h>
+
+struct dma_fence;
+struct dma_fence_ops;
+struct dma_fence_cb;
+
+/**
+ * struct dma_fence - software synchronization primitive
+ * @refcount: refcount for this fence
+ * @ops: dma_fence_ops associated with this fence
+ * @rcu: used for releasing fence with kfree_rcu
+ * @cb_list: list of all callbacks to call
+ * @lock: spin_lock_irqsave used for locking
+ * @context: execution context this fence belongs to, returned by
+ * dma_fence_context_alloc()
+ * @seqno: the sequence number of this fence inside the execution context,
+ * can be compared to decide which fence would be signaled later.
+ * @flags: A mask of DMA_FENCE_FLAG_* defined below
+ * @timestamp: Timestamp when the fence was signaled.
+ * @error: Optional, only valid if < 0, must be set before calling
+ * dma_fence_signal, indicates that the fence has completed with an error.
+ *
+ * the flags member must be manipulated and read using the appropriate
+ * atomic ops (bit_*), so taking the spinlock will not be needed most
+ * of the time.
+ *
+ * DMA_FENCE_FLAG_SIGNALED_BIT - fence is already signaled
+ * DMA_FENCE_FLAG_TIMESTAMP_BIT - timestamp recorded for fence signaling
+ * DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT - enable_signaling might have been called
+ * DMA_FENCE_FLAG_USER_BITS - start of the unused bits, can be used by the
+ * implementer of the fence for its own purposes. Can be used in different
+ * ways by different fence implementers, so do not rely on this.
+ *
+ * Since atomic bitops are used, this is not guaranteed to be the case.
+ * Particularly, if the bit was set, but dma_fence_signal was called right
+ * before this bit was set, it would have been able to set the
+ * DMA_FENCE_FLAG_SIGNALED_BIT, before enable_signaling was called.
+ * Adding a check for DMA_FENCE_FLAG_SIGNALED_BIT after setting
+ * DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT closes this race, and makes sure that
+ * after dma_fence_signal was called, any enable_signaling call will have either
+ * been completed, or never called at all.
+ */
+struct dma_fence {
+ struct kref refcount;
+ const struct dma_fence_ops *ops;
+ struct rcu_head rcu;
+ struct list_head cb_list;
+ spinlock_t *lock;
+ u64 context;
+ u64 seqno;
+ unsigned long flags;
+ ktime_t timestamp;
+ int error;
+};
+
+enum dma_fence_flag_bits {
+ DMA_FENCE_FLAG_SIGNALED_BIT,
+ DMA_FENCE_FLAG_TIMESTAMP_BIT,
+ DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT,
+ DMA_FENCE_FLAG_USER_BITS, /* must always be last member */
+};
+
+typedef void (*dma_fence_func_t)(struct dma_fence *fence,
+ struct dma_fence_cb *cb);
+
+/**
+ * struct dma_fence_cb - callback for dma_fence_add_callback()
+ * @node: used by dma_fence_add_callback() to append this struct to fence::cb_list
+ * @func: dma_fence_func_t to call
+ *
+ * This struct will be initialized by dma_fence_add_callback(), additional
+ * data can be passed along by embedding dma_fence_cb in another struct.
+ */
+struct dma_fence_cb {
+ struct list_head node;
+ dma_fence_func_t func;
+};
+
+/**
+ * struct dma_fence_ops - operations implemented for fence
+ *
+ */
+struct dma_fence_ops {
+ /**
+ * @use_64bit_seqno:
+ *
+ * True if this dma_fence implementation uses 64bit seqno, false
+ * otherwise.
+ */
+ bool use_64bit_seqno;
+
+ /**
+ * @get_driver_name:
+ *
+ * Returns the driver name. This is a callback to allow drivers to
+ * compute the name at runtime, without having it to store permanently
+ * for each fence, or build a cache of some sort.
+ *
+ * This callback is mandatory.
+ */
+ const char * (*get_driver_name)(struct dma_fence *fence);
+
+ /**
+ * @get_timeline_name:
+ *
+ * Return the name of the context this fence belongs to. This is a
+ * callback to allow drivers to compute the name at runtime, without
+ * having it to store permanently for each fence, or build a cache of
+ * some sort.
+ *
+ * This callback is mandatory.
+ */
+ const char * (*get_timeline_name)(struct dma_fence *fence);
+
+ /**
+ * @enable_signaling:
+ *
+ * Enable software signaling of fence.
+ *
+ * For fence implementations that have the capability for hw->hw
+ * signaling, they can implement this op to enable the necessary
+ * interrupts, or insert commands into cmdstream, etc, to avoid these
+ * costly operations for the common case where only hw->hw
+ * synchronization is required. This is called in the first
+ * dma_fence_wait() or dma_fence_add_callback() path to let the fence
+ * implementation know that there is another driver waiting on the
+ * signal (ie. hw->sw case).
+ *
+ * This function can be called from atomic context, but not
+ * from irq context, so normal spinlocks can be used.
+ *
+ * A return value of false indicates the fence already passed,
+ * or some failure occurred that made it impossible to enable
+ * signaling. True indicates successful enabling.
+ *
+ * &dma_fence.error may be set in enable_signaling, but only when false
+ * is returned.
+ *
+ * Since many implementations can call dma_fence_signal() even when before
+ * @enable_signaling has been called there's a race window, where the
+ * dma_fence_signal() might result in the final fence reference being
+ * released and its memory freed. To avoid this, implementations of this
+ * callback should grab their own reference using dma_fence_get(), to be
+ * released when the fence is signalled (through e.g. the interrupt
+ * handler).
+ *
+ * This callback is optional. If this callback is not present, then the
+ * driver must always have signaling enabled.
+ */
+ bool (*enable_signaling)(struct dma_fence *fence);
+
+ /**
+ * @signaled:
+ *
+ * Peek whether the fence is signaled, as a fastpath optimization for
+ * e.g. dma_fence_wait() or dma_fence_add_callback(). Note that this
+ * callback does not need to make any guarantees beyond that a fence
+ * once indicates as signalled must always return true from this
+ * callback. This callback may return false even if the fence has
+ * completed already, in this case information hasn't propogated throug
+ * the system yet. See also dma_fence_is_signaled().
+ *
+ * May set &dma_fence.error if returning true.
+ *
+ * This callback is optional.
+ */
+ bool (*signaled)(struct dma_fence *fence);
+
+ /**
+ * @wait:
+ *
+ * Custom wait implementation, defaults to dma_fence_default_wait() if
+ * not set.
+ *
+ * The dma_fence_default_wait implementation should work for any fence, as long
+ * as @enable_signaling works correctly. This hook allows drivers to
+ * have an optimized version for the case where a process context is
+ * already available, e.g. if @enable_signaling for the general case
+ * needs to set up a worker thread.
+ *
+ * Must return -ERESTARTSYS if the wait is intr = true and the wait was
+ * interrupted, and remaining jiffies if fence has signaled, or 0 if wait
+ * timed out. Can also return other error values on custom implementations,
+ * which should be treated as if the fence is signaled. For example a hardware
+ * lockup could be reported like that.
+ *
+ * This callback is optional.
+ */
+ signed long (*wait)(struct dma_fence *fence,
+ bool intr, signed long timeout);
+
+ /**
+ * @release:
+ *
+ * Called on destruction of fence to release additional resources.
+ * Can be called from irq context. This callback is optional. If it is
+ * NULL, then dma_fence_free() is instead called as the default
+ * implementation.
+ */
+ void (*release)(struct dma_fence *fence);
+
+ /**
+ * @fence_value_str:
+ *
+ * Callback to fill in free-form debug info specific to this fence, like
+ * the sequence number.
+ *
+ * This callback is optional.
+ */
+ void (*fence_value_str)(struct dma_fence *fence, char *str, int size);
+
+ /**
+ * @timeline_value_str:
+ *
+ * Fills in the current value of the timeline as a string, like the
+ * sequence number. Note that the specific fence passed to this function
+ * should not matter, drivers should only use it to look up the
+ * corresponding timeline structures.
+ */
+ void (*timeline_value_str)(struct dma_fence *fence,
+ char *str, int size);
+};
+
+#endif /* __LINUX_DMA_FENCE_TYPES_H */
diff --git a/include/linux/dma-fence.h b/include/linux/dma-fence.h
index 974717d6ac0c..142eb67e695f 100644
--- a/include/linux/dma-fence.h
+++ b/include/linux/dma-fence.h
@@ -21,6 +21,7 @@
#ifndef __LINUX_DMA_FENCE_H
#define __LINUX_DMA_FENCE_H
+#include <linux/dma-fence-types.h>
#include <linux/err.h>
#include <linux/wait.h>
#include <linux/list.h>
@@ -30,226 +31,7 @@
#include <linux/printk.h>
#include <linux/rcupdate.h>
-struct dma_fence;
-struct dma_fence_ops;
-struct dma_fence_cb;
-
-/**
- * struct dma_fence - software synchronization primitive
- * @refcount: refcount for this fence
- * @ops: dma_fence_ops associated with this fence
- * @rcu: used for releasing fence with kfree_rcu
- * @cb_list: list of all callbacks to call
- * @lock: spin_lock_irqsave used for locking
- * @context: execution context this fence belongs to, returned by
- * dma_fence_context_alloc()
- * @seqno: the sequence number of this fence inside the execution context,
- * can be compared to decide which fence would be signaled later.
- * @flags: A mask of DMA_FENCE_FLAG_* defined below
- * @timestamp: Timestamp when the fence was signaled.
- * @error: Optional, only valid if < 0, must be set before calling
- * dma_fence_signal, indicates that the fence has completed with an error.
- *
- * the flags member must be manipulated and read using the appropriate
- * atomic ops (bit_*), so taking the spinlock will not be needed most
- * of the time.
- *
- * DMA_FENCE_FLAG_SIGNALED_BIT - fence is already signaled
- * DMA_FENCE_FLAG_TIMESTAMP_BIT - timestamp recorded for fence signaling
- * DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT - enable_signaling might have been called
- * DMA_FENCE_FLAG_USER_BITS - start of the unused bits, can be used by the
- * implementer of the fence for its own purposes. Can be used in different
- * ways by different fence implementers, so do not rely on this.
- *
- * Since atomic bitops are used, this is not guaranteed to be the case.
- * Particularly, if the bit was set, but dma_fence_signal was called right
- * before this bit was set, it would have been able to set the
- * DMA_FENCE_FLAG_SIGNALED_BIT, before enable_signaling was called.
- * Adding a check for DMA_FENCE_FLAG_SIGNALED_BIT after setting
- * DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT closes this race, and makes sure that
- * after dma_fence_signal was called, any enable_signaling call will have either
- * been completed, or never called at all.
- */
-struct dma_fence {
- struct kref refcount;
- const struct dma_fence_ops *ops;
- struct rcu_head rcu;
- struct list_head cb_list;
- spinlock_t *lock;
- u64 context;
- u64 seqno;
- unsigned long flags;
- ktime_t timestamp;
- int error;
-};
-
-enum dma_fence_flag_bits {
- DMA_FENCE_FLAG_SIGNALED_BIT,
- DMA_FENCE_FLAG_TIMESTAMP_BIT,
- DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT,
- DMA_FENCE_FLAG_USER_BITS, /* must always be last member */
-};
-
-typedef void (*dma_fence_func_t)(struct dma_fence *fence,
- struct dma_fence_cb *cb);
-
-/**
- * struct dma_fence_cb - callback for dma_fence_add_callback()
- * @node: used by dma_fence_add_callback() to append this struct to fence::cb_list
- * @func: dma_fence_func_t to call
- *
- * This struct will be initialized by dma_fence_add_callback(), additional
- * data can be passed along by embedding dma_fence_cb in another struct.
- */
-struct dma_fence_cb {
- struct list_head node;
- dma_fence_func_t func;
-};
-
-/**
- * struct dma_fence_ops - operations implemented for fence
- *
- */
-struct dma_fence_ops {
- /**
- * @use_64bit_seqno:
- *
- * True if this dma_fence implementation uses 64bit seqno, false
- * otherwise.
- */
- bool use_64bit_seqno;
-
- /**
- * @get_driver_name:
- *
- * Returns the driver name. This is a callback to allow drivers to
- * compute the name at runtime, without having it to store permanently
- * for each fence, or build a cache of some sort.
- *
- * This callback is mandatory.
- */
- const char * (*get_driver_name)(struct dma_fence *fence);
-
- /**
- * @get_timeline_name:
- *
- * Return the name of the context this fence belongs to. This is a
- * callback to allow drivers to compute the name at runtime, without
- * having it to store permanently for each fence, or build a cache of
- * some sort.
- *
- * This callback is mandatory.
- */
- const char * (*get_timeline_name)(struct dma_fence *fence);
-
- /**
- * @enable_signaling:
- *
- * Enable software signaling of fence.
- *
- * For fence implementations that have the capability for hw->hw
- * signaling, they can implement this op to enable the necessary
- * interrupts, or insert commands into cmdstream, etc, to avoid these
- * costly operations for the common case where only hw->hw
- * synchronization is required. This is called in the first
- * dma_fence_wait() or dma_fence_add_callback() path to let the fence
- * implementation know that there is another driver waiting on the
- * signal (ie. hw->sw case).
- *
- * This function can be called from atomic context, but not
- * from irq context, so normal spinlocks can be used.
- *
- * A return value of false indicates the fence already passed,
- * or some failure occurred that made it impossible to enable
- * signaling. True indicates successful enabling.
- *
- * &dma_fence.error may be set in enable_signaling, but only when false
- * is returned.
- *
- * Since many implementations can call dma_fence_signal() even when before
- * @enable_signaling has been called there's a race window, where the
- * dma_fence_signal() might result in the final fence reference being
- * released and its memory freed. To avoid this, implementations of this
- * callback should grab their own reference using dma_fence_get(), to be
- * released when the fence is signalled (through e.g. the interrupt
- * handler).
- *
- * This callback is optional. If this callback is not present, then the
- * driver must always have signaling enabled.
- */
- bool (*enable_signaling)(struct dma_fence *fence);
-
- /**
- * @signaled:
- *
- * Peek whether the fence is signaled, as a fastpath optimization for
- * e.g. dma_fence_wait() or dma_fence_add_callback(). Note that this
- * callback does not need to make any guarantees beyond that a fence
- * once indicates as signalled must always return true from this
- * callback. This callback may return false even if the fence has
- * completed already, in this case information hasn't propogated throug
- * the system yet. See also dma_fence_is_signaled().
- *
- * May set &dma_fence.error if returning true.
- *
- * This callback is optional.
- */
- bool (*signaled)(struct dma_fence *fence);
-
- /**
- * @wait:
- *
- * Custom wait implementation, defaults to dma_fence_default_wait() if
- * not set.
- *
- * The dma_fence_default_wait implementation should work for any fence, as long
- * as @enable_signaling works correctly. This hook allows drivers to
- * have an optimized version for the case where a process context is
- * already available, e.g. if @enable_signaling for the general case
- * needs to set up a worker thread.
- *
- * Must return -ERESTARTSYS if the wait is intr = true and the wait was
- * interrupted, and remaining jiffies if fence has signaled, or 0 if wait
- * timed out. Can also return other error values on custom implementations,
- * which should be treated as if the fence is signaled. For example a hardware
- * lockup could be reported like that.
- *
- * This callback is optional.
- */
- signed long (*wait)(struct dma_fence *fence,
- bool intr, signed long timeout);
-
- /**
- * @release:
- *
- * Called on destruction of fence to release additional resources.
- * Can be called from irq context. This callback is optional. If it is
- * NULL, then dma_fence_free() is instead called as the default
- * implementation.
- */
- void (*release)(struct dma_fence *fence);
-
- /**
- * @fence_value_str:
- *
- * Callback to fill in free-form debug info specific to this fence, like
- * the sequence number.
- *
- * This callback is optional.
- */
- void (*fence_value_str)(struct dma_fence *fence, char *str, int size);
-
- /**
- * @timeline_value_str:
- *
- * Fills in the current value of the timeline as a string, like the
- * sequence number. Note that the specific fence passed to this function
- * should not matter, drivers should only use it to look up the
- * corresponding timeline structures.
- */
- void (*timeline_value_str)(struct dma_fence *fence,
- char *str, int size);
-};
+#include <trace/events/dma_fence.h>
void dma_fence_init(struct dma_fence *fence, const struct dma_fence_ops *ops,
spinlock_t *lock, u64 context, u64 seqno);
@@ -561,6 +343,35 @@ static inline signed long dma_fence_wait(struct dma_fence *fence, bool intr)
struct dma_fence *dma_fence_get_stub(void);
u64 dma_fence_context_alloc(unsigned num);
+static inline bool
+__dma_fence_signal(struct dma_fence *fence)
+{
+ return !test_and_set_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags);
+}
+
+static inline void
+__dma_fence_signal__timestamp(struct dma_fence *fence, ktime_t timestamp)
+{
+ fence->timestamp = timestamp;
+ set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
+ trace_dma_fence_signaled(fence);
+}
+
+static inline void
+__dma_fence_signal__notify(struct dma_fence *fence)
+{
+ struct dma_fence_cb *cur, *tmp;
+
+ lockdep_assert_held(fence->lock);
+ lockdep_assert_irqs_disabled();
+
+ list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
+ INIT_LIST_HEAD(&cur->node);
+ cur->func(fence, cur);
+ }
+ INIT_LIST_HEAD(&fence->cb_list);
+}
+
#define DMA_FENCE_TRACE(f, fmt, args...) \
do { \
struct dma_fence *__ff = (f); \
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* Re: [PATCH] dma-fence: Refactor signaling for manual invocation
2019-05-08 11:25 ` [PATCH] " Chris Wilson
@ 2019-05-08 11:52 ` Tvrtko Ursulin
0 siblings, 0 replies; 71+ messages in thread
From: Tvrtko Ursulin @ 2019-05-08 11:52 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 08/05/2019 12:25, Chris Wilson wrote:
> Move the duplicated code within dma-fence.c into the header for wider
> reuse.
For this one I am not sure whether to go with static inlines or
EXPORT_SYMBOL the helpers.
Also you'll need to mention that in the same patch you are optimized the
whole-list-unlink. Presumably, when this goes to dri-devel one day.
But overall it makes sense to me to allow controlled fine control of the
fence signaling from dma-fence core.
Regards,
Tvrtko
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/dma-buf/Makefile | 10 +-
> drivers/dma-buf/dma-fence-trace.c | 28 +++
> drivers/dma-buf/dma-fence.c | 32 +--
> drivers/gpu/drm/i915/gt/intel_breadcrumbs.c | 30 ---
> include/linux/dma-fence-types.h | 248 +++++++++++++++++++
> include/linux/dma-fence.h | 251 +++-----------------
> 6 files changed, 321 insertions(+), 278 deletions(-)
> create mode 100644 drivers/dma-buf/dma-fence-trace.c
> create mode 100644 include/linux/dma-fence-types.h
>
> diff --git a/drivers/dma-buf/Makefile b/drivers/dma-buf/Makefile
> index 1f006e083eb9..56e579878f26 100644
> --- a/drivers/dma-buf/Makefile
> +++ b/drivers/dma-buf/Makefile
> @@ -1,5 +1,11 @@
> -obj-y := dma-buf.o dma-fence.o dma-fence-array.o dma-fence-chain.o \
> - reservation.o seqno-fence.o
> +obj-y := \
> + dma-buf.o \
> + dma-fence.o \
> + dma-fence-array.o \
> + dma-fence-chain.o \
> + dma-fence-trace.o \
> + reservation.o \
> + seqno-fence.o
> obj-$(CONFIG_SYNC_FILE) += sync_file.o
> obj-$(CONFIG_SW_SYNC) += sw_sync.o sync_debug.o
> obj-$(CONFIG_UDMABUF) += udmabuf.o
> diff --git a/drivers/dma-buf/dma-fence-trace.c b/drivers/dma-buf/dma-fence-trace.c
> new file mode 100644
> index 000000000000..eb6f282be4c0
> --- /dev/null
> +++ b/drivers/dma-buf/dma-fence-trace.c
> @@ -0,0 +1,28 @@
> +/*
> + * Fence mechanism for dma-buf and to allow for asynchronous dma access
> + *
> + * Copyright (C) 2012 Canonical Ltd
> + * Copyright (C) 2012 Texas Instruments
> + *
> + * Authors:
> + * Rob Clark <robdclark@gmail.com>
> + * Maarten Lankhorst <maarten.lankhorst@canonical.com>
> + *
> + * This program is free software; you can redistribute it and/or modify it
> + * under the terms of the GNU General Public License version 2 as published by
> + * the Free Software Foundation.
> + *
> + * This program is distributed in the hope that it will be useful, but WITHOUT
> + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
> + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
> + * more details.
> + */
> +
> +#include <linux/dma-fence-types.h>
> +
> +#define CREATE_TRACE_POINTS
> +#include <trace/events/dma_fence.h>
> +
> +EXPORT_TRACEPOINT_SYMBOL(dma_fence_emit);
> +EXPORT_TRACEPOINT_SYMBOL(dma_fence_enable_signal);
> +EXPORT_TRACEPOINT_SYMBOL(dma_fence_signaled);
> diff --git a/drivers/dma-buf/dma-fence.c b/drivers/dma-buf/dma-fence.c
> index 9bf06042619a..8196a179fdc2 100644
> --- a/drivers/dma-buf/dma-fence.c
> +++ b/drivers/dma-buf/dma-fence.c
> @@ -24,13 +24,6 @@
> #include <linux/dma-fence.h>
> #include <linux/sched/signal.h>
>
> -#define CREATE_TRACE_POINTS
> -#include <trace/events/dma_fence.h>
> -
> -EXPORT_TRACEPOINT_SYMBOL(dma_fence_emit);
> -EXPORT_TRACEPOINT_SYMBOL(dma_fence_enable_signal);
> -EXPORT_TRACEPOINT_SYMBOL(dma_fence_signaled);
> -
> static DEFINE_SPINLOCK(dma_fence_stub_lock);
> static struct dma_fence dma_fence_stub;
>
> @@ -136,7 +129,6 @@ EXPORT_SYMBOL(dma_fence_context_alloc);
> */
> int dma_fence_signal_locked(struct dma_fence *fence)
> {
> - struct dma_fence_cb *cur, *tmp;
> int ret = 0;
>
> lockdep_assert_held(fence->lock);
> @@ -144,7 +136,7 @@ int dma_fence_signal_locked(struct dma_fence *fence)
> if (WARN_ON(!fence))
> return -EINVAL;
>
> - if (test_and_set_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) {
> + if (!__dma_fence_signal(fence)) {
> ret = -EINVAL;
>
> /*
> @@ -152,15 +144,10 @@ int dma_fence_signal_locked(struct dma_fence *fence)
> * still run through all callbacks
> */
> } else {
> - fence->timestamp = ktime_get();
> - set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
> - trace_dma_fence_signaled(fence);
> + __dma_fence_signal__timestamp(fence, ktime_get());
> }
>
> - list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
> - list_del_init(&cur->node);
> - cur->func(fence, cur);
> - }
> + __dma_fence_signal__notify(fence);
> return ret;
> }
> EXPORT_SYMBOL(dma_fence_signal_locked);
> @@ -185,21 +172,14 @@ int dma_fence_signal(struct dma_fence *fence)
> if (!fence)
> return -EINVAL;
>
> - if (test_and_set_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
> + if (!__dma_fence_signal(fence))
> return -EINVAL;
>
> - fence->timestamp = ktime_get();
> - set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
> - trace_dma_fence_signaled(fence);
> + __dma_fence_signal__timestamp(fence, ktime_get());
>
> if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &fence->flags)) {
> - struct dma_fence_cb *cur, *tmp;
> -
> spin_lock_irqsave(fence->lock, flags);
> - list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
> - list_del_init(&cur->node);
> - cur->func(fence, cur);
> - }
> + __dma_fence_signal__notify(fence);
> spin_unlock_irqrestore(fence->lock, flags);
> }
> return 0;
> diff --git a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
> index c092bdf5f0bf..d1f8572100c3 100644
> --- a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
> +++ b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
> @@ -23,7 +23,6 @@
> */
>
> #include <linux/kthread.h>
> -#include <trace/events/dma_fence.h>
> #include <uapi/linux/sched/types.h>
>
> #include "i915_drv.h"
> @@ -97,35 +96,6 @@ check_signal_order(struct intel_context *ce, struct i915_request *rq)
> return true;
> }
>
> -static bool
> -__dma_fence_signal(struct dma_fence *fence)
> -{
> - return !test_and_set_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags);
> -}
> -
> -static void
> -__dma_fence_signal__timestamp(struct dma_fence *fence, ktime_t timestamp)
> -{
> - fence->timestamp = timestamp;
> - set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
> - trace_dma_fence_signaled(fence);
> -}
> -
> -static void
> -__dma_fence_signal__notify(struct dma_fence *fence)
> -{
> - struct dma_fence_cb *cur, *tmp;
> -
> - lockdep_assert_held(fence->lock);
> - lockdep_assert_irqs_disabled();
> -
> - list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
> - INIT_LIST_HEAD(&cur->node);
> - cur->func(fence, cur);
> - }
> - INIT_LIST_HEAD(&fence->cb_list);
> -}
> -
> void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine)
> {
> struct intel_breadcrumbs *b = &engine->breadcrumbs;
> diff --git a/include/linux/dma-fence-types.h b/include/linux/dma-fence-types.h
> new file mode 100644
> index 000000000000..18e7511c0eed
> --- /dev/null
> +++ b/include/linux/dma-fence-types.h
> @@ -0,0 +1,248 @@
> +/*
> + * Fence mechanism for dma-buf to allow for asynchronous dma access
> + *
> + * Copyright (C) 2012 Canonical Ltd
> + * Copyright (C) 2012 Texas Instruments
> + *
> + * Authors:
> + * Rob Clark <robdclark@gmail.com>
> + * Maarten Lankhorst <maarten.lankhorst@canonical.com>
> + *
> + * This program is free software; you can redistribute it and/or modify it
> + * under the terms of the GNU General Public License version 2 as published by
> + * the Free Software Foundation.
> + *
> + * This program is distributed in the hope that it will be useful, but WITHOUT
> + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
> + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
> + * more details.
> + */
> +
> +#ifndef __LINUX_DMA_FENCE_TYPES_H
> +#define __LINUX_DMA_FENCE_TYPES_H
> +
> +#include <linux/kref.h>
> +#include <linux/ktime.h>
> +
> +struct dma_fence;
> +struct dma_fence_ops;
> +struct dma_fence_cb;
> +
> +/**
> + * struct dma_fence - software synchronization primitive
> + * @refcount: refcount for this fence
> + * @ops: dma_fence_ops associated with this fence
> + * @rcu: used for releasing fence with kfree_rcu
> + * @cb_list: list of all callbacks to call
> + * @lock: spin_lock_irqsave used for locking
> + * @context: execution context this fence belongs to, returned by
> + * dma_fence_context_alloc()
> + * @seqno: the sequence number of this fence inside the execution context,
> + * can be compared to decide which fence would be signaled later.
> + * @flags: A mask of DMA_FENCE_FLAG_* defined below
> + * @timestamp: Timestamp when the fence was signaled.
> + * @error: Optional, only valid if < 0, must be set before calling
> + * dma_fence_signal, indicates that the fence has completed with an error.
> + *
> + * the flags member must be manipulated and read using the appropriate
> + * atomic ops (bit_*), so taking the spinlock will not be needed most
> + * of the time.
> + *
> + * DMA_FENCE_FLAG_SIGNALED_BIT - fence is already signaled
> + * DMA_FENCE_FLAG_TIMESTAMP_BIT - timestamp recorded for fence signaling
> + * DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT - enable_signaling might have been called
> + * DMA_FENCE_FLAG_USER_BITS - start of the unused bits, can be used by the
> + * implementer of the fence for its own purposes. Can be used in different
> + * ways by different fence implementers, so do not rely on this.
> + *
> + * Since atomic bitops are used, this is not guaranteed to be the case.
> + * Particularly, if the bit was set, but dma_fence_signal was called right
> + * before this bit was set, it would have been able to set the
> + * DMA_FENCE_FLAG_SIGNALED_BIT, before enable_signaling was called.
> + * Adding a check for DMA_FENCE_FLAG_SIGNALED_BIT after setting
> + * DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT closes this race, and makes sure that
> + * after dma_fence_signal was called, any enable_signaling call will have either
> + * been completed, or never called at all.
> + */
> +struct dma_fence {
> + struct kref refcount;
> + const struct dma_fence_ops *ops;
> + struct rcu_head rcu;
> + struct list_head cb_list;
> + spinlock_t *lock;
> + u64 context;
> + u64 seqno;
> + unsigned long flags;
> + ktime_t timestamp;
> + int error;
> +};
> +
> +enum dma_fence_flag_bits {
> + DMA_FENCE_FLAG_SIGNALED_BIT,
> + DMA_FENCE_FLAG_TIMESTAMP_BIT,
> + DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT,
> + DMA_FENCE_FLAG_USER_BITS, /* must always be last member */
> +};
> +
> +typedef void (*dma_fence_func_t)(struct dma_fence *fence,
> + struct dma_fence_cb *cb);
> +
> +/**
> + * struct dma_fence_cb - callback for dma_fence_add_callback()
> + * @node: used by dma_fence_add_callback() to append this struct to fence::cb_list
> + * @func: dma_fence_func_t to call
> + *
> + * This struct will be initialized by dma_fence_add_callback(), additional
> + * data can be passed along by embedding dma_fence_cb in another struct.
> + */
> +struct dma_fence_cb {
> + struct list_head node;
> + dma_fence_func_t func;
> +};
> +
> +/**
> + * struct dma_fence_ops - operations implemented for fence
> + *
> + */
> +struct dma_fence_ops {
> + /**
> + * @use_64bit_seqno:
> + *
> + * True if this dma_fence implementation uses 64bit seqno, false
> + * otherwise.
> + */
> + bool use_64bit_seqno;
> +
> + /**
> + * @get_driver_name:
> + *
> + * Returns the driver name. This is a callback to allow drivers to
> + * compute the name at runtime, without having it to store permanently
> + * for each fence, or build a cache of some sort.
> + *
> + * This callback is mandatory.
> + */
> + const char * (*get_driver_name)(struct dma_fence *fence);
> +
> + /**
> + * @get_timeline_name:
> + *
> + * Return the name of the context this fence belongs to. This is a
> + * callback to allow drivers to compute the name at runtime, without
> + * having it to store permanently for each fence, or build a cache of
> + * some sort.
> + *
> + * This callback is mandatory.
> + */
> + const char * (*get_timeline_name)(struct dma_fence *fence);
> +
> + /**
> + * @enable_signaling:
> + *
> + * Enable software signaling of fence.
> + *
> + * For fence implementations that have the capability for hw->hw
> + * signaling, they can implement this op to enable the necessary
> + * interrupts, or insert commands into cmdstream, etc, to avoid these
> + * costly operations for the common case where only hw->hw
> + * synchronization is required. This is called in the first
> + * dma_fence_wait() or dma_fence_add_callback() path to let the fence
> + * implementation know that there is another driver waiting on the
> + * signal (ie. hw->sw case).
> + *
> + * This function can be called from atomic context, but not
> + * from irq context, so normal spinlocks can be used.
> + *
> + * A return value of false indicates the fence already passed,
> + * or some failure occurred that made it impossible to enable
> + * signaling. True indicates successful enabling.
> + *
> + * &dma_fence.error may be set in enable_signaling, but only when false
> + * is returned.
> + *
> + * Since many implementations can call dma_fence_signal() even when before
> + * @enable_signaling has been called there's a race window, where the
> + * dma_fence_signal() might result in the final fence reference being
> + * released and its memory freed. To avoid this, implementations of this
> + * callback should grab their own reference using dma_fence_get(), to be
> + * released when the fence is signalled (through e.g. the interrupt
> + * handler).
> + *
> + * This callback is optional. If this callback is not present, then the
> + * driver must always have signaling enabled.
> + */
> + bool (*enable_signaling)(struct dma_fence *fence);
> +
> + /**
> + * @signaled:
> + *
> + * Peek whether the fence is signaled, as a fastpath optimization for
> + * e.g. dma_fence_wait() or dma_fence_add_callback(). Note that this
> + * callback does not need to make any guarantees beyond that a fence
> + * once indicates as signalled must always return true from this
> + * callback. This callback may return false even if the fence has
> + * completed already, in this case information hasn't propogated throug
> + * the system yet. See also dma_fence_is_signaled().
> + *
> + * May set &dma_fence.error if returning true.
> + *
> + * This callback is optional.
> + */
> + bool (*signaled)(struct dma_fence *fence);
> +
> + /**
> + * @wait:
> + *
> + * Custom wait implementation, defaults to dma_fence_default_wait() if
> + * not set.
> + *
> + * The dma_fence_default_wait implementation should work for any fence, as long
> + * as @enable_signaling works correctly. This hook allows drivers to
> + * have an optimized version for the case where a process context is
> + * already available, e.g. if @enable_signaling for the general case
> + * needs to set up a worker thread.
> + *
> + * Must return -ERESTARTSYS if the wait is intr = true and the wait was
> + * interrupted, and remaining jiffies if fence has signaled, or 0 if wait
> + * timed out. Can also return other error values on custom implementations,
> + * which should be treated as if the fence is signaled. For example a hardware
> + * lockup could be reported like that.
> + *
> + * This callback is optional.
> + */
> + signed long (*wait)(struct dma_fence *fence,
> + bool intr, signed long timeout);
> +
> + /**
> + * @release:
> + *
> + * Called on destruction of fence to release additional resources.
> + * Can be called from irq context. This callback is optional. If it is
> + * NULL, then dma_fence_free() is instead called as the default
> + * implementation.
> + */
> + void (*release)(struct dma_fence *fence);
> +
> + /**
> + * @fence_value_str:
> + *
> + * Callback to fill in free-form debug info specific to this fence, like
> + * the sequence number.
> + *
> + * This callback is optional.
> + */
> + void (*fence_value_str)(struct dma_fence *fence, char *str, int size);
> +
> + /**
> + * @timeline_value_str:
> + *
> + * Fills in the current value of the timeline as a string, like the
> + * sequence number. Note that the specific fence passed to this function
> + * should not matter, drivers should only use it to look up the
> + * corresponding timeline structures.
> + */
> + void (*timeline_value_str)(struct dma_fence *fence,
> + char *str, int size);
> +};
> +
> +#endif /* __LINUX_DMA_FENCE_TYPES_H */
> diff --git a/include/linux/dma-fence.h b/include/linux/dma-fence.h
> index 974717d6ac0c..142eb67e695f 100644
> --- a/include/linux/dma-fence.h
> +++ b/include/linux/dma-fence.h
> @@ -21,6 +21,7 @@
> #ifndef __LINUX_DMA_FENCE_H
> #define __LINUX_DMA_FENCE_H
>
> +#include <linux/dma-fence-types.h>
> #include <linux/err.h>
> #include <linux/wait.h>
> #include <linux/list.h>
> @@ -30,226 +31,7 @@
> #include <linux/printk.h>
> #include <linux/rcupdate.h>
>
> -struct dma_fence;
> -struct dma_fence_ops;
> -struct dma_fence_cb;
> -
> -/**
> - * struct dma_fence - software synchronization primitive
> - * @refcount: refcount for this fence
> - * @ops: dma_fence_ops associated with this fence
> - * @rcu: used for releasing fence with kfree_rcu
> - * @cb_list: list of all callbacks to call
> - * @lock: spin_lock_irqsave used for locking
> - * @context: execution context this fence belongs to, returned by
> - * dma_fence_context_alloc()
> - * @seqno: the sequence number of this fence inside the execution context,
> - * can be compared to decide which fence would be signaled later.
> - * @flags: A mask of DMA_FENCE_FLAG_* defined below
> - * @timestamp: Timestamp when the fence was signaled.
> - * @error: Optional, only valid if < 0, must be set before calling
> - * dma_fence_signal, indicates that the fence has completed with an error.
> - *
> - * the flags member must be manipulated and read using the appropriate
> - * atomic ops (bit_*), so taking the spinlock will not be needed most
> - * of the time.
> - *
> - * DMA_FENCE_FLAG_SIGNALED_BIT - fence is already signaled
> - * DMA_FENCE_FLAG_TIMESTAMP_BIT - timestamp recorded for fence signaling
> - * DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT - enable_signaling might have been called
> - * DMA_FENCE_FLAG_USER_BITS - start of the unused bits, can be used by the
> - * implementer of the fence for its own purposes. Can be used in different
> - * ways by different fence implementers, so do not rely on this.
> - *
> - * Since atomic bitops are used, this is not guaranteed to be the case.
> - * Particularly, if the bit was set, but dma_fence_signal was called right
> - * before this bit was set, it would have been able to set the
> - * DMA_FENCE_FLAG_SIGNALED_BIT, before enable_signaling was called.
> - * Adding a check for DMA_FENCE_FLAG_SIGNALED_BIT after setting
> - * DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT closes this race, and makes sure that
> - * after dma_fence_signal was called, any enable_signaling call will have either
> - * been completed, or never called at all.
> - */
> -struct dma_fence {
> - struct kref refcount;
> - const struct dma_fence_ops *ops;
> - struct rcu_head rcu;
> - struct list_head cb_list;
> - spinlock_t *lock;
> - u64 context;
> - u64 seqno;
> - unsigned long flags;
> - ktime_t timestamp;
> - int error;
> -};
> -
> -enum dma_fence_flag_bits {
> - DMA_FENCE_FLAG_SIGNALED_BIT,
> - DMA_FENCE_FLAG_TIMESTAMP_BIT,
> - DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT,
> - DMA_FENCE_FLAG_USER_BITS, /* must always be last member */
> -};
> -
> -typedef void (*dma_fence_func_t)(struct dma_fence *fence,
> - struct dma_fence_cb *cb);
> -
> -/**
> - * struct dma_fence_cb - callback for dma_fence_add_callback()
> - * @node: used by dma_fence_add_callback() to append this struct to fence::cb_list
> - * @func: dma_fence_func_t to call
> - *
> - * This struct will be initialized by dma_fence_add_callback(), additional
> - * data can be passed along by embedding dma_fence_cb in another struct.
> - */
> -struct dma_fence_cb {
> - struct list_head node;
> - dma_fence_func_t func;
> -};
> -
> -/**
> - * struct dma_fence_ops - operations implemented for fence
> - *
> - */
> -struct dma_fence_ops {
> - /**
> - * @use_64bit_seqno:
> - *
> - * True if this dma_fence implementation uses 64bit seqno, false
> - * otherwise.
> - */
> - bool use_64bit_seqno;
> -
> - /**
> - * @get_driver_name:
> - *
> - * Returns the driver name. This is a callback to allow drivers to
> - * compute the name at runtime, without having it to store permanently
> - * for each fence, or build a cache of some sort.
> - *
> - * This callback is mandatory.
> - */
> - const char * (*get_driver_name)(struct dma_fence *fence);
> -
> - /**
> - * @get_timeline_name:
> - *
> - * Return the name of the context this fence belongs to. This is a
> - * callback to allow drivers to compute the name at runtime, without
> - * having it to store permanently for each fence, or build a cache of
> - * some sort.
> - *
> - * This callback is mandatory.
> - */
> - const char * (*get_timeline_name)(struct dma_fence *fence);
> -
> - /**
> - * @enable_signaling:
> - *
> - * Enable software signaling of fence.
> - *
> - * For fence implementations that have the capability for hw->hw
> - * signaling, they can implement this op to enable the necessary
> - * interrupts, or insert commands into cmdstream, etc, to avoid these
> - * costly operations for the common case where only hw->hw
> - * synchronization is required. This is called in the first
> - * dma_fence_wait() or dma_fence_add_callback() path to let the fence
> - * implementation know that there is another driver waiting on the
> - * signal (ie. hw->sw case).
> - *
> - * This function can be called from atomic context, but not
> - * from irq context, so normal spinlocks can be used.
> - *
> - * A return value of false indicates the fence already passed,
> - * or some failure occurred that made it impossible to enable
> - * signaling. True indicates successful enabling.
> - *
> - * &dma_fence.error may be set in enable_signaling, but only when false
> - * is returned.
> - *
> - * Since many implementations can call dma_fence_signal() even when before
> - * @enable_signaling has been called there's a race window, where the
> - * dma_fence_signal() might result in the final fence reference being
> - * released and its memory freed. To avoid this, implementations of this
> - * callback should grab their own reference using dma_fence_get(), to be
> - * released when the fence is signalled (through e.g. the interrupt
> - * handler).
> - *
> - * This callback is optional. If this callback is not present, then the
> - * driver must always have signaling enabled.
> - */
> - bool (*enable_signaling)(struct dma_fence *fence);
> -
> - /**
> - * @signaled:
> - *
> - * Peek whether the fence is signaled, as a fastpath optimization for
> - * e.g. dma_fence_wait() or dma_fence_add_callback(). Note that this
> - * callback does not need to make any guarantees beyond that a fence
> - * once indicates as signalled must always return true from this
> - * callback. This callback may return false even if the fence has
> - * completed already, in this case information hasn't propogated throug
> - * the system yet. See also dma_fence_is_signaled().
> - *
> - * May set &dma_fence.error if returning true.
> - *
> - * This callback is optional.
> - */
> - bool (*signaled)(struct dma_fence *fence);
> -
> - /**
> - * @wait:
> - *
> - * Custom wait implementation, defaults to dma_fence_default_wait() if
> - * not set.
> - *
> - * The dma_fence_default_wait implementation should work for any fence, as long
> - * as @enable_signaling works correctly. This hook allows drivers to
> - * have an optimized version for the case where a process context is
> - * already available, e.g. if @enable_signaling for the general case
> - * needs to set up a worker thread.
> - *
> - * Must return -ERESTARTSYS if the wait is intr = true and the wait was
> - * interrupted, and remaining jiffies if fence has signaled, or 0 if wait
> - * timed out. Can also return other error values on custom implementations,
> - * which should be treated as if the fence is signaled. For example a hardware
> - * lockup could be reported like that.
> - *
> - * This callback is optional.
> - */
> - signed long (*wait)(struct dma_fence *fence,
> - bool intr, signed long timeout);
> -
> - /**
> - * @release:
> - *
> - * Called on destruction of fence to release additional resources.
> - * Can be called from irq context. This callback is optional. If it is
> - * NULL, then dma_fence_free() is instead called as the default
> - * implementation.
> - */
> - void (*release)(struct dma_fence *fence);
> -
> - /**
> - * @fence_value_str:
> - *
> - * Callback to fill in free-form debug info specific to this fence, like
> - * the sequence number.
> - *
> - * This callback is optional.
> - */
> - void (*fence_value_str)(struct dma_fence *fence, char *str, int size);
> -
> - /**
> - * @timeline_value_str:
> - *
> - * Fills in the current value of the timeline as a string, like the
> - * sequence number. Note that the specific fence passed to this function
> - * should not matter, drivers should only use it to look up the
> - * corresponding timeline structures.
> - */
> - void (*timeline_value_str)(struct dma_fence *fence,
> - char *str, int size);
> -};
> +#include <trace/events/dma_fence.h>
>
> void dma_fence_init(struct dma_fence *fence, const struct dma_fence_ops *ops,
> spinlock_t *lock, u64 context, u64 seqno);
> @@ -561,6 +343,35 @@ static inline signed long dma_fence_wait(struct dma_fence *fence, bool intr)
> struct dma_fence *dma_fence_get_stub(void);
> u64 dma_fence_context_alloc(unsigned num);
>
> +static inline bool
> +__dma_fence_signal(struct dma_fence *fence)
> +{
> + return !test_and_set_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags);
> +}
> +
> +static inline void
> +__dma_fence_signal__timestamp(struct dma_fence *fence, ktime_t timestamp)
> +{
> + fence->timestamp = timestamp;
> + set_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags);
> + trace_dma_fence_signaled(fence);
> +}
> +
> +static inline void
> +__dma_fence_signal__notify(struct dma_fence *fence)
> +{
> + struct dma_fence_cb *cur, *tmp;
> +
> + lockdep_assert_held(fence->lock);
> + lockdep_assert_irqs_disabled();
> +
> + list_for_each_entry_safe(cur, tmp, &fence->cb_list, node) {
> + INIT_LIST_HEAD(&cur->node);
> + cur->func(fence, cur);
> + }
> + INIT_LIST_HEAD(&fence->cb_list);
> +}
> +
> #define DMA_FENCE_TRACE(f, fmt, args...) \
> do { \
> struct dma_fence *__ff = (f); \
>
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread
* [PATCH 09/40] drm/i915: Restore control over ppgtt for context creation ABI
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (6 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 08/40] dma-fence: Refactor signaling for manual invocation Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 10:24 ` Tvrtko Ursulin
2019-05-08 8:06 ` [PATCH 10/40] drm/i915: Allow a context to define its set of engines Chris Wilson
` (37 subsequent siblings)
45 siblings, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Having hid the partially exposed new ABI from the PR, put it back again
for completion of context recovery. A significant part of context
recovery is the ability to reuse as much of the old context as is
feasible (to avoid expensive reconstruction). The biggest chunk kept
hidden at the moment is fine-control over the ctx->ppgtt (the GPU page
tables and associated translation tables and kernel maps), so make
control over the ctx->ppgtt explicit.
This allows userspace to create and share virtual memory address spaces
(within the limits of a single fd) between contexts they own, along with
the ability to query the contexts for the vm state.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_drv.c | 2 ++
drivers/gpu/drm/i915/i915_gem_context.c | 5 -----
include/uapi/drm/i915_drm.h | 15 +++++++++++++++
3 files changed, 17 insertions(+), 5 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 2c7a4318d13c..5061cb32856b 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -3164,6 +3164,8 @@ static const struct drm_ioctl_desc i915_ioctls[] = {
DRM_IOCTL_DEF_DRV(I915_PERF_ADD_CONFIG, i915_perf_add_config_ioctl, DRM_UNLOCKED|DRM_RENDER_ALLOW),
DRM_IOCTL_DEF_DRV(I915_PERF_REMOVE_CONFIG, i915_perf_remove_config_ioctl, DRM_UNLOCKED|DRM_RENDER_ALLOW),
DRM_IOCTL_DEF_DRV(I915_QUERY, i915_query_ioctl, DRM_UNLOCKED|DRM_RENDER_ALLOW),
+ DRM_IOCTL_DEF_DRV(I915_GEM_VM_CREATE, i915_gem_vm_create_ioctl, DRM_RENDER_ALLOW),
+ DRM_IOCTL_DEF_DRV(I915_GEM_VM_DESTROY, i915_gem_vm_destroy_ioctl, DRM_RENDER_ALLOW),
};
static struct drm_driver driver = {
diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
index 65cefc520e79..413c4529191d 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/i915_gem_context.c
@@ -98,7 +98,6 @@
#include "i915_user_extensions.h"
#define I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE (1 << 1)
-#define I915_CONTEXT_PARAM_VM 0x9
#define ALL_L3_SLICES(dev) (1 << NUM_L3_SLICES(dev)) - 1
@@ -966,8 +965,6 @@ static int get_ppgtt(struct drm_i915_file_private *file_priv,
struct i915_hw_ppgtt *ppgtt;
int ret;
- return -EINVAL; /* nothing to see here; please move along */
-
if (!ctx->ppgtt)
return -ENODEV;
@@ -1066,8 +1063,6 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv,
struct i915_hw_ppgtt *ppgtt, *old;
int err;
- return -EINVAL; /* nothing to see here; please move along */
-
if (args->size)
return -EINVAL;
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index 3a73f5316766..d6ad4a15b2b9 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -355,6 +355,8 @@ typedef struct _drm_i915_sarea {
#define DRM_I915_PERF_ADD_CONFIG 0x37
#define DRM_I915_PERF_REMOVE_CONFIG 0x38
#define DRM_I915_QUERY 0x39
+#define DRM_I915_GEM_VM_CREATE 0x3a
+#define DRM_I915_GEM_VM_DESTROY 0x3b
/* Must be kept compact -- no holes */
#define DRM_IOCTL_I915_INIT DRM_IOW( DRM_COMMAND_BASE + DRM_I915_INIT, drm_i915_init_t)
@@ -415,6 +417,8 @@ typedef struct _drm_i915_sarea {
#define DRM_IOCTL_I915_PERF_ADD_CONFIG DRM_IOW(DRM_COMMAND_BASE + DRM_I915_PERF_ADD_CONFIG, struct drm_i915_perf_oa_config)
#define DRM_IOCTL_I915_PERF_REMOVE_CONFIG DRM_IOW(DRM_COMMAND_BASE + DRM_I915_PERF_REMOVE_CONFIG, __u64)
#define DRM_IOCTL_I915_QUERY DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_QUERY, struct drm_i915_query)
+#define DRM_IOCTL_I915_GEM_VM_CREATE DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_VM_CREATE, struct drm_i915_gem_vm_control)
+#define DRM_IOCTL_I915_GEM_VM_DESTROY DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_VM_DESTROY, struct drm_i915_gem_vm_control)
/* Allow drivers to submit batchbuffers directly to hardware, relying
* on the security mechanisms provided by hardware.
@@ -1507,6 +1511,17 @@ struct drm_i915_gem_context_param {
* On creation, all new contexts are marked as recoverable.
*/
#define I915_CONTEXT_PARAM_RECOVERABLE 0x8
+
+ /*
+ * The id of the associated virtual memory address space (ppGTT) of
+ * this context. Can be retrieved and passed to another context
+ * (on the same fd) for both to use the same ppGTT and so share
+ * address layouts, and avoid reloading the page tables on context
+ * switches between themselves.
+ *
+ * See DRM_I915_GEM_VM_CREATE and DRM_I915_GEM_VM_DESTROY.
+ */
+#define I915_CONTEXT_PARAM_VM 0x9
/* Must be kept compact -- no holes and well documented */
__u64 value;
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* Re: [PATCH 09/40] drm/i915: Restore control over ppgtt for context creation ABI
2019-05-08 8:06 ` [PATCH 09/40] drm/i915: Restore control over ppgtt for context creation ABI Chris Wilson
@ 2019-05-08 10:24 ` Tvrtko Ursulin
0 siblings, 0 replies; 71+ messages in thread
From: Tvrtko Ursulin @ 2019-05-08 10:24 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 08/05/2019 09:06, Chris Wilson wrote:
> Having hid the partially exposed new ABI from the PR, put it back again
> for completion of context recovery. A significant part of context
> recovery is the ability to reuse as much of the old context as is
> feasible (to avoid expensive reconstruction). The biggest chunk kept
> hidden at the moment is fine-control over the ctx->ppgtt (the GPU page
> tables and associated translation tables and kernel maps), so make
> control over the ctx->ppgtt explicit.
>
> This allows userspace to create and share virtual memory address spaces
> (within the limits of a single fd) between contexts they own, along with
> the ability to query the contexts for the vm state.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/i915_drv.c | 2 ++
> drivers/gpu/drm/i915/i915_gem_context.c | 5 -----
> include/uapi/drm/i915_drm.h | 15 +++++++++++++++
> 3 files changed, 17 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
> index 2c7a4318d13c..5061cb32856b 100644
> --- a/drivers/gpu/drm/i915/i915_drv.c
> +++ b/drivers/gpu/drm/i915/i915_drv.c
> @@ -3164,6 +3164,8 @@ static const struct drm_ioctl_desc i915_ioctls[] = {
> DRM_IOCTL_DEF_DRV(I915_PERF_ADD_CONFIG, i915_perf_add_config_ioctl, DRM_UNLOCKED|DRM_RENDER_ALLOW),
> DRM_IOCTL_DEF_DRV(I915_PERF_REMOVE_CONFIG, i915_perf_remove_config_ioctl, DRM_UNLOCKED|DRM_RENDER_ALLOW),
> DRM_IOCTL_DEF_DRV(I915_QUERY, i915_query_ioctl, DRM_UNLOCKED|DRM_RENDER_ALLOW),
> + DRM_IOCTL_DEF_DRV(I915_GEM_VM_CREATE, i915_gem_vm_create_ioctl, DRM_RENDER_ALLOW),
> + DRM_IOCTL_DEF_DRV(I915_GEM_VM_DESTROY, i915_gem_vm_destroy_ioctl, DRM_RENDER_ALLOW),
> };
>
> static struct drm_driver driver = {
> diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
> index 65cefc520e79..413c4529191d 100644
> --- a/drivers/gpu/drm/i915/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/i915_gem_context.c
> @@ -98,7 +98,6 @@
> #include "i915_user_extensions.h"
>
> #define I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE (1 << 1)
> -#define I915_CONTEXT_PARAM_VM 0x9
>
> #define ALL_L3_SLICES(dev) (1 << NUM_L3_SLICES(dev)) - 1
>
> @@ -966,8 +965,6 @@ static int get_ppgtt(struct drm_i915_file_private *file_priv,
> struct i915_hw_ppgtt *ppgtt;
> int ret;
>
> - return -EINVAL; /* nothing to see here; please move along */
> -
> if (!ctx->ppgtt)
> return -ENODEV;
>
> @@ -1066,8 +1063,6 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv,
> struct i915_hw_ppgtt *ppgtt, *old;
> int err;
>
> - return -EINVAL; /* nothing to see here; please move along */
> -
> if (args->size)
> return -EINVAL;
>
> diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
> index 3a73f5316766..d6ad4a15b2b9 100644
> --- a/include/uapi/drm/i915_drm.h
> +++ b/include/uapi/drm/i915_drm.h
> @@ -355,6 +355,8 @@ typedef struct _drm_i915_sarea {
> #define DRM_I915_PERF_ADD_CONFIG 0x37
> #define DRM_I915_PERF_REMOVE_CONFIG 0x38
> #define DRM_I915_QUERY 0x39
> +#define DRM_I915_GEM_VM_CREATE 0x3a
> +#define DRM_I915_GEM_VM_DESTROY 0x3b
> /* Must be kept compact -- no holes */
>
> #define DRM_IOCTL_I915_INIT DRM_IOW( DRM_COMMAND_BASE + DRM_I915_INIT, drm_i915_init_t)
> @@ -415,6 +417,8 @@ typedef struct _drm_i915_sarea {
> #define DRM_IOCTL_I915_PERF_ADD_CONFIG DRM_IOW(DRM_COMMAND_BASE + DRM_I915_PERF_ADD_CONFIG, struct drm_i915_perf_oa_config)
> #define DRM_IOCTL_I915_PERF_REMOVE_CONFIG DRM_IOW(DRM_COMMAND_BASE + DRM_I915_PERF_REMOVE_CONFIG, __u64)
> #define DRM_IOCTL_I915_QUERY DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_QUERY, struct drm_i915_query)
> +#define DRM_IOCTL_I915_GEM_VM_CREATE DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_VM_CREATE, struct drm_i915_gem_vm_control)
> +#define DRM_IOCTL_I915_GEM_VM_DESTROY DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_VM_DESTROY, struct drm_i915_gem_vm_control)
>
> /* Allow drivers to submit batchbuffers directly to hardware, relying
> * on the security mechanisms provided by hardware.
> @@ -1507,6 +1511,17 @@ struct drm_i915_gem_context_param {
> * On creation, all new contexts are marked as recoverable.
> */
> #define I915_CONTEXT_PARAM_RECOVERABLE 0x8
> +
> + /*
> + * The id of the associated virtual memory address space (ppGTT) of
> + * this context. Can be retrieved and passed to another context
> + * (on the same fd) for both to use the same ppGTT and so share
> + * address layouts, and avoid reloading the page tables on context
> + * switches between themselves.
> + *
> + * See DRM_I915_GEM_VM_CREATE and DRM_I915_GEM_VM_DESTROY.
> + */
> +#define I915_CONTEXT_PARAM_VM 0x9
> /* Must be kept compact -- no holes and well documented */
>
> __u64 value;
>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread
* [PATCH 10/40] drm/i915: Allow a context to define its set of engines
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (7 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 09/40] drm/i915: Restore control over ppgtt for context creation ABI Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 11/40] drm/i915: Extend I915_CONTEXT_PARAM_SSEU to support local ctx->engine[] Chris Wilson
` (36 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Over the last few years, we have debated how to extend the user API to
support an increase in the number of engines, that may be sparse and
even be heterogeneous within a class (not all video decoders created
equal). We settled on using (class, instance) tuples to identify a
specific engine, with an API for the user to construct a map of engines
to capabilities. Into this picture, we then add a challenge of virtual
engines; one user engine that maps behind the scenes to any number of
physical engines. To keep it general, we want the user to have full
control over that mapping. To that end, we allow the user to constrain a
context to define the set of engines that it can access, order fully
controlled by the user via (class, instance). With such precise control
in context setup, we can continue to use the existing execbuf uABI of
specifying a single index; only now it doesn't automagically map onto
the engines, it uses the user defined engine map from the context.
v2: Fixup freeing of local on success of get_engines()
v3: Allow empty engines[]
v4: s/nengine/num_engines/
v5: Replace 64 limit on num_engines with a note that execbuf is
currently limited to only using the first 64 engines.
v6: Actually use the engines_mutex to guard the ctx->engines.
Testcase: igt/gem_ctx_engines
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/i915_gem_context.c | 239 +++++++++++++++++-
drivers/gpu/drm/i915/i915_gem_context.h | 18 ++
drivers/gpu/drm/i915/i915_gem_context_types.h | 1 +
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 5 +-
drivers/gpu/drm/i915/i915_utils.h | 34 +++
include/uapi/drm/i915_drm.h | 31 +++
6 files changed, 315 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
index 413c4529191d..21bfcd529097 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/i915_gem_context.c
@@ -90,7 +90,6 @@
#include <drm/i915_drm.h>
#include "gt/intel_lrc_reg.h"
-#include "gt/intel_workarounds.h"
#include "i915_drv.h"
#include "i915_globals.h"
@@ -141,15 +140,31 @@ static void lut_close(struct i915_gem_context *ctx)
}
static struct intel_context *
-lookup_user_engine(struct i915_gem_context *ctx, u16 class, u16 instance)
+lookup_user_engine(struct i915_gem_context *ctx,
+ unsigned long flags,
+ const struct i915_engine_class_instance *ci)
+#define LOOKUP_USER_INDEX BIT(0)
{
- struct intel_engine_cs *engine;
+ int idx;
- engine = intel_engine_lookup_user(ctx->i915, class, instance);
- if (!engine)
+ if (!!(flags & LOOKUP_USER_INDEX) != i915_gem_context_user_engines(ctx))
return ERR_PTR(-EINVAL);
- return i915_gem_context_get_engine(ctx, engine->id);
+ if (!i915_gem_context_user_engines(ctx)) {
+ struct intel_engine_cs *engine;
+
+ engine = intel_engine_lookup_user(ctx->i915,
+ ci->engine_class,
+ ci->engine_instance);
+ if (!engine)
+ return ERR_PTR(-EINVAL);
+
+ idx = engine->id;
+ } else {
+ idx = ci->engine_instance;
+ }
+
+ return i915_gem_context_get_engine(ctx, idx);
}
static inline int new_hw_id(struct drm_i915_private *i915, gfp_t gfp)
@@ -257,6 +272,17 @@ static void free_engines(struct i915_gem_engines *e)
__free_engines(e, e->num_engines);
}
+static void free_engines_rcu(struct work_struct *wrk)
+{
+ struct i915_gem_engines *e =
+ container_of(wrk, struct i915_gem_engines, rcu.work);
+ struct drm_i915_private *i915 = e->i915;
+
+ mutex_lock(&i915->drm.struct_mutex);
+ free_engines(e);
+ mutex_unlock(&i915->drm.struct_mutex);
+}
+
static struct i915_gem_engines *default_engines(struct i915_gem_context *ctx)
{
struct intel_engine_cs *engine;
@@ -1352,9 +1378,7 @@ static int set_sseu(struct i915_gem_context *ctx,
if (user_sseu.flags || user_sseu.rsvd)
return -EINVAL;
- ce = lookup_user_engine(ctx,
- user_sseu.engine.engine_class,
- user_sseu.engine.engine_instance);
+ ce = lookup_user_engine(ctx, 0, &user_sseu.engine);
if (IS_ERR(ce))
return PTR_ERR(ce);
@@ -1379,6 +1403,191 @@ static int set_sseu(struct i915_gem_context *ctx,
return ret;
}
+struct set_engines {
+ struct i915_gem_context *ctx;
+ struct i915_gem_engines *engines;
+};
+
+static const i915_user_extension_fn set_engines__extensions[] = {
+};
+
+static int
+set_engines(struct i915_gem_context *ctx,
+ const struct drm_i915_gem_context_param *args)
+{
+ struct i915_context_param_engines __user *user =
+ u64_to_user_ptr(args->value);
+ struct set_engines set = { .ctx = ctx };
+ unsigned int num_engines, n;
+ u64 extensions;
+ int err;
+
+ if (!args->size) { /* switch back to legacy user_ring_map */
+ if (!i915_gem_context_user_engines(ctx))
+ return 0;
+
+ set.engines = default_engines(ctx);
+ if (IS_ERR(set.engines))
+ return PTR_ERR(set.engines);
+
+ goto replace;
+ }
+
+ BUILD_BUG_ON(!IS_ALIGNED(sizeof(*user), sizeof(*user->engines)));
+ if (args->size < sizeof(*user) ||
+ !IS_ALIGNED(args->size, sizeof(*user->engines))) {
+ DRM_DEBUG("Invalid size for engine array: %d\n",
+ args->size);
+ return -EINVAL;
+ }
+
+ /*
+ * Note that I915_EXEC_RING_MASK limits execbuf to only using the
+ * first 64 engines defined here.
+ */
+ num_engines = (args->size - sizeof(*user)) / sizeof(*user->engines);
+
+ set.engines = kmalloc(struct_size(set.engines, engines, num_engines),
+ GFP_KERNEL);
+ if (!set.engines)
+ return -ENOMEM;
+
+ set.engines->i915 = ctx->i915;
+ for (n = 0; n < num_engines; n++) {
+ struct i915_engine_class_instance ci;
+ struct intel_engine_cs *engine;
+
+ if (copy_from_user(&ci, &user->engines[n], sizeof(ci))) {
+ __free_engines(set.engines, n);
+ return -EFAULT;
+ }
+
+ if (ci.engine_class == (u16)I915_ENGINE_CLASS_INVALID &&
+ ci.engine_instance == (u16)I915_ENGINE_CLASS_INVALID_NONE) {
+ set.engines->engines[n] = NULL;
+ continue;
+ }
+
+ engine = intel_engine_lookup_user(ctx->i915,
+ ci.engine_class,
+ ci.engine_instance);
+ if (!engine) {
+ DRM_DEBUG("Invalid engine[%d]: { class:%d, instance:%d }\n",
+ n, ci.engine_class, ci.engine_instance);
+ __free_engines(set.engines, n);
+ return -ENOENT;
+ }
+
+ set.engines->engines[n] = intel_context_create(ctx, engine);
+ if (!set.engines->engines[n]) {
+ __free_engines(set.engines, n);
+ return -ENOMEM;
+ }
+ }
+ set.engines->num_engines = num_engines;
+
+ err = -EFAULT;
+ if (!get_user(extensions, &user->extensions))
+ err = i915_user_extensions(u64_to_user_ptr(extensions),
+ set_engines__extensions,
+ ARRAY_SIZE(set_engines__extensions),
+ &set);
+ if (err) {
+ free_engines(set.engines);
+ return err;
+ }
+
+replace:
+ mutex_lock(&ctx->engines_mutex);
+ if (args->size)
+ i915_gem_context_set_user_engines(ctx);
+ else
+ i915_gem_context_clear_user_engines(ctx);
+ rcu_swap_protected(ctx->engines, set.engines, 1);
+ mutex_unlock(&ctx->engines_mutex);
+
+ INIT_RCU_WORK(&set.engines->rcu, free_engines_rcu);
+ queue_rcu_work(system_wq, &set.engines->rcu);
+
+ return 0;
+}
+
+static int
+get_engines(struct i915_gem_context *ctx,
+ struct drm_i915_gem_context_param *args)
+{
+ struct i915_context_param_engines __user *user;
+ struct i915_gem_engines *e;
+ size_t n, count, size;
+ int err = 0;
+
+ err = mutex_lock_interruptible(&ctx->engines_mutex);
+ if (err)
+ return err;
+
+ if (!i915_gem_context_user_engines(ctx)) {
+ args->size = 0;
+ goto unlock;
+ }
+
+ e = i915_gem_context_engines(ctx);
+ count = e->num_engines;
+
+ /* Be paranoid in case we have an impedance mismatch */
+ if (!check_struct_size(user, engines, count, &size)) {
+ err = -EINVAL;
+ goto unlock;
+ }
+ if (overflows_type(size, args->size)) {
+ err = -EINVAL;
+ goto unlock;
+ }
+
+ if (!args->size) {
+ args->size = size;
+ goto unlock;
+ }
+
+ if (args->size < size) {
+ err = -EINVAL;
+ goto unlock;
+ }
+
+ user = u64_to_user_ptr(args->value);
+ if (!access_ok(user, size)) {
+ err = -EFAULT;
+ goto unlock;
+ }
+
+ if (put_user(0, &user->extensions)) {
+ err = -EFAULT;
+ goto unlock;
+ }
+
+ for (n = 0; n < count; n++) {
+ struct i915_engine_class_instance ci = {
+ .engine_class = I915_ENGINE_CLASS_INVALID,
+ .engine_instance = I915_ENGINE_CLASS_INVALID_NONE,
+ };
+
+ if (e->engines[n]) {
+ ci.engine_class = e->engines[n]->engine->uabi_class;
+ ci.engine_instance = e->engines[n]->engine->instance;
+ }
+
+ if (copy_to_user(&user->engines[n], &ci, sizeof(ci))) {
+ err = -EFAULT;
+ goto unlock;
+ }
+ }
+
+ args->size = size;
+
+unlock:
+ mutex_unlock(&ctx->engines_mutex);
+ return err;
+}
+
static int ctx_setparam(struct drm_i915_file_private *fpriv,
struct i915_gem_context *ctx,
struct drm_i915_gem_context_param *args)
@@ -1452,6 +1661,10 @@ static int ctx_setparam(struct drm_i915_file_private *fpriv,
ret = set_ppgtt(fpriv, ctx, args);
break;
+ case I915_CONTEXT_PARAM_ENGINES:
+ ret = set_engines(ctx, args);
+ break;
+
case I915_CONTEXT_PARAM_BAN_PERIOD:
default:
ret = -EINVAL;
@@ -1596,9 +1809,7 @@ static int get_sseu(struct i915_gem_context *ctx,
if (user_sseu.flags || user_sseu.rsvd)
return -EINVAL;
- ce = lookup_user_engine(ctx,
- user_sseu.engine.engine_class,
- user_sseu.engine.engine_instance);
+ ce = lookup_user_engine(ctx, 0, &user_sseu.engine);
if (IS_ERR(ce))
return PTR_ERR(ce);
@@ -1682,6 +1893,10 @@ int i915_gem_context_getparam_ioctl(struct drm_device *dev, void *data,
ret = get_ppgtt(file_priv, ctx, args);
break;
+ case I915_CONTEXT_PARAM_ENGINES:
+ ret = get_engines(ctx, args);
+ break;
+
case I915_CONTEXT_PARAM_BAN_PERIOD:
default:
ret = -EINVAL;
diff --git a/drivers/gpu/drm/i915/i915_gem_context.h b/drivers/gpu/drm/i915/i915_gem_context.h
index 272e183ebc0c..9ad4a6362438 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.h
+++ b/drivers/gpu/drm/i915/i915_gem_context.h
@@ -112,6 +112,24 @@ static inline void i915_gem_context_set_force_single_submission(struct i915_gem_
__set_bit(CONTEXT_FORCE_SINGLE_SUBMISSION, &ctx->flags);
}
+static inline bool
+i915_gem_context_user_engines(const struct i915_gem_context *ctx)
+{
+ return test_bit(CONTEXT_USER_ENGINES, &ctx->flags);
+}
+
+static inline void
+i915_gem_context_set_user_engines(struct i915_gem_context *ctx)
+{
+ set_bit(CONTEXT_USER_ENGINES, &ctx->flags);
+}
+
+static inline void
+i915_gem_context_clear_user_engines(struct i915_gem_context *ctx)
+{
+ clear_bit(CONTEXT_USER_ENGINES, &ctx->flags);
+}
+
int __i915_gem_context_pin_hw_id(struct i915_gem_context *ctx);
static inline int i915_gem_context_pin_hw_id(struct i915_gem_context *ctx)
{
diff --git a/drivers/gpu/drm/i915/i915_gem_context_types.h b/drivers/gpu/drm/i915/i915_gem_context_types.h
index d5cb4f121aad..fb965ded2508 100644
--- a/drivers/gpu/drm/i915/i915_gem_context_types.h
+++ b/drivers/gpu/drm/i915/i915_gem_context_types.h
@@ -146,6 +146,7 @@ struct i915_gem_context {
#define CONTEXT_BANNED 0
#define CONTEXT_CLOSED 1
#define CONTEXT_FORCE_SINGLE_SUBMISSION 2
+#define CONTEXT_USER_ENGINES 3
/**
* @hw_id: - unique identifier for the context
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 679f7c1561ba..d6c5220addd0 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -2165,7 +2165,10 @@ eb_select_engine(struct i915_execbuffer *eb,
unsigned int idx;
int err;
- idx = eb_select_legacy_ring(eb, file, args);
+ if (i915_gem_context_user_engines(eb->gem_context))
+ idx = args->flags & I915_EXEC_RING_MASK;
+ else
+ idx = eb_select_legacy_ring(eb, file, args);
ce = i915_gem_context_get_engine(eb->gem_context, idx);
if (IS_ERR(ce))
diff --git a/drivers/gpu/drm/i915/i915_utils.h b/drivers/gpu/drm/i915/i915_utils.h
index c849cfa7cb28..edfc69acdaac 100644
--- a/drivers/gpu/drm/i915/i915_utils.h
+++ b/drivers/gpu/drm/i915/i915_utils.h
@@ -26,6 +26,7 @@
#define __I915_UTILS_H
#include <linux/list.h>
+#include <linux/overflow.h>
#include <linux/sched.h>
#include <linux/types.h>
#include <linux/workqueue.h>
@@ -78,6 +79,39 @@
#define overflows_type(x, T) \
(sizeof(x) > sizeof(T) && (x) >> BITS_PER_TYPE(T))
+static inline bool
+__check_struct_size(size_t base, size_t arr, size_t count, size_t *size)
+{
+ size_t sz;
+
+ if (check_mul_overflow(count, arr, &sz))
+ return false;
+
+ if (check_add_overflow(sz, base, &sz))
+ return false;
+
+ *size = sz;
+ return true;
+}
+
+/**
+ * check_struct_size() - Calculate size of structure with trailing array.
+ * @p: Pointer to the structure.
+ * @member: Name of the array member.
+ * @n: Number of elements in the array.
+ * @sz: Total size of structure and array
+ *
+ * Calculates size of memory needed for structure @p followed by an
+ * array of @n @member elements, like struct_size() but reports
+ * whether it overflowed, and the resultant size in @sz
+ *
+ * Return: false if the calculation overflowed.
+ */
+#define check_struct_size(p, member, n, sz) \
+ likely(__check_struct_size(sizeof(*(p)), \
+ sizeof(*(p)->member) + __must_be_array((p)->member), \
+ n, sz))
+
#define ptr_mask_bits(ptr, n) ({ \
unsigned long __v = (unsigned long)(ptr); \
(typeof(ptr))(__v & -BIT(n)); \
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index d6ad4a15b2b9..8e1bb22926e4 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -136,6 +136,7 @@ enum drm_i915_gem_engine_class {
struct i915_engine_class_instance {
__u16 engine_class; /* see enum drm_i915_gem_engine_class */
__u16 engine_instance;
+#define I915_ENGINE_CLASS_INVALID_NONE -1
};
/**
@@ -1522,6 +1523,26 @@ struct drm_i915_gem_context_param {
* See DRM_I915_GEM_VM_CREATE and DRM_I915_GEM_VM_DESTROY.
*/
#define I915_CONTEXT_PARAM_VM 0x9
+
+/*
+ * I915_CONTEXT_PARAM_ENGINES:
+ *
+ * Bind this context to operate on this subset of available engines. Henceforth,
+ * the I915_EXEC_RING selector for DRM_IOCTL_I915_GEM_EXECBUFFER2 operates as
+ * an index into this array of engines; I915_EXEC_DEFAULT selecting engine[0]
+ * and upwards. Slots 0...N are filled in using the specified (class, instance).
+ * Use
+ * engine_class: I915_ENGINE_CLASS_INVALID,
+ * engine_instance: I915_ENGINE_CLASS_INVALID_NONE
+ * to specify a gap in the array that can be filled in later, e.g. by a
+ * virtual engine used for load balancing.
+ *
+ * Setting the number of engines bound to the context to 0, by passing a zero
+ * sized argument, will revert back to default settings.
+ *
+ * See struct i915_context_param_engines.
+ */
+#define I915_CONTEXT_PARAM_ENGINES 0xa
/* Must be kept compact -- no holes and well documented */
__u64 value;
@@ -1585,6 +1606,16 @@ struct drm_i915_gem_context_param_sseu {
__u32 rsvd;
};
+struct i915_context_param_engines {
+ __u64 extensions; /* linked chain of extension blocks, 0 terminates */
+ struct i915_engine_class_instance engines[0];
+} __attribute__((packed));
+
+#define I915_DEFINE_CONTEXT_PARAM_ENGINES(name__, N__) struct { \
+ __u64 extensions; \
+ struct i915_engine_class_instance engines[N__]; \
+} __attribute__((packed)) name__
+
struct drm_i915_gem_context_create_ext_setparam {
#define I915_CONTEXT_CREATE_EXT_SETPARAM 0
struct i915_user_extension base;
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 11/40] drm/i915: Extend I915_CONTEXT_PARAM_SSEU to support local ctx->engine[]
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (8 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 10/40] drm/i915: Allow a context to define its set of engines Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 10:25 ` Tvrtko Ursulin
2019-05-08 8:06 ` [PATCH 12/40] drm/i915: Re-expose SINGLE_TIMELINE flags for context creation Chris Wilson
` (35 subsequent siblings)
45 siblings, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Allow the user to specify a local engine index (as opposed to
class:index) that they can use to refer to a preset engine inside the
ctx->engine[] array defined by an earlier I915_CONTEXT_PARAM_ENGINES.
This will be useful for setting SSEU parameters on virtual engines that
are local to the context and do not have a valid global class:instance
lookup.
Note that due to the ambiguity in using class:instance with
ctx->engines[], if a user supplied engine map is active the user must
specify the engine to alter by its index into the ctx->engines[].
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/i915_gem_context.c | 24 ++++++++++++++++++++----
include/uapi/drm/i915_drm.h | 3 ++-
2 files changed, 22 insertions(+), 5 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
index 21bfcd529097..5fdb44714a5c 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/i915_gem_context.c
@@ -1363,6 +1363,7 @@ static int set_sseu(struct i915_gem_context *ctx,
struct drm_i915_gem_context_param_sseu user_sseu;
struct intel_context *ce;
struct intel_sseu sseu;
+ unsigned long lookup;
int ret;
if (args->size < sizeof(user_sseu))
@@ -1375,10 +1376,17 @@ static int set_sseu(struct i915_gem_context *ctx,
sizeof(user_sseu)))
return -EFAULT;
- if (user_sseu.flags || user_sseu.rsvd)
+ if (user_sseu.rsvd)
return -EINVAL;
- ce = lookup_user_engine(ctx, 0, &user_sseu.engine);
+ if (user_sseu.flags & ~(I915_CONTEXT_SSEU_FLAG_ENGINE_INDEX))
+ return -EINVAL;
+
+ lookup = 0;
+ if (user_sseu.flags & I915_CONTEXT_SSEU_FLAG_ENGINE_INDEX)
+ lookup |= LOOKUP_USER_INDEX;
+
+ ce = lookup_user_engine(ctx, lookup, &user_sseu.engine);
if (IS_ERR(ce))
return PTR_ERR(ce);
@@ -1795,6 +1803,7 @@ static int get_sseu(struct i915_gem_context *ctx,
{
struct drm_i915_gem_context_param_sseu user_sseu;
struct intel_context *ce;
+ unsigned long lookup;
int err;
if (args->size == 0)
@@ -1806,10 +1815,17 @@ static int get_sseu(struct i915_gem_context *ctx,
sizeof(user_sseu)))
return -EFAULT;
- if (user_sseu.flags || user_sseu.rsvd)
+ if (user_sseu.rsvd)
return -EINVAL;
- ce = lookup_user_engine(ctx, 0, &user_sseu.engine);
+ if (user_sseu.flags & ~(I915_CONTEXT_SSEU_FLAG_ENGINE_INDEX))
+ return -EINVAL;
+
+ lookup = 0;
+ if (user_sseu.flags & I915_CONTEXT_SSEU_FLAG_ENGINE_INDEX)
+ lookup |= LOOKUP_USER_INDEX;
+
+ ce = lookup_user_engine(ctx, lookup, &user_sseu.engine);
if (IS_ERR(ce))
return PTR_ERR(ce);
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index 8e1bb22926e4..82bd488ed0d1 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -1576,9 +1576,10 @@ struct drm_i915_gem_context_param_sseu {
struct i915_engine_class_instance engine;
/*
- * Unused for now. Must be cleared to zero.
+ * Unknown flags must be cleared to zero.
*/
__u32 flags;
+#define I915_CONTEXT_SSEU_FLAG_ENGINE_INDEX (1u << 0)
/*
* Mask of slices to enable for the context. Valid values are a subset
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* Re: [PATCH 11/40] drm/i915: Extend I915_CONTEXT_PARAM_SSEU to support local ctx->engine[]
2019-05-08 8:06 ` [PATCH 11/40] drm/i915: Extend I915_CONTEXT_PARAM_SSEU to support local ctx->engine[] Chris Wilson
@ 2019-05-08 10:25 ` Tvrtko Ursulin
0 siblings, 0 replies; 71+ messages in thread
From: Tvrtko Ursulin @ 2019-05-08 10:25 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 08/05/2019 09:06, Chris Wilson wrote:
> Allow the user to specify a local engine index (as opposed to
> class:index) that they can use to refer to a preset engine inside the
> ctx->engine[] array defined by an earlier I915_CONTEXT_PARAM_ENGINES.
> This will be useful for setting SSEU parameters on virtual engines that
> are local to the context and do not have a valid global class:instance
> lookup.
>
> Note that due to the ambiguity in using class:instance with
> ctx->engines[], if a user supplied engine map is active the user must
> specify the engine to alter by its index into the ctx->engines[].
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> ---
> drivers/gpu/drm/i915/i915_gem_context.c | 24 ++++++++++++++++++++----
> include/uapi/drm/i915_drm.h | 3 ++-
> 2 files changed, 22 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
> index 21bfcd529097..5fdb44714a5c 100644
> --- a/drivers/gpu/drm/i915/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/i915_gem_context.c
> @@ -1363,6 +1363,7 @@ static int set_sseu(struct i915_gem_context *ctx,
> struct drm_i915_gem_context_param_sseu user_sseu;
> struct intel_context *ce;
> struct intel_sseu sseu;
> + unsigned long lookup;
> int ret;
>
> if (args->size < sizeof(user_sseu))
> @@ -1375,10 +1376,17 @@ static int set_sseu(struct i915_gem_context *ctx,
> sizeof(user_sseu)))
> return -EFAULT;
>
> - if (user_sseu.flags || user_sseu.rsvd)
> + if (user_sseu.rsvd)
> return -EINVAL;
>
> - ce = lookup_user_engine(ctx, 0, &user_sseu.engine);
> + if (user_sseu.flags & ~(I915_CONTEXT_SSEU_FLAG_ENGINE_INDEX))
> + return -EINVAL;
> +
> + lookup = 0;
> + if (user_sseu.flags & I915_CONTEXT_SSEU_FLAG_ENGINE_INDEX)
> + lookup |= LOOKUP_USER_INDEX;
> +
> + ce = lookup_user_engine(ctx, lookup, &user_sseu.engine);
> if (IS_ERR(ce))
> return PTR_ERR(ce);
>
> @@ -1795,6 +1803,7 @@ static int get_sseu(struct i915_gem_context *ctx,
> {
> struct drm_i915_gem_context_param_sseu user_sseu;
> struct intel_context *ce;
> + unsigned long lookup;
> int err;
>
> if (args->size == 0)
> @@ -1806,10 +1815,17 @@ static int get_sseu(struct i915_gem_context *ctx,
> sizeof(user_sseu)))
> return -EFAULT;
>
> - if (user_sseu.flags || user_sseu.rsvd)
> + if (user_sseu.rsvd)
> return -EINVAL;
>
> - ce = lookup_user_engine(ctx, 0, &user_sseu.engine);
> + if (user_sseu.flags & ~(I915_CONTEXT_SSEU_FLAG_ENGINE_INDEX))
> + return -EINVAL;
> +
> + lookup = 0;
> + if (user_sseu.flags & I915_CONTEXT_SSEU_FLAG_ENGINE_INDEX)
> + lookup |= LOOKUP_USER_INDEX;
> +
> + ce = lookup_user_engine(ctx, lookup, &user_sseu.engine);
> if (IS_ERR(ce))
> return PTR_ERR(ce);
>
> diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
> index 8e1bb22926e4..82bd488ed0d1 100644
> --- a/include/uapi/drm/i915_drm.h
> +++ b/include/uapi/drm/i915_drm.h
> @@ -1576,9 +1576,10 @@ struct drm_i915_gem_context_param_sseu {
> struct i915_engine_class_instance engine;
>
> /*
> - * Unused for now. Must be cleared to zero.
> + * Unknown flags must be cleared to zero.
> */
> __u32 flags;
> +#define I915_CONTEXT_SSEU_FLAG_ENGINE_INDEX (1u << 0)
>
> /*
> * Mask of slices to enable for the context. Valid values are a subset
>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread
* [PATCH 12/40] drm/i915: Re-expose SINGLE_TIMELINE flags for context creation
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (9 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 11/40] drm/i915: Extend I915_CONTEXT_PARAM_SSEU to support local ctx->engine[] Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 10:26 ` Tvrtko Ursulin
2019-05-08 8:06 ` [PATCH 13/40] drm/i915: Allow userspace to clone contexts on creation Chris Wilson
` (34 subsequent siblings)
45 siblings, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
The SINGLE_TIMELINE flag can be used to create a context such that all
engine instances within that context share a common timeline. This can
be useful for mixing operations between real and virtual engines, or
when using a composite context for a single client API context.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_gem_context.c | 4 ----
include/uapi/drm/i915_drm.h | 3 ++-
2 files changed, 2 insertions(+), 5 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
index 5fdb44714a5c..9cd671298daf 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/i915_gem_context.c
@@ -96,8 +96,6 @@
#include "i915_trace.h"
#include "i915_user_extensions.h"
-#define I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE (1 << 1)
-
#define ALL_L3_SLICES(dev) (1 << NUM_L3_SLICES(dev)) - 1
static struct i915_global_gem_context {
@@ -505,8 +503,6 @@ i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags)
lockdep_assert_held(&dev_priv->drm.struct_mutex);
- BUILD_BUG_ON(I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE &
- ~I915_CONTEXT_CREATE_FLAGS_UNKNOWN);
if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE &&
!HAS_EXECLISTS(dev_priv))
return ERR_PTR(-EINVAL);
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index 82bd488ed0d1..957ba8e60e02 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -1469,8 +1469,9 @@ struct drm_i915_gem_context_create_ext {
__u32 ctx_id; /* output: id of new context*/
__u32 flags;
#define I915_CONTEXT_CREATE_FLAGS_USE_EXTENSIONS (1u << 0)
+#define I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE (1u << 1)
#define I915_CONTEXT_CREATE_FLAGS_UNKNOWN \
- (-(I915_CONTEXT_CREATE_FLAGS_USE_EXTENSIONS << 1))
+ (-(I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE << 1))
__u64 extensions;
};
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* Re: [PATCH 12/40] drm/i915: Re-expose SINGLE_TIMELINE flags for context creation
2019-05-08 8:06 ` [PATCH 12/40] drm/i915: Re-expose SINGLE_TIMELINE flags for context creation Chris Wilson
@ 2019-05-08 10:26 ` Tvrtko Ursulin
0 siblings, 0 replies; 71+ messages in thread
From: Tvrtko Ursulin @ 2019-05-08 10:26 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 08/05/2019 09:06, Chris Wilson wrote:
> The SINGLE_TIMELINE flag can be used to create a context such that all
> engine instances within that context share a common timeline. This can
> be useful for mixing operations between real and virtual engines, or
> when using a composite context for a single client API context.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/i915_gem_context.c | 4 ----
> include/uapi/drm/i915_drm.h | 3 ++-
> 2 files changed, 2 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
> index 5fdb44714a5c..9cd671298daf 100644
> --- a/drivers/gpu/drm/i915/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/i915_gem_context.c
> @@ -96,8 +96,6 @@
> #include "i915_trace.h"
> #include "i915_user_extensions.h"
>
> -#define I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE (1 << 1)
> -
> #define ALL_L3_SLICES(dev) (1 << NUM_L3_SLICES(dev)) - 1
>
> static struct i915_global_gem_context {
> @@ -505,8 +503,6 @@ i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags)
>
> lockdep_assert_held(&dev_priv->drm.struct_mutex);
>
> - BUILD_BUG_ON(I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE &
> - ~I915_CONTEXT_CREATE_FLAGS_UNKNOWN);
> if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE &&
> !HAS_EXECLISTS(dev_priv))
> return ERR_PTR(-EINVAL);
> diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
> index 82bd488ed0d1..957ba8e60e02 100644
> --- a/include/uapi/drm/i915_drm.h
> +++ b/include/uapi/drm/i915_drm.h
> @@ -1469,8 +1469,9 @@ struct drm_i915_gem_context_create_ext {
> __u32 ctx_id; /* output: id of new context*/
> __u32 flags;
> #define I915_CONTEXT_CREATE_FLAGS_USE_EXTENSIONS (1u << 0)
> +#define I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE (1u << 1)
> #define I915_CONTEXT_CREATE_FLAGS_UNKNOWN \
> - (-(I915_CONTEXT_CREATE_FLAGS_USE_EXTENSIONS << 1))
> + (-(I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE << 1))
> __u64 extensions;
> };
>
>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread
* [PATCH 13/40] drm/i915: Allow userspace to clone contexts on creation
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (10 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 12/40] drm/i915: Re-expose SINGLE_TIMELINE flags for context creation Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 14/40] drm/i915: Load balancing across a virtual engine Chris Wilson
` (33 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
A usecase arose out of handling context recovery in mesa, whereby they
wish to recreate a context with fresh logical state but preserving all
other details of the original. Currently, they create a new context and
iterate over which bits they want to copy across, but it would much more
convenient if they were able to just pass in a target context to clone
during creation. This essentially extends the setparam during creation
to pull the details from a target context instead of the user supplied
parameters.
The ideal here is that we don't expose control over anything more than
can be obtained via CONTEXT_PARAM. That is userspace retains explicit
control over all features, and this api is just convenience.
For example, you could replace
struct context_param p = { .param = CONTEXT_PARAM_VM };
param.ctx_id = old_id;
gem_context_get_param(&p.param);
new_id = gem_context_create();
param.ctx_id = new_id;
gem_context_set_param(&p.param);
gem_vm_destroy(param.value); /* drop the ref to VM_ID handle */
with
struct create_ext_param p = {
{ .name = CONTEXT_CREATE_CLONE },
.clone_id = old_id,
.flags = CLONE_FLAGS_VM
}
new_id = gem_context_create_ext(&p);
and not have to worry about stray namespace pollution etc.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/i915_gem_context.c | 206 ++++++++++++++++++++++++
include/uapi/drm/i915_drm.h | 15 ++
2 files changed, 221 insertions(+)
diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
index 9cd671298daf..040c0b770c21 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/i915_gem_context.c
@@ -1697,8 +1697,214 @@ static int create_setparam(struct i915_user_extension __user *ext, void *data)
return ctx_setparam(arg->fpriv, arg->ctx, &local.param);
}
+static int clone_engines(struct i915_gem_context *dst,
+ struct i915_gem_context *src)
+{
+ struct i915_gem_engines *e = i915_gem_context_lock_engines(src);
+ struct i915_gem_engines *clone;
+ bool user_engines;
+ unsigned long n;
+
+ clone = kmalloc(struct_size(e, engines, e->num_engines), GFP_KERNEL);
+ if (!clone)
+ goto err_unlock;
+
+ clone->i915 = dst->i915;
+ for (n = 0; n < e->num_engines; n++) {
+ if (!e->engines[n]) {
+ clone->engines[n] = NULL;
+ continue;
+ }
+
+ clone->engines[n] =
+ intel_context_create(dst, e->engines[n]->engine);
+ if (!clone->engines[n]) {
+ __free_engines(clone, n);
+ goto err_unlock;
+ }
+ }
+ clone->num_engines = n;
+
+ user_engines = i915_gem_context_user_engines(src);
+ i915_gem_context_unlock_engines(src);
+
+ free_engines(dst->engines);
+ RCU_INIT_POINTER(dst->engines, clone);
+ if (user_engines)
+ i915_gem_context_set_user_engines(dst);
+ else
+ i915_gem_context_clear_user_engines(dst);
+ return 0;
+
+err_unlock:
+ i915_gem_context_unlock_engines(src);
+ return -ENOMEM;
+}
+
+static int clone_flags(struct i915_gem_context *dst,
+ struct i915_gem_context *src)
+{
+ dst->user_flags = src->user_flags;
+ return 0;
+}
+
+static int clone_schedattr(struct i915_gem_context *dst,
+ struct i915_gem_context *src)
+{
+ dst->sched = src->sched;
+ return 0;
+}
+
+static int clone_sseu(struct i915_gem_context *dst,
+ struct i915_gem_context *src)
+{
+ struct i915_gem_engines *e = i915_gem_context_lock_engines(src);
+ struct i915_gem_engines *clone;
+ unsigned long n;
+ int err;
+
+ clone = dst->engines; /* no locking required; sole access */
+ if (e->num_engines != clone->num_engines) {
+ err = -EINVAL;
+ goto unlock;
+ }
+
+ for (n = 0; n < e->num_engines; n++) {
+ struct intel_context *ce = e->engines[n];
+
+ if (clone->engines[n]->engine->class != ce->engine->class) {
+ /* Must have compatible engine maps! */
+ err = -EINVAL;
+ goto unlock;
+ }
+
+ /* serialises with set_sseu */
+ err = intel_context_lock_pinned(ce);
+ if (err)
+ goto unlock;
+
+ clone->engines[n]->sseu = ce->sseu;
+ intel_context_unlock_pinned(ce);
+ }
+
+ err = 0;
+unlock:
+ i915_gem_context_unlock_engines(src);
+ return err;
+}
+
+static int clone_timeline(struct i915_gem_context *dst,
+ struct i915_gem_context *src)
+{
+ if (src->timeline) {
+ GEM_BUG_ON(src->timeline == dst->timeline);
+
+ if (dst->timeline)
+ i915_timeline_put(dst->timeline);
+ dst->timeline = i915_timeline_get(src->timeline);
+ }
+
+ return 0;
+}
+
+static int clone_vm(struct i915_gem_context *dst,
+ struct i915_gem_context *src)
+{
+ struct i915_hw_ppgtt *ppgtt;
+
+ rcu_read_lock();
+ do {
+ ppgtt = READ_ONCE(src->ppgtt);
+ if (!ppgtt)
+ break;
+
+ if (!kref_get_unless_zero(&ppgtt->ref))
+ continue;
+
+ /*
+ * This ppgtt may have be reallocated between
+ * the read and the kref, and reassigned to a third
+ * context. In order to avoid inadvertent sharing
+ * of this ppgtt with that third context (and not
+ * src), we have to confirm that we have the same
+ * ppgtt after passing through the strong memory
+ * barrier implied by a successful
+ * kref_get_unless_zero().
+ *
+ * Once we have acquired the current ppgtt of src,
+ * we no longer care if it is released from src, as
+ * it cannot be reallocated elsewhere.
+ */
+
+ if (ppgtt == READ_ONCE(src->ppgtt))
+ break;
+
+ i915_ppgtt_put(ppgtt);
+ } while (1);
+ rcu_read_unlock();
+
+ if (ppgtt) {
+ __assign_ppgtt(dst, ppgtt);
+ i915_ppgtt_put(ppgtt);
+ }
+
+ return 0;
+}
+
+static int create_clone(struct i915_user_extension __user *ext, void *data)
+{
+ static int (* const fn[])(struct i915_gem_context *dst,
+ struct i915_gem_context *src) = {
+#define MAP(x, y) [ilog2(I915_CONTEXT_CLONE_##x)] = y
+ MAP(ENGINES, clone_engines),
+ MAP(FLAGS, clone_flags),
+ MAP(SCHEDATTR, clone_schedattr),
+ MAP(SSEU, clone_sseu),
+ MAP(TIMELINE, clone_timeline),
+ MAP(VM, clone_vm),
+#undef MAP
+ };
+ struct drm_i915_gem_context_create_ext_clone local;
+ const struct create_ext *arg = data;
+ struct i915_gem_context *dst = arg->ctx;
+ struct i915_gem_context *src;
+ int err, bit;
+
+ if (copy_from_user(&local, ext, sizeof(local)))
+ return -EFAULT;
+
+ BUILD_BUG_ON(GENMASK(BITS_PER_TYPE(local.flags) - 1, ARRAY_SIZE(fn)) !=
+ I915_CONTEXT_CLONE_UNKNOWN);
+
+ if (local.flags & I915_CONTEXT_CLONE_UNKNOWN)
+ return -EINVAL;
+
+ if (local.rsvd)
+ return -EINVAL;
+
+ rcu_read_lock();
+ src = __i915_gem_context_lookup_rcu(arg->fpriv, local.clone_id);
+ rcu_read_unlock();
+ if (!src)
+ return -ENOENT;
+
+ GEM_BUG_ON(src == dst);
+
+ for (bit = 0; bit < ARRAY_SIZE(fn); bit++) {
+ if (!(local.flags & BIT(bit)))
+ continue;
+
+ err = fn[bit](dst, src);
+ if (err)
+ return err;
+ }
+
+ return 0;
+}
+
static const i915_user_extension_fn create_extensions[] = {
[I915_CONTEXT_CREATE_EXT_SETPARAM] = create_setparam,
+ [I915_CONTEXT_CREATE_EXT_CLONE] = create_clone,
};
static bool client_is_banned(struct drm_i915_file_private *file_priv)
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index 957ba8e60e02..62396d575e28 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -1624,6 +1624,21 @@ struct drm_i915_gem_context_create_ext_setparam {
struct drm_i915_gem_context_param param;
};
+struct drm_i915_gem_context_create_ext_clone {
+#define I915_CONTEXT_CREATE_EXT_CLONE 1
+ struct i915_user_extension base;
+ __u32 clone_id;
+ __u32 flags;
+#define I915_CONTEXT_CLONE_ENGINES (1u << 0)
+#define I915_CONTEXT_CLONE_FLAGS (1u << 1)
+#define I915_CONTEXT_CLONE_SCHEDATTR (1u << 2)
+#define I915_CONTEXT_CLONE_SSEU (1u << 3)
+#define I915_CONTEXT_CLONE_TIMELINE (1u << 4)
+#define I915_CONTEXT_CLONE_VM (1u << 5)
+#define I915_CONTEXT_CLONE_UNKNOWN -(I915_CONTEXT_CLONE_VM << 1)
+ __u64 rsvd;
+};
+
struct drm_i915_gem_context_destroy {
__u32 ctx_id;
__u32 pad;
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 14/40] drm/i915: Load balancing across a virtual engine
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (11 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 13/40] drm/i915: Allow userspace to clone contexts on creation Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 10:29 ` Tvrtko Ursulin
2019-05-08 11:23 ` [PATCH] " Chris Wilson
2019-05-08 8:06 ` [PATCH 15/40] drm/i915: Apply an execution_mask to the virtual_engine Chris Wilson
` (32 subsequent siblings)
45 siblings, 2 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Having allowed the user to define a set of engines that they will want
to only use, we go one step further and allow them to bind those engines
into a single virtual instance. Submitting a batch to the virtual engine
will then forward it to any one of the set in a manner as best to
distribute load. The virtual engine has a single timeline across all
engines (it operates as a single queue), so it is not able to concurrently
run batches across multiple engines by itself; that is left up to the user
to submit multiple concurrent batches to multiple queues. Multiple users
will be load balanced across the system.
The mechanism used for load balancing in this patch is a late greedy
balancer. When a request is ready for execution, it is added to each
engine's queue, and when an engine is ready for its next request it
claims it from the virtual engine. The first engine to do so, wins, i.e.
the request is executed at the earliest opportunity (idle moment) in the
system.
As not all HW is created equal, the user is still able to skip the
virtual engine and execute the batch on a specific engine, all within the
same queue. It will then be executed in order on the correct engine,
with execution on other virtual engines being moved away due to the load
detection.
A couple of areas for potential improvement left!
- The virtual engine always take priority over equal-priority tasks.
Mostly broken up by applying FQ_CODEL rules for prioritising new clients,
and hopefully the virtual and real engines are not then congested (i.e.
all work is via virtual engines, or all work is to the real engine).
- We require the breadcrumb irq around every virtual engine request. For
normal engines, we eliminate the need for the slow round trip via
interrupt by using the submit fence and queueing in order. For virtual
engines, we have to allow any job to transfer to a new ring, and cannot
coalesce the submissions, so require the completion fence instead,
forcing the persistent use of interrupts.
- We only drip feed single requests through each virtual engine and onto
the physical engines, even if there was enough work to fill all ELSP,
leaving small stalls with an idle CS event at the end of every request.
Could we be greedy and fill both slots? Being lazy is virtuous for load
distribution on less-than-full workloads though.
Other areas of improvement are more general, such as reducing lock
contention, reducing dispatch overhead, looking at direct submission
rather than bouncing around tasklets etc.
sseu: Lift the restriction to allow sseu to be reconfigured on virtual
engines composed of RENDER_CLASS (rcs).
v2: macroize check_user_mbz()
v3: Cancel virtual engines on wedging
v4: Commence commenting
v5: Replace 64b sibling_mask with a list of class:instance
v6: Drop the one-element array in the uabi
v7: Assert it is an virtual engine in to_virtual_engine()
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/gt/intel_engine_types.h | 8 +
drivers/gpu/drm/i915/gt/intel_lrc.c | 683 ++++++++++++++++++-
drivers/gpu/drm/i915/gt/intel_lrc.h | 9 +
drivers/gpu/drm/i915/gt/selftest_lrc.c | 180 +++++
drivers/gpu/drm/i915/i915_gem.h | 5 +
drivers/gpu/drm/i915/i915_gem_context.c | 116 +++-
drivers/gpu/drm/i915/i915_scheduler.c | 19 +-
drivers/gpu/drm/i915/i915_timeline_types.h | 1 +
include/uapi/drm/i915_drm.h | 39 ++
9 files changed, 1032 insertions(+), 28 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
index e381c1c73902..7b47e00fa082 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
@@ -227,6 +227,7 @@ struct intel_engine_execlists {
* @queue: queue of requests, in priority lists
*/
struct rb_root_cached queue;
+ struct rb_root_cached virtual;
/**
* @csb_write: control register for Context Switch buffer
@@ -445,6 +446,7 @@ struct intel_engine_cs {
#define I915_ENGINE_HAS_PREEMPTION BIT(2)
#define I915_ENGINE_HAS_SEMAPHORES BIT(3)
#define I915_ENGINE_NEEDS_BREADCRUMB_TASKLET BIT(4)
+#define I915_ENGINE_IS_VIRTUAL BIT(5)
unsigned int flags;
/*
@@ -534,6 +536,12 @@ intel_engine_needs_breadcrumb_tasklet(const struct intel_engine_cs *engine)
return engine->flags & I915_ENGINE_NEEDS_BREADCRUMB_TASKLET;
}
+static inline bool
+intel_engine_is_virtual(const struct intel_engine_cs *engine)
+{
+ return engine->flags & I915_ENGINE_IS_VIRTUAL;
+}
+
#define instdone_slice_mask(dev_priv__) \
(IS_GEN(dev_priv__, 7) ? \
1 : RUNTIME_INFO(dev_priv__)->sseu.slice_mask)
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index f1d62746e066..bc388df39802 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -166,6 +166,42 @@
#define ACTIVE_PRIORITY (I915_PRIORITY_NEWCLIENT | I915_PRIORITY_NOSEMAPHORE)
+struct virtual_engine {
+ struct intel_engine_cs base;
+ struct intel_context context;
+
+ /*
+ * We allow only a single request through the virtual engine at a time
+ * (each request in the timeline waits for the completion fence of
+ * the previous before being submitted). By restricting ourselves to
+ * only submitting a single request, each request is placed on to a
+ * physical to maximise load spreading (by virtue of the late greedy
+ * scheduling -- each real engine takes the next available request
+ * upon idling).
+ */
+ struct i915_request *request;
+
+ /*
+ * We keep a rbtree of available virtual engines inside each physical
+ * engine, sorted by priority. Here we preallocate the nodes we need
+ * for the virtual engine, indexed by physical_engine->id.
+ */
+ struct ve_node {
+ struct rb_node rb;
+ int prio;
+ } nodes[I915_NUM_ENGINES];
+
+ /* And finally, which physical engines this virtual engine maps onto. */
+ unsigned int num_siblings;
+ struct intel_engine_cs *siblings[0];
+};
+
+static struct virtual_engine *to_virtual_engine(struct intel_engine_cs *engine)
+{
+ GEM_BUG_ON(!intel_engine_is_virtual(engine));
+ return container_of(engine, struct virtual_engine, base);
+}
+
static int execlists_context_deferred_alloc(struct intel_context *ce,
struct intel_engine_cs *engine);
static void execlists_init_reg_state(u32 *reg_state,
@@ -229,7 +265,8 @@ static int queue_prio(const struct intel_engine_execlists *execlists)
}
static inline bool need_preempt(const struct intel_engine_cs *engine,
- const struct i915_request *rq)
+ const struct i915_request *rq,
+ struct rb_node *rb)
{
int last_prio;
@@ -264,6 +301,25 @@ static inline bool need_preempt(const struct intel_engine_cs *engine,
rq_prio(list_next_entry(rq, link)) > last_prio)
return true;
+ if (rb) {
+ struct virtual_engine *ve =
+ rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
+ bool preempt = false;
+
+ if (engine == ve->siblings[0]) { /* only preempt one sibling */
+ struct i915_request *next;
+
+ rcu_read_lock();
+ next = READ_ONCE(ve->request);
+ if (next)
+ preempt = rq_prio(next) > last_prio;
+ rcu_read_unlock();
+ }
+
+ if (preempt)
+ return preempt;
+ }
+
/*
* If the inflight context did not trigger the preemption, then maybe
* it was the set of queued requests? Pick the highest priority in
@@ -382,6 +438,8 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
list_for_each_entry_safe_reverse(rq, rn,
&engine->timeline.requests,
link) {
+ struct intel_engine_cs *owner;
+
if (i915_request_completed(rq))
break;
@@ -390,16 +448,32 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
GEM_BUG_ON(rq->hw_context->active);
- GEM_BUG_ON(rq_prio(rq) == I915_PRIORITY_INVALID);
- if (rq_prio(rq) != prio) {
- prio = rq_prio(rq);
- pl = i915_sched_lookup_priolist(engine, prio);
- }
- GEM_BUG_ON(RB_EMPTY_ROOT(&engine->execlists.queue.rb_root));
+ /*
+ * Push the request back into the queue for later resubmission.
+ * If this request is not native to this physical engine (i.e.
+ * it came from a virtual source), push it back onto the virtual
+ * engine so that it can be moved across onto another physical
+ * engine as load dictates.
+ */
+ owner = rq->hw_context->engine;
+ if (likely(owner == engine)) {
+ GEM_BUG_ON(rq_prio(rq) == I915_PRIORITY_INVALID);
+ if (rq_prio(rq) != prio) {
+ prio = rq_prio(rq);
+ pl = i915_sched_lookup_priolist(engine, prio);
+ }
+ GEM_BUG_ON(RB_EMPTY_ROOT(&engine->execlists.queue.rb_root));
- list_add(&rq->sched.link, pl);
+ list_add(&rq->sched.link, pl);
+ active = rq;
+ } else {
+ if (__i915_request_has_started(rq))
+ rq->sched.attr.priority |= boost;
- active = rq;
+ rq->engine = owner;
+ owner->submit_request(rq);
+ active = NULL;
+ }
}
/*
@@ -419,7 +493,7 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
* in the priority queue, but they will not gain immediate access to
* the GPU.
*/
- if (~prio & boost && __i915_request_has_started(active)) {
+ if (~prio & boost && active && __i915_request_has_started(active)) {
prio |= boost;
GEM_BUG_ON(active->sched.attr.priority >= prio);
active->sched.attr.priority = prio;
@@ -661,6 +735,90 @@ static void complete_preempt_context(struct intel_engine_execlists *execlists)
ACTIVE_PRIORITY);
}
+static void virtual_update_register_offsets(u32 *regs,
+ struct intel_engine_cs *engine)
+{
+ u32 base = engine->mmio_base;
+
+ /* Must match execlists_init_reg_state()! */
+
+ regs[CTX_CONTEXT_CONTROL] =
+ i915_mmio_reg_offset(RING_CONTEXT_CONTROL(base));
+ regs[CTX_RING_HEAD] = i915_mmio_reg_offset(RING_HEAD(base));
+ regs[CTX_RING_TAIL] = i915_mmio_reg_offset(RING_TAIL(base));
+ regs[CTX_RING_BUFFER_START] = i915_mmio_reg_offset(RING_START(base));
+ regs[CTX_RING_BUFFER_CONTROL] = i915_mmio_reg_offset(RING_CTL(base));
+
+ regs[CTX_BB_HEAD_U] = i915_mmio_reg_offset(RING_BBADDR_UDW(base));
+ regs[CTX_BB_HEAD_L] = i915_mmio_reg_offset(RING_BBADDR(base));
+ regs[CTX_BB_STATE] = i915_mmio_reg_offset(RING_BBSTATE(base));
+ regs[CTX_SECOND_BB_HEAD_U] =
+ i915_mmio_reg_offset(RING_SBBADDR_UDW(base));
+ regs[CTX_SECOND_BB_HEAD_L] = i915_mmio_reg_offset(RING_SBBADDR(base));
+ regs[CTX_SECOND_BB_STATE] = i915_mmio_reg_offset(RING_SBBSTATE(base));
+
+ regs[CTX_CTX_TIMESTAMP] =
+ i915_mmio_reg_offset(RING_CTX_TIMESTAMP(base));
+ regs[CTX_PDP3_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 3));
+ regs[CTX_PDP3_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 3));
+ regs[CTX_PDP2_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 2));
+ regs[CTX_PDP2_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 2));
+ regs[CTX_PDP1_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 1));
+ regs[CTX_PDP1_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 1));
+ regs[CTX_PDP0_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 0));
+ regs[CTX_PDP0_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 0));
+
+ if (engine->class == RENDER_CLASS) {
+ regs[CTX_RCS_INDIRECT_CTX] =
+ i915_mmio_reg_offset(RING_INDIRECT_CTX(base));
+ regs[CTX_RCS_INDIRECT_CTX_OFFSET] =
+ i915_mmio_reg_offset(RING_INDIRECT_CTX_OFFSET(base));
+ regs[CTX_BB_PER_CTX_PTR] =
+ i915_mmio_reg_offset(RING_BB_PER_CTX_PTR(base));
+
+ regs[CTX_R_PWR_CLK_STATE] =
+ i915_mmio_reg_offset(GEN8_R_PWR_CLK_STATE);
+ }
+}
+
+static bool virtual_matches(const struct virtual_engine *ve,
+ const struct i915_request *rq,
+ const struct intel_engine_cs *engine)
+{
+ const struct intel_engine_cs *active;
+
+ /*
+ * We track when the HW has completed saving the context image
+ * (i.e. when we have seen the final CS event switching out of
+ * the context) and must not overwrite the context image before
+ * then. This restricts us to only using the active engine
+ * while the previous virtualized request is inflight (so
+ * we reuse the register offsets). This is a very small
+ * hystersis on the greedy seelction algorithm.
+ */
+ active = READ_ONCE(ve->context.active);
+ if (active && active != engine)
+ return false;
+
+ return true;
+}
+
+static void virtual_xfer_breadcrumbs(struct virtual_engine *ve,
+ struct intel_engine_cs *engine)
+{
+ struct intel_engine_cs *old = ve->siblings[0];
+
+ /* All unattached (rq->engine == old) must already be completed */
+
+ spin_lock(&old->breadcrumbs.irq_lock);
+ if (!list_empty(&ve->context.signal_link)) {
+ list_move_tail(&ve->context.signal_link,
+ &engine->breadcrumbs.signalers);
+ intel_engine_queue_breadcrumbs(engine);
+ }
+ spin_unlock(&old->breadcrumbs.irq_lock);
+}
+
static void execlists_dequeue(struct intel_engine_cs *engine)
{
struct intel_engine_execlists * const execlists = &engine->execlists;
@@ -693,6 +851,26 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
* and context switches) submission.
*/
+ for (rb = rb_first_cached(&execlists->virtual); rb; ) {
+ struct virtual_engine *ve =
+ rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
+ struct i915_request *rq = READ_ONCE(ve->request);
+
+ if (!rq) { /* lazily cleanup after another engine handled rq */
+ rb_erase_cached(rb, &execlists->virtual);
+ RB_CLEAR_NODE(rb);
+ rb = rb_first_cached(&execlists->virtual);
+ continue;
+ }
+
+ if (!virtual_matches(ve, rq, engine)) {
+ rb = rb_next(rb);
+ continue;
+ }
+
+ break;
+ }
+
if (last) {
/*
* Don't resubmit or switch until all outstanding
@@ -714,7 +892,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
if (!execlists_is_active(execlists, EXECLISTS_ACTIVE_HWACK))
return;
- if (need_preempt(engine, last)) {
+ if (need_preempt(engine, last, rb)) {
inject_preempt_context(engine);
return;
}
@@ -754,6 +932,92 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
last->tail = last->wa_tail;
}
+ while (rb) { /* XXX virtual is always taking precedence */
+ struct virtual_engine *ve =
+ rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
+ struct i915_request *rq;
+
+ spin_lock(&ve->base.timeline.lock);
+
+ rq = ve->request;
+ if (unlikely(!rq)) { /* lost the race to a sibling */
+ spin_unlock(&ve->base.timeline.lock);
+ rb_erase_cached(rb, &execlists->virtual);
+ RB_CLEAR_NODE(rb);
+ rb = rb_first_cached(&execlists->virtual);
+ continue;
+ }
+
+ GEM_BUG_ON(rq != ve->request);
+ GEM_BUG_ON(rq->engine != &ve->base);
+ GEM_BUG_ON(rq->hw_context != &ve->context);
+
+ if (rq_prio(rq) >= queue_prio(execlists)) {
+ if (!virtual_matches(ve, rq, engine)) {
+ spin_unlock(&ve->base.timeline.lock);
+ rb = rb_next(rb);
+ continue;
+ }
+
+ if (last && !can_merge_rq(last, rq)) {
+ spin_unlock(&ve->base.timeline.lock);
+ return; /* leave this rq for another engine */
+ }
+
+ GEM_TRACE("%s: virtual rq=%llx:%lld%s, new engine? %s\n",
+ engine->name,
+ rq->fence.context,
+ rq->fence.seqno,
+ i915_request_completed(rq) ? "!" :
+ i915_request_started(rq) ? "*" :
+ "",
+ yesno(engine != ve->siblings[0]));
+
+ ve->request = NULL;
+ ve->base.execlists.queue_priority_hint = INT_MIN;
+ rb_erase_cached(rb, &execlists->virtual);
+ RB_CLEAR_NODE(rb);
+
+ rq->engine = engine;
+
+ if (engine != ve->siblings[0]) {
+ u32 *regs = ve->context.lrc_reg_state;
+ unsigned int n;
+
+ GEM_BUG_ON(READ_ONCE(ve->context.active));
+ virtual_update_register_offsets(regs, engine);
+
+ if (!list_empty(&ve->context.signals))
+ virtual_xfer_breadcrumbs(ve, engine);
+
+ /*
+ * Move the bound engine to the top of the list
+ * for future execution. We then kick this
+ * tasklet first before checking others, so that
+ * we preferentially reuse this set of bound
+ * registers.
+ */
+ for (n = 1; n < ve->num_siblings; n++) {
+ if (ve->siblings[n] == engine) {
+ swap(ve->siblings[n],
+ ve->siblings[0]);
+ break;
+ }
+ }
+
+ GEM_BUG_ON(ve->siblings[0] != engine);
+ }
+
+ __i915_request_submit(rq);
+ trace_i915_request_in(rq, port_index(port, execlists));
+ submit = true;
+ last = rq;
+ }
+
+ spin_unlock(&ve->base.timeline.lock);
+ break;
+ }
+
while ((rb = rb_first_cached(&execlists->queue))) {
struct i915_priolist *p = to_priolist(rb);
struct i915_request *rq, *rn;
@@ -1877,6 +2141,25 @@ static void reset_csb_pointers(struct intel_engine_execlists *execlists)
&execlists->csb_status[reset_value]);
}
+static struct i915_request *active_request(struct i915_request *rq)
+{
+ const struct list_head * const list = &rq->engine->timeline.requests;
+ const struct intel_context * const context = rq->hw_context;
+ struct i915_request *active = NULL;
+
+ list_for_each_entry_from_reverse(rq, list, link) {
+ if (i915_request_completed(rq))
+ break;
+
+ if (rq->hw_context != context)
+ break;
+
+ active = rq;
+ }
+
+ return active;
+}
+
static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
{
struct intel_engine_execlists * const execlists = &engine->execlists;
@@ -1897,7 +2180,8 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
if (!port_isset(execlists->port))
goto out_clear;
- ce = port_request(execlists->port)->hw_context;
+ rq = port_request(execlists->port);
+ ce = rq->hw_context;
/*
* Catch up with any missed context-switch interrupts.
@@ -1910,16 +2194,10 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
*/
execlists_cancel_port_requests(execlists);
- /* Push back any incomplete requests for replay after the reset. */
- rq = __unwind_incomplete_requests(engine, 0);
+ rq = active_request(rq);
if (!rq)
goto out_replay;
- if (rq->hw_context != ce) { /* caught just before a CS event */
- rq = NULL;
- goto out_replay;
- }
-
/*
* If this request hasn't started yet, e.g. it is waiting on a
* semaphore, we need to avoid skipping the request or else we
@@ -1966,13 +2244,16 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
}
execlists_init_reg_state(regs, ce, engine, ce->ring);
- /* Rerun the request; its payload has been neutered (if guilty). */
out_replay:
+ /* Rerun the request; its payload has been neutered (if guilty). */
ce->ring->head =
rq ? intel_ring_wrap(ce->ring, rq->head) : ce->ring->tail;
intel_ring_update_space(ce->ring);
__execlists_update_reg_state(ce, engine);
+ /* Push back any incomplete requests for replay after the reset. */
+ __unwind_incomplete_requests(engine, 0);
+
out_clear:
execlists_clear_all_active(execlists);
}
@@ -2046,6 +2327,26 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine)
i915_priolist_free(p);
}
+ /* Cancel all attached virtual engines */
+ while ((rb = rb_first_cached(&execlists->virtual))) {
+ struct virtual_engine *ve =
+ rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
+
+ rb_erase_cached(rb, &execlists->virtual);
+ RB_CLEAR_NODE(rb);
+
+ spin_lock(&ve->base.timeline.lock);
+ if (ve->request) {
+ ve->request->engine = engine;
+ __i915_request_submit(ve->request);
+ dma_fence_set_error(&ve->request->fence, -EIO);
+ i915_request_mark_complete(ve->request);
+ ve->base.execlists.queue_priority_hint = INT_MIN;
+ ve->request = NULL;
+ }
+ spin_unlock(&ve->base.timeline.lock);
+ }
+
/* Remaining _unready_ requests will be nop'ed when submitted */
execlists->queue_priority_hint = INT_MIN;
@@ -2529,12 +2830,15 @@ static void execlists_init_reg_state(u32 *regs,
bool rcs = engine->class == RENDER_CLASS;
u32 base = engine->mmio_base;
- /* A context is actually a big batch buffer with several
+ /*
+ * A context is actually a big batch buffer with several
* MI_LOAD_REGISTER_IMM commands followed by (reg, value) pairs. The
* values we are setting here are only for the first context restore:
* on a subsequent save, the GPU will recreate this batchbuffer with new
* values (including all the missing MI_LOAD_REGISTER_IMM commands that
* we are not initializing here).
+ *
+ * Must keep consistent with virtual_update_register_offsets().
*/
regs[CTX_LRI_HEADER_0] = MI_LOAD_REGISTER_IMM(rcs ? 14 : 11) |
MI_LRI_FORCE_POSTED;
@@ -2753,6 +3057,320 @@ static int execlists_context_deferred_alloc(struct intel_context *ce,
return ret;
}
+static void virtual_context_destroy(struct kref *kref)
+{
+ struct virtual_engine *ve =
+ container_of(kref, typeof(*ve), context.ref);
+ unsigned int n;
+
+ GEM_BUG_ON(ve->request);
+ GEM_BUG_ON(ve->context.active);
+
+ for (n = 0; n < ve->num_siblings; n++) {
+ struct intel_engine_cs *sibling = ve->siblings[n];
+ struct rb_node *node = &ve->nodes[sibling->id].rb;
+
+ if (RB_EMPTY_NODE(node))
+ continue;
+
+ spin_lock_irq(&sibling->timeline.lock);
+
+ /* Detachment is lazily performed in the execlists tasklet */
+ if (!RB_EMPTY_NODE(node))
+ rb_erase_cached(node, &sibling->execlists.virtual);
+
+ spin_unlock_irq(&sibling->timeline.lock);
+ }
+ GEM_BUG_ON(__tasklet_is_scheduled(&ve->base.execlists.tasklet));
+
+ if (ve->context.state)
+ __execlists_context_fini(&ve->context);
+
+ i915_timeline_fini(&ve->base.timeline);
+ kfree(ve);
+}
+
+static void virtual_engine_initial_hint(struct virtual_engine *ve)
+{
+ int swp;
+
+ /*
+ * Pick a random sibling on starting to help spread the load around.
+ *
+ * New contexts are typically created with exactly the same order
+ * of siblings, and often started in batches. Due to the way we iterate
+ * the array of sibling when submitting requests, sibling[0] is
+ * prioritised for dequeuing. If we make sure that sibling[0] is fairly
+ * randomised across the system, we also help spread the load by the
+ * first engine we inspect being different each time.
+ *
+ * NB This does not force us to execute on this engine, it will just
+ * typically be the first we inspect for submission.
+ */
+ swp = prandom_u32_max(ve->num_siblings);
+ if (!swp)
+ return;
+
+ swap(ve->siblings[swp], ve->siblings[0]);
+ virtual_update_register_offsets(ve->context.lrc_reg_state,
+ ve->siblings[0]);
+}
+
+static int virtual_context_pin(struct intel_context *ce)
+{
+ struct virtual_engine *ve = container_of(ce, typeof(*ve), context);
+ int err;
+
+ /* Note: we must use a real engine class for setting up reg state */
+ err = __execlists_context_pin(ce, ve->siblings[0]);
+ if (err)
+ return err;
+
+ virtual_engine_initial_hint(ve);
+ return 0;
+}
+
+static void virtual_context_enter(struct intel_context *ce)
+{
+ struct virtual_engine *ve = container_of(ce, typeof(*ve), context);
+ unsigned int n;
+
+ for (n = 0; n < ve->num_siblings; n++)
+ intel_engine_pm_get(ve->siblings[n]);
+}
+
+static void virtual_context_exit(struct intel_context *ce)
+{
+ struct virtual_engine *ve = container_of(ce, typeof(*ve), context);
+ unsigned int n;
+
+ ce->saturated = 0;
+ for (n = 0; n < ve->num_siblings; n++)
+ intel_engine_pm_put(ve->siblings[n]);
+}
+
+static const struct intel_context_ops virtual_context_ops = {
+ .pin = virtual_context_pin,
+ .unpin = execlists_context_unpin,
+
+ .enter = virtual_context_enter,
+ .exit = virtual_context_exit,
+
+ .destroy = virtual_context_destroy,
+};
+
+static void virtual_submission_tasklet(unsigned long data)
+{
+ struct virtual_engine * const ve = (struct virtual_engine *)data;
+ const int prio = ve->base.execlists.queue_priority_hint;
+ unsigned int n;
+
+ local_irq_disable();
+ for (n = 0; READ_ONCE(ve->request) && n < ve->num_siblings; n++) {
+ struct intel_engine_cs *sibling = ve->siblings[n];
+ struct ve_node * const node = &ve->nodes[sibling->id];
+ struct rb_node **parent, *rb;
+ bool first;
+
+ spin_lock(&sibling->timeline.lock);
+
+ if (!RB_EMPTY_NODE(&node->rb)) {
+ /*
+ * Cheat and avoid rebalancing the tree if we can
+ * reuse this node in situ.
+ */
+ first = rb_first_cached(&sibling->execlists.virtual) ==
+ &node->rb;
+ if (prio == node->prio || (prio > node->prio && first))
+ goto submit_engine;
+
+ rb_erase_cached(&node->rb, &sibling->execlists.virtual);
+ }
+
+ rb = NULL;
+ first = true;
+ parent = &sibling->execlists.virtual.rb_root.rb_node;
+ while (*parent) {
+ struct ve_node *other;
+
+ rb = *parent;
+ other = rb_entry(rb, typeof(*other), rb);
+ if (prio > other->prio) {
+ parent = &rb->rb_left;
+ } else {
+ parent = &rb->rb_right;
+ first = false;
+ }
+ }
+
+ rb_link_node(&node->rb, rb, parent);
+ rb_insert_color_cached(&node->rb,
+ &sibling->execlists.virtual,
+ first);
+
+submit_engine:
+ GEM_BUG_ON(RB_EMPTY_NODE(&node->rb));
+ node->prio = prio;
+ if (first && prio > sibling->execlists.queue_priority_hint) {
+ sibling->execlists.queue_priority_hint = prio;
+ tasklet_hi_schedule(&sibling->execlists.tasklet);
+ }
+
+ spin_unlock(&sibling->timeline.lock);
+ }
+ local_irq_enable();
+}
+
+static void virtual_submit_request(struct i915_request *rq)
+{
+ struct virtual_engine *ve = to_virtual_engine(rq->engine);
+
+ GEM_TRACE("%s: rq=%llx:%lld\n",
+ ve->base.name,
+ rq->fence.context,
+ rq->fence.seqno);
+
+ GEM_BUG_ON(ve->base.submit_request != virtual_submit_request);
+
+ GEM_BUG_ON(ve->request);
+ ve->base.execlists.queue_priority_hint = rq_prio(rq);
+ WRITE_ONCE(ve->request, rq);
+
+ tasklet_schedule(&ve->base.execlists.tasklet);
+}
+
+struct intel_context *
+intel_execlists_create_virtual(struct i915_gem_context *ctx,
+ struct intel_engine_cs **siblings,
+ unsigned int count)
+{
+ struct virtual_engine *ve;
+ unsigned int n;
+ int err;
+
+ if (count == 0)
+ return ERR_PTR(-EINVAL);
+
+ if (count == 1)
+ return intel_context_create(ctx, siblings[0]);
+
+ ve = kzalloc(struct_size(ve, siblings, count), GFP_KERNEL);
+ if (!ve)
+ return ERR_PTR(-ENOMEM);
+
+ ve->base.i915 = ctx->i915;
+ ve->base.id = -1;
+ ve->base.class = OTHER_CLASS;
+ ve->base.uabi_class = I915_ENGINE_CLASS_INVALID;
+ ve->base.instance = I915_ENGINE_CLASS_INVALID_VIRTUAL;
+ ve->base.flags = I915_ENGINE_IS_VIRTUAL;
+
+ snprintf(ve->base.name, sizeof(ve->base.name), "virtual");
+
+ err = i915_timeline_init(ctx->i915, &ve->base.timeline, NULL);
+ if (err)
+ goto err_put;
+ i915_timeline_set_subclass(&ve->base.timeline, TIMELINE_VIRTUAL);
+
+ intel_engine_init_execlists(&ve->base);
+
+ ve->base.cops = &virtual_context_ops;
+ ve->base.request_alloc = execlists_request_alloc;
+
+ ve->base.schedule = i915_schedule;
+ ve->base.submit_request = virtual_submit_request;
+
+ ve->base.execlists.queue_priority_hint = INT_MIN;
+ tasklet_init(&ve->base.execlists.tasklet,
+ virtual_submission_tasklet,
+ (unsigned long)ve);
+
+ intel_context_init(&ve->context, ctx, &ve->base);
+
+ for (n = 0; n < count; n++) {
+ struct intel_engine_cs *sibling = siblings[n];
+
+ GEM_BUG_ON(!is_power_of_2(sibling->mask));
+ if (sibling->mask & ve->base.mask) {
+ DRM_DEBUG("duplicate %s entry in load balancer\n",
+ sibling->name);
+ err = -EINVAL;
+ goto err_put;
+ }
+
+ /*
+ * The virtual engine implementation is tightly coupled to
+ * the execlists backend -- we push out request directly
+ * into a tree inside each physical engine. We could support
+ * layering if we handle cloning of the requests and
+ * submitting a copy into each backend.
+ */
+ if (sibling->execlists.tasklet.func !=
+ execlists_submission_tasklet) {
+ err = -ENODEV;
+ goto err_put;
+ }
+
+ GEM_BUG_ON(RB_EMPTY_NODE(&ve->nodes[sibling->id].rb));
+ RB_CLEAR_NODE(&ve->nodes[sibling->id].rb);
+
+ ve->siblings[ve->num_siblings++] = sibling;
+ ve->base.mask |= sibling->mask;
+
+ /*
+ * All physical engines must be compatible for their emission
+ * functions (as we build the instructions during request
+ * construction and do not alter them before submission
+ * on the physical engine). We use the engine class as a guide
+ * here, although that could be refined.
+ */
+ if (ve->base.class != OTHER_CLASS) {
+ if (ve->base.class != sibling->class) {
+ DRM_DEBUG("invalid mixing of engine class, sibling %d, already %d\n",
+ sibling->class, ve->base.class);
+ err = -EINVAL;
+ goto err_put;
+ }
+ continue;
+ }
+
+ ve->base.class = sibling->class;
+ ve->base.uabi_class = sibling->uabi_class;
+ snprintf(ve->base.name, sizeof(ve->base.name),
+ "v%dx%d", ve->base.class, count);
+ ve->base.context_size = sibling->context_size;
+
+ ve->base.emit_bb_start = sibling->emit_bb_start;
+ ve->base.emit_flush = sibling->emit_flush;
+ ve->base.emit_init_breadcrumb = sibling->emit_init_breadcrumb;
+ ve->base.emit_fini_breadcrumb = sibling->emit_fini_breadcrumb;
+ ve->base.emit_fini_breadcrumb_dw =
+ sibling->emit_fini_breadcrumb_dw;
+ }
+
+ return &ve->context;
+
+err_put:
+ intel_context_put(&ve->context);
+ return ERR_PTR(err);
+}
+
+struct intel_context *
+intel_execlists_clone_virtual(struct i915_gem_context *ctx,
+ struct intel_engine_cs *src)
+{
+ struct virtual_engine *se = to_virtual_engine(src);
+ struct intel_context *dst;
+
+ dst = intel_execlists_create_virtual(ctx,
+ se->siblings,
+ se->num_siblings);
+ if (IS_ERR(dst))
+ return dst;
+
+ return dst;
+}
+
void intel_execlists_show_requests(struct intel_engine_cs *engine,
struct drm_printer *m,
void (*show_request)(struct drm_printer *m,
@@ -2810,6 +3428,29 @@ void intel_execlists_show_requests(struct intel_engine_cs *engine,
show_request(m, last, "\t\tQ ");
}
+ last = NULL;
+ count = 0;
+ for (rb = rb_first_cached(&execlists->virtual); rb; rb = rb_next(rb)) {
+ struct virtual_engine *ve =
+ rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
+ struct i915_request *rq = READ_ONCE(ve->request);
+
+ if (rq) {
+ if (count++ < max - 1)
+ show_request(m, rq, "\t\tV ");
+ else
+ last = rq;
+ }
+ }
+ if (last) {
+ if (count > max) {
+ drm_printf(m,
+ "\t\t...skipping %d virtual requests...\n",
+ count - max);
+ }
+ show_request(m, last, "\t\tV ");
+ }
+
spin_unlock_irqrestore(&engine->timeline.lock, flags);
}
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.h b/drivers/gpu/drm/i915/gt/intel_lrc.h
index a0dc907a7249..5530606052e5 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.h
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.h
@@ -114,4 +114,13 @@ void intel_execlists_show_requests(struct intel_engine_cs *engine,
const char *prefix),
unsigned int max);
+struct intel_context *
+intel_execlists_create_virtual(struct i915_gem_context *ctx,
+ struct intel_engine_cs **siblings,
+ unsigned int count);
+
+struct intel_context *
+intel_execlists_clone_virtual(struct i915_gem_context *ctx,
+ struct intel_engine_cs *src);
+
#endif /* _INTEL_LRC_H_ */
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index 5b3d8e33f1cf..ccc0b6350123 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -1310,6 +1310,185 @@ static int live_preempt_smoke(void *arg)
return err;
}
+static int nop_virtual_engine(struct drm_i915_private *i915,
+ struct intel_engine_cs **siblings,
+ unsigned int nsibling,
+ unsigned int nctx,
+ unsigned int flags)
+#define CHAIN BIT(0)
+{
+ IGT_TIMEOUT(end_time);
+ struct i915_request *request[16];
+ struct i915_gem_context *ctx[16];
+ struct intel_context *ve[16];
+ unsigned long n, prime, nc;
+ struct igt_live_test t;
+ ktime_t times[2] = {};
+ int err;
+
+ GEM_BUG_ON(!nctx || nctx > ARRAY_SIZE(ctx));
+
+ for (n = 0; n < nctx; n++) {
+ ctx[n] = kernel_context(i915);
+ if (!ctx[n]) {
+ err = -ENOMEM;
+ nctx = n;
+ goto out;
+ }
+
+ ve[n] = intel_execlists_create_virtual(ctx[n],
+ siblings, nsibling);
+ if (IS_ERR(ve[n])) {
+ kernel_context_close(ctx[n]);
+ err = PTR_ERR(ve[n]);
+ nctx = n;
+ goto out;
+ }
+
+ err = intel_context_pin(ve[n]);
+ if (err) {
+ intel_context_put(ve[n]);
+ kernel_context_close(ctx[n]);
+ nctx = n;
+ goto out;
+ }
+ }
+
+ err = igt_live_test_begin(&t, i915, __func__, ve[0]->engine->name);
+ if (err)
+ goto out;
+
+ for_each_prime_number_from(prime, 1, 8192) {
+ times[1] = ktime_get_raw();
+
+ if (flags & CHAIN) {
+ for (nc = 0; nc < nctx; nc++) {
+ for (n = 0; n < prime; n++) {
+ request[nc] =
+ i915_request_create(ve[nc]);
+ if (IS_ERR(request[nc])) {
+ err = PTR_ERR(request[nc]);
+ goto out;
+ }
+
+ i915_request_add(request[nc]);
+ }
+ }
+ } else {
+ for (n = 0; n < prime; n++) {
+ for (nc = 0; nc < nctx; nc++) {
+ request[nc] =
+ i915_request_create(ve[nc]);
+ if (IS_ERR(request[nc])) {
+ err = PTR_ERR(request[nc]);
+ goto out;
+ }
+
+ i915_request_add(request[nc]);
+ }
+ }
+ }
+
+ for (nc = 0; nc < nctx; nc++) {
+ if (i915_request_wait(request[nc],
+ I915_WAIT_LOCKED,
+ HZ / 10) < 0) {
+ pr_err("%s(%s): wait for %llx:%lld timed out\n",
+ __func__, ve[0]->engine->name,
+ request[nc]->fence.context,
+ request[nc]->fence.seqno);
+
+ GEM_TRACE("%s(%s) failed at request %llx:%lld\n",
+ __func__, ve[0]->engine->name,
+ request[nc]->fence.context,
+ request[nc]->fence.seqno);
+ GEM_TRACE_DUMP();
+ i915_gem_set_wedged(i915);
+ break;
+ }
+ }
+
+ times[1] = ktime_sub(ktime_get_raw(), times[1]);
+ if (prime == 1)
+ times[0] = times[1];
+
+ if (__igt_timeout(end_time, NULL))
+ break;
+ }
+
+ err = igt_live_test_end(&t);
+ if (err)
+ goto out;
+
+ pr_info("Requestx%d latencies on %s: 1 = %lluns, %lu = %lluns\n",
+ nctx, ve[0]->engine->name, ktime_to_ns(times[0]),
+ prime, div64_u64(ktime_to_ns(times[1]), prime));
+
+out:
+ if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ err = -EIO;
+
+ for (nc = 0; nc < nctx; nc++) {
+ intel_context_unpin(ve[nc]);
+ intel_context_put(ve[nc]);
+ kernel_context_close(ctx[nc]);
+ }
+ return err;
+}
+
+static int live_virtual_engine(void *arg)
+{
+ struct drm_i915_private *i915 = arg;
+ struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1];
+ struct intel_engine_cs *engine;
+ enum intel_engine_id id;
+ unsigned int class, inst;
+ int err = -ENODEV;
+
+ if (USES_GUC_SUBMISSION(i915))
+ return 0;
+
+ mutex_lock(&i915->drm.struct_mutex);
+
+ for_each_engine(engine, i915, id) {
+ err = nop_virtual_engine(i915, &engine, 1, 1, 0);
+ if (err) {
+ pr_err("Failed to wrap engine %s: err=%d\n",
+ engine->name, err);
+ goto out_unlock;
+ }
+ }
+
+ for (class = 0; class <= MAX_ENGINE_CLASS; class++) {
+ int nsibling, n;
+
+ nsibling = 0;
+ for (inst = 0; inst <= MAX_ENGINE_INSTANCE; inst++) {
+ if (!i915->engine_class[class][inst])
+ break;
+
+ siblings[nsibling++] = i915->engine_class[class][inst];
+ }
+ if (nsibling < 2)
+ continue;
+
+ for (n = 1; n <= nsibling + 1; n++) {
+ err = nop_virtual_engine(i915, siblings, nsibling,
+ n, 0);
+ if (err)
+ goto out_unlock;
+ }
+
+ err = nop_virtual_engine(i915, siblings, nsibling, n, CHAIN);
+ if (err)
+ goto out_unlock;
+ }
+
+out_unlock:
+ mutex_unlock(&i915->drm.struct_mutex);
+ return err;
+}
+
int intel_execlists_live_selftests(struct drm_i915_private *i915)
{
static const struct i915_subtest tests[] = {
@@ -1322,6 +1501,7 @@ int intel_execlists_live_selftests(struct drm_i915_private *i915)
SUBTEST(live_chain_preempt),
SUBTEST(live_preempt_hang),
SUBTEST(live_preempt_smoke),
+ SUBTEST(live_virtual_engine),
};
if (!HAS_EXECLISTS(i915))
diff --git a/drivers/gpu/drm/i915/i915_gem.h b/drivers/gpu/drm/i915/i915_gem.h
index 67f8a4a807a0..fe82d3571072 100644
--- a/drivers/gpu/drm/i915/i915_gem.h
+++ b/drivers/gpu/drm/i915/i915_gem.h
@@ -91,4 +91,9 @@ static inline bool __tasklet_enable(struct tasklet_struct *t)
return atomic_dec_and_test(&t->count);
}
+static inline bool __tasklet_is_scheduled(struct tasklet_struct *t)
+{
+ return test_bit(TASKLET_STATE_SCHED, &t->state);
+}
+
#endif /* __I915_GEM_H__ */
diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
index 040c0b770c21..3ae1d27130a8 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/i915_gem_context.c
@@ -86,6 +86,7 @@
*/
#include <linux/log2.h>
+#include <linux/nospec.h>
#include <drm/i915_drm.h>
@@ -1218,7 +1219,6 @@ __intel_context_reconfigure_sseu(struct intel_context *ce,
int ret;
GEM_BUG_ON(INTEL_GEN(ce->gem_context->i915) < 8);
- GEM_BUG_ON(ce->engine->id != RCS0);
ret = intel_context_lock_pinned(ce);
if (ret)
@@ -1412,7 +1412,100 @@ struct set_engines {
struct i915_gem_engines *engines;
};
+static int
+set_engines__load_balance(struct i915_user_extension __user *base, void *data)
+{
+ struct i915_context_engines_load_balance __user *ext =
+ container_of_user(base, typeof(*ext), base);
+ const struct set_engines *set = data;
+ struct intel_engine_cs *stack[16];
+ struct intel_engine_cs **siblings;
+ struct intel_context *ce;
+ u16 num_siblings, idx;
+ unsigned int n;
+ int err;
+
+ if (!HAS_EXECLISTS(set->ctx->i915))
+ return -ENODEV;
+
+ if (USES_GUC_SUBMISSION(set->ctx->i915))
+ return -ENODEV; /* not implement yet */
+
+ if (get_user(idx, &ext->engine_index))
+ return -EFAULT;
+
+ if (idx >= set->engines->num_engines) {
+ DRM_DEBUG("Invalid placement value, %d >= %d\n",
+ idx, set->engines->num_engines);
+ return -EINVAL;
+ }
+
+ idx = array_index_nospec(idx, set->engines->num_engines);
+ if (set->engines->engines[idx]) {
+ DRM_DEBUG("Invalid placement[%d], already occupied\n", idx);
+ return -EEXIST;
+ }
+
+ if (get_user(num_siblings, &ext->num_siblings))
+ return -EFAULT;
+
+ err = check_user_mbz(&ext->flags);
+ if (err)
+ return err;
+
+ err = check_user_mbz(&ext->mbz64);
+ if (err)
+ return err;
+
+ siblings = stack;
+ if (num_siblings > ARRAY_SIZE(stack)) {
+ siblings = kmalloc_array(num_siblings,
+ sizeof(*siblings),
+ GFP_KERNEL);
+ if (!siblings)
+ return -ENOMEM;
+ }
+
+ for (n = 0; n < num_siblings; n++) {
+ struct i915_engine_class_instance ci;
+
+ if (copy_from_user(&ci, &ext->engines[n], sizeof(ci))) {
+ err = -EFAULT;
+ goto out_siblings;
+ }
+
+ siblings[n] = intel_engine_lookup_user(set->ctx->i915,
+ ci.engine_class,
+ ci.engine_instance);
+ if (!siblings[n]) {
+ DRM_DEBUG("Invalid sibling[%d]: { class:%d, inst:%d }\n",
+ n, ci.engine_class, ci.engine_instance);
+ err = -EINVAL;
+ goto out_siblings;
+ }
+ }
+
+ ce = intel_execlists_create_virtual(set->ctx, siblings, n);
+ if (IS_ERR(ce)) {
+ err = PTR_ERR(ce);
+ goto out_siblings;
+ }
+
+ if (cmpxchg(&set->engines->engines[idx], NULL, ce)) {
+ intel_context_put(ce);
+ err = -EEXIST;
+ goto out_siblings;
+ }
+
+out_siblings:
+ if (siblings != stack)
+ kfree(siblings);
+
+ return err;
+}
+
static const i915_user_extension_fn set_engines__extensions[] = {
+ [I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE] = set_engines__load_balance,
};
static int
@@ -1711,14 +1804,29 @@ static int clone_engines(struct i915_gem_context *dst,
clone->i915 = dst->i915;
for (n = 0; n < e->num_engines; n++) {
+ struct intel_engine_cs *engine;
+
if (!e->engines[n]) {
clone->engines[n] = NULL;
continue;
}
+ engine = e->engines[n]->engine;
- clone->engines[n] =
- intel_context_create(dst, e->engines[n]->engine);
- if (!clone->engines[n]) {
+ /*
+ * Virtual engines are singletons; they can only exist
+ * inside a single context, because they embed their
+ * HW context... As each virtual context implies a single
+ * timeline (each engine can only dequeue a single request
+ * at any time), it would be surprising for two contexts
+ * to use the same engine. So let's create a copy of
+ * the virtual engine instead.
+ */
+ if (intel_engine_is_virtual(engine))
+ clone->engines[n] =
+ intel_execlists_clone_virtual(dst, engine);
+ else
+ clone->engines[n] = intel_context_create(dst, engine);
+ if (IS_ERR_OR_NULL(clone->engines[n])) {
__free_engines(clone, n);
goto err_unlock;
}
diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c
index d215dcdf0b1e..78ceb56d7801 100644
--- a/drivers/gpu/drm/i915/i915_scheduler.c
+++ b/drivers/gpu/drm/i915/i915_scheduler.c
@@ -150,17 +150,26 @@ sched_lock_engine(const struct i915_sched_node *node,
struct intel_engine_cs *locked,
struct sched_cache *cache)
{
- struct intel_engine_cs *engine = node_to_request(node)->engine;
+ const struct i915_request *rq = node_to_request(node);
+ struct intel_engine_cs *engine;
GEM_BUG_ON(!locked);
- if (engine != locked) {
+ /*
+ * Virtual engines complicate acquiring the engine timeline lock,
+ * as their rq->engine pointer is not stable until under that
+ * engine lock. The simple ploy we use is to take the lock then
+ * check that the rq still belongs to the newly locked engine.
+ */
+ while (locked != (engine = READ_ONCE(rq->engine))) {
spin_unlock(&locked->timeline.lock);
memset(cache, 0, sizeof(*cache));
spin_lock(&engine->timeline.lock);
+ locked = engine;
}
- return engine;
+ GEM_BUG_ON(locked != engine);
+ return locked;
}
static inline int rq_prio(const struct i915_request *rq)
@@ -272,6 +281,7 @@ static void __i915_schedule(struct i915_sched_node *node,
spin_lock(&engine->timeline.lock);
/* Fifo and depth-first replacement ensure our deps execute before us */
+ engine = sched_lock_engine(node, engine, &cache);
list_for_each_entry_safe_reverse(dep, p, &dfs, dfs_link) {
INIT_LIST_HEAD(&dep->dfs_link);
@@ -283,8 +293,11 @@ static void __i915_schedule(struct i915_sched_node *node,
if (prio <= node->attr.priority || node_signaled(node))
continue;
+ GEM_BUG_ON(node_to_request(node)->engine != engine);
+
node->attr.priority = prio;
if (!list_empty(&node->link)) {
+ GEM_BUG_ON(intel_engine_is_virtual(engine));
if (!cache.priolist)
cache.priolist =
i915_sched_lookup_priolist(engine,
diff --git a/drivers/gpu/drm/i915/i915_timeline_types.h b/drivers/gpu/drm/i915/i915_timeline_types.h
index 5256a0b5c5f7..1688705f4a2b 100644
--- a/drivers/gpu/drm/i915/i915_timeline_types.h
+++ b/drivers/gpu/drm/i915/i915_timeline_types.h
@@ -26,6 +26,7 @@ struct i915_timeline {
spinlock_t lock;
#define TIMELINE_CLIENT 0 /* default subclass */
#define TIMELINE_ENGINE 1
+#define TIMELINE_VIRTUAL 2
struct mutex mutex; /* protects the flow of requests */
unsigned int pin_count;
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index 62396d575e28..f9770948161c 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -137,6 +137,7 @@ struct i915_engine_class_instance {
__u16 engine_class; /* see enum drm_i915_gem_engine_class */
__u16 engine_instance;
#define I915_ENGINE_CLASS_INVALID_NONE -1
+#define I915_ENGINE_CLASS_INVALID_VIRTUAL -2
};
/**
@@ -1608,8 +1609,46 @@ struct drm_i915_gem_context_param_sseu {
__u32 rsvd;
};
+/*
+ * i915_context_engines_load_balance:
+ *
+ * Enable load balancing across this set of engines.
+ *
+ * Into the I915_EXEC_DEFAULT slot [0], a virtual engine is created that when
+ * used will proxy the execbuffer request onto one of the set of engines
+ * in such a way as to distribute the load evenly across the set.
+ *
+ * The set of engines must be compatible (e.g. the same HW class) as they
+ * will share the same logical GPU context and ring.
+ *
+ * To intermix rendering with the virtual engine and direct rendering onto
+ * the backing engines (bypassing the load balancing proxy), the context must
+ * be defined to use a single timeline for all engines.
+ */
+struct i915_context_engines_load_balance {
+ struct i915_user_extension base;
+
+ __u16 engine_index;
+ __u16 num_siblings;
+ __u32 flags; /* all undefined flags must be zero */
+
+ __u64 mbz64; /* reserved for future use; must be zero */
+
+ struct i915_engine_class_instance engines[0];
+} __attribute__((packed));
+
+#define I915_DEFINE_CONTEXT_ENGINES_LOAD_BALANCE(name__, N__) struct { \
+ struct i915_user_extension base; \
+ __u16 engine_index; \
+ __u16 num_siblings; \
+ __u32 flags; \
+ __u64 mbz64; \
+ struct i915_engine_class_instance engines[N__]; \
+} __attribute__((packed)) name__
+
struct i915_context_param_engines {
__u64 extensions; /* linked chain of extension blocks, 0 terminates */
+#define I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE 0 /* see i915_context_engines_load_balance */
struct i915_engine_class_instance engines[0];
} __attribute__((packed));
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* Re: [PATCH 14/40] drm/i915: Load balancing across a virtual engine
2019-05-08 8:06 ` [PATCH 14/40] drm/i915: Load balancing across a virtual engine Chris Wilson
@ 2019-05-08 10:29 ` Tvrtko Ursulin
2019-05-08 11:17 ` Chris Wilson
2019-05-08 11:23 ` [PATCH] " Chris Wilson
1 sibling, 1 reply; 71+ messages in thread
From: Tvrtko Ursulin @ 2019-05-08 10:29 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 08/05/2019 09:06, Chris Wilson wrote:
> Having allowed the user to define a set of engines that they will want
> to only use, we go one step further and allow them to bind those engines
> into a single virtual instance. Submitting a batch to the virtual engine
> will then forward it to any one of the set in a manner as best to
> distribute load. The virtual engine has a single timeline across all
> engines (it operates as a single queue), so it is not able to concurrently
> run batches across multiple engines by itself; that is left up to the user
> to submit multiple concurrent batches to multiple queues. Multiple users
> will be load balanced across the system.
>
> The mechanism used for load balancing in this patch is a late greedy
> balancer. When a request is ready for execution, it is added to each
> engine's queue, and when an engine is ready for its next request it
> claims it from the virtual engine. The first engine to do so, wins, i.e.
> the request is executed at the earliest opportunity (idle moment) in the
> system.
>
> As not all HW is created equal, the user is still able to skip the
> virtual engine and execute the batch on a specific engine, all within the
> same queue. It will then be executed in order on the correct engine,
> with execution on other virtual engines being moved away due to the load
> detection.
>
> A couple of areas for potential improvement left!
>
> - The virtual engine always take priority over equal-priority tasks.
> Mostly broken up by applying FQ_CODEL rules for prioritising new clients,
> and hopefully the virtual and real engines are not then congested (i.e.
> all work is via virtual engines, or all work is to the real engine).
>
> - We require the breadcrumb irq around every virtual engine request. For
> normal engines, we eliminate the need for the slow round trip via
> interrupt by using the submit fence and queueing in order. For virtual
> engines, we have to allow any job to transfer to a new ring, and cannot
> coalesce the submissions, so require the completion fence instead,
> forcing the persistent use of interrupts.
>
> - We only drip feed single requests through each virtual engine and onto
> the physical engines, even if there was enough work to fill all ELSP,
> leaving small stalls with an idle CS event at the end of every request.
> Could we be greedy and fill both slots? Being lazy is virtuous for load
> distribution on less-than-full workloads though.
>
> Other areas of improvement are more general, such as reducing lock
> contention, reducing dispatch overhead, looking at direct submission
> rather than bouncing around tasklets etc.
>
> sseu: Lift the restriction to allow sseu to be reconfigured on virtual
> engines composed of RENDER_CLASS (rcs).
>
> v2: macroize check_user_mbz()
> v3: Cancel virtual engines on wedging
> v4: Commence commenting
> v5: Replace 64b sibling_mask with a list of class:instance
> v6: Drop the one-element array in the uabi
> v7: Assert it is an virtual engine in to_virtual_engine()
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> ---
> drivers/gpu/drm/i915/gt/intel_engine_types.h | 8 +
> drivers/gpu/drm/i915/gt/intel_lrc.c | 683 ++++++++++++++++++-
> drivers/gpu/drm/i915/gt/intel_lrc.h | 9 +
> drivers/gpu/drm/i915/gt/selftest_lrc.c | 180 +++++
> drivers/gpu/drm/i915/i915_gem.h | 5 +
> drivers/gpu/drm/i915/i915_gem_context.c | 116 +++-
> drivers/gpu/drm/i915/i915_scheduler.c | 19 +-
> drivers/gpu/drm/i915/i915_timeline_types.h | 1 +
> include/uapi/drm/i915_drm.h | 39 ++
> 9 files changed, 1032 insertions(+), 28 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
> index e381c1c73902..7b47e00fa082 100644
> --- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
> +++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
> @@ -227,6 +227,7 @@ struct intel_engine_execlists {
> * @queue: queue of requests, in priority lists
> */
> struct rb_root_cached queue;
> + struct rb_root_cached virtual;
>
> /**
> * @csb_write: control register for Context Switch buffer
> @@ -445,6 +446,7 @@ struct intel_engine_cs {
> #define I915_ENGINE_HAS_PREEMPTION BIT(2)
> #define I915_ENGINE_HAS_SEMAPHORES BIT(3)
> #define I915_ENGINE_NEEDS_BREADCRUMB_TASKLET BIT(4)
> +#define I915_ENGINE_IS_VIRTUAL BIT(5)
> unsigned int flags;
>
> /*
> @@ -534,6 +536,12 @@ intel_engine_needs_breadcrumb_tasklet(const struct intel_engine_cs *engine)
> return engine->flags & I915_ENGINE_NEEDS_BREADCRUMB_TASKLET;
> }
>
> +static inline bool
> +intel_engine_is_virtual(const struct intel_engine_cs *engine)
> +{
> + return engine->flags & I915_ENGINE_IS_VIRTUAL;
> +}
> +
> #define instdone_slice_mask(dev_priv__) \
> (IS_GEN(dev_priv__, 7) ? \
> 1 : RUNTIME_INFO(dev_priv__)->sseu.slice_mask)
> diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
> index f1d62746e066..bc388df39802 100644
> --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
> +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
> @@ -166,6 +166,42 @@
>
> #define ACTIVE_PRIORITY (I915_PRIORITY_NEWCLIENT | I915_PRIORITY_NOSEMAPHORE)
>
> +struct virtual_engine {
> + struct intel_engine_cs base;
> + struct intel_context context;
> +
> + /*
> + * We allow only a single request through the virtual engine at a time
> + * (each request in the timeline waits for the completion fence of
> + * the previous before being submitted). By restricting ourselves to
> + * only submitting a single request, each request is placed on to a
> + * physical to maximise load spreading (by virtue of the late greedy
> + * scheduling -- each real engine takes the next available request
> + * upon idling).
> + */
> + struct i915_request *request;
> +
> + /*
> + * We keep a rbtree of available virtual engines inside each physical
> + * engine, sorted by priority. Here we preallocate the nodes we need
> + * for the virtual engine, indexed by physical_engine->id.
> + */
> + struct ve_node {
> + struct rb_node rb;
> + int prio;
> + } nodes[I915_NUM_ENGINES];
> +
> + /* And finally, which physical engines this virtual engine maps onto. */
> + unsigned int num_siblings;
> + struct intel_engine_cs *siblings[0];
> +};
> +
> +static struct virtual_engine *to_virtual_engine(struct intel_engine_cs *engine)
> +{
> + GEM_BUG_ON(!intel_engine_is_virtual(engine));
> + return container_of(engine, struct virtual_engine, base);
> +}
> +
> static int execlists_context_deferred_alloc(struct intel_context *ce,
> struct intel_engine_cs *engine);
> static void execlists_init_reg_state(u32 *reg_state,
> @@ -229,7 +265,8 @@ static int queue_prio(const struct intel_engine_execlists *execlists)
> }
>
> static inline bool need_preempt(const struct intel_engine_cs *engine,
> - const struct i915_request *rq)
> + const struct i915_request *rq,
> + struct rb_node *rb)
> {
> int last_prio;
>
> @@ -264,6 +301,25 @@ static inline bool need_preempt(const struct intel_engine_cs *engine,
> rq_prio(list_next_entry(rq, link)) > last_prio)
> return true;
>
> + if (rb) {
> + struct virtual_engine *ve =
> + rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
> + bool preempt = false;
> +
> + if (engine == ve->siblings[0]) { /* only preempt one sibling */
> + struct i915_request *next;
> +
> + rcu_read_lock();
> + next = READ_ONCE(ve->request);
> + if (next)
> + preempt = rq_prio(next) > last_prio;
> + rcu_read_unlock();
> + }
> +
> + if (preempt)
> + return preempt;
> + }
> +
> /*
> * If the inflight context did not trigger the preemption, then maybe
> * it was the set of queued requests? Pick the highest priority in
> @@ -382,6 +438,8 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
> list_for_each_entry_safe_reverse(rq, rn,
> &engine->timeline.requests,
> link) {
> + struct intel_engine_cs *owner;
> +
> if (i915_request_completed(rq))
> break;
>
> @@ -390,16 +448,32 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
>
> GEM_BUG_ON(rq->hw_context->active);
>
> - GEM_BUG_ON(rq_prio(rq) == I915_PRIORITY_INVALID);
> - if (rq_prio(rq) != prio) {
> - prio = rq_prio(rq);
> - pl = i915_sched_lookup_priolist(engine, prio);
> - }
> - GEM_BUG_ON(RB_EMPTY_ROOT(&engine->execlists.queue.rb_root));
> + /*
> + * Push the request back into the queue for later resubmission.
> + * If this request is not native to this physical engine (i.e.
> + * it came from a virtual source), push it back onto the virtual
> + * engine so that it can be moved across onto another physical
> + * engine as load dictates.
> + */
> + owner = rq->hw_context->engine;
> + if (likely(owner == engine)) {
> + GEM_BUG_ON(rq_prio(rq) == I915_PRIORITY_INVALID);
> + if (rq_prio(rq) != prio) {
> + prio = rq_prio(rq);
> + pl = i915_sched_lookup_priolist(engine, prio);
> + }
> + GEM_BUG_ON(RB_EMPTY_ROOT(&engine->execlists.queue.rb_root));
>
> - list_add(&rq->sched.link, pl);
> + list_add(&rq->sched.link, pl);
> + active = rq;
> + } else {
> + if (__i915_request_has_started(rq))
> + rq->sched.attr.priority |= boost;
>
> - active = rq;
> + rq->engine = owner;
> + owner->submit_request(rq);
> + active = NULL;
> + }
> }
>
> /*
> @@ -419,7 +493,7 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
> * in the priority queue, but they will not gain immediate access to
> * the GPU.
> */
> - if (~prio & boost && __i915_request_has_started(active)) {
> + if (~prio & boost && active && __i915_request_has_started(active)) {
> prio |= boost;
> GEM_BUG_ON(active->sched.attr.priority >= prio);
> active->sched.attr.priority = prio;
> @@ -661,6 +735,90 @@ static void complete_preempt_context(struct intel_engine_execlists *execlists)
> ACTIVE_PRIORITY);
> }
>
> +static void virtual_update_register_offsets(u32 *regs,
> + struct intel_engine_cs *engine)
> +{
> + u32 base = engine->mmio_base;
> +
> + /* Must match execlists_init_reg_state()! */
> +
> + regs[CTX_CONTEXT_CONTROL] =
> + i915_mmio_reg_offset(RING_CONTEXT_CONTROL(base));
> + regs[CTX_RING_HEAD] = i915_mmio_reg_offset(RING_HEAD(base));
> + regs[CTX_RING_TAIL] = i915_mmio_reg_offset(RING_TAIL(base));
> + regs[CTX_RING_BUFFER_START] = i915_mmio_reg_offset(RING_START(base));
> + regs[CTX_RING_BUFFER_CONTROL] = i915_mmio_reg_offset(RING_CTL(base));
> +
> + regs[CTX_BB_HEAD_U] = i915_mmio_reg_offset(RING_BBADDR_UDW(base));
> + regs[CTX_BB_HEAD_L] = i915_mmio_reg_offset(RING_BBADDR(base));
> + regs[CTX_BB_STATE] = i915_mmio_reg_offset(RING_BBSTATE(base));
> + regs[CTX_SECOND_BB_HEAD_U] =
> + i915_mmio_reg_offset(RING_SBBADDR_UDW(base));
> + regs[CTX_SECOND_BB_HEAD_L] = i915_mmio_reg_offset(RING_SBBADDR(base));
> + regs[CTX_SECOND_BB_STATE] = i915_mmio_reg_offset(RING_SBBSTATE(base));
> +
> + regs[CTX_CTX_TIMESTAMP] =
> + i915_mmio_reg_offset(RING_CTX_TIMESTAMP(base));
> + regs[CTX_PDP3_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 3));
> + regs[CTX_PDP3_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 3));
> + regs[CTX_PDP2_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 2));
> + regs[CTX_PDP2_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 2));
> + regs[CTX_PDP1_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 1));
> + regs[CTX_PDP1_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 1));
> + regs[CTX_PDP0_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 0));
> + regs[CTX_PDP0_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 0));
> +
> + if (engine->class == RENDER_CLASS) {
> + regs[CTX_RCS_INDIRECT_CTX] =
> + i915_mmio_reg_offset(RING_INDIRECT_CTX(base));
> + regs[CTX_RCS_INDIRECT_CTX_OFFSET] =
> + i915_mmio_reg_offset(RING_INDIRECT_CTX_OFFSET(base));
> + regs[CTX_BB_PER_CTX_PTR] =
> + i915_mmio_reg_offset(RING_BB_PER_CTX_PTR(base));
> +
> + regs[CTX_R_PWR_CLK_STATE] =
> + i915_mmio_reg_offset(GEN8_R_PWR_CLK_STATE);
> + }
> +}
> +
> +static bool virtual_matches(const struct virtual_engine *ve,
> + const struct i915_request *rq,
> + const struct intel_engine_cs *engine)
> +{
> + const struct intel_engine_cs *active;
> +
> + /*
> + * We track when the HW has completed saving the context image
> + * (i.e. when we have seen the final CS event switching out of
> + * the context) and must not overwrite the context image before
> + * then. This restricts us to only using the active engine
> + * while the previous virtualized request is inflight (so
> + * we reuse the register offsets). This is a very small
> + * hystersis on the greedy seelction algorithm.
> + */
> + active = READ_ONCE(ve->context.active);
> + if (active && active != engine)
> + return false;
> +
> + return true;
> +}
> +
> +static void virtual_xfer_breadcrumbs(struct virtual_engine *ve,
> + struct intel_engine_cs *engine)
> +{
> + struct intel_engine_cs *old = ve->siblings[0];
> +
> + /* All unattached (rq->engine == old) must already be completed */
> +
> + spin_lock(&old->breadcrumbs.irq_lock);
> + if (!list_empty(&ve->context.signal_link)) {
> + list_move_tail(&ve->context.signal_link,
> + &engine->breadcrumbs.signalers);
> + intel_engine_queue_breadcrumbs(engine);
> + }
> + spin_unlock(&old->breadcrumbs.irq_lock);
> +}
> +
> static void execlists_dequeue(struct intel_engine_cs *engine)
> {
> struct intel_engine_execlists * const execlists = &engine->execlists;
> @@ -693,6 +851,26 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
> * and context switches) submission.
> */
>
> + for (rb = rb_first_cached(&execlists->virtual); rb; ) {
> + struct virtual_engine *ve =
> + rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
> + struct i915_request *rq = READ_ONCE(ve->request);
> +
> + if (!rq) { /* lazily cleanup after another engine handled rq */
> + rb_erase_cached(rb, &execlists->virtual);
> + RB_CLEAR_NODE(rb);
> + rb = rb_first_cached(&execlists->virtual);
> + continue;
> + }
> +
> + if (!virtual_matches(ve, rq, engine)) {
> + rb = rb_next(rb);
> + continue;
> + }
> +
> + break;
> + }
> +
> if (last) {
> /*
> * Don't resubmit or switch until all outstanding
> @@ -714,7 +892,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
> if (!execlists_is_active(execlists, EXECLISTS_ACTIVE_HWACK))
> return;
>
> - if (need_preempt(engine, last)) {
> + if (need_preempt(engine, last, rb)) {
> inject_preempt_context(engine);
> return;
> }
> @@ -754,6 +932,92 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
> last->tail = last->wa_tail;
> }
>
> + while (rb) { /* XXX virtual is always taking precedence */
> + struct virtual_engine *ve =
> + rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
> + struct i915_request *rq;
> +
> + spin_lock(&ve->base.timeline.lock);
> +
> + rq = ve->request;
> + if (unlikely(!rq)) { /* lost the race to a sibling */
> + spin_unlock(&ve->base.timeline.lock);
> + rb_erase_cached(rb, &execlists->virtual);
> + RB_CLEAR_NODE(rb);
> + rb = rb_first_cached(&execlists->virtual);
> + continue;
> + }
> +
> + GEM_BUG_ON(rq != ve->request);
> + GEM_BUG_ON(rq->engine != &ve->base);
> + GEM_BUG_ON(rq->hw_context != &ve->context);
> +
> + if (rq_prio(rq) >= queue_prio(execlists)) {
> + if (!virtual_matches(ve, rq, engine)) {
> + spin_unlock(&ve->base.timeline.lock);
> + rb = rb_next(rb);
> + continue;
> + }
> +
> + if (last && !can_merge_rq(last, rq)) {
> + spin_unlock(&ve->base.timeline.lock);
> + return; /* leave this rq for another engine */
> + }
> +
> + GEM_TRACE("%s: virtual rq=%llx:%lld%s, new engine? %s\n",
> + engine->name,
> + rq->fence.context,
> + rq->fence.seqno,
> + i915_request_completed(rq) ? "!" :
> + i915_request_started(rq) ? "*" :
> + "",
> + yesno(engine != ve->siblings[0]));
> +
> + ve->request = NULL;
> + ve->base.execlists.queue_priority_hint = INT_MIN;
> + rb_erase_cached(rb, &execlists->virtual);
> + RB_CLEAR_NODE(rb);
> +
> + rq->engine = engine;
> +
> + if (engine != ve->siblings[0]) {
> + u32 *regs = ve->context.lrc_reg_state;
> + unsigned int n;
> +
> + GEM_BUG_ON(READ_ONCE(ve->context.active));
> + virtual_update_register_offsets(regs, engine);
> +
> + if (!list_empty(&ve->context.signals))
> + virtual_xfer_breadcrumbs(ve, engine);
> +
> + /*
> + * Move the bound engine to the top of the list
> + * for future execution. We then kick this
> + * tasklet first before checking others, so that
> + * we preferentially reuse this set of bound
> + * registers.
> + */
> + for (n = 1; n < ve->num_siblings; n++) {
> + if (ve->siblings[n] == engine) {
> + swap(ve->siblings[n],
> + ve->siblings[0]);
> + break;
> + }
> + }
> +
> + GEM_BUG_ON(ve->siblings[0] != engine);
> + }
> +
> + __i915_request_submit(rq);
> + trace_i915_request_in(rq, port_index(port, execlists));
> + submit = true;
> + last = rq;
> + }
> +
> + spin_unlock(&ve->base.timeline.lock);
> + break;
> + }
> +
> while ((rb = rb_first_cached(&execlists->queue))) {
> struct i915_priolist *p = to_priolist(rb);
> struct i915_request *rq, *rn;
> @@ -1877,6 +2141,25 @@ static void reset_csb_pointers(struct intel_engine_execlists *execlists)
> &execlists->csb_status[reset_value]);
> }
>
> +static struct i915_request *active_request(struct i915_request *rq)
> +{
> + const struct list_head * const list = &rq->engine->timeline.requests;
> + const struct intel_context * const context = rq->hw_context;
> + struct i915_request *active = NULL;
> +
> + list_for_each_entry_from_reverse(rq, list, link) {
> + if (i915_request_completed(rq))
> + break;
> +
> + if (rq->hw_context != context)
> + break;
> +
> + active = rq;
> + }
> +
> + return active;
> +}
> +
> static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
> {
> struct intel_engine_execlists * const execlists = &engine->execlists;
> @@ -1897,7 +2180,8 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
> if (!port_isset(execlists->port))
> goto out_clear;
>
> - ce = port_request(execlists->port)->hw_context;
> + rq = port_request(execlists->port);
> + ce = rq->hw_context;
>
> /*
> * Catch up with any missed context-switch interrupts.
> @@ -1910,16 +2194,10 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
> */
> execlists_cancel_port_requests(execlists);
>
> - /* Push back any incomplete requests for replay after the reset. */
> - rq = __unwind_incomplete_requests(engine, 0);
> + rq = active_request(rq);
> if (!rq)
> goto out_replay;
>
> - if (rq->hw_context != ce) { /* caught just before a CS event */
> - rq = NULL;
> - goto out_replay;
> - }
> -
> /*
> * If this request hasn't started yet, e.g. it is waiting on a
> * semaphore, we need to avoid skipping the request or else we
> @@ -1966,13 +2244,16 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
> }
> execlists_init_reg_state(regs, ce, engine, ce->ring);
>
> - /* Rerun the request; its payload has been neutered (if guilty). */
> out_replay:
> + /* Rerun the request; its payload has been neutered (if guilty). */
> ce->ring->head =
> rq ? intel_ring_wrap(ce->ring, rq->head) : ce->ring->tail;
> intel_ring_update_space(ce->ring);
> __execlists_update_reg_state(ce, engine);
>
> + /* Push back any incomplete requests for replay after the reset. */
> + __unwind_incomplete_requests(engine, 0);
> +
> out_clear:
> execlists_clear_all_active(execlists);
> }
> @@ -2046,6 +2327,26 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine)
> i915_priolist_free(p);
> }
>
> + /* Cancel all attached virtual engines */
> + while ((rb = rb_first_cached(&execlists->virtual))) {
> + struct virtual_engine *ve =
> + rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
> +
> + rb_erase_cached(rb, &execlists->virtual);
> + RB_CLEAR_NODE(rb);
> +
> + spin_lock(&ve->base.timeline.lock);
> + if (ve->request) {
> + ve->request->engine = engine;
> + __i915_request_submit(ve->request);
> + dma_fence_set_error(&ve->request->fence, -EIO);
> + i915_request_mark_complete(ve->request);
> + ve->base.execlists.queue_priority_hint = INT_MIN;
> + ve->request = NULL;
> + }
> + spin_unlock(&ve->base.timeline.lock);
> + }
> +
> /* Remaining _unready_ requests will be nop'ed when submitted */
>
> execlists->queue_priority_hint = INT_MIN;
> @@ -2529,12 +2830,15 @@ static void execlists_init_reg_state(u32 *regs,
> bool rcs = engine->class == RENDER_CLASS;
> u32 base = engine->mmio_base;
>
> - /* A context is actually a big batch buffer with several
> + /*
> + * A context is actually a big batch buffer with several
> * MI_LOAD_REGISTER_IMM commands followed by (reg, value) pairs. The
> * values we are setting here are only for the first context restore:
> * on a subsequent save, the GPU will recreate this batchbuffer with new
> * values (including all the missing MI_LOAD_REGISTER_IMM commands that
> * we are not initializing here).
> + *
> + * Must keep consistent with virtual_update_register_offsets().
> */
> regs[CTX_LRI_HEADER_0] = MI_LOAD_REGISTER_IMM(rcs ? 14 : 11) |
> MI_LRI_FORCE_POSTED;
> @@ -2753,6 +3057,320 @@ static int execlists_context_deferred_alloc(struct intel_context *ce,
> return ret;
> }
>
> +static void virtual_context_destroy(struct kref *kref)
> +{
> + struct virtual_engine *ve =
> + container_of(kref, typeof(*ve), context.ref);
> + unsigned int n;
> +
> + GEM_BUG_ON(ve->request);
> + GEM_BUG_ON(ve->context.active);
> +
> + for (n = 0; n < ve->num_siblings; n++) {
> + struct intel_engine_cs *sibling = ve->siblings[n];
> + struct rb_node *node = &ve->nodes[sibling->id].rb;
> +
> + if (RB_EMPTY_NODE(node))
> + continue;
> +
> + spin_lock_irq(&sibling->timeline.lock);
> +
> + /* Detachment is lazily performed in the execlists tasklet */
> + if (!RB_EMPTY_NODE(node))
> + rb_erase_cached(node, &sibling->execlists.virtual);
> +
> + spin_unlock_irq(&sibling->timeline.lock);
> + }
> + GEM_BUG_ON(__tasklet_is_scheduled(&ve->base.execlists.tasklet));
> +
> + if (ve->context.state)
> + __execlists_context_fini(&ve->context);
> +
> + i915_timeline_fini(&ve->base.timeline);
> + kfree(ve);
> +}
> +
> +static void virtual_engine_initial_hint(struct virtual_engine *ve)
> +{
> + int swp;
> +
> + /*
> + * Pick a random sibling on starting to help spread the load around.
> + *
> + * New contexts are typically created with exactly the same order
> + * of siblings, and often started in batches. Due to the way we iterate
> + * the array of sibling when submitting requests, sibling[0] is
> + * prioritised for dequeuing. If we make sure that sibling[0] is fairly
> + * randomised across the system, we also help spread the load by the
> + * first engine we inspect being different each time.
> + *
> + * NB This does not force us to execute on this engine, it will just
> + * typically be the first we inspect for submission.
> + */
> + swp = prandom_u32_max(ve->num_siblings);
> + if (!swp)
> + return;
> +
> + swap(ve->siblings[swp], ve->siblings[0]);
> + virtual_update_register_offsets(ve->context.lrc_reg_state,
> + ve->siblings[0]);
> +}
> +
> +static int virtual_context_pin(struct intel_context *ce)
> +{
> + struct virtual_engine *ve = container_of(ce, typeof(*ve), context);
> + int err;
> +
> + /* Note: we must use a real engine class for setting up reg state */
> + err = __execlists_context_pin(ce, ve->siblings[0]);
> + if (err)
> + return err;
> +
> + virtual_engine_initial_hint(ve);
> + return 0;
> +}
> +
> +static void virtual_context_enter(struct intel_context *ce)
> +{
> + struct virtual_engine *ve = container_of(ce, typeof(*ve), context);
> + unsigned int n;
> +
> + for (n = 0; n < ve->num_siblings; n++)
> + intel_engine_pm_get(ve->siblings[n]);
> +}
> +
> +static void virtual_context_exit(struct intel_context *ce)
> +{
> + struct virtual_engine *ve = container_of(ce, typeof(*ve), context);
> + unsigned int n;
> +
> + ce->saturated = 0;
> + for (n = 0; n < ve->num_siblings; n++)
> + intel_engine_pm_put(ve->siblings[n]);
> +}
> +
> +static const struct intel_context_ops virtual_context_ops = {
> + .pin = virtual_context_pin,
> + .unpin = execlists_context_unpin,
> +
> + .enter = virtual_context_enter,
> + .exit = virtual_context_exit,
> +
> + .destroy = virtual_context_destroy,
> +};
> +
> +static void virtual_submission_tasklet(unsigned long data)
> +{
> + struct virtual_engine * const ve = (struct virtual_engine *)data;
> + const int prio = ve->base.execlists.queue_priority_hint;
> + unsigned int n;
> +
> + local_irq_disable();
> + for (n = 0; READ_ONCE(ve->request) && n < ve->num_siblings; n++) {
> + struct intel_engine_cs *sibling = ve->siblings[n];
> + struct ve_node * const node = &ve->nodes[sibling->id];
> + struct rb_node **parent, *rb;
> + bool first;
> +
> + spin_lock(&sibling->timeline.lock);
> +
> + if (!RB_EMPTY_NODE(&node->rb)) {
> + /*
> + * Cheat and avoid rebalancing the tree if we can
> + * reuse this node in situ.
> + */
> + first = rb_first_cached(&sibling->execlists.virtual) ==
> + &node->rb;
> + if (prio == node->prio || (prio > node->prio && first))
> + goto submit_engine;
> +
> + rb_erase_cached(&node->rb, &sibling->execlists.virtual);
> + }
> +
> + rb = NULL;
> + first = true;
> + parent = &sibling->execlists.virtual.rb_root.rb_node;
> + while (*parent) {
> + struct ve_node *other;
> +
> + rb = *parent;
> + other = rb_entry(rb, typeof(*other), rb);
> + if (prio > other->prio) {
> + parent = &rb->rb_left;
> + } else {
> + parent = &rb->rb_right;
> + first = false;
> + }
> + }
> +
> + rb_link_node(&node->rb, rb, parent);
> + rb_insert_color_cached(&node->rb,
> + &sibling->execlists.virtual,
> + first);
> +
> +submit_engine:
> + GEM_BUG_ON(RB_EMPTY_NODE(&node->rb));
> + node->prio = prio;
> + if (first && prio > sibling->execlists.queue_priority_hint) {
> + sibling->execlists.queue_priority_hint = prio;
> + tasklet_hi_schedule(&sibling->execlists.tasklet);
> + }
> +
> + spin_unlock(&sibling->timeline.lock);
> + }
> + local_irq_enable();
> +}
> +
> +static void virtual_submit_request(struct i915_request *rq)
> +{
> + struct virtual_engine *ve = to_virtual_engine(rq->engine);
> +
> + GEM_TRACE("%s: rq=%llx:%lld\n",
> + ve->base.name,
> + rq->fence.context,
> + rq->fence.seqno);
> +
> + GEM_BUG_ON(ve->base.submit_request != virtual_submit_request);
> +
> + GEM_BUG_ON(ve->request);
> + ve->base.execlists.queue_priority_hint = rq_prio(rq);
> + WRITE_ONCE(ve->request, rq);
> +
> + tasklet_schedule(&ve->base.execlists.tasklet);
> +}
> +
> +struct intel_context *
> +intel_execlists_create_virtual(struct i915_gem_context *ctx,
> + struct intel_engine_cs **siblings,
> + unsigned int count)
> +{
> + struct virtual_engine *ve;
> + unsigned int n;
> + int err;
> +
> + if (count == 0)
> + return ERR_PTR(-EINVAL);
> +
> + if (count == 1)
> + return intel_context_create(ctx, siblings[0]);
> +
> + ve = kzalloc(struct_size(ve, siblings, count), GFP_KERNEL);
> + if (!ve)
> + return ERR_PTR(-ENOMEM);
> +
> + ve->base.i915 = ctx->i915;
> + ve->base.id = -1;
> + ve->base.class = OTHER_CLASS;
> + ve->base.uabi_class = I915_ENGINE_CLASS_INVALID;
> + ve->base.instance = I915_ENGINE_CLASS_INVALID_VIRTUAL;
> + ve->base.flags = I915_ENGINE_IS_VIRTUAL;
> +
> + snprintf(ve->base.name, sizeof(ve->base.name), "virtual");
> +
> + err = i915_timeline_init(ctx->i915, &ve->base.timeline, NULL);
> + if (err)
> + goto err_put;
> + i915_timeline_set_subclass(&ve->base.timeline, TIMELINE_VIRTUAL);
> +
> + intel_engine_init_execlists(&ve->base);
> +
> + ve->base.cops = &virtual_context_ops;
> + ve->base.request_alloc = execlists_request_alloc;
> +
> + ve->base.schedule = i915_schedule;
> + ve->base.submit_request = virtual_submit_request;
> +
> + ve->base.execlists.queue_priority_hint = INT_MIN;
> + tasklet_init(&ve->base.execlists.tasklet,
> + virtual_submission_tasklet,
> + (unsigned long)ve);
> +
> + intel_context_init(&ve->context, ctx, &ve->base);
> +
> + for (n = 0; n < count; n++) {
> + struct intel_engine_cs *sibling = siblings[n];
> +
> + GEM_BUG_ON(!is_power_of_2(sibling->mask));
> + if (sibling->mask & ve->base.mask) {
> + DRM_DEBUG("duplicate %s entry in load balancer\n",
> + sibling->name);
> + err = -EINVAL;
> + goto err_put;
> + }
> +
> + /*
> + * The virtual engine implementation is tightly coupled to
> + * the execlists backend -- we push out request directly
> + * into a tree inside each physical engine. We could support
> + * layering if we handle cloning of the requests and
> + * submitting a copy into each backend.
> + */
> + if (sibling->execlists.tasklet.func !=
> + execlists_submission_tasklet) {
> + err = -ENODEV;
> + goto err_put;
> + }
> +
> + GEM_BUG_ON(RB_EMPTY_NODE(&ve->nodes[sibling->id].rb));
> + RB_CLEAR_NODE(&ve->nodes[sibling->id].rb);
> +
> + ve->siblings[ve->num_siblings++] = sibling;
> + ve->base.mask |= sibling->mask;
> +
> + /*
> + * All physical engines must be compatible for their emission
> + * functions (as we build the instructions during request
> + * construction and do not alter them before submission
> + * on the physical engine). We use the engine class as a guide
> + * here, although that could be refined.
> + */
> + if (ve->base.class != OTHER_CLASS) {
> + if (ve->base.class != sibling->class) {
> + DRM_DEBUG("invalid mixing of engine class, sibling %d, already %d\n",
> + sibling->class, ve->base.class);
> + err = -EINVAL;
> + goto err_put;
> + }
> + continue;
> + }
> +
> + ve->base.class = sibling->class;
> + ve->base.uabi_class = sibling->uabi_class;
> + snprintf(ve->base.name, sizeof(ve->base.name),
> + "v%dx%d", ve->base.class, count);
> + ve->base.context_size = sibling->context_size;
> +
> + ve->base.emit_bb_start = sibling->emit_bb_start;
> + ve->base.emit_flush = sibling->emit_flush;
> + ve->base.emit_init_breadcrumb = sibling->emit_init_breadcrumb;
> + ve->base.emit_fini_breadcrumb = sibling->emit_fini_breadcrumb;
> + ve->base.emit_fini_breadcrumb_dw =
> + sibling->emit_fini_breadcrumb_dw;
> + }
> +
> + return &ve->context;
> +
> +err_put:
> + intel_context_put(&ve->context);
> + return ERR_PTR(err);
> +}
> +
> +struct intel_context *
> +intel_execlists_clone_virtual(struct i915_gem_context *ctx,
> + struct intel_engine_cs *src)
> +{
> + struct virtual_engine *se = to_virtual_engine(src);
> + struct intel_context *dst;
> +
> + dst = intel_execlists_create_virtual(ctx,
> + se->siblings,
> + se->num_siblings);
> + if (IS_ERR(dst))
> + return dst;
> +
> + return dst;
> +}
> +
> void intel_execlists_show_requests(struct intel_engine_cs *engine,
> struct drm_printer *m,
> void (*show_request)(struct drm_printer *m,
> @@ -2810,6 +3428,29 @@ void intel_execlists_show_requests(struct intel_engine_cs *engine,
> show_request(m, last, "\t\tQ ");
> }
>
> + last = NULL;
> + count = 0;
> + for (rb = rb_first_cached(&execlists->virtual); rb; rb = rb_next(rb)) {
> + struct virtual_engine *ve =
> + rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
> + struct i915_request *rq = READ_ONCE(ve->request);
> +
> + if (rq) {
> + if (count++ < max - 1)
> + show_request(m, rq, "\t\tV ");
> + else
> + last = rq;
> + }
> + }
> + if (last) {
> + if (count > max) {
> + drm_printf(m,
> + "\t\t...skipping %d virtual requests...\n",
> + count - max);
> + }
> + show_request(m, last, "\t\tV ");
> + }
> +
> spin_unlock_irqrestore(&engine->timeline.lock, flags);
> }
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.h b/drivers/gpu/drm/i915/gt/intel_lrc.h
> index a0dc907a7249..5530606052e5 100644
> --- a/drivers/gpu/drm/i915/gt/intel_lrc.h
> +++ b/drivers/gpu/drm/i915/gt/intel_lrc.h
> @@ -114,4 +114,13 @@ void intel_execlists_show_requests(struct intel_engine_cs *engine,
> const char *prefix),
> unsigned int max);
>
> +struct intel_context *
> +intel_execlists_create_virtual(struct i915_gem_context *ctx,
> + struct intel_engine_cs **siblings,
> + unsigned int count);
> +
> +struct intel_context *
> +intel_execlists_clone_virtual(struct i915_gem_context *ctx,
> + struct intel_engine_cs *src);
> +
> #endif /* _INTEL_LRC_H_ */
> diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
> index 5b3d8e33f1cf..ccc0b6350123 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
> @@ -1310,6 +1310,185 @@ static int live_preempt_smoke(void *arg)
> return err;
> }
>
> +static int nop_virtual_engine(struct drm_i915_private *i915,
> + struct intel_engine_cs **siblings,
> + unsigned int nsibling,
> + unsigned int nctx,
> + unsigned int flags)
> +#define CHAIN BIT(0)
> +{
> + IGT_TIMEOUT(end_time);
> + struct i915_request *request[16];
> + struct i915_gem_context *ctx[16];
> + struct intel_context *ve[16];
> + unsigned long n, prime, nc;
> + struct igt_live_test t;
> + ktime_t times[2] = {};
> + int err;
> +
> + GEM_BUG_ON(!nctx || nctx > ARRAY_SIZE(ctx));
> +
> + for (n = 0; n < nctx; n++) {
> + ctx[n] = kernel_context(i915);
> + if (!ctx[n]) {
> + err = -ENOMEM;
> + nctx = n;
> + goto out;
> + }
> +
> + ve[n] = intel_execlists_create_virtual(ctx[n],
> + siblings, nsibling);
> + if (IS_ERR(ve[n])) {
> + kernel_context_close(ctx[n]);
> + err = PTR_ERR(ve[n]);
> + nctx = n;
> + goto out;
> + }
> +
> + err = intel_context_pin(ve[n]);
> + if (err) {
> + intel_context_put(ve[n]);
> + kernel_context_close(ctx[n]);
> + nctx = n;
> + goto out;
> + }
> + }
> +
> + err = igt_live_test_begin(&t, i915, __func__, ve[0]->engine->name);
> + if (err)
> + goto out;
> +
> + for_each_prime_number_from(prime, 1, 8192) {
> + times[1] = ktime_get_raw();
> +
> + if (flags & CHAIN) {
> + for (nc = 0; nc < nctx; nc++) {
> + for (n = 0; n < prime; n++) {
> + request[nc] =
> + i915_request_create(ve[nc]);
> + if (IS_ERR(request[nc])) {
> + err = PTR_ERR(request[nc]);
> + goto out;
> + }
> +
> + i915_request_add(request[nc]);
> + }
> + }
> + } else {
> + for (n = 0; n < prime; n++) {
> + for (nc = 0; nc < nctx; nc++) {
> + request[nc] =
> + i915_request_create(ve[nc]);
> + if (IS_ERR(request[nc])) {
> + err = PTR_ERR(request[nc]);
> + goto out;
> + }
> +
> + i915_request_add(request[nc]);
> + }
> + }
> + }
> +
> + for (nc = 0; nc < nctx; nc++) {
> + if (i915_request_wait(request[nc],
> + I915_WAIT_LOCKED,
> + HZ / 10) < 0) {
> + pr_err("%s(%s): wait for %llx:%lld timed out\n",
> + __func__, ve[0]->engine->name,
> + request[nc]->fence.context,
> + request[nc]->fence.seqno);
> +
> + GEM_TRACE("%s(%s) failed at request %llx:%lld\n",
> + __func__, ve[0]->engine->name,
> + request[nc]->fence.context,
> + request[nc]->fence.seqno);
> + GEM_TRACE_DUMP();
> + i915_gem_set_wedged(i915);
> + break;
> + }
> + }
> +
> + times[1] = ktime_sub(ktime_get_raw(), times[1]);
> + if (prime == 1)
> + times[0] = times[1];
> +
> + if (__igt_timeout(end_time, NULL))
> + break;
> + }
> +
> + err = igt_live_test_end(&t);
> + if (err)
> + goto out;
> +
> + pr_info("Requestx%d latencies on %s: 1 = %lluns, %lu = %lluns\n",
> + nctx, ve[0]->engine->name, ktime_to_ns(times[0]),
> + prime, div64_u64(ktime_to_ns(times[1]), prime));
> +
> +out:
> + if (igt_flush_test(i915, I915_WAIT_LOCKED))
> + err = -EIO;
> +
> + for (nc = 0; nc < nctx; nc++) {
> + intel_context_unpin(ve[nc]);
> + intel_context_put(ve[nc]);
> + kernel_context_close(ctx[nc]);
> + }
> + return err;
> +}
> +
> +static int live_virtual_engine(void *arg)
> +{
> + struct drm_i915_private *i915 = arg;
> + struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1];
> + struct intel_engine_cs *engine;
> + enum intel_engine_id id;
> + unsigned int class, inst;
> + int err = -ENODEV;
> +
> + if (USES_GUC_SUBMISSION(i915))
> + return 0;
> +
> + mutex_lock(&i915->drm.struct_mutex);
> +
> + for_each_engine(engine, i915, id) {
> + err = nop_virtual_engine(i915, &engine, 1, 1, 0);
> + if (err) {
> + pr_err("Failed to wrap engine %s: err=%d\n",
> + engine->name, err);
> + goto out_unlock;
> + }
> + }
> +
> + for (class = 0; class <= MAX_ENGINE_CLASS; class++) {
> + int nsibling, n;
> +
> + nsibling = 0;
> + for (inst = 0; inst <= MAX_ENGINE_INSTANCE; inst++) {
> + if (!i915->engine_class[class][inst])
> + break;
I previous review I said I think this should be continue instead of
break so vcs0 + vcs2 skus can also be tested.
> +
> + siblings[nsibling++] = i915->engine_class[class][inst];
> + }
> + if (nsibling < 2)
> + continue;
And also that single engine VE could be tested just as well, unless I am
missing something.
That's all remaining feedback I have on this patch.
Regards,
Tvrtko
> +
> + for (n = 1; n <= nsibling + 1; n++) {
> + err = nop_virtual_engine(i915, siblings, nsibling,
> + n, 0);
> + if (err)
> + goto out_unlock;
> + }
> +
> + err = nop_virtual_engine(i915, siblings, nsibling, n, CHAIN);
> + if (err)
> + goto out_unlock;
> + }
> +
> +out_unlock:
> + mutex_unlock(&i915->drm.struct_mutex);
> + return err;
> +}
> +
> int intel_execlists_live_selftests(struct drm_i915_private *i915)
> {
> static const struct i915_subtest tests[] = {
> @@ -1322,6 +1501,7 @@ int intel_execlists_live_selftests(struct drm_i915_private *i915)
> SUBTEST(live_chain_preempt),
> SUBTEST(live_preempt_hang),
> SUBTEST(live_preempt_smoke),
> + SUBTEST(live_virtual_engine),
> };
>
> if (!HAS_EXECLISTS(i915))
> diff --git a/drivers/gpu/drm/i915/i915_gem.h b/drivers/gpu/drm/i915/i915_gem.h
> index 67f8a4a807a0..fe82d3571072 100644
> --- a/drivers/gpu/drm/i915/i915_gem.h
> +++ b/drivers/gpu/drm/i915/i915_gem.h
> @@ -91,4 +91,9 @@ static inline bool __tasklet_enable(struct tasklet_struct *t)
> return atomic_dec_and_test(&t->count);
> }
>
> +static inline bool __tasklet_is_scheduled(struct tasklet_struct *t)
> +{
> + return test_bit(TASKLET_STATE_SCHED, &t->state);
> +}
> +
> #endif /* __I915_GEM_H__ */
> diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
> index 040c0b770c21..3ae1d27130a8 100644
> --- a/drivers/gpu/drm/i915/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/i915_gem_context.c
> @@ -86,6 +86,7 @@
> */
>
> #include <linux/log2.h>
> +#include <linux/nospec.h>
>
> #include <drm/i915_drm.h>
>
> @@ -1218,7 +1219,6 @@ __intel_context_reconfigure_sseu(struct intel_context *ce,
> int ret;
>
> GEM_BUG_ON(INTEL_GEN(ce->gem_context->i915) < 8);
> - GEM_BUG_ON(ce->engine->id != RCS0);
>
> ret = intel_context_lock_pinned(ce);
> if (ret)
> @@ -1412,7 +1412,100 @@ struct set_engines {
> struct i915_gem_engines *engines;
> };
>
> +static int
> +set_engines__load_balance(struct i915_user_extension __user *base, void *data)
> +{
> + struct i915_context_engines_load_balance __user *ext =
> + container_of_user(base, typeof(*ext), base);
> + const struct set_engines *set = data;
> + struct intel_engine_cs *stack[16];
> + struct intel_engine_cs **siblings;
> + struct intel_context *ce;
> + u16 num_siblings, idx;
> + unsigned int n;
> + int err;
> +
> + if (!HAS_EXECLISTS(set->ctx->i915))
> + return -ENODEV;
> +
> + if (USES_GUC_SUBMISSION(set->ctx->i915))
> + return -ENODEV; /* not implement yet */
> +
> + if (get_user(idx, &ext->engine_index))
> + return -EFAULT;
> +
> + if (idx >= set->engines->num_engines) {
> + DRM_DEBUG("Invalid placement value, %d >= %d\n",
> + idx, set->engines->num_engines);
> + return -EINVAL;
> + }
> +
> + idx = array_index_nospec(idx, set->engines->num_engines);
> + if (set->engines->engines[idx]) {
> + DRM_DEBUG("Invalid placement[%d], already occupied\n", idx);
> + return -EEXIST;
> + }
> +
> + if (get_user(num_siblings, &ext->num_siblings))
> + return -EFAULT;
> +
> + err = check_user_mbz(&ext->flags);
> + if (err)
> + return err;
> +
> + err = check_user_mbz(&ext->mbz64);
> + if (err)
> + return err;
> +
> + siblings = stack;
> + if (num_siblings > ARRAY_SIZE(stack)) {
> + siblings = kmalloc_array(num_siblings,
> + sizeof(*siblings),
> + GFP_KERNEL);
> + if (!siblings)
> + return -ENOMEM;
> + }
> +
> + for (n = 0; n < num_siblings; n++) {
> + struct i915_engine_class_instance ci;
> +
> + if (copy_from_user(&ci, &ext->engines[n], sizeof(ci))) {
> + err = -EFAULT;
> + goto out_siblings;
> + }
> +
> + siblings[n] = intel_engine_lookup_user(set->ctx->i915,
> + ci.engine_class,
> + ci.engine_instance);
> + if (!siblings[n]) {
> + DRM_DEBUG("Invalid sibling[%d]: { class:%d, inst:%d }\n",
> + n, ci.engine_class, ci.engine_instance);
> + err = -EINVAL;
> + goto out_siblings;
> + }
> + }
> +
> + ce = intel_execlists_create_virtual(set->ctx, siblings, n);
> + if (IS_ERR(ce)) {
> + err = PTR_ERR(ce);
> + goto out_siblings;
> + }
> +
> + if (cmpxchg(&set->engines->engines[idx], NULL, ce)) {
> + intel_context_put(ce);
> + err = -EEXIST;
> + goto out_siblings;
> + }
> +
> +out_siblings:
> + if (siblings != stack)
> + kfree(siblings);
> +
> + return err;
> +}
> +
> static const i915_user_extension_fn set_engines__extensions[] = {
> + [I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE] = set_engines__load_balance,
> };
>
> static int
> @@ -1711,14 +1804,29 @@ static int clone_engines(struct i915_gem_context *dst,
>
> clone->i915 = dst->i915;
> for (n = 0; n < e->num_engines; n++) {
> + struct intel_engine_cs *engine;
> +
> if (!e->engines[n]) {
> clone->engines[n] = NULL;
> continue;
> }
> + engine = e->engines[n]->engine;
>
> - clone->engines[n] =
> - intel_context_create(dst, e->engines[n]->engine);
> - if (!clone->engines[n]) {
> + /*
> + * Virtual engines are singletons; they can only exist
> + * inside a single context, because they embed their
> + * HW context... As each virtual context implies a single
> + * timeline (each engine can only dequeue a single request
> + * at any time), it would be surprising for two contexts
> + * to use the same engine. So let's create a copy of
> + * the virtual engine instead.
> + */
> + if (intel_engine_is_virtual(engine))
> + clone->engines[n] =
> + intel_execlists_clone_virtual(dst, engine);
> + else
> + clone->engines[n] = intel_context_create(dst, engine);
> + if (IS_ERR_OR_NULL(clone->engines[n])) {
> __free_engines(clone, n);
> goto err_unlock;
> }
> diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c
> index d215dcdf0b1e..78ceb56d7801 100644
> --- a/drivers/gpu/drm/i915/i915_scheduler.c
> +++ b/drivers/gpu/drm/i915/i915_scheduler.c
> @@ -150,17 +150,26 @@ sched_lock_engine(const struct i915_sched_node *node,
> struct intel_engine_cs *locked,
> struct sched_cache *cache)
> {
> - struct intel_engine_cs *engine = node_to_request(node)->engine;
> + const struct i915_request *rq = node_to_request(node);
> + struct intel_engine_cs *engine;
>
> GEM_BUG_ON(!locked);
>
> - if (engine != locked) {
> + /*
> + * Virtual engines complicate acquiring the engine timeline lock,
> + * as their rq->engine pointer is not stable until under that
> + * engine lock. The simple ploy we use is to take the lock then
> + * check that the rq still belongs to the newly locked engine.
> + */
> + while (locked != (engine = READ_ONCE(rq->engine))) {
> spin_unlock(&locked->timeline.lock);
> memset(cache, 0, sizeof(*cache));
> spin_lock(&engine->timeline.lock);
> + locked = engine;
> }
>
> - return engine;
> + GEM_BUG_ON(locked != engine);
> + return locked;
> }
>
> static inline int rq_prio(const struct i915_request *rq)
> @@ -272,6 +281,7 @@ static void __i915_schedule(struct i915_sched_node *node,
> spin_lock(&engine->timeline.lock);
>
> /* Fifo and depth-first replacement ensure our deps execute before us */
> + engine = sched_lock_engine(node, engine, &cache);
> list_for_each_entry_safe_reverse(dep, p, &dfs, dfs_link) {
> INIT_LIST_HEAD(&dep->dfs_link);
>
> @@ -283,8 +293,11 @@ static void __i915_schedule(struct i915_sched_node *node,
> if (prio <= node->attr.priority || node_signaled(node))
> continue;
>
> + GEM_BUG_ON(node_to_request(node)->engine != engine);
> +
> node->attr.priority = prio;
> if (!list_empty(&node->link)) {
> + GEM_BUG_ON(intel_engine_is_virtual(engine));
> if (!cache.priolist)
> cache.priolist =
> i915_sched_lookup_priolist(engine,
> diff --git a/drivers/gpu/drm/i915/i915_timeline_types.h b/drivers/gpu/drm/i915/i915_timeline_types.h
> index 5256a0b5c5f7..1688705f4a2b 100644
> --- a/drivers/gpu/drm/i915/i915_timeline_types.h
> +++ b/drivers/gpu/drm/i915/i915_timeline_types.h
> @@ -26,6 +26,7 @@ struct i915_timeline {
> spinlock_t lock;
> #define TIMELINE_CLIENT 0 /* default subclass */
> #define TIMELINE_ENGINE 1
> +#define TIMELINE_VIRTUAL 2
> struct mutex mutex; /* protects the flow of requests */
>
> unsigned int pin_count;
> diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
> index 62396d575e28..f9770948161c 100644
> --- a/include/uapi/drm/i915_drm.h
> +++ b/include/uapi/drm/i915_drm.h
> @@ -137,6 +137,7 @@ struct i915_engine_class_instance {
> __u16 engine_class; /* see enum drm_i915_gem_engine_class */
> __u16 engine_instance;
> #define I915_ENGINE_CLASS_INVALID_NONE -1
> +#define I915_ENGINE_CLASS_INVALID_VIRTUAL -2
> };
>
> /**
> @@ -1608,8 +1609,46 @@ struct drm_i915_gem_context_param_sseu {
> __u32 rsvd;
> };
>
> +/*
> + * i915_context_engines_load_balance:
> + *
> + * Enable load balancing across this set of engines.
> + *
> + * Into the I915_EXEC_DEFAULT slot [0], a virtual engine is created that when
> + * used will proxy the execbuffer request onto one of the set of engines
> + * in such a way as to distribute the load evenly across the set.
> + *
> + * The set of engines must be compatible (e.g. the same HW class) as they
> + * will share the same logical GPU context and ring.
> + *
> + * To intermix rendering with the virtual engine and direct rendering onto
> + * the backing engines (bypassing the load balancing proxy), the context must
> + * be defined to use a single timeline for all engines.
> + */
> +struct i915_context_engines_load_balance {
> + struct i915_user_extension base;
> +
> + __u16 engine_index;
> + __u16 num_siblings;
> + __u32 flags; /* all undefined flags must be zero */
> +
> + __u64 mbz64; /* reserved for future use; must be zero */
> +
> + struct i915_engine_class_instance engines[0];
> +} __attribute__((packed));
> +
> +#define I915_DEFINE_CONTEXT_ENGINES_LOAD_BALANCE(name__, N__) struct { \
> + struct i915_user_extension base; \
> + __u16 engine_index; \
> + __u16 num_siblings; \
> + __u32 flags; \
> + __u64 mbz64; \
> + struct i915_engine_class_instance engines[N__]; \
> +} __attribute__((packed)) name__
> +
> struct i915_context_param_engines {
> __u64 extensions; /* linked chain of extension blocks, 0 terminates */
> +#define I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE 0 /* see i915_context_engines_load_balance */
> struct i915_engine_class_instance engines[0];
> } __attribute__((packed));
>
>
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread* Re: [PATCH 14/40] drm/i915: Load balancing across a virtual engine
2019-05-08 10:29 ` Tvrtko Ursulin
@ 2019-05-08 11:17 ` Chris Wilson
2019-05-08 11:36 ` Tvrtko Ursulin
0 siblings, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 11:17 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Tvrtko Ursulin (2019-05-08 11:29:34)
>
> On 08/05/2019 09:06, Chris Wilson wrote:
> > +static int live_virtual_engine(void *arg)
> > +{
> > + struct drm_i915_private *i915 = arg;
> > + struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1];
> > + struct intel_engine_cs *engine;
> > + enum intel_engine_id id;
> > + unsigned int class, inst;
> > + int err = -ENODEV;
> > +
> > + if (USES_GUC_SUBMISSION(i915))
> > + return 0;
> > +
> > + mutex_lock(&i915->drm.struct_mutex);
> > +
> > + for_each_engine(engine, i915, id) {
> > + err = nop_virtual_engine(i915, &engine, 1, 1, 0);
> > + if (err) {
> > + pr_err("Failed to wrap engine %s: err=%d\n",
> > + engine->name, err);
> > + goto out_unlock;
> > + }
> > + }
> > +
> > + for (class = 0; class <= MAX_ENGINE_CLASS; class++) {
> > + int nsibling, n;
> > +
> > + nsibling = 0;
> > + for (inst = 0; inst <= MAX_ENGINE_INSTANCE; inst++) {
> > + if (!i915->engine_class[class][inst])
> > + break;
>
> I previous review I said I think this should be continue instead of
> break so vcs0 + vcs2 skus can also be tested.
Completely missed that, sorry.
> > +
> > + siblings[nsibling++] = i915->engine_class[class][inst];
> > + }
> > + if (nsibling < 2)
> > + continue;
>
> And also that single engine VE could be tested just as well, unless I am
> missing something.
There's no such thing as single engine VE. The current design requires
that this type of struct virtual_engine encompasses more than one engine
(failing that we break the single request scheduling, although might be
able to lift that with timeslicing but the early results were not
favourable); the single engine being a regular intel_context instance.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread* Re: [PATCH 14/40] drm/i915: Load balancing across a virtual engine
2019-05-08 11:17 ` Chris Wilson
@ 2019-05-08 11:36 ` Tvrtko Ursulin
0 siblings, 0 replies; 71+ messages in thread
From: Tvrtko Ursulin @ 2019-05-08 11:36 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 08/05/2019 12:17, Chris Wilson wrote:
> Quoting Tvrtko Ursulin (2019-05-08 11:29:34)
>>
>> On 08/05/2019 09:06, Chris Wilson wrote:
>>> +static int live_virtual_engine(void *arg)
>>> +{
>>> + struct drm_i915_private *i915 = arg;
>>> + struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1];
>>> + struct intel_engine_cs *engine;
>>> + enum intel_engine_id id;
>>> + unsigned int class, inst;
>>> + int err = -ENODEV;
>>> +
>>> + if (USES_GUC_SUBMISSION(i915))
>>> + return 0;
>>> +
>>> + mutex_lock(&i915->drm.struct_mutex);
>>> +
>>> + for_each_engine(engine, i915, id) {
>>> + err = nop_virtual_engine(i915, &engine, 1, 1, 0);
>>> + if (err) {
>>> + pr_err("Failed to wrap engine %s: err=%d\n",
>>> + engine->name, err);
>>> + goto out_unlock;
>>> + }
>>> + }
>>> +
>>> + for (class = 0; class <= MAX_ENGINE_CLASS; class++) {
>>> + int nsibling, n;
>>> +
>>> + nsibling = 0;
>>> + for (inst = 0; inst <= MAX_ENGINE_INSTANCE; inst++) {
>>> + if (!i915->engine_class[class][inst])
>>> + break;
>>
>> I previous review I said I think this should be continue instead of
>> break so vcs0 + vcs2 skus can also be tested.
>
> Completely missed that, sorry.
>
>>> +
>>> + siblings[nsibling++] = i915->engine_class[class][inst];
>>> + }
>>> + if (nsibling < 2)
>>> + continue;
>>
>> And also that single engine VE could be tested just as well, unless I am
>> missing something.
>
> There's no such thing as single engine VE. The current design requires
> that this type of struct virtual_engine encompasses more than one engine
> (failing that we break the single request scheduling, although might be
> able to lift that with timeslicing but the early results were not
> favourable); the single engine being a regular intel_context instance.
Yeah my bad, the auto-magic replacement with physical engine happens one
level higher than what this selftest is operating on.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread
* [PATCH] drm/i915: Load balancing across a virtual engine
2019-05-08 8:06 ` [PATCH 14/40] drm/i915: Load balancing across a virtual engine Chris Wilson
2019-05-08 10:29 ` Tvrtko Ursulin
@ 2019-05-08 11:23 ` Chris Wilson
2019-05-08 11:35 ` Tvrtko Ursulin
1 sibling, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 11:23 UTC (permalink / raw)
To: intel-gfx
Having allowed the user to define a set of engines that they will want
to only use, we go one step further and allow them to bind those engines
into a single virtual instance. Submitting a batch to the virtual engine
will then forward it to any one of the set in a manner as best to
distribute load. The virtual engine has a single timeline across all
engines (it operates as a single queue), so it is not able to concurrently
run batches across multiple engines by itself; that is left up to the user
to submit multiple concurrent batches to multiple queues. Multiple users
will be load balanced across the system.
The mechanism used for load balancing in this patch is a late greedy
balancer. When a request is ready for execution, it is added to each
engine's queue, and when an engine is ready for its next request it
claims it from the virtual engine. The first engine to do so, wins, i.e.
the request is executed at the earliest opportunity (idle moment) in the
system.
As not all HW is created equal, the user is still able to skip the
virtual engine and execute the batch on a specific engine, all within the
same queue. It will then be executed in order on the correct engine,
with execution on other virtual engines being moved away due to the load
detection.
A couple of areas for potential improvement left!
- The virtual engine always take priority over equal-priority tasks.
Mostly broken up by applying FQ_CODEL rules for prioritising new clients,
and hopefully the virtual and real engines are not then congested (i.e.
all work is via virtual engines, or all work is to the real engine).
- We require the breadcrumb irq around every virtual engine request. For
normal engines, we eliminate the need for the slow round trip via
interrupt by using the submit fence and queueing in order. For virtual
engines, we have to allow any job to transfer to a new ring, and cannot
coalesce the submissions, so require the completion fence instead,
forcing the persistent use of interrupts.
- We only drip feed single requests through each virtual engine and onto
the physical engines, even if there was enough work to fill all ELSP,
leaving small stalls with an idle CS event at the end of every request.
Could we be greedy and fill both slots? Being lazy is virtuous for load
distribution on less-than-full workloads though.
Other areas of improvement are more general, such as reducing lock
contention, reducing dispatch overhead, looking at direct submission
rather than bouncing around tasklets etc.
sseu: Lift the restriction to allow sseu to be reconfigured on virtual
engines composed of RENDER_CLASS (rcs).
v2: macroize check_user_mbz()
v3: Cancel virtual engines on wedging
v4: Commence commenting
v5: Replace 64b sibling_mask with a list of class:instance
v6: Drop the one-element array in the uabi
v7: Assert it is an virtual engine in to_virtual_engine()
v8: Skip over holes in [class][inst] so we can selftest with (vcs0, vcs2)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/gt/intel_engine_types.h | 8 +
drivers/gpu/drm/i915/gt/intel_lrc.c | 683 ++++++++++++++++++-
drivers/gpu/drm/i915/gt/intel_lrc.h | 9 +
drivers/gpu/drm/i915/gt/selftest_lrc.c | 180 +++++
drivers/gpu/drm/i915/i915_gem.h | 5 +
drivers/gpu/drm/i915/i915_gem_context.c | 116 +++-
drivers/gpu/drm/i915/i915_scheduler.c | 19 +-
drivers/gpu/drm/i915/i915_timeline_types.h | 1 +
include/uapi/drm/i915_drm.h | 39 ++
9 files changed, 1032 insertions(+), 28 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
index e381c1c73902..7b47e00fa082 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
@@ -227,6 +227,7 @@ struct intel_engine_execlists {
* @queue: queue of requests, in priority lists
*/
struct rb_root_cached queue;
+ struct rb_root_cached virtual;
/**
* @csb_write: control register for Context Switch buffer
@@ -445,6 +446,7 @@ struct intel_engine_cs {
#define I915_ENGINE_HAS_PREEMPTION BIT(2)
#define I915_ENGINE_HAS_SEMAPHORES BIT(3)
#define I915_ENGINE_NEEDS_BREADCRUMB_TASKLET BIT(4)
+#define I915_ENGINE_IS_VIRTUAL BIT(5)
unsigned int flags;
/*
@@ -534,6 +536,12 @@ intel_engine_needs_breadcrumb_tasklet(const struct intel_engine_cs *engine)
return engine->flags & I915_ENGINE_NEEDS_BREADCRUMB_TASKLET;
}
+static inline bool
+intel_engine_is_virtual(const struct intel_engine_cs *engine)
+{
+ return engine->flags & I915_ENGINE_IS_VIRTUAL;
+}
+
#define instdone_slice_mask(dev_priv__) \
(IS_GEN(dev_priv__, 7) ? \
1 : RUNTIME_INFO(dev_priv__)->sseu.slice_mask)
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index f1d62746e066..bc388df39802 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -166,6 +166,42 @@
#define ACTIVE_PRIORITY (I915_PRIORITY_NEWCLIENT | I915_PRIORITY_NOSEMAPHORE)
+struct virtual_engine {
+ struct intel_engine_cs base;
+ struct intel_context context;
+
+ /*
+ * We allow only a single request through the virtual engine at a time
+ * (each request in the timeline waits for the completion fence of
+ * the previous before being submitted). By restricting ourselves to
+ * only submitting a single request, each request is placed on to a
+ * physical to maximise load spreading (by virtue of the late greedy
+ * scheduling -- each real engine takes the next available request
+ * upon idling).
+ */
+ struct i915_request *request;
+
+ /*
+ * We keep a rbtree of available virtual engines inside each physical
+ * engine, sorted by priority. Here we preallocate the nodes we need
+ * for the virtual engine, indexed by physical_engine->id.
+ */
+ struct ve_node {
+ struct rb_node rb;
+ int prio;
+ } nodes[I915_NUM_ENGINES];
+
+ /* And finally, which physical engines this virtual engine maps onto. */
+ unsigned int num_siblings;
+ struct intel_engine_cs *siblings[0];
+};
+
+static struct virtual_engine *to_virtual_engine(struct intel_engine_cs *engine)
+{
+ GEM_BUG_ON(!intel_engine_is_virtual(engine));
+ return container_of(engine, struct virtual_engine, base);
+}
+
static int execlists_context_deferred_alloc(struct intel_context *ce,
struct intel_engine_cs *engine);
static void execlists_init_reg_state(u32 *reg_state,
@@ -229,7 +265,8 @@ static int queue_prio(const struct intel_engine_execlists *execlists)
}
static inline bool need_preempt(const struct intel_engine_cs *engine,
- const struct i915_request *rq)
+ const struct i915_request *rq,
+ struct rb_node *rb)
{
int last_prio;
@@ -264,6 +301,25 @@ static inline bool need_preempt(const struct intel_engine_cs *engine,
rq_prio(list_next_entry(rq, link)) > last_prio)
return true;
+ if (rb) {
+ struct virtual_engine *ve =
+ rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
+ bool preempt = false;
+
+ if (engine == ve->siblings[0]) { /* only preempt one sibling */
+ struct i915_request *next;
+
+ rcu_read_lock();
+ next = READ_ONCE(ve->request);
+ if (next)
+ preempt = rq_prio(next) > last_prio;
+ rcu_read_unlock();
+ }
+
+ if (preempt)
+ return preempt;
+ }
+
/*
* If the inflight context did not trigger the preemption, then maybe
* it was the set of queued requests? Pick the highest priority in
@@ -382,6 +438,8 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
list_for_each_entry_safe_reverse(rq, rn,
&engine->timeline.requests,
link) {
+ struct intel_engine_cs *owner;
+
if (i915_request_completed(rq))
break;
@@ -390,16 +448,32 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
GEM_BUG_ON(rq->hw_context->active);
- GEM_BUG_ON(rq_prio(rq) == I915_PRIORITY_INVALID);
- if (rq_prio(rq) != prio) {
- prio = rq_prio(rq);
- pl = i915_sched_lookup_priolist(engine, prio);
- }
- GEM_BUG_ON(RB_EMPTY_ROOT(&engine->execlists.queue.rb_root));
+ /*
+ * Push the request back into the queue for later resubmission.
+ * If this request is not native to this physical engine (i.e.
+ * it came from a virtual source), push it back onto the virtual
+ * engine so that it can be moved across onto another physical
+ * engine as load dictates.
+ */
+ owner = rq->hw_context->engine;
+ if (likely(owner == engine)) {
+ GEM_BUG_ON(rq_prio(rq) == I915_PRIORITY_INVALID);
+ if (rq_prio(rq) != prio) {
+ prio = rq_prio(rq);
+ pl = i915_sched_lookup_priolist(engine, prio);
+ }
+ GEM_BUG_ON(RB_EMPTY_ROOT(&engine->execlists.queue.rb_root));
- list_add(&rq->sched.link, pl);
+ list_add(&rq->sched.link, pl);
+ active = rq;
+ } else {
+ if (__i915_request_has_started(rq))
+ rq->sched.attr.priority |= boost;
- active = rq;
+ rq->engine = owner;
+ owner->submit_request(rq);
+ active = NULL;
+ }
}
/*
@@ -419,7 +493,7 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
* in the priority queue, but they will not gain immediate access to
* the GPU.
*/
- if (~prio & boost && __i915_request_has_started(active)) {
+ if (~prio & boost && active && __i915_request_has_started(active)) {
prio |= boost;
GEM_BUG_ON(active->sched.attr.priority >= prio);
active->sched.attr.priority = prio;
@@ -661,6 +735,90 @@ static void complete_preempt_context(struct intel_engine_execlists *execlists)
ACTIVE_PRIORITY);
}
+static void virtual_update_register_offsets(u32 *regs,
+ struct intel_engine_cs *engine)
+{
+ u32 base = engine->mmio_base;
+
+ /* Must match execlists_init_reg_state()! */
+
+ regs[CTX_CONTEXT_CONTROL] =
+ i915_mmio_reg_offset(RING_CONTEXT_CONTROL(base));
+ regs[CTX_RING_HEAD] = i915_mmio_reg_offset(RING_HEAD(base));
+ regs[CTX_RING_TAIL] = i915_mmio_reg_offset(RING_TAIL(base));
+ regs[CTX_RING_BUFFER_START] = i915_mmio_reg_offset(RING_START(base));
+ regs[CTX_RING_BUFFER_CONTROL] = i915_mmio_reg_offset(RING_CTL(base));
+
+ regs[CTX_BB_HEAD_U] = i915_mmio_reg_offset(RING_BBADDR_UDW(base));
+ regs[CTX_BB_HEAD_L] = i915_mmio_reg_offset(RING_BBADDR(base));
+ regs[CTX_BB_STATE] = i915_mmio_reg_offset(RING_BBSTATE(base));
+ regs[CTX_SECOND_BB_HEAD_U] =
+ i915_mmio_reg_offset(RING_SBBADDR_UDW(base));
+ regs[CTX_SECOND_BB_HEAD_L] = i915_mmio_reg_offset(RING_SBBADDR(base));
+ regs[CTX_SECOND_BB_STATE] = i915_mmio_reg_offset(RING_SBBSTATE(base));
+
+ regs[CTX_CTX_TIMESTAMP] =
+ i915_mmio_reg_offset(RING_CTX_TIMESTAMP(base));
+ regs[CTX_PDP3_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 3));
+ regs[CTX_PDP3_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 3));
+ regs[CTX_PDP2_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 2));
+ regs[CTX_PDP2_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 2));
+ regs[CTX_PDP1_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 1));
+ regs[CTX_PDP1_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 1));
+ regs[CTX_PDP0_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 0));
+ regs[CTX_PDP0_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 0));
+
+ if (engine->class == RENDER_CLASS) {
+ regs[CTX_RCS_INDIRECT_CTX] =
+ i915_mmio_reg_offset(RING_INDIRECT_CTX(base));
+ regs[CTX_RCS_INDIRECT_CTX_OFFSET] =
+ i915_mmio_reg_offset(RING_INDIRECT_CTX_OFFSET(base));
+ regs[CTX_BB_PER_CTX_PTR] =
+ i915_mmio_reg_offset(RING_BB_PER_CTX_PTR(base));
+
+ regs[CTX_R_PWR_CLK_STATE] =
+ i915_mmio_reg_offset(GEN8_R_PWR_CLK_STATE);
+ }
+}
+
+static bool virtual_matches(const struct virtual_engine *ve,
+ const struct i915_request *rq,
+ const struct intel_engine_cs *engine)
+{
+ const struct intel_engine_cs *active;
+
+ /*
+ * We track when the HW has completed saving the context image
+ * (i.e. when we have seen the final CS event switching out of
+ * the context) and must not overwrite the context image before
+ * then. This restricts us to only using the active engine
+ * while the previous virtualized request is inflight (so
+ * we reuse the register offsets). This is a very small
+ * hystersis on the greedy seelction algorithm.
+ */
+ active = READ_ONCE(ve->context.active);
+ if (active && active != engine)
+ return false;
+
+ return true;
+}
+
+static void virtual_xfer_breadcrumbs(struct virtual_engine *ve,
+ struct intel_engine_cs *engine)
+{
+ struct intel_engine_cs *old = ve->siblings[0];
+
+ /* All unattached (rq->engine == old) must already be completed */
+
+ spin_lock(&old->breadcrumbs.irq_lock);
+ if (!list_empty(&ve->context.signal_link)) {
+ list_move_tail(&ve->context.signal_link,
+ &engine->breadcrumbs.signalers);
+ intel_engine_queue_breadcrumbs(engine);
+ }
+ spin_unlock(&old->breadcrumbs.irq_lock);
+}
+
static void execlists_dequeue(struct intel_engine_cs *engine)
{
struct intel_engine_execlists * const execlists = &engine->execlists;
@@ -693,6 +851,26 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
* and context switches) submission.
*/
+ for (rb = rb_first_cached(&execlists->virtual); rb; ) {
+ struct virtual_engine *ve =
+ rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
+ struct i915_request *rq = READ_ONCE(ve->request);
+
+ if (!rq) { /* lazily cleanup after another engine handled rq */
+ rb_erase_cached(rb, &execlists->virtual);
+ RB_CLEAR_NODE(rb);
+ rb = rb_first_cached(&execlists->virtual);
+ continue;
+ }
+
+ if (!virtual_matches(ve, rq, engine)) {
+ rb = rb_next(rb);
+ continue;
+ }
+
+ break;
+ }
+
if (last) {
/*
* Don't resubmit or switch until all outstanding
@@ -714,7 +892,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
if (!execlists_is_active(execlists, EXECLISTS_ACTIVE_HWACK))
return;
- if (need_preempt(engine, last)) {
+ if (need_preempt(engine, last, rb)) {
inject_preempt_context(engine);
return;
}
@@ -754,6 +932,92 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
last->tail = last->wa_tail;
}
+ while (rb) { /* XXX virtual is always taking precedence */
+ struct virtual_engine *ve =
+ rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
+ struct i915_request *rq;
+
+ spin_lock(&ve->base.timeline.lock);
+
+ rq = ve->request;
+ if (unlikely(!rq)) { /* lost the race to a sibling */
+ spin_unlock(&ve->base.timeline.lock);
+ rb_erase_cached(rb, &execlists->virtual);
+ RB_CLEAR_NODE(rb);
+ rb = rb_first_cached(&execlists->virtual);
+ continue;
+ }
+
+ GEM_BUG_ON(rq != ve->request);
+ GEM_BUG_ON(rq->engine != &ve->base);
+ GEM_BUG_ON(rq->hw_context != &ve->context);
+
+ if (rq_prio(rq) >= queue_prio(execlists)) {
+ if (!virtual_matches(ve, rq, engine)) {
+ spin_unlock(&ve->base.timeline.lock);
+ rb = rb_next(rb);
+ continue;
+ }
+
+ if (last && !can_merge_rq(last, rq)) {
+ spin_unlock(&ve->base.timeline.lock);
+ return; /* leave this rq for another engine */
+ }
+
+ GEM_TRACE("%s: virtual rq=%llx:%lld%s, new engine? %s\n",
+ engine->name,
+ rq->fence.context,
+ rq->fence.seqno,
+ i915_request_completed(rq) ? "!" :
+ i915_request_started(rq) ? "*" :
+ "",
+ yesno(engine != ve->siblings[0]));
+
+ ve->request = NULL;
+ ve->base.execlists.queue_priority_hint = INT_MIN;
+ rb_erase_cached(rb, &execlists->virtual);
+ RB_CLEAR_NODE(rb);
+
+ rq->engine = engine;
+
+ if (engine != ve->siblings[0]) {
+ u32 *regs = ve->context.lrc_reg_state;
+ unsigned int n;
+
+ GEM_BUG_ON(READ_ONCE(ve->context.active));
+ virtual_update_register_offsets(regs, engine);
+
+ if (!list_empty(&ve->context.signals))
+ virtual_xfer_breadcrumbs(ve, engine);
+
+ /*
+ * Move the bound engine to the top of the list
+ * for future execution. We then kick this
+ * tasklet first before checking others, so that
+ * we preferentially reuse this set of bound
+ * registers.
+ */
+ for (n = 1; n < ve->num_siblings; n++) {
+ if (ve->siblings[n] == engine) {
+ swap(ve->siblings[n],
+ ve->siblings[0]);
+ break;
+ }
+ }
+
+ GEM_BUG_ON(ve->siblings[0] != engine);
+ }
+
+ __i915_request_submit(rq);
+ trace_i915_request_in(rq, port_index(port, execlists));
+ submit = true;
+ last = rq;
+ }
+
+ spin_unlock(&ve->base.timeline.lock);
+ break;
+ }
+
while ((rb = rb_first_cached(&execlists->queue))) {
struct i915_priolist *p = to_priolist(rb);
struct i915_request *rq, *rn;
@@ -1877,6 +2141,25 @@ static void reset_csb_pointers(struct intel_engine_execlists *execlists)
&execlists->csb_status[reset_value]);
}
+static struct i915_request *active_request(struct i915_request *rq)
+{
+ const struct list_head * const list = &rq->engine->timeline.requests;
+ const struct intel_context * const context = rq->hw_context;
+ struct i915_request *active = NULL;
+
+ list_for_each_entry_from_reverse(rq, list, link) {
+ if (i915_request_completed(rq))
+ break;
+
+ if (rq->hw_context != context)
+ break;
+
+ active = rq;
+ }
+
+ return active;
+}
+
static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
{
struct intel_engine_execlists * const execlists = &engine->execlists;
@@ -1897,7 +2180,8 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
if (!port_isset(execlists->port))
goto out_clear;
- ce = port_request(execlists->port)->hw_context;
+ rq = port_request(execlists->port);
+ ce = rq->hw_context;
/*
* Catch up with any missed context-switch interrupts.
@@ -1910,16 +2194,10 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
*/
execlists_cancel_port_requests(execlists);
- /* Push back any incomplete requests for replay after the reset. */
- rq = __unwind_incomplete_requests(engine, 0);
+ rq = active_request(rq);
if (!rq)
goto out_replay;
- if (rq->hw_context != ce) { /* caught just before a CS event */
- rq = NULL;
- goto out_replay;
- }
-
/*
* If this request hasn't started yet, e.g. it is waiting on a
* semaphore, we need to avoid skipping the request or else we
@@ -1966,13 +2244,16 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
}
execlists_init_reg_state(regs, ce, engine, ce->ring);
- /* Rerun the request; its payload has been neutered (if guilty). */
out_replay:
+ /* Rerun the request; its payload has been neutered (if guilty). */
ce->ring->head =
rq ? intel_ring_wrap(ce->ring, rq->head) : ce->ring->tail;
intel_ring_update_space(ce->ring);
__execlists_update_reg_state(ce, engine);
+ /* Push back any incomplete requests for replay after the reset. */
+ __unwind_incomplete_requests(engine, 0);
+
out_clear:
execlists_clear_all_active(execlists);
}
@@ -2046,6 +2327,26 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine)
i915_priolist_free(p);
}
+ /* Cancel all attached virtual engines */
+ while ((rb = rb_first_cached(&execlists->virtual))) {
+ struct virtual_engine *ve =
+ rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
+
+ rb_erase_cached(rb, &execlists->virtual);
+ RB_CLEAR_NODE(rb);
+
+ spin_lock(&ve->base.timeline.lock);
+ if (ve->request) {
+ ve->request->engine = engine;
+ __i915_request_submit(ve->request);
+ dma_fence_set_error(&ve->request->fence, -EIO);
+ i915_request_mark_complete(ve->request);
+ ve->base.execlists.queue_priority_hint = INT_MIN;
+ ve->request = NULL;
+ }
+ spin_unlock(&ve->base.timeline.lock);
+ }
+
/* Remaining _unready_ requests will be nop'ed when submitted */
execlists->queue_priority_hint = INT_MIN;
@@ -2529,12 +2830,15 @@ static void execlists_init_reg_state(u32 *regs,
bool rcs = engine->class == RENDER_CLASS;
u32 base = engine->mmio_base;
- /* A context is actually a big batch buffer with several
+ /*
+ * A context is actually a big batch buffer with several
* MI_LOAD_REGISTER_IMM commands followed by (reg, value) pairs. The
* values we are setting here are only for the first context restore:
* on a subsequent save, the GPU will recreate this batchbuffer with new
* values (including all the missing MI_LOAD_REGISTER_IMM commands that
* we are not initializing here).
+ *
+ * Must keep consistent with virtual_update_register_offsets().
*/
regs[CTX_LRI_HEADER_0] = MI_LOAD_REGISTER_IMM(rcs ? 14 : 11) |
MI_LRI_FORCE_POSTED;
@@ -2753,6 +3057,320 @@ static int execlists_context_deferred_alloc(struct intel_context *ce,
return ret;
}
+static void virtual_context_destroy(struct kref *kref)
+{
+ struct virtual_engine *ve =
+ container_of(kref, typeof(*ve), context.ref);
+ unsigned int n;
+
+ GEM_BUG_ON(ve->request);
+ GEM_BUG_ON(ve->context.active);
+
+ for (n = 0; n < ve->num_siblings; n++) {
+ struct intel_engine_cs *sibling = ve->siblings[n];
+ struct rb_node *node = &ve->nodes[sibling->id].rb;
+
+ if (RB_EMPTY_NODE(node))
+ continue;
+
+ spin_lock_irq(&sibling->timeline.lock);
+
+ /* Detachment is lazily performed in the execlists tasklet */
+ if (!RB_EMPTY_NODE(node))
+ rb_erase_cached(node, &sibling->execlists.virtual);
+
+ spin_unlock_irq(&sibling->timeline.lock);
+ }
+ GEM_BUG_ON(__tasklet_is_scheduled(&ve->base.execlists.tasklet));
+
+ if (ve->context.state)
+ __execlists_context_fini(&ve->context);
+
+ i915_timeline_fini(&ve->base.timeline);
+ kfree(ve);
+}
+
+static void virtual_engine_initial_hint(struct virtual_engine *ve)
+{
+ int swp;
+
+ /*
+ * Pick a random sibling on starting to help spread the load around.
+ *
+ * New contexts are typically created with exactly the same order
+ * of siblings, and often started in batches. Due to the way we iterate
+ * the array of sibling when submitting requests, sibling[0] is
+ * prioritised for dequeuing. If we make sure that sibling[0] is fairly
+ * randomised across the system, we also help spread the load by the
+ * first engine we inspect being different each time.
+ *
+ * NB This does not force us to execute on this engine, it will just
+ * typically be the first we inspect for submission.
+ */
+ swp = prandom_u32_max(ve->num_siblings);
+ if (!swp)
+ return;
+
+ swap(ve->siblings[swp], ve->siblings[0]);
+ virtual_update_register_offsets(ve->context.lrc_reg_state,
+ ve->siblings[0]);
+}
+
+static int virtual_context_pin(struct intel_context *ce)
+{
+ struct virtual_engine *ve = container_of(ce, typeof(*ve), context);
+ int err;
+
+ /* Note: we must use a real engine class for setting up reg state */
+ err = __execlists_context_pin(ce, ve->siblings[0]);
+ if (err)
+ return err;
+
+ virtual_engine_initial_hint(ve);
+ return 0;
+}
+
+static void virtual_context_enter(struct intel_context *ce)
+{
+ struct virtual_engine *ve = container_of(ce, typeof(*ve), context);
+ unsigned int n;
+
+ for (n = 0; n < ve->num_siblings; n++)
+ intel_engine_pm_get(ve->siblings[n]);
+}
+
+static void virtual_context_exit(struct intel_context *ce)
+{
+ struct virtual_engine *ve = container_of(ce, typeof(*ve), context);
+ unsigned int n;
+
+ ce->saturated = 0;
+ for (n = 0; n < ve->num_siblings; n++)
+ intel_engine_pm_put(ve->siblings[n]);
+}
+
+static const struct intel_context_ops virtual_context_ops = {
+ .pin = virtual_context_pin,
+ .unpin = execlists_context_unpin,
+
+ .enter = virtual_context_enter,
+ .exit = virtual_context_exit,
+
+ .destroy = virtual_context_destroy,
+};
+
+static void virtual_submission_tasklet(unsigned long data)
+{
+ struct virtual_engine * const ve = (struct virtual_engine *)data;
+ const int prio = ve->base.execlists.queue_priority_hint;
+ unsigned int n;
+
+ local_irq_disable();
+ for (n = 0; READ_ONCE(ve->request) && n < ve->num_siblings; n++) {
+ struct intel_engine_cs *sibling = ve->siblings[n];
+ struct ve_node * const node = &ve->nodes[sibling->id];
+ struct rb_node **parent, *rb;
+ bool first;
+
+ spin_lock(&sibling->timeline.lock);
+
+ if (!RB_EMPTY_NODE(&node->rb)) {
+ /*
+ * Cheat and avoid rebalancing the tree if we can
+ * reuse this node in situ.
+ */
+ first = rb_first_cached(&sibling->execlists.virtual) ==
+ &node->rb;
+ if (prio == node->prio || (prio > node->prio && first))
+ goto submit_engine;
+
+ rb_erase_cached(&node->rb, &sibling->execlists.virtual);
+ }
+
+ rb = NULL;
+ first = true;
+ parent = &sibling->execlists.virtual.rb_root.rb_node;
+ while (*parent) {
+ struct ve_node *other;
+
+ rb = *parent;
+ other = rb_entry(rb, typeof(*other), rb);
+ if (prio > other->prio) {
+ parent = &rb->rb_left;
+ } else {
+ parent = &rb->rb_right;
+ first = false;
+ }
+ }
+
+ rb_link_node(&node->rb, rb, parent);
+ rb_insert_color_cached(&node->rb,
+ &sibling->execlists.virtual,
+ first);
+
+submit_engine:
+ GEM_BUG_ON(RB_EMPTY_NODE(&node->rb));
+ node->prio = prio;
+ if (first && prio > sibling->execlists.queue_priority_hint) {
+ sibling->execlists.queue_priority_hint = prio;
+ tasklet_hi_schedule(&sibling->execlists.tasklet);
+ }
+
+ spin_unlock(&sibling->timeline.lock);
+ }
+ local_irq_enable();
+}
+
+static void virtual_submit_request(struct i915_request *rq)
+{
+ struct virtual_engine *ve = to_virtual_engine(rq->engine);
+
+ GEM_TRACE("%s: rq=%llx:%lld\n",
+ ve->base.name,
+ rq->fence.context,
+ rq->fence.seqno);
+
+ GEM_BUG_ON(ve->base.submit_request != virtual_submit_request);
+
+ GEM_BUG_ON(ve->request);
+ ve->base.execlists.queue_priority_hint = rq_prio(rq);
+ WRITE_ONCE(ve->request, rq);
+
+ tasklet_schedule(&ve->base.execlists.tasklet);
+}
+
+struct intel_context *
+intel_execlists_create_virtual(struct i915_gem_context *ctx,
+ struct intel_engine_cs **siblings,
+ unsigned int count)
+{
+ struct virtual_engine *ve;
+ unsigned int n;
+ int err;
+
+ if (count == 0)
+ return ERR_PTR(-EINVAL);
+
+ if (count == 1)
+ return intel_context_create(ctx, siblings[0]);
+
+ ve = kzalloc(struct_size(ve, siblings, count), GFP_KERNEL);
+ if (!ve)
+ return ERR_PTR(-ENOMEM);
+
+ ve->base.i915 = ctx->i915;
+ ve->base.id = -1;
+ ve->base.class = OTHER_CLASS;
+ ve->base.uabi_class = I915_ENGINE_CLASS_INVALID;
+ ve->base.instance = I915_ENGINE_CLASS_INVALID_VIRTUAL;
+ ve->base.flags = I915_ENGINE_IS_VIRTUAL;
+
+ snprintf(ve->base.name, sizeof(ve->base.name), "virtual");
+
+ err = i915_timeline_init(ctx->i915, &ve->base.timeline, NULL);
+ if (err)
+ goto err_put;
+ i915_timeline_set_subclass(&ve->base.timeline, TIMELINE_VIRTUAL);
+
+ intel_engine_init_execlists(&ve->base);
+
+ ve->base.cops = &virtual_context_ops;
+ ve->base.request_alloc = execlists_request_alloc;
+
+ ve->base.schedule = i915_schedule;
+ ve->base.submit_request = virtual_submit_request;
+
+ ve->base.execlists.queue_priority_hint = INT_MIN;
+ tasklet_init(&ve->base.execlists.tasklet,
+ virtual_submission_tasklet,
+ (unsigned long)ve);
+
+ intel_context_init(&ve->context, ctx, &ve->base);
+
+ for (n = 0; n < count; n++) {
+ struct intel_engine_cs *sibling = siblings[n];
+
+ GEM_BUG_ON(!is_power_of_2(sibling->mask));
+ if (sibling->mask & ve->base.mask) {
+ DRM_DEBUG("duplicate %s entry in load balancer\n",
+ sibling->name);
+ err = -EINVAL;
+ goto err_put;
+ }
+
+ /*
+ * The virtual engine implementation is tightly coupled to
+ * the execlists backend -- we push out request directly
+ * into a tree inside each physical engine. We could support
+ * layering if we handle cloning of the requests and
+ * submitting a copy into each backend.
+ */
+ if (sibling->execlists.tasklet.func !=
+ execlists_submission_tasklet) {
+ err = -ENODEV;
+ goto err_put;
+ }
+
+ GEM_BUG_ON(RB_EMPTY_NODE(&ve->nodes[sibling->id].rb));
+ RB_CLEAR_NODE(&ve->nodes[sibling->id].rb);
+
+ ve->siblings[ve->num_siblings++] = sibling;
+ ve->base.mask |= sibling->mask;
+
+ /*
+ * All physical engines must be compatible for their emission
+ * functions (as we build the instructions during request
+ * construction and do not alter them before submission
+ * on the physical engine). We use the engine class as a guide
+ * here, although that could be refined.
+ */
+ if (ve->base.class != OTHER_CLASS) {
+ if (ve->base.class != sibling->class) {
+ DRM_DEBUG("invalid mixing of engine class, sibling %d, already %d\n",
+ sibling->class, ve->base.class);
+ err = -EINVAL;
+ goto err_put;
+ }
+ continue;
+ }
+
+ ve->base.class = sibling->class;
+ ve->base.uabi_class = sibling->uabi_class;
+ snprintf(ve->base.name, sizeof(ve->base.name),
+ "v%dx%d", ve->base.class, count);
+ ve->base.context_size = sibling->context_size;
+
+ ve->base.emit_bb_start = sibling->emit_bb_start;
+ ve->base.emit_flush = sibling->emit_flush;
+ ve->base.emit_init_breadcrumb = sibling->emit_init_breadcrumb;
+ ve->base.emit_fini_breadcrumb = sibling->emit_fini_breadcrumb;
+ ve->base.emit_fini_breadcrumb_dw =
+ sibling->emit_fini_breadcrumb_dw;
+ }
+
+ return &ve->context;
+
+err_put:
+ intel_context_put(&ve->context);
+ return ERR_PTR(err);
+}
+
+struct intel_context *
+intel_execlists_clone_virtual(struct i915_gem_context *ctx,
+ struct intel_engine_cs *src)
+{
+ struct virtual_engine *se = to_virtual_engine(src);
+ struct intel_context *dst;
+
+ dst = intel_execlists_create_virtual(ctx,
+ se->siblings,
+ se->num_siblings);
+ if (IS_ERR(dst))
+ return dst;
+
+ return dst;
+}
+
void intel_execlists_show_requests(struct intel_engine_cs *engine,
struct drm_printer *m,
void (*show_request)(struct drm_printer *m,
@@ -2810,6 +3428,29 @@ void intel_execlists_show_requests(struct intel_engine_cs *engine,
show_request(m, last, "\t\tQ ");
}
+ last = NULL;
+ count = 0;
+ for (rb = rb_first_cached(&execlists->virtual); rb; rb = rb_next(rb)) {
+ struct virtual_engine *ve =
+ rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
+ struct i915_request *rq = READ_ONCE(ve->request);
+
+ if (rq) {
+ if (count++ < max - 1)
+ show_request(m, rq, "\t\tV ");
+ else
+ last = rq;
+ }
+ }
+ if (last) {
+ if (count > max) {
+ drm_printf(m,
+ "\t\t...skipping %d virtual requests...\n",
+ count - max);
+ }
+ show_request(m, last, "\t\tV ");
+ }
+
spin_unlock_irqrestore(&engine->timeline.lock, flags);
}
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.h b/drivers/gpu/drm/i915/gt/intel_lrc.h
index a0dc907a7249..5530606052e5 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.h
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.h
@@ -114,4 +114,13 @@ void intel_execlists_show_requests(struct intel_engine_cs *engine,
const char *prefix),
unsigned int max);
+struct intel_context *
+intel_execlists_create_virtual(struct i915_gem_context *ctx,
+ struct intel_engine_cs **siblings,
+ unsigned int count);
+
+struct intel_context *
+intel_execlists_clone_virtual(struct i915_gem_context *ctx,
+ struct intel_engine_cs *src);
+
#endif /* _INTEL_LRC_H_ */
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index 5b3d8e33f1cf..f880271fb9ba 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -1310,6 +1310,185 @@ static int live_preempt_smoke(void *arg)
return err;
}
+static int nop_virtual_engine(struct drm_i915_private *i915,
+ struct intel_engine_cs **siblings,
+ unsigned int nsibling,
+ unsigned int nctx,
+ unsigned int flags)
+#define CHAIN BIT(0)
+{
+ IGT_TIMEOUT(end_time);
+ struct i915_request *request[16];
+ struct i915_gem_context *ctx[16];
+ struct intel_context *ve[16];
+ unsigned long n, prime, nc;
+ struct igt_live_test t;
+ ktime_t times[2] = {};
+ int err;
+
+ GEM_BUG_ON(!nctx || nctx > ARRAY_SIZE(ctx));
+
+ for (n = 0; n < nctx; n++) {
+ ctx[n] = kernel_context(i915);
+ if (!ctx[n]) {
+ err = -ENOMEM;
+ nctx = n;
+ goto out;
+ }
+
+ ve[n] = intel_execlists_create_virtual(ctx[n],
+ siblings, nsibling);
+ if (IS_ERR(ve[n])) {
+ kernel_context_close(ctx[n]);
+ err = PTR_ERR(ve[n]);
+ nctx = n;
+ goto out;
+ }
+
+ err = intel_context_pin(ve[n]);
+ if (err) {
+ intel_context_put(ve[n]);
+ kernel_context_close(ctx[n]);
+ nctx = n;
+ goto out;
+ }
+ }
+
+ err = igt_live_test_begin(&t, i915, __func__, ve[0]->engine->name);
+ if (err)
+ goto out;
+
+ for_each_prime_number_from(prime, 1, 8192) {
+ times[1] = ktime_get_raw();
+
+ if (flags & CHAIN) {
+ for (nc = 0; nc < nctx; nc++) {
+ for (n = 0; n < prime; n++) {
+ request[nc] =
+ i915_request_create(ve[nc]);
+ if (IS_ERR(request[nc])) {
+ err = PTR_ERR(request[nc]);
+ goto out;
+ }
+
+ i915_request_add(request[nc]);
+ }
+ }
+ } else {
+ for (n = 0; n < prime; n++) {
+ for (nc = 0; nc < nctx; nc++) {
+ request[nc] =
+ i915_request_create(ve[nc]);
+ if (IS_ERR(request[nc])) {
+ err = PTR_ERR(request[nc]);
+ goto out;
+ }
+
+ i915_request_add(request[nc]);
+ }
+ }
+ }
+
+ for (nc = 0; nc < nctx; nc++) {
+ if (i915_request_wait(request[nc],
+ I915_WAIT_LOCKED,
+ HZ / 10) < 0) {
+ pr_err("%s(%s): wait for %llx:%lld timed out\n",
+ __func__, ve[0]->engine->name,
+ request[nc]->fence.context,
+ request[nc]->fence.seqno);
+
+ GEM_TRACE("%s(%s) failed at request %llx:%lld\n",
+ __func__, ve[0]->engine->name,
+ request[nc]->fence.context,
+ request[nc]->fence.seqno);
+ GEM_TRACE_DUMP();
+ i915_gem_set_wedged(i915);
+ break;
+ }
+ }
+
+ times[1] = ktime_sub(ktime_get_raw(), times[1]);
+ if (prime == 1)
+ times[0] = times[1];
+
+ if (__igt_timeout(end_time, NULL))
+ break;
+ }
+
+ err = igt_live_test_end(&t);
+ if (err)
+ goto out;
+
+ pr_info("Requestx%d latencies on %s: 1 = %lluns, %lu = %lluns\n",
+ nctx, ve[0]->engine->name, ktime_to_ns(times[0]),
+ prime, div64_u64(ktime_to_ns(times[1]), prime));
+
+out:
+ if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ err = -EIO;
+
+ for (nc = 0; nc < nctx; nc++) {
+ intel_context_unpin(ve[nc]);
+ intel_context_put(ve[nc]);
+ kernel_context_close(ctx[nc]);
+ }
+ return err;
+}
+
+static int live_virtual_engine(void *arg)
+{
+ struct drm_i915_private *i915 = arg;
+ struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1];
+ struct intel_engine_cs *engine;
+ enum intel_engine_id id;
+ unsigned int class, inst;
+ int err = -ENODEV;
+
+ if (USES_GUC_SUBMISSION(i915))
+ return 0;
+
+ mutex_lock(&i915->drm.struct_mutex);
+
+ for_each_engine(engine, i915, id) {
+ err = nop_virtual_engine(i915, &engine, 1, 1, 0);
+ if (err) {
+ pr_err("Failed to wrap engine %s: err=%d\n",
+ engine->name, err);
+ goto out_unlock;
+ }
+ }
+
+ for (class = 0; class <= MAX_ENGINE_CLASS; class++) {
+ int nsibling, n;
+
+ nsibling = 0;
+ for (inst = 0; inst <= MAX_ENGINE_INSTANCE; inst++) {
+ if (!i915->engine_class[class][inst])
+ continue;
+
+ siblings[nsibling++] = i915->engine_class[class][inst];
+ }
+ if (nsibling < 2)
+ continue;
+
+ for (n = 1; n <= nsibling + 1; n++) {
+ err = nop_virtual_engine(i915, siblings, nsibling,
+ n, 0);
+ if (err)
+ goto out_unlock;
+ }
+
+ err = nop_virtual_engine(i915, siblings, nsibling, n, CHAIN);
+ if (err)
+ goto out_unlock;
+ }
+
+out_unlock:
+ mutex_unlock(&i915->drm.struct_mutex);
+ return err;
+}
+
int intel_execlists_live_selftests(struct drm_i915_private *i915)
{
static const struct i915_subtest tests[] = {
@@ -1322,6 +1501,7 @@ int intel_execlists_live_selftests(struct drm_i915_private *i915)
SUBTEST(live_chain_preempt),
SUBTEST(live_preempt_hang),
SUBTEST(live_preempt_smoke),
+ SUBTEST(live_virtual_engine),
};
if (!HAS_EXECLISTS(i915))
diff --git a/drivers/gpu/drm/i915/i915_gem.h b/drivers/gpu/drm/i915/i915_gem.h
index 67f8a4a807a0..fe82d3571072 100644
--- a/drivers/gpu/drm/i915/i915_gem.h
+++ b/drivers/gpu/drm/i915/i915_gem.h
@@ -91,4 +91,9 @@ static inline bool __tasklet_enable(struct tasklet_struct *t)
return atomic_dec_and_test(&t->count);
}
+static inline bool __tasklet_is_scheduled(struct tasklet_struct *t)
+{
+ return test_bit(TASKLET_STATE_SCHED, &t->state);
+}
+
#endif /* __I915_GEM_H__ */
diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
index 040c0b770c21..3ae1d27130a8 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/i915_gem_context.c
@@ -86,6 +86,7 @@
*/
#include <linux/log2.h>
+#include <linux/nospec.h>
#include <drm/i915_drm.h>
@@ -1218,7 +1219,6 @@ __intel_context_reconfigure_sseu(struct intel_context *ce,
int ret;
GEM_BUG_ON(INTEL_GEN(ce->gem_context->i915) < 8);
- GEM_BUG_ON(ce->engine->id != RCS0);
ret = intel_context_lock_pinned(ce);
if (ret)
@@ -1412,7 +1412,100 @@ struct set_engines {
struct i915_gem_engines *engines;
};
+static int
+set_engines__load_balance(struct i915_user_extension __user *base, void *data)
+{
+ struct i915_context_engines_load_balance __user *ext =
+ container_of_user(base, typeof(*ext), base);
+ const struct set_engines *set = data;
+ struct intel_engine_cs *stack[16];
+ struct intel_engine_cs **siblings;
+ struct intel_context *ce;
+ u16 num_siblings, idx;
+ unsigned int n;
+ int err;
+
+ if (!HAS_EXECLISTS(set->ctx->i915))
+ return -ENODEV;
+
+ if (USES_GUC_SUBMISSION(set->ctx->i915))
+ return -ENODEV; /* not implement yet */
+
+ if (get_user(idx, &ext->engine_index))
+ return -EFAULT;
+
+ if (idx >= set->engines->num_engines) {
+ DRM_DEBUG("Invalid placement value, %d >= %d\n",
+ idx, set->engines->num_engines);
+ return -EINVAL;
+ }
+
+ idx = array_index_nospec(idx, set->engines->num_engines);
+ if (set->engines->engines[idx]) {
+ DRM_DEBUG("Invalid placement[%d], already occupied\n", idx);
+ return -EEXIST;
+ }
+
+ if (get_user(num_siblings, &ext->num_siblings))
+ return -EFAULT;
+
+ err = check_user_mbz(&ext->flags);
+ if (err)
+ return err;
+
+ err = check_user_mbz(&ext->mbz64);
+ if (err)
+ return err;
+
+ siblings = stack;
+ if (num_siblings > ARRAY_SIZE(stack)) {
+ siblings = kmalloc_array(num_siblings,
+ sizeof(*siblings),
+ GFP_KERNEL);
+ if (!siblings)
+ return -ENOMEM;
+ }
+
+ for (n = 0; n < num_siblings; n++) {
+ struct i915_engine_class_instance ci;
+
+ if (copy_from_user(&ci, &ext->engines[n], sizeof(ci))) {
+ err = -EFAULT;
+ goto out_siblings;
+ }
+
+ siblings[n] = intel_engine_lookup_user(set->ctx->i915,
+ ci.engine_class,
+ ci.engine_instance);
+ if (!siblings[n]) {
+ DRM_DEBUG("Invalid sibling[%d]: { class:%d, inst:%d }\n",
+ n, ci.engine_class, ci.engine_instance);
+ err = -EINVAL;
+ goto out_siblings;
+ }
+ }
+
+ ce = intel_execlists_create_virtual(set->ctx, siblings, n);
+ if (IS_ERR(ce)) {
+ err = PTR_ERR(ce);
+ goto out_siblings;
+ }
+
+ if (cmpxchg(&set->engines->engines[idx], NULL, ce)) {
+ intel_context_put(ce);
+ err = -EEXIST;
+ goto out_siblings;
+ }
+
+out_siblings:
+ if (siblings != stack)
+ kfree(siblings);
+
+ return err;
+}
+
static const i915_user_extension_fn set_engines__extensions[] = {
+ [I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE] = set_engines__load_balance,
};
static int
@@ -1711,14 +1804,29 @@ static int clone_engines(struct i915_gem_context *dst,
clone->i915 = dst->i915;
for (n = 0; n < e->num_engines; n++) {
+ struct intel_engine_cs *engine;
+
if (!e->engines[n]) {
clone->engines[n] = NULL;
continue;
}
+ engine = e->engines[n]->engine;
- clone->engines[n] =
- intel_context_create(dst, e->engines[n]->engine);
- if (!clone->engines[n]) {
+ /*
+ * Virtual engines are singletons; they can only exist
+ * inside a single context, because they embed their
+ * HW context... As each virtual context implies a single
+ * timeline (each engine can only dequeue a single request
+ * at any time), it would be surprising for two contexts
+ * to use the same engine. So let's create a copy of
+ * the virtual engine instead.
+ */
+ if (intel_engine_is_virtual(engine))
+ clone->engines[n] =
+ intel_execlists_clone_virtual(dst, engine);
+ else
+ clone->engines[n] = intel_context_create(dst, engine);
+ if (IS_ERR_OR_NULL(clone->engines[n])) {
__free_engines(clone, n);
goto err_unlock;
}
diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c
index d215dcdf0b1e..78ceb56d7801 100644
--- a/drivers/gpu/drm/i915/i915_scheduler.c
+++ b/drivers/gpu/drm/i915/i915_scheduler.c
@@ -150,17 +150,26 @@ sched_lock_engine(const struct i915_sched_node *node,
struct intel_engine_cs *locked,
struct sched_cache *cache)
{
- struct intel_engine_cs *engine = node_to_request(node)->engine;
+ const struct i915_request *rq = node_to_request(node);
+ struct intel_engine_cs *engine;
GEM_BUG_ON(!locked);
- if (engine != locked) {
+ /*
+ * Virtual engines complicate acquiring the engine timeline lock,
+ * as their rq->engine pointer is not stable until under that
+ * engine lock. The simple ploy we use is to take the lock then
+ * check that the rq still belongs to the newly locked engine.
+ */
+ while (locked != (engine = READ_ONCE(rq->engine))) {
spin_unlock(&locked->timeline.lock);
memset(cache, 0, sizeof(*cache));
spin_lock(&engine->timeline.lock);
+ locked = engine;
}
- return engine;
+ GEM_BUG_ON(locked != engine);
+ return locked;
}
static inline int rq_prio(const struct i915_request *rq)
@@ -272,6 +281,7 @@ static void __i915_schedule(struct i915_sched_node *node,
spin_lock(&engine->timeline.lock);
/* Fifo and depth-first replacement ensure our deps execute before us */
+ engine = sched_lock_engine(node, engine, &cache);
list_for_each_entry_safe_reverse(dep, p, &dfs, dfs_link) {
INIT_LIST_HEAD(&dep->dfs_link);
@@ -283,8 +293,11 @@ static void __i915_schedule(struct i915_sched_node *node,
if (prio <= node->attr.priority || node_signaled(node))
continue;
+ GEM_BUG_ON(node_to_request(node)->engine != engine);
+
node->attr.priority = prio;
if (!list_empty(&node->link)) {
+ GEM_BUG_ON(intel_engine_is_virtual(engine));
if (!cache.priolist)
cache.priolist =
i915_sched_lookup_priolist(engine,
diff --git a/drivers/gpu/drm/i915/i915_timeline_types.h b/drivers/gpu/drm/i915/i915_timeline_types.h
index 5256a0b5c5f7..1688705f4a2b 100644
--- a/drivers/gpu/drm/i915/i915_timeline_types.h
+++ b/drivers/gpu/drm/i915/i915_timeline_types.h
@@ -26,6 +26,7 @@ struct i915_timeline {
spinlock_t lock;
#define TIMELINE_CLIENT 0 /* default subclass */
#define TIMELINE_ENGINE 1
+#define TIMELINE_VIRTUAL 2
struct mutex mutex; /* protects the flow of requests */
unsigned int pin_count;
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index 62396d575e28..f9770948161c 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -137,6 +137,7 @@ struct i915_engine_class_instance {
__u16 engine_class; /* see enum drm_i915_gem_engine_class */
__u16 engine_instance;
#define I915_ENGINE_CLASS_INVALID_NONE -1
+#define I915_ENGINE_CLASS_INVALID_VIRTUAL -2
};
/**
@@ -1608,8 +1609,46 @@ struct drm_i915_gem_context_param_sseu {
__u32 rsvd;
};
+/*
+ * i915_context_engines_load_balance:
+ *
+ * Enable load balancing across this set of engines.
+ *
+ * Into the I915_EXEC_DEFAULT slot [0], a virtual engine is created that when
+ * used will proxy the execbuffer request onto one of the set of engines
+ * in such a way as to distribute the load evenly across the set.
+ *
+ * The set of engines must be compatible (e.g. the same HW class) as they
+ * will share the same logical GPU context and ring.
+ *
+ * To intermix rendering with the virtual engine and direct rendering onto
+ * the backing engines (bypassing the load balancing proxy), the context must
+ * be defined to use a single timeline for all engines.
+ */
+struct i915_context_engines_load_balance {
+ struct i915_user_extension base;
+
+ __u16 engine_index;
+ __u16 num_siblings;
+ __u32 flags; /* all undefined flags must be zero */
+
+ __u64 mbz64; /* reserved for future use; must be zero */
+
+ struct i915_engine_class_instance engines[0];
+} __attribute__((packed));
+
+#define I915_DEFINE_CONTEXT_ENGINES_LOAD_BALANCE(name__, N__) struct { \
+ struct i915_user_extension base; \
+ __u16 engine_index; \
+ __u16 num_siblings; \
+ __u32 flags; \
+ __u64 mbz64; \
+ struct i915_engine_class_instance engines[N__]; \
+} __attribute__((packed)) name__
+
struct i915_context_param_engines {
__u64 extensions; /* linked chain of extension blocks, 0 terminates */
+#define I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE 0 /* see i915_context_engines_load_balance */
struct i915_engine_class_instance engines[0];
} __attribute__((packed));
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* Re: [PATCH] drm/i915: Load balancing across a virtual engine
2019-05-08 11:23 ` [PATCH] " Chris Wilson
@ 2019-05-08 11:35 ` Tvrtko Ursulin
0 siblings, 0 replies; 71+ messages in thread
From: Tvrtko Ursulin @ 2019-05-08 11:35 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 08/05/2019 12:23, Chris Wilson wrote:
> Having allowed the user to define a set of engines that they will want
> to only use, we go one step further and allow them to bind those engines
> into a single virtual instance. Submitting a batch to the virtual engine
> will then forward it to any one of the set in a manner as best to
> distribute load. The virtual engine has a single timeline across all
> engines (it operates as a single queue), so it is not able to concurrently
> run batches across multiple engines by itself; that is left up to the user
> to submit multiple concurrent batches to multiple queues. Multiple users
> will be load balanced across the system.
>
> The mechanism used for load balancing in this patch is a late greedy
> balancer. When a request is ready for execution, it is added to each
> engine's queue, and when an engine is ready for its next request it
> claims it from the virtual engine. The first engine to do so, wins, i.e.
> the request is executed at the earliest opportunity (idle moment) in the
> system.
>
> As not all HW is created equal, the user is still able to skip the
> virtual engine and execute the batch on a specific engine, all within the
> same queue. It will then be executed in order on the correct engine,
> with execution on other virtual engines being moved away due to the load
> detection.
>
> A couple of areas for potential improvement left!
>
> - The virtual engine always take priority over equal-priority tasks.
> Mostly broken up by applying FQ_CODEL rules for prioritising new clients,
> and hopefully the virtual and real engines are not then congested (i.e.
> all work is via virtual engines, or all work is to the real engine).
>
> - We require the breadcrumb irq around every virtual engine request. For
> normal engines, we eliminate the need for the slow round trip via
> interrupt by using the submit fence and queueing in order. For virtual
> engines, we have to allow any job to transfer to a new ring, and cannot
> coalesce the submissions, so require the completion fence instead,
> forcing the persistent use of interrupts.
>
> - We only drip feed single requests through each virtual engine and onto
> the physical engines, even if there was enough work to fill all ELSP,
> leaving small stalls with an idle CS event at the end of every request.
> Could we be greedy and fill both slots? Being lazy is virtuous for load
> distribution on less-than-full workloads though.
>
> Other areas of improvement are more general, such as reducing lock
> contention, reducing dispatch overhead, looking at direct submission
> rather than bouncing around tasklets etc.
>
> sseu: Lift the restriction to allow sseu to be reconfigured on virtual
> engines composed of RENDER_CLASS (rcs).
>
> v2: macroize check_user_mbz()
> v3: Cancel virtual engines on wedging
> v4: Commence commenting
> v5: Replace 64b sibling_mask with a list of class:instance
> v6: Drop the one-element array in the uabi
> v7: Assert it is an virtual engine in to_virtual_engine()
> v8: Skip over holes in [class][inst] so we can selftest with (vcs0, vcs2)
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> ---
> drivers/gpu/drm/i915/gt/intel_engine_types.h | 8 +
> drivers/gpu/drm/i915/gt/intel_lrc.c | 683 ++++++++++++++++++-
> drivers/gpu/drm/i915/gt/intel_lrc.h | 9 +
> drivers/gpu/drm/i915/gt/selftest_lrc.c | 180 +++++
> drivers/gpu/drm/i915/i915_gem.h | 5 +
> drivers/gpu/drm/i915/i915_gem_context.c | 116 +++-
> drivers/gpu/drm/i915/i915_scheduler.c | 19 +-
> drivers/gpu/drm/i915/i915_timeline_types.h | 1 +
> include/uapi/drm/i915_drm.h | 39 ++
> 9 files changed, 1032 insertions(+), 28 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
> index e381c1c73902..7b47e00fa082 100644
> --- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
> +++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
> @@ -227,6 +227,7 @@ struct intel_engine_execlists {
> * @queue: queue of requests, in priority lists
> */
> struct rb_root_cached queue;
> + struct rb_root_cached virtual;
>
> /**
> * @csb_write: control register for Context Switch buffer
> @@ -445,6 +446,7 @@ struct intel_engine_cs {
> #define I915_ENGINE_HAS_PREEMPTION BIT(2)
> #define I915_ENGINE_HAS_SEMAPHORES BIT(3)
> #define I915_ENGINE_NEEDS_BREADCRUMB_TASKLET BIT(4)
> +#define I915_ENGINE_IS_VIRTUAL BIT(5)
> unsigned int flags;
>
> /*
> @@ -534,6 +536,12 @@ intel_engine_needs_breadcrumb_tasklet(const struct intel_engine_cs *engine)
> return engine->flags & I915_ENGINE_NEEDS_BREADCRUMB_TASKLET;
> }
>
> +static inline bool
> +intel_engine_is_virtual(const struct intel_engine_cs *engine)
> +{
> + return engine->flags & I915_ENGINE_IS_VIRTUAL;
> +}
> +
> #define instdone_slice_mask(dev_priv__) \
> (IS_GEN(dev_priv__, 7) ? \
> 1 : RUNTIME_INFO(dev_priv__)->sseu.slice_mask)
> diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
> index f1d62746e066..bc388df39802 100644
> --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
> +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
> @@ -166,6 +166,42 @@
>
> #define ACTIVE_PRIORITY (I915_PRIORITY_NEWCLIENT | I915_PRIORITY_NOSEMAPHORE)
>
> +struct virtual_engine {
> + struct intel_engine_cs base;
> + struct intel_context context;
> +
> + /*
> + * We allow only a single request through the virtual engine at a time
> + * (each request in the timeline waits for the completion fence of
> + * the previous before being submitted). By restricting ourselves to
> + * only submitting a single request, each request is placed on to a
> + * physical to maximise load spreading (by virtue of the late greedy
> + * scheduling -- each real engine takes the next available request
> + * upon idling).
> + */
> + struct i915_request *request;
> +
> + /*
> + * We keep a rbtree of available virtual engines inside each physical
> + * engine, sorted by priority. Here we preallocate the nodes we need
> + * for the virtual engine, indexed by physical_engine->id.
> + */
> + struct ve_node {
> + struct rb_node rb;
> + int prio;
> + } nodes[I915_NUM_ENGINES];
> +
> + /* And finally, which physical engines this virtual engine maps onto. */
> + unsigned int num_siblings;
> + struct intel_engine_cs *siblings[0];
> +};
> +
> +static struct virtual_engine *to_virtual_engine(struct intel_engine_cs *engine)
> +{
> + GEM_BUG_ON(!intel_engine_is_virtual(engine));
> + return container_of(engine, struct virtual_engine, base);
> +}
> +
> static int execlists_context_deferred_alloc(struct intel_context *ce,
> struct intel_engine_cs *engine);
> static void execlists_init_reg_state(u32 *reg_state,
> @@ -229,7 +265,8 @@ static int queue_prio(const struct intel_engine_execlists *execlists)
> }
>
> static inline bool need_preempt(const struct intel_engine_cs *engine,
> - const struct i915_request *rq)
> + const struct i915_request *rq,
> + struct rb_node *rb)
> {
> int last_prio;
>
> @@ -264,6 +301,25 @@ static inline bool need_preempt(const struct intel_engine_cs *engine,
> rq_prio(list_next_entry(rq, link)) > last_prio)
> return true;
>
> + if (rb) {
> + struct virtual_engine *ve =
> + rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
> + bool preempt = false;
> +
> + if (engine == ve->siblings[0]) { /* only preempt one sibling */
> + struct i915_request *next;
> +
> + rcu_read_lock();
> + next = READ_ONCE(ve->request);
> + if (next)
> + preempt = rq_prio(next) > last_prio;
> + rcu_read_unlock();
> + }
> +
> + if (preempt)
> + return preempt;
> + }
> +
> /*
> * If the inflight context did not trigger the preemption, then maybe
> * it was the set of queued requests? Pick the highest priority in
> @@ -382,6 +438,8 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
> list_for_each_entry_safe_reverse(rq, rn,
> &engine->timeline.requests,
> link) {
> + struct intel_engine_cs *owner;
> +
> if (i915_request_completed(rq))
> break;
>
> @@ -390,16 +448,32 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
>
> GEM_BUG_ON(rq->hw_context->active);
>
> - GEM_BUG_ON(rq_prio(rq) == I915_PRIORITY_INVALID);
> - if (rq_prio(rq) != prio) {
> - prio = rq_prio(rq);
> - pl = i915_sched_lookup_priolist(engine, prio);
> - }
> - GEM_BUG_ON(RB_EMPTY_ROOT(&engine->execlists.queue.rb_root));
> + /*
> + * Push the request back into the queue for later resubmission.
> + * If this request is not native to this physical engine (i.e.
> + * it came from a virtual source), push it back onto the virtual
> + * engine so that it can be moved across onto another physical
> + * engine as load dictates.
> + */
> + owner = rq->hw_context->engine;
> + if (likely(owner == engine)) {
> + GEM_BUG_ON(rq_prio(rq) == I915_PRIORITY_INVALID);
> + if (rq_prio(rq) != prio) {
> + prio = rq_prio(rq);
> + pl = i915_sched_lookup_priolist(engine, prio);
> + }
> + GEM_BUG_ON(RB_EMPTY_ROOT(&engine->execlists.queue.rb_root));
>
> - list_add(&rq->sched.link, pl);
> + list_add(&rq->sched.link, pl);
> + active = rq;
> + } else {
> + if (__i915_request_has_started(rq))
> + rq->sched.attr.priority |= boost;
>
> - active = rq;
> + rq->engine = owner;
> + owner->submit_request(rq);
> + active = NULL;
> + }
> }
>
> /*
> @@ -419,7 +493,7 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
> * in the priority queue, but they will not gain immediate access to
> * the GPU.
> */
> - if (~prio & boost && __i915_request_has_started(active)) {
> + if (~prio & boost && active && __i915_request_has_started(active)) {
> prio |= boost;
> GEM_BUG_ON(active->sched.attr.priority >= prio);
> active->sched.attr.priority = prio;
> @@ -661,6 +735,90 @@ static void complete_preempt_context(struct intel_engine_execlists *execlists)
> ACTIVE_PRIORITY);
> }
>
> +static void virtual_update_register_offsets(u32 *regs,
> + struct intel_engine_cs *engine)
> +{
> + u32 base = engine->mmio_base;
> +
> + /* Must match execlists_init_reg_state()! */
> +
> + regs[CTX_CONTEXT_CONTROL] =
> + i915_mmio_reg_offset(RING_CONTEXT_CONTROL(base));
> + regs[CTX_RING_HEAD] = i915_mmio_reg_offset(RING_HEAD(base));
> + regs[CTX_RING_TAIL] = i915_mmio_reg_offset(RING_TAIL(base));
> + regs[CTX_RING_BUFFER_START] = i915_mmio_reg_offset(RING_START(base));
> + regs[CTX_RING_BUFFER_CONTROL] = i915_mmio_reg_offset(RING_CTL(base));
> +
> + regs[CTX_BB_HEAD_U] = i915_mmio_reg_offset(RING_BBADDR_UDW(base));
> + regs[CTX_BB_HEAD_L] = i915_mmio_reg_offset(RING_BBADDR(base));
> + regs[CTX_BB_STATE] = i915_mmio_reg_offset(RING_BBSTATE(base));
> + regs[CTX_SECOND_BB_HEAD_U] =
> + i915_mmio_reg_offset(RING_SBBADDR_UDW(base));
> + regs[CTX_SECOND_BB_HEAD_L] = i915_mmio_reg_offset(RING_SBBADDR(base));
> + regs[CTX_SECOND_BB_STATE] = i915_mmio_reg_offset(RING_SBBSTATE(base));
> +
> + regs[CTX_CTX_TIMESTAMP] =
> + i915_mmio_reg_offset(RING_CTX_TIMESTAMP(base));
> + regs[CTX_PDP3_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 3));
> + regs[CTX_PDP3_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 3));
> + regs[CTX_PDP2_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 2));
> + regs[CTX_PDP2_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 2));
> + regs[CTX_PDP1_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 1));
> + regs[CTX_PDP1_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 1));
> + regs[CTX_PDP0_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 0));
> + regs[CTX_PDP0_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 0));
> +
> + if (engine->class == RENDER_CLASS) {
> + regs[CTX_RCS_INDIRECT_CTX] =
> + i915_mmio_reg_offset(RING_INDIRECT_CTX(base));
> + regs[CTX_RCS_INDIRECT_CTX_OFFSET] =
> + i915_mmio_reg_offset(RING_INDIRECT_CTX_OFFSET(base));
> + regs[CTX_BB_PER_CTX_PTR] =
> + i915_mmio_reg_offset(RING_BB_PER_CTX_PTR(base));
> +
> + regs[CTX_R_PWR_CLK_STATE] =
> + i915_mmio_reg_offset(GEN8_R_PWR_CLK_STATE);
> + }
> +}
> +
> +static bool virtual_matches(const struct virtual_engine *ve,
> + const struct i915_request *rq,
> + const struct intel_engine_cs *engine)
> +{
> + const struct intel_engine_cs *active;
> +
> + /*
> + * We track when the HW has completed saving the context image
> + * (i.e. when we have seen the final CS event switching out of
> + * the context) and must not overwrite the context image before
> + * then. This restricts us to only using the active engine
> + * while the previous virtualized request is inflight (so
> + * we reuse the register offsets). This is a very small
> + * hystersis on the greedy seelction algorithm.
> + */
> + active = READ_ONCE(ve->context.active);
> + if (active && active != engine)
> + return false;
> +
> + return true;
> +}
> +
> +static void virtual_xfer_breadcrumbs(struct virtual_engine *ve,
> + struct intel_engine_cs *engine)
> +{
> + struct intel_engine_cs *old = ve->siblings[0];
> +
> + /* All unattached (rq->engine == old) must already be completed */
> +
> + spin_lock(&old->breadcrumbs.irq_lock);
> + if (!list_empty(&ve->context.signal_link)) {
> + list_move_tail(&ve->context.signal_link,
> + &engine->breadcrumbs.signalers);
> + intel_engine_queue_breadcrumbs(engine);
> + }
> + spin_unlock(&old->breadcrumbs.irq_lock);
> +}
> +
> static void execlists_dequeue(struct intel_engine_cs *engine)
> {
> struct intel_engine_execlists * const execlists = &engine->execlists;
> @@ -693,6 +851,26 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
> * and context switches) submission.
> */
>
> + for (rb = rb_first_cached(&execlists->virtual); rb; ) {
> + struct virtual_engine *ve =
> + rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
> + struct i915_request *rq = READ_ONCE(ve->request);
> +
> + if (!rq) { /* lazily cleanup after another engine handled rq */
> + rb_erase_cached(rb, &execlists->virtual);
> + RB_CLEAR_NODE(rb);
> + rb = rb_first_cached(&execlists->virtual);
> + continue;
> + }
> +
> + if (!virtual_matches(ve, rq, engine)) {
> + rb = rb_next(rb);
> + continue;
> + }
> +
> + break;
> + }
> +
> if (last) {
> /*
> * Don't resubmit or switch until all outstanding
> @@ -714,7 +892,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
> if (!execlists_is_active(execlists, EXECLISTS_ACTIVE_HWACK))
> return;
>
> - if (need_preempt(engine, last)) {
> + if (need_preempt(engine, last, rb)) {
> inject_preempt_context(engine);
> return;
> }
> @@ -754,6 +932,92 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
> last->tail = last->wa_tail;
> }
>
> + while (rb) { /* XXX virtual is always taking precedence */
> + struct virtual_engine *ve =
> + rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
> + struct i915_request *rq;
> +
> + spin_lock(&ve->base.timeline.lock);
> +
> + rq = ve->request;
> + if (unlikely(!rq)) { /* lost the race to a sibling */
> + spin_unlock(&ve->base.timeline.lock);
> + rb_erase_cached(rb, &execlists->virtual);
> + RB_CLEAR_NODE(rb);
> + rb = rb_first_cached(&execlists->virtual);
> + continue;
> + }
> +
> + GEM_BUG_ON(rq != ve->request);
> + GEM_BUG_ON(rq->engine != &ve->base);
> + GEM_BUG_ON(rq->hw_context != &ve->context);
> +
> + if (rq_prio(rq) >= queue_prio(execlists)) {
> + if (!virtual_matches(ve, rq, engine)) {
> + spin_unlock(&ve->base.timeline.lock);
> + rb = rb_next(rb);
> + continue;
> + }
> +
> + if (last && !can_merge_rq(last, rq)) {
> + spin_unlock(&ve->base.timeline.lock);
> + return; /* leave this rq for another engine */
> + }
> +
> + GEM_TRACE("%s: virtual rq=%llx:%lld%s, new engine? %s\n",
> + engine->name,
> + rq->fence.context,
> + rq->fence.seqno,
> + i915_request_completed(rq) ? "!" :
> + i915_request_started(rq) ? "*" :
> + "",
> + yesno(engine != ve->siblings[0]));
> +
> + ve->request = NULL;
> + ve->base.execlists.queue_priority_hint = INT_MIN;
> + rb_erase_cached(rb, &execlists->virtual);
> + RB_CLEAR_NODE(rb);
> +
> + rq->engine = engine;
> +
> + if (engine != ve->siblings[0]) {
> + u32 *regs = ve->context.lrc_reg_state;
> + unsigned int n;
> +
> + GEM_BUG_ON(READ_ONCE(ve->context.active));
> + virtual_update_register_offsets(regs, engine);
> +
> + if (!list_empty(&ve->context.signals))
> + virtual_xfer_breadcrumbs(ve, engine);
> +
> + /*
> + * Move the bound engine to the top of the list
> + * for future execution. We then kick this
> + * tasklet first before checking others, so that
> + * we preferentially reuse this set of bound
> + * registers.
> + */
> + for (n = 1; n < ve->num_siblings; n++) {
> + if (ve->siblings[n] == engine) {
> + swap(ve->siblings[n],
> + ve->siblings[0]);
> + break;
> + }
> + }
> +
> + GEM_BUG_ON(ve->siblings[0] != engine);
> + }
> +
> + __i915_request_submit(rq);
> + trace_i915_request_in(rq, port_index(port, execlists));
> + submit = true;
> + last = rq;
> + }
> +
> + spin_unlock(&ve->base.timeline.lock);
> + break;
> + }
> +
> while ((rb = rb_first_cached(&execlists->queue))) {
> struct i915_priolist *p = to_priolist(rb);
> struct i915_request *rq, *rn;
> @@ -1877,6 +2141,25 @@ static void reset_csb_pointers(struct intel_engine_execlists *execlists)
> &execlists->csb_status[reset_value]);
> }
>
> +static struct i915_request *active_request(struct i915_request *rq)
> +{
> + const struct list_head * const list = &rq->engine->timeline.requests;
> + const struct intel_context * const context = rq->hw_context;
> + struct i915_request *active = NULL;
> +
> + list_for_each_entry_from_reverse(rq, list, link) {
> + if (i915_request_completed(rq))
> + break;
> +
> + if (rq->hw_context != context)
> + break;
> +
> + active = rq;
> + }
> +
> + return active;
> +}
> +
> static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
> {
> struct intel_engine_execlists * const execlists = &engine->execlists;
> @@ -1897,7 +2180,8 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
> if (!port_isset(execlists->port))
> goto out_clear;
>
> - ce = port_request(execlists->port)->hw_context;
> + rq = port_request(execlists->port);
> + ce = rq->hw_context;
>
> /*
> * Catch up with any missed context-switch interrupts.
> @@ -1910,16 +2194,10 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
> */
> execlists_cancel_port_requests(execlists);
>
> - /* Push back any incomplete requests for replay after the reset. */
> - rq = __unwind_incomplete_requests(engine, 0);
> + rq = active_request(rq);
> if (!rq)
> goto out_replay;
>
> - if (rq->hw_context != ce) { /* caught just before a CS event */
> - rq = NULL;
> - goto out_replay;
> - }
> -
> /*
> * If this request hasn't started yet, e.g. it is waiting on a
> * semaphore, we need to avoid skipping the request or else we
> @@ -1966,13 +2244,16 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
> }
> execlists_init_reg_state(regs, ce, engine, ce->ring);
>
> - /* Rerun the request; its payload has been neutered (if guilty). */
> out_replay:
> + /* Rerun the request; its payload has been neutered (if guilty). */
> ce->ring->head =
> rq ? intel_ring_wrap(ce->ring, rq->head) : ce->ring->tail;
> intel_ring_update_space(ce->ring);
> __execlists_update_reg_state(ce, engine);
>
> + /* Push back any incomplete requests for replay after the reset. */
> + __unwind_incomplete_requests(engine, 0);
> +
> out_clear:
> execlists_clear_all_active(execlists);
> }
> @@ -2046,6 +2327,26 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine)
> i915_priolist_free(p);
> }
>
> + /* Cancel all attached virtual engines */
> + while ((rb = rb_first_cached(&execlists->virtual))) {
> + struct virtual_engine *ve =
> + rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
> +
> + rb_erase_cached(rb, &execlists->virtual);
> + RB_CLEAR_NODE(rb);
> +
> + spin_lock(&ve->base.timeline.lock);
> + if (ve->request) {
> + ve->request->engine = engine;
> + __i915_request_submit(ve->request);
> + dma_fence_set_error(&ve->request->fence, -EIO);
> + i915_request_mark_complete(ve->request);
> + ve->base.execlists.queue_priority_hint = INT_MIN;
> + ve->request = NULL;
> + }
> + spin_unlock(&ve->base.timeline.lock);
> + }
> +
> /* Remaining _unready_ requests will be nop'ed when submitted */
>
> execlists->queue_priority_hint = INT_MIN;
> @@ -2529,12 +2830,15 @@ static void execlists_init_reg_state(u32 *regs,
> bool rcs = engine->class == RENDER_CLASS;
> u32 base = engine->mmio_base;
>
> - /* A context is actually a big batch buffer with several
> + /*
> + * A context is actually a big batch buffer with several
> * MI_LOAD_REGISTER_IMM commands followed by (reg, value) pairs. The
> * values we are setting here are only for the first context restore:
> * on a subsequent save, the GPU will recreate this batchbuffer with new
> * values (including all the missing MI_LOAD_REGISTER_IMM commands that
> * we are not initializing here).
> + *
> + * Must keep consistent with virtual_update_register_offsets().
> */
> regs[CTX_LRI_HEADER_0] = MI_LOAD_REGISTER_IMM(rcs ? 14 : 11) |
> MI_LRI_FORCE_POSTED;
> @@ -2753,6 +3057,320 @@ static int execlists_context_deferred_alloc(struct intel_context *ce,
> return ret;
> }
>
> +static void virtual_context_destroy(struct kref *kref)
> +{
> + struct virtual_engine *ve =
> + container_of(kref, typeof(*ve), context.ref);
> + unsigned int n;
> +
> + GEM_BUG_ON(ve->request);
> + GEM_BUG_ON(ve->context.active);
> +
> + for (n = 0; n < ve->num_siblings; n++) {
> + struct intel_engine_cs *sibling = ve->siblings[n];
> + struct rb_node *node = &ve->nodes[sibling->id].rb;
> +
> + if (RB_EMPTY_NODE(node))
> + continue;
> +
> + spin_lock_irq(&sibling->timeline.lock);
> +
> + /* Detachment is lazily performed in the execlists tasklet */
> + if (!RB_EMPTY_NODE(node))
> + rb_erase_cached(node, &sibling->execlists.virtual);
> +
> + spin_unlock_irq(&sibling->timeline.lock);
> + }
> + GEM_BUG_ON(__tasklet_is_scheduled(&ve->base.execlists.tasklet));
> +
> + if (ve->context.state)
> + __execlists_context_fini(&ve->context);
> +
> + i915_timeline_fini(&ve->base.timeline);
> + kfree(ve);
> +}
> +
> +static void virtual_engine_initial_hint(struct virtual_engine *ve)
> +{
> + int swp;
> +
> + /*
> + * Pick a random sibling on starting to help spread the load around.
> + *
> + * New contexts are typically created with exactly the same order
> + * of siblings, and often started in batches. Due to the way we iterate
> + * the array of sibling when submitting requests, sibling[0] is
> + * prioritised for dequeuing. If we make sure that sibling[0] is fairly
> + * randomised across the system, we also help spread the load by the
> + * first engine we inspect being different each time.
> + *
> + * NB This does not force us to execute on this engine, it will just
> + * typically be the first we inspect for submission.
> + */
> + swp = prandom_u32_max(ve->num_siblings);
> + if (!swp)
> + return;
> +
> + swap(ve->siblings[swp], ve->siblings[0]);
> + virtual_update_register_offsets(ve->context.lrc_reg_state,
> + ve->siblings[0]);
> +}
> +
> +static int virtual_context_pin(struct intel_context *ce)
> +{
> + struct virtual_engine *ve = container_of(ce, typeof(*ve), context);
> + int err;
> +
> + /* Note: we must use a real engine class for setting up reg state */
> + err = __execlists_context_pin(ce, ve->siblings[0]);
> + if (err)
> + return err;
> +
> + virtual_engine_initial_hint(ve);
> + return 0;
> +}
> +
> +static void virtual_context_enter(struct intel_context *ce)
> +{
> + struct virtual_engine *ve = container_of(ce, typeof(*ve), context);
> + unsigned int n;
> +
> + for (n = 0; n < ve->num_siblings; n++)
> + intel_engine_pm_get(ve->siblings[n]);
> +}
> +
> +static void virtual_context_exit(struct intel_context *ce)
> +{
> + struct virtual_engine *ve = container_of(ce, typeof(*ve), context);
> + unsigned int n;
> +
> + ce->saturated = 0;
> + for (n = 0; n < ve->num_siblings; n++)
> + intel_engine_pm_put(ve->siblings[n]);
> +}
> +
> +static const struct intel_context_ops virtual_context_ops = {
> + .pin = virtual_context_pin,
> + .unpin = execlists_context_unpin,
> +
> + .enter = virtual_context_enter,
> + .exit = virtual_context_exit,
> +
> + .destroy = virtual_context_destroy,
> +};
> +
> +static void virtual_submission_tasklet(unsigned long data)
> +{
> + struct virtual_engine * const ve = (struct virtual_engine *)data;
> + const int prio = ve->base.execlists.queue_priority_hint;
> + unsigned int n;
> +
> + local_irq_disable();
> + for (n = 0; READ_ONCE(ve->request) && n < ve->num_siblings; n++) {
> + struct intel_engine_cs *sibling = ve->siblings[n];
> + struct ve_node * const node = &ve->nodes[sibling->id];
> + struct rb_node **parent, *rb;
> + bool first;
> +
> + spin_lock(&sibling->timeline.lock);
> +
> + if (!RB_EMPTY_NODE(&node->rb)) {
> + /*
> + * Cheat and avoid rebalancing the tree if we can
> + * reuse this node in situ.
> + */
> + first = rb_first_cached(&sibling->execlists.virtual) ==
> + &node->rb;
> + if (prio == node->prio || (prio > node->prio && first))
> + goto submit_engine;
> +
> + rb_erase_cached(&node->rb, &sibling->execlists.virtual);
> + }
> +
> + rb = NULL;
> + first = true;
> + parent = &sibling->execlists.virtual.rb_root.rb_node;
> + while (*parent) {
> + struct ve_node *other;
> +
> + rb = *parent;
> + other = rb_entry(rb, typeof(*other), rb);
> + if (prio > other->prio) {
> + parent = &rb->rb_left;
> + } else {
> + parent = &rb->rb_right;
> + first = false;
> + }
> + }
> +
> + rb_link_node(&node->rb, rb, parent);
> + rb_insert_color_cached(&node->rb,
> + &sibling->execlists.virtual,
> + first);
> +
> +submit_engine:
> + GEM_BUG_ON(RB_EMPTY_NODE(&node->rb));
> + node->prio = prio;
> + if (first && prio > sibling->execlists.queue_priority_hint) {
> + sibling->execlists.queue_priority_hint = prio;
> + tasklet_hi_schedule(&sibling->execlists.tasklet);
> + }
> +
> + spin_unlock(&sibling->timeline.lock);
> + }
> + local_irq_enable();
> +}
> +
> +static void virtual_submit_request(struct i915_request *rq)
> +{
> + struct virtual_engine *ve = to_virtual_engine(rq->engine);
> +
> + GEM_TRACE("%s: rq=%llx:%lld\n",
> + ve->base.name,
> + rq->fence.context,
> + rq->fence.seqno);
> +
> + GEM_BUG_ON(ve->base.submit_request != virtual_submit_request);
> +
> + GEM_BUG_ON(ve->request);
> + ve->base.execlists.queue_priority_hint = rq_prio(rq);
> + WRITE_ONCE(ve->request, rq);
> +
> + tasklet_schedule(&ve->base.execlists.tasklet);
> +}
> +
> +struct intel_context *
> +intel_execlists_create_virtual(struct i915_gem_context *ctx,
> + struct intel_engine_cs **siblings,
> + unsigned int count)
> +{
> + struct virtual_engine *ve;
> + unsigned int n;
> + int err;
> +
> + if (count == 0)
> + return ERR_PTR(-EINVAL);
> +
> + if (count == 1)
> + return intel_context_create(ctx, siblings[0]);
> +
> + ve = kzalloc(struct_size(ve, siblings, count), GFP_KERNEL);
> + if (!ve)
> + return ERR_PTR(-ENOMEM);
> +
> + ve->base.i915 = ctx->i915;
> + ve->base.id = -1;
> + ve->base.class = OTHER_CLASS;
> + ve->base.uabi_class = I915_ENGINE_CLASS_INVALID;
> + ve->base.instance = I915_ENGINE_CLASS_INVALID_VIRTUAL;
> + ve->base.flags = I915_ENGINE_IS_VIRTUAL;
> +
> + snprintf(ve->base.name, sizeof(ve->base.name), "virtual");
> +
> + err = i915_timeline_init(ctx->i915, &ve->base.timeline, NULL);
> + if (err)
> + goto err_put;
> + i915_timeline_set_subclass(&ve->base.timeline, TIMELINE_VIRTUAL);
> +
> + intel_engine_init_execlists(&ve->base);
> +
> + ve->base.cops = &virtual_context_ops;
> + ve->base.request_alloc = execlists_request_alloc;
> +
> + ve->base.schedule = i915_schedule;
> + ve->base.submit_request = virtual_submit_request;
> +
> + ve->base.execlists.queue_priority_hint = INT_MIN;
> + tasklet_init(&ve->base.execlists.tasklet,
> + virtual_submission_tasklet,
> + (unsigned long)ve);
> +
> + intel_context_init(&ve->context, ctx, &ve->base);
> +
> + for (n = 0; n < count; n++) {
> + struct intel_engine_cs *sibling = siblings[n];
> +
> + GEM_BUG_ON(!is_power_of_2(sibling->mask));
> + if (sibling->mask & ve->base.mask) {
> + DRM_DEBUG("duplicate %s entry in load balancer\n",
> + sibling->name);
> + err = -EINVAL;
> + goto err_put;
> + }
> +
> + /*
> + * The virtual engine implementation is tightly coupled to
> + * the execlists backend -- we push out request directly
> + * into a tree inside each physical engine. We could support
> + * layering if we handle cloning of the requests and
> + * submitting a copy into each backend.
> + */
> + if (sibling->execlists.tasklet.func !=
> + execlists_submission_tasklet) {
> + err = -ENODEV;
> + goto err_put;
> + }
> +
> + GEM_BUG_ON(RB_EMPTY_NODE(&ve->nodes[sibling->id].rb));
> + RB_CLEAR_NODE(&ve->nodes[sibling->id].rb);
> +
> + ve->siblings[ve->num_siblings++] = sibling;
> + ve->base.mask |= sibling->mask;
> +
> + /*
> + * All physical engines must be compatible for their emission
> + * functions (as we build the instructions during request
> + * construction and do not alter them before submission
> + * on the physical engine). We use the engine class as a guide
> + * here, although that could be refined.
> + */
> + if (ve->base.class != OTHER_CLASS) {
> + if (ve->base.class != sibling->class) {
> + DRM_DEBUG("invalid mixing of engine class, sibling %d, already %d\n",
> + sibling->class, ve->base.class);
> + err = -EINVAL;
> + goto err_put;
> + }
> + continue;
> + }
> +
> + ve->base.class = sibling->class;
> + ve->base.uabi_class = sibling->uabi_class;
> + snprintf(ve->base.name, sizeof(ve->base.name),
> + "v%dx%d", ve->base.class, count);
> + ve->base.context_size = sibling->context_size;
> +
> + ve->base.emit_bb_start = sibling->emit_bb_start;
> + ve->base.emit_flush = sibling->emit_flush;
> + ve->base.emit_init_breadcrumb = sibling->emit_init_breadcrumb;
> + ve->base.emit_fini_breadcrumb = sibling->emit_fini_breadcrumb;
> + ve->base.emit_fini_breadcrumb_dw =
> + sibling->emit_fini_breadcrumb_dw;
> + }
> +
> + return &ve->context;
> +
> +err_put:
> + intel_context_put(&ve->context);
> + return ERR_PTR(err);
> +}
> +
> +struct intel_context *
> +intel_execlists_clone_virtual(struct i915_gem_context *ctx,
> + struct intel_engine_cs *src)
> +{
> + struct virtual_engine *se = to_virtual_engine(src);
> + struct intel_context *dst;
> +
> + dst = intel_execlists_create_virtual(ctx,
> + se->siblings,
> + se->num_siblings);
> + if (IS_ERR(dst))
> + return dst;
> +
> + return dst;
> +}
> +
> void intel_execlists_show_requests(struct intel_engine_cs *engine,
> struct drm_printer *m,
> void (*show_request)(struct drm_printer *m,
> @@ -2810,6 +3428,29 @@ void intel_execlists_show_requests(struct intel_engine_cs *engine,
> show_request(m, last, "\t\tQ ");
> }
>
> + last = NULL;
> + count = 0;
> + for (rb = rb_first_cached(&execlists->virtual); rb; rb = rb_next(rb)) {
> + struct virtual_engine *ve =
> + rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
> + struct i915_request *rq = READ_ONCE(ve->request);
> +
> + if (rq) {
> + if (count++ < max - 1)
> + show_request(m, rq, "\t\tV ");
> + else
> + last = rq;
> + }
> + }
> + if (last) {
> + if (count > max) {
> + drm_printf(m,
> + "\t\t...skipping %d virtual requests...\n",
> + count - max);
> + }
> + show_request(m, last, "\t\tV ");
> + }
> +
> spin_unlock_irqrestore(&engine->timeline.lock, flags);
> }
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.h b/drivers/gpu/drm/i915/gt/intel_lrc.h
> index a0dc907a7249..5530606052e5 100644
> --- a/drivers/gpu/drm/i915/gt/intel_lrc.h
> +++ b/drivers/gpu/drm/i915/gt/intel_lrc.h
> @@ -114,4 +114,13 @@ void intel_execlists_show_requests(struct intel_engine_cs *engine,
> const char *prefix),
> unsigned int max);
>
> +struct intel_context *
> +intel_execlists_create_virtual(struct i915_gem_context *ctx,
> + struct intel_engine_cs **siblings,
> + unsigned int count);
> +
> +struct intel_context *
> +intel_execlists_clone_virtual(struct i915_gem_context *ctx,
> + struct intel_engine_cs *src);
> +
> #endif /* _INTEL_LRC_H_ */
> diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
> index 5b3d8e33f1cf..f880271fb9ba 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
> @@ -1310,6 +1310,185 @@ static int live_preempt_smoke(void *arg)
> return err;
> }
>
> +static int nop_virtual_engine(struct drm_i915_private *i915,
> + struct intel_engine_cs **siblings,
> + unsigned int nsibling,
> + unsigned int nctx,
> + unsigned int flags)
> +#define CHAIN BIT(0)
> +{
> + IGT_TIMEOUT(end_time);
> + struct i915_request *request[16];
> + struct i915_gem_context *ctx[16];
> + struct intel_context *ve[16];
> + unsigned long n, prime, nc;
> + struct igt_live_test t;
> + ktime_t times[2] = {};
> + int err;
> +
> + GEM_BUG_ON(!nctx || nctx > ARRAY_SIZE(ctx));
> +
> + for (n = 0; n < nctx; n++) {
> + ctx[n] = kernel_context(i915);
> + if (!ctx[n]) {
> + err = -ENOMEM;
> + nctx = n;
> + goto out;
> + }
> +
> + ve[n] = intel_execlists_create_virtual(ctx[n],
> + siblings, nsibling);
> + if (IS_ERR(ve[n])) {
> + kernel_context_close(ctx[n]);
> + err = PTR_ERR(ve[n]);
> + nctx = n;
> + goto out;
> + }
> +
> + err = intel_context_pin(ve[n]);
> + if (err) {
> + intel_context_put(ve[n]);
> + kernel_context_close(ctx[n]);
> + nctx = n;
> + goto out;
> + }
> + }
> +
> + err = igt_live_test_begin(&t, i915, __func__, ve[0]->engine->name);
> + if (err)
> + goto out;
> +
> + for_each_prime_number_from(prime, 1, 8192) {
> + times[1] = ktime_get_raw();
> +
> + if (flags & CHAIN) {
> + for (nc = 0; nc < nctx; nc++) {
> + for (n = 0; n < prime; n++) {
> + request[nc] =
> + i915_request_create(ve[nc]);
> + if (IS_ERR(request[nc])) {
> + err = PTR_ERR(request[nc]);
> + goto out;
> + }
> +
> + i915_request_add(request[nc]);
> + }
> + }
> + } else {
> + for (n = 0; n < prime; n++) {
> + for (nc = 0; nc < nctx; nc++) {
> + request[nc] =
> + i915_request_create(ve[nc]);
> + if (IS_ERR(request[nc])) {
> + err = PTR_ERR(request[nc]);
> + goto out;
> + }
> +
> + i915_request_add(request[nc]);
> + }
> + }
> + }
> +
> + for (nc = 0; nc < nctx; nc++) {
> + if (i915_request_wait(request[nc],
> + I915_WAIT_LOCKED,
> + HZ / 10) < 0) {
> + pr_err("%s(%s): wait for %llx:%lld timed out\n",
> + __func__, ve[0]->engine->name,
> + request[nc]->fence.context,
> + request[nc]->fence.seqno);
> +
> + GEM_TRACE("%s(%s) failed at request %llx:%lld\n",
> + __func__, ve[0]->engine->name,
> + request[nc]->fence.context,
> + request[nc]->fence.seqno);
> + GEM_TRACE_DUMP();
> + i915_gem_set_wedged(i915);
> + break;
> + }
> + }
> +
> + times[1] = ktime_sub(ktime_get_raw(), times[1]);
> + if (prime == 1)
> + times[0] = times[1];
> +
> + if (__igt_timeout(end_time, NULL))
> + break;
> + }
> +
> + err = igt_live_test_end(&t);
> + if (err)
> + goto out;
> +
> + pr_info("Requestx%d latencies on %s: 1 = %lluns, %lu = %lluns\n",
> + nctx, ve[0]->engine->name, ktime_to_ns(times[0]),
> + prime, div64_u64(ktime_to_ns(times[1]), prime));
> +
> +out:
> + if (igt_flush_test(i915, I915_WAIT_LOCKED))
> + err = -EIO;
> +
> + for (nc = 0; nc < nctx; nc++) {
> + intel_context_unpin(ve[nc]);
> + intel_context_put(ve[nc]);
> + kernel_context_close(ctx[nc]);
> + }
> + return err;
> +}
> +
> +static int live_virtual_engine(void *arg)
> +{
> + struct drm_i915_private *i915 = arg;
> + struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1];
> + struct intel_engine_cs *engine;
> + enum intel_engine_id id;
> + unsigned int class, inst;
> + int err = -ENODEV;
> +
> + if (USES_GUC_SUBMISSION(i915))
> + return 0;
> +
> + mutex_lock(&i915->drm.struct_mutex);
> +
> + for_each_engine(engine, i915, id) {
> + err = nop_virtual_engine(i915, &engine, 1, 1, 0);
> + if (err) {
> + pr_err("Failed to wrap engine %s: err=%d\n",
> + engine->name, err);
> + goto out_unlock;
> + }
> + }
> +
> + for (class = 0; class <= MAX_ENGINE_CLASS; class++) {
> + int nsibling, n;
> +
> + nsibling = 0;
> + for (inst = 0; inst <= MAX_ENGINE_INSTANCE; inst++) {
> + if (!i915->engine_class[class][inst])
> + continue;
> +
> + siblings[nsibling++] = i915->engine_class[class][inst];
> + }
> + if (nsibling < 2)
> + continue;
> +
> + for (n = 1; n <= nsibling + 1; n++) {
> + err = nop_virtual_engine(i915, siblings, nsibling,
> + n, 0);
> + if (err)
> + goto out_unlock;
> + }
> +
> + err = nop_virtual_engine(i915, siblings, nsibling, n, CHAIN);
> + if (err)
> + goto out_unlock;
> + }
> +
> +out_unlock:
> + mutex_unlock(&i915->drm.struct_mutex);
> + return err;
> +}
> +
> int intel_execlists_live_selftests(struct drm_i915_private *i915)
> {
> static const struct i915_subtest tests[] = {
> @@ -1322,6 +1501,7 @@ int intel_execlists_live_selftests(struct drm_i915_private *i915)
> SUBTEST(live_chain_preempt),
> SUBTEST(live_preempt_hang),
> SUBTEST(live_preempt_smoke),
> + SUBTEST(live_virtual_engine),
> };
>
> if (!HAS_EXECLISTS(i915))
> diff --git a/drivers/gpu/drm/i915/i915_gem.h b/drivers/gpu/drm/i915/i915_gem.h
> index 67f8a4a807a0..fe82d3571072 100644
> --- a/drivers/gpu/drm/i915/i915_gem.h
> +++ b/drivers/gpu/drm/i915/i915_gem.h
> @@ -91,4 +91,9 @@ static inline bool __tasklet_enable(struct tasklet_struct *t)
> return atomic_dec_and_test(&t->count);
> }
>
> +static inline bool __tasklet_is_scheduled(struct tasklet_struct *t)
> +{
> + return test_bit(TASKLET_STATE_SCHED, &t->state);
> +}
> +
> #endif /* __I915_GEM_H__ */
> diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
> index 040c0b770c21..3ae1d27130a8 100644
> --- a/drivers/gpu/drm/i915/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/i915_gem_context.c
> @@ -86,6 +86,7 @@
> */
>
> #include <linux/log2.h>
> +#include <linux/nospec.h>
>
> #include <drm/i915_drm.h>
>
> @@ -1218,7 +1219,6 @@ __intel_context_reconfigure_sseu(struct intel_context *ce,
> int ret;
>
> GEM_BUG_ON(INTEL_GEN(ce->gem_context->i915) < 8);
> - GEM_BUG_ON(ce->engine->id != RCS0);
>
> ret = intel_context_lock_pinned(ce);
> if (ret)
> @@ -1412,7 +1412,100 @@ struct set_engines {
> struct i915_gem_engines *engines;
> };
>
> +static int
> +set_engines__load_balance(struct i915_user_extension __user *base, void *data)
> +{
> + struct i915_context_engines_load_balance __user *ext =
> + container_of_user(base, typeof(*ext), base);
> + const struct set_engines *set = data;
> + struct intel_engine_cs *stack[16];
> + struct intel_engine_cs **siblings;
> + struct intel_context *ce;
> + u16 num_siblings, idx;
> + unsigned int n;
> + int err;
> +
> + if (!HAS_EXECLISTS(set->ctx->i915))
> + return -ENODEV;
> +
> + if (USES_GUC_SUBMISSION(set->ctx->i915))
> + return -ENODEV; /* not implement yet */
> +
> + if (get_user(idx, &ext->engine_index))
> + return -EFAULT;
> +
> + if (idx >= set->engines->num_engines) {
> + DRM_DEBUG("Invalid placement value, %d >= %d\n",
> + idx, set->engines->num_engines);
> + return -EINVAL;
> + }
> +
> + idx = array_index_nospec(idx, set->engines->num_engines);
> + if (set->engines->engines[idx]) {
> + DRM_DEBUG("Invalid placement[%d], already occupied\n", idx);
> + return -EEXIST;
> + }
> +
> + if (get_user(num_siblings, &ext->num_siblings))
> + return -EFAULT;
> +
> + err = check_user_mbz(&ext->flags);
> + if (err)
> + return err;
> +
> + err = check_user_mbz(&ext->mbz64);
> + if (err)
> + return err;
> +
> + siblings = stack;
> + if (num_siblings > ARRAY_SIZE(stack)) {
> + siblings = kmalloc_array(num_siblings,
> + sizeof(*siblings),
> + GFP_KERNEL);
> + if (!siblings)
> + return -ENOMEM;
> + }
> +
> + for (n = 0; n < num_siblings; n++) {
> + struct i915_engine_class_instance ci;
> +
> + if (copy_from_user(&ci, &ext->engines[n], sizeof(ci))) {
> + err = -EFAULT;
> + goto out_siblings;
> + }
> +
> + siblings[n] = intel_engine_lookup_user(set->ctx->i915,
> + ci.engine_class,
> + ci.engine_instance);
> + if (!siblings[n]) {
> + DRM_DEBUG("Invalid sibling[%d]: { class:%d, inst:%d }\n",
> + n, ci.engine_class, ci.engine_instance);
> + err = -EINVAL;
> + goto out_siblings;
> + }
> + }
> +
> + ce = intel_execlists_create_virtual(set->ctx, siblings, n);
> + if (IS_ERR(ce)) {
> + err = PTR_ERR(ce);
> + goto out_siblings;
> + }
> +
> + if (cmpxchg(&set->engines->engines[idx], NULL, ce)) {
> + intel_context_put(ce);
> + err = -EEXIST;
> + goto out_siblings;
> + }
> +
> +out_siblings:
> + if (siblings != stack)
> + kfree(siblings);
> +
> + return err;
> +}
> +
> static const i915_user_extension_fn set_engines__extensions[] = {
> + [I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE] = set_engines__load_balance,
> };
>
> static int
> @@ -1711,14 +1804,29 @@ static int clone_engines(struct i915_gem_context *dst,
>
> clone->i915 = dst->i915;
> for (n = 0; n < e->num_engines; n++) {
> + struct intel_engine_cs *engine;
> +
> if (!e->engines[n]) {
> clone->engines[n] = NULL;
> continue;
> }
> + engine = e->engines[n]->engine;
>
> - clone->engines[n] =
> - intel_context_create(dst, e->engines[n]->engine);
> - if (!clone->engines[n]) {
> + /*
> + * Virtual engines are singletons; they can only exist
> + * inside a single context, because they embed their
> + * HW context... As each virtual context implies a single
> + * timeline (each engine can only dequeue a single request
> + * at any time), it would be surprising for two contexts
> + * to use the same engine. So let's create a copy of
> + * the virtual engine instead.
> + */
> + if (intel_engine_is_virtual(engine))
> + clone->engines[n] =
> + intel_execlists_clone_virtual(dst, engine);
> + else
> + clone->engines[n] = intel_context_create(dst, engine);
> + if (IS_ERR_OR_NULL(clone->engines[n])) {
> __free_engines(clone, n);
> goto err_unlock;
> }
> diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c
> index d215dcdf0b1e..78ceb56d7801 100644
> --- a/drivers/gpu/drm/i915/i915_scheduler.c
> +++ b/drivers/gpu/drm/i915/i915_scheduler.c
> @@ -150,17 +150,26 @@ sched_lock_engine(const struct i915_sched_node *node,
> struct intel_engine_cs *locked,
> struct sched_cache *cache)
> {
> - struct intel_engine_cs *engine = node_to_request(node)->engine;
> + const struct i915_request *rq = node_to_request(node);
> + struct intel_engine_cs *engine;
>
> GEM_BUG_ON(!locked);
>
> - if (engine != locked) {
> + /*
> + * Virtual engines complicate acquiring the engine timeline lock,
> + * as their rq->engine pointer is not stable until under that
> + * engine lock. The simple ploy we use is to take the lock then
> + * check that the rq still belongs to the newly locked engine.
> + */
> + while (locked != (engine = READ_ONCE(rq->engine))) {
> spin_unlock(&locked->timeline.lock);
> memset(cache, 0, sizeof(*cache));
> spin_lock(&engine->timeline.lock);
> + locked = engine;
> }
>
> - return engine;
> + GEM_BUG_ON(locked != engine);
> + return locked;
> }
>
> static inline int rq_prio(const struct i915_request *rq)
> @@ -272,6 +281,7 @@ static void __i915_schedule(struct i915_sched_node *node,
> spin_lock(&engine->timeline.lock);
>
> /* Fifo and depth-first replacement ensure our deps execute before us */
> + engine = sched_lock_engine(node, engine, &cache);
> list_for_each_entry_safe_reverse(dep, p, &dfs, dfs_link) {
> INIT_LIST_HEAD(&dep->dfs_link);
>
> @@ -283,8 +293,11 @@ static void __i915_schedule(struct i915_sched_node *node,
> if (prio <= node->attr.priority || node_signaled(node))
> continue;
>
> + GEM_BUG_ON(node_to_request(node)->engine != engine);
> +
> node->attr.priority = prio;
> if (!list_empty(&node->link)) {
> + GEM_BUG_ON(intel_engine_is_virtual(engine));
> if (!cache.priolist)
> cache.priolist =
> i915_sched_lookup_priolist(engine,
> diff --git a/drivers/gpu/drm/i915/i915_timeline_types.h b/drivers/gpu/drm/i915/i915_timeline_types.h
> index 5256a0b5c5f7..1688705f4a2b 100644
> --- a/drivers/gpu/drm/i915/i915_timeline_types.h
> +++ b/drivers/gpu/drm/i915/i915_timeline_types.h
> @@ -26,6 +26,7 @@ struct i915_timeline {
> spinlock_t lock;
> #define TIMELINE_CLIENT 0 /* default subclass */
> #define TIMELINE_ENGINE 1
> +#define TIMELINE_VIRTUAL 2
> struct mutex mutex; /* protects the flow of requests */
>
> unsigned int pin_count;
> diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
> index 62396d575e28..f9770948161c 100644
> --- a/include/uapi/drm/i915_drm.h
> +++ b/include/uapi/drm/i915_drm.h
> @@ -137,6 +137,7 @@ struct i915_engine_class_instance {
> __u16 engine_class; /* see enum drm_i915_gem_engine_class */
> __u16 engine_instance;
> #define I915_ENGINE_CLASS_INVALID_NONE -1
> +#define I915_ENGINE_CLASS_INVALID_VIRTUAL -2
> };
>
> /**
> @@ -1608,8 +1609,46 @@ struct drm_i915_gem_context_param_sseu {
> __u32 rsvd;
> };
>
> +/*
> + * i915_context_engines_load_balance:
> + *
> + * Enable load balancing across this set of engines.
> + *
> + * Into the I915_EXEC_DEFAULT slot [0], a virtual engine is created that when
> + * used will proxy the execbuffer request onto one of the set of engines
> + * in such a way as to distribute the load evenly across the set.
> + *
> + * The set of engines must be compatible (e.g. the same HW class) as they
> + * will share the same logical GPU context and ring.
> + *
> + * To intermix rendering with the virtual engine and direct rendering onto
> + * the backing engines (bypassing the load balancing proxy), the context must
> + * be defined to use a single timeline for all engines.
> + */
> +struct i915_context_engines_load_balance {
> + struct i915_user_extension base;
> +
> + __u16 engine_index;
> + __u16 num_siblings;
> + __u32 flags; /* all undefined flags must be zero */
> +
> + __u64 mbz64; /* reserved for future use; must be zero */
> +
> + struct i915_engine_class_instance engines[0];
> +} __attribute__((packed));
> +
> +#define I915_DEFINE_CONTEXT_ENGINES_LOAD_BALANCE(name__, N__) struct { \
> + struct i915_user_extension base; \
> + __u16 engine_index; \
> + __u16 num_siblings; \
> + __u32 flags; \
> + __u64 mbz64; \
> + struct i915_engine_class_instance engines[N__]; \
> +} __attribute__((packed)) name__
> +
> struct i915_context_param_engines {
> __u64 extensions; /* linked chain of extension blocks, 0 terminates */
> +#define I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE 0 /* see i915_context_engines_load_balance */
> struct i915_engine_class_instance engines[0];
> } __attribute__((packed));
>
>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread
* [PATCH 15/40] drm/i915: Apply an execution_mask to the virtual_engine
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (12 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 14/40] drm/i915: Load balancing across a virtual engine Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 10:13 ` Tvrtko Ursulin
2019-05-08 8:06 ` [PATCH 16/40] drm/i915: Extend execution fence to support a callback Chris Wilson
` (31 subsequent siblings)
45 siblings, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Allow the user to direct which physical engines of the virtual engine
they wish to execute one, as sometimes it is necessary to override the
load balancing algorithm.
v2: Only kick the virtual engines on context-out if required
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/gt/intel_lrc.c | 67 +++++++++++++
drivers/gpu/drm/i915/gt/selftest_lrc.c | 131 +++++++++++++++++++++++++
drivers/gpu/drm/i915/i915_request.c | 1 +
drivers/gpu/drm/i915/i915_request.h | 3 +
4 files changed, 202 insertions(+)
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index bc388df39802..69849ffb9c82 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -550,6 +550,15 @@ execlists_context_schedule_in(struct i915_request *rq)
rq->hw_context->active = rq->engine;
}
+static void kick_siblings(struct i915_request *rq)
+{
+ struct virtual_engine *ve = to_virtual_engine(rq->hw_context->engine);
+ struct i915_request *next = READ_ONCE(ve->request);
+
+ if (next && next->execution_mask & ~rq->execution_mask)
+ tasklet_schedule(&ve->base.execlists.tasklet);
+}
+
static inline void
execlists_context_schedule_out(struct i915_request *rq, unsigned long status)
{
@@ -557,6 +566,18 @@ execlists_context_schedule_out(struct i915_request *rq, unsigned long status)
intel_engine_context_out(rq->engine);
execlists_context_status_change(rq, status);
trace_i915_request_out(rq);
+
+ /*
+ * If this is part of a virtual engine, its next request may have
+ * been blocked waiting for access to the active context. We have
+ * to kick all the siblings again in case we need to switch (e.g.
+ * the next request is not runnable on this engine). Hopefully,
+ * we will already have submitted the next request before the
+ * tasklet runs and do not need to rebuild each virtual tree
+ * and kick everyone again.
+ */
+ if (rq->engine != rq->hw_context->engine)
+ kick_siblings(rq);
}
static u64 execlists_update_context(struct i915_request *rq)
@@ -787,6 +808,9 @@ static bool virtual_matches(const struct virtual_engine *ve,
{
const struct intel_engine_cs *active;
+ if (!(rq->execution_mask & engine->mask)) /* We peeked too soon! */
+ return false;
+
/*
* We track when the HW has completed saving the context image
* (i.e. when we have seen the final CS event switching out of
@@ -3159,12 +3183,44 @@ static const struct intel_context_ops virtual_context_ops = {
.destroy = virtual_context_destroy,
};
+static intel_engine_mask_t virtual_submission_mask(struct virtual_engine *ve)
+{
+ struct i915_request *rq;
+ intel_engine_mask_t mask;
+
+ rq = READ_ONCE(ve->request);
+ if (!rq)
+ return 0;
+
+ /* The rq is ready for submission; rq->execution_mask is now stable. */
+ mask = rq->execution_mask;
+ if (unlikely(!mask)) {
+ /* Invalid selection, submit to a random engine in error */
+ i915_request_skip(rq, -ENODEV);
+ mask = ve->siblings[0]->mask;
+ }
+
+ GEM_TRACE("%s: rq=%llx:%lld, mask=%x, prio=%d\n",
+ ve->base.name,
+ rq->fence.context, rq->fence.seqno,
+ mask, ve->base.execlists.queue_priority_hint);
+
+ return mask;
+}
+
static void virtual_submission_tasklet(unsigned long data)
{
struct virtual_engine * const ve = (struct virtual_engine *)data;
const int prio = ve->base.execlists.queue_priority_hint;
+ intel_engine_mask_t mask;
unsigned int n;
+ rcu_read_lock();
+ mask = virtual_submission_mask(ve);
+ rcu_read_unlock();
+ if (unlikely(!mask))
+ return;
+
local_irq_disable();
for (n = 0; READ_ONCE(ve->request) && n < ve->num_siblings; n++) {
struct intel_engine_cs *sibling = ve->siblings[n];
@@ -3172,6 +3228,17 @@ static void virtual_submission_tasklet(unsigned long data)
struct rb_node **parent, *rb;
bool first;
+ if (unlikely(!(mask & sibling->mask))) {
+ if (!RB_EMPTY_NODE(&node->rb)) {
+ spin_lock(&sibling->timeline.lock);
+ rb_erase_cached(&node->rb,
+ &sibling->execlists.virtual);
+ RB_CLEAR_NODE(&node->rb);
+ spin_unlock(&sibling->timeline.lock);
+ }
+ continue;
+ }
+
spin_lock(&sibling->timeline.lock);
if (!RB_EMPTY_NODE(&node->rb)) {
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index ccc0b6350123..2ef7639ee665 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -1489,6 +1489,136 @@ static int live_virtual_engine(void *arg)
return err;
}
+static int mask_virtual_engine(struct drm_i915_private *i915,
+ struct intel_engine_cs **siblings,
+ unsigned int nsibling)
+{
+ struct i915_request *request[MAX_ENGINE_INSTANCE + 1];
+ struct i915_gem_context *ctx;
+ struct intel_context *ve;
+ struct igt_live_test t;
+ unsigned int n;
+ int err;
+
+ /*
+ * Check that by setting the execution mask on a request, we can
+ * restrict it to our desired engine within the virtual engine.
+ */
+
+ ctx = kernel_context(i915);
+ if (!ctx)
+ return -ENOMEM;
+
+ ve = intel_execlists_create_virtual(ctx, siblings, nsibling);
+ if (IS_ERR(ve)) {
+ err = PTR_ERR(ve);
+ goto out_close;
+ }
+
+ err = intel_context_pin(ve);
+ if (err)
+ goto out_put;
+
+ err = igt_live_test_begin(&t, i915, __func__, ve->engine->name);
+ if (err)
+ goto out_unpin;
+
+ for (n = 0; n < nsibling; n++) {
+ request[n] = i915_request_create(ve);
+ if (IS_ERR(request)) {
+ err = PTR_ERR(request);
+ nsibling = n;
+ goto out;
+ }
+
+ /* Reverse order as it's more likely to be unnatural */
+ request[n]->execution_mask = siblings[nsibling - n - 1]->mask;
+
+ i915_request_get(request[n]);
+ i915_request_add(request[n]);
+ }
+
+ for (n = 0; n < nsibling; n++) {
+ if (i915_request_wait(request[n], I915_WAIT_LOCKED, HZ / 10) < 0) {
+ pr_err("%s(%s): wait for %llx:%lld timed out\n",
+ __func__, ve->engine->name,
+ request[n]->fence.context,
+ request[n]->fence.seqno);
+
+ GEM_TRACE("%s(%s) failed at request %llx:%lld\n",
+ __func__, ve->engine->name,
+ request[n]->fence.context,
+ request[n]->fence.seqno);
+ GEM_TRACE_DUMP();
+ i915_gem_set_wedged(i915);
+ err = -EIO;
+ goto out;
+ }
+
+ if (request[n]->engine != siblings[nsibling - n - 1]) {
+ pr_err("Executed on wrong sibling '%s', expected '%s'\n",
+ request[n]->engine->name,
+ siblings[nsibling - n - 1]->name);
+ err = -EINVAL;
+ goto out;
+ }
+ }
+
+ err = igt_live_test_end(&t);
+ if (err)
+ goto out;
+
+out:
+ if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ err = -EIO;
+
+ for (n = 0; n < nsibling; n++)
+ i915_request_put(request[n]);
+
+out_unpin:
+ intel_context_unpin(ve);
+out_put:
+ intel_context_put(ve);
+out_close:
+ kernel_context_close(ctx);
+ return err;
+}
+
+static int live_virtual_mask(void *arg)
+{
+ struct drm_i915_private *i915 = arg;
+ struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1];
+ unsigned int class, inst;
+ int err = 0;
+
+ if (USES_GUC_SUBMISSION(i915))
+ return 0;
+
+ mutex_lock(&i915->drm.struct_mutex);
+
+ for (class = 0; class <= MAX_ENGINE_CLASS; class++) {
+ unsigned int nsibling;
+
+ nsibling = 0;
+ for (inst = 0; inst <= MAX_ENGINE_INSTANCE; inst++) {
+ if (!i915->engine_class[class][inst])
+ break;
+
+ siblings[nsibling++] = i915->engine_class[class][inst];
+ }
+ if (nsibling < 2)
+ continue;
+
+ err = mask_virtual_engine(i915, siblings, nsibling);
+ if (err)
+ goto out_unlock;
+ }
+
+out_unlock:
+ mutex_unlock(&i915->drm.struct_mutex);
+ return err;
+}
+
int intel_execlists_live_selftests(struct drm_i915_private *i915)
{
static const struct i915_subtest tests[] = {
@@ -1502,6 +1632,7 @@ int intel_execlists_live_selftests(struct drm_i915_private *i915)
SUBTEST(live_preempt_hang),
SUBTEST(live_preempt_smoke),
SUBTEST(live_virtual_engine),
+ SUBTEST(live_virtual_mask),
};
if (!HAS_EXECLISTS(i915))
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index cbe4098cd6ec..3beced7daa15 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -699,6 +699,7 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp)
rq->batch = NULL;
rq->capture_list = NULL;
rq->waitboost = false;
+ rq->execution_mask = ALL_ENGINES;
INIT_LIST_HEAD(&rq->active_list);
INIT_LIST_HEAD(&rq->execute_cb);
diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
index 8025a89b5999..d7f9b2194568 100644
--- a/drivers/gpu/drm/i915/i915_request.h
+++ b/drivers/gpu/drm/i915/i915_request.h
@@ -28,6 +28,8 @@
#include <linux/dma-fence.h>
#include <linux/lockdep.h>
+#include "gt/intel_engine_types.h"
+
#include "i915_gem.h"
#include "i915_scheduler.h"
#include "i915_selftest.h"
@@ -156,6 +158,7 @@ struct i915_request {
*/
struct i915_sched_node sched;
struct i915_dependency dep;
+ intel_engine_mask_t execution_mask;
/*
* A convenience pointer to the current breadcrumb value stored in
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* Re: [PATCH 15/40] drm/i915: Apply an execution_mask to the virtual_engine
2019-05-08 8:06 ` [PATCH 15/40] drm/i915: Apply an execution_mask to the virtual_engine Chris Wilson
@ 2019-05-08 10:13 ` Tvrtko Ursulin
0 siblings, 0 replies; 71+ messages in thread
From: Tvrtko Ursulin @ 2019-05-08 10:13 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 08/05/2019 09:06, Chris Wilson wrote:
> Allow the user to direct which physical engines of the virtual engine
> they wish to execute one, as sometimes it is necessary to override the
> load balancing algorithm.
>
> v2: Only kick the virtual engines on context-out if required
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> ---
> drivers/gpu/drm/i915/gt/intel_lrc.c | 67 +++++++++++++
> drivers/gpu/drm/i915/gt/selftest_lrc.c | 131 +++++++++++++++++++++++++
> drivers/gpu/drm/i915/i915_request.c | 1 +
> drivers/gpu/drm/i915/i915_request.h | 3 +
> 4 files changed, 202 insertions(+)
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
> index bc388df39802..69849ffb9c82 100644
> --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
> +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
> @@ -550,6 +550,15 @@ execlists_context_schedule_in(struct i915_request *rq)
> rq->hw_context->active = rq->engine;
> }
>
> +static void kick_siblings(struct i915_request *rq)
> +{
> + struct virtual_engine *ve = to_virtual_engine(rq->hw_context->engine);
> + struct i915_request *next = READ_ONCE(ve->request);
> +
> + if (next && next->execution_mask & ~rq->execution_mask)
> + tasklet_schedule(&ve->base.execlists.tasklet);
> +}
> +
> static inline void
> execlists_context_schedule_out(struct i915_request *rq, unsigned long status)
> {
> @@ -557,6 +566,18 @@ execlists_context_schedule_out(struct i915_request *rq, unsigned long status)
> intel_engine_context_out(rq->engine);
> execlists_context_status_change(rq, status);
> trace_i915_request_out(rq);
> +
> + /*
> + * If this is part of a virtual engine, its next request may have
> + * been blocked waiting for access to the active context. We have
> + * to kick all the siblings again in case we need to switch (e.g.
> + * the next request is not runnable on this engine). Hopefully,
> + * we will already have submitted the next request before the
> + * tasklet runs and do not need to rebuild each virtual tree
> + * and kick everyone again.
> + */
> + if (rq->engine != rq->hw_context->engine)
> + kick_siblings(rq);
> }
>
> static u64 execlists_update_context(struct i915_request *rq)
> @@ -787,6 +808,9 @@ static bool virtual_matches(const struct virtual_engine *ve,
> {
> const struct intel_engine_cs *active;
>
> + if (!(rq->execution_mask & engine->mask)) /* We peeked too soon! */
> + return false;
> +
> /*
> * We track when the HW has completed saving the context image
> * (i.e. when we have seen the final CS event switching out of
> @@ -3159,12 +3183,44 @@ static const struct intel_context_ops virtual_context_ops = {
> .destroy = virtual_context_destroy,
> };
>
> +static intel_engine_mask_t virtual_submission_mask(struct virtual_engine *ve)
> +{
> + struct i915_request *rq;
> + intel_engine_mask_t mask;
> +
> + rq = READ_ONCE(ve->request);
> + if (!rq)
> + return 0;
> +
> + /* The rq is ready for submission; rq->execution_mask is now stable. */
> + mask = rq->execution_mask;
> + if (unlikely(!mask)) {
> + /* Invalid selection, submit to a random engine in error */
> + i915_request_skip(rq, -ENODEV);
> + mask = ve->siblings[0]->mask;
> + }
> +
> + GEM_TRACE("%s: rq=%llx:%lld, mask=%x, prio=%d\n",
> + ve->base.name,
> + rq->fence.context, rq->fence.seqno,
> + mask, ve->base.execlists.queue_priority_hint);
> +
> + return mask;
> +}
> +
> static void virtual_submission_tasklet(unsigned long data)
> {
> struct virtual_engine * const ve = (struct virtual_engine *)data;
> const int prio = ve->base.execlists.queue_priority_hint;
> + intel_engine_mask_t mask;
> unsigned int n;
>
> + rcu_read_lock();
> + mask = virtual_submission_mask(ve);
> + rcu_read_unlock();
> + if (unlikely(!mask))
> + return;
> +
> local_irq_disable();
> for (n = 0; READ_ONCE(ve->request) && n < ve->num_siblings; n++) {
> struct intel_engine_cs *sibling = ve->siblings[n];
> @@ -3172,6 +3228,17 @@ static void virtual_submission_tasklet(unsigned long data)
> struct rb_node **parent, *rb;
> bool first;
>
> + if (unlikely(!(mask & sibling->mask))) {
> + if (!RB_EMPTY_NODE(&node->rb)) {
> + spin_lock(&sibling->timeline.lock);
> + rb_erase_cached(&node->rb,
> + &sibling->execlists.virtual);
> + RB_CLEAR_NODE(&node->rb);
> + spin_unlock(&sibling->timeline.lock);
> + }
> + continue;
> + }
> +
> spin_lock(&sibling->timeline.lock);
>
> if (!RB_EMPTY_NODE(&node->rb)) {
> diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
> index ccc0b6350123..2ef7639ee665 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
> @@ -1489,6 +1489,136 @@ static int live_virtual_engine(void *arg)
> return err;
> }
>
> +static int mask_virtual_engine(struct drm_i915_private *i915,
> + struct intel_engine_cs **siblings,
> + unsigned int nsibling)
> +{
> + struct i915_request *request[MAX_ENGINE_INSTANCE + 1];
> + struct i915_gem_context *ctx;
> + struct intel_context *ve;
> + struct igt_live_test t;
> + unsigned int n;
> + int err;
> +
> + /*
> + * Check that by setting the execution mask on a request, we can
> + * restrict it to our desired engine within the virtual engine.
> + */
> +
> + ctx = kernel_context(i915);
> + if (!ctx)
> + return -ENOMEM;
> +
> + ve = intel_execlists_create_virtual(ctx, siblings, nsibling);
> + if (IS_ERR(ve)) {
> + err = PTR_ERR(ve);
> + goto out_close;
> + }
> +
> + err = intel_context_pin(ve);
> + if (err)
> + goto out_put;
> +
> + err = igt_live_test_begin(&t, i915, __func__, ve->engine->name);
> + if (err)
> + goto out_unpin;
> +
> + for (n = 0; n < nsibling; n++) {
> + request[n] = i915_request_create(ve);
> + if (IS_ERR(request)) {
> + err = PTR_ERR(request);
> + nsibling = n;
> + goto out;
> + }
> +
> + /* Reverse order as it's more likely to be unnatural */
> + request[n]->execution_mask = siblings[nsibling - n - 1]->mask;
> +
> + i915_request_get(request[n]);
> + i915_request_add(request[n]);
> + }
> +
> + for (n = 0; n < nsibling; n++) {
> + if (i915_request_wait(request[n], I915_WAIT_LOCKED, HZ / 10) < 0) {
> + pr_err("%s(%s): wait for %llx:%lld timed out\n",
> + __func__, ve->engine->name,
> + request[n]->fence.context,
> + request[n]->fence.seqno);
> +
> + GEM_TRACE("%s(%s) failed at request %llx:%lld\n",
> + __func__, ve->engine->name,
> + request[n]->fence.context,
> + request[n]->fence.seqno);
> + GEM_TRACE_DUMP();
> + i915_gem_set_wedged(i915);
> + err = -EIO;
> + goto out;
> + }
> +
> + if (request[n]->engine != siblings[nsibling - n - 1]) {
> + pr_err("Executed on wrong sibling '%s', expected '%s'\n",
> + request[n]->engine->name,
> + siblings[nsibling - n - 1]->name);
> + err = -EINVAL;
> + goto out;
> + }
> + }
> +
> + err = igt_live_test_end(&t);
> + if (err)
> + goto out;
> +
> +out:
> + if (igt_flush_test(i915, I915_WAIT_LOCKED))
> + err = -EIO;
> +
> + for (n = 0; n < nsibling; n++)
> + i915_request_put(request[n]);
> +
> +out_unpin:
> + intel_context_unpin(ve);
> +out_put:
> + intel_context_put(ve);
> +out_close:
> + kernel_context_close(ctx);
> + return err;
> +}
> +
> +static int live_virtual_mask(void *arg)
> +{
> + struct drm_i915_private *i915 = arg;
> + struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1];
> + unsigned int class, inst;
> + int err = 0;
> +
> + if (USES_GUC_SUBMISSION(i915))
> + return 0;
> +
> + mutex_lock(&i915->drm.struct_mutex);
> +
> + for (class = 0; class <= MAX_ENGINE_CLASS; class++) {
> + unsigned int nsibling;
> +
> + nsibling = 0;
> + for (inst = 0; inst <= MAX_ENGINE_INSTANCE; inst++) {
> + if (!i915->engine_class[class][inst])
> + break;
> +
> + siblings[nsibling++] = i915->engine_class[class][inst];
> + }
> + if (nsibling < 2)
> + continue;
> +
> + err = mask_virtual_engine(i915, siblings, nsibling);
> + if (err)
> + goto out_unlock;
> + }
> +
> +out_unlock:
> + mutex_unlock(&i915->drm.struct_mutex);
> + return err;
> +}
> +
> int intel_execlists_live_selftests(struct drm_i915_private *i915)
> {
> static const struct i915_subtest tests[] = {
> @@ -1502,6 +1632,7 @@ int intel_execlists_live_selftests(struct drm_i915_private *i915)
> SUBTEST(live_preempt_hang),
> SUBTEST(live_preempt_smoke),
> SUBTEST(live_virtual_engine),
> + SUBTEST(live_virtual_mask),
> };
>
> if (!HAS_EXECLISTS(i915))
> diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
> index cbe4098cd6ec..3beced7daa15 100644
> --- a/drivers/gpu/drm/i915/i915_request.c
> +++ b/drivers/gpu/drm/i915/i915_request.c
> @@ -699,6 +699,7 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp)
> rq->batch = NULL;
> rq->capture_list = NULL;
> rq->waitboost = false;
> + rq->execution_mask = ALL_ENGINES;
>
> INIT_LIST_HEAD(&rq->active_list);
> INIT_LIST_HEAD(&rq->execute_cb);
> diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
> index 8025a89b5999..d7f9b2194568 100644
> --- a/drivers/gpu/drm/i915/i915_request.h
> +++ b/drivers/gpu/drm/i915/i915_request.h
> @@ -28,6 +28,8 @@
> #include <linux/dma-fence.h>
> #include <linux/lockdep.h>
>
> +#include "gt/intel_engine_types.h"
> +
> #include "i915_gem.h"
> #include "i915_scheduler.h"
> #include "i915_selftest.h"
> @@ -156,6 +158,7 @@ struct i915_request {
> */
> struct i915_sched_node sched;
> struct i915_dependency dep;
> + intel_engine_mask_t execution_mask;
>
> /*
> * A convenience pointer to the current breadcrumb value stored in
>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread
* [PATCH 16/40] drm/i915: Extend execution fence to support a callback
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (13 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 15/40] drm/i915: Apply an execution_mask to the virtual_engine Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 17/40] drm/i915/execlists: Virtual engine bonding Chris Wilson
` (30 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
In the next patch, we will want to configure the slave request
depending on which physical engine the master request is executed on.
For this, we introduce a callback from the execute fence to convey this
information.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/i915_request.c | 84 +++++++++++++++++++++++++++--
drivers/gpu/drm/i915/i915_request.h | 4 ++
2 files changed, 83 insertions(+), 5 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 3beced7daa15..1188a450804f 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -38,6 +38,8 @@ struct execute_cb {
struct list_head link;
struct irq_work work;
struct i915_sw_fence *fence;
+ void (*hook)(struct i915_request *rq, struct dma_fence *signal);
+ struct i915_request *signal;
};
static struct i915_global_request {
@@ -329,6 +331,17 @@ static void irq_execute_cb(struct irq_work *wrk)
kmem_cache_free(global.slab_execute_cbs, cb);
}
+static void irq_execute_cb_hook(struct irq_work *wrk)
+{
+ struct execute_cb *cb = container_of(wrk, typeof(*cb), work);
+
+ cb->hook(container_of(cb->fence, struct i915_request, submit),
+ &cb->signal->fence);
+ i915_request_put(cb->signal);
+
+ irq_execute_cb(wrk);
+}
+
static void __notify_execute_cb(struct i915_request *rq)
{
struct execute_cb *cb;
@@ -355,14 +368,19 @@ static void __notify_execute_cb(struct i915_request *rq)
}
static int
-i915_request_await_execution(struct i915_request *rq,
- struct i915_request *signal,
- gfp_t gfp)
+__i915_request_await_execution(struct i915_request *rq,
+ struct i915_request *signal,
+ void (*hook)(struct i915_request *rq,
+ struct dma_fence *signal),
+ gfp_t gfp)
{
struct execute_cb *cb;
- if (i915_request_is_active(signal))
+ if (i915_request_is_active(signal)) {
+ if (hook)
+ hook(rq, &signal->fence);
return 0;
+ }
cb = kmem_cache_alloc(global.slab_execute_cbs, gfp);
if (!cb)
@@ -372,8 +390,18 @@ i915_request_await_execution(struct i915_request *rq,
i915_sw_fence_await(cb->fence);
init_irq_work(&cb->work, irq_execute_cb);
+ if (hook) {
+ cb->hook = hook;
+ cb->signal = i915_request_get(signal);
+ cb->work.func = irq_execute_cb_hook;
+ }
+
spin_lock_irq(&signal->lock);
if (i915_request_is_active(signal)) {
+ if (hook) {
+ hook(rq, &signal->fence);
+ i915_request_put(signal);
+ }
i915_sw_fence_complete(cb->fence);
kmem_cache_free(global.slab_execute_cbs, cb);
} else {
@@ -845,7 +873,7 @@ emit_semaphore_wait(struct i915_request *to,
return err;
/* Only submit our spinner after the signaler is running! */
- err = i915_request_await_execution(to, from, gfp);
+ err = __i915_request_await_execution(to, from, NULL, gfp);
if (err)
return err;
@@ -971,6 +999,52 @@ i915_request_await_dma_fence(struct i915_request *rq, struct dma_fence *fence)
return 0;
}
+int
+i915_request_await_execution(struct i915_request *rq,
+ struct dma_fence *fence,
+ void (*hook)(struct i915_request *rq,
+ struct dma_fence *signal))
+{
+ struct dma_fence **child = &fence;
+ unsigned int nchild = 1;
+ int ret;
+
+ if (dma_fence_is_array(fence)) {
+ struct dma_fence_array *array = to_dma_fence_array(fence);
+
+ /* XXX Error for signal-on-any fence arrays */
+
+ child = array->fences;
+ nchild = array->num_fences;
+ GEM_BUG_ON(!nchild);
+ }
+
+ do {
+ fence = *child++;
+ if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
+ continue;
+
+ /*
+ * We don't squash repeated fence dependencies here as we
+ * want to run our callback in all cases.
+ */
+
+ if (dma_fence_is_i915(fence))
+ ret = __i915_request_await_execution(rq,
+ to_request(fence),
+ hook,
+ I915_FENCE_GFP);
+ else
+ ret = i915_sw_fence_await_dma_fence(&rq->submit, fence,
+ I915_FENCE_TIMEOUT,
+ GFP_KERNEL);
+ if (ret < 0)
+ return ret;
+ } while (--nchild);
+
+ return 0;
+}
+
/**
* i915_request_await_object - set this request to (async) wait upon a bo
* @to: request we are wishing to use
diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
index d7f9b2194568..c9f7d07991c8 100644
--- a/drivers/gpu/drm/i915/i915_request.h
+++ b/drivers/gpu/drm/i915/i915_request.h
@@ -283,6 +283,10 @@ int i915_request_await_object(struct i915_request *to,
bool write);
int i915_request_await_dma_fence(struct i915_request *rq,
struct dma_fence *fence);
+int i915_request_await_execution(struct i915_request *rq,
+ struct dma_fence *fence,
+ void (*hook)(struct i915_request *rq,
+ struct dma_fence *signal));
void i915_request_add(struct i915_request *rq);
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 17/40] drm/i915/execlists: Virtual engine bonding
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (14 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 16/40] drm/i915: Extend execution fence to support a callback Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 18/40] drm/i915: Allow specification of parallel execbuf Chris Wilson
` (29 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Some users require that when a master batch is executed on one particular
engine, a companion batch is run simultaneously on a specific slave
engine. For this purpose, we introduce virtual engine bonding, allowing
maps of master:slaves to be constructed to constrain which physical
engines a virtual engine may select given a fence on a master engine.
For the moment, we continue to ignore the issue of preemption deferring
the master request for later. Ideally, we would like to then also remove
the slave and run something else rather than have it stall the pipeline.
With load balancing, we should be able to move workload around it, but
there is a similar stall on the master pipeline while it may wait for
the slave to be executed. At the cost of more latency for the bonded
request, it may be interesting to launch both on their engines in
lockstep. (Bubbles abound.)
Opens: Also what about bonding an engine as its own master? It doesn't
break anything internally, so allow the silliness.
v2: Emancipate the bonds
v3: Couple in delayed scheduling for the selftests
v4: Handle invalid mutually exclusive bonding
v5: Mention what the uapi does
v6: s/nbond/num_bonds/
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/gt/intel_engine_types.h | 7 +
drivers/gpu/drm/i915/gt/intel_lrc.c | 98 +++++++++
drivers/gpu/drm/i915/gt/intel_lrc.h | 4 +
drivers/gpu/drm/i915/gt/selftest_lrc.c | 191 ++++++++++++++++++
drivers/gpu/drm/i915/i915_gem_context.c | 84 ++++++++
drivers/gpu/drm/i915/selftests/lib_sw_fence.c | 3 +
include/uapi/drm/i915_drm.h | 44 ++++
7 files changed, 431 insertions(+)
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
index 7b47e00fa082..f3fc2e8acc90 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
@@ -405,6 +405,13 @@ struct intel_engine_cs {
*/
void (*submit_request)(struct i915_request *rq);
+ /*
+ * Called on signaling of a SUBMIT_FENCE, passing along the signaling
+ * request down to the bonded pairs.
+ */
+ void (*bond_execute)(struct i915_request *rq,
+ struct dma_fence *signal);
+
/*
* Call when the priority on a request has changed and it and its
* dependencies may need rescheduling. Note the request itself may
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 69849ffb9c82..5d80f99661bc 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -191,6 +191,18 @@ struct virtual_engine {
int prio;
} nodes[I915_NUM_ENGINES];
+ /*
+ * Keep track of bonded pairs -- restrictions upon on our selection
+ * of physical engines any particular request may be submitted to.
+ * If we receive a submit-fence from a master engine, we will only
+ * use one of sibling_mask physical engines.
+ */
+ struct ve_bond {
+ const struct intel_engine_cs *master;
+ intel_engine_mask_t sibling_mask;
+ } *bonds;
+ unsigned int num_bonds;
+
/* And finally, which physical engines this virtual engine maps onto. */
unsigned int num_siblings;
struct intel_engine_cs *siblings[0];
@@ -1002,6 +1014,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
rb_erase_cached(rb, &execlists->virtual);
RB_CLEAR_NODE(rb);
+ GEM_BUG_ON(!(rq->execution_mask & engine->mask));
rq->engine = engine;
if (engine != ve->siblings[0]) {
@@ -3110,6 +3123,8 @@ static void virtual_context_destroy(struct kref *kref)
if (ve->context.state)
__execlists_context_fini(&ve->context);
+ kfree(ve->bonds);
+
i915_timeline_fini(&ve->base.timeline);
kfree(ve);
}
@@ -3306,6 +3321,38 @@ static void virtual_submit_request(struct i915_request *rq)
tasklet_schedule(&ve->base.execlists.tasklet);
}
+static struct ve_bond *
+virtual_find_bond(struct virtual_engine *ve,
+ const struct intel_engine_cs *master)
+{
+ int i;
+
+ for (i = 0; i < ve->num_bonds; i++) {
+ if (ve->bonds[i].master == master)
+ return &ve->bonds[i];
+ }
+
+ return NULL;
+}
+
+static void
+virtual_bond_execute(struct i915_request *rq, struct dma_fence *signal)
+{
+ struct virtual_engine *ve = to_virtual_engine(rq->engine);
+ struct ve_bond *bond;
+
+ bond = virtual_find_bond(ve, to_request(signal)->engine);
+ if (bond) {
+ intel_engine_mask_t old, new, cmp;
+
+ cmp = READ_ONCE(rq->execution_mask);
+ do {
+ old = cmp;
+ new = cmp & bond->sibling_mask;
+ } while ((cmp = cmpxchg(&rq->execution_mask, old, new)) != old);
+ }
+}
+
struct intel_context *
intel_execlists_create_virtual(struct i915_gem_context *ctx,
struct intel_engine_cs **siblings,
@@ -3346,6 +3393,7 @@ intel_execlists_create_virtual(struct i915_gem_context *ctx,
ve->base.schedule = i915_schedule;
ve->base.submit_request = virtual_submit_request;
+ ve->base.bond_execute = virtual_bond_execute;
ve->base.execlists.queue_priority_hint = INT_MIN;
tasklet_init(&ve->base.execlists.tasklet,
@@ -3435,9 +3483,59 @@ intel_execlists_clone_virtual(struct i915_gem_context *ctx,
if (IS_ERR(dst))
return dst;
+ if (se->num_bonds) {
+ struct virtual_engine *de = to_virtual_engine(dst->engine);
+
+ de->bonds = kmemdup(se->bonds,
+ sizeof(*se->bonds) * se->num_bonds,
+ GFP_KERNEL);
+ if (!de->bonds) {
+ intel_context_put(dst);
+ return ERR_PTR(-ENOMEM);
+ }
+
+ de->num_bonds = se->num_bonds;
+ }
+
return dst;
}
+int intel_virtual_engine_attach_bond(struct intel_engine_cs *engine,
+ const struct intel_engine_cs *master,
+ const struct intel_engine_cs *sibling)
+{
+ struct virtual_engine *ve = to_virtual_engine(engine);
+ struct ve_bond *bond;
+ int n;
+
+ /* Sanity check the sibling is part of the virtual engine */
+ for (n = 0; n < ve->num_siblings; n++)
+ if (sibling == ve->siblings[n])
+ break;
+ if (n == ve->num_siblings)
+ return -EINVAL;
+
+ bond = virtual_find_bond(ve, master);
+ if (bond) {
+ bond->sibling_mask |= sibling->mask;
+ return 0;
+ }
+
+ bond = krealloc(ve->bonds,
+ sizeof(*bond) * (ve->num_bonds + 1),
+ GFP_KERNEL);
+ if (!bond)
+ return -ENOMEM;
+
+ bond[ve->num_bonds].master = master;
+ bond[ve->num_bonds].sibling_mask = sibling->mask;
+
+ ve->bonds = bond;
+ ve->num_bonds++;
+
+ return 0;
+}
+
void intel_execlists_show_requests(struct intel_engine_cs *engine,
struct drm_printer *m,
void (*show_request)(struct drm_printer *m,
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.h b/drivers/gpu/drm/i915/gt/intel_lrc.h
index 5530606052e5..e029aee87adf 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.h
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.h
@@ -123,4 +123,8 @@ struct intel_context *
intel_execlists_clone_virtual(struct i915_gem_context *ctx,
struct intel_engine_cs *src);
+int intel_virtual_engine_attach_bond(struct intel_engine_cs *engine,
+ const struct intel_engine_cs *master,
+ const struct intel_engine_cs *sibling);
+
#endif /* _INTEL_LRC_H_ */
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index 2ef7639ee665..475d0406f7ce 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -13,6 +13,7 @@
#include "selftests/igt_gem_utils.h"
#include "selftests/igt_live_test.h"
#include "selftests/igt_spinner.h"
+#include "selftests/lib_sw_fence.h"
#include "selftests/mock_context.h"
static int live_sanitycheck(void *arg)
@@ -1619,6 +1620,195 @@ static int live_virtual_mask(void *arg)
return err;
}
+static int bond_virtual_engine(struct drm_i915_private *i915,
+ unsigned int class,
+ struct intel_engine_cs **siblings,
+ unsigned int nsibling,
+ unsigned int flags)
+#define BOND_SCHEDULE BIT(0)
+{
+ struct intel_engine_cs *master;
+ struct i915_gem_context *ctx;
+ struct i915_request *rq[16];
+ enum intel_engine_id id;
+ unsigned long n;
+ int err;
+
+ GEM_BUG_ON(nsibling >= ARRAY_SIZE(rq) - 1);
+
+ ctx = kernel_context(i915);
+ if (!ctx)
+ return -ENOMEM;
+
+ err = 0;
+ rq[0] = ERR_PTR(-ENOMEM);
+ for_each_engine(master, i915, id) {
+ struct i915_sw_fence fence = {};
+
+ if (master->class == class)
+ continue;
+
+ memset_p((void *)rq, ERR_PTR(-EINVAL), ARRAY_SIZE(rq));
+
+ rq[0] = igt_request_alloc(ctx, master);
+ if (IS_ERR(rq[0])) {
+ err = PTR_ERR(rq[0]);
+ goto out;
+ }
+ i915_request_get(rq[0]);
+
+ if (flags & BOND_SCHEDULE) {
+ onstack_fence_init(&fence);
+ err = i915_sw_fence_await_sw_fence_gfp(&rq[0]->submit,
+ &fence,
+ GFP_KERNEL);
+ }
+ i915_request_add(rq[0]);
+ if (err < 0)
+ goto out;
+
+ for (n = 0; n < nsibling; n++) {
+ struct intel_context *ve;
+
+ ve = intel_execlists_create_virtual(ctx,
+ siblings,
+ nsibling);
+ if (IS_ERR(ve)) {
+ err = PTR_ERR(ve);
+ onstack_fence_fini(&fence);
+ goto out;
+ }
+
+ err = intel_virtual_engine_attach_bond(ve->engine,
+ master,
+ siblings[n]);
+ if (err) {
+ intel_context_put(ve);
+ onstack_fence_fini(&fence);
+ goto out;
+ }
+
+ err = intel_context_pin(ve);
+ intel_context_put(ve);
+ if (err) {
+ onstack_fence_fini(&fence);
+ goto out;
+ }
+
+ rq[n + 1] = i915_request_create(ve);
+ intel_context_unpin(ve);
+ if (IS_ERR(rq[n + 1])) {
+ err = PTR_ERR(rq[n + 1]);
+ onstack_fence_fini(&fence);
+ goto out;
+ }
+ i915_request_get(rq[n + 1]);
+
+ err = i915_request_await_execution(rq[n + 1],
+ &rq[0]->fence,
+ ve->engine->bond_execute);
+ i915_request_add(rq[n + 1]);
+ if (err < 0) {
+ onstack_fence_fini(&fence);
+ goto out;
+ }
+ }
+ onstack_fence_fini(&fence);
+
+ if (i915_request_wait(rq[0],
+ I915_WAIT_LOCKED,
+ HZ / 10) < 0) {
+ pr_err("Master request did not execute (on %s)!\n",
+ rq[0]->engine->name);
+ err = -EIO;
+ goto out;
+ }
+
+ for (n = 0; n < nsibling; n++) {
+ if (i915_request_wait(rq[n + 1],
+ I915_WAIT_LOCKED,
+ MAX_SCHEDULE_TIMEOUT) < 0) {
+ err = -EIO;
+ goto out;
+ }
+
+ if (rq[n + 1]->engine != siblings[n]) {
+ pr_err("Bonded request did not execute on target engine: expected %s, used %s; master was %s\n",
+ siblings[n]->name,
+ rq[n + 1]->engine->name,
+ rq[0]->engine->name);
+ err = -EINVAL;
+ goto out;
+ }
+ }
+
+ for (n = 0; !IS_ERR(rq[n]); n++)
+ i915_request_put(rq[n]);
+ rq[0] = ERR_PTR(-ENOMEM);
+ }
+
+out:
+ for (n = 0; !IS_ERR(rq[n]); n++)
+ i915_request_put(rq[n]);
+ if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ err = -EIO;
+
+ kernel_context_close(ctx);
+ return err;
+}
+
+static int live_virtual_bond(void *arg)
+{
+ static const struct phase {
+ const char *name;
+ unsigned int flags;
+ } phases[] = {
+ { "", 0 },
+ { "schedule", BOND_SCHEDULE },
+ { },
+ };
+ struct drm_i915_private *i915 = arg;
+ struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1];
+ unsigned int class, inst;
+ int err = 0;
+
+ if (USES_GUC_SUBMISSION(i915))
+ return 0;
+
+ mutex_lock(&i915->drm.struct_mutex);
+
+ for (class = 0; class <= MAX_ENGINE_CLASS; class++) {
+ const struct phase *p;
+ int nsibling;
+
+ nsibling = 0;
+ for (inst = 0; inst <= MAX_ENGINE_INSTANCE; inst++) {
+ if (!i915->engine_class[class][inst])
+ break;
+
+ GEM_BUG_ON(nsibling == ARRAY_SIZE(siblings));
+ siblings[nsibling++] = i915->engine_class[class][inst];
+ }
+ if (nsibling < 2)
+ continue;
+
+ for (p = phases; p->name; p++) {
+ err = bond_virtual_engine(i915,
+ class, siblings, nsibling,
+ p->flags);
+ if (err) {
+ pr_err("%s(%s): failed class=%d, nsibling=%d, err=%d\n",
+ __func__, p->name, class, nsibling, err);
+ goto out_unlock;
+ }
+ }
+ }
+
+out_unlock:
+ mutex_unlock(&i915->drm.struct_mutex);
+ return err;
+}
+
int intel_execlists_live_selftests(struct drm_i915_private *i915)
{
static const struct i915_subtest tests[] = {
@@ -1633,6 +1823,7 @@ int intel_execlists_live_selftests(struct drm_i915_private *i915)
SUBTEST(live_preempt_smoke),
SUBTEST(live_virtual_engine),
SUBTEST(live_virtual_mask),
+ SUBTEST(live_virtual_bond),
};
if (!HAS_EXECLISTS(i915))
diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
index 3ae1d27130a8..1825479864d6 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/i915_gem_context.c
@@ -1504,8 +1504,92 @@ set_engines__load_balance(struct i915_user_extension __user *base, void *data)
return err;
}
+static int
+set_engines__bond(struct i915_user_extension __user *base, void *data)
+{
+ struct i915_context_engines_bond __user *ext =
+ container_of_user(base, typeof(*ext), base);
+ const struct set_engines *set = data;
+ struct i915_engine_class_instance ci;
+ struct intel_engine_cs *virtual;
+ struct intel_engine_cs *master;
+ u16 idx, num_bonds;
+ int err, n;
+
+ if (get_user(idx, &ext->virtual_index))
+ return -EFAULT;
+
+ if (idx >= set->engines->num_engines) {
+ DRM_DEBUG("Invalid index for virtual engine: %d >= %d\n",
+ idx, set->engines->num_engines);
+ return -EINVAL;
+ }
+
+ idx = array_index_nospec(idx, set->engines->num_engines);
+ if (!set->engines->engines[idx]) {
+ DRM_DEBUG("Invalid engine at %d\n", idx);
+ return -EINVAL;
+ }
+ virtual = set->engines->engines[idx]->engine;
+
+ err = check_user_mbz(&ext->flags);
+ if (err)
+ return err;
+
+ for (n = 0; n < ARRAY_SIZE(ext->mbz64); n++) {
+ err = check_user_mbz(&ext->mbz64[n]);
+ if (err)
+ return err;
+ }
+
+ if (copy_from_user(&ci, &ext->master, sizeof(ci)))
+ return -EFAULT;
+
+ master = intel_engine_lookup_user(set->ctx->i915,
+ ci.engine_class, ci.engine_instance);
+ if (!master) {
+ DRM_DEBUG("Unrecognised master engine: { class:%u, instance:%u }\n",
+ ci.engine_class, ci.engine_instance);
+ return -EINVAL;
+ }
+
+ if (get_user(num_bonds, &ext->num_bonds))
+ return -EFAULT;
+
+ for (n = 0; n < num_bonds; n++) {
+ struct intel_engine_cs *bond;
+
+ if (copy_from_user(&ci, &ext->engines[n], sizeof(ci)))
+ return -EFAULT;
+
+ bond = intel_engine_lookup_user(set->ctx->i915,
+ ci.engine_class,
+ ci.engine_instance);
+ if (!bond) {
+ DRM_DEBUG("Unrecognised engine[%d] for bonding: { class:%d, instance: %d }\n",
+ n, ci.engine_class, ci.engine_instance);
+ return -EINVAL;
+ }
+
+ /*
+ * A non-virtual engine has no siblings to choose between; and
+ * a submit fence will always be directed to the one engine.
+ */
+ if (intel_engine_is_virtual(virtual)) {
+ err = intel_virtual_engine_attach_bond(virtual,
+ master,
+ bond);
+ if (err)
+ return err;
+ }
+ }
+
+ return 0;
+}
+
static const i915_user_extension_fn set_engines__extensions[] = {
[I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE] = set_engines__load_balance,
+ [I915_CONTEXT_ENGINES_EXT_BOND] = set_engines__bond,
};
static int
diff --git a/drivers/gpu/drm/i915/selftests/lib_sw_fence.c b/drivers/gpu/drm/i915/selftests/lib_sw_fence.c
index 2bfa72c1654b..b976c12817c5 100644
--- a/drivers/gpu/drm/i915/selftests/lib_sw_fence.c
+++ b/drivers/gpu/drm/i915/selftests/lib_sw_fence.c
@@ -45,6 +45,9 @@ void __onstack_fence_init(struct i915_sw_fence *fence,
void onstack_fence_fini(struct i915_sw_fence *fence)
{
+ if (!fence->flags)
+ return;
+
i915_sw_fence_commit(fence);
i915_sw_fence_fini(fence);
}
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index f9770948161c..e2da9027bcdf 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -1543,6 +1543,10 @@ struct drm_i915_gem_context_param {
* sized argument, will revert back to default settings.
*
* See struct i915_context_param_engines.
+ *
+ * Extensions:
+ * i915_context_engines_load_balance (I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE)
+ * i915_context_engines_bond (I915_CONTEXT_ENGINES_EXT_BOND)
*/
#define I915_CONTEXT_PARAM_ENGINES 0xa
/* Must be kept compact -- no holes and well documented */
@@ -1646,9 +1650,49 @@ struct i915_context_engines_load_balance {
struct i915_engine_class_instance engines[N__]; \
} __attribute__((packed)) name__
+/*
+ * i915_context_engines_bond:
+ *
+ * Constructed bonded pairs for execution within a virtual engine.
+ *
+ * All engines are equal, but some are more equal than others. Given
+ * the distribution of resources in the HW, it may be preferable to run
+ * a request on a given subset of engines in parallel to a request on a
+ * specific engine. We enable this selection of engines within a virtual
+ * engine by specifying bonding pairs, for any given master engine we will
+ * only execute on one of the corresponding siblings within the virtual engine.
+ *
+ * To execute a request in parallel on the master engine and a sibling requires
+ * coordination with a I915_EXEC_FENCE_SUBMIT.
+ */
+struct i915_context_engines_bond {
+ struct i915_user_extension base;
+
+ struct i915_engine_class_instance master;
+
+ __u16 virtual_index; /* index of virtual engine in ctx->engines[] */
+ __u16 num_bonds;
+
+ __u64 flags; /* all undefined flags must be zero */
+ __u64 mbz64[4]; /* reserved for future use; must be zero */
+
+ struct i915_engine_class_instance engines[0];
+} __attribute__((packed));
+
+#define I915_DEFINE_CONTEXT_ENGINES_BOND(name__, N__) struct { \
+ struct i915_user_extension base; \
+ struct i915_engine_class_instance master; \
+ __u16 virtual_index; \
+ __u16 num_bonds; \
+ __u64 flags; \
+ __u64 mbz64[4]; \
+ struct i915_engine_class_instance engines[N__]; \
+} __attribute__((packed)) name__
+
struct i915_context_param_engines {
__u64 extensions; /* linked chain of extension blocks, 0 terminates */
#define I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE 0 /* see i915_context_engines_load_balance */
+#define I915_CONTEXT_ENGINES_EXT_BOND 1 /* see i915_context_engines_bond */
struct i915_engine_class_instance engines[0];
} __attribute__((packed));
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 18/40] drm/i915: Allow specification of parallel execbuf
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (15 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 17/40] drm/i915/execlists: Virtual engine bonding Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 19/40] drm/i915: Split GEM object type definition to its own header Chris Wilson
` (28 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
There is a desire to split a task onto two engines and have them run at
the same time, e.g. scanline interleaving to spread the workload evenly.
Through the use of the out-fence from the first execbuf, we can
coordinate secondary execbuf to only become ready simultaneously with
the first, so that with all things idle the second execbufs are executed
in parallel with the first. The key difference here between the new
EXEC_FENCE_SUBMIT and the existing EXEC_FENCE_IN is that the in-fence
waits for the completion of the first request (so that all of its
rendering results are visible to the second execbuf, the more common
userspace fence requirement).
Since we only have a single input fence slot, userspace cannot mix an
in-fence and a submit-fence. It has to use one or the other! This is not
such a harsh requirement, since by virtue of the submit-fence, the
secondary execbuf inherit all of the dependencies from the first
request, and for the application the dependencies should be common
between the primary and secondary execbuf.
Suggested-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Testcase: igt/gem_exec_fence/parallel
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/i915_drv.c | 1 +
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 25 +++++++++++++++++++++-
include/uapi/drm/i915_drm.h | 17 ++++++++++++++-
3 files changed, 41 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 5061cb32856b..83d2eb9e74cb 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -443,6 +443,7 @@ static int i915_getparam_ioctl(struct drm_device *dev, void *data,
case I915_PARAM_HAS_EXEC_CAPTURE:
case I915_PARAM_HAS_EXEC_BATCH_FIRST:
case I915_PARAM_HAS_EXEC_FENCE_ARRAY:
+ case I915_PARAM_HAS_EXEC_SUBMIT_FENCE:
/* For the time being all of these are always true;
* if some supported hardware does not have one of these
* features this value needs to be provided from
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index d6c5220addd0..7ce25b54c57b 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -2318,6 +2318,7 @@ i915_gem_do_execbuffer(struct drm_device *dev,
{
struct i915_execbuffer eb;
struct dma_fence *in_fence = NULL;
+ struct dma_fence *exec_fence = NULL;
struct sync_file *out_fence = NULL;
int out_fence_fd = -1;
int err;
@@ -2360,11 +2361,24 @@ i915_gem_do_execbuffer(struct drm_device *dev,
return -EINVAL;
}
+ if (args->flags & I915_EXEC_FENCE_SUBMIT) {
+ if (in_fence) {
+ err = -EINVAL;
+ goto err_in_fence;
+ }
+
+ exec_fence = sync_file_get_fence(lower_32_bits(args->rsvd2));
+ if (!exec_fence) {
+ err = -EINVAL;
+ goto err_in_fence;
+ }
+ }
+
if (args->flags & I915_EXEC_FENCE_OUT) {
out_fence_fd = get_unused_fd_flags(O_CLOEXEC);
if (out_fence_fd < 0) {
err = out_fence_fd;
- goto err_in_fence;
+ goto err_exec_fence;
}
}
@@ -2494,6 +2508,13 @@ i915_gem_do_execbuffer(struct drm_device *dev,
goto err_request;
}
+ if (exec_fence) {
+ err = i915_request_await_execution(eb.request, exec_fence,
+ eb.engine->bond_execute);
+ if (err < 0)
+ goto err_request;
+ }
+
if (fences) {
err = await_fence_array(&eb, fences);
if (err)
@@ -2555,6 +2576,8 @@ i915_gem_do_execbuffer(struct drm_device *dev,
err_out_fence:
if (out_fence_fd != -1)
put_unused_fd(out_fence_fd);
+err_exec_fence:
+ dma_fence_put(exec_fence);
err_in_fence:
dma_fence_put(in_fence);
return err;
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index e2da9027bcdf..bdb00ec1f8be 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -604,6 +604,12 @@ typedef struct drm_i915_irq_wait {
*/
#define I915_PARAM_MMAP_GTT_COHERENT 52
+/*
+ * Query whether DRM_I915_GEM_EXECBUFFER2 supports coordination of parallel
+ * execution through use of explicit fence support.
+ * See I915_EXEC_FENCE_OUT and I915_EXEC_FENCE_SUBMIT.
+ */
+#define I915_PARAM_HAS_EXEC_SUBMIT_FENCE 53
/* Must be kept compact -- no holes and well documented */
typedef struct drm_i915_getparam {
@@ -1126,7 +1132,16 @@ struct drm_i915_gem_execbuffer2 {
*/
#define I915_EXEC_FENCE_ARRAY (1<<19)
-#define __I915_EXEC_UNKNOWN_FLAGS (-(I915_EXEC_FENCE_ARRAY<<1))
+/*
+ * Setting I915_EXEC_FENCE_SUBMIT implies that lower_32_bits(rsvd2) represent
+ * a sync_file fd to wait upon (in a nonblocking manner) prior to executing
+ * the batch.
+ *
+ * Returns -EINVAL if the sync_file fd cannot be found.
+ */
+#define I915_EXEC_FENCE_SUBMIT (1 << 20)
+
+#define __I915_EXEC_UNKNOWN_FLAGS (-(I915_EXEC_FENCE_SUBMIT << 1))
#define I915_EXEC_CONTEXT_ID_MASK (0xffffffff)
#define i915_execbuffer2_set_context_id(eb2, context) \
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 19/40] drm/i915: Split GEM object type definition to its own header
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (16 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 18/40] drm/i915: Allow specification of parallel execbuf Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 20/40] drm/i915: Pull GEM ioctls interface to its own file Chris Wilson
` (27 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx; +Cc: Matthew Auld
For convenience in avoiding inline spaghetti, keep the type definition
as a separate header.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
---
drivers/gpu/drm/i915/Makefile | 1 +
drivers/gpu/drm/i915/gem/Makefile | 1 +
| 16 +
.../gpu/drm/i915/gem/i915_gem_object_types.h | 285 +++++++++++++++++
drivers/gpu/drm/i915/gt/intel_engine_types.h | 1 +
drivers/gpu/drm/i915/i915_drv.h | 3 +-
drivers/gpu/drm/i915/i915_gem_batch_pool.h | 3 +-
drivers/gpu/drm/i915/i915_gem_gtt.h | 1 +
drivers/gpu/drm/i915/i915_gem_object.h | 295 +-----------------
9 files changed, 312 insertions(+), 294 deletions(-)
create mode 100644 drivers/gpu/drm/i915/gem/Makefile
create mode 100644 drivers/gpu/drm/i915/gem/Makefile.header-test
create mode 100644 drivers/gpu/drm/i915/gem/i915_gem_object_types.h
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index 68106fe35a04..96344c9a0726 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -85,6 +85,7 @@ gt-$(CONFIG_DRM_I915_SELFTEST) += \
i915-y += $(gt-y)
# GEM (Graphics Execution Management) code
+obj-y += gem/
i915-y += \
i915_active.o \
i915_cmd_parser.o \
diff --git a/drivers/gpu/drm/i915/gem/Makefile b/drivers/gpu/drm/i915/gem/Makefile
new file mode 100644
index 000000000000..07e7b8b840ea
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/Makefile
@@ -0,0 +1 @@
+include $(src)/Makefile.header-test # Extra header tests
--git a/drivers/gpu/drm/i915/gem/Makefile.header-test b/drivers/gpu/drm/i915/gem/Makefile.header-test
new file mode 100644
index 000000000000..61e06cbb4b32
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/Makefile.header-test
@@ -0,0 +1,16 @@
+# SPDX-License-Identifier: MIT
+# Copyright © 2019 Intel Corporation
+
+# Test the headers are compilable as standalone units
+header_test := $(notdir $(wildcard $(src)/*.h))
+
+quiet_cmd_header_test = HDRTEST $@
+ cmd_header_test = echo "\#include \"$(<F)\"" > $@
+
+header_test_%.c: %.h
+ $(call cmd,header_test)
+
+extra-$(CONFIG_DRM_I915_WERROR) += \
+ $(foreach h,$(header_test),$(patsubst %.h,header_test_%.o,$(h)))
+
+clean-files += $(foreach h,$(header_test),$(patsubst %.h,header_test_%.c,$(h)))
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
new file mode 100644
index 000000000000..e4b50944f553
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
@@ -0,0 +1,285 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2016 Intel Corporation
+ */
+
+#ifndef __I915_GEM_OBJECT_TYPES_H__
+#define __I915_GEM_OBJECT_TYPES_H__
+
+#include <linux/reservation.h>
+
+#include <drm/drm_gem.h>
+
+#include "../i915_active.h"
+#include "../i915_selftest.h"
+
+struct drm_i915_gem_object;
+
+/*
+ * struct i915_lut_handle tracks the fast lookups from handle to vma used
+ * for execbuf. Although we use a radixtree for that mapping, in order to
+ * remove them as the object or context is closed, we need a secondary list
+ * and a translation entry (i915_lut_handle).
+ */
+struct i915_lut_handle {
+ struct list_head obj_link;
+ struct list_head ctx_link;
+ struct i915_gem_context *ctx;
+ u32 handle;
+};
+
+struct drm_i915_gem_object_ops {
+ unsigned int flags;
+#define I915_GEM_OBJECT_HAS_STRUCT_PAGE BIT(0)
+#define I915_GEM_OBJECT_IS_SHRINKABLE BIT(1)
+#define I915_GEM_OBJECT_IS_PROXY BIT(2)
+#define I915_GEM_OBJECT_ASYNC_CANCEL BIT(3)
+
+ /* Interface between the GEM object and its backing storage.
+ * get_pages() is called once prior to the use of the associated set
+ * of pages before to binding them into the GTT, and put_pages() is
+ * called after we no longer need them. As we expect there to be
+ * associated cost with migrating pages between the backing storage
+ * and making them available for the GPU (e.g. clflush), we may hold
+ * onto the pages after they are no longer referenced by the GPU
+ * in case they may be used again shortly (for example migrating the
+ * pages to a different memory domain within the GTT). put_pages()
+ * will therefore most likely be called when the object itself is
+ * being released or under memory pressure (where we attempt to
+ * reap pages for the shrinker).
+ */
+ int (*get_pages)(struct drm_i915_gem_object *obj);
+ void (*put_pages)(struct drm_i915_gem_object *obj,
+ struct sg_table *pages);
+
+ int (*pwrite)(struct drm_i915_gem_object *obj,
+ const struct drm_i915_gem_pwrite *arg);
+
+ int (*dmabuf_export)(struct drm_i915_gem_object *obj);
+ void (*release)(struct drm_i915_gem_object *obj);
+};
+
+struct drm_i915_gem_object {
+ struct drm_gem_object base;
+
+ const struct drm_i915_gem_object_ops *ops;
+
+ struct {
+ /**
+ * @vma.lock: protect the list/tree of vmas
+ */
+ spinlock_t lock;
+
+ /**
+ * @vma.list: List of VMAs backed by this object
+ *
+ * The VMA on this list are ordered by type, all GGTT vma are
+ * placed at the head and all ppGTT vma are placed at the tail.
+ * The different types of GGTT vma are unordered between
+ * themselves, use the @vma.tree (which has a defined order
+ * between all VMA) to quickly find an exact match.
+ */
+ struct list_head list;
+
+ /**
+ * @vma.tree: Ordered tree of VMAs backed by this object
+ *
+ * All VMA created for this object are placed in the @vma.tree
+ * for fast retrieval via a binary search in
+ * i915_vma_instance(). They are also added to @vma.list for
+ * easy iteration.
+ */
+ struct rb_root tree;
+ } vma;
+
+ /**
+ * @lut_list: List of vma lookup entries in use for this object.
+ *
+ * If this object is closed, we need to remove all of its VMA from
+ * the fast lookup index in associated contexts; @lut_list provides
+ * this translation from object to context->handles_vma.
+ */
+ struct list_head lut_list;
+
+ /** Stolen memory for this object, instead of being backed by shmem. */
+ struct drm_mm_node *stolen;
+ union {
+ struct rcu_head rcu;
+ struct llist_node freed;
+ };
+
+ /**
+ * Whether the object is currently in the GGTT mmap.
+ */
+ unsigned int userfault_count;
+ struct list_head userfault_link;
+
+ struct list_head batch_pool_link;
+ I915_SELFTEST_DECLARE(struct list_head st_link);
+
+ unsigned long flags;
+
+ /**
+ * Have we taken a reference for the object for incomplete GPU
+ * activity?
+ */
+#define I915_BO_ACTIVE_REF 0
+
+ /*
+ * Is the object to be mapped as read-only to the GPU
+ * Only honoured if hardware has relevant pte bit
+ */
+ unsigned int cache_level:3;
+ unsigned int cache_coherent:2;
+#define I915_BO_CACHE_COHERENT_FOR_READ BIT(0)
+#define I915_BO_CACHE_COHERENT_FOR_WRITE BIT(1)
+ unsigned int cache_dirty:1;
+
+ /**
+ * @read_domains: Read memory domains.
+ *
+ * These monitor which caches contain read/write data related to the
+ * object. When transitioning from one set of domains to another,
+ * the driver is called to ensure that caches are suitably flushed and
+ * invalidated.
+ */
+ u16 read_domains;
+
+ /**
+ * @write_domain: Corresponding unique write memory domain.
+ */
+ u16 write_domain;
+
+ atomic_t frontbuffer_bits;
+ unsigned int frontbuffer_ggtt_origin; /* write once */
+ struct i915_active_request frontbuffer_write;
+
+ /** Current tiling stride for the object, if it's tiled. */
+ unsigned int tiling_and_stride;
+#define FENCE_MINIMUM_STRIDE 128 /* See i915_tiling_ok() */
+#define TILING_MASK (FENCE_MINIMUM_STRIDE - 1)
+#define STRIDE_MASK (~TILING_MASK)
+
+ /** Count of VMA actually bound by this object */
+ unsigned int bind_count;
+ unsigned int active_count;
+ /** Count of how many global VMA are currently pinned for use by HW */
+ unsigned int pin_global;
+
+ struct {
+ struct mutex lock; /* protects the pages and their use */
+ atomic_t pages_pin_count;
+
+ struct sg_table *pages;
+ void *mapping;
+
+ /* TODO: whack some of this into the error state */
+ struct i915_page_sizes {
+ /**
+ * The sg mask of the pages sg_table. i.e the mask of
+ * of the lengths for each sg entry.
+ */
+ unsigned int phys;
+
+ /**
+ * The gtt page sizes we are allowed to use given the
+ * sg mask and the supported page sizes. This will
+ * express the smallest unit we can use for the whole
+ * object, as well as the larger sizes we may be able
+ * to use opportunistically.
+ */
+ unsigned int sg;
+
+ /**
+ * The actual gtt page size usage. Since we can have
+ * multiple vma associated with this object we need to
+ * prevent any trampling of state, hence a copy of this
+ * struct also lives in each vma, therefore the gtt
+ * value here should only be read/write through the vma.
+ */
+ unsigned int gtt;
+ } page_sizes;
+
+ I915_SELFTEST_DECLARE(unsigned int page_mask);
+
+ struct i915_gem_object_page_iter {
+ struct scatterlist *sg_pos;
+ unsigned int sg_idx; /* in pages, but 32bit eek! */
+
+ struct radix_tree_root radix;
+ struct mutex lock; /* protects this cache */
+ } get_page;
+
+ /**
+ * Element within i915->mm.unbound_list or i915->mm.bound_list,
+ * locked by i915->mm.obj_lock.
+ */
+ struct list_head link;
+
+ /**
+ * Advice: are the backing pages purgeable?
+ */
+ unsigned int madv:2;
+
+ /**
+ * This is set if the object has been written to since the
+ * pages were last acquired.
+ */
+ bool dirty:1;
+
+ /**
+ * This is set if the object has been pinned due to unknown
+ * swizzling.
+ */
+ bool quirked:1;
+ } mm;
+
+ /** Breadcrumb of last rendering to the buffer.
+ * There can only be one writer, but we allow for multiple readers.
+ * If there is a writer that necessarily implies that all other
+ * read requests are complete - but we may only be lazily clearing
+ * the read requests. A read request is naturally the most recent
+ * request on a ring, so we may have two different write and read
+ * requests on one ring where the write request is older than the
+ * read request. This allows for the CPU to read from an active
+ * buffer by only waiting for the write to complete.
+ */
+ struct reservation_object *resv;
+
+ /** References from framebuffers, locks out tiling changes. */
+ unsigned int framebuffer_references;
+
+ /** Record of address bit 17 of each page at last unbind. */
+ unsigned long *bit_17;
+
+ union {
+ struct i915_gem_userptr {
+ uintptr_t ptr;
+
+ struct i915_mm_struct *mm;
+ struct i915_mmu_object *mmu_object;
+ struct work_struct *work;
+ } userptr;
+
+ unsigned long scratch;
+
+ void *gvt_info;
+ };
+
+ /** for phys allocated objects */
+ struct drm_dma_handle *phys_handle;
+
+ struct reservation_object __builtin_resv;
+};
+
+static inline struct drm_i915_gem_object *
+to_intel_bo(struct drm_gem_object *gem)
+{
+ /* Assert that to_intel_bo(NULL) == NULL */
+ BUILD_BUG_ON(offsetof(struct drm_i915_gem_object, base));
+
+ return container_of(gem, struct drm_i915_gem_object, base);
+}
+
+#endif
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
index f3fc2e8acc90..2c232501d973 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
@@ -29,6 +29,7 @@
#define I915_CMD_HASH_ORDER 9
struct dma_fence;
+struct drm_i915_gem_object;
struct drm_i915_reg_table;
struct i915_gem_context;
struct i915_request;
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 0a6ec61496f1..0b9ec8b3e5fa 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -81,7 +81,6 @@
#include "i915_gem.h"
#include "i915_gem_context.h"
#include "i915_gem_fence_reg.h"
-#include "i915_gem_object.h"
#include "i915_gem_gtt.h"
#include "i915_gpu_error.h"
#include "i915_request.h"
@@ -136,6 +135,8 @@ bool i915_error_injected(void);
__i915_printk(i915, i915_error_injected() ? KERN_DEBUG : KERN_ERR, \
fmt, ##__VA_ARGS__)
+struct drm_i915_gem_object;
+
enum hpd_pin {
HPD_NONE = 0,
HPD_TV = HPD_NONE, /* TV is known to be unreliable */
diff --git a/drivers/gpu/drm/i915/i915_gem_batch_pool.h b/drivers/gpu/drm/i915/i915_gem_batch_pool.h
index 56947daaaf65..feeeeeaa54d8 100644
--- a/drivers/gpu/drm/i915/i915_gem_batch_pool.h
+++ b/drivers/gpu/drm/i915/i915_gem_batch_pool.h
@@ -9,6 +9,7 @@
#include <linux/types.h>
+struct drm_i915_gem_object;
struct intel_engine_cs;
struct i915_gem_batch_pool {
@@ -19,7 +20,7 @@ struct i915_gem_batch_pool {
void i915_gem_batch_pool_init(struct i915_gem_batch_pool *pool,
struct intel_engine_cs *engine);
void i915_gem_batch_pool_fini(struct i915_gem_batch_pool *pool);
-struct drm_i915_gem_object*
+struct drm_i915_gem_object *
i915_gem_batch_pool_get(struct i915_gem_batch_pool *pool, size_t size);
#endif /* I915_GEM_BATCH_POOL_H */
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 2fafa04c45ec..593079f30fbe 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -61,6 +61,7 @@
struct drm_i915_file_private;
struct drm_i915_fence_reg;
+struct drm_i915_gem_object;
struct i915_vma;
typedef u32 gen6_pte_t;
diff --git a/drivers/gpu/drm/i915/i915_gem_object.h b/drivers/gpu/drm/i915/i915_gem_object.h
index ca93a40c0c87..3666b0c5f6ca 100644
--- a/drivers/gpu/drm/i915/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/i915_gem_object.h
@@ -1,308 +1,19 @@
/*
- * Copyright © 2016 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2016 Intel Corporation
*/
#ifndef __I915_GEM_OBJECT_H__
#define __I915_GEM_OBJECT_H__
-#include <linux/reservation.h>
-
-#include <drm/drm_vma_manager.h>
#include <drm/drm_gem.h>
#include <drm/drm_file.h>
#include <drm/drm_device.h>
#include <drm/i915_drm.h>
-#include "i915_request.h"
-#include "i915_selftest.h"
-
-struct drm_i915_gem_object;
-
-/*
- * struct i915_lut_handle tracks the fast lookups from handle to vma used
- * for execbuf. Although we use a radixtree for that mapping, in order to
- * remove them as the object or context is closed, we need a secondary list
- * and a translation entry (i915_lut_handle).
- */
-struct i915_lut_handle {
- struct list_head obj_link;
- struct list_head ctx_link;
- struct i915_gem_context *ctx;
- u32 handle;
-};
-
-struct drm_i915_gem_object_ops {
- unsigned int flags;
-#define I915_GEM_OBJECT_HAS_STRUCT_PAGE BIT(0)
-#define I915_GEM_OBJECT_IS_SHRINKABLE BIT(1)
-#define I915_GEM_OBJECT_IS_PROXY BIT(2)
-#define I915_GEM_OBJECT_ASYNC_CANCEL BIT(3)
-
- /* Interface between the GEM object and its backing storage.
- * get_pages() is called once prior to the use of the associated set
- * of pages before to binding them into the GTT, and put_pages() is
- * called after we no longer need them. As we expect there to be
- * associated cost with migrating pages between the backing storage
- * and making them available for the GPU (e.g. clflush), we may hold
- * onto the pages after they are no longer referenced by the GPU
- * in case they may be used again shortly (for example migrating the
- * pages to a different memory domain within the GTT). put_pages()
- * will therefore most likely be called when the object itself is
- * being released or under memory pressure (where we attempt to
- * reap pages for the shrinker).
- */
- int (*get_pages)(struct drm_i915_gem_object *);
- void (*put_pages)(struct drm_i915_gem_object *, struct sg_table *);
-
- int (*pwrite)(struct drm_i915_gem_object *,
- const struct drm_i915_gem_pwrite *);
-
- int (*dmabuf_export)(struct drm_i915_gem_object *);
- void (*release)(struct drm_i915_gem_object *);
-};
-
-struct drm_i915_gem_object {
- struct drm_gem_object base;
-
- const struct drm_i915_gem_object_ops *ops;
-
- struct {
- /**
- * @vma.lock: protect the list/tree of vmas
- */
- spinlock_t lock;
-
- /**
- * @vma.list: List of VMAs backed by this object
- *
- * The VMA on this list are ordered by type, all GGTT vma are
- * placed at the head and all ppGTT vma are placed at the tail.
- * The different types of GGTT vma are unordered between
- * themselves, use the @vma.tree (which has a defined order
- * between all VMA) to quickly find an exact match.
- */
- struct list_head list;
-
- /**
- * @vma.tree: Ordered tree of VMAs backed by this object
- *
- * All VMA created for this object are placed in the @vma.tree
- * for fast retrieval via a binary search in
- * i915_vma_instance(). They are also added to @vma.list for
- * easy iteration.
- */
- struct rb_root tree;
- } vma;
-
- /**
- * @lut_list: List of vma lookup entries in use for this object.
- *
- * If this object is closed, we need to remove all of its VMA from
- * the fast lookup index in associated contexts; @lut_list provides
- * this translation from object to context->handles_vma.
- */
- struct list_head lut_list;
-
- /** Stolen memory for this object, instead of being backed by shmem. */
- struct drm_mm_node *stolen;
- union {
- struct rcu_head rcu;
- struct llist_node freed;
- };
-
- /**
- * Whether the object is currently in the GGTT mmap.
- */
- unsigned int userfault_count;
- struct list_head userfault_link;
-
- struct list_head batch_pool_link;
- I915_SELFTEST_DECLARE(struct list_head st_link);
-
- unsigned long flags;
-
- /**
- * Have we taken a reference for the object for incomplete GPU
- * activity?
- */
-#define I915_BO_ACTIVE_REF 0
-
- /*
- * Is the object to be mapped as read-only to the GPU
- * Only honoured if hardware has relevant pte bit
- */
- unsigned int cache_level:3;
- unsigned int cache_coherent:2;
-#define I915_BO_CACHE_COHERENT_FOR_READ BIT(0)
-#define I915_BO_CACHE_COHERENT_FOR_WRITE BIT(1)
- unsigned int cache_dirty:1;
-
- /**
- * @read_domains: Read memory domains.
- *
- * These monitor which caches contain read/write data related to the
- * object. When transitioning from one set of domains to another,
- * the driver is called to ensure that caches are suitably flushed and
- * invalidated.
- */
- u16 read_domains;
-
- /**
- * @write_domain: Corresponding unique write memory domain.
- */
- u16 write_domain;
-
- atomic_t frontbuffer_bits;
- unsigned int frontbuffer_ggtt_origin; /* write once */
- struct i915_active_request frontbuffer_write;
-
- /** Current tiling stride for the object, if it's tiled. */
- unsigned int tiling_and_stride;
-#define FENCE_MINIMUM_STRIDE 128 /* See i915_tiling_ok() */
-#define TILING_MASK (FENCE_MINIMUM_STRIDE-1)
-#define STRIDE_MASK (~TILING_MASK)
-
- /** Count of VMA actually bound by this object */
- unsigned int bind_count;
- unsigned int active_count;
- /** Count of how many global VMA are currently pinned for use by HW */
- unsigned int pin_global;
-
- struct {
- struct mutex lock; /* protects the pages and their use */
- atomic_t pages_pin_count;
-
- struct sg_table *pages;
- void *mapping;
-
- /* TODO: whack some of this into the error state */
- struct i915_page_sizes {
- /**
- * The sg mask of the pages sg_table. i.e the mask of
- * of the lengths for each sg entry.
- */
- unsigned int phys;
-
- /**
- * The gtt page sizes we are allowed to use given the
- * sg mask and the supported page sizes. This will
- * express the smallest unit we can use for the whole
- * object, as well as the larger sizes we may be able
- * to use opportunistically.
- */
- unsigned int sg;
-
- /**
- * The actual gtt page size usage. Since we can have
- * multiple vma associated with this object we need to
- * prevent any trampling of state, hence a copy of this
- * struct also lives in each vma, therefore the gtt
- * value here should only be read/write through the vma.
- */
- unsigned int gtt;
- } page_sizes;
-
- I915_SELFTEST_DECLARE(unsigned int page_mask);
-
- struct i915_gem_object_page_iter {
- struct scatterlist *sg_pos;
- unsigned int sg_idx; /* in pages, but 32bit eek! */
-
- struct radix_tree_root radix;
- struct mutex lock; /* protects this cache */
- } get_page;
-
- /**
- * Element within i915->mm.unbound_list or i915->mm.bound_list,
- * locked by i915->mm.obj_lock.
- */
- struct list_head link;
-
- /**
- * Advice: are the backing pages purgeable?
- */
- unsigned int madv:2;
-
- /**
- * This is set if the object has been written to since the
- * pages were last acquired.
- */
- bool dirty:1;
-
- /**
- * This is set if the object has been pinned due to unknown
- * swizzling.
- */
- bool quirked:1;
- } mm;
-
- /** Breadcrumb of last rendering to the buffer.
- * There can only be one writer, but we allow for multiple readers.
- * If there is a writer that necessarily implies that all other
- * read requests are complete - but we may only be lazily clearing
- * the read requests. A read request is naturally the most recent
- * request on a ring, so we may have two different write and read
- * requests on one ring where the write request is older than the
- * read request. This allows for the CPU to read from an active
- * buffer by only waiting for the write to complete.
- */
- struct reservation_object *resv;
-
- /** References from framebuffers, locks out tiling changes. */
- unsigned int framebuffer_references;
-
- /** Record of address bit 17 of each page at last unbind. */
- unsigned long *bit_17;
-
- union {
- struct i915_gem_userptr {
- uintptr_t ptr;
-
- struct i915_mm_struct *mm;
- struct i915_mmu_object *mmu_object;
- struct work_struct *work;
- } userptr;
-
- unsigned long scratch;
-
- void *gvt_info;
- };
-
- /** for phys allocated objects */
- struct drm_dma_handle *phys_handle;
-
- struct reservation_object __builtin_resv;
-};
-
-static inline struct drm_i915_gem_object *
-to_intel_bo(struct drm_gem_object *gem)
-{
- /* Assert that to_intel_bo(NULL) == NULL */
- BUILD_BUG_ON(offsetof(struct drm_i915_gem_object, base));
-
- return container_of(gem, struct drm_i915_gem_object, base);
-}
+#include "gem/i915_gem_object_types.h"
struct drm_i915_gem_object *i915_gem_object_alloc(void);
void i915_gem_object_free(struct drm_i915_gem_object *obj);
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 20/40] drm/i915: Pull GEM ioctls interface to its own file
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (17 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 19/40] drm/i915: Split GEM object type definition to its own header Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 21/40] drm/i915: Move object->pages API to i915_gem_object.[ch] Chris Wilson
` (26 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx; +Cc: Matthew Auld
Declutter i915_drv/gem.h by moving the ioctl API into its own header.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
---
drivers/gpu/drm/i915/gem/i915_gem_ioctls.h | 52 ++++++++++++++++++++++
drivers/gpu/drm/i915/i915_drv.c | 1 +
drivers/gpu/drm/i915/i915_drv.h | 38 ----------------
drivers/gpu/drm/i915/i915_gem.c | 1 +
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 1 +
drivers/gpu/drm/i915/i915_gem_tiling.c | 3 ++
drivers/gpu/drm/i915/i915_gem_userptr.c | 12 +++--
7 files changed, 66 insertions(+), 42 deletions(-)
create mode 100644 drivers/gpu/drm/i915/gem/i915_gem_ioctls.h
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ioctls.h b/drivers/gpu/drm/i915/gem/i915_gem_ioctls.h
new file mode 100644
index 000000000000..ddc7f2a52b3e
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/i915_gem_ioctls.h
@@ -0,0 +1,52 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2019 Intel Corporation
+ */
+
+#ifndef I915_GEM_IOCTLS_H
+#define I915_GEM_IOCTLS_H
+
+struct drm_device;
+struct drm_file;
+
+int i915_gem_busy_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_create_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_execbuffer_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_execbuffer2_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_get_aperture_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_get_caching_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_get_tiling_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_madvise_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_mmap_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_mmap_gtt_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_pread_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_set_caching_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_set_tiling_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_sw_finish_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_throttle_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_userptr_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+int i915_gem_wait_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file);
+
+#endif
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 83d2eb9e74cb..30afd8aef737 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -47,6 +47,7 @@
#include <drm/drm_probe_helper.h>
#include <drm/i915_drm.h>
+#include "gem/i915_gem_ioctls.h"
#include "gt/intel_gt_pm.h"
#include "gt/intel_reset.h"
#include "gt/intel_workarounds.h"
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 0b9ec8b3e5fa..95e9b2c69f17 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2736,46 +2736,8 @@ static inline bool intel_vgpu_active(struct drm_i915_private *dev_priv)
}
/* i915_gem.c */
-int i915_gem_create_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
-int i915_gem_pread_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
-int i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
-int i915_gem_mmap_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
-int i915_gem_mmap_gtt_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
-int i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
-int i915_gem_sw_finish_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
-int i915_gem_execbuffer_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
-int i915_gem_execbuffer2_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
-int i915_gem_busy_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
-int i915_gem_get_caching_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file);
-int i915_gem_set_caching_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file);
-int i915_gem_throttle_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
-int i915_gem_madvise_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
-int i915_gem_set_tiling_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
-int i915_gem_get_tiling_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
int i915_gem_init_userptr(struct drm_i915_private *dev_priv);
void i915_gem_cleanup_userptr(struct drm_i915_private *dev_priv);
-int i915_gem_userptr_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file);
-int i915_gem_get_aperture_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
-int i915_gem_wait_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv);
void i915_gem_sanitize(struct drm_i915_private *i915);
int i915_gem_init_early(struct drm_i915_private *dev_priv);
void i915_gem_cleanup_early(struct drm_i915_private *dev_priv);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 2fcec1bfb038..78ca69dcfaa6 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -39,6 +39,7 @@
#include <linux/dma-buf.h>
#include <linux/mman.h>
+#include "gem/i915_gem_ioctls.h"
#include "gt/intel_engine_pm.h"
#include "gt/intel_gt_pm.h"
#include "gt/intel_mocs.h"
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 7ce25b54c57b..27391300507e 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -34,6 +34,7 @@
#include <drm/drm_syncobj.h>
#include <drm/i915_drm.h>
+#include "gem/i915_gem_ioctls.h"
#include "gt/intel_gt_pm.h"
#include "i915_drv.h"
diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c
index a9b5329dae3b..7d5cc9ccf6dd 100644
--- a/drivers/gpu/drm/i915/i915_gem_tiling.c
+++ b/drivers/gpu/drm/i915/i915_gem_tiling.c
@@ -28,6 +28,9 @@
#include <linux/string.h>
#include <linux/bitops.h>
#include <drm/i915_drm.h>
+
+#include "gem/i915_gem_ioctls.h"
+
#include "i915_drv.h"
/**
diff --git a/drivers/gpu/drm/i915/i915_gem_userptr.c b/drivers/gpu/drm/i915/i915_gem_userptr.c
index 215bf3fef10c..b9c753f413da 100644
--- a/drivers/gpu/drm/i915/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/i915_gem_userptr.c
@@ -22,16 +22,20 @@
*
*/
-#include <drm/i915_drm.h>
-#include "i915_drv.h"
-#include "i915_trace.h"
-#include "intel_drv.h"
#include <linux/mmu_context.h>
#include <linux/mmu_notifier.h>
#include <linux/mempolicy.h>
#include <linux/swap.h>
#include <linux/sched/mm.h>
+#include <drm/i915_drm.h>
+
+#include "gem/i915_gem_ioctls.h"
+
+#include "i915_drv.h"
+#include "i915_trace.h"
+#include "intel_drv.h"
+
struct i915_mm_struct {
struct mm_struct *mm;
struct drm_i915_private *i915;
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 21/40] drm/i915: Move object->pages API to i915_gem_object.[ch]
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (18 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 20/40] drm/i915: Pull GEM ioctls interface to its own file Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 22/40] drm/i915: Move shmem object setup to its own file Chris Wilson
` (25 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx; +Cc: Matthew Auld
Currently the code for manipulating the pages on an object is still
residing in i915_gem.c, move it to i915_gem_object.c
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
---
drivers/gpu/drm/i915/Makefile | 4 +-
.../gpu/drm/i915/{ => gem}/i915_gem_object.c | 5 +-
.../gpu/drm/i915/{ => gem}/i915_gem_object.h | 127 +++++++++++++++-
drivers/gpu/drm/i915/i915_drv.h | 137 +-----------------
drivers/gpu/drm/i915/i915_globals.c | 2 +-
drivers/gpu/drm/i915/i915_vma.h | 2 +-
drivers/gpu/drm/i915/intel_frontbuffer.h | 2 +-
7 files changed, 141 insertions(+), 138 deletions(-)
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_object.c (97%)
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_object.h (59%)
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index 96344c9a0726..1c8bd0a5212d 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -86,7 +86,10 @@ i915-y += $(gt-y)
# GEM (Graphics Execution Management) code
obj-y += gem/
+gem-y += \
+ gem/i915_gem_object.o
i915-y += \
+ $(gem-y) \
i915_active.o \
i915_cmd_parser.o \
i915_gem_batch_pool.o \
@@ -99,7 +102,6 @@ i915-y += \
i915_gem_gtt.o \
i915_gem_internal.o \
i915_gem.o \
- i915_gem_object.o \
i915_gem_pm.o \
i915_gem_render_state.o \
i915_gem_shrinker.o \
diff --git a/drivers/gpu/drm/i915/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
similarity index 97%
rename from drivers/gpu/drm/i915/i915_gem_object.c
rename to drivers/gpu/drm/i915/gem/i915_gem_object.c
index ac6a5ab84586..8179252bb39b 100644
--- a/drivers/gpu/drm/i915/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
@@ -22,9 +22,10 @@
*
*/
-#include "i915_drv.h"
#include "i915_gem_object.h"
-#include "i915_globals.h"
+
+#include "../i915_drv.h"
+#include "../i915_globals.h"
static struct i915_global_object {
struct i915_global base;
diff --git a/drivers/gpu/drm/i915/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
similarity index 59%
rename from drivers/gpu/drm/i915/i915_gem_object.h
rename to drivers/gpu/drm/i915/gem/i915_gem_object.h
index 3666b0c5f6ca..061b20c3da5b 100644
--- a/drivers/gpu/drm/i915/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -13,7 +13,7 @@
#include <drm/i915_drm.h>
-#include "gem/i915_gem_object_types.h"
+#include "i915_gem_object_types.h"
struct drm_i915_gem_object *i915_gem_object_alloc(void);
void i915_gem_object_free(struct drm_i915_gem_object *obj);
@@ -192,6 +192,131 @@ i915_gem_object_get_tile_row_size(const struct drm_i915_gem_object *obj)
int i915_gem_object_set_tiling(struct drm_i915_gem_object *obj,
unsigned int tiling, unsigned int stride);
+struct scatterlist *
+i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
+ unsigned int n, unsigned int *offset);
+
+struct page *
+i915_gem_object_get_page(struct drm_i915_gem_object *obj,
+ unsigned int n);
+
+struct page *
+i915_gem_object_get_dirty_page(struct drm_i915_gem_object *obj,
+ unsigned int n);
+
+dma_addr_t
+i915_gem_object_get_dma_address(struct drm_i915_gem_object *obj,
+ unsigned long n);
+
+void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj,
+ struct sg_table *pages,
+ unsigned int sg_page_sizes);
+int __i915_gem_object_get_pages(struct drm_i915_gem_object *obj);
+
+static inline int __must_check
+i915_gem_object_pin_pages(struct drm_i915_gem_object *obj)
+{
+ might_lock(&obj->mm.lock);
+
+ if (atomic_inc_not_zero(&obj->mm.pages_pin_count))
+ return 0;
+
+ return __i915_gem_object_get_pages(obj);
+}
+
+static inline bool
+i915_gem_object_has_pages(struct drm_i915_gem_object *obj)
+{
+ return !IS_ERR_OR_NULL(READ_ONCE(obj->mm.pages));
+}
+
+static inline void
+__i915_gem_object_pin_pages(struct drm_i915_gem_object *obj)
+{
+ GEM_BUG_ON(!i915_gem_object_has_pages(obj));
+
+ atomic_inc(&obj->mm.pages_pin_count);
+}
+
+static inline bool
+i915_gem_object_has_pinned_pages(struct drm_i915_gem_object *obj)
+{
+ return atomic_read(&obj->mm.pages_pin_count);
+}
+
+static inline void
+__i915_gem_object_unpin_pages(struct drm_i915_gem_object *obj)
+{
+ GEM_BUG_ON(!i915_gem_object_has_pages(obj));
+ GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj));
+
+ atomic_dec(&obj->mm.pages_pin_count);
+}
+
+static inline void
+i915_gem_object_unpin_pages(struct drm_i915_gem_object *obj)
+{
+ __i915_gem_object_unpin_pages(obj);
+}
+
+enum i915_mm_subclass { /* lockdep subclass for obj->mm.lock/struct_mutex */
+ I915_MM_NORMAL = 0,
+ I915_MM_SHRINKER /* called "recursively" from direct-reclaim-esque */
+};
+
+int __i915_gem_object_put_pages(struct drm_i915_gem_object *obj,
+ enum i915_mm_subclass subclass);
+void __i915_gem_object_truncate(struct drm_i915_gem_object *obj);
+
+enum i915_map_type {
+ I915_MAP_WB = 0,
+ I915_MAP_WC,
+#define I915_MAP_OVERRIDE BIT(31)
+ I915_MAP_FORCE_WB = I915_MAP_WB | I915_MAP_OVERRIDE,
+ I915_MAP_FORCE_WC = I915_MAP_WC | I915_MAP_OVERRIDE,
+};
+
+/**
+ * i915_gem_object_pin_map - return a contiguous mapping of the entire object
+ * @obj: the object to map into kernel address space
+ * @type: the type of mapping, used to select pgprot_t
+ *
+ * Calls i915_gem_object_pin_pages() to prevent reaping of the object's
+ * pages and then returns a contiguous mapping of the backing storage into
+ * the kernel address space. Based on the @type of mapping, the PTE will be
+ * set to either WriteBack or WriteCombine (via pgprot_t).
+ *
+ * The caller is responsible for calling i915_gem_object_unpin_map() when the
+ * mapping is no longer required.
+ *
+ * Returns the pointer through which to access the mapped object, or an
+ * ERR_PTR() on error.
+ */
+void *__must_check i915_gem_object_pin_map(struct drm_i915_gem_object *obj,
+ enum i915_map_type type);
+
+void __i915_gem_object_flush_map(struct drm_i915_gem_object *obj,
+ unsigned long offset,
+ unsigned long size);
+static inline void i915_gem_object_flush_map(struct drm_i915_gem_object *obj)
+{
+ __i915_gem_object_flush_map(obj, 0, obj->base.size);
+}
+
+/**
+ * i915_gem_object_unpin_map - releases an earlier mapping
+ * @obj: the object to unmap
+ *
+ * After pinning the object and mapping its pages, once you are finished
+ * with your access, call i915_gem_object_unpin_map() to release the pin
+ * upon the mapping. Once the pin count reaches zero, that mapping may be
+ * removed.
+ */
+static inline void i915_gem_object_unpin_map(struct drm_i915_gem_object *obj)
+{
+ i915_gem_object_unpin_pages(obj);
+}
+
static inline struct intel_engine_cs *
i915_gem_object_last_write_engine(struct drm_i915_gem_object *obj)
{
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 95e9b2c69f17..fbb86d74a019 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2809,137 +2809,6 @@ static inline int __sg_page_count(const struct scatterlist *sg)
return sg->length >> PAGE_SHIFT;
}
-struct scatterlist *
-i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
- unsigned int n, unsigned int *offset);
-
-struct page *
-i915_gem_object_get_page(struct drm_i915_gem_object *obj,
- unsigned int n);
-
-struct page *
-i915_gem_object_get_dirty_page(struct drm_i915_gem_object *obj,
- unsigned int n);
-
-dma_addr_t
-i915_gem_object_get_dma_address(struct drm_i915_gem_object *obj,
- unsigned long n);
-
-void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj,
- struct sg_table *pages,
- unsigned int sg_page_sizes);
-int __i915_gem_object_get_pages(struct drm_i915_gem_object *obj);
-
-static inline int __must_check
-i915_gem_object_pin_pages(struct drm_i915_gem_object *obj)
-{
- might_lock(&obj->mm.lock);
-
- if (atomic_inc_not_zero(&obj->mm.pages_pin_count))
- return 0;
-
- return __i915_gem_object_get_pages(obj);
-}
-
-static inline bool
-i915_gem_object_has_pages(struct drm_i915_gem_object *obj)
-{
- return !IS_ERR_OR_NULL(READ_ONCE(obj->mm.pages));
-}
-
-static inline void
-__i915_gem_object_pin_pages(struct drm_i915_gem_object *obj)
-{
- GEM_BUG_ON(!i915_gem_object_has_pages(obj));
-
- atomic_inc(&obj->mm.pages_pin_count);
-}
-
-static inline bool
-i915_gem_object_has_pinned_pages(struct drm_i915_gem_object *obj)
-{
- return atomic_read(&obj->mm.pages_pin_count);
-}
-
-static inline void
-__i915_gem_object_unpin_pages(struct drm_i915_gem_object *obj)
-{
- GEM_BUG_ON(!i915_gem_object_has_pages(obj));
- GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj));
-
- atomic_dec(&obj->mm.pages_pin_count);
-}
-
-static inline void
-i915_gem_object_unpin_pages(struct drm_i915_gem_object *obj)
-{
- __i915_gem_object_unpin_pages(obj);
-}
-
-enum i915_mm_subclass { /* lockdep subclass for obj->mm.lock/struct_mutex */
- I915_MM_NORMAL = 0,
- I915_MM_SHRINKER /* called "recursively" from direct-reclaim-esque */
-};
-
-int __i915_gem_object_put_pages(struct drm_i915_gem_object *obj,
- enum i915_mm_subclass subclass);
-void __i915_gem_object_truncate(struct drm_i915_gem_object *obj);
-
-enum i915_map_type {
- I915_MAP_WB = 0,
- I915_MAP_WC,
-#define I915_MAP_OVERRIDE BIT(31)
- I915_MAP_FORCE_WB = I915_MAP_WB | I915_MAP_OVERRIDE,
- I915_MAP_FORCE_WC = I915_MAP_WC | I915_MAP_OVERRIDE,
-};
-
-static inline enum i915_map_type
-i915_coherent_map_type(struct drm_i915_private *i915)
-{
- return HAS_LLC(i915) ? I915_MAP_WB : I915_MAP_WC;
-}
-
-/**
- * i915_gem_object_pin_map - return a contiguous mapping of the entire object
- * @obj: the object to map into kernel address space
- * @type: the type of mapping, used to select pgprot_t
- *
- * Calls i915_gem_object_pin_pages() to prevent reaping of the object's
- * pages and then returns a contiguous mapping of the backing storage into
- * the kernel address space. Based on the @type of mapping, the PTE will be
- * set to either WriteBack or WriteCombine (via pgprot_t).
- *
- * The caller is responsible for calling i915_gem_object_unpin_map() when the
- * mapping is no longer required.
- *
- * Returns the pointer through which to access the mapped object, or an
- * ERR_PTR() on error.
- */
-void *__must_check i915_gem_object_pin_map(struct drm_i915_gem_object *obj,
- enum i915_map_type type);
-
-void __i915_gem_object_flush_map(struct drm_i915_gem_object *obj,
- unsigned long offset,
- unsigned long size);
-static inline void i915_gem_object_flush_map(struct drm_i915_gem_object *obj)
-{
- __i915_gem_object_flush_map(obj, 0, obj->base.size);
-}
-
-/**
- * i915_gem_object_unpin_map - releases an earlier mapping
- * @obj: the object to unmap
- *
- * After pinning the object and mapping its pages, once you are finished
- * with your access, call i915_gem_object_unpin_map() to release the pin
- * upon the mapping. Once the pin count reaches zero, that mapping may be
- * removed.
- */
-static inline void i915_gem_object_unpin_map(struct drm_i915_gem_object *obj)
-{
- i915_gem_object_unpin_pages(obj);
-}
-
int i915_gem_obj_prepare_shmem_read(struct drm_i915_gem_object *obj,
unsigned int *needs_clflush);
int i915_gem_obj_prepare_shmem_write(struct drm_i915_gem_object *obj,
@@ -3338,4 +3207,10 @@ static inline u32 i915_scratch_offset(const struct drm_i915_private *i915)
return i915_ggtt_offset(i915->gt.scratch);
}
+static inline enum i915_map_type
+i915_coherent_map_type(struct drm_i915_private *i915)
+{
+ return HAS_LLC(i915) ? I915_MAP_WB : I915_MAP_WC;
+}
+
#endif
diff --git a/drivers/gpu/drm/i915/i915_globals.c b/drivers/gpu/drm/i915/i915_globals.c
index 81e5c2ce336b..db52a58eadcc 100644
--- a/drivers/gpu/drm/i915/i915_globals.c
+++ b/drivers/gpu/drm/i915/i915_globals.c
@@ -9,7 +9,7 @@
#include "i915_active.h"
#include "i915_gem_context.h"
-#include "i915_gem_object.h"
+#include "gem/i915_gem_object.h"
#include "i915_globals.h"
#include "i915_request.h"
#include "i915_scheduler.h"
diff --git a/drivers/gpu/drm/i915/i915_vma.h b/drivers/gpu/drm/i915/i915_vma.h
index 6eab70953a57..7950fa96ee86 100644
--- a/drivers/gpu/drm/i915/i915_vma.h
+++ b/drivers/gpu/drm/i915/i915_vma.h
@@ -32,7 +32,7 @@
#include "i915_gem_gtt.h"
#include "i915_gem_fence_reg.h"
-#include "i915_gem_object.h"
+#include "gem/i915_gem_object.h"
#include "i915_active.h"
#include "i915_request.h"
diff --git a/drivers/gpu/drm/i915/intel_frontbuffer.h b/drivers/gpu/drm/i915/intel_frontbuffer.h
index d5894666f658..5727320c8084 100644
--- a/drivers/gpu/drm/i915/intel_frontbuffer.h
+++ b/drivers/gpu/drm/i915/intel_frontbuffer.h
@@ -24,7 +24,7 @@
#ifndef __INTEL_FRONTBUFFER_H__
#define __INTEL_FRONTBUFFER_H__
-#include "i915_gem_object.h"
+#include "gem/i915_gem_object.h"
struct drm_i915_private;
struct drm_i915_gem_object;
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 22/40] drm/i915: Move shmem object setup to its own file
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (19 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 21/40] drm/i915: Move object->pages API to i915_gem_object.[ch] Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 23/40] drm/i915: Move phys objects " Chris Wilson
` (24 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx; +Cc: Matthew Auld
Split the plain old shmem object into its own file to start decluttering
i915_gem.c
v2: Lose the confusing, hysterical raisins, suffix of _gtt.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
---
drivers/gpu/drm/i915/Makefile | 3 +-
drivers/gpu/drm/i915/gem/i915_gem_object.c | 298 +++++++
drivers/gpu/drm/i915/gem/i915_gem_object.h | 41 +-
drivers/gpu/drm/i915/gem/i915_gem_shmem.c | 512 +++++++++++
drivers/gpu/drm/i915/gt/intel_lrc.c | 4 +-
drivers/gpu/drm/i915/gt/intel_ringbuffer.c | 2 +-
drivers/gpu/drm/i915/gvt/cmd_parser.c | 21 +-
drivers/gpu/drm/i915/i915_drv.h | 10 -
drivers/gpu/drm/i915/i915_gem.c | 826 +-----------------
drivers/gpu/drm/i915/i915_perf.c | 2 +-
drivers/gpu/drm/i915/intel_fbdev.c | 2 +-
drivers/gpu/drm/i915/intel_guc.c | 2 +-
drivers/gpu/drm/i915/intel_uc_fw.c | 3 +-
drivers/gpu/drm/i915/selftests/huge_pages.c | 6 +-
.../gpu/drm/i915/selftests/i915_gem_dmabuf.c | 8 +-
.../gpu/drm/i915/selftests/i915_gem_object.c | 4 +-
16 files changed, 883 insertions(+), 861 deletions(-)
create mode 100644 drivers/gpu/drm/i915/gem/i915_gem_shmem.c
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index 1c8bd0a5212d..625f9749355b 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -87,7 +87,8 @@ i915-y += $(gt-y)
# GEM (Graphics Execution Management) code
obj-y += gem/
gem-y += \
- gem/i915_gem_object.o
+ gem/i915_gem_object.o \
+ gem/i915_gem_shmem.o
i915-y += \
$(gem-y) \
i915_active.o \
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
index 8179252bb39b..86e7e88817af 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
@@ -26,6 +26,7 @@
#include "../i915_drv.h"
#include "../i915_globals.h"
+#include "../intel_frontbuffer.h"
static struct i915_global_object {
struct i915_global base;
@@ -42,6 +43,64 @@ void i915_gem_object_free(struct drm_i915_gem_object *obj)
return kmem_cache_free(global.slab_objects, obj);
}
+/* some bookkeeping */
+static void i915_gem_info_add_obj(struct drm_i915_private *i915,
+ u64 size)
+{
+ spin_lock(&i915->mm.object_stat_lock);
+ i915->mm.object_count++;
+ i915->mm.object_memory += size;
+ spin_unlock(&i915->mm.object_stat_lock);
+}
+
+static void i915_gem_info_remove_obj(struct drm_i915_private *i915,
+ u64 size)
+{
+ spin_lock(&i915->mm.object_stat_lock);
+ i915->mm.object_count--;
+ i915->mm.object_memory -= size;
+ spin_unlock(&i915->mm.object_stat_lock);
+}
+
+static void
+frontbuffer_retire(struct i915_active_request *active,
+ struct i915_request *request)
+{
+ struct drm_i915_gem_object *obj =
+ container_of(active, typeof(*obj), frontbuffer_write);
+
+ intel_fb_obj_flush(obj, ORIGIN_CS);
+}
+
+void i915_gem_object_init(struct drm_i915_gem_object *obj,
+ const struct drm_i915_gem_object_ops *ops)
+{
+ mutex_init(&obj->mm.lock);
+
+ spin_lock_init(&obj->vma.lock);
+ INIT_LIST_HEAD(&obj->vma.list);
+
+ INIT_LIST_HEAD(&obj->lut_list);
+ INIT_LIST_HEAD(&obj->batch_pool_link);
+
+ init_rcu_head(&obj->rcu);
+
+ obj->ops = ops;
+
+ reservation_object_init(&obj->__builtin_resv);
+ obj->resv = &obj->__builtin_resv;
+
+ obj->frontbuffer_ggtt_origin = ORIGIN_GTT;
+ i915_active_request_init(&obj->frontbuffer_write,
+ NULL, frontbuffer_retire);
+
+ obj->mm.madv = I915_MADV_WILLNEED;
+ INIT_RADIX_TREE(&obj->mm.get_page.radix, GFP_KERNEL | __GFP_NOWARN);
+ mutex_init(&obj->mm.get_page.lock);
+
+ i915_gem_info_add_obj(to_i915(obj->base.dev), obj->base.size);
+}
+
/**
* Mark up the object's coherency levels for a given cache_level
* @obj: #drm_i915_gem_object
@@ -64,6 +123,245 @@ void i915_gem_object_set_cache_coherency(struct drm_i915_gem_object *obj,
!(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE);
}
+void i915_gem_close_object(struct drm_gem_object *gem, struct drm_file *file)
+{
+ struct drm_i915_private *i915 = to_i915(gem->dev);
+ struct drm_i915_gem_object *obj = to_intel_bo(gem);
+ struct drm_i915_file_private *fpriv = file->driver_priv;
+ struct i915_lut_handle *lut, *ln;
+
+ mutex_lock(&i915->drm.struct_mutex);
+
+ list_for_each_entry_safe(lut, ln, &obj->lut_list, obj_link) {
+ struct i915_gem_context *ctx = lut->ctx;
+ struct i915_vma *vma;
+
+ GEM_BUG_ON(ctx->file_priv == ERR_PTR(-EBADF));
+ if (ctx->file_priv != fpriv)
+ continue;
+
+ vma = radix_tree_delete(&ctx->handles_vma, lut->handle);
+ GEM_BUG_ON(vma->obj != obj);
+
+ /* We allow the process to have multiple handles to the same
+ * vma, in the same fd namespace, by virtue of flink/open.
+ */
+ GEM_BUG_ON(!vma->open_count);
+ if (!--vma->open_count && !i915_vma_is_ggtt(vma))
+ i915_vma_close(vma);
+
+ list_del(&lut->obj_link);
+ list_del(&lut->ctx_link);
+
+ i915_lut_handle_free(lut);
+ __i915_gem_object_release_unless_active(obj);
+ }
+
+ mutex_unlock(&i915->drm.struct_mutex);
+}
+
+static bool discard_backing_storage(struct drm_i915_gem_object *obj)
+{
+ /* If we are the last user of the backing storage (be it shmemfs
+ * pages or stolen etc), we know that the pages are going to be
+ * immediately released. In this case, we can then skip copying
+ * back the contents from the GPU.
+ */
+
+ if (obj->mm.madv != I915_MADV_WILLNEED)
+ return false;
+
+ if (!obj->base.filp)
+ return true;
+
+ /* At first glance, this looks racy, but then again so would be
+ * userspace racing mmap against close. However, the first external
+ * reference to the filp can only be obtained through the
+ * i915_gem_mmap_ioctl() which safeguards us against the user
+ * acquiring such a reference whilst we are in the middle of
+ * freeing the object.
+ */
+ return atomic_long_read(&obj->base.filp->f_count) == 1;
+}
+
+static void __i915_gem_free_objects(struct drm_i915_private *i915,
+ struct llist_node *freed)
+{
+ struct drm_i915_gem_object *obj, *on;
+ intel_wakeref_t wakeref;
+
+ wakeref = intel_runtime_pm_get(i915);
+ llist_for_each_entry_safe(obj, on, freed, freed) {
+ struct i915_vma *vma, *vn;
+
+ trace_i915_gem_object_destroy(obj);
+
+ mutex_lock(&i915->drm.struct_mutex);
+
+ GEM_BUG_ON(i915_gem_object_is_active(obj));
+ list_for_each_entry_safe(vma, vn, &obj->vma.list, obj_link) {
+ GEM_BUG_ON(i915_vma_is_active(vma));
+ vma->flags &= ~I915_VMA_PIN_MASK;
+ i915_vma_destroy(vma);
+ }
+ GEM_BUG_ON(!list_empty(&obj->vma.list));
+ GEM_BUG_ON(!RB_EMPTY_ROOT(&obj->vma.tree));
+
+ /* This serializes freeing with the shrinker. Since the free
+ * is delayed, first by RCU then by the workqueue, we want the
+ * shrinker to be able to free pages of unreferenced objects,
+ * or else we may oom whilst there are plenty of deferred
+ * freed objects.
+ */
+ if (i915_gem_object_has_pages(obj)) {
+ spin_lock(&i915->mm.obj_lock);
+ list_del_init(&obj->mm.link);
+ spin_unlock(&i915->mm.obj_lock);
+ }
+
+ mutex_unlock(&i915->drm.struct_mutex);
+
+ GEM_BUG_ON(obj->bind_count);
+ GEM_BUG_ON(obj->userfault_count);
+ GEM_BUG_ON(atomic_read(&obj->frontbuffer_bits));
+ GEM_BUG_ON(!list_empty(&obj->lut_list));
+
+ if (obj->ops->release)
+ obj->ops->release(obj);
+
+ if (WARN_ON(i915_gem_object_has_pinned_pages(obj)))
+ atomic_set(&obj->mm.pages_pin_count, 0);
+ __i915_gem_object_put_pages(obj, I915_MM_NORMAL);
+ GEM_BUG_ON(i915_gem_object_has_pages(obj));
+
+ if (obj->base.import_attach)
+ drm_prime_gem_destroy(&obj->base, NULL);
+
+ reservation_object_fini(&obj->__builtin_resv);
+ drm_gem_object_release(&obj->base);
+ i915_gem_info_remove_obj(i915, obj->base.size);
+
+ bitmap_free(obj->bit_17);
+ i915_gem_object_free(obj);
+
+ GEM_BUG_ON(!atomic_read(&i915->mm.free_count));
+ atomic_dec(&i915->mm.free_count);
+
+ if (on)
+ cond_resched();
+ }
+ intel_runtime_pm_put(i915, wakeref);
+}
+
+void i915_gem_flush_free_objects(struct drm_i915_private *i915)
+{
+ struct llist_node *freed;
+
+ /* Free the oldest, most stale object to keep the free_list short */
+ freed = NULL;
+ if (!llist_empty(&i915->mm.free_list)) { /* quick test for hotpath */
+ /* Only one consumer of llist_del_first() allowed */
+ spin_lock(&i915->mm.free_lock);
+ freed = llist_del_first(&i915->mm.free_list);
+ spin_unlock(&i915->mm.free_lock);
+ }
+ if (unlikely(freed)) {
+ freed->next = NULL;
+ __i915_gem_free_objects(i915, freed);
+ }
+}
+
+static void __i915_gem_free_work(struct work_struct *work)
+{
+ struct drm_i915_private *i915 =
+ container_of(work, struct drm_i915_private, mm.free_work);
+ struct llist_node *freed;
+
+ /*
+ * All file-owned VMA should have been released by this point through
+ * i915_gem_close_object(), or earlier by i915_gem_context_close().
+ * However, the object may also be bound into the global GTT (e.g.
+ * older GPUs without per-process support, or for direct access through
+ * the GTT either for the user or for scanout). Those VMA still need to
+ * unbound now.
+ */
+
+ spin_lock(&i915->mm.free_lock);
+ while ((freed = llist_del_all(&i915->mm.free_list))) {
+ spin_unlock(&i915->mm.free_lock);
+
+ __i915_gem_free_objects(i915, freed);
+ if (need_resched())
+ return;
+
+ spin_lock(&i915->mm.free_lock);
+ }
+ spin_unlock(&i915->mm.free_lock);
+}
+
+static void __i915_gem_free_object_rcu(struct rcu_head *head)
+{
+ struct drm_i915_gem_object *obj =
+ container_of(head, typeof(*obj), rcu);
+ struct drm_i915_private *i915 = to_i915(obj->base.dev);
+
+ /*
+ * We reuse obj->rcu for the freed list, so we had better not treat
+ * it like a rcu_head from this point forwards. And we expect all
+ * objects to be freed via this path.
+ */
+ destroy_rcu_head(&obj->rcu);
+
+ /*
+ * Since we require blocking on struct_mutex to unbind the freed
+ * object from the GPU before releasing resources back to the
+ * system, we can not do that directly from the RCU callback (which may
+ * be a softirq context), but must instead then defer that work onto a
+ * kthread. We use the RCU callback rather than move the freed object
+ * directly onto the work queue so that we can mix between using the
+ * worker and performing frees directly from subsequent allocations for
+ * crude but effective memory throttling.
+ */
+ if (llist_add(&obj->freed, &i915->mm.free_list))
+ queue_work(i915->wq, &i915->mm.free_work);
+}
+
+void i915_gem_free_object(struct drm_gem_object *gem_obj)
+{
+ struct drm_i915_gem_object *obj = to_intel_bo(gem_obj);
+
+ if (obj->mm.quirked)
+ __i915_gem_object_unpin_pages(obj);
+
+ if (discard_backing_storage(obj))
+ obj->mm.madv = I915_MADV_DONTNEED;
+
+ /*
+ * Before we free the object, make sure any pure RCU-only
+ * read-side critical sections are complete, e.g.
+ * i915_gem_busy_ioctl(). For the corresponding synchronized
+ * lookup see i915_gem_object_lookup_rcu().
+ */
+ atomic_inc(&to_i915(obj->base.dev)->mm.free_count);
+ call_rcu(&obj->rcu, __i915_gem_free_object_rcu);
+}
+
+void __i915_gem_object_release_unless_active(struct drm_i915_gem_object *obj)
+{
+ lockdep_assert_held(&obj->base.dev->struct_mutex);
+
+ if (!i915_gem_object_has_active_reference(obj) &&
+ i915_gem_object_is_active(obj))
+ i915_gem_object_set_active_reference(obj);
+ else
+ i915_gem_object_put(obj);
+}
+
+void i915_gem_init__objects(struct drm_i915_private *i915)
+{
+ INIT_WORK(&i915->mm.free_work, __i915_gem_free_work);
+}
+
static void i915_global_objects_shrink(void)
{
kmem_cache_shrink(global.slab_objects);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index 061b20c3da5b..ad82303f741a 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -15,9 +15,29 @@
#include "i915_gem_object_types.h"
+void i915_gem_init__objects(struct drm_i915_private *i915);
+
struct drm_i915_gem_object *i915_gem_object_alloc(void);
void i915_gem_object_free(struct drm_i915_gem_object *obj);
+void i915_gem_object_init(struct drm_i915_gem_object *obj,
+ const struct drm_i915_gem_object_ops *ops);
+struct drm_i915_gem_object *
+i915_gem_object_create_shmem(struct drm_i915_private *i915, u64 size);
+struct drm_i915_gem_object *
+i915_gem_object_create_shmem_from_data(struct drm_i915_private *i915,
+ const void *data, size_t size);
+
+extern const struct drm_i915_gem_object_ops i915_gem_shmem_ops;
+void __i915_gem_object_release_shmem(struct drm_i915_gem_object *obj,
+ struct sg_table *pages,
+ bool needs_clflush);
+
+void i915_gem_close_object(struct drm_gem_object *gem, struct drm_file *file);
+void i915_gem_free_object(struct drm_gem_object *obj);
+
+void i915_gem_flush_free_objects(struct drm_i915_private *i915);
+
/**
* i915_gem_object_lookup_rcu - look up a temporary GEM object from its handle
* @filp: DRM file private date
@@ -338,8 +358,23 @@ void i915_gem_object_set_cache_coherency(struct drm_i915_gem_object *obj,
unsigned int cache_level);
void i915_gem_object_flush_if_display(struct drm_i915_gem_object *obj);
-void __i915_gem_object_release_shmem(struct drm_i915_gem_object *obj,
- struct sg_table *pages,
- bool needs_clflush);
+static inline bool cpu_write_needs_clflush(struct drm_i915_gem_object *obj)
+{
+ if (obj->cache_dirty)
+ return false;
+
+ if (!(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE))
+ return true;
+
+ return obj->pin_global; /* currently in use by HW, keep flushed */
+}
+
+static inline void __start_cpu_write(struct drm_i915_gem_object *obj)
+{
+ obj->read_domains = I915_GEM_DOMAIN_CPU;
+ obj->write_domain = I915_GEM_DOMAIN_CPU;
+ if (cpu_write_needs_clflush(obj))
+ obj->cache_dirty = true;
+}
#endif
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
new file mode 100644
index 000000000000..9d885b5e6643
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
@@ -0,0 +1,512 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2014-2016 Intel Corporation
+ */
+
+#include <linux/pagevec.h>
+#include <linux/swap.h>
+
+#include "i915_gem_object.h"
+
+#include "../i915_drv.h"
+
+/*
+ * Move pages to appropriate lru and release the pagevec, decrementing the
+ * ref count of those pages.
+ */
+static void check_release_pagevec(struct pagevec *pvec)
+{
+ check_move_unevictable_pages(pvec);
+ __pagevec_release(pvec);
+ cond_resched();
+}
+
+static int shmem_get_pages(struct drm_i915_gem_object *obj)
+{
+ struct drm_i915_private *i915 = to_i915(obj->base.dev);
+ const unsigned long page_count = obj->base.size / PAGE_SIZE;
+ unsigned long i;
+ struct address_space *mapping;
+ struct sg_table *st;
+ struct scatterlist *sg;
+ struct sgt_iter sgt_iter;
+ struct page *page;
+ unsigned long last_pfn = 0; /* suppress gcc warning */
+ unsigned int max_segment = i915_sg_segment_size();
+ unsigned int sg_page_sizes;
+ struct pagevec pvec;
+ gfp_t noreclaim;
+ int ret;
+
+ /*
+ * Assert that the object is not currently in any GPU domain. As it
+ * wasn't in the GTT, there shouldn't be any way it could have been in
+ * a GPU cache
+ */
+ GEM_BUG_ON(obj->read_domains & I915_GEM_GPU_DOMAINS);
+ GEM_BUG_ON(obj->write_domain & I915_GEM_GPU_DOMAINS);
+
+ /*
+ * If there's no chance of allocating enough pages for the whole
+ * object, bail early.
+ */
+ if (page_count > totalram_pages())
+ return -ENOMEM;
+
+ st = kmalloc(sizeof(*st), GFP_KERNEL);
+ if (!st)
+ return -ENOMEM;
+
+rebuild_st:
+ if (sg_alloc_table(st, page_count, GFP_KERNEL)) {
+ kfree(st);
+ return -ENOMEM;
+ }
+
+ /*
+ * Get the list of pages out of our struct file. They'll be pinned
+ * at this point until we release them.
+ *
+ * Fail silently without starting the shrinker
+ */
+ mapping = obj->base.filp->f_mapping;
+ mapping_set_unevictable(mapping);
+ noreclaim = mapping_gfp_constraint(mapping, ~__GFP_RECLAIM);
+ noreclaim |= __GFP_NORETRY | __GFP_NOWARN;
+
+ sg = st->sgl;
+ st->nents = 0;
+ sg_page_sizes = 0;
+ for (i = 0; i < page_count; i++) {
+ const unsigned int shrink[] = {
+ (I915_SHRINK_BOUND |
+ I915_SHRINK_UNBOUND |
+ I915_SHRINK_PURGEABLE),
+ 0,
+ }, *s = shrink;
+ gfp_t gfp = noreclaim;
+
+ do {
+ cond_resched();
+ page = shmem_read_mapping_page_gfp(mapping, i, gfp);
+ if (!IS_ERR(page))
+ break;
+
+ if (!*s) {
+ ret = PTR_ERR(page);
+ goto err_sg;
+ }
+
+ i915_gem_shrink(i915, 2 * page_count, NULL, *s++);
+
+ /*
+ * We've tried hard to allocate the memory by reaping
+ * our own buffer, now let the real VM do its job and
+ * go down in flames if truly OOM.
+ *
+ * However, since graphics tend to be disposable,
+ * defer the oom here by reporting the ENOMEM back
+ * to userspace.
+ */
+ if (!*s) {
+ /* reclaim and warn, but no oom */
+ gfp = mapping_gfp_mask(mapping);
+
+ /*
+ * Our bo are always dirty and so we require
+ * kswapd to reclaim our pages (direct reclaim
+ * does not effectively begin pageout of our
+ * buffers on its own). However, direct reclaim
+ * only waits for kswapd when under allocation
+ * congestion. So as a result __GFP_RECLAIM is
+ * unreliable and fails to actually reclaim our
+ * dirty pages -- unless you try over and over
+ * again with !__GFP_NORETRY. However, we still
+ * want to fail this allocation rather than
+ * trigger the out-of-memory killer and for
+ * this we want __GFP_RETRY_MAYFAIL.
+ */
+ gfp |= __GFP_RETRY_MAYFAIL;
+ }
+ } while (1);
+
+ if (!i ||
+ sg->length >= max_segment ||
+ page_to_pfn(page) != last_pfn + 1) {
+ if (i) {
+ sg_page_sizes |= sg->length;
+ sg = sg_next(sg);
+ }
+ st->nents++;
+ sg_set_page(sg, page, PAGE_SIZE, 0);
+ } else {
+ sg->length += PAGE_SIZE;
+ }
+ last_pfn = page_to_pfn(page);
+
+ /* Check that the i965g/gm workaround works. */
+ WARN_ON((gfp & __GFP_DMA32) && (last_pfn >= 0x00100000UL));
+ }
+ if (sg) { /* loop terminated early; short sg table */
+ sg_page_sizes |= sg->length;
+ sg_mark_end(sg);
+ }
+
+ /* Trim unused sg entries to avoid wasting memory. */
+ i915_sg_trim(st);
+
+ ret = i915_gem_gtt_prepare_pages(obj, st);
+ if (ret) {
+ /*
+ * DMA remapping failed? One possible cause is that
+ * it could not reserve enough large entries, asking
+ * for PAGE_SIZE chunks instead may be helpful.
+ */
+ if (max_segment > PAGE_SIZE) {
+ for_each_sgt_page(page, sgt_iter, st)
+ put_page(page);
+ sg_free_table(st);
+
+ max_segment = PAGE_SIZE;
+ goto rebuild_st;
+ } else {
+ dev_warn(&i915->drm.pdev->dev,
+ "Failed to DMA remap %lu pages\n",
+ page_count);
+ goto err_pages;
+ }
+ }
+
+ if (i915_gem_object_needs_bit17_swizzle(obj))
+ i915_gem_object_do_bit_17_swizzle(obj, st);
+
+ __i915_gem_object_set_pages(obj, st, sg_page_sizes);
+
+ return 0;
+
+err_sg:
+ sg_mark_end(sg);
+err_pages:
+ mapping_clear_unevictable(mapping);
+ pagevec_init(&pvec);
+ for_each_sgt_page(page, sgt_iter, st) {
+ if (!pagevec_add(&pvec, page))
+ check_release_pagevec(&pvec);
+ }
+ if (pagevec_count(&pvec))
+ check_release_pagevec(&pvec);
+ sg_free_table(st);
+ kfree(st);
+
+ /*
+ * shmemfs first checks if there is enough memory to allocate the page
+ * and reports ENOSPC should there be insufficient, along with the usual
+ * ENOMEM for a genuine allocation failure.
+ *
+ * We use ENOSPC in our driver to mean that we have run out of aperture
+ * space and so want to translate the error from shmemfs back to our
+ * usual understanding of ENOMEM.
+ */
+ if (ret == -ENOSPC)
+ ret = -ENOMEM;
+
+ return ret;
+}
+
+void
+__i915_gem_object_release_shmem(struct drm_i915_gem_object *obj,
+ struct sg_table *pages,
+ bool needs_clflush)
+{
+ GEM_BUG_ON(obj->mm.madv == __I915_MADV_PURGED);
+
+ if (obj->mm.madv == I915_MADV_DONTNEED)
+ obj->mm.dirty = false;
+
+ if (needs_clflush &&
+ (obj->read_domains & I915_GEM_DOMAIN_CPU) == 0 &&
+ !(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ))
+ drm_clflush_sg(pages);
+
+ __start_cpu_write(obj);
+}
+
+static void
+shmem_put_pages(struct drm_i915_gem_object *obj, struct sg_table *pages)
+{
+ struct sgt_iter sgt_iter;
+ struct pagevec pvec;
+ struct page *page;
+
+ __i915_gem_object_release_shmem(obj, pages, true);
+
+ i915_gem_gtt_finish_pages(obj, pages);
+
+ if (i915_gem_object_needs_bit17_swizzle(obj))
+ i915_gem_object_save_bit_17_swizzle(obj, pages);
+
+ mapping_clear_unevictable(file_inode(obj->base.filp)->i_mapping);
+
+ pagevec_init(&pvec);
+ for_each_sgt_page(page, sgt_iter, pages) {
+ if (obj->mm.dirty)
+ set_page_dirty(page);
+
+ if (obj->mm.madv == I915_MADV_WILLNEED)
+ mark_page_accessed(page);
+
+ if (!pagevec_add(&pvec, page))
+ check_release_pagevec(&pvec);
+ }
+ if (pagevec_count(&pvec))
+ check_release_pagevec(&pvec);
+ obj->mm.dirty = false;
+
+ sg_free_table(pages);
+ kfree(pages);
+}
+
+static int
+shmem_pwrite(struct drm_i915_gem_object *obj,
+ const struct drm_i915_gem_pwrite *arg)
+{
+ struct address_space *mapping = obj->base.filp->f_mapping;
+ char __user *user_data = u64_to_user_ptr(arg->data_ptr);
+ u64 remain, offset;
+ unsigned int pg;
+
+ /* Caller already validated user args */
+ GEM_BUG_ON(!access_ok(user_data, arg->size));
+
+ /*
+ * Before we instantiate/pin the backing store for our use, we
+ * can prepopulate the shmemfs filp efficiently using a write into
+ * the pagecache. We avoid the penalty of instantiating all the
+ * pages, important if the user is just writing to a few and never
+ * uses the object on the GPU, and using a direct write into shmemfs
+ * allows it to avoid the cost of retrieving a page (either swapin
+ * or clearing-before-use) before it is overwritten.
+ */
+ if (i915_gem_object_has_pages(obj))
+ return -ENODEV;
+
+ if (obj->mm.madv != I915_MADV_WILLNEED)
+ return -EFAULT;
+
+ /*
+ * Before the pages are instantiated the object is treated as being
+ * in the CPU domain. The pages will be clflushed as required before
+ * use, and we can freely write into the pages directly. If userspace
+ * races pwrite with any other operation; corruption will ensue -
+ * that is userspace's prerogative!
+ */
+
+ remain = arg->size;
+ offset = arg->offset;
+ pg = offset_in_page(offset);
+
+ do {
+ unsigned int len, unwritten;
+ struct page *page;
+ void *data, *vaddr;
+ int err;
+ char c;
+
+ len = PAGE_SIZE - pg;
+ if (len > remain)
+ len = remain;
+
+ /* Prefault the user page to reduce potential recursion */
+ err = __get_user(c, user_data);
+ if (err)
+ return err;
+
+ err = __get_user(c, user_data + len - 1);
+ if (err)
+ return err;
+
+ err = pagecache_write_begin(obj->base.filp, mapping,
+ offset, len, 0,
+ &page, &data);
+ if (err < 0)
+ return err;
+
+ vaddr = kmap_atomic(page);
+ unwritten = __copy_from_user_inatomic(vaddr + pg,
+ user_data,
+ len);
+ kunmap_atomic(vaddr);
+
+ err = pagecache_write_end(obj->base.filp, mapping,
+ offset, len, len - unwritten,
+ page, data);
+ if (err < 0)
+ return err;
+
+ /* We don't handle -EFAULT, leave it to the caller to check */
+ if (unwritten)
+ return -ENODEV;
+
+ remain -= len;
+ user_data += len;
+ offset += len;
+ pg = 0;
+ } while (remain);
+
+ return 0;
+}
+
+const struct drm_i915_gem_object_ops i915_gem_shmem_ops = {
+ .flags = I915_GEM_OBJECT_HAS_STRUCT_PAGE |
+ I915_GEM_OBJECT_IS_SHRINKABLE,
+
+ .get_pages = shmem_get_pages,
+ .put_pages = shmem_put_pages,
+
+ .pwrite = shmem_pwrite,
+};
+
+static int create_shmem(struct drm_i915_private *i915,
+ struct drm_gem_object *obj,
+ size_t size)
+{
+ unsigned long flags = VM_NORESERVE;
+ struct file *filp;
+
+ drm_gem_private_object_init(&i915->drm, obj, size);
+
+ if (i915->mm.gemfs)
+ filp = shmem_file_setup_with_mnt(i915->mm.gemfs, "i915", size,
+ flags);
+ else
+ filp = shmem_file_setup("i915", size, flags);
+ if (IS_ERR(filp))
+ return PTR_ERR(filp);
+
+ obj->filp = filp;
+ return 0;
+}
+
+struct drm_i915_gem_object *
+i915_gem_object_create_shmem(struct drm_i915_private *i915, u64 size)
+{
+ struct drm_i915_gem_object *obj;
+ struct address_space *mapping;
+ unsigned int cache_level;
+ gfp_t mask;
+ int ret;
+
+ /* There is a prevalence of the assumption that we fit the object's
+ * page count inside a 32bit _signed_ variable. Let's document this and
+ * catch if we ever need to fix it. In the meantime, if you do spot
+ * such a local variable, please consider fixing!
+ */
+ if (size >> PAGE_SHIFT > INT_MAX)
+ return ERR_PTR(-E2BIG);
+
+ if (overflows_type(size, obj->base.size))
+ return ERR_PTR(-E2BIG);
+
+ obj = i915_gem_object_alloc();
+ if (!obj)
+ return ERR_PTR(-ENOMEM);
+
+ ret = create_shmem(i915, &obj->base, size);
+ if (ret)
+ goto fail;
+
+ mask = GFP_HIGHUSER | __GFP_RECLAIMABLE;
+ if (IS_I965GM(i915) || IS_I965G(i915)) {
+ /* 965gm cannot relocate objects above 4GiB. */
+ mask &= ~__GFP_HIGHMEM;
+ mask |= __GFP_DMA32;
+ }
+
+ mapping = obj->base.filp->f_mapping;
+ mapping_set_gfp_mask(mapping, mask);
+ GEM_BUG_ON(!(mapping_gfp_mask(mapping) & __GFP_RECLAIM));
+
+ i915_gem_object_init(obj, &i915_gem_shmem_ops);
+
+ obj->write_domain = I915_GEM_DOMAIN_CPU;
+ obj->read_domains = I915_GEM_DOMAIN_CPU;
+
+ if (HAS_LLC(i915))
+ /* On some devices, we can have the GPU use the LLC (the CPU
+ * cache) for about a 10% performance improvement
+ * compared to uncached. Graphics requests other than
+ * display scanout are coherent with the CPU in
+ * accessing this cache. This means in this mode we
+ * don't need to clflush on the CPU side, and on the
+ * GPU side we only need to flush internal caches to
+ * get data visible to the CPU.
+ *
+ * However, we maintain the display planes as UC, and so
+ * need to rebind when first used as such.
+ */
+ cache_level = I915_CACHE_LLC;
+ else
+ cache_level = I915_CACHE_NONE;
+
+ i915_gem_object_set_cache_coherency(obj, cache_level);
+
+ trace_i915_gem_object_create(obj);
+
+ return obj;
+
+fail:
+ i915_gem_object_free(obj);
+ return ERR_PTR(ret);
+}
+
+/* Allocate a new GEM object and fill it with the supplied data */
+struct drm_i915_gem_object *
+i915_gem_object_create_shmem_from_data(struct drm_i915_private *dev_priv,
+ const void *data, size_t size)
+{
+ struct drm_i915_gem_object *obj;
+ struct file *file;
+ size_t offset;
+ int err;
+
+ obj = i915_gem_object_create_shmem(dev_priv, round_up(size, PAGE_SIZE));
+ if (IS_ERR(obj))
+ return obj;
+
+ GEM_BUG_ON(obj->write_domain != I915_GEM_DOMAIN_CPU);
+
+ file = obj->base.filp;
+ offset = 0;
+ do {
+ unsigned int len = min_t(typeof(size), size, PAGE_SIZE);
+ struct page *page;
+ void *pgdata, *vaddr;
+
+ err = pagecache_write_begin(file, file->f_mapping,
+ offset, len, 0,
+ &page, &pgdata);
+ if (err < 0)
+ goto fail;
+
+ vaddr = kmap(page);
+ memcpy(vaddr, data, len);
+ kunmap(page);
+
+ err = pagecache_write_end(file, file->f_mapping,
+ offset, len, len,
+ page, pgdata);
+ if (err < 0)
+ goto fail;
+
+ size -= len;
+ data += len;
+ offset += len;
+ } while (size);
+
+ return obj;
+
+fail:
+ i915_gem_object_put(obj);
+ return ERR_PTR(err);
+}
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 5d80f99661bc..5b6e6b54abf9 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -1961,7 +1961,7 @@ static int lrc_setup_wa_ctx(struct intel_engine_cs *engine)
struct i915_vma *vma;
int err;
- obj = i915_gem_object_create(engine->i915, CTX_WA_BB_OBJ_SIZE);
+ obj = i915_gem_object_create_shmem(engine->i915, CTX_WA_BB_OBJ_SIZE);
if (IS_ERR(obj))
return PTR_ERR(obj);
@@ -3051,7 +3051,7 @@ static int execlists_context_deferred_alloc(struct intel_context *ce,
*/
context_size += LRC_HEADER_PAGES * PAGE_SIZE;
- ctx_obj = i915_gem_object_create(engine->i915, context_size);
+ ctx_obj = i915_gem_object_create_shmem(engine->i915, context_size);
if (IS_ERR(ctx_obj))
return PTR_ERR(ctx_obj);
diff --git a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
index f0d60affdba3..ac93080bd863 100644
--- a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
@@ -1396,7 +1396,7 @@ alloc_context_vma(struct intel_engine_cs *engine)
struct i915_vma *vma;
int err;
- obj = i915_gem_object_create(i915, engine->context_size);
+ obj = i915_gem_object_create_shmem(i915, engine->context_size);
if (IS_ERR(obj))
return ERR_CAST(obj);
diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c
index ab002cfd3cab..fe66285f0236 100644
--- a/drivers/gpu/drm/i915/gvt/cmd_parser.c
+++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c
@@ -1721,7 +1721,7 @@ static int perform_bb_shadow(struct parser_exec_state *s)
int ret = 0;
struct intel_vgpu_mm *mm = (s->buf_addr_type == GTT_BUFFER) ?
s->vgpu->gtt.ggtt_mm : s->workload->shadow_mm;
- unsigned long gma_start_offset = 0;
+ unsigned long start_offset = 0;
/* get the start gm address of the batch buffer */
gma = get_gma_bb_from_cmd(s, 1);
@@ -1738,7 +1738,7 @@ static int perform_bb_shadow(struct parser_exec_state *s)
bb->ppgtt = (s->buf_addr_type == GTT_BUFFER) ? false : true;
- /* the gma_start_offset stores the batch buffer's start gma's
+ /* the start_offset stores the batch buffer's start gma's
* offset relative to page boundary. so for non-privileged batch
* buffer, the shadowed gem object holds exactly the same page
* layout as original gem object. This is for the convience of
@@ -1750,10 +1750,11 @@ static int perform_bb_shadow(struct parser_exec_state *s)
* that of shadowed page.
*/
if (bb->ppgtt)
- gma_start_offset = gma & ~I915_GTT_PAGE_MASK;
+ start_offset = gma & ~I915_GTT_PAGE_MASK;
- bb->obj = i915_gem_object_create(s->vgpu->gvt->dev_priv,
- roundup(bb_size + gma_start_offset, PAGE_SIZE));
+ bb->obj = i915_gem_object_create_shmem(s->vgpu->gvt->dev_priv,
+ round_up(bb_size + start_offset,
+ PAGE_SIZE));
if (IS_ERR(bb->obj)) {
ret = PTR_ERR(bb->obj);
goto err_free_bb;
@@ -1776,7 +1777,7 @@ static int perform_bb_shadow(struct parser_exec_state *s)
ret = copy_gma_to_hva(s->vgpu, mm,
gma, gma + bb_size,
- bb->va + gma_start_offset);
+ bb->va + start_offset);
if (ret < 0) {
gvt_vgpu_err("fail to copy guest ring buffer\n");
ret = -EFAULT;
@@ -1802,7 +1803,7 @@ static int perform_bb_shadow(struct parser_exec_state *s)
* buffer's gma in pair. After all, we don't want to pin the shadow
* buffer here (too early).
*/
- s->ip_va = bb->va + gma_start_offset;
+ s->ip_va = bb->va + start_offset;
s->ip_gma = gma;
return 0;
err_unmap:
@@ -2825,9 +2826,9 @@ static int shadow_indirect_ctx(struct intel_shadow_wa_ctx *wa_ctx)
int ret = 0;
void *map;
- obj = i915_gem_object_create(workload->vgpu->gvt->dev_priv,
- roundup(ctx_size + CACHELINE_BYTES,
- PAGE_SIZE));
+ obj = i915_gem_object_create_shmem(workload->vgpu->gvt->dev_priv,
+ roundup(ctx_size + CACHELINE_BYTES,
+ PAGE_SIZE));
if (IS_ERR(obj))
return PTR_ERR(obj);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index fbb86d74a019..6de9b641751d 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2745,16 +2745,6 @@ void i915_gem_load_init_fences(struct drm_i915_private *dev_priv);
int i915_gem_freeze(struct drm_i915_private *dev_priv);
int i915_gem_freeze_late(struct drm_i915_private *dev_priv);
-void i915_gem_object_init(struct drm_i915_gem_object *obj,
- const struct drm_i915_gem_object_ops *ops);
-struct drm_i915_gem_object *
-i915_gem_object_create(struct drm_i915_private *dev_priv, u64 size);
-struct drm_i915_gem_object *
-i915_gem_object_create_from_data(struct drm_i915_private *dev_priv,
- const void *data, size_t size);
-void i915_gem_close_object(struct drm_gem_object *gem, struct drm_file *file);
-void i915_gem_free_object(struct drm_gem_object *obj);
-
static inline void i915_gem_drain_freed_objects(struct drm_i915_private *i915)
{
if (!atomic_read(&i915->mm.free_count))
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 78ca69dcfaa6..4c5fba0c7bf6 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -57,19 +57,6 @@
#include "intel_frontbuffer.h"
#include "intel_pm.h"
-static void i915_gem_flush_free_objects(struct drm_i915_private *i915);
-
-static bool cpu_write_needs_clflush(struct drm_i915_gem_object *obj)
-{
- if (obj->cache_dirty)
- return false;
-
- if (!(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE))
- return true;
-
- return obj->pin_global; /* currently in use by HW, keep flushed */
-}
-
static int
insert_mappable_node(struct i915_ggtt *ggtt,
struct drm_mm_node *node, u32 size)
@@ -87,25 +74,6 @@ remove_mappable_node(struct drm_mm_node *node)
drm_mm_remove_node(node);
}
-/* some bookkeeping */
-static void i915_gem_info_add_obj(struct drm_i915_private *dev_priv,
- u64 size)
-{
- spin_lock(&dev_priv->mm.object_stat_lock);
- dev_priv->mm.object_count++;
- dev_priv->mm.object_memory += size;
- spin_unlock(&dev_priv->mm.object_stat_lock);
-}
-
-static void i915_gem_info_remove_obj(struct drm_i915_private *dev_priv,
- u64 size)
-{
- spin_lock(&dev_priv->mm.object_stat_lock);
- dev_priv->mm.object_count--;
- dev_priv->mm.object_memory -= size;
- spin_unlock(&dev_priv->mm.object_stat_lock);
-}
-
int
i915_gem_get_aperture_ioctl(struct drm_device *dev, void *data,
struct drm_file *file)
@@ -206,32 +174,6 @@ static int i915_gem_object_get_pages_phys(struct drm_i915_gem_object *obj)
return err;
}
-static void __start_cpu_write(struct drm_i915_gem_object *obj)
-{
- obj->read_domains = I915_GEM_DOMAIN_CPU;
- obj->write_domain = I915_GEM_DOMAIN_CPU;
- if (cpu_write_needs_clflush(obj))
- obj->cache_dirty = true;
-}
-
-void
-__i915_gem_object_release_shmem(struct drm_i915_gem_object *obj,
- struct sg_table *pages,
- bool needs_clflush)
-{
- GEM_BUG_ON(obj->mm.madv == __I915_MADV_PURGED);
-
- if (obj->mm.madv == I915_MADV_DONTNEED)
- obj->mm.dirty = false;
-
- if (needs_clflush &&
- (obj->read_domains & I915_GEM_DOMAIN_CPU) == 0 &&
- !(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ))
- drm_clflush_sg(pages);
-
- __start_cpu_write(obj);
-}
-
static void
i915_gem_object_put_pages_phys(struct drm_i915_gem_object *obj,
struct sg_table *pages)
@@ -283,8 +225,6 @@ static const struct drm_i915_gem_object_ops i915_gem_phys_ops = {
.release = i915_gem_object_release_phys,
};
-static const struct drm_i915_gem_object_ops i915_gem_object_ops;
-
int i915_gem_object_unbind(struct drm_i915_gem_object *obj)
{
struct i915_vma *vma;
@@ -541,7 +481,7 @@ i915_gem_create(struct drm_file *file,
return -EINVAL;
/* Allocate the new object */
- obj = i915_gem_object_create(dev_priv, size);
+ obj = i915_gem_object_create_shmem(dev_priv, size);
if (IS_ERR(obj))
return PTR_ERR(obj);
@@ -2065,52 +2005,6 @@ void __i915_gem_object_truncate(struct drm_i915_gem_object *obj)
obj->mm.pages = ERR_PTR(-EFAULT);
}
-/*
- * Move pages to appropriate lru and release the pagevec, decrementing the
- * ref count of those pages.
- */
-static void check_release_pagevec(struct pagevec *pvec)
-{
- check_move_unevictable_pages(pvec);
- __pagevec_release(pvec);
- cond_resched();
-}
-
-static void
-i915_gem_object_put_pages_gtt(struct drm_i915_gem_object *obj,
- struct sg_table *pages)
-{
- struct sgt_iter sgt_iter;
- struct pagevec pvec;
- struct page *page;
-
- __i915_gem_object_release_shmem(obj, pages, true);
- i915_gem_gtt_finish_pages(obj, pages);
-
- if (i915_gem_object_needs_bit17_swizzle(obj))
- i915_gem_object_save_bit_17_swizzle(obj, pages);
-
- mapping_clear_unevictable(file_inode(obj->base.filp)->i_mapping);
-
- pagevec_init(&pvec);
- for_each_sgt_page(page, sgt_iter, pages) {
- if (obj->mm.dirty)
- set_page_dirty(page);
-
- if (obj->mm.madv == I915_MADV_WILLNEED)
- mark_page_accessed(page);
-
- if (!pagevec_add(&pvec, page))
- check_release_pagevec(&pvec);
- }
- if (pagevec_count(&pvec))
- check_release_pagevec(&pvec);
- obj->mm.dirty = false;
-
- sg_free_table(pages);
- kfree(pages);
-}
-
static void __i915_gem_object_reset_page_iter(struct drm_i915_gem_object *obj)
{
struct radix_tree_iter iter;
@@ -2226,196 +2120,6 @@ bool i915_sg_trim(struct sg_table *orig_st)
return true;
}
-static int i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj)
-{
- struct drm_i915_private *dev_priv = to_i915(obj->base.dev);
- const unsigned long page_count = obj->base.size / PAGE_SIZE;
- unsigned long i;
- struct address_space *mapping;
- struct sg_table *st;
- struct scatterlist *sg;
- struct sgt_iter sgt_iter;
- struct page *page;
- unsigned long last_pfn = 0; /* suppress gcc warning */
- unsigned int max_segment = i915_sg_segment_size();
- unsigned int sg_page_sizes;
- struct pagevec pvec;
- gfp_t noreclaim;
- int ret;
-
- /*
- * Assert that the object is not currently in any GPU domain. As it
- * wasn't in the GTT, there shouldn't be any way it could have been in
- * a GPU cache
- */
- GEM_BUG_ON(obj->read_domains & I915_GEM_GPU_DOMAINS);
- GEM_BUG_ON(obj->write_domain & I915_GEM_GPU_DOMAINS);
-
- /*
- * If there's no chance of allocating enough pages for the whole
- * object, bail early.
- */
- if (page_count > totalram_pages())
- return -ENOMEM;
-
- st = kmalloc(sizeof(*st), GFP_KERNEL);
- if (st == NULL)
- return -ENOMEM;
-
-rebuild_st:
- if (sg_alloc_table(st, page_count, GFP_KERNEL)) {
- kfree(st);
- return -ENOMEM;
- }
-
- /*
- * Get the list of pages out of our struct file. They'll be pinned
- * at this point until we release them.
- *
- * Fail silently without starting the shrinker
- */
- mapping = obj->base.filp->f_mapping;
- mapping_set_unevictable(mapping);
- noreclaim = mapping_gfp_constraint(mapping, ~__GFP_RECLAIM);
- noreclaim |= __GFP_NORETRY | __GFP_NOWARN;
-
- sg = st->sgl;
- st->nents = 0;
- sg_page_sizes = 0;
- for (i = 0; i < page_count; i++) {
- const unsigned int shrink[] = {
- I915_SHRINK_BOUND | I915_SHRINK_UNBOUND | I915_SHRINK_PURGEABLE,
- 0,
- }, *s = shrink;
- gfp_t gfp = noreclaim;
-
- do {
- cond_resched();
- page = shmem_read_mapping_page_gfp(mapping, i, gfp);
- if (!IS_ERR(page))
- break;
-
- if (!*s) {
- ret = PTR_ERR(page);
- goto err_sg;
- }
-
- i915_gem_shrink(dev_priv, 2 * page_count, NULL, *s++);
-
- /*
- * We've tried hard to allocate the memory by reaping
- * our own buffer, now let the real VM do its job and
- * go down in flames if truly OOM.
- *
- * However, since graphics tend to be disposable,
- * defer the oom here by reporting the ENOMEM back
- * to userspace.
- */
- if (!*s) {
- /* reclaim and warn, but no oom */
- gfp = mapping_gfp_mask(mapping);
-
- /*
- * Our bo are always dirty and so we require
- * kswapd to reclaim our pages (direct reclaim
- * does not effectively begin pageout of our
- * buffers on its own). However, direct reclaim
- * only waits for kswapd when under allocation
- * congestion. So as a result __GFP_RECLAIM is
- * unreliable and fails to actually reclaim our
- * dirty pages -- unless you try over and over
- * again with !__GFP_NORETRY. However, we still
- * want to fail this allocation rather than
- * trigger the out-of-memory killer and for
- * this we want __GFP_RETRY_MAYFAIL.
- */
- gfp |= __GFP_RETRY_MAYFAIL;
- }
- } while (1);
-
- if (!i ||
- sg->length >= max_segment ||
- page_to_pfn(page) != last_pfn + 1) {
- if (i) {
- sg_page_sizes |= sg->length;
- sg = sg_next(sg);
- }
- st->nents++;
- sg_set_page(sg, page, PAGE_SIZE, 0);
- } else {
- sg->length += PAGE_SIZE;
- }
- last_pfn = page_to_pfn(page);
-
- /* Check that the i965g/gm workaround works. */
- WARN_ON((gfp & __GFP_DMA32) && (last_pfn >= 0x00100000UL));
- }
- if (sg) { /* loop terminated early; short sg table */
- sg_page_sizes |= sg->length;
- sg_mark_end(sg);
- }
-
- /* Trim unused sg entries to avoid wasting memory. */
- i915_sg_trim(st);
-
- ret = i915_gem_gtt_prepare_pages(obj, st);
- if (ret) {
- /*
- * DMA remapping failed? One possible cause is that
- * it could not reserve enough large entries, asking
- * for PAGE_SIZE chunks instead may be helpful.
- */
- if (max_segment > PAGE_SIZE) {
- for_each_sgt_page(page, sgt_iter, st)
- put_page(page);
- sg_free_table(st);
-
- max_segment = PAGE_SIZE;
- goto rebuild_st;
- } else {
- dev_warn(&dev_priv->drm.pdev->dev,
- "Failed to DMA remap %lu pages\n",
- page_count);
- goto err_pages;
- }
- }
-
- if (i915_gem_object_needs_bit17_swizzle(obj))
- i915_gem_object_do_bit_17_swizzle(obj, st);
-
- __i915_gem_object_set_pages(obj, st, sg_page_sizes);
-
- return 0;
-
-err_sg:
- sg_mark_end(sg);
-err_pages:
- mapping_clear_unevictable(mapping);
- pagevec_init(&pvec);
- for_each_sgt_page(page, sgt_iter, st) {
- if (!pagevec_add(&pvec, page))
- check_release_pagevec(&pvec);
- }
- if (pagevec_count(&pvec))
- check_release_pagevec(&pvec);
- sg_free_table(st);
- kfree(st);
-
- /*
- * shmemfs first checks if there is enough memory to allocate the page
- * and reports ENOSPC should there be insufficient, along with the usual
- * ENOMEM for a genuine allocation failure.
- *
- * We use ENOSPC in our driver to mean that we have run out of aperture
- * space and so want to translate the error from shmemfs back to our
- * usual understanding of ENOMEM.
- */
- if (ret == -ENOSPC)
- ret = -ENOMEM;
-
- return ret;
-}
-
void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj,
struct sg_table *pages,
unsigned int sg_page_sizes)
@@ -2662,133 +2366,6 @@ void __i915_gem_object_flush_map(struct drm_i915_gem_object *obj,
}
}
-static int
-i915_gem_object_pwrite_gtt(struct drm_i915_gem_object *obj,
- const struct drm_i915_gem_pwrite *arg)
-{
- struct address_space *mapping = obj->base.filp->f_mapping;
- char __user *user_data = u64_to_user_ptr(arg->data_ptr);
- u64 remain, offset;
- unsigned int pg;
-
- /* Caller already validated user args */
- GEM_BUG_ON(!access_ok(user_data, arg->size));
-
- /*
- * Before we instantiate/pin the backing store for our use, we
- * can prepopulate the shmemfs filp efficiently using a write into
- * the pagecache. We avoid the penalty of instantiating all the
- * pages, important if the user is just writing to a few and never
- * uses the object on the GPU, and using a direct write into shmemfs
- * allows it to avoid the cost of retrieving a page (either swapin
- * or clearing-before-use) before it is overwritten.
- */
- if (i915_gem_object_has_pages(obj))
- return -ENODEV;
-
- if (obj->mm.madv != I915_MADV_WILLNEED)
- return -EFAULT;
-
- /*
- * Before the pages are instantiated the object is treated as being
- * in the CPU domain. The pages will be clflushed as required before
- * use, and we can freely write into the pages directly. If userspace
- * races pwrite with any other operation; corruption will ensue -
- * that is userspace's prerogative!
- */
-
- remain = arg->size;
- offset = arg->offset;
- pg = offset_in_page(offset);
-
- do {
- unsigned int len, unwritten;
- struct page *page;
- void *data, *vaddr;
- int err;
- char c;
-
- len = PAGE_SIZE - pg;
- if (len > remain)
- len = remain;
-
- /* Prefault the user page to reduce potential recursion */
- err = __get_user(c, user_data);
- if (err)
- return err;
-
- err = __get_user(c, user_data + len - 1);
- if (err)
- return err;
-
- err = pagecache_write_begin(obj->base.filp, mapping,
- offset, len, 0,
- &page, &data);
- if (err < 0)
- return err;
-
- vaddr = kmap_atomic(page);
- unwritten = __copy_from_user_inatomic(vaddr + pg,
- user_data,
- len);
- kunmap_atomic(vaddr);
-
- err = pagecache_write_end(obj->base.filp, mapping,
- offset, len, len - unwritten,
- page, data);
- if (err < 0)
- return err;
-
- /* We don't handle -EFAULT, leave it to the caller to check */
- if (unwritten)
- return -ENODEV;
-
- remain -= len;
- user_data += len;
- offset += len;
- pg = 0;
- } while (remain);
-
- return 0;
-}
-
-void i915_gem_close_object(struct drm_gem_object *gem, struct drm_file *file)
-{
- struct drm_i915_private *i915 = to_i915(gem->dev);
- struct drm_i915_gem_object *obj = to_intel_bo(gem);
- struct drm_i915_file_private *fpriv = file->driver_priv;
- struct i915_lut_handle *lut, *ln;
-
- mutex_lock(&i915->drm.struct_mutex);
-
- list_for_each_entry_safe(lut, ln, &obj->lut_list, obj_link) {
- struct i915_gem_context *ctx = lut->ctx;
- struct i915_vma *vma;
-
- GEM_BUG_ON(ctx->file_priv == ERR_PTR(-EBADF));
- if (ctx->file_priv != fpriv)
- continue;
-
- vma = radix_tree_delete(&ctx->handles_vma, lut->handle);
- GEM_BUG_ON(vma->obj != obj);
-
- /* We allow the process to have multiple handles to the same
- * vma, in the same fd namespace, by virtue of flink/open.
- */
- GEM_BUG_ON(!vma->open_count);
- if (!--vma->open_count && !i915_vma_is_ggtt(vma))
- i915_vma_close(vma);
-
- list_del(&lut->obj_link);
- list_del(&lut->ctx_link);
-
- i915_lut_handle_free(lut);
- __i915_gem_object_release_unless_active(obj);
- }
-
- mutex_unlock(&i915->drm.struct_mutex);
-}
-
static unsigned long to_wait_timeout(s64 timeout_ns)
{
if (timeout_ns < 0)
@@ -3787,348 +3364,6 @@ i915_gem_madvise_ioctl(struct drm_device *dev, void *data,
return err;
}
-static void
-frontbuffer_retire(struct i915_active_request *active,
- struct i915_request *request)
-{
- struct drm_i915_gem_object *obj =
- container_of(active, typeof(*obj), frontbuffer_write);
-
- intel_fb_obj_flush(obj, ORIGIN_CS);
-}
-
-void i915_gem_object_init(struct drm_i915_gem_object *obj,
- const struct drm_i915_gem_object_ops *ops)
-{
- mutex_init(&obj->mm.lock);
-
- spin_lock_init(&obj->vma.lock);
- INIT_LIST_HEAD(&obj->vma.list);
-
- INIT_LIST_HEAD(&obj->lut_list);
- INIT_LIST_HEAD(&obj->batch_pool_link);
-
- init_rcu_head(&obj->rcu);
-
- obj->ops = ops;
-
- reservation_object_init(&obj->__builtin_resv);
- obj->resv = &obj->__builtin_resv;
-
- obj->frontbuffer_ggtt_origin = ORIGIN_GTT;
- i915_active_request_init(&obj->frontbuffer_write,
- NULL, frontbuffer_retire);
-
- obj->mm.madv = I915_MADV_WILLNEED;
- INIT_RADIX_TREE(&obj->mm.get_page.radix, GFP_KERNEL | __GFP_NOWARN);
- mutex_init(&obj->mm.get_page.lock);
-
- i915_gem_info_add_obj(to_i915(obj->base.dev), obj->base.size);
-}
-
-static const struct drm_i915_gem_object_ops i915_gem_object_ops = {
- .flags = I915_GEM_OBJECT_HAS_STRUCT_PAGE |
- I915_GEM_OBJECT_IS_SHRINKABLE,
-
- .get_pages = i915_gem_object_get_pages_gtt,
- .put_pages = i915_gem_object_put_pages_gtt,
-
- .pwrite = i915_gem_object_pwrite_gtt,
-};
-
-static int i915_gem_object_create_shmem(struct drm_device *dev,
- struct drm_gem_object *obj,
- size_t size)
-{
- struct drm_i915_private *i915 = to_i915(dev);
- unsigned long flags = VM_NORESERVE;
- struct file *filp;
-
- drm_gem_private_object_init(dev, obj, size);
-
- if (i915->mm.gemfs)
- filp = shmem_file_setup_with_mnt(i915->mm.gemfs, "i915", size,
- flags);
- else
- filp = shmem_file_setup("i915", size, flags);
-
- if (IS_ERR(filp))
- return PTR_ERR(filp);
-
- obj->filp = filp;
-
- return 0;
-}
-
-struct drm_i915_gem_object *
-i915_gem_object_create(struct drm_i915_private *dev_priv, u64 size)
-{
- struct drm_i915_gem_object *obj;
- struct address_space *mapping;
- unsigned int cache_level;
- gfp_t mask;
- int ret;
-
- /* There is a prevalence of the assumption that we fit the object's
- * page count inside a 32bit _signed_ variable. Let's document this and
- * catch if we ever need to fix it. In the meantime, if you do spot
- * such a local variable, please consider fixing!
- */
- if (size >> PAGE_SHIFT > INT_MAX)
- return ERR_PTR(-E2BIG);
-
- if (overflows_type(size, obj->base.size))
- return ERR_PTR(-E2BIG);
-
- obj = i915_gem_object_alloc();
- if (obj == NULL)
- return ERR_PTR(-ENOMEM);
-
- ret = i915_gem_object_create_shmem(&dev_priv->drm, &obj->base, size);
- if (ret)
- goto fail;
-
- mask = GFP_HIGHUSER | __GFP_RECLAIMABLE;
- if (IS_I965GM(dev_priv) || IS_I965G(dev_priv)) {
- /* 965gm cannot relocate objects above 4GiB. */
- mask &= ~__GFP_HIGHMEM;
- mask |= __GFP_DMA32;
- }
-
- mapping = obj->base.filp->f_mapping;
- mapping_set_gfp_mask(mapping, mask);
- GEM_BUG_ON(!(mapping_gfp_mask(mapping) & __GFP_RECLAIM));
-
- i915_gem_object_init(obj, &i915_gem_object_ops);
-
- obj->write_domain = I915_GEM_DOMAIN_CPU;
- obj->read_domains = I915_GEM_DOMAIN_CPU;
-
- if (HAS_LLC(dev_priv))
- /* On some devices, we can have the GPU use the LLC (the CPU
- * cache) for about a 10% performance improvement
- * compared to uncached. Graphics requests other than
- * display scanout are coherent with the CPU in
- * accessing this cache. This means in this mode we
- * don't need to clflush on the CPU side, and on the
- * GPU side we only need to flush internal caches to
- * get data visible to the CPU.
- *
- * However, we maintain the display planes as UC, and so
- * need to rebind when first used as such.
- */
- cache_level = I915_CACHE_LLC;
- else
- cache_level = I915_CACHE_NONE;
-
- i915_gem_object_set_cache_coherency(obj, cache_level);
-
- trace_i915_gem_object_create(obj);
-
- return obj;
-
-fail:
- i915_gem_object_free(obj);
- return ERR_PTR(ret);
-}
-
-static bool discard_backing_storage(struct drm_i915_gem_object *obj)
-{
- /* If we are the last user of the backing storage (be it shmemfs
- * pages or stolen etc), we know that the pages are going to be
- * immediately released. In this case, we can then skip copying
- * back the contents from the GPU.
- */
-
- if (obj->mm.madv != I915_MADV_WILLNEED)
- return false;
-
- if (obj->base.filp == NULL)
- return true;
-
- /* At first glance, this looks racy, but then again so would be
- * userspace racing mmap against close. However, the first external
- * reference to the filp can only be obtained through the
- * i915_gem_mmap_ioctl() which safeguards us against the user
- * acquiring such a reference whilst we are in the middle of
- * freeing the object.
- */
- return atomic_long_read(&obj->base.filp->f_count) == 1;
-}
-
-static void __i915_gem_free_objects(struct drm_i915_private *i915,
- struct llist_node *freed)
-{
- struct drm_i915_gem_object *obj, *on;
- intel_wakeref_t wakeref;
-
- wakeref = intel_runtime_pm_get(i915);
- llist_for_each_entry_safe(obj, on, freed, freed) {
- struct i915_vma *vma, *vn;
-
- trace_i915_gem_object_destroy(obj);
-
- mutex_lock(&i915->drm.struct_mutex);
-
- GEM_BUG_ON(i915_gem_object_is_active(obj));
- list_for_each_entry_safe(vma, vn, &obj->vma.list, obj_link) {
- GEM_BUG_ON(i915_vma_is_active(vma));
- vma->flags &= ~I915_VMA_PIN_MASK;
- i915_vma_destroy(vma);
- }
- GEM_BUG_ON(!list_empty(&obj->vma.list));
- GEM_BUG_ON(!RB_EMPTY_ROOT(&obj->vma.tree));
-
- /* This serializes freeing with the shrinker. Since the free
- * is delayed, first by RCU then by the workqueue, we want the
- * shrinker to be able to free pages of unreferenced objects,
- * or else we may oom whilst there are plenty of deferred
- * freed objects.
- */
- if (i915_gem_object_has_pages(obj)) {
- spin_lock(&i915->mm.obj_lock);
- list_del_init(&obj->mm.link);
- spin_unlock(&i915->mm.obj_lock);
- }
-
- mutex_unlock(&i915->drm.struct_mutex);
-
- GEM_BUG_ON(obj->bind_count);
- GEM_BUG_ON(obj->userfault_count);
- GEM_BUG_ON(atomic_read(&obj->frontbuffer_bits));
- GEM_BUG_ON(!list_empty(&obj->lut_list));
-
- if (obj->ops->release)
- obj->ops->release(obj);
-
- if (WARN_ON(i915_gem_object_has_pinned_pages(obj)))
- atomic_set(&obj->mm.pages_pin_count, 0);
- __i915_gem_object_put_pages(obj, I915_MM_NORMAL);
- GEM_BUG_ON(i915_gem_object_has_pages(obj));
-
- if (obj->base.import_attach)
- drm_prime_gem_destroy(&obj->base, NULL);
-
- reservation_object_fini(&obj->__builtin_resv);
- drm_gem_object_release(&obj->base);
- i915_gem_info_remove_obj(i915, obj->base.size);
-
- bitmap_free(obj->bit_17);
- i915_gem_object_free(obj);
-
- GEM_BUG_ON(!atomic_read(&i915->mm.free_count));
- atomic_dec(&i915->mm.free_count);
-
- if (on)
- cond_resched();
- }
- intel_runtime_pm_put(i915, wakeref);
-}
-
-static void i915_gem_flush_free_objects(struct drm_i915_private *i915)
-{
- struct llist_node *freed;
-
- /* Free the oldest, most stale object to keep the free_list short */
- freed = NULL;
- if (!llist_empty(&i915->mm.free_list)) { /* quick test for hotpath */
- /* Only one consumer of llist_del_first() allowed */
- spin_lock(&i915->mm.free_lock);
- freed = llist_del_first(&i915->mm.free_list);
- spin_unlock(&i915->mm.free_lock);
- }
- if (unlikely(freed)) {
- freed->next = NULL;
- __i915_gem_free_objects(i915, freed);
- }
-}
-
-static void __i915_gem_free_work(struct work_struct *work)
-{
- struct drm_i915_private *i915 =
- container_of(work, struct drm_i915_private, mm.free_work);
- struct llist_node *freed;
-
- /*
- * All file-owned VMA should have been released by this point through
- * i915_gem_close_object(), or earlier by i915_gem_context_close().
- * However, the object may also be bound into the global GTT (e.g.
- * older GPUs without per-process support, or for direct access through
- * the GTT either for the user or for scanout). Those VMA still need to
- * unbound now.
- */
-
- spin_lock(&i915->mm.free_lock);
- while ((freed = llist_del_all(&i915->mm.free_list))) {
- spin_unlock(&i915->mm.free_lock);
-
- __i915_gem_free_objects(i915, freed);
- if (need_resched())
- return;
-
- spin_lock(&i915->mm.free_lock);
- }
- spin_unlock(&i915->mm.free_lock);
-}
-
-static void __i915_gem_free_object_rcu(struct rcu_head *head)
-{
- struct drm_i915_gem_object *obj =
- container_of(head, typeof(*obj), rcu);
- struct drm_i915_private *i915 = to_i915(obj->base.dev);
-
- /*
- * We reuse obj->rcu for the freed list, so we had better not treat
- * it like a rcu_head from this point forwards. And we expect all
- * objects to be freed via this path.
- */
- destroy_rcu_head(&obj->rcu);
-
- /*
- * Since we require blocking on struct_mutex to unbind the freed
- * object from the GPU before releasing resources back to the
- * system, we can not do that directly from the RCU callback (which may
- * be a softirq context), but must instead then defer that work onto a
- * kthread. We use the RCU callback rather than move the freed object
- * directly onto the work queue so that we can mix between using the
- * worker and performing frees directly from subsequent allocations for
- * crude but effective memory throttling.
- */
- if (llist_add(&obj->freed, &i915->mm.free_list))
- queue_work(i915->wq, &i915->mm.free_work);
-}
-
-void i915_gem_free_object(struct drm_gem_object *gem_obj)
-{
- struct drm_i915_gem_object *obj = to_intel_bo(gem_obj);
-
- if (obj->mm.quirked)
- __i915_gem_object_unpin_pages(obj);
-
- if (discard_backing_storage(obj))
- obj->mm.madv = I915_MADV_DONTNEED;
-
- /*
- * Before we free the object, make sure any pure RCU-only
- * read-side critical sections are complete, e.g.
- * i915_gem_busy_ioctl(). For the corresponding synchronized
- * lookup see i915_gem_object_lookup_rcu().
- */
- atomic_inc(&to_i915(obj->base.dev)->mm.free_count);
- call_rcu(&obj->rcu, __i915_gem_free_object_rcu);
-}
-
-void __i915_gem_object_release_unless_active(struct drm_i915_gem_object *obj)
-{
- lockdep_assert_held(&obj->base.dev->struct_mutex);
-
- if (!i915_gem_object_has_active_reference(obj) &&
- i915_gem_object_is_active(obj))
- i915_gem_object_set_active_reference(obj);
- else
- i915_gem_object_put(obj);
-}
-
void i915_gem_sanitize(struct drm_i915_private *i915)
{
intel_wakeref_t wakeref;
@@ -4725,7 +3960,7 @@ static void i915_gem_init__mm(struct drm_i915_private *i915)
INIT_LIST_HEAD(&i915->mm.fence_list);
INIT_LIST_HEAD(&i915->mm.userfault_list);
- INIT_WORK(&i915->mm.free_work, __i915_gem_free_work);
+ i915_gem_init__objects(i915);
}
int i915_gem_init_early(struct drm_i915_private *dev_priv)
@@ -4891,57 +4126,6 @@ void i915_gem_track_fb(struct drm_i915_gem_object *old,
}
}
-/* Allocate a new GEM object and fill it with the supplied data */
-struct drm_i915_gem_object *
-i915_gem_object_create_from_data(struct drm_i915_private *dev_priv,
- const void *data, size_t size)
-{
- struct drm_i915_gem_object *obj;
- struct file *file;
- size_t offset;
- int err;
-
- obj = i915_gem_object_create(dev_priv, round_up(size, PAGE_SIZE));
- if (IS_ERR(obj))
- return obj;
-
- GEM_BUG_ON(obj->write_domain != I915_GEM_DOMAIN_CPU);
-
- file = obj->base.filp;
- offset = 0;
- do {
- unsigned int len = min_t(typeof(size), size, PAGE_SIZE);
- struct page *page;
- void *pgdata, *vaddr;
-
- err = pagecache_write_begin(file, file->f_mapping,
- offset, len, 0,
- &page, &pgdata);
- if (err < 0)
- goto fail;
-
- vaddr = kmap(page);
- memcpy(vaddr, data, len);
- kunmap(page);
-
- err = pagecache_write_end(file, file->f_mapping,
- offset, len, len,
- page, pgdata);
- if (err < 0)
- goto fail;
-
- size -= len;
- data += len;
- offset += len;
- } while (size);
-
- return obj;
-
-fail:
- i915_gem_object_put(obj);
- return ERR_PTR(err);
-}
-
struct scatterlist *
i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
unsigned int n,
@@ -5103,7 +4287,7 @@ int i915_gem_object_attach_phys(struct drm_i915_gem_object *obj, int align)
if (obj->ops == &i915_gem_phys_ops)
return 0;
- if (obj->ops != &i915_gem_object_ops)
+ if (obj->ops != &i915_gem_shmem_ops)
return -EINVAL;
err = i915_gem_object_unbind(obj);
@@ -5139,12 +4323,12 @@ int i915_gem_object_attach_phys(struct drm_i915_gem_object *obj, int align)
__i915_gem_object_pin_pages(obj);
if (!IS_ERR_OR_NULL(pages))
- i915_gem_object_ops.put_pages(obj, pages);
+ i915_gem_shmem_ops.put_pages(obj, pages);
mutex_unlock(&obj->mm.lock);
return 0;
err_xfer:
- obj->ops = &i915_gem_object_ops;
+ obj->ops = &i915_gem_shmem_ops;
if (!IS_ERR_OR_NULL(pages)) {
unsigned int sg_page_sizes = i915_sg_page_sizes(pages->sgl);
diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index c4995d5a16d2..379fd89a180f 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -1510,7 +1510,7 @@ static int alloc_oa_buffer(struct drm_i915_private *dev_priv)
BUILD_BUG_ON_NOT_POWER_OF_2(OA_BUFFER_SIZE);
BUILD_BUG_ON(OA_BUFFER_SIZE < SZ_128K || OA_BUFFER_SIZE > SZ_16M);
- bo = i915_gem_object_create(dev_priv, OA_BUFFER_SIZE);
+ bo = i915_gem_object_create_shmem(dev_priv, OA_BUFFER_SIZE);
if (IS_ERR(bo)) {
DRM_ERROR("Failed to allocate OA buffer\n");
ret = PTR_ERR(bo);
diff --git a/drivers/gpu/drm/i915/intel_fbdev.c b/drivers/gpu/drm/i915/intel_fbdev.c
index 89db71996148..0d3a6fa674e6 100644
--- a/drivers/gpu/drm/i915/intel_fbdev.c
+++ b/drivers/gpu/drm/i915/intel_fbdev.c
@@ -144,7 +144,7 @@ static int intelfb_alloc(struct drm_fb_helper *helper,
if (size * 2 < dev_priv->stolen_usable_size)
obj = i915_gem_object_create_stolen(dev_priv, size);
if (obj == NULL)
- obj = i915_gem_object_create(dev_priv, size);
+ obj = i915_gem_object_create_shmem(dev_priv, size);
if (IS_ERR(obj)) {
DRM_ERROR("failed to allocate framebuffer\n");
ret = PTR_ERR(obj);
diff --git a/drivers/gpu/drm/i915/intel_guc.c b/drivers/gpu/drm/i915/intel_guc.c
index c4ac29309fcc..98e45a500a15 100644
--- a/drivers/gpu/drm/i915/intel_guc.c
+++ b/drivers/gpu/drm/i915/intel_guc.c
@@ -690,7 +690,7 @@ struct i915_vma *intel_guc_allocate_vma(struct intel_guc *guc, u32 size)
u64 flags;
int ret;
- obj = i915_gem_object_create(dev_priv, size);
+ obj = i915_gem_object_create_shmem(dev_priv, size);
if (IS_ERR(obj))
return ERR_CAST(obj);
diff --git a/drivers/gpu/drm/i915/intel_uc_fw.c b/drivers/gpu/drm/i915/intel_uc_fw.c
index b9cb6fea9332..3257a054cb0b 100644
--- a/drivers/gpu/drm/i915/intel_uc_fw.c
+++ b/drivers/gpu/drm/i915/intel_uc_fw.c
@@ -159,7 +159,8 @@ void intel_uc_fw_fetch(struct drm_i915_private *dev_priv,
goto fail;
}
- obj = i915_gem_object_create_from_data(dev_priv, fw->data, fw->size);
+ obj = i915_gem_object_create_shmem_from_data(dev_priv,
+ fw->data, fw->size);
if (IS_ERR(obj)) {
err = PTR_ERR(obj);
DRM_DEBUG_DRIVER("%s fw object_create err=%d\n",
diff --git a/drivers/gpu/drm/i915/selftests/huge_pages.c b/drivers/gpu/drm/i915/selftests/huge_pages.c
index 1e1f83326a96..ce4ec87698f6 100644
--- a/drivers/gpu/drm/i915/selftests/huge_pages.c
+++ b/drivers/gpu/drm/i915/selftests/huge_pages.c
@@ -1390,7 +1390,7 @@ static int igt_ppgtt_gemfs_huge(void *arg)
for (i = 0; i < ARRAY_SIZE(sizes); ++i) {
unsigned int size = sizes[i];
- obj = i915_gem_object_create(i915, size);
+ obj = i915_gem_object_create_shmem(i915, size);
if (IS_ERR(obj))
return PTR_ERR(obj);
@@ -1568,7 +1568,7 @@ static int igt_tmpfs_fallback(void *arg)
i915->mm.gemfs = NULL;
- obj = i915_gem_object_create(i915, PAGE_SIZE);
+ obj = i915_gem_object_create_shmem(i915, PAGE_SIZE);
if (IS_ERR(obj)) {
err = PTR_ERR(obj);
goto out_restore;
@@ -1628,7 +1628,7 @@ static int igt_shrink_thp(void *arg)
return 0;
}
- obj = i915_gem_object_create(i915, SZ_2M);
+ obj = i915_gem_object_create_shmem(i915, SZ_2M);
if (IS_ERR(obj))
return PTR_ERR(obj);
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_dmabuf.c b/drivers/gpu/drm/i915/selftests/i915_gem_dmabuf.c
index 2b943ee246c9..cc65a503e2f0 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_dmabuf.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_dmabuf.c
@@ -33,7 +33,7 @@ static int igt_dmabuf_export(void *arg)
struct drm_i915_gem_object *obj;
struct dma_buf *dmabuf;
- obj = i915_gem_object_create(i915, PAGE_SIZE);
+ obj = i915_gem_object_create_shmem(i915, PAGE_SIZE);
if (IS_ERR(obj))
return PTR_ERR(obj);
@@ -57,7 +57,7 @@ static int igt_dmabuf_import_self(void *arg)
struct dma_buf *dmabuf;
int err;
- obj = i915_gem_object_create(i915, PAGE_SIZE);
+ obj = i915_gem_object_create_shmem(i915, PAGE_SIZE);
if (IS_ERR(obj))
return PTR_ERR(obj);
@@ -232,7 +232,7 @@ static int igt_dmabuf_export_vmap(void *arg)
void *ptr;
int err;
- obj = i915_gem_object_create(i915, PAGE_SIZE);
+ obj = i915_gem_object_create_shmem(i915, PAGE_SIZE);
if (IS_ERR(obj))
return PTR_ERR(obj);
@@ -279,7 +279,7 @@ static int igt_dmabuf_export_kmap(void *arg)
void *ptr;
int err;
- obj = i915_gem_object_create(i915, 2*PAGE_SIZE);
+ obj = i915_gem_object_create_shmem(i915, 2 * PAGE_SIZE);
if (IS_ERR(obj))
return PTR_ERR(obj);
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_object.c b/drivers/gpu/drm/i915/selftests/i915_gem_object.c
index b926d1cd165d..a0f59fb0d701 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_object.c
@@ -36,7 +36,7 @@ static int igt_gem_object(void *arg)
/* Basic test to ensure we can create an object */
- obj = i915_gem_object_create(i915, PAGE_SIZE);
+ obj = i915_gem_object_create_shmem(i915, PAGE_SIZE);
if (IS_ERR(obj)) {
err = PTR_ERR(obj);
pr_err("i915_gem_object_create failed, err=%d\n", err);
@@ -59,7 +59,7 @@ static int igt_phys_object(void *arg)
* i.e. exercise the i915_gem_object_phys API.
*/
- obj = i915_gem_object_create(i915, PAGE_SIZE);
+ obj = i915_gem_object_create_shmem(i915, PAGE_SIZE);
if (IS_ERR(obj)) {
err = PTR_ERR(obj);
pr_err("i915_gem_object_create failed, err=%d\n", err);
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 23/40] drm/i915: Move phys objects to its own file
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (20 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 22/40] drm/i915: Move shmem object setup to its own file Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 24/40] drm/i915: Move mmap and friends " Chris Wilson
` (23 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx; +Cc: Matthew Auld
Continuing the decluttering of i915_gem.c, this time the legacy physical
object.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
---
drivers/gpu/drm/i915/Makefile | 2 +
drivers/gpu/drm/i915/gem/i915_gem_object.h | 11 +-
.../gpu/drm/i915/gem/i915_gem_object_types.h | 2 +
drivers/gpu/drm/i915/gem/i915_gem_pages.c | 510 +++++++++++++
drivers/gpu/drm/i915/gem/i915_gem_phys.c | 212 ++++++
drivers/gpu/drm/i915/gem/i915_gem_shmem.c | 61 ++
.../drm/i915/gem/selftests/i915_gem_phys.c | 80 ++
drivers/gpu/drm/i915/i915_drv.h | 2 -
drivers/gpu/drm/i915/i915_gem.c | 699 +-----------------
drivers/gpu/drm/i915/i915_gem_shrinker.c | 59 +-
.../gpu/drm/i915/selftests/i915_gem_object.c | 54 --
.../drm/i915/selftests/i915_mock_selftests.h | 1 +
12 files changed, 885 insertions(+), 808 deletions(-)
create mode 100644 drivers/gpu/drm/i915/gem/i915_gem_pages.c
create mode 100644 drivers/gpu/drm/i915/gem/i915_gem_phys.c
create mode 100644 drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index 625f9749355b..ba3b82f3cd49 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -88,6 +88,8 @@ i915-y += $(gt-y)
obj-y += gem/
gem-y += \
gem/i915_gem_object.o \
+ gem/i915_gem_pages.o \
+ gem/i915_gem_phys.o \
gem/i915_gem_shmem.o
i915-y += \
$(gem-y) \
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index ad82303f741a..2e963a593245 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -33,11 +33,17 @@ void __i915_gem_object_release_shmem(struct drm_i915_gem_object *obj,
struct sg_table *pages,
bool needs_clflush);
+int i915_gem_object_attach_phys(struct drm_i915_gem_object *obj, int align);
+
void i915_gem_close_object(struct drm_gem_object *gem, struct drm_file *file);
void i915_gem_free_object(struct drm_gem_object *obj);
void i915_gem_flush_free_objects(struct drm_i915_private *i915);
+struct sg_table *
+__i915_gem_object_unset_pages(struct drm_i915_gem_object *obj);
+void i915_gem_object_truncate(struct drm_i915_gem_object *obj);
+
/**
* i915_gem_object_lookup_rcu - look up a temporary GEM object from its handle
* @filp: DRM file private date
@@ -231,6 +237,8 @@ i915_gem_object_get_dma_address(struct drm_i915_gem_object *obj,
void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj,
struct sg_table *pages,
unsigned int sg_page_sizes);
+
+int ____i915_gem_object_get_pages(struct drm_i915_gem_object *obj);
int __i915_gem_object_get_pages(struct drm_i915_gem_object *obj);
static inline int __must_check
@@ -286,7 +294,8 @@ enum i915_mm_subclass { /* lockdep subclass for obj->mm.lock/struct_mutex */
int __i915_gem_object_put_pages(struct drm_i915_gem_object *obj,
enum i915_mm_subclass subclass);
-void __i915_gem_object_truncate(struct drm_i915_gem_object *obj);
+void i915_gem_object_truncate(struct drm_i915_gem_object *obj);
+void i915_gem_object_writeback(struct drm_i915_gem_object *obj);
enum i915_map_type {
I915_MAP_WB = 0,
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
index e4b50944f553..24abdd6999b5 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
@@ -52,6 +52,8 @@ struct drm_i915_gem_object_ops {
int (*get_pages)(struct drm_i915_gem_object *obj);
void (*put_pages)(struct drm_i915_gem_object *obj,
struct sg_table *pages);
+ void (*truncate)(struct drm_i915_gem_object *obj);
+ void (*writeback)(struct drm_i915_gem_object *obj);
int (*pwrite)(struct drm_i915_gem_object *obj,
const struct drm_i915_gem_pwrite *arg);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pages.c b/drivers/gpu/drm/i915/gem/i915_gem_pages.c
new file mode 100644
index 000000000000..452aba0a3359
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pages.c
@@ -0,0 +1,510 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2014-2016 Intel Corporation
+ */
+
+#include "i915_gem_object.h"
+
+#include "../i915_drv.h"
+
+void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj,
+ struct sg_table *pages,
+ unsigned int sg_page_sizes)
+{
+ struct drm_i915_private *i915 = to_i915(obj->base.dev);
+ unsigned long supported = INTEL_INFO(i915)->page_sizes;
+ int i;
+
+ lockdep_assert_held(&obj->mm.lock);
+
+ /* Make the pages coherent with the GPU (flushing any swapin). */
+ if (obj->cache_dirty) {
+ obj->write_domain = 0;
+ if (i915_gem_object_has_struct_page(obj))
+ drm_clflush_sg(pages);
+ obj->cache_dirty = false;
+ }
+
+ obj->mm.get_page.sg_pos = pages->sgl;
+ obj->mm.get_page.sg_idx = 0;
+
+ obj->mm.pages = pages;
+
+ if (i915_gem_object_is_tiled(obj) &&
+ i915->quirks & QUIRK_PIN_SWIZZLED_PAGES) {
+ GEM_BUG_ON(obj->mm.quirked);
+ __i915_gem_object_pin_pages(obj);
+ obj->mm.quirked = true;
+ }
+
+ GEM_BUG_ON(!sg_page_sizes);
+ obj->mm.page_sizes.phys = sg_page_sizes;
+
+ /*
+ * Calculate the supported page-sizes which fit into the given
+ * sg_page_sizes. This will give us the page-sizes which we may be able
+ * to use opportunistically when later inserting into the GTT. For
+ * example if phys=2G, then in theory we should be able to use 1G, 2M,
+ * 64K or 4K pages, although in practice this will depend on a number of
+ * other factors.
+ */
+ obj->mm.page_sizes.sg = 0;
+ for_each_set_bit(i, &supported, ilog2(I915_GTT_MAX_PAGE_SIZE) + 1) {
+ if (obj->mm.page_sizes.phys & ~0u << i)
+ obj->mm.page_sizes.sg |= BIT(i);
+ }
+ GEM_BUG_ON(!HAS_PAGE_SIZES(i915, obj->mm.page_sizes.sg));
+
+ spin_lock(&i915->mm.obj_lock);
+ list_add(&obj->mm.link, &i915->mm.unbound_list);
+ spin_unlock(&i915->mm.obj_lock);
+}
+
+int ____i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
+{
+ int err;
+
+ if (unlikely(obj->mm.madv != I915_MADV_WILLNEED)) {
+ DRM_DEBUG("Attempting to obtain a purgeable object\n");
+ return -EFAULT;
+ }
+
+ err = obj->ops->get_pages(obj);
+ GEM_BUG_ON(!err && !i915_gem_object_has_pages(obj));
+
+ return err;
+}
+
+/* Ensure that the associated pages are gathered from the backing storage
+ * and pinned into our object. i915_gem_object_pin_pages() may be called
+ * multiple times before they are released by a single call to
+ * i915_gem_object_unpin_pages() - once the pages are no longer referenced
+ * either as a result of memory pressure (reaping pages under the shrinker)
+ * or as the object is itself released.
+ */
+int __i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
+{
+ int err;
+
+ err = mutex_lock_interruptible(&obj->mm.lock);
+ if (err)
+ return err;
+
+ if (unlikely(!i915_gem_object_has_pages(obj))) {
+ GEM_BUG_ON(i915_gem_object_has_pinned_pages(obj));
+
+ err = ____i915_gem_object_get_pages(obj);
+ if (err)
+ goto unlock;
+
+ smp_mb__before_atomic();
+ }
+ atomic_inc(&obj->mm.pages_pin_count);
+
+unlock:
+ mutex_unlock(&obj->mm.lock);
+ return err;
+}
+
+/* Immediately discard the backing storage */
+void i915_gem_object_truncate(struct drm_i915_gem_object *obj)
+{
+ drm_gem_free_mmap_offset(&obj->base);
+ if (obj->ops->truncate)
+ obj->ops->truncate(obj);
+}
+
+/* Try to discard unwanted pages */
+void i915_gem_object_writeback(struct drm_i915_gem_object *obj)
+{
+ lockdep_assert_held(&obj->mm.lock);
+ GEM_BUG_ON(i915_gem_object_has_pages(obj));
+
+ if (obj->ops->writeback)
+ obj->ops->writeback(obj);
+}
+
+static void __i915_gem_object_reset_page_iter(struct drm_i915_gem_object *obj)
+{
+ struct radix_tree_iter iter;
+ void __rcu **slot;
+
+ rcu_read_lock();
+ radix_tree_for_each_slot(slot, &obj->mm.get_page.radix, &iter, 0)
+ radix_tree_delete(&obj->mm.get_page.radix, iter.index);
+ rcu_read_unlock();
+}
+
+struct sg_table *
+__i915_gem_object_unset_pages(struct drm_i915_gem_object *obj)
+{
+ struct drm_i915_private *i915 = to_i915(obj->base.dev);
+ struct sg_table *pages;
+
+ pages = fetch_and_zero(&obj->mm.pages);
+ if (IS_ERR_OR_NULL(pages))
+ return pages;
+
+ spin_lock(&i915->mm.obj_lock);
+ list_del(&obj->mm.link);
+ spin_unlock(&i915->mm.obj_lock);
+
+ if (obj->mm.mapping) {
+ void *ptr;
+
+ ptr = page_mask_bits(obj->mm.mapping);
+ if (is_vmalloc_addr(ptr))
+ vunmap(ptr);
+ else
+ kunmap(kmap_to_page(ptr));
+
+ obj->mm.mapping = NULL;
+ }
+
+ __i915_gem_object_reset_page_iter(obj);
+ obj->mm.page_sizes.phys = obj->mm.page_sizes.sg = 0;
+
+ return pages;
+}
+
+int __i915_gem_object_put_pages(struct drm_i915_gem_object *obj,
+ enum i915_mm_subclass subclass)
+{
+ struct sg_table *pages;
+ int err;
+
+ if (i915_gem_object_has_pinned_pages(obj))
+ return -EBUSY;
+
+ GEM_BUG_ON(obj->bind_count);
+
+ /* May be called by shrinker from within get_pages() (on another bo) */
+ mutex_lock_nested(&obj->mm.lock, subclass);
+ if (unlikely(atomic_read(&obj->mm.pages_pin_count))) {
+ err = -EBUSY;
+ goto unlock;
+ }
+
+ /*
+ * ->put_pages might need to allocate memory for the bit17 swizzle
+ * array, hence protect them from being reaped by removing them from gtt
+ * lists early.
+ */
+ pages = __i915_gem_object_unset_pages(obj);
+
+ /*
+ * XXX Temporary hijinx to avoid updating all backends to handle
+ * NULL pages. In the future, when we have more asynchronous
+ * get_pages backends we should be better able to handle the
+ * cancellation of the async task in a more uniform manner.
+ */
+ if (!pages && !i915_gem_object_needs_async_cancel(obj))
+ pages = ERR_PTR(-EINVAL);
+
+ if (!IS_ERR(pages))
+ obj->ops->put_pages(obj, pages);
+
+ err = 0;
+unlock:
+ mutex_unlock(&obj->mm.lock);
+
+ return err;
+}
+
+/* The 'mapping' part of i915_gem_object_pin_map() below */
+static void *i915_gem_object_map(const struct drm_i915_gem_object *obj,
+ enum i915_map_type type)
+{
+ unsigned long n_pages = obj->base.size >> PAGE_SHIFT;
+ struct sg_table *sgt = obj->mm.pages;
+ struct sgt_iter sgt_iter;
+ struct page *page;
+ struct page *stack_pages[32];
+ struct page **pages = stack_pages;
+ unsigned long i = 0;
+ pgprot_t pgprot;
+ void *addr;
+
+ /* A single page can always be kmapped */
+ if (n_pages == 1 && type == I915_MAP_WB)
+ return kmap(sg_page(sgt->sgl));
+
+ if (n_pages > ARRAY_SIZE(stack_pages)) {
+ /* Too big for stack -- allocate temporary array instead */
+ pages = kvmalloc_array(n_pages, sizeof(*pages), GFP_KERNEL);
+ if (!pages)
+ return NULL;
+ }
+
+ for_each_sgt_page(page, sgt_iter, sgt)
+ pages[i++] = page;
+
+ /* Check that we have the expected number of pages */
+ GEM_BUG_ON(i != n_pages);
+
+ switch (type) {
+ default:
+ MISSING_CASE(type);
+ /* fallthrough to use PAGE_KERNEL anyway */
+ case I915_MAP_WB:
+ pgprot = PAGE_KERNEL;
+ break;
+ case I915_MAP_WC:
+ pgprot = pgprot_writecombine(PAGE_KERNEL_IO);
+ break;
+ }
+ addr = vmap(pages, n_pages, 0, pgprot);
+
+ if (pages != stack_pages)
+ kvfree(pages);
+
+ return addr;
+}
+
+/* get, pin, and map the pages of the object into kernel space */
+void *i915_gem_object_pin_map(struct drm_i915_gem_object *obj,
+ enum i915_map_type type)
+{
+ enum i915_map_type has_type;
+ bool pinned;
+ void *ptr;
+ int err;
+
+ if (unlikely(!i915_gem_object_has_struct_page(obj)))
+ return ERR_PTR(-ENXIO);
+
+ err = mutex_lock_interruptible(&obj->mm.lock);
+ if (err)
+ return ERR_PTR(err);
+
+ pinned = !(type & I915_MAP_OVERRIDE);
+ type &= ~I915_MAP_OVERRIDE;
+
+ if (!atomic_inc_not_zero(&obj->mm.pages_pin_count)) {
+ if (unlikely(!i915_gem_object_has_pages(obj))) {
+ GEM_BUG_ON(i915_gem_object_has_pinned_pages(obj));
+
+ err = ____i915_gem_object_get_pages(obj);
+ if (err)
+ goto err_unlock;
+
+ smp_mb__before_atomic();
+ }
+ atomic_inc(&obj->mm.pages_pin_count);
+ pinned = false;
+ }
+ GEM_BUG_ON(!i915_gem_object_has_pages(obj));
+
+ ptr = page_unpack_bits(obj->mm.mapping, &has_type);
+ if (ptr && has_type != type) {
+ if (pinned) {
+ err = -EBUSY;
+ goto err_unpin;
+ }
+
+ if (is_vmalloc_addr(ptr))
+ vunmap(ptr);
+ else
+ kunmap(kmap_to_page(ptr));
+
+ ptr = obj->mm.mapping = NULL;
+ }
+
+ if (!ptr) {
+ ptr = i915_gem_object_map(obj, type);
+ if (!ptr) {
+ err = -ENOMEM;
+ goto err_unpin;
+ }
+
+ obj->mm.mapping = page_pack_bits(ptr, type);
+ }
+
+out_unlock:
+ mutex_unlock(&obj->mm.lock);
+ return ptr;
+
+err_unpin:
+ atomic_dec(&obj->mm.pages_pin_count);
+err_unlock:
+ ptr = ERR_PTR(err);
+ goto out_unlock;
+}
+
+void __i915_gem_object_flush_map(struct drm_i915_gem_object *obj,
+ unsigned long offset,
+ unsigned long size)
+{
+ enum i915_map_type has_type;
+ void *ptr;
+
+ GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj));
+ GEM_BUG_ON(range_overflows_t(typeof(obj->base.size),
+ offset, size, obj->base.size));
+
+ obj->mm.dirty = true;
+
+ if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE)
+ return;
+
+ ptr = page_unpack_bits(obj->mm.mapping, &has_type);
+ if (has_type == I915_MAP_WC)
+ return;
+
+ drm_clflush_virt_range(ptr + offset, size);
+ if (size == obj->base.size) {
+ obj->write_domain &= ~I915_GEM_DOMAIN_CPU;
+ obj->cache_dirty = false;
+ }
+}
+
+struct scatterlist *
+i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
+ unsigned int n,
+ unsigned int *offset)
+{
+ struct i915_gem_object_page_iter *iter = &obj->mm.get_page;
+ struct scatterlist *sg;
+ unsigned int idx, count;
+
+ might_sleep();
+ GEM_BUG_ON(n >= obj->base.size >> PAGE_SHIFT);
+ GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj));
+
+ /* As we iterate forward through the sg, we record each entry in a
+ * radixtree for quick repeated (backwards) lookups. If we have seen
+ * this index previously, we will have an entry for it.
+ *
+ * Initial lookup is O(N), but this is amortized to O(1) for
+ * sequential page access (where each new request is consecutive
+ * to the previous one). Repeated lookups are O(lg(obj->base.size)),
+ * i.e. O(1) with a large constant!
+ */
+ if (n < READ_ONCE(iter->sg_idx))
+ goto lookup;
+
+ mutex_lock(&iter->lock);
+
+ /* We prefer to reuse the last sg so that repeated lookup of this
+ * (or the subsequent) sg are fast - comparing against the last
+ * sg is faster than going through the radixtree.
+ */
+
+ sg = iter->sg_pos;
+ idx = iter->sg_idx;
+ count = __sg_page_count(sg);
+
+ while (idx + count <= n) {
+ void *entry;
+ unsigned long i;
+ int ret;
+
+ /* If we cannot allocate and insert this entry, or the
+ * individual pages from this range, cancel updating the
+ * sg_idx so that on this lookup we are forced to linearly
+ * scan onwards, but on future lookups we will try the
+ * insertion again (in which case we need to be careful of
+ * the error return reporting that we have already inserted
+ * this index).
+ */
+ ret = radix_tree_insert(&iter->radix, idx, sg);
+ if (ret && ret != -EEXIST)
+ goto scan;
+
+ entry = xa_mk_value(idx);
+ for (i = 1; i < count; i++) {
+ ret = radix_tree_insert(&iter->radix, idx + i, entry);
+ if (ret && ret != -EEXIST)
+ goto scan;
+ }
+
+ idx += count;
+ sg = ____sg_next(sg);
+ count = __sg_page_count(sg);
+ }
+
+scan:
+ iter->sg_pos = sg;
+ iter->sg_idx = idx;
+
+ mutex_unlock(&iter->lock);
+
+ if (unlikely(n < idx)) /* insertion completed by another thread */
+ goto lookup;
+
+ /* In case we failed to insert the entry into the radixtree, we need
+ * to look beyond the current sg.
+ */
+ while (idx + count <= n) {
+ idx += count;
+ sg = ____sg_next(sg);
+ count = __sg_page_count(sg);
+ }
+
+ *offset = n - idx;
+ return sg;
+
+lookup:
+ rcu_read_lock();
+
+ sg = radix_tree_lookup(&iter->radix, n);
+ GEM_BUG_ON(!sg);
+
+ /* If this index is in the middle of multi-page sg entry,
+ * the radix tree will contain a value entry that points
+ * to the start of that range. We will return the pointer to
+ * the base page and the offset of this page within the
+ * sg entry's range.
+ */
+ *offset = 0;
+ if (unlikely(xa_is_value(sg))) {
+ unsigned long base = xa_to_value(sg);
+
+ sg = radix_tree_lookup(&iter->radix, base);
+ GEM_BUG_ON(!sg);
+
+ *offset = n - base;
+ }
+
+ rcu_read_unlock();
+
+ return sg;
+}
+
+struct page *
+i915_gem_object_get_page(struct drm_i915_gem_object *obj, unsigned int n)
+{
+ struct scatterlist *sg;
+ unsigned int offset;
+
+ GEM_BUG_ON(!i915_gem_object_has_struct_page(obj));
+
+ sg = i915_gem_object_get_sg(obj, n, &offset);
+ return nth_page(sg_page(sg), offset);
+}
+
+/* Like i915_gem_object_get_page(), but mark the returned page dirty */
+struct page *
+i915_gem_object_get_dirty_page(struct drm_i915_gem_object *obj,
+ unsigned int n)
+{
+ struct page *page;
+
+ page = i915_gem_object_get_page(obj, n);
+ if (!obj->mm.dirty)
+ set_page_dirty(page);
+
+ return page;
+}
+
+dma_addr_t
+i915_gem_object_get_dma_address(struct drm_i915_gem_object *obj,
+ unsigned long n)
+{
+ struct scatterlist *sg;
+ unsigned int offset;
+
+ sg = i915_gem_object_get_sg(obj, n, &offset);
+ return sg_dma_address(sg) + (offset << PAGE_SHIFT);
+}
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_phys.c b/drivers/gpu/drm/i915/gem/i915_gem_phys.c
new file mode 100644
index 000000000000..1bf3e0afcba2
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/i915_gem_phys.c
@@ -0,0 +1,212 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2014-2016 Intel Corporation
+ */
+
+#include <linux/highmem.h>
+#include <linux/shmem_fs.h>
+#include <linux/swap.h>
+
+#include <drm/drm.h> /* for drm_legacy.h! */
+#include <drm/drm_cache.h>
+#include <drm/drm_legacy.h> /* for drm_pci.h! */
+#include <drm/drm_pci.h>
+
+#include "i915_gem_object.h"
+
+#include "../i915_drv.h"
+
+static int i915_gem_object_get_pages_phys(struct drm_i915_gem_object *obj)
+{
+ struct address_space *mapping = obj->base.filp->f_mapping;
+ struct drm_dma_handle *phys;
+ struct sg_table *st;
+ struct scatterlist *sg;
+ char *vaddr;
+ int i;
+ int err;
+
+ if (WARN_ON(i915_gem_object_needs_bit17_swizzle(obj)))
+ return -EINVAL;
+
+ /* Always aligning to the object size, allows a single allocation
+ * to handle all possible callers, and given typical object sizes,
+ * the alignment of the buddy allocation will naturally match.
+ */
+ phys = drm_pci_alloc(obj->base.dev,
+ roundup_pow_of_two(obj->base.size),
+ roundup_pow_of_two(obj->base.size));
+ if (!phys)
+ return -ENOMEM;
+
+ vaddr = phys->vaddr;
+ for (i = 0; i < obj->base.size / PAGE_SIZE; i++) {
+ struct page *page;
+ char *src;
+
+ page = shmem_read_mapping_page(mapping, i);
+ if (IS_ERR(page)) {
+ err = PTR_ERR(page);
+ goto err_phys;
+ }
+
+ src = kmap_atomic(page);
+ memcpy(vaddr, src, PAGE_SIZE);
+ drm_clflush_virt_range(vaddr, PAGE_SIZE);
+ kunmap_atomic(src);
+
+ put_page(page);
+ vaddr += PAGE_SIZE;
+ }
+
+ i915_gem_chipset_flush(to_i915(obj->base.dev));
+
+ st = kmalloc(sizeof(*st), GFP_KERNEL);
+ if (!st) {
+ err = -ENOMEM;
+ goto err_phys;
+ }
+
+ if (sg_alloc_table(st, 1, GFP_KERNEL)) {
+ kfree(st);
+ err = -ENOMEM;
+ goto err_phys;
+ }
+
+ sg = st->sgl;
+ sg->offset = 0;
+ sg->length = obj->base.size;
+
+ sg_dma_address(sg) = phys->busaddr;
+ sg_dma_len(sg) = obj->base.size;
+
+ obj->phys_handle = phys;
+
+ __i915_gem_object_set_pages(obj, st, sg->length);
+
+ return 0;
+
+err_phys:
+ drm_pci_free(obj->base.dev, phys);
+
+ return err;
+}
+
+static void
+i915_gem_object_put_pages_phys(struct drm_i915_gem_object *obj,
+ struct sg_table *pages)
+{
+ __i915_gem_object_release_shmem(obj, pages, false);
+
+ if (obj->mm.dirty) {
+ struct address_space *mapping = obj->base.filp->f_mapping;
+ char *vaddr = obj->phys_handle->vaddr;
+ int i;
+
+ for (i = 0; i < obj->base.size / PAGE_SIZE; i++) {
+ struct page *page;
+ char *dst;
+
+ page = shmem_read_mapping_page(mapping, i);
+ if (IS_ERR(page))
+ continue;
+
+ dst = kmap_atomic(page);
+ drm_clflush_virt_range(vaddr, PAGE_SIZE);
+ memcpy(dst, vaddr, PAGE_SIZE);
+ kunmap_atomic(dst);
+
+ set_page_dirty(page);
+ if (obj->mm.madv == I915_MADV_WILLNEED)
+ mark_page_accessed(page);
+ put_page(page);
+ vaddr += PAGE_SIZE;
+ }
+ obj->mm.dirty = false;
+ }
+
+ sg_free_table(pages);
+ kfree(pages);
+
+ drm_pci_free(obj->base.dev, obj->phys_handle);
+}
+
+static void
+i915_gem_object_release_phys(struct drm_i915_gem_object *obj)
+{
+ i915_gem_object_unpin_pages(obj);
+}
+
+static const struct drm_i915_gem_object_ops i915_gem_phys_ops = {
+ .get_pages = i915_gem_object_get_pages_phys,
+ .put_pages = i915_gem_object_put_pages_phys,
+ .release = i915_gem_object_release_phys,
+};
+
+int i915_gem_object_attach_phys(struct drm_i915_gem_object *obj, int align)
+{
+ struct sg_table *pages;
+ int err;
+
+ if (align > obj->base.size)
+ return -EINVAL;
+
+ if (obj->ops == &i915_gem_phys_ops)
+ return 0;
+
+ if (obj->ops != &i915_gem_shmem_ops)
+ return -EINVAL;
+
+ err = i915_gem_object_unbind(obj);
+ if (err)
+ return err;
+
+ mutex_lock(&obj->mm.lock);
+
+ if (obj->mm.madv != I915_MADV_WILLNEED) {
+ err = -EFAULT;
+ goto err_unlock;
+ }
+
+ if (obj->mm.quirked) {
+ err = -EFAULT;
+ goto err_unlock;
+ }
+
+ if (obj->mm.mapping) {
+ err = -EBUSY;
+ goto err_unlock;
+ }
+
+ pages = __i915_gem_object_unset_pages(obj);
+
+ obj->ops = &i915_gem_phys_ops;
+
+ err = ____i915_gem_object_get_pages(obj);
+ if (err)
+ goto err_xfer;
+
+ /* Perma-pin (until release) the physical set of pages */
+ __i915_gem_object_pin_pages(obj);
+
+ if (!IS_ERR_OR_NULL(pages))
+ i915_gem_shmem_ops.put_pages(obj, pages);
+ mutex_unlock(&obj->mm.lock);
+ return 0;
+
+err_xfer:
+ obj->ops = &i915_gem_shmem_ops;
+ if (!IS_ERR_OR_NULL(pages)) {
+ unsigned int sg_page_sizes = i915_sg_page_sizes(pages->sgl);
+
+ __i915_gem_object_set_pages(obj, pages, sg_page_sizes);
+ }
+err_unlock:
+ mutex_unlock(&obj->mm.lock);
+ return err;
+}
+
+#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
+#include "selftests/i915_gem_phys.c"
+#endif
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
index 9d885b5e6643..78df86d1db7a 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
@@ -214,6 +214,65 @@ static int shmem_get_pages(struct drm_i915_gem_object *obj)
return ret;
}
+static void
+shmem_truncate(struct drm_i915_gem_object *obj)
+{
+ /*
+ * Our goal here is to return as much of the memory as
+ * is possible back to the system as we are called from OOM.
+ * To do this we must instruct the shmfs to drop all of its
+ * backing pages, *now*.
+ */
+ shmem_truncate_range(file_inode(obj->base.filp), 0, (loff_t)-1);
+ obj->mm.madv = __I915_MADV_PURGED;
+ obj->mm.pages = ERR_PTR(-EFAULT);
+}
+
+static void
+shmem_writeback(struct drm_i915_gem_object *obj)
+{
+ struct address_space *mapping;
+ struct writeback_control wbc = {
+ .sync_mode = WB_SYNC_NONE,
+ .nr_to_write = SWAP_CLUSTER_MAX,
+ .range_start = 0,
+ .range_end = LLONG_MAX,
+ .for_reclaim = 1,
+ };
+ unsigned long i;
+
+ /*
+ * Leave mmapings intact (GTT will have been revoked on unbinding,
+ * leaving only CPU mmapings around) and add those pages to the LRU
+ * instead of invoking writeback so they are aged and paged out
+ * as normal.
+ */
+ mapping = obj->base.filp->f_mapping;
+
+ /* Begin writeback on each dirty page */
+ for (i = 0; i < obj->base.size >> PAGE_SHIFT; i++) {
+ struct page *page;
+
+ page = find_lock_entry(mapping, i);
+ if (!page || xa_is_value(page))
+ continue;
+
+ if (!page_mapped(page) && clear_page_dirty_for_io(page)) {
+ int ret;
+
+ SetPageReclaim(page);
+ ret = mapping->a_ops->writepage(page, &wbc);
+ if (!PageWriteback(page))
+ ClearPageReclaim(page);
+ if (!ret)
+ goto put;
+ }
+ unlock_page(page);
+put:
+ put_page(page);
+ }
+}
+
void
__i915_gem_object_release_shmem(struct drm_i915_gem_object *obj,
struct sg_table *pages,
@@ -363,6 +422,8 @@ const struct drm_i915_gem_object_ops i915_gem_shmem_ops = {
.get_pages = shmem_get_pages,
.put_pages = shmem_put_pages,
+ .truncate = shmem_truncate,
+ .writeback = shmem_writeback,
.pwrite = shmem_pwrite,
};
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c
new file mode 100644
index 000000000000..b76b503b3999
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c
@@ -0,0 +1,80 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2016 Intel Corporation
+ */
+
+#include "../../i915_selftest.h"
+
+#include "../../selftests/mock_gem_device.h"
+
+static int mock_phys_object(void *arg)
+{
+ struct drm_i915_private *i915 = arg;
+ struct drm_i915_gem_object *obj;
+ int err;
+
+ /* Create an object and bind it to a contiguous set of physical pages,
+ * i.e. exercise the i915_gem_object_phys API.
+ */
+
+ obj = i915_gem_object_create_shmem(i915, PAGE_SIZE);
+ if (IS_ERR(obj)) {
+ err = PTR_ERR(obj);
+ pr_err("i915_gem_object_create failed, err=%d\n", err);
+ goto out;
+ }
+
+ mutex_lock(&i915->drm.struct_mutex);
+ err = i915_gem_object_attach_phys(obj, PAGE_SIZE);
+ mutex_unlock(&i915->drm.struct_mutex);
+ if (err) {
+ pr_err("i915_gem_object_attach_phys failed, err=%d\n", err);
+ goto out_obj;
+ }
+
+ if (obj->ops != &i915_gem_phys_ops) {
+ pr_err("i915_gem_object_attach_phys did not create a phys object\n");
+ err = -EINVAL;
+ goto out_obj;
+ }
+
+ if (!atomic_read(&obj->mm.pages_pin_count)) {
+ pr_err("i915_gem_object_attach_phys did not pin its phys pages\n");
+ err = -EINVAL;
+ goto out_obj;
+ }
+
+ /* Make the object dirty so that put_pages must do copy back the data */
+ mutex_lock(&i915->drm.struct_mutex);
+ err = i915_gem_object_set_to_gtt_domain(obj, true);
+ mutex_unlock(&i915->drm.struct_mutex);
+ if (err) {
+ pr_err("i915_gem_object_set_to_gtt_domain failed with err=%d\n",
+ err);
+ goto out_obj;
+ }
+
+out_obj:
+ i915_gem_object_put(obj);
+out:
+ return err;
+}
+
+int i915_gem_phys_mock_selftests(void)
+{
+ static const struct i915_subtest tests[] = {
+ SUBTEST(mock_phys_object),
+ };
+ struct drm_i915_private *i915;
+ int err;
+
+ i915 = mock_gem_device();
+ if (!i915)
+ return -ENOMEM;
+
+ err = i915_subtests(tests, i915);
+
+ drm_dev_put(&i915->drm);
+ return err;
+}
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 6de9b641751d..da9f5a375613 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2887,8 +2887,6 @@ i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,
const struct i915_ggtt_view *view,
unsigned int flags);
void i915_gem_object_unpin_from_display_plane(struct i915_vma *vma);
-int i915_gem_object_attach_phys(struct drm_i915_gem_object *obj,
- int align);
int i915_gem_open(struct drm_i915_private *i915, struct drm_file *file);
void i915_gem_release(struct drm_device *dev, struct drm_file *file);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 4c5fba0c7bf6..0549b73f01ff 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -26,7 +26,6 @@
*/
#include <drm/drm_vma_manager.h>
-#include <drm/drm_pci.h>
#include <drm/i915_drm.h>
#include <linux/dma-fence-array.h>
#include <linux/kthread.h>
@@ -98,133 +97,6 @@ i915_gem_get_aperture_ioctl(struct drm_device *dev, void *data,
return 0;
}
-static int i915_gem_object_get_pages_phys(struct drm_i915_gem_object *obj)
-{
- struct address_space *mapping = obj->base.filp->f_mapping;
- drm_dma_handle_t *phys;
- struct sg_table *st;
- struct scatterlist *sg;
- char *vaddr;
- int i;
- int err;
-
- if (WARN_ON(i915_gem_object_needs_bit17_swizzle(obj)))
- return -EINVAL;
-
- /* Always aligning to the object size, allows a single allocation
- * to handle all possible callers, and given typical object sizes,
- * the alignment of the buddy allocation will naturally match.
- */
- phys = drm_pci_alloc(obj->base.dev,
- roundup_pow_of_two(obj->base.size),
- roundup_pow_of_two(obj->base.size));
- if (!phys)
- return -ENOMEM;
-
- vaddr = phys->vaddr;
- for (i = 0; i < obj->base.size / PAGE_SIZE; i++) {
- struct page *page;
- char *src;
-
- page = shmem_read_mapping_page(mapping, i);
- if (IS_ERR(page)) {
- err = PTR_ERR(page);
- goto err_phys;
- }
-
- src = kmap_atomic(page);
- memcpy(vaddr, src, PAGE_SIZE);
- drm_clflush_virt_range(vaddr, PAGE_SIZE);
- kunmap_atomic(src);
-
- put_page(page);
- vaddr += PAGE_SIZE;
- }
-
- i915_gem_chipset_flush(to_i915(obj->base.dev));
-
- st = kmalloc(sizeof(*st), GFP_KERNEL);
- if (!st) {
- err = -ENOMEM;
- goto err_phys;
- }
-
- if (sg_alloc_table(st, 1, GFP_KERNEL)) {
- kfree(st);
- err = -ENOMEM;
- goto err_phys;
- }
-
- sg = st->sgl;
- sg->offset = 0;
- sg->length = obj->base.size;
-
- sg_dma_address(sg) = phys->busaddr;
- sg_dma_len(sg) = obj->base.size;
-
- obj->phys_handle = phys;
-
- __i915_gem_object_set_pages(obj, st, sg->length);
-
- return 0;
-
-err_phys:
- drm_pci_free(obj->base.dev, phys);
-
- return err;
-}
-
-static void
-i915_gem_object_put_pages_phys(struct drm_i915_gem_object *obj,
- struct sg_table *pages)
-{
- __i915_gem_object_release_shmem(obj, pages, false);
-
- if (obj->mm.dirty) {
- struct address_space *mapping = obj->base.filp->f_mapping;
- char *vaddr = obj->phys_handle->vaddr;
- int i;
-
- for (i = 0; i < obj->base.size / PAGE_SIZE; i++) {
- struct page *page;
- char *dst;
-
- page = shmem_read_mapping_page(mapping, i);
- if (IS_ERR(page))
- continue;
-
- dst = kmap_atomic(page);
- drm_clflush_virt_range(vaddr, PAGE_SIZE);
- memcpy(dst, vaddr, PAGE_SIZE);
- kunmap_atomic(dst);
-
- set_page_dirty(page);
- if (obj->mm.madv == I915_MADV_WILLNEED)
- mark_page_accessed(page);
- put_page(page);
- vaddr += PAGE_SIZE;
- }
- obj->mm.dirty = false;
- }
-
- sg_free_table(pages);
- kfree(pages);
-
- drm_pci_free(obj->base.dev, obj->phys_handle);
-}
-
-static void
-i915_gem_object_release_phys(struct drm_i915_gem_object *obj)
-{
- i915_gem_object_unpin_pages(obj);
-}
-
-static const struct drm_i915_gem_object_ops i915_gem_phys_ops = {
- .get_pages = i915_gem_object_get_pages_phys,
- .put_pages = i915_gem_object_put_pages_phys,
- .release = i915_gem_object_release_phys,
-};
-
int i915_gem_object_unbind(struct drm_i915_gem_object *obj)
{
struct i915_vma *vma;
@@ -1937,11 +1809,6 @@ static int i915_gem_object_create_mmap_offset(struct drm_i915_gem_object *obj)
return err;
}
-static void i915_gem_object_free_mmap_offset(struct drm_i915_gem_object *obj)
-{
- drm_gem_free_mmap_offset(&obj->base);
-}
-
int
i915_gem_mmap_gtt(struct drm_file *file,
struct drm_device *dev,
@@ -1987,111 +1854,6 @@ i915_gem_mmap_gtt_ioctl(struct drm_device *dev, void *data,
return i915_gem_mmap_gtt(file, dev, args->handle, &args->offset);
}
-/* Immediately discard the backing storage */
-void __i915_gem_object_truncate(struct drm_i915_gem_object *obj)
-{
- i915_gem_object_free_mmap_offset(obj);
-
- if (obj->base.filp == NULL)
- return;
-
- /* Our goal here is to return as much of the memory as
- * is possible back to the system as we are called from OOM.
- * To do this we must instruct the shmfs to drop all of its
- * backing pages, *now*.
- */
- shmem_truncate_range(file_inode(obj->base.filp), 0, (loff_t)-1);
- obj->mm.madv = __I915_MADV_PURGED;
- obj->mm.pages = ERR_PTR(-EFAULT);
-}
-
-static void __i915_gem_object_reset_page_iter(struct drm_i915_gem_object *obj)
-{
- struct radix_tree_iter iter;
- void __rcu **slot;
-
- rcu_read_lock();
- radix_tree_for_each_slot(slot, &obj->mm.get_page.radix, &iter, 0)
- radix_tree_delete(&obj->mm.get_page.radix, iter.index);
- rcu_read_unlock();
-}
-
-static struct sg_table *
-__i915_gem_object_unset_pages(struct drm_i915_gem_object *obj)
-{
- struct drm_i915_private *i915 = to_i915(obj->base.dev);
- struct sg_table *pages;
-
- pages = fetch_and_zero(&obj->mm.pages);
- if (IS_ERR_OR_NULL(pages))
- return pages;
-
- spin_lock(&i915->mm.obj_lock);
- list_del(&obj->mm.link);
- spin_unlock(&i915->mm.obj_lock);
-
- if (obj->mm.mapping) {
- void *ptr;
-
- ptr = page_mask_bits(obj->mm.mapping);
- if (is_vmalloc_addr(ptr))
- vunmap(ptr);
- else
- kunmap(kmap_to_page(ptr));
-
- obj->mm.mapping = NULL;
- }
-
- __i915_gem_object_reset_page_iter(obj);
- obj->mm.page_sizes.phys = obj->mm.page_sizes.sg = 0;
-
- return pages;
-}
-
-int __i915_gem_object_put_pages(struct drm_i915_gem_object *obj,
- enum i915_mm_subclass subclass)
-{
- struct sg_table *pages;
- int ret;
-
- if (i915_gem_object_has_pinned_pages(obj))
- return -EBUSY;
-
- GEM_BUG_ON(obj->bind_count);
-
- /* May be called by shrinker from within get_pages() (on another bo) */
- mutex_lock_nested(&obj->mm.lock, subclass);
- if (unlikely(atomic_read(&obj->mm.pages_pin_count))) {
- ret = -EBUSY;
- goto unlock;
- }
-
- /*
- * ->put_pages might need to allocate memory for the bit17 swizzle
- * array, hence protect them from being reaped by removing them from gtt
- * lists early.
- */
- pages = __i915_gem_object_unset_pages(obj);
-
- /*
- * XXX Temporary hijinx to avoid updating all backends to handle
- * NULL pages. In the future, when we have more asynchronous
- * get_pages backends we should be better able to handle the
- * cancellation of the async task in a more uniform manner.
- */
- if (!pages && !i915_gem_object_needs_async_cancel(obj))
- pages = ERR_PTR(-EINVAL);
-
- if (!IS_ERR(pages))
- obj->ops->put_pages(obj, pages);
-
- ret = 0;
-unlock:
- mutex_unlock(&obj->mm.lock);
-
- return ret;
-}
-
bool i915_sg_trim(struct sg_table *orig_st)
{
struct sg_table new_st;
@@ -2120,252 +1882,6 @@ bool i915_sg_trim(struct sg_table *orig_st)
return true;
}
-void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj,
- struct sg_table *pages,
- unsigned int sg_page_sizes)
-{
- struct drm_i915_private *i915 = to_i915(obj->base.dev);
- unsigned long supported = INTEL_INFO(i915)->page_sizes;
- int i;
-
- lockdep_assert_held(&obj->mm.lock);
-
- /* Make the pages coherent with the GPU (flushing any swapin). */
- if (obj->cache_dirty) {
- obj->write_domain = 0;
- if (i915_gem_object_has_struct_page(obj))
- drm_clflush_sg(pages);
- obj->cache_dirty = false;
- }
-
- obj->mm.get_page.sg_pos = pages->sgl;
- obj->mm.get_page.sg_idx = 0;
-
- obj->mm.pages = pages;
-
- if (i915_gem_object_is_tiled(obj) &&
- i915->quirks & QUIRK_PIN_SWIZZLED_PAGES) {
- GEM_BUG_ON(obj->mm.quirked);
- __i915_gem_object_pin_pages(obj);
- obj->mm.quirked = true;
- }
-
- GEM_BUG_ON(!sg_page_sizes);
- obj->mm.page_sizes.phys = sg_page_sizes;
-
- /*
- * Calculate the supported page-sizes which fit into the given
- * sg_page_sizes. This will give us the page-sizes which we may be able
- * to use opportunistically when later inserting into the GTT. For
- * example if phys=2G, then in theory we should be able to use 1G, 2M,
- * 64K or 4K pages, although in practice this will depend on a number of
- * other factors.
- */
- obj->mm.page_sizes.sg = 0;
- for_each_set_bit(i, &supported, ilog2(I915_GTT_MAX_PAGE_SIZE) + 1) {
- if (obj->mm.page_sizes.phys & ~0u << i)
- obj->mm.page_sizes.sg |= BIT(i);
- }
- GEM_BUG_ON(!HAS_PAGE_SIZES(i915, obj->mm.page_sizes.sg));
-
- spin_lock(&i915->mm.obj_lock);
- list_add(&obj->mm.link, &i915->mm.unbound_list);
- spin_unlock(&i915->mm.obj_lock);
-}
-
-static int ____i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
-{
- int err;
-
- if (unlikely(obj->mm.madv != I915_MADV_WILLNEED)) {
- DRM_DEBUG("Attempting to obtain a purgeable object\n");
- return -EFAULT;
- }
-
- err = obj->ops->get_pages(obj);
- GEM_BUG_ON(!err && !i915_gem_object_has_pages(obj));
-
- return err;
-}
-
-/* Ensure that the associated pages are gathered from the backing storage
- * and pinned into our object. i915_gem_object_pin_pages() may be called
- * multiple times before they are released by a single call to
- * i915_gem_object_unpin_pages() - once the pages are no longer referenced
- * either as a result of memory pressure (reaping pages under the shrinker)
- * or as the object is itself released.
- */
-int __i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
-{
- int err;
-
- err = mutex_lock_interruptible(&obj->mm.lock);
- if (err)
- return err;
-
- if (unlikely(!i915_gem_object_has_pages(obj))) {
- GEM_BUG_ON(i915_gem_object_has_pinned_pages(obj));
-
- err = ____i915_gem_object_get_pages(obj);
- if (err)
- goto unlock;
-
- smp_mb__before_atomic();
- }
- atomic_inc(&obj->mm.pages_pin_count);
-
-unlock:
- mutex_unlock(&obj->mm.lock);
- return err;
-}
-
-/* The 'mapping' part of i915_gem_object_pin_map() below */
-static void *i915_gem_object_map(const struct drm_i915_gem_object *obj,
- enum i915_map_type type)
-{
- unsigned long n_pages = obj->base.size >> PAGE_SHIFT;
- struct sg_table *sgt = obj->mm.pages;
- struct sgt_iter sgt_iter;
- struct page *page;
- struct page *stack_pages[32];
- struct page **pages = stack_pages;
- unsigned long i = 0;
- pgprot_t pgprot;
- void *addr;
-
- /* A single page can always be kmapped */
- if (n_pages == 1 && type == I915_MAP_WB)
- return kmap(sg_page(sgt->sgl));
-
- if (n_pages > ARRAY_SIZE(stack_pages)) {
- /* Too big for stack -- allocate temporary array instead */
- pages = kvmalloc_array(n_pages, sizeof(*pages), GFP_KERNEL);
- if (!pages)
- return NULL;
- }
-
- for_each_sgt_page(page, sgt_iter, sgt)
- pages[i++] = page;
-
- /* Check that we have the expected number of pages */
- GEM_BUG_ON(i != n_pages);
-
- switch (type) {
- default:
- MISSING_CASE(type);
- /* fallthrough to use PAGE_KERNEL anyway */
- case I915_MAP_WB:
- pgprot = PAGE_KERNEL;
- break;
- case I915_MAP_WC:
- pgprot = pgprot_writecombine(PAGE_KERNEL_IO);
- break;
- }
- addr = vmap(pages, n_pages, 0, pgprot);
-
- if (pages != stack_pages)
- kvfree(pages);
-
- return addr;
-}
-
-/* get, pin, and map the pages of the object into kernel space */
-void *i915_gem_object_pin_map(struct drm_i915_gem_object *obj,
- enum i915_map_type type)
-{
- enum i915_map_type has_type;
- bool pinned;
- void *ptr;
- int ret;
-
- if (unlikely(!i915_gem_object_has_struct_page(obj)))
- return ERR_PTR(-ENXIO);
-
- ret = mutex_lock_interruptible(&obj->mm.lock);
- if (ret)
- return ERR_PTR(ret);
-
- pinned = !(type & I915_MAP_OVERRIDE);
- type &= ~I915_MAP_OVERRIDE;
-
- if (!atomic_inc_not_zero(&obj->mm.pages_pin_count)) {
- if (unlikely(!i915_gem_object_has_pages(obj))) {
- GEM_BUG_ON(i915_gem_object_has_pinned_pages(obj));
-
- ret = ____i915_gem_object_get_pages(obj);
- if (ret)
- goto err_unlock;
-
- smp_mb__before_atomic();
- }
- atomic_inc(&obj->mm.pages_pin_count);
- pinned = false;
- }
- GEM_BUG_ON(!i915_gem_object_has_pages(obj));
-
- ptr = page_unpack_bits(obj->mm.mapping, &has_type);
- if (ptr && has_type != type) {
- if (pinned) {
- ret = -EBUSY;
- goto err_unpin;
- }
-
- if (is_vmalloc_addr(ptr))
- vunmap(ptr);
- else
- kunmap(kmap_to_page(ptr));
-
- ptr = obj->mm.mapping = NULL;
- }
-
- if (!ptr) {
- ptr = i915_gem_object_map(obj, type);
- if (!ptr) {
- ret = -ENOMEM;
- goto err_unpin;
- }
-
- obj->mm.mapping = page_pack_bits(ptr, type);
- }
-
-out_unlock:
- mutex_unlock(&obj->mm.lock);
- return ptr;
-
-err_unpin:
- atomic_dec(&obj->mm.pages_pin_count);
-err_unlock:
- ptr = ERR_PTR(ret);
- goto out_unlock;
-}
-
-void __i915_gem_object_flush_map(struct drm_i915_gem_object *obj,
- unsigned long offset,
- unsigned long size)
-{
- enum i915_map_type has_type;
- void *ptr;
-
- GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj));
- GEM_BUG_ON(range_overflows_t(typeof(obj->base.size),
- offset, size, obj->base.size));
-
- obj->mm.dirty = true;
-
- if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE)
- return;
-
- ptr = page_unpack_bits(obj->mm.mapping, &has_type);
- if (has_type == I915_MAP_WC)
- return;
-
- drm_clflush_virt_range(ptr + offset, size);
- if (size == obj->base.size) {
- obj->write_domain &= ~I915_GEM_DOMAIN_CPU;
- obj->cache_dirty = false;
- }
-}
-
static unsigned long to_wait_timeout(s64 timeout_ns)
{
if (timeout_ns < 0)
@@ -3354,7 +2870,7 @@ i915_gem_madvise_ioctl(struct drm_device *dev, void *data,
/* if the object is no longer attached, discard its backing storage */
if (obj->mm.madv == I915_MADV_DONTNEED &&
!i915_gem_object_has_pages(obj))
- __i915_gem_object_truncate(obj);
+ i915_gem_object_truncate(obj);
args->retained = obj->mm.madv != __I915_MADV_PURGED;
mutex_unlock(&obj->mm.lock);
@@ -4126,219 +3642,6 @@ void i915_gem_track_fb(struct drm_i915_gem_object *old,
}
}
-struct scatterlist *
-i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
- unsigned int n,
- unsigned int *offset)
-{
- struct i915_gem_object_page_iter *iter = &obj->mm.get_page;
- struct scatterlist *sg;
- unsigned int idx, count;
-
- might_sleep();
- GEM_BUG_ON(n >= obj->base.size >> PAGE_SHIFT);
- GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj));
-
- /* As we iterate forward through the sg, we record each entry in a
- * radixtree for quick repeated (backwards) lookups. If we have seen
- * this index previously, we will have an entry for it.
- *
- * Initial lookup is O(N), but this is amortized to O(1) for
- * sequential page access (where each new request is consecutive
- * to the previous one). Repeated lookups are O(lg(obj->base.size)),
- * i.e. O(1) with a large constant!
- */
- if (n < READ_ONCE(iter->sg_idx))
- goto lookup;
-
- mutex_lock(&iter->lock);
-
- /* We prefer to reuse the last sg so that repeated lookup of this
- * (or the subsequent) sg are fast - comparing against the last
- * sg is faster than going through the radixtree.
- */
-
- sg = iter->sg_pos;
- idx = iter->sg_idx;
- count = __sg_page_count(sg);
-
- while (idx + count <= n) {
- void *entry;
- unsigned long i;
- int ret;
-
- /* If we cannot allocate and insert this entry, or the
- * individual pages from this range, cancel updating the
- * sg_idx so that on this lookup we are forced to linearly
- * scan onwards, but on future lookups we will try the
- * insertion again (in which case we need to be careful of
- * the error return reporting that we have already inserted
- * this index).
- */
- ret = radix_tree_insert(&iter->radix, idx, sg);
- if (ret && ret != -EEXIST)
- goto scan;
-
- entry = xa_mk_value(idx);
- for (i = 1; i < count; i++) {
- ret = radix_tree_insert(&iter->radix, idx + i, entry);
- if (ret && ret != -EEXIST)
- goto scan;
- }
-
- idx += count;
- sg = ____sg_next(sg);
- count = __sg_page_count(sg);
- }
-
-scan:
- iter->sg_pos = sg;
- iter->sg_idx = idx;
-
- mutex_unlock(&iter->lock);
-
- if (unlikely(n < idx)) /* insertion completed by another thread */
- goto lookup;
-
- /* In case we failed to insert the entry into the radixtree, we need
- * to look beyond the current sg.
- */
- while (idx + count <= n) {
- idx += count;
- sg = ____sg_next(sg);
- count = __sg_page_count(sg);
- }
-
- *offset = n - idx;
- return sg;
-
-lookup:
- rcu_read_lock();
-
- sg = radix_tree_lookup(&iter->radix, n);
- GEM_BUG_ON(!sg);
-
- /* If this index is in the middle of multi-page sg entry,
- * the radix tree will contain a value entry that points
- * to the start of that range. We will return the pointer to
- * the base page and the offset of this page within the
- * sg entry's range.
- */
- *offset = 0;
- if (unlikely(xa_is_value(sg))) {
- unsigned long base = xa_to_value(sg);
-
- sg = radix_tree_lookup(&iter->radix, base);
- GEM_BUG_ON(!sg);
-
- *offset = n - base;
- }
-
- rcu_read_unlock();
-
- return sg;
-}
-
-struct page *
-i915_gem_object_get_page(struct drm_i915_gem_object *obj, unsigned int n)
-{
- struct scatterlist *sg;
- unsigned int offset;
-
- GEM_BUG_ON(!i915_gem_object_has_struct_page(obj));
-
- sg = i915_gem_object_get_sg(obj, n, &offset);
- return nth_page(sg_page(sg), offset);
-}
-
-/* Like i915_gem_object_get_page(), but mark the returned page dirty */
-struct page *
-i915_gem_object_get_dirty_page(struct drm_i915_gem_object *obj,
- unsigned int n)
-{
- struct page *page;
-
- page = i915_gem_object_get_page(obj, n);
- if (!obj->mm.dirty)
- set_page_dirty(page);
-
- return page;
-}
-
-dma_addr_t
-i915_gem_object_get_dma_address(struct drm_i915_gem_object *obj,
- unsigned long n)
-{
- struct scatterlist *sg;
- unsigned int offset;
-
- sg = i915_gem_object_get_sg(obj, n, &offset);
- return sg_dma_address(sg) + (offset << PAGE_SHIFT);
-}
-
-int i915_gem_object_attach_phys(struct drm_i915_gem_object *obj, int align)
-{
- struct sg_table *pages;
- int err;
-
- if (align > obj->base.size)
- return -EINVAL;
-
- if (obj->ops == &i915_gem_phys_ops)
- return 0;
-
- if (obj->ops != &i915_gem_shmem_ops)
- return -EINVAL;
-
- err = i915_gem_object_unbind(obj);
- if (err)
- return err;
-
- mutex_lock(&obj->mm.lock);
-
- if (obj->mm.madv != I915_MADV_WILLNEED) {
- err = -EFAULT;
- goto err_unlock;
- }
-
- if (obj->mm.quirked) {
- err = -EFAULT;
- goto err_unlock;
- }
-
- if (obj->mm.mapping) {
- err = -EBUSY;
- goto err_unlock;
- }
-
- pages = __i915_gem_object_unset_pages(obj);
-
- obj->ops = &i915_gem_phys_ops;
-
- err = ____i915_gem_object_get_pages(obj);
- if (err)
- goto err_xfer;
-
- /* Perma-pin (until release) the physical set of pages */
- __i915_gem_object_pin_pages(obj);
-
- if (!IS_ERR_OR_NULL(pages))
- i915_gem_shmem_ops.put_pages(obj, pages);
- mutex_unlock(&obj->mm.lock);
- return 0;
-
-err_xfer:
- obj->ops = &i915_gem_shmem_ops;
- if (!IS_ERR_OR_NULL(pages)) {
- unsigned int sg_page_sizes = i915_sg_page_sizes(pages->sgl);
-
- __i915_gem_object_set_pages(obj, pages, sg_page_sizes);
- }
-err_unlock:
- mutex_unlock(&obj->mm.lock);
- return err;
-}
-
#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
#include "selftests/scatterlist.c"
#include "selftests/mock_gem_device.c"
diff --git a/drivers/gpu/drm/i915/i915_gem_shrinker.c b/drivers/gpu/drm/i915/i915_gem_shrinker.c
index 588e3898b120..2c7aefb3e101 100644
--- a/drivers/gpu/drm/i915/i915_gem_shrinker.c
+++ b/drivers/gpu/drm/i915/i915_gem_shrinker.c
@@ -114,65 +114,18 @@ static bool unsafe_drop_pages(struct drm_i915_gem_object *obj)
return !i915_gem_object_has_pages(obj);
}
-static void __start_writeback(struct drm_i915_gem_object *obj,
- unsigned int flags)
+static void try_to_writeback(struct drm_i915_gem_object *obj,
+ unsigned int flags)
{
- struct address_space *mapping;
- struct writeback_control wbc = {
- .sync_mode = WB_SYNC_NONE,
- .nr_to_write = SWAP_CLUSTER_MAX,
- .range_start = 0,
- .range_end = LLONG_MAX,
- .for_reclaim = 1,
- };
- unsigned long i;
-
- lockdep_assert_held(&obj->mm.lock);
- GEM_BUG_ON(i915_gem_object_has_pages(obj));
-
switch (obj->mm.madv) {
case I915_MADV_DONTNEED:
- __i915_gem_object_truncate(obj);
+ i915_gem_object_truncate(obj);
case __I915_MADV_PURGED:
return;
}
- if (!obj->base.filp)
- return;
-
- if (!(flags & I915_SHRINK_WRITEBACK))
- return;
-
- /*
- * Leave mmapings intact (GTT will have been revoked on unbinding,
- * leaving only CPU mmapings around) and add those pages to the LRU
- * instead of invoking writeback so they are aged and paged out
- * as normal.
- */
- mapping = obj->base.filp->f_mapping;
-
- /* Begin writeback on each dirty page */
- for (i = 0; i < obj->base.size >> PAGE_SHIFT; i++) {
- struct page *page;
-
- page = find_lock_entry(mapping, i);
- if (!page || xa_is_value(page))
- continue;
-
- if (!page_mapped(page) && clear_page_dirty_for_io(page)) {
- int ret;
-
- SetPageReclaim(page);
- ret = mapping->a_ops->writepage(page, &wbc);
- if (!PageWriteback(page))
- ClearPageReclaim(page);
- if (!ret)
- goto put;
- }
- unlock_page(page);
-put:
- put_page(page);
- }
+ if (flags & I915_SHRINK_WRITEBACK)
+ i915_gem_object_writeback(obj);
}
/**
@@ -315,7 +268,7 @@ i915_gem_shrink(struct drm_i915_private *i915,
mutex_lock_nested(&obj->mm.lock,
I915_MM_SHRINKER);
if (!i915_gem_object_has_pages(obj)) {
- __start_writeback(obj, flags);
+ try_to_writeback(obj, flags);
count += obj->base.size >> PAGE_SHIFT;
}
mutex_unlock(&obj->mm.lock);
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_object.c b/drivers/gpu/drm/i915/selftests/i915_gem_object.c
index a0f59fb0d701..b98a286a8be5 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_object.c
@@ -49,59 +49,6 @@ static int igt_gem_object(void *arg)
return err;
}
-static int igt_phys_object(void *arg)
-{
- struct drm_i915_private *i915 = arg;
- struct drm_i915_gem_object *obj;
- int err;
-
- /* Create an object and bind it to a contiguous set of physical pages,
- * i.e. exercise the i915_gem_object_phys API.
- */
-
- obj = i915_gem_object_create_shmem(i915, PAGE_SIZE);
- if (IS_ERR(obj)) {
- err = PTR_ERR(obj);
- pr_err("i915_gem_object_create failed, err=%d\n", err);
- goto out;
- }
-
- mutex_lock(&i915->drm.struct_mutex);
- err = i915_gem_object_attach_phys(obj, PAGE_SIZE);
- mutex_unlock(&i915->drm.struct_mutex);
- if (err) {
- pr_err("i915_gem_object_attach_phys failed, err=%d\n", err);
- goto out_obj;
- }
-
- if (obj->ops != &i915_gem_phys_ops) {
- pr_err("i915_gem_object_attach_phys did not create a phys object\n");
- err = -EINVAL;
- goto out_obj;
- }
-
- if (!atomic_read(&obj->mm.pages_pin_count)) {
- pr_err("i915_gem_object_attach_phys did not pin its phys pages\n");
- err = -EINVAL;
- goto out_obj;
- }
-
- /* Make the object dirty so that put_pages must do copy back the data */
- mutex_lock(&i915->drm.struct_mutex);
- err = i915_gem_object_set_to_gtt_domain(obj, true);
- mutex_unlock(&i915->drm.struct_mutex);
- if (err) {
- pr_err("i915_gem_object_set_to_gtt_domain failed with err=%d\n",
- err);
- goto out_obj;
- }
-
-out_obj:
- i915_gem_object_put(obj);
-out:
- return err;
-}
-
static int igt_gem_huge(void *arg)
{
const unsigned int nreal = 509; /* just to be awkward */
@@ -631,7 +578,6 @@ int i915_gem_object_mock_selftests(void)
{
static const struct i915_subtest tests[] = {
SUBTEST(igt_gem_object),
- SUBTEST(igt_phys_object),
};
struct drm_i915_private *i915;
int err;
diff --git a/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h b/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h
index 88e5ab586337..510eb176bb2c 100644
--- a/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h
+++ b/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h
@@ -18,6 +18,7 @@ selftest(engine, intel_engine_cs_mock_selftests)
selftest(timelines, i915_timeline_mock_selftests)
selftest(requests, i915_request_mock_selftests)
selftest(objects, i915_gem_object_mock_selftests)
+selftest(phys, i915_gem_phys_mock_selftests)
selftest(dmabuf, i915_gem_dmabuf_mock_selftests)
selftest(vma, i915_vma_mock_selftests)
selftest(evict, i915_gem_evict_mock_selftests)
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 24/40] drm/i915: Move mmap and friends to its own file
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (21 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 23/40] drm/i915: Move phys objects " Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 25/40] drm/i915: Move GEM domain management " Chris Wilson
` (22 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx; +Cc: Matthew Auld
Continuing the decluttering of i915_gem.c, now the turn of do_mmap and
the faulthandlers
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
---
drivers/gpu/drm/i915/Makefile | 1 +
drivers/gpu/drm/i915/gem/i915_gem_mman.c | 505 ++++++++++++++++
drivers/gpu/drm/i915/gem/i915_gem_object.c | 56 ++
drivers/gpu/drm/i915/gem/i915_gem_object.h | 7 +
.../drm/i915/gem/selftests/i915_gem_mman.c | 503 ++++++++++++++++
drivers/gpu/drm/i915/i915_drv.h | 1 -
drivers/gpu/drm/i915/i915_gem.c | 561 +-----------------
drivers/gpu/drm/i915/i915_gem_tiling.c | 2 +-
.../gpu/drm/i915/selftests/i915_gem_object.c | 487 ---------------
.../drm/i915/selftests/i915_live_selftests.h | 1 +
10 files changed, 1088 insertions(+), 1036 deletions(-)
create mode 100644 drivers/gpu/drm/i915/gem/i915_gem_mman.c
create mode 100644 drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index ba3b82f3cd49..d05757c52492 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -88,6 +88,7 @@ i915-y += $(gt-y)
obj-y += gem/
gem-y += \
gem/i915_gem_object.o \
+ gem/i915_gem_mman.o \
gem/i915_gem_pages.o \
gem/i915_gem_phys.o \
gem/i915_gem_shmem.o
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
new file mode 100644
index 000000000000..1bcc6e1091e9
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
@@ -0,0 +1,505 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2014-2016 Intel Corporation
+ */
+
+#include <linux/mman.h>
+#include <linux/sizes.h>
+
+#include "i915_gem_ioctls.h"
+#include "i915_gem_object.h"
+
+#include "../i915_gem_gtt.h"
+#include "../i915_vma.h"
+#include "../i915_drv.h"
+#include "../intel_drv.h"
+
+static inline bool
+__vma_matches(struct vm_area_struct *vma, struct file *filp,
+ unsigned long addr, unsigned long size)
+{
+ if (vma->vm_file != filp)
+ return false;
+
+ return vma->vm_start == addr &&
+ (vma->vm_end - vma->vm_start) == PAGE_ALIGN(size);
+}
+
+/**
+ * i915_gem_mmap_ioctl - Maps the contents of an object, returning the address
+ * it is mapped to.
+ * @dev: drm device
+ * @data: ioctl data blob
+ * @file: drm file
+ *
+ * While the mapping holds a reference on the contents of the object, it doesn't
+ * imply a ref on the object itself.
+ *
+ * IMPORTANT:
+ *
+ * DRM driver writers who look a this function as an example for how to do GEM
+ * mmap support, please don't implement mmap support like here. The modern way
+ * to implement DRM mmap support is with an mmap offset ioctl (like
+ * i915_gem_mmap_gtt) and then using the mmap syscall on the DRM fd directly.
+ * That way debug tooling like valgrind will understand what's going on, hiding
+ * the mmap call in a driver private ioctl will break that. The i915 driver only
+ * does cpu mmaps this way because we didn't know better.
+ */
+int
+i915_gem_mmap_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file)
+{
+ struct drm_i915_gem_mmap *args = data;
+ struct drm_i915_gem_object *obj;
+ unsigned long addr;
+
+ if (args->flags & ~(I915_MMAP_WC))
+ return -EINVAL;
+
+ if (args->flags & I915_MMAP_WC && !boot_cpu_has(X86_FEATURE_PAT))
+ return -ENODEV;
+
+ obj = i915_gem_object_lookup(file, args->handle);
+ if (!obj)
+ return -ENOENT;
+
+ /* prime objects have no backing filp to GEM mmap
+ * pages from.
+ */
+ if (!obj->base.filp) {
+ addr = -ENXIO;
+ goto err;
+ }
+
+ if (range_overflows(args->offset, args->size, (u64)obj->base.size)) {
+ addr = -EINVAL;
+ goto err;
+ }
+
+ addr = vm_mmap(obj->base.filp, 0, args->size,
+ PROT_READ | PROT_WRITE, MAP_SHARED,
+ args->offset);
+ if (IS_ERR_VALUE(addr))
+ goto err;
+
+ if (args->flags & I915_MMAP_WC) {
+ struct mm_struct *mm = current->mm;
+ struct vm_area_struct *vma;
+
+ if (down_write_killable(&mm->mmap_sem)) {
+ addr = -EINTR;
+ goto err;
+ }
+ vma = find_vma(mm, addr);
+ if (vma && __vma_matches(vma, obj->base.filp, addr, args->size))
+ vma->vm_page_prot =
+ pgprot_writecombine(vm_get_page_prot(vma->vm_flags));
+ else
+ addr = -ENOMEM;
+ up_write(&mm->mmap_sem);
+ if (IS_ERR_VALUE(addr))
+ goto err;
+
+ /* This may race, but that's ok, it only gets set */
+ WRITE_ONCE(obj->frontbuffer_ggtt_origin, ORIGIN_CPU);
+ }
+ i915_gem_object_put(obj);
+
+ args->addr_ptr = (u64)addr;
+ return 0;
+
+err:
+ i915_gem_object_put(obj);
+ return addr;
+}
+
+static unsigned int tile_row_pages(const struct drm_i915_gem_object *obj)
+{
+ return i915_gem_object_get_tile_row_size(obj) >> PAGE_SHIFT;
+}
+
+/**
+ * i915_gem_mmap_gtt_version - report the current feature set for GTT mmaps
+ *
+ * A history of the GTT mmap interface:
+ *
+ * 0 - Everything had to fit into the GTT. Both parties of a memcpy had to
+ * aligned and suitable for fencing, and still fit into the available
+ * mappable space left by the pinned display objects. A classic problem
+ * we called the page-fault-of-doom where we would ping-pong between
+ * two objects that could not fit inside the GTT and so the memcpy
+ * would page one object in at the expense of the other between every
+ * single byte.
+ *
+ * 1 - Objects can be any size, and have any compatible fencing (X Y, or none
+ * as set via i915_gem_set_tiling() [DRM_I915_GEM_SET_TILING]). If the
+ * object is too large for the available space (or simply too large
+ * for the mappable aperture!), a view is created instead and faulted
+ * into userspace. (This view is aligned and sized appropriately for
+ * fenced access.)
+ *
+ * 2 - Recognise WC as a separate cache domain so that we can flush the
+ * delayed writes via GTT before performing direct access via WC.
+ *
+ * 3 - Remove implicit set-domain(GTT) and synchronisation on initial
+ * pagefault; swapin remains transparent.
+ *
+ * Restrictions:
+ *
+ * * snoopable objects cannot be accessed via the GTT. It can cause machine
+ * hangs on some architectures, corruption on others. An attempt to service
+ * a GTT page fault from a snoopable object will generate a SIGBUS.
+ *
+ * * the object must be able to fit into RAM (physical memory, though no
+ * limited to the mappable aperture).
+ *
+ *
+ * Caveats:
+ *
+ * * a new GTT page fault will synchronize rendering from the GPU and flush
+ * all data to system memory. Subsequent access will not be synchronized.
+ *
+ * * all mappings are revoked on runtime device suspend.
+ *
+ * * there are only 8, 16 or 32 fence registers to share between all users
+ * (older machines require fence register for display and blitter access
+ * as well). Contention of the fence registers will cause the previous users
+ * to be unmapped and any new access will generate new page faults.
+ *
+ * * running out of memory while servicing a fault may generate a SIGBUS,
+ * rather than the expected SIGSEGV.
+ */
+int i915_gem_mmap_gtt_version(void)
+{
+ return 3;
+}
+
+static inline struct i915_ggtt_view
+compute_partial_view(const struct drm_i915_gem_object *obj,
+ pgoff_t page_offset,
+ unsigned int chunk)
+{
+ struct i915_ggtt_view view;
+
+ if (i915_gem_object_is_tiled(obj))
+ chunk = roundup(chunk, tile_row_pages(obj));
+
+ view.type = I915_GGTT_VIEW_PARTIAL;
+ view.partial.offset = rounddown(page_offset, chunk);
+ view.partial.size =
+ min_t(unsigned int, chunk,
+ (obj->base.size >> PAGE_SHIFT) - view.partial.offset);
+
+ /* If the partial covers the entire object, just create a normal VMA. */
+ if (chunk >= obj->base.size >> PAGE_SHIFT)
+ view.type = I915_GGTT_VIEW_NORMAL;
+
+ return view;
+}
+
+/**
+ * i915_gem_fault - fault a page into the GTT
+ * @vmf: fault info
+ *
+ * The fault handler is set up by drm_gem_mmap() when a object is GTT mapped
+ * from userspace. The fault handler takes care of binding the object to
+ * the GTT (if needed), allocating and programming a fence register (again,
+ * only if needed based on whether the old reg is still valid or the object
+ * is tiled) and inserting a new PTE into the faulting process.
+ *
+ * Note that the faulting process may involve evicting existing objects
+ * from the GTT and/or fence registers to make room. So performance may
+ * suffer if the GTT working set is large or there are few fence registers
+ * left.
+ *
+ * The current feature set supported by i915_gem_fault() and thus GTT mmaps
+ * is exposed via I915_PARAM_MMAP_GTT_VERSION (see i915_gem_mmap_gtt_version).
+ */
+vm_fault_t i915_gem_fault(struct vm_fault *vmf)
+{
+#define MIN_CHUNK_PAGES (SZ_1M >> PAGE_SHIFT)
+ struct vm_area_struct *area = vmf->vma;
+ struct drm_i915_gem_object *obj = to_intel_bo(area->vm_private_data);
+ struct drm_device *dev = obj->base.dev;
+ struct drm_i915_private *i915 = to_i915(dev);
+ struct i915_ggtt *ggtt = &i915->ggtt;
+ bool write = area->vm_flags & VM_WRITE;
+ intel_wakeref_t wakeref;
+ struct i915_vma *vma;
+ pgoff_t page_offset;
+ int srcu;
+ int ret;
+
+ /* Sanity check that we allow writing into this object */
+ if (i915_gem_object_is_readonly(obj) && write)
+ return VM_FAULT_SIGBUS;
+
+ /* We don't use vmf->pgoff since that has the fake offset */
+ page_offset = (vmf->address - area->vm_start) >> PAGE_SHIFT;
+
+ trace_i915_gem_object_fault(obj, page_offset, true, write);
+
+ ret = i915_gem_object_pin_pages(obj);
+ if (ret)
+ goto err;
+
+ wakeref = intel_runtime_pm_get(i915);
+
+ srcu = i915_reset_trylock(i915);
+ if (srcu < 0) {
+ ret = srcu;
+ goto err_rpm;
+ }
+
+ ret = i915_mutex_lock_interruptible(dev);
+ if (ret)
+ goto err_reset;
+
+ /* Access to snoopable pages through the GTT is incoherent. */
+ if (obj->cache_level != I915_CACHE_NONE && !HAS_LLC(i915)) {
+ ret = -EFAULT;
+ goto err_unlock;
+ }
+
+ /* Now pin it into the GTT as needed */
+ vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0,
+ PIN_MAPPABLE |
+ PIN_NONBLOCK |
+ PIN_NONFAULT);
+ if (IS_ERR(vma)) {
+ /* Use a partial view if it is bigger than available space */
+ struct i915_ggtt_view view =
+ compute_partial_view(obj, page_offset, MIN_CHUNK_PAGES);
+ unsigned int flags;
+
+ flags = PIN_MAPPABLE;
+ if (view.type == I915_GGTT_VIEW_NORMAL)
+ flags |= PIN_NONBLOCK; /* avoid warnings for pinned */
+
+ /*
+ * Userspace is now writing through an untracked VMA, abandon
+ * all hope that the hardware is able to track future writes.
+ */
+ obj->frontbuffer_ggtt_origin = ORIGIN_CPU;
+
+ vma = i915_gem_object_ggtt_pin(obj, &view, 0, 0, flags);
+ if (IS_ERR(vma) && !view.type) {
+ flags = PIN_MAPPABLE;
+ view.type = I915_GGTT_VIEW_PARTIAL;
+ vma = i915_gem_object_ggtt_pin(obj, &view, 0, 0, flags);
+ }
+ }
+ if (IS_ERR(vma)) {
+ ret = PTR_ERR(vma);
+ goto err_unlock;
+ }
+
+ ret = i915_vma_pin_fence(vma);
+ if (ret)
+ goto err_unpin;
+
+ /* Finally, remap it using the new GTT offset */
+ ret = remap_io_mapping(area,
+ area->vm_start + (vma->ggtt_view.partial.offset << PAGE_SHIFT),
+ (ggtt->gmadr.start + vma->node.start) >> PAGE_SHIFT,
+ min_t(u64, vma->size, area->vm_end - area->vm_start),
+ &ggtt->iomap);
+ if (ret)
+ goto err_fence;
+
+ /* Mark as being mmapped into userspace for later revocation */
+ assert_rpm_wakelock_held(i915);
+ if (!i915_vma_set_userfault(vma) && !obj->userfault_count++)
+ list_add(&obj->userfault_link, &i915->mm.userfault_list);
+ GEM_BUG_ON(!obj->userfault_count);
+
+ i915_vma_set_ggtt_write(vma);
+
+err_fence:
+ i915_vma_unpin_fence(vma);
+err_unpin:
+ __i915_vma_unpin(vma);
+err_unlock:
+ mutex_unlock(&dev->struct_mutex);
+err_reset:
+ i915_reset_unlock(i915, srcu);
+err_rpm:
+ intel_runtime_pm_put(i915, wakeref);
+ i915_gem_object_unpin_pages(obj);
+err:
+ switch (ret) {
+ case -EIO:
+ /*
+ * We eat errors when the gpu is terminally wedged to avoid
+ * userspace unduly crashing (gl has no provisions for mmaps to
+ * fail). But any other -EIO isn't ours (e.g. swap in failure)
+ * and so needs to be reported.
+ */
+ if (!i915_terminally_wedged(i915))
+ return VM_FAULT_SIGBUS;
+ /* else: fall through */
+ case -EAGAIN:
+ /*
+ * EAGAIN means the gpu is hung and we'll wait for the error
+ * handler to reset everything when re-faulting in
+ * i915_mutex_lock_interruptible.
+ */
+ case 0:
+ case -ERESTARTSYS:
+ case -EINTR:
+ case -EBUSY:
+ /*
+ * EBUSY is ok: this just means that another thread
+ * already did the job.
+ */
+ return VM_FAULT_NOPAGE;
+ case -ENOMEM:
+ return VM_FAULT_OOM;
+ case -ENOSPC:
+ case -EFAULT:
+ return VM_FAULT_SIGBUS;
+ default:
+ WARN_ONCE(ret, "unhandled error in %s: %i\n", __func__, ret);
+ return VM_FAULT_SIGBUS;
+ }
+}
+
+void __i915_gem_object_release_mmap(struct drm_i915_gem_object *obj)
+{
+ struct i915_vma *vma;
+
+ GEM_BUG_ON(!obj->userfault_count);
+
+ obj->userfault_count = 0;
+ list_del(&obj->userfault_link);
+ drm_vma_node_unmap(&obj->base.vma_node,
+ obj->base.dev->anon_inode->i_mapping);
+
+ for_each_ggtt_vma(vma, obj)
+ i915_vma_unset_userfault(vma);
+}
+
+/**
+ * i915_gem_object_release_mmap - remove physical page mappings
+ * @obj: obj in question
+ *
+ * Preserve the reservation of the mmapping with the DRM core code, but
+ * relinquish ownership of the pages back to the system.
+ *
+ * It is vital that we remove the page mapping if we have mapped a tiled
+ * object through the GTT and then lose the fence register due to
+ * resource pressure. Similarly if the object has been moved out of the
+ * aperture, than pages mapped into userspace must be revoked. Removing the
+ * mapping will then trigger a page fault on the next user access, allowing
+ * fixup by i915_gem_fault().
+ */
+void i915_gem_object_release_mmap(struct drm_i915_gem_object *obj)
+{
+ struct drm_i915_private *i915 = to_i915(obj->base.dev);
+ intel_wakeref_t wakeref;
+
+ /* Serialisation between user GTT access and our code depends upon
+ * revoking the CPU's PTE whilst the mutex is held. The next user
+ * pagefault then has to wait until we release the mutex.
+ *
+ * Note that RPM complicates somewhat by adding an additional
+ * requirement that operations to the GGTT be made holding the RPM
+ * wakeref.
+ */
+ lockdep_assert_held(&i915->drm.struct_mutex);
+ wakeref = intel_runtime_pm_get(i915);
+
+ if (!obj->userfault_count)
+ goto out;
+
+ __i915_gem_object_release_mmap(obj);
+
+ /* Ensure that the CPU's PTE are revoked and there are not outstanding
+ * memory transactions from userspace before we return. The TLB
+ * flushing implied above by changing the PTE above *should* be
+ * sufficient, an extra barrier here just provides us with a bit
+ * of paranoid documentation about our requirement to serialise
+ * memory writes before touching registers / GSM.
+ */
+ wmb();
+
+out:
+ intel_runtime_pm_put(i915, wakeref);
+}
+
+static int create_mmap_offset(struct drm_i915_gem_object *obj)
+{
+ struct drm_i915_private *i915 = to_i915(obj->base.dev);
+ int err;
+
+ err = drm_gem_create_mmap_offset(&obj->base);
+ if (likely(!err))
+ return 0;
+
+ /* Attempt to reap some mmap space from dead objects */
+ do {
+ err = i915_gem_wait_for_idle(i915,
+ I915_WAIT_INTERRUPTIBLE,
+ MAX_SCHEDULE_TIMEOUT);
+ if (err)
+ break;
+
+ i915_gem_drain_freed_objects(i915);
+ err = drm_gem_create_mmap_offset(&obj->base);
+ if (!err)
+ break;
+
+ } while (flush_delayed_work(&i915->gem.retire_work));
+
+ return err;
+}
+
+int
+i915_gem_mmap_gtt(struct drm_file *file,
+ struct drm_device *dev,
+ u32 handle,
+ u64 *offset)
+{
+ struct drm_i915_gem_object *obj;
+ int ret;
+
+ obj = i915_gem_object_lookup(file, handle);
+ if (!obj)
+ return -ENOENT;
+
+ ret = create_mmap_offset(obj);
+ if (ret == 0)
+ *offset = drm_vma_node_offset_addr(&obj->base.vma_node);
+
+ i915_gem_object_put(obj);
+ return ret;
+}
+
+/**
+ * i915_gem_mmap_gtt_ioctl - prepare an object for GTT mmap'ing
+ * @dev: DRM device
+ * @data: GTT mapping ioctl data
+ * @file: GEM object info
+ *
+ * Simply returns the fake offset to userspace so it can mmap it.
+ * The mmap call will end up in drm_gem_mmap(), which will set things
+ * up so we can get faults in the handler above.
+ *
+ * The fault handler will take care of binding the object into the GTT
+ * (since it may have been evicted to make room for something), allocating
+ * a fence register, and mapping the appropriate aperture address into
+ * userspace.
+ */
+int
+i915_gem_mmap_gtt_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file)
+{
+ struct drm_i915_gem_mmap_gtt *args = data;
+
+ return i915_gem_mmap_gtt(file, dev, args->handle, &args->offset);
+}
+
+#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
+#include "selftests/i915_gem_mman.c"
+#endif
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
index 86e7e88817af..005e17b3acce 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
@@ -26,6 +26,7 @@
#include "../i915_drv.h"
#include "../i915_globals.h"
+#include "../i915_gem_clflush.h"
#include "../intel_frontbuffer.h"
static struct i915_global_object {
@@ -357,6 +358,61 @@ void __i915_gem_object_release_unless_active(struct drm_i915_gem_object *obj)
i915_gem_object_put(obj);
}
+static inline enum fb_op_origin
+fb_write_origin(struct drm_i915_gem_object *obj, unsigned int domain)
+{
+ return (domain == I915_GEM_DOMAIN_GTT ?
+ obj->frontbuffer_ggtt_origin : ORIGIN_CPU);
+}
+
+static bool gpu_write_needs_clflush(struct drm_i915_gem_object *obj)
+{
+ return !(obj->cache_level == I915_CACHE_NONE ||
+ obj->cache_level == I915_CACHE_WT);
+}
+
+void
+i915_gem_object_flush_write_domain(struct drm_i915_gem_object *obj,
+ unsigned int flush_domains)
+{
+ struct drm_i915_private *dev_priv = to_i915(obj->base.dev);
+ struct i915_vma *vma;
+
+ if (!(obj->write_domain & flush_domains))
+ return;
+
+ switch (obj->write_domain) {
+ case I915_GEM_DOMAIN_GTT:
+ i915_gem_flush_ggtt_writes(dev_priv);
+
+ intel_fb_obj_flush(obj,
+ fb_write_origin(obj, I915_GEM_DOMAIN_GTT));
+
+ for_each_ggtt_vma(vma, obj) {
+ if (vma->iomap)
+ continue;
+
+ i915_vma_unset_ggtt_write(vma);
+ }
+ break;
+
+ case I915_GEM_DOMAIN_WC:
+ wmb();
+ break;
+
+ case I915_GEM_DOMAIN_CPU:
+ i915_gem_clflush_object(obj, I915_CLFLUSH_SYNC);
+ break;
+
+ case I915_GEM_DOMAIN_RENDER:
+ if (gpu_write_needs_clflush(obj))
+ obj->cache_dirty = true;
+ break;
+ }
+
+ obj->write_domain = 0;
+}
+
void i915_gem_init__objects(struct drm_i915_private *i915)
{
INIT_WORK(&i915->mm.free_work, __i915_gem_free_work);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index 2e963a593245..9bf8155d27de 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -346,6 +346,13 @@ static inline void i915_gem_object_unpin_map(struct drm_i915_gem_object *obj)
i915_gem_object_unpin_pages(obj);
}
+void __i915_gem_object_release_mmap(struct drm_i915_gem_object *obj);
+void i915_gem_object_release_mmap(struct drm_i915_gem_object *obj);
+
+void
+i915_gem_object_flush_write_domain(struct drm_i915_gem_object *obj,
+ unsigned int flush_domains);
+
static inline struct intel_engine_cs *
i915_gem_object_last_write_engine(struct drm_i915_gem_object *obj)
{
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
new file mode 100644
index 000000000000..87da01230179
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
@@ -0,0 +1,503 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2016 Intel Corporation
+ */
+
+#include <linux/prime_numbers.h>
+
+#include "gt/intel_gt_pm.h"
+#include "i915_selftest.h"
+#include "selftests/huge_gem_object.h"
+#include "selftests/igt_flush_test.h"
+
+struct tile {
+ unsigned int width;
+ unsigned int height;
+ unsigned int stride;
+ unsigned int size;
+ unsigned int tiling;
+ unsigned int swizzle;
+};
+
+static u64 swizzle_bit(unsigned int bit, u64 offset)
+{
+ return (offset & BIT_ULL(bit)) >> (bit - 6);
+}
+
+static u64 tiled_offset(const struct tile *tile, u64 v)
+{
+ u64 x, y;
+
+ if (tile->tiling == I915_TILING_NONE)
+ return v;
+
+ y = div64_u64_rem(v, tile->stride, &x);
+ v = div64_u64_rem(y, tile->height, &y) * tile->stride * tile->height;
+
+ if (tile->tiling == I915_TILING_X) {
+ v += y * tile->width;
+ v += div64_u64_rem(x, tile->width, &x) << tile->size;
+ v += x;
+ } else if (tile->width == 128) {
+ const unsigned int ytile_span = 16;
+ const unsigned int ytile_height = 512;
+
+ v += y * ytile_span;
+ v += div64_u64_rem(x, ytile_span, &x) * ytile_height;
+ v += x;
+ } else {
+ const unsigned int ytile_span = 32;
+ const unsigned int ytile_height = 256;
+
+ v += y * ytile_span;
+ v += div64_u64_rem(x, ytile_span, &x) * ytile_height;
+ v += x;
+ }
+
+ switch (tile->swizzle) {
+ case I915_BIT_6_SWIZZLE_9:
+ v ^= swizzle_bit(9, v);
+ break;
+ case I915_BIT_6_SWIZZLE_9_10:
+ v ^= swizzle_bit(9, v) ^ swizzle_bit(10, v);
+ break;
+ case I915_BIT_6_SWIZZLE_9_11:
+ v ^= swizzle_bit(9, v) ^ swizzle_bit(11, v);
+ break;
+ case I915_BIT_6_SWIZZLE_9_10_11:
+ v ^= swizzle_bit(9, v) ^ swizzle_bit(10, v) ^ swizzle_bit(11, v);
+ break;
+ }
+
+ return v;
+}
+
+static int check_partial_mapping(struct drm_i915_gem_object *obj,
+ const struct tile *tile,
+ unsigned long end_time)
+{
+ const unsigned int nreal = obj->scratch / PAGE_SIZE;
+ const unsigned long npages = obj->base.size / PAGE_SIZE;
+ struct i915_vma *vma;
+ unsigned long page;
+ int err;
+
+ if (igt_timeout(end_time,
+ "%s: timed out before tiling=%d stride=%d\n",
+ __func__, tile->tiling, tile->stride))
+ return -EINTR;
+
+ err = i915_gem_object_set_tiling(obj, tile->tiling, tile->stride);
+ if (err) {
+ pr_err("Failed to set tiling mode=%u, stride=%u, err=%d\n",
+ tile->tiling, tile->stride, err);
+ return err;
+ }
+
+ GEM_BUG_ON(i915_gem_object_get_tiling(obj) != tile->tiling);
+ GEM_BUG_ON(i915_gem_object_get_stride(obj) != tile->stride);
+
+ for_each_prime_number_from(page, 1, npages) {
+ struct i915_ggtt_view view =
+ compute_partial_view(obj, page, MIN_CHUNK_PAGES);
+ u32 __iomem *io;
+ struct page *p;
+ unsigned int n;
+ u64 offset;
+ u32 *cpu;
+
+ GEM_BUG_ON(view.partial.size > nreal);
+ cond_resched();
+
+ err = i915_gem_object_set_to_gtt_domain(obj, true);
+ if (err) {
+ pr_err("Failed to flush to GTT write domain; err=%d\n",
+ err);
+ return err;
+ }
+
+ vma = i915_gem_object_ggtt_pin(obj, &view, 0, 0, PIN_MAPPABLE);
+ if (IS_ERR(vma)) {
+ pr_err("Failed to pin partial view: offset=%lu; err=%d\n",
+ page, (int)PTR_ERR(vma));
+ return PTR_ERR(vma);
+ }
+
+ n = page - view.partial.offset;
+ GEM_BUG_ON(n >= view.partial.size);
+
+ io = i915_vma_pin_iomap(vma);
+ i915_vma_unpin(vma);
+ if (IS_ERR(io)) {
+ pr_err("Failed to iomap partial view: offset=%lu; err=%d\n",
+ page, (int)PTR_ERR(io));
+ return PTR_ERR(io);
+ }
+
+ iowrite32(page, io + n * PAGE_SIZE / sizeof(*io));
+ i915_vma_unpin_iomap(vma);
+
+ offset = tiled_offset(tile, page << PAGE_SHIFT);
+ if (offset >= obj->base.size)
+ continue;
+
+ i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+
+ p = i915_gem_object_get_page(obj, offset >> PAGE_SHIFT);
+ cpu = kmap(p) + offset_in_page(offset);
+ drm_clflush_virt_range(cpu, sizeof(*cpu));
+ if (*cpu != (u32)page) {
+ pr_err("Partial view for %lu [%u] (offset=%llu, size=%u [%llu, row size %u], fence=%d, tiling=%d, stride=%d) misalignment, expected write to page (%llu + %u [0x%llx]) of 0x%x, found 0x%x\n",
+ page, n,
+ view.partial.offset,
+ view.partial.size,
+ vma->size >> PAGE_SHIFT,
+ tile->tiling ? tile_row_pages(obj) : 0,
+ vma->fence ? vma->fence->id : -1, tile->tiling, tile->stride,
+ offset >> PAGE_SHIFT,
+ (unsigned int)offset_in_page(offset),
+ offset,
+ (u32)page, *cpu);
+ err = -EINVAL;
+ }
+ *cpu = 0;
+ drm_clflush_virt_range(cpu, sizeof(*cpu));
+ kunmap(p);
+ if (err)
+ return err;
+
+ i915_vma_destroy(vma);
+ }
+
+ return 0;
+}
+
+static int igt_partial_tiling(void *arg)
+{
+ const unsigned int nreal = 1 << 12; /* largest tile row x2 */
+ struct drm_i915_private *i915 = arg;
+ struct drm_i915_gem_object *obj;
+ intel_wakeref_t wakeref;
+ int tiling;
+ int err;
+
+ /* We want to check the page mapping and fencing of a large object
+ * mmapped through the GTT. The object we create is larger than can
+ * possibly be mmaped as a whole, and so we must use partial GGTT vma.
+ * We then check that a write through each partial GGTT vma ends up
+ * in the right set of pages within the object, and with the expected
+ * tiling, which we verify by manual swizzling.
+ */
+
+ obj = huge_gem_object(i915,
+ nreal << PAGE_SHIFT,
+ (1 + next_prime_number(i915->ggtt.vm.total >> PAGE_SHIFT)) << PAGE_SHIFT);
+ if (IS_ERR(obj))
+ return PTR_ERR(obj);
+
+ err = i915_gem_object_pin_pages(obj);
+ if (err) {
+ pr_err("Failed to allocate %u pages (%lu total), err=%d\n",
+ nreal, obj->base.size / PAGE_SIZE, err);
+ goto out;
+ }
+
+ mutex_lock(&i915->drm.struct_mutex);
+ wakeref = intel_runtime_pm_get(i915);
+
+ if (1) {
+ IGT_TIMEOUT(end);
+ struct tile tile;
+
+ tile.height = 1;
+ tile.width = 1;
+ tile.size = 0;
+ tile.stride = 0;
+ tile.swizzle = I915_BIT_6_SWIZZLE_NONE;
+ tile.tiling = I915_TILING_NONE;
+
+ err = check_partial_mapping(obj, &tile, end);
+ if (err && err != -EINTR)
+ goto out_unlock;
+ }
+
+ for (tiling = I915_TILING_X; tiling <= I915_TILING_Y; tiling++) {
+ IGT_TIMEOUT(end);
+ unsigned int max_pitch;
+ unsigned int pitch;
+ struct tile tile;
+
+ if (i915->quirks & QUIRK_PIN_SWIZZLED_PAGES)
+ /*
+ * The swizzling pattern is actually unknown as it
+ * varies based on physical address of each page.
+ * See i915_gem_detect_bit_6_swizzle().
+ */
+ break;
+
+ tile.tiling = tiling;
+ switch (tiling) {
+ case I915_TILING_X:
+ tile.swizzle = i915->mm.bit_6_swizzle_x;
+ break;
+ case I915_TILING_Y:
+ tile.swizzle = i915->mm.bit_6_swizzle_y;
+ break;
+ }
+
+ GEM_BUG_ON(tile.swizzle == I915_BIT_6_SWIZZLE_UNKNOWN);
+ if (tile.swizzle == I915_BIT_6_SWIZZLE_9_17 ||
+ tile.swizzle == I915_BIT_6_SWIZZLE_9_10_17)
+ continue;
+
+ if (INTEL_GEN(i915) <= 2) {
+ tile.height = 16;
+ tile.width = 128;
+ tile.size = 11;
+ } else if (tile.tiling == I915_TILING_Y &&
+ HAS_128_BYTE_Y_TILING(i915)) {
+ tile.height = 32;
+ tile.width = 128;
+ tile.size = 12;
+ } else {
+ tile.height = 8;
+ tile.width = 512;
+ tile.size = 12;
+ }
+
+ if (INTEL_GEN(i915) < 4)
+ max_pitch = 8192 / tile.width;
+ else if (INTEL_GEN(i915) < 7)
+ max_pitch = 128 * I965_FENCE_MAX_PITCH_VAL / tile.width;
+ else
+ max_pitch = 128 * GEN7_FENCE_MAX_PITCH_VAL / tile.width;
+
+ for (pitch = max_pitch; pitch; pitch >>= 1) {
+ tile.stride = tile.width * pitch;
+ err = check_partial_mapping(obj, &tile, end);
+ if (err == -EINTR)
+ goto next_tiling;
+ if (err)
+ goto out_unlock;
+
+ if (pitch > 2 && INTEL_GEN(i915) >= 4) {
+ tile.stride = tile.width * (pitch - 1);
+ err = check_partial_mapping(obj, &tile, end);
+ if (err == -EINTR)
+ goto next_tiling;
+ if (err)
+ goto out_unlock;
+ }
+
+ if (pitch < max_pitch && INTEL_GEN(i915) >= 4) {
+ tile.stride = tile.width * (pitch + 1);
+ err = check_partial_mapping(obj, &tile, end);
+ if (err == -EINTR)
+ goto next_tiling;
+ if (err)
+ goto out_unlock;
+ }
+ }
+
+ if (INTEL_GEN(i915) >= 4) {
+ for_each_prime_number(pitch, max_pitch) {
+ tile.stride = tile.width * pitch;
+ err = check_partial_mapping(obj, &tile, end);
+ if (err == -EINTR)
+ goto next_tiling;
+ if (err)
+ goto out_unlock;
+ }
+ }
+
+next_tiling: ;
+ }
+
+out_unlock:
+ intel_runtime_pm_put(i915, wakeref);
+ mutex_unlock(&i915->drm.struct_mutex);
+ i915_gem_object_unpin_pages(obj);
+out:
+ i915_gem_object_put(obj);
+ return err;
+}
+
+static int make_obj_busy(struct drm_i915_gem_object *obj)
+{
+ struct drm_i915_private *i915 = to_i915(obj->base.dev);
+ struct i915_request *rq;
+ struct i915_vma *vma;
+ int err;
+
+ vma = i915_vma_instance(obj, &i915->ggtt.vm, NULL);
+ if (IS_ERR(vma))
+ return PTR_ERR(vma);
+
+ err = i915_vma_pin(vma, 0, 0, PIN_USER);
+ if (err)
+ return err;
+
+ rq = i915_request_create(i915->engine[RCS0]->kernel_context);
+ if (IS_ERR(rq)) {
+ i915_vma_unpin(vma);
+ return PTR_ERR(rq);
+ }
+
+ err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+
+ i915_request_add(rq);
+
+ __i915_gem_object_release_unless_active(obj);
+ i915_vma_unpin(vma);
+
+ return err;
+}
+
+static bool assert_mmap_offset(struct drm_i915_private *i915,
+ unsigned long size,
+ int expected)
+{
+ struct drm_i915_gem_object *obj;
+ int err;
+
+ obj = i915_gem_object_create_internal(i915, size);
+ if (IS_ERR(obj))
+ return PTR_ERR(obj);
+
+ err = create_mmap_offset(obj);
+ i915_gem_object_put(obj);
+
+ return err == expected;
+}
+
+static void disable_retire_worker(struct drm_i915_private *i915)
+{
+ i915_gem_shrinker_unregister(i915);
+
+ intel_gt_pm_get(i915);
+
+ cancel_delayed_work_sync(&i915->gem.retire_work);
+ flush_work(&i915->gem.idle_work);
+}
+
+static void restore_retire_worker(struct drm_i915_private *i915)
+{
+ intel_gt_pm_put(i915);
+
+ mutex_lock(&i915->drm.struct_mutex);
+ igt_flush_test(i915, I915_WAIT_LOCKED);
+ mutex_unlock(&i915->drm.struct_mutex);
+
+ i915_gem_shrinker_register(i915);
+}
+
+static int igt_mmap_offset_exhaustion(void *arg)
+{
+ struct drm_i915_private *i915 = arg;
+ struct drm_mm *mm = &i915->drm.vma_offset_manager->vm_addr_space_mm;
+ struct drm_i915_gem_object *obj;
+ struct drm_mm_node resv, *hole;
+ u64 hole_start, hole_end;
+ int loop, err;
+
+ /* Disable background reaper */
+ disable_retire_worker(i915);
+ GEM_BUG_ON(!i915->gt.awake);
+
+ /* Trim the device mmap space to only a page */
+ memset(&resv, 0, sizeof(resv));
+ drm_mm_for_each_hole(hole, mm, hole_start, hole_end) {
+ resv.start = hole_start;
+ resv.size = hole_end - hole_start - 1; /* PAGE_SIZE units */
+ err = drm_mm_reserve_node(mm, &resv);
+ if (err) {
+ pr_err("Failed to trim VMA manager, err=%d\n", err);
+ goto out_park;
+ }
+ break;
+ }
+
+ /* Just fits! */
+ if (!assert_mmap_offset(i915, PAGE_SIZE, 0)) {
+ pr_err("Unable to insert object into single page hole\n");
+ err = -EINVAL;
+ goto out;
+ }
+
+ /* Too large */
+ if (!assert_mmap_offset(i915, 2 * PAGE_SIZE, -ENOSPC)) {
+ pr_err("Unexpectedly succeeded in inserting too large object into single page hole\n");
+ err = -EINVAL;
+ goto out;
+ }
+
+ /* Fill the hole, further allocation attempts should then fail */
+ obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
+ if (IS_ERR(obj)) {
+ err = PTR_ERR(obj);
+ goto out;
+ }
+
+ err = create_mmap_offset(obj);
+ if (err) {
+ pr_err("Unable to insert object into reclaimed hole\n");
+ goto err_obj;
+ }
+
+ if (!assert_mmap_offset(i915, PAGE_SIZE, -ENOSPC)) {
+ pr_err("Unexpectedly succeeded in inserting object into no holes!\n");
+ err = -EINVAL;
+ goto err_obj;
+ }
+
+ i915_gem_object_put(obj);
+
+ /* Now fill with busy dead objects that we expect to reap */
+ for (loop = 0; loop < 3; loop++) {
+ if (i915_terminally_wedged(i915))
+ break;
+
+ obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
+ if (IS_ERR(obj)) {
+ err = PTR_ERR(obj);
+ goto out;
+ }
+
+ mutex_lock(&i915->drm.struct_mutex);
+ err = make_obj_busy(obj);
+ mutex_unlock(&i915->drm.struct_mutex);
+ if (err) {
+ pr_err("[loop %d] Failed to busy the object\n", loop);
+ goto err_obj;
+ }
+
+ /* NB we rely on the _active_ reference to access obj now */
+ GEM_BUG_ON(!i915_gem_object_is_active(obj));
+ err = create_mmap_offset(obj);
+ if (err) {
+ pr_err("[loop %d] create_mmap_offset failed with err=%d\n",
+ loop, err);
+ goto out;
+ }
+ }
+
+out:
+ drm_mm_remove_node(&resv);
+out_park:
+ restore_retire_worker(i915);
+ return err;
+err_obj:
+ i915_gem_object_put(obj);
+ goto out;
+}
+
+int i915_gem_mman_live_selftests(struct drm_i915_private *i915)
+{
+ static const struct i915_subtest tests[] = {
+ SUBTEST(igt_partial_tiling),
+ SUBTEST(igt_mmap_offset_exhaustion),
+ };
+
+ return i915_subtests(tests, i915);
+}
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index da9f5a375613..e8fa6f4b100c 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2790,7 +2790,6 @@ i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj,
u64 flags);
int i915_gem_object_unbind(struct drm_i915_gem_object *obj);
-void i915_gem_release_mmap(struct drm_i915_gem_object *obj);
void i915_gem_runtime_suspend(struct drm_i915_private *dev_priv);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 0549b73f01ff..78d99841cf76 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -380,12 +380,6 @@ i915_gem_dumb_create(struct drm_file *file,
&args->size, &args->handle);
}
-static bool gpu_write_needs_clflush(struct drm_i915_gem_object *obj)
-{
- return !(obj->cache_level == I915_CACHE_NONE ||
- obj->cache_level == I915_CACHE_WT);
-}
-
/**
* Creates a new mm object and returns a handle to it.
* @dev: drm device pointer
@@ -405,13 +399,6 @@ i915_gem_create_ioctl(struct drm_device *dev, void *data,
&args->size, &args->handle);
}
-static inline enum fb_op_origin
-fb_write_origin(struct drm_i915_gem_object *obj, unsigned int domain)
-{
- return (domain == I915_GEM_DOMAIN_GTT ?
- obj->frontbuffer_ggtt_origin : ORIGIN_CPU);
-}
-
void i915_gem_flush_ggtt_writes(struct drm_i915_private *dev_priv)
{
intel_wakeref_t wakeref;
@@ -451,47 +438,6 @@ void i915_gem_flush_ggtt_writes(struct drm_i915_private *dev_priv)
}
}
-static void
-flush_write_domain(struct drm_i915_gem_object *obj, unsigned int flush_domains)
-{
- struct drm_i915_private *dev_priv = to_i915(obj->base.dev);
- struct i915_vma *vma;
-
- if (!(obj->write_domain & flush_domains))
- return;
-
- switch (obj->write_domain) {
- case I915_GEM_DOMAIN_GTT:
- i915_gem_flush_ggtt_writes(dev_priv);
-
- intel_fb_obj_flush(obj,
- fb_write_origin(obj, I915_GEM_DOMAIN_GTT));
-
- for_each_ggtt_vma(vma, obj) {
- if (vma->iomap)
- continue;
-
- i915_vma_unset_ggtt_write(vma);
- }
- break;
-
- case I915_GEM_DOMAIN_WC:
- wmb();
- break;
-
- case I915_GEM_DOMAIN_CPU:
- i915_gem_clflush_object(obj, I915_CLFLUSH_SYNC);
- break;
-
- case I915_GEM_DOMAIN_RENDER:
- if (gpu_write_needs_clflush(obj))
- obj->cache_dirty = true;
- break;
- }
-
- obj->write_domain = 0;
-}
-
/*
* Pins the specified object's pages and synchronizes the object with
* GPU accesses. Sets needs_clflush to non-zero if the caller should
@@ -528,7 +474,7 @@ int i915_gem_obj_prepare_shmem_read(struct drm_i915_gem_object *obj,
goto out;
}
- flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+ i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
/* If we're not in the cpu read domain, set ourself into the gtt
* read domain and manually flush cachelines (if required). This
@@ -580,7 +526,7 @@ int i915_gem_obj_prepare_shmem_write(struct drm_i915_gem_object *obj,
goto out;
}
- flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+ i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
/* If we're not in the cpu write domain, set ourself into the
* gtt write domain and manually flush cachelines (as required).
@@ -1183,6 +1129,13 @@ static void i915_gem_object_bump_inactive_ggtt(struct drm_i915_gem_object *obj)
spin_unlock(&i915->mm.obj_lock);
}
+static inline enum fb_op_origin
+fb_write_origin(struct drm_i915_gem_object *obj, unsigned int domain)
+{
+ return (domain == I915_GEM_DOMAIN_GTT ?
+ obj->frontbuffer_ggtt_origin : ORIGIN_CPU);
+}
+
/**
* Called when user space prepares to use an object with the CPU, either
* through the mmap ioctl's mapping or a GTT mapping.
@@ -1326,420 +1279,6 @@ i915_gem_sw_finish_ioctl(struct drm_device *dev, void *data,
return 0;
}
-static inline bool
-__vma_matches(struct vm_area_struct *vma, struct file *filp,
- unsigned long addr, unsigned long size)
-{
- if (vma->vm_file != filp)
- return false;
-
- return vma->vm_start == addr &&
- (vma->vm_end - vma->vm_start) == PAGE_ALIGN(size);
-}
-
-/**
- * i915_gem_mmap_ioctl - Maps the contents of an object, returning the address
- * it is mapped to.
- * @dev: drm device
- * @data: ioctl data blob
- * @file: drm file
- *
- * While the mapping holds a reference on the contents of the object, it doesn't
- * imply a ref on the object itself.
- *
- * IMPORTANT:
- *
- * DRM driver writers who look a this function as an example for how to do GEM
- * mmap support, please don't implement mmap support like here. The modern way
- * to implement DRM mmap support is with an mmap offset ioctl (like
- * i915_gem_mmap_gtt) and then using the mmap syscall on the DRM fd directly.
- * That way debug tooling like valgrind will understand what's going on, hiding
- * the mmap call in a driver private ioctl will break that. The i915 driver only
- * does cpu mmaps this way because we didn't know better.
- */
-int
-i915_gem_mmap_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file)
-{
- struct drm_i915_gem_mmap *args = data;
- struct drm_i915_gem_object *obj;
- unsigned long addr;
-
- if (args->flags & ~(I915_MMAP_WC))
- return -EINVAL;
-
- if (args->flags & I915_MMAP_WC && !boot_cpu_has(X86_FEATURE_PAT))
- return -ENODEV;
-
- obj = i915_gem_object_lookup(file, args->handle);
- if (!obj)
- return -ENOENT;
-
- /* prime objects have no backing filp to GEM mmap
- * pages from.
- */
- if (!obj->base.filp) {
- addr = -ENXIO;
- goto err;
- }
-
- if (range_overflows(args->offset, args->size, (u64)obj->base.size)) {
- addr = -EINVAL;
- goto err;
- }
-
- addr = vm_mmap(obj->base.filp, 0, args->size,
- PROT_READ | PROT_WRITE, MAP_SHARED,
- args->offset);
- if (IS_ERR_VALUE(addr))
- goto err;
-
- if (args->flags & I915_MMAP_WC) {
- struct mm_struct *mm = current->mm;
- struct vm_area_struct *vma;
-
- if (down_write_killable(&mm->mmap_sem)) {
- addr = -EINTR;
- goto err;
- }
- vma = find_vma(mm, addr);
- if (vma && __vma_matches(vma, obj->base.filp, addr, args->size))
- vma->vm_page_prot =
- pgprot_writecombine(vm_get_page_prot(vma->vm_flags));
- else
- addr = -ENOMEM;
- up_write(&mm->mmap_sem);
- if (IS_ERR_VALUE(addr))
- goto err;
-
- /* This may race, but that's ok, it only gets set */
- WRITE_ONCE(obj->frontbuffer_ggtt_origin, ORIGIN_CPU);
- }
- i915_gem_object_put(obj);
-
- args->addr_ptr = (u64)addr;
- return 0;
-
-err:
- i915_gem_object_put(obj);
- return addr;
-}
-
-static unsigned int tile_row_pages(const struct drm_i915_gem_object *obj)
-{
- return i915_gem_object_get_tile_row_size(obj) >> PAGE_SHIFT;
-}
-
-/**
- * i915_gem_mmap_gtt_version - report the current feature set for GTT mmaps
- *
- * A history of the GTT mmap interface:
- *
- * 0 - Everything had to fit into the GTT. Both parties of a memcpy had to
- * aligned and suitable for fencing, and still fit into the available
- * mappable space left by the pinned display objects. A classic problem
- * we called the page-fault-of-doom where we would ping-pong between
- * two objects that could not fit inside the GTT and so the memcpy
- * would page one object in at the expense of the other between every
- * single byte.
- *
- * 1 - Objects can be any size, and have any compatible fencing (X Y, or none
- * as set via i915_gem_set_tiling() [DRM_I915_GEM_SET_TILING]). If the
- * object is too large for the available space (or simply too large
- * for the mappable aperture!), a view is created instead and faulted
- * into userspace. (This view is aligned and sized appropriately for
- * fenced access.)
- *
- * 2 - Recognise WC as a separate cache domain so that we can flush the
- * delayed writes via GTT before performing direct access via WC.
- *
- * 3 - Remove implicit set-domain(GTT) and synchronisation on initial
- * pagefault; swapin remains transparent.
- *
- * Restrictions:
- *
- * * snoopable objects cannot be accessed via the GTT. It can cause machine
- * hangs on some architectures, corruption on others. An attempt to service
- * a GTT page fault from a snoopable object will generate a SIGBUS.
- *
- * * the object must be able to fit into RAM (physical memory, though no
- * limited to the mappable aperture).
- *
- *
- * Caveats:
- *
- * * a new GTT page fault will synchronize rendering from the GPU and flush
- * all data to system memory. Subsequent access will not be synchronized.
- *
- * * all mappings are revoked on runtime device suspend.
- *
- * * there are only 8, 16 or 32 fence registers to share between all users
- * (older machines require fence register for display and blitter access
- * as well). Contention of the fence registers will cause the previous users
- * to be unmapped and any new access will generate new page faults.
- *
- * * running out of memory while servicing a fault may generate a SIGBUS,
- * rather than the expected SIGSEGV.
- */
-int i915_gem_mmap_gtt_version(void)
-{
- return 3;
-}
-
-static inline struct i915_ggtt_view
-compute_partial_view(const struct drm_i915_gem_object *obj,
- pgoff_t page_offset,
- unsigned int chunk)
-{
- struct i915_ggtt_view view;
-
- if (i915_gem_object_is_tiled(obj))
- chunk = roundup(chunk, tile_row_pages(obj));
-
- view.type = I915_GGTT_VIEW_PARTIAL;
- view.partial.offset = rounddown(page_offset, chunk);
- view.partial.size =
- min_t(unsigned int, chunk,
- (obj->base.size >> PAGE_SHIFT) - view.partial.offset);
-
- /* If the partial covers the entire object, just create a normal VMA. */
- if (chunk >= obj->base.size >> PAGE_SHIFT)
- view.type = I915_GGTT_VIEW_NORMAL;
-
- return view;
-}
-
-/**
- * i915_gem_fault - fault a page into the GTT
- * @vmf: fault info
- *
- * The fault handler is set up by drm_gem_mmap() when a object is GTT mapped
- * from userspace. The fault handler takes care of binding the object to
- * the GTT (if needed), allocating and programming a fence register (again,
- * only if needed based on whether the old reg is still valid or the object
- * is tiled) and inserting a new PTE into the faulting process.
- *
- * Note that the faulting process may involve evicting existing objects
- * from the GTT and/or fence registers to make room. So performance may
- * suffer if the GTT working set is large or there are few fence registers
- * left.
- *
- * The current feature set supported by i915_gem_fault() and thus GTT mmaps
- * is exposed via I915_PARAM_MMAP_GTT_VERSION (see i915_gem_mmap_gtt_version).
- */
-vm_fault_t i915_gem_fault(struct vm_fault *vmf)
-{
-#define MIN_CHUNK_PAGES (SZ_1M >> PAGE_SHIFT)
- struct vm_area_struct *area = vmf->vma;
- struct drm_i915_gem_object *obj = to_intel_bo(area->vm_private_data);
- struct drm_device *dev = obj->base.dev;
- struct drm_i915_private *dev_priv = to_i915(dev);
- struct i915_ggtt *ggtt = &dev_priv->ggtt;
- bool write = area->vm_flags & VM_WRITE;
- intel_wakeref_t wakeref;
- struct i915_vma *vma;
- pgoff_t page_offset;
- int srcu;
- int ret;
-
- /* Sanity check that we allow writing into this object */
- if (i915_gem_object_is_readonly(obj) && write)
- return VM_FAULT_SIGBUS;
-
- /* We don't use vmf->pgoff since that has the fake offset */
- page_offset = (vmf->address - area->vm_start) >> PAGE_SHIFT;
-
- trace_i915_gem_object_fault(obj, page_offset, true, write);
-
- ret = i915_gem_object_pin_pages(obj);
- if (ret)
- goto err;
-
- wakeref = intel_runtime_pm_get(dev_priv);
-
- srcu = i915_reset_trylock(dev_priv);
- if (srcu < 0) {
- ret = srcu;
- goto err_rpm;
- }
-
- ret = i915_mutex_lock_interruptible(dev);
- if (ret)
- goto err_reset;
-
- /* Access to snoopable pages through the GTT is incoherent. */
- if (obj->cache_level != I915_CACHE_NONE && !HAS_LLC(dev_priv)) {
- ret = -EFAULT;
- goto err_unlock;
- }
-
- /* Now pin it into the GTT as needed */
- vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0,
- PIN_MAPPABLE |
- PIN_NONBLOCK |
- PIN_NONFAULT);
- if (IS_ERR(vma)) {
- /* Use a partial view if it is bigger than available space */
- struct i915_ggtt_view view =
- compute_partial_view(obj, page_offset, MIN_CHUNK_PAGES);
- unsigned int flags;
-
- flags = PIN_MAPPABLE;
- if (view.type == I915_GGTT_VIEW_NORMAL)
- flags |= PIN_NONBLOCK; /* avoid warnings for pinned */
-
- /*
- * Userspace is now writing through an untracked VMA, abandon
- * all hope that the hardware is able to track future writes.
- */
- obj->frontbuffer_ggtt_origin = ORIGIN_CPU;
-
- vma = i915_gem_object_ggtt_pin(obj, &view, 0, 0, flags);
- if (IS_ERR(vma) && !view.type) {
- flags = PIN_MAPPABLE;
- view.type = I915_GGTT_VIEW_PARTIAL;
- vma = i915_gem_object_ggtt_pin(obj, &view, 0, 0, flags);
- }
- }
- if (IS_ERR(vma)) {
- ret = PTR_ERR(vma);
- goto err_unlock;
- }
-
- ret = i915_vma_pin_fence(vma);
- if (ret)
- goto err_unpin;
-
- /* Finally, remap it using the new GTT offset */
- ret = remap_io_mapping(area,
- area->vm_start + (vma->ggtt_view.partial.offset << PAGE_SHIFT),
- (ggtt->gmadr.start + vma->node.start) >> PAGE_SHIFT,
- min_t(u64, vma->size, area->vm_end - area->vm_start),
- &ggtt->iomap);
- if (ret)
- goto err_fence;
-
- /* Mark as being mmapped into userspace for later revocation */
- assert_rpm_wakelock_held(dev_priv);
- if (!i915_vma_set_userfault(vma) && !obj->userfault_count++)
- list_add(&obj->userfault_link, &dev_priv->mm.userfault_list);
- GEM_BUG_ON(!obj->userfault_count);
-
- i915_vma_set_ggtt_write(vma);
-
-err_fence:
- i915_vma_unpin_fence(vma);
-err_unpin:
- __i915_vma_unpin(vma);
-err_unlock:
- mutex_unlock(&dev->struct_mutex);
-err_reset:
- i915_reset_unlock(dev_priv, srcu);
-err_rpm:
- intel_runtime_pm_put(dev_priv, wakeref);
- i915_gem_object_unpin_pages(obj);
-err:
- switch (ret) {
- case -EIO:
- /*
- * We eat errors when the gpu is terminally wedged to avoid
- * userspace unduly crashing (gl has no provisions for mmaps to
- * fail). But any other -EIO isn't ours (e.g. swap in failure)
- * and so needs to be reported.
- */
- if (!i915_terminally_wedged(dev_priv))
- return VM_FAULT_SIGBUS;
- /* else: fall through */
- case -EAGAIN:
- /*
- * EAGAIN means the gpu is hung and we'll wait for the error
- * handler to reset everything when re-faulting in
- * i915_mutex_lock_interruptible.
- */
- case 0:
- case -ERESTARTSYS:
- case -EINTR:
- case -EBUSY:
- /*
- * EBUSY is ok: this just means that another thread
- * already did the job.
- */
- return VM_FAULT_NOPAGE;
- case -ENOMEM:
- return VM_FAULT_OOM;
- case -ENOSPC:
- case -EFAULT:
- return VM_FAULT_SIGBUS;
- default:
- WARN_ONCE(ret, "unhandled error in i915_gem_fault: %i\n", ret);
- return VM_FAULT_SIGBUS;
- }
-}
-
-static void __i915_gem_object_release_mmap(struct drm_i915_gem_object *obj)
-{
- struct i915_vma *vma;
-
- GEM_BUG_ON(!obj->userfault_count);
-
- obj->userfault_count = 0;
- list_del(&obj->userfault_link);
- drm_vma_node_unmap(&obj->base.vma_node,
- obj->base.dev->anon_inode->i_mapping);
-
- for_each_ggtt_vma(vma, obj)
- i915_vma_unset_userfault(vma);
-}
-
-/**
- * i915_gem_release_mmap - remove physical page mappings
- * @obj: obj in question
- *
- * Preserve the reservation of the mmapping with the DRM core code, but
- * relinquish ownership of the pages back to the system.
- *
- * It is vital that we remove the page mapping if we have mapped a tiled
- * object through the GTT and then lose the fence register due to
- * resource pressure. Similarly if the object has been moved out of the
- * aperture, than pages mapped into userspace must be revoked. Removing the
- * mapping will then trigger a page fault on the next user access, allowing
- * fixup by i915_gem_fault().
- */
-void
-i915_gem_release_mmap(struct drm_i915_gem_object *obj)
-{
- struct drm_i915_private *i915 = to_i915(obj->base.dev);
- intel_wakeref_t wakeref;
-
- /* Serialisation between user GTT access and our code depends upon
- * revoking the CPU's PTE whilst the mutex is held. The next user
- * pagefault then has to wait until we release the mutex.
- *
- * Note that RPM complicates somewhat by adding an additional
- * requirement that operations to the GGTT be made holding the RPM
- * wakeref.
- */
- lockdep_assert_held(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(i915);
-
- if (!obj->userfault_count)
- goto out;
-
- __i915_gem_object_release_mmap(obj);
-
- /* Ensure that the CPU's PTE are revoked and there are not outstanding
- * memory transactions from userspace before we return. The TLB
- * flushing implied above by changing the PTE above *should* be
- * sufficient, an extra barrier here just provides us with a bit
- * of paranoid documentation about our requirement to serialise
- * memory writes before touching registers / GSM.
- */
- wmb();
-
-out:
- intel_runtime_pm_put(i915, wakeref);
-}
-
void i915_gem_runtime_suspend(struct drm_i915_private *dev_priv)
{
struct drm_i915_gem_object *obj, *on;
@@ -1782,78 +1321,6 @@ void i915_gem_runtime_suspend(struct drm_i915_private *dev_priv)
}
}
-static int i915_gem_object_create_mmap_offset(struct drm_i915_gem_object *obj)
-{
- struct drm_i915_private *dev_priv = to_i915(obj->base.dev);
- int err;
-
- err = drm_gem_create_mmap_offset(&obj->base);
- if (likely(!err))
- return 0;
-
- /* Attempt to reap some mmap space from dead objects */
- do {
- err = i915_gem_wait_for_idle(dev_priv,
- I915_WAIT_INTERRUPTIBLE,
- MAX_SCHEDULE_TIMEOUT);
- if (err)
- break;
-
- i915_gem_drain_freed_objects(dev_priv);
- err = drm_gem_create_mmap_offset(&obj->base);
- if (!err)
- break;
-
- } while (flush_delayed_work(&dev_priv->gem.retire_work));
-
- return err;
-}
-
-int
-i915_gem_mmap_gtt(struct drm_file *file,
- struct drm_device *dev,
- u32 handle,
- u64 *offset)
-{
- struct drm_i915_gem_object *obj;
- int ret;
-
- obj = i915_gem_object_lookup(file, handle);
- if (!obj)
- return -ENOENT;
-
- ret = i915_gem_object_create_mmap_offset(obj);
- if (ret == 0)
- *offset = drm_vma_node_offset_addr(&obj->base.vma_node);
-
- i915_gem_object_put(obj);
- return ret;
-}
-
-/**
- * i915_gem_mmap_gtt_ioctl - prepare an object for GTT mmap'ing
- * @dev: DRM device
- * @data: GTT mapping ioctl data
- * @file: GEM object info
- *
- * Simply returns the fake offset to userspace so it can mmap it.
- * The mmap call will end up in drm_gem_mmap(), which will set things
- * up so we can get faults in the handler above.
- *
- * The fault handler will take care of binding the object into the GTT
- * (since it may have been evicted to make room for something), allocating
- * a fence register, and mapping the appropriate aperture address into
- * userspace.
- */
-int
-i915_gem_mmap_gtt_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file)
-{
- struct drm_i915_gem_mmap_gtt *args = data;
-
- return i915_gem_mmap_gtt(file, dev, args->handle, &args->offset);
-}
-
bool i915_sg_trim(struct sg_table *orig_st)
{
struct sg_table new_st;
@@ -2057,7 +1524,7 @@ static void __i915_gem_object_flush_for_display(struct drm_i915_gem_object *obj)
* We manually flush the CPU domain so that we can override and
* force the flush for the display, and perform it asyncrhonously.
*/
- flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+ i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
if (obj->cache_dirty)
i915_gem_clflush_object(obj, I915_CLFLUSH_FORCE);
obj->write_domain = 0;
@@ -2111,7 +1578,7 @@ i915_gem_object_set_to_wc_domain(struct drm_i915_gem_object *obj, bool write)
if (ret)
return ret;
- flush_write_domain(obj, ~I915_GEM_DOMAIN_WC);
+ i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_WC);
/* Serialise direct access to this object with the barriers for
* coherent writes from the GPU, by effectively invalidating the
@@ -2173,7 +1640,7 @@ i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write)
if (ret)
return ret;
- flush_write_domain(obj, ~I915_GEM_DOMAIN_GTT);
+ i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_GTT);
/* Serialise direct access to this object with the barriers for
* coherent writes from the GPU, by effectively invalidating the
@@ -2282,7 +1749,7 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
* then double check if the GTT mapping is still
* valid for that pointer access.
*/
- i915_gem_release_mmap(obj);
+ i915_gem_object_release_mmap(obj);
/* As we no longer need a fence for GTT access,
* we can relinquish it now (and so prevent having
@@ -2537,7 +2004,7 @@ i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write)
if (ret)
return ret;
- flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+ i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
/* Flush the CPU cache if it's still invalid. */
if ((obj->read_domains & I915_GEM_DOMAIN_CPU) == 0) {
diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c
index 7d5cc9ccf6dd..86d6d92ccbc9 100644
--- a/drivers/gpu/drm/i915/i915_gem_tiling.c
+++ b/drivers/gpu/drm/i915/i915_gem_tiling.c
@@ -299,7 +299,7 @@ i915_gem_object_set_tiling(struct drm_i915_gem_object *obj,
i915_gem_object_unlock(obj);
/* Force the fence to be reacquired for GTT access */
- i915_gem_release_mmap(obj);
+ i915_gem_object_release_mmap(obj);
/* Try to preallocate memory required to save swizzling on put-pages */
if (i915_gem_object_needs_bit17_swizzle(obj)) {
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_object.c b/drivers/gpu/drm/i915/selftests/i915_gem_object.c
index b98a286a8be5..a3dd2f1be95b 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_object.c
@@ -89,491 +89,6 @@ static int igt_gem_huge(void *arg)
return err;
}
-struct tile {
- unsigned int width;
- unsigned int height;
- unsigned int stride;
- unsigned int size;
- unsigned int tiling;
- unsigned int swizzle;
-};
-
-static u64 swizzle_bit(unsigned int bit, u64 offset)
-{
- return (offset & BIT_ULL(bit)) >> (bit - 6);
-}
-
-static u64 tiled_offset(const struct tile *tile, u64 v)
-{
- u64 x, y;
-
- if (tile->tiling == I915_TILING_NONE)
- return v;
-
- y = div64_u64_rem(v, tile->stride, &x);
- v = div64_u64_rem(y, tile->height, &y) * tile->stride * tile->height;
-
- if (tile->tiling == I915_TILING_X) {
- v += y * tile->width;
- v += div64_u64_rem(x, tile->width, &x) << tile->size;
- v += x;
- } else if (tile->width == 128) {
- const unsigned int ytile_span = 16;
- const unsigned int ytile_height = 512;
-
- v += y * ytile_span;
- v += div64_u64_rem(x, ytile_span, &x) * ytile_height;
- v += x;
- } else {
- const unsigned int ytile_span = 32;
- const unsigned int ytile_height = 256;
-
- v += y * ytile_span;
- v += div64_u64_rem(x, ytile_span, &x) * ytile_height;
- v += x;
- }
-
- switch (tile->swizzle) {
- case I915_BIT_6_SWIZZLE_9:
- v ^= swizzle_bit(9, v);
- break;
- case I915_BIT_6_SWIZZLE_9_10:
- v ^= swizzle_bit(9, v) ^ swizzle_bit(10, v);
- break;
- case I915_BIT_6_SWIZZLE_9_11:
- v ^= swizzle_bit(9, v) ^ swizzle_bit(11, v);
- break;
- case I915_BIT_6_SWIZZLE_9_10_11:
- v ^= swizzle_bit(9, v) ^ swizzle_bit(10, v) ^ swizzle_bit(11, v);
- break;
- }
-
- return v;
-}
-
-static int check_partial_mapping(struct drm_i915_gem_object *obj,
- const struct tile *tile,
- unsigned long end_time)
-{
- const unsigned int nreal = obj->scratch / PAGE_SIZE;
- const unsigned long npages = obj->base.size / PAGE_SIZE;
- struct i915_vma *vma;
- unsigned long page;
- int err;
-
- if (igt_timeout(end_time,
- "%s: timed out before tiling=%d stride=%d\n",
- __func__, tile->tiling, tile->stride))
- return -EINTR;
-
- err = i915_gem_object_set_tiling(obj, tile->tiling, tile->stride);
- if (err) {
- pr_err("Failed to set tiling mode=%u, stride=%u, err=%d\n",
- tile->tiling, tile->stride, err);
- return err;
- }
-
- GEM_BUG_ON(i915_gem_object_get_tiling(obj) != tile->tiling);
- GEM_BUG_ON(i915_gem_object_get_stride(obj) != tile->stride);
-
- for_each_prime_number_from(page, 1, npages) {
- struct i915_ggtt_view view =
- compute_partial_view(obj, page, MIN_CHUNK_PAGES);
- u32 __iomem *io;
- struct page *p;
- unsigned int n;
- u64 offset;
- u32 *cpu;
-
- GEM_BUG_ON(view.partial.size > nreal);
- cond_resched();
-
- err = i915_gem_object_set_to_gtt_domain(obj, true);
- if (err) {
- pr_err("Failed to flush to GTT write domain; err=%d\n",
- err);
- return err;
- }
-
- vma = i915_gem_object_ggtt_pin(obj, &view, 0, 0, PIN_MAPPABLE);
- if (IS_ERR(vma)) {
- pr_err("Failed to pin partial view: offset=%lu; err=%d\n",
- page, (int)PTR_ERR(vma));
- return PTR_ERR(vma);
- }
-
- n = page - view.partial.offset;
- GEM_BUG_ON(n >= view.partial.size);
-
- io = i915_vma_pin_iomap(vma);
- i915_vma_unpin(vma);
- if (IS_ERR(io)) {
- pr_err("Failed to iomap partial view: offset=%lu; err=%d\n",
- page, (int)PTR_ERR(io));
- return PTR_ERR(io);
- }
-
- iowrite32(page, io + n * PAGE_SIZE/sizeof(*io));
- i915_vma_unpin_iomap(vma);
-
- offset = tiled_offset(tile, page << PAGE_SHIFT);
- if (offset >= obj->base.size)
- continue;
-
- flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
-
- p = i915_gem_object_get_page(obj, offset >> PAGE_SHIFT);
- cpu = kmap(p) + offset_in_page(offset);
- drm_clflush_virt_range(cpu, sizeof(*cpu));
- if (*cpu != (u32)page) {
- pr_err("Partial view for %lu [%u] (offset=%llu, size=%u [%llu, row size %u], fence=%d, tiling=%d, stride=%d) misalignment, expected write to page (%llu + %u [0x%llx]) of 0x%x, found 0x%x\n",
- page, n,
- view.partial.offset,
- view.partial.size,
- vma->size >> PAGE_SHIFT,
- tile->tiling ? tile_row_pages(obj) : 0,
- vma->fence ? vma->fence->id : -1, tile->tiling, tile->stride,
- offset >> PAGE_SHIFT,
- (unsigned int)offset_in_page(offset),
- offset,
- (u32)page, *cpu);
- err = -EINVAL;
- }
- *cpu = 0;
- drm_clflush_virt_range(cpu, sizeof(*cpu));
- kunmap(p);
- if (err)
- return err;
-
- i915_vma_destroy(vma);
- }
-
- return 0;
-}
-
-static int igt_partial_tiling(void *arg)
-{
- const unsigned int nreal = 1 << 12; /* largest tile row x2 */
- struct drm_i915_private *i915 = arg;
- struct drm_i915_gem_object *obj;
- intel_wakeref_t wakeref;
- int tiling;
- int err;
-
- /* We want to check the page mapping and fencing of a large object
- * mmapped through the GTT. The object we create is larger than can
- * possibly be mmaped as a whole, and so we must use partial GGTT vma.
- * We then check that a write through each partial GGTT vma ends up
- * in the right set of pages within the object, and with the expected
- * tiling, which we verify by manual swizzling.
- */
-
- obj = huge_gem_object(i915,
- nreal << PAGE_SHIFT,
- (1 + next_prime_number(i915->ggtt.vm.total >> PAGE_SHIFT)) << PAGE_SHIFT);
- if (IS_ERR(obj))
- return PTR_ERR(obj);
-
- err = i915_gem_object_pin_pages(obj);
- if (err) {
- pr_err("Failed to allocate %u pages (%lu total), err=%d\n",
- nreal, obj->base.size / PAGE_SIZE, err);
- goto out;
- }
-
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(i915);
-
- if (1) {
- IGT_TIMEOUT(end);
- struct tile tile;
-
- tile.height = 1;
- tile.width = 1;
- tile.size = 0;
- tile.stride = 0;
- tile.swizzle = I915_BIT_6_SWIZZLE_NONE;
- tile.tiling = I915_TILING_NONE;
-
- err = check_partial_mapping(obj, &tile, end);
- if (err && err != -EINTR)
- goto out_unlock;
- }
-
- for (tiling = I915_TILING_X; tiling <= I915_TILING_Y; tiling++) {
- IGT_TIMEOUT(end);
- unsigned int max_pitch;
- unsigned int pitch;
- struct tile tile;
-
- if (i915->quirks & QUIRK_PIN_SWIZZLED_PAGES)
- /*
- * The swizzling pattern is actually unknown as it
- * varies based on physical address of each page.
- * See i915_gem_detect_bit_6_swizzle().
- */
- break;
-
- tile.tiling = tiling;
- switch (tiling) {
- case I915_TILING_X:
- tile.swizzle = i915->mm.bit_6_swizzle_x;
- break;
- case I915_TILING_Y:
- tile.swizzle = i915->mm.bit_6_swizzle_y;
- break;
- }
-
- GEM_BUG_ON(tile.swizzle == I915_BIT_6_SWIZZLE_UNKNOWN);
- if (tile.swizzle == I915_BIT_6_SWIZZLE_9_17 ||
- tile.swizzle == I915_BIT_6_SWIZZLE_9_10_17)
- continue;
-
- if (INTEL_GEN(i915) <= 2) {
- tile.height = 16;
- tile.width = 128;
- tile.size = 11;
- } else if (tile.tiling == I915_TILING_Y &&
- HAS_128_BYTE_Y_TILING(i915)) {
- tile.height = 32;
- tile.width = 128;
- tile.size = 12;
- } else {
- tile.height = 8;
- tile.width = 512;
- tile.size = 12;
- }
-
- if (INTEL_GEN(i915) < 4)
- max_pitch = 8192 / tile.width;
- else if (INTEL_GEN(i915) < 7)
- max_pitch = 128 * I965_FENCE_MAX_PITCH_VAL / tile.width;
- else
- max_pitch = 128 * GEN7_FENCE_MAX_PITCH_VAL / tile.width;
-
- for (pitch = max_pitch; pitch; pitch >>= 1) {
- tile.stride = tile.width * pitch;
- err = check_partial_mapping(obj, &tile, end);
- if (err == -EINTR)
- goto next_tiling;
- if (err)
- goto out_unlock;
-
- if (pitch > 2 && INTEL_GEN(i915) >= 4) {
- tile.stride = tile.width * (pitch - 1);
- err = check_partial_mapping(obj, &tile, end);
- if (err == -EINTR)
- goto next_tiling;
- if (err)
- goto out_unlock;
- }
-
- if (pitch < max_pitch && INTEL_GEN(i915) >= 4) {
- tile.stride = tile.width * (pitch + 1);
- err = check_partial_mapping(obj, &tile, end);
- if (err == -EINTR)
- goto next_tiling;
- if (err)
- goto out_unlock;
- }
- }
-
- if (INTEL_GEN(i915) >= 4) {
- for_each_prime_number(pitch, max_pitch) {
- tile.stride = tile.width * pitch;
- err = check_partial_mapping(obj, &tile, end);
- if (err == -EINTR)
- goto next_tiling;
- if (err)
- goto out_unlock;
- }
- }
-
-next_tiling: ;
- }
-
-out_unlock:
- intel_runtime_pm_put(i915, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
- i915_gem_object_unpin_pages(obj);
-out:
- i915_gem_object_put(obj);
- return err;
-}
-
-static int make_obj_busy(struct drm_i915_gem_object *obj)
-{
- struct drm_i915_private *i915 = to_i915(obj->base.dev);
- struct i915_request *rq;
- struct i915_vma *vma;
- int err;
-
- vma = i915_vma_instance(obj, &i915->ggtt.vm, NULL);
- if (IS_ERR(vma))
- return PTR_ERR(vma);
-
- err = i915_vma_pin(vma, 0, 0, PIN_USER);
- if (err)
- return err;
-
- rq = i915_request_create(i915->engine[RCS0]->kernel_context);
- if (IS_ERR(rq)) {
- i915_vma_unpin(vma);
- return PTR_ERR(rq);
- }
-
- err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
-
- i915_request_add(rq);
-
- __i915_gem_object_release_unless_active(obj);
- i915_vma_unpin(vma);
-
- return err;
-}
-
-static bool assert_mmap_offset(struct drm_i915_private *i915,
- unsigned long size,
- int expected)
-{
- struct drm_i915_gem_object *obj;
- int err;
-
- obj = i915_gem_object_create_internal(i915, size);
- if (IS_ERR(obj))
- return PTR_ERR(obj);
-
- err = i915_gem_object_create_mmap_offset(obj);
- i915_gem_object_put(obj);
-
- return err == expected;
-}
-
-static void disable_retire_worker(struct drm_i915_private *i915)
-{
- i915_gem_shrinker_unregister(i915);
-
- intel_gt_pm_get(i915);
-
- cancel_delayed_work_sync(&i915->gem.retire_work);
- flush_work(&i915->gem.idle_work);
-}
-
-static void restore_retire_worker(struct drm_i915_private *i915)
-{
- intel_gt_pm_put(i915);
-
- mutex_lock(&i915->drm.struct_mutex);
- igt_flush_test(i915, I915_WAIT_LOCKED);
- mutex_unlock(&i915->drm.struct_mutex);
-
- i915_gem_shrinker_register(i915);
-}
-
-static int igt_mmap_offset_exhaustion(void *arg)
-{
- struct drm_i915_private *i915 = arg;
- struct drm_mm *mm = &i915->drm.vma_offset_manager->vm_addr_space_mm;
- struct drm_i915_gem_object *obj;
- struct drm_mm_node resv, *hole;
- u64 hole_start, hole_end;
- int loop, err;
-
- /* Disable background reaper */
- disable_retire_worker(i915);
- GEM_BUG_ON(!i915->gt.awake);
-
- /* Trim the device mmap space to only a page */
- memset(&resv, 0, sizeof(resv));
- drm_mm_for_each_hole(hole, mm, hole_start, hole_end) {
- resv.start = hole_start;
- resv.size = hole_end - hole_start - 1; /* PAGE_SIZE units */
- err = drm_mm_reserve_node(mm, &resv);
- if (err) {
- pr_err("Failed to trim VMA manager, err=%d\n", err);
- goto out_park;
- }
- break;
- }
-
- /* Just fits! */
- if (!assert_mmap_offset(i915, PAGE_SIZE, 0)) {
- pr_err("Unable to insert object into single page hole\n");
- err = -EINVAL;
- goto out;
- }
-
- /* Too large */
- if (!assert_mmap_offset(i915, 2*PAGE_SIZE, -ENOSPC)) {
- pr_err("Unexpectedly succeeded in inserting too large object into single page hole\n");
- err = -EINVAL;
- goto out;
- }
-
- /* Fill the hole, further allocation attempts should then fail */
- obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
- if (IS_ERR(obj)) {
- err = PTR_ERR(obj);
- goto out;
- }
-
- err = i915_gem_object_create_mmap_offset(obj);
- if (err) {
- pr_err("Unable to insert object into reclaimed hole\n");
- goto err_obj;
- }
-
- if (!assert_mmap_offset(i915, PAGE_SIZE, -ENOSPC)) {
- pr_err("Unexpectedly succeeded in inserting object into no holes!\n");
- err = -EINVAL;
- goto err_obj;
- }
-
- i915_gem_object_put(obj);
-
- /* Now fill with busy dead objects that we expect to reap */
- for (loop = 0; loop < 3; loop++) {
- intel_wakeref_t wakeref;
-
- if (i915_terminally_wedged(i915))
- break;
-
- obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
- if (IS_ERR(obj)) {
- err = PTR_ERR(obj);
- goto out;
- }
-
- err = 0;
- mutex_lock(&i915->drm.struct_mutex);
- with_intel_runtime_pm(i915, wakeref)
- err = make_obj_busy(obj);
- mutex_unlock(&i915->drm.struct_mutex);
- if (err) {
- pr_err("[loop %d] Failed to busy the object\n", loop);
- goto err_obj;
- }
-
- /* NB we rely on the _active_ reference to access obj now */
- GEM_BUG_ON(!i915_gem_object_is_active(obj));
- err = i915_gem_object_create_mmap_offset(obj);
- if (err) {
- pr_err("[loop %d] i915_gem_object_create_mmap_offset failed with err=%d\n",
- loop, err);
- goto out;
- }
- }
-
-out:
- drm_mm_remove_node(&resv);
-out_park:
- restore_retire_worker(i915);
- return err;
-err_obj:
- i915_gem_object_put(obj);
- goto out;
-}
-
int i915_gem_object_mock_selftests(void)
{
static const struct i915_subtest tests[] = {
@@ -596,8 +111,6 @@ int i915_gem_object_live_selftests(struct drm_i915_private *i915)
{
static const struct i915_subtest tests[] = {
SUBTEST(igt_gem_huge),
- SUBTEST(igt_partial_tiling),
- SUBTEST(igt_mmap_offset_exhaustion),
};
return i915_subtests(tests, i915);
diff --git a/drivers/gpu/drm/i915/selftests/i915_live_selftests.h b/drivers/gpu/drm/i915/selftests/i915_live_selftests.h
index 6d766925ad04..bbf387de8db3 100644
--- a/drivers/gpu/drm/i915/selftests/i915_live_selftests.h
+++ b/drivers/gpu/drm/i915/selftests/i915_live_selftests.h
@@ -16,6 +16,7 @@ selftest(timelines, i915_timeline_live_selftests)
selftest(requests, i915_request_live_selftests)
selftest(active, i915_active_live_selftests)
selftest(objects, i915_gem_object_live_selftests)
+selftest(mman, i915_gem_mman_live_selftests)
selftest(dmabuf, i915_gem_dmabuf_live_selftests)
selftest(coherency, i915_gem_coherency_live_selftests)
selftest(gtt, i915_gem_gtt_live_selftests)
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 25/40] drm/i915: Move GEM domain management to its own file
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (22 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 24/40] drm/i915: Move mmap and friends " Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 26/40] drm/i915: Move more GEM objects under gem/ Chris Wilson
` (21 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx; +Cc: Matthew Auld
Continuing the decluttering of i915_gem.c, that of the read/write
domains, perhaps the biggest of GEM's follies?
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
---
drivers/gpu/drm/i915/Makefile | 1 +
drivers/gpu/drm/i915/gem/i915_gem_domain.c | 784 ++++++++++++++++++
drivers/gpu/drm/i915/gem/i915_gem_object.h | 29 +
drivers/gpu/drm/i915/gvt/cmd_parser.c | 4 +-
drivers/gpu/drm/i915/gvt/scheduler.c | 6 +-
drivers/gpu/drm/i915/i915_cmd_parser.c | 8 +-
drivers/gpu/drm/i915/i915_drv.h | 26 -
drivers/gpu/drm/i915/i915_gem.c | 777 +----------------
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 4 +-
drivers/gpu/drm/i915/i915_gem_render_state.c | 4 +-
drivers/gpu/drm/i915/selftests/huge_pages.c | 4 +-
.../drm/i915/selftests/i915_gem_coherency.c | 8 +-
.../gpu/drm/i915/selftests/i915_gem_context.c | 8 +-
13 files changed, 841 insertions(+), 822 deletions(-)
create mode 100644 drivers/gpu/drm/i915/gem/i915_gem_domain.c
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index d05757c52492..29fc924ba819 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -87,6 +87,7 @@ i915-y += $(gt-y)
# GEM (Graphics Execution Management) code
obj-y += gem/
gem-y += \
+ gem/i915_gem_domain.o \
gem/i915_gem_object.o \
gem/i915_gem_mman.o \
gem/i915_gem_pages.o \
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_domain.c b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
new file mode 100644
index 000000000000..eee421e3021c
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
@@ -0,0 +1,784 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2014-2016 Intel Corporation
+ */
+
+#include "i915_gem_ioctls.h"
+#include "i915_gem_object.h"
+
+#include "../i915_drv.h"
+#include "../i915_gem_clflush.h"
+#include "../i915_gem_gtt.h"
+#include "../i915_vma.h"
+
+#include "../intel_frontbuffer.h"
+
+static void __i915_gem_object_flush_for_display(struct drm_i915_gem_object *obj)
+{
+ /*
+ * We manually flush the CPU domain so that we can override and
+ * force the flush for the display, and perform it asyncrhonously.
+ */
+ i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+ if (obj->cache_dirty)
+ i915_gem_clflush_object(obj, I915_CLFLUSH_FORCE);
+ obj->write_domain = 0;
+}
+
+void i915_gem_object_flush_if_display(struct drm_i915_gem_object *obj)
+{
+ if (!READ_ONCE(obj->pin_global))
+ return;
+
+ mutex_lock(&obj->base.dev->struct_mutex);
+ __i915_gem_object_flush_for_display(obj);
+ mutex_unlock(&obj->base.dev->struct_mutex);
+}
+
+/**
+ * Moves a single object to the WC read, and possibly write domain.
+ * @obj: object to act on
+ * @write: ask for write access or read only
+ *
+ * This function returns when the move is complete, including waiting on
+ * flushes to occur.
+ */
+int
+i915_gem_object_set_to_wc_domain(struct drm_i915_gem_object *obj, bool write)
+{
+ int ret;
+
+ lockdep_assert_held(&obj->base.dev->struct_mutex);
+
+ ret = i915_gem_object_wait(obj,
+ I915_WAIT_INTERRUPTIBLE |
+ I915_WAIT_LOCKED |
+ (write ? I915_WAIT_ALL : 0),
+ MAX_SCHEDULE_TIMEOUT);
+ if (ret)
+ return ret;
+
+ if (obj->write_domain == I915_GEM_DOMAIN_WC)
+ return 0;
+
+ /* Flush and acquire obj->pages so that we are coherent through
+ * direct access in memory with previous cached writes through
+ * shmemfs and that our cache domain tracking remains valid.
+ * For example, if the obj->filp was moved to swap without us
+ * being notified and releasing the pages, we would mistakenly
+ * continue to assume that the obj remained out of the CPU cached
+ * domain.
+ */
+ ret = i915_gem_object_pin_pages(obj);
+ if (ret)
+ return ret;
+
+ i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_WC);
+
+ /* Serialise direct access to this object with the barriers for
+ * coherent writes from the GPU, by effectively invalidating the
+ * WC domain upon first access.
+ */
+ if ((obj->read_domains & I915_GEM_DOMAIN_WC) == 0)
+ mb();
+
+ /* It should now be out of any other write domains, and we can update
+ * the domain values for our changes.
+ */
+ GEM_BUG_ON((obj->write_domain & ~I915_GEM_DOMAIN_WC) != 0);
+ obj->read_domains |= I915_GEM_DOMAIN_WC;
+ if (write) {
+ obj->read_domains = I915_GEM_DOMAIN_WC;
+ obj->write_domain = I915_GEM_DOMAIN_WC;
+ obj->mm.dirty = true;
+ }
+
+ i915_gem_object_unpin_pages(obj);
+ return 0;
+}
+
+/**
+ * Moves a single object to the GTT read, and possibly write domain.
+ * @obj: object to act on
+ * @write: ask for write access or read only
+ *
+ * This function returns when the move is complete, including waiting on
+ * flushes to occur.
+ */
+int
+i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write)
+{
+ int ret;
+
+ lockdep_assert_held(&obj->base.dev->struct_mutex);
+
+ ret = i915_gem_object_wait(obj,
+ I915_WAIT_INTERRUPTIBLE |
+ I915_WAIT_LOCKED |
+ (write ? I915_WAIT_ALL : 0),
+ MAX_SCHEDULE_TIMEOUT);
+ if (ret)
+ return ret;
+
+ if (obj->write_domain == I915_GEM_DOMAIN_GTT)
+ return 0;
+
+ /* Flush and acquire obj->pages so that we are coherent through
+ * direct access in memory with previous cached writes through
+ * shmemfs and that our cache domain tracking remains valid.
+ * For example, if the obj->filp was moved to swap without us
+ * being notified and releasing the pages, we would mistakenly
+ * continue to assume that the obj remained out of the CPU cached
+ * domain.
+ */
+ ret = i915_gem_object_pin_pages(obj);
+ if (ret)
+ return ret;
+
+ i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_GTT);
+
+ /* Serialise direct access to this object with the barriers for
+ * coherent writes from the GPU, by effectively invalidating the
+ * GTT domain upon first access.
+ */
+ if ((obj->read_domains & I915_GEM_DOMAIN_GTT) == 0)
+ mb();
+
+ /* It should now be out of any other write domains, and we can update
+ * the domain values for our changes.
+ */
+ GEM_BUG_ON((obj->write_domain & ~I915_GEM_DOMAIN_GTT) != 0);
+ obj->read_domains |= I915_GEM_DOMAIN_GTT;
+ if (write) {
+ obj->read_domains = I915_GEM_DOMAIN_GTT;
+ obj->write_domain = I915_GEM_DOMAIN_GTT;
+ obj->mm.dirty = true;
+ }
+
+ i915_gem_object_unpin_pages(obj);
+ return 0;
+}
+
+/**
+ * Changes the cache-level of an object across all VMA.
+ * @obj: object to act on
+ * @cache_level: new cache level to set for the object
+ *
+ * After this function returns, the object will be in the new cache-level
+ * across all GTT and the contents of the backing storage will be coherent,
+ * with respect to the new cache-level. In order to keep the backing storage
+ * coherent for all users, we only allow a single cache level to be set
+ * globally on the object and prevent it from being changed whilst the
+ * hardware is reading from the object. That is if the object is currently
+ * on the scanout it will be set to uncached (or equivalent display
+ * cache coherency) and all non-MOCS GPU access will also be uncached so
+ * that all direct access to the scanout remains coherent.
+ */
+int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
+ enum i915_cache_level cache_level)
+{
+ struct i915_vma *vma;
+ int ret;
+
+ lockdep_assert_held(&obj->base.dev->struct_mutex);
+
+ if (obj->cache_level == cache_level)
+ return 0;
+
+ /* Inspect the list of currently bound VMA and unbind any that would
+ * be invalid given the new cache-level. This is principally to
+ * catch the issue of the CS prefetch crossing page boundaries and
+ * reading an invalid PTE on older architectures.
+ */
+restart:
+ list_for_each_entry(vma, &obj->vma.list, obj_link) {
+ if (!drm_mm_node_allocated(&vma->node))
+ continue;
+
+ if (i915_vma_is_pinned(vma)) {
+ DRM_DEBUG("can not change the cache level of pinned objects\n");
+ return -EBUSY;
+ }
+
+ if (!i915_vma_is_closed(vma) &&
+ i915_gem_valid_gtt_space(vma, cache_level))
+ continue;
+
+ ret = i915_vma_unbind(vma);
+ if (ret)
+ return ret;
+
+ /* As unbinding may affect other elements in the
+ * obj->vma_list (due to side-effects from retiring
+ * an active vma), play safe and restart the iterator.
+ */
+ goto restart;
+ }
+
+ /* We can reuse the existing drm_mm nodes but need to change the
+ * cache-level on the PTE. We could simply unbind them all and
+ * rebind with the correct cache-level on next use. However since
+ * we already have a valid slot, dma mapping, pages etc, we may as
+ * rewrite the PTE in the belief that doing so tramples upon less
+ * state and so involves less work.
+ */
+ if (obj->bind_count) {
+ /* Before we change the PTE, the GPU must not be accessing it.
+ * If we wait upon the object, we know that all the bound
+ * VMA are no longer active.
+ */
+ ret = i915_gem_object_wait(obj,
+ I915_WAIT_INTERRUPTIBLE |
+ I915_WAIT_LOCKED |
+ I915_WAIT_ALL,
+ MAX_SCHEDULE_TIMEOUT);
+ if (ret)
+ return ret;
+
+ if (!HAS_LLC(to_i915(obj->base.dev)) &&
+ cache_level != I915_CACHE_NONE) {
+ /* Access to snoopable pages through the GTT is
+ * incoherent and on some machines causes a hard
+ * lockup. Relinquish the CPU mmaping to force
+ * userspace to refault in the pages and we can
+ * then double check if the GTT mapping is still
+ * valid for that pointer access.
+ */
+ i915_gem_object_release_mmap(obj);
+
+ /* As we no longer need a fence for GTT access,
+ * we can relinquish it now (and so prevent having
+ * to steal a fence from someone else on the next
+ * fence request). Note GPU activity would have
+ * dropped the fence as all snoopable access is
+ * supposed to be linear.
+ */
+ for_each_ggtt_vma(vma, obj) {
+ ret = i915_vma_put_fence(vma);
+ if (ret)
+ return ret;
+ }
+ } else {
+ /* We either have incoherent backing store and
+ * so no GTT access or the architecture is fully
+ * coherent. In such cases, existing GTT mmaps
+ * ignore the cache bit in the PTE and we can
+ * rewrite it without confusing the GPU or having
+ * to force userspace to fault back in its mmaps.
+ */
+ }
+
+ list_for_each_entry(vma, &obj->vma.list, obj_link) {
+ if (!drm_mm_node_allocated(&vma->node))
+ continue;
+
+ ret = i915_vma_bind(vma, cache_level, PIN_UPDATE);
+ if (ret)
+ return ret;
+ }
+ }
+
+ list_for_each_entry(vma, &obj->vma.list, obj_link)
+ vma->node.color = cache_level;
+ i915_gem_object_set_cache_coherency(obj, cache_level);
+ obj->cache_dirty = true; /* Always invalidate stale cachelines */
+
+ return 0;
+}
+
+int i915_gem_get_caching_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file)
+{
+ struct drm_i915_gem_caching *args = data;
+ struct drm_i915_gem_object *obj;
+ int err = 0;
+
+ rcu_read_lock();
+ obj = i915_gem_object_lookup_rcu(file, args->handle);
+ if (!obj) {
+ err = -ENOENT;
+ goto out;
+ }
+
+ switch (obj->cache_level) {
+ case I915_CACHE_LLC:
+ case I915_CACHE_L3_LLC:
+ args->caching = I915_CACHING_CACHED;
+ break;
+
+ case I915_CACHE_WT:
+ args->caching = I915_CACHING_DISPLAY;
+ break;
+
+ default:
+ args->caching = I915_CACHING_NONE;
+ break;
+ }
+out:
+ rcu_read_unlock();
+ return err;
+}
+
+int i915_gem_set_caching_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file)
+{
+ struct drm_i915_private *i915 = to_i915(dev);
+ struct drm_i915_gem_caching *args = data;
+ struct drm_i915_gem_object *obj;
+ enum i915_cache_level level;
+ int ret = 0;
+
+ switch (args->caching) {
+ case I915_CACHING_NONE:
+ level = I915_CACHE_NONE;
+ break;
+ case I915_CACHING_CACHED:
+ /*
+ * Due to a HW issue on BXT A stepping, GPU stores via a
+ * snooped mapping may leave stale data in a corresponding CPU
+ * cacheline, whereas normally such cachelines would get
+ * invalidated.
+ */
+ if (!HAS_LLC(i915) && !HAS_SNOOP(i915))
+ return -ENODEV;
+
+ level = I915_CACHE_LLC;
+ break;
+ case I915_CACHING_DISPLAY:
+ level = HAS_WT(i915) ? I915_CACHE_WT : I915_CACHE_NONE;
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ obj = i915_gem_object_lookup(file, args->handle);
+ if (!obj)
+ return -ENOENT;
+
+ /*
+ * The caching mode of proxy object is handled by its generator, and
+ * not allowed to be changed by userspace.
+ */
+ if (i915_gem_object_is_proxy(obj)) {
+ ret = -ENXIO;
+ goto out;
+ }
+
+ if (obj->cache_level == level)
+ goto out;
+
+ ret = i915_gem_object_wait(obj,
+ I915_WAIT_INTERRUPTIBLE,
+ MAX_SCHEDULE_TIMEOUT);
+ if (ret)
+ goto out;
+
+ ret = i915_mutex_lock_interruptible(dev);
+ if (ret)
+ goto out;
+
+ ret = i915_gem_object_set_cache_level(obj, level);
+ mutex_unlock(&dev->struct_mutex);
+
+out:
+ i915_gem_object_put(obj);
+ return ret;
+}
+
+/*
+ * Prepare buffer for display plane (scanout, cursors, etc). Can be called from
+ * an uninterruptible phase (modesetting) and allows any flushes to be pipelined
+ * (for pageflips). We only flush the caches while preparing the buffer for
+ * display, the callers are responsible for frontbuffer flush.
+ */
+struct i915_vma *
+i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,
+ u32 alignment,
+ const struct i915_ggtt_view *view,
+ unsigned int flags)
+{
+ struct i915_vma *vma;
+ int ret;
+
+ lockdep_assert_held(&obj->base.dev->struct_mutex);
+
+ /* Mark the global pin early so that we account for the
+ * display coherency whilst setting up the cache domains.
+ */
+ obj->pin_global++;
+
+ /* The display engine is not coherent with the LLC cache on gen6. As
+ * a result, we make sure that the pinning that is about to occur is
+ * done with uncached PTEs. This is lowest common denominator for all
+ * chipsets.
+ *
+ * However for gen6+, we could do better by using the GFDT bit instead
+ * of uncaching, which would allow us to flush all the LLC-cached data
+ * with that bit in the PTE to main memory with just one PIPE_CONTROL.
+ */
+ ret = i915_gem_object_set_cache_level(obj,
+ HAS_WT(to_i915(obj->base.dev)) ?
+ I915_CACHE_WT : I915_CACHE_NONE);
+ if (ret) {
+ vma = ERR_PTR(ret);
+ goto err_unpin_global;
+ }
+
+ /* As the user may map the buffer once pinned in the display plane
+ * (e.g. libkms for the bootup splash), we have to ensure that we
+ * always use map_and_fenceable for all scanout buffers. However,
+ * it may simply be too big to fit into mappable, in which case
+ * put it anyway and hope that userspace can cope (but always first
+ * try to preserve the existing ABI).
+ */
+ vma = ERR_PTR(-ENOSPC);
+ if ((flags & PIN_MAPPABLE) == 0 &&
+ (!view || view->type == I915_GGTT_VIEW_NORMAL))
+ vma = i915_gem_object_ggtt_pin(obj, view, 0, alignment,
+ flags |
+ PIN_MAPPABLE |
+ PIN_NONBLOCK);
+ if (IS_ERR(vma))
+ vma = i915_gem_object_ggtt_pin(obj, view, 0, alignment, flags);
+ if (IS_ERR(vma))
+ goto err_unpin_global;
+
+ vma->display_alignment = max_t(u64, vma->display_alignment, alignment);
+
+ __i915_gem_object_flush_for_display(obj);
+
+ /* It should now be out of any other write domains, and we can update
+ * the domain values for our changes.
+ */
+ obj->read_domains |= I915_GEM_DOMAIN_GTT;
+
+ return vma;
+
+err_unpin_global:
+ obj->pin_global--;
+ return vma;
+}
+
+static void i915_gem_object_bump_inactive_ggtt(struct drm_i915_gem_object *obj)
+{
+ struct drm_i915_private *i915 = to_i915(obj->base.dev);
+ struct list_head *list;
+ struct i915_vma *vma;
+
+ GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj));
+
+ mutex_lock(&i915->ggtt.vm.mutex);
+ for_each_ggtt_vma(vma, obj) {
+ if (!drm_mm_node_allocated(&vma->node))
+ continue;
+
+ list_move_tail(&vma->vm_link, &vma->vm->bound_list);
+ }
+ mutex_unlock(&i915->ggtt.vm.mutex);
+
+ spin_lock(&i915->mm.obj_lock);
+ list = obj->bind_count ? &i915->mm.bound_list : &i915->mm.unbound_list;
+ list_move_tail(&obj->mm.link, list);
+ spin_unlock(&i915->mm.obj_lock);
+}
+
+void
+i915_gem_object_unpin_from_display_plane(struct i915_vma *vma)
+{
+ lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
+
+ if (WARN_ON(vma->obj->pin_global == 0))
+ return;
+
+ if (--vma->obj->pin_global == 0)
+ vma->display_alignment = I915_GTT_MIN_ALIGNMENT;
+
+ /* Bump the LRU to try and avoid premature eviction whilst flipping */
+ i915_gem_object_bump_inactive_ggtt(vma->obj);
+
+ i915_vma_unpin(vma);
+}
+
+/**
+ * Moves a single object to the CPU read, and possibly write domain.
+ * @obj: object to act on
+ * @write: requesting write or read-only access
+ *
+ * This function returns when the move is complete, including waiting on
+ * flushes to occur.
+ */
+int
+i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write)
+{
+ int ret;
+
+ lockdep_assert_held(&obj->base.dev->struct_mutex);
+
+ ret = i915_gem_object_wait(obj,
+ I915_WAIT_INTERRUPTIBLE |
+ I915_WAIT_LOCKED |
+ (write ? I915_WAIT_ALL : 0),
+ MAX_SCHEDULE_TIMEOUT);
+ if (ret)
+ return ret;
+
+ i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+
+ /* Flush the CPU cache if it's still invalid. */
+ if ((obj->read_domains & I915_GEM_DOMAIN_CPU) == 0) {
+ i915_gem_clflush_object(obj, I915_CLFLUSH_SYNC);
+ obj->read_domains |= I915_GEM_DOMAIN_CPU;
+ }
+
+ /* It should now be out of any other write domains, and we can update
+ * the domain values for our changes.
+ */
+ GEM_BUG_ON(obj->write_domain & ~I915_GEM_DOMAIN_CPU);
+
+ /* If we're writing through the CPU, then the GPU read domains will
+ * need to be invalidated at next use.
+ */
+ if (write)
+ __start_cpu_write(obj);
+
+ return 0;
+}
+
+static inline enum fb_op_origin
+fb_write_origin(struct drm_i915_gem_object *obj, unsigned int domain)
+{
+ return (domain == I915_GEM_DOMAIN_GTT ?
+ obj->frontbuffer_ggtt_origin : ORIGIN_CPU);
+}
+
+/**
+ * Called when user space prepares to use an object with the CPU, either
+ * through the mmap ioctl's mapping or a GTT mapping.
+ * @dev: drm device
+ * @data: ioctl data blob
+ * @file: drm file
+ */
+int
+i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file)
+{
+ struct drm_i915_gem_set_domain *args = data;
+ struct drm_i915_gem_object *obj;
+ u32 read_domains = args->read_domains;
+ u32 write_domain = args->write_domain;
+ int err;
+
+ /* Only handle setting domains to types used by the CPU. */
+ if ((write_domain | read_domains) & I915_GEM_GPU_DOMAINS)
+ return -EINVAL;
+
+ /*
+ * Having something in the write domain implies it's in the read
+ * domain, and only that read domain. Enforce that in the request.
+ */
+ if (write_domain && read_domains != write_domain)
+ return -EINVAL;
+
+ if (!read_domains)
+ return 0;
+
+ obj = i915_gem_object_lookup(file, args->handle);
+ if (!obj)
+ return -ENOENT;
+
+ /*
+ * Already in the desired write domain? Nothing for us to do!
+ *
+ * We apply a little bit of cunning here to catch a broader set of
+ * no-ops. If obj->write_domain is set, we must be in the same
+ * obj->read_domains, and only that domain. Therefore, if that
+ * obj->write_domain matches the request read_domains, we are
+ * already in the same read/write domain and can skip the operation,
+ * without having to further check the requested write_domain.
+ */
+ if (READ_ONCE(obj->write_domain) == read_domains) {
+ err = 0;
+ goto out;
+ }
+
+ /*
+ * Try to flush the object off the GPU without holding the lock.
+ * We will repeat the flush holding the lock in the normal manner
+ * to catch cases where we are gazumped.
+ */
+ err = i915_gem_object_wait(obj,
+ I915_WAIT_INTERRUPTIBLE |
+ I915_WAIT_PRIORITY |
+ (write_domain ? I915_WAIT_ALL : 0),
+ MAX_SCHEDULE_TIMEOUT);
+ if (err)
+ goto out;
+
+ /*
+ * Proxy objects do not control access to the backing storage, ergo
+ * they cannot be used as a means to manipulate the cache domain
+ * tracking for that backing storage. The proxy object is always
+ * considered to be outside of any cache domain.
+ */
+ if (i915_gem_object_is_proxy(obj)) {
+ err = -ENXIO;
+ goto out;
+ }
+
+ /*
+ * Flush and acquire obj->pages so that we are coherent through
+ * direct access in memory with previous cached writes through
+ * shmemfs and that our cache domain tracking remains valid.
+ * For example, if the obj->filp was moved to swap without us
+ * being notified and releasing the pages, we would mistakenly
+ * continue to assume that the obj remained out of the CPU cached
+ * domain.
+ */
+ err = i915_gem_object_pin_pages(obj);
+ if (err)
+ goto out;
+
+ err = i915_mutex_lock_interruptible(dev);
+ if (err)
+ goto out_unpin;
+
+ if (read_domains & I915_GEM_DOMAIN_WC)
+ err = i915_gem_object_set_to_wc_domain(obj, write_domain);
+ else if (read_domains & I915_GEM_DOMAIN_GTT)
+ err = i915_gem_object_set_to_gtt_domain(obj, write_domain);
+ else
+ err = i915_gem_object_set_to_cpu_domain(obj, write_domain);
+
+ /* And bump the LRU for this access */
+ i915_gem_object_bump_inactive_ggtt(obj);
+
+ mutex_unlock(&dev->struct_mutex);
+
+ if (write_domain != 0)
+ intel_fb_obj_invalidate(obj,
+ fb_write_origin(obj, write_domain));
+
+out_unpin:
+ i915_gem_object_unpin_pages(obj);
+out:
+ i915_gem_object_put(obj);
+ return err;
+}
+
+/*
+ * Pins the specified object's pages and synchronizes the object with
+ * GPU accesses. Sets needs_clflush to non-zero if the caller should
+ * flush the object from the CPU cache.
+ */
+int i915_gem_object_prepare_read(struct drm_i915_gem_object *obj,
+ unsigned int *needs_clflush)
+{
+ int ret;
+
+ lockdep_assert_held(&obj->base.dev->struct_mutex);
+
+ *needs_clflush = 0;
+ if (!i915_gem_object_has_struct_page(obj))
+ return -ENODEV;
+
+ ret = i915_gem_object_wait(obj,
+ I915_WAIT_INTERRUPTIBLE |
+ I915_WAIT_LOCKED,
+ MAX_SCHEDULE_TIMEOUT);
+ if (ret)
+ return ret;
+
+ ret = i915_gem_object_pin_pages(obj);
+ if (ret)
+ return ret;
+
+ if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ ||
+ !static_cpu_has(X86_FEATURE_CLFLUSH)) {
+ ret = i915_gem_object_set_to_cpu_domain(obj, false);
+ if (ret)
+ goto err_unpin;
+ else
+ goto out;
+ }
+
+ i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+
+ /* If we're not in the cpu read domain, set ourself into the gtt
+ * read domain and manually flush cachelines (if required). This
+ * optimizes for the case when the gpu will dirty the data
+ * anyway again before the next pread happens.
+ */
+ if (!obj->cache_dirty &&
+ !(obj->read_domains & I915_GEM_DOMAIN_CPU))
+ *needs_clflush = CLFLUSH_BEFORE;
+
+out:
+ /* return with the pages pinned */
+ return 0;
+
+err_unpin:
+ i915_gem_object_unpin_pages(obj);
+ return ret;
+}
+
+int i915_gem_object_prepare_write(struct drm_i915_gem_object *obj,
+ unsigned int *needs_clflush)
+{
+ int ret;
+
+ lockdep_assert_held(&obj->base.dev->struct_mutex);
+
+ *needs_clflush = 0;
+ if (!i915_gem_object_has_struct_page(obj))
+ return -ENODEV;
+
+ ret = i915_gem_object_wait(obj,
+ I915_WAIT_INTERRUPTIBLE |
+ I915_WAIT_LOCKED |
+ I915_WAIT_ALL,
+ MAX_SCHEDULE_TIMEOUT);
+ if (ret)
+ return ret;
+
+ ret = i915_gem_object_pin_pages(obj);
+ if (ret)
+ return ret;
+
+ if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE ||
+ !static_cpu_has(X86_FEATURE_CLFLUSH)) {
+ ret = i915_gem_object_set_to_cpu_domain(obj, true);
+ if (ret)
+ goto err_unpin;
+ else
+ goto out;
+ }
+
+ i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+
+ /* If we're not in the cpu write domain, set ourself into the
+ * gtt write domain and manually flush cachelines (as required).
+ * This optimizes for the case when the gpu will use the data
+ * right away and we therefore have to clflush anyway.
+ */
+ if (!obj->cache_dirty) {
+ *needs_clflush |= CLFLUSH_AFTER;
+
+ /*
+ * Same trick applies to invalidate partially written
+ * cachelines read before writing.
+ */
+ if (!(obj->read_domains & I915_GEM_DOMAIN_CPU))
+ *needs_clflush |= CLFLUSH_BEFORE;
+ }
+
+out:
+ intel_fb_obj_invalidate(obj, ORIGIN_CPU);
+ obj->mm.dirty = true;
+ /* return with the pages pinned */
+ return 0;
+
+err_unpin:
+ i915_gem_object_unpin_pages(obj);
+ return ret;
+}
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index 9bf8155d27de..980587e420e6 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -15,6 +15,8 @@
#include "i915_gem_object_types.h"
+#include "../i915_gem_gtt.h"
+
void i915_gem_init__objects(struct drm_i915_private *i915);
struct drm_i915_gem_object *i915_gem_object_alloc(void);
@@ -353,6 +355,20 @@ void
i915_gem_object_flush_write_domain(struct drm_i915_gem_object *obj,
unsigned int flush_domains);
+int i915_gem_object_prepare_read(struct drm_i915_gem_object *obj,
+ unsigned int *needs_clflush);
+int i915_gem_object_prepare_write(struct drm_i915_gem_object *obj,
+ unsigned int *needs_clflush);
+#define CLFLUSH_BEFORE BIT(0)
+#define CLFLUSH_AFTER BIT(1)
+#define CLFLUSH_FLAGS (CLFLUSH_BEFORE | CLFLUSH_AFTER)
+
+static inline void
+i915_gem_object_finish_access(struct drm_i915_gem_object *obj)
+{
+ i915_gem_object_unpin_pages(obj);
+}
+
static inline struct intel_engine_cs *
i915_gem_object_last_write_engine(struct drm_i915_gem_object *obj)
{
@@ -374,6 +390,19 @@ void i915_gem_object_set_cache_coherency(struct drm_i915_gem_object *obj,
unsigned int cache_level);
void i915_gem_object_flush_if_display(struct drm_i915_gem_object *obj);
+int __must_check
+i915_gem_object_set_to_wc_domain(struct drm_i915_gem_object *obj, bool write);
+int __must_check
+i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write);
+int __must_check
+i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write);
+struct i915_vma * __must_check
+i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,
+ u32 alignment,
+ const struct i915_ggtt_view *view,
+ unsigned int flags);
+void i915_gem_object_unpin_from_display_plane(struct i915_vma *vma);
+
static inline bool cpu_write_needs_clflush(struct drm_i915_gem_object *obj)
{
if (obj->cache_dirty)
diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c
index fe66285f0236..1ee881825e38 100644
--- a/drivers/gpu/drm/i915/gvt/cmd_parser.c
+++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c
@@ -1760,7 +1760,7 @@ static int perform_bb_shadow(struct parser_exec_state *s)
goto err_free_bb;
}
- ret = i915_gem_obj_prepare_shmem_write(bb->obj, &bb->clflush);
+ ret = i915_gem_object_prepare_write(bb->obj, &bb->clflush);
if (ret)
goto err_free_obj;
@@ -1809,7 +1809,7 @@ static int perform_bb_shadow(struct parser_exec_state *s)
err_unmap:
i915_gem_object_unpin_map(bb->obj);
err_finish_shmem_access:
- i915_gem_obj_finish_shmem_access(bb->obj);
+ i915_gem_object_finish_access(bb->obj);
err_free_obj:
i915_gem_object_put(bb->obj);
err_free_bb:
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index d97d9355bf15..f0507658c154 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -465,7 +465,7 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
bb->obj->base.size);
bb->clflush &= ~CLFLUSH_AFTER;
}
- i915_gem_obj_finish_shmem_access(bb->obj);
+ i915_gem_object_finish_access(bb->obj);
bb->accessing = false;
} else {
@@ -493,7 +493,7 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
if (ret)
goto err;
- i915_gem_obj_finish_shmem_access(bb->obj);
+ i915_gem_object_finish_access(bb->obj);
bb->accessing = false;
ret = i915_vma_move_to_active(bb->vma,
@@ -571,7 +571,7 @@ static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload)
list_for_each_entry_safe(bb, pos, &workload->shadow_bb, list) {
if (bb->obj) {
if (bb->accessing)
- i915_gem_obj_finish_shmem_access(bb->obj);
+ i915_gem_object_finish_access(bb->obj);
if (bb->va && !IS_ERR(bb->va))
i915_gem_object_unpin_map(bb->obj);
diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index e9fadcb4d592..c893bd4eb2c8 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -1058,11 +1058,11 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
void *dst, *src;
int ret;
- ret = i915_gem_obj_prepare_shmem_read(src_obj, &src_needs_clflush);
+ ret = i915_gem_object_prepare_read(src_obj, &src_needs_clflush);
if (ret)
return ERR_PTR(ret);
- ret = i915_gem_obj_prepare_shmem_write(dst_obj, &dst_needs_clflush);
+ ret = i915_gem_object_prepare_write(dst_obj, &dst_needs_clflush);
if (ret) {
dst = ERR_PTR(ret);
goto unpin_src;
@@ -1120,9 +1120,9 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
*needs_clflush_after = dst_needs_clflush & CLFLUSH_AFTER;
unpin_dst:
- i915_gem_obj_finish_shmem_access(dst_obj);
+ i915_gem_object_finish_access(dst_obj);
unpin_src:
- i915_gem_obj_finish_shmem_access(src_obj);
+ i915_gem_object_finish_access(src_obj);
return dst;
}
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index e8fa6f4b100c..569443dd217e 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2798,20 +2798,6 @@ static inline int __sg_page_count(const struct scatterlist *sg)
return sg->length >> PAGE_SHIFT;
}
-int i915_gem_obj_prepare_shmem_read(struct drm_i915_gem_object *obj,
- unsigned int *needs_clflush);
-int i915_gem_obj_prepare_shmem_write(struct drm_i915_gem_object *obj,
- unsigned int *needs_clflush);
-#define CLFLUSH_BEFORE BIT(0)
-#define CLFLUSH_AFTER BIT(1)
-#define CLFLUSH_FLAGS (CLFLUSH_BEFORE | CLFLUSH_AFTER)
-
-static inline void
-i915_gem_obj_finish_shmem_access(struct drm_i915_gem_object *obj)
-{
- i915_gem_object_unpin_pages(obj);
-}
-
static inline int __must_check
i915_mutex_lock_interruptible(struct drm_device *dev)
{
@@ -2874,18 +2860,6 @@ int i915_gem_object_wait_priority(struct drm_i915_gem_object *obj,
const struct i915_sched_attr *attr);
#define I915_PRIORITY_DISPLAY I915_USER_PRIORITY(I915_PRIORITY_MAX)
-int __must_check
-i915_gem_object_set_to_wc_domain(struct drm_i915_gem_object *obj, bool write);
-int __must_check
-i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write);
-int __must_check
-i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write);
-struct i915_vma * __must_check
-i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,
- u32 alignment,
- const struct i915_ggtt_view *view,
- unsigned int flags);
-void i915_gem_object_unpin_from_display_plane(struct i915_vma *vma);
int i915_gem_open(struct drm_i915_private *i915, struct drm_file *file);
void i915_gem_release(struct drm_device *dev, struct drm_file *file);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 78d99841cf76..fc9c9776999a 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -438,123 +438,6 @@ void i915_gem_flush_ggtt_writes(struct drm_i915_private *dev_priv)
}
}
-/*
- * Pins the specified object's pages and synchronizes the object with
- * GPU accesses. Sets needs_clflush to non-zero if the caller should
- * flush the object from the CPU cache.
- */
-int i915_gem_obj_prepare_shmem_read(struct drm_i915_gem_object *obj,
- unsigned int *needs_clflush)
-{
- int ret;
-
- lockdep_assert_held(&obj->base.dev->struct_mutex);
-
- *needs_clflush = 0;
- if (!i915_gem_object_has_struct_page(obj))
- return -ENODEV;
-
- ret = i915_gem_object_wait(obj,
- I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED,
- MAX_SCHEDULE_TIMEOUT);
- if (ret)
- return ret;
-
- ret = i915_gem_object_pin_pages(obj);
- if (ret)
- return ret;
-
- if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ ||
- !static_cpu_has(X86_FEATURE_CLFLUSH)) {
- ret = i915_gem_object_set_to_cpu_domain(obj, false);
- if (ret)
- goto err_unpin;
- else
- goto out;
- }
-
- i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
-
- /* If we're not in the cpu read domain, set ourself into the gtt
- * read domain and manually flush cachelines (if required). This
- * optimizes for the case when the gpu will dirty the data
- * anyway again before the next pread happens.
- */
- if (!obj->cache_dirty &&
- !(obj->read_domains & I915_GEM_DOMAIN_CPU))
- *needs_clflush = CLFLUSH_BEFORE;
-
-out:
- /* return with the pages pinned */
- return 0;
-
-err_unpin:
- i915_gem_object_unpin_pages(obj);
- return ret;
-}
-
-int i915_gem_obj_prepare_shmem_write(struct drm_i915_gem_object *obj,
- unsigned int *needs_clflush)
-{
- int ret;
-
- lockdep_assert_held(&obj->base.dev->struct_mutex);
-
- *needs_clflush = 0;
- if (!i915_gem_object_has_struct_page(obj))
- return -ENODEV;
-
- ret = i915_gem_object_wait(obj,
- I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED |
- I915_WAIT_ALL,
- MAX_SCHEDULE_TIMEOUT);
- if (ret)
- return ret;
-
- ret = i915_gem_object_pin_pages(obj);
- if (ret)
- return ret;
-
- if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE ||
- !static_cpu_has(X86_FEATURE_CLFLUSH)) {
- ret = i915_gem_object_set_to_cpu_domain(obj, true);
- if (ret)
- goto err_unpin;
- else
- goto out;
- }
-
- i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
-
- /* If we're not in the cpu write domain, set ourself into the
- * gtt write domain and manually flush cachelines (as required).
- * This optimizes for the case when the gpu will use the data
- * right away and we therefore have to clflush anyway.
- */
- if (!obj->cache_dirty) {
- *needs_clflush |= CLFLUSH_AFTER;
-
- /*
- * Same trick applies to invalidate partially written
- * cachelines read before writing.
- */
- if (!(obj->read_domains & I915_GEM_DOMAIN_CPU))
- *needs_clflush |= CLFLUSH_BEFORE;
- }
-
-out:
- intel_fb_obj_invalidate(obj, ORIGIN_CPU);
- obj->mm.dirty = true;
- /* return with the pages pinned */
- return 0;
-
-err_unpin:
- i915_gem_object_unpin_pages(obj);
- return ret;
-}
-
static int
shmem_pread(struct page *page, int offset, int len, char __user *user_data,
bool needs_clflush)
@@ -588,7 +471,7 @@ i915_gem_shmem_pread(struct drm_i915_gem_object *obj,
if (ret)
return ret;
- ret = i915_gem_obj_prepare_shmem_read(obj, &needs_clflush);
+ ret = i915_gem_object_prepare_read(obj, &needs_clflush);
mutex_unlock(&obj->base.dev->struct_mutex);
if (ret)
return ret;
@@ -610,7 +493,7 @@ i915_gem_shmem_pread(struct drm_i915_gem_object *obj,
offset = 0;
}
- i915_gem_obj_finish_shmem_access(obj);
+ i915_gem_object_finish_access(obj);
return ret;
}
@@ -985,7 +868,7 @@ i915_gem_shmem_pwrite(struct drm_i915_gem_object *obj,
if (ret)
return ret;
- ret = i915_gem_obj_prepare_shmem_write(obj, &needs_clflush);
+ ret = i915_gem_object_prepare_write(obj, &needs_clflush);
mutex_unlock(&i915->drm.struct_mutex);
if (ret)
return ret;
@@ -1017,7 +900,7 @@ i915_gem_shmem_pwrite(struct drm_i915_gem_object *obj,
}
intel_fb_obj_flush(obj, ORIGIN_CPU);
- i915_gem_obj_finish_shmem_access(obj);
+ i915_gem_object_finish_access(obj);
return ret;
}
@@ -1106,150 +989,6 @@ i915_gem_pwrite_ioctl(struct drm_device *dev, void *data,
return ret;
}
-static void i915_gem_object_bump_inactive_ggtt(struct drm_i915_gem_object *obj)
-{
- struct drm_i915_private *i915 = to_i915(obj->base.dev);
- struct list_head *list;
- struct i915_vma *vma;
-
- GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj));
-
- mutex_lock(&i915->ggtt.vm.mutex);
- for_each_ggtt_vma(vma, obj) {
- if (!drm_mm_node_allocated(&vma->node))
- continue;
-
- list_move_tail(&vma->vm_link, &vma->vm->bound_list);
- }
- mutex_unlock(&i915->ggtt.vm.mutex);
-
- spin_lock(&i915->mm.obj_lock);
- list = obj->bind_count ? &i915->mm.bound_list : &i915->mm.unbound_list;
- list_move_tail(&obj->mm.link, list);
- spin_unlock(&i915->mm.obj_lock);
-}
-
-static inline enum fb_op_origin
-fb_write_origin(struct drm_i915_gem_object *obj, unsigned int domain)
-{
- return (domain == I915_GEM_DOMAIN_GTT ?
- obj->frontbuffer_ggtt_origin : ORIGIN_CPU);
-}
-
-/**
- * Called when user space prepares to use an object with the CPU, either
- * through the mmap ioctl's mapping or a GTT mapping.
- * @dev: drm device
- * @data: ioctl data blob
- * @file: drm file
- */
-int
-i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file)
-{
- struct drm_i915_gem_set_domain *args = data;
- struct drm_i915_gem_object *obj;
- u32 read_domains = args->read_domains;
- u32 write_domain = args->write_domain;
- int err;
-
- /* Only handle setting domains to types used by the CPU. */
- if ((write_domain | read_domains) & I915_GEM_GPU_DOMAINS)
- return -EINVAL;
-
- /*
- * Having something in the write domain implies it's in the read
- * domain, and only that read domain. Enforce that in the request.
- */
- if (write_domain && read_domains != write_domain)
- return -EINVAL;
-
- if (!read_domains)
- return 0;
-
- obj = i915_gem_object_lookup(file, args->handle);
- if (!obj)
- return -ENOENT;
-
- /*
- * Already in the desired write domain? Nothing for us to do!
- *
- * We apply a little bit of cunning here to catch a broader set of
- * no-ops. If obj->write_domain is set, we must be in the same
- * obj->read_domains, and only that domain. Therefore, if that
- * obj->write_domain matches the request read_domains, we are
- * already in the same read/write domain and can skip the operation,
- * without having to further check the requested write_domain.
- */
- if (READ_ONCE(obj->write_domain) == read_domains) {
- err = 0;
- goto out;
- }
-
- /*
- * Try to flush the object off the GPU without holding the lock.
- * We will repeat the flush holding the lock in the normal manner
- * to catch cases where we are gazumped.
- */
- err = i915_gem_object_wait(obj,
- I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_PRIORITY |
- (write_domain ? I915_WAIT_ALL : 0),
- MAX_SCHEDULE_TIMEOUT);
- if (err)
- goto out;
-
- /*
- * Proxy objects do not control access to the backing storage, ergo
- * they cannot be used as a means to manipulate the cache domain
- * tracking for that backing storage. The proxy object is always
- * considered to be outside of any cache domain.
- */
- if (i915_gem_object_is_proxy(obj)) {
- err = -ENXIO;
- goto out;
- }
-
- /*
- * Flush and acquire obj->pages so that we are coherent through
- * direct access in memory with previous cached writes through
- * shmemfs and that our cache domain tracking remains valid.
- * For example, if the obj->filp was moved to swap without us
- * being notified and releasing the pages, we would mistakenly
- * continue to assume that the obj remained out of the CPU cached
- * domain.
- */
- err = i915_gem_object_pin_pages(obj);
- if (err)
- goto out;
-
- err = i915_mutex_lock_interruptible(dev);
- if (err)
- goto out_unpin;
-
- if (read_domains & I915_GEM_DOMAIN_WC)
- err = i915_gem_object_set_to_wc_domain(obj, write_domain);
- else if (read_domains & I915_GEM_DOMAIN_GTT)
- err = i915_gem_object_set_to_gtt_domain(obj, write_domain);
- else
- err = i915_gem_object_set_to_cpu_domain(obj, write_domain);
-
- /* And bump the LRU for this access */
- i915_gem_object_bump_inactive_ggtt(obj);
-
- mutex_unlock(&dev->struct_mutex);
-
- if (write_domain != 0)
- intel_fb_obj_invalidate(obj,
- fb_write_origin(obj, write_domain));
-
-out_unpin:
- i915_gem_object_unpin_pages(obj);
-out:
- i915_gem_object_put(obj);
- return err;
-}
-
/**
* Called when user space has done writes to this buffer
* @dev: drm device
@@ -1518,514 +1257,6 @@ int i915_gem_wait_for_idle(struct drm_i915_private *i915,
return 0;
}
-static void __i915_gem_object_flush_for_display(struct drm_i915_gem_object *obj)
-{
- /*
- * We manually flush the CPU domain so that we can override and
- * force the flush for the display, and perform it asyncrhonously.
- */
- i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
- if (obj->cache_dirty)
- i915_gem_clflush_object(obj, I915_CLFLUSH_FORCE);
- obj->write_domain = 0;
-}
-
-void i915_gem_object_flush_if_display(struct drm_i915_gem_object *obj)
-{
- if (!READ_ONCE(obj->pin_global))
- return;
-
- mutex_lock(&obj->base.dev->struct_mutex);
- __i915_gem_object_flush_for_display(obj);
- mutex_unlock(&obj->base.dev->struct_mutex);
-}
-
-/**
- * Moves a single object to the WC read, and possibly write domain.
- * @obj: object to act on
- * @write: ask for write access or read only
- *
- * This function returns when the move is complete, including waiting on
- * flushes to occur.
- */
-int
-i915_gem_object_set_to_wc_domain(struct drm_i915_gem_object *obj, bool write)
-{
- int ret;
-
- lockdep_assert_held(&obj->base.dev->struct_mutex);
-
- ret = i915_gem_object_wait(obj,
- I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED |
- (write ? I915_WAIT_ALL : 0),
- MAX_SCHEDULE_TIMEOUT);
- if (ret)
- return ret;
-
- if (obj->write_domain == I915_GEM_DOMAIN_WC)
- return 0;
-
- /* Flush and acquire obj->pages so that we are coherent through
- * direct access in memory with previous cached writes through
- * shmemfs and that our cache domain tracking remains valid.
- * For example, if the obj->filp was moved to swap without us
- * being notified and releasing the pages, we would mistakenly
- * continue to assume that the obj remained out of the CPU cached
- * domain.
- */
- ret = i915_gem_object_pin_pages(obj);
- if (ret)
- return ret;
-
- i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_WC);
-
- /* Serialise direct access to this object with the barriers for
- * coherent writes from the GPU, by effectively invalidating the
- * WC domain upon first access.
- */
- if ((obj->read_domains & I915_GEM_DOMAIN_WC) == 0)
- mb();
-
- /* It should now be out of any other write domains, and we can update
- * the domain values for our changes.
- */
- GEM_BUG_ON((obj->write_domain & ~I915_GEM_DOMAIN_WC) != 0);
- obj->read_domains |= I915_GEM_DOMAIN_WC;
- if (write) {
- obj->read_domains = I915_GEM_DOMAIN_WC;
- obj->write_domain = I915_GEM_DOMAIN_WC;
- obj->mm.dirty = true;
- }
-
- i915_gem_object_unpin_pages(obj);
- return 0;
-}
-
-/**
- * Moves a single object to the GTT read, and possibly write domain.
- * @obj: object to act on
- * @write: ask for write access or read only
- *
- * This function returns when the move is complete, including waiting on
- * flushes to occur.
- */
-int
-i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write)
-{
- int ret;
-
- lockdep_assert_held(&obj->base.dev->struct_mutex);
-
- ret = i915_gem_object_wait(obj,
- I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED |
- (write ? I915_WAIT_ALL : 0),
- MAX_SCHEDULE_TIMEOUT);
- if (ret)
- return ret;
-
- if (obj->write_domain == I915_GEM_DOMAIN_GTT)
- return 0;
-
- /* Flush and acquire obj->pages so that we are coherent through
- * direct access in memory with previous cached writes through
- * shmemfs and that our cache domain tracking remains valid.
- * For example, if the obj->filp was moved to swap without us
- * being notified and releasing the pages, we would mistakenly
- * continue to assume that the obj remained out of the CPU cached
- * domain.
- */
- ret = i915_gem_object_pin_pages(obj);
- if (ret)
- return ret;
-
- i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_GTT);
-
- /* Serialise direct access to this object with the barriers for
- * coherent writes from the GPU, by effectively invalidating the
- * GTT domain upon first access.
- */
- if ((obj->read_domains & I915_GEM_DOMAIN_GTT) == 0)
- mb();
-
- /* It should now be out of any other write domains, and we can update
- * the domain values for our changes.
- */
- GEM_BUG_ON((obj->write_domain & ~I915_GEM_DOMAIN_GTT) != 0);
- obj->read_domains |= I915_GEM_DOMAIN_GTT;
- if (write) {
- obj->read_domains = I915_GEM_DOMAIN_GTT;
- obj->write_domain = I915_GEM_DOMAIN_GTT;
- obj->mm.dirty = true;
- }
-
- i915_gem_object_unpin_pages(obj);
- return 0;
-}
-
-/**
- * Changes the cache-level of an object across all VMA.
- * @obj: object to act on
- * @cache_level: new cache level to set for the object
- *
- * After this function returns, the object will be in the new cache-level
- * across all GTT and the contents of the backing storage will be coherent,
- * with respect to the new cache-level. In order to keep the backing storage
- * coherent for all users, we only allow a single cache level to be set
- * globally on the object and prevent it from being changed whilst the
- * hardware is reading from the object. That is if the object is currently
- * on the scanout it will be set to uncached (or equivalent display
- * cache coherency) and all non-MOCS GPU access will also be uncached so
- * that all direct access to the scanout remains coherent.
- */
-int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
- enum i915_cache_level cache_level)
-{
- struct i915_vma *vma;
- int ret;
-
- lockdep_assert_held(&obj->base.dev->struct_mutex);
-
- if (obj->cache_level == cache_level)
- return 0;
-
- /* Inspect the list of currently bound VMA and unbind any that would
- * be invalid given the new cache-level. This is principally to
- * catch the issue of the CS prefetch crossing page boundaries and
- * reading an invalid PTE on older architectures.
- */
-restart:
- list_for_each_entry(vma, &obj->vma.list, obj_link) {
- if (!drm_mm_node_allocated(&vma->node))
- continue;
-
- if (i915_vma_is_pinned(vma)) {
- DRM_DEBUG("can not change the cache level of pinned objects\n");
- return -EBUSY;
- }
-
- if (!i915_vma_is_closed(vma) &&
- i915_gem_valid_gtt_space(vma, cache_level))
- continue;
-
- ret = i915_vma_unbind(vma);
- if (ret)
- return ret;
-
- /* As unbinding may affect other elements in the
- * obj->vma_list (due to side-effects from retiring
- * an active vma), play safe and restart the iterator.
- */
- goto restart;
- }
-
- /* We can reuse the existing drm_mm nodes but need to change the
- * cache-level on the PTE. We could simply unbind them all and
- * rebind with the correct cache-level on next use. However since
- * we already have a valid slot, dma mapping, pages etc, we may as
- * rewrite the PTE in the belief that doing so tramples upon less
- * state and so involves less work.
- */
- if (obj->bind_count) {
- /* Before we change the PTE, the GPU must not be accessing it.
- * If we wait upon the object, we know that all the bound
- * VMA are no longer active.
- */
- ret = i915_gem_object_wait(obj,
- I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED |
- I915_WAIT_ALL,
- MAX_SCHEDULE_TIMEOUT);
- if (ret)
- return ret;
-
- if (!HAS_LLC(to_i915(obj->base.dev)) &&
- cache_level != I915_CACHE_NONE) {
- /* Access to snoopable pages through the GTT is
- * incoherent and on some machines causes a hard
- * lockup. Relinquish the CPU mmaping to force
- * userspace to refault in the pages and we can
- * then double check if the GTT mapping is still
- * valid for that pointer access.
- */
- i915_gem_object_release_mmap(obj);
-
- /* As we no longer need a fence for GTT access,
- * we can relinquish it now (and so prevent having
- * to steal a fence from someone else on the next
- * fence request). Note GPU activity would have
- * dropped the fence as all snoopable access is
- * supposed to be linear.
- */
- for_each_ggtt_vma(vma, obj) {
- ret = i915_vma_put_fence(vma);
- if (ret)
- return ret;
- }
- } else {
- /* We either have incoherent backing store and
- * so no GTT access or the architecture is fully
- * coherent. In such cases, existing GTT mmaps
- * ignore the cache bit in the PTE and we can
- * rewrite it without confusing the GPU or having
- * to force userspace to fault back in its mmaps.
- */
- }
-
- list_for_each_entry(vma, &obj->vma.list, obj_link) {
- if (!drm_mm_node_allocated(&vma->node))
- continue;
-
- ret = i915_vma_bind(vma, cache_level, PIN_UPDATE);
- if (ret)
- return ret;
- }
- }
-
- list_for_each_entry(vma, &obj->vma.list, obj_link)
- vma->node.color = cache_level;
- i915_gem_object_set_cache_coherency(obj, cache_level);
- obj->cache_dirty = true; /* Always invalidate stale cachelines */
-
- return 0;
-}
-
-int i915_gem_get_caching_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file)
-{
- struct drm_i915_gem_caching *args = data;
- struct drm_i915_gem_object *obj;
- int err = 0;
-
- rcu_read_lock();
- obj = i915_gem_object_lookup_rcu(file, args->handle);
- if (!obj) {
- err = -ENOENT;
- goto out;
- }
-
- switch (obj->cache_level) {
- case I915_CACHE_LLC:
- case I915_CACHE_L3_LLC:
- args->caching = I915_CACHING_CACHED;
- break;
-
- case I915_CACHE_WT:
- args->caching = I915_CACHING_DISPLAY;
- break;
-
- default:
- args->caching = I915_CACHING_NONE;
- break;
- }
-out:
- rcu_read_unlock();
- return err;
-}
-
-int i915_gem_set_caching_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file)
-{
- struct drm_i915_private *i915 = to_i915(dev);
- struct drm_i915_gem_caching *args = data;
- struct drm_i915_gem_object *obj;
- enum i915_cache_level level;
- int ret = 0;
-
- switch (args->caching) {
- case I915_CACHING_NONE:
- level = I915_CACHE_NONE;
- break;
- case I915_CACHING_CACHED:
- /*
- * Due to a HW issue on BXT A stepping, GPU stores via a
- * snooped mapping may leave stale data in a corresponding CPU
- * cacheline, whereas normally such cachelines would get
- * invalidated.
- */
- if (!HAS_LLC(i915) && !HAS_SNOOP(i915))
- return -ENODEV;
-
- level = I915_CACHE_LLC;
- break;
- case I915_CACHING_DISPLAY:
- level = HAS_WT(i915) ? I915_CACHE_WT : I915_CACHE_NONE;
- break;
- default:
- return -EINVAL;
- }
-
- obj = i915_gem_object_lookup(file, args->handle);
- if (!obj)
- return -ENOENT;
-
- /*
- * The caching mode of proxy object is handled by its generator, and
- * not allowed to be changed by userspace.
- */
- if (i915_gem_object_is_proxy(obj)) {
- ret = -ENXIO;
- goto out;
- }
-
- if (obj->cache_level == level)
- goto out;
-
- ret = i915_gem_object_wait(obj,
- I915_WAIT_INTERRUPTIBLE,
- MAX_SCHEDULE_TIMEOUT);
- if (ret)
- goto out;
-
- ret = i915_mutex_lock_interruptible(dev);
- if (ret)
- goto out;
-
- ret = i915_gem_object_set_cache_level(obj, level);
- mutex_unlock(&dev->struct_mutex);
-
-out:
- i915_gem_object_put(obj);
- return ret;
-}
-
-/*
- * Prepare buffer for display plane (scanout, cursors, etc). Can be called from
- * an uninterruptible phase (modesetting) and allows any flushes to be pipelined
- * (for pageflips). We only flush the caches while preparing the buffer for
- * display, the callers are responsible for frontbuffer flush.
- */
-struct i915_vma *
-i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,
- u32 alignment,
- const struct i915_ggtt_view *view,
- unsigned int flags)
-{
- struct i915_vma *vma;
- int ret;
-
- lockdep_assert_held(&obj->base.dev->struct_mutex);
-
- /* Mark the global pin early so that we account for the
- * display coherency whilst setting up the cache domains.
- */
- obj->pin_global++;
-
- /* The display engine is not coherent with the LLC cache on gen6. As
- * a result, we make sure that the pinning that is about to occur is
- * done with uncached PTEs. This is lowest common denominator for all
- * chipsets.
- *
- * However for gen6+, we could do better by using the GFDT bit instead
- * of uncaching, which would allow us to flush all the LLC-cached data
- * with that bit in the PTE to main memory with just one PIPE_CONTROL.
- */
- ret = i915_gem_object_set_cache_level(obj,
- HAS_WT(to_i915(obj->base.dev)) ?
- I915_CACHE_WT : I915_CACHE_NONE);
- if (ret) {
- vma = ERR_PTR(ret);
- goto err_unpin_global;
- }
-
- /* As the user may map the buffer once pinned in the display plane
- * (e.g. libkms for the bootup splash), we have to ensure that we
- * always use map_and_fenceable for all scanout buffers. However,
- * it may simply be too big to fit into mappable, in which case
- * put it anyway and hope that userspace can cope (but always first
- * try to preserve the existing ABI).
- */
- vma = ERR_PTR(-ENOSPC);
- if ((flags & PIN_MAPPABLE) == 0 &&
- (!view || view->type == I915_GGTT_VIEW_NORMAL))
- vma = i915_gem_object_ggtt_pin(obj, view, 0, alignment,
- flags |
- PIN_MAPPABLE |
- PIN_NONBLOCK);
- if (IS_ERR(vma))
- vma = i915_gem_object_ggtt_pin(obj, view, 0, alignment, flags);
- if (IS_ERR(vma))
- goto err_unpin_global;
-
- vma->display_alignment = max_t(u64, vma->display_alignment, alignment);
-
- __i915_gem_object_flush_for_display(obj);
-
- /* It should now be out of any other write domains, and we can update
- * the domain values for our changes.
- */
- obj->read_domains |= I915_GEM_DOMAIN_GTT;
-
- return vma;
-
-err_unpin_global:
- obj->pin_global--;
- return vma;
-}
-
-void
-i915_gem_object_unpin_from_display_plane(struct i915_vma *vma)
-{
- lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
-
- if (WARN_ON(vma->obj->pin_global == 0))
- return;
-
- if (--vma->obj->pin_global == 0)
- vma->display_alignment = I915_GTT_MIN_ALIGNMENT;
-
- /* Bump the LRU to try and avoid premature eviction whilst flipping */
- i915_gem_object_bump_inactive_ggtt(vma->obj);
-
- i915_vma_unpin(vma);
-}
-
-/**
- * Moves a single object to the CPU read, and possibly write domain.
- * @obj: object to act on
- * @write: requesting write or read-only access
- *
- * This function returns when the move is complete, including waiting on
- * flushes to occur.
- */
-int
-i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write)
-{
- int ret;
-
- lockdep_assert_held(&obj->base.dev->struct_mutex);
-
- ret = i915_gem_object_wait(obj,
- I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED |
- (write ? I915_WAIT_ALL : 0),
- MAX_SCHEDULE_TIMEOUT);
- if (ret)
- return ret;
-
- i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
-
- /* Flush the CPU cache if it's still invalid. */
- if ((obj->read_domains & I915_GEM_DOMAIN_CPU) == 0) {
- i915_gem_clflush_object(obj, I915_CLFLUSH_SYNC);
- obj->read_domains |= I915_GEM_DOMAIN_CPU;
- }
-
- /* It should now be out of any other write domains, and we can update
- * the domain values for our changes.
- */
- GEM_BUG_ON(obj->write_domain & ~I915_GEM_DOMAIN_CPU);
-
- /* If we're writing through the CPU, then the GPU read domains will
- * need to be invalidated at next use.
- */
- if (write)
- __start_cpu_write(obj);
-
- return 0;
-}
-
/* Throttle our rendering by waiting until the ring has completed our requests
* emitted over 20 msec ago.
*
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 27391300507e..fc68c9096e94 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -1026,7 +1026,7 @@ static void reloc_cache_reset(struct reloc_cache *cache)
mb();
kunmap_atomic(vaddr);
- i915_gem_obj_finish_shmem_access((struct drm_i915_gem_object *)cache->node.mm);
+ i915_gem_object_finish_access((struct drm_i915_gem_object *)cache->node.mm);
} else {
wmb();
io_mapping_unmap_atomic((void __iomem *)vaddr);
@@ -1058,7 +1058,7 @@ static void *reloc_kmap(struct drm_i915_gem_object *obj,
unsigned int flushes;
int err;
- err = i915_gem_obj_prepare_shmem_write(obj, &flushes);
+ err = i915_gem_object_prepare_write(obj, &flushes);
if (err)
return ERR_PTR(err);
diff --git a/drivers/gpu/drm/i915/i915_gem_render_state.c b/drivers/gpu/drm/i915/i915_gem_render_state.c
index 9440024c763f..f3b42b026fff 100644
--- a/drivers/gpu/drm/i915/i915_gem_render_state.c
+++ b/drivers/gpu/drm/i915/i915_gem_render_state.c
@@ -84,7 +84,7 @@ static int render_state_setup(struct intel_render_state *so,
u32 *d;
int ret;
- ret = i915_gem_obj_prepare_shmem_write(so->obj, &needs_clflush);
+ ret = i915_gem_object_prepare_write(so->obj, &needs_clflush);
if (ret)
return ret;
@@ -166,7 +166,7 @@ static int render_state_setup(struct intel_render_state *so,
ret = 0;
out:
- i915_gem_obj_finish_shmem_access(so->obj);
+ i915_gem_object_finish_access(so->obj);
return ret;
err:
diff --git a/drivers/gpu/drm/i915/selftests/huge_pages.c b/drivers/gpu/drm/i915/selftests/huge_pages.c
index ce4ec87698f6..b22b8249dfbd 100644
--- a/drivers/gpu/drm/i915/selftests/huge_pages.c
+++ b/drivers/gpu/drm/i915/selftests/huge_pages.c
@@ -1017,7 +1017,7 @@ static int cpu_check(struct drm_i915_gem_object *obj, u32 dword, u32 val)
unsigned long n;
int err;
- err = i915_gem_obj_prepare_shmem_read(obj, &needs_flush);
+ err = i915_gem_object_prepare_read(obj, &needs_flush);
if (err)
return err;
@@ -1038,7 +1038,7 @@ static int cpu_check(struct drm_i915_gem_object *obj, u32 dword, u32 val)
kunmap_atomic(ptr);
}
- i915_gem_obj_finish_shmem_access(obj);
+ i915_gem_object_finish_access(obj);
return err;
}
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_coherency.c b/drivers/gpu/drm/i915/selftests/i915_gem_coherency.c
index 046a38743152..cb25b5fc8027 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_coherency.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_coherency.c
@@ -37,7 +37,7 @@ static int cpu_set(struct drm_i915_gem_object *obj,
u32 *cpu;
int err;
- err = i915_gem_obj_prepare_shmem_write(obj, &needs_clflush);
+ err = i915_gem_object_prepare_write(obj, &needs_clflush);
if (err)
return err;
@@ -54,7 +54,7 @@ static int cpu_set(struct drm_i915_gem_object *obj,
drm_clflush_virt_range(cpu, sizeof(*cpu));
kunmap_atomic(map);
- i915_gem_obj_finish_shmem_access(obj);
+ i915_gem_object_finish_access(obj);
return 0;
}
@@ -69,7 +69,7 @@ static int cpu_get(struct drm_i915_gem_object *obj,
u32 *cpu;
int err;
- err = i915_gem_obj_prepare_shmem_read(obj, &needs_clflush);
+ err = i915_gem_object_prepare_read(obj, &needs_clflush);
if (err)
return err;
@@ -83,7 +83,7 @@ static int cpu_get(struct drm_i915_gem_object *obj,
*v = *cpu;
kunmap_atomic(map);
- i915_gem_obj_finish_shmem_access(obj);
+ i915_gem_object_finish_access(obj);
return 0;
}
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/selftests/i915_gem_context.c
index 34ac5cc6d59f..c69c6d9a998b 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_context.c
@@ -354,7 +354,7 @@ static int cpu_fill(struct drm_i915_gem_object *obj, u32 value)
unsigned int n, m, need_flush;
int err;
- err = i915_gem_obj_prepare_shmem_write(obj, &need_flush);
+ err = i915_gem_object_prepare_write(obj, &need_flush);
if (err)
return err;
@@ -369,7 +369,7 @@ static int cpu_fill(struct drm_i915_gem_object *obj, u32 value)
kunmap_atomic(map);
}
- i915_gem_obj_finish_shmem_access(obj);
+ i915_gem_object_finish_access(obj);
obj->read_domains = I915_GEM_DOMAIN_GTT | I915_GEM_DOMAIN_CPU;
obj->write_domain = 0;
return 0;
@@ -381,7 +381,7 @@ static noinline int cpu_check(struct drm_i915_gem_object *obj,
unsigned int n, m, needs_flush;
int err;
- err = i915_gem_obj_prepare_shmem_read(obj, &needs_flush);
+ err = i915_gem_object_prepare_read(obj, &needs_flush);
if (err)
return err;
@@ -419,7 +419,7 @@ static noinline int cpu_check(struct drm_i915_gem_object *obj,
break;
}
- i915_gem_obj_finish_shmem_access(obj);
+ i915_gem_object_finish_access(obj);
return err;
}
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 26/40] drm/i915: Move more GEM objects under gem/
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (23 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 25/40] drm/i915: Move GEM domain management " Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 27/40] drm/i915: Pull scatterlist utils out of i915_gem.h Chris Wilson
` (20 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Continuing the theme of separating out the GEM clutter.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/Makefile | 26 +++++------
| 2 -
.../gpu/drm/i915/{ => gem}/i915_gem_clflush.c | 27 +++--------
drivers/gpu/drm/i915/gem/i915_gem_clflush.h | 20 +++++++++
.../gpu/drm/i915/{ => gem}/i915_gem_context.c | 27 ++---------
.../gpu/drm/i915/{ => gem}/i915_gem_context.h | 22 +--------
.../i915/{ => gem}/i915_gem_context_types.h | 0
.../gpu/drm/i915/{ => gem}/i915_gem_dmabuf.c | 28 +++---------
drivers/gpu/drm/i915/gem/i915_gem_domain.c | 2 +-
.../drm/i915/{ => gem}/i915_gem_execbuffer.c | 37 ++++-----------
.../drm/i915/{ => gem}/i915_gem_internal.c | 33 +++++---------
drivers/gpu/drm/i915/gem/i915_gem_object.c | 10 ++++-
drivers/gpu/drm/i915/{ => gem}/i915_gem_pm.c | 2 +-
drivers/gpu/drm/i915/{ => gem}/i915_gem_pm.h | 0
.../drm/i915/{ => gem}/i915_gem_shrinker.c | 25 ++---------
.../gpu/drm/i915/{ => gem}/i915_gem_stolen.c | 33 ++++----------
.../gpu/drm/i915/{ => gem}/i915_gem_tiling.c | 31 +++----------
.../gpu/drm/i915/{ => gem}/i915_gem_userptr.c | 30 +++----------
drivers/gpu/drm/i915/{ => gem}/i915_gemfs.c | 25 ++---------
drivers/gpu/drm/i915/gem/i915_gemfs.h | 16 +++++++
.../{ => gem}/selftests/huge_gem_object.c | 22 +--------
.../drm/i915/gem/selftests/huge_gem_object.h | 27 +++++++++++
.../drm/i915/{ => gem}/selftests/huge_pages.c | 34 +++++---------
.../{ => gem}/selftests/i915_gem_coherency.c | 26 ++---------
.../{ => gem}/selftests/i915_gem_context.c | 40 +++++------------
.../{ => gem}/selftests/i915_gem_dmabuf.c | 26 ++---------
.../drm/i915/gem/selftests/i915_gem_mman.c | 2 +-
.../{ => gem}/selftests/i915_gem_object.c | 28 +++---------
.../i915/{ => gem}/selftests/igt_gem_utils.c | 2 +-
.../i915/{ => gem}/selftests/igt_gem_utils.h | 0
.../i915/{ => gem}/selftests/mock_context.c | 24 ++--------
.../gpu/drm/i915/gem/selftests/mock_context.h | 24 ++++++++++
.../i915/{ => gem}/selftests/mock_dmabuf.c | 22 +--------
.../gpu/drm/i915/gem/selftests/mock_dmabuf.h | 22 +++++++++
.../{ => gem}/selftests/mock_gem_object.h | 7 ++-
drivers/gpu/drm/i915/gt/intel_context.c | 4 +-
drivers/gpu/drm/i915/gt/intel_engine_cs.c | 3 ++
drivers/gpu/drm/i915/gt/intel_lrc.c | 2 +
drivers/gpu/drm/i915/gt/intel_lrc.h | 14 +++---
drivers/gpu/drm/i915/gt/intel_reset.c | 2 +
drivers/gpu/drm/i915/gt/intel_ringbuffer.c | 3 ++
drivers/gpu/drm/i915/gt/mock_engine.c | 3 +-
drivers/gpu/drm/i915/gt/selftest_hangcheck.c | 7 +--
drivers/gpu/drm/i915/gt/selftest_lrc.c | 7 ++-
.../gpu/drm/i915/gt/selftest_workarounds.c | 6 ++-
drivers/gpu/drm/i915/gvt/mmio_context.c | 1 +
drivers/gpu/drm/i915/gvt/scheduler.c | 5 ++-
drivers/gpu/drm/i915/i915_debugfs.c | 2 +-
drivers/gpu/drm/i915/i915_drv.c | 1 +
drivers/gpu/drm/i915/i915_drv.h | 2 +-
drivers/gpu/drm/i915/i915_gem.c | 11 ++---
drivers/gpu/drm/i915/i915_gem_clflush.h | 36 ---------------
drivers/gpu/drm/i915/i915_gem_evict.c | 2 +
drivers/gpu/drm/i915/i915_gemfs.h | 34 --------------
drivers/gpu/drm/i915/i915_globals.c | 2 +-
drivers/gpu/drm/i915/i915_gpu_error.c | 2 +
drivers/gpu/drm/i915/i915_perf.c | 2 +
drivers/gpu/drm/i915/i915_request.c | 3 ++
drivers/gpu/drm/i915/intel_display.c | 1 -
drivers/gpu/drm/i915/intel_guc_submission.c | 2 +
drivers/gpu/drm/i915/intel_overlay.c | 2 +
.../gpu/drm/i915/selftests/huge_gem_object.h | 45 -------------------
drivers/gpu/drm/i915/selftests/i915_active.c | 2 +
drivers/gpu/drm/i915/selftests/i915_gem.c | 6 ++-
.../gpu/drm/i915/selftests/i915_gem_evict.c | 6 ++-
drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 3 +-
drivers/gpu/drm/i915/selftests/i915_request.c | 5 ++-
.../gpu/drm/i915/selftests/i915_timeline.c | 1 +
drivers/gpu/drm/i915/selftests/i915_vma.c | 3 +-
.../gpu/drm/i915/selftests/igt_flush_test.c | 2 +
drivers/gpu/drm/i915/selftests/igt_spinner.c | 3 +-
drivers/gpu/drm/i915/selftests/igt_spinner.h | 2 +-
drivers/gpu/drm/i915/selftests/intel_guc.c | 1 +
drivers/gpu/drm/i915/selftests/mock_context.h | 42 -----------------
drivers/gpu/drm/i915/selftests/mock_dmabuf.h | 41 -----------------
.../gpu/drm/i915/selftests/mock_gem_device.c | 5 ++-
drivers/gpu/drm/i915/selftests/mock_request.c | 2 +-
77 files changed, 338 insertions(+), 692 deletions(-)
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_clflush.c (77%)
create mode 100644 drivers/gpu/drm/i915/gem/i915_gem_clflush.h
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_context.c (98%)
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_context.h (85%)
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_context_types.h (100%)
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_dmabuf.c (86%)
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_execbuffer.c (98%)
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_internal.c (81%)
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_pm.c (99%)
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_pm.h (100%)
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_shrinker.c (93%)
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_stolen.c (93%)
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_tiling.c (90%)
rename drivers/gpu/drm/i915/{ => gem}/i915_gem_userptr.c (94%)
rename drivers/gpu/drm/i915/{ => gem}/i915_gemfs.c (50%)
create mode 100644 drivers/gpu/drm/i915/gem/i915_gemfs.h
rename drivers/gpu/drm/i915/{ => gem}/selftests/huge_gem_object.c (70%)
create mode 100644 drivers/gpu/drm/i915/gem/selftests/huge_gem_object.h
rename drivers/gpu/drm/i915/{ => gem}/selftests/huge_pages.c (97%)
rename drivers/gpu/drm/i915/{ => gem}/selftests/i915_gem_coherency.c (87%)
rename drivers/gpu/drm/i915/{ => gem}/selftests/i915_gem_context.c (96%)
rename drivers/gpu/drm/i915/{ => gem}/selftests/i915_gem_dmabuf.c (87%)
rename drivers/gpu/drm/i915/{ => gem}/selftests/i915_gem_object.c (61%)
rename drivers/gpu/drm/i915/{ => gem}/selftests/igt_gem_utils.c (95%)
rename drivers/gpu/drm/i915/{ => gem}/selftests/igt_gem_utils.h (100%)
rename drivers/gpu/drm/i915/{ => gem}/selftests/mock_context.c (63%)
create mode 100644 drivers/gpu/drm/i915/gem/selftests/mock_context.h
rename drivers/gpu/drm/i915/{ => gem}/selftests/mock_dmabuf.c (73%)
create mode 100644 drivers/gpu/drm/i915/gem/selftests/mock_dmabuf.h
rename drivers/gpu/drm/i915/{ => gem}/selftests/mock_gem_object.h (65%)
delete mode 100644 drivers/gpu/drm/i915/i915_gem_clflush.h
delete mode 100644 drivers/gpu/drm/i915/i915_gemfs.h
delete mode 100644 drivers/gpu/drm/i915/selftests/huge_gem_object.h
delete mode 100644 drivers/gpu/drm/i915/selftests/mock_context.h
delete mode 100644 drivers/gpu/drm/i915/selftests/mock_dmabuf.h
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index 29fc924ba819..0c39ef967cf0 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -87,33 +87,33 @@ i915-y += $(gt-y)
# GEM (Graphics Execution Management) code
obj-y += gem/
gem-y += \
+ gem/i915_gem_clflush.o \
+ gem/i915_gem_context.o \
+ gem/i915_gem_dmabuf.o \
gem/i915_gem_domain.o \
+ gem/i915_gem_execbuffer.o \
+ gem/i915_gem_internal.o \
gem/i915_gem_object.o \
gem/i915_gem_mman.o \
gem/i915_gem_pages.o \
gem/i915_gem_phys.o \
- gem/i915_gem_shmem.o
+ gem/i915_gem_pm.o \
+ gem/i915_gem_shmem.o \
+ gem/i915_gem_shrinker.o \
+ gem/i915_gem_stolen.o \
+ gem/i915_gem_tiling.o \
+ gem/i915_gem_userptr.o \
+ gem/i915_gemfs.o
i915-y += \
$(gem-y) \
i915_active.o \
i915_cmd_parser.o \
i915_gem_batch_pool.o \
- i915_gem_clflush.o \
- i915_gem_context.o \
- i915_gem_dmabuf.o \
i915_gem_evict.o \
- i915_gem_execbuffer.o \
i915_gem_fence_reg.o \
i915_gem_gtt.o \
- i915_gem_internal.o \
i915_gem.o \
- i915_gem_pm.o \
i915_gem_render_state.o \
- i915_gem_shrinker.o \
- i915_gem_stolen.o \
- i915_gem_tiling.o \
- i915_gem_userptr.o \
- i915_gemfs.o \
i915_globals.o \
i915_query.o \
i915_request.o \
@@ -198,10 +198,10 @@ i915-y += dvo_ch7017.o \
# Post-mortem debug and GPU hang state capture
i915-$(CONFIG_DRM_I915_CAPTURE_ERROR) += i915_gpu_error.o
i915-$(CONFIG_DRM_I915_SELFTEST) += \
+ gem/selftests/igt_gem_utils.o \
selftests/i915_random.o \
selftests/i915_selftest.o \
selftests/igt_flush_test.o \
- selftests/igt_gem_utils.o \
selftests/igt_live_test.o \
selftests/igt_reset.o \
selftests/igt_spinner.o
--git a/drivers/gpu/drm/i915/Makefile.header-test b/drivers/gpu/drm/i915/Makefile.header-test
index 2ca4a5fa3186..752ad8ce2411 100644
--- a/drivers/gpu/drm/i915/Makefile.header-test
+++ b/drivers/gpu/drm/i915/Makefile.header-test
@@ -6,8 +6,6 @@ header_test := \
i915_active_types.h \
i915_debugfs.h \
i915_drv.h \
- i915_gem_context_types.h \
- i915_gem_pm.h \
i915_irq.h \
i915_params.h \
i915_priolist_types.h \
diff --git a/drivers/gpu/drm/i915/i915_gem_clflush.c b/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
similarity index 77%
rename from drivers/gpu/drm/i915/i915_gem_clflush.c
rename to drivers/gpu/drm/i915/gem/i915_gem_clflush.c
index 8e74c23cbd91..093bfff55a96 100644
--- a/drivers/gpu/drm/i915/i915_gem_clflush.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
@@ -1,31 +1,14 @@
/*
- * Copyright © 2016 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2016 Intel Corporation
*/
-#include "i915_drv.h"
-#include "intel_frontbuffer.h"
#include "i915_gem_clflush.h"
+#include "../i915_drv.h"
+#include "../intel_frontbuffer.h"
+
static DEFINE_SPINLOCK(clflush_lock);
struct clflush {
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_clflush.h b/drivers/gpu/drm/i915/gem/i915_gem_clflush.h
new file mode 100644
index 000000000000..e6c382973129
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/i915_gem_clflush.h
@@ -0,0 +1,20 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2016 Intel Corporation
+ */
+
+#ifndef __I915_GEM_CLFLUSH_H__
+#define __I915_GEM_CLFLUSH_H__
+
+#include <linux/types.h>
+
+struct drm_i915_private;
+struct drm_i915_gem_object;
+
+bool i915_gem_clflush_object(struct drm_i915_gem_object *obj,
+ unsigned int flags);
+#define I915_CLFLUSH_FORCE BIT(0)
+#define I915_CLFLUSH_SYNC BIT(1)
+
+#endif /* __I915_GEM_CLFLUSH_H__ */
diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
similarity index 98%
rename from drivers/gpu/drm/i915/i915_gem_context.c
rename to drivers/gpu/drm/i915/gem/i915_gem_context.c
index 1825479864d6..217d87c13cd8 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -1,28 +1,7 @@
/*
- * Copyright © 2011-2012 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
- *
- * Authors:
- * Ben Widawsky <ben@bwidawsk.net>
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2011-2012 Intel Corporation
*/
/*
@@ -92,7 +71,7 @@
#include "gt/intel_lrc_reg.h"
-#include "i915_drv.h"
+#include "i915_gem_context.h"
#include "i915_globals.h"
#include "i915_trace.h"
#include "i915_user_extensions.h"
diff --git a/drivers/gpu/drm/i915/i915_gem_context.h b/drivers/gpu/drm/i915/gem/i915_gem_context.h
similarity index 85%
rename from drivers/gpu/drm/i915/i915_gem_context.h
rename to drivers/gpu/drm/i915/gem/i915_gem_context.h
index 9ad4a6362438..630392c77e48 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.h
@@ -1,25 +1,7 @@
/*
- * Copyright © 2016 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2016 Intel Corporation
*/
#ifndef __I915_GEM_CONTEXT_H__
diff --git a/drivers/gpu/drm/i915/i915_gem_context_types.h b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
similarity index 100%
rename from drivers/gpu/drm/i915/i915_gem_context_types.h
rename to drivers/gpu/drm/i915/gem/i915_gem_context_types.h
diff --git a/drivers/gpu/drm/i915/i915_gem_dmabuf.c b/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
similarity index 86%
rename from drivers/gpu/drm/i915/i915_gem_dmabuf.c
rename to drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
index 5a101a9462d8..9b75dd8c267d 100644
--- a/drivers/gpu/drm/i915/i915_gem_dmabuf.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
@@ -1,34 +1,16 @@
/*
- * Copyright 2012 Red Hat Inc
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
+ * SPDX-License-Identifier: MIT
*
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
- * DEALINGS IN THE SOFTWARE.
- *
- * Authors:
- * Dave Airlie <airlied@redhat.com>
+ * Copyright 2012 Red Hat Inc
*/
#include <linux/dma-buf.h>
+#include <linux/highmem.h>
#include <linux/reservation.h>
+#include "i915_gem_object.h"
-#include "i915_drv.h"
+#include "../i915_drv.h"
static struct drm_i915_gem_object *dma_buf_to_obj(struct dma_buf *buf)
{
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_domain.c b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
index eee421e3021c..5e1429f8e910 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_domain.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
@@ -4,11 +4,11 @@
* Copyright © 2014-2016 Intel Corporation
*/
+#include "i915_gem_clflush.h"
#include "i915_gem_ioctls.h"
#include "i915_gem_object.h"
#include "../i915_drv.h"
-#include "../i915_gem_clflush.h"
#include "../i915_gem_gtt.h"
#include "../i915_vma.h"
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
similarity index 98%
rename from drivers/gpu/drm/i915/i915_gem_execbuffer.c
rename to drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index fc68c9096e94..a2761029f0c3 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -1,29 +1,7 @@
/*
- * Copyright © 2008,2010 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
- *
- * Authors:
- * Eric Anholt <eric@anholt.net>
- * Chris Wilson <chris@chris-wilson.co.uk>
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2008,2010 Intel Corporation
*/
#include <linux/intel-iommu.h>
@@ -35,13 +13,16 @@
#include <drm/i915_drm.h>
#include "gem/i915_gem_ioctls.h"
+#include "gt/intel_context.h"
#include "gt/intel_gt_pm.h"
-#include "i915_drv.h"
+#include "i915_gem_ioctls.h"
#include "i915_gem_clflush.h"
-#include "i915_trace.h"
-#include "intel_drv.h"
-#include "intel_frontbuffer.h"
+#include "i915_gem_context.h"
+
+#include "../i915_trace.h"
+#include "../intel_drv.h"
+#include "../intel_frontbuffer.h"
enum {
FORCE_CPU_RELOC = 1,
diff --git a/drivers/gpu/drm/i915/i915_gem_internal.c b/drivers/gpu/drm/i915/gem/i915_gem_internal.c
similarity index 81%
rename from drivers/gpu/drm/i915/i915_gem_internal.c
rename to drivers/gpu/drm/i915/gem/i915_gem_internal.c
index ab627ed1269c..d40adb3bbe29 100644
--- a/drivers/gpu/drm/i915/i915_gem_internal.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_internal.c
@@ -1,29 +1,20 @@
/*
- * Copyright © 2014-2016 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2014-2016 Intel Corporation
*/
+#include <linux/scatterlist.h>
+#include <linux/slab.h>
+#include <linux/swiotlb.h>
+
#include <drm/i915_drm.h>
-#include "i915_drv.h"
+
+#include "i915_gem_object.h"
+
+#include "../i915_drv.h"
+#include "../i915_gem.h"
+#include "../i915_utils.h"
#define QUIET (__GFP_NORETRY | __GFP_NOWARN)
#define MAYFAIL (__GFP_RETRY_MAYFAIL | __GFP_NOWARN)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
index 005e17b3acce..04928ae415eb 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
@@ -22,11 +22,12 @@
*
*/
+#include "i915_gem_context.h"
+#include "i915_gem_clflush.h"
#include "i915_gem_object.h"
#include "../i915_drv.h"
#include "../i915_globals.h"
-#include "../i915_gem_clflush.h"
#include "../intel_frontbuffer.h"
static struct i915_global_object {
@@ -443,3 +444,10 @@ int __init i915_global_objects_init(void)
i915_global_register(&global.base);
return 0;
}
+
+#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
+#include "selftests/huge_gem_object.c"
+#include "selftests/huge_pages.c"
+#include "selftests/i915_gem_object.c"
+#include "selftests/i915_gem_coherency.c"
+#endif
diff --git a/drivers/gpu/drm/i915/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
similarity index 99%
rename from drivers/gpu/drm/i915/i915_gem_pm.c
rename to drivers/gpu/drm/i915/gem/i915_gem_pm.c
index fa9c2ebd966a..be75a52a7ab0 100644
--- a/drivers/gpu/drm/i915/i915_gem_pm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
@@ -4,10 +4,10 @@
* Copyright © 2019 Intel Corporation
*/
+#include "gem/i915_gem_pm.h"
#include "gt/intel_gt_pm.h"
#include "i915_drv.h"
-#include "i915_gem_pm.h"
#include "i915_globals.h"
static void i915_gem_park(struct drm_i915_private *i915)
diff --git a/drivers/gpu/drm/i915/i915_gem_pm.h b/drivers/gpu/drm/i915/gem/i915_gem_pm.h
similarity index 100%
rename from drivers/gpu/drm/i915/i915_gem_pm.h
rename to drivers/gpu/drm/i915/gem/i915_gem_pm.h
diff --git a/drivers/gpu/drm/i915/i915_gem_shrinker.c b/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
similarity index 93%
rename from drivers/gpu/drm/i915/i915_gem_shrinker.c
rename to drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
index 2c7aefb3e101..d551b78af65d 100644
--- a/drivers/gpu/drm/i915/i915_gem_shrinker.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
@@ -1,25 +1,7 @@
/*
- * Copyright © 2008-2015 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2008-2015 Intel Corporation
*/
#include <linux/oom.h>
@@ -32,8 +14,7 @@
#include <linux/vmalloc.h>
#include <drm/i915_drm.h>
-#include "i915_drv.h"
-#include "i915_trace.h"
+#include "../i915_trace.h"
static bool shrinker_lock(struct drm_i915_private *i915,
unsigned int flags,
diff --git a/drivers/gpu/drm/i915/i915_gem_stolen.c b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
similarity index 93%
rename from drivers/gpu/drm/i915/i915_gem_stolen.c
rename to drivers/gpu/drm/i915/gem/i915_gem_stolen.c
index 0a8082cfc761..b0961f2c568b 100644
--- a/drivers/gpu/drm/i915/i915_gem_stolen.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
@@ -1,33 +1,16 @@
/*
- * Copyright © 2008-2012 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
- *
- * Authors:
- * Eric Anholt <eric@anholt.net>
- * Chris Wilson <chris@chris-wilson.co.uk>
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2008-2012 Intel Corporation
*/
+#include <linux/errno.h>
+#include <linux/mutex.h>
+
+#include <drm/drm_mm.h>
#include <drm/i915_drm.h>
-#include "i915_drv.h"
+
+#include "../i915_drv.h"
/*
* The BIOS typically reserves some of the system's memory for the exclusive
diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/gem/i915_gem_tiling.c
similarity index 90%
rename from drivers/gpu/drm/i915/i915_gem_tiling.c
rename to drivers/gpu/drm/i915/gem/i915_gem_tiling.c
index 86d6d92ccbc9..90f6305a8aac 100644
--- a/drivers/gpu/drm/i915/i915_gem_tiling.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_tiling.c
@@ -1,37 +1,18 @@
/*
- * Copyright © 2008 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
- *
- * Authors:
- * Eric Anholt <eric@anholt.net>
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2008 Intel Corporation
*/
#include <linux/string.h>
#include <linux/bitops.h>
#include <drm/i915_drm.h>
-#include "gem/i915_gem_ioctls.h"
+#include "i915_gem_ioctls.h"
+#include "i915_gem_object.h"
-#include "i915_drv.h"
+#include "../i915_drv.h"
+#include "../i915_gem.h"
/**
* DOC: buffer object tiling
diff --git a/drivers/gpu/drm/i915/i915_gem_userptr.c b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
similarity index 94%
rename from drivers/gpu/drm/i915/i915_gem_userptr.c
rename to drivers/gpu/drm/i915/gem/i915_gem_userptr.c
index b9c753f413da..007a2b8cac8b 100644
--- a/drivers/gpu/drm/i915/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
@@ -1,25 +1,7 @@
/*
- * Copyright © 2012-2014 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2012-2014 Intel Corporation
*/
#include <linux/mmu_context.h>
@@ -30,11 +12,11 @@
#include <drm/i915_drm.h>
-#include "gem/i915_gem_ioctls.h"
+#include "i915_gem_ioctls.h"
+#include "i915_gem_object.h"
-#include "i915_drv.h"
-#include "i915_trace.h"
-#include "intel_drv.h"
+#include "../i915_trace.h"
+#include "../intel_drv.h"
struct i915_mm_struct {
struct mm_struct *mm;
diff --git a/drivers/gpu/drm/i915/i915_gemfs.c b/drivers/gpu/drm/i915/gem/i915_gemfs.c
similarity index 50%
rename from drivers/gpu/drm/i915/i915_gemfs.c
rename to drivers/gpu/drm/i915/gem/i915_gemfs.c
index 888b7d3f04c3..fecdbc19def4 100644
--- a/drivers/gpu/drm/i915/i915_gemfs.c
+++ b/drivers/gpu/drm/i915/gem/i915_gemfs.c
@@ -1,34 +1,17 @@
/*
- * Copyright © 2017 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2017 Intel Corporation
*/
#include <linux/fs.h>
#include <linux/mount.h>
#include <linux/pagemap.h>
-#include "i915_drv.h"
#include "i915_gemfs.h"
+#include "../i915_drv.h"
+
int i915_gemfs_init(struct drm_i915_private *i915)
{
struct file_system_type *type;
diff --git a/drivers/gpu/drm/i915/gem/i915_gemfs.h b/drivers/gpu/drm/i915/gem/i915_gemfs.h
new file mode 100644
index 000000000000..2a1e59af3e4a
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/i915_gemfs.h
@@ -0,0 +1,16 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2017 Intel Corporation
+ */
+
+#ifndef __I915_GEMFS_H__
+#define __I915_GEMFS_H__
+
+struct drm_i915_private;
+
+int i915_gemfs_init(struct drm_i915_private *i915);
+
+void i915_gemfs_fini(struct drm_i915_private *i915);
+
+#endif
diff --git a/drivers/gpu/drm/i915/selftests/huge_gem_object.c b/drivers/gpu/drm/i915/gem/selftests/huge_gem_object.c
similarity index 70%
rename from drivers/gpu/drm/i915/selftests/huge_gem_object.c
rename to drivers/gpu/drm/i915/gem/selftests/huge_gem_object.c
index 419fd4d6a8f0..824f3761314c 100644
--- a/drivers/gpu/drm/i915/selftests/huge_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/selftests/huge_gem_object.c
@@ -1,25 +1,7 @@
/*
- * Copyright © 2016 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2016 Intel Corporation
*/
#include "huge_gem_object.h"
diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_gem_object.h b/drivers/gpu/drm/i915/gem/selftests/huge_gem_object.h
new file mode 100644
index 000000000000..549c1394bcdc
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/selftests/huge_gem_object.h
@@ -0,0 +1,27 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2016 Intel Corporation
+ */
+
+#ifndef __HUGE_GEM_OBJECT_H
+#define __HUGE_GEM_OBJECT_H
+
+struct drm_i915_gem_object *
+huge_gem_object(struct drm_i915_private *i915,
+ phys_addr_t phys_size,
+ dma_addr_t dma_size);
+
+static inline phys_addr_t
+huge_gem_object_phys_size(struct drm_i915_gem_object *obj)
+{
+ return obj->scratch;
+}
+
+static inline dma_addr_t
+huge_gem_object_dma_size(struct drm_i915_gem_object *obj)
+{
+ return obj->base.size;
+}
+
+#endif /* !__HUGE_GEM_OBJECT_H */
diff --git a/drivers/gpu/drm/i915/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
similarity index 97%
rename from drivers/gpu/drm/i915/selftests/huge_pages.c
rename to drivers/gpu/drm/i915/gem/selftests/huge_pages.c
index b22b8249dfbd..88ca4ed430ee 100644
--- a/drivers/gpu/drm/i915/selftests/huge_pages.c
+++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
@@ -1,34 +1,20 @@
/*
- * Copyright © 2017 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2017 Intel Corporation
*/
-#include "../i915_selftest.h"
-
#include <linux/prime_numbers.h>
+#include "../../i915_selftest.h"
+#include "../i915_gem_pm.h"
+
#include "igt_gem_utils.h"
-#include "mock_drm.h"
-#include "i915_random.h"
+#include "mock_context.h"
+
+#include "../../selftests/mock_drm.h"
+#include "../../selftests/mock_gem_device.h"
+#include "../../selftests/i915_random.h"
static const unsigned int page_sizes[] = {
I915_GTT_PAGE_SIZE_2M,
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_coherency.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
similarity index 87%
rename from drivers/gpu/drm/i915/selftests/i915_gem_coherency.c
rename to drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
index cb25b5fc8027..b6613caa6f45 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_coherency.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
@@ -1,31 +1,13 @@
/*
- * Copyright © 2017 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2017 Intel Corporation
*/
#include <linux/prime_numbers.h>
-#include "../i915_selftest.h"
-#include "i915_random.h"
+#include "../../i915_selftest.h"
+#include "../../selftests/i915_random.h"
static int cpu_set(struct drm_i915_gem_object *obj,
unsigned long offset,
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
similarity index 96%
rename from drivers/gpu/drm/i915/selftests/i915_gem_context.c
rename to drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
index c69c6d9a998b..653ae08a277f 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
@@ -1,42 +1,26 @@
/*
- * Copyright © 2017 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2017 Intel Corporation
*/
#include <linux/prime_numbers.h>
+#include "gem/i915_gem_pm.h"
#include "gt/intel_reset.h"
#include "i915_selftest.h"
-#include "i915_random.h"
-#include "igt_flush_test.h"
-#include "igt_gem_utils.h"
-#include "igt_live_test.h"
-#include "igt_reset.h"
-#include "igt_spinner.h"
+#include "gem/selftests/igt_gem_utils.h"
+#include "selftests/i915_random.h"
+#include "selftests/igt_flush_test.h"
+#include "selftests/igt_live_test.h"
+#include "selftests/igt_reset.h"
+#include "selftests/igt_spinner.h"
+#include "selftests/mock_drm.h"
+#include "selftests/mock_gem_device.h"
-#include "mock_drm.h"
-#include "mock_gem_device.h"
#include "huge_gem_object.h"
+#include "igt_gem_utils.h"
#define DW_PER_PAGE (PAGE_SIZE / sizeof(u32))
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_dmabuf.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_dmabuf.c
similarity index 87%
rename from drivers/gpu/drm/i915/selftests/i915_gem_dmabuf.c
rename to drivers/gpu/drm/i915/gem/selftests/i915_gem_dmabuf.c
index cc65a503e2f0..fc4dd92f4496 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_dmabuf.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_dmabuf.c
@@ -1,30 +1,12 @@
/*
- * Copyright © 2016 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2016 Intel Corporation
*/
-#include "../i915_selftest.h"
+#include "../../i915_selftest.h"
-#include "mock_gem_device.h"
+#include "../../selftests/mock_gem_device.h"
#include "mock_dmabuf.h"
static int igt_dmabuf_export(void *arg)
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
index 87da01230179..12c90d8fe0fb 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
@@ -7,8 +7,8 @@
#include <linux/prime_numbers.h>
#include "gt/intel_gt_pm.h"
+#include "huge_gem_object.h"
#include "i915_selftest.h"
-#include "selftests/huge_gem_object.h"
#include "selftests/igt_flush_test.h"
struct tile {
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_object.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_object.c
similarity index 61%
rename from drivers/gpu/drm/i915/selftests/i915_gem_object.c
rename to drivers/gpu/drm/i915/gem/selftests/i915_gem_object.c
index a3dd2f1be95b..4266ad03ee47 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_object.c
@@ -1,32 +1,14 @@
/*
- * Copyright © 2016 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2016 Intel Corporation
*/
-#include "../i915_selftest.h"
+#include "../../i915_selftest.h"
-#include "igt_flush_test.h"
-#include "mock_gem_device.h"
#include "huge_gem_object.h"
+#include "selftests/igt_flush_test.h"
+#include "selftests/mock_gem_device.h"
static int igt_gem_object(void *arg)
{
diff --git a/drivers/gpu/drm/i915/selftests/igt_gem_utils.c b/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c
similarity index 95%
rename from drivers/gpu/drm/i915/selftests/igt_gem_utils.c
rename to drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c
index 16891b1a3e50..b96713ed13e8 100644
--- a/drivers/gpu/drm/i915/selftests/igt_gem_utils.c
+++ b/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c
@@ -7,10 +7,10 @@
#include "igt_gem_utils.h"
#include "gt/intel_context.h"
+#include "i915_request.h"
#include "../i915_gem_context.h"
#include "../i915_gem_pm.h"
-#include "../i915_request.h"
struct i915_request *
igt_request_alloc(struct i915_gem_context *ctx, struct intel_engine_cs *engine)
diff --git a/drivers/gpu/drm/i915/selftests/igt_gem_utils.h b/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.h
similarity index 100%
rename from drivers/gpu/drm/i915/selftests/igt_gem_utils.h
rename to drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.h
diff --git a/drivers/gpu/drm/i915/selftests/mock_context.c b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
similarity index 63%
rename from drivers/gpu/drm/i915/selftests/mock_context.c
rename to drivers/gpu/drm/i915/gem/selftests/mock_context.c
index 10e67c931ed1..d373d7238622 100644
--- a/drivers/gpu/drm/i915/selftests/mock_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
@@ -1,29 +1,11 @@
/*
- * Copyright © 2016 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2016 Intel Corporation
*/
#include "mock_context.h"
-#include "mock_gtt.h"
+#include "../../selftests/mock_gtt.h"
struct i915_gem_context *
mock_context(struct drm_i915_private *i915,
diff --git a/drivers/gpu/drm/i915/gem/selftests/mock_context.h b/drivers/gpu/drm/i915/gem/selftests/mock_context.h
new file mode 100644
index 000000000000..0b926653914f
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/selftests/mock_context.h
@@ -0,0 +1,24 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2016 Intel Corporation
+ */
+
+#ifndef __MOCK_CONTEXT_H
+#define __MOCK_CONTEXT_H
+
+void mock_init_contexts(struct drm_i915_private *i915);
+
+struct i915_gem_context *
+mock_context(struct drm_i915_private *i915,
+ const char *name);
+
+void mock_context_close(struct i915_gem_context *ctx);
+
+struct i915_gem_context *
+live_context(struct drm_i915_private *i915, struct drm_file *file);
+
+struct i915_gem_context *kernel_context(struct drm_i915_private *i915);
+void kernel_context_close(struct i915_gem_context *ctx);
+
+#endif /* !__MOCK_CONTEXT_H */
diff --git a/drivers/gpu/drm/i915/selftests/mock_dmabuf.c b/drivers/gpu/drm/i915/gem/selftests/mock_dmabuf.c
similarity index 73%
rename from drivers/gpu/drm/i915/selftests/mock_dmabuf.c
rename to drivers/gpu/drm/i915/gem/selftests/mock_dmabuf.c
index ca682caf1062..b9e059d4328a 100644
--- a/drivers/gpu/drm/i915/selftests/mock_dmabuf.c
+++ b/drivers/gpu/drm/i915/gem/selftests/mock_dmabuf.c
@@ -1,25 +1,7 @@
/*
- * Copyright © 2016 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * SPDX-License-Identifier: MIT
*
+ * Copyright © 2016 Intel Corporation
*/
#include "mock_dmabuf.h"
diff --git a/drivers/gpu/drm/i915/gem/selftests/mock_dmabuf.h b/drivers/gpu/drm/i915/gem/selftests/mock_dmabuf.h
new file mode 100644
index 000000000000..f0f8bbd82dfc
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/selftests/mock_dmabuf.h
@@ -0,0 +1,22 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2016 Intel Corporation
+ */
+
+#ifndef __MOCK_DMABUF_H__
+#define __MOCK_DMABUF_H__
+
+#include <linux/dma-buf.h>
+
+struct mock_dmabuf {
+ int npages;
+ struct page *pages[];
+};
+
+static struct mock_dmabuf *to_mock(struct dma_buf *buf)
+{
+ return buf->priv;
+}
+
+#endif /* !__MOCK_DMABUF_H__ */
diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_object.h b/drivers/gpu/drm/i915/gem/selftests/mock_gem_object.h
similarity index 65%
rename from drivers/gpu/drm/i915/selftests/mock_gem_object.h
rename to drivers/gpu/drm/i915/gem/selftests/mock_gem_object.h
index 20acdbee7bd0..370360b4a148 100644
--- a/drivers/gpu/drm/i915/selftests/mock_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/selftests/mock_gem_object.h
@@ -1,4 +1,9 @@
-/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2016 Intel Corporation
+ */
+
#ifndef __MOCK_GEM_OBJECT_H__
#define __MOCK_GEM_OBJECT_H__
diff --git a/drivers/gpu/drm/i915/gt/intel_context.c b/drivers/gpu/drm/i915/gt/intel_context.c
index 5b31e1e05ddd..c78ec0b58e77 100644
--- a/drivers/gpu/drm/i915/gt/intel_context.c
+++ b/drivers/gpu/drm/i915/gt/intel_context.c
@@ -4,8 +4,10 @@
* Copyright © 2019 Intel Corporation
*/
+#include "gem/i915_gem_context.h"
+#include "gem/i915_gem_pm.h"
+
#include "i915_drv.h"
-#include "i915_gem_context.h"
#include "i915_globals.h"
#include "intel_context.h"
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
index 6434170ea4b6..c397ec1b3df9 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
@@ -24,10 +24,13 @@
#include <drm/drm_print.h>
+#include "gem/i915_gem_context.h"
+
#include "i915_drv.h"
#include "intel_engine.h"
#include "intel_engine_pm.h"
+#include "intel_context.h"
#include "intel_lrc.h"
#include "intel_reset.h"
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 5b6e6b54abf9..64bd25a9e6f5 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -133,6 +133,8 @@
*/
#include <linux/interrupt.h>
+#include "gem/i915_gem_context.h"
+
#include "i915_drv.h"
#include "i915_gem_render_state.h"
#include "i915_vgpu.h"
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.h b/drivers/gpu/drm/i915/gt/intel_lrc.h
index e029aee87adf..c2bba82bcc16 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.h
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.h
@@ -24,7 +24,15 @@
#ifndef _INTEL_LRC_H_
#define _INTEL_LRC_H_
-#include "intel_engine.h"
+#include <linux/types.h>
+
+struct drm_printer;
+
+struct drm_i915_private;
+struct i915_gem_context;
+struct i915_request;
+struct intel_context;
+struct intel_engine_cs;
/* Execlists regs */
#define RING_ELSP(base) _MMIO((base) + 0x230)
@@ -96,10 +104,6 @@ int intel_execlists_submission_init(struct intel_engine_cs *engine);
*/
#define LRC_HEADER_PAGES LRC_PPHWSP_PN
-struct drm_printer;
-
-struct drm_i915_private;
-
void intel_execlists_set_default_submission(struct intel_engine_cs *engine);
void intel_lr_context_reset(struct intel_engine_cs *engine,
diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c
index 419b3415370b..aa605d79fa94 100644
--- a/drivers/gpu/drm/i915/gt/intel_reset.c
+++ b/drivers/gpu/drm/i915/gt/intel_reset.c
@@ -7,6 +7,8 @@
#include <linux/sched/mm.h>
#include <linux/stop_machine.h>
+#include "gem/i915_gem_context.h"
+
#include "i915_drv.h"
#include "i915_gpu_error.h"
#include "i915_irq.h"
diff --git a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
index ac93080bd863..66d5a52d505c 100644
--- a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
@@ -31,9 +31,12 @@
#include <drm/i915_drm.h>
+#include "gem/i915_gem_context.h"
+
#include "i915_drv.h"
#include "i915_gem_render_state.h"
#include "i915_trace.h"
+#include "intel_context.h"
#include "intel_reset.h"
#include "intel_workarounds.h"
diff --git a/drivers/gpu/drm/i915/gt/mock_engine.c b/drivers/gpu/drm/i915/gt/mock_engine.c
index 2941916b37bf..6d7562769eb2 100644
--- a/drivers/gpu/drm/i915/gt/mock_engine.c
+++ b/drivers/gpu/drm/i915/gt/mock_engine.c
@@ -22,8 +22,9 @@
*
*/
+#include "gem/i915_gem_context.h"
+
#include "i915_drv.h"
-#include "i915_gem_context.h"
#include "intel_context.h"
#include "intel_engine_pm.h"
diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
index dab3d30c9c73..4b4589b92521 100644
--- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
+++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
@@ -24,18 +24,19 @@
#include <linux/kthread.h>
+#include "gem/i915_gem_context.h"
#include "intel_engine_pm.h"
#include "i915_selftest.h"
#include "selftests/i915_random.h"
#include "selftests/igt_flush_test.h"
-#include "selftests/igt_gem_utils.h"
#include "selftests/igt_reset.h"
#include "selftests/igt_wedge_me.h"
-
-#include "selftests/mock_context.h"
#include "selftests/mock_drm.h"
+#include "gem/selftests/mock_context.h"
+#include "gem/selftests/igt_gem_utils.h"
+
#define IGT_IDLE_TIMEOUT 50 /* ms; time to wait after flushing between tests */
struct hang {
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index 475d0406f7ce..8e5ddb42fa3e 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -6,15 +6,18 @@
#include <linux/prime_numbers.h>
+#include "gem/i915_gem_pm.h"
#include "gt/intel_reset.h"
+
#include "i915_selftest.h"
#include "selftests/i915_random.h"
#include "selftests/igt_flush_test.h"
-#include "selftests/igt_gem_utils.h"
#include "selftests/igt_live_test.h"
#include "selftests/igt_spinner.h"
#include "selftests/lib_sw_fence.h"
-#include "selftests/mock_context.h"
+
+#include "gem/selftests/igt_gem_utils.h"
+#include "gem/selftests/mock_context.h"
static int live_sanitycheck(void *arg)
{
diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
index 9f7680b9984b..40b6df911d8d 100644
--- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
@@ -4,17 +4,19 @@
* Copyright © 2018 Intel Corporation
*/
+#include "gem/i915_gem_pm.h"
#include "i915_selftest.h"
#include "intel_reset.h"
#include "selftests/igt_flush_test.h"
-#include "selftests/igt_gem_utils.h"
#include "selftests/igt_reset.h"
#include "selftests/igt_spinner.h"
#include "selftests/igt_wedge_me.h"
-#include "selftests/mock_context.h"
#include "selftests/mock_drm.h"
+#include "gem/selftests/igt_gem_utils.h"
+#include "gem/selftests/mock_context.h"
+
static const struct wo_register {
enum intel_platform platform;
u32 reg;
diff --git a/drivers/gpu/drm/i915/gvt/mmio_context.c b/drivers/gpu/drm/i915/gvt/mmio_context.c
index a27bdd3951f6..12b99910029d 100644
--- a/drivers/gpu/drm/i915/gvt/mmio_context.c
+++ b/drivers/gpu/drm/i915/gvt/mmio_context.c
@@ -34,6 +34,7 @@
*/
#include "i915_drv.h"
+#include "gt/intel_context.h"
#include "gvt.h"
#include "trace.h"
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index f0507658c154..0c3724d0b7b5 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -35,8 +35,11 @@
#include <linux/kthread.h>
+#include "gem/i915_gem_context.h"
+#include "gem/i915_gem_pm.h"
+#include "gt/intel_context.h"
+
#include "i915_drv.h"
-#include "i915_gem_pm.h"
#include "gvt.h"
#define RING_CTX_OFF(x) \
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index cb9b56fb6a8a..371b5bbdde3f 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -32,10 +32,10 @@
#include <drm/drm_debugfs.h>
#include <drm/drm_fourcc.h>
+#include "gem/i915_gem_context.h"
#include "gt/intel_reset.h"
#include "i915_debugfs.h"
-#include "i915_gem_context.h"
#include "i915_irq.h"
#include "intel_csr.h"
#include "intel_dp.h"
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 30afd8aef737..15975519407b 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -47,6 +47,7 @@
#include <drm/drm_probe_helper.h>
#include <drm/i915_drm.h>
+#include "gem/i915_gem_context.h"
#include "gem/i915_gem_ioctls.h"
#include "gt/intel_gt_pm.h"
#include "gt/intel_reset.h"
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 569443dd217e..76a72843e29b 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -79,7 +79,7 @@
#include "intel_wopcm.h"
#include "i915_gem.h"
-#include "i915_gem_context.h"
+#include "gem/i915_gem_context_types.h"
#include "i915_gem_fence_reg.h"
#include "i915_gem_gtt.h"
#include "i915_gpu_error.h"
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index fc9c9776999a..343a8a837363 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -38,7 +38,11 @@
#include <linux/dma-buf.h>
#include <linux/mman.h>
+#include "gem/i915_gem_clflush.h"
+#include "gem/i915_gem_context.h"
#include "gem/i915_gem_ioctls.h"
+#include "gem/i915_gem_pm.h"
+#include "gem/i915_gemfs.h"
#include "gt/intel_engine_pm.h"
#include "gt/intel_gt_pm.h"
#include "gt/intel_mocs.h"
@@ -46,9 +50,6 @@
#include "gt/intel_workarounds.h"
#include "i915_drv.h"
-#include "i915_gem_clflush.h"
-#include "i915_gemfs.h"
-#include "i915_gem_pm.h"
#include "i915_trace.h"
#include "i915_vgpu.h"
@@ -2343,9 +2344,5 @@ void i915_gem_track_fb(struct drm_i915_gem_object *old,
#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
#include "selftests/scatterlist.c"
#include "selftests/mock_gem_device.c"
-#include "selftests/huge_gem_object.c"
-#include "selftests/huge_pages.c"
-#include "selftests/i915_gem_object.c"
-#include "selftests/i915_gem_coherency.c"
#include "selftests/i915_gem.c"
#endif
diff --git a/drivers/gpu/drm/i915/i915_gem_clflush.h b/drivers/gpu/drm/i915/i915_gem_clflush.h
deleted file mode 100644
index f390247561b3..000000000000
--- a/drivers/gpu/drm/i915/i915_gem_clflush.h
+++ /dev/null
@@ -1,36 +0,0 @@
-/*
- * Copyright © 2016 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
- *
- */
-
-#ifndef __I915_GEM_CLFLUSH_H__
-#define __I915_GEM_CLFLUSH_H__
-
-struct drm_i915_private;
-struct drm_i915_gem_object;
-
-bool i915_gem_clflush_object(struct drm_i915_gem_object *obj,
- unsigned int flags);
-#define I915_CLFLUSH_FORCE BIT(0)
-#define I915_CLFLUSH_SYNC BIT(1)
-
-#endif /* __I915_GEM_CLFLUSH_H__ */
diff --git a/drivers/gpu/drm/i915/i915_gem_evict.c b/drivers/gpu/drm/i915/i915_gem_evict.c
index 0bdb3e072ba5..a5783c4cb98b 100644
--- a/drivers/gpu/drm/i915/i915_gem_evict.c
+++ b/drivers/gpu/drm/i915/i915_gem_evict.c
@@ -28,6 +28,8 @@
#include <drm/i915_drm.h>
+#include "gem/i915_gem_context.h"
+
#include "i915_drv.h"
#include "intel_drv.h"
#include "i915_trace.h"
diff --git a/drivers/gpu/drm/i915/i915_gemfs.h b/drivers/gpu/drm/i915/i915_gemfs.h
deleted file mode 100644
index cca8bdc5b93e..000000000000
--- a/drivers/gpu/drm/i915/i915_gemfs.h
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Copyright © 2017 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
- *
- */
-
-#ifndef __I915_GEMFS_H__
-#define __I915_GEMFS_H__
-
-struct drm_i915_private;
-
-int i915_gemfs_init(struct drm_i915_private *i915);
-
-void i915_gemfs_fini(struct drm_i915_private *i915);
-
-#endif
diff --git a/drivers/gpu/drm/i915/i915_globals.c b/drivers/gpu/drm/i915/i915_globals.c
index db52a58eadcc..2d5fcba98841 100644
--- a/drivers/gpu/drm/i915/i915_globals.c
+++ b/drivers/gpu/drm/i915/i915_globals.c
@@ -8,7 +8,7 @@
#include <linux/workqueue.h>
#include "i915_active.h"
-#include "i915_gem_context.h"
+#include "gem/i915_gem_context.h"
#include "gem/i915_gem_object.h"
#include "i915_globals.h"
#include "i915_request.h"
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index 4f85cbdddb0d..c86865a34972 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -36,6 +36,8 @@
#include <drm/drm_print.h>
+#include "gem/i915_gem_context.h"
+
#include "i915_drv.h"
#include "i915_gpu_error.h"
#include "intel_atomic.h"
diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index 379fd89a180f..2e33a9b4eae7 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -195,6 +195,8 @@
#include <linux/sizes.h>
#include <linux/uuid.h>
+#include "gem/i915_gem_context.h"
+#include "gem/i915_gem_pm.h"
#include "gt/intel_lrc_reg.h"
#include "i915_drv.h"
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 1188a450804f..4864fd6116f4 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -29,6 +29,9 @@
#include <linux/sched/clock.h>
#include <linux/sched/signal.h>
+#include "gem/i915_gem_context.h"
+#include "gt/intel_context.h"
+
#include "i915_active.h"
#include "i915_drv.h"
#include "i915_globals.h"
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index a351c8e219ba..e63f290ccc26 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -45,7 +45,6 @@
#include <drm/i915_drm.h>
#include "i915_drv.h"
-#include "i915_gem_clflush.h"
#include "i915_trace.h"
#include "intel_acpi.h"
#include "intel_atomic.h"
diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c
index ea0e3734d37c..389a9131d413 100644
--- a/drivers/gpu/drm/i915/intel_guc_submission.c
+++ b/drivers/gpu/drm/i915/intel_guc_submission.c
@@ -26,6 +26,8 @@
#include "gt/intel_engine_pm.h"
#include "gt/intel_lrc_reg.h"
+#include "gt/intel_context.h"
+#include "gem/i915_gem_context.h"
#include "intel_guc_submission.h"
#include "i915_drv.h"
diff --git a/drivers/gpu/drm/i915/intel_overlay.c b/drivers/gpu/drm/i915/intel_overlay.c
index b64b45d9b538..80dcd879fc58 100644
--- a/drivers/gpu/drm/i915/intel_overlay.c
+++ b/drivers/gpu/drm/i915/intel_overlay.c
@@ -29,6 +29,8 @@
#include <drm/drm_fourcc.h>
#include <drm/i915_drm.h>
+#include "gem/i915_gem_pm.h"
+
#include "i915_drv.h"
#include "i915_reg.h"
#include "intel_drv.h"
diff --git a/drivers/gpu/drm/i915/selftests/huge_gem_object.h b/drivers/gpu/drm/i915/selftests/huge_gem_object.h
deleted file mode 100644
index a6133a9e8029..000000000000
--- a/drivers/gpu/drm/i915/selftests/huge_gem_object.h
+++ /dev/null
@@ -1,45 +0,0 @@
-/*
- * Copyright © 2016 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
- *
- */
-
-#ifndef __HUGE_GEM_OBJECT_H
-#define __HUGE_GEM_OBJECT_H
-
-struct drm_i915_gem_object *
-huge_gem_object(struct drm_i915_private *i915,
- phys_addr_t phys_size,
- dma_addr_t dma_size);
-
-static inline phys_addr_t
-huge_gem_object_phys_size(struct drm_i915_gem_object *obj)
-{
- return obj->scratch;
-}
-
-static inline dma_addr_t
-huge_gem_object_dma_size(struct drm_i915_gem_object *obj)
-{
- return obj->base.size;
-}
-
-#endif /* !__HUGE_GEM_OBJECT_H */
diff --git a/drivers/gpu/drm/i915/selftests/i915_active.c b/drivers/gpu/drm/i915/selftests/i915_active.c
index eee838dc0634..2af2b0ba9147 100644
--- a/drivers/gpu/drm/i915/selftests/i915_active.c
+++ b/drivers/gpu/drm/i915/selftests/i915_active.c
@@ -9,6 +9,8 @@
#include "igt_flush_test.h"
#include "lib_sw_fence.h"
+#include "../gem/i915_gem_pm.h"
+
struct live_active {
struct i915_active base;
bool retired;
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem.c b/drivers/gpu/drm/i915/selftests/i915_gem.c
index c6a9bff85311..96221972c3f8 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem.c
@@ -8,9 +8,11 @@
#include "../i915_selftest.h"
-#include "igt_gem_utils.h"
#include "igt_flush_test.h"
-#include "mock_context.h"
+#include "mock_drm.h"
+
+#include "../gem/selftests/igt_gem_utils.h"
+#include "../gem/selftests/mock_context.h"
static int switch_to_context(struct drm_i915_private *i915,
struct i915_gem_context *ctx)
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
index 4fc6e5445dd1..f0c320694277 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
@@ -24,12 +24,14 @@
#include "../i915_selftest.h"
-#include "igt_gem_utils.h"
#include "lib_sw_fence.h"
-#include "mock_context.h"
#include "mock_drm.h"
#include "mock_gem_device.h"
+#include "../gem/i915_gem_pm.h"
+#include "../gem/selftests/igt_gem_utils.h"
+#include "../gem/selftests/mock_context.h"
+
static void quirk_add(struct drm_i915_gem_object *obj,
struct list_head *objects)
{
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
index 9cca66e4420a..e1c8e34b8405 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
@@ -28,10 +28,11 @@
#include "../i915_selftest.h"
#include "i915_random.h"
-#include "mock_context.h"
#include "mock_drm.h"
#include "mock_gem_device.h"
+#include "../gem/selftests/mock_context.h"
+
static void cleanup_freed_objects(struct drm_i915_private *i915)
{
/*
diff --git a/drivers/gpu/drm/i915/selftests/i915_request.c b/drivers/gpu/drm/i915/selftests/i915_request.c
index b60591531e4a..92710353982c 100644
--- a/drivers/gpu/drm/i915/selftests/i915_request.c
+++ b/drivers/gpu/drm/i915/selftests/i915_request.c
@@ -29,10 +29,13 @@
#include "igt_live_test.h"
#include "lib_sw_fence.h"
-#include "mock_context.h"
+#include "gem/i915_gem_pm.h"
+
#include "mock_drm.h"
#include "mock_gem_device.h"
+#include "../gem/selftests/mock_context.h"
+
static int igt_add_request(void *arg)
{
struct drm_i915_private *i915 = arg;
diff --git a/drivers/gpu/drm/i915/selftests/i915_timeline.c b/drivers/gpu/drm/i915/selftests/i915_timeline.c
index ff9ebe50fae8..82c9be5be92f 100644
--- a/drivers/gpu/drm/i915/selftests/i915_timeline.c
+++ b/drivers/gpu/drm/i915/selftests/i915_timeline.c
@@ -6,6 +6,7 @@
#include <linux/prime_numbers.h>
+#include "../gem/i915_gem_pm.h"
#include "../i915_selftest.h"
#include "i915_random.h"
diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c
index fc594b030f5a..131d0d9a6cd1 100644
--- a/drivers/gpu/drm/i915/selftests/i915_vma.c
+++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
@@ -27,9 +27,10 @@
#include "../i915_selftest.h"
#include "mock_gem_device.h"
-#include "mock_context.h"
#include "mock_gtt.h"
+#include "../gem/selftests/mock_context.h"
+
static bool assert_vma(struct i915_vma *vma,
struct drm_i915_gem_object *obj,
struct i915_gem_context *ctx)
diff --git a/drivers/gpu/drm/i915/selftests/igt_flush_test.c b/drivers/gpu/drm/i915/selftests/igt_flush_test.c
index e42f3c58536a..34af282c3c93 100644
--- a/drivers/gpu/drm/i915/selftests/igt_flush_test.c
+++ b/drivers/gpu/drm/i915/selftests/igt_flush_test.c
@@ -9,6 +9,8 @@
#include "../i915_selftest.h"
#include "igt_flush_test.h"
+#include "../gem/i915_gem_context.h"
+
int igt_flush_test(struct drm_i915_private *i915, unsigned int flags)
{
int ret = i915_terminally_wedged(i915) ? -EIO : 0;
diff --git a/drivers/gpu/drm/i915/selftests/igt_spinner.c b/drivers/gpu/drm/i915/selftests/igt_spinner.c
index ece8a8a0d3b0..45688e6698b7 100644
--- a/drivers/gpu/drm/i915/selftests/igt_spinner.c
+++ b/drivers/gpu/drm/i915/selftests/igt_spinner.c
@@ -4,9 +4,10 @@
* Copyright © 2018 Intel Corporation
*/
-#include "igt_gem_utils.h"
#include "igt_spinner.h"
+#include "../gem/selftests/igt_gem_utils.h"
+
int igt_spinner_init(struct igt_spinner *spin, struct drm_i915_private *i915)
{
unsigned int mode;
diff --git a/drivers/gpu/drm/i915/selftests/igt_spinner.h b/drivers/gpu/drm/i915/selftests/igt_spinner.h
index d312e7cdab68..2b10c78da26b 100644
--- a/drivers/gpu/drm/i915/selftests/igt_spinner.h
+++ b/drivers/gpu/drm/i915/selftests/igt_spinner.h
@@ -9,11 +9,11 @@
#include "../i915_selftest.h"
+#include "gem/i915_gem_context.h"
#include "gt/intel_engine.h"
#include "../i915_drv.h"
#include "../i915_request.h"
-#include "../i915_gem_context.h"
struct igt_spinner {
struct drm_i915_private *i915;
diff --git a/drivers/gpu/drm/i915/selftests/intel_guc.c b/drivers/gpu/drm/i915/selftests/intel_guc.c
index b05a21eaa8f4..f7eb483f9eae 100644
--- a/drivers/gpu/drm/i915/selftests/intel_guc.c
+++ b/drivers/gpu/drm/i915/selftests/intel_guc.c
@@ -23,6 +23,7 @@
*/
#include "../i915_selftest.h"
+#include "../gem/i915_gem_pm.h"
/* max doorbell number + negative test for each client type */
#define ATTEMPTS (GUC_NUM_DOORBELLS + GUC_CLIENT_PRIORITY_NUM)
diff --git a/drivers/gpu/drm/i915/selftests/mock_context.h b/drivers/gpu/drm/i915/selftests/mock_context.h
deleted file mode 100644
index 29b9d60a158b..000000000000
--- a/drivers/gpu/drm/i915/selftests/mock_context.h
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * Copyright © 2016 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
- *
- */
-
-#ifndef __MOCK_CONTEXT_H
-#define __MOCK_CONTEXT_H
-
-void mock_init_contexts(struct drm_i915_private *i915);
-
-struct i915_gem_context *
-mock_context(struct drm_i915_private *i915,
- const char *name);
-
-void mock_context_close(struct i915_gem_context *ctx);
-
-struct i915_gem_context *
-live_context(struct drm_i915_private *i915, struct drm_file *file);
-
-struct i915_gem_context *kernel_context(struct drm_i915_private *i915);
-void kernel_context_close(struct i915_gem_context *ctx);
-
-#endif /* !__MOCK_CONTEXT_H */
diff --git a/drivers/gpu/drm/i915/selftests/mock_dmabuf.h b/drivers/gpu/drm/i915/selftests/mock_dmabuf.h
deleted file mode 100644
index ec80613159b9..000000000000
--- a/drivers/gpu/drm/i915/selftests/mock_dmabuf.h
+++ /dev/null
@@ -1,41 +0,0 @@
-
-/*
- * Copyright © 2016 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
- *
- */
-
-#ifndef __MOCK_DMABUF_H__
-#define __MOCK_DMABUF_H__
-
-#include <linux/dma-buf.h>
-
-struct mock_dmabuf {
- int npages;
- struct page *pages[];
-};
-
-static struct mock_dmabuf *to_mock(struct dma_buf *buf)
-{
- return buf->priv;
-}
-
-#endif /* !__MOCK_DMABUF_H__ */
diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
index 9fd02025d382..e25b74a27f83 100644
--- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
+++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
@@ -27,13 +27,14 @@
#include "gt/mock_engine.h"
-#include "mock_context.h"
#include "mock_request.h"
#include "mock_gem_device.h"
-#include "mock_gem_object.h"
#include "mock_gtt.h"
#include "mock_uncore.h"
+#include "gem/selftests/mock_context.h"
+#include "gem/selftests/mock_gem_object.h"
+
void mock_device_flush(struct drm_i915_private *i915)
{
struct intel_engine_cs *engine;
diff --git a/drivers/gpu/drm/i915/selftests/mock_request.c b/drivers/gpu/drm/i915/selftests/mock_request.c
index b99f7576153c..9390fc09984b 100644
--- a/drivers/gpu/drm/i915/selftests/mock_request.c
+++ b/drivers/gpu/drm/i915/selftests/mock_request.c
@@ -22,9 +22,9 @@
*
*/
+#include "gem/selftests/igt_gem_utils.h"
#include "gt/mock_engine.h"
-#include "igt_gem_utils.h"
#include "mock_request.h"
struct i915_request *
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 27/40] drm/i915: Pull scatterlist utils out of i915_gem.h
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (24 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 26/40] drm/i915: Move more GEM objects under gem/ Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 28/40] drm/i915: Move GEM object domain management from struct_mutex to local Chris Wilson
` (19 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx; +Cc: Matthew Auld
Out scatterlist utility routines can be pulled out of i915_gem.h for a
bit more decluttering.
v2: Push I915_GTT_PAGE_SIZE out of i915_scatterlist itself and into the
caller.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
---
drivers/gpu/drm/i915/Makefile | 1 +
drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c | 1 +
drivers/gpu/drm/i915/gem/i915_gem_internal.c | 1 +
drivers/gpu/drm/i915/gem/i915_gem_pages.c | 1 +
drivers/gpu/drm/i915/gem/i915_gem_phys.c | 1 +
drivers/gpu/drm/i915/gem/i915_gem_shmem.c | 1 +
drivers/gpu/drm/i915/gem/i915_gem_userptr.c | 1 +
.../drm/i915/gem/selftests/huge_gem_object.c | 2 +
.../drm/i915/gem/selftests/i915_gem_dmabuf.c | 2 +
drivers/gpu/drm/i915/i915_drv.h | 110 ---------------
drivers/gpu/drm/i915/i915_gem.c | 30 +----
drivers/gpu/drm/i915/i915_gem_fence_reg.c | 2 +
drivers/gpu/drm/i915/i915_gem_gtt.c | 3 +-
drivers/gpu/drm/i915/i915_gem_gtt.h | 4 +-
drivers/gpu/drm/i915/i915_gpu_error.c | 1 +
drivers/gpu/drm/i915/i915_scatterlist.c | 39 ++++++
drivers/gpu/drm/i915/i915_scatterlist.h | 127 ++++++++++++++++++
drivers/gpu/drm/i915/selftests/i915_vma.c | 1 +
drivers/gpu/drm/i915/selftests/scatterlist.c | 1 +
19 files changed, 188 insertions(+), 141 deletions(-)
create mode 100644 drivers/gpu/drm/i915/i915_scatterlist.c
create mode 100644 drivers/gpu/drm/i915/i915_scatterlist.h
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index 0c39ef967cf0..9f5f4acacae5 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -44,6 +44,7 @@ i915-y += i915_drv.o \
i915_irq.o \
i915_params.o \
i915_pci.o \
+ i915_scatterlist.o \
i915_suspend.o \
i915_sysfs.o \
intel_csr.o \
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c b/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
index 9b75dd8c267d..4e7efe159531 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
@@ -11,6 +11,7 @@
#include "i915_gem_object.h"
#include "../i915_drv.h"
+#include "../i915_scatterlist.h"
static struct drm_i915_gem_object *dma_buf_to_obj(struct dma_buf *buf)
{
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_internal.c b/drivers/gpu/drm/i915/gem/i915_gem_internal.c
index d40adb3bbe29..d072d0cbce06 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_internal.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_internal.c
@@ -14,6 +14,7 @@
#include "../i915_drv.h"
#include "../i915_gem.h"
+#include "../i915_scatterlist.h"
#include "../i915_utils.h"
#define QUIET (__GFP_NORETRY | __GFP_NOWARN)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pages.c b/drivers/gpu/drm/i915/gem/i915_gem_pages.c
index 452aba0a3359..21635d0ac95e 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pages.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pages.c
@@ -7,6 +7,7 @@
#include "i915_gem_object.h"
#include "../i915_drv.h"
+#include "../i915_scatterlist.h"
void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj,
struct sg_table *pages,
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_phys.c b/drivers/gpu/drm/i915/gem/i915_gem_phys.c
index 1bf3e0afcba2..22d185301578 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_phys.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_phys.c
@@ -16,6 +16,7 @@
#include "i915_gem_object.h"
#include "../i915_drv.h"
+#include "../i915_scatterlist.h"
static int i915_gem_object_get_pages_phys(struct drm_i915_gem_object *obj)
{
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
index 78df86d1db7a..2eee1af88bcb 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
@@ -10,6 +10,7 @@
#include "i915_gem_object.h"
#include "../i915_drv.h"
+#include "../i915_scatterlist.h"
/*
* Move pages to appropriate lru and release the pagevec, decrementing the
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
index 007a2b8cac8b..0137baa409e1 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
@@ -15,6 +15,7 @@
#include "i915_gem_ioctls.h"
#include "i915_gem_object.h"
+#include "../i915_scatterlist.h"
#include "../i915_trace.h"
#include "../intel_drv.h"
diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_gem_object.c b/drivers/gpu/drm/i915/gem/selftests/huge_gem_object.c
index 824f3761314c..c03781f8b435 100644
--- a/drivers/gpu/drm/i915/gem/selftests/huge_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/selftests/huge_gem_object.c
@@ -4,6 +4,8 @@
* Copyright © 2016 Intel Corporation
*/
+#include "../../i915_scatterlist.h"
+
#include "huge_gem_object.h"
static void huge_free_pages(struct drm_i915_gem_object *obj,
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_dmabuf.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_dmabuf.c
index fc4dd92f4496..da6eb6ecdf68 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_dmabuf.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_dmabuf.c
@@ -9,6 +9,8 @@
#include "../../selftests/mock_gem_device.h"
#include "mock_dmabuf.h"
+#include "../../i915_drv.h"
+
static int igt_dmabuf_export(void *arg)
{
struct drm_i915_private *i915 = arg;
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 76a72843e29b..6f8ddfbe7d85 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2153,111 +2153,6 @@ enum hdmi_force_audio {
GENMASK(INTEL_FRONTBUFFER_BITS_PER_PIPE * ((pipe) + 1) - 1, \
INTEL_FRONTBUFFER_BITS_PER_PIPE * (pipe))
-/*
- * Optimised SGL iterator for GEM objects
- */
-static __always_inline struct sgt_iter {
- struct scatterlist *sgp;
- union {
- unsigned long pfn;
- dma_addr_t dma;
- };
- unsigned int curr;
- unsigned int max;
-} __sgt_iter(struct scatterlist *sgl, bool dma) {
- struct sgt_iter s = { .sgp = sgl };
-
- if (s.sgp) {
- s.max = s.curr = s.sgp->offset;
- s.max += s.sgp->length;
- if (dma)
- s.dma = sg_dma_address(s.sgp);
- else
- s.pfn = page_to_pfn(sg_page(s.sgp));
- }
-
- return s;
-}
-
-static inline struct scatterlist *____sg_next(struct scatterlist *sg)
-{
- ++sg;
- if (unlikely(sg_is_chain(sg)))
- sg = sg_chain_ptr(sg);
- return sg;
-}
-
-/**
- * __sg_next - return the next scatterlist entry in a list
- * @sg: The current sg entry
- *
- * Description:
- * If the entry is the last, return NULL; otherwise, step to the next
- * element in the array (@sg@+1). If that's a chain pointer, follow it;
- * otherwise just return the pointer to the current element.
- **/
-static inline struct scatterlist *__sg_next(struct scatterlist *sg)
-{
- return sg_is_last(sg) ? NULL : ____sg_next(sg);
-}
-
-/**
- * for_each_sgt_dma - iterate over the DMA addresses of the given sg_table
- * @__dmap: DMA address (output)
- * @__iter: 'struct sgt_iter' (iterator state, internal)
- * @__sgt: sg_table to iterate over (input)
- */
-#define for_each_sgt_dma(__dmap, __iter, __sgt) \
- for ((__iter) = __sgt_iter((__sgt)->sgl, true); \
- ((__dmap) = (__iter).dma + (__iter).curr); \
- (((__iter).curr += I915_GTT_PAGE_SIZE) >= (__iter).max) ? \
- (__iter) = __sgt_iter(__sg_next((__iter).sgp), true), 0 : 0)
-
-/**
- * for_each_sgt_page - iterate over the pages of the given sg_table
- * @__pp: page pointer (output)
- * @__iter: 'struct sgt_iter' (iterator state, internal)
- * @__sgt: sg_table to iterate over (input)
- */
-#define for_each_sgt_page(__pp, __iter, __sgt) \
- for ((__iter) = __sgt_iter((__sgt)->sgl, false); \
- ((__pp) = (__iter).pfn == 0 ? NULL : \
- pfn_to_page((__iter).pfn + ((__iter).curr >> PAGE_SHIFT))); \
- (((__iter).curr += PAGE_SIZE) >= (__iter).max) ? \
- (__iter) = __sgt_iter(__sg_next((__iter).sgp), false), 0 : 0)
-
-bool i915_sg_trim(struct sg_table *orig_st);
-
-static inline unsigned int i915_sg_page_sizes(struct scatterlist *sg)
-{
- unsigned int page_sizes;
-
- page_sizes = 0;
- while (sg) {
- GEM_BUG_ON(sg->offset);
- GEM_BUG_ON(!IS_ALIGNED(sg->length, PAGE_SIZE));
- page_sizes |= sg->length;
- sg = __sg_next(sg);
- }
-
- return page_sizes;
-}
-
-static inline unsigned int i915_sg_segment_size(void)
-{
- unsigned int size = swiotlb_max_segment();
-
- if (size == 0)
- return SCATTERLIST_MAX_SEGMENT;
-
- size = rounddown(size, PAGE_SIZE);
- /* swiotlb_max_segment_size can return 1 byte when it means one page. */
- if (size < PAGE_SIZE)
- size = PAGE_SIZE;
-
- return size;
-}
-
#define INTEL_INFO(dev_priv) (&(dev_priv)->__info)
#define RUNTIME_INFO(dev_priv) (&(dev_priv)->__runtime)
#define DRIVER_CAPS(dev_priv) (&(dev_priv)->caps)
@@ -2793,11 +2688,6 @@ int i915_gem_object_unbind(struct drm_i915_gem_object *obj);
void i915_gem_runtime_suspend(struct drm_i915_private *dev_priv);
-static inline int __sg_page_count(const struct scatterlist *sg)
-{
- return sg->length >> PAGE_SHIFT;
-}
-
static inline int __must_check
i915_mutex_lock_interruptible(struct drm_device *dev)
{
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 343a8a837363..7fa1236677b9 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -50,6 +50,7 @@
#include "gt/intel_workarounds.h"
#include "i915_drv.h"
+#include "i915_scatterlist.h"
#include "i915_trace.h"
#include "i915_vgpu.h"
@@ -1061,34 +1062,6 @@ void i915_gem_runtime_suspend(struct drm_i915_private *dev_priv)
}
}
-bool i915_sg_trim(struct sg_table *orig_st)
-{
- struct sg_table new_st;
- struct scatterlist *sg, *new_sg;
- unsigned int i;
-
- if (orig_st->nents == orig_st->orig_nents)
- return false;
-
- if (sg_alloc_table(&new_st, orig_st->nents, GFP_KERNEL | __GFP_NOWARN))
- return false;
-
- new_sg = new_st.sgl;
- for_each_sg(orig_st->sgl, sg, orig_st->nents, i) {
- sg_set_page(new_sg, sg_page(sg), sg->length, 0);
- sg_dma_address(new_sg) = sg_dma_address(sg);
- sg_dma_len(new_sg) = sg_dma_len(sg);
-
- new_sg = sg_next(new_sg);
- }
- GEM_BUG_ON(new_sg); /* Should walk exactly nents and hit the end */
-
- sg_free_table(orig_st);
-
- *orig_st = new_st;
- return true;
-}
-
static unsigned long to_wait_timeout(s64 timeout_ns)
{
if (timeout_ns < 0)
@@ -2342,7 +2315,6 @@ void i915_gem_track_fb(struct drm_i915_gem_object *old,
}
#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
-#include "selftests/scatterlist.c"
#include "selftests/mock_gem_device.c"
#include "selftests/i915_gem.c"
#endif
diff --git a/drivers/gpu/drm/i915/i915_gem_fence_reg.c b/drivers/gpu/drm/i915/i915_gem_fence_reg.c
index 3084f52e3372..2e9e32330aaa 100644
--- a/drivers/gpu/drm/i915/i915_gem_fence_reg.c
+++ b/drivers/gpu/drm/i915/i915_gem_fence_reg.c
@@ -22,7 +22,9 @@
*/
#include <drm/i915_drm.h>
+
#include "i915_drv.h"
+#include "i915_scatterlist.h"
/**
* DOC: fence register handling
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 8f5db787b7f2..d3eba67d4bf9 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -36,8 +36,9 @@
#include <drm/i915_drm.h>
#include "i915_drv.h"
-#include "i915_vgpu.h"
+#include "i915_scatterlist.h"
#include "i915_trace.h"
+#include "i915_vgpu.h"
#include "intel_drv.h"
#include "intel_frontbuffer.h"
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 593079f30fbe..8e661c76d235 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -40,6 +40,7 @@
#include "gt/intel_reset.h"
#include "i915_request.h"
+#include "i915_scatterlist.h"
#include "i915_selftest.h"
#include "i915_timeline.h"
@@ -162,7 +163,8 @@ typedef u64 gen8_ppgtt_pml4e_t;
#define GEN8_PDE_IPS_64K BIT(11)
#define GEN8_PDE_PS_2M BIT(7)
-struct sg_table;
+#define for_each_sgt_dma(__dmap, __iter, __sgt) \
+ __for_each_sgt_dma(__dmap, __iter, __sgt, I915_GTT_PAGE_SIZE)
struct intel_rotation_info {
struct intel_rotation_plane_info {
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index c86865a34972..707811256501 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -40,6 +40,7 @@
#include "i915_drv.h"
#include "i915_gpu_error.h"
+#include "i915_scatterlist.h"
#include "intel_atomic.h"
#include "intel_csr.h"
#include "intel_overlay.h"
diff --git a/drivers/gpu/drm/i915/i915_scatterlist.c b/drivers/gpu/drm/i915/i915_scatterlist.c
new file mode 100644
index 000000000000..cc6b3846a8c7
--- /dev/null
+++ b/drivers/gpu/drm/i915/i915_scatterlist.c
@@ -0,0 +1,39 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2016 Intel Corporation
+ */
+
+#include "i915_scatterlist.h"
+
+bool i915_sg_trim(struct sg_table *orig_st)
+{
+ struct sg_table new_st;
+ struct scatterlist *sg, *new_sg;
+ unsigned int i;
+
+ if (orig_st->nents == orig_st->orig_nents)
+ return false;
+
+ if (sg_alloc_table(&new_st, orig_st->nents, GFP_KERNEL | __GFP_NOWARN))
+ return false;
+
+ new_sg = new_st.sgl;
+ for_each_sg(orig_st->sgl, sg, orig_st->nents, i) {
+ sg_set_page(new_sg, sg_page(sg), sg->length, 0);
+ sg_dma_address(new_sg) = sg_dma_address(sg);
+ sg_dma_len(new_sg) = sg_dma_len(sg);
+
+ new_sg = sg_next(new_sg);
+ }
+ GEM_BUG_ON(new_sg); /* Should walk exactly nents and hit the end */
+
+ sg_free_table(orig_st);
+
+ *orig_st = new_st;
+ return true;
+}
+
+#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
+#include "selftests/scatterlist.c"
+#endif
diff --git a/drivers/gpu/drm/i915/i915_scatterlist.h b/drivers/gpu/drm/i915/i915_scatterlist.h
new file mode 100644
index 000000000000..6617963df9ed
--- /dev/null
+++ b/drivers/gpu/drm/i915/i915_scatterlist.h
@@ -0,0 +1,127 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2016 Intel Corporation
+ */
+
+#ifndef I915_SCATTERLIST_H
+#define I915_SCATTERLIST_H
+
+#include <linux/pfn.h>
+#include <linux/scatterlist.h>
+#include <linux/swiotlb.h>
+
+#include "i915_gem.h"
+
+/*
+ * Optimised SGL iterator for GEM objects
+ */
+static __always_inline struct sgt_iter {
+ struct scatterlist *sgp;
+ union {
+ unsigned long pfn;
+ dma_addr_t dma;
+ };
+ unsigned int curr;
+ unsigned int max;
+} __sgt_iter(struct scatterlist *sgl, bool dma) {
+ struct sgt_iter s = { .sgp = sgl };
+
+ if (s.sgp) {
+ s.max = s.curr = s.sgp->offset;
+ s.max += s.sgp->length;
+ if (dma)
+ s.dma = sg_dma_address(s.sgp);
+ else
+ s.pfn = page_to_pfn(sg_page(s.sgp));
+ }
+
+ return s;
+}
+
+static inline int __sg_page_count(const struct scatterlist *sg)
+{
+ return sg->length >> PAGE_SHIFT;
+}
+
+static inline struct scatterlist *____sg_next(struct scatterlist *sg)
+{
+ ++sg;
+ if (unlikely(sg_is_chain(sg)))
+ sg = sg_chain_ptr(sg);
+ return sg;
+}
+
+/**
+ * __sg_next - return the next scatterlist entry in a list
+ * @sg: The current sg entry
+ *
+ * Description:
+ * If the entry is the last, return NULL; otherwise, step to the next
+ * element in the array (@sg@+1). If that's a chain pointer, follow it;
+ * otherwise just return the pointer to the current element.
+ **/
+static inline struct scatterlist *__sg_next(struct scatterlist *sg)
+{
+ return sg_is_last(sg) ? NULL : ____sg_next(sg);
+}
+
+/**
+ * __for_each_sgt_dma - iterate over the DMA addresses of the given sg_table
+ * @__dmap: DMA address (output)
+ * @__iter: 'struct sgt_iter' (iterator state, internal)
+ * @__sgt: sg_table to iterate over (input)
+ * @__step: step size
+ */
+#define __for_each_sgt_dma(__dmap, __iter, __sgt, __step) \
+ for ((__iter) = __sgt_iter((__sgt)->sgl, true); \
+ ((__dmap) = (__iter).dma + (__iter).curr); \
+ (((__iter).curr += (__step)) >= (__iter).max) ? \
+ (__iter) = __sgt_iter(__sg_next((__iter).sgp), true), 0 : 0)
+
+/**
+ * for_each_sgt_page - iterate over the pages of the given sg_table
+ * @__pp: page pointer (output)
+ * @__iter: 'struct sgt_iter' (iterator state, internal)
+ * @__sgt: sg_table to iterate over (input)
+ */
+#define for_each_sgt_page(__pp, __iter, __sgt) \
+ for ((__iter) = __sgt_iter((__sgt)->sgl, false); \
+ ((__pp) = (__iter).pfn == 0 ? NULL : \
+ pfn_to_page((__iter).pfn + ((__iter).curr >> PAGE_SHIFT))); \
+ (((__iter).curr += PAGE_SIZE) >= (__iter).max) ? \
+ (__iter) = __sgt_iter(__sg_next((__iter).sgp), false), 0 : 0)
+
+static inline unsigned int i915_sg_page_sizes(struct scatterlist *sg)
+{
+ unsigned int page_sizes;
+
+ page_sizes = 0;
+ while (sg) {
+ GEM_BUG_ON(sg->offset);
+ GEM_BUG_ON(!IS_ALIGNED(sg->length, PAGE_SIZE));
+ page_sizes |= sg->length;
+ sg = __sg_next(sg);
+ }
+
+ return page_sizes;
+}
+
+static inline unsigned int i915_sg_segment_size(void)
+{
+ unsigned int size = swiotlb_max_segment();
+
+ if (size == 0)
+ return SCATTERLIST_MAX_SEGMENT;
+
+ size = rounddown(size, PAGE_SIZE);
+ /* swiotlb_max_segment_size can return 1 byte when it means one page. */
+ if (size < PAGE_SIZE)
+ size = PAGE_SIZE;
+
+ return size;
+}
+
+bool i915_sg_trim(struct sg_table *orig_st);
+
+#endif
diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c
index 131d0d9a6cd1..b0a4296a0504 100644
--- a/drivers/gpu/drm/i915/selftests/i915_vma.c
+++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
@@ -30,6 +30,7 @@
#include "mock_gtt.h"
#include "../gem/selftests/mock_context.h"
+#include "../i915_scatterlist.h"
static bool assert_vma(struct i915_vma *vma,
struct drm_i915_gem_object *obj,
diff --git a/drivers/gpu/drm/i915/selftests/scatterlist.c b/drivers/gpu/drm/i915/selftests/scatterlist.c
index cd6d2a16071f..5f0c48a9f051 100644
--- a/drivers/gpu/drm/i915/selftests/scatterlist.c
+++ b/drivers/gpu/drm/i915/selftests/scatterlist.c
@@ -25,6 +25,7 @@
#include <linux/random.h>
#include "../i915_selftest.h"
+#include "../i915_utils.h"
#define PFN_BIAS (1 << 10)
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 28/40] drm/i915: Move GEM object domain management from struct_mutex to local
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (25 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 27/40] drm/i915: Pull scatterlist utils out of i915_gem.h Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 29/40] drm/i915: Move GEM object waiting to its own file Chris Wilson
` (18 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx; +Cc: Matthew Auld
Use the per-object local lock to control the cache domain of the
individual GEM objects, not struct_mutex. This is a huge leap forward
for us in terms of object-level synchronisation; execbuffers are
coordinated using the ww_mutex and pread/pwrite is finally fully
serialised again.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
---
drivers/gpu/drm/i915/Makefile | 1 +
drivers/gpu/drm/i915/gem/i915_gem_clflush.c | 4 +-
drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c | 10 +-
drivers/gpu/drm/i915/gem/i915_gem_domain.c | 70 +++++-----
.../gpu/drm/i915/gem/i915_gem_execbuffer.c | 123 ++++++++++++------
drivers/gpu/drm/i915/gem/i915_gem_fence.c | 97 ++++++++++++++
drivers/gpu/drm/i915/gem/i915_gem_object.c | 2 +
drivers/gpu/drm/i915/gem/i915_gem_object.h | 14 ++
drivers/gpu/drm/i915/gem/i915_gem_pm.c | 7 +-
.../gpu/drm/i915/gem/selftests/huge_pages.c | 12 +-
.../i915/gem/selftests/i915_gem_coherency.c | 12 ++
.../drm/i915/gem/selftests/i915_gem_context.c | 20 +++
.../drm/i915/gem/selftests/i915_gem_mman.c | 6 +
.../drm/i915/gem/selftests/i915_gem_phys.c | 4 +-
drivers/gpu/drm/i915/gt/selftest_hangcheck.c | 4 +
drivers/gpu/drm/i915/gt/selftest_lrc.c | 2 +
.../gpu/drm/i915/gt/selftest_workarounds.c | 6 +
drivers/gpu/drm/i915/gvt/cmd_parser.c | 2 +
drivers/gpu/drm/i915/gvt/scheduler.c | 8 +-
drivers/gpu/drm/i915/i915_cmd_parser.c | 23 ++--
drivers/gpu/drm/i915/i915_gem.c | 122 +++++++++--------
drivers/gpu/drm/i915/i915_gem_gtt.c | 5 +-
drivers/gpu/drm/i915/i915_gem_render_state.c | 2 +
drivers/gpu/drm/i915/i915_vma.c | 8 +-
drivers/gpu/drm/i915/i915_vma.h | 12 ++
drivers/gpu/drm/i915/intel_display.c | 5 +
drivers/gpu/drm/i915/intel_guc_log.c | 6 +-
drivers/gpu/drm/i915/intel_overlay.c | 25 ++--
drivers/gpu/drm/i915/intel_uc_fw.c | 6 +-
drivers/gpu/drm/i915/selftests/i915_request.c | 4 +
drivers/gpu/drm/i915/selftests/igt_spinner.c | 2 +
31 files changed, 444 insertions(+), 180 deletions(-)
create mode 100644 drivers/gpu/drm/i915/gem/i915_gem_fence.c
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index 9f5f4acacae5..e5348c355987 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -93,6 +93,7 @@ gem-y += \
gem/i915_gem_dmabuf.o \
gem/i915_gem_domain.o \
gem/i915_gem_execbuffer.o \
+ gem/i915_gem_fence.o \
gem/i915_gem_internal.o \
gem/i915_gem_object.o \
gem/i915_gem_mman.o \
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_clflush.c b/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
index 093bfff55a96..efab47250588 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
@@ -96,6 +96,8 @@ bool i915_gem_clflush_object(struct drm_i915_gem_object *obj,
{
struct clflush *clflush;
+ assert_object_held(obj);
+
/*
* Stolen memory is always coherent with the GPU as it is explicitly
* marked as wc by the system, or the system is cache-coherent.
@@ -145,9 +147,7 @@ bool i915_gem_clflush_object(struct drm_i915_gem_object *obj,
true, I915_FENCE_TIMEOUT,
I915_FENCE_GFP);
- reservation_object_lock(obj->resv, NULL);
reservation_object_add_excl_fence(obj->resv, &clflush->dma);
- reservation_object_unlock(obj->resv);
i915_sw_fence_commit(&clflush->wait);
} else if (obj->mm.pages) {
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c b/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
index 4e7efe159531..50981ea513f0 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
@@ -152,7 +152,6 @@ static int i915_gem_dmabuf_mmap(struct dma_buf *dma_buf, struct vm_area_struct *
static int i915_gem_begin_cpu_access(struct dma_buf *dma_buf, enum dma_data_direction direction)
{
struct drm_i915_gem_object *obj = dma_buf_to_obj(dma_buf);
- struct drm_device *dev = obj->base.dev;
bool write = (direction == DMA_BIDIRECTIONAL || direction == DMA_TO_DEVICE);
int err;
@@ -160,12 +159,12 @@ static int i915_gem_begin_cpu_access(struct dma_buf *dma_buf, enum dma_data_dire
if (err)
return err;
- err = i915_mutex_lock_interruptible(dev);
+ err = i915_gem_object_lock_interruptible(obj);
if (err)
goto out;
err = i915_gem_object_set_to_cpu_domain(obj, write);
- mutex_unlock(&dev->struct_mutex);
+ i915_gem_object_unlock(obj);
out:
i915_gem_object_unpin_pages(obj);
@@ -175,19 +174,18 @@ static int i915_gem_begin_cpu_access(struct dma_buf *dma_buf, enum dma_data_dire
static int i915_gem_end_cpu_access(struct dma_buf *dma_buf, enum dma_data_direction direction)
{
struct drm_i915_gem_object *obj = dma_buf_to_obj(dma_buf);
- struct drm_device *dev = obj->base.dev;
int err;
err = i915_gem_object_pin_pages(obj);
if (err)
return err;
- err = i915_mutex_lock_interruptible(dev);
+ err = i915_gem_object_lock_interruptible(obj);
if (err)
goto out;
err = i915_gem_object_set_to_gtt_domain(obj, false);
- mutex_unlock(&dev->struct_mutex);
+ i915_gem_object_unlock(obj);
out:
i915_gem_object_unpin_pages(obj);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_domain.c b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
index 5e1429f8e910..4c84bb911d6c 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_domain.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
@@ -31,9 +31,9 @@ void i915_gem_object_flush_if_display(struct drm_i915_gem_object *obj)
if (!READ_ONCE(obj->pin_global))
return;
- mutex_lock(&obj->base.dev->struct_mutex);
+ i915_gem_object_lock(obj);
__i915_gem_object_flush_for_display(obj);
- mutex_unlock(&obj->base.dev->struct_mutex);
+ i915_gem_object_unlock(obj);
}
/**
@@ -49,11 +49,10 @@ i915_gem_object_set_to_wc_domain(struct drm_i915_gem_object *obj, bool write)
{
int ret;
- lockdep_assert_held(&obj->base.dev->struct_mutex);
+ assert_object_held(obj);
ret = i915_gem_object_wait(obj,
I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED |
(write ? I915_WAIT_ALL : 0),
MAX_SCHEDULE_TIMEOUT);
if (ret)
@@ -111,11 +110,10 @@ i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write)
{
int ret;
- lockdep_assert_held(&obj->base.dev->struct_mutex);
+ assert_object_held(obj);
ret = i915_gem_object_wait(obj,
I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED |
(write ? I915_WAIT_ALL : 0),
MAX_SCHEDULE_TIMEOUT);
if (ret)
@@ -181,7 +179,7 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
struct i915_vma *vma;
int ret;
- lockdep_assert_held(&obj->base.dev->struct_mutex);
+ assert_object_held(obj);
if (obj->cache_level == cache_level)
return 0;
@@ -230,7 +228,6 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
*/
ret = i915_gem_object_wait(obj,
I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED |
I915_WAIT_ALL,
MAX_SCHEDULE_TIMEOUT);
if (ret)
@@ -374,12 +371,16 @@ int i915_gem_set_caching_ioctl(struct drm_device *dev, void *data,
if (ret)
goto out;
- ret = i915_mutex_lock_interruptible(dev);
+ ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
if (ret)
goto out;
- ret = i915_gem_object_set_cache_level(obj, level);
- mutex_unlock(&dev->struct_mutex);
+ ret = i915_gem_object_lock_interruptible(obj);
+ if (ret == 0) {
+ ret = i915_gem_object_set_cache_level(obj, level);
+ i915_gem_object_unlock(obj);
+ }
+ mutex_unlock(&i915->drm.struct_mutex);
out:
i915_gem_object_put(obj);
@@ -401,7 +402,7 @@ i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,
struct i915_vma *vma;
int ret;
- lockdep_assert_held(&obj->base.dev->struct_mutex);
+ assert_object_held(obj);
/* Mark the global pin early so that we account for the
* display coherency whilst setting up the cache domains.
@@ -486,16 +487,18 @@ static void i915_gem_object_bump_inactive_ggtt(struct drm_i915_gem_object *obj)
void
i915_gem_object_unpin_from_display_plane(struct i915_vma *vma)
{
- lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
+ struct drm_i915_gem_object *obj = vma->obj;
+
+ assert_object_held(obj);
- if (WARN_ON(vma->obj->pin_global == 0))
+ if (WARN_ON(obj->pin_global == 0))
return;
- if (--vma->obj->pin_global == 0)
+ if (--obj->pin_global == 0)
vma->display_alignment = I915_GTT_MIN_ALIGNMENT;
/* Bump the LRU to try and avoid premature eviction whilst flipping */
- i915_gem_object_bump_inactive_ggtt(vma->obj);
+ i915_gem_object_bump_inactive_ggtt(obj);
i915_vma_unpin(vma);
}
@@ -513,11 +516,10 @@ i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write)
{
int ret;
- lockdep_assert_held(&obj->base.dev->struct_mutex);
+ assert_object_held(obj);
ret = i915_gem_object_wait(obj,
I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED |
(write ? I915_WAIT_ALL : 0),
MAX_SCHEDULE_TIMEOUT);
if (ret)
@@ -639,7 +641,7 @@ i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
if (err)
goto out;
- err = i915_mutex_lock_interruptible(dev);
+ err = i915_gem_object_lock_interruptible(obj);
if (err)
goto out_unpin;
@@ -653,7 +655,7 @@ i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
/* And bump the LRU for this access */
i915_gem_object_bump_inactive_ggtt(obj);
- mutex_unlock(&dev->struct_mutex);
+ i915_gem_object_unlock(obj);
if (write_domain != 0)
intel_fb_obj_invalidate(obj,
@@ -676,22 +678,23 @@ int i915_gem_object_prepare_read(struct drm_i915_gem_object *obj,
{
int ret;
- lockdep_assert_held(&obj->base.dev->struct_mutex);
-
*needs_clflush = 0;
if (!i915_gem_object_has_struct_page(obj))
return -ENODEV;
+ ret = i915_gem_object_lock_interruptible(obj);
+ if (ret)
+ return ret;
+
ret = i915_gem_object_wait(obj,
- I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED,
+ I915_WAIT_INTERRUPTIBLE,
MAX_SCHEDULE_TIMEOUT);
if (ret)
- return ret;
+ goto err_unlock;
ret = i915_gem_object_pin_pages(obj);
if (ret)
- return ret;
+ goto err_unlock;
if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ ||
!static_cpu_has(X86_FEATURE_CLFLUSH)) {
@@ -719,6 +722,8 @@ int i915_gem_object_prepare_read(struct drm_i915_gem_object *obj,
err_unpin:
i915_gem_object_unpin_pages(obj);
+err_unlock:
+ i915_gem_object_unlock(obj);
return ret;
}
@@ -727,23 +732,24 @@ int i915_gem_object_prepare_write(struct drm_i915_gem_object *obj,
{
int ret;
- lockdep_assert_held(&obj->base.dev->struct_mutex);
-
*needs_clflush = 0;
if (!i915_gem_object_has_struct_page(obj))
return -ENODEV;
+ ret = i915_gem_object_lock_interruptible(obj);
+ if (ret)
+ return ret;
+
ret = i915_gem_object_wait(obj,
I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED |
I915_WAIT_ALL,
MAX_SCHEDULE_TIMEOUT);
if (ret)
- return ret;
+ goto err_unlock;
ret = i915_gem_object_pin_pages(obj);
if (ret)
- return ret;
+ goto err_unlock;
if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE ||
!static_cpu_has(X86_FEATURE_CLFLUSH)) {
@@ -780,5 +786,7 @@ int i915_gem_object_prepare_write(struct drm_i915_gem_object *obj,
err_unpin:
i915_gem_object_unpin_pages(obj);
+err_unlock:
+ i915_gem_object_unlock(obj);
return ret;
}
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index a2761029f0c3..080f69358224 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -1076,7 +1076,9 @@ static void *reloc_iomap(struct drm_i915_gem_object *obj,
if (use_cpu_reloc(cache, obj))
return NULL;
+ i915_gem_object_lock(obj);
err = i915_gem_object_set_to_gtt_domain(obj, true);
+ i915_gem_object_unlock(obj);
if (err)
return ERR_PTR(err);
@@ -1165,6 +1167,26 @@ static void clflush_write32(u32 *addr, u32 value, unsigned int flushes)
*addr = value;
}
+static int reloc_move_to_gpu(struct i915_request *rq, struct i915_vma *vma)
+{
+ struct drm_i915_gem_object *obj = vma->obj;
+ int err;
+
+ i915_vma_lock(vma);
+
+ if (obj->cache_dirty & ~obj->cache_coherent)
+ i915_gem_clflush_object(obj, 0);
+ obj->write_domain = 0;
+
+ err = i915_request_await_object(rq, vma->obj, true);
+ if (err == 0)
+ err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+
+ i915_vma_unlock(vma);
+
+ return err;
+}
+
static int __reloc_gpu_alloc(struct i915_execbuffer *eb,
struct i915_vma *vma,
unsigned int len)
@@ -1176,15 +1198,6 @@ static int __reloc_gpu_alloc(struct i915_execbuffer *eb,
u32 *cmd;
int err;
- if (DBG_FORCE_RELOC == FORCE_GPU_RELOC) {
- obj = vma->obj;
- if (obj->cache_dirty & ~obj->cache_coherent)
- i915_gem_clflush_object(obj, 0);
- obj->write_domain = 0;
- }
-
- GEM_BUG_ON(vma->obj->write_domain & I915_GEM_DOMAIN_CPU);
-
obj = i915_gem_batch_pool_get(&eb->engine->batch_pool, PAGE_SIZE);
if (IS_ERR(obj))
return PTR_ERR(obj);
@@ -1213,7 +1226,7 @@ static int __reloc_gpu_alloc(struct i915_execbuffer *eb,
goto err_unpin;
}
- err = i915_request_await_object(rq, vma->obj, true);
+ err = reloc_move_to_gpu(rq, vma);
if (err)
goto err_request;
@@ -1221,14 +1234,12 @@ static int __reloc_gpu_alloc(struct i915_execbuffer *eb,
batch->node.start, PAGE_SIZE,
cache->gen > 5 ? 0 : I915_DISPATCH_SECURE);
if (err)
- goto err_request;
+ goto skip_request;
+ i915_vma_lock(batch);
GEM_BUG_ON(!reservation_object_test_signaled_rcu(batch->resv, true));
err = i915_vma_move_to_active(batch, rq, 0);
- if (err)
- goto skip_request;
-
- err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+ i915_vma_unlock(batch);
if (err)
goto skip_request;
@@ -1837,24 +1848,59 @@ static int eb_relocate(struct i915_execbuffer *eb)
static int eb_move_to_gpu(struct i915_execbuffer *eb)
{
const unsigned int count = eb->buffer_count;
+ struct ww_acquire_ctx acquire;
unsigned int i;
- int err;
+ int err = 0;
+
+ ww_acquire_init(&acquire, &reservation_ww_class);
for (i = 0; i < count; i++) {
+ struct i915_vma *vma = eb->vma[i];
+
+ err = ww_mutex_lock_interruptible(&vma->resv->lock, &acquire);
+ if (!err)
+ continue;
+
+ GEM_BUG_ON(err == -EALREADY); /* No duplicate vma */
+
+ if (err == -EDEADLK) {
+ GEM_BUG_ON(i == 0);
+ do {
+ int j = i - 1;
+
+ ww_mutex_unlock(&eb->vma[j]->resv->lock);
+
+ swap(eb->flags[i], eb->flags[j]);
+ swap(eb->vma[i], eb->vma[j]);
+ eb->vma[i]->exec_flags = &eb->flags[i];
+ } while (--i);
+ GEM_BUG_ON(vma != eb->vma[0]);
+ vma->exec_flags = &eb->flags[0];
+
+ err = ww_mutex_lock_slow_interruptible(&vma->resv->lock,
+ &acquire);
+ }
+ if (err)
+ break;
+ }
+ ww_acquire_done(&acquire);
+
+ while (i--) {
unsigned int flags = eb->flags[i];
struct i915_vma *vma = eb->vma[i];
struct drm_i915_gem_object *obj = vma->obj;
+ assert_vma_held(vma);
+
if (flags & EXEC_OBJECT_CAPTURE) {
struct i915_capture_list *capture;
capture = kmalloc(sizeof(*capture), GFP_KERNEL);
- if (unlikely(!capture))
- return -ENOMEM;
-
- capture->next = eb->request->capture_list;
- capture->vma = eb->vma[i];
- eb->request->capture_list = capture;
+ if (capture) {
+ capture->next = eb->request->capture_list;
+ capture->vma = vma;
+ eb->request->capture_list = capture;
+ }
}
/*
@@ -1874,24 +1920,15 @@ static int eb_move_to_gpu(struct i915_execbuffer *eb)
flags &= ~EXEC_OBJECT_ASYNC;
}
- if (flags & EXEC_OBJECT_ASYNC)
- continue;
-
- err = i915_request_await_object
- (eb->request, obj, flags & EXEC_OBJECT_WRITE);
- if (err)
- return err;
- }
+ if (err == 0 && !(flags & EXEC_OBJECT_ASYNC)) {
+ err = i915_request_await_object
+ (eb->request, obj, flags & EXEC_OBJECT_WRITE);
+ }
- for (i = 0; i < count; i++) {
- unsigned int flags = eb->flags[i];
- struct i915_vma *vma = eb->vma[i];
+ if (err == 0)
+ err = i915_vma_move_to_active(vma, eb->request, flags);
- err = i915_vma_move_to_active(vma, eb->request, flags);
- if (unlikely(err)) {
- i915_request_skip(eb->request, err);
- return err;
- }
+ i915_vma_unlock(vma);
__eb_unreserve_vma(vma, flags);
vma->exec_flags = NULL;
@@ -1899,12 +1936,20 @@ static int eb_move_to_gpu(struct i915_execbuffer *eb)
if (unlikely(flags & __EXEC_OBJECT_HAS_REF))
i915_vma_put(vma);
}
+ ww_acquire_fini(&acquire);
+
+ if (unlikely(err))
+ goto err_skip;
+
eb->exec = NULL;
/* Unconditionally flush any chipset caches (for streaming writes). */
i915_gem_chipset_flush(eb->i915);
-
return 0;
+
+err_skip:
+ i915_request_skip(eb->request, err);
+ return err;
}
static bool i915_gem_check_execbuffer(struct drm_i915_gem_execbuffer2 *exec)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_fence.c b/drivers/gpu/drm/i915/gem/i915_gem_fence.c
new file mode 100644
index 000000000000..aa2efdc57c92
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/i915_gem_fence.c
@@ -0,0 +1,97 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2019 Intel Corporation
+ */
+
+#include "i915_gem_object.h"
+
+#include "../i915_drv.h"
+
+struct stub_fence {
+ struct dma_fence dma;
+ struct i915_sw_fence chain;
+};
+
+static int __i915_sw_fence_call
+stub_notify(struct i915_sw_fence *fence, enum i915_sw_fence_notify state)
+{
+ struct stub_fence *stub = container_of(fence, typeof(*stub), chain);
+
+ switch (state) {
+ case FENCE_COMPLETE:
+ dma_fence_signal(&stub->dma);
+ break;
+
+ case FENCE_FREE:
+ dma_fence_put(&stub->dma);
+ break;
+ }
+
+ return NOTIFY_DONE;
+}
+
+static const char *stub_driver_name(struct dma_fence *fence)
+{
+ return DRIVER_NAME;
+}
+
+static const char *stub_timeline_name(struct dma_fence *fence)
+{
+ return "object";
+}
+
+static void stub_release(struct dma_fence *fence)
+{
+ struct stub_fence *stub = container_of(fence, typeof(*stub), dma);
+
+ i915_sw_fence_fini(&stub->chain);
+
+ BUILD_BUG_ON(offsetof(typeof(*stub), dma));
+ dma_fence_free(&stub->dma);
+}
+
+static const struct dma_fence_ops stub_fence_ops = {
+ .get_driver_name = stub_driver_name,
+ .get_timeline_name = stub_timeline_name,
+ .release = stub_release,
+};
+
+struct dma_fence *
+i915_gem_object_lock_fence(struct drm_i915_gem_object *obj)
+{
+ struct stub_fence *stub;
+
+ assert_object_held(obj);
+
+ stub = kmalloc(sizeof(*stub), GFP_KERNEL);
+ if (!stub)
+ return NULL;
+
+ i915_sw_fence_init(&stub->chain, stub_notify);
+ dma_fence_init(&stub->dma, &stub_fence_ops, &stub->chain.wait.lock,
+ to_i915(obj->base.dev)->mm.unordered_timeline,
+ 0);
+
+ if (i915_sw_fence_await_reservation(&stub->chain,
+ obj->resv, NULL,
+ true, I915_FENCE_TIMEOUT,
+ I915_FENCE_GFP) < 0)
+ goto err;
+
+ reservation_object_add_excl_fence(obj->resv, &stub->dma);
+
+ return &stub->dma;
+
+err:
+ stub_release(&stub->dma);
+ return NULL;
+}
+
+void i915_gem_object_unlock_fence(struct drm_i915_gem_object *obj,
+ struct dma_fence *fence)
+{
+ struct stub_fence *stub = container_of(fence, typeof(*stub), dma);
+
+ i915_sw_fence_commit(&stub->chain);
+}
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
index 04928ae415eb..df79e2eead62 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
@@ -379,6 +379,8 @@ i915_gem_object_flush_write_domain(struct drm_i915_gem_object *obj,
struct drm_i915_private *dev_priv = to_i915(obj->base.dev);
struct i915_vma *vma;
+ assert_object_held(obj);
+
if (!(obj->write_domain & flush_domains))
return;
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index 980587e420e6..509d145d808a 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -99,16 +99,29 @@ i915_gem_object_put(struct drm_i915_gem_object *obj)
__drm_gem_object_put(&obj->base);
}
+#define assert_object_held(obj) reservation_object_assert_held((obj)->resv)
+
static inline void i915_gem_object_lock(struct drm_i915_gem_object *obj)
{
reservation_object_lock(obj->resv, NULL);
}
+static inline int
+i915_gem_object_lock_interruptible(struct drm_i915_gem_object *obj)
+{
+ return reservation_object_lock_interruptible(obj->resv, NULL);
+}
+
static inline void i915_gem_object_unlock(struct drm_i915_gem_object *obj)
{
reservation_object_unlock(obj->resv);
}
+struct dma_fence *
+i915_gem_object_lock_fence(struct drm_i915_gem_object *obj);
+void i915_gem_object_unlock_fence(struct drm_i915_gem_object *obj,
+ struct dma_fence *fence);
+
static inline void
i915_gem_object_set_readonly(struct drm_i915_gem_object *obj)
{
@@ -367,6 +380,7 @@ static inline void
i915_gem_object_finish_access(struct drm_i915_gem_object *obj)
{
i915_gem_object_unpin_pages(obj);
+ i915_gem_object_unlock(obj);
}
static inline struct intel_engine_cs *
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
index be75a52a7ab0..111b687f14b5 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
@@ -186,12 +186,13 @@ void i915_gem_suspend_late(struct drm_i915_private *i915)
* machine in an unusable condition.
*/
- mutex_lock(&i915->drm.struct_mutex);
for (phase = phases; *phase; phase++) {
- list_for_each_entry(obj, *phase, mm.link)
+ list_for_each_entry(obj, *phase, mm.link) {
+ i915_gem_object_lock(obj);
WARN_ON(i915_gem_object_set_to_gtt_domain(obj, false));
+ i915_gem_object_unlock(obj);
+ }
}
- mutex_unlock(&i915->drm.struct_mutex);
intel_uc_sanitize(i915);
i915_gem_sanitize(i915);
diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
index 88ca4ed430ee..5a89cbab3b2e 100644
--- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
+++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
@@ -959,10 +959,6 @@ static int gpu_write(struct i915_vma *vma,
GEM_BUG_ON(!intel_engine_can_store_dword(engine));
- err = i915_gem_object_set_to_gtt_domain(vma->obj, true);
- if (err)
- return err;
-
batch = gpu_write_dw(vma, dword * sizeof(u32), value);
if (IS_ERR(batch))
return PTR_ERR(batch);
@@ -973,13 +969,19 @@ static int gpu_write(struct i915_vma *vma,
goto err_batch;
}
+ i915_vma_lock(batch);
err = i915_vma_move_to_active(batch, rq, 0);
+ i915_vma_unlock(batch);
if (err)
goto err_request;
i915_gem_object_set_active_reference(batch->obj);
- err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+ i915_vma_lock(vma);
+ err = i915_gem_object_set_to_gtt_domain(vma->obj, false);
+ if (err == 0)
+ err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+ i915_vma_unlock(vma);
if (err)
goto err_request;
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
index b6613caa6f45..22fe4b6fd506 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
@@ -78,7 +78,9 @@ static int gtt_set(struct drm_i915_gem_object *obj,
u32 __iomem *map;
int err;
+ i915_gem_object_lock(obj);
err = i915_gem_object_set_to_gtt_domain(obj, true);
+ i915_gem_object_unlock(obj);
if (err)
return err;
@@ -105,7 +107,9 @@ static int gtt_get(struct drm_i915_gem_object *obj,
u32 __iomem *map;
int err;
+ i915_gem_object_lock(obj);
err = i915_gem_object_set_to_gtt_domain(obj, false);
+ i915_gem_object_unlock(obj);
if (err)
return err;
@@ -131,7 +135,9 @@ static int wc_set(struct drm_i915_gem_object *obj,
u32 *map;
int err;
+ i915_gem_object_lock(obj);
err = i915_gem_object_set_to_wc_domain(obj, true);
+ i915_gem_object_unlock(obj);
if (err)
return err;
@@ -152,7 +158,9 @@ static int wc_get(struct drm_i915_gem_object *obj,
u32 *map;
int err;
+ i915_gem_object_lock(obj);
err = i915_gem_object_set_to_wc_domain(obj, false);
+ i915_gem_object_unlock(obj);
if (err)
return err;
@@ -176,7 +184,9 @@ static int gpu_set(struct drm_i915_gem_object *obj,
u32 *cs;
int err;
+ i915_gem_object_lock(obj);
err = i915_gem_object_set_to_gtt_domain(obj, true);
+ i915_gem_object_unlock(obj);
if (err)
return err;
@@ -215,7 +225,9 @@ static int gpu_set(struct drm_i915_gem_object *obj,
}
intel_ring_advance(rq, cs);
+ i915_vma_lock(vma);
err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+ i915_vma_unlock(vma);
i915_vma_unpin(vma);
i915_request_add(rq);
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
index 653ae08a277f..72eedd6c2a0a 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
@@ -209,7 +209,9 @@ gpu_fill_dw(struct i915_vma *vma, u64 offset, unsigned long count, u32 value)
i915_gem_object_flush_map(obj);
i915_gem_object_unpin_map(obj);
+ i915_gem_object_lock(obj);
err = i915_gem_object_set_to_gtt_domain(obj, false);
+ i915_gem_object_unlock(obj);
if (err)
goto err;
@@ -261,7 +263,9 @@ static int gpu_fill(struct drm_i915_gem_object *obj,
if (IS_ERR(vma))
return PTR_ERR(vma);
+ i915_gem_object_lock(obj);
err = i915_gem_object_set_to_gtt_domain(obj, false);
+ i915_gem_object_unlock(obj);
if (err)
return err;
@@ -302,11 +306,15 @@ static int gpu_fill(struct drm_i915_gem_object *obj,
if (err)
goto err_request;
+ i915_vma_lock(batch);
err = i915_vma_move_to_active(batch, rq, 0);
+ i915_vma_unlock(batch);
if (err)
goto skip_request;
+ i915_vma_lock(vma);
err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+ i915_vma_unlock(vma);
if (err)
goto skip_request;
@@ -754,7 +762,9 @@ emit_rpcs_query(struct drm_i915_gem_object *obj,
if (IS_ERR(vma))
return PTR_ERR(vma);
+ i915_gem_object_lock(obj);
err = i915_gem_object_set_to_gtt_domain(obj, false);
+ i915_gem_object_unlock(obj);
if (err)
return err;
@@ -780,11 +790,15 @@ emit_rpcs_query(struct drm_i915_gem_object *obj,
if (err)
goto err_request;
+ i915_vma_lock(batch);
err = i915_vma_move_to_active(batch, rq, 0);
+ i915_vma_unlock(batch);
if (err)
goto skip_request;
+ i915_vma_lock(vma);
err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+ i915_vma_unlock(vma);
if (err)
goto skip_request;
@@ -1345,7 +1359,9 @@ static int write_to_scratch(struct i915_gem_context *ctx,
if (err)
goto err_request;
+ i915_vma_lock(vma);
err = i915_vma_move_to_active(vma, rq, 0);
+ i915_vma_unlock(vma);
if (err)
goto skip_request;
@@ -1440,7 +1456,9 @@ static int read_from_scratch(struct i915_gem_context *ctx,
if (err)
goto err_request;
+ i915_vma_lock(vma);
err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+ i915_vma_unlock(vma);
if (err)
goto skip_request;
@@ -1449,7 +1467,9 @@ static int read_from_scratch(struct i915_gem_context *ctx,
i915_request_add(rq);
+ i915_gem_object_lock(obj);
err = i915_gem_object_set_to_cpu_domain(obj, false);
+ i915_gem_object_unlock(obj);
if (err)
goto err;
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
index 12c90d8fe0fb..297f8864d392 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
@@ -110,7 +110,9 @@ static int check_partial_mapping(struct drm_i915_gem_object *obj,
GEM_BUG_ON(view.partial.size > nreal);
cond_resched();
+ i915_gem_object_lock(obj);
err = i915_gem_object_set_to_gtt_domain(obj, true);
+ i915_gem_object_unlock(obj);
if (err) {
pr_err("Failed to flush to GTT write domain; err=%d\n",
err);
@@ -142,7 +144,9 @@ static int check_partial_mapping(struct drm_i915_gem_object *obj,
if (offset >= obj->base.size)
continue;
+ i915_gem_object_lock(obj);
i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+ i915_gem_object_unlock(obj);
p = i915_gem_object_get_page(obj, offset >> PAGE_SHIFT);
cpu = kmap(p) + offset_in_page(offset);
@@ -344,7 +348,9 @@ static int make_obj_busy(struct drm_i915_gem_object *obj)
return PTR_ERR(rq);
}
+ i915_vma_lock(vma);
err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+ i915_vma_unlock(vma);
i915_request_add(rq);
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c
index b76b503b3999..8ecf8d7bd083 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c
@@ -46,9 +46,9 @@ static int mock_phys_object(void *arg)
}
/* Make the object dirty so that put_pages must do copy back the data */
- mutex_lock(&i915->drm.struct_mutex);
+ i915_gem_object_lock(obj);
err = i915_gem_object_set_to_gtt_domain(obj, true);
- mutex_unlock(&i915->drm.struct_mutex);
+ i915_gem_object_unlock(obj);
if (err) {
pr_err("i915_gem_object_set_to_gtt_domain failed with err=%d\n",
err);
diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
index 4b4589b92521..45745e36b5f4 100644
--- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
+++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
@@ -115,7 +115,9 @@ static int move_to_active(struct i915_vma *vma,
{
int err;
+ i915_vma_lock(vma);
err = i915_vma_move_to_active(vma, rq, flags);
+ i915_vma_unlock(vma);
if (err)
return err;
@@ -1298,7 +1300,9 @@ static int __igt_reset_evict_vma(struct drm_i915_private *i915,
}
}
+ i915_vma_lock(arg.vma);
err = i915_vma_move_to_active(arg.vma, rq, flags);
+ i915_vma_unlock(arg.vma);
if (flags & EXEC_OBJECT_NEEDS_FENCE)
i915_vma_unpin_fence(arg.vma);
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index 8e5ddb42fa3e..c96cd69f9800 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -1108,11 +1108,13 @@ static int smoke_submit(struct preempt_smoke *smoke,
}
if (vma) {
+ i915_vma_lock(vma);
err = rq->engine->emit_bb_start(rq,
vma->node.start,
PAGE_SIZE, 0);
if (!err)
err = i915_vma_move_to_active(vma, rq, 0);
+ i915_vma_unlock(vma);
}
i915_request_add(rq);
diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
index 40b6df911d8d..d4ba5296b2c8 100644
--- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
@@ -111,7 +111,9 @@ read_nonprivs(struct i915_gem_context *ctx, struct intel_engine_cs *engine)
goto err_pin;
}
+ i915_vma_lock(vma);
err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+ i915_vma_unlock(vma);
if (err)
goto err_req;
@@ -188,8 +190,10 @@ static int check_whitelist(struct i915_gem_context *ctx,
return PTR_ERR(results);
err = 0;
+ i915_gem_object_lock(results);
igt_wedge_on_timeout(&wedge, ctx->i915, HZ / 5) /* a safety net! */
err = i915_gem_object_set_to_cpu_domain(results, false);
+ i915_gem_object_unlock(results);
if (i915_terminally_wedged(ctx->i915))
err = -EIO;
if (err)
@@ -360,7 +364,9 @@ static struct i915_vma *create_batch(struct i915_gem_context *ctx)
if (err)
goto err_obj;
+ i915_gem_object_lock(obj);
err = i915_gem_object_set_to_wc_domain(obj, true);
+ i915_gem_object_unlock(obj);
if (err)
goto err_obj;
diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c
index 1ee881825e38..f6d15683868a 100644
--- a/drivers/gpu/drm/i915/gvt/cmd_parser.c
+++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c
@@ -2840,7 +2840,9 @@ static int shadow_indirect_ctx(struct intel_shadow_wa_ctx *wa_ctx)
goto put_obj;
}
+ i915_gem_object_lock(obj);
ret = i915_gem_object_set_to_cpu_domain(obj, false);
+ i915_gem_object_unlock(obj);
if (ret) {
gvt_vgpu_err("failed to set shadow indirect ctx to CPU\n");
goto unmap_src;
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index 0c3724d0b7b5..840cac06add7 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -492,18 +492,18 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
}
ret = i915_gem_object_set_to_gtt_domain(bb->obj,
- false);
+ false);
if (ret)
goto err;
- i915_gem_object_finish_access(bb->obj);
- bb->accessing = false;
-
ret = i915_vma_move_to_active(bb->vma,
workload->req,
0);
if (ret)
goto err;
+
+ i915_gem_object_finish_access(bb->obj);
+ bb->accessing = false;
}
}
return 0;
diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index c893bd4eb2c8..a28bcd2d7c09 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -1058,19 +1058,20 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
void *dst, *src;
int ret;
- ret = i915_gem_object_prepare_read(src_obj, &src_needs_clflush);
+ ret = i915_gem_object_prepare_write(dst_obj, &dst_needs_clflush);
if (ret)
return ERR_PTR(ret);
- ret = i915_gem_object_prepare_write(dst_obj, &dst_needs_clflush);
- if (ret) {
- dst = ERR_PTR(ret);
- goto unpin_src;
- }
-
dst = i915_gem_object_pin_map(dst_obj, I915_MAP_FORCE_WB);
+ i915_gem_object_finish_access(dst_obj);
if (IS_ERR(dst))
- goto unpin_dst;
+ return dst;
+
+ ret = i915_gem_object_prepare_read(src_obj, &src_needs_clflush);
+ if (ret) {
+ i915_gem_object_unpin_map(dst_obj);
+ return ERR_PTR(ret);
+ }
src = ERR_PTR(-ENODEV);
if (src_needs_clflush &&
@@ -1116,13 +1117,11 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
}
}
+ i915_gem_object_finish_access(src_obj);
+
/* dst_obj is returned with vmap pinned */
*needs_clflush_after = dst_needs_clflush & CLFLUSH_AFTER;
-unpin_dst:
- i915_gem_object_finish_access(dst_obj);
-unpin_src:
- i915_gem_object_finish_access(src_obj);
return dst;
}
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 7fa1236677b9..32fdc1977afe 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -103,19 +103,10 @@ int i915_gem_object_unbind(struct drm_i915_gem_object *obj)
{
struct i915_vma *vma;
LIST_HEAD(still_in_list);
- int ret;
+ int ret = 0;
lockdep_assert_held(&obj->base.dev->struct_mutex);
- /* Closed vma are removed from the obj->vma_list - but they may
- * still have an active binding on the object. To remove those we
- * must wait for all rendering to complete to the object (as unbinding
- * must anyway), and retire the requests.
- */
- ret = i915_gem_object_set_to_cpu_domain(obj, false);
- if (ret)
- return ret;
-
spin_lock(&obj->vma.lock);
while (!ret && (vma = list_first_entry_or_null(&obj->vma.list,
struct i915_vma,
@@ -138,29 +129,17 @@ i915_gem_object_wait_fence(struct dma_fence *fence,
unsigned int flags,
long timeout)
{
- struct i915_request *rq;
-
BUILD_BUG_ON(I915_WAIT_INTERRUPTIBLE != 0x1);
if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
return timeout;
- if (!dma_fence_is_i915(fence))
- return dma_fence_wait_timeout(fence,
- flags & I915_WAIT_INTERRUPTIBLE,
- timeout);
+ if (dma_fence_is_i915(fence))
+ return i915_request_wait(to_request(fence), flags, timeout);
- rq = to_request(fence);
- if (i915_request_completed(rq))
- goto out;
-
- timeout = i915_request_wait(rq, flags, timeout);
-
-out:
- if (flags & I915_WAIT_LOCKED && i915_request_completed(rq))
- i915_request_retire_upto(rq);
-
- return timeout;
+ return dma_fence_wait_timeout(fence,
+ flags & I915_WAIT_INTERRUPTIBLE,
+ timeout);
}
static long
@@ -463,21 +442,22 @@ static int
i915_gem_shmem_pread(struct drm_i915_gem_object *obj,
struct drm_i915_gem_pread *args)
{
- char __user *user_data;
- u64 remain;
unsigned int needs_clflush;
unsigned int idx, offset;
+ struct dma_fence *fence;
+ char __user *user_data;
+ u64 remain;
int ret;
- ret = mutex_lock_interruptible(&obj->base.dev->struct_mutex);
- if (ret)
- return ret;
-
ret = i915_gem_object_prepare_read(obj, &needs_clflush);
- mutex_unlock(&obj->base.dev->struct_mutex);
if (ret)
return ret;
+ fence = i915_gem_object_lock_fence(obj);
+ i915_gem_object_finish_access(obj);
+ if (!fence)
+ return -ENOMEM;
+
remain = args->size;
user_data = u64_to_user_ptr(args->data_ptr);
offset = offset_in_page(args->offset);
@@ -495,7 +475,7 @@ i915_gem_shmem_pread(struct drm_i915_gem_object *obj,
offset = 0;
}
- i915_gem_object_finish_access(obj);
+ i915_gem_object_unlock_fence(obj, fence);
return ret;
}
@@ -531,8 +511,9 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
struct i915_ggtt *ggtt = &i915->ggtt;
intel_wakeref_t wakeref;
struct drm_mm_node node;
- struct i915_vma *vma;
+ struct dma_fence *fence;
void __user *user_data;
+ struct i915_vma *vma;
u64 remain, offset;
int ret;
@@ -561,11 +542,24 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
GEM_BUG_ON(!node.allocated);
}
- ret = i915_gem_object_set_to_gtt_domain(obj, false);
+ mutex_unlock(&i915->drm.struct_mutex);
+
+ ret = i915_gem_object_lock_interruptible(obj);
if (ret)
goto out_unpin;
- mutex_unlock(&i915->drm.struct_mutex);
+ ret = i915_gem_object_set_to_gtt_domain(obj, false);
+ if (ret) {
+ i915_gem_object_unlock(obj);
+ goto out_unpin;
+ }
+
+ fence = i915_gem_object_lock_fence(obj);
+ i915_gem_object_unlock(obj);
+ if (!fence) {
+ ret = -ENOMEM;
+ goto out_unpin;
+ }
user_data = u64_to_user_ptr(args->data_ptr);
remain = args->size;
@@ -603,8 +597,9 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
offset += page_length;
}
- mutex_lock(&i915->drm.struct_mutex);
+ i915_gem_object_unlock_fence(obj, fence);
out_unpin:
+ mutex_lock(&i915->drm.struct_mutex);
if (node.allocated) {
wmb();
ggtt->vm.clear_range(&ggtt->vm, node.start, node.size);
@@ -715,6 +710,7 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
struct i915_ggtt *ggtt = &i915->ggtt;
intel_wakeref_t wakeref;
struct drm_mm_node node;
+ struct dma_fence *fence;
struct i915_vma *vma;
u64 remain, offset;
void __user *user_data;
@@ -762,11 +758,24 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
GEM_BUG_ON(!node.allocated);
}
- ret = i915_gem_object_set_to_gtt_domain(obj, true);
+ mutex_unlock(&i915->drm.struct_mutex);
+
+ ret = i915_gem_object_lock_interruptible(obj);
if (ret)
goto out_unpin;
- mutex_unlock(&i915->drm.struct_mutex);
+ ret = i915_gem_object_set_to_gtt_domain(obj, true);
+ if (ret) {
+ i915_gem_object_unlock(obj);
+ goto out_unpin;
+ }
+
+ fence = i915_gem_object_lock_fence(obj);
+ i915_gem_object_unlock(obj);
+ if (!fence) {
+ ret = -ENOMEM;
+ goto out_unpin;
+ }
intel_fb_obj_invalidate(obj, ORIGIN_CPU);
@@ -811,8 +820,9 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
}
intel_fb_obj_flush(obj, ORIGIN_CPU);
- mutex_lock(&i915->drm.struct_mutex);
+ i915_gem_object_unlock_fence(obj, fence);
out_unpin:
+ mutex_lock(&i915->drm.struct_mutex);
if (node.allocated) {
wmb();
ggtt->vm.clear_range(&ggtt->vm, node.start, node.size);
@@ -858,23 +868,23 @@ static int
i915_gem_shmem_pwrite(struct drm_i915_gem_object *obj,
const struct drm_i915_gem_pwrite *args)
{
- struct drm_i915_private *i915 = to_i915(obj->base.dev);
- void __user *user_data;
- u64 remain;
unsigned int partial_cacheline_write;
unsigned int needs_clflush;
unsigned int offset, idx;
+ struct dma_fence *fence;
+ void __user *user_data;
+ u64 remain;
int ret;
- ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
- if (ret)
- return ret;
-
ret = i915_gem_object_prepare_write(obj, &needs_clflush);
- mutex_unlock(&i915->drm.struct_mutex);
if (ret)
return ret;
+ fence = i915_gem_object_lock_fence(obj);
+ i915_gem_object_finish_access(obj);
+ if (!fence)
+ return -ENOMEM;
+
/* If we don't overwrite a cacheline completely we need to be
* careful to have up-to-date data by first clflushing. Don't
* overcomplicate things and flush the entire patch.
@@ -902,7 +912,8 @@ i915_gem_shmem_pwrite(struct drm_i915_gem_object *obj,
}
intel_fb_obj_flush(obj, ORIGIN_CPU);
- i915_gem_object_finish_access(obj);
+ i915_gem_object_unlock_fence(obj, fence);
+
return ret;
}
@@ -1781,7 +1792,9 @@ static int __intel_engines_record_defaults(struct drm_i915_private *i915)
if (err)
goto err_active;
+ i915_gem_object_lock(state->obj);
err = i915_gem_object_set_to_cpu_domain(state->obj, false);
+ i915_gem_object_unlock(state->obj);
if (err)
goto err_active;
@@ -2228,12 +2241,13 @@ int i915_gem_freeze_late(struct drm_i915_private *i915)
i915_gem_shrink(i915, -1UL, NULL, I915_SHRINK_UNBOUND);
i915_gem_drain_freed_objects(i915);
- mutex_lock(&i915->drm.struct_mutex);
for (phase = phases; *phase; phase++) {
- list_for_each_entry(obj, *phase, mm.link)
+ list_for_each_entry(obj, *phase, mm.link) {
+ i915_gem_object_lock(obj);
WARN_ON(i915_gem_object_set_to_cpu_domain(obj, true));
+ i915_gem_object_unlock(obj);
+ }
}
- mutex_unlock(&i915->drm.struct_mutex);
return 0;
}
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index d3eba67d4bf9..56be4b091bc7 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -3519,8 +3519,11 @@ void i915_gem_restore_gtt_mappings(struct drm_i915_private *dev_priv)
WARN_ON(i915_vma_bind(vma,
obj ? obj->cache_level : 0,
PIN_UPDATE));
- if (obj)
+ if (obj) {
+ i915_gem_object_lock(obj);
WARN_ON(i915_gem_object_set_to_gtt_domain(obj, false));
+ i915_gem_object_unlock(obj);
+ }
lock:
mutex_lock(&ggtt->vm.mutex);
diff --git a/drivers/gpu/drm/i915/i915_gem_render_state.c b/drivers/gpu/drm/i915/i915_gem_render_state.c
index f3b42b026fff..706ed71468e8 100644
--- a/drivers/gpu/drm/i915/i915_gem_render_state.c
+++ b/drivers/gpu/drm/i915/i915_gem_render_state.c
@@ -222,7 +222,9 @@ int i915_gem_render_state_emit(struct i915_request *rq)
goto err_unpin;
}
+ i915_vma_lock(so.vma);
err = i915_vma_move_to_active(so.vma, rq, 0);
+ i915_vma_unlock(so.vma);
err_unpin:
i915_vma_unpin(so.vma);
err_vma:
diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
index d4d308b6d1d8..3620d39c0bc8 100644
--- a/drivers/gpu/drm/i915/i915_vma.c
+++ b/drivers/gpu/drm/i915/i915_vma.c
@@ -847,13 +847,14 @@ void i915_vma_destroy(struct i915_vma *vma)
{
lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
- GEM_BUG_ON(i915_vma_is_active(vma));
GEM_BUG_ON(i915_vma_is_pinned(vma));
if (i915_vma_is_closed(vma))
list_del(&vma->closed_link);
WARN_ON(i915_vma_unbind(vma));
+ GEM_BUG_ON(i915_vma_is_active(vma));
+
__i915_vma_destroy(vma);
}
@@ -915,12 +916,10 @@ static void export_fence(struct i915_vma *vma,
* handle an error right now. Worst case should be missed
* synchronisation leading to rendering corruption.
*/
- reservation_object_lock(resv, NULL);
if (flags & EXEC_OBJECT_WRITE)
reservation_object_add_excl_fence(resv, &rq->fence);
else if (reservation_object_reserve_shared(resv, 1) == 0)
reservation_object_add_shared_fence(resv, &rq->fence);
- reservation_object_unlock(resv);
}
int i915_vma_move_to_active(struct i915_vma *vma,
@@ -929,7 +928,8 @@ int i915_vma_move_to_active(struct i915_vma *vma,
{
struct drm_i915_gem_object *obj = vma->obj;
- lockdep_assert_held(&rq->i915->drm.struct_mutex);
+ assert_vma_held(vma);
+ assert_object_held(obj);
GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
/*
diff --git a/drivers/gpu/drm/i915/i915_vma.h b/drivers/gpu/drm/i915/i915_vma.h
index 7950fa96ee86..71ac7ee8620a 100644
--- a/drivers/gpu/drm/i915/i915_vma.h
+++ b/drivers/gpu/drm/i915/i915_vma.h
@@ -295,6 +295,18 @@ void i915_vma_close(struct i915_vma *vma);
void i915_vma_reopen(struct i915_vma *vma);
void i915_vma_destroy(struct i915_vma *vma);
+#define assert_vma_held(vma) reservation_object_assert_held((vma)->resv)
+
+static inline void i915_vma_lock(struct i915_vma *vma)
+{
+ reservation_object_lock(vma->resv, NULL);
+}
+
+static inline void i915_vma_unlock(struct i915_vma *vma)
+{
+ reservation_object_unlock(vma->resv);
+}
+
int __i915_vma_do_pin(struct i915_vma *vma,
u64 size, u64 alignment, u64 flags);
static inline int __must_check
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index e63f290ccc26..ee2ffa5a03df 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -2102,6 +2102,7 @@ intel_pin_and_fence_fb_obj(struct drm_framebuffer *fb,
* pin/unpin/fence and not more.
*/
wakeref = intel_runtime_pm_get(dev_priv);
+ i915_gem_object_lock(obj);
atomic_inc(&dev_priv->gpu_error.pending_fb_pin);
@@ -2156,6 +2157,7 @@ intel_pin_and_fence_fb_obj(struct drm_framebuffer *fb,
err:
atomic_dec(&dev_priv->gpu_error.pending_fb_pin);
+ i915_gem_object_unlock(obj);
intel_runtime_pm_put(dev_priv, wakeref);
return vma;
}
@@ -2164,9 +2166,12 @@ void intel_unpin_fb_vma(struct i915_vma *vma, unsigned long flags)
{
lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
+ i915_gem_object_lock(vma->obj);
if (flags & PLANE_HAS_FENCE)
i915_vma_unpin_fence(vma);
i915_gem_object_unpin_from_display_plane(vma);
+ i915_gem_object_unlock(vma->obj);
+
i915_vma_put(vma);
}
diff --git a/drivers/gpu/drm/i915/intel_guc_log.c b/drivers/gpu/drm/i915/intel_guc_log.c
index 7146524264dd..67eadc82c396 100644
--- a/drivers/gpu/drm/i915/intel_guc_log.c
+++ b/drivers/gpu/drm/i915/intel_guc_log.c
@@ -343,8 +343,6 @@ static void capture_logs_work(struct work_struct *work)
static int guc_log_map(struct intel_guc_log *log)
{
- struct intel_guc *guc = log_to_guc(log);
- struct drm_i915_private *dev_priv = guc_to_i915(guc);
void *vaddr;
int ret;
@@ -353,9 +351,9 @@ static int guc_log_map(struct intel_guc_log *log)
if (!log->vma)
return -ENODEV;
- mutex_lock(&dev_priv->drm.struct_mutex);
+ i915_gem_object_lock(log->vma->obj);
ret = i915_gem_object_set_to_wc_domain(log->vma->obj, true);
- mutex_unlock(&dev_priv->drm.struct_mutex);
+ i915_gem_object_unlock(log->vma->obj);
if (ret)
return ret;
diff --git a/drivers/gpu/drm/i915/intel_overlay.c b/drivers/gpu/drm/i915/intel_overlay.c
index 80dcd879fc58..a2ac06a08715 100644
--- a/drivers/gpu/drm/i915/intel_overlay.c
+++ b/drivers/gpu/drm/i915/intel_overlay.c
@@ -765,8 +765,10 @@ static int intel_overlay_do_put_image(struct intel_overlay *overlay,
atomic_inc(&dev_priv->gpu_error.pending_fb_pin);
+ i915_gem_object_lock(new_bo);
vma = i915_gem_object_pin_to_display_plane(new_bo,
0, NULL, PIN_MAPPABLE);
+ i915_gem_object_unlock(new_bo);
if (IS_ERR(vma)) {
ret = PTR_ERR(vma);
goto out_pin_section;
@@ -1305,15 +1307,20 @@ int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data,
static int get_registers(struct intel_overlay *overlay, bool use_phys)
{
+ struct drm_i915_private *i915 = overlay->i915;
struct drm_i915_gem_object *obj;
struct i915_vma *vma;
int err;
- obj = i915_gem_object_create_stolen(overlay->i915, PAGE_SIZE);
+ mutex_lock(&i915->drm.struct_mutex);
+
+ obj = i915_gem_object_create_stolen(i915, PAGE_SIZE);
if (obj == NULL)
- obj = i915_gem_object_create_internal(overlay->i915, PAGE_SIZE);
- if (IS_ERR(obj))
- return PTR_ERR(obj);
+ obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
+ if (IS_ERR(obj)) {
+ err = PTR_ERR(obj);
+ goto err_unlock;
+ }
vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0, PIN_MAPPABLE);
if (IS_ERR(vma)) {
@@ -1334,10 +1341,13 @@ static int get_registers(struct intel_overlay *overlay, bool use_phys)
}
overlay->reg_bo = obj;
+ mutex_unlock(&i915->drm.struct_mutex);
return 0;
err_put_bo:
i915_gem_object_put(obj);
+err_unlock:
+ mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -1363,18 +1373,16 @@ void intel_overlay_setup(struct drm_i915_private *dev_priv)
INIT_ACTIVE_REQUEST(&overlay->last_flip);
- mutex_lock(&dev_priv->drm.struct_mutex);
-
ret = get_registers(overlay, OVERLAY_NEEDS_PHYSICAL(dev_priv));
if (ret)
goto out_free;
+ i915_gem_object_lock(overlay->reg_bo);
ret = i915_gem_object_set_to_gtt_domain(overlay->reg_bo, true);
+ i915_gem_object_unlock(overlay->reg_bo);
if (ret)
goto out_reg_bo;
- mutex_unlock(&dev_priv->drm.struct_mutex);
-
memset_io(overlay->regs, 0, sizeof(struct overlay_registers));
update_polyphase_filter(overlay->regs);
update_reg_attrs(overlay, overlay->regs);
@@ -1386,7 +1394,6 @@ void intel_overlay_setup(struct drm_i915_private *dev_priv)
out_reg_bo:
i915_gem_object_put(overlay->reg_bo);
out_free:
- mutex_unlock(&dev_priv->drm.struct_mutex);
kfree(overlay);
}
diff --git a/drivers/gpu/drm/i915/intel_uc_fw.c b/drivers/gpu/drm/i915/intel_uc_fw.c
index 3257a054cb0b..e0042650726a 100644
--- a/drivers/gpu/drm/i915/intel_uc_fw.c
+++ b/drivers/gpu/drm/i915/intel_uc_fw.c
@@ -246,15 +246,13 @@ int intel_uc_fw_upload(struct intel_uc_fw *uc_fw,
intel_uc_fw_type_repr(uc_fw->type),
intel_uc_fw_status_repr(uc_fw->load_status));
- intel_uc_fw_ggtt_bind(uc_fw);
-
/* Call custom loader */
+ intel_uc_fw_ggtt_bind(uc_fw);
err = xfer(uc_fw);
+ intel_uc_fw_ggtt_unbind(uc_fw);
if (err)
goto fail;
- intel_uc_fw_ggtt_unbind(uc_fw);
-
uc_fw->load_status = INTEL_UC_FIRMWARE_SUCCESS;
DRM_DEBUG_DRIVER("%s fw load %s\n",
intel_uc_fw_type_repr(uc_fw->type),
diff --git a/drivers/gpu/drm/i915/selftests/i915_request.c b/drivers/gpu/drm/i915/selftests/i915_request.c
index 92710353982c..1b9c183920aa 100644
--- a/drivers/gpu/drm/i915/selftests/i915_request.c
+++ b/drivers/gpu/drm/i915/selftests/i915_request.c
@@ -875,7 +875,9 @@ static int live_all_engines(void *arg)
i915_gem_object_set_active_reference(batch->obj);
}
+ i915_vma_lock(batch);
err = i915_vma_move_to_active(batch, request[id], 0);
+ i915_vma_unlock(batch);
GEM_BUG_ON(err);
i915_request_get(request[id]);
@@ -990,7 +992,9 @@ static int live_sequential_engines(void *arg)
GEM_BUG_ON(err);
request[id]->batch = batch;
+ i915_vma_lock(batch);
err = i915_vma_move_to_active(batch, request[id], 0);
+ i915_vma_unlock(batch);
GEM_BUG_ON(err);
i915_gem_object_set_active_reference(batch->obj);
diff --git a/drivers/gpu/drm/i915/selftests/igt_spinner.c b/drivers/gpu/drm/i915/selftests/igt_spinner.c
index 45688e6698b7..d845d1dc7ce7 100644
--- a/drivers/gpu/drm/i915/selftests/igt_spinner.c
+++ b/drivers/gpu/drm/i915/selftests/igt_spinner.c
@@ -76,7 +76,9 @@ static int move_to_active(struct i915_vma *vma,
{
int err;
+ i915_vma_lock(vma);
err = i915_vma_move_to_active(vma, rq, flags);
+ i915_vma_unlock(vma);
if (err)
return err;
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 29/40] drm/i915: Move GEM object waiting to its own file
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (26 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 28/40] drm/i915: Move GEM object domain management from struct_mutex to local Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-10 14:17 ` Mika Kuoppala
2019-05-08 8:06 ` [PATCH 30/40] drm/i915: Move GEM object busy checking " Chris Wilson
` (17 subsequent siblings)
45 siblings, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Continuing the decluttering of i915_gem.c by moving the object wait
decomposition into its own file.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/Makefile | 1 +
drivers/gpu/drm/i915/gem/i915_gem_object.h | 8 +
drivers/gpu/drm/i915/gem/i915_gem_wait.c | 277 +++++++++++++++++++++
drivers/gpu/drm/i915/i915_drv.h | 7 -
drivers/gpu/drm/i915/i915_gem.c | 254 -------------------
drivers/gpu/drm/i915/i915_utils.h | 10 -
6 files changed, 286 insertions(+), 271 deletions(-)
create mode 100644 drivers/gpu/drm/i915/gem/i915_gem_wait.c
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index e5348c355987..a4cc2f7f9bc6 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -105,6 +105,7 @@ gem-y += \
gem/i915_gem_stolen.o \
gem/i915_gem_tiling.o \
gem/i915_gem_userptr.o \
+ gem/i915_gem_wait.o \
gem/i915_gemfs.o
i915-y += \
$(gem-y) \
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index 509d145d808a..23bca003fbfb 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -436,4 +436,12 @@ static inline void __start_cpu_write(struct drm_i915_gem_object *obj)
obj->cache_dirty = true;
}
+int i915_gem_object_wait(struct drm_i915_gem_object *obj,
+ unsigned int flags,
+ long timeout);
+int i915_gem_object_wait_priority(struct drm_i915_gem_object *obj,
+ unsigned int flags,
+ const struct i915_sched_attr *attr);
+#define I915_PRIORITY_DISPLAY I915_USER_PRIORITY(I915_PRIORITY_MAX)
+
#endif
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_wait.c b/drivers/gpu/drm/i915/gem/i915_gem_wait.c
new file mode 100644
index 000000000000..fed5c751ef37
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/i915_gem_wait.c
@@ -0,0 +1,277 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2016 Intel Corporation
+ */
+
+#include <linux/dma-fence-array.h>
+#include <linux/jiffies.h>
+
+#include "gt/intel_engine.h"
+
+#include "i915_gem_ioctls.h"
+#include "i915_gem_object.h"
+
+static long
+i915_gem_object_wait_fence(struct dma_fence *fence,
+ unsigned int flags,
+ long timeout)
+{
+ BUILD_BUG_ON(I915_WAIT_INTERRUPTIBLE != 0x1);
+
+ if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
+ return timeout;
+
+ if (dma_fence_is_i915(fence))
+ return i915_request_wait(to_request(fence), flags, timeout);
+
+ return dma_fence_wait_timeout(fence,
+ flags & I915_WAIT_INTERRUPTIBLE,
+ timeout);
+}
+
+static long
+i915_gem_object_wait_reservation(struct reservation_object *resv,
+ unsigned int flags,
+ long timeout)
+{
+ unsigned int seq = __read_seqcount_begin(&resv->seq);
+ struct dma_fence *excl;
+ bool prune_fences = false;
+
+ if (flags & I915_WAIT_ALL) {
+ struct dma_fence **shared;
+ unsigned int count, i;
+ int ret;
+
+ ret = reservation_object_get_fences_rcu(resv,
+ &excl, &count, &shared);
+ if (ret)
+ return ret;
+
+ for (i = 0; i < count; i++) {
+ timeout = i915_gem_object_wait_fence(shared[i],
+ flags, timeout);
+ if (timeout < 0)
+ break;
+
+ dma_fence_put(shared[i]);
+ }
+
+ for (; i < count; i++)
+ dma_fence_put(shared[i]);
+ kfree(shared);
+
+ /*
+ * If both shared fences and an exclusive fence exist,
+ * then by construction the shared fences must be later
+ * than the exclusive fence. If we successfully wait for
+ * all the shared fences, we know that the exclusive fence
+ * must all be signaled. If all the shared fences are
+ * signaled, we can prune the array and recover the
+ * floating references on the fences/requests.
+ */
+ prune_fences = count && timeout >= 0;
+ } else {
+ excl = reservation_object_get_excl_rcu(resv);
+ }
+
+ if (excl && timeout >= 0)
+ timeout = i915_gem_object_wait_fence(excl, flags, timeout);
+
+ dma_fence_put(excl);
+
+ /*
+ * Opportunistically prune the fences iff we know they have *all* been
+ * signaled and that the reservation object has not been changed (i.e.
+ * no new fences have been added).
+ */
+ if (prune_fences && !__read_seqcount_retry(&resv->seq, seq)) {
+ if (reservation_object_trylock(resv)) {
+ if (!__read_seqcount_retry(&resv->seq, seq))
+ reservation_object_add_excl_fence(resv, NULL);
+ reservation_object_unlock(resv);
+ }
+ }
+
+ return timeout;
+}
+
+static void __fence_set_priority(struct dma_fence *fence,
+ const struct i915_sched_attr *attr)
+{
+ struct i915_request *rq;
+ struct intel_engine_cs *engine;
+
+ if (dma_fence_is_signaled(fence) || !dma_fence_is_i915(fence))
+ return;
+
+ rq = to_request(fence);
+ engine = rq->engine;
+
+ local_bh_disable();
+ rcu_read_lock(); /* RCU serialisation for set-wedged protection */
+ if (engine->schedule)
+ engine->schedule(rq, attr);
+ rcu_read_unlock();
+ local_bh_enable(); /* kick the tasklets if queues were reprioritised */
+}
+
+static void fence_set_priority(struct dma_fence *fence,
+ const struct i915_sched_attr *attr)
+{
+ /* Recurse once into a fence-array */
+ if (dma_fence_is_array(fence)) {
+ struct dma_fence_array *array = to_dma_fence_array(fence);
+ int i;
+
+ for (i = 0; i < array->num_fences; i++)
+ __fence_set_priority(array->fences[i], attr);
+ } else {
+ __fence_set_priority(fence, attr);
+ }
+}
+
+int
+i915_gem_object_wait_priority(struct drm_i915_gem_object *obj,
+ unsigned int flags,
+ const struct i915_sched_attr *attr)
+{
+ struct dma_fence *excl;
+
+ if (flags & I915_WAIT_ALL) {
+ struct dma_fence **shared;
+ unsigned int count, i;
+ int ret;
+
+ ret = reservation_object_get_fences_rcu(obj->resv,
+ &excl, &count, &shared);
+ if (ret)
+ return ret;
+
+ for (i = 0; i < count; i++) {
+ fence_set_priority(shared[i], attr);
+ dma_fence_put(shared[i]);
+ }
+
+ kfree(shared);
+ } else {
+ excl = reservation_object_get_excl_rcu(obj->resv);
+ }
+
+ if (excl) {
+ fence_set_priority(excl, attr);
+ dma_fence_put(excl);
+ }
+ return 0;
+}
+
+/**
+ * Waits for rendering to the object to be completed
+ * @obj: i915 gem object
+ * @flags: how to wait (under a lock, for all rendering or just for writes etc)
+ * @timeout: how long to wait
+ */
+int
+i915_gem_object_wait(struct drm_i915_gem_object *obj,
+ unsigned int flags,
+ long timeout)
+{
+ might_sleep();
+ GEM_BUG_ON(timeout < 0);
+
+ timeout = i915_gem_object_wait_reservation(obj->resv, flags, timeout);
+ return timeout < 0 ? timeout : 0;
+}
+
+static inline unsigned long nsecs_to_jiffies_timeout(const u64 n)
+{
+ /* nsecs_to_jiffies64() does not guard against overflow */
+ if (NSEC_PER_SEC % HZ &&
+ div_u64(n, NSEC_PER_SEC) >= MAX_JIFFY_OFFSET / HZ)
+ return MAX_JIFFY_OFFSET;
+
+ return min_t(u64, MAX_JIFFY_OFFSET, nsecs_to_jiffies64(n) + 1);
+}
+
+static unsigned long to_wait_timeout(s64 timeout_ns)
+{
+ if (timeout_ns < 0)
+ return MAX_SCHEDULE_TIMEOUT;
+
+ if (timeout_ns == 0)
+ return 0;
+
+ return nsecs_to_jiffies_timeout(timeout_ns);
+}
+
+/**
+ * i915_gem_wait_ioctl - implements DRM_IOCTL_I915_GEM_WAIT
+ * @dev: drm device pointer
+ * @data: ioctl data blob
+ * @file: drm file pointer
+ *
+ * Returns 0 if successful, else an error is returned with the remaining time in
+ * the timeout parameter.
+ * -ETIME: object is still busy after timeout
+ * -ERESTARTSYS: signal interrupted the wait
+ * -ENONENT: object doesn't exist
+ * Also possible, but rare:
+ * -EAGAIN: incomplete, restart syscall
+ * -ENOMEM: damn
+ * -ENODEV: Internal IRQ fail
+ * -E?: The add request failed
+ *
+ * The wait ioctl with a timeout of 0 reimplements the busy ioctl. With any
+ * non-zero timeout parameter the wait ioctl will wait for the given number of
+ * nanoseconds on an object becoming unbusy. Since the wait itself does so
+ * without holding struct_mutex the object may become re-busied before this
+ * function completes. A similar but shorter * race condition exists in the busy
+ * ioctl
+ */
+int
+i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
+{
+ struct drm_i915_gem_wait *args = data;
+ struct drm_i915_gem_object *obj;
+ ktime_t start;
+ long ret;
+
+ if (args->flags != 0)
+ return -EINVAL;
+
+ obj = i915_gem_object_lookup(file, args->bo_handle);
+ if (!obj)
+ return -ENOENT;
+
+ start = ktime_get();
+
+ ret = i915_gem_object_wait(obj,
+ I915_WAIT_INTERRUPTIBLE |
+ I915_WAIT_PRIORITY |
+ I915_WAIT_ALL,
+ to_wait_timeout(args->timeout_ns));
+
+ if (args->timeout_ns > 0) {
+ args->timeout_ns -= ktime_to_ns(ktime_sub(ktime_get(), start));
+ if (args->timeout_ns < 0)
+ args->timeout_ns = 0;
+
+ /*
+ * Apparently ktime isn't accurate enough and occasionally has a
+ * bit of mismatch in the jiffies<->nsecs<->ktime loop. So patch
+ * things up to make the test happy. We allow up to 1 jiffy.
+ *
+ * This is a regression from the timespec->ktime conversion.
+ */
+ if (ret == -ETIME && !nsecs_to_jiffies(args->timeout_ns))
+ args->timeout_ns = 0;
+
+ /* Asked to wait beyond the jiffie/scheduler precision? */
+ if (ret == -ETIME && args->timeout_ns)
+ ret = -EAGAIN;
+ }
+
+ i915_gem_object_put(obj);
+ return ret;
+}
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 6f8ddfbe7d85..8eb01b1b3e0e 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2742,13 +2742,6 @@ void i915_gem_suspend(struct drm_i915_private *dev_priv);
void i915_gem_suspend_late(struct drm_i915_private *dev_priv);
void i915_gem_resume(struct drm_i915_private *dev_priv);
vm_fault_t i915_gem_fault(struct vm_fault *vmf);
-int i915_gem_object_wait(struct drm_i915_gem_object *obj,
- unsigned int flags,
- long timeout);
-int i915_gem_object_wait_priority(struct drm_i915_gem_object *obj,
- unsigned int flags,
- const struct i915_sched_attr *attr);
-#define I915_PRIORITY_DISPLAY I915_USER_PRIORITY(I915_PRIORITY_MAX)
int i915_gem_open(struct drm_i915_private *i915, struct drm_file *file);
void i915_gem_release(struct drm_device *dev, struct drm_file *file);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 32fdc1977afe..467273dd2d4a 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -124,178 +124,6 @@ int i915_gem_object_unbind(struct drm_i915_gem_object *obj)
return ret;
}
-static long
-i915_gem_object_wait_fence(struct dma_fence *fence,
- unsigned int flags,
- long timeout)
-{
- BUILD_BUG_ON(I915_WAIT_INTERRUPTIBLE != 0x1);
-
- if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
- return timeout;
-
- if (dma_fence_is_i915(fence))
- return i915_request_wait(to_request(fence), flags, timeout);
-
- return dma_fence_wait_timeout(fence,
- flags & I915_WAIT_INTERRUPTIBLE,
- timeout);
-}
-
-static long
-i915_gem_object_wait_reservation(struct reservation_object *resv,
- unsigned int flags,
- long timeout)
-{
- unsigned int seq = __read_seqcount_begin(&resv->seq);
- struct dma_fence *excl;
- bool prune_fences = false;
-
- if (flags & I915_WAIT_ALL) {
- struct dma_fence **shared;
- unsigned int count, i;
- int ret;
-
- ret = reservation_object_get_fences_rcu(resv,
- &excl, &count, &shared);
- if (ret)
- return ret;
-
- for (i = 0; i < count; i++) {
- timeout = i915_gem_object_wait_fence(shared[i],
- flags, timeout);
- if (timeout < 0)
- break;
-
- dma_fence_put(shared[i]);
- }
-
- for (; i < count; i++)
- dma_fence_put(shared[i]);
- kfree(shared);
-
- /*
- * If both shared fences and an exclusive fence exist,
- * then by construction the shared fences must be later
- * than the exclusive fence. If we successfully wait for
- * all the shared fences, we know that the exclusive fence
- * must all be signaled. If all the shared fences are
- * signaled, we can prune the array and recover the
- * floating references on the fences/requests.
- */
- prune_fences = count && timeout >= 0;
- } else {
- excl = reservation_object_get_excl_rcu(resv);
- }
-
- if (excl && timeout >= 0)
- timeout = i915_gem_object_wait_fence(excl, flags, timeout);
-
- dma_fence_put(excl);
-
- /*
- * Opportunistically prune the fences iff we know they have *all* been
- * signaled and that the reservation object has not been changed (i.e.
- * no new fences have been added).
- */
- if (prune_fences && !__read_seqcount_retry(&resv->seq, seq)) {
- if (reservation_object_trylock(resv)) {
- if (!__read_seqcount_retry(&resv->seq, seq))
- reservation_object_add_excl_fence(resv, NULL);
- reservation_object_unlock(resv);
- }
- }
-
- return timeout;
-}
-
-static void __fence_set_priority(struct dma_fence *fence,
- const struct i915_sched_attr *attr)
-{
- struct i915_request *rq;
- struct intel_engine_cs *engine;
-
- if (dma_fence_is_signaled(fence) || !dma_fence_is_i915(fence))
- return;
-
- rq = to_request(fence);
- engine = rq->engine;
-
- local_bh_disable();
- rcu_read_lock(); /* RCU serialisation for set-wedged protection */
- if (engine->schedule)
- engine->schedule(rq, attr);
- rcu_read_unlock();
- local_bh_enable(); /* kick the tasklets if queues were reprioritised */
-}
-
-static void fence_set_priority(struct dma_fence *fence,
- const struct i915_sched_attr *attr)
-{
- /* Recurse once into a fence-array */
- if (dma_fence_is_array(fence)) {
- struct dma_fence_array *array = to_dma_fence_array(fence);
- int i;
-
- for (i = 0; i < array->num_fences; i++)
- __fence_set_priority(array->fences[i], attr);
- } else {
- __fence_set_priority(fence, attr);
- }
-}
-
-int
-i915_gem_object_wait_priority(struct drm_i915_gem_object *obj,
- unsigned int flags,
- const struct i915_sched_attr *attr)
-{
- struct dma_fence *excl;
-
- if (flags & I915_WAIT_ALL) {
- struct dma_fence **shared;
- unsigned int count, i;
- int ret;
-
- ret = reservation_object_get_fences_rcu(obj->resv,
- &excl, &count, &shared);
- if (ret)
- return ret;
-
- for (i = 0; i < count; i++) {
- fence_set_priority(shared[i], attr);
- dma_fence_put(shared[i]);
- }
-
- kfree(shared);
- } else {
- excl = reservation_object_get_excl_rcu(obj->resv);
- }
-
- if (excl) {
- fence_set_priority(excl, attr);
- dma_fence_put(excl);
- }
- return 0;
-}
-
-/**
- * Waits for rendering to the object to be completed
- * @obj: i915 gem object
- * @flags: how to wait (under a lock, for all rendering or just for writes etc)
- * @timeout: how long to wait
- */
-int
-i915_gem_object_wait(struct drm_i915_gem_object *obj,
- unsigned int flags,
- long timeout)
-{
- might_sleep();
- GEM_BUG_ON(timeout < 0);
-
- timeout = i915_gem_object_wait_reservation(obj->resv, flags, timeout);
- return timeout < 0 ? timeout : 0;
-}
-
static int
i915_gem_phys_pwrite(struct drm_i915_gem_object *obj,
struct drm_i915_gem_pwrite *args,
@@ -1073,88 +901,6 @@ void i915_gem_runtime_suspend(struct drm_i915_private *dev_priv)
}
}
-static unsigned long to_wait_timeout(s64 timeout_ns)
-{
- if (timeout_ns < 0)
- return MAX_SCHEDULE_TIMEOUT;
-
- if (timeout_ns == 0)
- return 0;
-
- return nsecs_to_jiffies_timeout(timeout_ns);
-}
-
-/**
- * i915_gem_wait_ioctl - implements DRM_IOCTL_I915_GEM_WAIT
- * @dev: drm device pointer
- * @data: ioctl data blob
- * @file: drm file pointer
- *
- * Returns 0 if successful, else an error is returned with the remaining time in
- * the timeout parameter.
- * -ETIME: object is still busy after timeout
- * -ERESTARTSYS: signal interrupted the wait
- * -ENONENT: object doesn't exist
- * Also possible, but rare:
- * -EAGAIN: incomplete, restart syscall
- * -ENOMEM: damn
- * -ENODEV: Internal IRQ fail
- * -E?: The add request failed
- *
- * The wait ioctl with a timeout of 0 reimplements the busy ioctl. With any
- * non-zero timeout parameter the wait ioctl will wait for the given number of
- * nanoseconds on an object becoming unbusy. Since the wait itself does so
- * without holding struct_mutex the object may become re-busied before this
- * function completes. A similar but shorter * race condition exists in the busy
- * ioctl
- */
-int
-i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
-{
- struct drm_i915_gem_wait *args = data;
- struct drm_i915_gem_object *obj;
- ktime_t start;
- long ret;
-
- if (args->flags != 0)
- return -EINVAL;
-
- obj = i915_gem_object_lookup(file, args->bo_handle);
- if (!obj)
- return -ENOENT;
-
- start = ktime_get();
-
- ret = i915_gem_object_wait(obj,
- I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_PRIORITY |
- I915_WAIT_ALL,
- to_wait_timeout(args->timeout_ns));
-
- if (args->timeout_ns > 0) {
- args->timeout_ns -= ktime_to_ns(ktime_sub(ktime_get(), start));
- if (args->timeout_ns < 0)
- args->timeout_ns = 0;
-
- /*
- * Apparently ktime isn't accurate enough and occasionally has a
- * bit of mismatch in the jiffies<->nsecs<->ktime loop. So patch
- * things up to make the test happy. We allow up to 1 jiffy.
- *
- * This is a regression from the timespec->ktime conversion.
- */
- if (ret == -ETIME && !nsecs_to_jiffies(args->timeout_ns))
- args->timeout_ns = 0;
-
- /* Asked to wait beyond the jiffie/scheduler precision? */
- if (ret == -ETIME && args->timeout_ns)
- ret = -EAGAIN;
- }
-
- i915_gem_object_put(obj);
- return ret;
-}
-
static int wait_for_engines(struct drm_i915_private *i915)
{
if (wait_for(intel_engines_are_idle(i915), I915_IDLE_ENGINES_TIMEOUT)) {
diff --git a/drivers/gpu/drm/i915/i915_utils.h b/drivers/gpu/drm/i915/i915_utils.h
index edfc69acdaac..9911f53382a5 100644
--- a/drivers/gpu/drm/i915/i915_utils.h
+++ b/drivers/gpu/drm/i915/i915_utils.h
@@ -218,16 +218,6 @@ static inline unsigned long msecs_to_jiffies_timeout(const unsigned int m)
return min_t(unsigned long, MAX_JIFFY_OFFSET, j + 1);
}
-static inline unsigned long nsecs_to_jiffies_timeout(const u64 n)
-{
- /* nsecs_to_jiffies64() does not guard against overflow */
- if (NSEC_PER_SEC % HZ &&
- div_u64(n, NSEC_PER_SEC) >= MAX_JIFFY_OFFSET / HZ)
- return MAX_JIFFY_OFFSET;
-
- return min_t(u64, MAX_JIFFY_OFFSET, nsecs_to_jiffies64(n) + 1);
-}
-
/*
* If you need to wait X milliseconds between events A and B, but event B
* doesn't happen exactly after event A, you record the timestamp (jiffies) of
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* Re: [PATCH 29/40] drm/i915: Move GEM object waiting to its own file
2019-05-08 8:06 ` [PATCH 29/40] drm/i915: Move GEM object waiting to its own file Chris Wilson
@ 2019-05-10 14:17 ` Mika Kuoppala
2019-05-10 14:33 ` Chris Wilson
0 siblings, 1 reply; 71+ messages in thread
From: Mika Kuoppala @ 2019-05-10 14:17 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
Chris Wilson <chris@chris-wilson.co.uk> writes:
> Continuing the decluttering of i915_gem.c by moving the object wait
> decomposition into its own file.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/Makefile | 1 +
> drivers/gpu/drm/i915/gem/i915_gem_object.h | 8 +
> drivers/gpu/drm/i915/gem/i915_gem_wait.c | 277 +++++++++++++++++++++
> drivers/gpu/drm/i915/i915_drv.h | 7 -
> drivers/gpu/drm/i915/i915_gem.c | 254 -------------------
> drivers/gpu/drm/i915/i915_utils.h | 10 -
> 6 files changed, 286 insertions(+), 271 deletions(-)
> create mode 100644 drivers/gpu/drm/i915/gem/i915_gem_wait.c
>
> diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
> index e5348c355987..a4cc2f7f9bc6 100644
> --- a/drivers/gpu/drm/i915/Makefile
> +++ b/drivers/gpu/drm/i915/Makefile
> @@ -105,6 +105,7 @@ gem-y += \
> gem/i915_gem_stolen.o \
> gem/i915_gem_tiling.o \
> gem/i915_gem_userptr.o \
> + gem/i915_gem_wait.o \
> gem/i915_gemfs.o
> i915-y += \
> $(gem-y) \
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
> index 509d145d808a..23bca003fbfb 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
> @@ -436,4 +436,12 @@ static inline void __start_cpu_write(struct drm_i915_gem_object *obj)
> obj->cache_dirty = true;
> }
>
> +int i915_gem_object_wait(struct drm_i915_gem_object *obj,
> + unsigned int flags,
> + long timeout);
> +int i915_gem_object_wait_priority(struct drm_i915_gem_object *obj,
> + unsigned int flags,
> + const struct i915_sched_attr *attr);
> +#define I915_PRIORITY_DISPLAY I915_USER_PRIORITY(I915_PRIORITY_MAX)
> +
> #endif
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_wait.c b/drivers/gpu/drm/i915/gem/i915_gem_wait.c
> new file mode 100644
> index 000000000000..fed5c751ef37
> --- /dev/null
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_wait.c
> @@ -0,0 +1,277 @@
> +/*
> + * SPDX-License-Identifier: MIT
> + *
> + * Copyright © 2016 Intel Corporation
> + */
> +
> +#include <linux/dma-fence-array.h>
> +#include <linux/jiffies.h>
> +
> +#include "gt/intel_engine.h"
> +
> +#include "i915_gem_ioctls.h"
> +#include "i915_gem_object.h"
> +
> +static long
> +i915_gem_object_wait_fence(struct dma_fence *fence,
> + unsigned int flags,
> + long timeout)
> +{
> + BUILD_BUG_ON(I915_WAIT_INTERRUPTIBLE != 0x1);
> +
> + if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
> + return timeout;
> +
> + if (dma_fence_is_i915(fence))
> + return i915_request_wait(to_request(fence), flags, timeout);
> +
> + return dma_fence_wait_timeout(fence,
> + flags & I915_WAIT_INTERRUPTIBLE,
> + timeout);
> +}
> +
> +static long
> +i915_gem_object_wait_reservation(struct reservation_object *resv,
> + unsigned int flags,
> + long timeout)
> +{
> + unsigned int seq = __read_seqcount_begin(&resv->seq);
> + struct dma_fence *excl;
> + bool prune_fences = false;
> +
> + if (flags & I915_WAIT_ALL) {
> + struct dma_fence **shared;
> + unsigned int count, i;
> + int ret;
> +
> + ret = reservation_object_get_fences_rcu(resv,
> + &excl, &count, &shared);
> + if (ret)
> + return ret;
> +
> + for (i = 0; i < count; i++) {
> + timeout = i915_gem_object_wait_fence(shared[i],
> + flags, timeout);
> + if (timeout < 0)
> + break;
> +
> + dma_fence_put(shared[i]);
> + }
> +
> + for (; i < count; i++)
> + dma_fence_put(shared[i]);
> + kfree(shared);
> +
> + /*
> + * If both shared fences and an exclusive fence exist,
> + * then by construction the shared fences must be later
> + * than the exclusive fence. If we successfully wait for
> + * all the shared fences, we know that the exclusive fence
> + * must all be signaled. If all the shared fences are
> + * signaled, we can prune the array and recover the
> + * floating references on the fences/requests.
> + */
> + prune_fences = count && timeout >= 0;
> + } else {
> + excl = reservation_object_get_excl_rcu(resv);
> + }
> +
> + if (excl && timeout >= 0)
> + timeout = i915_gem_object_wait_fence(excl, flags, timeout);
> +
> + dma_fence_put(excl);
> +
> + /*
> + * Opportunistically prune the fences iff we know they have *all* been
> + * signaled and that the reservation object has not been changed (i.e.
> + * no new fences have been added).
> + */
> + if (prune_fences && !__read_seqcount_retry(&resv->seq, seq)) {
> + if (reservation_object_trylock(resv)) {
> + if (!__read_seqcount_retry(&resv->seq, seq))
> + reservation_object_add_excl_fence(resv, NULL);
> + reservation_object_unlock(resv);
> + }
> + }
> +
> + return timeout;
> +}
> +
> +static void __fence_set_priority(struct dma_fence *fence,
> + const struct i915_sched_attr *attr)
> +{
> + struct i915_request *rq;
> + struct intel_engine_cs *engine;
> +
> + if (dma_fence_is_signaled(fence) || !dma_fence_is_i915(fence))
> + return;
> +
> + rq = to_request(fence);
> + engine = rq->engine;
> +
> + local_bh_disable();
> + rcu_read_lock(); /* RCU serialisation for set-wedged protection */
> + if (engine->schedule)
> + engine->schedule(rq, attr);
> + rcu_read_unlock();
> + local_bh_enable(); /* kick the tasklets if queues were reprioritised */
> +}
> +
> +static void fence_set_priority(struct dma_fence *fence,
> + const struct i915_sched_attr *attr)
> +{
> + /* Recurse once into a fence-array */
> + if (dma_fence_is_array(fence)) {
> + struct dma_fence_array *array = to_dma_fence_array(fence);
> + int i;
> +
> + for (i = 0; i < array->num_fences; i++)
> + __fence_set_priority(array->fences[i], attr);
> + } else {
> + __fence_set_priority(fence, attr);
> + }
> +}
> +
> +int
> +i915_gem_object_wait_priority(struct drm_i915_gem_object *obj,
> + unsigned int flags,
> + const struct i915_sched_attr *attr)
> +{
> + struct dma_fence *excl;
> +
> + if (flags & I915_WAIT_ALL) {
> + struct dma_fence **shared;
> + unsigned int count, i;
> + int ret;
> +
> + ret = reservation_object_get_fences_rcu(obj->resv,
> + &excl, &count, &shared);
> + if (ret)
> + return ret;
> +
> + for (i = 0; i < count; i++) {
> + fence_set_priority(shared[i], attr);
> + dma_fence_put(shared[i]);
> + }
> +
> + kfree(shared);
> + } else {
> + excl = reservation_object_get_excl_rcu(obj->resv);
> + }
> +
> + if (excl) {
> + fence_set_priority(excl, attr);
> + dma_fence_put(excl);
> + }
> + return 0;
> +}
> +
> +/**
> + * Waits for rendering to the object to be completed
> + * @obj: i915 gem object
> + * @flags: how to wait (under a lock, for all rendering or just for writes etc)
> + * @timeout: how long to wait
> + */
> +int
> +i915_gem_object_wait(struct drm_i915_gem_object *obj,
> + unsigned int flags,
> + long timeout)
> +{
> + might_sleep();
> + GEM_BUG_ON(timeout < 0);
> +
> + timeout = i915_gem_object_wait_reservation(obj->resv, flags, timeout);
> + return timeout < 0 ? timeout : 0;
> +}
> +
> +static inline unsigned long nsecs_to_jiffies_timeout(const u64 n)
> +{
> + /* nsecs_to_jiffies64() does not guard against overflow */
> + if (NSEC_PER_SEC % HZ &&
> + div_u64(n, NSEC_PER_SEC) >= MAX_JIFFY_OFFSET / HZ)
> + return MAX_JIFFY_OFFSET;
> +
> + return min_t(u64, MAX_JIFFY_OFFSET, nsecs_to_jiffies64(n) + 1);
> +}
> +
> +static unsigned long to_wait_timeout(s64 timeout_ns)
> +{
> + if (timeout_ns < 0)
> + return MAX_SCHEDULE_TIMEOUT;
> +
> + if (timeout_ns == 0)
> + return 0;
> +
> + return nsecs_to_jiffies_timeout(timeout_ns);
> +}
> +
> +/**
> + * i915_gem_wait_ioctl - implements DRM_IOCTL_I915_GEM_WAIT
> + * @dev: drm device pointer
> + * @data: ioctl data blob
> + * @file: drm file pointer
> + *
> + * Returns 0 if successful, else an error is returned with the remaining time in
> + * the timeout parameter.
> + * -ETIME: object is still busy after timeout
> + * -ERESTARTSYS: signal interrupted the wait
> + * -ENONENT: object doesn't exist
> + * Also possible, but rare:
> + * -EAGAIN: incomplete, restart syscall
> + * -ENOMEM: damn
> + * -ENODEV: Internal IRQ fail
> + * -E?: The add request failed
> + *
> + * The wait ioctl with a timeout of 0 reimplements the busy ioctl. With any
> + * non-zero timeout parameter the wait ioctl will wait for the given number of
> + * nanoseconds on an object becoming unbusy. Since the wait itself does so
> + * without holding struct_mutex the object may become re-busied before this
> + * function completes. A similar but shorter * race condition exists in the busy
> + * ioctl
> + */
> +int
> +i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
> +{
> + struct drm_i915_gem_wait *args = data;
> + struct drm_i915_gem_object *obj;
> + ktime_t start;
> + long ret;
> +
> + if (args->flags != 0)
> + return -EINVAL;
> +
> + obj = i915_gem_object_lookup(file, args->bo_handle);
> + if (!obj)
> + return -ENOENT;
> +
> + start = ktime_get();
> +
> + ret = i915_gem_object_wait(obj,
> + I915_WAIT_INTERRUPTIBLE |
> + I915_WAIT_PRIORITY |
> + I915_WAIT_ALL,
> + to_wait_timeout(args->timeout_ns));
> +
> + if (args->timeout_ns > 0) {
> + args->timeout_ns -= ktime_to_ns(ktime_sub(ktime_get(), start));
> + if (args->timeout_ns < 0)
> + args->timeout_ns = 0;
> +
> + /*
> + * Apparently ktime isn't accurate enough and occasionally has a
> + * bit of mismatch in the jiffies<->nsecs<->ktime loop. So patch
> + * things up to make the test happy. We allow up to 1 jiffy.
> + *
> + * This is a regression from the timespec->ktime conversion.
> + */
> + if (ret == -ETIME && !nsecs_to_jiffies(args->timeout_ns))
> + args->timeout_ns = 0;
> +
> + /* Asked to wait beyond the jiffie/scheduler precision? */
> + if (ret == -ETIME && args->timeout_ns)
> + ret = -EAGAIN;
> + }
> +
> + i915_gem_object_put(obj);
> + return ret;
> +}
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index 6f8ddfbe7d85..8eb01b1b3e0e 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -2742,13 +2742,6 @@ void i915_gem_suspend(struct drm_i915_private *dev_priv);
> void i915_gem_suspend_late(struct drm_i915_private *dev_priv);
> void i915_gem_resume(struct drm_i915_private *dev_priv);
> vm_fault_t i915_gem_fault(struct vm_fault *vmf);
> -int i915_gem_object_wait(struct drm_i915_gem_object *obj,
> - unsigned int flags,
> - long timeout);
> -int i915_gem_object_wait_priority(struct drm_i915_gem_object *obj,
> - unsigned int flags,
> - const struct i915_sched_attr *attr);
> -#define I915_PRIORITY_DISPLAY I915_USER_PRIORITY(I915_PRIORITY_MAX)
>
> int i915_gem_open(struct drm_i915_private *i915, struct drm_file *file);
> void i915_gem_release(struct drm_device *dev, struct drm_file *file);
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index 32fdc1977afe..467273dd2d4a 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -124,178 +124,6 @@ int i915_gem_object_unbind(struct drm_i915_gem_object *obj)
> return ret;
> }
>
> -static long
> -i915_gem_object_wait_fence(struct dma_fence *fence,
> - unsigned int flags,
> - long timeout)
> -{
> - BUILD_BUG_ON(I915_WAIT_INTERRUPTIBLE != 0x1);
> -
> - if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
> - return timeout;
> -
> - if (dma_fence_is_i915(fence))
> - return i915_request_wait(to_request(fence), flags, timeout);
> -
> - return dma_fence_wait_timeout(fence,
> - flags & I915_WAIT_INTERRUPTIBLE,
> - timeout);
> -}
> -
> -static long
> -i915_gem_object_wait_reservation(struct reservation_object *resv,
> - unsigned int flags,
> - long timeout)
> -{
> - unsigned int seq = __read_seqcount_begin(&resv->seq);
> - struct dma_fence *excl;
> - bool prune_fences = false;
> -
> - if (flags & I915_WAIT_ALL) {
> - struct dma_fence **shared;
> - unsigned int count, i;
> - int ret;
> -
> - ret = reservation_object_get_fences_rcu(resv,
> - &excl, &count, &shared);
> - if (ret)
> - return ret;
> -
> - for (i = 0; i < count; i++) {
> - timeout = i915_gem_object_wait_fence(shared[i],
> - flags, timeout);
> - if (timeout < 0)
> - break;
> -
> - dma_fence_put(shared[i]);
> - }
> -
> - for (; i < count; i++)
> - dma_fence_put(shared[i]);
> - kfree(shared);
> -
> - /*
> - * If both shared fences and an exclusive fence exist,
> - * then by construction the shared fences must be later
> - * than the exclusive fence. If we successfully wait for
> - * all the shared fences, we know that the exclusive fence
> - * must all be signaled. If all the shared fences are
> - * signaled, we can prune the array and recover the
> - * floating references on the fences/requests.
> - */
> - prune_fences = count && timeout >= 0;
> - } else {
> - excl = reservation_object_get_excl_rcu(resv);
> - }
> -
> - if (excl && timeout >= 0)
> - timeout = i915_gem_object_wait_fence(excl, flags, timeout);
> -
> - dma_fence_put(excl);
> -
> - /*
> - * Opportunistically prune the fences iff we know they have *all* been
> - * signaled and that the reservation object has not been changed (i.e.
> - * no new fences have been added).
> - */
> - if (prune_fences && !__read_seqcount_retry(&resv->seq, seq)) {
> - if (reservation_object_trylock(resv)) {
> - if (!__read_seqcount_retry(&resv->seq, seq))
> - reservation_object_add_excl_fence(resv, NULL);
> - reservation_object_unlock(resv);
> - }
> - }
> -
> - return timeout;
> -}
> -
> -static void __fence_set_priority(struct dma_fence *fence,
> - const struct i915_sched_attr *attr)
> -{
> - struct i915_request *rq;
> - struct intel_engine_cs *engine;
> -
> - if (dma_fence_is_signaled(fence) || !dma_fence_is_i915(fence))
> - return;
> -
> - rq = to_request(fence);
> - engine = rq->engine;
> -
> - local_bh_disable();
> - rcu_read_lock(); /* RCU serialisation for set-wedged protection */
> - if (engine->schedule)
> - engine->schedule(rq, attr);
> - rcu_read_unlock();
> - local_bh_enable(); /* kick the tasklets if queues were reprioritised */
> -}
> -
> -static void fence_set_priority(struct dma_fence *fence,
> - const struct i915_sched_attr *attr)
> -{
> - /* Recurse once into a fence-array */
> - if (dma_fence_is_array(fence)) {
> - struct dma_fence_array *array = to_dma_fence_array(fence);
> - int i;
> -
> - for (i = 0; i < array->num_fences; i++)
> - __fence_set_priority(array->fences[i], attr);
> - } else {
> - __fence_set_priority(fence, attr);
> - }
> -}
> -
> -int
> -i915_gem_object_wait_priority(struct drm_i915_gem_object *obj,
> - unsigned int flags,
> - const struct i915_sched_attr *attr)
> -{
> - struct dma_fence *excl;
> -
> - if (flags & I915_WAIT_ALL) {
> - struct dma_fence **shared;
> - unsigned int count, i;
> - int ret;
> -
> - ret = reservation_object_get_fences_rcu(obj->resv,
> - &excl, &count, &shared);
> - if (ret)
> - return ret;
> -
> - for (i = 0; i < count; i++) {
> - fence_set_priority(shared[i], attr);
> - dma_fence_put(shared[i]);
> - }
> -
> - kfree(shared);
> - } else {
> - excl = reservation_object_get_excl_rcu(obj->resv);
> - }
> -
> - if (excl) {
> - fence_set_priority(excl, attr);
> - dma_fence_put(excl);
> - }
> - return 0;
> -}
> -
> -/**
> - * Waits for rendering to the object to be completed
> - * @obj: i915 gem object
> - * @flags: how to wait (under a lock, for all rendering or just for writes etc)
> - * @timeout: how long to wait
> - */
> -int
> -i915_gem_object_wait(struct drm_i915_gem_object *obj,
> - unsigned int flags,
> - long timeout)
> -{
> - might_sleep();
> - GEM_BUG_ON(timeout < 0);
> -
> - timeout = i915_gem_object_wait_reservation(obj->resv, flags, timeout);
> - return timeout < 0 ? timeout : 0;
> -}
> -
> static int
> i915_gem_phys_pwrite(struct drm_i915_gem_object *obj,
> struct drm_i915_gem_pwrite *args,
> @@ -1073,88 +901,6 @@ void i915_gem_runtime_suspend(struct drm_i915_private *dev_priv)
> }
> }
>
> -static unsigned long to_wait_timeout(s64 timeout_ns)
> -{
> - if (timeout_ns < 0)
> - return MAX_SCHEDULE_TIMEOUT;
> -
> - if (timeout_ns == 0)
> - return 0;
> -
> - return nsecs_to_jiffies_timeout(timeout_ns);
> -}
> -
> -/**
> - * i915_gem_wait_ioctl - implements DRM_IOCTL_I915_GEM_WAIT
> - * @dev: drm device pointer
> - * @data: ioctl data blob
> - * @file: drm file pointer
> - *
> - * Returns 0 if successful, else an error is returned with the remaining time in
> - * the timeout parameter.
> - * -ETIME: object is still busy after timeout
> - * -ERESTARTSYS: signal interrupted the wait
> - * -ENONENT: object doesn't exist
> - * Also possible, but rare:
> - * -EAGAIN: incomplete, restart syscall
> - * -ENOMEM: damn
> - * -ENODEV: Internal IRQ fail
> - * -E?: The add request failed
> - *
> - * The wait ioctl with a timeout of 0 reimplements the busy ioctl. With any
> - * non-zero timeout parameter the wait ioctl will wait for the given number of
> - * nanoseconds on an object becoming unbusy. Since the wait itself does so
> - * without holding struct_mutex the object may become re-busied before this
> - * function completes. A similar but shorter * race condition exists in the busy
> - * ioctl
> - */
> -int
> -i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
> -{
> - struct drm_i915_gem_wait *args = data;
> - struct drm_i915_gem_object *obj;
> - ktime_t start;
> - long ret;
> -
> - if (args->flags != 0)
> - return -EINVAL;
> -
> - obj = i915_gem_object_lookup(file, args->bo_handle);
> - if (!obj)
> - return -ENOENT;
> -
> - start = ktime_get();
> -
> - ret = i915_gem_object_wait(obj,
> - I915_WAIT_INTERRUPTIBLE |
> - I915_WAIT_PRIORITY |
> - I915_WAIT_ALL,
> - to_wait_timeout(args->timeout_ns));
> -
> - if (args->timeout_ns > 0) {
> - args->timeout_ns -= ktime_to_ns(ktime_sub(ktime_get(), start));
> - if (args->timeout_ns < 0)
> - args->timeout_ns = 0;
> -
> - /*
> - * Apparently ktime isn't accurate enough and occasionally has a
> - * bit of mismatch in the jiffies<->nsecs<->ktime loop. So patch
> - * things up to make the test happy. We allow up to 1 jiffy.
> - *
> - * This is a regression from the timespec->ktime conversion.
> - */
> - if (ret == -ETIME && !nsecs_to_jiffies(args->timeout_ns))
> - args->timeout_ns = 0;
> -
> - /* Asked to wait beyond the jiffie/scheduler precision? */
> - if (ret == -ETIME && args->timeout_ns)
> - ret = -EAGAIN;
> - }
> -
> - i915_gem_object_put(obj);
> - return ret;
> -}
> -
> static int wait_for_engines(struct drm_i915_private *i915)
> {
> if (wait_for(intel_engines_are_idle(i915), I915_IDLE_ENGINES_TIMEOUT)) {
> diff --git a/drivers/gpu/drm/i915/i915_utils.h b/drivers/gpu/drm/i915/i915_utils.h
> index edfc69acdaac..9911f53382a5 100644
> --- a/drivers/gpu/drm/i915/i915_utils.h
> +++ b/drivers/gpu/drm/i915/i915_utils.h
> @@ -218,16 +218,6 @@ static inline unsigned long msecs_to_jiffies_timeout(const unsigned int m)
> return min_t(unsigned long, MAX_JIFFY_OFFSET, j + 1);
> }
>
> -static inline unsigned long nsecs_to_jiffies_timeout(const u64 n)
> -{
> - /* nsecs_to_jiffies64() does not guard against overflow */
> - if (NSEC_PER_SEC % HZ &&
> - div_u64(n, NSEC_PER_SEC) >= MAX_JIFFY_OFFSET / HZ)
> - return MAX_JIFFY_OFFSET;
> -
> - return min_t(u64, MAX_JIFFY_OFFSET, nsecs_to_jiffies64(n) + 1);
> -}
Seems that the wait stuff was only user so jiffiying the timeout. Just looks
generic for other usage too.
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread* Re: [PATCH 29/40] drm/i915: Move GEM object waiting to its own file
2019-05-10 14:17 ` Mika Kuoppala
@ 2019-05-10 14:33 ` Chris Wilson
0 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-10 14:33 UTC (permalink / raw)
To: Mika Kuoppala, intel-gfx
Quoting Mika Kuoppala (2019-05-10 15:17:12)
> > diff --git a/drivers/gpu/drm/i915/i915_utils.h b/drivers/gpu/drm/i915/i915_utils.h
> > index edfc69acdaac..9911f53382a5 100644
> > --- a/drivers/gpu/drm/i915/i915_utils.h
> > +++ b/drivers/gpu/drm/i915/i915_utils.h
> > @@ -218,16 +218,6 @@ static inline unsigned long msecs_to_jiffies_timeout(const unsigned int m)
> > return min_t(unsigned long, MAX_JIFFY_OFFSET, j + 1);
> > }
> >
> > -static inline unsigned long nsecs_to_jiffies_timeout(const u64 n)
> > -{
> > - /* nsecs_to_jiffies64() does not guard against overflow */
> > - if (NSEC_PER_SEC % HZ &&
> > - div_u64(n, NSEC_PER_SEC) >= MAX_JIFFY_OFFSET / HZ)
> > - return MAX_JIFFY_OFFSET;
> > -
> > - return min_t(u64, MAX_JIFFY_OFFSET, nsecs_to_jiffies64(n) + 1);
> > -}
>
> Seems that the wait stuff was only user so jiffiying the timeout. Just looks
> generic for other usage too.
But not generic enough to be in jiffies.h :)
I didn't want to clutter a common header with a one-off. But it can be
moved around on a whim, so not too concerned if this is not its final
resting place.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread
* [PATCH 30/40] drm/i915: Move GEM object busy checking to its own file
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (27 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 29/40] drm/i915: Move GEM object waiting to its own file Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-10 14:29 ` Mika Kuoppala
2019-05-08 8:06 ` [PATCH 31/40] drm/i915: Move GEM client throttling " Chris Wilson
` (16 subsequent siblings)
45 siblings, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Continuing the decluttering of i915_gem.c by moving the object busy
checking into its own file.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/Makefile | 1 +
drivers/gpu/drm/i915/gem/i915_gem_busy.c | 138 +++++++++++++++++++++++
drivers/gpu/drm/i915/i915_gem.c | 128 ---------------------
3 files changed, 139 insertions(+), 128 deletions(-)
create mode 100644 drivers/gpu/drm/i915/gem/i915_gem_busy.c
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index a4cc2f7f9bc6..865d7b51c297 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -88,6 +88,7 @@ i915-y += $(gt-y)
# GEM (Graphics Execution Management) code
obj-y += gem/
gem-y += \
+ gem/i915_gem_busy.o \
gem/i915_gem_clflush.o \
gem/i915_gem_context.o \
gem/i915_gem_dmabuf.o \
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_busy.c b/drivers/gpu/drm/i915/gem/i915_gem_busy.c
new file mode 100644
index 000000000000..5a5eda3003e9
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/i915_gem_busy.c
@@ -0,0 +1,138 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2014-2016 Intel Corporation
+ */
+
+#include "gt/intel_engine.h"
+
+#include "i915_gem_ioctls.h"
+#include "i915_gem_object.h"
+
+static __always_inline u32 __busy_read_flag(u8 id)
+{
+ if (id == (u8)I915_ENGINE_CLASS_INVALID)
+ return 0xffff0000u;
+
+ GEM_BUG_ON(id >= 16);
+ return 0x10000u << id;
+}
+
+static __always_inline u32 __busy_write_id(u8 id)
+{
+ /*
+ * The uABI guarantees an active writer is also amongst the read
+ * engines. This would be true if we accessed the activity tracking
+ * under the lock, but as we perform the lookup of the object and
+ * its activity locklessly we can not guarantee that the last_write
+ * being active implies that we have set the same engine flag from
+ * last_read - hence we always set both read and write busy for
+ * last_write.
+ */
+ if (id == (u8)I915_ENGINE_CLASS_INVALID)
+ return 0xffffffffu;
+
+ return (id + 1) | __busy_read_flag(id);
+}
+
+static __always_inline unsigned int
+__busy_set_if_active(const struct dma_fence *fence, u32 (*flag)(u8 id))
+{
+ const struct i915_request *rq;
+
+ /*
+ * We have to check the current hw status of the fence as the uABI
+ * guarantees forward progress. We could rely on the idle worker
+ * to eventually flush us, but to minimise latency just ask the
+ * hardware.
+ *
+ * Note we only report on the status of native fences.
+ */
+ if (!dma_fence_is_i915(fence))
+ return 0;
+
+ /* opencode to_request() in order to avoid const warnings */
+ rq = container_of(fence, const struct i915_request, fence);
+ if (i915_request_completed(rq))
+ return 0;
+
+ /* Beware type-expansion follies! */
+ BUILD_BUG_ON(!typecheck(u8, rq->engine->uabi_class));
+ return flag(rq->engine->uabi_class);
+}
+
+static __always_inline unsigned int
+busy_check_reader(const struct dma_fence *fence)
+{
+ return __busy_set_if_active(fence, __busy_read_flag);
+}
+
+static __always_inline unsigned int
+busy_check_writer(const struct dma_fence *fence)
+{
+ if (!fence)
+ return 0;
+
+ return __busy_set_if_active(fence, __busy_write_id);
+}
+
+int
+i915_gem_busy_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file)
+{
+ struct drm_i915_gem_busy *args = data;
+ struct drm_i915_gem_object *obj;
+ struct reservation_object_list *list;
+ unsigned int seq;
+ int err;
+
+ err = -ENOENT;
+ rcu_read_lock();
+ obj = i915_gem_object_lookup_rcu(file, args->handle);
+ if (!obj)
+ goto out;
+
+ /*
+ * A discrepancy here is that we do not report the status of
+ * non-i915 fences, i.e. even though we may report the object as idle,
+ * a call to set-domain may still stall waiting for foreign rendering.
+ * This also means that wait-ioctl may report an object as busy,
+ * where busy-ioctl considers it idle.
+ *
+ * We trade the ability to warn of foreign fences to report on which
+ * i915 engines are active for the object.
+ *
+ * Alternatively, we can trade that extra information on read/write
+ * activity with
+ * args->busy =
+ * !reservation_object_test_signaled_rcu(obj->resv, true);
+ * to report the overall busyness. This is what the wait-ioctl does.
+ *
+ */
+retry:
+ seq = raw_read_seqcount(&obj->resv->seq);
+
+ /* Translate the exclusive fence to the READ *and* WRITE engine */
+ args->busy = busy_check_writer(rcu_dereference(obj->resv->fence_excl));
+
+ /* Translate shared fences to READ set of engines */
+ list = rcu_dereference(obj->resv->fence);
+ if (list) {
+ unsigned int shared_count = list->shared_count, i;
+
+ for (i = 0; i < shared_count; ++i) {
+ struct dma_fence *fence =
+ rcu_dereference(list->shared[i]);
+
+ args->busy |= busy_check_reader(fence);
+ }
+ }
+
+ if (args->busy && read_seqcount_retry(&obj->resv->seq, seq))
+ goto retry;
+
+ err = 0;
+out:
+ rcu_read_unlock();
+ return err;
+}
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 467273dd2d4a..2f1e6dd78dc1 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1118,134 +1118,6 @@ i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj,
return vma;
}
-static __always_inline u32 __busy_read_flag(u8 id)
-{
- if (id == (u8)I915_ENGINE_CLASS_INVALID)
- return 0xffff0000u;
-
- GEM_BUG_ON(id >= 16);
- return 0x10000u << id;
-}
-
-static __always_inline u32 __busy_write_id(u8 id)
-{
- /*
- * The uABI guarantees an active writer is also amongst the read
- * engines. This would be true if we accessed the activity tracking
- * under the lock, but as we perform the lookup of the object and
- * its activity locklessly we can not guarantee that the last_write
- * being active implies that we have set the same engine flag from
- * last_read - hence we always set both read and write busy for
- * last_write.
- */
- if (id == (u8)I915_ENGINE_CLASS_INVALID)
- return 0xffffffffu;
-
- return (id + 1) | __busy_read_flag(id);
-}
-
-static __always_inline unsigned int
-__busy_set_if_active(const struct dma_fence *fence, u32 (*flag)(u8 id))
-{
- const struct i915_request *rq;
-
- /*
- * We have to check the current hw status of the fence as the uABI
- * guarantees forward progress. We could rely on the idle worker
- * to eventually flush us, but to minimise latency just ask the
- * hardware.
- *
- * Note we only report on the status of native fences.
- */
- if (!dma_fence_is_i915(fence))
- return 0;
-
- /* opencode to_request() in order to avoid const warnings */
- rq = container_of(fence, const struct i915_request, fence);
- if (i915_request_completed(rq))
- return 0;
-
- /* Beware type-expansion follies! */
- BUILD_BUG_ON(!typecheck(u8, rq->engine->uabi_class));
- return flag(rq->engine->uabi_class);
-}
-
-static __always_inline unsigned int
-busy_check_reader(const struct dma_fence *fence)
-{
- return __busy_set_if_active(fence, __busy_read_flag);
-}
-
-static __always_inline unsigned int
-busy_check_writer(const struct dma_fence *fence)
-{
- if (!fence)
- return 0;
-
- return __busy_set_if_active(fence, __busy_write_id);
-}
-
-int
-i915_gem_busy_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file)
-{
- struct drm_i915_gem_busy *args = data;
- struct drm_i915_gem_object *obj;
- struct reservation_object_list *list;
- unsigned int seq;
- int err;
-
- err = -ENOENT;
- rcu_read_lock();
- obj = i915_gem_object_lookup_rcu(file, args->handle);
- if (!obj)
- goto out;
-
- /*
- * A discrepancy here is that we do not report the status of
- * non-i915 fences, i.e. even though we may report the object as idle,
- * a call to set-domain may still stall waiting for foreign rendering.
- * This also means that wait-ioctl may report an object as busy,
- * where busy-ioctl considers it idle.
- *
- * We trade the ability to warn of foreign fences to report on which
- * i915 engines are active for the object.
- *
- * Alternatively, we can trade that extra information on read/write
- * activity with
- * args->busy =
- * !reservation_object_test_signaled_rcu(obj->resv, true);
- * to report the overall busyness. This is what the wait-ioctl does.
- *
- */
-retry:
- seq = raw_read_seqcount(&obj->resv->seq);
-
- /* Translate the exclusive fence to the READ *and* WRITE engine */
- args->busy = busy_check_writer(rcu_dereference(obj->resv->fence_excl));
-
- /* Translate shared fences to READ set of engines */
- list = rcu_dereference(obj->resv->fence);
- if (list) {
- unsigned int shared_count = list->shared_count, i;
-
- for (i = 0; i < shared_count; ++i) {
- struct dma_fence *fence =
- rcu_dereference(list->shared[i]);
-
- args->busy |= busy_check_reader(fence);
- }
- }
-
- if (args->busy && read_seqcount_retry(&obj->resv->seq, seq))
- goto retry;
-
- err = 0;
-out:
- rcu_read_unlock();
- return err;
-}
-
int
i915_gem_throttle_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv)
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 31/40] drm/i915: Move GEM client throttling to its own file
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (28 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 30/40] drm/i915: Move GEM object busy checking " Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-10 14:37 ` Mika Kuoppala
2019-05-08 8:06 ` [PATCH 32/40] drm/i915: Drop the deferred active reference Chris Wilson
` (15 subsequent siblings)
45 siblings, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Continuing the decluttering of i915_gem.c by moving the client self
throttling into its own file.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/Makefile | 1 +
drivers/gpu/drm/i915/gem/i915_gem_throttle.c | 74 ++++++++++++++++++++
drivers/gpu/drm/i915/i915_drv.h | 6 --
drivers/gpu/drm/i915/i915_gem.c | 58 ---------------
4 files changed, 75 insertions(+), 64 deletions(-)
create mode 100644 drivers/gpu/drm/i915/gem/i915_gem_throttle.c
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index 865d7b51c297..78d36eaff070 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -104,6 +104,7 @@ gem-y += \
gem/i915_gem_shmem.o \
gem/i915_gem_shrinker.o \
gem/i915_gem_stolen.o \
+ gem/i915_gem_throttle.o \
gem/i915_gem_tiling.o \
gem/i915_gem_userptr.o \
gem/i915_gem_wait.o \
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_throttle.c b/drivers/gpu/drm/i915/gem/i915_gem_throttle.c
new file mode 100644
index 000000000000..491bc28c175d
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/i915_gem_throttle.c
@@ -0,0 +1,74 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2014-2016 Intel Corporation
+ */
+
+#include <linux/jiffies.h>
+
+#include <drm/drm_file.h>
+
+#include "i915_gem_ioctls.h"
+#include "i915_gem_object.h"
+
+#include "../i915_drv.h"
+
+/*
+ * 20ms is a fairly arbitrary limit (greater than the average frame time)
+ * chosen to prevent the CPU getting more than a frame ahead of the GPU
+ * (when using lax throttling for the frontbuffer). We also use it to
+ * offer free GPU waitboosts for severely congested workloads.
+ */
+#define DRM_I915_THROTTLE_JIFFIES msecs_to_jiffies(20)
+
+/*
+ * Throttle our rendering by waiting until the ring has completed our requests
+ * emitted over 20 msec ago.
+ *
+ * Note that if we were to use the current jiffies each time around the loop,
+ * we wouldn't escape the function with any frames outstanding if the time to
+ * render a frame was over 20ms.
+ *
+ * This should get us reasonable parallelism between CPU and GPU but also
+ * relatively low latency when blocking on a particular request to finish.
+ */
+int
+i915_gem_throttle_ioctl(struct drm_device *dev, void *data,
+ struct drm_file *file)
+{
+ struct drm_i915_file_private *file_priv = file->driver_priv;
+ unsigned long recent_enough = jiffies - DRM_I915_THROTTLE_JIFFIES;
+ struct i915_request *request, *target = NULL;
+ long ret;
+
+ /* ABI: return -EIO if already wedged */
+ ret = i915_terminally_wedged(to_i915(dev));
+ if (ret)
+ return ret;
+
+ spin_lock(&file_priv->mm.lock);
+ list_for_each_entry(request, &file_priv->mm.request_list, client_link) {
+ if (time_after_eq(request->emitted_jiffies, recent_enough))
+ break;
+
+ if (target) {
+ list_del(&target->client_link);
+ target->file_priv = NULL;
+ }
+
+ target = request;
+ }
+ if (target)
+ i915_request_get(target);
+ spin_unlock(&file_priv->mm.lock);
+
+ if (!target)
+ return 0;
+
+ ret = i915_request_wait(target,
+ I915_WAIT_INTERRUPTIBLE,
+ MAX_SCHEDULE_TIMEOUT);
+ i915_request_put(target);
+
+ return ret < 0 ? ret : 0;
+}
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 8eb01b1b3e0e..b372ac47aa1c 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -212,12 +212,6 @@ struct drm_i915_file_private {
struct {
spinlock_t lock;
struct list_head request_list;
-/* 20ms is a fairly arbitrary limit (greater than the average frame time)
- * chosen to prevent the CPU getting more than a frame ahead of the GPU
- * (when using lax throttling for the frontbuffer). We also use it to
- * offer free GPU waitboosts for severely congested workloads.
- */
-#define DRM_I915_THROTTLE_JIFFIES msecs_to_jiffies(20)
} mm;
struct idr context_idr;
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 2f1e6dd78dc1..2a9e8ecf2926 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -988,57 +988,6 @@ int i915_gem_wait_for_idle(struct drm_i915_private *i915,
return 0;
}
-/* Throttle our rendering by waiting until the ring has completed our requests
- * emitted over 20 msec ago.
- *
- * Note that if we were to use the current jiffies each time around the loop,
- * we wouldn't escape the function with any frames outstanding if the time to
- * render a frame was over 20ms.
- *
- * This should get us reasonable parallelism between CPU and GPU but also
- * relatively low latency when blocking on a particular request to finish.
- */
-static int
-i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)
-{
- struct drm_i915_private *dev_priv = to_i915(dev);
- struct drm_i915_file_private *file_priv = file->driver_priv;
- unsigned long recent_enough = jiffies - DRM_I915_THROTTLE_JIFFIES;
- struct i915_request *request, *target = NULL;
- long ret;
-
- /* ABI: return -EIO if already wedged */
- ret = i915_terminally_wedged(dev_priv);
- if (ret)
- return ret;
-
- spin_lock(&file_priv->mm.lock);
- list_for_each_entry(request, &file_priv->mm.request_list, client_link) {
- if (time_after_eq(request->emitted_jiffies, recent_enough))
- break;
-
- if (target) {
- list_del(&target->client_link);
- target->file_priv = NULL;
- }
-
- target = request;
- }
- if (target)
- i915_request_get(target);
- spin_unlock(&file_priv->mm.lock);
-
- if (target == NULL)
- return 0;
-
- ret = i915_request_wait(target,
- I915_WAIT_INTERRUPTIBLE,
- MAX_SCHEDULE_TIMEOUT);
- i915_request_put(target);
-
- return ret < 0 ? ret : 0;
-}
-
struct i915_vma *
i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj,
const struct i915_ggtt_view *view,
@@ -1118,13 +1067,6 @@ i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj,
return vma;
}
-int
-i915_gem_throttle_ioctl(struct drm_device *dev, void *data,
- struct drm_file *file_priv)
-{
- return i915_gem_ring_throttle(dev, file_priv);
-}
-
int
i915_gem_madvise_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv)
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 32/40] drm/i915: Drop the deferred active reference
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (29 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 31/40] drm/i915: Move GEM client throttling " Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 33/40] drm/i915: Move object close under its own lock Chris Wilson
` (14 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx; +Cc: Matthew Auld
An old optimisation to reduce the number of atomics per batch sadly
relies on struct_mutex for coordination. In order to remove struct_mutex
from serialising object/context closing, always taking and releasing an
active reference on first use / last use greatly simplifies the locking.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
---
drivers/gpu/drm/i915/gem/i915_gem_context.c | 2 +-
drivers/gpu/drm/i915/gem/i915_gem_object.c | 13 +---------
drivers/gpu/drm/i915/gem/i915_gem_object.h | 24 +------------------
.../gpu/drm/i915/gem/i915_gem_object_types.h | 8 -------
.../gpu/drm/i915/gem/selftests/huge_pages.c | 3 +--
.../i915/gem/selftests/i915_gem_coherency.c | 4 ++--
.../drm/i915/gem/selftests/i915_gem_context.c | 11 +++++----
.../drm/i915/gem/selftests/i915_gem_mman.c | 2 +-
drivers/gpu/drm/i915/gt/intel_engine_cs.c | 2 +-
drivers/gpu/drm/i915/gt/intel_ringbuffer.c | 3 +--
drivers/gpu/drm/i915/gt/selftest_hangcheck.c | 9 +------
.../gpu/drm/i915/gt/selftest_workarounds.c | 3 ---
drivers/gpu/drm/i915/gvt/scheduler.c | 2 +-
drivers/gpu/drm/i915/i915_gem_batch_pool.c | 2 +-
drivers/gpu/drm/i915/i915_gem_render_state.c | 2 +-
drivers/gpu/drm/i915/i915_vma.c | 15 +++++-------
drivers/gpu/drm/i915/selftests/i915_request.c | 8 -------
drivers/gpu/drm/i915/selftests/igt_spinner.c | 9 +------
18 files changed, 26 insertions(+), 96 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index 217d87c13cd8..5016a3e1f863 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -112,7 +112,7 @@ static void lut_close(struct i915_gem_context *ctx)
radix_tree_iter_delete(&ctx->handles_vma, &iter, slot);
vma->open_count--;
- __i915_gem_object_release_unless_active(vma->obj);
+ i915_vma_put(vma);
}
rcu_read_unlock();
}
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
index df79e2eead62..c178cf7614c6 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
@@ -156,7 +156,7 @@ void i915_gem_close_object(struct drm_gem_object *gem, struct drm_file *file)
list_del(&lut->ctx_link);
i915_lut_handle_free(lut);
- __i915_gem_object_release_unless_active(obj);
+ i915_gem_object_put(obj);
}
mutex_unlock(&i915->drm.struct_mutex);
@@ -348,17 +348,6 @@ void i915_gem_free_object(struct drm_gem_object *gem_obj)
call_rcu(&obj->rcu, __i915_gem_free_object_rcu);
}
-void __i915_gem_object_release_unless_active(struct drm_i915_gem_object *obj)
-{
- lockdep_assert_held(&obj->base.dev->struct_mutex);
-
- if (!i915_gem_object_has_active_reference(obj) &&
- i915_gem_object_is_active(obj))
- i915_gem_object_set_active_reference(obj);
- else
- i915_gem_object_put(obj);
-}
-
static inline enum fb_op_origin
fb_write_origin(struct drm_i915_gem_object *obj, unsigned int domain)
{
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index 23bca003fbfb..dcb765cd97a3 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -161,31 +161,9 @@ i915_gem_object_needs_async_cancel(const struct drm_i915_gem_object *obj)
static inline bool
i915_gem_object_is_active(const struct drm_i915_gem_object *obj)
{
- return obj->active_count;
+ return READ_ONCE(obj->active_count);
}
-static inline bool
-i915_gem_object_has_active_reference(const struct drm_i915_gem_object *obj)
-{
- return test_bit(I915_BO_ACTIVE_REF, &obj->flags);
-}
-
-static inline void
-i915_gem_object_set_active_reference(struct drm_i915_gem_object *obj)
-{
- lockdep_assert_held(&obj->base.dev->struct_mutex);
- __set_bit(I915_BO_ACTIVE_REF, &obj->flags);
-}
-
-static inline void
-i915_gem_object_clear_active_reference(struct drm_i915_gem_object *obj)
-{
- lockdep_assert_held(&obj->base.dev->struct_mutex);
- __clear_bit(I915_BO_ACTIVE_REF, &obj->flags);
-}
-
-void __i915_gem_object_release_unless_active(struct drm_i915_gem_object *obj);
-
static inline bool
i915_gem_object_is_framebuffer(const struct drm_i915_gem_object *obj)
{
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
index 24abdd6999b5..133581339d81 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
@@ -120,14 +120,6 @@ struct drm_i915_gem_object {
struct list_head batch_pool_link;
I915_SELFTEST_DECLARE(struct list_head st_link);
- unsigned long flags;
-
- /**
- * Have we taken a reference for the object for incomplete GPU
- * activity?
- */
-#define I915_BO_ACTIVE_REF 0
-
/*
* Is the object to be mapped as read-only to the GPU
* Only honoured if hardware has relevant pte bit
diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
index 5a89cbab3b2e..9e211aa044ea 100644
--- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
+++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
@@ -975,8 +975,6 @@ static int gpu_write(struct i915_vma *vma,
if (err)
goto err_request;
- i915_gem_object_set_active_reference(batch->obj);
-
i915_vma_lock(vma);
err = i915_gem_object_set_to_gtt_domain(vma->obj, false);
if (err == 0)
@@ -995,6 +993,7 @@ static int gpu_write(struct i915_vma *vma,
err_batch:
i915_vma_unpin(batch);
i915_vma_close(batch);
+ i915_vma_put(batch);
return err;
}
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
index 22fe4b6fd506..ae9586645150 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
@@ -365,7 +365,7 @@ static int igt_gem_coherency(void *arg)
}
}
- __i915_gem_object_release_unless_active(obj);
+ i915_gem_object_put(obj);
}
}
}
@@ -377,7 +377,7 @@ static int igt_gem_coherency(void *arg)
return err;
put_object:
- __i915_gem_object_release_unless_active(obj);
+ i915_gem_object_put(obj);
goto unlock;
}
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
index 72eedd6c2a0a..1bc3b8026400 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
@@ -318,14 +318,14 @@ static int gpu_fill(struct drm_i915_gem_object *obj,
if (err)
goto skip_request;
- i915_gem_object_set_active_reference(batch->obj);
+ i915_request_add(rq);
+
i915_vma_unpin(batch);
i915_vma_close(batch);
+ i915_vma_put(batch);
i915_vma_unpin(vma);
- i915_request_add(rq);
-
return 0;
skip_request:
@@ -802,9 +802,9 @@ emit_rpcs_query(struct drm_i915_gem_object *obj,
if (err)
goto skip_request;
- i915_gem_object_set_active_reference(batch->obj);
i915_vma_unpin(batch);
i915_vma_close(batch);
+ i915_vma_put(batch);
i915_vma_unpin(vma);
@@ -820,6 +820,7 @@ emit_rpcs_query(struct drm_i915_gem_object *obj,
i915_request_add(rq);
err_batch:
i915_vma_unpin(batch);
+ i915_vma_put(batch);
err_vma:
i915_vma_unpin(vma);
@@ -1365,9 +1366,9 @@ static int write_to_scratch(struct i915_gem_context *ctx,
if (err)
goto skip_request;
- i915_gem_object_set_active_reference(obj);
i915_vma_unpin(vma);
i915_vma_close(vma);
+ i915_vma_put(vma);
i915_request_add(rq);
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
index 297f8864d392..5db3327958fb 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
@@ -354,8 +354,8 @@ static int make_obj_busy(struct drm_i915_gem_object *obj)
i915_request_add(rq);
- __i915_gem_object_release_unless_active(obj);
i915_vma_unpin(vma);
+ i915_gem_object_put(obj); /* leave it only alive via its active ref */
return err;
}
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
index c397ec1b3df9..4d911f60d1f8 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
@@ -477,7 +477,7 @@ static void cleanup_status_page(struct intel_engine_cs *engine)
i915_vma_unpin(vma);
i915_gem_object_unpin_map(vma->obj);
- __i915_gem_object_release_unless_active(vma->obj);
+ i915_gem_object_put(vma->obj);
}
static int pin_ggtt_status_page(struct intel_engine_cs *engine,
diff --git a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
index 66d5a52d505c..ff58d658e3e2 100644
--- a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
@@ -1302,10 +1302,9 @@ intel_engine_create_ring(struct intel_engine_cs *engine,
void intel_ring_free(struct kref *ref)
{
struct intel_ring *ring = container_of(ref, typeof(*ring), ref);
- struct drm_i915_gem_object *obj = ring->vma->obj;
i915_vma_close(ring->vma);
- __i915_gem_object_release_unless_active(obj);
+ i915_vma_put(ring->vma);
i915_timeline_put(ring->timeline);
kfree(ring);
diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
index 45745e36b5f4..910f96231934 100644
--- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
+++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
@@ -118,15 +118,8 @@ static int move_to_active(struct i915_vma *vma,
i915_vma_lock(vma);
err = i915_vma_move_to_active(vma, rq, flags);
i915_vma_unlock(vma);
- if (err)
- return err;
-
- if (!i915_gem_object_has_active_reference(vma->obj)) {
- i915_gem_object_get(vma->obj);
- i915_gem_object_set_active_reference(vma->obj);
- }
- return 0;
+ return err;
}
static struct i915_request *
diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
index d4ba5296b2c8..db40d4f805ea 100644
--- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
@@ -135,9 +135,6 @@ read_nonprivs(struct i915_gem_context *ctx, struct intel_engine_cs *engine)
}
intel_ring_advance(rq, cs);
- i915_gem_object_get(result);
- i915_gem_object_set_active_reference(result);
-
i915_request_add(rq);
i915_vma_unpin(vma);
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index 840cac06add7..d4dc0e189547 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -583,7 +583,7 @@ static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload)
i915_vma_unpin(bb->vma);
i915_vma_close(bb->vma);
}
- __i915_gem_object_release_unless_active(bb->obj);
+ i915_gem_object_put(bb->obj);
}
list_del(&bb->list);
kfree(bb);
diff --git a/drivers/gpu/drm/i915/i915_gem_batch_pool.c b/drivers/gpu/drm/i915/i915_gem_batch_pool.c
index f3890b664e3f..56adfdcaed3e 100644
--- a/drivers/gpu/drm/i915/i915_gem_batch_pool.c
+++ b/drivers/gpu/drm/i915/i915_gem_batch_pool.c
@@ -55,7 +55,7 @@ void i915_gem_batch_pool_fini(struct i915_gem_batch_pool *pool)
list_for_each_entry_safe(obj, next,
&pool->cache_list[n],
batch_pool_link)
- __i915_gem_object_release_unless_active(obj);
+ i915_gem_object_put(obj);
INIT_LIST_HEAD(&pool->cache_list[n]);
}
diff --git a/drivers/gpu/drm/i915/i915_gem_render_state.c b/drivers/gpu/drm/i915/i915_gem_render_state.c
index 706ed71468e8..4ee032072d4f 100644
--- a/drivers/gpu/drm/i915/i915_gem_render_state.c
+++ b/drivers/gpu/drm/i915/i915_gem_render_state.c
@@ -230,6 +230,6 @@ int i915_gem_render_state_emit(struct i915_request *rq)
err_vma:
i915_vma_close(so.vma);
err_obj:
- __i915_gem_object_release_unless_active(so.obj);
+ i915_gem_object_put(so.obj);
return err;
}
diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
index 3620d39c0bc8..16d47f1f645a 100644
--- a/drivers/gpu/drm/i915/i915_vma.c
+++ b/drivers/gpu/drm/i915/i915_vma.c
@@ -115,10 +115,7 @@ static void __i915_vma_retire(struct i915_active *ref)
*/
obj_bump_mru(obj);
- if (i915_gem_object_has_active_reference(obj)) {
- i915_gem_object_clear_active_reference(obj);
- i915_gem_object_put(obj);
- }
+ i915_gem_object_put(obj); /* and drop the active reference */
}
static struct i915_vma *
@@ -443,7 +440,7 @@ void i915_vma_unpin_and_release(struct i915_vma **p_vma, unsigned int flags)
if (flags & I915_VMA_RELEASE_MAP)
i915_gem_object_unpin_map(obj);
- __i915_gem_object_release_unless_active(obj);
+ i915_gem_object_put(obj);
}
bool i915_vma_misplaced(const struct i915_vma *vma,
@@ -940,12 +937,12 @@ int i915_vma_move_to_active(struct i915_vma *vma,
* add the active reference first and queue for it to be dropped
* *last*.
*/
- if (!vma->active.count)
- obj->active_count++;
+ if (!vma->active.count && !obj->active_count++)
+ i915_gem_object_get(obj); /* once more for the active ref */
if (unlikely(i915_active_ref(&vma->active, rq->fence.context, rq))) {
- if (!vma->active.count)
- obj->active_count--;
+ if (!vma->active.count && !--obj->active_count)
+ i915_gem_object_put(obj);
return -ENOMEM;
}
diff --git a/drivers/gpu/drm/i915/selftests/i915_request.c b/drivers/gpu/drm/i915/selftests/i915_request.c
index 1b9c183920aa..693ea097c05f 100644
--- a/drivers/gpu/drm/i915/selftests/i915_request.c
+++ b/drivers/gpu/drm/i915/selftests/i915_request.c
@@ -870,11 +870,6 @@ static int live_all_engines(void *arg)
GEM_BUG_ON(err);
request[id]->batch = batch;
- if (!i915_gem_object_has_active_reference(batch->obj)) {
- i915_gem_object_get(batch->obj);
- i915_gem_object_set_active_reference(batch->obj);
- }
-
i915_vma_lock(batch);
err = i915_vma_move_to_active(batch, request[id], 0);
i915_vma_unlock(batch);
@@ -997,9 +992,6 @@ static int live_sequential_engines(void *arg)
i915_vma_unlock(batch);
GEM_BUG_ON(err);
- i915_gem_object_set_active_reference(batch->obj);
- i915_vma_get(batch);
-
i915_request_get(request[id]);
i915_request_add(request[id]);
diff --git a/drivers/gpu/drm/i915/selftests/igt_spinner.c b/drivers/gpu/drm/i915/selftests/igt_spinner.c
index d845d1dc7ce7..02306aa83d2a 100644
--- a/drivers/gpu/drm/i915/selftests/igt_spinner.c
+++ b/drivers/gpu/drm/i915/selftests/igt_spinner.c
@@ -79,15 +79,8 @@ static int move_to_active(struct i915_vma *vma,
i915_vma_lock(vma);
err = i915_vma_move_to_active(vma, rq, flags);
i915_vma_unlock(vma);
- if (err)
- return err;
-
- if (!i915_gem_object_has_active_reference(vma->obj)) {
- i915_gem_object_get(vma->obj);
- i915_gem_object_set_active_reference(vma->obj);
- }
- return 0;
+ return err;
}
struct i915_request *
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 33/40] drm/i915: Move object close under its own lock
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (30 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 32/40] drm/i915: Drop the deferred active reference Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:06 ` [PATCH 34/40] drm/i915: Rename intel_context.active to .inflight Chris Wilson
` (13 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Use i915_gem_object_lock() to guard the LUT and active reference to
allow us to break free of struct_mutex for handling GEM_CLOSE.
Testcase: igt/gem_close_race
Testcase: igt/gem_exec_parallel
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/gem/i915_gem_context.c | 76 ++++++++++---------
.../gpu/drm/i915/gem/i915_gem_context_types.h | 12 +--
.../gpu/drm/i915/gem/i915_gem_execbuffer.c | 23 ++++--
drivers/gpu/drm/i915/gem/i915_gem_object.c | 38 ++++++----
.../gpu/drm/i915/gem/i915_gem_object_types.h | 1 -
.../gpu/drm/i915/gem/selftests/mock_context.c | 1 -
drivers/gpu/drm/i915/i915_drv.h | 4 +-
drivers/gpu/drm/i915/i915_gem.c | 1 +
drivers/gpu/drm/i915/i915_gem_gtt.c | 1 +
drivers/gpu/drm/i915/i915_timeline.c | 13 ++--
drivers/gpu/drm/i915/i915_vma.c | 42 ++++++----
drivers/gpu/drm/i915/i915_vma.h | 17 ++---
.../gpu/drm/i915/selftests/mock_gem_device.c | 1 +
13 files changed, 132 insertions(+), 98 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index 5016a3e1f863..a9608d9ced6a 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -95,24 +95,42 @@ void i915_lut_handle_free(struct i915_lut_handle *lut)
static void lut_close(struct i915_gem_context *ctx)
{
- struct i915_lut_handle *lut, *ln;
struct radix_tree_iter iter;
void __rcu **slot;
- list_for_each_entry_safe(lut, ln, &ctx->handles_list, ctx_link) {
- list_del(&lut->obj_link);
- i915_lut_handle_free(lut);
- }
- INIT_LIST_HEAD(&ctx->handles_list);
+ lockdep_assert_held(&ctx->mutex);
rcu_read_lock();
radix_tree_for_each_slot(slot, &ctx->handles_vma, &iter, 0) {
struct i915_vma *vma = rcu_dereference_raw(*slot);
-
- radix_tree_iter_delete(&ctx->handles_vma, &iter, slot);
-
- vma->open_count--;
- i915_vma_put(vma);
+ struct drm_i915_gem_object *obj = vma->obj;
+ struct i915_lut_handle *lut;
+ bool found = false;
+
+ rcu_read_unlock();
+ i915_gem_object_lock(obj);
+ list_for_each_entry(lut, &obj->lut_list, obj_link) {
+ if (lut->ctx != ctx)
+ continue;
+
+ if (lut->handle != iter.index)
+ continue;
+
+ list_del(&lut->obj_link);
+ i915_lut_handle_free(lut);
+ found = true;
+ break;
+ }
+ i915_gem_object_unlock(obj);
+ rcu_read_lock();
+
+ if (found) {
+ radix_tree_iter_delete(&ctx->handles_vma, &iter, slot);
+ if (atomic_dec_and_test(&vma->open_count) &&
+ !i915_vma_is_ggtt(vma))
+ i915_vma_close(vma);
+ i915_gem_object_put(obj);
+ }
}
rcu_read_unlock();
}
@@ -250,15 +268,9 @@ static void free_engines(struct i915_gem_engines *e)
__free_engines(e, e->num_engines);
}
-static void free_engines_rcu(struct work_struct *wrk)
+static void free_engines_rcu(struct rcu_head *rcu)
{
- struct i915_gem_engines *e =
- container_of(wrk, struct i915_gem_engines, rcu.work);
- struct drm_i915_private *i915 = e->i915;
-
- mutex_lock(&i915->drm.struct_mutex);
- free_engines(e);
- mutex_unlock(&i915->drm.struct_mutex);
+ free_engines(container_of(rcu, struct i915_gem_engines, rcu));
}
static struct i915_gem_engines *default_engines(struct i915_gem_context *ctx)
@@ -271,7 +283,7 @@ static struct i915_gem_engines *default_engines(struct i915_gem_context *ctx)
if (!e)
return ERR_PTR(-ENOMEM);
- e->i915 = ctx->i915;
+ init_rcu_head(&e->rcu);
for_each_engine(engine, ctx->i915, id) {
struct intel_context *ce;
@@ -359,7 +371,10 @@ void i915_gem_context_release(struct kref *ref)
static void context_close(struct i915_gem_context *ctx)
{
+ mutex_lock(&ctx->mutex);
+
i915_gem_context_set_closed(ctx);
+ ctx->file_priv = ERR_PTR(-EBADF);
/*
* This context will never again be assinged to HW, so we can
@@ -374,7 +389,7 @@ static void context_close(struct i915_gem_context *ctx)
*/
lut_close(ctx);
- ctx->file_priv = ERR_PTR(-EBADF);
+ mutex_unlock(&ctx->mutex);
i915_gem_context_put(ctx);
}
@@ -429,7 +444,6 @@ __create_context(struct drm_i915_private *dev_priv)
RCU_INIT_POINTER(ctx->engines, e);
INIT_RADIX_TREE(&ctx->handles_vma, GFP_KERNEL);
- INIT_LIST_HEAD(&ctx->handles_list);
INIT_LIST_HEAD(&ctx->hw_id_link);
/* NB: Mark all slices as needing a remap so that when the context first
@@ -772,9 +786,7 @@ int i915_gem_context_open(struct drm_i915_private *i915,
return 0;
err_ctx:
- mutex_lock(&i915->drm.struct_mutex);
context_close(ctx);
- mutex_unlock(&i915->drm.struct_mutex);
err:
idr_destroy(&file_priv->vm_idr);
idr_destroy(&file_priv->context_idr);
@@ -787,8 +799,6 @@ void i915_gem_context_close(struct drm_file *file)
{
struct drm_i915_file_private *file_priv = file->driver_priv;
- lockdep_assert_held(&file_priv->dev_priv->drm.struct_mutex);
-
idr_for_each(&file_priv->context_idr, context_idr_cleanup, NULL);
idr_destroy(&file_priv->context_idr);
mutex_destroy(&file_priv->context_idr_lock);
@@ -1093,7 +1103,9 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv,
goto unlock;
/* Teardown the existing obj:vma cache, it will have to be rebuilt. */
+ mutex_lock(&ctx->mutex);
lut_close(ctx);
+ mutex_unlock(&ctx->mutex);
old = __set_ppgtt(ctx, ppgtt);
@@ -1612,7 +1624,7 @@ set_engines(struct i915_gem_context *ctx,
if (!set.engines)
return -ENOMEM;
- set.engines->i915 = ctx->i915;
+ init_rcu_head(&set.engines->rcu);
for (n = 0; n < num_engines; n++) {
struct i915_engine_class_instance ci;
struct intel_engine_cs *engine;
@@ -1666,8 +1678,7 @@ set_engines(struct i915_gem_context *ctx,
rcu_swap_protected(ctx->engines, set.engines, 1);
mutex_unlock(&ctx->engines_mutex);
- INIT_RCU_WORK(&set.engines->rcu, free_engines_rcu);
- queue_rcu_work(system_wq, &set.engines->rcu);
+ call_rcu(&set.engines->rcu, free_engines_rcu);
return 0;
}
@@ -1865,7 +1876,7 @@ static int clone_engines(struct i915_gem_context *dst,
if (!clone)
goto err_unlock;
- clone->i915 = dst->i915;
+ init_rcu_head(&clone->rcu);
for (n = 0; n < e->num_engines; n++) {
struct intel_engine_cs *engine;
@@ -2137,9 +2148,7 @@ int i915_gem_context_create_ioctl(struct drm_device *dev, void *data,
return 0;
err_ctx:
- mutex_lock(&dev->struct_mutex);
context_close(ext_data.ctx);
- mutex_unlock(&dev->struct_mutex);
return ret;
}
@@ -2164,10 +2173,7 @@ int i915_gem_context_destroy_ioctl(struct drm_device *dev, void *data,
if (!ctx)
return -ENOENT;
- mutex_lock(&dev->struct_mutex);
context_close(ctx);
- mutex_unlock(&dev->struct_mutex);
-
return 0;
}
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
index fb965ded2508..3db7448b9732 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
@@ -30,8 +30,7 @@ struct i915_timeline;
struct intel_ring;
struct i915_gem_engines {
- struct rcu_work rcu;
- struct drm_i915_private *i915;
+ struct rcu_head rcu;
unsigned int num_engines;
struct intel_context *engines[];
};
@@ -192,17 +191,12 @@ struct i915_gem_context {
/** remap_slice: Bitmask of cache lines that need remapping */
u8 remap_slice;
- /** handles_vma: rbtree to look up our context specific obj/vma for
+ /**
+ * handles_vma: rbtree to look up our context specific obj/vma for
* the user handle. (user handles are per fd, but the binding is
* per vm, which may be one per context or shared with the global GTT)
*/
struct radix_tree_root handles_vma;
-
- /** handles_list: reverse list of all the rbtree entries in use for
- * this context, which allows us to free all the allocations on
- * context close.
- */
- struct list_head handles_list;
};
#endif /* __I915_GEM_CONTEXT_TYPES_H__ */
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 080f69358224..5133787480ea 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -802,9 +802,6 @@ static int eb_lookup_vmas(struct i915_execbuffer *eb)
unsigned int i, batch;
int err;
- if (unlikely(i915_gem_context_is_closed(eb->gem_context)))
- return -ENOENT;
-
if (unlikely(i915_gem_context_is_banned(eb->gem_context)))
return -EIO;
@@ -813,6 +810,12 @@ static int eb_lookup_vmas(struct i915_execbuffer *eb)
batch = eb_batch_index(eb);
+ mutex_lock(&eb->gem_context->mutex);
+ if (unlikely(i915_gem_context_is_closed(eb->gem_context))) {
+ err = -ENOENT;
+ goto err_ctx;
+ }
+
for (i = 0; i < eb->buffer_count; i++) {
u32 handle = eb->exec[i].handle;
struct i915_lut_handle *lut;
@@ -847,12 +850,14 @@ static int eb_lookup_vmas(struct i915_execbuffer *eb)
}
/* transfer ref to ctx */
- if (!vma->open_count++)
+ if (!atomic_fetch_inc(&vma->open_count))
i915_vma_reopen(vma);
- list_add(&lut->obj_link, &obj->lut_list);
- list_add(&lut->ctx_link, &eb->gem_context->handles_list);
- lut->ctx = eb->gem_context;
lut->handle = handle;
+ lut->ctx = eb->gem_context;
+
+ i915_gem_object_lock(obj);
+ list_add(&lut->obj_link, &obj->lut_list);
+ i915_gem_object_unlock(obj);
add_vma:
err = eb_add_vma(eb, i, batch, vma);
@@ -865,6 +870,8 @@ static int eb_lookup_vmas(struct i915_execbuffer *eb)
eb_vma_misplaced(&eb->exec[i], vma, eb->flags[i]));
}
+ mutex_unlock(&eb->gem_context->mutex);
+
eb->args->flags |= __EXEC_VALIDATED;
return eb_reserve(eb);
@@ -872,6 +879,8 @@ static int eb_lookup_vmas(struct i915_execbuffer *eb)
i915_gem_object_put(obj);
err_vma:
eb->vma[i] = NULL;
+err_ctx:
+ mutex_unlock(&eb->gem_context->mutex);
return err;
}
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
index c178cf7614c6..29deae2eec93 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
@@ -127,39 +127,47 @@ void i915_gem_object_set_cache_coherency(struct drm_i915_gem_object *obj,
void i915_gem_close_object(struct drm_gem_object *gem, struct drm_file *file)
{
- struct drm_i915_private *i915 = to_i915(gem->dev);
struct drm_i915_gem_object *obj = to_intel_bo(gem);
struct drm_i915_file_private *fpriv = file->driver_priv;
struct i915_lut_handle *lut, *ln;
+ LIST_HEAD(close);
- mutex_lock(&i915->drm.struct_mutex);
-
+ i915_gem_object_lock(obj);
list_for_each_entry_safe(lut, ln, &obj->lut_list, obj_link) {
struct i915_gem_context *ctx = lut->ctx;
- struct i915_vma *vma;
- GEM_BUG_ON(ctx->file_priv == ERR_PTR(-EBADF));
if (ctx->file_priv != fpriv)
continue;
- vma = radix_tree_delete(&ctx->handles_vma, lut->handle);
- GEM_BUG_ON(vma->obj != obj);
+ i915_gem_context_get(ctx);
+ list_move(&lut->obj_link, &close);
+ }
+ i915_gem_object_unlock(obj);
+
+ list_for_each_entry_safe(lut, ln, &close, obj_link) {
+ struct i915_gem_context *ctx = lut->ctx;
+ struct i915_vma *vma;
- /* We allow the process to have multiple handles to the same
+ /*
+ * We allow the process to have multiple handles to the same
* vma, in the same fd namespace, by virtue of flink/open.
*/
- GEM_BUG_ON(!vma->open_count);
- if (!--vma->open_count && !i915_vma_is_ggtt(vma))
- i915_vma_close(vma);
- list_del(&lut->obj_link);
- list_del(&lut->ctx_link);
+ mutex_lock(&ctx->mutex);
+ vma = radix_tree_delete(&ctx->handles_vma, lut->handle);
+ if (vma) {
+ GEM_BUG_ON(vma->obj != obj);
+ GEM_BUG_ON(!atomic_read(&vma->open_count));
+ if (atomic_dec_and_test(&vma->open_count) &&
+ !i915_vma_is_ggtt(vma))
+ i915_vma_close(vma);
+ }
+ mutex_unlock(&ctx->mutex);
+ i915_gem_context_put(lut->ctx);
i915_lut_handle_free(lut);
i915_gem_object_put(obj);
}
-
- mutex_unlock(&i915->drm.struct_mutex);
}
static bool discard_backing_storage(struct drm_i915_gem_object *obj)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
index 133581339d81..4a5332662bf9 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
@@ -24,7 +24,6 @@ struct drm_i915_gem_object;
*/
struct i915_lut_handle {
struct list_head obj_link;
- struct list_head ctx_link;
struct i915_gem_context *ctx;
u32 handle;
};
diff --git a/drivers/gpu/drm/i915/gem/selftests/mock_context.c b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
index d373d7238622..080c44509ba9 100644
--- a/drivers/gpu/drm/i915/gem/selftests/mock_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
@@ -30,7 +30,6 @@ mock_context(struct drm_i915_private *i915,
RCU_INIT_POINTER(ctx->engines, e);
INIT_RADIX_TREE(&ctx->handles_vma, GFP_KERNEL);
- INIT_LIST_HEAD(&ctx->handles_list);
INIT_LIST_HEAD(&ctx->hw_id_link);
mutex_init(&ctx->mutex);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index b372ac47aa1c..9546a9544c82 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1986,10 +1986,12 @@ struct drm_i915_private {
} timelines;
struct list_head active_rings;
- struct list_head closed_vma;
struct intel_wakeref wakeref;
+ struct list_head closed_vma;
+ spinlock_t closed_lock; /* guards the list of closed_vma */
+
/**
* Is the GPU currently considered idle, or busy executing
* userspace requests? Whilst idle, we allow runtime power
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 2a9e8ecf2926..90923fa6603d 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1732,6 +1732,7 @@ int i915_gem_init_early(struct drm_i915_private *dev_priv)
INIT_LIST_HEAD(&dev_priv->gt.active_rings);
INIT_LIST_HEAD(&dev_priv->gt.closed_vma);
+ spin_lock_init(&dev_priv->gt.closed_lock);
i915_gem_init__mm(dev_priv);
i915_gem_init__pm(dev_priv);
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 56be4b091bc7..f26eaa81e7fa 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -1929,6 +1929,7 @@ static struct i915_vma *pd_vma_create(struct gen6_hw_ppgtt *ppgtt, int size)
vma->ggtt_view.type = I915_GGTT_VIEW_ROTATED; /* prevent fencing */
INIT_LIST_HEAD(&vma->obj_link);
+ INIT_LIST_HEAD(&vma->closed_link);
mutex_lock(&vma->vm->mutex);
list_add(&vma->vm_link, &vma->vm->unbound_list);
diff --git a/drivers/gpu/drm/i915/i915_timeline.c b/drivers/gpu/drm/i915/i915_timeline.c
index 5fbea0892f33..000e1a9b6750 100644
--- a/drivers/gpu/drm/i915/i915_timeline.c
+++ b/drivers/gpu/drm/i915/i915_timeline.c
@@ -61,7 +61,7 @@ hwsp_alloc(struct i915_timeline *timeline, unsigned int *cacheline)
BUILD_BUG_ON(BITS_PER_TYPE(u64) * CACHELINE_BYTES > PAGE_SIZE);
- spin_lock(>->hwsp_lock);
+ spin_lock_irq(>->hwsp_lock);
/* hwsp_free_list only contains HWSP that have available cachelines */
hwsp = list_first_entry_or_null(>->hwsp_free_list,
@@ -69,7 +69,7 @@ hwsp_alloc(struct i915_timeline *timeline, unsigned int *cacheline)
if (!hwsp) {
struct i915_vma *vma;
- spin_unlock(>->hwsp_lock);
+ spin_unlock_irq(>->hwsp_lock);
hwsp = kmalloc(sizeof(*hwsp), GFP_KERNEL);
if (!hwsp)
@@ -86,7 +86,7 @@ hwsp_alloc(struct i915_timeline *timeline, unsigned int *cacheline)
hwsp->free_bitmap = ~0ull;
hwsp->gt = gt;
- spin_lock(>->hwsp_lock);
+ spin_lock_irq(>->hwsp_lock);
list_add(&hwsp->free_link, >->hwsp_free_list);
}
@@ -96,7 +96,7 @@ hwsp_alloc(struct i915_timeline *timeline, unsigned int *cacheline)
if (!hwsp->free_bitmap)
list_del(&hwsp->free_link);
- spin_unlock(>->hwsp_lock);
+ spin_unlock_irq(>->hwsp_lock);
GEM_BUG_ON(hwsp->vma->private != hwsp);
return hwsp->vma;
@@ -105,8 +105,9 @@ hwsp_alloc(struct i915_timeline *timeline, unsigned int *cacheline)
static void __idle_hwsp_free(struct i915_timeline_hwsp *hwsp, int cacheline)
{
struct i915_gt_timelines *gt = hwsp->gt;
+ unsigned long flags;
- spin_lock(>->hwsp_lock);
+ spin_lock_irqsave(>->hwsp_lock, flags);
/* As a cacheline becomes available, publish the HWSP on the freelist */
if (!hwsp->free_bitmap)
@@ -122,7 +123,7 @@ static void __idle_hwsp_free(struct i915_timeline_hwsp *hwsp, int cacheline)
kfree(hwsp);
}
- spin_unlock(>->hwsp_lock);
+ spin_unlock_irqrestore(>->hwsp_lock, flags);
}
static void __idle_cacheline_free(struct i915_timeline_cacheline *cl)
diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
index 16d47f1f645a..392c2757c217 100644
--- a/drivers/gpu/drm/i915/i915_vma.c
+++ b/drivers/gpu/drm/i915/i915_vma.c
@@ -143,6 +143,8 @@ vma_create(struct drm_i915_gem_object *obj,
vma->size = obj->base.size;
vma->display_alignment = I915_GTT_MIN_ALIGNMENT;
+ INIT_LIST_HEAD(&vma->closed_link);
+
if (view && view->type != I915_GGTT_VIEW_NORMAL) {
vma->ggtt_view = *view;
if (view->type == I915_GGTT_VIEW_PARTIAL) {
@@ -785,10 +787,10 @@ int __i915_vma_do_pin(struct i915_vma *vma,
void i915_vma_close(struct i915_vma *vma)
{
- lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
+ struct drm_i915_private *i915 = vma->vm->i915;
+ unsigned long flags;
GEM_BUG_ON(i915_vma_is_closed(vma));
- vma->flags |= I915_VMA_CLOSED;
/*
* We defer actually closing, unbinding and destroying the VMA until
@@ -802,17 +804,26 @@ void i915_vma_close(struct i915_vma *vma)
* causing us to rebind the VMA once more. This ends up being a lot
* of wasted work for the steady state.
*/
- list_add_tail(&vma->closed_link, &vma->vm->i915->gt.closed_vma);
+ spin_lock_irqsave(&i915->gt.closed_lock, flags);
+ list_add(&vma->closed_link, &i915->gt.closed_vma);
+ spin_unlock_irqrestore(&i915->gt.closed_lock, flags);
}
-void i915_vma_reopen(struct i915_vma *vma)
+static void __i915_vma_remove_closed(struct i915_vma *vma)
{
- lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
+ struct drm_i915_private *i915 = vma->vm->i915;
- if (vma->flags & I915_VMA_CLOSED) {
- vma->flags &= ~I915_VMA_CLOSED;
- list_del(&vma->closed_link);
- }
+ if (!i915_vma_is_closed(vma))
+ return;
+
+ spin_lock_irq(&i915->gt.closed_lock);
+ list_del_init(&vma->closed_link);
+ spin_unlock_irq(&i915->gt.closed_lock);
+}
+
+void i915_vma_reopen(struct i915_vma *vma)
+{
+ __i915_vma_remove_closed(vma);
}
static void __i915_vma_destroy(struct i915_vma *vma)
@@ -846,8 +857,7 @@ void i915_vma_destroy(struct i915_vma *vma)
GEM_BUG_ON(i915_vma_is_pinned(vma));
- if (i915_vma_is_closed(vma))
- list_del(&vma->closed_link);
+ __i915_vma_remove_closed(vma);
WARN_ON(i915_vma_unbind(vma));
GEM_BUG_ON(i915_vma_is_active(vma));
@@ -859,12 +869,16 @@ void i915_vma_parked(struct drm_i915_private *i915)
{
struct i915_vma *vma, *next;
+ spin_lock_irq(&i915->gt.closed_lock);
list_for_each_entry_safe(vma, next, &i915->gt.closed_vma, closed_link) {
- GEM_BUG_ON(!i915_vma_is_closed(vma));
+ list_del_init(&vma->closed_link);
+ spin_unlock_irq(&i915->gt.closed_lock);
+
i915_vma_destroy(vma);
- }
- GEM_BUG_ON(!list_empty(&i915->gt.closed_vma));
+ spin_lock_irq(&i915->gt.closed_lock);
+ }
+ spin_unlock_irq(&i915->gt.closed_lock);
}
static void __i915_vma_iounmap(struct i915_vma *vma)
diff --git a/drivers/gpu/drm/i915/i915_vma.h b/drivers/gpu/drm/i915/i915_vma.h
index 71ac7ee8620a..f0884cc8b2d3 100644
--- a/drivers/gpu/drm/i915/i915_vma.h
+++ b/drivers/gpu/drm/i915/i915_vma.h
@@ -69,7 +69,7 @@ struct i915_vma {
* handles (but same file) for execbuf, i.e. the number of aliases
* that exist in the ctx->handle_vmas LUT for this vma.
*/
- unsigned int open_count;
+ atomic_t open_count;
unsigned long flags;
/**
* How many users have pinned this object in GTT space.
@@ -104,10 +104,9 @@ struct i915_vma {
#define I915_VMA_GGTT BIT(11)
#define I915_VMA_CAN_FENCE BIT(12)
-#define I915_VMA_CLOSED BIT(13)
-#define I915_VMA_USERFAULT_BIT 14
+#define I915_VMA_USERFAULT_BIT 13
#define I915_VMA_USERFAULT BIT(I915_VMA_USERFAULT_BIT)
-#define I915_VMA_GGTT_WRITE BIT(15)
+#define I915_VMA_GGTT_WRITE BIT(14)
struct i915_active active;
struct i915_active_request last_fence;
@@ -190,11 +189,6 @@ static inline bool i915_vma_is_map_and_fenceable(const struct i915_vma *vma)
return vma->flags & I915_VMA_CAN_FENCE;
}
-static inline bool i915_vma_is_closed(const struct i915_vma *vma)
-{
- return vma->flags & I915_VMA_CLOSED;
-}
-
static inline bool i915_vma_set_userfault(struct i915_vma *vma)
{
GEM_BUG_ON(!i915_vma_is_map_and_fenceable(vma));
@@ -211,6 +205,11 @@ static inline bool i915_vma_has_userfault(const struct i915_vma *vma)
return test_bit(I915_VMA_USERFAULT_BIT, &vma->flags);
}
+static inline bool i915_vma_is_closed(const struct i915_vma *vma)
+{
+ return !list_empty(&vma->closed_link);
+}
+
static inline u32 i915_ggtt_offset(const struct i915_vma *vma)
{
GEM_BUG_ON(!i915_vma_is_ggtt(vma));
diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
index e25b74a27f83..b7f3fbb4ae89 100644
--- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
+++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
@@ -203,6 +203,7 @@ struct drm_i915_private *mock_gem_device(void)
INIT_LIST_HEAD(&i915->gt.active_rings);
INIT_LIST_HEAD(&i915->gt.closed_vma);
+ spin_lock_init(&i915->gt.closed_lock);
mutex_lock(&i915->drm.struct_mutex);
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 34/40] drm/i915: Rename intel_context.active to .inflight
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (31 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 33/40] drm/i915: Move object close under its own lock Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-10 14:44 ` Mika Kuoppala
2019-05-08 8:06 ` [PATCH 35/40] drm/i915: Keep contexts pinned until after the next kernel context switch Chris Wilson
` (12 subsequent siblings)
45 siblings, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
Rename the engine this HW context is currently active upon (that we are
flying upon) to disambiguate between the mixture of different active
terms (and prevent conflict in future patches).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/gt/intel_context_types.h | 2 +-
drivers/gpu/drm/i915/gt/intel_lrc.c | 22 +++++++++----------
2 files changed, 12 insertions(+), 12 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_context_types.h b/drivers/gpu/drm/i915/gt/intel_context_types.h
index 963a312430e6..825fcf0ac9c4 100644
--- a/drivers/gpu/drm/i915/gt/intel_context_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_context_types.h
@@ -37,7 +37,7 @@ struct intel_context {
struct i915_gem_context *gem_context;
struct intel_engine_cs *engine;
- struct intel_engine_cs *active;
+ struct intel_engine_cs *inflight;
struct list_head signal_link;
struct list_head signals;
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 64bd25a9e6f5..5e418bf46c46 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -460,7 +460,7 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
__i915_request_unsubmit(rq);
unwind_wa_tail(rq);
- GEM_BUG_ON(rq->hw_context->active);
+ GEM_BUG_ON(rq->hw_context->inflight);
/*
* Push the request back into the queue for later resubmission.
@@ -557,11 +557,11 @@ execlists_user_end(struct intel_engine_execlists *execlists)
static inline void
execlists_context_schedule_in(struct i915_request *rq)
{
- GEM_BUG_ON(rq->hw_context->active);
+ GEM_BUG_ON(rq->hw_context->inflight);
execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN);
intel_engine_context_in(rq->engine);
- rq->hw_context->active = rq->engine;
+ rq->hw_context->inflight = rq->engine;
}
static void kick_siblings(struct i915_request *rq)
@@ -576,7 +576,7 @@ static void kick_siblings(struct i915_request *rq)
static inline void
execlists_context_schedule_out(struct i915_request *rq, unsigned long status)
{
- rq->hw_context->active = NULL;
+ rq->hw_context->inflight = NULL;
intel_engine_context_out(rq->engine);
execlists_context_status_change(rq, status);
trace_i915_request_out(rq);
@@ -820,7 +820,7 @@ static bool virtual_matches(const struct virtual_engine *ve,
const struct i915_request *rq,
const struct intel_engine_cs *engine)
{
- const struct intel_engine_cs *active;
+ const struct intel_engine_cs *inflight;
if (!(rq->execution_mask & engine->mask)) /* We peeked too soon! */
return false;
@@ -834,8 +834,8 @@ static bool virtual_matches(const struct virtual_engine *ve,
* we reuse the register offsets). This is a very small
* hystersis on the greedy seelction algorithm.
*/
- active = READ_ONCE(ve->context.active);
- if (active && active != engine)
+ inflight = READ_ONCE(ve->context.inflight);
+ if (inflight && inflight != engine)
return false;
return true;
@@ -1023,7 +1023,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
u32 *regs = ve->context.lrc_reg_state;
unsigned int n;
- GEM_BUG_ON(READ_ONCE(ve->context.active));
+ GEM_BUG_ON(READ_ONCE(ve->context.inflight));
virtual_update_register_offsets(regs, engine);
if (!list_empty(&ve->context.signals))
@@ -1501,7 +1501,7 @@ static void execlists_context_unpin(struct intel_context *ce)
* had the chance to run yet; let it run before we teardown the
* reference it may use.
*/
- engine = READ_ONCE(ce->active);
+ engine = READ_ONCE(ce->inflight);
if (unlikely(engine)) {
unsigned long flags;
@@ -1509,7 +1509,7 @@ static void execlists_context_unpin(struct intel_context *ce)
process_csb(engine);
spin_unlock_irqrestore(&engine->timeline.lock, flags);
- GEM_BUG_ON(READ_ONCE(ce->active));
+ GEM_BUG_ON(READ_ONCE(ce->inflight));
}
i915_gem_context_unpin_hw_id(ce->gem_context);
@@ -3103,7 +3103,7 @@ static void virtual_context_destroy(struct kref *kref)
unsigned int n;
GEM_BUG_ON(ve->request);
- GEM_BUG_ON(ve->context.active);
+ GEM_BUG_ON(ve->context.inflight);
for (n = 0; n < ve->num_siblings; n++) {
struct intel_engine_cs *sibling = ve->siblings[n];
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* Re: [PATCH 34/40] drm/i915: Rename intel_context.active to .inflight
2019-05-08 8:06 ` [PATCH 34/40] drm/i915: Rename intel_context.active to .inflight Chris Wilson
@ 2019-05-10 14:44 ` Mika Kuoppala
0 siblings, 0 replies; 71+ messages in thread
From: Mika Kuoppala @ 2019-05-10 14:44 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
Chris Wilson <chris@chris-wilson.co.uk> writes:
> Rename the engine this HW context is currently active upon (that we are
> flying upon) to disambiguate between the mixture of different active
> terms (and prevent conflict in future patches).
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/gt/intel_context_types.h | 2 +-
> drivers/gpu/drm/i915/gt/intel_lrc.c | 22 +++++++++----------
> 2 files changed, 12 insertions(+), 12 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_context_types.h b/drivers/gpu/drm/i915/gt/intel_context_types.h
> index 963a312430e6..825fcf0ac9c4 100644
> --- a/drivers/gpu/drm/i915/gt/intel_context_types.h
> +++ b/drivers/gpu/drm/i915/gt/intel_context_types.h
> @@ -37,7 +37,7 @@ struct intel_context {
>
> struct i915_gem_context *gem_context;
> struct intel_engine_cs *engine;
> - struct intel_engine_cs *active;
> + struct intel_engine_cs *inflight;
Active_on came to my mind when first reading this. As discussed
in irc 'active' is already quite overloaded so inflight seems
fitting.
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
>
> struct list_head signal_link;
> struct list_head signals;
> diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
> index 64bd25a9e6f5..5e418bf46c46 100644
> --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
> +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
> @@ -460,7 +460,7 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
> __i915_request_unsubmit(rq);
> unwind_wa_tail(rq);
>
> - GEM_BUG_ON(rq->hw_context->active);
> + GEM_BUG_ON(rq->hw_context->inflight);
>
> /*
> * Push the request back into the queue for later resubmission.
> @@ -557,11 +557,11 @@ execlists_user_end(struct intel_engine_execlists *execlists)
> static inline void
> execlists_context_schedule_in(struct i915_request *rq)
> {
> - GEM_BUG_ON(rq->hw_context->active);
> + GEM_BUG_ON(rq->hw_context->inflight);
>
> execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN);
> intel_engine_context_in(rq->engine);
> - rq->hw_context->active = rq->engine;
> + rq->hw_context->inflight = rq->engine;
> }
>
> static void kick_siblings(struct i915_request *rq)
> @@ -576,7 +576,7 @@ static void kick_siblings(struct i915_request *rq)
> static inline void
> execlists_context_schedule_out(struct i915_request *rq, unsigned long status)
> {
> - rq->hw_context->active = NULL;
> + rq->hw_context->inflight = NULL;
> intel_engine_context_out(rq->engine);
> execlists_context_status_change(rq, status);
> trace_i915_request_out(rq);
> @@ -820,7 +820,7 @@ static bool virtual_matches(const struct virtual_engine *ve,
> const struct i915_request *rq,
> const struct intel_engine_cs *engine)
> {
> - const struct intel_engine_cs *active;
> + const struct intel_engine_cs *inflight;
>
> if (!(rq->execution_mask & engine->mask)) /* We peeked too soon! */
> return false;
> @@ -834,8 +834,8 @@ static bool virtual_matches(const struct virtual_engine *ve,
> * we reuse the register offsets). This is a very small
> * hystersis on the greedy seelction algorithm.
> */
> - active = READ_ONCE(ve->context.active);
> - if (active && active != engine)
> + inflight = READ_ONCE(ve->context.inflight);
> + if (inflight && inflight != engine)
> return false;
>
> return true;
> @@ -1023,7 +1023,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
> u32 *regs = ve->context.lrc_reg_state;
> unsigned int n;
>
> - GEM_BUG_ON(READ_ONCE(ve->context.active));
> + GEM_BUG_ON(READ_ONCE(ve->context.inflight));
> virtual_update_register_offsets(regs, engine);
>
> if (!list_empty(&ve->context.signals))
> @@ -1501,7 +1501,7 @@ static void execlists_context_unpin(struct intel_context *ce)
> * had the chance to run yet; let it run before we teardown the
> * reference it may use.
> */
> - engine = READ_ONCE(ce->active);
> + engine = READ_ONCE(ce->inflight);
> if (unlikely(engine)) {
> unsigned long flags;
>
> @@ -1509,7 +1509,7 @@ static void execlists_context_unpin(struct intel_context *ce)
> process_csb(engine);
> spin_unlock_irqrestore(&engine->timeline.lock, flags);
>
> - GEM_BUG_ON(READ_ONCE(ce->active));
> + GEM_BUG_ON(READ_ONCE(ce->inflight));
> }
>
> i915_gem_context_unpin_hw_id(ce->gem_context);
> @@ -3103,7 +3103,7 @@ static void virtual_context_destroy(struct kref *kref)
> unsigned int n;
>
> GEM_BUG_ON(ve->request);
> - GEM_BUG_ON(ve->context.active);
> + GEM_BUG_ON(ve->context.inflight);
>
> for (n = 0; n < ve->num_siblings; n++) {
> struct intel_engine_cs *sibling = ve->siblings[n];
> --
> 2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread
* [PATCH 35/40] drm/i915: Keep contexts pinned until after the next kernel context switch
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (32 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 34/40] drm/i915: Rename intel_context.active to .inflight Chris Wilson
@ 2019-05-08 8:06 ` Chris Wilson
2019-05-08 8:07 ` [PATCH 36/40] drm/i915: Stop retiring along engine Chris Wilson
` (11 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:06 UTC (permalink / raw)
To: intel-gfx
We need to keep the context image pinned in memory until after the GPU
has finished writing into it. Since it continues to write as we signal
the final breadcrumb, we need to keep it pinned until the request after
it is complete. Currently we know the order in which requests execute on
each engine, and so to remove that presumption we need to identify a
request/context-switch we know must occur after our completion. Any
request queued after the signal must imply a context switch, for
simplicity we use a fresh request from the kernel context.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/gem/i915_gem_context.c | 24 ++----
drivers/gpu/drm/i915/gem/i915_gem_context.h | 1 -
drivers/gpu/drm/i915/gem/i915_gem_pm.c | 20 ++++-
drivers/gpu/drm/i915/gt/intel_context.c | 80 +++++++++++++++---
drivers/gpu/drm/i915/gt/intel_context.h | 3 +
drivers/gpu/drm/i915/gt/intel_context_types.h | 6 +-
drivers/gpu/drm/i915/gt/intel_engine.h | 2 -
drivers/gpu/drm/i915/gt/intel_engine_cs.c | 23 +-----
drivers/gpu/drm/i915/gt/intel_engine_pm.c | 2 +
drivers/gpu/drm/i915/gt/intel_engine_types.h | 13 +--
drivers/gpu/drm/i915/gt/intel_lrc.c | 62 ++------------
drivers/gpu/drm/i915/gt/intel_ringbuffer.c | 44 +---------
drivers/gpu/drm/i915/gt/mock_engine.c | 11 +--
drivers/gpu/drm/i915/i915_active.c | 81 ++++++++++++++++++-
drivers/gpu/drm/i915/i915_active.h | 5 ++
drivers/gpu/drm/i915/i915_active_types.h | 3 +
drivers/gpu/drm/i915/i915_gem.c | 4 -
drivers/gpu/drm/i915/i915_request.c | 15 ----
.../gpu/drm/i915/selftests/mock_gem_device.c | 1 -
19 files changed, 215 insertions(+), 185 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index a9608d9ced6a..aa2bd1d6ceb6 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -688,17 +688,6 @@ int i915_gem_contexts_init(struct drm_i915_private *dev_priv)
return 0;
}
-void i915_gem_contexts_lost(struct drm_i915_private *dev_priv)
-{
- struct intel_engine_cs *engine;
- enum intel_engine_id id;
-
- lockdep_assert_held(&dev_priv->drm.struct_mutex);
-
- for_each_engine(engine, dev_priv, id)
- intel_engine_lost_context(engine);
-}
-
void i915_gem_contexts_fini(struct drm_i915_private *i915)
{
lockdep_assert_held(&i915->drm.struct_mutex);
@@ -1183,10 +1172,6 @@ gen8_modify_rpcs(struct intel_context *ce, struct intel_sseu sseu)
if (ret)
goto out_add;
- ret = gen8_emit_rpcs_config(rq, ce, sseu);
- if (ret)
- goto out_add;
-
/*
* Guarantee context image and the timeline remains pinned until the
* modifying request is retired by setting the ce activity tracker.
@@ -1194,9 +1179,12 @@ gen8_modify_rpcs(struct intel_context *ce, struct intel_sseu sseu)
* But we only need to take one pin on the account of it. Or in other
* words transfer the pinned ce object to tracked active request.
*/
- if (!i915_active_request_isset(&ce->active_tracker))
- __intel_context_pin(ce);
- __i915_active_request_set(&ce->active_tracker, rq);
+ GEM_BUG_ON(i915_active_is_idle(&ce->active));
+ ret = i915_active_ref(&ce->active, rq->fence.context, rq);
+ if (ret)
+ goto out_add;
+
+ ret = gen8_emit_rpcs_config(rq, ce, sseu);
out_add:
i915_request_add(rq);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.h b/drivers/gpu/drm/i915/gem/i915_gem_context.h
index 630392c77e48..9691dd062f72 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.h
@@ -134,7 +134,6 @@ static inline bool i915_gem_context_is_kernel(struct i915_gem_context *ctx)
/* i915_gem_context.c */
int __must_check i915_gem_contexts_init(struct drm_i915_private *dev_priv);
-void i915_gem_contexts_lost(struct drm_i915_private *dev_priv);
void i915_gem_contexts_fini(struct drm_i915_private *dev_priv);
int i915_gem_context_open(struct drm_i915_private *i915,
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
index 111b687f14b5..1f2feddd9e20 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
@@ -10,6 +10,22 @@
#include "i915_drv.h"
#include "i915_globals.h"
+static void call_idle_barriers(struct intel_engine_cs *engine)
+{
+ struct llist_node *node, *next;
+
+ llist_for_each_safe(node, next, llist_del_all(&engine->barrier_tasks)) {
+ struct i915_active_request *active =
+ container_of((struct list_head *)node,
+ typeof(*active), link);
+
+ INIT_LIST_HEAD(&active->link);
+ RCU_INIT_POINTER(active->request, NULL);
+
+ active->retire(active, NULL);
+ }
+}
+
static void i915_gem_park(struct drm_i915_private *i915)
{
struct intel_engine_cs *engine;
@@ -17,8 +33,10 @@ static void i915_gem_park(struct drm_i915_private *i915)
lockdep_assert_held(&i915->drm.struct_mutex);
- for_each_engine(engine, i915, id)
+ for_each_engine(engine, i915, id) {
+ call_idle_barriers(engine); /* cleanup after wedging */
i915_gem_batch_pool_fini(&engine->batch_pool);
+ }
i915_timelines_park(i915);
i915_vma_parked(i915);
diff --git a/drivers/gpu/drm/i915/gt/intel_context.c b/drivers/gpu/drm/i915/gt/intel_context.c
index c78ec0b58e77..c10eb4904264 100644
--- a/drivers/gpu/drm/i915/gt/intel_context.c
+++ b/drivers/gpu/drm/i915/gt/intel_context.c
@@ -61,7 +61,6 @@ int __intel_context_do_pin(struct intel_context *ce)
i915_gem_context_get(ce->gem_context); /* for ctx->ppgtt */
- intel_context_get(ce);
smp_mb__before_atomic(); /* flush pin before it is visible */
}
@@ -89,20 +88,45 @@ void intel_context_unpin(struct intel_context *ce)
ce->ops->unpin(ce);
i915_gem_context_put(ce->gem_context);
- intel_context_put(ce);
+ intel_context_inactive(ce);
}
mutex_unlock(&ce->pin_mutex);
intel_context_put(ce);
}
-static void intel_context_retire(struct i915_active_request *active,
- struct i915_request *rq)
+static int __context_pin_state(struct i915_vma *vma, unsigned long flags)
{
- struct intel_context *ce =
- container_of(active, typeof(*ce), active_tracker);
+ int err;
- intel_context_unpin(ce);
+ err = i915_vma_pin(vma, 0, 0, flags | PIN_GLOBAL);
+ if (err)
+ return err;
+
+ /*
+ * And mark it as a globally pinned object to let the shrinker know
+ * it cannot reclaim the object until we release it.
+ */
+ vma->obj->pin_global++;
+ vma->obj->mm.dirty = true;
+
+ return 0;
+}
+
+static void __context_unpin_state(struct i915_vma *vma)
+{
+ vma->obj->pin_global--;
+ __i915_vma_unpin(vma);
+}
+
+static void intel_context_retire(struct i915_active *active)
+{
+ struct intel_context *ce = container_of(active, typeof(*ce), active);
+
+ if (ce->state)
+ __context_unpin_state(ce->state);
+
+ intel_context_put(ce);
}
void
@@ -125,8 +149,46 @@ intel_context_init(struct intel_context *ce,
mutex_init(&ce->pin_mutex);
- i915_active_request_init(&ce->active_tracker,
- NULL, intel_context_retire);
+ i915_active_init(ctx->i915, &ce->active, intel_context_retire);
+}
+
+int intel_context_active(struct intel_context *ce, unsigned long flags)
+{
+ int err;
+
+ if (!i915_active_acquire(&ce->active))
+ return 0;
+
+ intel_context_get(ce);
+
+ if (!ce->state)
+ return 0;
+
+ err = __context_pin_state(ce->state, flags);
+ if (err) {
+ i915_active_cancel(&ce->active);
+ intel_context_put(ce);
+ return err;
+ }
+
+ /* Preallocate tracking nodes */
+ if (!i915_gem_context_is_kernel(ce->gem_context)) {
+ err = i915_active_acquire_preallocate_barrier(&ce->active,
+ ce->engine);
+ if (err) {
+ i915_active_release(&ce->active);
+ return err;
+ }
+ }
+
+ return 0;
+}
+
+void intel_context_inactive(struct intel_context *ce)
+{
+ /* Nodes preallocated in intel_context_active() */
+ i915_active_acquire_barrier(&ce->active);
+ i915_active_release(&ce->active);
}
static void i915_global_context_shrink(void)
diff --git a/drivers/gpu/drm/i915/gt/intel_context.h b/drivers/gpu/drm/i915/gt/intel_context.h
index 63392c88cd98..e71629f7c2e0 100644
--- a/drivers/gpu/drm/i915/gt/intel_context.h
+++ b/drivers/gpu/drm/i915/gt/intel_context.h
@@ -102,6 +102,9 @@ static inline void intel_context_exit(struct intel_context *ce)
ce->ops->exit(ce);
}
+int intel_context_active(struct intel_context *ce, unsigned long flags);
+void intel_context_inactive(struct intel_context *ce);
+
static inline struct intel_context *intel_context_get(struct intel_context *ce)
{
kref_get(&ce->ref);
diff --git a/drivers/gpu/drm/i915/gt/intel_context_types.h b/drivers/gpu/drm/i915/gt/intel_context_types.h
index 825fcf0ac9c4..e95be4be9612 100644
--- a/drivers/gpu/drm/i915/gt/intel_context_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_context_types.h
@@ -56,10 +56,10 @@ struct intel_context {
intel_engine_mask_t saturated; /* submitting semaphores too late? */
/**
- * active_tracker: Active tracker for the external rq activity
- * on this intel_context object.
+ * active: Active tracker for the rq activity (inc. external) on this
+ * intel_context object.
*/
- struct i915_active_request active_tracker;
+ struct i915_active active;
const struct intel_context_ops *ops;
diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h
index 9359b3a7ad9c..1ae2934b7ecc 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine.h
@@ -448,8 +448,6 @@ static inline void intel_engine_reset(struct intel_engine_cs *engine,
bool intel_engine_is_idle(struct intel_engine_cs *engine);
bool intel_engines_are_idle(struct drm_i915_private *dev_priv);
-void intel_engine_lost_context(struct intel_engine_cs *engine);
-
void intel_engines_reset_default_submission(struct drm_i915_private *i915);
unsigned int intel_engines_has_context_isolation(struct drm_i915_private *i915);
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
index 4d911f60d1f8..794aa8b11742 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
@@ -561,6 +561,8 @@ static int intel_engine_setup_common(struct intel_engine_cs *engine)
{
int err;
+ init_llist_head(&engine->barrier_tasks);
+
err = init_status_page(engine);
if (err)
return err;
@@ -820,6 +822,7 @@ void intel_engine_cleanup_common(struct intel_engine_cs *engine)
if (engine->preempt_context)
intel_context_unpin(engine->preempt_context);
intel_context_unpin(engine->kernel_context);
+ GEM_BUG_ON(!llist_empty(&engine->barrier_tasks));
i915_timeline_fini(&engine->timeline);
@@ -1150,26 +1153,6 @@ void intel_engines_reset_default_submission(struct drm_i915_private *i915)
engine->set_default_submission(engine);
}
-/**
- * intel_engine_lost_context: called when the GPU is reset into unknown state
- * @engine: the engine
- *
- * We have either reset the GPU or otherwise about to lose state tracking of
- * the current GPU logical state (e.g. suspend). On next use, it is therefore
- * imperative that we make no presumptions about the current state and load
- * from scratch.
- */
-void intel_engine_lost_context(struct intel_engine_cs *engine)
-{
- struct intel_context *ce;
-
- lockdep_assert_held(&engine->i915->drm.struct_mutex);
-
- ce = fetch_and_zero(&engine->last_retired_context);
- if (ce)
- intel_context_unpin(ce);
-}
-
bool intel_engine_can_store_dword(struct intel_engine_cs *engine)
{
switch (INTEL_GEN(engine->i915)) {
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pm.c b/drivers/gpu/drm/i915/gt/intel_engine_pm.c
index ccf034764741..3c448a061abd 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_pm.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_pm.c
@@ -88,6 +88,8 @@ static bool switch_to_kernel_context(struct intel_engine_cs *engine)
/* Check again on the next retirement. */
engine->wakeref_serial = engine->serial + 1;
+
+ i915_request_add_barriers(rq);
__i915_request_commit(rq);
return false;
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
index 2c232501d973..d3d5a505f621 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
@@ -11,6 +11,7 @@
#include <linux/irq_work.h>
#include <linux/kref.h>
#include <linux/list.h>
+#include <linux/llist.h>
#include <linux/types.h>
#include "i915_gem.h"
@@ -286,6 +287,7 @@ struct intel_engine_cs {
struct intel_ring *buffer;
struct i915_timeline timeline;
+ struct llist_head barrier_tasks;
struct intel_context *kernel_context; /* pinned */
struct intel_context *preempt_context; /* pinned; optional */
@@ -433,17 +435,6 @@ struct intel_engine_cs {
struct intel_engine_execlists execlists;
- /* Contexts are pinned whilst they are active on the GPU. The last
- * context executed remains active whilst the GPU is idle - the
- * switch away and write to the context object only occurs on the
- * next execution. Contexts are only unpinned on retirement of the
- * following request ensuring that we can always write to the object
- * on the context switch even after idling. Across suspend, we switch
- * to the kernel context and trash it as the save may not happen
- * before the hardware is powered down.
- */
- struct intel_context *last_retired_context;
-
/* status_notifier: list of callbacks for context-switch changes */
struct atomic_notifier_head context_status_notifier;
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 5e418bf46c46..bc135a9e6d42 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -1464,60 +1464,11 @@ static void execlists_context_destroy(struct kref *kref)
intel_context_free(ce);
}
-static int __context_pin(struct i915_vma *vma)
-{
- unsigned int flags;
- int err;
-
- flags = PIN_GLOBAL | PIN_HIGH;
- flags |= PIN_OFFSET_BIAS | i915_ggtt_pin_bias(vma);
-
- err = i915_vma_pin(vma, 0, 0, flags);
- if (err)
- return err;
-
- vma->obj->pin_global++;
- vma->obj->mm.dirty = true;
-
- return 0;
-}
-
-static void __context_unpin(struct i915_vma *vma)
-{
- vma->obj->pin_global--;
- __i915_vma_unpin(vma);
-}
-
static void execlists_context_unpin(struct intel_context *ce)
{
- struct intel_engine_cs *engine;
-
- /*
- * The tasklet may still be using a pointer to our state, via an
- * old request. However, since we know we only unpin the context
- * on retirement of the following request, we know that the last
- * request referencing us will have had a completion CS interrupt.
- * If we see that it is still active, it means that the tasklet hasn't
- * had the chance to run yet; let it run before we teardown the
- * reference it may use.
- */
- engine = READ_ONCE(ce->inflight);
- if (unlikely(engine)) {
- unsigned long flags;
-
- spin_lock_irqsave(&engine->timeline.lock, flags);
- process_csb(engine);
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
-
- GEM_BUG_ON(READ_ONCE(ce->inflight));
- }
-
i915_gem_context_unpin_hw_id(ce->gem_context);
-
- intel_ring_unpin(ce->ring);
-
i915_gem_object_unpin_map(ce->state->obj);
- __context_unpin(ce->state);
+ intel_ring_unpin(ce->ring);
}
static void
@@ -1554,7 +1505,10 @@ __execlists_context_pin(struct intel_context *ce,
goto err;
GEM_BUG_ON(!ce->state);
- ret = __context_pin(ce->state);
+ ret = intel_context_active(ce,
+ engine->i915->ggtt.pin_bias |
+ PIN_OFFSET_BIAS |
+ PIN_HIGH);
if (ret)
goto err;
@@ -1563,7 +1517,7 @@ __execlists_context_pin(struct intel_context *ce,
I915_MAP_OVERRIDE);
if (IS_ERR(vaddr)) {
ret = PTR_ERR(vaddr);
- goto unpin_vma;
+ goto unpin_active;
}
ret = intel_ring_pin(ce->ring);
@@ -1584,8 +1538,8 @@ __execlists_context_pin(struct intel_context *ce,
intel_ring_unpin(ce->ring);
unpin_map:
i915_gem_object_unpin_map(ce->state->obj);
-unpin_vma:
- __context_unpin(ce->state);
+unpin_active:
+ intel_context_inactive(ce);
err:
return ret;
}
diff --git a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
index ff58d658e3e2..fc596adc637a 100644
--- a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
@@ -1349,45 +1349,9 @@ static void __context_unpin_ppgtt(struct i915_gem_context *ctx)
gen6_ppgtt_unpin(ppgtt);
}
-static int __context_pin(struct intel_context *ce)
-{
- struct i915_vma *vma;
- int err;
-
- vma = ce->state;
- if (!vma)
- return 0;
-
- err = i915_vma_pin(vma, 0, 0, PIN_GLOBAL | PIN_HIGH);
- if (err)
- return err;
-
- /*
- * And mark is as a globally pinned object to let the shrinker know
- * it cannot reclaim the object until we release it.
- */
- vma->obj->pin_global++;
- vma->obj->mm.dirty = true;
-
- return 0;
-}
-
-static void __context_unpin(struct intel_context *ce)
-{
- struct i915_vma *vma;
-
- vma = ce->state;
- if (!vma)
- return;
-
- vma->obj->pin_global--;
- i915_vma_unpin(vma);
-}
-
static void ring_context_unpin(struct intel_context *ce)
{
__context_unpin_ppgtt(ce->gem_context);
- __context_unpin(ce);
}
static struct i915_vma *
@@ -1477,18 +1441,18 @@ static int ring_context_pin(struct intel_context *ce)
ce->state = vma;
}
- err = __context_pin(ce);
+ err = intel_context_active(ce, PIN_HIGH);
if (err)
return err;
err = __context_pin_ppgtt(ce->gem_context);
if (err)
- goto err_unpin;
+ goto err_active;
return 0;
-err_unpin:
- __context_unpin(ce);
+err_active:
+ intel_context_inactive(ce);
return err;
}
diff --git a/drivers/gpu/drm/i915/gt/mock_engine.c b/drivers/gpu/drm/i915/gt/mock_engine.c
index 6d7562769eb2..b7675ef18523 100644
--- a/drivers/gpu/drm/i915/gt/mock_engine.c
+++ b/drivers/gpu/drm/i915/gt/mock_engine.c
@@ -146,12 +146,18 @@ static void mock_context_destroy(struct kref *ref)
static int mock_context_pin(struct intel_context *ce)
{
+ int ret;
+
if (!ce->ring) {
ce->ring = mock_ring(ce->engine);
if (!ce->ring)
return -ENOMEM;
}
+ ret = intel_context_active(ce, PIN_HIGH);
+ if (ret)
+ return ret;
+
mock_timeline_pin(ce->ring->timeline);
return 0;
}
@@ -328,14 +334,9 @@ void mock_engine_free(struct intel_engine_cs *engine)
{
struct mock_engine *mock =
container_of(engine, typeof(*mock), base);
- struct intel_context *ce;
GEM_BUG_ON(timer_pending(&mock->hw_delay));
- ce = fetch_and_zero(&engine->last_retired_context);
- if (ce)
- intel_context_unpin(ce);
-
intel_context_unpin(engine->kernel_context);
intel_engine_fini_breadcrumbs(engine);
diff --git a/drivers/gpu/drm/i915/i915_active.c b/drivers/gpu/drm/i915/i915_active.c
index 863ae12707ba..563299e50901 100644
--- a/drivers/gpu/drm/i915/i915_active.c
+++ b/drivers/gpu/drm/i915/i915_active.c
@@ -100,7 +100,7 @@ active_instance(struct i915_active *ref, u64 idx)
parent = *p;
node = rb_entry(parent, struct active_node, node);
- if (node->timeline == idx)
+ if (node->timeline == idx && !IS_ERR(node->base.request))
goto replace;
if (node->timeline < idx)
@@ -157,6 +157,7 @@ void i915_active_init(struct drm_i915_private *i915,
ref->retire = retire;
ref->tree = RB_ROOT;
i915_active_request_init(&ref->last, NULL, last_retire);
+ init_llist_head(&ref->barriers);
ref->count = 0;
}
@@ -263,6 +264,84 @@ void i915_active_fini(struct i915_active *ref)
}
#endif
+int i915_active_acquire_preallocate_barrier(struct i915_active *ref,
+ struct intel_engine_cs *engine)
+{
+ struct drm_i915_private *i915 = engine->i915;
+ unsigned long tmp;
+ int err = 0;
+
+ i915_active_acquire(ref);
+ for_each_engine_masked(engine, i915, engine->mask, tmp) {
+ struct intel_context *kctx = engine->kernel_context;
+ struct active_node *node;
+
+ node = kmem_cache_alloc(global.slab_cache, GFP_KERNEL);
+ if (unlikely(!node)) {
+ err = -ENOMEM;
+ break;
+ }
+
+ i915_active_request_init(&node->base,
+ (void *)engine, node_retire);
+ node->timeline = kctx->ring->timeline->fence_context;
+ node->ref = ref;
+ ref->count++;
+
+ llist_add((struct llist_node *)&node->base.link,
+ &ref->barriers);
+ }
+ i915_active_release(ref);
+
+ return err;
+}
+
+void i915_active_acquire_barrier(struct i915_active *ref)
+{
+ struct llist_node *pos, *next;
+
+ i915_active_acquire(ref);
+
+ llist_for_each_safe(pos, next, llist_del_all(&ref->barriers)) {
+ struct intel_engine_cs *engine;
+ struct active_node *node;
+ struct rb_node **p, *parent;
+
+ node = container_of((struct list_head *)pos,
+ typeof(*node), base.link);
+
+ engine = (void *)rcu_access_pointer(node->base.request);
+ RCU_INIT_POINTER(node->base.request, ERR_PTR(-EAGAIN));
+
+ parent = NULL;
+ p = &ref->tree.rb_node;
+ while (*p) {
+ parent = *p;
+ if (rb_entry(parent,
+ struct active_node,
+ node)->timeline < node->timeline)
+ p = &parent->rb_right;
+ else
+ p = &parent->rb_left;
+ }
+ rb_link_node(&node->node, parent, p);
+ rb_insert_color(&node->node, &ref->tree);
+
+ llist_add((struct llist_node *)&node->base.link,
+ &engine->barrier_tasks);
+ }
+ i915_active_release(ref);
+}
+
+void i915_request_add_barriers(struct i915_request *rq)
+{
+ struct intel_engine_cs *engine = rq->engine;
+ struct llist_node *node, *next;
+
+ llist_for_each_safe(node, next, llist_del_all(&engine->barrier_tasks))
+ list_add_tail((struct list_head *)node, &rq->active_list);
+}
+
int i915_active_request_set(struct i915_active_request *active,
struct i915_request *rq)
{
diff --git a/drivers/gpu/drm/i915/i915_active.h b/drivers/gpu/drm/i915/i915_active.h
index 7d758719ce39..d55d37673944 100644
--- a/drivers/gpu/drm/i915/i915_active.h
+++ b/drivers/gpu/drm/i915/i915_active.h
@@ -406,4 +406,9 @@ void i915_active_fini(struct i915_active *ref);
static inline void i915_active_fini(struct i915_active *ref) { }
#endif
+int i915_active_acquire_preallocate_barrier(struct i915_active *ref,
+ struct intel_engine_cs *engine);
+void i915_active_acquire_barrier(struct i915_active *ref);
+void i915_request_add_barriers(struct i915_request *rq);
+
#endif /* _I915_ACTIVE_H_ */
diff --git a/drivers/gpu/drm/i915/i915_active_types.h b/drivers/gpu/drm/i915/i915_active_types.h
index b679253b53a5..c025991b9233 100644
--- a/drivers/gpu/drm/i915/i915_active_types.h
+++ b/drivers/gpu/drm/i915/i915_active_types.h
@@ -7,6 +7,7 @@
#ifndef _I915_ACTIVE_TYPES_H_
#define _I915_ACTIVE_TYPES_H_
+#include <linux/llist.h>
#include <linux/rbtree.h>
#include <linux/rcupdate.h>
@@ -31,6 +32,8 @@ struct i915_active {
unsigned int count;
void (*retire)(struct i915_active *ref);
+
+ struct llist_head barriers;
};
#endif /* _I915_ACTIVE_TYPES_H_ */
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 90923fa6603d..ad944e12d8be 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1153,10 +1153,6 @@ void i915_gem_sanitize(struct drm_i915_private *i915)
intel_uncore_forcewake_put(&i915->uncore, FORCEWAKE_ALL);
intel_runtime_pm_put(i915, wakeref);
-
- mutex_lock(&i915->drm.struct_mutex);
- i915_gem_contexts_lost(i915);
- mutex_unlock(&i915->drm.struct_mutex);
}
void i915_gem_init_swizzling(struct drm_i915_private *dev_priv)
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 4864fd6116f4..ae4c96ba02a9 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -213,18 +213,6 @@ static void __retire_engine_request(struct intel_engine_cs *engine,
spin_unlock(&rq->lock);
local_irq_enable();
-
- /*
- * The backing object for the context is done after switching to the
- * *next* context. Therefore we cannot retire the previous context until
- * the next context has already started running. However, since we
- * cannot take the required locks at i915_request_submit() we
- * defer the unpinning of the active context to now, retirement of
- * the subsequent request.
- */
- if (engine->last_retired_context)
- intel_context_unpin(engine->last_retired_context);
- engine->last_retired_context = rq->hw_context;
}
static void __retire_engine_upto(struct intel_engine_cs *engine,
@@ -764,9 +752,6 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp)
rq->infix = rq->ring->emit; /* end of header; start of user payload */
- /* Keep a second pin for the dual retirement along engine and ring */
- __intel_context_pin(ce);
-
intel_context_mark_active(ce);
return rq;
diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
index b7f3fbb4ae89..a96d0c012d46 100644
--- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
+++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
@@ -56,7 +56,6 @@ static void mock_device_release(struct drm_device *dev)
mutex_lock(&i915->drm.struct_mutex);
mock_device_flush(i915);
- i915_gem_contexts_lost(i915);
mutex_unlock(&i915->drm.struct_mutex);
flush_work(&i915->gem.idle_work);
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 36/40] drm/i915: Stop retiring along engine
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (33 preceding siblings ...)
2019-05-08 8:06 ` [PATCH 35/40] drm/i915: Keep contexts pinned until after the next kernel context switch Chris Wilson
@ 2019-05-08 8:07 ` Chris Wilson
2019-05-08 8:07 ` [PATCH 37/40] drm/i915: Replace engine->timeline with a plain list Chris Wilson
` (10 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:07 UTC (permalink / raw)
To: intel-gfx
We no longer track the execution order along the engine and so no longer
need to enforce ordering of retire along the engine.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_request.c | 128 +++++++++++-----------------
1 file changed, 52 insertions(+), 76 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index ae4c96ba02a9..c6fc0e8c3876 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -183,72 +183,23 @@ static void free_capture_list(struct i915_request *request)
}
}
-static void __retire_engine_request(struct intel_engine_cs *engine,
- struct i915_request *rq)
-{
- GEM_TRACE("%s(%s) fence %llx:%lld, current %d\n",
- __func__, engine->name,
- rq->fence.context, rq->fence.seqno,
- hwsp_seqno(rq));
-
- GEM_BUG_ON(!i915_request_completed(rq));
-
- local_irq_disable();
-
- spin_lock(&engine->timeline.lock);
- GEM_BUG_ON(!list_is_first(&rq->link, &engine->timeline.requests));
- list_del_init(&rq->link);
- spin_unlock(&engine->timeline.lock);
-
- spin_lock(&rq->lock);
- i915_request_mark_complete(rq);
- if (!i915_request_signaled(rq))
- dma_fence_signal_locked(&rq->fence);
- if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &rq->fence.flags))
- i915_request_cancel_breadcrumb(rq);
- if (rq->waitboost) {
- GEM_BUG_ON(!atomic_read(&rq->i915->gt_pm.rps.num_waiters));
- atomic_dec(&rq->i915->gt_pm.rps.num_waiters);
- }
- spin_unlock(&rq->lock);
-
- local_irq_enable();
-}
-
-static void __retire_engine_upto(struct intel_engine_cs *engine,
- struct i915_request *rq)
-{
- struct i915_request *tmp;
-
- if (list_empty(&rq->link))
- return;
-
- do {
- tmp = list_first_entry(&engine->timeline.requests,
- typeof(*tmp), link);
-
- GEM_BUG_ON(tmp->engine != engine);
- __retire_engine_request(engine, tmp);
- } while (tmp != rq);
-}
-
-static void i915_request_retire(struct i915_request *request)
+static bool i915_request_retire(struct i915_request *rq)
{
struct i915_active_request *active, *next;
- GEM_TRACE("%s fence %llx:%lld, current %d\n",
- request->engine->name,
- request->fence.context, request->fence.seqno,
- hwsp_seqno(request));
+ lockdep_assert_held(&rq->i915->drm.struct_mutex);
+ if (!i915_request_completed(rq))
+ return false;
- lockdep_assert_held(&request->i915->drm.struct_mutex);
- GEM_BUG_ON(!i915_sw_fence_signaled(&request->submit));
- GEM_BUG_ON(!i915_request_completed(request));
+ GEM_TRACE("%s fence %llx:%lld, current %d\n",
+ rq->engine->name,
+ rq->fence.context, rq->fence.seqno,
+ hwsp_seqno(rq));
- trace_i915_request_retire(request);
+ GEM_BUG_ON(!i915_sw_fence_signaled(&rq->submit));
+ trace_i915_request_retire(rq);
- advance_ring(request);
- free_capture_list(request);
+ advance_ring(rq);
/*
* Walk through the active list, calling retire on each. This allows
@@ -260,7 +211,7 @@ static void i915_request_retire(struct i915_request *request)
* pass along the auxiliary information (to avoid dereferencing
* the node after the callback).
*/
- list_for_each_entry_safe(active, next, &request->active_list, link) {
+ list_for_each_entry_safe(active, next, &rq->active_list, link) {
/*
* In microbenchmarks or focusing upon time inside the kernel,
* we may spend an inordinate amount of time simply handling
@@ -276,18 +227,39 @@ static void i915_request_retire(struct i915_request *request)
INIT_LIST_HEAD(&active->link);
RCU_INIT_POINTER(active->request, NULL);
- active->retire(active, request);
+ active->retire(active, rq);
+ }
+
+ local_irq_disable();
+
+ spin_lock(&rq->engine->timeline.lock);
+ list_del(&rq->link);
+ spin_unlock(&rq->engine->timeline.lock);
+
+ spin_lock(&rq->lock);
+ i915_request_mark_complete(rq);
+ if (!i915_request_signaled(rq))
+ dma_fence_signal_locked(&rq->fence);
+ if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &rq->fence.flags))
+ i915_request_cancel_breadcrumb(rq);
+ if (rq->waitboost) {
+ GEM_BUG_ON(!atomic_read(&rq->i915->gt_pm.rps.num_waiters));
+ atomic_dec(&rq->i915->gt_pm.rps.num_waiters);
}
+ spin_unlock(&rq->lock);
+
+ local_irq_enable();
- i915_request_remove_from_client(request);
+ intel_context_exit(rq->hw_context);
+ intel_context_unpin(rq->hw_context);
- __retire_engine_upto(request->engine, request);
+ i915_request_remove_from_client(rq);
- intel_context_exit(request->hw_context);
- intel_context_unpin(request->hw_context);
+ free_capture_list(rq);
+ i915_sched_node_fini(&rq->sched);
+ i915_request_put(rq);
- i915_sched_node_fini(&request->sched);
- i915_request_put(request);
+ return true;
}
void i915_request_retire_upto(struct i915_request *rq)
@@ -309,9 +281,7 @@ void i915_request_retire_upto(struct i915_request *rq)
do {
tmp = list_first_entry(&ring->request_list,
typeof(*tmp), ring_link);
-
- i915_request_retire(tmp);
- } while (tmp != rq);
+ } while (i915_request_retire(tmp) && tmp != rq);
}
static void irq_execute_cb(struct irq_work *wrk)
@@ -605,12 +575,9 @@ static void ring_retire_requests(struct intel_ring *ring)
{
struct i915_request *rq, *rn;
- list_for_each_entry_safe(rq, rn, &ring->request_list, ring_link) {
- if (!i915_request_completed(rq))
+ list_for_each_entry_safe(rq, rn, &ring->request_list, ring_link)
+ if (!i915_request_retire(rq))
break;
-
- i915_request_retire(rq);
- }
}
static noinline struct i915_request *
@@ -625,6 +592,15 @@ request_alloc_slow(struct intel_context *ce, gfp_t gfp)
if (!gfpflags_allow_blocking(gfp))
goto out;
+ /* Move our oldest request to the slab-cache (if not in use!) */
+ rq = list_first_entry(&ring->request_list, typeof(*rq), ring_link);
+ i915_request_retire(rq);
+
+ rq = kmem_cache_alloc(global.slab_requests,
+ gfp | __GFP_RETRY_MAYFAIL | __GFP_NOWARN);
+ if (rq)
+ return rq;
+
/* Ratelimit ourselves to prevent oom from malicious clients */
rq = list_last_entry(&ring->request_list, typeof(*rq), ring_link);
cond_synchronize_rcu(rq->rcustate);
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 37/40] drm/i915: Replace engine->timeline with a plain list
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (34 preceding siblings ...)
2019-05-08 8:07 ` [PATCH 36/40] drm/i915: Stop retiring along engine Chris Wilson
@ 2019-05-08 8:07 ` Chris Wilson
2019-05-08 8:07 ` [PATCH 38/40] drm/i915: Flush the execution-callbacks on retiring Chris Wilson
` (9 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:07 UTC (permalink / raw)
To: intel-gfx
To continue the onslaught of removing the assumption of a global
execution ordering, another casualty is the engine->timeline. Without an
actual timeline to track, it is overkill and we can replace it with a
much less grand plain list. We still need a list of requests inflight,
for the simple purpose of finding inflight requests (for retiring,
resetting, preemption etc).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/gt/intel_engine.h | 6 ++
drivers/gpu/drm/i915/gt/intel_engine_cs.c | 62 ++++++------
drivers/gpu/drm/i915/gt/intel_engine_types.h | 6 +-
drivers/gpu/drm/i915/gt/intel_lrc.c | 95 ++++++++++---------
drivers/gpu/drm/i915/gt/intel_reset.c | 10 +-
drivers/gpu/drm/i915/gt/intel_ringbuffer.c | 15 ++-
drivers/gpu/drm/i915/gt/mock_engine.c | 18 ++--
drivers/gpu/drm/i915/i915_gpu_error.c | 5 +-
drivers/gpu/drm/i915/i915_request.c | 43 +++------
drivers/gpu/drm/i915/i915_request.h | 2 +-
drivers/gpu/drm/i915/i915_scheduler.c | 36 +++----
drivers/gpu/drm/i915/i915_timeline.c | 1 -
drivers/gpu/drm/i915/i915_timeline.h | 19 ----
drivers/gpu/drm/i915/i915_timeline_types.h | 4 -
drivers/gpu/drm/i915/intel_guc_submission.c | 16 ++--
.../gpu/drm/i915/selftests/mock_timeline.c | 1 -
16 files changed, 152 insertions(+), 187 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h
index 1ae2934b7ecc..b8f5c4e1d39e 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine.h
@@ -544,4 +544,10 @@ static inline bool inject_preempt_hang(struct intel_engine_execlists *execlists)
#endif
+void intel_engine_init_active(struct intel_engine_cs *engine,
+ unsigned int subclass);
+#define ENGINE_PHYSICAL 0
+#define ENGINE_MOCK 1
+#define ENGINE_VIRTUAL 2
+
#endif /* _INTEL_RINGBUFFER_H_ */
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
index 794aa8b11742..f2e28510433b 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
@@ -567,14 +567,7 @@ static int intel_engine_setup_common(struct intel_engine_cs *engine)
if (err)
return err;
- err = i915_timeline_init(engine->i915,
- &engine->timeline,
- engine->status_page.vma);
- if (err)
- goto err_hwsp;
-
- i915_timeline_set_subclass(&engine->timeline, TIMELINE_ENGINE);
-
+ intel_engine_init_active(engine, ENGINE_PHYSICAL);
intel_engine_init_breadcrumbs(engine);
intel_engine_init_execlists(engine);
intel_engine_init_hangcheck(engine);
@@ -587,10 +580,6 @@ static int intel_engine_setup_common(struct intel_engine_cs *engine)
intel_sseu_from_device_info(&RUNTIME_INFO(engine->i915)->sseu);
return 0;
-
-err_hwsp:
- cleanup_status_page(engine);
- return err;
}
/**
@@ -747,6 +736,27 @@ static int pin_context(struct i915_gem_context *ctx,
return 0;
}
+void
+intel_engine_init_active(struct intel_engine_cs *engine, unsigned int subclass)
+{
+ INIT_LIST_HEAD(&engine->active.requests);
+
+ spin_lock_init(&engine->active.lock);
+ lockdep_set_subclass(&engine->active.lock, subclass);
+
+ /*
+ * Due to an interesting quirk in lockdep's internal debug tracking,
+ * after setting a subclass we must ensure the lock is used. Otherwise,
+ * nr_unused_locks is incremented once too often.
+ */
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+ local_irq_disable();
+ lock_map_acquire(&engine->active.lock.dep_map);
+ lock_map_release(&engine->active.lock.dep_map);
+ local_irq_enable();
+#endif
+}
+
/**
* intel_engines_init_common - initialize cengine state which might require hw access
* @engine: Engine to initialize.
@@ -810,6 +820,8 @@ int intel_engine_init_common(struct intel_engine_cs *engine)
*/
void intel_engine_cleanup_common(struct intel_engine_cs *engine)
{
+ GEM_BUG_ON(!list_empty(&engine->active.requests));
+
cleanup_status_page(engine);
intel_engine_fini_breadcrumbs(engine);
@@ -824,8 +836,6 @@ void intel_engine_cleanup_common(struct intel_engine_cs *engine)
intel_context_unpin(engine->kernel_context);
GEM_BUG_ON(!llist_empty(&engine->barrier_tasks));
- i915_timeline_fini(&engine->timeline);
-
intel_wa_list_free(&engine->ctx_wa_list);
intel_wa_list_free(&engine->wa_list);
intel_wa_list_free(&engine->whitelist);
@@ -1431,16 +1441,6 @@ void intel_engine_dump(struct intel_engine_cs *engine,
drm_printf(m, "\tRequests:\n");
- rq = list_first_entry(&engine->timeline.requests,
- struct i915_request, link);
- if (&rq->link != &engine->timeline.requests)
- print_request(m, rq, "\t\tfirst ");
-
- rq = list_last_entry(&engine->timeline.requests,
- struct i915_request, link);
- if (&rq->link != &engine->timeline.requests)
- print_request(m, rq, "\t\tlast ");
-
rq = intel_engine_find_active_request(engine);
if (rq) {
print_request(m, rq, "\t\tactive ");
@@ -1521,7 +1521,7 @@ int intel_enable_engine_stats(struct intel_engine_cs *engine)
if (!intel_engine_supports_stats(engine))
return -ENODEV;
- spin_lock_irqsave(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
write_seqlock(&engine->stats.lock);
if (unlikely(engine->stats.enabled == ~0)) {
@@ -1547,7 +1547,7 @@ int intel_enable_engine_stats(struct intel_engine_cs *engine)
unlock:
write_sequnlock(&engine->stats.lock);
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
return err;
}
@@ -1632,22 +1632,22 @@ intel_engine_find_active_request(struct intel_engine_cs *engine)
* At all other times, we must assume the GPU is still running, but
* we only care about the snapshot of this moment.
*/
- spin_lock_irqsave(&engine->timeline.lock, flags);
- list_for_each_entry(request, &engine->timeline.requests, link) {
+ spin_lock_irqsave(&engine->active.lock, flags);
+ list_for_each_entry(request, &engine->active.requests, sched.link) {
if (i915_request_completed(request))
continue;
if (!i915_request_started(request))
- break;
+ continue;
/* More than one preemptible request may match! */
if (!match_ring(request))
- break;
+ continue;
active = request;
break;
}
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
return active;
}
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
index d3d5a505f621..e00443f53b59 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
@@ -286,7 +286,11 @@ struct intel_engine_cs {
struct intel_ring *buffer;
- struct i915_timeline timeline;
+ struct {
+ spinlock_t lock;
+ struct list_head requests;
+ } active;
+
struct llist_head barrier_tasks;
struct intel_context *kernel_context; /* pinned */
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index bc135a9e6d42..d350ceadffc5 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -311,8 +311,8 @@ static inline bool need_preempt(const struct intel_engine_cs *engine,
* Check against the first request in ELSP[1], it will, thanks to the
* power of PI, be the highest priority of that context.
*/
- if (!list_is_last(&rq->link, &engine->timeline.requests) &&
- rq_prio(list_next_entry(rq, link)) > last_prio)
+ if (!list_is_last(&rq->sched.link, &engine->active.requests) &&
+ rq_prio(list_next_entry(rq, sched.link)) > last_prio)
return true;
if (rb) {
@@ -447,11 +447,11 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
struct list_head *uninitialized_var(pl);
int prio = I915_PRIORITY_INVALID | boost;
- lockdep_assert_held(&engine->timeline.lock);
+ lockdep_assert_held(&engine->active.lock);
list_for_each_entry_safe_reverse(rq, rn,
- &engine->timeline.requests,
- link) {
+ &engine->active.requests,
+ sched.link) {
struct intel_engine_cs *owner;
if (i915_request_completed(rq))
@@ -478,7 +478,7 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
}
GEM_BUG_ON(RB_EMPTY_ROOT(&engine->execlists.queue.rb_root));
- list_add(&rq->sched.link, pl);
+ list_move(&rq->sched.link, pl);
active = rq;
} else {
if (__i915_request_has_started(rq))
@@ -975,11 +975,11 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
rb_entry(rb, typeof(*ve), nodes[engine->id].rb);
struct i915_request *rq;
- spin_lock(&ve->base.timeline.lock);
+ spin_lock(&ve->base.active.lock);
rq = ve->request;
if (unlikely(!rq)) { /* lost the race to a sibling */
- spin_unlock(&ve->base.timeline.lock);
+ spin_unlock(&ve->base.active.lock);
rb_erase_cached(rb, &execlists->virtual);
RB_CLEAR_NODE(rb);
rb = rb_first_cached(&execlists->virtual);
@@ -992,13 +992,13 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
if (rq_prio(rq) >= queue_prio(execlists)) {
if (!virtual_matches(ve, rq, engine)) {
- spin_unlock(&ve->base.timeline.lock);
+ spin_unlock(&ve->base.active.lock);
rb = rb_next(rb);
continue;
}
if (last && !can_merge_rq(last, rq)) {
- spin_unlock(&ve->base.timeline.lock);
+ spin_unlock(&ve->base.active.lock);
return; /* leave this rq for another engine */
}
@@ -1053,7 +1053,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
last = rq;
}
- spin_unlock(&ve->base.timeline.lock);
+ spin_unlock(&ve->base.active.lock);
break;
}
@@ -1110,8 +1110,6 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
GEM_BUG_ON(port_isset(port));
}
- list_del_init(&rq->sched.link);
-
__i915_request_submit(rq);
trace_i915_request_in(rq, port_index(port, execlists));
@@ -1212,7 +1210,7 @@ static void process_csb(struct intel_engine_cs *engine)
const u8 num_entries = execlists->csb_size;
u8 head, tail;
- lockdep_assert_held(&engine->timeline.lock);
+ lockdep_assert_held(&engine->active.lock);
/*
* Note that csb_write, csb_status may be either in HWSP or mmio.
@@ -1372,7 +1370,7 @@ static void process_csb(struct intel_engine_cs *engine)
static void __execlists_submission_tasklet(struct intel_engine_cs *const engine)
{
- lockdep_assert_held(&engine->timeline.lock);
+ lockdep_assert_held(&engine->active.lock);
process_csb(engine);
if (!execlists_is_active(&engine->execlists, EXECLISTS_ACTIVE_PREEMPT))
@@ -1393,15 +1391,16 @@ static void execlists_submission_tasklet(unsigned long data)
!!intel_wakeref_active(&engine->wakeref),
engine->execlists.active);
- spin_lock_irqsave(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
__execlists_submission_tasklet(engine);
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
}
static void queue_request(struct intel_engine_cs *engine,
struct i915_sched_node *node,
int prio)
{
+ GEM_BUG_ON(!list_empty(&node->link));
list_add_tail(&node->link, i915_sched_lookup_priolist(engine, prio));
}
@@ -1432,7 +1431,7 @@ static void execlists_submit_request(struct i915_request *request)
unsigned long flags;
/* Will be called from irq-context when using foreign fences. */
- spin_lock_irqsave(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
queue_request(engine, &request->sched, rq_prio(request));
@@ -1441,7 +1440,7 @@ static void execlists_submit_request(struct i915_request *request)
submit_queue(engine, rq_prio(request));
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
}
static void __execlists_context_fini(struct intel_context *ce)
@@ -2092,8 +2091,8 @@ static void execlists_reset_prepare(struct intel_engine_cs *engine)
intel_engine_stop_cs(engine);
/* And flush any current direct submission. */
- spin_lock_irqsave(&engine->timeline.lock, flags);
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
}
static bool lrc_regs_ok(const struct i915_request *rq)
@@ -2136,11 +2135,11 @@ static void reset_csb_pointers(struct intel_engine_execlists *execlists)
static struct i915_request *active_request(struct i915_request *rq)
{
- const struct list_head * const list = &rq->engine->timeline.requests;
+ const struct list_head * const list = &rq->engine->active.requests;
const struct intel_context * const context = rq->hw_context;
struct i915_request *active = NULL;
- list_for_each_entry_from_reverse(rq, list, link) {
+ list_for_each_entry_from_reverse(rq, list, sched.link) {
if (i915_request_completed(rq))
break;
@@ -2257,11 +2256,11 @@ static void execlists_reset(struct intel_engine_cs *engine, bool stalled)
GEM_TRACE("%s\n", engine->name);
- spin_lock_irqsave(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
__execlists_reset(engine, stalled);
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
}
static void nop_submission_tasklet(unsigned long data)
@@ -2292,12 +2291,12 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine)
* submission's irq state, we also wish to remind ourselves that
* it is irq state.)
*/
- spin_lock_irqsave(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
__execlists_reset(engine, true);
/* Mark all executing requests as skipped. */
- list_for_each_entry(rq, &engine->timeline.requests, link) {
+ list_for_each_entry(rq, &engine->active.requests, sched.link) {
if (!i915_request_signaled(rq))
dma_fence_set_error(&rq->fence, -EIO);
@@ -2328,7 +2327,7 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine)
rb_erase_cached(rb, &execlists->virtual);
RB_CLEAR_NODE(rb);
- spin_lock(&ve->base.timeline.lock);
+ spin_lock(&ve->base.active.lock);
if (ve->request) {
ve->request->engine = engine;
__i915_request_submit(ve->request);
@@ -2337,7 +2336,7 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine)
ve->base.execlists.queue_priority_hint = INT_MIN;
ve->request = NULL;
}
- spin_unlock(&ve->base.timeline.lock);
+ spin_unlock(&ve->base.active.lock);
}
/* Remaining _unready_ requests will be nop'ed when submitted */
@@ -2349,7 +2348,7 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine)
GEM_BUG_ON(__tasklet_is_enabled(&execlists->tasklet));
execlists->tasklet.func = nop_submission_tasklet;
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
}
static void execlists_reset_finish(struct intel_engine_cs *engine)
@@ -3050,12 +3049,18 @@ static int execlists_context_deferred_alloc(struct intel_context *ce,
return ret;
}
+static struct list_head *virtual_queue(struct virtual_engine *ve)
+{
+ return &ve->base.execlists.default_priolist.requests[0];
+}
+
static void virtual_context_destroy(struct kref *kref)
{
struct virtual_engine *ve =
container_of(kref, typeof(*ve), context.ref);
unsigned int n;
+ GEM_BUG_ON(!list_empty(virtual_queue(ve)));
GEM_BUG_ON(ve->request);
GEM_BUG_ON(ve->context.inflight);
@@ -3066,13 +3071,13 @@ static void virtual_context_destroy(struct kref *kref)
if (RB_EMPTY_NODE(node))
continue;
- spin_lock_irq(&sibling->timeline.lock);
+ spin_lock_irq(&sibling->active.lock);
/* Detachment is lazily performed in the execlists tasklet */
if (!RB_EMPTY_NODE(node))
rb_erase_cached(node, &sibling->execlists.virtual);
- spin_unlock_irq(&sibling->timeline.lock);
+ spin_unlock_irq(&sibling->active.lock);
}
GEM_BUG_ON(__tasklet_is_scheduled(&ve->base.execlists.tasklet));
@@ -3080,8 +3085,6 @@ static void virtual_context_destroy(struct kref *kref)
__execlists_context_fini(&ve->context);
kfree(ve->bonds);
-
- i915_timeline_fini(&ve->base.timeline);
kfree(ve);
}
@@ -3201,16 +3204,16 @@ static void virtual_submission_tasklet(unsigned long data)
if (unlikely(!(mask & sibling->mask))) {
if (!RB_EMPTY_NODE(&node->rb)) {
- spin_lock(&sibling->timeline.lock);
+ spin_lock(&sibling->active.lock);
rb_erase_cached(&node->rb,
&sibling->execlists.virtual);
RB_CLEAR_NODE(&node->rb);
- spin_unlock(&sibling->timeline.lock);
+ spin_unlock(&sibling->active.lock);
}
continue;
}
- spin_lock(&sibling->timeline.lock);
+ spin_lock(&sibling->active.lock);
if (!RB_EMPTY_NODE(&node->rb)) {
/*
@@ -3254,7 +3257,7 @@ static void virtual_submission_tasklet(unsigned long data)
tasklet_hi_schedule(&sibling->execlists.tasklet);
}
- spin_unlock(&sibling->timeline.lock);
+ spin_unlock(&sibling->active.lock);
}
local_irq_enable();
}
@@ -3271,9 +3274,13 @@ static void virtual_submit_request(struct i915_request *rq)
GEM_BUG_ON(ve->base.submit_request != virtual_submit_request);
GEM_BUG_ON(ve->request);
+ GEM_BUG_ON(!list_empty(virtual_queue(ve)));
+
ve->base.execlists.queue_priority_hint = rq_prio(rq);
WRITE_ONCE(ve->request, rq);
+ list_move_tail(&rq->sched.link, virtual_queue(ve));
+
tasklet_schedule(&ve->base.execlists.tasklet);
}
@@ -3337,10 +3344,7 @@ intel_execlists_create_virtual(struct i915_gem_context *ctx,
snprintf(ve->base.name, sizeof(ve->base.name), "virtual");
- err = i915_timeline_init(ctx->i915, &ve->base.timeline, NULL);
- if (err)
- goto err_put;
- i915_timeline_set_subclass(&ve->base.timeline, TIMELINE_VIRTUAL);
+ intel_engine_init_active(&ve->base, ENGINE_VIRTUAL);
intel_engine_init_execlists(&ve->base);
@@ -3351,6 +3355,7 @@ intel_execlists_create_virtual(struct i915_gem_context *ctx,
ve->base.submit_request = virtual_submit_request;
ve->base.bond_execute = virtual_bond_execute;
+ INIT_LIST_HEAD(virtual_queue(ve));
ve->base.execlists.queue_priority_hint = INT_MIN;
tasklet_init(&ve->base.execlists.tasklet,
virtual_submission_tasklet,
@@ -3505,11 +3510,11 @@ void intel_execlists_show_requests(struct intel_engine_cs *engine,
unsigned int count;
struct rb_node *rb;
- spin_lock_irqsave(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
last = NULL;
count = 0;
- list_for_each_entry(rq, &engine->timeline.requests, link) {
+ list_for_each_entry(rq, &engine->active.requests, sched.link) {
if (count++ < max - 1)
show_request(m, rq, "\t\tE ");
else
@@ -3572,7 +3577,7 @@ void intel_execlists_show_requests(struct intel_engine_cs *engine,
show_request(m, last, "\t\tV ");
}
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
}
void intel_lr_context_reset(struct intel_engine_cs *engine,
diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c
index aa605d79fa94..2b0b08096ea5 100644
--- a/drivers/gpu/drm/i915/gt/intel_reset.c
+++ b/drivers/gpu/drm/i915/gt/intel_reset.c
@@ -49,12 +49,12 @@ static void engine_skip_context(struct i915_request *rq)
struct intel_engine_cs *engine = rq->engine;
struct i915_gem_context *hung_ctx = rq->gem_context;
- lockdep_assert_held(&engine->timeline.lock);
+ lockdep_assert_held(&engine->active.lock);
if (!i915_request_is_active(rq))
return;
- list_for_each_entry_continue(rq, &engine->timeline.requests, link)
+ list_for_each_entry_continue(rq, &engine->active.requests, sched.link)
if (rq->gem_context == hung_ctx)
i915_request_skip(rq, -EIO);
}
@@ -130,7 +130,7 @@ void i915_reset_request(struct i915_request *rq, bool guilty)
rq->fence.seqno,
yesno(guilty));
- lockdep_assert_held(&rq->engine->timeline.lock);
+ lockdep_assert_held(&rq->engine->active.lock);
GEM_BUG_ON(i915_request_completed(rq));
if (guilty) {
@@ -785,10 +785,10 @@ static void nop_submit_request(struct i915_request *request)
engine->name, request->fence.context, request->fence.seqno);
dma_fence_set_error(&request->fence, -EIO);
- spin_lock_irqsave(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
__i915_request_submit(request);
i915_request_mark_complete(request);
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
intel_engine_queue_breadcrumbs(engine);
}
diff --git a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
index fc596adc637a..93b0893c736b 100644
--- a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
@@ -730,14 +730,13 @@ static void reset_prepare(struct intel_engine_cs *engine)
static void reset_ring(struct intel_engine_cs *engine, bool stalled)
{
- struct i915_timeline *tl = &engine->timeline;
struct i915_request *pos, *rq;
unsigned long flags;
u32 head;
rq = NULL;
- spin_lock_irqsave(&tl->lock, flags);
- list_for_each_entry(pos, &tl->requests, link) {
+ spin_lock_irqsave(&engine->active.lock, flags);
+ list_for_each_entry(pos, &engine->active.requests, sched.link) {
if (!i915_request_completed(pos)) {
rq = pos;
break;
@@ -791,7 +790,7 @@ static void reset_ring(struct intel_engine_cs *engine, bool stalled)
}
engine->buffer->head = intel_ring_wrap(engine->buffer, head);
- spin_unlock_irqrestore(&tl->lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
}
static void reset_finish(struct intel_engine_cs *engine)
@@ -877,10 +876,10 @@ static void cancel_requests(struct intel_engine_cs *engine)
struct i915_request *request;
unsigned long flags;
- spin_lock_irqsave(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
/* Mark all submitted requests as skipped. */
- list_for_each_entry(request, &engine->timeline.requests, link) {
+ list_for_each_entry(request, &engine->active.requests, sched.link) {
if (!i915_request_signaled(request))
dma_fence_set_error(&request->fence, -EIO);
@@ -889,7 +888,7 @@ static void cancel_requests(struct intel_engine_cs *engine)
/* Remaining _unready_ requests will be nop'ed when submitted */
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
}
static void i9xx_submit_request(struct i915_request *request)
@@ -1267,8 +1266,6 @@ intel_engine_create_ring(struct intel_engine_cs *engine,
GEM_BUG_ON(!is_power_of_2(size));
GEM_BUG_ON(RING_CTL_SIZE(size) & ~RING_NR_PAGES);
- GEM_BUG_ON(timeline == &engine->timeline);
- lockdep_assert_held(&engine->i915->drm.struct_mutex);
ring = kzalloc(sizeof(*ring), GFP_KERNEL);
if (!ring)
diff --git a/drivers/gpu/drm/i915/gt/mock_engine.c b/drivers/gpu/drm/i915/gt/mock_engine.c
index b7675ef18523..00c666d3e652 100644
--- a/drivers/gpu/drm/i915/gt/mock_engine.c
+++ b/drivers/gpu/drm/i915/gt/mock_engine.c
@@ -229,17 +229,17 @@ static void mock_cancel_requests(struct intel_engine_cs *engine)
struct i915_request *request;
unsigned long flags;
- spin_lock_irqsave(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
/* Mark all submitted requests as skipped. */
- list_for_each_entry(request, &engine->timeline.requests, sched.link) {
+ list_for_each_entry(request, &engine->active.requests, sched.link) {
if (!i915_request_signaled(request))
dma_fence_set_error(&request->fence, -EIO);
i915_request_mark_complete(request);
}
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
}
struct intel_engine_cs *mock_engine(struct drm_i915_private *i915,
@@ -285,28 +285,23 @@ int mock_engine_init(struct intel_engine_cs *engine)
struct drm_i915_private *i915 = engine->i915;
int err;
+ intel_engine_init_active(engine, ENGINE_MOCK);
intel_engine_init_breadcrumbs(engine);
intel_engine_init_execlists(engine);
intel_engine_init__pm(engine);
- if (i915_timeline_init(i915, &engine->timeline, NULL))
- goto err_breadcrumbs;
- i915_timeline_set_subclass(&engine->timeline, TIMELINE_ENGINE);
-
engine->kernel_context =
i915_gem_context_get_engine(i915->kernel_context, engine->id);
if (IS_ERR(engine->kernel_context))
- goto err_timeline;
+ goto err_breadcrumbs;
err = intel_context_pin(engine->kernel_context);
intel_context_put(engine->kernel_context);
if (err)
- goto err_timeline;
+ goto err_breadcrumbs;
return 0;
-err_timeline:
- i915_timeline_fini(&engine->timeline);
err_breadcrumbs:
intel_engine_fini_breadcrumbs(engine);
return -ENOMEM;
@@ -340,7 +335,6 @@ void mock_engine_free(struct intel_engine_cs *engine)
intel_context_unpin(engine->kernel_context);
intel_engine_fini_breadcrumbs(engine);
- i915_timeline_fini(&engine->timeline);
kfree(engine);
}
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index 707811256501..4cc900e432e2 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -1269,7 +1269,7 @@ static void engine_record_requests(struct intel_engine_cs *engine,
count = 0;
request = first;
- list_for_each_entry_from(request, &engine->timeline.requests, link)
+ list_for_each_entry_from(request, &engine->active.requests, sched.link)
count++;
if (!count)
return;
@@ -1282,7 +1282,8 @@ static void engine_record_requests(struct intel_engine_cs *engine,
count = 0;
request = first;
- list_for_each_entry_from(request, &engine->timeline.requests, link) {
+ list_for_each_entry_from(request,
+ &engine->active.requests, sched.link) {
if (count >= ee->num_requests) {
/*
* If the ring request list was changed in
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index c6fc0e8c3876..03f15ac63ec9 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -232,9 +232,9 @@ static bool i915_request_retire(struct i915_request *rq)
local_irq_disable();
- spin_lock(&rq->engine->timeline.lock);
- list_del(&rq->link);
- spin_unlock(&rq->engine->timeline.lock);
+ spin_lock(&rq->engine->active.lock);
+ list_del(&rq->sched.link);
+ spin_unlock(&rq->engine->active.lock);
spin_lock(&rq->lock);
i915_request_mark_complete(rq);
@@ -254,6 +254,7 @@ static bool i915_request_retire(struct i915_request *rq)
intel_context_unpin(rq->hw_context);
i915_request_remove_from_client(rq);
+ list_del(&rq->link);
free_capture_list(rq);
i915_sched_node_fini(&rq->sched);
@@ -373,28 +374,17 @@ __i915_request_await_execution(struct i915_request *rq,
return 0;
}
-static void move_to_timeline(struct i915_request *request,
- struct i915_timeline *timeline)
-{
- GEM_BUG_ON(request->timeline == &request->engine->timeline);
- lockdep_assert_held(&request->engine->timeline.lock);
-
- spin_lock(&request->timeline->lock);
- list_move_tail(&request->link, &timeline->requests);
- spin_unlock(&request->timeline->lock);
-}
-
void __i915_request_submit(struct i915_request *request)
{
struct intel_engine_cs *engine = request->engine;
- GEM_TRACE("%s fence %llx:%lld -> current %d\n",
+ GEM_TRACE("%s fence %llx:%lld, current %d\n",
engine->name,
request->fence.context, request->fence.seqno,
hwsp_seqno(request));
GEM_BUG_ON(!irqs_disabled());
- lockdep_assert_held(&engine->timeline.lock);
+ lockdep_assert_held(&engine->active.lock);
if (i915_gem_context_is_banned(request->gem_context))
i915_request_skip(request, -EIO);
@@ -422,6 +412,8 @@ void __i915_request_submit(struct i915_request *request)
/* We may be recursing from the signal callback of another i915 fence */
spin_lock_nested(&request->lock, SINGLE_DEPTH_NESTING);
+ list_move_tail(&request->sched.link, &engine->active.requests);
+
GEM_BUG_ON(test_bit(I915_FENCE_FLAG_ACTIVE, &request->fence.flags));
set_bit(I915_FENCE_FLAG_ACTIVE, &request->fence.flags);
@@ -437,9 +429,6 @@ void __i915_request_submit(struct i915_request *request)
engine->emit_fini_breadcrumb(request,
request->ring->vaddr + request->postfix);
- /* Transfer from per-context onto the global per-engine timeline */
- move_to_timeline(request, &engine->timeline);
-
engine->serial++;
trace_i915_request_execute(request);
@@ -451,11 +440,11 @@ void i915_request_submit(struct i915_request *request)
unsigned long flags;
/* Will be called from irq-context when using foreign fences. */
- spin_lock_irqsave(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
__i915_request_submit(request);
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
}
void __i915_request_unsubmit(struct i915_request *request)
@@ -468,7 +457,7 @@ void __i915_request_unsubmit(struct i915_request *request)
hwsp_seqno(request));
GEM_BUG_ON(!irqs_disabled());
- lockdep_assert_held(&engine->timeline.lock);
+ lockdep_assert_held(&engine->active.lock);
/*
* Only unwind in reverse order, required so that the per-context list
@@ -486,9 +475,6 @@ void __i915_request_unsubmit(struct i915_request *request)
spin_unlock(&request->lock);
- /* Transfer back from the global per-engine timeline to per-context */
- move_to_timeline(request, request->timeline);
-
/*
* We don't need to wake_up any waiters on request->execute, they
* will get woken by any other event or us re-adding this request
@@ -504,11 +490,11 @@ void i915_request_unsubmit(struct i915_request *request)
unsigned long flags;
/* Will be called from irq-context when using foreign fences. */
- spin_lock_irqsave(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
__i915_request_unsubmit(request);
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
}
static int __i915_sw_fence_call
@@ -674,7 +660,6 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp)
rq->engine = ce->engine;
rq->ring = ce->ring;
rq->timeline = tl;
- GEM_BUG_ON(rq->timeline == &ce->engine->timeline);
rq->hwsp_seqno = tl->hwsp_seqno;
rq->hwsp_cacheline = tl->hwsp_cacheline;
rq->rcustate = get_state_synchronize_rcu(); /* acts as smp_mb() */
@@ -1135,9 +1120,7 @@ __i915_request_add_to_timeline(struct i915_request *rq)
0);
}
- spin_lock_irq(&timeline->lock);
list_add_tail(&rq->link, &timeline->requests);
- spin_unlock_irq(&timeline->lock);
/*
* Make sure that no request gazumped us - if it was allocated after
diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
index c9f7d07991c8..edbbdfec24ab 100644
--- a/drivers/gpu/drm/i915/i915_request.h
+++ b/drivers/gpu/drm/i915/i915_request.h
@@ -217,7 +217,7 @@ struct i915_request {
bool waitboost;
- /** engine->request_list entry for this request */
+ /** timeline->request entry for this request */
struct list_head link;
/** ring->request_list entry for this request */
diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c
index 78ceb56d7801..2cba5b9b34bb 100644
--- a/drivers/gpu/drm/i915/i915_scheduler.c
+++ b/drivers/gpu/drm/i915/i915_scheduler.c
@@ -77,7 +77,7 @@ i915_sched_lookup_priolist(struct intel_engine_cs *engine, int prio)
bool first = true;
int idx, i;
- lockdep_assert_held(&engine->timeline.lock);
+ lockdep_assert_held(&engine->active.lock);
assert_priolists(execlists);
/* buckets sorted from highest [in slot 0] to lowest priority */
@@ -162,9 +162,9 @@ sched_lock_engine(const struct i915_sched_node *node,
* check that the rq still belongs to the newly locked engine.
*/
while (locked != (engine = READ_ONCE(rq->engine))) {
- spin_unlock(&locked->timeline.lock);
+ spin_unlock(&locked->active.lock);
memset(cache, 0, sizeof(*cache));
- spin_lock(&engine->timeline.lock);
+ spin_lock(&engine->active.lock);
locked = engine;
}
@@ -278,7 +278,7 @@ static void __i915_schedule(struct i915_sched_node *node,
memset(&cache, 0, sizeof(cache));
engine = node_to_request(node)->engine;
- spin_lock(&engine->timeline.lock);
+ spin_lock(&engine->active.lock);
/* Fifo and depth-first replacement ensure our deps execute before us */
engine = sched_lock_engine(node, engine, &cache);
@@ -287,7 +287,7 @@ static void __i915_schedule(struct i915_sched_node *node,
node = dep->signaler;
engine = sched_lock_engine(node, engine, &cache);
- lockdep_assert_held(&engine->timeline.lock);
+ lockdep_assert_held(&engine->active.lock);
/* Recheck after acquiring the engine->timeline.lock */
if (prio <= node->attr.priority || node_signaled(node))
@@ -296,14 +296,8 @@ static void __i915_schedule(struct i915_sched_node *node,
GEM_BUG_ON(node_to_request(node)->engine != engine);
node->attr.priority = prio;
- if (!list_empty(&node->link)) {
- GEM_BUG_ON(intel_engine_is_virtual(engine));
- if (!cache.priolist)
- cache.priolist =
- i915_sched_lookup_priolist(engine,
- prio);
- list_move_tail(&node->link, cache.priolist);
- } else {
+
+ if (list_empty(&node->link)) {
/*
* If the request is not in the priolist queue because
* it is not yet runnable, then it doesn't contribute
@@ -312,8 +306,16 @@ static void __i915_schedule(struct i915_sched_node *node,
* queue; but in that case we may still need to reorder
* the inflight requests.
*/
- if (!i915_sw_fence_done(&node_to_request(node)->submit))
- continue;
+ continue;
+ }
+
+ if (!intel_engine_is_virtual(engine) &&
+ !i915_request_is_active(node_to_request(node))) {
+ if (!cache.priolist)
+ cache.priolist =
+ i915_sched_lookup_priolist(engine,
+ prio);
+ list_move_tail(&node->link, cache.priolist);
}
if (prio <= engine->execlists.queue_priority_hint)
@@ -325,7 +327,7 @@ static void __i915_schedule(struct i915_sched_node *node,
kick_submission(engine, prio);
}
- spin_unlock(&engine->timeline.lock);
+ spin_unlock(&engine->active.lock);
}
void i915_schedule(struct i915_request *rq, const struct i915_sched_attr *attr)
@@ -439,8 +441,6 @@ void i915_sched_node_fini(struct i915_sched_node *node)
{
struct i915_dependency *dep, *tmp;
- GEM_BUG_ON(!list_empty(&node->link));
-
spin_lock_irq(&schedule_lock);
/*
diff --git a/drivers/gpu/drm/i915/i915_timeline.c b/drivers/gpu/drm/i915/i915_timeline.c
index 000e1a9b6750..c311ce9c6f9d 100644
--- a/drivers/gpu/drm/i915/i915_timeline.c
+++ b/drivers/gpu/drm/i915/i915_timeline.c
@@ -251,7 +251,6 @@ int i915_timeline_init(struct drm_i915_private *i915,
timeline->fence_context = dma_fence_context_alloc(1);
- spin_lock_init(&timeline->lock);
mutex_init(&timeline->mutex);
INIT_ACTIVE_REQUEST(&timeline->last_request);
diff --git a/drivers/gpu/drm/i915/i915_timeline.h b/drivers/gpu/drm/i915/i915_timeline.h
index 27668a1a69a3..36e5e5a65155 100644
--- a/drivers/gpu/drm/i915/i915_timeline.h
+++ b/drivers/gpu/drm/i915/i915_timeline.h
@@ -36,25 +36,6 @@ int i915_timeline_init(struct drm_i915_private *i915,
struct i915_vma *hwsp);
void i915_timeline_fini(struct i915_timeline *tl);
-static inline void
-i915_timeline_set_subclass(struct i915_timeline *timeline,
- unsigned int subclass)
-{
- lockdep_set_subclass(&timeline->lock, subclass);
-
- /*
- * Due to an interesting quirk in lockdep's internal debug tracking,
- * after setting a subclass we must ensure the lock is used. Otherwise,
- * nr_unused_locks is incremented once too often.
- */
-#ifdef CONFIG_DEBUG_LOCK_ALLOC
- local_irq_disable();
- lock_map_acquire(&timeline->lock.dep_map);
- lock_map_release(&timeline->lock.dep_map);
- local_irq_enable();
-#endif
-}
-
struct i915_timeline *
i915_timeline_create(struct drm_i915_private *i915,
struct i915_vma *global_hwsp);
diff --git a/drivers/gpu/drm/i915/i915_timeline_types.h b/drivers/gpu/drm/i915/i915_timeline_types.h
index 1688705f4a2b..fce5cb4f1090 100644
--- a/drivers/gpu/drm/i915/i915_timeline_types.h
+++ b/drivers/gpu/drm/i915/i915_timeline_types.h
@@ -23,10 +23,6 @@ struct i915_timeline {
u64 fence_context;
u32 seqno;
- spinlock_t lock;
-#define TIMELINE_CLIENT 0 /* default subclass */
-#define TIMELINE_ENGINE 1
-#define TIMELINE_VIRTUAL 2
struct mutex mutex; /* protects the flow of requests */
unsigned int pin_count;
diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c
index 389a9131d413..240c8a2aabc2 100644
--- a/drivers/gpu/drm/i915/intel_guc_submission.c
+++ b/drivers/gpu/drm/i915/intel_guc_submission.c
@@ -740,7 +740,7 @@ static bool __guc_dequeue(struct intel_engine_cs *engine)
bool submit = false;
struct rb_node *rb;
- lockdep_assert_held(&engine->timeline.lock);
+ lockdep_assert_held(&engine->active.lock);
if (port_isset(port)) {
if (intel_engine_has_preemption(engine)) {
@@ -822,7 +822,7 @@ static void guc_submission_tasklet(unsigned long data)
struct i915_request *rq;
unsigned long flags;
- spin_lock_irqsave(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
rq = port_request(port);
while (rq && i915_request_completed(rq)) {
@@ -847,7 +847,7 @@ static void guc_submission_tasklet(unsigned long data)
if (!execlists_is_active(execlists, EXECLISTS_ACTIVE_PREEMPT))
guc_dequeue(engine);
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
}
static void guc_reset_prepare(struct intel_engine_cs *engine)
@@ -884,7 +884,7 @@ static void guc_reset(struct intel_engine_cs *engine, bool stalled)
struct i915_request *rq;
unsigned long flags;
- spin_lock_irqsave(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
execlists_cancel_port_requests(execlists);
@@ -900,7 +900,7 @@ static void guc_reset(struct intel_engine_cs *engine, bool stalled)
intel_lr_context_reset(engine, rq->hw_context, rq->head, stalled);
out_unlock:
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
}
static void guc_cancel_requests(struct intel_engine_cs *engine)
@@ -926,13 +926,13 @@ static void guc_cancel_requests(struct intel_engine_cs *engine)
* submission's irq state, we also wish to remind ourselves that
* it is irq state.)
*/
- spin_lock_irqsave(&engine->timeline.lock, flags);
+ spin_lock_irqsave(&engine->active.lock, flags);
/* Cancel the requests on the HW and clear the ELSP tracker. */
execlists_cancel_port_requests(execlists);
/* Mark all executing requests as skipped. */
- list_for_each_entry(rq, &engine->timeline.requests, link) {
+ list_for_each_entry(rq, &engine->active.requests, sched.link) {
if (!i915_request_signaled(rq))
dma_fence_set_error(&rq->fence, -EIO);
@@ -961,7 +961,7 @@ static void guc_cancel_requests(struct intel_engine_cs *engine)
execlists->queue = RB_ROOT_CACHED;
GEM_BUG_ON(port_isset(execlists->port));
- spin_unlock_irqrestore(&engine->timeline.lock, flags);
+ spin_unlock_irqrestore(&engine->active.lock, flags);
}
static void guc_reset_finish(struct intel_engine_cs *engine)
diff --git a/drivers/gpu/drm/i915/selftests/mock_timeline.c b/drivers/gpu/drm/i915/selftests/mock_timeline.c
index e084476469ef..65b52be23d42 100644
--- a/drivers/gpu/drm/i915/selftests/mock_timeline.c
+++ b/drivers/gpu/drm/i915/selftests/mock_timeline.c
@@ -13,7 +13,6 @@ void mock_timeline_init(struct i915_timeline *timeline, u64 context)
timeline->i915 = NULL;
timeline->fence_context = context;
- spin_lock_init(&timeline->lock);
mutex_init(&timeline->mutex);
INIT_ACTIVE_REQUEST(&timeline->last_request);
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 38/40] drm/i915: Flush the execution-callbacks on retiring
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (35 preceding siblings ...)
2019-05-08 8:07 ` [PATCH 37/40] drm/i915: Replace engine->timeline with a plain list Chris Wilson
@ 2019-05-08 8:07 ` Chris Wilson
2019-05-08 8:07 ` [PATCH 39/40] drm/i915/execlists: Preempt-to-busy Chris Wilson
` (8 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:07 UTC (permalink / raw)
To: intel-gfx
In the unlikely case the request completes while we regard it as not even
executing on the GPU (see the next patch!), we have to flush any pending
execution callbacks at retirement and ensure that we do not add any
more.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_request.c | 93 +++++++++++++++--------------
1 file changed, 49 insertions(+), 44 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 03f15ac63ec9..b1491ce01a77 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -119,6 +119,50 @@ const struct dma_fence_ops i915_fence_ops = {
.release = i915_fence_release,
};
+static void irq_execute_cb(struct irq_work *wrk)
+{
+ struct execute_cb *cb = container_of(wrk, typeof(*cb), work);
+
+ i915_sw_fence_complete(cb->fence);
+ kmem_cache_free(global.slab_execute_cbs, cb);
+}
+
+static void irq_execute_cb_hook(struct irq_work *wrk)
+{
+ struct execute_cb *cb = container_of(wrk, typeof(*cb), work);
+
+ cb->hook(container_of(cb->fence, struct i915_request, submit),
+ &cb->signal->fence);
+ i915_request_put(cb->signal);
+
+ irq_execute_cb(wrk);
+}
+
+static void __notify_execute_cb(struct i915_request *rq)
+{
+ struct execute_cb *cb;
+
+ lockdep_assert_held(&rq->lock);
+
+ if (list_empty(&rq->execute_cb))
+ return;
+
+ list_for_each_entry(cb, &rq->execute_cb, link)
+ irq_work_queue(&cb->work);
+
+ /*
+ * XXX Rollback on __i915_request_unsubmit()
+ *
+ * In the future, perhaps when we have an active time-slicing scheduler,
+ * it will be interesting to unsubmit parallel execution and remove
+ * busywaits from the GPU until their master is restarted. This is
+ * quite hairy, we have to carefully rollback the fence and do a
+ * preempt-to-idle cycle on the target engine, all the while the
+ * master execute_cb may refire.
+ */
+ INIT_LIST_HEAD(&rq->execute_cb);
+}
+
static inline void
i915_request_remove_from_client(struct i915_request *request)
{
@@ -246,6 +290,11 @@ static bool i915_request_retire(struct i915_request *rq)
GEM_BUG_ON(!atomic_read(&rq->i915->gt_pm.rps.num_waiters));
atomic_dec(&rq->i915->gt_pm.rps.num_waiters);
}
+ if (!test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags)) {
+ set_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags);
+ __notify_execute_cb(rq);
+ }
+ GEM_BUG_ON(!list_empty(&rq->execute_cb));
spin_unlock(&rq->lock);
local_irq_enable();
@@ -285,50 +334,6 @@ void i915_request_retire_upto(struct i915_request *rq)
} while (i915_request_retire(tmp) && tmp != rq);
}
-static void irq_execute_cb(struct irq_work *wrk)
-{
- struct execute_cb *cb = container_of(wrk, typeof(*cb), work);
-
- i915_sw_fence_complete(cb->fence);
- kmem_cache_free(global.slab_execute_cbs, cb);
-}
-
-static void irq_execute_cb_hook(struct irq_work *wrk)
-{
- struct execute_cb *cb = container_of(wrk, typeof(*cb), work);
-
- cb->hook(container_of(cb->fence, struct i915_request, submit),
- &cb->signal->fence);
- i915_request_put(cb->signal);
-
- irq_execute_cb(wrk);
-}
-
-static void __notify_execute_cb(struct i915_request *rq)
-{
- struct execute_cb *cb;
-
- lockdep_assert_held(&rq->lock);
-
- if (list_empty(&rq->execute_cb))
- return;
-
- list_for_each_entry(cb, &rq->execute_cb, link)
- irq_work_queue(&cb->work);
-
- /*
- * XXX Rollback on __i915_request_unsubmit()
- *
- * In the future, perhaps when we have an active time-slicing scheduler,
- * it will be interesting to unsubmit parallel execution and remove
- * busywaits from the GPU until their master is restarted. This is
- * quite hairy, we have to carefully rollback the fence and do a
- * preempt-to-idle cycle on the target engine, all the while the
- * master execute_cb may refire.
- */
- INIT_LIST_HEAD(&rq->execute_cb);
-}
-
static int
__i915_request_await_execution(struct i915_request *rq,
struct i915_request *signal,
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 39/40] drm/i915/execlists: Preempt-to-busy
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (36 preceding siblings ...)
2019-05-08 8:07 ` [PATCH 38/40] drm/i915: Flush the execution-callbacks on retiring Chris Wilson
@ 2019-05-08 8:07 ` Chris Wilson
2019-05-08 8:07 ` [PATCH 40/40] drm/i915/execlists: Minimalistic timeslicing Chris Wilson
` (7 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:07 UTC (permalink / raw)
To: intel-gfx
When using a global seqno, we required a precise stop-the-workd event to
handle preemption and unwind the global seqno counter. To accomplish
this, we would preempt to a special out-of-band context and wait for the
machine to report that it was idle. Given an idle machine, we could very
precisely see which requests had completed and which we needed to feed
back into the run queue.
However, now that we have scrapped the global seqno, we no longer need
to precisely unwind the global counter and only track requests by their
per-context seqno. This allows us to loosely unwind inflight requests
while scheduling a preemption, with the enormous caveat that the
requests we put back on the run queue are still _inflight_ (until the
preemption request is complete). This makes request tracking much more
messy, as at any point then we can see a completed request that we
believe is not currently scheduled for execution. We also have to be
careful not to rewind RING_TAIL past RING_HEAD on preempting to the
running context, and for this we use a semaphore to prevent completion
of the request before continuing.
To accomplish this feat, we change how we track requests scheduled to
the HW. Instead of appending our requests onto a single list as we
submit, we track each submission to ELSP as its own block. Then upon
receiving the CS preemption event, we promote the pending block to the
inflight block (discarding what was previously being tracked). As normal
CS completion events arrive, we then remove stale entries from the
inflight tracker.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/gem/i915_gem_context.c | 2 +-
drivers/gpu/drm/i915/gt/intel_context_types.h | 5 +
drivers/gpu/drm/i915/gt/intel_engine.h | 61 +-
drivers/gpu/drm/i915/gt/intel_engine_cs.c | 61 +-
drivers/gpu/drm/i915/gt/intel_engine_types.h | 52 +-
drivers/gpu/drm/i915/gt/intel_lrc.c | 684 ++++++++----------
drivers/gpu/drm/i915/i915_gpu_error.c | 19 +-
drivers/gpu/drm/i915/i915_request.c | 6 +
drivers/gpu/drm/i915/i915_request.h | 1 +
drivers/gpu/drm/i915/i915_scheduler.c | 3 +-
drivers/gpu/drm/i915/i915_utils.h | 12 +
drivers/gpu/drm/i915/intel_guc_submission.c | 175 ++---
drivers/gpu/drm/i915/selftests/i915_request.c | 8 +-
13 files changed, 475 insertions(+), 614 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index aa2bd1d6ceb6..0f11c571bb04 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -642,7 +642,7 @@ static void init_contexts(struct drm_i915_private *i915)
static bool needs_preempt_context(struct drm_i915_private *i915)
{
- return HAS_EXECLISTS(i915);
+ return USES_GUC_SUBMISSION(i915);
}
int i915_gem_contexts_init(struct drm_i915_private *dev_priv)
diff --git a/drivers/gpu/drm/i915/gt/intel_context_types.h b/drivers/gpu/drm/i915/gt/intel_context_types.h
index e95be4be9612..b565c3ff4378 100644
--- a/drivers/gpu/drm/i915/gt/intel_context_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_context_types.h
@@ -13,6 +13,7 @@
#include <linux/types.h>
#include "i915_active_types.h"
+#include "i915_utils.h"
#include "intel_engine_types.h"
#include "intel_sseu.h"
@@ -38,6 +39,10 @@ struct intel_context {
struct i915_gem_context *gem_context;
struct intel_engine_cs *engine;
struct intel_engine_cs *inflight;
+#define intel_context_inflight(ce) ptr_mask_bits((ce)->inflight, 2)
+#define intel_context_inflight_count(ce) ptr_unmask_bits((ce)->inflight, 2)
+#define intel_context_inflight_inc(ce) ptr_count_inc(&(ce)->inflight)
+#define intel_context_inflight_dec(ce) ptr_count_dec(&(ce)->inflight)
struct list_head signal_link;
struct list_head signals;
diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h
index b8f5c4e1d39e..12b19b17e9b0 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine.h
@@ -106,71 +106,26 @@ hangcheck_action_to_str(const enum intel_engine_hangcheck_action a)
void intel_engines_set_scheduler_caps(struct drm_i915_private *i915);
-static inline void
-execlists_set_active(struct intel_engine_execlists *execlists,
- unsigned int bit)
-{
- __set_bit(bit, (unsigned long *)&execlists->active);
-}
-
-static inline bool
-execlists_set_active_once(struct intel_engine_execlists *execlists,
- unsigned int bit)
-{
- return !__test_and_set_bit(bit, (unsigned long *)&execlists->active);
-}
-
-static inline void
-execlists_clear_active(struct intel_engine_execlists *execlists,
- unsigned int bit)
-{
- __clear_bit(bit, (unsigned long *)&execlists->active);
-}
-
-static inline void
-execlists_clear_all_active(struct intel_engine_execlists *execlists)
+static inline unsigned int
+execlists_num_ports(const struct intel_engine_execlists * const execlists)
{
- execlists->active = 0;
+ return execlists->port_mask + 1;
}
-static inline bool
-execlists_is_active(const struct intel_engine_execlists *execlists,
- unsigned int bit)
+static inline struct i915_request *
+execlists_active(const struct intel_engine_execlists *execlists)
{
- return test_bit(bit, (unsigned long *)&execlists->active);
+ GEM_BUG_ON(execlists->active - execlists->inflight >
+ execlists_num_ports(execlists));
+ return READ_ONCE(*execlists->active);
}
-void execlists_user_begin(struct intel_engine_execlists *execlists,
- const struct execlist_port *port);
-void execlists_user_end(struct intel_engine_execlists *execlists);
-
void
execlists_cancel_port_requests(struct intel_engine_execlists * const execlists);
struct i915_request *
execlists_unwind_incomplete_requests(struct intel_engine_execlists *execlists);
-static inline unsigned int
-execlists_num_ports(const struct intel_engine_execlists * const execlists)
-{
- return execlists->port_mask + 1;
-}
-
-static inline struct execlist_port *
-execlists_port_complete(struct intel_engine_execlists * const execlists,
- struct execlist_port * const port)
-{
- const unsigned int m = execlists->port_mask;
-
- GEM_BUG_ON(port_index(port, execlists) != 0);
- GEM_BUG_ON(!execlists_is_active(execlists, EXECLISTS_ACTIVE_USER));
-
- memmove(port, port + 1, m * sizeof(struct execlist_port));
- memset(port + m, 0, sizeof(struct execlist_port));
-
- return port;
-}
-
static inline u32
intel_read_status_page(const struct intel_engine_cs *engine, int reg)
{
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
index f2e28510433b..261ceef5f4fd 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
@@ -458,6 +458,10 @@ void intel_engine_init_execlists(struct intel_engine_cs *engine)
GEM_BUG_ON(!is_power_of_2(execlists_num_ports(execlists)));
GEM_BUG_ON(execlists_num_ports(execlists) > EXECLIST_MAX_PORTS);
+ memset(execlists->pending, 0, sizeof(execlists->pending));
+ execlists->active =
+ memset(execlists->inflight, 0, sizeof(execlists->inflight));
+
execlists->queue_priority_hint = INT_MIN;
execlists->queue = RB_ROOT_CACHED;
}
@@ -1101,7 +1105,7 @@ bool intel_engine_is_idle(struct intel_engine_cs *engine)
return true;
/* Waiting to drain ELSP? */
- if (READ_ONCE(engine->execlists.active)) {
+ if (execlists_active(&engine->execlists)) {
struct tasklet_struct *t = &engine->execlists.tasklet;
synchronize_hardirq(engine->i915->drm.irq);
@@ -1118,7 +1122,7 @@ bool intel_engine_is_idle(struct intel_engine_cs *engine)
/* Otherwise flush the tasklet if it was on another cpu */
tasklet_unlock_wait(t);
- if (READ_ONCE(engine->execlists.active))
+ if (execlists_active(&engine->execlists))
return false;
}
@@ -1315,6 +1319,7 @@ static void intel_engine_print_registers(const struct intel_engine_cs *engine,
}
if (HAS_EXECLISTS(dev_priv)) {
+ struct i915_request * const *port, *rq;
const u32 *hws =
&engine->status_page.addr[I915_HWS_CSB_BUF0_INDEX];
const u8 num_entries = execlists->csb_size;
@@ -1347,26 +1352,28 @@ static void intel_engine_print_registers(const struct intel_engine_cs *engine,
}
rcu_read_lock();
- for (idx = 0; idx < execlists_num_ports(execlists); idx++) {
- struct i915_request *rq;
- unsigned int count;
-
- rq = port_unpack(&execlists->port[idx], &count);
- if (rq) {
- char hdr[80];
-
- snprintf(hdr, sizeof(hdr),
- "\t\tELSP[%d] count=%d, ring:{start:%08x, hwsp:%08x, seqno:%08x}, rq: ",
- idx, count,
- i915_ggtt_offset(rq->ring->vma),
- rq->timeline->hwsp_offset,
- hwsp_seqno(rq));
- print_request(m, rq, hdr);
- } else {
- drm_printf(m, "\t\tELSP[%d] idle\n", idx);
- }
+ for (port = execlists->active; (rq = *port); port++) {
+ char hdr[80];
+
+ snprintf(hdr, sizeof(hdr),
+ "\t\tActive[%d] ring:{start:%08x, hwsp:%08x, seqno:%08x}, rq: ",
+ (int)(port - execlists->active),
+ i915_ggtt_offset(rq->ring->vma),
+ rq->timeline->hwsp_offset,
+ hwsp_seqno(rq));
+ print_request(m, rq, hdr);
+ }
+ for (port = execlists->pending; (rq = *port); port++) {
+ char hdr[80];
+
+ snprintf(hdr, sizeof(hdr),
+ "\t\tPending[%d] ring:{start:%08x, hwsp:%08x, seqno:%08x}, rq: ",
+ (int)(port - execlists->pending),
+ i915_ggtt_offset(rq->ring->vma),
+ rq->timeline->hwsp_offset,
+ hwsp_seqno(rq));
+ print_request(m, rq, hdr);
}
- drm_printf(m, "\t\tHW active? 0x%x\n", execlists->active);
rcu_read_unlock();
} else if (INTEL_GEN(dev_priv) > 6) {
drm_printf(m, "\tPP_DIR_BASE: 0x%08x\n",
@@ -1530,15 +1537,19 @@ int intel_enable_engine_stats(struct intel_engine_cs *engine)
}
if (engine->stats.enabled++ == 0) {
- const struct execlist_port *port = execlists->port;
- unsigned int num_ports = execlists_num_ports(execlists);
+ struct i915_request * const *port;
+ struct i915_request *rq;
engine->stats.enabled_at = ktime_get();
/* XXX submission method oblivious? */
- while (num_ports-- && port_isset(port)) {
+ for (port = execlists->active; (rq = *port); port++)
engine->stats.active++;
- port++;
+
+ for (port = execlists->pending; (rq = *port); port++) {
+ /* Exclude any contexts already counted in active */
+ if (intel_context_inflight_count(rq->hw_context) == 1)
+ engine->stats.active++;
}
if (engine->stats.active)
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
index e00443f53b59..de43bb62c24a 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
@@ -160,51 +160,10 @@ struct intel_engine_execlists {
*/
u32 __iomem *ctrl_reg;
- /**
- * @port: execlist port states
- *
- * For each hardware ELSP (ExecList Submission Port) we keep
- * track of the last request and the number of times we submitted
- * that port to hw. We then count the number of times the hw reports
- * a context completion or preemption. As only one context can
- * be active on hw, we limit resubmission of context to port[0]. This
- * is called Lite Restore, of the context.
- */
- struct execlist_port {
- /**
- * @request_count: combined request and submission count
- */
- struct i915_request *request_count;
-#define EXECLIST_COUNT_BITS 2
-#define port_request(p) ptr_mask_bits((p)->request_count, EXECLIST_COUNT_BITS)
-#define port_count(p) ptr_unmask_bits((p)->request_count, EXECLIST_COUNT_BITS)
-#define port_pack(rq, count) ptr_pack_bits(rq, count, EXECLIST_COUNT_BITS)
-#define port_unpack(p, count) ptr_unpack_bits((p)->request_count, count, EXECLIST_COUNT_BITS)
-#define port_set(p, packed) ((p)->request_count = (packed))
-#define port_isset(p) ((p)->request_count)
-#define port_index(p, execlists) ((p) - (execlists)->port)
-
- /**
- * @context_id: context ID for port
- */
- GEM_DEBUG_DECL(u32 context_id);
-
#define EXECLIST_MAX_PORTS 2
- } port[EXECLIST_MAX_PORTS];
-
- /**
- * @active: is the HW active? We consider the HW as active after
- * submitting any context for execution and until we have seen the
- * last context completion event. After that, we do not expect any
- * more events until we submit, and so can park the HW.
- *
- * As we have a small number of different sources from which we feed
- * the HW, we track the state of each inside a single bitfield.
- */
- unsigned int active;
-#define EXECLISTS_ACTIVE_USER 0
-#define EXECLISTS_ACTIVE_PREEMPT 1
-#define EXECLISTS_ACTIVE_HWACK 2
+ struct i915_request * const *active;
+ struct i915_request *inflight[EXECLIST_MAX_PORTS + 1 /* sentinel */];
+ struct i915_request *pending[EXECLIST_MAX_PORTS + 1];
/**
* @port_mask: number of execlist ports - 1
@@ -245,11 +204,6 @@ struct intel_engine_execlists {
*/
u32 *csb_status;
- /**
- * @preempt_complete_status: expected CSB upon completing preemption
- */
- u32 preempt_complete_status;
-
/**
* @csb_size: context status buffer FIFO size
*/
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index d350ceadffc5..f08c2e223879 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -161,6 +161,8 @@
#define GEN8_CTX_STATUS_COMPLETED_MASK \
(GEN8_CTX_STATUS_COMPLETE | GEN8_CTX_STATUS_PREEMPTED)
+#define CTX_DESC_FORCE_RESTORE BIT_ULL(2)
+
/* Typical size of the average request (2 pipecontrols and a MI_BB) */
#define EXECLISTS_REQUEST_SIZE 64 /* bytes */
#define WA_TAIL_DWORDS 2
@@ -223,6 +225,14 @@ static void execlists_init_reg_state(u32 *reg_state,
struct intel_engine_cs *engine,
struct intel_ring *ring);
+static inline u32 intel_hws_preempt_address(struct intel_engine_cs *engine)
+{
+ return (i915_ggtt_offset(engine->status_page.vma) +
+ I915_GEM_HWS_PREEMPT_ADDR);
+}
+
+#define ring_suspend(E) ((E)->status_page.addr[I915_GEM_HWS_PREEMPT])
+
static inline struct i915_priolist *to_priolist(struct rb_node *rb)
{
return rb_entry(rb, struct i915_priolist, node);
@@ -284,12 +294,6 @@ static inline bool need_preempt(const struct intel_engine_cs *engine,
{
int last_prio;
- if (!engine->preempt_context)
- return false;
-
- if (i915_request_completed(rq))
- return false;
-
/*
* Check if the current priority hint merits a preemption attempt.
*
@@ -351,9 +355,6 @@ __maybe_unused static inline bool
assert_priority_queue(const struct i915_request *prev,
const struct i915_request *next)
{
- const struct intel_engine_execlists *execlists =
- &prev->engine->execlists;
-
/*
* Without preemption, the prev may refer to the still active element
* which we refuse to let go.
@@ -361,7 +362,7 @@ assert_priority_queue(const struct i915_request *prev,
* Even with preemption, there are times when we think it is better not
* to preempt and leave an ostensibly lower priority request in flight.
*/
- if (port_request(execlists->port) == prev)
+ if (i915_request_is_active(prev))
return true;
return rq_prio(prev) >= rq_prio(next);
@@ -455,13 +456,11 @@ __unwind_incomplete_requests(struct intel_engine_cs *engine, int boost)
struct intel_engine_cs *owner;
if (i915_request_completed(rq))
- break;
+ continue; /* XXX */
__i915_request_unsubmit(rq);
unwind_wa_tail(rq);
- GEM_BUG_ON(rq->hw_context->inflight);
-
/*
* Push the request back into the queue for later resubmission.
* If this request is not native to this physical engine (i.e.
@@ -541,32 +540,32 @@ execlists_context_status_change(struct i915_request *rq, unsigned long status)
status, rq);
}
-inline void
-execlists_user_begin(struct intel_engine_execlists *execlists,
- const struct execlist_port *port)
+static inline struct i915_request *
+execlists_schedule_in(struct i915_request *rq, int idx)
{
- execlists_set_active_once(execlists, EXECLISTS_ACTIVE_USER);
-}
+ struct intel_context *ce = rq->hw_context;
+ int count;
-inline void
-execlists_user_end(struct intel_engine_execlists *execlists)
-{
- execlists_clear_active(execlists, EXECLISTS_ACTIVE_USER);
-}
+ trace_i915_request_in(rq, idx);
-static inline void
-execlists_context_schedule_in(struct i915_request *rq)
-{
- GEM_BUG_ON(rq->hw_context->inflight);
+ count = intel_context_inflight_count(ce);
+ if (!count) {
+ intel_context_get(ce);
+ ce->inflight = rq->engine;
+
+ execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN);
+ intel_engine_context_in(ce->inflight);
+ }
- execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN);
- intel_engine_context_in(rq->engine);
- rq->hw_context->inflight = rq->engine;
+ intel_context_inflight_inc(ce);
+ GEM_BUG_ON(intel_context_inflight(ce) != rq->engine);
+
+ return i915_request_get(rq);
}
-static void kick_siblings(struct i915_request *rq)
+static void kick_siblings(struct i915_request *rq, struct intel_context *ce)
{
- struct virtual_engine *ve = to_virtual_engine(rq->hw_context->engine);
+ struct virtual_engine *ve = container_of(ce, typeof(*ve), context);
struct i915_request *next = READ_ONCE(ve->request);
if (next && next->execution_mask & ~rq->execution_mask)
@@ -574,29 +573,42 @@ static void kick_siblings(struct i915_request *rq)
}
static inline void
-execlists_context_schedule_out(struct i915_request *rq, unsigned long status)
+execlists_schedule_out(struct i915_request *rq)
{
- rq->hw_context->inflight = NULL;
- intel_engine_context_out(rq->engine);
- execlists_context_status_change(rq, status);
+ struct intel_context *ce = rq->hw_context;
+
+ GEM_BUG_ON(!intel_context_inflight_count(ce));
+
trace_i915_request_out(rq);
- /*
- * If this is part of a virtual engine, its next request may have
- * been blocked waiting for access to the active context. We have
- * to kick all the siblings again in case we need to switch (e.g.
- * the next request is not runnable on this engine). Hopefully,
- * we will already have submitted the next request before the
- * tasklet runs and do not need to rebuild each virtual tree
- * and kick everyone again.
- */
- if (rq->engine != rq->hw_context->engine)
- kick_siblings(rq);
+ intel_context_inflight_dec(ce);
+ if (!intel_context_inflight_count(ce)) {
+ intel_engine_context_out(ce->inflight);
+ execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_OUT);
+
+ ce->inflight = NULL;
+ intel_context_put(ce);
+
+ /*
+ * If this is part of a virtual engine, its next request may
+ * have been blocked waiting for access to the active context.
+ * We have to kick all the siblings again in case we need to
+ * switch (e.g. the next request is not runnable on this
+ * engine). Hopefully, we will already have submitted the next
+ * request before the tasklet runs and do not need to rebuild
+ * each virtual tree and kick everyone again.
+ */
+ if (rq->engine != ce->engine)
+ kick_siblings(rq, ce);
+ }
+
+ i915_request_put(rq);
}
-static u64 execlists_update_context(struct i915_request *rq)
+static u64 execlists_update_context(const struct i915_request *rq)
{
struct intel_context *ce = rq->hw_context;
+ u64 desc;
ce->lrc_reg_state[CTX_RING_TAIL + 1] =
intel_ring_set_tail(rq->ring, rq->tail);
@@ -617,7 +629,11 @@ static u64 execlists_update_context(struct i915_request *rq)
* wmb).
*/
mb();
- return ce->lrc_desc;
+
+ desc = ce->lrc_desc;
+ ce->lrc_desc &= ~CTX_DESC_FORCE_RESTORE;
+
+ return desc;
}
static inline void write_desc(struct intel_engine_execlists *execlists, u64 desc, u32 port)
@@ -631,12 +647,54 @@ static inline void write_desc(struct intel_engine_execlists *execlists, u64 desc
}
}
+static __maybe_unused void
+trace_ports(const struct intel_engine_execlists *execlists,
+ const char *msg,
+ struct i915_request * const *ports)
+{
+ const struct intel_engine_cs *engine =
+ container_of(execlists, typeof(*engine), execlists);
+
+ GEM_TRACE("%s: %s { %llx:%lld%s, %llx:%lld }\n",
+ engine->name, msg,
+ ports[0]->fence.context,
+ ports[0]->fence.seqno,
+ i915_request_completed(ports[0]) ? "!" :
+ i915_request_started(ports[0]) ? "*" :
+ "",
+ ports[1] ? ports[1]->fence.context : 0,
+ ports[1] ? ports[1]->fence.seqno : 0);
+}
+
+static __maybe_unused bool
+assert_pending_valid(const struct intel_engine_execlists *execlists,
+ const char *msg)
+{
+ struct i915_request * const *port, *rq;
+ struct intel_context *ce = NULL;
+
+ trace_ports(execlists, msg, execlists->pending);
+
+ if (execlists->pending[execlists_num_ports(execlists)])
+ return false;
+
+ for (port = execlists->pending; (rq = *port); port++) {
+ if (ce == rq->hw_context)
+ return false;
+
+ ce = rq->hw_context;
+ }
+
+ return ce;
+}
+
static void execlists_submit_ports(struct intel_engine_cs *engine)
{
struct intel_engine_execlists *execlists = &engine->execlists;
- struct execlist_port *port = execlists->port;
unsigned int n;
+ GEM_BUG_ON(!assert_pending_valid(execlists, "submit"));
+
/*
* We can skip acquiring intel_runtime_pm_get() here as it was taken
* on our behalf by the request (see i915_gem_mark_busy()) and it will
@@ -654,38 +712,16 @@ static void execlists_submit_ports(struct intel_engine_cs *engine)
* of elsq entries, keep this in mind before changing the loop below.
*/
for (n = execlists_num_ports(execlists); n--; ) {
- struct i915_request *rq;
- unsigned int count;
- u64 desc;
-
- rq = port_unpack(&port[n], &count);
- if (rq) {
- GEM_BUG_ON(count > !n);
- if (!count++)
- execlists_context_schedule_in(rq);
- port_set(&port[n], port_pack(rq, count));
- desc = execlists_update_context(rq);
- GEM_DEBUG_EXEC(port[n].context_id = upper_32_bits(desc));
-
- GEM_TRACE("%s in[%d]: ctx=%d.%d, fence %llx:%lld (current %d), prio=%d\n",
- engine->name, n,
- port[n].context_id, count,
- rq->fence.context, rq->fence.seqno,
- hwsp_seqno(rq),
- rq_prio(rq));
- } else {
- GEM_BUG_ON(!n);
- desc = 0;
- }
+ struct i915_request *rq = execlists->pending[n];
- write_desc(execlists, desc, n);
+ write_desc(execlists,
+ rq ? execlists_update_context(rq) : 0,
+ n);
}
/* we need to manually load the submit queue */
if (execlists->ctrl_reg)
writel(EL_CTRL_LOAD, execlists->ctrl_reg);
-
- execlists_clear_active(execlists, EXECLISTS_ACTIVE_HWACK);
}
static bool ctx_single_port_submission(const struct intel_context *ce)
@@ -709,6 +745,7 @@ static bool can_merge_ctx(const struct intel_context *prev,
static bool can_merge_rq(const struct i915_request *prev,
const struct i915_request *next)
{
+ GEM_BUG_ON(prev == next);
GEM_BUG_ON(!assert_priority_queue(prev, next));
if (!can_merge_ctx(prev->hw_context, next->hw_context))
@@ -717,59 +754,6 @@ static bool can_merge_rq(const struct i915_request *prev,
return true;
}
-static void port_assign(struct execlist_port *port, struct i915_request *rq)
-{
- GEM_BUG_ON(rq == port_request(port));
-
- if (port_isset(port))
- i915_request_put(port_request(port));
-
- port_set(port, port_pack(i915_request_get(rq), port_count(port)));
-}
-
-static void inject_preempt_context(struct intel_engine_cs *engine)
-{
- struct intel_engine_execlists *execlists = &engine->execlists;
- struct intel_context *ce = engine->preempt_context;
- unsigned int n;
-
- GEM_BUG_ON(execlists->preempt_complete_status !=
- upper_32_bits(ce->lrc_desc));
-
- /*
- * Switch to our empty preempt context so
- * the state of the GPU is known (idle).
- */
- GEM_TRACE("%s\n", engine->name);
- for (n = execlists_num_ports(execlists); --n; )
- write_desc(execlists, 0, n);
-
- write_desc(execlists, ce->lrc_desc, n);
-
- /* we need to manually load the submit queue */
- if (execlists->ctrl_reg)
- writel(EL_CTRL_LOAD, execlists->ctrl_reg);
-
- execlists_clear_active(execlists, EXECLISTS_ACTIVE_HWACK);
- execlists_set_active(execlists, EXECLISTS_ACTIVE_PREEMPT);
-
- (void)I915_SELFTEST_ONLY(execlists->preempt_hang.count++);
-}
-
-static void complete_preempt_context(struct intel_engine_execlists *execlists)
-{
- GEM_BUG_ON(!execlists_is_active(execlists, EXECLISTS_ACTIVE_PREEMPT));
-
- if (inject_preempt_hang(execlists))
- return;
-
- execlists_cancel_port_requests(execlists);
- __unwind_incomplete_requests(container_of(execlists,
- struct intel_engine_cs,
- execlists),
- ACTIVE_PRIORITY);
-}
-
static void virtual_update_register_offsets(u32 *regs,
struct intel_engine_cs *engine)
{
@@ -834,7 +818,7 @@ static bool virtual_matches(const struct virtual_engine *ve,
* we reuse the register offsets). This is a very small
* hystersis on the greedy seelction algorithm.
*/
- inflight = READ_ONCE(ve->context.inflight);
+ inflight = intel_context_inflight(&ve->context);
if (inflight && inflight != engine)
return false;
@@ -857,13 +841,23 @@ static void virtual_xfer_breadcrumbs(struct virtual_engine *ve,
spin_unlock(&old->breadcrumbs.irq_lock);
}
-static void execlists_dequeue(struct intel_engine_cs *engine)
+static struct i915_request *
+last_active(const struct intel_engine_execlists *execlists)
+{
+ struct i915_request * const *last = execlists->active;
+
+ while (*last && i915_request_completed(*last))
+ last++;
+
+ return *last;
+}
+
+static bool execlists_dequeue(struct intel_engine_cs *engine)
{
struct intel_engine_execlists * const execlists = &engine->execlists;
- struct execlist_port *port = execlists->port;
- const struct execlist_port * const last_port =
- &execlists->port[execlists->port_mask];
- struct i915_request *last = port_request(port);
+ struct i915_request **port = execlists->pending;
+ struct i915_request ** const last_port = port + execlists->port_mask;
+ struct i915_request *last;
struct rb_node *rb;
bool submit = false;
@@ -909,65 +903,72 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
break;
}
+ /*
+ * If the queue is higher priority than the last
+ * request in the currently active context, submit afresh.
+ * We will resubmit again afterwards in case we need to split
+ * the active context to interject the preemption request,
+ * i.e. we will retrigger preemption following the ack in case
+ * of trouble.
+ */
+ last = last_active(execlists);
if (last) {
- /*
- * Don't resubmit or switch until all outstanding
- * preemptions (lite-restore) are seen. Then we
- * know the next preemption status we see corresponds
- * to this ELSP update.
- */
- GEM_BUG_ON(!execlists_is_active(execlists,
- EXECLISTS_ACTIVE_USER));
- GEM_BUG_ON(!port_count(&port[0]));
-
- /*
- * If we write to ELSP a second time before the HW has had
- * a chance to respond to the previous write, we can confuse
- * the HW and hit "undefined behaviour". After writing to ELSP,
- * we must then wait until we see a context-switch event from
- * the HW to indicate that it has had a chance to respond.
- */
- if (!execlists_is_active(execlists, EXECLISTS_ACTIVE_HWACK))
- return;
-
if (need_preempt(engine, last, rb)) {
- inject_preempt_context(engine);
- return;
- }
+ GEM_TRACE("%s: preempting last=%llx:%lld, prio=%d, hint=%d\n",
+ engine->name,
+ last->fence.context,
+ last->fence.seqno,
+ last->sched.attr.priority,
+ execlists->queue_priority_hint);
+ /*
+ * Don't let the RING_HEAD advance past the breadcrumb
+ * as we unwind (and until we resubmit) so that we do
+ * not accidentally tell it to go backwards.
+ */
+ ring_suspend(engine) = 1;
- /*
- * In theory, we could coalesce more requests onto
- * the second port (the first port is active, with
- * no preemptions pending). However, that means we
- * then have to deal with the possible lite-restore
- * of the second port (as we submit the ELSP, there
- * may be a context-switch) but also we may complete
- * the resubmission before the context-switch. Ergo,
- * coalescing onto the second port will cause a
- * preemption event, but we cannot predict whether
- * that will affect port[0] or port[1].
- *
- * If the second port is already active, we can wait
- * until the next context-switch before contemplating
- * new requests. The GPU will be busy and we should be
- * able to resubmit the new ELSP before it idles,
- * avoiding pipeline bubbles (momentary pauses where
- * the driver is unable to keep up the supply of new
- * work). However, we have to double check that the
- * priorities of the ports haven't been switch.
- */
- if (port_count(&port[1]))
- return;
+ /*
+ * Note that we have not stopped the GPU at this point,
+ * so we are unwinding the incomplete requests as they
+ * remain inflight and so by the time we do complete
+ * the preemption, some of the unwound requests may
+ * complete!
+ */
+ __unwind_incomplete_requests(engine, ACTIVE_PRIORITY);
- /*
- * WaIdleLiteRestore:bdw,skl
- * Apply the wa NOOPs to prevent
- * ring:HEAD == rq:TAIL as we resubmit the
- * request. See gen8_emit_fini_breadcrumb() for
- * where we prepare the padding after the
- * end of the request.
- */
- last->tail = last->wa_tail;
+ /*
+ * If we need to return to the preempted context, we
+ * need to skip the lite-restore and force it to
+ * reload the RING_TAIL. Otherwise, the HW has a
+ * tendency to ignore us rewinding the TAIL to the
+ * end of an earlier request.
+ */
+ last->hw_context->lrc_desc |= CTX_DESC_FORCE_RESTORE;
+ last = NULL;
+ } else {
+ /*
+ * Otherwise if we already have a request pending
+ * for execution after the current one, we can
+ * just wait until the next CS event before
+ * queuing more. In either case we will force a
+ * lite-restore preemption event, but if we wait
+ * we hopefully coalesce several updates into a single
+ * submission.
+ */
+ if (!list_is_last(&last->sched.link,
+ &engine->active.requests))
+ return false;
+
+ /*
+ * WaIdleLiteRestore:bdw,skl
+ * Apply the wa NOOPs to prevent
+ * ring:HEAD == rq:TAIL as we resubmit the
+ * request. See gen8_emit_fini_breadcrumb() for
+ * where we prepare the padding after the
+ * end of the request.
+ */
+ last->tail = last->wa_tail;
+ }
}
while (rb) { /* XXX virtual is always taking precedence */
@@ -997,9 +998,24 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
continue;
}
+ if (i915_request_completed(rq)) {
+ ve->request = NULL;
+ ve->base.execlists.queue_priority_hint = INT_MIN;
+ rb_erase_cached(rb, &execlists->virtual);
+ RB_CLEAR_NODE(rb);
+
+ rq->engine = engine;
+ __i915_request_submit(rq);
+
+ spin_unlock(&ve->base.active.lock);
+
+ rb = rb_first_cached(&execlists->virtual);
+ continue;
+ }
+
if (last && !can_merge_rq(last, rq)) {
spin_unlock(&ve->base.active.lock);
- return; /* leave this rq for another engine */
+ return false; /* leave this for another */
}
GEM_TRACE("%s: virtual rq=%llx:%lld%s, new engine? %s\n",
@@ -1048,9 +1064,10 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
}
__i915_request_submit(rq);
- trace_i915_request_in(rq, port_index(port, execlists));
- submit = true;
- last = rq;
+ if (!i915_request_completed(rq)) {
+ submit = true;
+ last = rq;
+ }
}
spin_unlock(&ve->base.active.lock);
@@ -1063,6 +1080,9 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
int i;
priolist_for_each_request_consume(rq, rn, p, i) {
+ if (i915_request_completed(rq))
+ goto skip;
+
/*
* Can we combine this request with the current port?
* It has to be the same context/ringbuffer and not
@@ -1102,19 +1122,14 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
ctx_single_port_submission(rq->hw_context))
goto done;
-
- if (submit)
- port_assign(port, last);
+ *port = execlists_schedule_in(last, port - execlists->pending);
port++;
-
- GEM_BUG_ON(port_isset(port));
}
- __i915_request_submit(rq);
- trace_i915_request_in(rq, port_index(port, execlists));
-
last = rq;
submit = true;
+skip:
+ __i915_request_submit(rq);
}
rb_erase_cached(&p->node, &execlists->queue);
@@ -1139,54 +1154,32 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
* interrupt for secondary ports).
*/
execlists->queue_priority_hint = queue_prio(execlists);
+ GEM_TRACE("%s: queue_priority_hint:%d, submit:%s\n",
+ engine->name, execlists->queue_priority_hint,
+ yesno(submit));
if (submit) {
- port_assign(port, last);
+ *port = execlists_schedule_in(last, port - execlists->pending);
+ memset(port + 1, 0, (last_port - port) * sizeof(*port));
execlists_submit_ports(engine);
}
- /* We must always keep the beast fed if we have work piled up */
- GEM_BUG_ON(rb_first_cached(&execlists->queue) &&
- !port_isset(execlists->port));
-
- /* Re-evaluate the executing context setup after each preemptive kick */
- if (last)
- execlists_user_begin(execlists, execlists->port);
-
- /* If the engine is now idle, so should be the flag; and vice versa. */
- GEM_BUG_ON(execlists_is_active(&engine->execlists,
- EXECLISTS_ACTIVE_USER) ==
- !port_isset(engine->execlists.port));
+ return submit;
}
void
execlists_cancel_port_requests(struct intel_engine_execlists * const execlists)
{
- struct execlist_port *port = execlists->port;
- unsigned int num_ports = execlists_num_ports(execlists);
-
- while (num_ports-- && port_isset(port)) {
- struct i915_request *rq = port_request(port);
-
- GEM_TRACE("%s:port%u fence %llx:%lld, (current %d)\n",
- rq->engine->name,
- (unsigned int)(port - execlists->port),
- rq->fence.context, rq->fence.seqno,
- hwsp_seqno(rq));
+ struct i915_request * const *port, *rq;
- GEM_BUG_ON(!execlists->active);
- execlists_context_schedule_out(rq,
- i915_request_completed(rq) ?
- INTEL_CONTEXT_SCHEDULE_OUT :
- INTEL_CONTEXT_SCHEDULE_PREEMPTED);
-
- i915_request_put(rq);
-
- memset(port, 0, sizeof(*port));
- port++;
- }
+ for (port = execlists->pending; (rq = *port); port++)
+ execlists_schedule_out(rq);
+ memset(execlists->pending, 0, sizeof(execlists->pending));
- execlists_clear_all_active(execlists);
+ for (port = execlists->active; (rq = *port); port++)
+ execlists_schedule_out(rq);
+ execlists->active =
+ memset(execlists->inflight, 0, sizeof(execlists->inflight));
}
static inline void
@@ -1205,7 +1198,6 @@ reset_in_progress(const struct intel_engine_execlists *execlists)
static void process_csb(struct intel_engine_cs *engine)
{
struct intel_engine_execlists * const execlists = &engine->execlists;
- struct execlist_port *port = execlists->port;
const u32 * const buf = execlists->csb_status;
const u8 num_entries = execlists->csb_size;
u8 head, tail;
@@ -1239,9 +1231,7 @@ static void process_csb(struct intel_engine_cs *engine)
rmb();
do {
- struct i915_request *rq;
unsigned int status;
- unsigned int count;
if (++head == num_entries)
head = 0;
@@ -1264,68 +1254,37 @@ static void process_csb(struct intel_engine_cs *engine)
* status notifier.
*/
- GEM_TRACE("%s csb[%d]: status=0x%08x:0x%08x, active=0x%x\n",
+ GEM_TRACE("%s csb[%d]: status=0x%08x:0x%08x\n",
engine->name, head,
- buf[2 * head + 0], buf[2 * head + 1],
- execlists->active);
+ buf[2 * head + 0], buf[2 * head + 1]);
status = buf[2 * head];
- if (status & (GEN8_CTX_STATUS_IDLE_ACTIVE |
- GEN8_CTX_STATUS_PREEMPTED))
- execlists_set_active(execlists,
- EXECLISTS_ACTIVE_HWACK);
- if (status & GEN8_CTX_STATUS_ACTIVE_IDLE)
- execlists_clear_active(execlists,
- EXECLISTS_ACTIVE_HWACK);
-
- if (!(status & GEN8_CTX_STATUS_COMPLETED_MASK))
- continue;
-
- /* We should never get a COMPLETED | IDLE_ACTIVE! */
- GEM_BUG_ON(status & GEN8_CTX_STATUS_IDLE_ACTIVE);
+ if (status & GEN8_CTX_STATUS_IDLE_ACTIVE) {
+promote:
+ GEM_BUG_ON(!assert_pending_valid(execlists, "promote"));
+ execlists->active =
+ memcpy(execlists->inflight,
+ execlists->pending,
+ execlists_num_ports(execlists) *
+ sizeof(*execlists->pending));
+ execlists->pending[0] = NULL;
- if (status & GEN8_CTX_STATUS_COMPLETE &&
- buf[2*head + 1] == execlists->preempt_complete_status) {
- GEM_TRACE("%s preempt-idle\n", engine->name);
- complete_preempt_context(execlists);
- continue;
- }
+ if (!inject_preempt_hang(execlists))
+ ring_suspend(engine) = 0;
+ } else if (status & GEN8_CTX_STATUS_PREEMPTED) {
+ struct i915_request * const *port = execlists->active;
- if (status & GEN8_CTX_STATUS_PREEMPTED &&
- execlists_is_active(execlists,
- EXECLISTS_ACTIVE_PREEMPT))
- continue;
+ trace_ports(execlists, "preempted", execlists->active);
- GEM_BUG_ON(!execlists_is_active(execlists,
- EXECLISTS_ACTIVE_USER));
+ while (*port)
+ execlists_schedule_out(*port++);
- rq = port_unpack(port, &count);
- GEM_TRACE("%s out[0]: ctx=%d.%d, fence %llx:%lld (current %d), prio=%d\n",
- engine->name,
- port->context_id, count,
- rq ? rq->fence.context : 0,
- rq ? rq->fence.seqno : 0,
- rq ? hwsp_seqno(rq) : 0,
- rq ? rq_prio(rq) : 0);
+ goto promote;
+ } else if (*execlists->active) {
+ struct i915_request *rq = *execlists->active++;
- /* Check the context/desc id for this event matches */
- GEM_DEBUG_BUG_ON(buf[2 * head + 1] != port->context_id);
-
- GEM_BUG_ON(count == 0);
- if (--count == 0) {
- /*
- * On the final event corresponding to the
- * submission of this context, we expect either
- * an element-switch event or a completion
- * event (and on completion, the active-idle
- * marker). No more preemptions, lite-restore
- * or otherwise.
- */
- GEM_BUG_ON(status & GEN8_CTX_STATUS_PREEMPTED);
- GEM_BUG_ON(port_isset(&port[1]) &&
- !(status & GEN8_CTX_STATUS_ELEMENT_SWITCH));
- GEM_BUG_ON(!port_isset(&port[1]) &&
- !(status & GEN8_CTX_STATUS_ACTIVE_IDLE));
+ trace_ports(execlists, "completed",
+ execlists->active - 1);
/*
* We rely on the hardware being strongly
@@ -1334,21 +1293,10 @@ static void process_csb(struct intel_engine_cs *engine)
* user interrupt and CSB is processed.
*/
GEM_BUG_ON(!i915_request_completed(rq));
+ execlists_schedule_out(rq);
- execlists_context_schedule_out(rq,
- INTEL_CONTEXT_SCHEDULE_OUT);
- i915_request_put(rq);
-
- GEM_TRACE("%s completed ctx=%d\n",
- engine->name, port->context_id);
-
- port = execlists_port_complete(execlists, port);
- if (port_isset(port))
- execlists_user_begin(execlists, port);
- else
- execlists_user_end(execlists);
- } else {
- port_set(port, port_pack(rq, count));
+ GEM_BUG_ON(execlists->active - execlists->inflight >
+ execlists_num_ports(execlists));
}
} while (head != tail);
@@ -1370,11 +1318,17 @@ static void process_csb(struct intel_engine_cs *engine)
static void __execlists_submission_tasklet(struct intel_engine_cs *const engine)
{
+ bool submit;
+
lockdep_assert_held(&engine->active.lock);
- process_csb(engine);
- if (!execlists_is_active(&engine->execlists, EXECLISTS_ACTIVE_PREEMPT))
- execlists_dequeue(engine);
+ do {
+ process_csb(engine);
+
+ submit = false;
+ if (!engine->execlists.pending[0])
+ submit = execlists_dequeue(engine);
+ } while (submit);
}
/*
@@ -1386,11 +1340,6 @@ static void execlists_submission_tasklet(unsigned long data)
struct intel_engine_cs * const engine = (struct intel_engine_cs *)data;
unsigned long flags;
- GEM_TRACE("%s awake?=%d, active=%x\n",
- engine->name,
- !!intel_wakeref_active(&engine->wakeref),
- engine->execlists.active);
-
spin_lock_irqsave(&engine->active.lock, flags);
__execlists_submission_tasklet(engine);
spin_unlock_irqrestore(&engine->active.lock, flags);
@@ -1417,12 +1366,16 @@ static void __submit_queue_imm(struct intel_engine_cs *engine)
tasklet_hi_schedule(&execlists->tasklet);
}
-static void submit_queue(struct intel_engine_cs *engine, int prio)
+static void submit_queue(struct intel_engine_cs *engine,
+ const struct i915_request *rq)
{
- if (prio > engine->execlists.queue_priority_hint) {
- engine->execlists.queue_priority_hint = prio;
- __submit_queue_imm(engine);
- }
+ struct intel_engine_execlists *execlists = &engine->execlists;
+
+ if (rq_prio(rq) <= execlists->queue_priority_hint)
+ return;
+
+ execlists->queue_priority_hint = rq_prio(rq);
+ __submit_queue_imm(engine);
}
static void execlists_submit_request(struct i915_request *request)
@@ -1438,7 +1391,7 @@ static void execlists_submit_request(struct i915_request *request)
GEM_BUG_ON(RB_EMPTY_ROOT(&engine->execlists.queue.rb_root));
GEM_BUG_ON(list_empty(&request->sched.link));
- submit_queue(engine, rq_prio(request));
+ submit_queue(engine, request);
spin_unlock_irqrestore(&engine->active.lock, flags);
}
@@ -2095,27 +2048,13 @@ static void execlists_reset_prepare(struct intel_engine_cs *engine)
spin_unlock_irqrestore(&engine->active.lock, flags);
}
-static bool lrc_regs_ok(const struct i915_request *rq)
-{
- const struct intel_ring *ring = rq->ring;
- const u32 *regs = rq->hw_context->lrc_reg_state;
-
- /* Quick spot check for the common signs of context corruption */
-
- if (regs[CTX_RING_BUFFER_CONTROL + 1] !=
- (RING_CTL_SIZE(ring->size) | RING_VALID))
- return false;
-
- if (regs[CTX_RING_BUFFER_START + 1] != i915_ggtt_offset(ring->vma))
- return false;
-
- return true;
-}
-
-static void reset_csb_pointers(struct intel_engine_execlists *execlists)
+static void reset_csb_pointers(struct intel_engine_cs *engine)
{
+ struct intel_engine_execlists * const execlists = &engine->execlists;
const unsigned int reset_value = execlists->csb_size - 1;
+ ring_suspend(engine) = 0;
+
/*
* After a reset, the HW starts writing into CSB entry [0]. We
* therefore have to set our HEAD pointer back one entry so that
@@ -2162,18 +2101,19 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
process_csb(engine); /* drain preemption events */
/* Following the reset, we need to reload the CSB read/write pointers */
- reset_csb_pointers(&engine->execlists);
+ reset_csb_pointers(engine);
/*
* Save the currently executing context, even if we completed
* its request, it was still running at the time of the
* reset and will have been clobbered.
*/
- if (!port_isset(execlists->port))
- goto out_clear;
+ rq = execlists_active(execlists);
+ if (!rq)
+ return;
- rq = port_request(execlists->port);
ce = rq->hw_context;
+ rq = active_request(rq);
/*
* Catch up with any missed context-switch interrupts.
@@ -2186,9 +2126,12 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
*/
execlists_cancel_port_requests(execlists);
- rq = active_request(rq);
- if (!rq)
+ if (!rq) {
+ ce->ring->head = ce->ring->tail;
goto out_replay;
+ }
+
+ ce->ring->head = intel_ring_wrap(ce->ring, rq->head);
/*
* If this request hasn't started yet, e.g. it is waiting on a
@@ -2202,7 +2145,7 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
* Otherwise, if we have not started yet, the request should replay
* perfectly and we do not need to flag the result as being erroneous.
*/
- if (!i915_request_started(rq) && lrc_regs_ok(rq))
+ if (!i915_request_started(rq))
goto out_replay;
/*
@@ -2217,7 +2160,7 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
* image back to the expected values to skip over the guilty request.
*/
i915_reset_request(rq, stalled);
- if (!stalled && lrc_regs_ok(rq))
+ if (!stalled)
goto out_replay;
/*
@@ -2237,17 +2180,13 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
execlists_init_reg_state(regs, ce, engine, ce->ring);
out_replay:
- /* Rerun the request; its payload has been neutered (if guilty). */
- ce->ring->head =
- rq ? intel_ring_wrap(ce->ring, rq->head) : ce->ring->tail;
+ GEM_TRACE("%s replay {head:%04x, tail:%04x\n",
+ engine->name, ce->ring->head, ce->ring->tail);
intel_ring_update_space(ce->ring);
__execlists_update_reg_state(ce, engine);
/* Push back any incomplete requests for replay after the reset. */
__unwind_incomplete_requests(engine, 0);
-
-out_clear:
- execlists_clear_all_active(execlists);
}
static void execlists_reset(struct intel_engine_cs *engine, bool stalled)
@@ -2343,7 +2282,6 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine)
execlists->queue_priority_hint = INT_MIN;
execlists->queue = RB_ROOT_CACHED;
- GEM_BUG_ON(port_isset(execlists->port));
GEM_BUG_ON(__tasklet_is_enabled(&execlists->tasklet));
execlists->tasklet.func = nop_submission_tasklet;
@@ -2561,15 +2499,29 @@ static u32 *gen8_emit_wa_tail(struct i915_request *request, u32 *cs)
return cs;
}
+static u32 *emit_preempt_busywait(struct i915_request *request, u32 *cs)
+{
+ *cs++ = MI_SEMAPHORE_WAIT |
+ MI_SEMAPHORE_GLOBAL_GTT |
+ MI_SEMAPHORE_POLL |
+ MI_SEMAPHORE_SAD_EQ_SDD;
+ *cs++ = 0;
+ *cs++ = intel_hws_preempt_address(request->engine);
+ *cs++ = 0;
+
+ return cs;
+}
+
static u32 *gen8_emit_fini_breadcrumb(struct i915_request *request, u32 *cs)
{
cs = gen8_emit_ggtt_write(cs,
request->fence.seqno,
request->timeline->hwsp_offset,
0);
-
*cs++ = MI_USER_INTERRUPT;
+
*cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
+ cs = emit_preempt_busywait(request, cs);
request->tail = intel_ring_offset(request, cs);
assert_ring_tail_valid(request->ring, request->tail);
@@ -2589,9 +2541,10 @@ static u32 *gen8_emit_fini_breadcrumb_rcs(struct i915_request *request, u32 *cs)
PIPE_CONTROL_FLUSH_ENABLE |
PIPE_CONTROL_CS_STALL,
0);
-
*cs++ = MI_USER_INTERRUPT;
+
*cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
+ cs = emit_preempt_busywait(request, cs);
request->tail = intel_ring_offset(request, cs);
assert_ring_tail_valid(request->ring, request->tail);
@@ -2640,8 +2593,7 @@ void intel_execlists_set_default_submission(struct intel_engine_cs *engine)
engine->flags |= I915_ENGINE_SUPPORTS_STATS;
if (!intel_vgpu_active(engine->i915))
engine->flags |= I915_ENGINE_HAS_SEMAPHORES;
- if (engine->preempt_context &&
- HAS_LOGICAL_RING_PREEMPTION(engine->i915))
+ if (HAS_LOGICAL_RING_PREEMPTION(engine->i915))
engine->flags |= I915_ENGINE_HAS_PREEMPTION;
}
@@ -2763,11 +2715,6 @@ int intel_execlists_submission_init(struct intel_engine_cs *engine)
i915_mmio_reg_offset(RING_ELSP(base));
}
- execlists->preempt_complete_status = ~0u;
- if (engine->preempt_context)
- execlists->preempt_complete_status =
- upper_32_bits(engine->preempt_context->lrc_desc);
-
execlists->csb_status =
&engine->status_page.addr[I915_HWS_CSB_BUF0_INDEX];
@@ -2779,7 +2726,7 @@ int intel_execlists_submission_init(struct intel_engine_cs *engine)
else
execlists->csb_size = GEN11_CSB_ENTRIES;
- reset_csb_pointers(execlists);
+ reset_csb_pointers(engine);
return 0;
}
@@ -2962,11 +2909,6 @@ populate_lr_context(struct intel_context *ce,
if (!engine->default_state)
regs[CTX_CONTEXT_CONTROL + 1] |=
_MASKED_BIT_ENABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT);
- if (ce->gem_context == engine->i915->preempt_context &&
- INTEL_GEN(engine->i915) < 11)
- regs[CTX_CONTEXT_CONTROL + 1] |=
- _MASKED_BIT_ENABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT |
- CTX_CTRL_ENGINE_CTX_SAVE_INHIBIT);
ret = 0;
err_unpin_ctx:
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index 4cc900e432e2..67f56b4a4e3a 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -1241,10 +1241,10 @@ static void error_record_engine_registers(struct i915_gpu_state *error,
}
}
-static void record_request(struct i915_request *request,
+static void record_request(const struct i915_request *request,
struct drm_i915_error_request *erq)
{
- struct i915_gem_context *ctx = request->gem_context;
+ const struct i915_gem_context *ctx = request->gem_context;
erq->flags = request->fence.flags;
erq->context = request->fence.context;
@@ -1308,20 +1308,15 @@ static void engine_record_requests(struct intel_engine_cs *engine,
ee->num_requests = count;
}
-static void error_record_engine_execlists(struct intel_engine_cs *engine,
+static void error_record_engine_execlists(const struct intel_engine_cs *engine,
struct drm_i915_error_engine *ee)
{
const struct intel_engine_execlists * const execlists = &engine->execlists;
- unsigned int n;
+ struct i915_request * const *port = execlists->active;
+ unsigned int n = 0;
- for (n = 0; n < execlists_num_ports(execlists); n++) {
- struct i915_request *rq = port_request(&execlists->port[n]);
-
- if (!rq)
- break;
-
- record_request(rq, &ee->execlist[n]);
- }
+ while (*port)
+ record_request(*port++, &ee->execlist[n++]);
ee->num_ports = n;
}
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index b1491ce01a77..d69a1a75c8c7 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -276,6 +276,12 @@ static bool i915_request_retire(struct i915_request *rq)
local_irq_disable();
+ /*
+ * We only loosely track inflight requests across preemption,
+ * and so we may find ourselves attempting to retire a _completed_
+ * request that we have removed from the HW and put back on a run
+ * queue.
+ */
spin_lock(&rq->engine->active.lock);
list_del(&rq->sched.link);
spin_unlock(&rq->engine->active.lock);
diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
index edbbdfec24ab..bebc1e9b4a5e 100644
--- a/drivers/gpu/drm/i915/i915_request.h
+++ b/drivers/gpu/drm/i915/i915_request.h
@@ -28,6 +28,7 @@
#include <linux/dma-fence.h>
#include <linux/lockdep.h>
+#include "gt/intel_context_types.h"
#include "gt/intel_engine_types.h"
#include "i915_gem.h"
diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c
index 2cba5b9b34bb..7aa892ff6550 100644
--- a/drivers/gpu/drm/i915/i915_scheduler.c
+++ b/drivers/gpu/drm/i915/i915_scheduler.c
@@ -179,8 +179,7 @@ static inline int rq_prio(const struct i915_request *rq)
static void kick_submission(struct intel_engine_cs *engine, int prio)
{
- const struct i915_request *inflight =
- port_request(engine->execlists.port);
+ const struct i915_request *inflight = *engine->execlists.active;
/*
* If we are already the currently executing context, don't
diff --git a/drivers/gpu/drm/i915/i915_utils.h b/drivers/gpu/drm/i915/i915_utils.h
index 9911f53382a5..6e0179e76dd9 100644
--- a/drivers/gpu/drm/i915/i915_utils.h
+++ b/drivers/gpu/drm/i915/i915_utils.h
@@ -131,6 +131,18 @@ __check_struct_size(size_t base, size_t arr, size_t count, size_t *size)
((typeof(ptr))((unsigned long)(ptr) | __bits)); \
})
+#define ptr_count_dec(p_ptr) do { \
+ typeof(p_ptr) __p = (p_ptr); \
+ unsigned long __v = (unsigned long)(*__p); \
+ *__p = (typeof(*p_ptr))(--__v); \
+} while (0)
+
+#define ptr_count_inc(p_ptr) do { \
+ typeof(p_ptr) __p = (p_ptr); \
+ unsigned long __v = (unsigned long)(*__p); \
+ *__p = (typeof(*p_ptr))(++__v); \
+} while (0)
+
#define page_mask_bits(ptr) ptr_mask_bits(ptr, PAGE_SHIFT)
#define page_unmask_bits(ptr) ptr_unmask_bits(ptr, PAGE_SHIFT)
#define page_pack_bits(ptr, bits) ptr_pack_bits(ptr, bits, PAGE_SHIFT)
diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c
index 240c8a2aabc2..32b336d5d897 100644
--- a/drivers/gpu/drm/i915/intel_guc_submission.c
+++ b/drivers/gpu/drm/i915/intel_guc_submission.c
@@ -32,7 +32,11 @@
#include "intel_guc_submission.h"
#include "i915_drv.h"
-#define GUC_PREEMPT_FINISHED 0x1
+enum {
+ GUC_PREEMPT_NONE = 0,
+ GUC_PREEMPT_INPROGRESS,
+ GUC_PREEMPT_FINISHED,
+};
#define GUC_PREEMPT_BREADCRUMB_DWORDS 0x8
#define GUC_PREEMPT_BREADCRUMB_BYTES \
(sizeof(u32) * GUC_PREEMPT_BREADCRUMB_DWORDS)
@@ -537,15 +541,11 @@ static void guc_add_request(struct intel_guc *guc, struct i915_request *rq)
u32 ctx_desc = lower_32_bits(rq->hw_context->lrc_desc);
u32 ring_tail = intel_ring_set_tail(rq->ring, rq->tail) / sizeof(u64);
- spin_lock(&client->wq_lock);
-
guc_wq_item_append(client, engine->guc_id, ctx_desc,
ring_tail, rq->fence.seqno);
guc_ring_doorbell(client);
client->submissions[engine->id] += 1;
-
- spin_unlock(&client->wq_lock);
}
/*
@@ -631,8 +631,9 @@ static void inject_preempt_context(struct work_struct *work)
data[6] = intel_guc_ggtt_offset(guc, guc->shared_data);
if (WARN_ON(intel_guc_send(guc, data, ARRAY_SIZE(data)))) {
- execlists_clear_active(&engine->execlists,
- EXECLISTS_ACTIVE_PREEMPT);
+ intel_write_status_page(engine,
+ I915_GEM_HWS_PREEMPT,
+ GUC_PREEMPT_NONE);
tasklet_schedule(&engine->execlists.tasklet);
}
@@ -672,8 +673,6 @@ static void complete_preempt_context(struct intel_engine_cs *engine)
{
struct intel_engine_execlists *execlists = &engine->execlists;
- GEM_BUG_ON(!execlists_is_active(execlists, EXECLISTS_ACTIVE_PREEMPT));
-
if (inject_preempt_hang(execlists))
return;
@@ -681,89 +680,90 @@ static void complete_preempt_context(struct intel_engine_cs *engine)
execlists_unwind_incomplete_requests(execlists);
wait_for_guc_preempt_report(engine);
- intel_write_status_page(engine, I915_GEM_HWS_PREEMPT, 0);
+ intel_write_status_page(engine, I915_GEM_HWS_PREEMPT, GUC_PREEMPT_NONE);
}
-/**
- * guc_submit() - Submit commands through GuC
- * @engine: engine associated with the commands
- *
- * The only error here arises if the doorbell hardware isn't functioning
- * as expected, which really shouln't happen.
- */
-static void guc_submit(struct intel_engine_cs *engine)
+static void guc_submit(struct intel_engine_cs *engine,
+ struct i915_request **out,
+ struct i915_request **end)
{
struct intel_guc *guc = &engine->i915->guc;
- struct intel_engine_execlists * const execlists = &engine->execlists;
- struct execlist_port *port = execlists->port;
- unsigned int n;
+ struct intel_guc_client *client = guc->execbuf_client;
- for (n = 0; n < execlists_num_ports(execlists); n++) {
- struct i915_request *rq;
- unsigned int count;
+ spin_lock(&client->wq_lock);
- rq = port_unpack(&port[n], &count);
- if (rq && count == 0) {
- port_set(&port[n], port_pack(rq, ++count));
+ do {
+ struct i915_request *rq = *out++;
- flush_ggtt_writes(rq->ring->vma);
+ flush_ggtt_writes(rq->ring->vma);
+ guc_add_request(guc, rq);
+ } while (out != end);
- guc_add_request(guc, rq);
- }
- }
+ spin_unlock(&client->wq_lock);
}
-static void port_assign(struct execlist_port *port, struct i915_request *rq)
+static inline int rq_prio(const struct i915_request *rq)
{
- GEM_BUG_ON(port_isset(port));
-
- port_set(port, i915_request_get(rq));
+ return rq->sched.attr.priority | __NO_PREEMPTION;
}
-static inline int rq_prio(const struct i915_request *rq)
+static struct i915_request *schedule_in(struct i915_request *rq, int idx)
{
- return rq->sched.attr.priority;
+ trace_i915_request_in(rq, idx);
+
+ if (!rq->hw_context->inflight)
+ rq->hw_context->inflight = rq->engine;
+ intel_context_inflight_inc(rq->hw_context);
+
+ return i915_request_get(rq);
}
-static inline int port_prio(const struct execlist_port *port)
+static void schedule_out(struct i915_request *rq)
{
- return rq_prio(port_request(port)) | __NO_PREEMPTION;
+ trace_i915_request_out(rq);
+
+ intel_context_inflight_dec(rq->hw_context);
+ if (!intel_context_inflight_count(rq->hw_context))
+ rq->hw_context->inflight = NULL;
+
+ i915_request_put(rq);
}
-static bool __guc_dequeue(struct intel_engine_cs *engine)
+static void __guc_dequeue(struct intel_engine_cs *engine)
{
struct intel_engine_execlists * const execlists = &engine->execlists;
- struct execlist_port *port = execlists->port;
- struct i915_request *last = NULL;
- const struct execlist_port * const last_port =
- &execlists->port[execlists->port_mask];
+ struct i915_request **first = execlists->inflight;
+ struct i915_request ** const last_port = first + execlists->port_mask;
+ struct i915_request *last = first[0];
+ struct i915_request **port;
bool submit = false;
struct rb_node *rb;
lockdep_assert_held(&engine->active.lock);
- if (port_isset(port)) {
+ if (last) {
if (intel_engine_has_preemption(engine)) {
struct guc_preempt_work *preempt_work =
&engine->i915->guc.preempt_work[engine->id];
int prio = execlists->queue_priority_hint;
- if (i915_scheduler_need_preempt(prio,
- port_prio(port))) {
- execlists_set_active(execlists,
- EXECLISTS_ACTIVE_PREEMPT);
+ if (i915_scheduler_need_preempt(prio, rq_prio(last))) {
+ intel_write_status_page(engine,
+ I915_GEM_HWS_PREEMPT,
+ GUC_PREEMPT_INPROGRESS);
queue_work(engine->i915->guc.preempt_wq,
&preempt_work->work);
- return false;
+ return;
}
}
- port++;
- if (port_isset(port))
- return false;
+ if (*++first)
+ return;
+
+ last = NULL;
}
- GEM_BUG_ON(port_isset(port));
+ port = first;
while ((rb = rb_first_cached(&execlists->queue))) {
struct i915_priolist *p = to_priolist(rb);
struct i915_request *rq, *rn;
@@ -774,18 +774,15 @@ static bool __guc_dequeue(struct intel_engine_cs *engine)
if (port == last_port)
goto done;
- if (submit)
- port_assign(port, last);
+ *port = schedule_in(last,
+ port - execlists->inflight);
port++;
}
list_del_init(&rq->sched.link);
-
__i915_request_submit(rq);
- trace_i915_request_in(rq, port_index(port, execlists));
-
- last = rq;
submit = true;
+ last = rq;
}
rb_erase_cached(&p->node, &execlists->queue);
@@ -794,58 +791,41 @@ static bool __guc_dequeue(struct intel_engine_cs *engine)
done:
execlists->queue_priority_hint =
rb ? to_priolist(rb)->priority : INT_MIN;
- if (submit)
- port_assign(port, last);
- if (last)
- execlists_user_begin(execlists, execlists->port);
-
- /* We must always keep the beast fed if we have work piled up */
- GEM_BUG_ON(port_isset(execlists->port) &&
- !execlists_is_active(execlists, EXECLISTS_ACTIVE_USER));
- GEM_BUG_ON(rb_first_cached(&execlists->queue) &&
- !port_isset(execlists->port));
-
- return submit;
-}
-
-static void guc_dequeue(struct intel_engine_cs *engine)
-{
- if (__guc_dequeue(engine))
- guc_submit(engine);
+ if (submit) {
+ *port = schedule_in(last, port - execlists->inflight);
+ *++port = NULL;
+ guc_submit(engine, first, port);
+ }
+ execlists->active = execlists->inflight;
}
static void guc_submission_tasklet(unsigned long data)
{
struct intel_engine_cs * const engine = (struct intel_engine_cs *)data;
struct intel_engine_execlists * const execlists = &engine->execlists;
- struct execlist_port *port = execlists->port;
- struct i915_request *rq;
+ struct i915_request **port, *rq;
unsigned long flags;
spin_lock_irqsave(&engine->active.lock, flags);
- rq = port_request(port);
- while (rq && i915_request_completed(rq)) {
- trace_i915_request_out(rq);
- i915_request_put(rq);
+ for (port = execlists->inflight; (rq = *port); port++) {
+ if (!i915_request_completed(rq))
+ break;
- port = execlists_port_complete(execlists, port);
- if (port_isset(port)) {
- execlists_user_begin(execlists, port);
- rq = port_request(port);
- } else {
- execlists_user_end(execlists);
- rq = NULL;
- }
+ schedule_out(rq);
+ }
+ if (port != execlists->inflight) {
+ int idx = port - execlists->inflight;
+ int rem = ARRAY_SIZE(execlists->inflight) - idx;
+ memmove(execlists->inflight, port, rem * sizeof(*port));
}
- if (execlists_is_active(execlists, EXECLISTS_ACTIVE_PREEMPT) &&
- intel_read_status_page(engine, I915_GEM_HWS_PREEMPT) ==
+ if (intel_read_status_page(engine, I915_GEM_HWS_PREEMPT) ==
GUC_PREEMPT_FINISHED)
complete_preempt_context(engine);
- if (!execlists_is_active(execlists, EXECLISTS_ACTIVE_PREEMPT))
- guc_dequeue(engine);
+ if (!intel_read_status_page(engine, I915_GEM_HWS_PREEMPT))
+ __guc_dequeue(engine);
spin_unlock_irqrestore(&engine->active.lock, flags);
}
@@ -959,7 +939,6 @@ static void guc_cancel_requests(struct intel_engine_cs *engine)
execlists->queue_priority_hint = INT_MIN;
execlists->queue = RB_ROOT_CACHED;
- GEM_BUG_ON(port_isset(execlists->port));
spin_unlock_irqrestore(&engine->active.lock, flags);
}
@@ -1422,7 +1401,7 @@ int intel_guc_submission_enable(struct intel_guc *guc)
* and it is guaranteed that it will remove the work item from the
* queue before our request is completed.
*/
- BUILD_BUG_ON(ARRAY_SIZE(engine->execlists.port) *
+ BUILD_BUG_ON(ARRAY_SIZE(engine->execlists.inflight) *
sizeof(struct guc_wq_item) *
I915_NUM_ENGINES > GUC_WQ_SIZE);
diff --git a/drivers/gpu/drm/i915/selftests/i915_request.c b/drivers/gpu/drm/i915/selftests/i915_request.c
index 693ea097c05f..218671b250d2 100644
--- a/drivers/gpu/drm/i915/selftests/i915_request.c
+++ b/drivers/gpu/drm/i915/selftests/i915_request.c
@@ -367,13 +367,15 @@ static int __igt_breadcrumbs_smoketest(void *arg)
if (!wait_event_timeout(wait->wait,
i915_sw_fence_done(wait),
- HZ / 2)) {
+ 5 * HZ)) {
struct i915_request *rq = requests[count - 1];
- pr_err("waiting for %d fences (last %llx:%lld) on %s timed out!\n",
- count,
+ pr_err("waiting for %d/%d fences (last %llx:%lld) on %s timed out!\n",
+ atomic_read(&wait->pending), count,
rq->fence.context, rq->fence.seqno,
t->engine->name);
+ GEM_TRACE_DUMP();
+
i915_gem_set_wedged(t->engine->i915);
GEM_BUG_ON(!i915_request_completed(rq));
i915_sw_fence_wait(wait);
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* [PATCH 40/40] drm/i915/execlists: Minimalistic timeslicing
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (37 preceding siblings ...)
2019-05-08 8:07 ` [PATCH 39/40] drm/i915/execlists: Preempt-to-busy Chris Wilson
@ 2019-05-08 8:07 ` Chris Wilson
2019-05-08 10:47 ` ✗ Fi.CI.CHECKPATCH: warning for series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Patchwork
` (6 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 8:07 UTC (permalink / raw)
To: intel-gfx
If we have multiple contexts of equal priority pending execution,
activate a timer to demote the currently executing context in favour of
the next in the queue when that timeslice expires. This enforces
fairness between contexts (so long as they allow preemption -- forced
preemption, in the future, will kick those who do not obey) and allows
us to avoid userspace blocking forward progress with e.g. unbounded
MI_SEMAPHORE_WAIT.
For the starting point here, we use the jiffie as our timeslice so that
we should be reasonably efficient wrt frequent CPU wakeups.
Testcase: igt/gem_exec_scheduler/semaphore-resolve
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/gt/intel_engine_types.h | 6 +
drivers/gpu/drm/i915/gt/intel_lrc.c | 111 +++++++++
drivers/gpu/drm/i915/gt/selftest_lrc.c | 223 +++++++++++++++++++
drivers/gpu/drm/i915/i915_scheduler.c | 1 +
drivers/gpu/drm/i915/i915_scheduler_types.h | 1 +
5 files changed, 342 insertions(+)
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
index de43bb62c24a..7f093f4b0af0 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
@@ -12,6 +12,7 @@
#include <linux/kref.h>
#include <linux/list.h>
#include <linux/llist.h>
+#include <linux/timer.h>
#include <linux/types.h>
#include "i915_gem.h"
@@ -137,6 +138,11 @@ struct intel_engine_execlists {
*/
struct tasklet_struct tasklet;
+ /**
+ * @timer: kick the current context if its timeslice expires
+ */
+ struct timer_list timer;
+
/**
* @default_priolist: priority list for I915_PRIORITY_NORMAL
*/
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index f08c2e223879..ea869f77a3ee 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -268,6 +268,7 @@ static int effective_prio(const struct i915_request *rq)
}
/* Restrict mere WAIT boosts from triggering preemption */
+ BUILD_BUG_ON(__NO_PREEMPTION & ~I915_PRIORITY_MASK); /* only internal */
return prio | __NO_PREEMPTION;
}
@@ -852,6 +853,81 @@ last_active(const struct intel_engine_execlists *execlists)
return *last;
}
+static void
+defer_request(struct i915_request * const rq, struct list_head * const pl)
+{
+ struct i915_dependency *p;
+
+ /*
+ * We want to move the interrupted request to the back of
+ * the round-robin list (i.e. its priority level), but
+ * in doing so, we must then move all requests that were in
+ * flight and were waiting for the interrupted request to
+ * be run after it again.
+ */
+ list_move_tail(&rq->sched.link, pl);
+
+ list_for_each_entry(p, &rq->sched.waiters_list, wait_link) {
+ struct i915_request *w =
+ container_of(p->waiter, typeof(*w), sched);
+
+ if (!i915_sw_fence_done(&w->submit))
+ continue;
+
+ /* Leave semaphores spinning on the other engines */
+ if (w->engine != rq->engine)
+ continue;
+
+ /* No waiter should start before the active request completed */
+ GEM_BUG_ON(i915_request_started(w));
+
+ GEM_BUG_ON(rq_prio(w) > rq_prio(rq));
+ if (rq_prio(w) < rq_prio(rq))
+ continue;
+
+ /*
+ * This should be very shallow as it is limited by the
+ * number of requests that can fit in a ring (<64) and
+ * the number of contexts that can be in flight on this
+ * engine.
+ */
+ defer_request(w, pl);
+ }
+}
+
+static void defer_active(struct intel_engine_cs *engine)
+{
+ struct i915_request *rq;
+
+ rq = __unwind_incomplete_requests(engine, 0);
+ if (!rq)
+ return;
+
+ defer_request(rq, i915_sched_lookup_priolist(engine, rq_prio(rq)));
+}
+
+static bool
+need_timeslice(struct intel_engine_cs *engine, const struct i915_request *rq)
+{
+ int hint;
+
+ if (list_is_last(&rq->sched.link, &engine->active.requests))
+ return false;
+
+ hint = max(rq_prio(list_next_entry(rq, sched.link)),
+ engine->execlists.queue_priority_hint);
+
+ return hint >= rq_prio(rq);
+}
+
+static bool
+enable_timeslice(struct intel_engine_cs *engine)
+{
+ struct i915_request *last = last_active(&engine->execlists);
+
+ return last && need_timeslice(engine, last);
+}
+
static bool execlists_dequeue(struct intel_engine_cs *engine)
{
struct intel_engine_execlists * const execlists = &engine->execlists;
@@ -945,6 +1021,27 @@ static bool execlists_dequeue(struct intel_engine_cs *engine)
*/
last->hw_context->lrc_desc |= CTX_DESC_FORCE_RESTORE;
last = NULL;
+ } else if (need_timeslice(engine, last) &&
+ !timer_pending(&engine->execlists.timer)) {
+ GEM_TRACE("%s: expired last=%llx:%lld, prio=%d, hint=%d\n",
+ engine->name,
+ last->fence.context,
+ last->fence.seqno,
+ last->sched.attr.priority,
+ execlists->queue_priority_hint);
+
+ ring_suspend(engine) = 1;
+ defer_active(engine);
+
+ /*
+ * Unlike for preemption, if we rewind and continue
+ * executing the same context as previously active,
+ * the order of execution will remain the same and
+ * the tail will only advance. We do not need to
+ * force a full context restore, as a lite-restore
+ * is sufficient to resample the monotonic TAIL.
+ */
+ last = NULL;
} else {
/*
* Otherwise if we already have a request pending
@@ -1269,6 +1366,9 @@ static void process_csb(struct intel_engine_cs *engine)
sizeof(*execlists->pending));
execlists->pending[0] = NULL;
+ if (enable_timeslice(engine))
+ mod_timer(&execlists->timer, jiffies + 1);
+
if (!inject_preempt_hang(execlists))
ring_suspend(engine) = 0;
} else if (status & GEN8_CTX_STATUS_PREEMPTED) {
@@ -1345,6 +1445,15 @@ static void execlists_submission_tasklet(unsigned long data)
spin_unlock_irqrestore(&engine->active.lock, flags);
}
+static void execlists_submission_timer(struct timer_list *timer)
+{
+ struct intel_engine_cs *engine =
+ from_timer(engine, timer, execlists.timer);
+
+ /* Kick the tasklet for some interrupt coalescing and reset handling */
+ tasklet_hi_schedule(&engine->execlists.tasklet);
+}
+
static void queue_request(struct intel_engine_cs *engine,
struct i915_sched_node *node,
int prio)
@@ -2573,6 +2682,7 @@ static int gen8_init_rcs_context(struct i915_request *rq)
static void execlists_park(struct intel_engine_cs *engine)
{
+ del_timer_sync(&engine->execlists.timer);
intel_engine_park(engine);
}
@@ -2670,6 +2780,7 @@ int intel_execlists_submission_setup(struct intel_engine_cs *engine)
tasklet_init(&engine->execlists.tasklet,
execlists_submission_tasklet, (unsigned long)engine);
+ timer_setup(&engine->execlists.timer, execlists_submission_timer, 0);
logical_ring_default_vfuncs(engine);
logical_ring_default_irqs(engine);
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index c96cd69f9800..3641123fb7e3 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -79,6 +79,225 @@ static int live_sanitycheck(void *arg)
return err;
}
+static int
+emit_semaphore_chain(struct i915_request *rq, struct i915_vma *vma, int idx)
+{
+ u32 *cs;
+
+ cs = intel_ring_begin(rq, 10);
+ if (IS_ERR(cs))
+ return PTR_ERR(cs);
+
+ *cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
+
+ *cs++ = MI_SEMAPHORE_WAIT |
+ MI_SEMAPHORE_GLOBAL_GTT |
+ MI_SEMAPHORE_POLL |
+ MI_SEMAPHORE_SAD_NEQ_SDD;
+ *cs++ = 0;
+ *cs++ = i915_ggtt_offset(vma) + 4 * idx;
+ *cs++ = 0;
+
+ if (idx > 0) {
+ *cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT;
+ *cs++ = i915_ggtt_offset(vma) + 4 * (idx - 1);
+ *cs++ = 0;
+ *cs++ = 1;
+ } else {
+ *cs++ = MI_NOOP;
+ *cs++ = MI_NOOP;
+ *cs++ = MI_NOOP;
+ *cs++ = MI_NOOP;
+ }
+
+ *cs++ = MI_ARB_ON_OFF | MI_ARB_DISABLE;
+
+ intel_ring_advance(rq, cs);
+ return 0;
+}
+
+static struct i915_request *
+semaphore_queue(struct intel_engine_cs *engine, struct i915_vma *vma, int idx)
+{
+ struct i915_gem_context *ctx;
+ struct i915_request *rq;
+ int err;
+
+ ctx = kernel_context(engine->i915);
+ if (!ctx)
+ return ERR_PTR(-ENOMEM);
+
+ rq = igt_request_alloc(ctx, engine);
+ if (IS_ERR(rq))
+ goto out_ctx;
+
+ err = emit_semaphore_chain(rq, vma, idx);
+ i915_request_add(rq);
+ if (err)
+ rq = ERR_PTR(err);
+
+out_ctx:
+ kernel_context_close(ctx);
+ return rq;
+}
+
+static int
+release_queue(struct intel_engine_cs *engine,
+ struct i915_vma *vma,
+ int idx)
+{
+ struct i915_sched_attr attr = {
+ .priority = I915_USER_PRIORITY(I915_PRIORITY_MAX),
+ };
+ struct i915_request *rq;
+ u32 *cs;
+
+ rq = i915_request_create(engine->kernel_context);
+ if (IS_ERR(rq))
+ return PTR_ERR(rq);
+
+ cs = intel_ring_begin(rq, 4);
+ if (IS_ERR(cs)) {
+ i915_request_add(rq);
+ return PTR_ERR(cs);
+ }
+
+ *cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT;
+ *cs++ = i915_ggtt_offset(vma) + 4 * (idx - 1);
+ *cs++ = 0;
+ *cs++ = 1;
+
+ intel_ring_advance(rq, cs);
+ i915_request_add(rq);
+
+ engine->schedule(rq, &attr);
+
+ return 0;
+}
+
+static int
+slice_semaphore_queue(struct intel_engine_cs *outer,
+ struct i915_vma *vma,
+ int count)
+{
+ struct intel_engine_cs *engine;
+ struct i915_request *head;
+ enum intel_engine_id id;
+ int err, i, n = 0;
+
+ head = semaphore_queue(outer, vma, n++);
+ if (IS_ERR(head))
+ return PTR_ERR(head);
+
+ i915_request_get(head);
+ for_each_engine(engine, outer->i915, id) {
+ for (i = 0; i < count; i++) {
+ struct i915_request *rq;
+
+ rq = semaphore_queue(engine, vma, n++);
+ if (IS_ERR(rq)) {
+ err = PTR_ERR(rq);
+ goto out;
+ }
+ }
+ }
+
+ err = release_queue(outer, vma, n);
+ if (err)
+ goto out;
+
+ if (i915_request_wait(head,
+ I915_WAIT_LOCKED,
+ 2 * RUNTIME_INFO(outer->i915)->num_engines * (count + 2) * (count + 3)) < 0) {
+ pr_err("Failed to slice along semaphore chain of length (%d, %d)!\n",
+ count, n);
+ GEM_TRACE_DUMP();
+ i915_gem_set_wedged(outer->i915);
+ err = -EIO;
+ }
+
+out:
+ i915_request_put(head);
+ return err;
+}
+
+static int live_timeslice_preempt(void *arg)
+{
+ struct drm_i915_private *i915 = arg;
+ struct drm_i915_gem_object *obj;
+ intel_wakeref_t wakeref;
+ struct i915_vma *vma;
+ void *vaddr;
+ int err = 0;
+ int count;
+
+ /*
+ * If a request takes too long, we would like to give other users
+ * a fair go on the GPU. In particular, users may create batches
+ * that wait upon external input, where that input may even be
+ * supplied by another GPU job. To avoid blocking forever, we
+ * need to preempt the current task and replace it with another
+ * ready task.
+ */
+
+ mutex_lock(&i915->drm.struct_mutex);
+ wakeref = intel_runtime_pm_get(i915);
+
+ obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
+ if (IS_ERR(obj)) {
+ err = PTR_ERR(obj);
+ goto err_unlock;
+ }
+
+ vma = i915_vma_instance(obj, &i915->ggtt.vm, NULL);
+ if (IS_ERR(vma)) {
+ err = PTR_ERR(vma);
+ goto err_obj;
+ }
+
+ vaddr = i915_gem_object_pin_map(obj, I915_MAP_WC);
+ if (IS_ERR(vaddr)) {
+ err = PTR_ERR(vaddr);
+ goto err_obj;
+ }
+
+ err = i915_vma_pin(vma, 0, 0, PIN_GLOBAL);
+ if (err)
+ goto err_map;
+
+ for_each_prime_number_from(count, 1, 16) {
+ struct intel_engine_cs *engine;
+ enum intel_engine_id id;
+
+ for_each_engine(engine, i915, id) {
+ memset(vaddr, 0, PAGE_SIZE);
+
+ err = slice_semaphore_queue(engine, vma, count);
+ if (err)
+ goto err_pin;
+
+ if (igt_flush_test(i915, I915_WAIT_LOCKED)) {
+ err = -EIO;
+ goto err_pin;
+ }
+ }
+ }
+
+err_pin:
+ i915_vma_unpin(vma);
+err_map:
+ i915_gem_object_unpin_map(obj);
+err_obj:
+ i915_gem_object_put(obj);
+err_unlock:
+ if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ err = -EIO;
+ intel_runtime_pm_put(i915, wakeref);
+ mutex_unlock(&i915->drm.struct_mutex);
+
+ return err;
+}
+
static int live_busywait_preempt(void *arg)
{
struct drm_i915_private *i915 = arg;
@@ -398,6 +617,9 @@ static int live_late_preempt(void *arg)
if (!ctx_lo)
goto err_ctx_hi;
+ /* Make sure ctx_lo stays before ctx_hi until we trigger preemption. */
+ ctx_lo->sched.priority = I915_USER_PRIORITY(1);
+
for_each_engine(engine, i915, id) {
struct igt_live_test t;
struct i915_request *rq;
@@ -1818,6 +2040,7 @@ int intel_execlists_live_selftests(struct drm_i915_private *i915)
{
static const struct i915_subtest tests[] = {
SUBTEST(live_sanitycheck),
+ SUBTEST(live_timeslice_preempt),
SUBTEST(live_busywait_preempt),
SUBTEST(live_preempt),
SUBTEST(live_late_preempt),
diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c
index 7aa892ff6550..eb91d2d94ee6 100644
--- a/drivers/gpu/drm/i915/i915_scheduler.c
+++ b/drivers/gpu/drm/i915/i915_scheduler.c
@@ -394,6 +394,7 @@ bool __i915_sched_node_add_dependency(struct i915_sched_node *node,
list_add(&dep->wait_link, &signal->waiters_list);
list_add(&dep->signal_link, &node->signalers_list);
dep->signaler = signal;
+ dep->waiter = node;
dep->flags = flags;
/* Keep track of whether anyone on this chain has a semaphore */
diff --git a/drivers/gpu/drm/i915/i915_scheduler_types.h b/drivers/gpu/drm/i915/i915_scheduler_types.h
index 3e309631bd0b..aad81acba9dc 100644
--- a/drivers/gpu/drm/i915/i915_scheduler_types.h
+++ b/drivers/gpu/drm/i915/i915_scheduler_types.h
@@ -62,6 +62,7 @@ struct i915_sched_node {
struct i915_dependency {
struct i915_sched_node *signaler;
+ struct i915_sched_node *waiter;
struct list_head signal_link;
struct list_head wait_link;
struct list_head dfs_link;
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 71+ messages in thread* ✗ Fi.CI.CHECKPATCH: warning for series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (38 preceding siblings ...)
2019-05-08 8:07 ` [PATCH 40/40] drm/i915/execlists: Minimalistic timeslicing Chris Wilson
@ 2019-05-08 10:47 ` Patchwork
2019-05-08 10:59 ` ✗ Fi.CI.SPARSE: " Patchwork
` (5 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Patchwork @ 2019-05-08 10:47 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
URL : https://patchwork.freedesktop.org/series/60403/
State : warning
== Summary ==
$ dim checkpatch origin/drm-tip
67b421614964 drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
5f78625e4a23 drm/i915: Rearrange i915_scheduler.c
fea1b40e0ff4 drm/i915: Pass i915_sched_node around internally
fad4daf02033 drm/i915: Check for no-op priority changes first
4c36ecd8e737 drm/i915: Bump signaler priority on adding a waiter
483a8a8f05cf drm/i915: Convert inconsistent static engine tables into an init error
0f169a3e02f6 drm/i915: Seal races between async GPU cancellation, retirement and signaling
eb869674ac86 dma-fence: Refactor signaling for manual invocation
-:30: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#30:
new file mode 100644
-:35: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#35: FILE: drivers/dma-buf/dma-fence-trace.c:1:
+/*
-:173: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#173: FILE: include/linux/dma-fence-types.h:1:
+/*
-:244: CHECK:UNCOMMENTED_DEFINITION: spinlock_t definition without comment
#244: FILE: include/linux/dma-fence-types.h:72:
+ spinlock_t *lock;
total: 0 errors, 3 warnings, 1 checks, 641 lines checked
1415c169395e drm/i915: Restore control over ppgtt for context creation ABI
-:81: WARNING:LONG_LINE: line over 100 characters
#81: FILE: include/uapi/drm/i915_drm.h:420:
+#define DRM_IOCTL_I915_GEM_VM_CREATE DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_VM_CREATE, struct drm_i915_gem_vm_control)
-:82: WARNING:LONG_LINE: line over 100 characters
#82: FILE: include/uapi/drm/i915_drm.h:421:
+#define DRM_IOCTL_I915_GEM_VM_DESTROY DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_VM_DESTROY, struct drm_i915_gem_vm_control)
-:82: WARNING:SPACING: space prohibited between function name and open parenthesis '('
#82: FILE: include/uapi/drm/i915_drm.h:421:
+#define DRM_IOCTL_I915_GEM_VM_DESTROY DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_VM_DESTROY, struct drm_i915_gem_vm_control)
-:82: ERROR:COMPLEX_MACRO: Macros with complex values should be enclosed in parentheses
#82: FILE: include/uapi/drm/i915_drm.h:421:
+#define DRM_IOCTL_I915_GEM_VM_DESTROY DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_VM_DESTROY, struct drm_i915_gem_vm_control)
total: 1 errors, 3 warnings, 0 checks, 64 lines checked
d257fe3e40a0 drm/i915: Allow a context to define its set of engines
-:437: CHECK:MACRO_ARG_REUSE: Macro argument reuse 'p' - possible side-effects?
#437: FILE: drivers/gpu/drm/i915/i915_utils.h:110:
+#define check_struct_size(p, member, n, sz) \
+ likely(__check_struct_size(sizeof(*(p)), \
+ sizeof(*(p)->member) + __must_be_array((p)->member), \
+ n, sz))
-:437: CHECK:MACRO_ARG_REUSE: Macro argument reuse 'member' - possible side-effects?
#437: FILE: drivers/gpu/drm/i915/i915_utils.h:110:
+#define check_struct_size(p, member, n, sz) \
+ likely(__check_struct_size(sizeof(*(p)), \
+ sizeof(*(p)->member) + __must_be_array((p)->member), \
+ n, sz))
-:437: CHECK:MACRO_ARG_PRECEDENCE: Macro argument 'member' may be better as '(member)' to avoid precedence issues
#437: FILE: drivers/gpu/drm/i915/i915_utils.h:110:
+#define check_struct_size(p, member, n, sz) \
+ likely(__check_struct_size(sizeof(*(p)), \
+ sizeof(*(p)->member) + __must_be_array((p)->member), \
+ n, sz))
total: 0 errors, 0 warnings, 3 checks, 428 lines checked
7af4301e09f7 drm/i915: Extend I915_CONTEXT_PARAM_SSEU to support local ctx->engine[]
5bbcdc2dae1c drm/i915: Re-expose SINGLE_TIMELINE flags for context creation
fc7ed4a4ef81 drm/i915: Allow userspace to clone contexts on creation
-:213: ERROR:BRACKET_SPACE: space prohibited before open square bracket '['
#213: FILE: drivers/gpu/drm/i915/i915_gem_context.c:1858:
+#define MAP(x, y) [ilog2(I915_CONTEXT_CLONE_##x)] = y
total: 1 errors, 0 warnings, 0 checks, 235 lines checked
69947c56c90e drm/i915: Load balancing across a virtual engine
266984fb2212 drm/i915: Apply an execution_mask to the virtual_engine
f0417a49d538 drm/i915: Extend execution fence to support a callback
7afded600522 drm/i915/execlists: Virtual engine bonding
00c6c2e19f78 drm/i915: Allow specification of parallel execbuf
b688a0332163 drm/i915: Split GEM object type definition to its own header
-:25: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#25:
new file mode 100644
-:59: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#59: FILE: drivers/gpu/drm/i915/gem/i915_gem_object_types.h:1:
+/*
-:60: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#60: FILE: drivers/gpu/drm/i915/gem/i915_gem_object_types.h:2:
+ * SPDX-License-Identifier: MIT
-:177: WARNING:LINE_SPACING: Missing a blank line after declarations
#177: FILE: drivers/gpu/drm/i915/gem/i915_gem_object_types.h:119:
+ struct list_head batch_pool_link;
+ I915_SELFTEST_DECLARE(struct list_head st_link);
-:436: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#436: FILE: drivers/gpu/drm/i915/i915_gem_object.h:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 5 warnings, 0 checks, 664 lines checked
39f76472714f drm/i915: Pull GEM ioctls interface to its own file
-:12: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#12:
new file mode 100644
-:17: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#17: FILE: drivers/gpu/drm/i915/gem/i915_gem_ioctls.h:1:
+/*
-:18: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#18: FILE: drivers/gpu/drm/i915/gem/i915_gem_ioctls.h:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 3 warnings, 0 checks, 152 lines checked
3483581bdd02 drm/i915: Move object->pages API to i915_gem_object.[ch]
-:38: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#38:
rename from drivers/gpu/drm/i915/i915_gem_object.c
total: 0 errors, 1 warnings, 0 checks, 339 lines checked
fb33bd9448ad drm/i915: Move shmem object setup to its own file
-:413: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#413:
new file mode 100644
-:418: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#418: FILE: drivers/gpu/drm/i915/gem/i915_gem_shmem.c:1:
+/*
-:419: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#419: FILE: drivers/gpu/drm/i915/gem/i915_gem_shmem.c:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 3 warnings, 0 checks, 2003 lines checked
35b7895848a6 drm/i915: Move phys objects to its own file
-:80: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#80:
new file mode 100644
-:85: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#85: FILE: drivers/gpu/drm/i915/gem/i915_gem_pages.c:1:
+/*
-:86: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#86: FILE: drivers/gpu/drm/i915/gem/i915_gem_pages.c:2:
+ * SPDX-License-Identifier: MIT
-:185: WARNING:MEMORY_BARRIER: memory barrier without comment
#185: FILE: drivers/gpu/drm/i915/gem/i915_gem_pages.c:101:
+ smp_mb__before_atomic();
-:250: CHECK:MULTIPLE_ASSIGNMENTS: multiple assignments should be avoided
#250: FILE: drivers/gpu/drm/i915/gem/i915_gem_pages.c:166:
+ obj->mm.page_sizes.phys = obj->mm.page_sizes.sg = 0;
-:376: WARNING:MEMORY_BARRIER: memory barrier without comment
#376: FILE: drivers/gpu/drm/i915/gem/i915_gem_pages.c:292:
+ smp_mb__before_atomic();
-:395: CHECK:MULTIPLE_ASSIGNMENTS: multiple assignments should be avoided
#395: FILE: drivers/gpu/drm/i915/gem/i915_gem_pages.c:311:
+ ptr = obj->mm.mapping = NULL;
-:601: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#601: FILE: drivers/gpu/drm/i915/gem/i915_gem_phys.c:1:
+/*
-:602: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#602: FILE: drivers/gpu/drm/i915/gem/i915_gem_phys.c:2:
+ * SPDX-License-Identifier: MIT
-:898: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#898: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c:1:
+/*
-:899: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#899: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 9 warnings, 2 checks, 1825 lines checked
4734074d136c drm/i915: Move mmap and friends to its own file
-:25: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#25:
new file mode 100644
-:30: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#30: FILE: drivers/gpu/drm/i915/gem/i915_gem_mman.c:1:
+/*
-:31: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#31: FILE: drivers/gpu/drm/i915/gem/i915_gem_mman.c:2:
+ * SPDX-License-Identifier: MIT
-:590: WARNING:MEMORY_BARRIER: memory barrier without comment
#590: FILE: drivers/gpu/drm/i915/gem/i915_gem_object.c:400:
+ wmb();
-:633: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#633: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c:1:
+/*
-:634: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#634: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c:2:
+ * SPDX-License-Identifier: MIT
-:827: WARNING:LONG_LINE: line over 100 characters
#827: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c:195:
+ (1 + next_prime_number(i915->ggtt.vm.total >> PAGE_SHIFT)) << PAGE_SHIFT);
-:946: WARNING:SPACING: space prohibited before semicolon
#946: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c:314:
+next_tiling: ;
total: 0 errors, 8 warnings, 0 checks, 2256 lines checked
1786d7cb9bc6 drm/i915: Move GEM domain management to its own file
-:25: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#25:
new file mode 100644
-:30: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#30: FILE: drivers/gpu/drm/i915/gem/i915_gem_domain.c:1:
+/*
-:31: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#31: FILE: drivers/gpu/drm/i915/gem/i915_gem_domain.c:2:
+ * SPDX-License-Identifier: MIT
-:113: WARNING:MEMORY_BARRIER: memory barrier without comment
#113: FILE: drivers/gpu/drm/i915/gem/i915_gem_domain.c:84:
+ mb();
-:175: WARNING:MEMORY_BARRIER: memory barrier without comment
#175: FILE: drivers/gpu/drm/i915/gem/i915_gem_domain.c:146:
+ mb();
total: 0 errors, 5 warnings, 0 checks, 1871 lines checked
8be66d73a04c drm/i915: Move more GEM objects under gem/
-:87: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#87:
rename from drivers/gpu/drm/i915/i915_gem_clflush.c
-:114: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#114: FILE: drivers/gpu/drm/i915/gem/i915_gem_clflush.c:2:
+ * SPDX-License-Identifier: MIT
-:135: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#135: FILE: drivers/gpu/drm/i915/gem/i915_gem_clflush.h:1:
+/*
-:136: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#136: FILE: drivers/gpu/drm/i915/gem/i915_gem_clflush.h:2:
+ * SPDX-License-Identifier: MIT
-:187: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#187: FILE: drivers/gpu/drm/i915/gem/i915_gem_context.c:2:
+ * SPDX-License-Identifier: MIT
-:231: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#231: FILE: drivers/gpu/drm/i915/gem/i915_gem_context.h:2:
+ * SPDX-License-Identifier: MIT
-:262: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#262: FILE: drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c:2:
+ * SPDX-License-Identifier: MIT
-:338: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#338: FILE: drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c:2:
+ * SPDX-License-Identifier: MIT
-:394: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#394: FILE: drivers/gpu/drm/i915/gem/i915_gem_internal.c:2:
+ * SPDX-License-Identifier: MIT
-:495: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#495: FILE: drivers/gpu/drm/i915/gem/i915_gem_shrinker.c:2:
+ * SPDX-License-Identifier: MIT
-:544: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#544: FILE: drivers/gpu/drm/i915/gem/i915_gem_stolen.c:2:
+ * SPDX-License-Identifier: MIT
-:592: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#592: FILE: drivers/gpu/drm/i915/gem/i915_gem_tiling.c:2:
+ * SPDX-License-Identifier: MIT
-:640: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#640: FILE: drivers/gpu/drm/i915/gem/i915_gem_userptr.c:2:
+ * SPDX-License-Identifier: MIT
-:691: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#691: FILE: drivers/gpu/drm/i915/gem/i915_gemfs.c:2:
+ * SPDX-License-Identifier: MIT
-:714: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#714: FILE: drivers/gpu/drm/i915/gem/i915_gemfs.h:1:
+/*
-:715: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#715: FILE: drivers/gpu/drm/i915/gem/i915_gemfs.h:2:
+ * SPDX-License-Identifier: MIT
-:759: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#759: FILE: drivers/gpu/drm/i915/gem/selftests/huge_gem_object.c:2:
+ * SPDX-License-Identifier: MIT
-:771: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#771: FILE: drivers/gpu/drm/i915/gem/selftests/huge_gem_object.h:1:
+/*
-:772: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#772: FILE: drivers/gpu/drm/i915/gem/selftests/huge_gem_object.h:2:
+ * SPDX-License-Identifier: MIT
-:827: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#827: FILE: drivers/gpu/drm/i915/gem/selftests/huge_pages.c:2:
+ * SPDX-License-Identifier: MIT
-:879: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#879: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c:2:
+ * SPDX-License-Identifier: MIT
-:922: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#922: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c:2:
+ * SPDX-License-Identifier: MIT
-:984: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#984: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_dmabuf.c:2:
+ * SPDX-License-Identifier: MIT
-:1040: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#1040: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_object.c:2:
+ * SPDX-License-Identifier: MIT
-:1108: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#1108: FILE: drivers/gpu/drm/i915/gem/selftests/mock_context.c:2:
+ * SPDX-License-Identifier: MIT
-:1125: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#1125: FILE: drivers/gpu/drm/i915/gem/selftests/mock_context.h:1:
+/*
-:1126: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#1126: FILE: drivers/gpu/drm/i915/gem/selftests/mock_context.h:2:
+ * SPDX-License-Identifier: MIT
-:1178: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#1178: FILE: drivers/gpu/drm/i915/gem/selftests/mock_dmabuf.c:2:
+ * SPDX-License-Identifier: MIT
-:1190: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#1190: FILE: drivers/gpu/drm/i915/gem/selftests/mock_dmabuf.h:1:
+/*
-:1191: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#1191: FILE: drivers/gpu/drm/i915/gem/selftests/mock_dmabuf.h:2:
+ * SPDX-License-Identifier: MIT
-:1221: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#1221: FILE: drivers/gpu/drm/i915/gem/selftests/mock_gem_object.h:1:
+/*
-:1222: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#1222: FILE: drivers/gpu/drm/i915/gem/selftests/mock_gem_object.h:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 32 warnings, 0 checks, 1405 lines checked
e95a11a1ae00 drm/i915: Pull scatterlist utils out of i915_gem.h
-:372: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#372:
new file mode 100644
-:377: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#377: FILE: drivers/gpu/drm/i915/i915_scatterlist.c:1:
+/*
-:378: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#378: FILE: drivers/gpu/drm/i915/i915_scatterlist.c:2:
+ * SPDX-License-Identifier: MIT
-:422: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#422: FILE: drivers/gpu/drm/i915/i915_scatterlist.h:1:
+/*
-:423: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#423: FILE: drivers/gpu/drm/i915/i915_scatterlist.h:2:
+ * SPDX-License-Identifier: MIT
-:452: CHECK:MULTIPLE_ASSIGNMENTS: multiple assignments should be avoided
#452: FILE: drivers/gpu/drm/i915/i915_scatterlist.h:31:
+ s.max = s.curr = s.sgp->offset;
-:497: CHECK:MACRO_ARG_REUSE: Macro argument reuse '__iter' - possible side-effects?
#497: FILE: drivers/gpu/drm/i915/i915_scatterlist.h:76:
+#define __for_each_sgt_dma(__dmap, __iter, __sgt, __step) \
+ for ((__iter) = __sgt_iter((__sgt)->sgl, true); \
+ ((__dmap) = (__iter).dma + (__iter).curr); \
+ (((__iter).curr += (__step)) >= (__iter).max) ? \
+ (__iter) = __sgt_iter(__sg_next((__iter).sgp), true), 0 : 0)
-:509: CHECK:MACRO_ARG_REUSE: Macro argument reuse '__iter' - possible side-effects?
#509: FILE: drivers/gpu/drm/i915/i915_scatterlist.h:88:
+#define for_each_sgt_page(__pp, __iter, __sgt) \
+ for ((__iter) = __sgt_iter((__sgt)->sgl, false); \
+ ((__pp) = (__iter).pfn == 0 ? NULL : \
+ pfn_to_page((__iter).pfn + ((__iter).curr >> PAGE_SHIFT))); \
+ (((__iter).curr += PAGE_SIZE) >= (__iter).max) ? \
+ (__iter) = __sgt_iter(__sg_next((__iter).sgp), false), 0 : 0)
total: 0 errors, 5 warnings, 3 checks, 457 lines checked
82fa4ffba0ac drm/i915: Move GEM object domain management from struct_mutex to local
-:525: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#525:
new file mode 100644
-:530: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#530: FILE: drivers/gpu/drm/i915/gem/i915_gem_fence.c:1:
+/*
-:531: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#531: FILE: drivers/gpu/drm/i915/gem/i915_gem_fence.c:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 3 warnings, 0 checks, 1438 lines checked
37093e2cfcb7 drm/i915: Move GEM object waiting to its own file
-:41: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#41:
new file mode 100644
-:46: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#46: FILE: drivers/gpu/drm/i915/gem/i915_gem_wait.c:1:
+/*
-:47: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#47: FILE: drivers/gpu/drm/i915/gem/i915_gem_wait.c:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 3 warnings, 0 checks, 591 lines checked
3853ef7d5efd drm/i915: Move GEM object busy checking to its
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread* ✗ Fi.CI.SPARSE: warning for series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (39 preceding siblings ...)
2019-05-08 10:47 ` ✗ Fi.CI.CHECKPATCH: warning for series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Patchwork
@ 2019-05-08 10:59 ` Patchwork
2019-05-08 11:09 ` ✓ Fi.CI.BAT: success " Patchwork
` (4 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Patchwork @ 2019-05-08 10:59 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
URL : https://patchwork.freedesktop.org/series/60403/
State : warning
== Summary ==
$ dim sparse origin/drm-tip
Sparse version: v0.5.2
Commit: drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
Okay!
Commit: drm/i915: Rearrange i915_scheduler.c
Okay!
Commit: drm/i915: Pass i915_sched_node around internally
Okay!
Commit: drm/i915: Check for no-op priority changes first
Okay!
Commit: drm/i915: Bump signaler priority on adding a waiter
Okay!
Commit: drm/i915: Convert inconsistent static engine tables into an init error
Okay!
Commit: drm/i915: Seal races between async GPU cancellation, retirement and signaling
Okay!
Commit: dma-fence: Refactor signaling for manual invocation
Okay!
Commit: drm/i915: Restore control over ppgtt for context creation ABI
Okay!
Commit: drm/i915: Allow a context to define its set of engines
+drivers/gpu/drm/i915/i915_utils.h:87:13: error: incorrect type in conditional
+drivers/gpu/drm/i915/i915_utils.h:87:13: error: undefined identifier '__builtin_mul_overflow'
+drivers/gpu/drm/i915/i915_utils.h:87:13: got void
+drivers/gpu/drm/i915/i915_utils.h:87:13: warning: call with no type!
+drivers/gpu/drm/i915/i915_utils.h:90:13: error: incorrect type in conditional
+drivers/gpu/drm/i915/i915_utils.h:90:13: error: undefined identifier '__builtin_add_overflow'
+drivers/gpu/drm/i915/i915_utils.h:90:13: got void
+drivers/gpu/drm/i915/i915_utils.h:90:13: warning: call with no type!
-drivers/gpu/drm/i915/selftests/../i915_utils.h:184:16: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/selftests/../i915_utils.h:218:16: warning: expression using sizeof(void)
+./include/linux/overflow.h:285:13: error: incorrect type in conditional
+./include/linux/overflow.h:285:13: error: not a function <noident>
+./include/linux/overflow.h:285:13: got void
+./include/linux/overflow.h:287:13: error: incorrect type in conditional
+./include/linux/overflow.h:287:13: error: not a function <noident>
+./include/linux/overflow.h:287:13: got void
Commit: drm/i915: Extend I915_CONTEXT_PARAM_SSEU to support local ctx->engine[]
Okay!
Commit: drm/i915: Re-expose SINGLE_TIMELINE flags for context creation
Okay!
Commit: drm/i915: Allow userspace to clone contexts on creation
+drivers/gpu/drm/i915/i915_gem_context.c:1859:17: error: bad integer constant expression
+drivers/gpu/drm/i915/i915_gem_context.c:1860:17: error: bad integer constant expression
+drivers/gpu/drm/i915/i915_gem_context.c:1861:17: error: bad integer constant expression
+drivers/gpu/drm/i915/i915_gem_context.c:1862:17: error: bad integer constant expression
+drivers/gpu/drm/i915/i915_gem_context.c:1863:17: error: bad integer constant expression
+drivers/gpu/drm/i915/i915_gem_context.c:1864:17: error: bad integer constant expression
-drivers/gpu/drm/i915/i915_utils.h:87:13: warning: call with no type!
-drivers/gpu/drm/i915/i915_utils.h:90:13: warning: call with no type!
-drivers/gpu/drm/i915/selftests/i915_gem_context.c:1266:25: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/selftests/i915_gem_context.c:1266:25: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/selftests/i915_gem_context.c:454:16: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/selftests/i915_gem_context.c:571:33: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/selftests/i915_gem_context.c:571:33: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/selftests/i915_gem_context.c:693:33: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/selftests/i915_gem_context.c:693:33: warning: expression using sizeof(void)
+./include/linux/overflow.h:285:13: error: incorrect type in conditional
+./include/linux/overflow.h:285:13: error: not a function <noident>
-./include/linux/overflow.h:285:13: warning: call with no type!
+./include/linux/overflow.h:285:13: got void
+./include/linux/overflow.h:287:13: error: incorrect type in conditional
+./include/linux/overflow.h:287:13: error: not a function <noident>
-./include/linux/overflow.h:287:13: warning: call with no type!
+./include/linux/overflow.h:287:13: got void
-./include/linux/slab.h:666:13: warning: call with no type!
Commit: drm/i915: Load balancing across a virtual engine
+./include/linux/overflow.h:285:13: error: incorrect type in conditional
+./include/linux/overflow.h:285:13: error: undefined identifier '__builtin_mul_overflow'
+./include/linux/overflow.h:285:13: got void
+./include/linux/overflow.h:285:13: warning: call with no type!
+./include/linux/overflow.h:287:13: error: incorrect type in conditional
+./include/linux/overflow.h:287:13: error: undefined identifier '__builtin_add_overflow'
+./include/linux/overflow.h:287:13: got void
+./include/linux/overflow.h:287:13: warning: call with no type!
+./include/linux/slab.h:666:13: error: not a function <noident>
Commit: drm/i915: Apply an execution_mask to the virtual_engine
Okay!
Commit: drm/i915: Extend execution fence to support a callback
Okay!
Commit: drm/i915/execlists: Virtual engine bonding
Okay!
Commit: drm/i915: Allow specification of parallel execbuf
Okay!
Commit: drm/i915: Split GEM object type definition to its own header
+./include/uapi/linux/perf_event.h:147:56: warning: cast truncates bits from constant value (8000000000000000 becomes 0)
Commit: drm/i915: Pull GEM ioctls interface to its own file
Okay!
Commit: drm/i915: Move object->pages API to i915_gem_object.[ch]
+./include/uapi/linux/perf_event.h:147:56: warning: cast truncates bits from constant value (8000000000000000 becomes 0)
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread* ✓ Fi.CI.BAT: success for series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (40 preceding siblings ...)
2019-05-08 10:59 ` ✗ Fi.CI.SPARSE: " Patchwork
@ 2019-05-08 11:09 ` Patchwork
2019-05-08 11:49 ` ✗ Fi.CI.CHECKPATCH: warning for series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD (rev4) Patchwork
` (3 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Patchwork @ 2019-05-08 11:09 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
URL : https://patchwork.freedesktop.org/series/60403/
State : success
== Summary ==
CI Bug Log - changes from CI_DRM_6063 -> Patchwork_12984
====================================================
Summary
-------
**SUCCESS**
No regressions found.
External URL: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_12984/
New tests
---------
New tests have been introduced between CI_DRM_6063 and Patchwork_12984:
### New IGT tests (1) ###
* igt@i915_selftest@live_mman:
- Statuses : 42 pass(s)
- Exec time: [4.50, 57.25] s
Known issues
------------
Here are the changes found in Patchwork_12984 that come from known issues:
### IGT changes ###
#### Issues hit ####
* igt@i915_pm_rpm@module-reload:
- fi-skl-6770hq: [PASS][1] -> [FAIL][2] ([fdo#108511])
[1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6063/fi-skl-6770hq/igt@i915_pm_rpm@module-reload.html
[2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_12984/fi-skl-6770hq/igt@i915_pm_rpm@module-reload.html
#### Possible fixes ####
* igt@gem_exec_basic@readonly-render:
- {fi-icl-y}: [INCOMPLETE][3] ([fdo#107713]) -> [PASS][4]
[3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6063/fi-icl-y/igt@gem_exec_basic@readonly-render.html
[4]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_12984/fi-icl-y/igt@gem_exec_basic@readonly-render.html
{name}: This element is suppressed. This means it is ignored when computing
the status of the difference (SUCCESS, WARNING, or FAILURE).
[fdo#107713]: https://bugs.freedesktop.org/show_bug.cgi?id=107713
[fdo#108511]: https://bugs.freedesktop.org/show_bug.cgi?id=108511
Participating hosts (53 -> 43)
------------------------------
Missing (10): fi-kbl-soraka fi-ilk-m540 fi-hsw-4200u fi-byt-squawks fi-bsw-cyan fi-apl-guc fi-ctg-p8600 fi-pnv-d510 fi-byt-clapper fi-bdw-samus
Build changes
-------------
* Linux: CI_DRM_6063 -> Patchwork_12984
CI_DRM_6063: 44ae4003d35743cbc7883825c5fe777d136b5247 @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_4972: f052e49a43cc9704ea5f240df15dd9d3dfed68ab @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
Patchwork_12984: 9bf3a26c75f2ad0800b50d5eaf9a6f40d90254bd @ git://anongit.freedesktop.org/gfx-ci/linux
== Linux commits ==
9bf3a26c75f2 drm/i915/execlists: Minimalistic timeslicing
c3f5d7c6f6cc drm/i915/execlists: Preempt-to-busy
3ab6fd267bc8 drm/i915: Flush the execution-callbacks on retiring
f29e29f63664 drm/i915: Replace engine->timeline with a plain list
4bcdd4c21eeb drm/i915: Stop retiring along engine
16ad5ab86e98 drm/i915: Keep contexts pinned until after the next kernel context switch
ff7e06f46eef drm/i915: Rename intel_context.active to .inflight
06e6fa4897ee drm/i915: Move object close under its own lock
5ec5ae3c8224 drm/i915: Drop the deferred active reference
a0342667999c drm/i915: Move GEM client throttling to its own file
3853ef7d5efd drm/i915: Move GEM object busy checking to its own file
37093e2cfcb7 drm/i915: Move GEM object waiting to its own file
82fa4ffba0ac drm/i915: Move GEM object domain management from struct_mutex to local
e95a11a1ae00 drm/i915: Pull scatterlist utils out of i915_gem.h
8be66d73a04c drm/i915: Move more GEM objects under gem/
1786d7cb9bc6 drm/i915: Move GEM domain management to its own file
4734074d136c drm/i915: Move mmap and friends to its own file
35b7895848a6 drm/i915: Move phys objects to its own file
fb33bd9448ad drm/i915: Move shmem object setup to its own file
3483581bdd02 drm/i915: Move object->pages API to i915_gem_object.[ch]
39f76472714f drm/i915: Pull GEM ioctls interface to its own file
b688a0332163 drm/i915: Split GEM object type definition to its own header
00c6c2e19f78 drm/i915: Allow specification of parallel execbuf
7afded600522 drm/i915/execlists: Virtual engine bonding
f0417a49d538 drm/i915: Extend execution fence to support a callback
266984fb2212 drm/i915: Apply an execution_mask to the virtual_engine
69947c56c90e drm/i915: Load balancing across a virtual engine
fc7ed4a4ef81 drm/i915: Allow userspace to clone contexts on creation
5bbcdc2dae1c drm/i915: Re-expose SINGLE_TIMELINE flags for context creation
7af4301e09f7 drm/i915: Extend I915_CONTEXT_PARAM_SSEU to support local ctx->engine[]
d257fe3e40a0 drm/i915: Allow a context to define its set of engines
1415c169395e drm/i915: Restore control over ppgtt for context creation ABI
eb869674ac86 dma-fence: Refactor signaling for manual invocation
0f169a3e02f6 drm/i915: Seal races between async GPU cancellation, retirement and signaling
483a8a8f05cf drm/i915: Convert inconsistent static engine tables into an init error
4c36ecd8e737 drm/i915: Bump signaler priority on adding a waiter
fad4daf02033 drm/i915: Check for no-op priority changes first
fea1b40e0ff4 drm/i915: Pass i915_sched_node around internally
5f78625e4a23 drm/i915: Rearrange i915_scheduler.c
67b421614964 drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_12984/
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread* ✗ Fi.CI.CHECKPATCH: warning for series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD (rev4)
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (41 preceding siblings ...)
2019-05-08 11:09 ` ✓ Fi.CI.BAT: success " Patchwork
@ 2019-05-08 11:49 ` Patchwork
2019-05-08 12:06 ` ✗ Fi.CI.SPARSE: " Patchwork
` (2 subsequent siblings)
45 siblings, 0 replies; 71+ messages in thread
From: Patchwork @ 2019-05-08 11:49 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD (rev4)
URL : https://patchwork.freedesktop.org/series/60403/
State : warning
== Summary ==
$ dim checkpatch origin/drm-tip
e7ad7ff7e483 drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
91a757861669 drm/i915: Rearrange i915_scheduler.c
388db2b0c7a6 drm/i915: Pass i915_sched_node around internally
f4522862f19a drm/i915: Check for no-op priority changes first
007c8b9dcc4b drm/i915: Bump signaler priority on adding a waiter
bfcd51b8596a drm/i915: Convert inconsistent static engine tables into an init error
753e000c7a33 drm/i915: Seal races between async GPU cancellation, retirement and signaling
559d9dee1ecb dma-fence: Refactor signaling for manual invocation
-:30: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#30:
new file mode 100644
-:35: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#35: FILE: drivers/dma-buf/dma-fence-trace.c:1:
+/*
-:195: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#195: FILE: include/linux/dma-fence-types.h:1:
+/*
-:266: CHECK:UNCOMMENTED_DEFINITION: spinlock_t definition without comment
#266: FILE: include/linux/dma-fence-types.h:72:
+ spinlock_t *lock;
total: 0 errors, 3 warnings, 1 checks, 669 lines checked
692fdcbeb38c drm/i915: Restore control over ppgtt for context creation ABI
-:81: WARNING:LONG_LINE: line over 100 characters
#81: FILE: include/uapi/drm/i915_drm.h:420:
+#define DRM_IOCTL_I915_GEM_VM_CREATE DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_VM_CREATE, struct drm_i915_gem_vm_control)
-:82: WARNING:LONG_LINE: line over 100 characters
#82: FILE: include/uapi/drm/i915_drm.h:421:
+#define DRM_IOCTL_I915_GEM_VM_DESTROY DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_VM_DESTROY, struct drm_i915_gem_vm_control)
-:82: WARNING:SPACING: space prohibited between function name and open parenthesis '('
#82: FILE: include/uapi/drm/i915_drm.h:421:
+#define DRM_IOCTL_I915_GEM_VM_DESTROY DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_VM_DESTROY, struct drm_i915_gem_vm_control)
-:82: ERROR:COMPLEX_MACRO: Macros with complex values should be enclosed in parentheses
#82: FILE: include/uapi/drm/i915_drm.h:421:
+#define DRM_IOCTL_I915_GEM_VM_DESTROY DRM_IOW (DRM_COMMAND_BASE + DRM_I915_GEM_VM_DESTROY, struct drm_i915_gem_vm_control)
total: 1 errors, 3 warnings, 0 checks, 64 lines checked
82ba6f07ab8b drm/i915: Allow a context to define its set of engines
-:437: CHECK:MACRO_ARG_REUSE: Macro argument reuse 'p' - possible side-effects?
#437: FILE: drivers/gpu/drm/i915/i915_utils.h:110:
+#define check_struct_size(p, member, n, sz) \
+ likely(__check_struct_size(sizeof(*(p)), \
+ sizeof(*(p)->member) + __must_be_array((p)->member), \
+ n, sz))
-:437: CHECK:MACRO_ARG_REUSE: Macro argument reuse 'member' - possible side-effects?
#437: FILE: drivers/gpu/drm/i915/i915_utils.h:110:
+#define check_struct_size(p, member, n, sz) \
+ likely(__check_struct_size(sizeof(*(p)), \
+ sizeof(*(p)->member) + __must_be_array((p)->member), \
+ n, sz))
-:437: CHECK:MACRO_ARG_PRECEDENCE: Macro argument 'member' may be better as '(member)' to avoid precedence issues
#437: FILE: drivers/gpu/drm/i915/i915_utils.h:110:
+#define check_struct_size(p, member, n, sz) \
+ likely(__check_struct_size(sizeof(*(p)), \
+ sizeof(*(p)->member) + __must_be_array((p)->member), \
+ n, sz))
total: 0 errors, 0 warnings, 3 checks, 428 lines checked
94d7eb79a37a drm/i915: Extend I915_CONTEXT_PARAM_SSEU to support local ctx->engine[]
a699c113a8e2 drm/i915: Re-expose SINGLE_TIMELINE flags for context creation
b60419b1ad85 drm/i915: Allow userspace to clone contexts on creation
-:213: ERROR:BRACKET_SPACE: space prohibited before open square bracket '['
#213: FILE: drivers/gpu/drm/i915/i915_gem_context.c:1858:
+#define MAP(x, y) [ilog2(I915_CONTEXT_CLONE_##x)] = y
total: 1 errors, 0 warnings, 0 checks, 235 lines checked
263014777cd1 drm/i915: Load balancing across a virtual engine
ff8dbcf5aaa5 drm/i915: Apply an execution_mask to the virtual_engine
b4069da09690 drm/i915: Extend execution fence to support a callback
75381837ffb3 drm/i915/execlists: Virtual engine bonding
d1ad066b91b5 drm/i915: Allow specification of parallel execbuf
d5d4ec7726a0 drm/i915: Split GEM object type definition to its own header
-:25: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#25:
new file mode 100644
-:59: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#59: FILE: drivers/gpu/drm/i915/gem/i915_gem_object_types.h:1:
+/*
-:60: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#60: FILE: drivers/gpu/drm/i915/gem/i915_gem_object_types.h:2:
+ * SPDX-License-Identifier: MIT
-:177: WARNING:LINE_SPACING: Missing a blank line after declarations
#177: FILE: drivers/gpu/drm/i915/gem/i915_gem_object_types.h:119:
+ struct list_head batch_pool_link;
+ I915_SELFTEST_DECLARE(struct list_head st_link);
-:436: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#436: FILE: drivers/gpu/drm/i915/i915_gem_object.h:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 5 warnings, 0 checks, 664 lines checked
667eb2ffc9de drm/i915: Pull GEM ioctls interface to its own file
-:12: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#12:
new file mode 100644
-:17: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#17: FILE: drivers/gpu/drm/i915/gem/i915_gem_ioctls.h:1:
+/*
-:18: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#18: FILE: drivers/gpu/drm/i915/gem/i915_gem_ioctls.h:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 3 warnings, 0 checks, 152 lines checked
582d82e30dc2 drm/i915: Move object->pages API to i915_gem_object.[ch]
-:38: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#38:
rename from drivers/gpu/drm/i915/i915_gem_object.c
total: 0 errors, 1 warnings, 0 checks, 339 lines checked
007fcb909112 drm/i915: Move shmem object setup to its own file
-:413: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#413:
new file mode 100644
-:418: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#418: FILE: drivers/gpu/drm/i915/gem/i915_gem_shmem.c:1:
+/*
-:419: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#419: FILE: drivers/gpu/drm/i915/gem/i915_gem_shmem.c:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 3 warnings, 0 checks, 2003 lines checked
37e93800d435 drm/i915: Move phys objects to its own file
-:80: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#80:
new file mode 100644
-:85: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#85: FILE: drivers/gpu/drm/i915/gem/i915_gem_pages.c:1:
+/*
-:86: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#86: FILE: drivers/gpu/drm/i915/gem/i915_gem_pages.c:2:
+ * SPDX-License-Identifier: MIT
-:185: WARNING:MEMORY_BARRIER: memory barrier without comment
#185: FILE: drivers/gpu/drm/i915/gem/i915_gem_pages.c:101:
+ smp_mb__before_atomic();
-:250: CHECK:MULTIPLE_ASSIGNMENTS: multiple assignments should be avoided
#250: FILE: drivers/gpu/drm/i915/gem/i915_gem_pages.c:166:
+ obj->mm.page_sizes.phys = obj->mm.page_sizes.sg = 0;
-:376: WARNING:MEMORY_BARRIER: memory barrier without comment
#376: FILE: drivers/gpu/drm/i915/gem/i915_gem_pages.c:292:
+ smp_mb__before_atomic();
-:395: CHECK:MULTIPLE_ASSIGNMENTS: multiple assignments should be avoided
#395: FILE: drivers/gpu/drm/i915/gem/i915_gem_pages.c:311:
+ ptr = obj->mm.mapping = NULL;
-:601: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#601: FILE: drivers/gpu/drm/i915/gem/i915_gem_phys.c:1:
+/*
-:602: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#602: FILE: drivers/gpu/drm/i915/gem/i915_gem_phys.c:2:
+ * SPDX-License-Identifier: MIT
-:898: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#898: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c:1:
+/*
-:899: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#899: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 9 warnings, 2 checks, 1825 lines checked
12effeeee050 drm/i915: Move mmap and friends to its own file
-:25: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#25:
new file mode 100644
-:30: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#30: FILE: drivers/gpu/drm/i915/gem/i915_gem_mman.c:1:
+/*
-:31: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#31: FILE: drivers/gpu/drm/i915/gem/i915_gem_mman.c:2:
+ * SPDX-License-Identifier: MIT
-:590: WARNING:MEMORY_BARRIER: memory barrier without comment
#590: FILE: drivers/gpu/drm/i915/gem/i915_gem_object.c:400:
+ wmb();
-:633: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#633: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c:1:
+/*
-:634: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#634: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c:2:
+ * SPDX-License-Identifier: MIT
-:827: WARNING:LONG_LINE: line over 100 characters
#827: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c:195:
+ (1 + next_prime_number(i915->ggtt.vm.total >> PAGE_SHIFT)) << PAGE_SHIFT);
-:946: WARNING:SPACING: space prohibited before semicolon
#946: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c:314:
+next_tiling: ;
total: 0 errors, 8 warnings, 0 checks, 2256 lines checked
eb861cac35cf drm/i915: Move GEM domain management to its own file
-:25: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#25:
new file mode 100644
-:30: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#30: FILE: drivers/gpu/drm/i915/gem/i915_gem_domain.c:1:
+/*
-:31: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#31: FILE: drivers/gpu/drm/i915/gem/i915_gem_domain.c:2:
+ * SPDX-License-Identifier: MIT
-:113: WARNING:MEMORY_BARRIER: memory barrier without comment
#113: FILE: drivers/gpu/drm/i915/gem/i915_gem_domain.c:84:
+ mb();
-:175: WARNING:MEMORY_BARRIER: memory barrier without comment
#175: FILE: drivers/gpu/drm/i915/gem/i915_gem_domain.c:146:
+ mb();
total: 0 errors, 5 warnings, 0 checks, 1871 lines checked
23ba65f065cb drm/i915: Move more GEM objects under gem/
-:87: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#87:
rename from drivers/gpu/drm/i915/i915_gem_clflush.c
-:114: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#114: FILE: drivers/gpu/drm/i915/gem/i915_gem_clflush.c:2:
+ * SPDX-License-Identifier: MIT
-:135: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#135: FILE: drivers/gpu/drm/i915/gem/i915_gem_clflush.h:1:
+/*
-:136: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#136: FILE: drivers/gpu/drm/i915/gem/i915_gem_clflush.h:2:
+ * SPDX-License-Identifier: MIT
-:187: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#187: FILE: drivers/gpu/drm/i915/gem/i915_gem_context.c:2:
+ * SPDX-License-Identifier: MIT
-:231: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#231: FILE: drivers/gpu/drm/i915/gem/i915_gem_context.h:2:
+ * SPDX-License-Identifier: MIT
-:262: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#262: FILE: drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c:2:
+ * SPDX-License-Identifier: MIT
-:338: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#338: FILE: drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c:2:
+ * SPDX-License-Identifier: MIT
-:394: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#394: FILE: drivers/gpu/drm/i915/gem/i915_gem_internal.c:2:
+ * SPDX-License-Identifier: MIT
-:495: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#495: FILE: drivers/gpu/drm/i915/gem/i915_gem_shrinker.c:2:
+ * SPDX-License-Identifier: MIT
-:544: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#544: FILE: drivers/gpu/drm/i915/gem/i915_gem_stolen.c:2:
+ * SPDX-License-Identifier: MIT
-:592: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#592: FILE: drivers/gpu/drm/i915/gem/i915_gem_tiling.c:2:
+ * SPDX-License-Identifier: MIT
-:640: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#640: FILE: drivers/gpu/drm/i915/gem/i915_gem_userptr.c:2:
+ * SPDX-License-Identifier: MIT
-:691: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#691: FILE: drivers/gpu/drm/i915/gem/i915_gemfs.c:2:
+ * SPDX-License-Identifier: MIT
-:714: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#714: FILE: drivers/gpu/drm/i915/gem/i915_gemfs.h:1:
+/*
-:715: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#715: FILE: drivers/gpu/drm/i915/gem/i915_gemfs.h:2:
+ * SPDX-License-Identifier: MIT
-:759: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#759: FILE: drivers/gpu/drm/i915/gem/selftests/huge_gem_object.c:2:
+ * SPDX-License-Identifier: MIT
-:771: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#771: FILE: drivers/gpu/drm/i915/gem/selftests/huge_gem_object.h:1:
+/*
-:772: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#772: FILE: drivers/gpu/drm/i915/gem/selftests/huge_gem_object.h:2:
+ * SPDX-License-Identifier: MIT
-:827: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#827: FILE: drivers/gpu/drm/i915/gem/selftests/huge_pages.c:2:
+ * SPDX-License-Identifier: MIT
-:879: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#879: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c:2:
+ * SPDX-License-Identifier: MIT
-:922: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#922: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c:2:
+ * SPDX-License-Identifier: MIT
-:984: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#984: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_dmabuf.c:2:
+ * SPDX-License-Identifier: MIT
-:1040: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#1040: FILE: drivers/gpu/drm/i915/gem/selftests/i915_gem_object.c:2:
+ * SPDX-License-Identifier: MIT
-:1108: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#1108: FILE: drivers/gpu/drm/i915/gem/selftests/mock_context.c:2:
+ * SPDX-License-Identifier: MIT
-:1125: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#1125: FILE: drivers/gpu/drm/i915/gem/selftests/mock_context.h:1:
+/*
-:1126: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#1126: FILE: drivers/gpu/drm/i915/gem/selftests/mock_context.h:2:
+ * SPDX-License-Identifier: MIT
-:1178: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#1178: FILE: drivers/gpu/drm/i915/gem/selftests/mock_dmabuf.c:2:
+ * SPDX-License-Identifier: MIT
-:1190: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#1190: FILE: drivers/gpu/drm/i915/gem/selftests/mock_dmabuf.h:1:
+/*
-:1191: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#1191: FILE: drivers/gpu/drm/i915/gem/selftests/mock_dmabuf.h:2:
+ * SPDX-License-Identifier: MIT
-:1221: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#1221: FILE: drivers/gpu/drm/i915/gem/selftests/mock_gem_object.h:1:
+/*
-:1222: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#1222: FILE: drivers/gpu/drm/i915/gem/selftests/mock_gem_object.h:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 32 warnings, 0 checks, 1405 lines checked
f1297620a3f4 drm/i915: Pull scatterlist utils out of i915_gem.h
-:372: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#372:
new file mode 100644
-:377: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#377: FILE: drivers/gpu/drm/i915/i915_scatterlist.c:1:
+/*
-:378: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#378: FILE: drivers/gpu/drm/i915/i915_scatterlist.c:2:
+ * SPDX-License-Identifier: MIT
-:422: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#422: FILE: drivers/gpu/drm/i915/i915_scatterlist.h:1:
+/*
-:423: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#423: FILE: drivers/gpu/drm/i915/i915_scatterlist.h:2:
+ * SPDX-License-Identifier: MIT
-:452: CHECK:MULTIPLE_ASSIGNMENTS: multiple assignments should be avoided
#452: FILE: drivers/gpu/drm/i915/i915_scatterlist.h:31:
+ s.max = s.curr = s.sgp->offset;
-:497: CHECK:MACRO_ARG_REUSE: Macro argument reuse '__iter' - possible side-effects?
#497: FILE: drivers/gpu/drm/i915/i915_scatterlist.h:76:
+#define __for_each_sgt_dma(__dmap, __iter, __sgt, __step) \
+ for ((__iter) = __sgt_iter((__sgt)->sgl, true); \
+ ((__dmap) = (__iter).dma + (__iter).curr); \
+ (((__iter).curr += (__step)) >= (__iter).max) ? \
+ (__iter) = __sgt_iter(__sg_next((__iter).sgp), true), 0 : 0)
-:509: CHECK:MACRO_ARG_REUSE: Macro argument reuse '__iter' - possible side-effects?
#509: FILE: drivers/gpu/drm/i915/i915_scatterlist.h:88:
+#define for_each_sgt_page(__pp, __iter, __sgt) \
+ for ((__iter) = __sgt_iter((__sgt)->sgl, false); \
+ ((__pp) = (__iter).pfn == 0 ? NULL : \
+ pfn_to_page((__iter).pfn + ((__iter).curr >> PAGE_SHIFT))); \
+ (((__iter).curr += PAGE_SIZE) >= (__iter).max) ? \
+ (__iter) = __sgt_iter(__sg_next((__iter).sgp), false), 0 : 0)
total: 0 errors, 5 warnings, 3 checks, 457 lines checked
354f0876da89 drm/i915: Move GEM object domain management from struct_mutex to local
-:525: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#525:
new file mode 100644
-:530: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#530: FILE: drivers/gpu/drm/i915/gem/i915_gem_fence.c:1:
+/*
-:531: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#531: FILE: drivers/gpu/drm/i915/gem/i915_gem_fence.c:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 3 warnings, 0 checks, 1438 lines checked
31b1d008b2d8 drm/i915: Move GEM object waiting to its own file
-:41: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#41:
new file mode 100644
-:46: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#46: FILE: drivers/gpu/drm/i915/gem/i915_gem_wait.c:1:
+/*
-:47: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#47: FILE: drivers/gpu/drm/i915/gem/i915_gem_wait.c:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 3 warnings, 0 checks, 591 lines checked
2025c4b388cd drm/i915: Move GEM object busy checking to its
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread* ✗ Fi.CI.SPARSE: warning for series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD (rev4)
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (42 preceding siblings ...)
2019-05-08 11:49 ` ✗ Fi.CI.CHECKPATCH: warning for series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD (rev4) Patchwork
@ 2019-05-08 12:06 ` Patchwork
2019-05-08 12:10 ` ✗ Fi.CI.BAT: failure " Patchwork
2019-05-08 12:30 ` [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Mika Kuoppala
45 siblings, 0 replies; 71+ messages in thread
From: Patchwork @ 2019-05-08 12:06 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD (rev4)
URL : https://patchwork.freedesktop.org/series/60403/
State : warning
== Summary ==
$ dim sparse origin/drm-tip
Sparse version: v0.5.2
Commit: drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
Okay!
Commit: drm/i915: Rearrange i915_scheduler.c
Okay!
Commit: drm/i915: Pass i915_sched_node around internally
Okay!
Commit: drm/i915: Check for no-op priority changes first
Okay!
Commit: drm/i915: Bump signaler priority on adding a waiter
Okay!
Commit: drm/i915: Convert inconsistent static engine tables into an init error
Okay!
Commit: drm/i915: Seal races between async GPU cancellation, retirement and signaling
Okay!
Commit: dma-fence: Refactor signaling for manual invocation
Okay!
Commit: drm/i915: Restore control over ppgtt for context creation ABI
Okay!
Commit: drm/i915: Allow a context to define its set of engines
+drivers/gpu/drm/i915/i915_utils.h:87:13: error: incorrect type in conditional
+drivers/gpu/drm/i915/i915_utils.h:87:13: error: undefined identifier '__builtin_mul_overflow'
+drivers/gpu/drm/i915/i915_utils.h:87:13: got void
+drivers/gpu/drm/i915/i915_utils.h:87:13: warning: call with no type!
+drivers/gpu/drm/i915/i915_utils.h:90:13: error: incorrect type in conditional
+drivers/gpu/drm/i915/i915_utils.h:90:13: error: undefined identifier '__builtin_add_overflow'
+drivers/gpu/drm/i915/i915_utils.h:90:13: got void
+drivers/gpu/drm/i915/i915_utils.h:90:13: warning: call with no type!
-drivers/gpu/drm/i915/selftests/../i915_utils.h:184:16: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/selftests/../i915_utils.h:218:16: warning: expression using sizeof(void)
+./include/linux/overflow.h:285:13: error: incorrect type in conditional
+./include/linux/overflow.h:285:13: error: not a function <noident>
+./include/linux/overflow.h:285:13: got void
+./include/linux/overflow.h:287:13: error: incorrect type in conditional
+./include/linux/overflow.h:287:13: error: not a function <noident>
+./include/linux/overflow.h:287:13: got void
Commit: drm/i915: Extend I915_CONTEXT_PARAM_SSEU to support local ctx->engine[]
Okay!
Commit: drm/i915: Re-expose SINGLE_TIMELINE flags for context creation
Okay!
Commit: drm/i915: Allow userspace to clone contexts on creation
+drivers/gpu/drm/i915/i915_gem_context.c:1859:17: error: bad integer constant expression
+drivers/gpu/drm/i915/i915_gem_context.c:1860:17: error: bad integer constant expression
+drivers/gpu/drm/i915/i915_gem_context.c:1861:17: error: bad integer constant expression
+drivers/gpu/drm/i915/i915_gem_context.c:1862:17: error: bad integer constant expression
+drivers/gpu/drm/i915/i915_gem_context.c:1863:17: error: bad integer constant expression
+drivers/gpu/drm/i915/i915_gem_context.c:1864:17: error: bad integer constant expression
-drivers/gpu/drm/i915/i915_utils.h:87:13: warning: call with no type!
-drivers/gpu/drm/i915/i915_utils.h:90:13: warning: call with no type!
-drivers/gpu/drm/i915/selftests/i915_gem_context.c:1266:25: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/selftests/i915_gem_context.c:1266:25: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/selftests/i915_gem_context.c:454:16: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/selftests/i915_gem_context.c:571:33: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/selftests/i915_gem_context.c:571:33: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/selftests/i915_gem_context.c:693:33: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/selftests/i915_gem_context.c:693:33: warning: expression using sizeof(void)
+./include/linux/overflow.h:285:13: error: incorrect type in conditional
+./include/linux/overflow.h:285:13: error: not a function <noident>
-./include/linux/overflow.h:285:13: warning: call with no type!
+./include/linux/overflow.h:285:13: got void
+./include/linux/overflow.h:287:13: error: incorrect type in conditional
+./include/linux/overflow.h:287:13: error: not a function <noident>
-./include/linux/overflow.h:287:13: warning: call with no type!
+./include/linux/overflow.h:287:13: got void
-./include/linux/slab.h:666:13: warning: call with no type!
Commit: drm/i915: Load balancing across a virtual engine
+./include/linux/overflow.h:285:13: error: incorrect type in conditional
+./include/linux/overflow.h:285:13: error: undefined identifier '__builtin_mul_overflow'
+./include/linux/overflow.h:285:13: got void
+./include/linux/overflow.h:285:13: warning: call with no type!
+./include/linux/overflow.h:287:13: error: incorrect type in conditional
+./include/linux/overflow.h:287:13: error: undefined identifier '__builtin_add_overflow'
+./include/linux/overflow.h:287:13: got void
+./include/linux/overflow.h:287:13: warning: call with no type!
+./include/linux/slab.h:666:13: error: not a function <noident>
Commit: drm/i915: Apply an execution_mask to the virtual_engine
Okay!
Commit: drm/i915: Extend execution fence to support a callback
Okay!
Commit: drm/i915/execlists: Virtual engine bonding
Okay!
Commit: drm/i915: Allow specification of parallel execbuf
Okay!
Commit: drm/i915: Split GEM object type definition to its own header
+./include/uapi/linux/perf_event.h:147:56: warning: cast truncates bits from constant value (8000000000000000 becomes 0)
Commit: drm/i915: Pull GEM ioctls interface to its own file
Okay!
Commit: drm/i915: Move object->pages API to i915_gem_object.[ch]
+./include/uapi/linux/perf_event.h:147:56: warning: cast truncates bits from constant value (8000000000000000 becomes 0)
Commit: drm/i915: Move shmem object setup to its own file
+drivers/gpu/drm/i915/gem/i915_gem_shmem.c:482:36: warning: expression using sizeof(void)
-O:drivers/gpu/drm/i915/i915_gem.c:4913:36: warning: expression using sizeof(void)
+./include/uapi/linux/perf_event.h:147:56: warning: cast truncates bits from constant value (8000000000000000 becomes 0)
Commit: drm/i915: Move phys objects to its own file
+./include/uapi/linux/perf_event.h:147:56: warning: cast truncates bits from constant value (8000000000000000 becomes 0)
+./include/uapi/linux/perf_event.h:147:56: warning: cast truncates bits from constant value (8000000000000000 becomes 0)
Commit: drm/i915: Move mmap and friends to its own file
+drivers/gpu/drm/i915/gem/i915_gem_mman.c:191:17: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/gem/i915_gem_mman.c:191:17: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/gem/i915_gem_mman.c:191:17: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/gem/i915_gem_mman.c:191:17: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/gem/i915_gem_mman.c:306:32: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/gem/i915_gem_mman.c:306:32: warning: expression using sizeof(void)
-O:drivers/gpu/drm/i915/i915_gem.c:1502:17: warning: expression using sizeof(void)
-O:drivers/gpu/drm/i915/i915_gem.c:1502:17: warning: expression using sizeof(void)
-O:drivers/gpu/drm/i915/i915_gem.c:1502:17: warning: expression using sizeof(void)
-O:drivers/gpu/drm/i915/i915_gem.c:1502:17: warning: expression using sizeof(void)
-O:drivers/gpu/drm/i915/i915_gem.c:1617:32: warning: expression using sizeof(void)
-O:drivers/gpu/drm/i915/i915_gem.c:1617:32: warning: expression using sizeof(void)
+./include/uapi/linux/perf_event.h:147:56: warning: cast truncates bits from constant value (8000000000000000 becomes 0)
Commit: drm/i915: Move GEM domain management to its own file
+drivers/gpu/drm/i915/gem/i915_gem_domain.c:447:34: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/gem/i915_gem_domain.c:447:34: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/i915_gem.c:889:39: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/i915_gem.c:889:39: warning: expression using sizeof(void)
-O:drivers/gpu/drm/i915/i915_gem.c:1951:34: warning: expression using sizeof(void)
-O:drivers/gpu/drm/i915/i915_gem.c:1951:34: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/i915_gem.c:889:39: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/i915_gem.c:889:39: warning: expression using sizeof(void)
+./include/uapi/linux/perf_event.h:147:56: warning: cast truncates bits from constant value (8000000000000000 becomes 0)
Commit: drm/i915: Move more GEM objects under gem/
+drivers/gpu/drm/i915/gem/i915_gem_context.c:2030:17: error: bad integer constant expression
+drivers/gpu/drm/i915/gem/i915_gem_context.c:2031:17: error: bad integer constant expression
+drivers/gpu/drm/i915/gem/i915_gem_context.c:2032:17: error: bad integer constant expression
+drivers/gpu/drm/i915/gem/i915_gem_context.c:2033:17: error: bad integer constant expression
+drivers/gpu/drm/i915/gem/i915_gem_context.c:2034:17: error: bad integer constant expression
+drivers/gpu/drm/i915/gem/i915_gem_context.c:2035:17: error: bad integer constant expression
+drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c:1496:25: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/gem/i915_gem_internal.c:55:39: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/gem/i915_gem_internal.c:57:37: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/gem/i915_gem_internal.c:57:37: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/gem/i915_gem_internal.c:85:29: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/gem/i915_gem_internal.c:85:29: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/gem/i915_gem_shrinker.c:339:25: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/gem/selftests/huge_pages.c:202:36: warning: expression using sizeof(void)
+drivers/gpu/drm/i915/gem/selftests/huge_pages.c:202:36: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/gem/i915_gem_context.c:2030:17: error: bad integer constant expression
-drivers/gpu/drm/i915/gem/i915_gem_context.c:2031:17: error: bad integer constant expression
-drivers/gpu/drm/i915/gem/i915_gem_context.c:2032:17: error: bad integer constant expression
-drivers/gpu/drm/i915/gem/i915_gem_context.c:2033:17: error: bad integer constant expression
-drivers/gpu/drm/i915/gem/i915_gem_context.c:2034:17: error: bad integer constant expression
-drivers/gpu/drm/i915/gem/i915_gem_context.c:2035:17: error: bad integer constant expression
-drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c:1496:25: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/gem/i915_gem_internal.c:55:39: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/gem/i915_gem_internal.c:57:37: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/gem/i915_gem_internal.c:57:37: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/gem/i915_gem_internal.c:85:29: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/gem/i915_gem_internal.c:85:29: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/gem/i915_gem_shrinker.c:339:25: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/gem/selftests/huge_pages.c:202:36: warning: expression using sizeof(void)
-drivers/gpu/drm/i915/gem/selftests/huge_pages.c:202:36: warning: expression using sizeof(void)
+./include/uapi/linux/perf_event.h:147:56: warning: cast truncates bits from constant value (8000000000000000 becomes 0)
+./include/uapi/linux/perf_event.h:147:56: warning: cast truncates bits from constant value (8000000000000000 becomes 0)
Commit: drm/i915: Pull scatterlist utils out of i915_gem.h
Okay!
Commit: drm/i915: Move GEM object domain management from struct_mutex to local
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:20: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/linux/reservation.h:220:45: warning: dereference of noderef expression
+./include/uapi/linux/perf_event.h:147:56: warning: cast truncates bits from constant value (8000000000000000 becomes 0)
Commit: drm/i915: Move GEM object waiting to its own file
+drivers/gpu/drm/i915/gem/i915_gem_wait.c:194:16:
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread* ✗ Fi.CI.BAT: failure for series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD (rev4)
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (43 preceding siblings ...)
2019-05-08 12:06 ` ✗ Fi.CI.SPARSE: " Patchwork
@ 2019-05-08 12:10 ` Patchwork
2019-05-08 12:30 ` [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Mika Kuoppala
45 siblings, 0 replies; 71+ messages in thread
From: Patchwork @ 2019-05-08 12:10 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: series starting with [01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD (rev4)
URL : https://patchwork.freedesktop.org/series/60403/
State : failure
== Summary ==
CI Bug Log - changes from CI_DRM_6063 -> Patchwork_12985
====================================================
Summary
-------
**FAILURE**
Serious unknown changes coming with Patchwork_12985 absolutely need to be
verified manually.
If you think the reported changes have nothing to do with the changes
introduced in Patchwork_12985, please notify your bug team to allow them
to document this new failure mode, which will reduce false positives in CI.
External URL: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_12985/
Possible new issues
-------------------
Here are the unknown changes that may have been introduced in Patchwork_12985:
### IGT changes ###
#### Possible regressions ####
* igt@runner@aborted:
- fi-icl-u3: NOTRUN -> [FAIL][1]
[1]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_12985/fi-icl-u3/igt@runner@aborted.html
New tests
---------
New tests have been introduced between CI_DRM_6063 and Patchwork_12985:
### New IGT tests (1) ###
* igt@i915_selftest@live_mman:
- Statuses : 38 pass(s)
- Exec time: [4.46, 57.68] s
Known issues
------------
Here are the changes found in Patchwork_12985 that come from known issues:
### IGT changes ###
#### Issues hit ####
* igt@gem_exec_parallel@basic:
- fi-apl-guc: [PASS][2] -> [FAIL][3] ([fdo#109474] / [fdo#110512])
[2]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6063/fi-apl-guc/igt@gem_exec_parallel@basic.html
[3]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_12985/fi-apl-guc/igt@gem_exec_parallel@basic.html
* igt@gem_exec_reloc@basic-gtt:
- fi-apl-guc: [PASS][4] -> [FAIL][5] ([fdo#109474])
[4]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6063/fi-apl-guc/igt@gem_exec_reloc@basic-gtt.html
[5]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_12985/fi-apl-guc/igt@gem_exec_reloc@basic-gtt.html
* igt@i915_selftest@live_contexts:
- fi-bdw-gvtdvm: [PASS][6] -> [DMESG-FAIL][7] ([fdo#110235])
[6]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6063/fi-bdw-gvtdvm/igt@i915_selftest@live_contexts.html
[7]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_12985/fi-bdw-gvtdvm/igt@i915_selftest@live_contexts.html
* igt@i915_selftest@live_execlists:
- fi-icl-u3: [PASS][8] -> [INCOMPLETE][9] ([fdo#107713])
[8]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6063/fi-icl-u3/igt@i915_selftest@live_execlists.html
[9]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_12985/fi-icl-u3/igt@i915_selftest@live_execlists.html
* igt@i915_selftest@live_hangcheck:
- fi-skl-iommu: [PASS][10] -> [INCOMPLETE][11] ([fdo#108602] / [fdo#108744])
[10]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6063/fi-skl-iommu/igt@i915_selftest@live_hangcheck.html
[11]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_12985/fi-skl-iommu/igt@i915_selftest@live_hangcheck.html
{name}: This element is suppressed. This means it is ignored when computing
the status of the difference (SUCCESS, WARNING, or FAILURE).
[fdo#107713]: https://bugs.freedesktop.org/show_bug.cgi?id=107713
[fdo#108602]: https://bugs.freedesktop.org/show_bug.cgi?id=108602
[fdo#108744]: https://bugs.freedesktop.org/show_bug.cgi?id=108744
[fdo#109100]: https://bugs.freedesktop.org/show_bug.cgi?id=109100
[fdo#109474]: https://bugs.freedesktop.org/show_bug.cgi?id=109474
[fdo#110235]: https://bugs.freedesktop.org/show_bug.cgi?id=110235
[fdo#110512]: https://bugs.freedesktop.org/show_bug.cgi?id=110512
Participating hosts (53 -> 42)
------------------------------
Missing (11): fi-kbl-soraka fi-ilk-m540 fi-hsw-4200u fi-byt-squawks fi-bsw-cyan fi-skl-6260u fi-kbl-guc fi-ctg-p8600 fi-pnv-d510 fi-byt-clapper fi-bdw-samus
Build changes
-------------
* Linux: CI_DRM_6063 -> Patchwork_12985
CI_DRM_6063: 44ae4003d35743cbc7883825c5fe777d136b5247 @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_4972: f052e49a43cc9704ea5f240df15dd9d3dfed68ab @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
Patchwork_12985: f3a02b00de822331e44bd7f08aa53ffbfcdf8f4c @ git://anongit.freedesktop.org/gfx-ci/linux
== Linux commits ==
f3a02b00de82 drm/i915/execlists: Minimalistic timeslicing
fac27574d0e9 drm/i915/execlists: Preempt-to-busy
fa40ebec0d94 drm/i915: Flush the execution-callbacks on retiring
fa5d8af7b06b drm/i915: Replace engine->timeline with a plain list
c964292d3f9a drm/i915: Stop retiring along engine
1637bc4b6181 drm/i915: Keep contexts pinned until after the next kernel context switch
583bed163977 drm/i915: Rename intel_context.active to .inflight
242b3ceead01 drm/i915: Move object close under its own lock
ce78a3d23c6b drm/i915: Drop the deferred active reference
600a838dbb91 drm/i915: Move GEM client throttling to its own file
2025c4b388cd drm/i915: Move GEM object busy checking to its own file
31b1d008b2d8 drm/i915: Move GEM object waiting to its own file
354f0876da89 drm/i915: Move GEM object domain management from struct_mutex to local
f1297620a3f4 drm/i915: Pull scatterlist utils out of i915_gem.h
23ba65f065cb drm/i915: Move more GEM objects under gem/
eb861cac35cf drm/i915: Move GEM domain management to its own file
12effeeee050 drm/i915: Move mmap and friends to its own file
37e93800d435 drm/i915: Move phys objects to its own file
007fcb909112 drm/i915: Move shmem object setup to its own file
582d82e30dc2 drm/i915: Move object->pages API to i915_gem_object.[ch]
667eb2ffc9de drm/i915: Pull GEM ioctls interface to its own file
d5d4ec7726a0 drm/i915: Split GEM object type definition to its own header
d1ad066b91b5 drm/i915: Allow specification of parallel execbuf
75381837ffb3 drm/i915/execlists: Virtual engine bonding
b4069da09690 drm/i915: Extend execution fence to support a callback
ff8dbcf5aaa5 drm/i915: Apply an execution_mask to the virtual_engine
263014777cd1 drm/i915: Load balancing across a virtual engine
b60419b1ad85 drm/i915: Allow userspace to clone contexts on creation
a699c113a8e2 drm/i915: Re-expose SINGLE_TIMELINE flags for context creation
94d7eb79a37a drm/i915: Extend I915_CONTEXT_PARAM_SSEU to support local ctx->engine[]
82ba6f07ab8b drm/i915: Allow a context to define its set of engines
692fdcbeb38c drm/i915: Restore control over ppgtt for context creation ABI
559d9dee1ecb dma-fence: Refactor signaling for manual invocation
753e000c7a33 drm/i915: Seal races between async GPU cancellation, retirement and signaling
bfcd51b8596a drm/i915: Convert inconsistent static engine tables into an init error
007c8b9dcc4b drm/i915: Bump signaler priority on adding a waiter
f4522862f19a drm/i915: Check for no-op priority changes first
388db2b0c7a6 drm/i915: Pass i915_sched_node around internally
91a757861669 drm/i915: Rearrange i915_scheduler.c
e7ad7ff7e483 drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_12985/
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread* Re: [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
2019-05-08 8:06 [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Chris Wilson
` (44 preceding siblings ...)
2019-05-08 12:10 ` ✗ Fi.CI.BAT: failure " Patchwork
@ 2019-05-08 12:30 ` Mika Kuoppala
2019-05-08 12:40 ` Chris Wilson
45 siblings, 1 reply; 71+ messages in thread
From: Mika Kuoppala @ 2019-05-08 12:30 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
Chris Wilson <chris@chris-wilson.co.uk> writes:
> After realising we need to sample RING_START to detect context switches
> from preemption events that do not allow for the seqno to advance, we
> can also realise that the seqno itself is just a distance along the ring
> and so can be replaced by sampling RING_HEAD.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
> ---
> drivers/gpu/drm/i915/gt/intel_engine.h | 15 ---------
> drivers/gpu/drm/i915/gt/intel_engine_cs.c | 5 ++-
> drivers/gpu/drm/i915/gt/intel_engine_types.h | 3 +-
> drivers/gpu/drm/i915/gt/intel_hangcheck.c | 8 ++---
> drivers/gpu/drm/i915/gt/intel_lrc.c | 19 +++---------
> drivers/gpu/drm/i915/gt/intel_ringbuffer.c | 32 ++------------------
> drivers/gpu/drm/i915/i915_debugfs.c | 12 ++------
> 7 files changed, 17 insertions(+), 77 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h
> index 06d785533502..9359b3a7ad9c 100644
> --- a/drivers/gpu/drm/i915/gt/intel_engine.h
> +++ b/drivers/gpu/drm/i915/gt/intel_engine.h
> @@ -215,8 +215,6 @@ intel_write_status_page(struct intel_engine_cs *engine, int reg, u32 value)
> */
> #define I915_GEM_HWS_PREEMPT 0x32
> #define I915_GEM_HWS_PREEMPT_ADDR (I915_GEM_HWS_PREEMPT * sizeof(u32))
> -#define I915_GEM_HWS_HANGCHECK 0x34
> -#define I915_GEM_HWS_HANGCHECK_ADDR (I915_GEM_HWS_HANGCHECK * sizeof(u32))
> #define I915_GEM_HWS_SEQNO 0x40
> #define I915_GEM_HWS_SEQNO_ADDR (I915_GEM_HWS_SEQNO * sizeof(u32))
> #define I915_GEM_HWS_SCRATCH 0x80
> @@ -548,17 +546,4 @@ static inline bool inject_preempt_hang(struct intel_engine_execlists *execlists)
>
> #endif
>
> -static inline u32
> -intel_engine_next_hangcheck_seqno(struct intel_engine_cs *engine)
> -{
> - return engine->hangcheck.next_seqno =
> - next_pseudo_random32(engine->hangcheck.next_seqno);
> -}
> -
> -static inline u32
> -intel_engine_get_hangcheck_seqno(struct intel_engine_cs *engine)
> -{
> - return intel_read_status_page(engine, I915_GEM_HWS_HANGCHECK);
> -}
> -
> #endif /* _INTEL_RINGBUFFER_H_ */
> diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
> index 416d7e2e6f8c..4c3753c1b573 100644
> --- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c
> +++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
> @@ -721,6 +721,7 @@ static int measure_breadcrumb_dw(struct intel_engine_cs *engine)
> goto out_timeline;
>
> dw = engine->emit_fini_breadcrumb(&frame->rq, frame->cs) - frame->cs;
> + GEM_BUG_ON(dw & 1); /* RING_TAIL must be qword aligned */
>
> i915_timeline_unpin(&frame->timeline);
>
> @@ -1444,9 +1445,7 @@ void intel_engine_dump(struct intel_engine_cs *engine,
> drm_printf(m, "*** WEDGED ***\n");
>
> drm_printf(m, "\tAwake? %d\n", atomic_read(&engine->wakeref.count));
> - drm_printf(m, "\tHangcheck %x:%x [%d ms]\n",
> - engine->hangcheck.last_seqno,
> - engine->hangcheck.next_seqno,
> + drm_printf(m, "\tHangcheck: %d ms ago\n",
> jiffies_to_msecs(jiffies - engine->hangcheck.action_timestamp));
> drm_printf(m, "\tReset count: %d (global %d)\n",
> i915_reset_engine_count(error, engine),
> diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
> index c0ab11b12e14..e381c1c73902 100644
> --- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
> +++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
> @@ -54,8 +54,7 @@ struct intel_instdone {
> struct intel_engine_hangcheck {
> u64 acthd;
> u32 last_ring;
> - u32 last_seqno;
> - u32 next_seqno;
> + u32 last_head;
> unsigned long action_timestamp;
> struct intel_instdone instdone;
> };
> diff --git a/drivers/gpu/drm/i915/gt/intel_hangcheck.c b/drivers/gpu/drm/i915/gt/intel_hangcheck.c
> index 721ab74a382f..3a4d09b80fa0 100644
> --- a/drivers/gpu/drm/i915/gt/intel_hangcheck.c
> +++ b/drivers/gpu/drm/i915/gt/intel_hangcheck.c
> @@ -28,7 +28,7 @@
> struct hangcheck {
> u64 acthd;
> u32 ring;
> - u32 seqno;
> + u32 head;
> enum intel_engine_hangcheck_action action;
> unsigned long action_timestamp;
> int deadlock;
> @@ -134,16 +134,16 @@ static void hangcheck_load_sample(struct intel_engine_cs *engine,
> struct hangcheck *hc)
> {
> hc->acthd = intel_engine_get_active_head(engine);
> - hc->seqno = intel_engine_get_hangcheck_seqno(engine);
> hc->ring = ENGINE_READ(engine, RING_START);
> + hc->head = ENGINE_READ(engine, RING_HEAD);
> }
>
> static void hangcheck_store_sample(struct intel_engine_cs *engine,
> const struct hangcheck *hc)
> {
> engine->hangcheck.acthd = hc->acthd;
> - engine->hangcheck.last_seqno = hc->seqno;
> engine->hangcheck.last_ring = hc->ring;
> + engine->hangcheck.last_head = hc->head;
> }
>
> static enum intel_engine_hangcheck_action
> @@ -156,7 +156,7 @@ hangcheck_get_action(struct intel_engine_cs *engine,
> if (engine->hangcheck.last_ring != hc->ring)
> return ENGINE_ACTIVE_SEQNO;
>
> - if (engine->hangcheck.last_seqno != hc->seqno)
> + if (engine->hangcheck.last_head != hc->head)
> return ENGINE_ACTIVE_SEQNO;
Change the enum also?
>
> return engine_stuck(engine, hc->acthd);
> diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
> index d1a54d2c3d5d..f1d62746e066 100644
> --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
> +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
> @@ -2275,12 +2275,6 @@ static u32 *gen8_emit_fini_breadcrumb(struct i915_request *request, u32 *cs)
> request->timeline->hwsp_offset,
> 0);
>
> - cs = gen8_emit_ggtt_write(cs,
> - intel_engine_next_hangcheck_seqno(request->engine),
> - I915_GEM_HWS_HANGCHECK_ADDR,
> - MI_FLUSH_DW_STORE_INDEX);
> -
> -
> *cs++ = MI_USER_INTERRUPT;
> *cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
>
> @@ -2297,14 +2291,11 @@ static u32 *gen8_emit_fini_breadcrumb_rcs(struct i915_request *request, u32 *cs)
> request->timeline->hwsp_offset,
> PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH |
> PIPE_CONTROL_DEPTH_CACHE_FLUSH |
> - PIPE_CONTROL_DC_FLUSH_ENABLE |
> - PIPE_CONTROL_FLUSH_ENABLE |
> - PIPE_CONTROL_CS_STALL);
???
-Mika
> -
> - cs = gen8_emit_ggtt_write_rcs(cs,
> - intel_engine_next_hangcheck_seqno(request->engine),
> - I915_GEM_HWS_HANGCHECK_ADDR,
> - PIPE_CONTROL_STORE_DATA_INDEX);
> + PIPE_CONTROL_DC_FLUSH_ENABLE);
> + cs = gen8_emit_pipe_control(cs,
> + PIPE_CONTROL_FLUSH_ENABLE |
> + PIPE_CONTROL_CS_STALL,
> + 0);
>
> *cs++ = MI_USER_INTERRUPT;
> *cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
> diff --git a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
> index 6fbc5ddbc896..f0d60affdba3 100644
> --- a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
> +++ b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
> @@ -309,11 +309,6 @@ static u32 *gen6_rcs_emit_breadcrumb(struct i915_request *rq, u32 *cs)
> *cs++ = rq->timeline->hwsp_offset | PIPE_CONTROL_GLOBAL_GTT;
> *cs++ = rq->fence.seqno;
>
> - *cs++ = GFX_OP_PIPE_CONTROL(4);
> - *cs++ = PIPE_CONTROL_QW_WRITE | PIPE_CONTROL_STORE_DATA_INDEX;
> - *cs++ = I915_GEM_HWS_HANGCHECK_ADDR | PIPE_CONTROL_GLOBAL_GTT;
> - *cs++ = intel_engine_next_hangcheck_seqno(rq->engine);
> -
> *cs++ = MI_USER_INTERRUPT;
> *cs++ = MI_NOOP;
>
> @@ -415,13 +410,6 @@ static u32 *gen7_rcs_emit_breadcrumb(struct i915_request *rq, u32 *cs)
> *cs++ = rq->timeline->hwsp_offset;
> *cs++ = rq->fence.seqno;
>
> - *cs++ = GFX_OP_PIPE_CONTROL(4);
> - *cs++ = (PIPE_CONTROL_QW_WRITE |
> - PIPE_CONTROL_STORE_DATA_INDEX |
> - PIPE_CONTROL_GLOBAL_GTT_IVB);
> - *cs++ = I915_GEM_HWS_HANGCHECK_ADDR;
> - *cs++ = intel_engine_next_hangcheck_seqno(rq->engine);
> -
> *cs++ = MI_USER_INTERRUPT;
> *cs++ = MI_NOOP;
>
> @@ -440,12 +428,7 @@ static u32 *gen6_xcs_emit_breadcrumb(struct i915_request *rq, u32 *cs)
> *cs++ = I915_GEM_HWS_SEQNO_ADDR | MI_FLUSH_DW_USE_GTT;
> *cs++ = rq->fence.seqno;
>
> - *cs++ = MI_FLUSH_DW | MI_FLUSH_DW_OP_STOREDW | MI_FLUSH_DW_STORE_INDEX;
> - *cs++ = I915_GEM_HWS_HANGCHECK_ADDR | MI_FLUSH_DW_USE_GTT;
> - *cs++ = intel_engine_next_hangcheck_seqno(rq->engine);
> -
> *cs++ = MI_USER_INTERRUPT;
> - *cs++ = MI_NOOP;
>
> rq->tail = intel_ring_offset(rq, cs);
> assert_ring_tail_valid(rq->ring, rq->tail);
> @@ -465,10 +448,6 @@ static u32 *gen7_xcs_emit_breadcrumb(struct i915_request *rq, u32 *cs)
> *cs++ = I915_GEM_HWS_SEQNO_ADDR | MI_FLUSH_DW_USE_GTT;
> *cs++ = rq->fence.seqno;
>
> - *cs++ = MI_FLUSH_DW | MI_FLUSH_DW_OP_STOREDW | MI_FLUSH_DW_STORE_INDEX;
> - *cs++ = I915_GEM_HWS_HANGCHECK_ADDR | MI_FLUSH_DW_USE_GTT;
> - *cs++ = intel_engine_next_hangcheck_seqno(rq->engine);
> -
> for (i = 0; i < GEN7_XCS_WA; i++) {
> *cs++ = MI_STORE_DWORD_INDEX;
> *cs++ = I915_GEM_HWS_SEQNO_ADDR;
> @@ -480,6 +459,7 @@ static u32 *gen7_xcs_emit_breadcrumb(struct i915_request *rq, u32 *cs)
> *cs++ = 0;
>
> *cs++ = MI_USER_INTERRUPT;
> + *cs++ = MI_NOOP;
>
> rq->tail = intel_ring_offset(rq, cs);
> assert_ring_tail_valid(rq->ring, rq->tail);
> @@ -928,11 +908,8 @@ static u32 *i9xx_emit_breadcrumb(struct i915_request *rq, u32 *cs)
> *cs++ = I915_GEM_HWS_SEQNO_ADDR;
> *cs++ = rq->fence.seqno;
>
> - *cs++ = MI_STORE_DWORD_INDEX;
> - *cs++ = I915_GEM_HWS_HANGCHECK_ADDR;
> - *cs++ = intel_engine_next_hangcheck_seqno(rq->engine);
> -
> *cs++ = MI_USER_INTERRUPT;
> + *cs++ = MI_NOOP;
>
> rq->tail = intel_ring_offset(rq, cs);
> assert_ring_tail_valid(rq->ring, rq->tail);
> @@ -950,10 +927,6 @@ static u32 *gen5_emit_breadcrumb(struct i915_request *rq, u32 *cs)
>
> *cs++ = MI_FLUSH;
>
> - *cs++ = MI_STORE_DWORD_INDEX;
> - *cs++ = I915_GEM_HWS_HANGCHECK_ADDR;
> - *cs++ = intel_engine_next_hangcheck_seqno(rq->engine);
> -
> BUILD_BUG_ON(GEN5_WA_STORES < 1);
> for (i = 0; i < GEN5_WA_STORES; i++) {
> *cs++ = MI_STORE_DWORD_INDEX;
> @@ -962,7 +935,6 @@ static u32 *gen5_emit_breadcrumb(struct i915_request *rq, u32 *cs)
> }
>
> *cs++ = MI_USER_INTERRUPT;
> - *cs++ = MI_NOOP;
>
> rq->tail = intel_ring_offset(rq, cs);
> assert_ring_tail_valid(rq->ring, rq->tail);
> diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
> index b6094063be9b..cb9b56fb6a8a 100644
> --- a/drivers/gpu/drm/i915/i915_debugfs.c
> +++ b/drivers/gpu/drm/i915/i915_debugfs.c
> @@ -1288,7 +1288,6 @@ static int i915_hangcheck_info(struct seq_file *m, void *unused)
> struct drm_i915_private *dev_priv = node_to_i915(m->private);
> struct intel_engine_cs *engine;
> u64 acthd[I915_NUM_ENGINES];
> - u32 seqno[I915_NUM_ENGINES];
> struct intel_instdone instdone;
> intel_wakeref_t wakeref;
> enum intel_engine_id id;
> @@ -1305,10 +1304,8 @@ static int i915_hangcheck_info(struct seq_file *m, void *unused)
> }
>
> with_intel_runtime_pm(dev_priv, wakeref) {
> - for_each_engine(engine, dev_priv, id) {
> + for_each_engine(engine, dev_priv, id)
> acthd[id] = intel_engine_get_active_head(engine);
> - seqno[id] = intel_engine_get_hangcheck_seqno(engine);
> - }
>
> intel_engine_get_instdone(dev_priv->engine[RCS0], &instdone);
> }
> @@ -1325,11 +1322,8 @@ static int i915_hangcheck_info(struct seq_file *m, void *unused)
> seq_printf(m, "GT active? %s\n", yesno(dev_priv->gt.awake));
>
> for_each_engine(engine, dev_priv, id) {
> - seq_printf(m, "%s:\n", engine->name);
> - seq_printf(m, "\tseqno = %x [current %x, last %x], %dms ago\n",
> - engine->hangcheck.last_seqno,
> - seqno[id],
> - engine->hangcheck.next_seqno,
> + seq_printf(m, "%s: %d ms ago\n",
> + engine->name,
> jiffies_to_msecs(jiffies -
> engine->hangcheck.action_timestamp));
>
> --
> 2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread* Re: [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
2019-05-08 12:30 ` [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD Mika Kuoppala
@ 2019-05-08 12:40 ` Chris Wilson
2019-05-08 14:00 ` Mika Kuoppala
0 siblings, 1 reply; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 12:40 UTC (permalink / raw)
To: Mika Kuoppala, intel-gfx
Quoting Mika Kuoppala (2019-05-08 13:30:46)
> Chris Wilson <chris@chris-wilson.co.uk> writes:
>
> > After realising we need to sample RING_START to detect context switches
> > from preemption events that do not allow for the seqno to advance, we
> > can also realise that the seqno itself is just a distance along the ring
> > and so can be replaced by sampling RING_HEAD.
> >
> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> > Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
> > ---
> > static enum intel_engine_hangcheck_action
> > @@ -156,7 +156,7 @@ hangcheck_get_action(struct intel_engine_cs *engine,
> > if (engine->hangcheck.last_ring != hc->ring)
> > return ENGINE_ACTIVE_SEQNO;
> >
> > - if (engine->hangcheck.last_seqno != hc->seqno)
> > + if (engine->hangcheck.last_head != hc->head)
> > return ENGINE_ACTIVE_SEQNO;
>
> Change the enum also?
Pffifle. As far as hangcheck goes RING_START:RING_HEAD comprise its
seqno.
Makes for a good talking point in a few years time :)
> > diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
> > index d1a54d2c3d5d..f1d62746e066 100644
> > --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
> > +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
> > @@ -2275,12 +2275,6 @@ static u32 *gen8_emit_fini_breadcrumb(struct i915_request *request, u32 *cs)
> > request->timeline->hwsp_offset,
> > 0);
> >
> > - cs = gen8_emit_ggtt_write(cs,
> > - intel_engine_next_hangcheck_seqno(request->engine),
> > - I915_GEM_HWS_HANGCHECK_ADDR,
> > - MI_FLUSH_DW_STORE_INDEX);
> > -
> > -
> > *cs++ = MI_USER_INTERRUPT;
> > *cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
> >
> > @@ -2297,14 +2291,11 @@ static u32 *gen8_emit_fini_breadcrumb_rcs(struct i915_request *request, u32 *cs)
> > request->timeline->hwsp_offset,
> > PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH |
> > PIPE_CONTROL_DEPTH_CACHE_FLUSH |
> > - PIPE_CONTROL_DC_FLUSH_ENABLE |
> > - PIPE_CONTROL_FLUSH_ENABLE |
> > - PIPE_CONTROL_CS_STALL);
>
> ???
Kabylake sends the interrupt too early otherwise. The hangcheck write
saved us by pure accident.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread
* Re: [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
2019-05-08 12:40 ` Chris Wilson
@ 2019-05-08 14:00 ` Mika Kuoppala
2019-05-08 14:10 ` Chris Wilson
0 siblings, 1 reply; 71+ messages in thread
From: Mika Kuoppala @ 2019-05-08 14:00 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
Chris Wilson <chris@chris-wilson.co.uk> writes:
> Quoting Mika Kuoppala (2019-05-08 13:30:46)
>> Chris Wilson <chris@chris-wilson.co.uk> writes:
>>
>> > After realising we need to sample RING_START to detect context switches
>> > from preemption events that do not allow for the seqno to advance, we
>> > can also realise that the seqno itself is just a distance along the ring
>> > and so can be replaced by sampling RING_HEAD.
>> >
>> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
>> > Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
>> > ---
>> > static enum intel_engine_hangcheck_action
>> > @@ -156,7 +156,7 @@ hangcheck_get_action(struct intel_engine_cs *engine,
>> > if (engine->hangcheck.last_ring != hc->ring)
>> > return ENGINE_ACTIVE_SEQNO;
>> >
>> > - if (engine->hangcheck.last_seqno != hc->seqno)
>> > + if (engine->hangcheck.last_head != hc->head)
>> > return ENGINE_ACTIVE_SEQNO;
>>
>> Change the enum also?
>
> Pffifle. As far as hangcheck goes RING_START:RING_HEAD comprise its
> seqno.
>
> Makes for a good talking point in a few years time :)
Fair enough.
>
>> > diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
>> > index d1a54d2c3d5d..f1d62746e066 100644
>> > --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
>> > +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
>> > @@ -2275,12 +2275,6 @@ static u32 *gen8_emit_fini_breadcrumb(struct i915_request *request, u32 *cs)
>> > request->timeline->hwsp_offset,
>> > 0);
>> >
>> > - cs = gen8_emit_ggtt_write(cs,
>> > - intel_engine_next_hangcheck_seqno(request->engine),
>> > - I915_GEM_HWS_HANGCHECK_ADDR,
>> > - MI_FLUSH_DW_STORE_INDEX);
>> > -
>> > -
>> > *cs++ = MI_USER_INTERRUPT;
>> > *cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
>> >
>> > @@ -2297,14 +2291,11 @@ static u32 *gen8_emit_fini_breadcrumb_rcs(struct i915_request *request, u32 *cs)
>> > request->timeline->hwsp_offset,
>> > PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH |
>> > PIPE_CONTROL_DEPTH_CACHE_FLUSH |
>> > - PIPE_CONTROL_DC_FLUSH_ENABLE |
>> > - PIPE_CONTROL_FLUSH_ENABLE |
>> > - PIPE_CONTROL_CS_STALL);
>>
>> ???
>
> Kabylake sends the interrupt too early otherwise. The hangcheck write
> saved us by pure accident.
I read the diff wrong at first try also, was concerned that we lost cs stall.
Regardless this could benefit from a comment explaining the need to
force sync for the intr.
Was happy to see measure_breadcrumb_dw() paying off. Also leaning to
it for alignment forcing.
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread
* Re: [PATCH 01/40] drm/i915/hangcheck: Replace hangcheck.seqno with RING_HEAD
2019-05-08 14:00 ` Mika Kuoppala
@ 2019-05-08 14:10 ` Chris Wilson
0 siblings, 0 replies; 71+ messages in thread
From: Chris Wilson @ 2019-05-08 14:10 UTC (permalink / raw)
To: Mika Kuoppala, intel-gfx
Quoting Mika Kuoppala (2019-05-08 15:00:11)
> Chris Wilson <chris@chris-wilson.co.uk> writes:
>
> > Quoting Mika Kuoppala (2019-05-08 13:30:46)
> >> Chris Wilson <chris@chris-wilson.co.uk> writes:
> >>
> >> > After realising we need to sample RING_START to detect context switches
> >> > from preemption events that do not allow for the seqno to advance, we
> >> > can also realise that the seqno itself is just a distance along the ring
> >> > and so can be replaced by sampling RING_HEAD.
> >> >
> >> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> >> > Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
> >> > ---
> >> > static enum intel_engine_hangcheck_action
> >> > @@ -156,7 +156,7 @@ hangcheck_get_action(struct intel_engine_cs *engine,
> >> > if (engine->hangcheck.last_ring != hc->ring)
> >> > return ENGINE_ACTIVE_SEQNO;
> >> >
> >> > - if (engine->hangcheck.last_seqno != hc->seqno)
> >> > + if (engine->hangcheck.last_head != hc->head)
> >> > return ENGINE_ACTIVE_SEQNO;
> >>
> >> Change the enum also?
> >
> > Pffifle. As far as hangcheck goes RING_START:RING_HEAD comprise its
> > seqno.
> >
> > Makes for a good talking point in a few years time :)
>
> Fair enough.
>
> >
> >> > diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
> >> > index d1a54d2c3d5d..f1d62746e066 100644
> >> > --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
> >> > +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
> >> > @@ -2275,12 +2275,6 @@ static u32 *gen8_emit_fini_breadcrumb(struct i915_request *request, u32 *cs)
> >> > request->timeline->hwsp_offset,
> >> > 0);
> >> >
> >> > - cs = gen8_emit_ggtt_write(cs,
> >> > - intel_engine_next_hangcheck_seqno(request->engine),
> >> > - I915_GEM_HWS_HANGCHECK_ADDR,
> >> > - MI_FLUSH_DW_STORE_INDEX);
> >> > -
> >> > -
> >> > *cs++ = MI_USER_INTERRUPT;
> >> > *cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
> >> >
> >> > @@ -2297,14 +2291,11 @@ static u32 *gen8_emit_fini_breadcrumb_rcs(struct i915_request *request, u32 *cs)
> >> > request->timeline->hwsp_offset,
> >> > PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH |
> >> > PIPE_CONTROL_DEPTH_CACHE_FLUSH |
> >> > - PIPE_CONTROL_DC_FLUSH_ENABLE |
> >> > - PIPE_CONTROL_FLUSH_ENABLE |
> >> > - PIPE_CONTROL_CS_STALL);
> >>
> >> ???
> >
> > Kabylake sends the interrupt too early otherwise. The hangcheck write
> > saved us by pure accident.
>
> I read the diff wrong at first try also, was concerned that we lost cs stall.
> Regardless this could benefit from a comment explaining the need to
> force sync for the intr.
Indeed, that is sensible since it's an empirical result and worth
validating again later. Done and pushed, thanks.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 71+ messages in thread