* [PATCH 01/27] dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling)
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:01 ` [PATCH 02/27] drm/mm: Pack allocated/scanned boolean into a bitfield Chris Wilson
` (29 subsequent siblings)
30 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
Make dma_fence_enable_sw_signaling() behave like its
dma_fence_add_callback() and dma_fence_default_wait() counterparts and
perform the test to enable signaling under the fence->lock, along with
the action to do so. This ensure that should an implementation be trying
to flush the cb_list (by signaling) on retirement before freeing the
fence, it can do so in a race-free manner.
See also 0fc89b6802ba ("dma-fence: Simply wrap dma_fence_signal_locked
with dma_fence_signal").
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/dma-buf/dma-fence.c | 11 +++++------
1 file changed, 5 insertions(+), 6 deletions(-)
diff --git a/drivers/dma-buf/dma-fence.c b/drivers/dma-buf/dma-fence.c
index 2c136aee3e79..587727089134 100644
--- a/drivers/dma-buf/dma-fence.c
+++ b/drivers/dma-buf/dma-fence.c
@@ -285,19 +285,18 @@ void dma_fence_enable_sw_signaling(struct dma_fence *fence)
{
unsigned long flags;
+ if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
+ return;
+
+ spin_lock_irqsave(fence->lock, flags);
if (!test_and_set_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT,
&fence->flags) &&
- !test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags) &&
fence->ops->enable_signaling) {
trace_dma_fence_enable_signal(fence);
-
- spin_lock_irqsave(fence->lock, flags);
-
if (!fence->ops->enable_signaling(fence))
dma_fence_signal_locked(fence);
-
- spin_unlock_irqrestore(fence->lock, flags);
}
+ spin_unlock_irqrestore(fence->lock, flags);
}
EXPORT_SYMBOL(dma_fence_enable_sw_signaling);
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* [PATCH 02/27] drm/mm: Pack allocated/scanned boolean into a bitfield
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
2019-09-25 10:01 ` [PATCH 01/27] dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling) Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:01 ` [PATCH 03/27] drm/i915: Only track bound elements of the GTT Chris Wilson
` (28 subsequent siblings)
30 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
The ulterior motive to switching the booleans over to bitops is to
allow use of the allocated flag as a bitlock.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/drm_mm.c | 36 +++++++++++--------
.../gpu/drm/i915/gem/i915_gem_execbuffer.c | 6 ++--
drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c | 2 +-
drivers/gpu/drm/i915/i915_gem.c | 16 ++++-----
drivers/gpu/drm/i915/i915_gem_evict.c | 2 +-
drivers/gpu/drm/i915/i915_vma.c | 2 +-
drivers/gpu/drm/i915/i915_vma.h | 4 +--
drivers/gpu/drm/selftests/test-drm_mm.c | 14 ++++----
drivers/gpu/drm/vc4/vc4_crtc.c | 2 +-
drivers/gpu/drm/vc4/vc4_hvs.c | 2 +-
drivers/gpu/drm/vc4/vc4_plane.c | 4 +--
include/drm/drm_mm.h | 7 ++--
12 files changed, 52 insertions(+), 45 deletions(-)
diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c
index 4581c5387372..211967006cec 100644
--- a/drivers/gpu/drm/drm_mm.c
+++ b/drivers/gpu/drm/drm_mm.c
@@ -174,7 +174,7 @@ static void drm_mm_interval_tree_add_node(struct drm_mm_node *hole_node,
node->__subtree_last = LAST(node);
- if (hole_node->allocated) {
+ if (drm_mm_node_allocated(hole_node)) {
rb = &hole_node->rb;
while (rb) {
parent = rb_entry(rb, struct drm_mm_node, rb);
@@ -424,9 +424,9 @@ int drm_mm_reserve_node(struct drm_mm *mm, struct drm_mm_node *node)
node->mm = mm;
+ __set_bit(DRM_MM_NODE_ALLOCATED_BIT, &node->flags);
list_add(&node->node_list, &hole->node_list);
drm_mm_interval_tree_add_node(hole, node);
- node->allocated = true;
node->hole_size = 0;
rm_hole(hole);
@@ -543,9 +543,9 @@ int drm_mm_insert_node_in_range(struct drm_mm * const mm,
node->color = color;
node->hole_size = 0;
+ __set_bit(DRM_MM_NODE_ALLOCATED_BIT, &node->flags);
list_add(&node->node_list, &hole->node_list);
drm_mm_interval_tree_add_node(hole, node);
- node->allocated = true;
rm_hole(hole);
if (adj_start > hole_start)
@@ -561,6 +561,11 @@ int drm_mm_insert_node_in_range(struct drm_mm * const mm,
}
EXPORT_SYMBOL(drm_mm_insert_node_in_range);
+static inline bool drm_mm_node_scanned_block(const struct drm_mm_node *node)
+{
+ return test_bit(DRM_MM_NODE_SCANNED_BIT, &node->flags);
+}
+
/**
* drm_mm_remove_node - Remove a memory node from the allocator.
* @node: drm_mm_node to remove
@@ -574,8 +579,8 @@ void drm_mm_remove_node(struct drm_mm_node *node)
struct drm_mm *mm = node->mm;
struct drm_mm_node *prev_node;
- DRM_MM_BUG_ON(!node->allocated);
- DRM_MM_BUG_ON(node->scanned_block);
+ DRM_MM_BUG_ON(!drm_mm_node_allocated(node));
+ DRM_MM_BUG_ON(drm_mm_node_scanned_block(node));
prev_node = list_prev_entry(node, node_list);
@@ -584,11 +589,12 @@ void drm_mm_remove_node(struct drm_mm_node *node)
drm_mm_interval_tree_remove(node, &mm->interval_tree);
list_del(&node->node_list);
- node->allocated = false;
if (drm_mm_hole_follows(prev_node))
rm_hole(prev_node);
add_hole(prev_node);
+
+ clear_bit_unlock(DRM_MM_NODE_ALLOCATED_BIT, &node->flags);
}
EXPORT_SYMBOL(drm_mm_remove_node);
@@ -605,7 +611,7 @@ void drm_mm_replace_node(struct drm_mm_node *old, struct drm_mm_node *new)
{
struct drm_mm *mm = old->mm;
- DRM_MM_BUG_ON(!old->allocated);
+ DRM_MM_BUG_ON(!drm_mm_node_allocated(old));
*new = *old;
@@ -622,8 +628,7 @@ void drm_mm_replace_node(struct drm_mm_node *old, struct drm_mm_node *new)
&mm->holes_addr);
}
- old->allocated = false;
- new->allocated = true;
+ clear_bit_unlock(DRM_MM_NODE_ALLOCATED_BIT, &old->flags);
}
EXPORT_SYMBOL(drm_mm_replace_node);
@@ -731,9 +736,9 @@ bool drm_mm_scan_add_block(struct drm_mm_scan *scan,
u64 adj_start, adj_end;
DRM_MM_BUG_ON(node->mm != mm);
- DRM_MM_BUG_ON(!node->allocated);
- DRM_MM_BUG_ON(node->scanned_block);
- node->scanned_block = true;
+ DRM_MM_BUG_ON(!drm_mm_node_allocated(node));
+ DRM_MM_BUG_ON(drm_mm_node_scanned_block(node));
+ __set_bit(DRM_MM_NODE_SCANNED_BIT, &node->flags);
mm->scan_active++;
/* Remove this block from the node_list so that we enlarge the hole
@@ -818,8 +823,7 @@ bool drm_mm_scan_remove_block(struct drm_mm_scan *scan,
struct drm_mm_node *prev_node;
DRM_MM_BUG_ON(node->mm != scan->mm);
- DRM_MM_BUG_ON(!node->scanned_block);
- node->scanned_block = false;
+ DRM_MM_BUG_ON(!drm_mm_node_scanned_block(node));
DRM_MM_BUG_ON(!node->mm->scan_active);
node->mm->scan_active--;
@@ -837,6 +841,8 @@ bool drm_mm_scan_remove_block(struct drm_mm_scan *scan,
list_next_entry(node, node_list));
list_add(&node->node_list, &prev_node->node_list);
+ __clear_bit(DRM_MM_NODE_SCANNED_BIT, &node->flags);
+
return (node->start + node->size > scan->hit_start &&
node->start < scan->hit_end);
}
@@ -917,7 +923,7 @@ void drm_mm_init(struct drm_mm *mm, u64 start, u64 size)
/* Clever trick to avoid a special case in the free hole tracking. */
INIT_LIST_HEAD(&mm->head_node.node_list);
- mm->head_node.allocated = false;
+ mm->head_node.flags = 0;
mm->head_node.mm = mm;
mm->head_node.start = start + size;
mm->head_node.size = -size;
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 27dbcb508055..c049199a1df5 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -906,7 +906,7 @@ static void reloc_cache_init(struct reloc_cache *cache,
cache->use_64bit_reloc = HAS_64BIT_RELOC(i915);
cache->has_fence = cache->gen < 4;
cache->needs_unfenced = INTEL_INFO(i915)->unfenced_needs_alignment;
- cache->node.allocated = false;
+ cache->node.flags = 0;
cache->ce = NULL;
cache->rq = NULL;
cache->rq_size = 0;
@@ -968,7 +968,7 @@ static void reloc_cache_reset(struct reloc_cache *cache)
intel_gt_flush_ggtt_writes(ggtt->vm.gt);
io_mapping_unmap_atomic((void __iomem *)vaddr);
- if (cache->node.allocated) {
+ if (drm_mm_node_allocated(&cache->node)) {
ggtt->vm.clear_range(&ggtt->vm,
cache->node.start,
cache->node.size);
@@ -1061,7 +1061,7 @@ static void *reloc_iomap(struct drm_i915_gem_object *obj,
}
offset = cache->node.start;
- if (cache->node.allocated) {
+ if (drm_mm_node_allocated(&cache->node)) {
ggtt->vm.insert_page(&ggtt->vm,
i915_gem_object_get_dma_address(obj, page),
offset, I915_CACHE_NONE, 0);
diff --git a/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c
index ea9a807abd4f..4ff80a9f4005 100644
--- a/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c
+++ b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c
@@ -402,7 +402,7 @@ static u32 uc_fw_ggtt_offset(struct intel_uc_fw *uc_fw, struct i915_ggtt *ggtt)
{
struct drm_mm_node *node = &ggtt->uc_fw;
- GEM_BUG_ON(!node->allocated);
+ GEM_BUG_ON(!drm_mm_node_allocated(node));
GEM_BUG_ON(upper_32_bits(node->start));
GEM_BUG_ON(upper_32_bits(node->start + node->size - 1));
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 2da9544fa9a4..b8f7fe311b0a 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -351,12 +351,12 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
PIN_NOEVICT);
if (!IS_ERR(vma)) {
node.start = i915_ggtt_offset(vma);
- node.allocated = false;
+ node.flags = 0;
} else {
ret = insert_mappable_node(ggtt, &node, PAGE_SIZE);
if (ret)
goto out_unlock;
- GEM_BUG_ON(!node.allocated);
+ GEM_BUG_ON(!drm_mm_node_allocated(&node));
}
mutex_unlock(&i915->drm.struct_mutex);
@@ -393,7 +393,7 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
unsigned page_offset = offset_in_page(offset);
unsigned page_length = PAGE_SIZE - page_offset;
page_length = remain < page_length ? remain : page_length;
- if (node.allocated) {
+ if (drm_mm_node_allocated(&node)) {
ggtt->vm.insert_page(&ggtt->vm,
i915_gem_object_get_dma_address(obj, offset >> PAGE_SHIFT),
node.start, I915_CACHE_NONE, 0);
@@ -415,7 +415,7 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
i915_gem_object_unlock_fence(obj, fence);
out_unpin:
mutex_lock(&i915->drm.struct_mutex);
- if (node.allocated) {
+ if (drm_mm_node_allocated(&node)) {
ggtt->vm.clear_range(&ggtt->vm, node.start, node.size);
remove_mappable_node(&node);
} else {
@@ -561,12 +561,12 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
PIN_NOEVICT);
if (!IS_ERR(vma)) {
node.start = i915_ggtt_offset(vma);
- node.allocated = false;
+ node.flags = 0;
} else {
ret = insert_mappable_node(ggtt, &node, PAGE_SIZE);
if (ret)
goto out_rpm;
- GEM_BUG_ON(!node.allocated);
+ GEM_BUG_ON(!drm_mm_node_allocated(&node));
}
mutex_unlock(&i915->drm.struct_mutex);
@@ -604,7 +604,7 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
unsigned int page_offset = offset_in_page(offset);
unsigned int page_length = PAGE_SIZE - page_offset;
page_length = remain < page_length ? remain : page_length;
- if (node.allocated) {
+ if (drm_mm_node_allocated(&node)) {
/* flush the write before we modify the GGTT */
intel_gt_flush_ggtt_writes(ggtt->vm.gt);
ggtt->vm.insert_page(&ggtt->vm,
@@ -636,7 +636,7 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
out_unpin:
mutex_lock(&i915->drm.struct_mutex);
intel_gt_flush_ggtt_writes(ggtt->vm.gt);
- if (node.allocated) {
+ if (drm_mm_node_allocated(&node)) {
ggtt->vm.clear_range(&ggtt->vm, node.start, node.size);
remove_mappable_node(&node);
} else {
diff --git a/drivers/gpu/drm/i915/i915_gem_evict.c b/drivers/gpu/drm/i915/i915_gem_evict.c
index e76c9da9992d..8c1e04f402bc 100644
--- a/drivers/gpu/drm/i915/i915_gem_evict.c
+++ b/drivers/gpu/drm/i915/i915_gem_evict.c
@@ -299,7 +299,7 @@ int i915_gem_evict_for_node(struct i915_address_space *vm,
break;
}
- GEM_BUG_ON(!node->allocated);
+ GEM_BUG_ON(!drm_mm_node_allocated(node));
vma = container_of(node, typeof(*vma), node);
/* If we are using coloring to insert guard pages between
diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
index 9d5b0f87c210..68c34b1a20e4 100644
--- a/drivers/gpu/drm/i915/i915_vma.c
+++ b/drivers/gpu/drm/i915/i915_vma.c
@@ -795,7 +795,7 @@ void i915_vma_reopen(struct i915_vma *vma)
static void __i915_vma_destroy(struct i915_vma *vma)
{
- GEM_BUG_ON(vma->node.allocated);
+ GEM_BUG_ON(drm_mm_node_allocated(&vma->node));
GEM_BUG_ON(vma->fence);
mutex_lock(&vma->vm->mutex);
diff --git a/drivers/gpu/drm/i915/i915_vma.h b/drivers/gpu/drm/i915/i915_vma.h
index 8bcb5812c446..e49b199f7de7 100644
--- a/drivers/gpu/drm/i915/i915_vma.h
+++ b/drivers/gpu/drm/i915/i915_vma.h
@@ -228,7 +228,7 @@ static inline bool i915_vma_is_closed(const struct i915_vma *vma)
static inline u32 i915_ggtt_offset(const struct i915_vma *vma)
{
GEM_BUG_ON(!i915_vma_is_ggtt(vma));
- GEM_BUG_ON(!vma->node.allocated);
+ GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
GEM_BUG_ON(upper_32_bits(vma->node.start));
GEM_BUG_ON(upper_32_bits(vma->node.start + vma->node.size - 1));
return lower_32_bits(vma->node.start);
@@ -390,7 +390,7 @@ static inline bool i915_vma_is_bound(const struct i915_vma *vma,
static inline bool i915_node_color_differs(const struct drm_mm_node *node,
unsigned long color)
{
- return node->allocated && node->color != color;
+ return drm_mm_node_allocated(node) && node->color != color;
}
/**
diff --git a/drivers/gpu/drm/selftests/test-drm_mm.c b/drivers/gpu/drm/selftests/test-drm_mm.c
index 388f9844f4ba..9aabe82dcd3a 100644
--- a/drivers/gpu/drm/selftests/test-drm_mm.c
+++ b/drivers/gpu/drm/selftests/test-drm_mm.c
@@ -854,7 +854,7 @@ static bool assert_contiguous_in_range(struct drm_mm *mm,
if (start > 0) {
node = __drm_mm_interval_first(mm, 0, start - 1);
- if (node->allocated) {
+ if (drm_mm_node_allocated(node)) {
pr_err("node before start: node=%llx+%llu, start=%llx\n",
node->start, node->size, start);
return false;
@@ -863,7 +863,7 @@ static bool assert_contiguous_in_range(struct drm_mm *mm,
if (end < U64_MAX) {
node = __drm_mm_interval_first(mm, end, U64_MAX);
- if (node->allocated) {
+ if (drm_mm_node_allocated(node)) {
pr_err("node after end: node=%llx+%llu, end=%llx\n",
node->start, node->size, end);
return false;
@@ -1156,12 +1156,12 @@ static void show_holes(const struct drm_mm *mm, int count)
struct drm_mm_node *next = list_next_entry(hole, node_list);
const char *node1 = NULL, *node2 = NULL;
- if (hole->allocated)
+ if (drm_mm_node_allocated(hole))
node1 = kasprintf(GFP_KERNEL,
"[%llx + %lld, color=%ld], ",
hole->start, hole->size, hole->color);
- if (next->allocated)
+ if (drm_mm_node_allocated(next))
node2 = kasprintf(GFP_KERNEL,
", [%llx + %lld, color=%ld]",
next->start, next->size, next->color);
@@ -1900,18 +1900,18 @@ static void separate_adjacent_colors(const struct drm_mm_node *node,
u64 *start,
u64 *end)
{
- if (node->allocated && node->color != color)
+ if (drm_mm_node_allocated(node) && node->color != color)
++*start;
node = list_next_entry(node, node_list);
- if (node->allocated && node->color != color)
+ if (drm_mm_node_allocated(node) && node->color != color)
--*end;
}
static bool colors_abutt(const struct drm_mm_node *node)
{
if (!drm_mm_hole_follows(node) &&
- list_next_entry(node, node_list)->allocated) {
+ drm_mm_node_allocated(list_next_entry(node, node_list))) {
pr_err("colors abutt; %ld [%llx + %llx] is next to %ld [%llx + %llx]!\n",
node->color, node->start, node->size,
list_next_entry(node, node_list)->color,
diff --git a/drivers/gpu/drm/vc4/vc4_crtc.c b/drivers/gpu/drm/vc4/vc4_crtc.c
index f1f0a7c87771..b00e20f5ce05 100644
--- a/drivers/gpu/drm/vc4/vc4_crtc.c
+++ b/drivers/gpu/drm/vc4/vc4_crtc.c
@@ -994,7 +994,7 @@ static void vc4_crtc_destroy_state(struct drm_crtc *crtc,
struct vc4_dev *vc4 = to_vc4_dev(crtc->dev);
struct vc4_crtc_state *vc4_state = to_vc4_crtc_state(state);
- if (vc4_state->mm.allocated) {
+ if (drm_mm_node_allocated(&vc4_state->mm)) {
unsigned long flags;
spin_lock_irqsave(&vc4->hvs->mm_lock, flags);
diff --git a/drivers/gpu/drm/vc4/vc4_hvs.c b/drivers/gpu/drm/vc4/vc4_hvs.c
index 9936b15d0bf1..5a43659da319 100644
--- a/drivers/gpu/drm/vc4/vc4_hvs.c
+++ b/drivers/gpu/drm/vc4/vc4_hvs.c
@@ -315,7 +315,7 @@ static void vc4_hvs_unbind(struct device *dev, struct device *master,
struct drm_device *drm = dev_get_drvdata(master);
struct vc4_dev *vc4 = drm->dev_private;
- if (vc4->hvs->mitchell_netravali_filter.allocated)
+ if (drm_mm_node_allocated(&vc4->hvs->mitchell_netravali_filter))
drm_mm_remove_node(&vc4->hvs->mitchell_netravali_filter);
drm_mm_takedown(&vc4->hvs->dlist_mm);
diff --git a/drivers/gpu/drm/vc4/vc4_plane.c b/drivers/gpu/drm/vc4/vc4_plane.c
index 5e5f90810aca..4934127f0d76 100644
--- a/drivers/gpu/drm/vc4/vc4_plane.c
+++ b/drivers/gpu/drm/vc4/vc4_plane.c
@@ -178,7 +178,7 @@ static void vc4_plane_destroy_state(struct drm_plane *plane,
struct vc4_dev *vc4 = to_vc4_dev(plane->dev);
struct vc4_plane_state *vc4_state = to_vc4_plane_state(state);
- if (vc4_state->lbm.allocated) {
+ if (drm_mm_node_allocated(&vc4_state->lbm)) {
unsigned long irqflags;
spin_lock_irqsave(&vc4->hvs->mm_lock, irqflags);
@@ -557,7 +557,7 @@ static int vc4_plane_allocate_lbm(struct drm_plane_state *state)
/* Allocate the LBM memory that the HVS will use for temporary
* storage due to our scaling/format conversion.
*/
- if (!vc4_state->lbm.allocated) {
+ if (!drm_mm_node_allocated(&vc4_state->lbm)) {
int ret;
spin_lock_irqsave(&vc4->hvs->mm_lock, irqflags);
diff --git a/include/drm/drm_mm.h b/include/drm/drm_mm.h
index 2c3bbb43c7d1..d7939c054259 100644
--- a/include/drm/drm_mm.h
+++ b/include/drm/drm_mm.h
@@ -168,8 +168,9 @@ struct drm_mm_node {
struct rb_node rb_hole_addr;
u64 __subtree_last;
u64 hole_size;
- bool allocated : 1;
- bool scanned_block : 1;
+ unsigned long flags;
+#define DRM_MM_NODE_ALLOCATED_BIT 0
+#define DRM_MM_NODE_SCANNED_BIT 1
#ifdef CONFIG_DRM_DEBUG_MM
depot_stack_handle_t stack;
#endif
@@ -253,7 +254,7 @@ struct drm_mm_scan {
*/
static inline bool drm_mm_node_allocated(const struct drm_mm_node *node)
{
- return node->allocated;
+ return test_bit(DRM_MM_NODE_ALLOCATED_BIT, &node->flags);
}
/**
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* [PATCH 03/27] drm/i915: Only track bound elements of the GTT
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
2019-09-25 10:01 ` [PATCH 01/27] dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling) Chris Wilson
2019-09-25 10:01 ` [PATCH 02/27] drm/mm: Pack allocated/scanned boolean into a bitfield Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:01 ` [PATCH 04/27] drm/i915: Mark up address spaces that may need to allocate Chris Wilson
` (27 subsequent siblings)
30 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx; +Cc: Matthew Auld
The premise here is to simply avoiding having to acquire the vm->mutex
inside vma create/destroy to update the vm->unbound_lists, to avoid some
nasty lock recursions later.
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_stolen.c | 2 +-
drivers/gpu/drm/i915/i915_gem_gtt.c | 23 ++++---------------
drivers/gpu/drm/i915/i915_gem_gtt.h | 5 ----
drivers/gpu/drm/i915/i915_vma.c | 12 ++--------
drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 2 +-
5 files changed, 8 insertions(+), 36 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
index bfbc3e3daf92..e45eb8721850 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
@@ -692,7 +692,7 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_i915_private *dev_priv
__i915_vma_set_map_and_fenceable(vma);
mutex_lock(&ggtt->vm.mutex);
- list_move_tail(&vma->vm_link, &ggtt->vm.bound_list);
+ list_add_tail(&vma->vm_link, &ggtt->vm.bound_list);
mutex_unlock(&ggtt->vm.mutex);
GEM_BUG_ON(i915_gem_object_is_shrinkable(obj));
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index e62e9d1a1307..ad9eb2d68f3f 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -505,19 +505,12 @@ static void i915_address_space_fini(struct i915_address_space *vm)
static void ppgtt_destroy_vma(struct i915_address_space *vm)
{
- struct list_head *phases[] = {
- &vm->bound_list,
- &vm->unbound_list,
- NULL,
- }, **phase;
+ struct i915_vma *vma, *vn;
mutex_lock(&vm->i915->drm.struct_mutex);
- for (phase = phases; *phase; phase++) {
- struct i915_vma *vma, *vn;
-
- list_for_each_entry_safe(vma, vn, *phase, vm_link)
- i915_vma_destroy(vma);
- }
+ list_for_each_entry_safe(vma, vn, &vm->bound_list, vm_link)
+ i915_vma_destroy(vma);
+ GEM_BUG_ON(!list_empty(&vm->bound_list));
mutex_unlock(&vm->i915->drm.struct_mutex);
}
@@ -528,9 +521,6 @@ static void __i915_vm_release(struct work_struct *work)
ppgtt_destroy_vma(vm);
- GEM_BUG_ON(!list_empty(&vm->bound_list));
- GEM_BUG_ON(!list_empty(&vm->unbound_list));
-
vm->cleanup(vm);
i915_address_space_fini(vm);
@@ -569,7 +559,6 @@ static void i915_address_space_init(struct i915_address_space *vm, int subclass)
stash_init(&vm->free_pages);
- INIT_LIST_HEAD(&vm->unbound_list);
INIT_LIST_HEAD(&vm->bound_list);
}
@@ -1887,10 +1876,6 @@ static struct i915_vma *pd_vma_create(struct gen6_ppgtt *ppgtt, int size)
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);
- mutex_unlock(&vma->vm->mutex);
-
return vma;
}
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 8fd2234ba0bf..bbdc735466c1 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -320,11 +320,6 @@ struct i915_address_space {
*/
struct list_head bound_list;
- /**
- * List of vma that are not unbound.
- */
- struct list_head unbound_list;
-
struct pagestash free_pages;
/* Global GTT */
diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
index 68c34b1a20e4..d097f77890ba 100644
--- a/drivers/gpu/drm/i915/i915_vma.c
+++ b/drivers/gpu/drm/i915/i915_vma.c
@@ -218,10 +218,6 @@ vma_create(struct drm_i915_gem_object *obj,
spin_unlock(&obj->vma.lock);
- mutex_lock(&vm->mutex);
- list_add(&vma->vm_link, &vm->unbound_list);
- mutex_unlock(&vm->mutex);
-
return vma;
err_vma:
@@ -657,7 +653,7 @@ i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
GEM_BUG_ON(!i915_gem_valid_gtt_space(vma, color));
mutex_lock(&vma->vm->mutex);
- list_move_tail(&vma->vm_link, &vma->vm->bound_list);
+ list_add_tail(&vma->vm_link, &vma->vm->bound_list);
mutex_unlock(&vma->vm->mutex);
if (vma->obj) {
@@ -685,7 +681,7 @@ i915_vma_remove(struct i915_vma *vma)
mutex_lock(&vma->vm->mutex);
drm_mm_remove_node(&vma->node);
- list_move_tail(&vma->vm_link, &vma->vm->unbound_list);
+ list_del(&vma->vm_link);
mutex_unlock(&vma->vm->mutex);
/*
@@ -798,10 +794,6 @@ static void __i915_vma_destroy(struct i915_vma *vma)
GEM_BUG_ON(drm_mm_node_allocated(&vma->node));
GEM_BUG_ON(vma->fence);
- mutex_lock(&vma->vm->mutex);
- list_del(&vma->vm_link);
- mutex_unlock(&vma->vm->mutex);
-
if (vma->obj) {
struct drm_i915_gem_object *obj = vma->obj;
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
index 0d40e0b42923..84d94eabc7dd 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
@@ -1242,7 +1242,7 @@ static void track_vma_bind(struct i915_vma *vma)
vma->pages = obj->mm.pages;
mutex_lock(&vma->vm->mutex);
- list_move_tail(&vma->vm_link, &vma->vm->bound_list);
+ list_add_tail(&vma->vm_link, &vma->vm->bound_list);
mutex_unlock(&vma->vm->mutex);
}
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* [PATCH 04/27] drm/i915: Mark up address spaces that may need to allocate
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (2 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 03/27] drm/i915: Only track bound elements of the GTT Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:01 ` [PATCH 05/27] drm/i915: Pull i915_vma_pin under the vm->mutex Chris Wilson
` (26 subsequent siblings)
30 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
Since we cannot allocate underneath the vm->mutex (it is used in the
direct-reclaim paths), we need to shift the allocations off into a
mutexless worker with fence recursion prevention. To know when we need
this protection, we mark up the address spaces that do allocate before
insertion.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_gem_gtt.c | 3 +++
drivers/gpu/drm/i915/i915_gem_gtt.h | 2 ++
2 files changed, 5 insertions(+)
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index ad9eb2d68f3f..483eeff8730e 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -1502,6 +1502,7 @@ static struct i915_ppgtt *gen8_ppgtt_create(struct drm_i915_private *i915)
goto err_free_pd;
}
+ ppgtt->vm.bind_alloc = I915_VMA_LOCAL_BIND;
ppgtt->vm.insert_entries = gen8_ppgtt_insert;
ppgtt->vm.allocate_va_range = gen8_ppgtt_alloc;
ppgtt->vm.clear_range = gen8_ppgtt_clear;
@@ -1950,6 +1951,7 @@ static struct i915_ppgtt *gen6_ppgtt_create(struct drm_i915_private *i915)
ppgtt_init(&ppgtt->base, &i915->gt);
ppgtt->base.vm.top = 1;
+ ppgtt->base.vm.bind_alloc = I915_VMA_LOCAL_BIND;
ppgtt->base.vm.allocate_va_range = gen6_alloc_va_range;
ppgtt->base.vm.clear_range = gen6_ppgtt_clear_range;
ppgtt->base.vm.insert_entries = gen6_ppgtt_insert_entries;
@@ -2581,6 +2583,7 @@ static int init_aliasing_ppgtt(struct i915_ggtt *ggtt)
goto err_ppgtt;
ggtt->alias = ppgtt;
+ ggtt->vm.bind_alloc |= ppgtt->vm.bind_alloc;
GEM_BUG_ON(ggtt->vm.vma_ops.bind_vma != ggtt_bind_vma);
ggtt->vm.vma_ops.bind_vma = aliasing_gtt_bind_vma;
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index bbdc735466c1..d8dfa3811e54 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -305,6 +305,8 @@ struct i915_address_space {
u64 total; /* size addr space maps (ex. 2GB for ggtt) */
u64 reserved; /* size addr space reserved */
+ unsigned int bind_alloc;
+
bool closed;
struct mutex mutex; /* protects vma and our lists */
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* [PATCH 05/27] drm/i915: Pull i915_vma_pin under the vm->mutex
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (3 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 04/27] drm/i915: Mark up address spaces that may need to allocate Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-27 8:47 ` Tvrtko Ursulin
2019-09-25 10:01 ` [PATCH 06/27] drm/i915: Push the i915_active.retire into a worker Chris Wilson
` (25 subsequent siblings)
30 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
Replace the struct_mutex requirement for pinning the i915_vma with the
local vm->mutex instead. Note that the vm->mutex is tainted by the
shrinker (we require unbinding from inside fs-reclaim) and so we cannot
allocate while holding that mutex. Instead we have to preallocate
workers to do allocate and apply the PTE updates after we have we
reserved their slot in the drm_mm (using fences to order the PTE writes
with the GPU work and with later unbind).
In adding the asynchronous vma binding, one subtle requirement is to
avoid coupling the binding fence into the backing object->resv. That is
the asynchronous binding only applies to the vma timeline itself and not
to the pages as that is a more global timeline (the binding of one vma
does not need to be ordered with another vma, nor does the implicit GEM
fencing depend on a vma, only on writes to the backing store). Keeping
the vma binding distinct from the backing store timelines is verified by
a number of async gem_exec_fence and gem_exec_schedule tests. The way we
do this is quite simple, we keep the fence for the vma binding separate
and only wait on it as required, and never add it to the obj->resv
itself.
Another consequence in reducing the locking around the vma is the
destruction of the vma is no longer globally serialised by struct_mutex.
A natural solution would be to add a kref to i915_vma, but that requires
decoupling the reference cycles, possibly by introducing a new
i915_mm_pages object that is own by both obj->mm and vma->pages.
However, we have not taken that route due to the overshadowing lmem/ttm
discussions, and instead play a series of complicated games with
trylocks to (hopefully) ensure that only one destruction path is called!
v2: Add some commentary, and some helpers to reduce patch churn.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/display/intel_display.c | 29 +-
drivers/gpu/drm/i915/display/intel_fbdev.c | 8 +-
drivers/gpu/drm/i915/display/intel_overlay.c | 11 +-
.../gpu/drm/i915/gem/i915_gem_client_blt.c | 2 +-
drivers/gpu/drm/i915/gem/i915_gem_context.c | 20 +-
drivers/gpu/drm/i915/gem/i915_gem_domain.c | 13 +-
.../gpu/drm/i915/gem/i915_gem_execbuffer.c | 43 +-
drivers/gpu/drm/i915/gem/i915_gem_mman.c | 20 +-
drivers/gpu/drm/i915/gem/i915_gem_object.c | 33 +-
drivers/gpu/drm/i915/gem/i915_gem_object.h | 5 +
drivers/gpu/drm/i915/gem/i915_gem_shrinker.c | 73 +--
drivers/gpu/drm/i915/gem/i915_gem_stolen.c | 8 +-
drivers/gpu/drm/i915/gem/i915_gem_tiling.c | 27 +-
drivers/gpu/drm/i915/gem/i915_gem_userptr.c | 27 +-
.../gpu/drm/i915/gem/selftests/huge_pages.c | 9 +-
.../drm/i915/gem/selftests/i915_gem_context.c | 12 +-
.../drm/i915/gem/selftests/i915_gem_mman.c | 4 -
.../drm/i915/gem/selftests/igt_gem_utils.c | 7 +-
drivers/gpu/drm/i915/gt/intel_gt.c | 5 +-
drivers/gpu/drm/i915/gt/intel_ringbuffer.c | 4 +-
drivers/gpu/drm/i915/gt/selftest_hangcheck.c | 19 +-
drivers/gpu/drm/i915/gvt/aperture_gm.c | 12 +-
drivers/gpu/drm/i915/i915_active.c | 95 +++-
drivers/gpu/drm/i915/i915_active.h | 7 +
drivers/gpu/drm/i915/i915_active_types.h | 5 +
drivers/gpu/drm/i915/i915_drv.c | 2 -
drivers/gpu/drm/i915/i915_gem.c | 83 ++-
drivers/gpu/drm/i915/i915_gem_evict.c | 28 +-
drivers/gpu/drm/i915/i915_gem_fence_reg.c | 9 +-
drivers/gpu/drm/i915/i915_gem_gtt.c | 109 ++--
drivers/gpu/drm/i915/i915_gem_gtt.h | 45 +-
drivers/gpu/drm/i915/i915_perf.c | 32 +-
drivers/gpu/drm/i915/i915_vma.c | 524 ++++++++++++------
drivers/gpu/drm/i915/i915_vma.h | 84 +--
drivers/gpu/drm/i915/selftests/i915_gem.c | 2 -
.../gpu/drm/i915/selftests/i915_gem_evict.c | 36 +-
drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 58 +-
drivers/gpu/drm/i915/selftests/i915_request.c | 7 +
drivers/gpu/drm/i915/selftests/i915_vma.c | 6 +-
39 files changed, 823 insertions(+), 700 deletions(-)
diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
index 5ecf54270181..12abcd1e471b 100644
--- a/drivers/gpu/drm/i915/display/intel_display.c
+++ b/drivers/gpu/drm/i915/display/intel_display.c
@@ -2079,7 +2079,6 @@ intel_pin_and_fence_fb_obj(struct drm_framebuffer *fb,
unsigned int pinctl;
u32 alignment;
- WARN_ON(!mutex_is_locked(&dev->struct_mutex));
if (WARN_ON(!i915_gem_object_is_framebuffer(obj)))
return ERR_PTR(-EINVAL);
@@ -2163,8 +2162,6 @@ intel_pin_and_fence_fb_obj(struct drm_framebuffer *fb,
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);
@@ -3065,12 +3062,10 @@ intel_alloc_initial_plane_obj(struct intel_crtc *crtc,
return false;
}
- mutex_lock(&dev->struct_mutex);
obj = i915_gem_object_create_stolen_for_preallocated(dev_priv,
base_aligned,
base_aligned,
size_aligned);
- mutex_unlock(&dev->struct_mutex);
if (!obj)
return false;
@@ -3232,13 +3227,11 @@ intel_find_initial_plane_obj(struct intel_crtc *intel_crtc,
intel_state->color_plane[0].stride =
intel_fb_pitch(fb, 0, intel_state->base.rotation);
- mutex_lock(&dev->struct_mutex);
intel_state->vma =
intel_pin_and_fence_fb_obj(fb,
&intel_state->view,
intel_plane_uses_fence(intel_state),
&intel_state->flags);
- mutex_unlock(&dev->struct_mutex);
if (IS_ERR(intel_state->vma)) {
DRM_ERROR("failed to pin boot fb on pipe %d: %li\n",
intel_crtc->pipe, PTR_ERR(intel_state->vma));
@@ -14368,8 +14361,6 @@ static void fb_obj_bump_render_priority(struct drm_i915_gem_object *obj)
* bits. Some older platforms need special physical address handling for
* cursor planes.
*
- * Must be called with struct_mutex held.
- *
* Returns 0 on success, negative error code on failure.
*/
int
@@ -14426,15 +14417,8 @@ intel_prepare_plane_fb(struct drm_plane *plane,
if (ret)
return ret;
- ret = mutex_lock_interruptible(&dev_priv->drm.struct_mutex);
- if (ret) {
- i915_gem_object_unpin_pages(obj);
- return ret;
- }
-
ret = intel_plane_pin_fb(to_intel_plane_state(new_state));
- mutex_unlock(&dev_priv->drm.struct_mutex);
i915_gem_object_unpin_pages(obj);
if (ret)
return ret;
@@ -14483,8 +14467,6 @@ intel_prepare_plane_fb(struct drm_plane *plane,
* @old_state: the state from the previous modeset
*
* Cleans up a framebuffer that has just been removed from a plane.
- *
- * Must be called with struct_mutex held.
*/
void
intel_cleanup_plane_fb(struct drm_plane *plane,
@@ -14500,9 +14482,7 @@ intel_cleanup_plane_fb(struct drm_plane *plane,
}
/* Should only be called after a successful intel_prepare_plane_fb()! */
- mutex_lock(&dev_priv->drm.struct_mutex);
intel_plane_unpin_fb(to_intel_plane_state(old_state));
- mutex_unlock(&dev_priv->drm.struct_mutex);
}
int
@@ -14705,7 +14685,6 @@ intel_legacy_cursor_update(struct drm_plane *plane,
u32 src_w, u32 src_h,
struct drm_modeset_acquire_ctx *ctx)
{
- struct drm_i915_private *dev_priv = to_i915(crtc->dev);
struct drm_plane_state *old_plane_state, *new_plane_state;
struct intel_plane *intel_plane = to_intel_plane(plane);
struct intel_crtc_state *crtc_state =
@@ -14771,13 +14750,9 @@ intel_legacy_cursor_update(struct drm_plane *plane,
if (ret)
goto out_free;
- ret = mutex_lock_interruptible(&dev_priv->drm.struct_mutex);
- if (ret)
- goto out_free;
-
ret = intel_plane_pin_fb(to_intel_plane_state(new_plane_state));
if (ret)
- goto out_unlock;
+ goto out_free;
intel_frontbuffer_flush(to_intel_frontbuffer(fb), ORIGIN_FLIP);
intel_frontbuffer_track(to_intel_frontbuffer(old_plane_state->fb),
@@ -14807,8 +14782,6 @@ intel_legacy_cursor_update(struct drm_plane *plane,
intel_plane_unpin_fb(to_intel_plane_state(old_plane_state));
-out_unlock:
- mutex_unlock(&dev_priv->drm.struct_mutex);
out_free:
if (new_crtc_state)
intel_crtc_destroy_state(crtc, &new_crtc_state->base);
diff --git a/drivers/gpu/drm/i915/display/intel_fbdev.c b/drivers/gpu/drm/i915/display/intel_fbdev.c
index 68338669f054..97cde017670a 100644
--- a/drivers/gpu/drm/i915/display/intel_fbdev.c
+++ b/drivers/gpu/drm/i915/display/intel_fbdev.c
@@ -204,7 +204,6 @@ static int intelfb_create(struct drm_fb_helper *helper,
sizes->fb_height = intel_fb->base.height;
}
- mutex_lock(&dev->struct_mutex);
wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm);
/* Pin the GGTT vma for our access via info->screen_base.
@@ -266,7 +265,6 @@ static int intelfb_create(struct drm_fb_helper *helper,
ifbdev->vma_flags = flags;
intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref);
- mutex_unlock(&dev->struct_mutex);
vga_switcheroo_client_fb_set(pdev, info);
return 0;
@@ -274,7 +272,6 @@ static int intelfb_create(struct drm_fb_helper *helper,
intel_unpin_fb_vma(vma, flags);
out_unlock:
intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref);
- mutex_unlock(&dev->struct_mutex);
return ret;
}
@@ -291,11 +288,8 @@ static void intel_fbdev_destroy(struct intel_fbdev *ifbdev)
drm_fb_helper_fini(&ifbdev->helper);
- if (ifbdev->vma) {
- mutex_lock(&ifbdev->helper.dev->struct_mutex);
+ if (ifbdev->vma)
intel_unpin_fb_vma(ifbdev->vma, ifbdev->vma_flags);
- mutex_unlock(&ifbdev->helper.dev->struct_mutex);
- }
if (ifbdev->fb)
drm_framebuffer_remove(&ifbdev->fb->base);
diff --git a/drivers/gpu/drm/i915/display/intel_overlay.c b/drivers/gpu/drm/i915/display/intel_overlay.c
index 5efef9babadb..3f4ac1ee7668 100644
--- a/drivers/gpu/drm/i915/display/intel_overlay.c
+++ b/drivers/gpu/drm/i915/display/intel_overlay.c
@@ -1303,15 +1303,11 @@ static int get_registers(struct intel_overlay *overlay, bool use_phys)
struct i915_vma *vma;
int err;
- mutex_lock(&i915->drm.struct_mutex);
-
obj = i915_gem_object_create_stolen(i915, PAGE_SIZE);
if (obj == NULL)
obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
- if (IS_ERR(obj)) {
- err = PTR_ERR(obj);
- goto err_unlock;
- }
+ if (IS_ERR(obj))
+ return PTR_ERR(obj);
vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0, PIN_MAPPABLE);
if (IS_ERR(vma)) {
@@ -1332,13 +1328,10 @@ 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;
}
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c b/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c
index 7f61a8024133..c1fca5728e6e 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c
@@ -211,7 +211,7 @@ static void clear_pages_worker(struct work_struct *work)
* keep track of the GPU activity within this vma/request, and
* propagate the signal from the request to w->dma.
*/
- err = i915_active_add_request(&vma->active, rq);
+ err = __i915_vma_move_to_active(vma, rq);
if (err)
goto out_request;
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index f7ba0935ed67..95f8e66e45db 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -313,8 +313,6 @@ static void i915_gem_context_free(struct i915_gem_context *ctx)
GEM_BUG_ON(!i915_gem_context_is_closed(ctx));
release_hw_id(ctx);
- if (ctx->vm)
- i915_vm_put(ctx->vm);
free_engines(rcu_access_pointer(ctx->engines));
mutex_destroy(&ctx->engines_mutex);
@@ -379,9 +377,13 @@ void i915_gem_context_release(struct kref *ref)
static void context_close(struct i915_gem_context *ctx)
{
+ i915_gem_context_set_closed(ctx);
+
+ if (ctx->vm)
+ i915_vm_close(ctx->vm);
+
mutex_lock(&ctx->mutex);
- i915_gem_context_set_closed(ctx);
ctx->file_priv = ERR_PTR(-EBADF);
/*
@@ -474,7 +476,7 @@ __set_ppgtt(struct i915_gem_context *ctx, struct i915_address_space *vm)
GEM_BUG_ON(old && i915_vm_is_4lvl(vm) != i915_vm_is_4lvl(old));
- ctx->vm = i915_vm_get(vm);
+ ctx->vm = i915_vm_open(vm);
context_apply_all(ctx, __apply_ppgtt, vm);
return old;
@@ -488,7 +490,7 @@ static void __assign_ppgtt(struct i915_gem_context *ctx,
vm = __set_ppgtt(ctx, vm);
if (vm)
- i915_vm_put(vm);
+ i915_vm_close(vm);
}
static void __set_timeline(struct intel_timeline **dst,
@@ -953,7 +955,7 @@ static int get_ppgtt(struct drm_i915_file_private *file_priv,
if (ret < 0)
goto err_unlock;
- i915_vm_get(vm);
+ i915_vm_open(vm);
args->size = 0;
args->value = ret;
@@ -973,7 +975,7 @@ static void set_ppgtt_barrier(void *data)
if (INTEL_GEN(old->i915) < 8)
gen6_ppgtt_unpin_all(i915_vm_to_ppgtt(old));
- i915_vm_put(old);
+ i915_vm_close(old);
}
static int emit_ppgtt_update(struct i915_request *rq, void *data)
@@ -1090,8 +1092,8 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv,
set_ppgtt_barrier,
old);
if (err) {
- i915_vm_put(__set_ppgtt(ctx, old));
- i915_vm_put(old);
+ i915_vm_close(__set_ppgtt(ctx, old));
+ i915_vm_close(old);
}
unlock:
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_domain.c b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
index 55c3ab59e3aa..9937b4c341f1 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_domain.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
@@ -288,7 +288,12 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
if (!drm_mm_node_allocated(&vma->node))
continue;
- ret = i915_vma_bind(vma, cache_level, PIN_UPDATE);
+ /* Wait for an earlier async bind, need to rewrite it */
+ ret = i915_vma_sync(vma);
+ if (ret)
+ return ret;
+
+ ret = i915_vma_bind(vma, cache_level, PIN_UPDATE, NULL);
if (ret)
return ret;
}
@@ -391,16 +396,11 @@ int i915_gem_set_caching_ioctl(struct drm_device *dev, void *data,
if (ret)
goto out;
- ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
- if (ret)
- goto out;
-
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);
@@ -485,6 +485,7 @@ static void i915_gem_object_bump_inactive_ggtt(struct drm_i915_gem_object *obj)
if (!drm_mm_node_allocated(&vma->node))
continue;
+ GEM_BUG_ON(vma->vm != &i915->ggtt.vm);
list_move_tail(&vma->vm_link, &vma->vm->bound_list);
}
mutex_unlock(&i915->ggtt.vm.mutex);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index c049199a1df5..88a881be12ec 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -698,7 +698,9 @@ static int eb_reserve(struct i915_execbuffer *eb)
case 1:
/* Too fragmented, unbind everything and retry */
+ mutex_lock(&eb->context->vm->mutex);
err = i915_gem_evict_vm(eb->context->vm);
+ mutex_unlock(&eb->context->vm->mutex);
if (err)
return err;
break;
@@ -972,7 +974,9 @@ static void reloc_cache_reset(struct reloc_cache *cache)
ggtt->vm.clear_range(&ggtt->vm,
cache->node.start,
cache->node.size);
+ mutex_lock(&ggtt->vm.mutex);
drm_mm_remove_node(&cache->node);
+ mutex_unlock(&ggtt->vm.mutex);
} else {
i915_vma_unpin((struct i915_vma *)cache->node.mm);
}
@@ -1047,11 +1051,13 @@ static void *reloc_iomap(struct drm_i915_gem_object *obj,
PIN_NOEVICT);
if (IS_ERR(vma)) {
memset(&cache->node, 0, sizeof(cache->node));
+ mutex_lock(&ggtt->vm.mutex);
err = drm_mm_insert_node_in_range
(&ggtt->vm.mm, &cache->node,
PAGE_SIZE, 0, I915_COLOR_UNEVICTABLE,
0, ggtt->mappable_end,
DRM_MM_INSERT_LOW);
+ mutex_unlock(&ggtt->vm.mutex);
if (err) /* no inactive aperture space, use cpu reloc */
return NULL;
} else {
@@ -1416,7 +1422,7 @@ eb_relocate_entry(struct i915_execbuffer *eb,
if (reloc->write_domain == I915_GEM_DOMAIN_INSTRUCTION &&
IS_GEN(eb->i915, 6)) {
err = i915_vma_bind(target, target->obj->cache_level,
- PIN_GLOBAL);
+ PIN_GLOBAL, NULL);
if (WARN_ONCE(err,
"Unexpected failure to bind target VMA!"))
return err;
@@ -2140,35 +2146,6 @@ static struct i915_request *eb_throttle(struct intel_context *ce)
return i915_request_get(rq);
}
-static int
-__eb_pin_context(struct i915_execbuffer *eb, struct intel_context *ce)
-{
- int err;
-
- if (likely(atomic_inc_not_zero(&ce->pin_count)))
- return 0;
-
- err = mutex_lock_interruptible(&eb->i915->drm.struct_mutex);
- if (err)
- return err;
-
- err = __intel_context_do_pin(ce);
- mutex_unlock(&eb->i915->drm.struct_mutex);
-
- return err;
-}
-
-static void
-__eb_unpin_context(struct i915_execbuffer *eb, struct intel_context *ce)
-{
- if (likely(atomic_add_unless(&ce->pin_count, -1, 1)))
- return;
-
- mutex_lock(&eb->i915->drm.struct_mutex);
- intel_context_unpin(ce);
- mutex_unlock(&eb->i915->drm.struct_mutex);
-}
-
static int __eb_pin_engine(struct i915_execbuffer *eb, struct intel_context *ce)
{
struct intel_timeline *tl;
@@ -2188,7 +2165,7 @@ static int __eb_pin_engine(struct i915_execbuffer *eb, struct intel_context *ce)
* GGTT space, so do this first before we reserve a seqno for
* ourselves.
*/
- err = __eb_pin_context(eb, ce);
+ err = intel_context_pin(ce);
if (err)
return err;
@@ -2232,7 +2209,7 @@ static int __eb_pin_engine(struct i915_execbuffer *eb, struct intel_context *ce)
intel_context_exit(ce);
intel_context_timeline_unlock(tl);
err_unpin:
- __eb_unpin_context(eb, ce);
+ intel_context_unpin(ce);
return err;
}
@@ -2245,7 +2222,7 @@ static void eb_unpin_engine(struct i915_execbuffer *eb)
intel_context_exit(ce);
mutex_unlock(&tl->mutex);
- __eb_unpin_context(eb, ce);
+ intel_context_unpin(ce);
}
static unsigned int
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
index 860b751c51f1..c44614895487 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_mman.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
@@ -249,16 +249,6 @@ vm_fault_t i915_gem_fault(struct vm_fault *vmf)
if (ret)
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 |
@@ -291,7 +281,13 @@ vm_fault_t i915_gem_fault(struct vm_fault *vmf)
}
if (IS_ERR(vma)) {
ret = PTR_ERR(vma);
- goto err_unlock;
+ 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_unpin;
}
ret = i915_vma_pin_fence(vma);
@@ -329,8 +325,6 @@ vm_fault_t i915_gem_fault(struct vm_fault *vmf)
i915_vma_unpin_fence(vma);
err_unpin:
__i915_vma_unpin(vma);
-err_unlock:
- mutex_unlock(&dev->struct_mutex);
err_reset:
intel_gt_reset_unlock(ggtt->vm.gt, srcu);
err_rpm:
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
index 0ef60dae23a7..dbf9be9a79f4 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
@@ -155,21 +155,30 @@ static void __i915_gem_free_objects(struct drm_i915_private *i915,
wakeref = intel_runtime_pm_get(&i915->runtime_pm);
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);
-
- list_for_each_entry_safe(vma, vn, &obj->vma.list, obj_link) {
- GEM_BUG_ON(i915_vma_is_active(vma));
- atomic_and(~I915_VMA_PIN_MASK, &vma->flags);
- i915_vma_destroy(vma);
+ if (!list_empty(&obj->vma.list)) {
+ struct i915_vma *vma;
+
+ /*
+ * Note that the vma keeps an object reference while
+ * it is active, so it *should* not sleep while we
+ * destroy it. Our debug code errs insits it *might*.
+ * For the moment, play along.
+ */
+ spin_lock(&obj->vma.lock);
+ while ((vma = list_first_entry_or_null(&obj->vma.list,
+ struct i915_vma,
+ obj_link))) {
+ GEM_BUG_ON(vma->obj != obj);
+ spin_unlock(&obj->vma.lock);
+
+ i915_vma_destroy(vma);
+
+ spin_lock(&obj->vma.lock);
+ }
+ spin_unlock(&obj->vma.lock);
}
- GEM_BUG_ON(!list_empty(&obj->vma.list));
- GEM_BUG_ON(!RB_EMPTY_ROOT(&obj->vma.tree));
-
- mutex_unlock(&i915->drm.struct_mutex);
GEM_BUG_ON(atomic_read(&obj->bind_count));
GEM_BUG_ON(obj->userfault_count);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index 29b9eddc4c7f..a78af25dce36 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -106,6 +106,11 @@ static inline void i915_gem_object_lock(struct drm_i915_gem_object *obj)
dma_resv_lock(obj->base.resv, NULL);
}
+static inline bool i915_gem_object_trylock(struct drm_i915_gem_object *obj)
+{
+ return dma_resv_trylock(obj->base.resv);
+}
+
static inline int
i915_gem_object_lock_interruptible(struct drm_i915_gem_object *obj)
{
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c b/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
index d2c05d752909..fd3ce6da8497 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
@@ -16,40 +16,6 @@
#include "i915_trace.h"
-static bool shrinker_lock(struct drm_i915_private *i915,
- unsigned int flags,
- bool *unlock)
-{
- struct mutex *m = &i915->drm.struct_mutex;
-
- switch (mutex_trylock_recursive(m)) {
- case MUTEX_TRYLOCK_RECURSIVE:
- *unlock = false;
- return true;
-
- case MUTEX_TRYLOCK_FAILED:
- *unlock = false;
- if (flags & I915_SHRINK_ACTIVE &&
- mutex_lock_killable_nested(m, I915_MM_SHRINKER) == 0)
- *unlock = true;
- return *unlock;
-
- case MUTEX_TRYLOCK_SUCCESS:
- *unlock = true;
- return true;
- }
-
- BUG();
-}
-
-static void shrinker_unlock(struct drm_i915_private *i915, bool unlock)
-{
- if (!unlock)
- return;
-
- mutex_unlock(&i915->drm.struct_mutex);
-}
-
static bool swap_available(void)
{
return get_nr_swap_pages() > 0;
@@ -155,10 +121,6 @@ i915_gem_shrink(struct drm_i915_private *i915,
intel_wakeref_t wakeref = 0;
unsigned long count = 0;
unsigned long scanned = 0;
- bool unlock;
-
- if (!shrinker_lock(i915, shrink, &unlock))
- return 0;
/*
* When shrinking the active list, we should also consider active
@@ -268,8 +230,6 @@ i915_gem_shrink(struct drm_i915_private *i915,
if (shrink & I915_SHRINK_BOUND)
intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- shrinker_unlock(i915, unlock);
-
if (nr_scanned)
*nr_scanned += scanned;
return count;
@@ -339,19 +299,14 @@ i915_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc)
struct drm_i915_private *i915 =
container_of(shrinker, struct drm_i915_private, mm.shrinker);
unsigned long freed;
- bool unlock;
sc->nr_scanned = 0;
- if (!shrinker_lock(i915, 0, &unlock))
- return SHRINK_STOP;
-
freed = i915_gem_shrink(i915,
sc->nr_to_scan,
&sc->nr_scanned,
I915_SHRINK_BOUND |
- I915_SHRINK_UNBOUND |
- I915_SHRINK_WRITEBACK);
+ I915_SHRINK_UNBOUND);
if (sc->nr_scanned < sc->nr_to_scan && current_is_kswapd()) {
intel_wakeref_t wakeref;
@@ -366,8 +321,6 @@ i915_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc)
}
}
- shrinker_unlock(i915, unlock);
-
return sc->nr_scanned ? freed : SHRINK_STOP;
}
@@ -384,6 +337,7 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr)
freed_pages = 0;
with_intel_runtime_pm(&i915->runtime_pm, wakeref)
freed_pages += i915_gem_shrink(i915, -1UL, NULL,
+ I915_SHRINK_ACTIVE |
I915_SHRINK_BOUND |
I915_SHRINK_UNBOUND |
I915_SHRINK_WRITEBACK);
@@ -419,10 +373,6 @@ i915_gem_shrinker_vmap(struct notifier_block *nb, unsigned long event, void *ptr
struct i915_vma *vma, *next;
unsigned long freed_pages = 0;
intel_wakeref_t wakeref;
- bool unlock;
-
- if (!shrinker_lock(i915, 0, &unlock))
- return NOTIFY_DONE;
with_intel_runtime_pm(&i915->runtime_pm, wakeref)
freed_pages += i915_gem_shrink(i915, -1UL, NULL,
@@ -439,15 +389,11 @@ i915_gem_shrinker_vmap(struct notifier_block *nb, unsigned long event, void *ptr
if (!vma->iomap || i915_vma_is_active(vma))
continue;
- mutex_unlock(&i915->ggtt.vm.mutex);
- if (i915_vma_unbind(vma) == 0)
+ if (__i915_vma_unbind(vma) == 0)
freed_pages += count;
- mutex_lock(&i915->ggtt.vm.mutex);
}
mutex_unlock(&i915->ggtt.vm.mutex);
- shrinker_unlock(i915, unlock);
-
*(unsigned long *)ptr += freed_pages;
return NOTIFY_DONE;
}
@@ -490,22 +436,9 @@ void i915_gem_shrinker_taints_mutex(struct drm_i915_private *i915,
fs_reclaim_acquire(GFP_KERNEL);
- /*
- * As we invariably rely on the struct_mutex within the shrinker,
- * but have a complicated recursion dance, taint all the mutexes used
- * within the shrinker with the struct_mutex. For completeness, we
- * taint with all subclass of struct_mutex, even though we should
- * only need tainting by I915_MM_NORMAL to catch possible ABBA
- * deadlocks from using struct_mutex inside @mutex.
- */
- mutex_acquire(&i915->drm.struct_mutex.dep_map,
- I915_MM_SHRINKER, 0, _RET_IP_);
-
mutex_acquire(&mutex->dep_map, 0, 0, _RET_IP_);
mutex_release(&mutex->dep_map, 0, _RET_IP_);
- mutex_release(&i915->drm.struct_mutex.dep_map, 0, _RET_IP_);
-
fs_reclaim_release(GFP_KERNEL);
if (unlock)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
index e45eb8721850..fad98a921cde 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
@@ -624,8 +624,6 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_i915_private *dev_priv
if (!drm_mm_initialized(&dev_priv->mm.stolen))
return NULL;
- lockdep_assert_held(&dev_priv->drm.struct_mutex);
-
DRM_DEBUG_DRIVER("creating preallocated stolen object: stolen_offset=%pa, gtt_offset=%pa, size=%pa\n",
&stolen_offset, >t_offset, &size);
@@ -677,21 +675,25 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_i915_private *dev_priv
* setting up the GTT space. The actual reservation will occur
* later.
*/
+ mutex_lock(&ggtt->vm.mutex);
ret = i915_gem_gtt_reserve(&ggtt->vm, &vma->node,
size, gtt_offset, obj->cache_level,
0);
if (ret) {
DRM_DEBUG_DRIVER("failed to allocate stolen GTT space\n");
+ mutex_unlock(&ggtt->vm.mutex);
goto err_pages;
}
GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
+ GEM_BUG_ON(vma->pages);
vma->pages = obj->mm.pages;
+ atomic_set(&vma->pages_count, I915_VMA_PAGES_ACTIVE);
+
set_bit(I915_VMA_GLOBAL_BIND_BIT, __i915_vma_flags(vma));
__i915_vma_set_map_and_fenceable(vma);
- mutex_lock(&ggtt->vm.mutex);
list_add_tail(&vma->vm_link, &ggtt->vm.bound_list);
mutex_unlock(&ggtt->vm.mutex);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_tiling.c b/drivers/gpu/drm/i915/gem/i915_gem_tiling.c
index e5d1ae8d4dba..dc2a83ce44d5 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_tiling.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_tiling.c
@@ -181,22 +181,25 @@ static int
i915_gem_object_fence_prepare(struct drm_i915_gem_object *obj,
int tiling_mode, unsigned int stride)
{
+ struct i915_ggtt *ggtt = &to_i915(obj->base.dev)->ggtt;
struct i915_vma *vma;
- int ret;
+ int ret = 0;
if (tiling_mode == I915_TILING_NONE)
return 0;
+ mutex_lock(&ggtt->vm.mutex);
for_each_ggtt_vma(vma, obj) {
if (i915_vma_fence_prepare(vma, tiling_mode, stride))
continue;
- ret = i915_vma_unbind(vma);
+ ret = __i915_vma_unbind(vma);
if (ret)
- return ret;
+ break;
}
+ mutex_unlock(&ggtt->vm.mutex);
- return 0;
+ return ret;
}
int
@@ -212,7 +215,6 @@ i915_gem_object_set_tiling(struct drm_i915_gem_object *obj,
GEM_BUG_ON(!i915_tiling_ok(obj, tiling, stride));
GEM_BUG_ON(!stride ^ (tiling == I915_TILING_NONE));
- lockdep_assert_held(&i915->drm.struct_mutex);
if ((tiling | stride) == obj->tiling_and_stride)
return 0;
@@ -233,16 +235,18 @@ i915_gem_object_set_tiling(struct drm_i915_gem_object *obj,
* whilst executing a fenced command for an untiled object.
*/
- err = i915_gem_object_fence_prepare(obj, tiling, stride);
- if (err)
- return err;
-
i915_gem_object_lock(obj);
if (i915_gem_object_is_framebuffer(obj)) {
i915_gem_object_unlock(obj);
return -EBUSY;
}
+ err = i915_gem_object_fence_prepare(obj, tiling, stride);
+ if (err) {
+ i915_gem_object_unlock(obj);
+ return err;
+ }
+
/* If the memory has unknown (i.e. varying) swizzling, we pin the
* pages to prevent them being swapped out and causing corruption
* due to the change in swizzling.
@@ -368,12 +372,7 @@ i915_gem_set_tiling_ioctl(struct drm_device *dev, void *data,
}
}
- err = mutex_lock_interruptible(&dev->struct_mutex);
- if (err)
- goto err;
-
err = i915_gem_object_set_tiling(obj, args->tiling_mode, args->stride);
- mutex_unlock(&dev->struct_mutex);
/* We have to maintain this existing ABI... */
args->stride = i915_gem_object_get_stride(obj);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
index 11b231c187c5..27573ef903b4 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
@@ -92,7 +92,6 @@ userptr_mn_invalidate_range_start(struct mmu_notifier *_mn,
struct i915_mmu_notifier *mn =
container_of(_mn, struct i915_mmu_notifier, mn);
struct interval_tree_node *it;
- struct mutex *unlock = NULL;
unsigned long end;
int ret = 0;
@@ -129,33 +128,13 @@ userptr_mn_invalidate_range_start(struct mmu_notifier *_mn,
}
spin_unlock(&mn->lock);
- if (!unlock) {
- unlock = &mn->mm->i915->drm.struct_mutex;
-
- switch (mutex_trylock_recursive(unlock)) {
- default:
- case MUTEX_TRYLOCK_FAILED:
- if (mutex_lock_killable_nested(unlock, I915_MM_SHRINKER)) {
- i915_gem_object_put(obj);
- return -EINTR;
- }
- /* fall through */
- case MUTEX_TRYLOCK_SUCCESS:
- break;
-
- case MUTEX_TRYLOCK_RECURSIVE:
- unlock = ERR_PTR(-EEXIST);
- break;
- }
- }
-
ret = i915_gem_object_unbind(obj,
I915_GEM_OBJECT_UNBIND_ACTIVE);
if (ret == 0)
ret = __i915_gem_object_put_pages(obj, I915_MM_SHRINKER);
i915_gem_object_put(obj);
if (ret)
- goto unlock;
+ return ret;
spin_lock(&mn->lock);
@@ -168,10 +147,6 @@ userptr_mn_invalidate_range_start(struct mmu_notifier *_mn,
}
spin_unlock(&mn->lock);
-unlock:
- if (!IS_ERR_OR_NULL(unlock))
- mutex_unlock(unlock);
-
return ret;
}
diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
index c5cea4379216..98b2a6ccfcc1 100644
--- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
+++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
@@ -333,7 +333,12 @@ static int igt_check_page_sizes(struct i915_vma *vma)
struct drm_i915_private *i915 = vma->vm->i915;
unsigned int supported = INTEL_INFO(i915)->page_sizes;
struct drm_i915_gem_object *obj = vma->obj;
- int err = 0;
+ int err;
+
+ /* We have to wait for the async bind to complete before our asserts */
+ err = i915_vma_sync(vma);
+ if (err)
+ return err;
if (!HAS_PAGE_SIZES(i915, vma->page_sizes.sg)) {
pr_err("unsupported page_sizes.sg=%u, supported=%u\n",
@@ -1390,7 +1395,7 @@ static int igt_ppgtt_pin_update(void *arg)
goto out_unpin;
}
- err = i915_vma_bind(vma, I915_CACHE_NONE, PIN_UPDATE);
+ err = i915_vma_bind(vma, I915_CACHE_NONE, PIN_UPDATE, NULL);
if (err)
goto out_unpin;
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 dc25bcc3e372..6ef3d745dc8d 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
@@ -747,10 +747,7 @@ emit_rpcs_query(struct drm_i915_gem_object *obj,
if (err)
goto skip_request;
- i915_vma_unpin(batch);
- i915_vma_close(batch);
- i915_vma_put(batch);
-
+ i915_vma_unpin_and_release(&batch, 0);
i915_vma_unpin(vma);
*rq_out = i915_request_get(rq);
@@ -764,8 +761,7 @@ emit_rpcs_query(struct drm_i915_gem_object *obj,
err_request:
i915_request_add(rq);
err_batch:
- i915_vma_unpin(batch);
- i915_vma_put(batch);
+ i915_vma_unpin_and_release(&batch, 0);
err_vma:
i915_vma_unpin(vma);
@@ -1309,9 +1305,7 @@ static int write_to_scratch(struct i915_gem_context *ctx,
if (err)
goto skip_request;
- i915_vma_unpin(vma);
- i915_vma_close(vma);
- i915_vma_put(vma);
+ i915_vma_unpin_and_release(&vma, 0);
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 aefe557527f8..36aca1c172e7 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
@@ -322,7 +322,6 @@ static int igt_partial_tiling(void *arg)
goto out;
}
- mutex_lock(&i915->drm.struct_mutex);
wakeref = intel_runtime_pm_get(&i915->runtime_pm);
if (1) {
@@ -415,7 +414,6 @@ next_tiling: ;
out_unlock:
intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
i915_gem_object_unpin_pages(obj);
out:
i915_gem_object_put(obj);
@@ -458,7 +456,6 @@ static int igt_smoke_tiling(void *arg)
goto out;
}
- mutex_lock(&i915->drm.struct_mutex);
wakeref = intel_runtime_pm_get(&i915->runtime_pm);
count = 0;
@@ -508,7 +505,6 @@ static int igt_smoke_tiling(void *arg)
pr_info("%s: Completed %lu trials\n", __func__, count);
intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
i915_gem_object_unpin_pages(obj);
out:
i915_gem_object_put(obj);
diff --git a/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c b/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c
index ee5dc13a30b3..6718da20f35d 100644
--- a/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c
+++ b/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c
@@ -154,9 +154,7 @@ int igt_gpu_fill_dw(struct intel_context *ce,
i915_request_add(rq);
- i915_vma_unpin(batch);
- i915_vma_close(batch);
- i915_vma_put(batch);
+ i915_vma_unpin_and_release(&batch, 0);
return 0;
@@ -165,7 +163,6 @@ int igt_gpu_fill_dw(struct intel_context *ce,
err_request:
i915_request_add(rq);
err_batch:
- i915_vma_unpin(batch);
- i915_vma_put(batch);
+ i915_vma_unpin_and_release(&batch, 0);
return err;
}
diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c b/drivers/gpu/drm/i915/gt/intel_gt.c
index eef9bdae8ebb..95e67db52668 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt.c
+++ b/drivers/gpu/drm/i915/gt/intel_gt.c
@@ -301,11 +301,12 @@ void intel_gt_flush_ggtt_writes(struct intel_gt *gt)
with_intel_runtime_pm(&i915->runtime_pm, wakeref) {
struct intel_uncore *uncore = gt->uncore;
+ unsigned long flags;
- spin_lock_irq(&uncore->lock);
+ spin_lock_irqsave(&uncore->lock, flags);
intel_uncore_posting_read_fw(uncore,
RING_HEAD(RENDER_RING_BASE));
- spin_unlock_irq(&uncore->lock);
+ spin_unlock_irqrestore(&uncore->lock, flags);
}
}
diff --git a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
index 0747b8c9f768..ec32996254c0 100644
--- a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
@@ -1338,15 +1338,13 @@ void intel_ring_free(struct kref *ref)
{
struct intel_ring *ring = container_of(ref, typeof(*ring), ref);
- i915_vma_close(ring->vma);
i915_vma_put(ring->vma);
-
kfree(ring);
}
static void __ring_context_fini(struct intel_context *ce)
{
- i915_gem_object_put(ce->state->obj);
+ i915_vma_put(ce->state);
}
static void ring_context_destroy(struct kref *ref)
diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
index a0098fc35921..e53eea1050f8 100644
--- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
+++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
@@ -1127,15 +1127,14 @@ static int evict_vma(void *data)
{
struct evict_vma *arg = data;
struct i915_address_space *vm = arg->vma->vm;
- struct drm_i915_private *i915 = vm->i915;
struct drm_mm_node evict = arg->vma->node;
int err;
complete(&arg->completion);
- mutex_lock(&i915->drm.struct_mutex);
+ mutex_lock(&vm->mutex);
err = i915_gem_evict_for_node(vm, &evict, 0);
- mutex_unlock(&i915->drm.struct_mutex);
+ mutex_unlock(&vm->mutex);
return err;
}
@@ -1143,39 +1142,33 @@ static int evict_vma(void *data)
static int evict_fence(void *data)
{
struct evict_vma *arg = data;
- struct drm_i915_private *i915 = arg->vma->vm->i915;
int err;
complete(&arg->completion);
- mutex_lock(&i915->drm.struct_mutex);
-
/* Mark the fence register as dirty to force the mmio update. */
err = i915_gem_object_set_tiling(arg->vma->obj, I915_TILING_Y, 512);
if (err) {
pr_err("Invalid Y-tiling settings; err:%d\n", err);
- goto out_unlock;
+ return err;
}
err = i915_vma_pin(arg->vma, 0, 0, PIN_GLOBAL | PIN_MAPPABLE);
if (err) {
pr_err("Unable to pin vma for Y-tiled fence; err:%d\n", err);
- goto out_unlock;
+ return err;
}
err = i915_vma_pin_fence(arg->vma);
i915_vma_unpin(arg->vma);
if (err) {
pr_err("Unable to pin Y-tiled fence; err:%d\n", err);
- goto out_unlock;
+ return err;
}
i915_vma_unpin_fence(arg->vma);
-out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
-
- return err;
+ return 0;
}
static int __igt_reset_evict_vma(struct intel_gt *gt,
diff --git a/drivers/gpu/drm/i915/gvt/aperture_gm.c b/drivers/gpu/drm/i915/gvt/aperture_gm.c
index 5ff2437b2998..d996bbc7ea59 100644
--- a/drivers/gpu/drm/i915/gvt/aperture_gm.c
+++ b/drivers/gpu/drm/i915/gvt/aperture_gm.c
@@ -61,14 +61,14 @@ static int alloc_gm(struct intel_vgpu *vgpu, bool high_gm)
flags = PIN_MAPPABLE;
}
- mutex_lock(&dev_priv->drm.struct_mutex);
+ mutex_lock(&dev_priv->ggtt.vm.mutex);
mmio_hw_access_pre(dev_priv);
ret = i915_gem_gtt_insert(&dev_priv->ggtt.vm, node,
size, I915_GTT_PAGE_SIZE,
I915_COLOR_UNEVICTABLE,
start, end, flags);
mmio_hw_access_post(dev_priv);
- mutex_unlock(&dev_priv->drm.struct_mutex);
+ mutex_unlock(&dev_priv->ggtt.vm.mutex);
if (ret)
gvt_err("fail to alloc %s gm space from host\n",
high_gm ? "high" : "low");
@@ -98,9 +98,9 @@ static int alloc_vgpu_gm(struct intel_vgpu *vgpu)
return 0;
out_free_aperture:
- mutex_lock(&dev_priv->drm.struct_mutex);
+ mutex_lock(&dev_priv->ggtt.vm.mutex);
drm_mm_remove_node(&vgpu->gm.low_gm_node);
- mutex_unlock(&dev_priv->drm.struct_mutex);
+ mutex_unlock(&dev_priv->ggtt.vm.mutex);
return ret;
}
@@ -108,10 +108,10 @@ static void free_vgpu_gm(struct intel_vgpu *vgpu)
{
struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
- mutex_lock(&dev_priv->drm.struct_mutex);
+ mutex_lock(&dev_priv->ggtt.vm.mutex);
drm_mm_remove_node(&vgpu->gm.low_gm_node);
drm_mm_remove_node(&vgpu->gm.high_gm_node);
- mutex_unlock(&dev_priv->drm.struct_mutex);
+ mutex_unlock(&dev_priv->ggtt.vm.mutex);
}
/**
diff --git a/drivers/gpu/drm/i915/i915_active.c b/drivers/gpu/drm/i915/i915_active.c
index d5aac6ff803a..0791736a08fd 100644
--- a/drivers/gpu/drm/i915/i915_active.c
+++ b/drivers/gpu/drm/i915/i915_active.c
@@ -146,6 +146,7 @@ __active_retire(struct i915_active *ref)
if (!retire)
return;
+ GEM_BUG_ON(rcu_access_pointer(ref->excl));
rbtree_postorder_for_each_entry_safe(it, n, &root, node) {
GEM_BUG_ON(i915_active_request_isset(&it->base));
kmem_cache_free(global.slab_cache, it);
@@ -245,6 +246,8 @@ void __i915_active_init(struct drm_i915_private *i915,
ref->flags = 0;
ref->active = active;
ref->retire = retire;
+
+ ref->excl = NULL;
ref->tree = RB_ROOT;
ref->cache = NULL;
init_llist_head(&ref->preallocated_barriers);
@@ -341,6 +344,46 @@ int i915_active_ref(struct i915_active *ref,
return err;
}
+static void excl_cb(struct dma_fence *f, struct dma_fence_cb *cb)
+{
+ struct i915_active *ref = container_of(cb, typeof(*ref), excl_cb);
+
+ RCU_INIT_POINTER(ref->excl, NULL);
+ dma_fence_put(f);
+
+ active_retire(ref);
+}
+
+void i915_active_set_exclusive(struct i915_active *ref, struct dma_fence *f)
+{
+ /* We expect the caller to manage the exclusive timeline ordering */
+ GEM_BUG_ON(i915_active_is_idle(ref));
+
+ dma_fence_get(f);
+
+ rcu_read_lock();
+ if (rcu_access_pointer(ref->excl)) {
+ struct dma_fence *old;
+
+ old = dma_fence_get_rcu_safe(&ref->excl);
+ if (old) {
+ if (dma_fence_remove_callback(old, &ref->excl_cb))
+ atomic_dec(&ref->count);
+ dma_fence_put(old);
+ }
+ }
+ rcu_read_unlock();
+
+ atomic_inc(&ref->count);
+ rcu_assign_pointer(ref->excl, f);
+
+ if (dma_fence_add_callback(f, &ref->excl_cb, excl_cb)) {
+ RCU_INIT_POINTER(ref->excl, NULL);
+ atomic_dec(&ref->count);
+ dma_fence_put(f);
+ }
+}
+
int i915_active_acquire(struct i915_active *ref)
{
int err;
@@ -399,6 +442,25 @@ void i915_active_ungrab(struct i915_active *ref)
__active_ungrab(ref);
}
+static int excl_wait(struct i915_active *ref)
+{
+ struct dma_fence *old;
+ int err = 0;
+
+ if (!rcu_access_pointer(ref->excl))
+ return 0;
+
+ rcu_read_lock();
+ old = dma_fence_get_rcu_safe(&ref->excl);
+ rcu_read_unlock();
+ if (old) {
+ err = dma_fence_wait(old, true);
+ dma_fence_put(old);
+ }
+
+ return err;
+}
+
int i915_active_wait(struct i915_active *ref)
{
struct active_node *it, *n;
@@ -419,6 +481,10 @@ int i915_active_wait(struct i915_active *ref)
return 0;
}
+ err = excl_wait(ref);
+ if (err)
+ goto out;
+
rbtree_postorder_for_each_entry_safe(it, n, &ref->tree, node) {
if (is_barrier(&it->base)) { /* unconnected idle-barrier */
err = -EBUSY;
@@ -430,6 +496,7 @@ int i915_active_wait(struct i915_active *ref)
break;
}
+out:
__active_retire(ref);
if (err)
return err;
@@ -454,26 +521,22 @@ int i915_request_await_active_request(struct i915_request *rq,
int i915_request_await_active(struct i915_request *rq, struct i915_active *ref)
{
- struct active_node *it, *n;
- int err;
-
- if (RB_EMPTY_ROOT(&ref->tree))
- return 0;
+ int err = 0;
- /* await allocates and so we need to avoid hitting the shrinker */
- err = i915_active_acquire(ref);
- if (err)
- return err;
+ if (rcu_access_pointer(ref->excl)) {
+ struct dma_fence *fence;
- mutex_lock(&ref->mutex);
- rbtree_postorder_for_each_entry_safe(it, n, &ref->tree, node) {
- err = i915_request_await_active_request(rq, &it->base);
- if (err)
- break;
+ rcu_read_lock();
+ fence = dma_fence_get_rcu_safe(&ref->excl);
+ rcu_read_unlock();
+ if (fence) {
+ err = i915_request_await_dma_fence(rq, fence);
+ dma_fence_put(fence);
+ }
}
- mutex_unlock(&ref->mutex);
- i915_active_release(ref);
+ /* In the future we may choose to await on all fences */
+
return err;
}
diff --git a/drivers/gpu/drm/i915/i915_active.h b/drivers/gpu/drm/i915/i915_active.h
index 949c6835335b..90034f61b7c2 100644
--- a/drivers/gpu/drm/i915/i915_active.h
+++ b/drivers/gpu/drm/i915/i915_active.h
@@ -379,6 +379,13 @@ i915_active_add_request(struct i915_active *ref, struct i915_request *rq)
return i915_active_ref(ref, i915_request_timeline(rq), rq);
}
+void i915_active_set_exclusive(struct i915_active *ref, struct dma_fence *f);
+
+static inline bool i915_active_has_exclusive(struct i915_active *ref)
+{
+ return rcu_access_pointer(ref->excl);
+}
+
int i915_active_wait(struct i915_active *ref);
int i915_request_await_active(struct i915_request *rq,
diff --git a/drivers/gpu/drm/i915/i915_active_types.h b/drivers/gpu/drm/i915/i915_active_types.h
index 1854e7d168c1..86e7a232ea3c 100644
--- a/drivers/gpu/drm/i915/i915_active_types.h
+++ b/drivers/gpu/drm/i915/i915_active_types.h
@@ -8,6 +8,7 @@
#define _I915_ACTIVE_TYPES_H_
#include <linux/atomic.h>
+#include <linux/dma-fence.h>
#include <linux/llist.h>
#include <linux/mutex.h>
#include <linux/rbtree.h>
@@ -51,6 +52,10 @@ struct i915_active {
struct mutex mutex;
atomic_t count;
+ /* Preallocated "exclusive" node */
+ struct dma_fence __rcu *excl;
+ struct dma_fence_cb excl_cb;
+
unsigned long flags;
#define I915_ACTIVE_GRAB_BIT 0
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index a9ee73b61f4d..ab00ba6f695b 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -1881,10 +1881,8 @@ static int i915_drm_resume(struct drm_device *dev)
if (ret)
DRM_ERROR("failed to re-enable GGTT\n");
- mutex_lock(&dev_priv->drm.struct_mutex);
i915_gem_restore_gtt_mappings(dev_priv);
i915_gem_restore_fences(dev_priv);
- mutex_unlock(&dev_priv->drm.struct_mutex);
intel_csr_ucode_resume(dev_priv);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index b8f7fe311b0a..934ea776a46a 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -62,20 +62,31 @@
#include "intel_pm.h"
static int
-insert_mappable_node(struct i915_ggtt *ggtt,
- struct drm_mm_node *node, u32 size)
+insert_mappable_node(struct i915_ggtt *ggtt, struct drm_mm_node *node, u32 size)
{
+ int err;
+
+ err = mutex_lock_interruptible(&ggtt->vm.mutex);
+ if (err)
+ return err;
+
memset(node, 0, sizeof(*node));
- return drm_mm_insert_node_in_range(&ggtt->vm.mm, node,
- size, 0, I915_COLOR_UNEVICTABLE,
- 0, ggtt->mappable_end,
- DRM_MM_INSERT_LOW);
+ err = drm_mm_insert_node_in_range(&ggtt->vm.mm, node,
+ size, 0, I915_COLOR_UNEVICTABLE,
+ 0, ggtt->mappable_end,
+ DRM_MM_INSERT_LOW);
+
+ mutex_unlock(&ggtt->vm.mutex);
+
+ return err;
}
static void
-remove_mappable_node(struct drm_mm_node *node)
+remove_mappable_node(struct i915_ggtt *ggtt, struct drm_mm_node *node)
{
+ mutex_lock(&ggtt->vm.mutex);
drm_mm_remove_node(node);
+ mutex_unlock(&ggtt->vm.mutex);
}
int
@@ -87,7 +98,8 @@ i915_gem_get_aperture_ioctl(struct drm_device *dev, void *data,
struct i915_vma *vma;
u64 pinned;
- mutex_lock(&ggtt->vm.mutex);
+ if (mutex_lock_interruptible(&ggtt->vm.mutex))
+ return -EINTR;
pinned = ggtt->vm.reserved;
list_for_each_entry(vma, &ggtt->vm.bound_list, vm_link)
@@ -109,20 +121,24 @@ int i915_gem_object_unbind(struct drm_i915_gem_object *obj,
LIST_HEAD(still_in_list);
int ret = 0;
- lockdep_assert_held(&obj->base.dev->struct_mutex);
-
spin_lock(&obj->vma.lock);
while (!ret && (vma = list_first_entry_or_null(&obj->vma.list,
struct i915_vma,
obj_link))) {
+ struct i915_address_space *vm = vma->vm;
+
+ ret = -EBUSY;
+ if (!i915_vm_tryopen(vm))
+ break;
+
list_move_tail(&vma->obj_link, &still_in_list);
spin_unlock(&obj->vma.lock);
- ret = -EBUSY;
if (flags & I915_GEM_OBJECT_UNBIND_ACTIVE ||
!i915_vma_is_active(vma))
ret = i915_vma_unbind(vma);
+ i915_vm_close(vm);
spin_lock(&obj->vma.lock);
}
list_splice(&still_in_list, &obj->vma.list);
@@ -338,10 +354,6 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
u64 remain, offset;
int ret;
- ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
- if (ret)
- return ret;
-
wakeref = intel_runtime_pm_get(&i915->runtime_pm);
vma = ERR_PTR(-ENODEV);
if (!i915_gem_object_is_tiled(obj))
@@ -355,12 +367,10 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
} else {
ret = insert_mappable_node(ggtt, &node, PAGE_SIZE);
if (ret)
- goto out_unlock;
+ goto out_rpm;
GEM_BUG_ON(!drm_mm_node_allocated(&node));
}
- mutex_unlock(&i915->drm.struct_mutex);
-
ret = i915_gem_object_lock_interruptible(obj);
if (ret)
goto out_unpin;
@@ -414,17 +424,14 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
i915_gem_object_unlock_fence(obj, fence);
out_unpin:
- mutex_lock(&i915->drm.struct_mutex);
if (drm_mm_node_allocated(&node)) {
ggtt->vm.clear_range(&ggtt->vm, node.start, node.size);
- remove_mappable_node(&node);
+ remove_mappable_node(ggtt, &node);
} else {
i915_vma_unpin(vma);
}
-out_unlock:
+out_rpm:
intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
-
return ret;
}
@@ -531,10 +538,6 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
void __user *user_data;
int ret;
- ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
- if (ret)
- return ret;
-
if (i915_gem_object_has_struct_page(obj)) {
/*
* Avoid waking the device up if we can fallback, as
@@ -544,10 +547,8 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
* using the cache bypass of indirect GGTT access.
*/
wakeref = intel_runtime_pm_get_if_in_use(rpm);
- if (!wakeref) {
- ret = -EFAULT;
- goto out_unlock;
- }
+ if (!wakeref)
+ return -EFAULT;
} else {
/* No backing pages, no fallback, we must force GGTT access */
wakeref = intel_runtime_pm_get(rpm);
@@ -569,8 +570,6 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
GEM_BUG_ON(!drm_mm_node_allocated(&node));
}
- mutex_unlock(&i915->drm.struct_mutex);
-
ret = i915_gem_object_lock_interruptible(obj);
if (ret)
goto out_unpin;
@@ -634,18 +633,15 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
i915_gem_object_unlock_fence(obj, fence);
out_unpin:
- mutex_lock(&i915->drm.struct_mutex);
intel_gt_flush_ggtt_writes(ggtt->vm.gt);
if (drm_mm_node_allocated(&node)) {
ggtt->vm.clear_range(&ggtt->vm, node.start, node.size);
- remove_mappable_node(&node);
+ remove_mappable_node(ggtt, &node);
} else {
i915_vma_unpin(vma);
}
out_rpm:
intel_runtime_pm_put(rpm, wakeref);
-out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
return ret;
}
@@ -968,8 +964,6 @@ i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj,
struct i915_vma *vma;
int ret;
- lockdep_assert_held(&obj->base.dev->struct_mutex);
-
if (flags & PIN_MAPPABLE &&
(!view || view->type == I915_GGTT_VIEW_NORMAL)) {
/* If the required space is larger than the available
@@ -1016,13 +1010,6 @@ i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj,
return ERR_PTR(-ENOSPC);
}
- WARN(i915_vma_is_pinned(vma),
- "bo is already pinned in ggtt with incorrect alignment:"
- " offset=%08x, req.alignment=%llx,"
- " req.map_and_fenceable=%d, vma->map_and_fenceable=%d\n",
- i915_ggtt_offset(vma), alignment,
- !!(flags & PIN_MAPPABLE),
- i915_vma_is_map_and_fenceable(vma));
ret = i915_vma_unbind(vma);
if (ret)
return ERR_PTR(ret);
@@ -1441,8 +1428,6 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
}
if (ret == -EIO) {
- mutex_lock(&dev_priv->drm.struct_mutex);
-
/*
* Allow engines or uC initialisation to fail by marking the GPU
* as wedged. But we only want to do this when the GPU is angry,
@@ -1459,8 +1444,6 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
i915_gem_restore_gtt_mappings(dev_priv);
i915_gem_restore_fences(dev_priv);
intel_init_clock_gating(dev_priv);
-
- mutex_unlock(&dev_priv->drm.struct_mutex);
}
i915_gem_drain_freed_objects(dev_priv);
diff --git a/drivers/gpu/drm/i915/i915_gem_evict.c b/drivers/gpu/drm/i915/i915_gem_evict.c
index 8c1e04f402bc..0552bf93eea3 100644
--- a/drivers/gpu/drm/i915/i915_gem_evict.c
+++ b/drivers/gpu/drm/i915/i915_gem_evict.c
@@ -47,8 +47,7 @@ static int ggtt_flush(struct drm_i915_private *i915)
* bound by their active reference.
*/
return i915_gem_wait_for_idle(i915,
- I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED,
+ I915_WAIT_INTERRUPTIBLE,
MAX_SCHEDULE_TIMEOUT);
}
@@ -104,7 +103,7 @@ i915_gem_evict_something(struct i915_address_space *vm,
struct i915_vma *active;
int ret;
- lockdep_assert_held(&vm->i915->drm.struct_mutex);
+ lockdep_assert_held(&vm->mutex);
trace_i915_gem_evict(vm, min_size, alignment, flags);
/*
@@ -127,15 +126,6 @@ i915_gem_evict_something(struct i915_address_space *vm,
min_size, alignment, color,
start, end, mode);
- /*
- * Retire before we search the active list. Although we have
- * reasonable accuracy in our retirement lists, we may have
- * a stray pin (preventing eviction) that can only be resolved by
- * retiring.
- */
- if (!(flags & PIN_NONBLOCK))
- i915_retire_requests(dev_priv);
-
search_again:
active = NULL;
INIT_LIST_HEAD(&eviction_list);
@@ -235,12 +225,12 @@ i915_gem_evict_something(struct i915_address_space *vm,
list_for_each_entry_safe(vma, next, &eviction_list, evict_link) {
__i915_vma_unpin(vma);
if (ret == 0)
- ret = i915_vma_unbind(vma);
+ ret = __i915_vma_unbind(vma);
}
while (ret == 0 && (node = drm_mm_scan_color_evict(&scan))) {
vma = container_of(node, struct i915_vma, node);
- ret = i915_vma_unbind(vma);
+ ret = __i915_vma_unbind(vma);
}
return ret;
@@ -268,7 +258,7 @@ int i915_gem_evict_for_node(struct i915_address_space *vm,
struct i915_vma *vma, *next;
int ret = 0;
- lockdep_assert_held(&vm->i915->drm.struct_mutex);
+ lockdep_assert_held(&vm->mutex);
GEM_BUG_ON(!IS_ALIGNED(start, I915_GTT_PAGE_SIZE));
GEM_BUG_ON(!IS_ALIGNED(end, I915_GTT_PAGE_SIZE));
@@ -349,7 +339,7 @@ int i915_gem_evict_for_node(struct i915_address_space *vm,
list_for_each_entry_safe(vma, next, &eviction_list, evict_link) {
__i915_vma_unpin(vma);
if (ret == 0)
- ret = i915_vma_unbind(vma);
+ ret = __i915_vma_unbind(vma);
}
return ret;
@@ -373,7 +363,7 @@ int i915_gem_evict_vm(struct i915_address_space *vm)
struct i915_vma *vma, *next;
int ret;
- lockdep_assert_held(&vm->i915->drm.struct_mutex);
+ lockdep_assert_held(&vm->mutex);
trace_i915_gem_evict_vm(vm);
/* Switch back to the default context in order to unpin
@@ -388,7 +378,6 @@ int i915_gem_evict_vm(struct i915_address_space *vm)
}
INIT_LIST_HEAD(&eviction_list);
- mutex_lock(&vm->mutex);
list_for_each_entry(vma, &vm->bound_list, vm_link) {
if (i915_vma_is_pinned(vma))
continue;
@@ -396,13 +385,12 @@ int i915_gem_evict_vm(struct i915_address_space *vm)
__i915_vma_pin(vma);
list_add(&vma->evict_link, &eviction_list);
}
- mutex_unlock(&vm->mutex);
ret = 0;
list_for_each_entry_safe(vma, next, &eviction_list, evict_link) {
__i915_vma_unpin(vma);
if (ret == 0)
- ret = i915_vma_unbind(vma);
+ ret = __i915_vma_unbind(vma);
}
return ret;
}
diff --git a/drivers/gpu/drm/i915/i915_gem_fence_reg.c b/drivers/gpu/drm/i915/i915_gem_fence_reg.c
index 615a9f4ef30c..487b7261f7ed 100644
--- a/drivers/gpu/drm/i915/i915_gem_fence_reg.c
+++ b/drivers/gpu/drm/i915/i915_gem_fence_reg.c
@@ -230,14 +230,15 @@ static int fence_update(struct i915_fence_reg *fence,
i915_gem_object_get_tiling(vma->obj)))
return -EINVAL;
- ret = i915_active_wait(&vma->active);
+ ret = i915_vma_sync(vma);
if (ret)
return ret;
}
old = xchg(&fence->vma, NULL);
if (old) {
- ret = i915_active_wait(&old->active);
+ /* XXX Ideally we would move the waiting to outside the mutex */
+ ret = i915_vma_sync(old);
if (ret) {
fence->vma = old;
return ret;
@@ -331,13 +332,15 @@ static struct i915_fence_reg *fence_find(struct drm_i915_private *i915)
return ERR_PTR(-EDEADLK);
}
-static int __i915_vma_pin_fence(struct i915_vma *vma)
+int __i915_vma_pin_fence(struct i915_vma *vma)
{
struct i915_ggtt *ggtt = i915_vm_to_ggtt(vma->vm);
struct i915_fence_reg *fence;
struct i915_vma *set = i915_gem_object_is_tiled(vma->obj) ? vma : NULL;
int err;
+ lockdep_assert_held(&vma->vm->mutex);
+
/* Just update our place in the LRU if our fence is getting reused. */
if (vma->fence) {
fence = vma->fence;
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 483eeff8730e..a351b98afe03 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -150,16 +150,18 @@ static void gmch_ggtt_invalidate(struct i915_ggtt *ggtt)
static int ppgtt_bind_vma(struct i915_vma *vma,
enum i915_cache_level cache_level,
- u32 unused)
+ u32 flags)
{
u32 pte_flags;
int err;
- if (!i915_vma_is_bound(vma, I915_VMA_LOCAL_BIND)) {
+ if (flags & I915_VMA_ALLOC) {
err = vma->vm->allocate_va_range(vma->vm,
vma->node.start, vma->size);
if (err)
return err;
+
+ set_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma));
}
/* Applicable to VLV, and gen8+ */
@@ -167,6 +169,7 @@ static int ppgtt_bind_vma(struct i915_vma *vma,
if (i915_gem_object_is_readonly(vma->obj))
pte_flags |= PTE_READ_ONLY;
+ GEM_BUG_ON(!test_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma)));
vma->vm->insert_entries(vma->vm, vma, cache_level, pte_flags);
wmb();
@@ -175,7 +178,8 @@ static int ppgtt_bind_vma(struct i915_vma *vma,
static void ppgtt_unbind_vma(struct i915_vma *vma)
{
- vma->vm->clear_range(vma->vm, vma->node.start, vma->size);
+ if (test_and_clear_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma)))
+ vma->vm->clear_range(vma->vm, vma->node.start, vma->size);
}
static int ppgtt_set_pages(struct i915_vma *vma)
@@ -503,15 +507,26 @@ static void i915_address_space_fini(struct i915_address_space *vm)
mutex_destroy(&vm->mutex);
}
-static void ppgtt_destroy_vma(struct i915_address_space *vm)
+void __i915_vm_close(struct i915_address_space *vm)
{
struct i915_vma *vma, *vn;
- mutex_lock(&vm->i915->drm.struct_mutex);
- list_for_each_entry_safe(vma, vn, &vm->bound_list, vm_link)
+ mutex_lock(&vm->mutex);
+ list_for_each_entry_safe(vma, vn, &vm->bound_list, vm_link) {
+ struct drm_i915_gem_object *obj = vma->obj;
+
+ /* Keep the obj (and hence the vma) alive as _we_ destroy it */
+ if (!kref_get_unless_zero(&obj->base.refcount))
+ continue;
+
+ atomic_and(~I915_VMA_PIN_MASK, &vma->flags);
+ WARN_ON(__i915_vma_unbind(vma));
i915_vma_destroy(vma);
+
+ i915_gem_object_put(obj);
+ }
GEM_BUG_ON(!list_empty(&vm->bound_list));
- mutex_unlock(&vm->i915->drm.struct_mutex);
+ mutex_unlock(&vm->mutex);
}
static void __i915_vm_release(struct work_struct *work)
@@ -519,8 +534,6 @@ static void __i915_vm_release(struct work_struct *work)
struct i915_address_space *vm =
container_of(work, struct i915_address_space, rcu.work);
- ppgtt_destroy_vma(vm);
-
vm->cleanup(vm);
i915_address_space_fini(vm);
@@ -535,7 +548,6 @@ void i915_vm_release(struct kref *kref)
GEM_BUG_ON(i915_is_ggtt(vm));
trace_i915_ppgtt_release(vm);
- vm->closed = true;
queue_rcu_work(vm->i915->wq, &vm->rcu);
}
@@ -543,6 +555,7 @@ static void i915_address_space_init(struct i915_address_space *vm, int subclass)
{
kref_init(&vm->ref);
INIT_RCU_WORK(&vm->rcu, __i915_vm_release);
+ atomic_set(&vm->open, 1);
/*
* The vm->mutex must be reclaim safe (for use in the shrinker).
@@ -1771,12 +1784,8 @@ static void gen6_ppgtt_free_pd(struct gen6_ppgtt *ppgtt)
static void gen6_ppgtt_cleanup(struct i915_address_space *vm)
{
struct gen6_ppgtt *ppgtt = to_gen6_ppgtt(i915_vm_to_ppgtt(vm));
- struct drm_i915_private *i915 = vm->i915;
- /* FIXME remove the struct_mutex to bring the locking under control */
- mutex_lock(&i915->drm.struct_mutex);
i915_vma_destroy(ppgtt->vma);
- mutex_unlock(&i915->drm.struct_mutex);
gen6_ppgtt_free_pd(ppgtt);
free_scratch(vm);
@@ -1865,7 +1874,8 @@ static struct i915_vma *pd_vma_create(struct gen6_ppgtt *ppgtt, int size)
i915_active_init(i915, &vma->active, NULL, NULL);
- vma->vm = &ggtt->vm;
+ mutex_init(&vma->pages_mutex);
+ vma->vm = i915_vm_get(&ggtt->vm);
vma->ops = &pd_vma_ops;
vma->private = ppgtt;
@@ -1885,7 +1895,7 @@ int gen6_ppgtt_pin(struct i915_ppgtt *base)
struct gen6_ppgtt *ppgtt = to_gen6_ppgtt(base);
int err = 0;
- GEM_BUG_ON(ppgtt->base.vm.closed);
+ GEM_BUG_ON(!atomic_read(&ppgtt->base.vm.open));
/*
* Workaround the limited maximum vma->pin_count and the aliasing_ppgtt
@@ -2463,14 +2473,18 @@ static int aliasing_gtt_bind_vma(struct i915_vma *vma,
if (flags & I915_VMA_LOCAL_BIND) {
struct i915_ppgtt *alias = i915_vm_to_ggtt(vma->vm)->alias;
- if (!i915_vma_is_bound(vma, I915_VMA_LOCAL_BIND)) {
+ if (flags & I915_VMA_ALLOC) {
ret = alias->vm.allocate_va_range(&alias->vm,
vma->node.start,
vma->size);
if (ret)
return ret;
+
+ set_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma));
}
+ GEM_BUG_ON(!test_bit(I915_VMA_ALLOC_BIT,
+ __i915_vma_flags(vma)));
alias->vm.insert_entries(&alias->vm, vma,
cache_level, pte_flags);
}
@@ -2499,7 +2513,7 @@ static void aliasing_gtt_unbind_vma(struct i915_vma *vma)
vm->clear_range(vm, vma->node.start, vma->size);
}
- if (i915_vma_is_bound(vma, I915_VMA_LOCAL_BIND)) {
+ if (test_and_clear_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma))) {
struct i915_address_space *vm =
&i915_vm_to_ggtt(vma->vm)->alias->vm;
@@ -2602,22 +2616,16 @@ static int init_aliasing_ppgtt(struct i915_ggtt *ggtt)
static void fini_aliasing_ppgtt(struct i915_ggtt *ggtt)
{
- struct drm_i915_private *i915 = ggtt->vm.i915;
struct i915_ppgtt *ppgtt;
- mutex_lock(&i915->drm.struct_mutex);
-
ppgtt = fetch_and_zero(&ggtt->alias);
if (!ppgtt)
- goto out;
+ return;
i915_vm_put(&ppgtt->vm);
ggtt->vm.vma_ops.bind_vma = ggtt_bind_vma;
ggtt->vm.vma_ops.unbind_vma = ggtt_unbind_vma;
-
-out:
- mutex_unlock(&i915->drm.struct_mutex);
}
static int ggtt_reserve_guc_top(struct i915_ggtt *ggtt)
@@ -2734,32 +2742,28 @@ int i915_init_ggtt(struct drm_i915_private *i915)
static void ggtt_cleanup_hw(struct i915_ggtt *ggtt)
{
- struct drm_i915_private *i915 = ggtt->vm.i915;
struct i915_vma *vma, *vn;
- ggtt->vm.closed = true;
+ atomic_set(&ggtt->vm.open, 0);
rcu_barrier(); /* flush the RCU'ed__i915_vm_release */
- flush_workqueue(i915->wq);
+ flush_workqueue(ggtt->vm.i915->wq);
- mutex_lock(&i915->drm.struct_mutex);
+ mutex_lock(&ggtt->vm.mutex);
list_for_each_entry_safe(vma, vn, &ggtt->vm.bound_list, vm_link)
- WARN_ON(i915_vma_unbind(vma));
+ WARN_ON(__i915_vma_unbind(vma));
if (drm_mm_node_allocated(&ggtt->error_capture))
drm_mm_remove_node(&ggtt->error_capture);
ggtt_release_guc_top(ggtt);
-
- if (drm_mm_initialized(&ggtt->vm.mm)) {
- intel_vgt_deballoon(ggtt);
- i915_address_space_fini(&ggtt->vm);
- }
+ intel_vgt_deballoon(ggtt);
ggtt->vm.cleanup(&ggtt->vm);
- mutex_unlock(&i915->drm.struct_mutex);
+ mutex_unlock(&ggtt->vm.mutex);
+ i915_address_space_fini(&ggtt->vm);
arch_phys_wc_del(ggtt->mtrr);
io_mapping_fini(&ggtt->iomap);
@@ -3188,9 +3192,6 @@ int i915_ggtt_probe_hw(struct drm_i915_private *i915)
static int ggtt_init_hw(struct i915_ggtt *ggtt)
{
struct drm_i915_private *i915 = ggtt->vm.i915;
- int ret = 0;
-
- mutex_lock(&i915->drm.struct_mutex);
i915_address_space_init(&ggtt->vm, VM_CLASS_GGTT);
@@ -3206,18 +3207,14 @@ static int ggtt_init_hw(struct i915_ggtt *ggtt)
ggtt->gmadr.start,
ggtt->mappable_end)) {
ggtt->vm.cleanup(&ggtt->vm);
- ret = -EIO;
- goto out;
+ return -EIO;
}
ggtt->mtrr = arch_phys_wc_add(ggtt->gmadr.start, ggtt->mappable_end);
i915_ggtt_init_fences(ggtt);
-out:
- mutex_unlock(&i915->drm.struct_mutex);
-
- return ret;
+ return 0;
}
/**
@@ -3289,6 +3286,7 @@ static void ggtt_restore_mappings(struct i915_ggtt *ggtt)
{
struct i915_vma *vma, *vn;
bool flush = false;
+ int open;
intel_gt_check_and_clear_faults(ggtt->vm.gt);
@@ -3296,7 +3294,9 @@ static void ggtt_restore_mappings(struct i915_ggtt *ggtt)
/* First fill our portion of the GTT with scratch pages */
ggtt->vm.clear_range(&ggtt->vm, 0, ggtt->vm.total);
- ggtt->vm.closed = true; /* skip rewriting PTE on VMA unbind */
+
+ /* Skip rewriting PTE on VMA unbind. */
+ open = atomic_xchg(&ggtt->vm.open, 0);
/* clflush objects bound into the GGTT and rebind them. */
list_for_each_entry_safe(vma, vn, &ggtt->vm.bound_list, vm_link) {
@@ -3305,24 +3305,20 @@ static void ggtt_restore_mappings(struct i915_ggtt *ggtt)
if (!i915_vma_is_bound(vma, I915_VMA_GLOBAL_BIND))
continue;
- mutex_unlock(&ggtt->vm.mutex);
-
- if (!i915_vma_unbind(vma))
- goto lock;
+ if (!__i915_vma_unbind(vma))
+ continue;
+ clear_bit(I915_VMA_GLOBAL_BIND_BIT, __i915_vma_flags(vma));
WARN_ON(i915_vma_bind(vma,
obj ? obj->cache_level : 0,
- PIN_UPDATE));
+ PIN_GLOBAL, NULL));
if (obj) { /* only used during resume => exclusive access */
flush |= fetch_and_zero(&obj->write_domain);
obj->read_domains |= I915_GEM_DOMAIN_GTT;
}
-
-lock:
- mutex_lock(&ggtt->vm.mutex);
}
- ggtt->vm.closed = false;
+ atomic_set(&ggtt->vm.open, open);
ggtt->invalidate(ggtt);
mutex_unlock(&ggtt->vm.mutex);
@@ -3714,7 +3710,8 @@ int i915_gem_gtt_insert(struct i915_address_space *vm,
u64 offset;
int err;
- lockdep_assert_held(&vm->i915->drm.struct_mutex);
+ lockdep_assert_held(&vm->mutex);
+
GEM_BUG_ON(!size);
GEM_BUG_ON(!IS_ALIGNED(size, I915_GTT_PAGE_SIZE));
GEM_BUG_ON(alignment && !is_power_of_2(alignment));
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index d8dfa3811e54..03418dede503 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -307,7 +307,14 @@ struct i915_address_space {
unsigned int bind_alloc;
- bool closed;
+ /*
+ * Each active user context has its own address space (in full-ppgtt).
+ * Since the vm may be shared between multiple contexts, we count how
+ * many contexts keep us "open". Once open hits zero, we are closed
+ * and do not allow any new attachments, and proceed to shutdown our
+ * vma and page directories.
+ */
+ atomic_t open;
struct mutex mutex; /* protects vma and our lists */
#define VM_CLASS_GGTT 0
@@ -581,6 +588,35 @@ static inline void i915_vm_put(struct i915_address_space *vm)
kref_put(&vm->ref, i915_vm_release);
}
+static inline struct i915_address_space *
+i915_vm_open(struct i915_address_space *vm)
+{
+ GEM_BUG_ON(!atomic_read(&vm->open));
+ atomic_inc(&vm->open);
+ return i915_vm_get(vm);
+}
+
+static inline bool
+i915_vm_tryopen(struct i915_address_space *vm)
+{
+ if (atomic_add_unless(&vm->open, 1, 0))
+ return i915_vm_get(vm);
+
+ return false;
+}
+
+void __i915_vm_close(struct i915_address_space *vm);
+
+static inline void
+i915_vm_close(struct i915_address_space *vm)
+{
+ GEM_BUG_ON(!atomic_read(&vm->open));
+ if (atomic_dec_and_test(&vm->open))
+ __i915_vm_close(vm);
+
+ i915_vm_put(vm);
+}
+
int gen6_ppgtt_pin(struct i915_ppgtt *base);
void gen6_ppgtt_unpin(struct i915_ppgtt *base);
void gen6_ppgtt_unpin_all(struct i915_ppgtt *base);
@@ -613,10 +649,9 @@ int i915_gem_gtt_insert(struct i915_address_space *vm,
#define PIN_OFFSET_BIAS BIT_ULL(6)
#define PIN_OFFSET_FIXED BIT_ULL(7)
-#define PIN_MBZ BIT_ULL(8) /* I915_VMA_PIN_OVERFLOW */
-#define PIN_GLOBAL BIT_ULL(9) /* I915_VMA_GLOBAL_BIND */
-#define PIN_USER BIT_ULL(10) /* I915_VMA_LOCAL_BIND */
-#define PIN_UPDATE BIT_ULL(11)
+#define PIN_UPDATE BIT_ULL(9)
+#define PIN_GLOBAL BIT_ULL(10) /* I915_VMA_GLOBAL_BIND */
+#define PIN_USER BIT_ULL(11) /* I915_VMA_LOCAL_BIND */
#define PIN_OFFSET_MASK (-I915_GTT_PAGE_SIZE)
diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index 524f6710b7aa..80055501eccb 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -1204,15 +1204,10 @@ static int i915_oa_read(struct i915_perf_stream *stream,
static struct intel_context *oa_pin_context(struct i915_perf_stream *stream)
{
struct i915_gem_engines_iter it;
- struct drm_i915_private *i915 = stream->dev_priv;
struct i915_gem_context *ctx = stream->ctx;
struct intel_context *ce;
int err;
- err = i915_mutex_lock_interruptible(&i915->drm);
- if (err)
- return ERR_PTR(err);
-
for_each_gem_engine(ce, i915_gem_context_lock_engines(ctx), it) {
if (ce->engine->class != RENDER_CLASS)
continue;
@@ -1229,10 +1224,6 @@ static struct intel_context *oa_pin_context(struct i915_perf_stream *stream)
}
i915_gem_context_unlock_engines(ctx);
- mutex_unlock(&i915->drm.struct_mutex);
- if (err)
- return ERR_PTR(err);
-
return stream->pinned_ctx;
}
@@ -1331,32 +1322,22 @@ static int oa_get_render_ctx_id(struct i915_perf_stream *stream)
*/
static void oa_put_render_ctx_id(struct i915_perf_stream *stream)
{
- struct drm_i915_private *dev_priv = stream->dev_priv;
struct intel_context *ce;
stream->specific_ctx_id = INVALID_CTX_ID;
stream->specific_ctx_id_mask = 0;
ce = fetch_and_zero(&stream->pinned_ctx);
- if (ce) {
- mutex_lock(&dev_priv->drm.struct_mutex);
+ if (ce)
intel_context_unpin(ce);
- mutex_unlock(&dev_priv->drm.struct_mutex);
- }
}
static void
free_oa_buffer(struct i915_perf_stream *stream)
{
- struct drm_i915_private *i915 = stream->dev_priv;
-
- mutex_lock(&i915->drm.struct_mutex);
-
i915_vma_unpin_and_release(&stream->oa_buffer.vma,
I915_VMA_RELEASE_MAP);
- mutex_unlock(&i915->drm.struct_mutex);
-
stream->oa_buffer.vaddr = NULL;
}
@@ -1511,18 +1492,13 @@ static int alloc_oa_buffer(struct i915_perf_stream *stream)
if (WARN_ON(stream->oa_buffer.vma))
return -ENODEV;
- ret = i915_mutex_lock_interruptible(&dev_priv->drm);
- if (ret)
- return ret;
-
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_shmem(dev_priv, OA_BUFFER_SIZE);
if (IS_ERR(bo)) {
DRM_ERROR("Failed to allocate OA buffer\n");
- ret = PTR_ERR(bo);
- goto unlock;
+ return PTR_ERR(bo);
}
i915_gem_object_set_cache_coherency(bo, I915_CACHE_LLC);
@@ -1546,7 +1522,7 @@ static int alloc_oa_buffer(struct i915_perf_stream *stream)
i915_ggtt_offset(stream->oa_buffer.vma),
stream->oa_buffer.vaddr);
- goto unlock;
+ return 0;
err_unpin:
__i915_vma_unpin(vma);
@@ -1557,8 +1533,6 @@ static int alloc_oa_buffer(struct i915_perf_stream *stream)
stream->oa_buffer.vaddr = NULL;
stream->oa_buffer.vma = NULL;
-unlock:
- mutex_unlock(&dev_priv->drm.struct_mutex);
return ret;
}
diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
index d097f77890ba..defeb9bb1daa 100644
--- a/drivers/gpu/drm/i915/i915_vma.c
+++ b/drivers/gpu/drm/i915/i915_vma.c
@@ -32,6 +32,7 @@
#include "i915_drv.h"
#include "i915_globals.h"
+#include "i915_sw_fence_work.h"
#include "i915_trace.h"
#include "i915_vma.h"
@@ -110,7 +111,8 @@ vma_create(struct drm_i915_gem_object *obj,
if (vma == NULL)
return ERR_PTR(-ENOMEM);
- vma->vm = vm;
+ mutex_init(&vma->pages_mutex);
+ vma->vm = i915_vm_get(vm);
vma->ops = &vm->vma_ops;
vma->obj = obj;
vma->resv = obj->base.resv;
@@ -261,8 +263,6 @@ vma_lookup(struct drm_i915_gem_object *obj,
* Once created, the VMA is kept until either the object is freed, or the
* address space is closed.
*
- * Must be called with struct_mutex held.
- *
* Returns the vma, or an error pointer.
*/
struct i915_vma *
@@ -273,7 +273,7 @@ i915_vma_instance(struct drm_i915_gem_object *obj,
struct i915_vma *vma;
GEM_BUG_ON(view && !i915_is_ggtt(vm));
- GEM_BUG_ON(vm->closed);
+ GEM_BUG_ON(!atomic_read(&vm->open));
spin_lock(&obj->vma.lock);
vma = vma_lookup(obj, vm, view);
@@ -287,18 +287,63 @@ i915_vma_instance(struct drm_i915_gem_object *obj,
return vma;
}
+struct i915_vma_work {
+ struct dma_fence_work base;
+ struct i915_vma *vma;
+ enum i915_cache_level cache_level;
+ unsigned int flags;
+};
+
+static int __vma_bind(struct dma_fence_work *work)
+{
+ struct i915_vma_work *vw = container_of(work, typeof(*vw), base);
+ struct i915_vma *vma = vw->vma;
+ int err;
+
+ err = vma->ops->bind_vma(vma, vw->cache_level, vw->flags);
+ if (err)
+ atomic_or(I915_VMA_ERROR, &vma->flags);
+
+ if (vma->obj)
+ __i915_gem_object_unpin_pages(vma->obj);
+
+ return err;
+}
+
+static const struct dma_fence_work_ops bind_ops = {
+ .name = "bind",
+ .work = __vma_bind,
+};
+
+struct i915_vma_work *i915_vma_work(void)
+{
+ struct i915_vma_work *vw;
+
+ vw = kzalloc(sizeof(*vw), GFP_KERNEL);
+ if (!vw)
+ return NULL;
+
+ dma_fence_work_init(&vw->base, &bind_ops);
+ vw->base.dma.error = -EAGAIN; /* disable the worker by default */
+
+ return vw;
+}
+
/**
* i915_vma_bind - Sets up PTEs for an VMA in it's corresponding address space.
* @vma: VMA to map
* @cache_level: mapping cache level
* @flags: flags like global or local mapping
+ * @work: preallocated worker for allocating and binding the PTE
*
* DMA addresses are taken from the scatter-gather table of this object (or of
* this VMA in case of non-default GGTT views) and PTE entries set up.
* Note that DMA addresses are also the only part of the SG table we care about.
*/
-int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level,
- u32 flags)
+int i915_vma_bind(struct i915_vma *vma,
+ enum i915_cache_level cache_level,
+ u32 flags,
+ struct i915_vma_work *work)
{
u32 bind_flags;
u32 vma_flags;
@@ -315,11 +360,8 @@ int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level,
if (GEM_DEBUG_WARN_ON(!flags))
return -EINVAL;
- bind_flags = 0;
- if (flags & PIN_GLOBAL)
- bind_flags |= I915_VMA_GLOBAL_BIND;
- if (flags & PIN_USER)
- bind_flags |= I915_VMA_LOCAL_BIND;
+ bind_flags = flags;
+ bind_flags &= I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND;
vma_flags = atomic_read(&vma->flags);
vma_flags &= I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND;
@@ -333,9 +375,32 @@ int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level,
GEM_BUG_ON(!vma->pages);
trace_i915_vma_bind(vma, bind_flags);
- ret = vma->ops->bind_vma(vma, cache_level, bind_flags);
- if (ret)
- return ret;
+ if (work && (bind_flags & ~vma_flags) & vma->vm->bind_alloc) {
+ work->vma = vma;
+ work->cache_level = cache_level;
+ work->flags = bind_flags | I915_VMA_ALLOC;
+
+ /*
+ * Note we only want to chain up to the migration fence on
+ * the pages (not the object itself). As we don't track that,
+ * yet, we have to use the exclusive fence instead.
+ *
+ * Also note that we do not want to track the async vma as
+ * part of the obj->resv->excl_fence as it only affects
+ * execution and not content or object's backing store lifetime.
+ */
+ GEM_BUG_ON(i915_active_has_exclusive(&vma->active));
+ i915_active_set_exclusive(&vma->active, &work->base.dma);
+ work->base.dma.error = 0; /* enable the queue_work() */
+
+ if (vma->obj)
+ __i915_gem_object_pin_pages(vma->obj);
+ } else {
+ GEM_BUG_ON((bind_flags & ~vma_flags) & vma->vm->bind_alloc);
+ ret = vma->ops->bind_vma(vma, cache_level, bind_flags);
+ if (ret)
+ return ret;
+ }
atomic_or(bind_flags, &vma->flags);
return 0;
@@ -348,9 +413,7 @@ void __iomem *i915_vma_pin_iomap(struct i915_vma *vma)
/* Access through the GTT requires the device to be awake. */
assert_rpm_wakelock_held(&vma->vm->i915->runtime_pm);
-
- lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
- if (WARN_ON(!i915_vma_is_map_and_fenceable(vma))) {
+ if (GEM_WARN_ON(!i915_vma_is_map_and_fenceable(vma))) {
err = -ENODEV;
goto err;
}
@@ -358,7 +421,7 @@ void __iomem *i915_vma_pin_iomap(struct i915_vma *vma)
GEM_BUG_ON(!i915_vma_is_ggtt(vma));
GEM_BUG_ON(!i915_vma_is_bound(vma, I915_VMA_GLOBAL_BIND));
- ptr = vma->iomap;
+ ptr = READ_ONCE(vma->iomap);
if (ptr == NULL) {
ptr = io_mapping_map_wc(&i915_vm_to_ggtt(vma->vm)->iomap,
vma->node.start,
@@ -368,7 +431,10 @@ void __iomem *i915_vma_pin_iomap(struct i915_vma *vma)
goto err;
}
- vma->iomap = ptr;
+ if (unlikely(cmpxchg(&vma->iomap, NULL, ptr))) {
+ io_mapping_unmap(ptr);
+ ptr = vma->iomap;
+ }
}
__i915_vma_pin(vma);
@@ -388,18 +454,12 @@ void __iomem *i915_vma_pin_iomap(struct i915_vma *vma)
void i915_vma_flush_writes(struct i915_vma *vma)
{
- if (!i915_vma_has_ggtt_write(vma))
- return;
-
- intel_gt_flush_ggtt_writes(vma->vm->gt);
-
- i915_vma_unset_ggtt_write(vma);
+ if (i915_vma_unset_ggtt_write(vma))
+ intel_gt_flush_ggtt_writes(vma->vm->gt);
}
void i915_vma_unpin_iomap(struct i915_vma *vma)
{
- lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
-
GEM_BUG_ON(vma->iomap == NULL);
i915_vma_flush_writes(vma);
@@ -435,6 +495,9 @@ bool i915_vma_misplaced(const struct i915_vma *vma,
if (!drm_mm_node_allocated(&vma->node))
return false;
+ if (test_bit(I915_VMA_ERROR_BIT, __i915_vma_flags(vma)))
+ return true;
+
if (vma->node.size < size)
return true;
@@ -535,7 +598,6 @@ static void assert_bind_count(const struct drm_i915_gem_object *obj)
static int
i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
{
- struct drm_i915_private *dev_priv = vma->vm->i915;
unsigned long color;
u64 start, end;
int ret;
@@ -561,7 +623,7 @@ i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
end = vma->vm->total;
if (flags & PIN_MAPPABLE)
- end = min_t(u64, end, dev_priv->ggtt.mappable_end);
+ end = min_t(u64, end, i915_vm_to_ggtt(vma->vm)->mappable_end);
if (flags & PIN_ZONE_4G)
end = min_t(u64, end, (1ULL << 32) - I915_GTT_PAGE_SIZE);
GEM_BUG_ON(!IS_ALIGNED(end, I915_GTT_PAGE_SIZE));
@@ -578,34 +640,20 @@ i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
}
color = 0;
- if (vma->obj) {
- ret = i915_gem_object_pin_pages(vma->obj);
- if (ret)
- return ret;
-
- if (i915_vm_has_cache_coloring(vma->vm))
- color = vma->obj->cache_level;
- }
-
- GEM_BUG_ON(vma->pages);
-
- ret = vma->ops->set_pages(vma);
- if (ret)
- goto err_unpin;
+ if (vma->obj && i915_vm_has_cache_coloring(vma->vm))
+ color = vma->obj->cache_level;
if (flags & PIN_OFFSET_FIXED) {
u64 offset = flags & PIN_OFFSET_MASK;
if (!IS_ALIGNED(offset, alignment) ||
- range_overflows(offset, size, end)) {
- ret = -EINVAL;
- goto err_clear;
- }
+ range_overflows(offset, size, end))
+ return -EINVAL;
ret = i915_gem_gtt_reserve(vma->vm, &vma->node,
size, offset, color,
flags);
if (ret)
- goto err_clear;
+ return ret;
} else {
/*
* We only support huge gtt pages through the 48b PPGTT,
@@ -644,7 +692,7 @@ i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
size, alignment, color,
start, end, flags);
if (ret)
- goto err_clear;
+ return ret;
GEM_BUG_ON(vma->node.start < start);
GEM_BUG_ON(vma->node.start + vma->node.size > end);
@@ -652,23 +700,15 @@ i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
GEM_BUG_ON(!i915_gem_valid_gtt_space(vma, color));
- mutex_lock(&vma->vm->mutex);
list_add_tail(&vma->vm_link, &vma->vm->bound_list);
- mutex_unlock(&vma->vm->mutex);
if (vma->obj) {
+ atomic_inc(&vma->obj->mm.pages_pin_count);
atomic_inc(&vma->obj->bind_count);
assert_bind_count(vma->obj);
}
return 0;
-
-err_clear:
- vma->ops->clear_pages(vma);
-err_unpin:
- if (vma->obj)
- i915_gem_object_unpin_pages(vma->obj);
- return ret;
}
static void
@@ -677,12 +717,7 @@ i915_vma_remove(struct i915_vma *vma)
GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
GEM_BUG_ON(i915_vma_is_bound(vma, I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND));
- vma->ops->clear_pages(vma);
-
- mutex_lock(&vma->vm->mutex);
- drm_mm_remove_node(&vma->node);
list_del(&vma->vm_link);
- mutex_unlock(&vma->vm->mutex);
/*
* Since the unbound list is global, only move to that list if
@@ -701,51 +736,211 @@ i915_vma_remove(struct i915_vma *vma)
i915_gem_object_unpin_pages(obj);
assert_bind_count(obj);
}
+
+ drm_mm_remove_node(&vma->node);
}
-int __i915_vma_do_pin(struct i915_vma *vma,
- u64 size, u64 alignment, u64 flags)
+static bool try_qad_pin(struct i915_vma *vma, unsigned int flags)
{
- const unsigned int bound = atomic_read(&vma->flags);
- int ret;
+ unsigned int bound;
+ bool pinned = true;
- lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
- GEM_BUG_ON((flags & (PIN_GLOBAL | PIN_USER)) == 0);
- GEM_BUG_ON((flags & PIN_GLOBAL) && !i915_vma_is_ggtt(vma));
+ bound = atomic_read(&vma->flags);
+ do {
+ if (unlikely(flags & ~bound))
+ return false;
- if (WARN_ON(bound & I915_VMA_PIN_OVERFLOW)) {
- ret = -EBUSY;
- goto err_unpin;
+ if (unlikely(bound & (I915_VMA_OVERFLOW | I915_VMA_ERROR)))
+ return false;
+
+ if (!(bound & I915_VMA_PIN_MASK))
+ goto unpinned;
+
+ GEM_BUG_ON(((bound + 1) & I915_VMA_PIN_MASK) == 0);
+ } while (!atomic_try_cmpxchg(&vma->flags, &bound, bound + 1));
+
+ return true;
+
+unpinned:
+ /*
+ * If pin_count==0, but we are bound, check under the lock to avoid
+ * racing with a concurrent i915_vma_unbind().
+ */
+ mutex_lock(&vma->vm->mutex);
+ do {
+ if (unlikely(bound & (I915_VMA_OVERFLOW | I915_VMA_ERROR))) {
+ pinned = false;
+ break;
+ }
+
+ if (unlikely(flags & ~bound)) {
+ pinned = false;
+ break;
+ }
+ } while (!atomic_try_cmpxchg(&vma->flags, &bound, bound + 1));
+ mutex_unlock(&vma->vm->mutex);
+
+ return pinned;
+}
+
+static int vma_get_pages(struct i915_vma *vma)
+{
+ int err = 0;
+
+ if (atomic_add_unless(&vma->pages_count, 1, 0))
+ return 0;
+
+ /* Allocations ahoy! */
+ if (mutex_lock_interruptible(&vma->pages_mutex))
+ return -EINTR;
+
+ if (!atomic_read(&vma->pages_count)) {
+ if (vma->obj) {
+ err = i915_gem_object_pin_pages(vma->obj);
+ if (err)
+ goto unlock;
+ }
+
+ err = vma->ops->set_pages(vma);
+ if (err)
+ goto unlock;
}
+ atomic_inc(&vma->pages_count);
- if ((bound & I915_VMA_BIND_MASK) == 0) {
- ret = i915_vma_insert(vma, size, alignment, flags);
- if (ret)
- goto err_unpin;
+unlock:
+ mutex_unlock(&vma->pages_mutex);
+
+ return err;
+}
+
+static void __vma_put_pages(struct i915_vma *vma, unsigned int count)
+{
+ /* We allocate under vma_get_pages, so beware the shrinker */
+ mutex_lock_nested(&vma->pages_mutex, SINGLE_DEPTH_NESTING);
+ GEM_BUG_ON(atomic_read(&vma->pages_count) < count);
+ if (atomic_sub_return(count, &vma->pages_count) == 0) {
+ vma->ops->clear_pages(vma);
+ GEM_BUG_ON(vma->pages);
+ if (vma->obj)
+ i915_gem_object_unpin_pages(vma->obj);
}
- GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
+ mutex_unlock(&vma->pages_mutex);
+}
- ret = i915_vma_bind(vma, vma->obj ? vma->obj->cache_level : 0, flags);
- if (ret)
- goto err_remove;
+static void vma_put_pages(struct i915_vma *vma)
+{
+ if (atomic_add_unless(&vma->pages_count, -1, 1))
+ return;
+
+ __vma_put_pages(vma, 1);
+}
+
+static void vma_unbind_pages(struct i915_vma *vma)
+{
+ unsigned int count;
+
+ lockdep_assert_held(&vma->vm->mutex);
+
+ /* The upper portion of pages_count is the number of bindings */
+ count = atomic_read(&vma->pages_count);
+ count >>= I915_VMA_PAGES_BIAS;
+ GEM_BUG_ON(!count);
+
+ __vma_put_pages(vma, count | count << I915_VMA_PAGES_BIAS);
+}
+
+int i915_vma_pin(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
+{
+ struct i915_vma_work *work = NULL;
+ unsigned int bound;
+ int err;
+
+ BUILD_BUG_ON(PIN_GLOBAL != I915_VMA_GLOBAL_BIND);
+ BUILD_BUG_ON(PIN_USER != I915_VMA_LOCAL_BIND);
+
+ GEM_BUG_ON(flags & PIN_UPDATE);
+ GEM_BUG_ON(!(flags & (PIN_USER | PIN_GLOBAL)));
+
+ /* First try and grab the pin without rebinding the vma */
+ if (try_qad_pin(vma, flags & I915_VMA_BIND_MASK))
+ return 0;
+
+ err = vma_get_pages(vma);
+ if (err)
+ return err;
+
+ if (flags & vma->vm->bind_alloc) {
+ work = i915_vma_work();
+ if (!work) {
+ err = -ENOMEM;
+ goto err_pages;
+ }
+ }
+
+ /* No more allocations allowed once we hold vm->mutex */
+ err = mutex_lock_interruptible(&vma->vm->mutex);
+ if (err)
+ goto err_fence;
+
+ bound = atomic_read(&vma->flags);
+ if (unlikely(bound & I915_VMA_ERROR)) {
+ err = -ENOMEM;
+ goto err_unlock;
+ }
+
+ if (unlikely(!((bound + 1) & I915_VMA_PIN_MASK))) {
+ err = -EAGAIN; /* pins are meant to be fairly temporary */
+ goto err_unlock;
+ }
+
+ if (unlikely(!(flags & ~bound & I915_VMA_BIND_MASK))) {
+ __i915_vma_pin(vma);
+ goto err_unlock;
+ }
+
+ err = i915_active_acquire(&vma->active);
+ if (err)
+ goto err_unlock;
+
+ if (!(bound & I915_VMA_BIND_MASK)) {
+ err = i915_vma_insert(vma, size, alignment, flags);
+ if (err)
+ goto err_active;
+
+ if (i915_is_ggtt(vma->vm))
+ __i915_vma_set_map_and_fenceable(vma);
+ }
- GEM_BUG_ON(!i915_vma_is_bound(vma, I915_VMA_BIND_MASK));
+ GEM_BUG_ON(!vma->pages);
+ err = i915_vma_bind(vma,
+ vma->obj ? vma->obj->cache_level : 0,
+ flags, work);
+ if (err)
+ goto err_remove;
- if ((bound ^ atomic_read(&vma->flags)) & I915_VMA_GLOBAL_BIND)
- __i915_vma_set_map_and_fenceable(vma);
+ /* There should only be at most 2 active bindings (user, global) */
+ GEM_BUG_ON(bound + I915_VMA_PAGES_ACTIVE < bound);
+ atomic_add(I915_VMA_PAGES_ACTIVE, &vma->pages_count);
+ list_move_tail(&vma->vm_link, &vma->vm->bound_list);
+ __i915_vma_pin(vma);
+ GEM_BUG_ON(!i915_vma_is_pinned(vma));
+ GEM_BUG_ON(!i915_vma_is_bound(vma, flags));
GEM_BUG_ON(i915_vma_misplaced(vma, size, alignment, flags));
- return 0;
err_remove:
- if ((bound & I915_VMA_BIND_MASK) == 0) {
+ if (!i915_vma_is_bound(vma, I915_VMA_BIND_MASK))
i915_vma_remove(vma);
- GEM_BUG_ON(vma->pages);
- GEM_BUG_ON(atomic_read(&vma->flags) & I915_VMA_BIND_MASK);
- }
-err_unpin:
- __i915_vma_unpin(vma);
- return ret;
+err_active:
+ i915_active_release(&vma->active);
+err_unlock:
+ mutex_unlock(&vma->vm->mutex);
+err_fence:
+ if (work)
+ dma_fence_work_commit(&work->base);
+err_pages:
+ vma_put_pages(vma);
+ return err;
}
void i915_vma_close(struct i915_vma *vma)
@@ -776,9 +971,6 @@ static void __i915_vma_remove_closed(struct i915_vma *vma)
{
struct drm_i915_private *i915 = vma->vm->i915;
- 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);
@@ -786,40 +978,35 @@ static void __i915_vma_remove_closed(struct i915_vma *vma)
void i915_vma_reopen(struct i915_vma *vma)
{
- __i915_vma_remove_closed(vma);
+ if (i915_vma_is_closed(vma))
+ __i915_vma_remove_closed(vma);
}
-static void __i915_vma_destroy(struct i915_vma *vma)
+void i915_vma_destroy(struct i915_vma *vma)
{
- GEM_BUG_ON(drm_mm_node_allocated(&vma->node));
- GEM_BUG_ON(vma->fence);
+ if (drm_mm_node_allocated(&vma->node)) {
+ mutex_lock(&vma->vm->mutex);
+ atomic_and(~I915_VMA_PIN_MASK, &vma->flags);
+ WARN_ON(__i915_vma_unbind(vma));
+ mutex_unlock(&vma->vm->mutex);
+ GEM_BUG_ON(drm_mm_node_allocated(&vma->node));
+ }
+ GEM_BUG_ON(i915_vma_is_active(vma));
if (vma->obj) {
struct drm_i915_gem_object *obj = vma->obj;
spin_lock(&obj->vma.lock);
list_del(&vma->obj_link);
- rb_erase(&vma->obj_node, &vma->obj->vma.tree);
+ rb_erase(&vma->obj_node, &obj->vma.tree);
spin_unlock(&obj->vma.lock);
}
- i915_active_fini(&vma->active);
-
- i915_vma_free(vma);
-}
-
-void i915_vma_destroy(struct i915_vma *vma)
-{
- lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
-
- GEM_BUG_ON(i915_vma_is_pinned(vma));
-
__i915_vma_remove_closed(vma);
+ i915_vm_put(vma->vm);
- WARN_ON(i915_vma_unbind(vma));
- GEM_BUG_ON(i915_vma_is_active(vma));
-
- __i915_vma_destroy(vma);
+ i915_active_fini(&vma->active);
+ i915_vma_free(vma);
}
void i915_vma_parked(struct drm_i915_private *i915)
@@ -828,12 +1015,32 @@ void i915_vma_parked(struct drm_i915_private *i915)
spin_lock_irq(&i915->gt.closed_lock);
list_for_each_entry_safe(vma, next, &i915->gt.closed_vma, closed_link) {
- list_del_init(&vma->closed_link);
+ struct drm_i915_gem_object *obj = vma->obj;
+ struct i915_address_space *vm = vma->vm;
+
+ /* XXX All to avoid keeping a reference on i915_vma itself */
+
+ if (!kref_get_unless_zero(&obj->base.refcount))
+ continue;
+
+ if (!i915_vm_tryopen(vm)) {
+ i915_gem_object_put(obj);
+ obj = NULL;
+ }
+
spin_unlock_irq(&i915->gt.closed_lock);
- i915_vma_destroy(vma);
+ if (obj) {
+ i915_vma_destroy(vma);
+ i915_gem_object_put(obj);
+ }
+ i915_vm_close(vm);
+
+ /* Restart after dropping lock */
spin_lock_irq(&i915->gt.closed_lock);
+ next = list_first_entry(&i915->gt.closed_vma,
+ typeof(*next), closed_link);
}
spin_unlock_irq(&i915->gt.closed_lock);
}
@@ -873,6 +1080,20 @@ void i915_vma_revoke_mmap(struct i915_vma *vma)
list_del(&vma->obj->userfault_link);
}
+int __i915_vma_move_to_active(struct i915_vma *vma, struct i915_request *rq)
+{
+ int err;
+
+ GEM_BUG_ON(!i915_vma_is_pinned(vma));
+
+ /* Wait for the vma to be bound before we start! */
+ err = i915_request_await_active(rq, &vma->active);
+ if (err)
+ return err;
+
+ return i915_active_add_request(&vma->active, rq);
+}
+
int i915_vma_move_to_active(struct i915_vma *vma,
struct i915_request *rq,
unsigned int flags)
@@ -880,19 +1101,9 @@ int i915_vma_move_to_active(struct i915_vma *vma,
struct drm_i915_gem_object *obj = vma->obj;
int err;
- assert_vma_held(vma);
assert_object_held(obj);
- GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
- /*
- * Add a reference if we're newly entering the active list.
- * The order in which we add operations to the retirement queue is
- * vital here: mark_active adds to the start of the callback list,
- * such that subsequent callbacks are called first. Therefore we
- * add the active reference first and queue for it to be dropped
- * *last*.
- */
- err = i915_active_add_request(&vma->active, rq);
+ err = __i915_vma_move_to_active(vma, rq);
if (unlikely(err))
return err;
@@ -918,38 +1129,23 @@ int i915_vma_move_to_active(struct i915_vma *vma,
return 0;
}
-int i915_vma_unbind(struct i915_vma *vma)
+int __i915_vma_unbind(struct i915_vma *vma)
{
int ret;
- lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
+ lockdep_assert_held(&vma->vm->mutex);
/*
* First wait upon any activity as retiring the request may
* have side-effects such as unpinning or even unbinding this vma.
+ *
+ * XXX Actually waiting under the vm->mutex is a hinderance and
+ * should be pipelined wherever possible. In cases where that is
+ * unavoidable, we should lift the wait to before the mutex.
*/
- might_sleep();
- if (i915_vma_is_active(vma)) {
- /*
- * When a closed VMA is retired, it is unbound - eek.
- * In order to prevent it from being recursively closed,
- * take a pin on the vma so that the second unbind is
- * aborted.
- *
- * Even more scary is that the retire callback may free
- * the object (last active vma). To prevent the explosion
- * we defer the actual object free to a worker that can
- * only proceed once it acquires the struct_mutex (which
- * we currently hold, therefore it cannot free this object
- * before we are finished).
- */
- __i915_vma_pin(vma);
- ret = i915_active_wait(&vma->active);
- __i915_vma_unpin(vma);
- if (ret)
- return ret;
- }
- GEM_BUG_ON(i915_vma_is_active(vma));
+ ret = i915_vma_sync(vma);
+ if (ret)
+ return ret;
if (i915_vma_is_pinned(vma)) {
vma_print_allocator(vma, "is pinned");
@@ -970,16 +1166,12 @@ int i915_vma_unbind(struct i915_vma *vma)
GEM_BUG_ON(i915_vma_has_ggtt_write(vma));
/* release the fence reg _after_ flushing */
- mutex_lock(&vma->vm->mutex);
ret = i915_vma_revoke_fence(vma);
- mutex_unlock(&vma->vm->mutex);
if (ret)
return ret;
/* Force a pagefault for domain tracking on next user access */
- mutex_lock(&vma->vm->mutex);
i915_vma_revoke_mmap(vma);
- mutex_unlock(&vma->vm->mutex);
__i915_vma_iounmap(vma);
clear_bit(I915_VMA_CAN_FENCE_BIT, __i915_vma_flags(vma));
@@ -987,17 +1179,33 @@ int i915_vma_unbind(struct i915_vma *vma)
GEM_BUG_ON(vma->fence);
GEM_BUG_ON(i915_vma_has_userfault(vma));
- if (likely(!vma->vm->closed)) {
+ if (likely(atomic_read(&vma->vm->open))) {
trace_i915_vma_unbind(vma);
vma->ops->unbind_vma(vma);
}
- atomic_and(~I915_VMA_BIND_MASK, &vma->flags);
+ atomic_and(~(I915_VMA_BIND_MASK | I915_VMA_ERROR), &vma->flags);
+ vma_unbind_pages(vma);
i915_vma_remove(vma);
return 0;
}
+int i915_vma_unbind(struct i915_vma *vma)
+{
+ struct i915_address_space *vm = vma->vm;
+ int err;
+
+ err = mutex_lock_interruptible(&vm->mutex);
+ if (err)
+ return err;
+
+ err = __i915_vma_unbind(vma);
+ mutex_unlock(&vm->mutex);
+
+ return err;
+}
+
struct i915_vma *i915_vma_make_unshrinkable(struct i915_vma *vma)
{
i915_gem_object_make_unshrinkable(vma->obj);
diff --git a/drivers/gpu/drm/i915/i915_vma.h b/drivers/gpu/drm/i915/i915_vma.h
index e49b199f7de7..858908e3d1cc 100644
--- a/drivers/gpu/drm/i915/i915_vma.h
+++ b/drivers/gpu/drm/i915/i915_vma.h
@@ -96,25 +96,28 @@ struct i915_vma {
* exclusive cachelines of a single page, so a maximum of 64 possible
* users.
*/
-#define I915_VMA_PIN_MASK 0xff
-#define I915_VMA_PIN_OVERFLOW_BIT 8
-#define I915_VMA_PIN_OVERFLOW ((int)BIT(I915_VMA_PIN_OVERFLOW_BIT))
+#define I915_VMA_PIN_MASK 0x3ff
+#define I915_VMA_OVERFLOW 0x200
/** Flags and address space this VMA is bound to */
-#define I915_VMA_GLOBAL_BIND_BIT 9
-#define I915_VMA_LOCAL_BIND_BIT 10
+#define I915_VMA_GLOBAL_BIND_BIT 10
+#define I915_VMA_LOCAL_BIND_BIT 11
#define I915_VMA_GLOBAL_BIND ((int)BIT(I915_VMA_GLOBAL_BIND_BIT))
#define I915_VMA_LOCAL_BIND ((int)BIT(I915_VMA_LOCAL_BIND_BIT))
-#define I915_VMA_BIND_MASK (I915_VMA_GLOBAL_BIND | \
- I915_VMA_LOCAL_BIND | \
- I915_VMA_PIN_OVERFLOW)
+#define I915_VMA_BIND_MASK (I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND)
-#define I915_VMA_GGTT_BIT 11
-#define I915_VMA_CAN_FENCE_BIT 12
-#define I915_VMA_USERFAULT_BIT 13
-#define I915_VMA_GGTT_WRITE_BIT 14
+#define I915_VMA_ALLOC_BIT 12
+#define I915_VMA_ALLOC ((int)BIT(I915_VMA_ALLOC_BIT))
+
+#define I915_VMA_ERROR_BIT 13
+#define I915_VMA_ERROR ((int)BIT(I915_VMA_ERROR_BIT))
+
+#define I915_VMA_GGTT_BIT 14
+#define I915_VMA_CAN_FENCE_BIT 15
+#define I915_VMA_USERFAULT_BIT 16
+#define I915_VMA_GGTT_WRITE_BIT 17
#define I915_VMA_GGTT ((int)BIT(I915_VMA_GGTT_BIT))
#define I915_VMA_CAN_FENCE ((int)BIT(I915_VMA_CAN_FENCE_BIT))
@@ -123,6 +126,11 @@ struct i915_vma {
struct i915_active active;
+#define I915_VMA_PAGES_BIAS 24
+#define I915_VMA_PAGES_ACTIVE (BIT(24) | 1)
+ atomic_t pages_count; /* number of active binds to the pages */
+ struct mutex pages_mutex; /* protect acquire/release of backing pages */
+
/**
* Support different GGTT views into the same object.
* This means there can be multiple VMA mappings per object and per VM.
@@ -169,6 +177,8 @@ static inline bool i915_vma_is_active(const struct i915_vma *vma)
return !i915_active_is_idle(&vma->active);
}
+int __must_check __i915_vma_move_to_active(struct i915_vma *vma,
+ struct i915_request *rq);
int __must_check i915_vma_move_to_active(struct i915_vma *vma,
struct i915_request *rq,
unsigned int flags);
@@ -307,13 +317,18 @@ i915_vma_compare(struct i915_vma *vma,
return memcmp(&vma->ggtt_view.partial, &view->partial, view->type);
}
-int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level,
- u32 flags);
+struct i915_vma_work *i915_vma_work(void);
+int i915_vma_bind(struct i915_vma *vma,
+ enum i915_cache_level cache_level,
+ u32 flags,
+ struct i915_vma_work *work);
+
bool i915_gem_valid_gtt_space(struct i915_vma *vma, unsigned long color);
bool i915_vma_misplaced(const struct i915_vma *vma,
u64 size, u64 alignment, u64 flags);
void __i915_vma_set_map_and_fenceable(struct i915_vma *vma);
void i915_vma_revoke_mmap(struct i915_vma *vma);
+int __i915_vma_unbind(struct i915_vma *vma);
int __must_check i915_vma_unbind(struct i915_vma *vma);
void i915_vma_unlink_ctx(struct i915_vma *vma);
void i915_vma_close(struct i915_vma *vma);
@@ -332,26 +347,8 @@ static inline void i915_vma_unlock(struct i915_vma *vma)
dma_resv_unlock(vma->resv);
}
-int __i915_vma_do_pin(struct i915_vma *vma,
- u64 size, u64 alignment, u64 flags);
-static inline int __must_check
-i915_vma_pin(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
-{
- BUILD_BUG_ON(PIN_MBZ != I915_VMA_PIN_OVERFLOW);
- BUILD_BUG_ON(PIN_GLOBAL != I915_VMA_GLOBAL_BIND);
- BUILD_BUG_ON(PIN_USER != I915_VMA_LOCAL_BIND);
-
- /* Pin early to prevent the shrinker/eviction logic from destroying
- * our vma as we insert and bind.
- */
- if (likely(((atomic_inc_return(&vma->flags) ^ flags) & I915_VMA_BIND_MASK) == 0)) {
- GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
- GEM_BUG_ON(i915_vma_misplaced(vma, size, alignment, flags));
- return 0;
- }
-
- return __i915_vma_do_pin(vma, size, alignment, flags);
-}
+int __must_check
+i915_vma_pin(struct i915_vma *vma, u64 size, u64 alignment, u64 flags);
static inline int i915_vma_pin_count(const struct i915_vma *vma)
{
@@ -366,17 +363,17 @@ static inline bool i915_vma_is_pinned(const struct i915_vma *vma)
static inline void __i915_vma_pin(struct i915_vma *vma)
{
atomic_inc(&vma->flags);
- GEM_BUG_ON(atomic_read(&vma->flags) & I915_VMA_PIN_OVERFLOW);
+ GEM_BUG_ON(!i915_vma_is_pinned(vma));
}
static inline void __i915_vma_unpin(struct i915_vma *vma)
{
+ GEM_BUG_ON(!i915_vma_is_pinned(vma));
atomic_dec(&vma->flags);
}
static inline void i915_vma_unpin(struct i915_vma *vma)
{
- GEM_BUG_ON(!i915_vma_is_pinned(vma));
GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
__i915_vma_unpin(vma);
}
@@ -402,8 +399,6 @@ static inline bool i915_node_color_differs(const struct drm_mm_node *node,
* the caller must call i915_vma_unpin_iomap to relinquish the pinning
* after the iomapping is no longer required.
*
- * Callers must hold the struct_mutex.
- *
* Returns a valid iomapped pointer or ERR_PTR.
*/
void __iomem *i915_vma_pin_iomap(struct i915_vma *vma);
@@ -415,8 +410,8 @@ void __iomem *i915_vma_pin_iomap(struct i915_vma *vma);
*
* Unpins the previously iomapped VMA from i915_vma_pin_iomap().
*
- * Callers must hold the struct_mutex. This function is only valid to be
- * called on a VMA previously iomapped by the caller with i915_vma_pin_iomap().
+ * This function is only valid to be called on a VMA previously
+ * iomapped by the caller with i915_vma_pin_iomap().
*/
void i915_vma_unpin_iomap(struct i915_vma *vma);
@@ -444,6 +439,8 @@ static inline struct page *i915_vma_first_page(struct i915_vma *vma)
int __must_check i915_vma_pin_fence(struct i915_vma *vma);
int __must_check i915_vma_revoke_fence(struct i915_vma *vma);
+int __i915_vma_pin_fence(struct i915_vma *vma);
+
static inline void __i915_vma_unpin_fence(struct i915_vma *vma)
{
GEM_BUG_ON(atomic_read(&vma->fence->pin_count) <= 0);
@@ -461,7 +458,6 @@ static inline void __i915_vma_unpin_fence(struct i915_vma *vma)
static inline void
i915_vma_unpin_fence(struct i915_vma *vma)
{
- /* lockdep_assert_held(&vma->vm->i915->drm.struct_mutex); */
if (vma->fence)
__i915_vma_unpin_fence(vma);
}
@@ -490,4 +486,10 @@ struct i915_vma *i915_vma_make_unshrinkable(struct i915_vma *vma);
void i915_vma_make_shrinkable(struct i915_vma *vma);
void i915_vma_make_purgeable(struct i915_vma *vma);
+static inline int i915_vma_sync(struct i915_vma *vma)
+{
+ /* Wait for the asynchronous bindings and pending GPU reads */
+ return i915_active_wait(&vma->active);
+}
+
#endif
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem.c b/drivers/gpu/drm/i915/selftests/i915_gem.c
index 37593831b539..0346c3e5b6b6 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem.c
@@ -119,10 +119,8 @@ static void pm_resume(struct drm_i915_private *i915)
intel_gt_sanitize(&i915->gt, false);
i915_gem_sanitize(i915);
- mutex_lock(&i915->drm.struct_mutex);
i915_gem_restore_gtt_mappings(i915);
i915_gem_restore_fences(i915);
- mutex_unlock(&i915->drm.struct_mutex);
i915_gem_resume(i915);
}
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
index 2905fb21d866..75a4695b82bb 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
@@ -106,14 +106,11 @@ static int populate_ggtt(struct drm_i915_private *i915,
static void unpin_ggtt(struct drm_i915_private *i915)
{
- struct i915_ggtt *ggtt = &i915->ggtt;
struct i915_vma *vma;
- mutex_lock(&ggtt->vm.mutex);
list_for_each_entry(vma, &i915->ggtt.vm.bound_list, vm_link)
if (vma->obj->mm.quirked)
i915_vma_unpin(vma);
- mutex_unlock(&ggtt->vm.mutex);
}
static void cleanup_objects(struct drm_i915_private *i915,
@@ -127,11 +124,7 @@ static void cleanup_objects(struct drm_i915_private *i915,
i915_gem_object_put(obj);
}
- mutex_unlock(&i915->drm.struct_mutex);
-
i915_gem_drain_freed_objects(i915);
-
- mutex_lock(&i915->drm.struct_mutex);
}
static int igt_evict_something(void *arg)
@@ -148,10 +141,12 @@ static int igt_evict_something(void *arg)
goto cleanup;
/* Everything is pinned, nothing should happen */
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_evict_something(&ggtt->vm,
I915_GTT_PAGE_SIZE, 0, 0,
0, U64_MAX,
0);
+ mutex_unlock(&ggtt->vm.mutex);
if (err != -ENOSPC) {
pr_err("i915_gem_evict_something failed on a full GGTT with err=%d\n",
err);
@@ -161,10 +156,12 @@ static int igt_evict_something(void *arg)
unpin_ggtt(i915);
/* Everything is unpinned, we should be able to evict something */
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_evict_something(&ggtt->vm,
I915_GTT_PAGE_SIZE, 0, 0,
0, U64_MAX,
0);
+ mutex_unlock(&ggtt->vm.mutex);
if (err) {
pr_err("i915_gem_evict_something failed on a full GGTT with err=%d\n",
err);
@@ -230,7 +227,9 @@ static int igt_evict_for_vma(void *arg)
goto cleanup;
/* Everything is pinned, nothing should happen */
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_evict_for_node(&ggtt->vm, &target, 0);
+ mutex_unlock(&ggtt->vm.mutex);
if (err != -ENOSPC) {
pr_err("i915_gem_evict_for_node on a full GGTT returned err=%d\n",
err);
@@ -240,7 +239,9 @@ static int igt_evict_for_vma(void *arg)
unpin_ggtt(i915);
/* Everything is unpinned, we should be able to evict the node */
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_evict_for_node(&ggtt->vm, &target, 0);
+ mutex_unlock(&ggtt->vm.mutex);
if (err) {
pr_err("i915_gem_evict_for_node returned err=%d\n",
err);
@@ -319,7 +320,9 @@ static int igt_evict_for_cache_color(void *arg)
i915_vma_unpin(vma);
/* Remove just the second vma */
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_evict_for_node(&ggtt->vm, &target, 0);
+ mutex_unlock(&ggtt->vm.mutex);
if (err) {
pr_err("[0]i915_gem_evict_for_node returned err=%d\n", err);
goto cleanup;
@@ -330,7 +333,9 @@ static int igt_evict_for_cache_color(void *arg)
*/
target.color = I915_CACHE_L3_LLC;
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_evict_for_node(&ggtt->vm, &target, 0);
+ mutex_unlock(&ggtt->vm.mutex);
if (!err) {
pr_err("[1]i915_gem_evict_for_node returned err=%d\n", err);
err = -EINVAL;
@@ -360,7 +365,9 @@ static int igt_evict_vm(void *arg)
goto cleanup;
/* Everything is pinned, nothing should happen */
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_evict_vm(&ggtt->vm);
+ mutex_unlock(&ggtt->vm.mutex);
if (err) {
pr_err("i915_gem_evict_vm on a full GGTT returned err=%d]\n",
err);
@@ -369,7 +376,9 @@ static int igt_evict_vm(void *arg)
unpin_ggtt(i915);
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_evict_vm(&ggtt->vm);
+ mutex_unlock(&ggtt->vm.mutex);
if (err) {
pr_err("i915_gem_evict_vm on a full GGTT returned err=%d]\n",
err);
@@ -410,11 +419,11 @@ static int igt_evict_contexts(void *arg)
if (!HAS_FULL_PPGTT(i915))
return 0;
- mutex_lock(&i915->drm.struct_mutex);
wakeref = intel_runtime_pm_get(&i915->runtime_pm);
/* Reserve a block so that we know we have enough to fit a few rq */
memset(&hole, 0, sizeof(hole));
+ mutex_lock(&i915->ggtt.vm.mutex);
err = i915_gem_gtt_insert(&i915->ggtt.vm, &hole,
PRETEND_GGTT_SIZE, 0, I915_COLOR_UNEVICTABLE,
0, i915->ggtt.vm.total,
@@ -427,7 +436,9 @@ static int igt_evict_contexts(void *arg)
do {
struct reserved *r;
+ mutex_unlock(&i915->ggtt.vm.mutex);
r = kcalloc(1, sizeof(*r), GFP_KERNEL);
+ mutex_lock(&i915->ggtt.vm.mutex);
if (!r) {
err = -ENOMEM;
goto out_locked;
@@ -447,7 +458,7 @@ static int igt_evict_contexts(void *arg)
count++;
} while (1);
drm_mm_remove_node(&hole);
- mutex_unlock(&i915->drm.struct_mutex);
+ mutex_unlock(&i915->ggtt.vm.mutex);
pr_info("Filled GGTT with %lu 1MiB nodes\n", count);
/* Overfill the GGTT with context objects and so try to evict one. */
@@ -510,7 +521,7 @@ static int igt_evict_contexts(void *arg)
break;
}
- mutex_lock(&i915->drm.struct_mutex);
+ mutex_lock(&i915->ggtt.vm.mutex);
out_locked:
if (igt_flush_test(i915, I915_WAIT_LOCKED))
err = -EIO;
@@ -524,8 +535,8 @@ static int igt_evict_contexts(void *arg)
}
if (drm_mm_node_allocated(&hole))
drm_mm_remove_node(&hole);
+ mutex_unlock(&i915->ggtt.vm.mutex);
intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -547,12 +558,9 @@ int i915_gem_evict_mock_selftests(void)
if (!i915)
return -ENOMEM;
- mutex_lock(&i915->drm.struct_mutex);
with_intel_runtime_pm(&i915->runtime_pm, wakeref)
err = i915_subtests(tests, i915);
- mutex_unlock(&i915->drm.struct_mutex);
-
drm_dev_put(&i915->drm);
return err;
}
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
index 84d94eabc7dd..373fbd58a17a 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
@@ -38,16 +38,7 @@
static void cleanup_freed_objects(struct drm_i915_private *i915)
{
- /*
- * As we may hold onto the struct_mutex for inordinate lengths of
- * time, the NMI khungtaskd detector may fire for the free objects
- * worker.
- */
- mutex_unlock(&i915->drm.struct_mutex);
-
i915_gem_drain_freed_objects(i915);
-
- mutex_lock(&i915->drm.struct_mutex);
}
static void fake_free_pages(struct drm_i915_gem_object *obj,
@@ -880,6 +871,15 @@ static int __shrink_hole(struct drm_i915_private *i915,
i915_vma_unpin(vma);
addr += size;
+ /*
+ * Since we are injecting allocation faults at random intervals,
+ * wait for this allocation to complete before we change the
+ * faultinjection.
+ */
+ err = i915_vma_sync(vma);
+ if (err)
+ break;
+
if (igt_timeout(end_time,
"%s timed out at ofset %llx [%llx - %llx]\n",
__func__, addr, hole_start, hole_end)) {
@@ -1013,21 +1013,19 @@ static int exercise_ppgtt(struct drm_i915_private *dev_priv,
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(&dev_priv->drm.struct_mutex);
ppgtt = i915_ppgtt_create(dev_priv);
if (IS_ERR(ppgtt)) {
err = PTR_ERR(ppgtt);
- goto out_unlock;
+ goto out_free;
}
GEM_BUG_ON(offset_in_page(ppgtt->vm.total));
- GEM_BUG_ON(ppgtt->vm.closed);
+ GEM_BUG_ON(!atomic_read(&ppgtt->vm.open));
err = func(dev_priv, &ppgtt->vm, 0, ppgtt->vm.total, end_time);
i915_vm_put(&ppgtt->vm);
-out_unlock:
- mutex_unlock(&dev_priv->drm.struct_mutex);
+out_free:
mock_file_free(dev_priv, file);
return err;
}
@@ -1090,7 +1088,6 @@ static int exercise_ggtt(struct drm_i915_private *i915,
IGT_TIMEOUT(end_time);
int err = 0;
- mutex_lock(&i915->drm.struct_mutex);
restart:
list_sort(NULL, &ggtt->vm.mm.hole_stack, sort_holes);
drm_mm_for_each_hole(node, &ggtt->vm.mm, hole_start, hole_end) {
@@ -1111,7 +1108,6 @@ static int exercise_ggtt(struct drm_i915_private *i915,
last = hole_end;
goto restart;
}
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -1153,13 +1149,9 @@ static int igt_ggtt_page(void *arg)
unsigned int *order, n;
int err;
- mutex_lock(&i915->drm.struct_mutex);
-
obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
- if (IS_ERR(obj)) {
- err = PTR_ERR(obj);
- goto out_unlock;
- }
+ if (IS_ERR(obj))
+ return PTR_ERR(obj);
err = i915_gem_object_pin_pages(obj);
if (err)
@@ -1227,8 +1219,6 @@ static int igt_ggtt_page(void *arg)
i915_gem_object_unpin_pages(obj);
out_free:
i915_gem_object_put(obj);
-out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -1239,6 +1229,9 @@ static void track_vma_bind(struct i915_vma *vma)
atomic_inc(&obj->bind_count); /* track for eviction later */
__i915_gem_object_pin_pages(obj);
+ GEM_BUG_ON(vma->pages);
+ atomic_set(&vma->pages_count, I915_VMA_PAGES_ACTIVE);
+ __i915_gem_object_pin_pages(obj);
vma->pages = obj->mm.pages;
mutex_lock(&vma->vm->mutex);
@@ -1335,11 +1328,13 @@ static int igt_gtt_reserve(void *arg)
goto out;
}
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_gtt_reserve(&ggtt->vm, &vma->node,
obj->base.size,
total,
obj->cache_level,
0);
+ mutex_unlock(&ggtt->vm.mutex);
if (err) {
pr_err("i915_gem_gtt_reserve (pass 1) failed at %llu/%llu with err=%d\n",
total, ggtt->vm.total, err);
@@ -1385,11 +1380,13 @@ static int igt_gtt_reserve(void *arg)
goto out;
}
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_gtt_reserve(&ggtt->vm, &vma->node,
obj->base.size,
total,
obj->cache_level,
0);
+ mutex_unlock(&ggtt->vm.mutex);
if (err) {
pr_err("i915_gem_gtt_reserve (pass 2) failed at %llu/%llu with err=%d\n",
total, ggtt->vm.total, err);
@@ -1429,11 +1426,13 @@ static int igt_gtt_reserve(void *arg)
2*I915_GTT_PAGE_SIZE,
I915_GTT_MIN_ALIGNMENT);
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_gtt_reserve(&ggtt->vm, &vma->node,
obj->base.size,
offset,
obj->cache_level,
0);
+ mutex_unlock(&ggtt->vm.mutex);
if (err) {
pr_err("i915_gem_gtt_reserve (pass 3) failed at %llu/%llu with err=%d\n",
total, ggtt->vm.total, err);
@@ -1502,11 +1501,13 @@ static int igt_gtt_insert(void *arg)
/* Check a couple of obviously invalid requests */
for (ii = invalid_insert; ii->size; ii++) {
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_gtt_insert(&ggtt->vm, &tmp,
ii->size, ii->alignment,
I915_COLOR_UNEVICTABLE,
ii->start, ii->end,
0);
+ mutex_unlock(&ggtt->vm.mutex);
if (err != -ENOSPC) {
pr_err("Invalid i915_gem_gtt_insert(.size=%llx, .alignment=%llx, .start=%llx, .end=%llx) succeeded (err=%d)\n",
ii->size, ii->alignment, ii->start, ii->end,
@@ -1542,10 +1543,12 @@ static int igt_gtt_insert(void *arg)
goto out;
}
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_gtt_insert(&ggtt->vm, &vma->node,
obj->base.size, 0, obj->cache_level,
0, ggtt->vm.total,
0);
+ mutex_unlock(&ggtt->vm.mutex);
if (err == -ENOSPC) {
/* maxed out the GGTT space */
i915_gem_object_put(obj);
@@ -1600,10 +1603,12 @@ static int igt_gtt_insert(void *arg)
goto out;
}
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_gtt_insert(&ggtt->vm, &vma->node,
obj->base.size, 0, obj->cache_level,
0, ggtt->vm.total,
0);
+ mutex_unlock(&ggtt->vm.mutex);
if (err) {
pr_err("i915_gem_gtt_insert (pass 2) failed at %llu/%llu with err=%d\n",
total, ggtt->vm.total, err);
@@ -1647,10 +1652,12 @@ static int igt_gtt_insert(void *arg)
goto out;
}
+ mutex_lock(&ggtt->vm.mutex);
err = i915_gem_gtt_insert(&ggtt->vm, &vma->node,
obj->base.size, 0, obj->cache_level,
0, ggtt->vm.total,
0);
+ mutex_unlock(&ggtt->vm.mutex);
if (err) {
pr_err("i915_gem_gtt_insert (pass 3) failed at %llu/%llu with err=%d\n",
total, ggtt->vm.total, err);
@@ -1694,8 +1701,9 @@ int i915_gem_gtt_mock_selftests(void)
}
mock_init_ggtt(i915, ggtt);
- mutex_lock(&i915->drm.struct_mutex);
err = i915_subtests(tests, ggtt);
+
+ mutex_lock(&i915->drm.struct_mutex);
mock_device_flush(i915);
mutex_unlock(&i915->drm.struct_mutex);
diff --git a/drivers/gpu/drm/i915/selftests/i915_request.c b/drivers/gpu/drm/i915/selftests/i915_request.c
index b3688543ed7d..6b5b03081880 100644
--- a/drivers/gpu/drm/i915/selftests/i915_request.c
+++ b/drivers/gpu/drm/i915/selftests/i915_request.c
@@ -647,8 +647,15 @@ static struct i915_vma *empty_batch(struct drm_i915_private *i915)
if (err)
goto err;
+ /* Force the wait wait now to avoid including it in the benchmark */
+ err = i915_vma_sync(vma);
+ if (err)
+ goto err_pin;
+
return vma;
+err_pin:
+ i915_vma_unpin(vma);
err:
i915_gem_object_put(obj);
return ERR_PTR(err);
diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c
index 97752deecccb..0e4f66312b39 100644
--- a/drivers/gpu/drm/i915/selftests/i915_vma.c
+++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
@@ -831,8 +831,9 @@ int i915_vma_mock_selftests(void)
}
mock_init_ggtt(i915, ggtt);
- mutex_lock(&i915->drm.struct_mutex);
err = i915_subtests(tests, ggtt);
+
+ mutex_lock(&i915->drm.struct_mutex);
mock_device_flush(i915);
mutex_unlock(&i915->drm.struct_mutex);
@@ -879,8 +880,6 @@ static int igt_vma_remapped_gtt(void *arg)
if (IS_ERR(obj))
return PTR_ERR(obj);
- mutex_lock(&i915->drm.struct_mutex);
-
wakeref = intel_runtime_pm_get(&i915->runtime_pm);
for (t = types; *t; t++) {
@@ -976,7 +975,6 @@ static int igt_vma_remapped_gtt(void *arg)
out:
intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
i915_gem_object_put(obj);
return err;
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* Re: [PATCH 05/27] drm/i915: Pull i915_vma_pin under the vm->mutex
2019-09-25 10:01 ` [PATCH 05/27] drm/i915: Pull i915_vma_pin under the vm->mutex Chris Wilson
@ 2019-09-27 8:47 ` Tvrtko Ursulin
2019-09-27 11:06 ` Chris Wilson
0 siblings, 1 reply; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-27 8:47 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 25/09/2019 11:01, Chris Wilson wrote:
> Replace the struct_mutex requirement for pinning the i915_vma with the
> local vm->mutex instead. Note that the vm->mutex is tainted by the
> shrinker (we require unbinding from inside fs-reclaim) and so we cannot
> allocate while holding that mutex. Instead we have to preallocate
> workers to do allocate and apply the PTE updates after we have we
> reserved their slot in the drm_mm (using fences to order the PTE writes
> with the GPU work and with later unbind).
>
> In adding the asynchronous vma binding, one subtle requirement is to
> avoid coupling the binding fence into the backing object->resv. That is
> the asynchronous binding only applies to the vma timeline itself and not
> to the pages as that is a more global timeline (the binding of one vma
> does not need to be ordered with another vma, nor does the implicit GEM
> fencing depend on a vma, only on writes to the backing store). Keeping
> the vma binding distinct from the backing store timelines is verified by
> a number of async gem_exec_fence and gem_exec_schedule tests. The way we
> do this is quite simple, we keep the fence for the vma binding separate
> and only wait on it as required, and never add it to the obj->resv
> itself.
>
> Another consequence in reducing the locking around the vma is the
> destruction of the vma is no longer globally serialised by struct_mutex.
> A natural solution would be to add a kref to i915_vma, but that requires
> decoupling the reference cycles, possibly by introducing a new
> i915_mm_pages object that is own by both obj->mm and vma->pages.
> However, we have not taken that route due to the overshadowing lmem/ttm
> discussions, and instead play a series of complicated games with
> trylocks to (hopefully) ensure that only one destruction path is called!
>
> v2: Add some commentary, and some helpers to reduce patch churn.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> ---
> drivers/gpu/drm/i915/display/intel_display.c | 29 +-
> drivers/gpu/drm/i915/display/intel_fbdev.c | 8 +-
> drivers/gpu/drm/i915/display/intel_overlay.c | 11 +-
> .../gpu/drm/i915/gem/i915_gem_client_blt.c | 2 +-
> drivers/gpu/drm/i915/gem/i915_gem_context.c | 20 +-
> drivers/gpu/drm/i915/gem/i915_gem_domain.c | 13 +-
> .../gpu/drm/i915/gem/i915_gem_execbuffer.c | 43 +-
> drivers/gpu/drm/i915/gem/i915_gem_mman.c | 20 +-
> drivers/gpu/drm/i915/gem/i915_gem_object.c | 33 +-
> drivers/gpu/drm/i915/gem/i915_gem_object.h | 5 +
> drivers/gpu/drm/i915/gem/i915_gem_shrinker.c | 73 +--
> drivers/gpu/drm/i915/gem/i915_gem_stolen.c | 8 +-
> drivers/gpu/drm/i915/gem/i915_gem_tiling.c | 27 +-
> drivers/gpu/drm/i915/gem/i915_gem_userptr.c | 27 +-
> .../gpu/drm/i915/gem/selftests/huge_pages.c | 9 +-
> .../drm/i915/gem/selftests/i915_gem_context.c | 12 +-
> .../drm/i915/gem/selftests/i915_gem_mman.c | 4 -
> .../drm/i915/gem/selftests/igt_gem_utils.c | 7 +-
> drivers/gpu/drm/i915/gt/intel_gt.c | 5 +-
> drivers/gpu/drm/i915/gt/intel_ringbuffer.c | 4 +-
> drivers/gpu/drm/i915/gt/selftest_hangcheck.c | 19 +-
> drivers/gpu/drm/i915/gvt/aperture_gm.c | 12 +-
> drivers/gpu/drm/i915/i915_active.c | 95 +++-
> drivers/gpu/drm/i915/i915_active.h | 7 +
> drivers/gpu/drm/i915/i915_active_types.h | 5 +
> drivers/gpu/drm/i915/i915_drv.c | 2 -
> drivers/gpu/drm/i915/i915_gem.c | 83 ++-
> drivers/gpu/drm/i915/i915_gem_evict.c | 28 +-
> drivers/gpu/drm/i915/i915_gem_fence_reg.c | 9 +-
> drivers/gpu/drm/i915/i915_gem_gtt.c | 109 ++--
> drivers/gpu/drm/i915/i915_gem_gtt.h | 45 +-
> drivers/gpu/drm/i915/i915_perf.c | 32 +-
> drivers/gpu/drm/i915/i915_vma.c | 524 ++++++++++++------
> drivers/gpu/drm/i915/i915_vma.h | 84 +--
> drivers/gpu/drm/i915/selftests/i915_gem.c | 2 -
> .../gpu/drm/i915/selftests/i915_gem_evict.c | 36 +-
> drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 58 +-
> drivers/gpu/drm/i915/selftests/i915_request.c | 7 +
> drivers/gpu/drm/i915/selftests/i915_vma.c | 6 +-
> 39 files changed, 823 insertions(+), 700 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
> index 5ecf54270181..12abcd1e471b 100644
> --- a/drivers/gpu/drm/i915/display/intel_display.c
> +++ b/drivers/gpu/drm/i915/display/intel_display.c
> @@ -2079,7 +2079,6 @@ intel_pin_and_fence_fb_obj(struct drm_framebuffer *fb,
> unsigned int pinctl;
> u32 alignment;
>
> - WARN_ON(!mutex_is_locked(&dev->struct_mutex));
> if (WARN_ON(!i915_gem_object_is_framebuffer(obj)))
> return ERR_PTR(-EINVAL);
>
> @@ -2163,8 +2162,6 @@ intel_pin_and_fence_fb_obj(struct drm_framebuffer *fb,
>
> 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);
> @@ -3065,12 +3062,10 @@ intel_alloc_initial_plane_obj(struct intel_crtc *crtc,
> return false;
> }
>
> - mutex_lock(&dev->struct_mutex);
> obj = i915_gem_object_create_stolen_for_preallocated(dev_priv,
> base_aligned,
> base_aligned,
> size_aligned);
> - mutex_unlock(&dev->struct_mutex);
> if (!obj)
> return false;
>
> @@ -3232,13 +3227,11 @@ intel_find_initial_plane_obj(struct intel_crtc *intel_crtc,
> intel_state->color_plane[0].stride =
> intel_fb_pitch(fb, 0, intel_state->base.rotation);
>
> - mutex_lock(&dev->struct_mutex);
> intel_state->vma =
> intel_pin_and_fence_fb_obj(fb,
> &intel_state->view,
> intel_plane_uses_fence(intel_state),
> &intel_state->flags);
> - mutex_unlock(&dev->struct_mutex);
> if (IS_ERR(intel_state->vma)) {
> DRM_ERROR("failed to pin boot fb on pipe %d: %li\n",
> intel_crtc->pipe, PTR_ERR(intel_state->vma));
> @@ -14368,8 +14361,6 @@ static void fb_obj_bump_render_priority(struct drm_i915_gem_object *obj)
> * bits. Some older platforms need special physical address handling for
> * cursor planes.
> *
> - * Must be called with struct_mutex held.
> - *
> * Returns 0 on success, negative error code on failure.
> */
> int
> @@ -14426,15 +14417,8 @@ intel_prepare_plane_fb(struct drm_plane *plane,
> if (ret)
> return ret;
>
> - ret = mutex_lock_interruptible(&dev_priv->drm.struct_mutex);
> - if (ret) {
> - i915_gem_object_unpin_pages(obj);
> - return ret;
> - }
> -
> ret = intel_plane_pin_fb(to_intel_plane_state(new_state));
>
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> i915_gem_object_unpin_pages(obj);
> if (ret)
> return ret;
> @@ -14483,8 +14467,6 @@ intel_prepare_plane_fb(struct drm_plane *plane,
> * @old_state: the state from the previous modeset
> *
> * Cleans up a framebuffer that has just been removed from a plane.
> - *
> - * Must be called with struct_mutex held.
> */
> void
> intel_cleanup_plane_fb(struct drm_plane *plane,
> @@ -14500,9 +14482,7 @@ intel_cleanup_plane_fb(struct drm_plane *plane,
> }
>
> /* Should only be called after a successful intel_prepare_plane_fb()! */
> - mutex_lock(&dev_priv->drm.struct_mutex);
> intel_plane_unpin_fb(to_intel_plane_state(old_state));
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> }
>
> int
> @@ -14705,7 +14685,6 @@ intel_legacy_cursor_update(struct drm_plane *plane,
> u32 src_w, u32 src_h,
> struct drm_modeset_acquire_ctx *ctx)
> {
> - struct drm_i915_private *dev_priv = to_i915(crtc->dev);
> struct drm_plane_state *old_plane_state, *new_plane_state;
> struct intel_plane *intel_plane = to_intel_plane(plane);
> struct intel_crtc_state *crtc_state =
> @@ -14771,13 +14750,9 @@ intel_legacy_cursor_update(struct drm_plane *plane,
> if (ret)
> goto out_free;
>
> - ret = mutex_lock_interruptible(&dev_priv->drm.struct_mutex);
> - if (ret)
> - goto out_free;
> -
> ret = intel_plane_pin_fb(to_intel_plane_state(new_plane_state));
> if (ret)
> - goto out_unlock;
> + goto out_free;
>
> intel_frontbuffer_flush(to_intel_frontbuffer(fb), ORIGIN_FLIP);
> intel_frontbuffer_track(to_intel_frontbuffer(old_plane_state->fb),
> @@ -14807,8 +14782,6 @@ intel_legacy_cursor_update(struct drm_plane *plane,
>
> intel_plane_unpin_fb(to_intel_plane_state(old_plane_state));
>
> -out_unlock:
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> out_free:
> if (new_crtc_state)
> intel_crtc_destroy_state(crtc, &new_crtc_state->base);
> diff --git a/drivers/gpu/drm/i915/display/intel_fbdev.c b/drivers/gpu/drm/i915/display/intel_fbdev.c
> index 68338669f054..97cde017670a 100644
> --- a/drivers/gpu/drm/i915/display/intel_fbdev.c
> +++ b/drivers/gpu/drm/i915/display/intel_fbdev.c
> @@ -204,7 +204,6 @@ static int intelfb_create(struct drm_fb_helper *helper,
> sizes->fb_height = intel_fb->base.height;
> }
>
> - mutex_lock(&dev->struct_mutex);
> wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm);
>
> /* Pin the GGTT vma for our access via info->screen_base.
> @@ -266,7 +265,6 @@ static int intelfb_create(struct drm_fb_helper *helper,
> ifbdev->vma_flags = flags;
>
> intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref);
> - mutex_unlock(&dev->struct_mutex);
> vga_switcheroo_client_fb_set(pdev, info);
> return 0;
>
> @@ -274,7 +272,6 @@ static int intelfb_create(struct drm_fb_helper *helper,
> intel_unpin_fb_vma(vma, flags);
> out_unlock:
> intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref);
> - mutex_unlock(&dev->struct_mutex);
> return ret;
> }
>
> @@ -291,11 +288,8 @@ static void intel_fbdev_destroy(struct intel_fbdev *ifbdev)
>
> drm_fb_helper_fini(&ifbdev->helper);
>
> - if (ifbdev->vma) {
> - mutex_lock(&ifbdev->helper.dev->struct_mutex);
> + if (ifbdev->vma)
> intel_unpin_fb_vma(ifbdev->vma, ifbdev->vma_flags);
> - mutex_unlock(&ifbdev->helper.dev->struct_mutex);
> - }
>
> if (ifbdev->fb)
> drm_framebuffer_remove(&ifbdev->fb->base);
> diff --git a/drivers/gpu/drm/i915/display/intel_overlay.c b/drivers/gpu/drm/i915/display/intel_overlay.c
> index 5efef9babadb..3f4ac1ee7668 100644
> --- a/drivers/gpu/drm/i915/display/intel_overlay.c
> +++ b/drivers/gpu/drm/i915/display/intel_overlay.c
> @@ -1303,15 +1303,11 @@ static int get_registers(struct intel_overlay *overlay, bool use_phys)
> struct i915_vma *vma;
> int err;
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> obj = i915_gem_object_create_stolen(i915, PAGE_SIZE);
> if (obj == NULL)
> obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
> - if (IS_ERR(obj)) {
> - err = PTR_ERR(obj);
> - goto err_unlock;
> - }
> + if (IS_ERR(obj))
> + return PTR_ERR(obj);
>
> vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0, PIN_MAPPABLE);
> if (IS_ERR(vma)) {
> @@ -1332,13 +1328,10 @@ 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;
> }
>
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c b/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c
> index 7f61a8024133..c1fca5728e6e 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c
> @@ -211,7 +211,7 @@ static void clear_pages_worker(struct work_struct *work)
> * keep track of the GPU activity within this vma/request, and
> * propagate the signal from the request to w->dma.
> */
> - err = i915_active_add_request(&vma->active, rq);
> + err = __i915_vma_move_to_active(vma, rq);
> if (err)
> goto out_request;
>
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
> index f7ba0935ed67..95f8e66e45db 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
> @@ -313,8 +313,6 @@ static void i915_gem_context_free(struct i915_gem_context *ctx)
> GEM_BUG_ON(!i915_gem_context_is_closed(ctx));
>
> release_hw_id(ctx);
> - if (ctx->vm)
> - i915_vm_put(ctx->vm);
>
> free_engines(rcu_access_pointer(ctx->engines));
> mutex_destroy(&ctx->engines_mutex);
> @@ -379,9 +377,13 @@ void i915_gem_context_release(struct kref *ref)
>
> static void context_close(struct i915_gem_context *ctx)
> {
> + i915_gem_context_set_closed(ctx);
> +
> + if (ctx->vm)
> + i915_vm_close(ctx->vm);
> +
> mutex_lock(&ctx->mutex);
>
> - i915_gem_context_set_closed(ctx);
> ctx->file_priv = ERR_PTR(-EBADF);
>
> /*
> @@ -474,7 +476,7 @@ __set_ppgtt(struct i915_gem_context *ctx, struct i915_address_space *vm)
>
> GEM_BUG_ON(old && i915_vm_is_4lvl(vm) != i915_vm_is_4lvl(old));
>
> - ctx->vm = i915_vm_get(vm);
> + ctx->vm = i915_vm_open(vm);
> context_apply_all(ctx, __apply_ppgtt, vm);
>
> return old;
> @@ -488,7 +490,7 @@ static void __assign_ppgtt(struct i915_gem_context *ctx,
>
> vm = __set_ppgtt(ctx, vm);
> if (vm)
> - i915_vm_put(vm);
> + i915_vm_close(vm);
> }
>
> static void __set_timeline(struct intel_timeline **dst,
> @@ -953,7 +955,7 @@ static int get_ppgtt(struct drm_i915_file_private *file_priv,
> if (ret < 0)
> goto err_unlock;
>
> - i915_vm_get(vm);
> + i915_vm_open(vm);
>
> args->size = 0;
> args->value = ret;
> @@ -973,7 +975,7 @@ static void set_ppgtt_barrier(void *data)
> if (INTEL_GEN(old->i915) < 8)
> gen6_ppgtt_unpin_all(i915_vm_to_ppgtt(old));
>
> - i915_vm_put(old);
> + i915_vm_close(old);
> }
>
> static int emit_ppgtt_update(struct i915_request *rq, void *data)
> @@ -1090,8 +1092,8 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv,
> set_ppgtt_barrier,
> old);
> if (err) {
> - i915_vm_put(__set_ppgtt(ctx, old));
> - i915_vm_put(old);
> + i915_vm_close(__set_ppgtt(ctx, old));
> + i915_vm_close(old);
> }
>
> unlock:
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_domain.c b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
> index 55c3ab59e3aa..9937b4c341f1 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_domain.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
> @@ -288,7 +288,12 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
> if (!drm_mm_node_allocated(&vma->node))
> continue;
>
> - ret = i915_vma_bind(vma, cache_level, PIN_UPDATE);
> + /* Wait for an earlier async bind, need to rewrite it */
> + ret = i915_vma_sync(vma);
> + if (ret)
> + return ret;
> +
> + ret = i915_vma_bind(vma, cache_level, PIN_UPDATE, NULL);
> if (ret)
> return ret;
> }
> @@ -391,16 +396,11 @@ int i915_gem_set_caching_ioctl(struct drm_device *dev, void *data,
> if (ret)
> goto out;
>
> - ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
> - if (ret)
> - goto out;
> -
> 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);
> @@ -485,6 +485,7 @@ static void i915_gem_object_bump_inactive_ggtt(struct drm_i915_gem_object *obj)
> if (!drm_mm_node_allocated(&vma->node))
> continue;
>
> + GEM_BUG_ON(vma->vm != &i915->ggtt.vm);
> list_move_tail(&vma->vm_link, &vma->vm->bound_list);
> }
> mutex_unlock(&i915->ggtt.vm.mutex);
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
> index c049199a1df5..88a881be12ec 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
> @@ -698,7 +698,9 @@ static int eb_reserve(struct i915_execbuffer *eb)
>
> case 1:
> /* Too fragmented, unbind everything and retry */
> + mutex_lock(&eb->context->vm->mutex);
> err = i915_gem_evict_vm(eb->context->vm);
> + mutex_unlock(&eb->context->vm->mutex);
> if (err)
> return err;
> break;
> @@ -972,7 +974,9 @@ static void reloc_cache_reset(struct reloc_cache *cache)
> ggtt->vm.clear_range(&ggtt->vm,
> cache->node.start,
> cache->node.size);
> + mutex_lock(&ggtt->vm.mutex);
> drm_mm_remove_node(&cache->node);
> + mutex_unlock(&ggtt->vm.mutex);
> } else {
> i915_vma_unpin((struct i915_vma *)cache->node.mm);
> }
> @@ -1047,11 +1051,13 @@ static void *reloc_iomap(struct drm_i915_gem_object *obj,
> PIN_NOEVICT);
> if (IS_ERR(vma)) {
> memset(&cache->node, 0, sizeof(cache->node));
> + mutex_lock(&ggtt->vm.mutex);
> err = drm_mm_insert_node_in_range
> (&ggtt->vm.mm, &cache->node,
> PAGE_SIZE, 0, I915_COLOR_UNEVICTABLE,
> 0, ggtt->mappable_end,
> DRM_MM_INSERT_LOW);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err) /* no inactive aperture space, use cpu reloc */
> return NULL;
> } else {
> @@ -1416,7 +1422,7 @@ eb_relocate_entry(struct i915_execbuffer *eb,
> if (reloc->write_domain == I915_GEM_DOMAIN_INSTRUCTION &&
> IS_GEN(eb->i915, 6)) {
> err = i915_vma_bind(target, target->obj->cache_level,
> - PIN_GLOBAL);
> + PIN_GLOBAL, NULL);
> if (WARN_ONCE(err,
> "Unexpected failure to bind target VMA!"))
> return err;
> @@ -2140,35 +2146,6 @@ static struct i915_request *eb_throttle(struct intel_context *ce)
> return i915_request_get(rq);
> }
>
> -static int
> -__eb_pin_context(struct i915_execbuffer *eb, struct intel_context *ce)
> -{
> - int err;
> -
> - if (likely(atomic_inc_not_zero(&ce->pin_count)))
> - return 0;
> -
> - err = mutex_lock_interruptible(&eb->i915->drm.struct_mutex);
> - if (err)
> - return err;
> -
> - err = __intel_context_do_pin(ce);
> - mutex_unlock(&eb->i915->drm.struct_mutex);
> -
> - return err;
> -}
> -
> -static void
> -__eb_unpin_context(struct i915_execbuffer *eb, struct intel_context *ce)
> -{
> - if (likely(atomic_add_unless(&ce->pin_count, -1, 1)))
> - return;
> -
> - mutex_lock(&eb->i915->drm.struct_mutex);
> - intel_context_unpin(ce);
> - mutex_unlock(&eb->i915->drm.struct_mutex);
> -}
> -
> static int __eb_pin_engine(struct i915_execbuffer *eb, struct intel_context *ce)
> {
> struct intel_timeline *tl;
> @@ -2188,7 +2165,7 @@ static int __eb_pin_engine(struct i915_execbuffer *eb, struct intel_context *ce)
> * GGTT space, so do this first before we reserve a seqno for
> * ourselves.
> */
> - err = __eb_pin_context(eb, ce);
> + err = intel_context_pin(ce);
> if (err)
> return err;
>
> @@ -2232,7 +2209,7 @@ static int __eb_pin_engine(struct i915_execbuffer *eb, struct intel_context *ce)
> intel_context_exit(ce);
> intel_context_timeline_unlock(tl);
> err_unpin:
> - __eb_unpin_context(eb, ce);
> + intel_context_unpin(ce);
> return err;
> }
>
> @@ -2245,7 +2222,7 @@ static void eb_unpin_engine(struct i915_execbuffer *eb)
> intel_context_exit(ce);
> mutex_unlock(&tl->mutex);
>
> - __eb_unpin_context(eb, ce);
> + intel_context_unpin(ce);
> }
>
> static unsigned int
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
> index 860b751c51f1..c44614895487 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_mman.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
> @@ -249,16 +249,6 @@ vm_fault_t i915_gem_fault(struct vm_fault *vmf)
> if (ret)
> 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 |
> @@ -291,7 +281,13 @@ vm_fault_t i915_gem_fault(struct vm_fault *vmf)
> }
> if (IS_ERR(vma)) {
> ret = PTR_ERR(vma);
> - goto err_unlock;
> + 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_unpin;
> }
>
> ret = i915_vma_pin_fence(vma);
> @@ -329,8 +325,6 @@ vm_fault_t i915_gem_fault(struct vm_fault *vmf)
> i915_vma_unpin_fence(vma);
> err_unpin:
> __i915_vma_unpin(vma);
> -err_unlock:
> - mutex_unlock(&dev->struct_mutex);
> err_reset:
> intel_gt_reset_unlock(ggtt->vm.gt, srcu);
> err_rpm:
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
> index 0ef60dae23a7..dbf9be9a79f4 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_object.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
> @@ -155,21 +155,30 @@ static void __i915_gem_free_objects(struct drm_i915_private *i915,
>
> wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> 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);
> -
> - list_for_each_entry_safe(vma, vn, &obj->vma.list, obj_link) {
> - GEM_BUG_ON(i915_vma_is_active(vma));
> - atomic_and(~I915_VMA_PIN_MASK, &vma->flags);
> - i915_vma_destroy(vma);
> + if (!list_empty(&obj->vma.list)) {
> + struct i915_vma *vma;
> +
> + /*
> + * Note that the vma keeps an object reference while
> + * it is active, so it *should* not sleep while we
> + * destroy it. Our debug code errs insits it *might*.
> + * For the moment, play along.
> + */
> + spin_lock(&obj->vma.lock);
> + while ((vma = list_first_entry_or_null(&obj->vma.list,
> + struct i915_vma,
> + obj_link))) {
> + GEM_BUG_ON(vma->obj != obj);
> + spin_unlock(&obj->vma.lock);
> +
> + i915_vma_destroy(vma);
> +
> + spin_lock(&obj->vma.lock);
> + }
> + spin_unlock(&obj->vma.lock);
> }
> - GEM_BUG_ON(!list_empty(&obj->vma.list));
> - GEM_BUG_ON(!RB_EMPTY_ROOT(&obj->vma.tree));
> -
> - mutex_unlock(&i915->drm.struct_mutex);
>
> GEM_BUG_ON(atomic_read(&obj->bind_count));
> GEM_BUG_ON(obj->userfault_count);
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
> index 29b9eddc4c7f..a78af25dce36 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
> @@ -106,6 +106,11 @@ static inline void i915_gem_object_lock(struct drm_i915_gem_object *obj)
> dma_resv_lock(obj->base.resv, NULL);
> }
>
> +static inline bool i915_gem_object_trylock(struct drm_i915_gem_object *obj)
> +{
> + return dma_resv_trylock(obj->base.resv);
> +}
> +
> static inline int
> i915_gem_object_lock_interruptible(struct drm_i915_gem_object *obj)
> {
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c b/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
> index d2c05d752909..fd3ce6da8497 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
> @@ -16,40 +16,6 @@
>
> #include "i915_trace.h"
>
> -static bool shrinker_lock(struct drm_i915_private *i915,
> - unsigned int flags,
> - bool *unlock)
> -{
> - struct mutex *m = &i915->drm.struct_mutex;
> -
> - switch (mutex_trylock_recursive(m)) {
> - case MUTEX_TRYLOCK_RECURSIVE:
> - *unlock = false;
> - return true;
> -
> - case MUTEX_TRYLOCK_FAILED:
> - *unlock = false;
> - if (flags & I915_SHRINK_ACTIVE &&
> - mutex_lock_killable_nested(m, I915_MM_SHRINKER) == 0)
> - *unlock = true;
> - return *unlock;
> -
> - case MUTEX_TRYLOCK_SUCCESS:
> - *unlock = true;
> - return true;
> - }
> -
> - BUG();
> -}
> -
> -static void shrinker_unlock(struct drm_i915_private *i915, bool unlock)
> -{
> - if (!unlock)
> - return;
> -
> - mutex_unlock(&i915->drm.struct_mutex);
> -}
> -
> static bool swap_available(void)
> {
> return get_nr_swap_pages() > 0;
> @@ -155,10 +121,6 @@ i915_gem_shrink(struct drm_i915_private *i915,
> intel_wakeref_t wakeref = 0;
> unsigned long count = 0;
> unsigned long scanned = 0;
> - bool unlock;
> -
> - if (!shrinker_lock(i915, shrink, &unlock))
> - return 0;
>
> /*
> * When shrinking the active list, we should also consider active
> @@ -268,8 +230,6 @@ i915_gem_shrink(struct drm_i915_private *i915,
> if (shrink & I915_SHRINK_BOUND)
> intel_runtime_pm_put(&i915->runtime_pm, wakeref);
>
> - shrinker_unlock(i915, unlock);
> -
> if (nr_scanned)
> *nr_scanned += scanned;
> return count;
> @@ -339,19 +299,14 @@ i915_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc)
> struct drm_i915_private *i915 =
> container_of(shrinker, struct drm_i915_private, mm.shrinker);
> unsigned long freed;
> - bool unlock;
>
> sc->nr_scanned = 0;
>
> - if (!shrinker_lock(i915, 0, &unlock))
> - return SHRINK_STOP;
> -
> freed = i915_gem_shrink(i915,
> sc->nr_to_scan,
> &sc->nr_scanned,
> I915_SHRINK_BOUND |
> - I915_SHRINK_UNBOUND |
> - I915_SHRINK_WRITEBACK);
> + I915_SHRINK_UNBOUND);
> if (sc->nr_scanned < sc->nr_to_scan && current_is_kswapd()) {
> intel_wakeref_t wakeref;
>
> @@ -366,8 +321,6 @@ i915_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc)
> }
> }
>
> - shrinker_unlock(i915, unlock);
> -
> return sc->nr_scanned ? freed : SHRINK_STOP;
> }
>
> @@ -384,6 +337,7 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr)
> freed_pages = 0;
> with_intel_runtime_pm(&i915->runtime_pm, wakeref)
> freed_pages += i915_gem_shrink(i915, -1UL, NULL,
> + I915_SHRINK_ACTIVE |
> I915_SHRINK_BOUND |
> I915_SHRINK_UNBOUND |
> I915_SHRINK_WRITEBACK);
> @@ -419,10 +373,6 @@ i915_gem_shrinker_vmap(struct notifier_block *nb, unsigned long event, void *ptr
> struct i915_vma *vma, *next;
> unsigned long freed_pages = 0;
> intel_wakeref_t wakeref;
> - bool unlock;
> -
> - if (!shrinker_lock(i915, 0, &unlock))
> - return NOTIFY_DONE;
>
> with_intel_runtime_pm(&i915->runtime_pm, wakeref)
> freed_pages += i915_gem_shrink(i915, -1UL, NULL,
> @@ -439,15 +389,11 @@ i915_gem_shrinker_vmap(struct notifier_block *nb, unsigned long event, void *ptr
> if (!vma->iomap || i915_vma_is_active(vma))
> continue;
>
> - mutex_unlock(&i915->ggtt.vm.mutex);
> - if (i915_vma_unbind(vma) == 0)
> + if (__i915_vma_unbind(vma) == 0)
> freed_pages += count;
> - mutex_lock(&i915->ggtt.vm.mutex);
> }
> mutex_unlock(&i915->ggtt.vm.mutex);
>
> - shrinker_unlock(i915, unlock);
> -
> *(unsigned long *)ptr += freed_pages;
> return NOTIFY_DONE;
> }
> @@ -490,22 +436,9 @@ void i915_gem_shrinker_taints_mutex(struct drm_i915_private *i915,
>
> fs_reclaim_acquire(GFP_KERNEL);
>
> - /*
> - * As we invariably rely on the struct_mutex within the shrinker,
> - * but have a complicated recursion dance, taint all the mutexes used
> - * within the shrinker with the struct_mutex. For completeness, we
> - * taint with all subclass of struct_mutex, even though we should
> - * only need tainting by I915_MM_NORMAL to catch possible ABBA
> - * deadlocks from using struct_mutex inside @mutex.
> - */
> - mutex_acquire(&i915->drm.struct_mutex.dep_map,
> - I915_MM_SHRINKER, 0, _RET_IP_);
> -
> mutex_acquire(&mutex->dep_map, 0, 0, _RET_IP_);
> mutex_release(&mutex->dep_map, 0, _RET_IP_);
>
> - mutex_release(&i915->drm.struct_mutex.dep_map, 0, _RET_IP_);
> -
> fs_reclaim_release(GFP_KERNEL);
>
> if (unlock)
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
> index e45eb8721850..fad98a921cde 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
> @@ -624,8 +624,6 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_i915_private *dev_priv
> if (!drm_mm_initialized(&dev_priv->mm.stolen))
> return NULL;
>
> - lockdep_assert_held(&dev_priv->drm.struct_mutex);
> -
> DRM_DEBUG_DRIVER("creating preallocated stolen object: stolen_offset=%pa, gtt_offset=%pa, size=%pa\n",
> &stolen_offset, >t_offset, &size);
>
> @@ -677,21 +675,25 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_i915_private *dev_priv
> * setting up the GTT space. The actual reservation will occur
> * later.
> */
> + mutex_lock(&ggtt->vm.mutex);
> ret = i915_gem_gtt_reserve(&ggtt->vm, &vma->node,
> size, gtt_offset, obj->cache_level,
> 0);
> if (ret) {
> DRM_DEBUG_DRIVER("failed to allocate stolen GTT space\n");
> + mutex_unlock(&ggtt->vm.mutex);
> goto err_pages;
> }
>
> GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
>
> + GEM_BUG_ON(vma->pages);
> vma->pages = obj->mm.pages;
> + atomic_set(&vma->pages_count, I915_VMA_PAGES_ACTIVE);
> +
> set_bit(I915_VMA_GLOBAL_BIND_BIT, __i915_vma_flags(vma));
> __i915_vma_set_map_and_fenceable(vma);
>
> - mutex_lock(&ggtt->vm.mutex);
> list_add_tail(&vma->vm_link, &ggtt->vm.bound_list);
> mutex_unlock(&ggtt->vm.mutex);
>
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_tiling.c b/drivers/gpu/drm/i915/gem/i915_gem_tiling.c
> index e5d1ae8d4dba..dc2a83ce44d5 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_tiling.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_tiling.c
> @@ -181,22 +181,25 @@ static int
> i915_gem_object_fence_prepare(struct drm_i915_gem_object *obj,
> int tiling_mode, unsigned int stride)
> {
> + struct i915_ggtt *ggtt = &to_i915(obj->base.dev)->ggtt;
> struct i915_vma *vma;
> - int ret;
> + int ret = 0;
>
> if (tiling_mode == I915_TILING_NONE)
> return 0;
>
> + mutex_lock(&ggtt->vm.mutex);
> for_each_ggtt_vma(vma, obj) {
> if (i915_vma_fence_prepare(vma, tiling_mode, stride))
> continue;
>
> - ret = i915_vma_unbind(vma);
> + ret = __i915_vma_unbind(vma);
> if (ret)
> - return ret;
> + break;
> }
> + mutex_unlock(&ggtt->vm.mutex);
>
> - return 0;
> + return ret;
> }
>
> int
> @@ -212,7 +215,6 @@ i915_gem_object_set_tiling(struct drm_i915_gem_object *obj,
>
> GEM_BUG_ON(!i915_tiling_ok(obj, tiling, stride));
> GEM_BUG_ON(!stride ^ (tiling == I915_TILING_NONE));
> - lockdep_assert_held(&i915->drm.struct_mutex);
>
> if ((tiling | stride) == obj->tiling_and_stride)
> return 0;
> @@ -233,16 +235,18 @@ i915_gem_object_set_tiling(struct drm_i915_gem_object *obj,
> * whilst executing a fenced command for an untiled object.
> */
>
> - err = i915_gem_object_fence_prepare(obj, tiling, stride);
> - if (err)
> - return err;
> -
> i915_gem_object_lock(obj);
> if (i915_gem_object_is_framebuffer(obj)) {
> i915_gem_object_unlock(obj);
> return -EBUSY;
> }
>
> + err = i915_gem_object_fence_prepare(obj, tiling, stride);
> + if (err) {
> + i915_gem_object_unlock(obj);
> + return err;
> + }
> +
> /* If the memory has unknown (i.e. varying) swizzling, we pin the
> * pages to prevent them being swapped out and causing corruption
> * due to the change in swizzling.
> @@ -368,12 +372,7 @@ i915_gem_set_tiling_ioctl(struct drm_device *dev, void *data,
> }
> }
>
> - err = mutex_lock_interruptible(&dev->struct_mutex);
> - if (err)
> - goto err;
> -
> err = i915_gem_object_set_tiling(obj, args->tiling_mode, args->stride);
> - mutex_unlock(&dev->struct_mutex);
>
> /* We have to maintain this existing ABI... */
> args->stride = i915_gem_object_get_stride(obj);
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
> index 11b231c187c5..27573ef903b4 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
> @@ -92,7 +92,6 @@ userptr_mn_invalidate_range_start(struct mmu_notifier *_mn,
> struct i915_mmu_notifier *mn =
> container_of(_mn, struct i915_mmu_notifier, mn);
> struct interval_tree_node *it;
> - struct mutex *unlock = NULL;
> unsigned long end;
> int ret = 0;
>
> @@ -129,33 +128,13 @@ userptr_mn_invalidate_range_start(struct mmu_notifier *_mn,
> }
> spin_unlock(&mn->lock);
>
> - if (!unlock) {
> - unlock = &mn->mm->i915->drm.struct_mutex;
> -
> - switch (mutex_trylock_recursive(unlock)) {
> - default:
> - case MUTEX_TRYLOCK_FAILED:
> - if (mutex_lock_killable_nested(unlock, I915_MM_SHRINKER)) {
> - i915_gem_object_put(obj);
> - return -EINTR;
> - }
> - /* fall through */
> - case MUTEX_TRYLOCK_SUCCESS:
> - break;
> -
> - case MUTEX_TRYLOCK_RECURSIVE:
> - unlock = ERR_PTR(-EEXIST);
> - break;
> - }
> - }
> -
> ret = i915_gem_object_unbind(obj,
> I915_GEM_OBJECT_UNBIND_ACTIVE);
> if (ret == 0)
> ret = __i915_gem_object_put_pages(obj, I915_MM_SHRINKER);
> i915_gem_object_put(obj);
> if (ret)
> - goto unlock;
> + return ret;
>
> spin_lock(&mn->lock);
>
> @@ -168,10 +147,6 @@ userptr_mn_invalidate_range_start(struct mmu_notifier *_mn,
> }
> spin_unlock(&mn->lock);
>
> -unlock:
> - if (!IS_ERR_OR_NULL(unlock))
> - mutex_unlock(unlock);
> -
> return ret;
>
> }
> diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
> index c5cea4379216..98b2a6ccfcc1 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
> @@ -333,7 +333,12 @@ static int igt_check_page_sizes(struct i915_vma *vma)
> struct drm_i915_private *i915 = vma->vm->i915;
> unsigned int supported = INTEL_INFO(i915)->page_sizes;
> struct drm_i915_gem_object *obj = vma->obj;
> - int err = 0;
> + int err;
> +
> + /* We have to wait for the async bind to complete before our asserts */
> + err = i915_vma_sync(vma);
> + if (err)
> + return err;
>
> if (!HAS_PAGE_SIZES(i915, vma->page_sizes.sg)) {
> pr_err("unsupported page_sizes.sg=%u, supported=%u\n",
> @@ -1390,7 +1395,7 @@ static int igt_ppgtt_pin_update(void *arg)
> goto out_unpin;
> }
>
> - err = i915_vma_bind(vma, I915_CACHE_NONE, PIN_UPDATE);
> + err = i915_vma_bind(vma, I915_CACHE_NONE, PIN_UPDATE, NULL);
> if (err)
> goto out_unpin;
>
> 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 dc25bcc3e372..6ef3d745dc8d 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
> @@ -747,10 +747,7 @@ emit_rpcs_query(struct drm_i915_gem_object *obj,
> if (err)
> goto skip_request;
>
> - i915_vma_unpin(batch);
> - i915_vma_close(batch);
> - i915_vma_put(batch);
> -
> + i915_vma_unpin_and_release(&batch, 0);
> i915_vma_unpin(vma);
>
> *rq_out = i915_request_get(rq);
> @@ -764,8 +761,7 @@ emit_rpcs_query(struct drm_i915_gem_object *obj,
> err_request:
> i915_request_add(rq);
> err_batch:
> - i915_vma_unpin(batch);
> - i915_vma_put(batch);
> + i915_vma_unpin_and_release(&batch, 0);
> err_vma:
> i915_vma_unpin(vma);
>
> @@ -1309,9 +1305,7 @@ static int write_to_scratch(struct i915_gem_context *ctx,
> if (err)
> goto skip_request;
>
> - i915_vma_unpin(vma);
> - i915_vma_close(vma);
> - i915_vma_put(vma);
> + i915_vma_unpin_and_release(&vma, 0);
>
> 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 aefe557527f8..36aca1c172e7 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
> @@ -322,7 +322,6 @@ static int igt_partial_tiling(void *arg)
> goto out;
> }
>
> - mutex_lock(&i915->drm.struct_mutex);
> wakeref = intel_runtime_pm_get(&i915->runtime_pm);
>
> if (1) {
> @@ -415,7 +414,6 @@ next_tiling: ;
>
> out_unlock:
> intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> i915_gem_object_unpin_pages(obj);
> out:
> i915_gem_object_put(obj);
> @@ -458,7 +456,6 @@ static int igt_smoke_tiling(void *arg)
> goto out;
> }
>
> - mutex_lock(&i915->drm.struct_mutex);
> wakeref = intel_runtime_pm_get(&i915->runtime_pm);
>
> count = 0;
> @@ -508,7 +505,6 @@ static int igt_smoke_tiling(void *arg)
> pr_info("%s: Completed %lu trials\n", __func__, count);
>
> intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> i915_gem_object_unpin_pages(obj);
> out:
> i915_gem_object_put(obj);
> diff --git a/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c b/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c
> index ee5dc13a30b3..6718da20f35d 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/igt_gem_utils.c
> @@ -154,9 +154,7 @@ int igt_gpu_fill_dw(struct intel_context *ce,
>
> i915_request_add(rq);
>
> - i915_vma_unpin(batch);
> - i915_vma_close(batch);
> - i915_vma_put(batch);
> + i915_vma_unpin_and_release(&batch, 0);
>
> return 0;
>
> @@ -165,7 +163,6 @@ int igt_gpu_fill_dw(struct intel_context *ce,
> err_request:
> i915_request_add(rq);
> err_batch:
> - i915_vma_unpin(batch);
> - i915_vma_put(batch);
> + i915_vma_unpin_and_release(&batch, 0);
> return err;
> }
> diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c b/drivers/gpu/drm/i915/gt/intel_gt.c
> index eef9bdae8ebb..95e67db52668 100644
> --- a/drivers/gpu/drm/i915/gt/intel_gt.c
> +++ b/drivers/gpu/drm/i915/gt/intel_gt.c
> @@ -301,11 +301,12 @@ void intel_gt_flush_ggtt_writes(struct intel_gt *gt)
>
> with_intel_runtime_pm(&i915->runtime_pm, wakeref) {
> struct intel_uncore *uncore = gt->uncore;
> + unsigned long flags;
>
> - spin_lock_irq(&uncore->lock);
> + spin_lock_irqsave(&uncore->lock, flags);
> intel_uncore_posting_read_fw(uncore,
> RING_HEAD(RENDER_RING_BASE));
> - spin_unlock_irq(&uncore->lock);
> + spin_unlock_irqrestore(&uncore->lock, flags);
> }
> }
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
> index 0747b8c9f768..ec32996254c0 100644
> --- a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
> +++ b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
> @@ -1338,15 +1338,13 @@ void intel_ring_free(struct kref *ref)
> {
> struct intel_ring *ring = container_of(ref, typeof(*ring), ref);
>
> - i915_vma_close(ring->vma);
> i915_vma_put(ring->vma);
> -
> kfree(ring);
> }
>
> static void __ring_context_fini(struct intel_context *ce)
> {
> - i915_gem_object_put(ce->state->obj);
> + i915_vma_put(ce->state);
> }
>
> static void ring_context_destroy(struct kref *ref)
> diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
> index a0098fc35921..e53eea1050f8 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
> @@ -1127,15 +1127,14 @@ static int evict_vma(void *data)
> {
> struct evict_vma *arg = data;
> struct i915_address_space *vm = arg->vma->vm;
> - struct drm_i915_private *i915 = vm->i915;
> struct drm_mm_node evict = arg->vma->node;
> int err;
>
> complete(&arg->completion);
>
> - mutex_lock(&i915->drm.struct_mutex);
> + mutex_lock(&vm->mutex);
> err = i915_gem_evict_for_node(vm, &evict, 0);
> - mutex_unlock(&i915->drm.struct_mutex);
> + mutex_unlock(&vm->mutex);
>
> return err;
> }
> @@ -1143,39 +1142,33 @@ static int evict_vma(void *data)
> static int evict_fence(void *data)
> {
> struct evict_vma *arg = data;
> - struct drm_i915_private *i915 = arg->vma->vm->i915;
> int err;
>
> complete(&arg->completion);
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> /* Mark the fence register as dirty to force the mmio update. */
> err = i915_gem_object_set_tiling(arg->vma->obj, I915_TILING_Y, 512);
> if (err) {
> pr_err("Invalid Y-tiling settings; err:%d\n", err);
> - goto out_unlock;
> + return err;
> }
>
> err = i915_vma_pin(arg->vma, 0, 0, PIN_GLOBAL | PIN_MAPPABLE);
> if (err) {
> pr_err("Unable to pin vma for Y-tiled fence; err:%d\n", err);
> - goto out_unlock;
> + return err;
> }
>
> err = i915_vma_pin_fence(arg->vma);
> i915_vma_unpin(arg->vma);
> if (err) {
> pr_err("Unable to pin Y-tiled fence; err:%d\n", err);
> - goto out_unlock;
> + return err;
> }
>
> i915_vma_unpin_fence(arg->vma);
>
> -out_unlock:
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> - return err;
> + return 0;
> }
>
> static int __igt_reset_evict_vma(struct intel_gt *gt,
> diff --git a/drivers/gpu/drm/i915/gvt/aperture_gm.c b/drivers/gpu/drm/i915/gvt/aperture_gm.c
> index 5ff2437b2998..d996bbc7ea59 100644
> --- a/drivers/gpu/drm/i915/gvt/aperture_gm.c
> +++ b/drivers/gpu/drm/i915/gvt/aperture_gm.c
> @@ -61,14 +61,14 @@ static int alloc_gm(struct intel_vgpu *vgpu, bool high_gm)
> flags = PIN_MAPPABLE;
> }
>
> - mutex_lock(&dev_priv->drm.struct_mutex);
> + mutex_lock(&dev_priv->ggtt.vm.mutex);
> mmio_hw_access_pre(dev_priv);
> ret = i915_gem_gtt_insert(&dev_priv->ggtt.vm, node,
> size, I915_GTT_PAGE_SIZE,
> I915_COLOR_UNEVICTABLE,
> start, end, flags);
> mmio_hw_access_post(dev_priv);
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> + mutex_unlock(&dev_priv->ggtt.vm.mutex);
> if (ret)
> gvt_err("fail to alloc %s gm space from host\n",
> high_gm ? "high" : "low");
> @@ -98,9 +98,9 @@ static int alloc_vgpu_gm(struct intel_vgpu *vgpu)
>
> return 0;
> out_free_aperture:
> - mutex_lock(&dev_priv->drm.struct_mutex);
> + mutex_lock(&dev_priv->ggtt.vm.mutex);
> drm_mm_remove_node(&vgpu->gm.low_gm_node);
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> + mutex_unlock(&dev_priv->ggtt.vm.mutex);
> return ret;
> }
>
> @@ -108,10 +108,10 @@ static void free_vgpu_gm(struct intel_vgpu *vgpu)
> {
> struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
>
> - mutex_lock(&dev_priv->drm.struct_mutex);
> + mutex_lock(&dev_priv->ggtt.vm.mutex);
> drm_mm_remove_node(&vgpu->gm.low_gm_node);
> drm_mm_remove_node(&vgpu->gm.high_gm_node);
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> + mutex_unlock(&dev_priv->ggtt.vm.mutex);
> }
>
> /**
> diff --git a/drivers/gpu/drm/i915/i915_active.c b/drivers/gpu/drm/i915/i915_active.c
> index d5aac6ff803a..0791736a08fd 100644
> --- a/drivers/gpu/drm/i915/i915_active.c
> +++ b/drivers/gpu/drm/i915/i915_active.c
> @@ -146,6 +146,7 @@ __active_retire(struct i915_active *ref)
> if (!retire)
> return;
>
> + GEM_BUG_ON(rcu_access_pointer(ref->excl));
> rbtree_postorder_for_each_entry_safe(it, n, &root, node) {
> GEM_BUG_ON(i915_active_request_isset(&it->base));
> kmem_cache_free(global.slab_cache, it);
> @@ -245,6 +246,8 @@ void __i915_active_init(struct drm_i915_private *i915,
> ref->flags = 0;
> ref->active = active;
> ref->retire = retire;
> +
> + ref->excl = NULL;
> ref->tree = RB_ROOT;
> ref->cache = NULL;
> init_llist_head(&ref->preallocated_barriers);
> @@ -341,6 +344,46 @@ int i915_active_ref(struct i915_active *ref,
> return err;
> }
>
> +static void excl_cb(struct dma_fence *f, struct dma_fence_cb *cb)
> +{
> + struct i915_active *ref = container_of(cb, typeof(*ref), excl_cb);
> +
> + RCU_INIT_POINTER(ref->excl, NULL);
> + dma_fence_put(f);
> +
> + active_retire(ref);
> +}
> +
> +void i915_active_set_exclusive(struct i915_active *ref, struct dma_fence *f)
> +{
> + /* We expect the caller to manage the exclusive timeline ordering */
> + GEM_BUG_ON(i915_active_is_idle(ref));
> +
> + dma_fence_get(f);
> +
> + rcu_read_lock();
> + if (rcu_access_pointer(ref->excl)) {
> + struct dma_fence *old;
> +
> + old = dma_fence_get_rcu_safe(&ref->excl);
> + if (old) {
> + if (dma_fence_remove_callback(old, &ref->excl_cb))
> + atomic_dec(&ref->count);
> + dma_fence_put(old);
> + }
> + }
> + rcu_read_unlock();
> +
> + atomic_inc(&ref->count);
> + rcu_assign_pointer(ref->excl, f);
> +
> + if (dma_fence_add_callback(f, &ref->excl_cb, excl_cb)) {
> + RCU_INIT_POINTER(ref->excl, NULL);
> + atomic_dec(&ref->count);
> + dma_fence_put(f);
> + }
> +}
> +
> int i915_active_acquire(struct i915_active *ref)
> {
> int err;
> @@ -399,6 +442,25 @@ void i915_active_ungrab(struct i915_active *ref)
> __active_ungrab(ref);
> }
>
> +static int excl_wait(struct i915_active *ref)
> +{
> + struct dma_fence *old;
> + int err = 0;
> +
> + if (!rcu_access_pointer(ref->excl))
> + return 0;
> +
> + rcu_read_lock();
> + old = dma_fence_get_rcu_safe(&ref->excl);
> + rcu_read_unlock();
> + if (old) {
> + err = dma_fence_wait(old, true);
> + dma_fence_put(old);
> + }
> +
> + return err;
> +}
> +
> int i915_active_wait(struct i915_active *ref)
> {
> struct active_node *it, *n;
> @@ -419,6 +481,10 @@ int i915_active_wait(struct i915_active *ref)
> return 0;
> }
>
> + err = excl_wait(ref);
> + if (err)
> + goto out;
> +
> rbtree_postorder_for_each_entry_safe(it, n, &ref->tree, node) {
> if (is_barrier(&it->base)) { /* unconnected idle-barrier */
> err = -EBUSY;
> @@ -430,6 +496,7 @@ int i915_active_wait(struct i915_active *ref)
> break;
> }
>
> +out:
> __active_retire(ref);
> if (err)
> return err;
> @@ -454,26 +521,22 @@ int i915_request_await_active_request(struct i915_request *rq,
>
> int i915_request_await_active(struct i915_request *rq, struct i915_active *ref)
> {
> - struct active_node *it, *n;
> - int err;
> -
> - if (RB_EMPTY_ROOT(&ref->tree))
> - return 0;
> + int err = 0;
>
> - /* await allocates and so we need to avoid hitting the shrinker */
> - err = i915_active_acquire(ref);
> - if (err)
> - return err;
> + if (rcu_access_pointer(ref->excl)) {
> + struct dma_fence *fence;
>
> - mutex_lock(&ref->mutex);
> - rbtree_postorder_for_each_entry_safe(it, n, &ref->tree, node) {
> - err = i915_request_await_active_request(rq, &it->base);
> - if (err)
> - break;
> + rcu_read_lock();
> + fence = dma_fence_get_rcu_safe(&ref->excl);
> + rcu_read_unlock();
> + if (fence) {
> + err = i915_request_await_dma_fence(rq, fence);
> + dma_fence_put(fence);
> + }
> }
> - mutex_unlock(&ref->mutex);
>
> - i915_active_release(ref);
> + /* In the future we may choose to await on all fences */
> +
> return err;
> }
>
> diff --git a/drivers/gpu/drm/i915/i915_active.h b/drivers/gpu/drm/i915/i915_active.h
> index 949c6835335b..90034f61b7c2 100644
> --- a/drivers/gpu/drm/i915/i915_active.h
> +++ b/drivers/gpu/drm/i915/i915_active.h
> @@ -379,6 +379,13 @@ i915_active_add_request(struct i915_active *ref, struct i915_request *rq)
> return i915_active_ref(ref, i915_request_timeline(rq), rq);
> }
>
> +void i915_active_set_exclusive(struct i915_active *ref, struct dma_fence *f);
> +
> +static inline bool i915_active_has_exclusive(struct i915_active *ref)
> +{
> + return rcu_access_pointer(ref->excl);
> +}
> +
> int i915_active_wait(struct i915_active *ref);
>
> int i915_request_await_active(struct i915_request *rq,
> diff --git a/drivers/gpu/drm/i915/i915_active_types.h b/drivers/gpu/drm/i915/i915_active_types.h
> index 1854e7d168c1..86e7a232ea3c 100644
> --- a/drivers/gpu/drm/i915/i915_active_types.h
> +++ b/drivers/gpu/drm/i915/i915_active_types.h
> @@ -8,6 +8,7 @@
> #define _I915_ACTIVE_TYPES_H_
>
> #include <linux/atomic.h>
> +#include <linux/dma-fence.h>
> #include <linux/llist.h>
> #include <linux/mutex.h>
> #include <linux/rbtree.h>
> @@ -51,6 +52,10 @@ struct i915_active {
> struct mutex mutex;
> atomic_t count;
>
> + /* Preallocated "exclusive" node */
> + struct dma_fence __rcu *excl;
> + struct dma_fence_cb excl_cb;
> +
> unsigned long flags;
> #define I915_ACTIVE_GRAB_BIT 0
>
> diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
> index a9ee73b61f4d..ab00ba6f695b 100644
> --- a/drivers/gpu/drm/i915/i915_drv.c
> +++ b/drivers/gpu/drm/i915/i915_drv.c
> @@ -1881,10 +1881,8 @@ static int i915_drm_resume(struct drm_device *dev)
> if (ret)
> DRM_ERROR("failed to re-enable GGTT\n");
>
> - mutex_lock(&dev_priv->drm.struct_mutex);
> i915_gem_restore_gtt_mappings(dev_priv);
> i915_gem_restore_fences(dev_priv);
> - mutex_unlock(&dev_priv->drm.struct_mutex);
>
> intel_csr_ucode_resume(dev_priv);
>
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index b8f7fe311b0a..934ea776a46a 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -62,20 +62,31 @@
> #include "intel_pm.h"
>
> static int
> -insert_mappable_node(struct i915_ggtt *ggtt,
> - struct drm_mm_node *node, u32 size)
> +insert_mappable_node(struct i915_ggtt *ggtt, struct drm_mm_node *node, u32 size)
> {
> + int err;
> +
> + err = mutex_lock_interruptible(&ggtt->vm.mutex);
> + if (err)
> + return err;
> +
> memset(node, 0, sizeof(*node));
> - return drm_mm_insert_node_in_range(&ggtt->vm.mm, node,
> - size, 0, I915_COLOR_UNEVICTABLE,
> - 0, ggtt->mappable_end,
> - DRM_MM_INSERT_LOW);
> + err = drm_mm_insert_node_in_range(&ggtt->vm.mm, node,
> + size, 0, I915_COLOR_UNEVICTABLE,
> + 0, ggtt->mappable_end,
> + DRM_MM_INSERT_LOW);
> +
> + mutex_unlock(&ggtt->vm.mutex);
> +
> + return err;
> }
>
> static void
> -remove_mappable_node(struct drm_mm_node *node)
> +remove_mappable_node(struct i915_ggtt *ggtt, struct drm_mm_node *node)
> {
> + mutex_lock(&ggtt->vm.mutex);
> drm_mm_remove_node(node);
> + mutex_unlock(&ggtt->vm.mutex);
> }
>
> int
> @@ -87,7 +98,8 @@ i915_gem_get_aperture_ioctl(struct drm_device *dev, void *data,
> struct i915_vma *vma;
> u64 pinned;
>
> - mutex_lock(&ggtt->vm.mutex);
> + if (mutex_lock_interruptible(&ggtt->vm.mutex))
> + return -EINTR;
>
> pinned = ggtt->vm.reserved;
> list_for_each_entry(vma, &ggtt->vm.bound_list, vm_link)
> @@ -109,20 +121,24 @@ int i915_gem_object_unbind(struct drm_i915_gem_object *obj,
> LIST_HEAD(still_in_list);
> int ret = 0;
>
> - lockdep_assert_held(&obj->base.dev->struct_mutex);
> -
> spin_lock(&obj->vma.lock);
> while (!ret && (vma = list_first_entry_or_null(&obj->vma.list,
> struct i915_vma,
> obj_link))) {
> + struct i915_address_space *vm = vma->vm;
> +
> + ret = -EBUSY;
> + if (!i915_vm_tryopen(vm))
> + break;
> +
> list_move_tail(&vma->obj_link, &still_in_list);
> spin_unlock(&obj->vma.lock);
>
> - ret = -EBUSY;
> if (flags & I915_GEM_OBJECT_UNBIND_ACTIVE ||
> !i915_vma_is_active(vma))
> ret = i915_vma_unbind(vma);
>
> + i915_vm_close(vm);
> spin_lock(&obj->vma.lock);
> }
> list_splice(&still_in_list, &obj->vma.list);
> @@ -338,10 +354,6 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
> u64 remain, offset;
> int ret;
>
> - ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
> - if (ret)
> - return ret;
> -
> wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> vma = ERR_PTR(-ENODEV);
> if (!i915_gem_object_is_tiled(obj))
> @@ -355,12 +367,10 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
> } else {
> ret = insert_mappable_node(ggtt, &node, PAGE_SIZE);
> if (ret)
> - goto out_unlock;
> + goto out_rpm;
> GEM_BUG_ON(!drm_mm_node_allocated(&node));
> }
>
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> ret = i915_gem_object_lock_interruptible(obj);
> if (ret)
> goto out_unpin;
> @@ -414,17 +424,14 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
>
> i915_gem_object_unlock_fence(obj, fence);
> out_unpin:
> - mutex_lock(&i915->drm.struct_mutex);
> if (drm_mm_node_allocated(&node)) {
> ggtt->vm.clear_range(&ggtt->vm, node.start, node.size);
> - remove_mappable_node(&node);
> + remove_mappable_node(ggtt, &node);
> } else {
> i915_vma_unpin(vma);
> }
> -out_unlock:
> +out_rpm:
> intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> return ret;
> }
>
> @@ -531,10 +538,6 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
> void __user *user_data;
> int ret;
>
> - ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
> - if (ret)
> - return ret;
> -
> if (i915_gem_object_has_struct_page(obj)) {
> /*
> * Avoid waking the device up if we can fallback, as
> @@ -544,10 +547,8 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
> * using the cache bypass of indirect GGTT access.
> */
> wakeref = intel_runtime_pm_get_if_in_use(rpm);
> - if (!wakeref) {
> - ret = -EFAULT;
> - goto out_unlock;
> - }
> + if (!wakeref)
> + return -EFAULT;
> } else {
> /* No backing pages, no fallback, we must force GGTT access */
> wakeref = intel_runtime_pm_get(rpm);
> @@ -569,8 +570,6 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
> GEM_BUG_ON(!drm_mm_node_allocated(&node));
> }
>
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> ret = i915_gem_object_lock_interruptible(obj);
> if (ret)
> goto out_unpin;
> @@ -634,18 +633,15 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
>
> i915_gem_object_unlock_fence(obj, fence);
> out_unpin:
> - mutex_lock(&i915->drm.struct_mutex);
> intel_gt_flush_ggtt_writes(ggtt->vm.gt);
> if (drm_mm_node_allocated(&node)) {
> ggtt->vm.clear_range(&ggtt->vm, node.start, node.size);
> - remove_mappable_node(&node);
> + remove_mappable_node(ggtt, &node);
> } else {
> i915_vma_unpin(vma);
> }
> out_rpm:
> intel_runtime_pm_put(rpm, wakeref);
> -out_unlock:
> - mutex_unlock(&i915->drm.struct_mutex);
> return ret;
> }
>
> @@ -968,8 +964,6 @@ i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj,
> struct i915_vma *vma;
> int ret;
>
> - lockdep_assert_held(&obj->base.dev->struct_mutex);
> -
> if (flags & PIN_MAPPABLE &&
> (!view || view->type == I915_GGTT_VIEW_NORMAL)) {
> /* If the required space is larger than the available
> @@ -1016,13 +1010,6 @@ i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj,
> return ERR_PTR(-ENOSPC);
> }
>
> - WARN(i915_vma_is_pinned(vma),
> - "bo is already pinned in ggtt with incorrect alignment:"
> - " offset=%08x, req.alignment=%llx,"
> - " req.map_and_fenceable=%d, vma->map_and_fenceable=%d\n",
> - i915_ggtt_offset(vma), alignment,
> - !!(flags & PIN_MAPPABLE),
> - i915_vma_is_map_and_fenceable(vma));
> ret = i915_vma_unbind(vma);
> if (ret)
> return ERR_PTR(ret);
> @@ -1441,8 +1428,6 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
> }
>
> if (ret == -EIO) {
> - mutex_lock(&dev_priv->drm.struct_mutex);
> -
> /*
> * Allow engines or uC initialisation to fail by marking the GPU
> * as wedged. But we only want to do this when the GPU is angry,
> @@ -1459,8 +1444,6 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
> i915_gem_restore_gtt_mappings(dev_priv);
> i915_gem_restore_fences(dev_priv);
> intel_init_clock_gating(dev_priv);
> -
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> }
>
> i915_gem_drain_freed_objects(dev_priv);
> diff --git a/drivers/gpu/drm/i915/i915_gem_evict.c b/drivers/gpu/drm/i915/i915_gem_evict.c
> index 8c1e04f402bc..0552bf93eea3 100644
> --- a/drivers/gpu/drm/i915/i915_gem_evict.c
> +++ b/drivers/gpu/drm/i915/i915_gem_evict.c
> @@ -47,8 +47,7 @@ static int ggtt_flush(struct drm_i915_private *i915)
> * bound by their active reference.
> */
> return i915_gem_wait_for_idle(i915,
> - I915_WAIT_INTERRUPTIBLE |
> - I915_WAIT_LOCKED,
> + I915_WAIT_INTERRUPTIBLE,
> MAX_SCHEDULE_TIMEOUT);
> }
>
> @@ -104,7 +103,7 @@ i915_gem_evict_something(struct i915_address_space *vm,
> struct i915_vma *active;
> int ret;
>
> - lockdep_assert_held(&vm->i915->drm.struct_mutex);
> + lockdep_assert_held(&vm->mutex);
> trace_i915_gem_evict(vm, min_size, alignment, flags);
>
> /*
> @@ -127,15 +126,6 @@ i915_gem_evict_something(struct i915_address_space *vm,
> min_size, alignment, color,
> start, end, mode);
>
> - /*
> - * Retire before we search the active list. Although we have
> - * reasonable accuracy in our retirement lists, we may have
> - * a stray pin (preventing eviction) that can only be resolved by
> - * retiring.
> - */
> - if (!(flags & PIN_NONBLOCK))
> - i915_retire_requests(dev_priv);
> -
> search_again:
> active = NULL;
> INIT_LIST_HEAD(&eviction_list);
> @@ -235,12 +225,12 @@ i915_gem_evict_something(struct i915_address_space *vm,
> list_for_each_entry_safe(vma, next, &eviction_list, evict_link) {
> __i915_vma_unpin(vma);
> if (ret == 0)
> - ret = i915_vma_unbind(vma);
> + ret = __i915_vma_unbind(vma);
> }
>
> while (ret == 0 && (node = drm_mm_scan_color_evict(&scan))) {
> vma = container_of(node, struct i915_vma, node);
> - ret = i915_vma_unbind(vma);
> + ret = __i915_vma_unbind(vma);
> }
>
> return ret;
> @@ -268,7 +258,7 @@ int i915_gem_evict_for_node(struct i915_address_space *vm,
> struct i915_vma *vma, *next;
> int ret = 0;
>
> - lockdep_assert_held(&vm->i915->drm.struct_mutex);
> + lockdep_assert_held(&vm->mutex);
> GEM_BUG_ON(!IS_ALIGNED(start, I915_GTT_PAGE_SIZE));
> GEM_BUG_ON(!IS_ALIGNED(end, I915_GTT_PAGE_SIZE));
>
> @@ -349,7 +339,7 @@ int i915_gem_evict_for_node(struct i915_address_space *vm,
> list_for_each_entry_safe(vma, next, &eviction_list, evict_link) {
> __i915_vma_unpin(vma);
> if (ret == 0)
> - ret = i915_vma_unbind(vma);
> + ret = __i915_vma_unbind(vma);
> }
>
> return ret;
> @@ -373,7 +363,7 @@ int i915_gem_evict_vm(struct i915_address_space *vm)
> struct i915_vma *vma, *next;
> int ret;
>
> - lockdep_assert_held(&vm->i915->drm.struct_mutex);
> + lockdep_assert_held(&vm->mutex);
> trace_i915_gem_evict_vm(vm);
>
> /* Switch back to the default context in order to unpin
> @@ -388,7 +378,6 @@ int i915_gem_evict_vm(struct i915_address_space *vm)
> }
>
> INIT_LIST_HEAD(&eviction_list);
> - mutex_lock(&vm->mutex);
> list_for_each_entry(vma, &vm->bound_list, vm_link) {
> if (i915_vma_is_pinned(vma))
> continue;
> @@ -396,13 +385,12 @@ int i915_gem_evict_vm(struct i915_address_space *vm)
> __i915_vma_pin(vma);
> list_add(&vma->evict_link, &eviction_list);
> }
> - mutex_unlock(&vm->mutex);
>
> ret = 0;
> list_for_each_entry_safe(vma, next, &eviction_list, evict_link) {
> __i915_vma_unpin(vma);
> if (ret == 0)
> - ret = i915_vma_unbind(vma);
> + ret = __i915_vma_unbind(vma);
> }
> return ret;
> }
> diff --git a/drivers/gpu/drm/i915/i915_gem_fence_reg.c b/drivers/gpu/drm/i915/i915_gem_fence_reg.c
> index 615a9f4ef30c..487b7261f7ed 100644
> --- a/drivers/gpu/drm/i915/i915_gem_fence_reg.c
> +++ b/drivers/gpu/drm/i915/i915_gem_fence_reg.c
> @@ -230,14 +230,15 @@ static int fence_update(struct i915_fence_reg *fence,
> i915_gem_object_get_tiling(vma->obj)))
> return -EINVAL;
>
> - ret = i915_active_wait(&vma->active);
> + ret = i915_vma_sync(vma);
> if (ret)
> return ret;
> }
>
> old = xchg(&fence->vma, NULL);
> if (old) {
> - ret = i915_active_wait(&old->active);
> + /* XXX Ideally we would move the waiting to outside the mutex */
> + ret = i915_vma_sync(old);
> if (ret) {
> fence->vma = old;
> return ret;
> @@ -331,13 +332,15 @@ static struct i915_fence_reg *fence_find(struct drm_i915_private *i915)
> return ERR_PTR(-EDEADLK);
> }
>
> -static int __i915_vma_pin_fence(struct i915_vma *vma)
> +int __i915_vma_pin_fence(struct i915_vma *vma)
> {
> struct i915_ggtt *ggtt = i915_vm_to_ggtt(vma->vm);
> struct i915_fence_reg *fence;
> struct i915_vma *set = i915_gem_object_is_tiled(vma->obj) ? vma : NULL;
> int err;
>
> + lockdep_assert_held(&vma->vm->mutex);
> +
> /* Just update our place in the LRU if our fence is getting reused. */
> if (vma->fence) {
> fence = vma->fence;
> diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
> index 483eeff8730e..a351b98afe03 100644
> --- a/drivers/gpu/drm/i915/i915_gem_gtt.c
> +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
> @@ -150,16 +150,18 @@ static void gmch_ggtt_invalidate(struct i915_ggtt *ggtt)
>
> static int ppgtt_bind_vma(struct i915_vma *vma,
> enum i915_cache_level cache_level,
> - u32 unused)
> + u32 flags)
> {
> u32 pte_flags;
> int err;
>
> - if (!i915_vma_is_bound(vma, I915_VMA_LOCAL_BIND)) {
> + if (flags & I915_VMA_ALLOC) {
> err = vma->vm->allocate_va_range(vma->vm,
> vma->node.start, vma->size);
> if (err)
> return err;
> +
> + set_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma));
> }
>
> /* Applicable to VLV, and gen8+ */
> @@ -167,6 +169,7 @@ static int ppgtt_bind_vma(struct i915_vma *vma,
> if (i915_gem_object_is_readonly(vma->obj))
> pte_flags |= PTE_READ_ONLY;
>
> + GEM_BUG_ON(!test_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma)));
> vma->vm->insert_entries(vma->vm, vma, cache_level, pte_flags);
> wmb();
>
> @@ -175,7 +178,8 @@ static int ppgtt_bind_vma(struct i915_vma *vma,
>
> static void ppgtt_unbind_vma(struct i915_vma *vma)
> {
> - vma->vm->clear_range(vma->vm, vma->node.start, vma->size);
> + if (test_and_clear_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma)))
> + vma->vm->clear_range(vma->vm, vma->node.start, vma->size);
> }
>
> static int ppgtt_set_pages(struct i915_vma *vma)
> @@ -503,15 +507,26 @@ static void i915_address_space_fini(struct i915_address_space *vm)
> mutex_destroy(&vm->mutex);
> }
>
> -static void ppgtt_destroy_vma(struct i915_address_space *vm)
> +void __i915_vm_close(struct i915_address_space *vm)
> {
> struct i915_vma *vma, *vn;
>
> - mutex_lock(&vm->i915->drm.struct_mutex);
> - list_for_each_entry_safe(vma, vn, &vm->bound_list, vm_link)
> + mutex_lock(&vm->mutex);
> + list_for_each_entry_safe(vma, vn, &vm->bound_list, vm_link) {
> + struct drm_i915_gem_object *obj = vma->obj;
> +
> + /* Keep the obj (and hence the vma) alive as _we_ destroy it */
> + if (!kref_get_unless_zero(&obj->base.refcount))
> + continue;
> +
> + atomic_and(~I915_VMA_PIN_MASK, &vma->flags);
> + WARN_ON(__i915_vma_unbind(vma));
> i915_vma_destroy(vma);
> +
> + i915_gem_object_put(obj);
> + }
> GEM_BUG_ON(!list_empty(&vm->bound_list));
> - mutex_unlock(&vm->i915->drm.struct_mutex);
> + mutex_unlock(&vm->mutex);
> }
>
> static void __i915_vm_release(struct work_struct *work)
> @@ -519,8 +534,6 @@ static void __i915_vm_release(struct work_struct *work)
> struct i915_address_space *vm =
> container_of(work, struct i915_address_space, rcu.work);
>
> - ppgtt_destroy_vma(vm);
> -
> vm->cleanup(vm);
> i915_address_space_fini(vm);
>
> @@ -535,7 +548,6 @@ void i915_vm_release(struct kref *kref)
> GEM_BUG_ON(i915_is_ggtt(vm));
> trace_i915_ppgtt_release(vm);
>
> - vm->closed = true;
> queue_rcu_work(vm->i915->wq, &vm->rcu);
> }
>
> @@ -543,6 +555,7 @@ static void i915_address_space_init(struct i915_address_space *vm, int subclass)
> {
> kref_init(&vm->ref);
> INIT_RCU_WORK(&vm->rcu, __i915_vm_release);
> + atomic_set(&vm->open, 1);
>
> /*
> * The vm->mutex must be reclaim safe (for use in the shrinker).
> @@ -1771,12 +1784,8 @@ static void gen6_ppgtt_free_pd(struct gen6_ppgtt *ppgtt)
> static void gen6_ppgtt_cleanup(struct i915_address_space *vm)
> {
> struct gen6_ppgtt *ppgtt = to_gen6_ppgtt(i915_vm_to_ppgtt(vm));
> - struct drm_i915_private *i915 = vm->i915;
>
> - /* FIXME remove the struct_mutex to bring the locking under control */
> - mutex_lock(&i915->drm.struct_mutex);
> i915_vma_destroy(ppgtt->vma);
> - mutex_unlock(&i915->drm.struct_mutex);
>
> gen6_ppgtt_free_pd(ppgtt);
> free_scratch(vm);
> @@ -1865,7 +1874,8 @@ static struct i915_vma *pd_vma_create(struct gen6_ppgtt *ppgtt, int size)
>
> i915_active_init(i915, &vma->active, NULL, NULL);
>
> - vma->vm = &ggtt->vm;
> + mutex_init(&vma->pages_mutex);
> + vma->vm = i915_vm_get(&ggtt->vm);
> vma->ops = &pd_vma_ops;
> vma->private = ppgtt;
>
> @@ -1885,7 +1895,7 @@ int gen6_ppgtt_pin(struct i915_ppgtt *base)
> struct gen6_ppgtt *ppgtt = to_gen6_ppgtt(base);
> int err = 0;
>
> - GEM_BUG_ON(ppgtt->base.vm.closed);
> + GEM_BUG_ON(!atomic_read(&ppgtt->base.vm.open));
>
> /*
> * Workaround the limited maximum vma->pin_count and the aliasing_ppgtt
> @@ -2463,14 +2473,18 @@ static int aliasing_gtt_bind_vma(struct i915_vma *vma,
> if (flags & I915_VMA_LOCAL_BIND) {
> struct i915_ppgtt *alias = i915_vm_to_ggtt(vma->vm)->alias;
>
> - if (!i915_vma_is_bound(vma, I915_VMA_LOCAL_BIND)) {
> + if (flags & I915_VMA_ALLOC) {
> ret = alias->vm.allocate_va_range(&alias->vm,
> vma->node.start,
> vma->size);
> if (ret)
> return ret;
> +
> + set_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma));
> }
>
> + GEM_BUG_ON(!test_bit(I915_VMA_ALLOC_BIT,
> + __i915_vma_flags(vma)));
> alias->vm.insert_entries(&alias->vm, vma,
> cache_level, pte_flags);
> }
> @@ -2499,7 +2513,7 @@ static void aliasing_gtt_unbind_vma(struct i915_vma *vma)
> vm->clear_range(vm, vma->node.start, vma->size);
> }
>
> - if (i915_vma_is_bound(vma, I915_VMA_LOCAL_BIND)) {
> + if (test_and_clear_bit(I915_VMA_ALLOC_BIT, __i915_vma_flags(vma))) {
> struct i915_address_space *vm =
> &i915_vm_to_ggtt(vma->vm)->alias->vm;
>
> @@ -2602,22 +2616,16 @@ static int init_aliasing_ppgtt(struct i915_ggtt *ggtt)
>
> static void fini_aliasing_ppgtt(struct i915_ggtt *ggtt)
> {
> - struct drm_i915_private *i915 = ggtt->vm.i915;
> struct i915_ppgtt *ppgtt;
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> ppgtt = fetch_and_zero(&ggtt->alias);
> if (!ppgtt)
> - goto out;
> + return;
>
> i915_vm_put(&ppgtt->vm);
>
> ggtt->vm.vma_ops.bind_vma = ggtt_bind_vma;
> ggtt->vm.vma_ops.unbind_vma = ggtt_unbind_vma;
> -
> -out:
> - mutex_unlock(&i915->drm.struct_mutex);
> }
>
> static int ggtt_reserve_guc_top(struct i915_ggtt *ggtt)
> @@ -2734,32 +2742,28 @@ int i915_init_ggtt(struct drm_i915_private *i915)
>
> static void ggtt_cleanup_hw(struct i915_ggtt *ggtt)
> {
> - struct drm_i915_private *i915 = ggtt->vm.i915;
> struct i915_vma *vma, *vn;
>
> - ggtt->vm.closed = true;
> + atomic_set(&ggtt->vm.open, 0);
>
> rcu_barrier(); /* flush the RCU'ed__i915_vm_release */
> - flush_workqueue(i915->wq);
> + flush_workqueue(ggtt->vm.i915->wq);
>
> - mutex_lock(&i915->drm.struct_mutex);
> + mutex_lock(&ggtt->vm.mutex);
>
> list_for_each_entry_safe(vma, vn, &ggtt->vm.bound_list, vm_link)
> - WARN_ON(i915_vma_unbind(vma));
> + WARN_ON(__i915_vma_unbind(vma));
>
> if (drm_mm_node_allocated(&ggtt->error_capture))
> drm_mm_remove_node(&ggtt->error_capture);
>
> ggtt_release_guc_top(ggtt);
> -
> - if (drm_mm_initialized(&ggtt->vm.mm)) {
> - intel_vgt_deballoon(ggtt);
> - i915_address_space_fini(&ggtt->vm);
> - }
> + intel_vgt_deballoon(ggtt);
>
> ggtt->vm.cleanup(&ggtt->vm);
>
> - mutex_unlock(&i915->drm.struct_mutex);
> + mutex_unlock(&ggtt->vm.mutex);
> + i915_address_space_fini(&ggtt->vm);
>
> arch_phys_wc_del(ggtt->mtrr);
> io_mapping_fini(&ggtt->iomap);
> @@ -3188,9 +3192,6 @@ int i915_ggtt_probe_hw(struct drm_i915_private *i915)
> static int ggtt_init_hw(struct i915_ggtt *ggtt)
> {
> struct drm_i915_private *i915 = ggtt->vm.i915;
> - int ret = 0;
> -
> - mutex_lock(&i915->drm.struct_mutex);
>
> i915_address_space_init(&ggtt->vm, VM_CLASS_GGTT);
>
> @@ -3206,18 +3207,14 @@ static int ggtt_init_hw(struct i915_ggtt *ggtt)
> ggtt->gmadr.start,
> ggtt->mappable_end)) {
> ggtt->vm.cleanup(&ggtt->vm);
> - ret = -EIO;
> - goto out;
> + return -EIO;
> }
>
> ggtt->mtrr = arch_phys_wc_add(ggtt->gmadr.start, ggtt->mappable_end);
>
> i915_ggtt_init_fences(ggtt);
>
> -out:
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> - return ret;
> + return 0;
> }
>
> /**
> @@ -3289,6 +3286,7 @@ static void ggtt_restore_mappings(struct i915_ggtt *ggtt)
> {
> struct i915_vma *vma, *vn;
> bool flush = false;
> + int open;
>
> intel_gt_check_and_clear_faults(ggtt->vm.gt);
>
> @@ -3296,7 +3294,9 @@ static void ggtt_restore_mappings(struct i915_ggtt *ggtt)
>
> /* First fill our portion of the GTT with scratch pages */
> ggtt->vm.clear_range(&ggtt->vm, 0, ggtt->vm.total);
> - ggtt->vm.closed = true; /* skip rewriting PTE on VMA unbind */
> +
> + /* Skip rewriting PTE on VMA unbind. */
> + open = atomic_xchg(&ggtt->vm.open, 0);
>
> /* clflush objects bound into the GGTT and rebind them. */
> list_for_each_entry_safe(vma, vn, &ggtt->vm.bound_list, vm_link) {
> @@ -3305,24 +3305,20 @@ static void ggtt_restore_mappings(struct i915_ggtt *ggtt)
> if (!i915_vma_is_bound(vma, I915_VMA_GLOBAL_BIND))
> continue;
>
> - mutex_unlock(&ggtt->vm.mutex);
> -
> - if (!i915_vma_unbind(vma))
> - goto lock;
> + if (!__i915_vma_unbind(vma))
> + continue;
>
> + clear_bit(I915_VMA_GLOBAL_BIND_BIT, __i915_vma_flags(vma));
> WARN_ON(i915_vma_bind(vma,
> obj ? obj->cache_level : 0,
> - PIN_UPDATE));
> + PIN_GLOBAL, NULL));
> if (obj) { /* only used during resume => exclusive access */
> flush |= fetch_and_zero(&obj->write_domain);
> obj->read_domains |= I915_GEM_DOMAIN_GTT;
> }
> -
> -lock:
> - mutex_lock(&ggtt->vm.mutex);
> }
>
> - ggtt->vm.closed = false;
> + atomic_set(&ggtt->vm.open, open);
> ggtt->invalidate(ggtt);
>
> mutex_unlock(&ggtt->vm.mutex);
> @@ -3714,7 +3710,8 @@ int i915_gem_gtt_insert(struct i915_address_space *vm,
> u64 offset;
> int err;
>
> - lockdep_assert_held(&vm->i915->drm.struct_mutex);
> + lockdep_assert_held(&vm->mutex);
> +
> GEM_BUG_ON(!size);
> GEM_BUG_ON(!IS_ALIGNED(size, I915_GTT_PAGE_SIZE));
> GEM_BUG_ON(alignment && !is_power_of_2(alignment));
> diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
> index d8dfa3811e54..03418dede503 100644
> --- a/drivers/gpu/drm/i915/i915_gem_gtt.h
> +++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
> @@ -307,7 +307,14 @@ struct i915_address_space {
>
> unsigned int bind_alloc;
>
> - bool closed;
> + /*
> + * Each active user context has its own address space (in full-ppgtt).
> + * Since the vm may be shared between multiple contexts, we count how
> + * many contexts keep us "open". Once open hits zero, we are closed
> + * and do not allow any new attachments, and proceed to shutdown our
> + * vma and page directories.
> + */
> + atomic_t open;
>
> struct mutex mutex; /* protects vma and our lists */
> #define VM_CLASS_GGTT 0
> @@ -581,6 +588,35 @@ static inline void i915_vm_put(struct i915_address_space *vm)
> kref_put(&vm->ref, i915_vm_release);
> }
>
> +static inline struct i915_address_space *
> +i915_vm_open(struct i915_address_space *vm)
> +{
> + GEM_BUG_ON(!atomic_read(&vm->open));
> + atomic_inc(&vm->open);
> + return i915_vm_get(vm);
> +}
> +
> +static inline bool
> +i915_vm_tryopen(struct i915_address_space *vm)
> +{
> + if (atomic_add_unless(&vm->open, 1, 0))
> + return i915_vm_get(vm);
> +
> + return false;
> +}
> +
> +void __i915_vm_close(struct i915_address_space *vm);
> +
> +static inline void
> +i915_vm_close(struct i915_address_space *vm)
> +{
> + GEM_BUG_ON(!atomic_read(&vm->open));
> + if (atomic_dec_and_test(&vm->open))
> + __i915_vm_close(vm);
> +
> + i915_vm_put(vm);
> +}
> +
> int gen6_ppgtt_pin(struct i915_ppgtt *base);
> void gen6_ppgtt_unpin(struct i915_ppgtt *base);
> void gen6_ppgtt_unpin_all(struct i915_ppgtt *base);
> @@ -613,10 +649,9 @@ int i915_gem_gtt_insert(struct i915_address_space *vm,
> #define PIN_OFFSET_BIAS BIT_ULL(6)
> #define PIN_OFFSET_FIXED BIT_ULL(7)
>
> -#define PIN_MBZ BIT_ULL(8) /* I915_VMA_PIN_OVERFLOW */
> -#define PIN_GLOBAL BIT_ULL(9) /* I915_VMA_GLOBAL_BIND */
> -#define PIN_USER BIT_ULL(10) /* I915_VMA_LOCAL_BIND */
> -#define PIN_UPDATE BIT_ULL(11)
> +#define PIN_UPDATE BIT_ULL(9)
> +#define PIN_GLOBAL BIT_ULL(10) /* I915_VMA_GLOBAL_BIND */
> +#define PIN_USER BIT_ULL(11) /* I915_VMA_LOCAL_BIND */
>
> #define PIN_OFFSET_MASK (-I915_GTT_PAGE_SIZE)
>
> diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
> index 524f6710b7aa..80055501eccb 100644
> --- a/drivers/gpu/drm/i915/i915_perf.c
> +++ b/drivers/gpu/drm/i915/i915_perf.c
> @@ -1204,15 +1204,10 @@ static int i915_oa_read(struct i915_perf_stream *stream,
> static struct intel_context *oa_pin_context(struct i915_perf_stream *stream)
> {
> struct i915_gem_engines_iter it;
> - struct drm_i915_private *i915 = stream->dev_priv;
> struct i915_gem_context *ctx = stream->ctx;
> struct intel_context *ce;
> int err;
>
> - err = i915_mutex_lock_interruptible(&i915->drm);
> - if (err)
> - return ERR_PTR(err);
> -
> for_each_gem_engine(ce, i915_gem_context_lock_engines(ctx), it) {
> if (ce->engine->class != RENDER_CLASS)
> continue;
> @@ -1229,10 +1224,6 @@ static struct intel_context *oa_pin_context(struct i915_perf_stream *stream)
> }
> i915_gem_context_unlock_engines(ctx);
>
> - mutex_unlock(&i915->drm.struct_mutex);
> - if (err)
> - return ERR_PTR(err);
> -
> return stream->pinned_ctx;
> }
>
> @@ -1331,32 +1322,22 @@ static int oa_get_render_ctx_id(struct i915_perf_stream *stream)
> */
> static void oa_put_render_ctx_id(struct i915_perf_stream *stream)
> {
> - struct drm_i915_private *dev_priv = stream->dev_priv;
> struct intel_context *ce;
>
> stream->specific_ctx_id = INVALID_CTX_ID;
> stream->specific_ctx_id_mask = 0;
>
> ce = fetch_and_zero(&stream->pinned_ctx);
> - if (ce) {
> - mutex_lock(&dev_priv->drm.struct_mutex);
> + if (ce)
> intel_context_unpin(ce);
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> - }
> }
>
> static void
> free_oa_buffer(struct i915_perf_stream *stream)
> {
> - struct drm_i915_private *i915 = stream->dev_priv;
> -
> - mutex_lock(&i915->drm.struct_mutex);
> -
> i915_vma_unpin_and_release(&stream->oa_buffer.vma,
> I915_VMA_RELEASE_MAP);
>
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> stream->oa_buffer.vaddr = NULL;
> }
>
> @@ -1511,18 +1492,13 @@ static int alloc_oa_buffer(struct i915_perf_stream *stream)
> if (WARN_ON(stream->oa_buffer.vma))
> return -ENODEV;
>
> - ret = i915_mutex_lock_interruptible(&dev_priv->drm);
> - if (ret)
> - return ret;
> -
> 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_shmem(dev_priv, OA_BUFFER_SIZE);
> if (IS_ERR(bo)) {
> DRM_ERROR("Failed to allocate OA buffer\n");
> - ret = PTR_ERR(bo);
> - goto unlock;
> + return PTR_ERR(bo);
> }
>
> i915_gem_object_set_cache_coherency(bo, I915_CACHE_LLC);
> @@ -1546,7 +1522,7 @@ static int alloc_oa_buffer(struct i915_perf_stream *stream)
> i915_ggtt_offset(stream->oa_buffer.vma),
> stream->oa_buffer.vaddr);
>
> - goto unlock;
> + return 0;
>
> err_unpin:
> __i915_vma_unpin(vma);
> @@ -1557,8 +1533,6 @@ static int alloc_oa_buffer(struct i915_perf_stream *stream)
> stream->oa_buffer.vaddr = NULL;
> stream->oa_buffer.vma = NULL;
>
> -unlock:
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> return ret;
> }
>
> diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
> index d097f77890ba..defeb9bb1daa 100644
> --- a/drivers/gpu/drm/i915/i915_vma.c
> +++ b/drivers/gpu/drm/i915/i915_vma.c
> @@ -32,6 +32,7 @@
>
> #include "i915_drv.h"
> #include "i915_globals.h"
> +#include "i915_sw_fence_work.h"
> #include "i915_trace.h"
> #include "i915_vma.h"
>
> @@ -110,7 +111,8 @@ vma_create(struct drm_i915_gem_object *obj,
> if (vma == NULL)
> return ERR_PTR(-ENOMEM);
>
> - vma->vm = vm;
> + mutex_init(&vma->pages_mutex);
> + vma->vm = i915_vm_get(vm);
> vma->ops = &vm->vma_ops;
> vma->obj = obj;
> vma->resv = obj->base.resv;
> @@ -261,8 +263,6 @@ vma_lookup(struct drm_i915_gem_object *obj,
> * Once created, the VMA is kept until either the object is freed, or the
> * address space is closed.
> *
> - * Must be called with struct_mutex held.
> - *
> * Returns the vma, or an error pointer.
> */
> struct i915_vma *
> @@ -273,7 +273,7 @@ i915_vma_instance(struct drm_i915_gem_object *obj,
> struct i915_vma *vma;
>
> GEM_BUG_ON(view && !i915_is_ggtt(vm));
> - GEM_BUG_ON(vm->closed);
> + GEM_BUG_ON(!atomic_read(&vm->open));
>
> spin_lock(&obj->vma.lock);
> vma = vma_lookup(obj, vm, view);
> @@ -287,18 +287,63 @@ i915_vma_instance(struct drm_i915_gem_object *obj,
> return vma;
> }
>
> +struct i915_vma_work {
> + struct dma_fence_work base;
> + struct i915_vma *vma;
> + enum i915_cache_level cache_level;
> + unsigned int flags;
> +};
> +
> +static int __vma_bind(struct dma_fence_work *work)
> +{
> + struct i915_vma_work *vw = container_of(work, typeof(*vw), base);
> + struct i915_vma *vma = vw->vma;
> + int err;
> +
> + err = vma->ops->bind_vma(vma, vw->cache_level, vw->flags);
> + if (err)
> + atomic_or(I915_VMA_ERROR, &vma->flags);
> +
> + if (vma->obj)
> + __i915_gem_object_unpin_pages(vma->obj);
> +
> + return err;
> +}
> +
> +static const struct dma_fence_work_ops bind_ops = {
> + .name = "bind",
> + .work = __vma_bind,
> +};
> +
> +struct i915_vma_work *i915_vma_work(void)
> +{
> + struct i915_vma_work *vw;
> +
> + vw = kzalloc(sizeof(*vw), GFP_KERNEL);
> + if (!vw)
> + return NULL;
> +
> + dma_fence_work_init(&vw->base, &bind_ops);
> + vw->base.dma.error = -EAGAIN; /* disable the worker by default */
> +
> + return vw;
> +}
> +
> /**
> * i915_vma_bind - Sets up PTEs for an VMA in it's corresponding address space.
> * @vma: VMA to map
> * @cache_level: mapping cache level
> * @flags: flags like global or local mapping
> + * @work: preallocated worker for allocating and binding the PTE
> *
> * DMA addresses are taken from the scatter-gather table of this object (or of
> * this VMA in case of non-default GGTT views) and PTE entries set up.
> * Note that DMA addresses are also the only part of the SG table we care about.
> */
> -int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level,
> - u32 flags)
> +int i915_vma_bind(struct i915_vma *vma,
> + enum i915_cache_level cache_level,
> + u32 flags,
> + struct i915_vma_work *work)
> {
> u32 bind_flags;
> u32 vma_flags;
> @@ -315,11 +360,8 @@ int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level,
> if (GEM_DEBUG_WARN_ON(!flags))
> return -EINVAL;
>
> - bind_flags = 0;
> - if (flags & PIN_GLOBAL)
> - bind_flags |= I915_VMA_GLOBAL_BIND;
> - if (flags & PIN_USER)
> - bind_flags |= I915_VMA_LOCAL_BIND;
> + bind_flags = flags;
> + bind_flags &= I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND;
>
> vma_flags = atomic_read(&vma->flags);
> vma_flags &= I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND;
> @@ -333,9 +375,32 @@ int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level,
> GEM_BUG_ON(!vma->pages);
>
> trace_i915_vma_bind(vma, bind_flags);
> - ret = vma->ops->bind_vma(vma, cache_level, bind_flags);
> - if (ret)
> - return ret;
> + if (work && (bind_flags & ~vma_flags) & vma->vm->bind_alloc) {
> + work->vma = vma;
> + work->cache_level = cache_level;
> + work->flags = bind_flags | I915_VMA_ALLOC;
> +
> + /*
> + * Note we only want to chain up to the migration fence on
> + * the pages (not the object itself). As we don't track that,
> + * yet, we have to use the exclusive fence instead.
> + *
> + * Also note that we do not want to track the async vma as
> + * part of the obj->resv->excl_fence as it only affects
> + * execution and not content or object's backing store lifetime.
> + */
> + GEM_BUG_ON(i915_active_has_exclusive(&vma->active));
> + i915_active_set_exclusive(&vma->active, &work->base.dma);
> + work->base.dma.error = 0; /* enable the queue_work() */
> +
> + if (vma->obj)
> + __i915_gem_object_pin_pages(vma->obj);
> + } else {
> + GEM_BUG_ON((bind_flags & ~vma_flags) & vma->vm->bind_alloc);
> + ret = vma->ops->bind_vma(vma, cache_level, bind_flags);
> + if (ret)
> + return ret;
> + }
>
> atomic_or(bind_flags, &vma->flags);
> return 0;
> @@ -348,9 +413,7 @@ void __iomem *i915_vma_pin_iomap(struct i915_vma *vma)
>
> /* Access through the GTT requires the device to be awake. */
> assert_rpm_wakelock_held(&vma->vm->i915->runtime_pm);
> -
> - lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
> - if (WARN_ON(!i915_vma_is_map_and_fenceable(vma))) {
> + if (GEM_WARN_ON(!i915_vma_is_map_and_fenceable(vma))) {
> err = -ENODEV;
> goto err;
> }
> @@ -358,7 +421,7 @@ void __iomem *i915_vma_pin_iomap(struct i915_vma *vma)
> GEM_BUG_ON(!i915_vma_is_ggtt(vma));
> GEM_BUG_ON(!i915_vma_is_bound(vma, I915_VMA_GLOBAL_BIND));
>
> - ptr = vma->iomap;
> + ptr = READ_ONCE(vma->iomap);
> if (ptr == NULL) {
> ptr = io_mapping_map_wc(&i915_vm_to_ggtt(vma->vm)->iomap,
> vma->node.start,
> @@ -368,7 +431,10 @@ void __iomem *i915_vma_pin_iomap(struct i915_vma *vma)
> goto err;
> }
>
> - vma->iomap = ptr;
> + if (unlikely(cmpxchg(&vma->iomap, NULL, ptr))) {
> + io_mapping_unmap(ptr);
> + ptr = vma->iomap;
> + }
> }
>
> __i915_vma_pin(vma);
> @@ -388,18 +454,12 @@ void __iomem *i915_vma_pin_iomap(struct i915_vma *vma)
>
> void i915_vma_flush_writes(struct i915_vma *vma)
> {
> - if (!i915_vma_has_ggtt_write(vma))
> - return;
> -
> - intel_gt_flush_ggtt_writes(vma->vm->gt);
> -
> - i915_vma_unset_ggtt_write(vma);
> + if (i915_vma_unset_ggtt_write(vma))
> + intel_gt_flush_ggtt_writes(vma->vm->gt);
> }
>
> void i915_vma_unpin_iomap(struct i915_vma *vma)
> {
> - lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
> -
> GEM_BUG_ON(vma->iomap == NULL);
>
> i915_vma_flush_writes(vma);
> @@ -435,6 +495,9 @@ bool i915_vma_misplaced(const struct i915_vma *vma,
> if (!drm_mm_node_allocated(&vma->node))
> return false;
>
> + if (test_bit(I915_VMA_ERROR_BIT, __i915_vma_flags(vma)))
> + return true;
> +
> if (vma->node.size < size)
> return true;
>
> @@ -535,7 +598,6 @@ static void assert_bind_count(const struct drm_i915_gem_object *obj)
> static int
> i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
> {
> - struct drm_i915_private *dev_priv = vma->vm->i915;
> unsigned long color;
> u64 start, end;
> int ret;
> @@ -561,7 +623,7 @@ i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
>
> end = vma->vm->total;
> if (flags & PIN_MAPPABLE)
> - end = min_t(u64, end, dev_priv->ggtt.mappable_end);
> + end = min_t(u64, end, i915_vm_to_ggtt(vma->vm)->mappable_end);
> if (flags & PIN_ZONE_4G)
> end = min_t(u64, end, (1ULL << 32) - I915_GTT_PAGE_SIZE);
> GEM_BUG_ON(!IS_ALIGNED(end, I915_GTT_PAGE_SIZE));
> @@ -578,34 +640,20 @@ i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
> }
>
> color = 0;
> - if (vma->obj) {
> - ret = i915_gem_object_pin_pages(vma->obj);
> - if (ret)
> - return ret;
> -
> - if (i915_vm_has_cache_coloring(vma->vm))
> - color = vma->obj->cache_level;
> - }
> -
> - GEM_BUG_ON(vma->pages);
> -
> - ret = vma->ops->set_pages(vma);
> - if (ret)
> - goto err_unpin;
> + if (vma->obj && i915_vm_has_cache_coloring(vma->vm))
> + color = vma->obj->cache_level;
>
> if (flags & PIN_OFFSET_FIXED) {
> u64 offset = flags & PIN_OFFSET_MASK;
> if (!IS_ALIGNED(offset, alignment) ||
> - range_overflows(offset, size, end)) {
> - ret = -EINVAL;
> - goto err_clear;
> - }
> + range_overflows(offset, size, end))
> + return -EINVAL;
>
> ret = i915_gem_gtt_reserve(vma->vm, &vma->node,
> size, offset, color,
> flags);
> if (ret)
> - goto err_clear;
> + return ret;
> } else {
> /*
> * We only support huge gtt pages through the 48b PPGTT,
> @@ -644,7 +692,7 @@ i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
> size, alignment, color,
> start, end, flags);
> if (ret)
> - goto err_clear;
> + return ret;
>
> GEM_BUG_ON(vma->node.start < start);
> GEM_BUG_ON(vma->node.start + vma->node.size > end);
> @@ -652,23 +700,15 @@ i915_vma_insert(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
> GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
> GEM_BUG_ON(!i915_gem_valid_gtt_space(vma, color));
>
> - mutex_lock(&vma->vm->mutex);
> list_add_tail(&vma->vm_link, &vma->vm->bound_list);
> - mutex_unlock(&vma->vm->mutex);
>
> if (vma->obj) {
> + atomic_inc(&vma->obj->mm.pages_pin_count);
> atomic_inc(&vma->obj->bind_count);
> assert_bind_count(vma->obj);
> }
>
> return 0;
> -
> -err_clear:
> - vma->ops->clear_pages(vma);
> -err_unpin:
> - if (vma->obj)
> - i915_gem_object_unpin_pages(vma->obj);
> - return ret;
> }
>
> static void
> @@ -677,12 +717,7 @@ i915_vma_remove(struct i915_vma *vma)
> GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
> GEM_BUG_ON(i915_vma_is_bound(vma, I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND));
>
> - vma->ops->clear_pages(vma);
> -
> - mutex_lock(&vma->vm->mutex);
> - drm_mm_remove_node(&vma->node);
> list_del(&vma->vm_link);
> - mutex_unlock(&vma->vm->mutex);
>
> /*
> * Since the unbound list is global, only move to that list if
> @@ -701,51 +736,211 @@ i915_vma_remove(struct i915_vma *vma)
> i915_gem_object_unpin_pages(obj);
> assert_bind_count(obj);
> }
> +
> + drm_mm_remove_node(&vma->node);
> }
>
> -int __i915_vma_do_pin(struct i915_vma *vma,
> - u64 size, u64 alignment, u64 flags)
> +static bool try_qad_pin(struct i915_vma *vma, unsigned int flags)
What does the QAD stand for?
> {
> - const unsigned int bound = atomic_read(&vma->flags);
> - int ret;
> + unsigned int bound;
> + bool pinned = true;
>
> - lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
> - GEM_BUG_ON((flags & (PIN_GLOBAL | PIN_USER)) == 0);
> - GEM_BUG_ON((flags & PIN_GLOBAL) && !i915_vma_is_ggtt(vma));
> + bound = atomic_read(&vma->flags);
> + do {
> + if (unlikely(flags & ~bound))
> + return false;
>
> - if (WARN_ON(bound & I915_VMA_PIN_OVERFLOW)) {
> - ret = -EBUSY;
> - goto err_unpin;
> + if (unlikely(bound & (I915_VMA_OVERFLOW | I915_VMA_ERROR)))
> + return false;
> +
> + if (!(bound & I915_VMA_PIN_MASK))
> + goto unpinned;
> +
> + GEM_BUG_ON(((bound + 1) & I915_VMA_PIN_MASK) == 0);
> + } while (!atomic_try_cmpxchg(&vma->flags, &bound, bound + 1));
> +
> + return true;
> +
> +unpinned:
> + /*
> + * If pin_count==0, but we are bound, check under the lock to avoid
> + * racing with a concurrent i915_vma_unbind().
> + */
> + mutex_lock(&vma->vm->mutex);
> + do {
> + if (unlikely(bound & (I915_VMA_OVERFLOW | I915_VMA_ERROR))) {
> + pinned = false;
> + break;
> + }
> +
> + if (unlikely(flags & ~bound)) {
> + pinned = false;
> + break;
> + }
> + } while (!atomic_try_cmpxchg(&vma->flags, &bound, bound + 1));
> + mutex_unlock(&vma->vm->mutex);
> +
> + return pinned;
> +}
> +
> +static int vma_get_pages(struct i915_vma *vma)
> +{
> + int err = 0;
> +
> + if (atomic_add_unless(&vma->pages_count, 1, 0))
> + return 0;
> +
> + /* Allocations ahoy! */
> + if (mutex_lock_interruptible(&vma->pages_mutex))
> + return -EINTR;
> +
> + if (!atomic_read(&vma->pages_count)) {
> + if (vma->obj) {
> + err = i915_gem_object_pin_pages(vma->obj);
> + if (err)
> + goto unlock;
> + }
> +
> + err = vma->ops->set_pages(vma);
> + if (err)
> + goto unlock;
> }
> + atomic_inc(&vma->pages_count);
>
> - if ((bound & I915_VMA_BIND_MASK) == 0) {
> - ret = i915_vma_insert(vma, size, alignment, flags);
> - if (ret)
> - goto err_unpin;
> +unlock:
> + mutex_unlock(&vma->pages_mutex);
> +
> + return err;
> +}
> +
> +static void __vma_put_pages(struct i915_vma *vma, unsigned int count)
> +{
> + /* We allocate under vma_get_pages, so beware the shrinker */
> + mutex_lock_nested(&vma->pages_mutex, SINGLE_DEPTH_NESTING);
> + GEM_BUG_ON(atomic_read(&vma->pages_count) < count);
> + if (atomic_sub_return(count, &vma->pages_count) == 0) {
> + vma->ops->clear_pages(vma);
> + GEM_BUG_ON(vma->pages);
> + if (vma->obj)
> + i915_gem_object_unpin_pages(vma->obj);
> }
> - GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
> + mutex_unlock(&vma->pages_mutex);
> +}
>
> - ret = i915_vma_bind(vma, vma->obj ? vma->obj->cache_level : 0, flags);
> - if (ret)
> - goto err_remove;
> +static void vma_put_pages(struct i915_vma *vma)
> +{
> + if (atomic_add_unless(&vma->pages_count, -1, 1))
> + return;
> +
> + __vma_put_pages(vma, 1);
> +}
> +
> +static void vma_unbind_pages(struct i915_vma *vma)
> +{
> + unsigned int count;
> +
> + lockdep_assert_held(&vma->vm->mutex);
> +
> + /* The upper portion of pages_count is the number of bindings */
> + count = atomic_read(&vma->pages_count);
> + count >>= I915_VMA_PAGES_BIAS;
> + GEM_BUG_ON(!count);
> +
> + __vma_put_pages(vma, count | count << I915_VMA_PAGES_BIAS);
> +}
> +
> +int i915_vma_pin(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
> +{
> + struct i915_vma_work *work = NULL;
> + unsigned int bound;
> + int err;
> +
> + BUILD_BUG_ON(PIN_GLOBAL != I915_VMA_GLOBAL_BIND);
> + BUILD_BUG_ON(PIN_USER != I915_VMA_LOCAL_BIND);
> +
> + GEM_BUG_ON(flags & PIN_UPDATE);
> + GEM_BUG_ON(!(flags & (PIN_USER | PIN_GLOBAL)));
> +
> + /* First try and grab the pin without rebinding the vma */
> + if (try_qad_pin(vma, flags & I915_VMA_BIND_MASK))
> + return 0;
> +
> + err = vma_get_pages(vma);
> + if (err)
> + return err;
> +
> + if (flags & vma->vm->bind_alloc) {
> + work = i915_vma_work();
> + if (!work) {
> + err = -ENOMEM;
> + goto err_pages;
> + }
> + }
> +
> + /* No more allocations allowed once we hold vm->mutex */
> + err = mutex_lock_interruptible(&vma->vm->mutex);
> + if (err)
> + goto err_fence;
> +
> + bound = atomic_read(&vma->flags);
> + if (unlikely(bound & I915_VMA_ERROR)) {
> + err = -ENOMEM;
> + goto err_unlock;
> + }
> +
> + if (unlikely(!((bound + 1) & I915_VMA_PIN_MASK))) {
> + err = -EAGAIN; /* pins are meant to be fairly temporary */
> + goto err_unlock;
> + }
> +
> + if (unlikely(!(flags & ~bound & I915_VMA_BIND_MASK))) {
> + __i915_vma_pin(vma);
> + goto err_unlock;
> + }
> +
> + err = i915_active_acquire(&vma->active);
> + if (err)
> + goto err_unlock;
> +
> + if (!(bound & I915_VMA_BIND_MASK)) {
> + err = i915_vma_insert(vma, size, alignment, flags);
> + if (err)
> + goto err_active;
> +
> + if (i915_is_ggtt(vma->vm))
> + __i915_vma_set_map_and_fenceable(vma);
> + }
>
> - GEM_BUG_ON(!i915_vma_is_bound(vma, I915_VMA_BIND_MASK));
> + GEM_BUG_ON(!vma->pages);
> + err = i915_vma_bind(vma,
> + vma->obj ? vma->obj->cache_level : 0,
> + flags, work);
> + if (err)
> + goto err_remove;
>
> - if ((bound ^ atomic_read(&vma->flags)) & I915_VMA_GLOBAL_BIND)
> - __i915_vma_set_map_and_fenceable(vma);
> + /* There should only be at most 2 active bindings (user, global) */
> + GEM_BUG_ON(bound + I915_VMA_PAGES_ACTIVE < bound);
> + atomic_add(I915_VMA_PAGES_ACTIVE, &vma->pages_count);
> + list_move_tail(&vma->vm_link, &vma->vm->bound_list);
>
> + __i915_vma_pin(vma);
> + GEM_BUG_ON(!i915_vma_is_pinned(vma));
> + GEM_BUG_ON(!i915_vma_is_bound(vma, flags));
> GEM_BUG_ON(i915_vma_misplaced(vma, size, alignment, flags));
> - return 0;
>
> err_remove:
> - if ((bound & I915_VMA_BIND_MASK) == 0) {
> + if (!i915_vma_is_bound(vma, I915_VMA_BIND_MASK))
> i915_vma_remove(vma);
> - GEM_BUG_ON(vma->pages);
> - GEM_BUG_ON(atomic_read(&vma->flags) & I915_VMA_BIND_MASK);
> - }
> -err_unpin:
> - __i915_vma_unpin(vma);
> - return ret;
> +err_active:
> + i915_active_release(&vma->active);
> +err_unlock:
> + mutex_unlock(&vma->vm->mutex);
> +err_fence:
> + if (work)
> + dma_fence_work_commit(&work->base);
> +err_pages:
> + vma_put_pages(vma);
> + return err;
> }
>
> void i915_vma_close(struct i915_vma *vma)
> @@ -776,9 +971,6 @@ static void __i915_vma_remove_closed(struct i915_vma *vma)
> {
> struct drm_i915_private *i915 = vma->vm->i915;
>
> - 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);
> @@ -786,40 +978,35 @@ static void __i915_vma_remove_closed(struct i915_vma *vma)
>
> void i915_vma_reopen(struct i915_vma *vma)
> {
> - __i915_vma_remove_closed(vma);
> + if (i915_vma_is_closed(vma))
> + __i915_vma_remove_closed(vma);
> }
>
> -static void __i915_vma_destroy(struct i915_vma *vma)
> +void i915_vma_destroy(struct i915_vma *vma)
> {
> - GEM_BUG_ON(drm_mm_node_allocated(&vma->node));
> - GEM_BUG_ON(vma->fence);
> + if (drm_mm_node_allocated(&vma->node)) {
> + mutex_lock(&vma->vm->mutex);
> + atomic_and(~I915_VMA_PIN_MASK, &vma->flags);
> + WARN_ON(__i915_vma_unbind(vma));
> + mutex_unlock(&vma->vm->mutex);
> + GEM_BUG_ON(drm_mm_node_allocated(&vma->node));
> + }
> + GEM_BUG_ON(i915_vma_is_active(vma));
>
> if (vma->obj) {
> struct drm_i915_gem_object *obj = vma->obj;
>
> spin_lock(&obj->vma.lock);
> list_del(&vma->obj_link);
> - rb_erase(&vma->obj_node, &vma->obj->vma.tree);
> + rb_erase(&vma->obj_node, &obj->vma.tree);
> spin_unlock(&obj->vma.lock);
> }
>
> - i915_active_fini(&vma->active);
> -
> - i915_vma_free(vma);
> -}
> -
> -void i915_vma_destroy(struct i915_vma *vma)
> -{
> - lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
> -
> - GEM_BUG_ON(i915_vma_is_pinned(vma));
> -
> __i915_vma_remove_closed(vma);
> + i915_vm_put(vma->vm);
>
> - WARN_ON(i915_vma_unbind(vma));
> - GEM_BUG_ON(i915_vma_is_active(vma));
> -
> - __i915_vma_destroy(vma);
> + i915_active_fini(&vma->active);
> + i915_vma_free(vma);
> }
>
> void i915_vma_parked(struct drm_i915_private *i915)
> @@ -828,12 +1015,32 @@ void i915_vma_parked(struct drm_i915_private *i915)
>
> spin_lock_irq(&i915->gt.closed_lock);
> list_for_each_entry_safe(vma, next, &i915->gt.closed_vma, closed_link) {
> - list_del_init(&vma->closed_link);
> + struct drm_i915_gem_object *obj = vma->obj;
> + struct i915_address_space *vm = vma->vm;
> +
> + /* XXX All to avoid keeping a reference on i915_vma itself */
> +
> + if (!kref_get_unless_zero(&obj->base.refcount))
> + continue;
> +
> + if (!i915_vm_tryopen(vm)) {
> + i915_gem_object_put(obj);
> + obj = NULL;
> + }
> +
> spin_unlock_irq(&i915->gt.closed_lock);
>
> - i915_vma_destroy(vma);
> + if (obj) {
> + i915_vma_destroy(vma);
> + i915_gem_object_put(obj);
> + }
>
> + i915_vm_close(vm);
> +
> + /* Restart after dropping lock */
> spin_lock_irq(&i915->gt.closed_lock);
> + next = list_first_entry(&i915->gt.closed_vma,
> + typeof(*next), closed_link);
> }
> spin_unlock_irq(&i915->gt.closed_lock);
> }
> @@ -873,6 +1080,20 @@ void i915_vma_revoke_mmap(struct i915_vma *vma)
> list_del(&vma->obj->userfault_link);
> }
>
> +int __i915_vma_move_to_active(struct i915_vma *vma, struct i915_request *rq)
> +{
> + int err;
> +
> + GEM_BUG_ON(!i915_vma_is_pinned(vma));
> +
> + /* Wait for the vma to be bound before we start! */
> + err = i915_request_await_active(rq, &vma->active);
> + if (err)
> + return err;
> +
> + return i915_active_add_request(&vma->active, rq);
> +}
> +
> int i915_vma_move_to_active(struct i915_vma *vma,
> struct i915_request *rq,
> unsigned int flags)
> @@ -880,19 +1101,9 @@ int i915_vma_move_to_active(struct i915_vma *vma,
> struct drm_i915_gem_object *obj = vma->obj;
> int err;
>
> - assert_vma_held(vma);
> assert_object_held(obj);
> - GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
>
> - /*
> - * Add a reference if we're newly entering the active list.
> - * The order in which we add operations to the retirement queue is
> - * vital here: mark_active adds to the start of the callback list,
> - * such that subsequent callbacks are called first. Therefore we
> - * add the active reference first and queue for it to be dropped
> - * *last*.
> - */
> - err = i915_active_add_request(&vma->active, rq);
> + err = __i915_vma_move_to_active(vma, rq);
> if (unlikely(err))
> return err;
>
> @@ -918,38 +1129,23 @@ int i915_vma_move_to_active(struct i915_vma *vma,
> return 0;
> }
>
> -int i915_vma_unbind(struct i915_vma *vma)
> +int __i915_vma_unbind(struct i915_vma *vma)
> {
> int ret;
>
> - lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
> + lockdep_assert_held(&vma->vm->mutex);
>
> /*
> * First wait upon any activity as retiring the request may
> * have side-effects such as unpinning or even unbinding this vma.
> + *
> + * XXX Actually waiting under the vm->mutex is a hinderance and
> + * should be pipelined wherever possible. In cases where that is
> + * unavoidable, we should lift the wait to before the mutex.
> */
> - might_sleep();
> - if (i915_vma_is_active(vma)) {
> - /*
> - * When a closed VMA is retired, it is unbound - eek.
> - * In order to prevent it from being recursively closed,
> - * take a pin on the vma so that the second unbind is
> - * aborted.
> - *
> - * Even more scary is that the retire callback may free
> - * the object (last active vma). To prevent the explosion
> - * we defer the actual object free to a worker that can
> - * only proceed once it acquires the struct_mutex (which
> - * we currently hold, therefore it cannot free this object
> - * before we are finished).
> - */
> - __i915_vma_pin(vma);
> - ret = i915_active_wait(&vma->active);
> - __i915_vma_unpin(vma);
> - if (ret)
> - return ret;
> - }
> - GEM_BUG_ON(i915_vma_is_active(vma));
> + ret = i915_vma_sync(vma);
> + if (ret)
> + return ret;
>
> if (i915_vma_is_pinned(vma)) {
> vma_print_allocator(vma, "is pinned");
> @@ -970,16 +1166,12 @@ int i915_vma_unbind(struct i915_vma *vma)
> GEM_BUG_ON(i915_vma_has_ggtt_write(vma));
>
> /* release the fence reg _after_ flushing */
> - mutex_lock(&vma->vm->mutex);
> ret = i915_vma_revoke_fence(vma);
> - mutex_unlock(&vma->vm->mutex);
> if (ret)
> return ret;
>
> /* Force a pagefault for domain tracking on next user access */
> - mutex_lock(&vma->vm->mutex);
> i915_vma_revoke_mmap(vma);
> - mutex_unlock(&vma->vm->mutex);
>
> __i915_vma_iounmap(vma);
> clear_bit(I915_VMA_CAN_FENCE_BIT, __i915_vma_flags(vma));
> @@ -987,17 +1179,33 @@ int i915_vma_unbind(struct i915_vma *vma)
> GEM_BUG_ON(vma->fence);
> GEM_BUG_ON(i915_vma_has_userfault(vma));
>
> - if (likely(!vma->vm->closed)) {
> + if (likely(atomic_read(&vma->vm->open))) {
> trace_i915_vma_unbind(vma);
> vma->ops->unbind_vma(vma);
> }
> - atomic_and(~I915_VMA_BIND_MASK, &vma->flags);
> + atomic_and(~(I915_VMA_BIND_MASK | I915_VMA_ERROR), &vma->flags);
>
> + vma_unbind_pages(vma);
> i915_vma_remove(vma);
>
> return 0;
> }
>
> +int i915_vma_unbind(struct i915_vma *vma)
> +{
> + struct i915_address_space *vm = vma->vm;
> + int err;
> +
> + err = mutex_lock_interruptible(&vm->mutex);
> + if (err)
> + return err;
> +
> + err = __i915_vma_unbind(vma);
> + mutex_unlock(&vm->mutex);
> +
> + return err;
> +}
> +
> struct i915_vma *i915_vma_make_unshrinkable(struct i915_vma *vma)
> {
> i915_gem_object_make_unshrinkable(vma->obj);
> diff --git a/drivers/gpu/drm/i915/i915_vma.h b/drivers/gpu/drm/i915/i915_vma.h
> index e49b199f7de7..858908e3d1cc 100644
> --- a/drivers/gpu/drm/i915/i915_vma.h
> +++ b/drivers/gpu/drm/i915/i915_vma.h
> @@ -96,25 +96,28 @@ struct i915_vma {
> * exclusive cachelines of a single page, so a maximum of 64 possible
> * users.
> */
> -#define I915_VMA_PIN_MASK 0xff
> -#define I915_VMA_PIN_OVERFLOW_BIT 8
> -#define I915_VMA_PIN_OVERFLOW ((int)BIT(I915_VMA_PIN_OVERFLOW_BIT))
> +#define I915_VMA_PIN_MASK 0x3ff
> +#define I915_VMA_OVERFLOW 0x200
>
> /** Flags and address space this VMA is bound to */
> -#define I915_VMA_GLOBAL_BIND_BIT 9
> -#define I915_VMA_LOCAL_BIND_BIT 10
> +#define I915_VMA_GLOBAL_BIND_BIT 10
> +#define I915_VMA_LOCAL_BIND_BIT 11
>
> #define I915_VMA_GLOBAL_BIND ((int)BIT(I915_VMA_GLOBAL_BIND_BIT))
> #define I915_VMA_LOCAL_BIND ((int)BIT(I915_VMA_LOCAL_BIND_BIT))
>
> -#define I915_VMA_BIND_MASK (I915_VMA_GLOBAL_BIND | \
> - I915_VMA_LOCAL_BIND | \
> - I915_VMA_PIN_OVERFLOW)
> +#define I915_VMA_BIND_MASK (I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND)
>
> -#define I915_VMA_GGTT_BIT 11
> -#define I915_VMA_CAN_FENCE_BIT 12
> -#define I915_VMA_USERFAULT_BIT 13
> -#define I915_VMA_GGTT_WRITE_BIT 14
> +#define I915_VMA_ALLOC_BIT 12
> +#define I915_VMA_ALLOC ((int)BIT(I915_VMA_ALLOC_BIT))
> +
> +#define I915_VMA_ERROR_BIT 13
> +#define I915_VMA_ERROR ((int)BIT(I915_VMA_ERROR_BIT))
> +
> +#define I915_VMA_GGTT_BIT 14
> +#define I915_VMA_CAN_FENCE_BIT 15
> +#define I915_VMA_USERFAULT_BIT 16
> +#define I915_VMA_GGTT_WRITE_BIT 17
>
> #define I915_VMA_GGTT ((int)BIT(I915_VMA_GGTT_BIT))
> #define I915_VMA_CAN_FENCE ((int)BIT(I915_VMA_CAN_FENCE_BIT))
> @@ -123,6 +126,11 @@ struct i915_vma {
>
> struct i915_active active;
>
> +#define I915_VMA_PAGES_BIAS 24
> +#define I915_VMA_PAGES_ACTIVE (BIT(24) | 1)
> + atomic_t pages_count; /* number of active binds to the pages */
> + struct mutex pages_mutex; /* protect acquire/release of backing pages */
> +
> /**
> * Support different GGTT views into the same object.
> * This means there can be multiple VMA mappings per object and per VM.
> @@ -169,6 +177,8 @@ static inline bool i915_vma_is_active(const struct i915_vma *vma)
> return !i915_active_is_idle(&vma->active);
> }
>
> +int __must_check __i915_vma_move_to_active(struct i915_vma *vma,
> + struct i915_request *rq);
> int __must_check i915_vma_move_to_active(struct i915_vma *vma,
> struct i915_request *rq,
> unsigned int flags);
> @@ -307,13 +317,18 @@ i915_vma_compare(struct i915_vma *vma,
> return memcmp(&vma->ggtt_view.partial, &view->partial, view->type);
> }
>
> -int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level,
> - u32 flags);
> +struct i915_vma_work *i915_vma_work(void);
> +int i915_vma_bind(struct i915_vma *vma,
> + enum i915_cache_level cache_level,
> + u32 flags,
> + struct i915_vma_work *work);
> +
> bool i915_gem_valid_gtt_space(struct i915_vma *vma, unsigned long color);
> bool i915_vma_misplaced(const struct i915_vma *vma,
> u64 size, u64 alignment, u64 flags);
> void __i915_vma_set_map_and_fenceable(struct i915_vma *vma);
> void i915_vma_revoke_mmap(struct i915_vma *vma);
> +int __i915_vma_unbind(struct i915_vma *vma);
> int __must_check i915_vma_unbind(struct i915_vma *vma);
> void i915_vma_unlink_ctx(struct i915_vma *vma);
> void i915_vma_close(struct i915_vma *vma);
> @@ -332,26 +347,8 @@ static inline void i915_vma_unlock(struct i915_vma *vma)
> dma_resv_unlock(vma->resv);
> }
>
> -int __i915_vma_do_pin(struct i915_vma *vma,
> - u64 size, u64 alignment, u64 flags);
> -static inline int __must_check
> -i915_vma_pin(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
> -{
> - BUILD_BUG_ON(PIN_MBZ != I915_VMA_PIN_OVERFLOW);
> - BUILD_BUG_ON(PIN_GLOBAL != I915_VMA_GLOBAL_BIND);
> - BUILD_BUG_ON(PIN_USER != I915_VMA_LOCAL_BIND);
> -
> - /* Pin early to prevent the shrinker/eviction logic from destroying
> - * our vma as we insert and bind.
> - */
> - if (likely(((atomic_inc_return(&vma->flags) ^ flags) & I915_VMA_BIND_MASK) == 0)) {
> - GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
> - GEM_BUG_ON(i915_vma_misplaced(vma, size, alignment, flags));
> - return 0;
> - }
> -
> - return __i915_vma_do_pin(vma, size, alignment, flags);
> -}
> +int __must_check
> +i915_vma_pin(struct i915_vma *vma, u64 size, u64 alignment, u64 flags);
>
> static inline int i915_vma_pin_count(const struct i915_vma *vma)
> {
> @@ -366,17 +363,17 @@ static inline bool i915_vma_is_pinned(const struct i915_vma *vma)
> static inline void __i915_vma_pin(struct i915_vma *vma)
> {
> atomic_inc(&vma->flags);
> - GEM_BUG_ON(atomic_read(&vma->flags) & I915_VMA_PIN_OVERFLOW);
> + GEM_BUG_ON(!i915_vma_is_pinned(vma));
> }
>
> static inline void __i915_vma_unpin(struct i915_vma *vma)
> {
> + GEM_BUG_ON(!i915_vma_is_pinned(vma));
> atomic_dec(&vma->flags);
> }
>
> static inline void i915_vma_unpin(struct i915_vma *vma)
> {
> - GEM_BUG_ON(!i915_vma_is_pinned(vma));
> GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
> __i915_vma_unpin(vma);
> }
> @@ -402,8 +399,6 @@ static inline bool i915_node_color_differs(const struct drm_mm_node *node,
> * the caller must call i915_vma_unpin_iomap to relinquish the pinning
> * after the iomapping is no longer required.
> *
> - * Callers must hold the struct_mutex.
> - *
> * Returns a valid iomapped pointer or ERR_PTR.
> */
> void __iomem *i915_vma_pin_iomap(struct i915_vma *vma);
> @@ -415,8 +410,8 @@ void __iomem *i915_vma_pin_iomap(struct i915_vma *vma);
> *
> * Unpins the previously iomapped VMA from i915_vma_pin_iomap().
> *
> - * Callers must hold the struct_mutex. This function is only valid to be
> - * called on a VMA previously iomapped by the caller with i915_vma_pin_iomap().
> + * This function is only valid to be called on a VMA previously
> + * iomapped by the caller with i915_vma_pin_iomap().
> */
> void i915_vma_unpin_iomap(struct i915_vma *vma);
>
> @@ -444,6 +439,8 @@ static inline struct page *i915_vma_first_page(struct i915_vma *vma)
> int __must_check i915_vma_pin_fence(struct i915_vma *vma);
> int __must_check i915_vma_revoke_fence(struct i915_vma *vma);
>
> +int __i915_vma_pin_fence(struct i915_vma *vma);
> +
> static inline void __i915_vma_unpin_fence(struct i915_vma *vma)
> {
> GEM_BUG_ON(atomic_read(&vma->fence->pin_count) <= 0);
> @@ -461,7 +458,6 @@ static inline void __i915_vma_unpin_fence(struct i915_vma *vma)
> static inline void
> i915_vma_unpin_fence(struct i915_vma *vma)
> {
> - /* lockdep_assert_held(&vma->vm->i915->drm.struct_mutex); */
> if (vma->fence)
> __i915_vma_unpin_fence(vma);
> }
> @@ -490,4 +486,10 @@ struct i915_vma *i915_vma_make_unshrinkable(struct i915_vma *vma);
> void i915_vma_make_shrinkable(struct i915_vma *vma);
> void i915_vma_make_purgeable(struct i915_vma *vma);
>
> +static inline int i915_vma_sync(struct i915_vma *vma)
> +{
> + /* Wait for the asynchronous bindings and pending GPU reads */
> + return i915_active_wait(&vma->active);
> +}
> +
> #endif
> diff --git a/drivers/gpu/drm/i915/selftests/i915_gem.c b/drivers/gpu/drm/i915/selftests/i915_gem.c
> index 37593831b539..0346c3e5b6b6 100644
> --- a/drivers/gpu/drm/i915/selftests/i915_gem.c
> +++ b/drivers/gpu/drm/i915/selftests/i915_gem.c
> @@ -119,10 +119,8 @@ static void pm_resume(struct drm_i915_private *i915)
> intel_gt_sanitize(&i915->gt, false);
> i915_gem_sanitize(i915);
>
> - mutex_lock(&i915->drm.struct_mutex);
> i915_gem_restore_gtt_mappings(i915);
> i915_gem_restore_fences(i915);
> - mutex_unlock(&i915->drm.struct_mutex);
>
> i915_gem_resume(i915);
> }
> diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
> index 2905fb21d866..75a4695b82bb 100644
> --- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
> +++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
> @@ -106,14 +106,11 @@ static int populate_ggtt(struct drm_i915_private *i915,
>
> static void unpin_ggtt(struct drm_i915_private *i915)
> {
> - struct i915_ggtt *ggtt = &i915->ggtt;
> struct i915_vma *vma;
>
> - mutex_lock(&ggtt->vm.mutex);
> list_for_each_entry(vma, &i915->ggtt.vm.bound_list, vm_link)
> if (vma->obj->mm.quirked)
> i915_vma_unpin(vma);
> - mutex_unlock(&ggtt->vm.mutex);
> }
>
> static void cleanup_objects(struct drm_i915_private *i915,
> @@ -127,11 +124,7 @@ static void cleanup_objects(struct drm_i915_private *i915,
> i915_gem_object_put(obj);
> }
>
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> i915_gem_drain_freed_objects(i915);
> -
> - mutex_lock(&i915->drm.struct_mutex);
> }
>
> static int igt_evict_something(void *arg)
> @@ -148,10 +141,12 @@ static int igt_evict_something(void *arg)
> goto cleanup;
>
> /* Everything is pinned, nothing should happen */
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_evict_something(&ggtt->vm,
> I915_GTT_PAGE_SIZE, 0, 0,
> 0, U64_MAX,
> 0);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err != -ENOSPC) {
> pr_err("i915_gem_evict_something failed on a full GGTT with err=%d\n",
> err);
> @@ -161,10 +156,12 @@ static int igt_evict_something(void *arg)
> unpin_ggtt(i915);
>
> /* Everything is unpinned, we should be able to evict something */
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_evict_something(&ggtt->vm,
> I915_GTT_PAGE_SIZE, 0, 0,
> 0, U64_MAX,
> 0);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err) {
> pr_err("i915_gem_evict_something failed on a full GGTT with err=%d\n",
> err);
> @@ -230,7 +227,9 @@ static int igt_evict_for_vma(void *arg)
> goto cleanup;
>
> /* Everything is pinned, nothing should happen */
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_evict_for_node(&ggtt->vm, &target, 0);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err != -ENOSPC) {
> pr_err("i915_gem_evict_for_node on a full GGTT returned err=%d\n",
> err);
> @@ -240,7 +239,9 @@ static int igt_evict_for_vma(void *arg)
> unpin_ggtt(i915);
>
> /* Everything is unpinned, we should be able to evict the node */
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_evict_for_node(&ggtt->vm, &target, 0);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err) {
> pr_err("i915_gem_evict_for_node returned err=%d\n",
> err);
> @@ -319,7 +320,9 @@ static int igt_evict_for_cache_color(void *arg)
> i915_vma_unpin(vma);
>
> /* Remove just the second vma */
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_evict_for_node(&ggtt->vm, &target, 0);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err) {
> pr_err("[0]i915_gem_evict_for_node returned err=%d\n", err);
> goto cleanup;
> @@ -330,7 +333,9 @@ static int igt_evict_for_cache_color(void *arg)
> */
> target.color = I915_CACHE_L3_LLC;
>
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_evict_for_node(&ggtt->vm, &target, 0);
> + mutex_unlock(&ggtt->vm.mutex);
> if (!err) {
> pr_err("[1]i915_gem_evict_for_node returned err=%d\n", err);
> err = -EINVAL;
> @@ -360,7 +365,9 @@ static int igt_evict_vm(void *arg)
> goto cleanup;
>
> /* Everything is pinned, nothing should happen */
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_evict_vm(&ggtt->vm);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err) {
> pr_err("i915_gem_evict_vm on a full GGTT returned err=%d]\n",
> err);
> @@ -369,7 +376,9 @@ static int igt_evict_vm(void *arg)
>
> unpin_ggtt(i915);
>
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_evict_vm(&ggtt->vm);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err) {
> pr_err("i915_gem_evict_vm on a full GGTT returned err=%d]\n",
> err);
> @@ -410,11 +419,11 @@ static int igt_evict_contexts(void *arg)
> if (!HAS_FULL_PPGTT(i915))
> return 0;
>
> - mutex_lock(&i915->drm.struct_mutex);
> wakeref = intel_runtime_pm_get(&i915->runtime_pm);
>
> /* Reserve a block so that we know we have enough to fit a few rq */
> memset(&hole, 0, sizeof(hole));
> + mutex_lock(&i915->ggtt.vm.mutex);
> err = i915_gem_gtt_insert(&i915->ggtt.vm, &hole,
> PRETEND_GGTT_SIZE, 0, I915_COLOR_UNEVICTABLE,
> 0, i915->ggtt.vm.total,
> @@ -427,7 +436,9 @@ static int igt_evict_contexts(void *arg)
> do {
> struct reserved *r;
>
> + mutex_unlock(&i915->ggtt.vm.mutex);
> r = kcalloc(1, sizeof(*r), GFP_KERNEL);
> + mutex_lock(&i915->ggtt.vm.mutex);
> if (!r) {
> err = -ENOMEM;
> goto out_locked;
> @@ -447,7 +458,7 @@ static int igt_evict_contexts(void *arg)
> count++;
> } while (1);
> drm_mm_remove_node(&hole);
> - mutex_unlock(&i915->drm.struct_mutex);
> + mutex_unlock(&i915->ggtt.vm.mutex);
> pr_info("Filled GGTT with %lu 1MiB nodes\n", count);
>
> /* Overfill the GGTT with context objects and so try to evict one. */
> @@ -510,7 +521,7 @@ static int igt_evict_contexts(void *arg)
> break;
> }
>
> - mutex_lock(&i915->drm.struct_mutex);
> + mutex_lock(&i915->ggtt.vm.mutex);
> out_locked:
> if (igt_flush_test(i915, I915_WAIT_LOCKED))
> err = -EIO;
> @@ -524,8 +535,8 @@ static int igt_evict_contexts(void *arg)
> }
> if (drm_mm_node_allocated(&hole))
> drm_mm_remove_node(&hole);
> + mutex_unlock(&i915->ggtt.vm.mutex);
> intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
>
> return err;
> }
> @@ -547,12 +558,9 @@ int i915_gem_evict_mock_selftests(void)
> if (!i915)
> return -ENOMEM;
>
> - mutex_lock(&i915->drm.struct_mutex);
> with_intel_runtime_pm(&i915->runtime_pm, wakeref)
> err = i915_subtests(tests, i915);
>
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> drm_dev_put(&i915->drm);
> return err;
> }
> diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
> index 84d94eabc7dd..373fbd58a17a 100644
> --- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
> +++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
> @@ -38,16 +38,7 @@
>
> static void cleanup_freed_objects(struct drm_i915_private *i915)
> {
> - /*
> - * As we may hold onto the struct_mutex for inordinate lengths of
> - * time, the NMI khungtaskd detector may fire for the free objects
> - * worker.
> - */
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> i915_gem_drain_freed_objects(i915);
> -
> - mutex_lock(&i915->drm.struct_mutex);
> }
>
> static void fake_free_pages(struct drm_i915_gem_object *obj,
> @@ -880,6 +871,15 @@ static int __shrink_hole(struct drm_i915_private *i915,
> i915_vma_unpin(vma);
> addr += size;
>
> + /*
> + * Since we are injecting allocation faults at random intervals,
> + * wait for this allocation to complete before we change the
> + * faultinjection.
> + */
> + err = i915_vma_sync(vma);
> + if (err)
> + break;
> +
> if (igt_timeout(end_time,
> "%s timed out at ofset %llx [%llx - %llx]\n",
> __func__, addr, hole_start, hole_end)) {
> @@ -1013,21 +1013,19 @@ static int exercise_ppgtt(struct drm_i915_private *dev_priv,
> if (IS_ERR(file))
> return PTR_ERR(file);
>
> - mutex_lock(&dev_priv->drm.struct_mutex);
> ppgtt = i915_ppgtt_create(dev_priv);
> if (IS_ERR(ppgtt)) {
> err = PTR_ERR(ppgtt);
> - goto out_unlock;
> + goto out_free;
> }
> GEM_BUG_ON(offset_in_page(ppgtt->vm.total));
> - GEM_BUG_ON(ppgtt->vm.closed);
> + GEM_BUG_ON(!atomic_read(&ppgtt->vm.open));
>
> err = func(dev_priv, &ppgtt->vm, 0, ppgtt->vm.total, end_time);
>
> i915_vm_put(&ppgtt->vm);
> -out_unlock:
> - mutex_unlock(&dev_priv->drm.struct_mutex);
>
> +out_free:
> mock_file_free(dev_priv, file);
> return err;
> }
> @@ -1090,7 +1088,6 @@ static int exercise_ggtt(struct drm_i915_private *i915,
> IGT_TIMEOUT(end_time);
> int err = 0;
>
> - mutex_lock(&i915->drm.struct_mutex);
> restart:
> list_sort(NULL, &ggtt->vm.mm.hole_stack, sort_holes);
> drm_mm_for_each_hole(node, &ggtt->vm.mm, hole_start, hole_end) {
> @@ -1111,7 +1108,6 @@ static int exercise_ggtt(struct drm_i915_private *i915,
> last = hole_end;
> goto restart;
> }
> - mutex_unlock(&i915->drm.struct_mutex);
>
> return err;
> }
> @@ -1153,13 +1149,9 @@ static int igt_ggtt_page(void *arg)
> unsigned int *order, n;
> int err;
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
> - if (IS_ERR(obj)) {
> - err = PTR_ERR(obj);
> - goto out_unlock;
> - }
> + if (IS_ERR(obj))
> + return PTR_ERR(obj);
>
> err = i915_gem_object_pin_pages(obj);
> if (err)
> @@ -1227,8 +1219,6 @@ static int igt_ggtt_page(void *arg)
> i915_gem_object_unpin_pages(obj);
> out_free:
> i915_gem_object_put(obj);
> -out_unlock:
> - mutex_unlock(&i915->drm.struct_mutex);
> return err;
> }
>
> @@ -1239,6 +1229,9 @@ static void track_vma_bind(struct i915_vma *vma)
> atomic_inc(&obj->bind_count); /* track for eviction later */
> __i915_gem_object_pin_pages(obj);
>
> + GEM_BUG_ON(vma->pages);
> + atomic_set(&vma->pages_count, I915_VMA_PAGES_ACTIVE);
> + __i915_gem_object_pin_pages(obj);
> vma->pages = obj->mm.pages;
>
> mutex_lock(&vma->vm->mutex);
> @@ -1335,11 +1328,13 @@ static int igt_gtt_reserve(void *arg)
> goto out;
> }
>
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_gtt_reserve(&ggtt->vm, &vma->node,
> obj->base.size,
> total,
> obj->cache_level,
> 0);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err) {
> pr_err("i915_gem_gtt_reserve (pass 1) failed at %llu/%llu with err=%d\n",
> total, ggtt->vm.total, err);
> @@ -1385,11 +1380,13 @@ static int igt_gtt_reserve(void *arg)
> goto out;
> }
>
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_gtt_reserve(&ggtt->vm, &vma->node,
> obj->base.size,
> total,
> obj->cache_level,
> 0);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err) {
> pr_err("i915_gem_gtt_reserve (pass 2) failed at %llu/%llu with err=%d\n",
> total, ggtt->vm.total, err);
> @@ -1429,11 +1426,13 @@ static int igt_gtt_reserve(void *arg)
> 2*I915_GTT_PAGE_SIZE,
> I915_GTT_MIN_ALIGNMENT);
>
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_gtt_reserve(&ggtt->vm, &vma->node,
> obj->base.size,
> offset,
> obj->cache_level,
> 0);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err) {
> pr_err("i915_gem_gtt_reserve (pass 3) failed at %llu/%llu with err=%d\n",
> total, ggtt->vm.total, err);
> @@ -1502,11 +1501,13 @@ static int igt_gtt_insert(void *arg)
>
> /* Check a couple of obviously invalid requests */
> for (ii = invalid_insert; ii->size; ii++) {
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_gtt_insert(&ggtt->vm, &tmp,
> ii->size, ii->alignment,
> I915_COLOR_UNEVICTABLE,
> ii->start, ii->end,
> 0);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err != -ENOSPC) {
> pr_err("Invalid i915_gem_gtt_insert(.size=%llx, .alignment=%llx, .start=%llx, .end=%llx) succeeded (err=%d)\n",
> ii->size, ii->alignment, ii->start, ii->end,
> @@ -1542,10 +1543,12 @@ static int igt_gtt_insert(void *arg)
> goto out;
> }
>
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_gtt_insert(&ggtt->vm, &vma->node,
> obj->base.size, 0, obj->cache_level,
> 0, ggtt->vm.total,
> 0);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err == -ENOSPC) {
> /* maxed out the GGTT space */
> i915_gem_object_put(obj);
> @@ -1600,10 +1603,12 @@ static int igt_gtt_insert(void *arg)
> goto out;
> }
>
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_gtt_insert(&ggtt->vm, &vma->node,
> obj->base.size, 0, obj->cache_level,
> 0, ggtt->vm.total,
> 0);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err) {
> pr_err("i915_gem_gtt_insert (pass 2) failed at %llu/%llu with err=%d\n",
> total, ggtt->vm.total, err);
> @@ -1647,10 +1652,12 @@ static int igt_gtt_insert(void *arg)
> goto out;
> }
>
> + mutex_lock(&ggtt->vm.mutex);
> err = i915_gem_gtt_insert(&ggtt->vm, &vma->node,
> obj->base.size, 0, obj->cache_level,
> 0, ggtt->vm.total,
> 0);
> + mutex_unlock(&ggtt->vm.mutex);
> if (err) {
> pr_err("i915_gem_gtt_insert (pass 3) failed at %llu/%llu with err=%d\n",
> total, ggtt->vm.total, err);
> @@ -1694,8 +1701,9 @@ int i915_gem_gtt_mock_selftests(void)
> }
> mock_init_ggtt(i915, ggtt);
>
> - mutex_lock(&i915->drm.struct_mutex);
> err = i915_subtests(tests, ggtt);
> +
> + mutex_lock(&i915->drm.struct_mutex);
> mock_device_flush(i915);
> mutex_unlock(&i915->drm.struct_mutex);
>
> diff --git a/drivers/gpu/drm/i915/selftests/i915_request.c b/drivers/gpu/drm/i915/selftests/i915_request.c
> index b3688543ed7d..6b5b03081880 100644
> --- a/drivers/gpu/drm/i915/selftests/i915_request.c
> +++ b/drivers/gpu/drm/i915/selftests/i915_request.c
> @@ -647,8 +647,15 @@ static struct i915_vma *empty_batch(struct drm_i915_private *i915)
> if (err)
> goto err;
>
> + /* Force the wait wait now to avoid including it in the benchmark */
> + err = i915_vma_sync(vma);
> + if (err)
> + goto err_pin;
> +
> return vma;
>
> +err_pin:
> + i915_vma_unpin(vma);
> err:
> i915_gem_object_put(obj);
> return ERR_PTR(err);
> diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c
> index 97752deecccb..0e4f66312b39 100644
> --- a/drivers/gpu/drm/i915/selftests/i915_vma.c
> +++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
> @@ -831,8 +831,9 @@ int i915_vma_mock_selftests(void)
> }
> mock_init_ggtt(i915, ggtt);
>
> - mutex_lock(&i915->drm.struct_mutex);
> err = i915_subtests(tests, ggtt);
> +
> + mutex_lock(&i915->drm.struct_mutex);
> mock_device_flush(i915);
> mutex_unlock(&i915->drm.struct_mutex);
>
> @@ -879,8 +880,6 @@ static int igt_vma_remapped_gtt(void *arg)
> if (IS_ERR(obj))
> return PTR_ERR(obj);
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> wakeref = intel_runtime_pm_get(&i915->runtime_pm);
>
> for (t = types; *t; t++) {
> @@ -976,7 +975,6 @@ static int igt_vma_remapped_gtt(void *arg)
>
> out:
> intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> i915_gem_object_put(obj);
>
> return err;
>
High level concept looks okay to me, and the low level details as well.
But it is so huge that I cannot guarantee I haven't missed something in
the middle. So for any hypothetical missed thing in that last third, I
put my faith that before release we catch it and fix it.
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] 70+ messages in thread* Re: [PATCH 05/27] drm/i915: Pull i915_vma_pin under the vm->mutex
2019-09-27 8:47 ` Tvrtko Ursulin
@ 2019-09-27 11:06 ` Chris Wilson
0 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-27 11:06 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Tvrtko Ursulin (2019-09-27 09:47:59)
>
> On 25/09/2019 11:01, Chris Wilson wrote:
> > -int __i915_vma_do_pin(struct i915_vma *vma,
> > - u64 size, u64 alignment, u64 flags)
> > +static bool try_qad_pin(struct i915_vma *vma, unsigned int flags)
>
> What does the QAD stand for?
quick and dirty
[snip]
> High level concept looks okay to me, and the low level details as well.
> But it is so huge that I cannot guarantee I haven't missed something in
> the middle. So for any hypothetical missed thing in that last third, I
> put my faith that before release we catch it and fix it.
It's such a huge seismic shift, that we will have to make again before
too long, that some little detail will have slipped through. All we can
do is to hope to capture anything that escapes the many machine months
of testing (but was it testing the right thing???) into a regression
test to make sure the next wave of changes is better.
Plus we can look forward to all the opportunities to optimise the code
all over again. :|
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* [PATCH 06/27] drm/i915: Push the i915_active.retire into a worker
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (4 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 05/27] drm/i915: Pull i915_vma_pin under the vm->mutex Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:01 ` [PATCH 07/27] drm/i915: Coordinate i915_active with its own mutex Chris Wilson
` (24 subsequent siblings)
30 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx; +Cc: Matthew Auld
As we need to use a mutex to serialise i915_active activation
(because we want to allow the callback to sleep), we need to push the
i915_active.retire into a worker callback in case we get need to retire
from an atomic context.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
---
.../gpu/drm/i915/display/intel_frontbuffer.c | 4 ++-
drivers/gpu/drm/i915/gem/i915_gem_context.c | 1 +
drivers/gpu/drm/i915/gt/intel_context.c | 2 ++
drivers/gpu/drm/i915/gt/intel_engine_pool.c | 1 +
drivers/gpu/drm/i915/gt/intel_timeline.c | 1 +
drivers/gpu/drm/i915/i915_active.c | 34 ++++++++++++++++---
drivers/gpu/drm/i915/i915_active_types.h | 13 ++++++-
drivers/gpu/drm/i915/i915_vma.c | 2 ++
drivers/gpu/drm/i915/selftests/i915_active.c | 6 ++--
9 files changed, 55 insertions(+), 9 deletions(-)
diff --git a/drivers/gpu/drm/i915/display/intel_frontbuffer.c b/drivers/gpu/drm/i915/display/intel_frontbuffer.c
index fc40dc1fdbcc..6428b8dd70d3 100644
--- a/drivers/gpu/drm/i915/display/intel_frontbuffer.c
+++ b/drivers/gpu/drm/i915/display/intel_frontbuffer.c
@@ -206,6 +206,7 @@ static int frontbuffer_active(struct i915_active *ref)
return 0;
}
+__i915_active_call
static void frontbuffer_retire(struct i915_active *ref)
{
struct intel_frontbuffer *front =
@@ -257,7 +258,8 @@ intel_frontbuffer_get(struct drm_i915_gem_object *obj)
kref_init(&front->ref);
atomic_set(&front->bits, 0);
i915_active_init(i915, &front->write,
- frontbuffer_active, frontbuffer_retire);
+ frontbuffer_active,
+ i915_active_may_sleep(frontbuffer_retire));
spin_lock(&i915->fb_tracking.lock);
if (obj->frontbuffer) {
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index 95f8e66e45db..4cd7d2ecf1d5 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -848,6 +848,7 @@ struct context_barrier_task {
void *data;
};
+__i915_active_call
static void cb_retire(struct i915_active *base)
{
struct context_barrier_task *cb = container_of(base, typeof(*cb), base);
diff --git a/drivers/gpu/drm/i915/gt/intel_context.c b/drivers/gpu/drm/i915/gt/intel_context.c
index 26cb838c272c..06fabdf205cf 100644
--- a/drivers/gpu/drm/i915/gt/intel_context.c
+++ b/drivers/gpu/drm/i915/gt/intel_context.c
@@ -138,6 +138,7 @@ static void __context_unpin_state(struct i915_vma *vma)
__i915_vma_unpin(vma);
}
+__i915_active_call
static void __intel_context_retire(struct i915_active *active)
{
struct intel_context *ce = container_of(active, typeof(*ce), active);
@@ -150,6 +151,7 @@ static void __intel_context_retire(struct i915_active *active)
intel_timeline_unpin(ce->timeline);
intel_ring_unpin(ce->ring);
+
intel_context_put(ce);
}
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pool.c b/drivers/gpu/drm/i915/gt/intel_engine_pool.c
index 97d36cca8ded..81fab101fdb4 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_pool.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_pool.c
@@ -61,6 +61,7 @@ static int pool_active(struct i915_active *ref)
return 0;
}
+__i915_active_call
static void pool_retire(struct i915_active *ref)
{
struct intel_engine_pool_node *node =
diff --git a/drivers/gpu/drm/i915/gt/intel_timeline.c b/drivers/gpu/drm/i915/gt/intel_timeline.c
index 9d436e14ea8d..653f60e78392 100644
--- a/drivers/gpu/drm/i915/gt/intel_timeline.c
+++ b/drivers/gpu/drm/i915/gt/intel_timeline.c
@@ -136,6 +136,7 @@ static void __idle_cacheline_free(struct intel_timeline_cacheline *cl)
kfree(cl);
}
+__i915_active_call
static void __cacheline_retire(struct i915_active *active)
{
struct intel_timeline_cacheline *cl =
diff --git a/drivers/gpu/drm/i915/i915_active.c b/drivers/gpu/drm/i915/i915_active.c
index 0791736a08fd..7ca066688b98 100644
--- a/drivers/gpu/drm/i915/i915_active.c
+++ b/drivers/gpu/drm/i915/i915_active.c
@@ -132,6 +132,7 @@ __active_retire(struct i915_active *ref)
bool retire = false;
lockdep_assert_held(&ref->mutex);
+ GEM_BUG_ON(i915_active_is_idle(ref));
/* return the unused nodes to our slabcache -- flushing the allocator */
if (atomic_dec_and_test(&ref->count)) {
@@ -157,6 +158,19 @@ __active_retire(struct i915_active *ref)
ref->retire(ref);
}
+static void
+active_work(struct work_struct *wrk)
+{
+ struct i915_active *ref = container_of(wrk, typeof(*ref), work);
+
+ GEM_BUG_ON(!atomic_read(&ref->count));
+ if (atomic_add_unless(&ref->count, -1, 1))
+ return;
+
+ mutex_lock(&ref->mutex);
+ __active_retire(ref);
+}
+
static void
active_retire(struct i915_active *ref)
{
@@ -164,8 +178,13 @@ active_retire(struct i915_active *ref)
if (atomic_add_unless(&ref->count, -1, 1))
return;
- /* One active may be flushed from inside the acquire of another */
- mutex_lock_nested(&ref->mutex, SINGLE_DEPTH_NESTING);
+ /* If we are inside interrupt context (fence signaling), defer */
+ if (ref->flags & I915_ACTIVE_RETIRE_SLEEPS ||
+ !mutex_trylock(&ref->mutex)) {
+ queue_work(system_unbound_wq, &ref->work);
+ return;
+ }
+
__active_retire(ref);
}
@@ -240,12 +259,16 @@ void __i915_active_init(struct drm_i915_private *i915,
void (*retire)(struct i915_active *ref),
struct lock_class_key *key)
{
+ unsigned long bits;
+
debug_active_init(ref);
ref->i915 = i915;
ref->flags = 0;
ref->active = active;
- ref->retire = retire;
+ ref->retire = ptr_unpack_bits(retire, &bits, 2);
+ if (bits & I915_ACTIVE_MAY_SLEEP)
+ ref->flags |= I915_ACTIVE_RETIRE_SLEEPS;
ref->excl = NULL;
ref->tree = RB_ROOT;
@@ -253,6 +276,7 @@ void __i915_active_init(struct drm_i915_private *i915,
init_llist_head(&ref->preallocated_barriers);
atomic_set(&ref->count, 0);
__mutex_init(&ref->mutex, "i915_active", key);
+ INIT_WORK(&ref->work, active_work);
}
static bool ____active_del_barrier(struct i915_active *ref,
@@ -504,6 +528,7 @@ int i915_active_wait(struct i915_active *ref)
if (wait_on_bit(&ref->flags, I915_ACTIVE_GRAB_BIT, TASK_KILLABLE))
return -EINTR;
+ flush_work(&ref->work);
if (!i915_active_is_idle(ref))
return -EBUSY;
@@ -544,8 +569,9 @@ int i915_request_await_active(struct i915_request *rq, struct i915_active *ref)
void i915_active_fini(struct i915_active *ref)
{
debug_active_fini(ref);
- GEM_BUG_ON(!RB_EMPTY_ROOT(&ref->tree));
GEM_BUG_ON(atomic_read(&ref->count));
+ GEM_BUG_ON(work_pending(&ref->work));
+ GEM_BUG_ON(!RB_EMPTY_ROOT(&ref->tree));
mutex_destroy(&ref->mutex);
}
#endif
diff --git a/drivers/gpu/drm/i915/i915_active_types.h b/drivers/gpu/drm/i915/i915_active_types.h
index 86e7a232ea3c..021167f0004d 100644
--- a/drivers/gpu/drm/i915/i915_active_types.h
+++ b/drivers/gpu/drm/i915/i915_active_types.h
@@ -13,6 +13,9 @@
#include <linux/mutex.h>
#include <linux/rbtree.h>
#include <linux/rcupdate.h>
+#include <linux/workqueue.h>
+
+#include "i915_utils.h"
struct drm_i915_private;
struct i915_active_request;
@@ -44,6 +47,11 @@ struct i915_active_request {
struct active_node;
+#define I915_ACTIVE_MAY_SLEEP BIT(0)
+
+#define __i915_active_call __aligned(4)
+#define i915_active_may_sleep(fn) ptr_pack_bits(&(fn), I915_ACTIVE_MAY_SLEEP, 2)
+
struct i915_active {
struct drm_i915_private *i915;
@@ -57,11 +65,14 @@ struct i915_active {
struct dma_fence_cb excl_cb;
unsigned long flags;
-#define I915_ACTIVE_GRAB_BIT 0
+#define I915_ACTIVE_RETIRE_SLEEPS BIT(0)
+#define I915_ACTIVE_GRAB_BIT 1
int (*active)(struct i915_active *ref);
void (*retire)(struct i915_active *ref);
+ struct work_struct work;
+
struct llist_head preallocated_barriers;
};
diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
index defeb9bb1daa..2717441a9f3a 100644
--- a/drivers/gpu/drm/i915/i915_vma.c
+++ b/drivers/gpu/drm/i915/i915_vma.c
@@ -91,6 +91,7 @@ static int __i915_vma_active(struct i915_active *ref)
return i915_vma_tryget(active_to_vma(ref)) ? 0 : -ENOENT;
}
+__i915_active_call
static void __i915_vma_retire(struct i915_active *ref)
{
i915_vma_put(active_to_vma(ref));
@@ -1152,6 +1153,7 @@ int __i915_vma_unbind(struct i915_vma *vma)
return -EBUSY;
}
+ GEM_BUG_ON(i915_vma_is_active(vma));
if (!drm_mm_node_allocated(&vma->node))
return 0;
diff --git a/drivers/gpu/drm/i915/selftests/i915_active.c b/drivers/gpu/drm/i915/selftests/i915_active.c
index afecfa081ff4..a41785822ed9 100644
--- a/drivers/gpu/drm/i915/selftests/i915_active.c
+++ b/drivers/gpu/drm/i915/selftests/i915_active.c
@@ -121,7 +121,7 @@ __live_active_setup(struct drm_i915_private *i915)
}
i915_active_release(&active->base);
- if (active->retired && count) {
+ if (READ_ONCE(active->retired) && count) {
pr_err("i915_active retired before submission!\n");
err = -EINVAL;
}
@@ -161,7 +161,7 @@ static int live_active_wait(void *arg)
}
i915_active_wait(&active->base);
- if (!active->retired) {
+ if (!READ_ONCE(active->retired)) {
pr_err("i915_active not retired after waiting!\n");
err = -EINVAL;
}
@@ -200,7 +200,7 @@ static int live_active_retire(void *arg)
if (igt_flush_test(i915, I915_WAIT_LOCKED))
err = -EIO;
- if (!active->retired) {
+ if (!READ_ONCE(active->retired)) {
pr_err("i915_active not retired after flushing!\n");
err = -EINVAL;
}
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* [PATCH 07/27] drm/i915: Coordinate i915_active with its own mutex
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (5 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 06/27] drm/i915: Push the i915_active.retire into a worker Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-27 11:10 ` Tvrtko Ursulin
2019-09-25 10:01 ` [PATCH 08/27] drm/i915: Move idle barrier cleanup into engine-pm Chris Wilson
` (23 subsequent siblings)
30 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
Forgo the struct_mutex serialisation for i915_active, and interpose its
own mutex handling for active/retire.
This is a multi-layered sleight-of-hand. First, we had to ensure that no
active/retire callbacks accidentally inverted the mutex ordering rules,
nor assumed that they were themselves serialised by struct_mutex. More
challenging though, is the rule over updating elements of the active
rbtree. Instead of the whole i915_active now being serialised by
struct_mutex, allocations/rotations of the tree are serialised by the
i915_active.mutex and individual nodes are serialised by the caller
using the i915_timeline.mutex (we need to use nested spinlocks to
interact with the dma_fence callback lists).
The pain point here is that instead of a single mutex around execbuf, we
now have to take a mutex for active tracker (one for each vma, context,
etc) and a couple of spinlocks for each fence update. The improvement in
fine grained locking allowing for multiple concurrent clients
(eventually!) should be worth it in typical loads.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
.../gpu/drm/i915/display/intel_frontbuffer.c | 2 +-
drivers/gpu/drm/i915/display/intel_overlay.c | 3 +-
drivers/gpu/drm/i915/gem/i915_gem_context.c | 4 +-
.../gpu/drm/i915/gem/i915_gem_object_types.h | 1 +
drivers/gpu/drm/i915/gem/i915_gem_pm.c | 9 +-
drivers/gpu/drm/i915/gt/intel_context.c | 4 +-
drivers/gpu/drm/i915/gt/intel_engine_pool.c | 2 +-
drivers/gpu/drm/i915/gt/intel_reset.c | 10 +-
drivers/gpu/drm/i915/gt/intel_timeline.c | 7 +-
.../gpu/drm/i915/gt/intel_timeline_types.h | 2 +-
drivers/gpu/drm/i915/gt/selftest_context.c | 16 +-
drivers/gpu/drm/i915/gt/selftest_lrc.c | 10 +-
.../gpu/drm/i915/gt/selftests/mock_timeline.c | 2 +-
drivers/gpu/drm/i915/gvt/scheduler.c | 3 -
drivers/gpu/drm/i915/i915_active.c | 289 +++++++---------
drivers/gpu/drm/i915/i915_active.h | 319 ++++--------------
drivers/gpu/drm/i915/i915_active_types.h | 23 +-
drivers/gpu/drm/i915/i915_gem.c | 42 ++-
drivers/gpu/drm/i915/i915_gem_gtt.c | 3 +-
drivers/gpu/drm/i915/i915_gpu_error.c | 4 +-
drivers/gpu/drm/i915/i915_request.c | 38 +--
drivers/gpu/drm/i915/i915_request.h | 1 -
drivers/gpu/drm/i915/i915_vma.c | 4 +-
drivers/gpu/drm/i915/selftests/i915_active.c | 32 +-
24 files changed, 264 insertions(+), 566 deletions(-)
diff --git a/drivers/gpu/drm/i915/display/intel_frontbuffer.c b/drivers/gpu/drm/i915/display/intel_frontbuffer.c
index 6428b8dd70d3..84b164f31895 100644
--- a/drivers/gpu/drm/i915/display/intel_frontbuffer.c
+++ b/drivers/gpu/drm/i915/display/intel_frontbuffer.c
@@ -257,7 +257,7 @@ intel_frontbuffer_get(struct drm_i915_gem_object *obj)
front->obj = obj;
kref_init(&front->ref);
atomic_set(&front->bits, 0);
- i915_active_init(i915, &front->write,
+ i915_active_init(&front->write,
frontbuffer_active,
i915_active_may_sleep(frontbuffer_retire));
diff --git a/drivers/gpu/drm/i915/display/intel_overlay.c b/drivers/gpu/drm/i915/display/intel_overlay.c
index 3f4ac1ee7668..e12e1a753af0 100644
--- a/drivers/gpu/drm/i915/display/intel_overlay.c
+++ b/drivers/gpu/drm/i915/display/intel_overlay.c
@@ -1360,8 +1360,7 @@ void intel_overlay_setup(struct drm_i915_private *dev_priv)
overlay->contrast = 75;
overlay->saturation = 146;
- i915_active_init(dev_priv,
- &overlay->last_flip,
+ i915_active_init(&overlay->last_flip,
NULL, intel_overlay_last_flip_retire);
ret = get_registers(overlay, OVERLAY_NEEDS_PHYSICAL(dev_priv));
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index 4cd7d2ecf1d5..9d85aab68d34 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -868,20 +868,18 @@ static int context_barrier_task(struct i915_gem_context *ctx,
void (*task)(void *data),
void *data)
{
- struct drm_i915_private *i915 = ctx->i915;
struct context_barrier_task *cb;
struct i915_gem_engines_iter it;
struct intel_context *ce;
int err = 0;
- lockdep_assert_held(&i915->drm.struct_mutex);
GEM_BUG_ON(!task);
cb = kmalloc(sizeof(*cb), GFP_KERNEL);
if (!cb)
return -ENOMEM;
- i915_active_init(i915, &cb->base, NULL, cb_retire);
+ i915_active_init(&cb->base, NULL, cb_retire);
err = i915_active_acquire(&cb->base);
if (err) {
kfree(cb);
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 d695f187b790..e291ddfeb7dd 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
@@ -8,6 +8,7 @@
#define __I915_GEM_OBJECT_TYPES_H__
#include <drm/drm_gem.h>
+#include <uapi/drm/i915_drm.h>
#include "i915_active.h"
#include "i915_selftest.h"
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
index a11ad4d914ca..3a47e1190ce1 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
@@ -16,14 +16,11 @@ 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 =
+ struct dma_fence_cb *cb =
container_of((struct list_head *)node,
- typeof(*active), link);
+ typeof(*cb), node);
- INIT_LIST_HEAD(&active->link);
- RCU_INIT_POINTER(active->request, NULL);
-
- active->retire(active, NULL);
+ cb->func(NULL, cb);
}
}
diff --git a/drivers/gpu/drm/i915/gt/intel_context.c b/drivers/gpu/drm/i915/gt/intel_context.c
index 06fabdf205cf..35a40c2820a2 100644
--- a/drivers/gpu/drm/i915/gt/intel_context.c
+++ b/drivers/gpu/drm/i915/gt/intel_context.c
@@ -240,7 +240,7 @@ intel_context_init(struct intel_context *ce,
mutex_init(&ce->pin_mutex);
- i915_active_init(ctx->i915, &ce->active,
+ i915_active_init(&ce->active,
__intel_context_active, __intel_context_retire);
}
@@ -307,7 +307,7 @@ int intel_context_prepare_remote_request(struct intel_context *ce,
return err;
/* Queue this switch after current activity by this context. */
- err = i915_active_request_set(&tl->last_request, rq);
+ err = i915_active_fence_set(&tl->last_request, rq);
mutex_unlock(&tl->mutex);
if (err)
return err;
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pool.c b/drivers/gpu/drm/i915/gt/intel_engine_pool.c
index 81fab101fdb4..3cdbd5f8b5be 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_pool.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_pool.c
@@ -95,7 +95,7 @@ node_create(struct intel_engine_pool *pool, size_t sz)
return ERR_PTR(-ENOMEM);
node->pool = pool;
- i915_active_init(engine->i915, &node->active, pool_active, pool_retire);
+ i915_active_init(&node->active, pool_active, pool_retire);
obj = i915_gem_object_create_internal(engine->i915, sz);
if (IS_ERR(obj)) {
diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c
index ae68c3786f63..e8c8d0041e8f 100644
--- a/drivers/gpu/drm/i915/gt/intel_reset.c
+++ b/drivers/gpu/drm/i915/gt/intel_reset.c
@@ -828,10 +828,10 @@ static bool __intel_gt_unset_wedged(struct intel_gt *gt)
*/
spin_lock_irqsave(&timelines->lock, flags);
list_for_each_entry(tl, &timelines->active_list, link) {
- struct i915_request *rq;
+ struct dma_fence *fence;
- rq = i915_active_request_get_unlocked(&tl->last_request);
- if (!rq)
+ fence = i915_active_fence_get(&tl->last_request);
+ if (!fence)
continue;
spin_unlock_irqrestore(&timelines->lock, flags);
@@ -843,8 +843,8 @@ static bool __intel_gt_unset_wedged(struct intel_gt *gt)
* (I915_FENCE_TIMEOUT) so this wait should not be unbounded
* in the worst case.
*/
- dma_fence_default_wait(&rq->fence, false, MAX_SCHEDULE_TIMEOUT);
- i915_request_put(rq);
+ dma_fence_default_wait(fence, false, MAX_SCHEDULE_TIMEOUT);
+ dma_fence_put(fence);
/* Restart iteration after droping lock */
spin_lock_irqsave(&timelines->lock, flags);
diff --git a/drivers/gpu/drm/i915/gt/intel_timeline.c b/drivers/gpu/drm/i915/gt/intel_timeline.c
index 653f60e78392..0f959694303c 100644
--- a/drivers/gpu/drm/i915/gt/intel_timeline.c
+++ b/drivers/gpu/drm/i915/gt/intel_timeline.c
@@ -178,8 +178,7 @@ cacheline_alloc(struct intel_timeline_hwsp *hwsp, unsigned int cacheline)
cl->hwsp = hwsp;
cl->vaddr = page_pack_bits(vaddr, cacheline);
- i915_active_init(hwsp->gt->i915, &cl->active,
- __cacheline_active, __cacheline_retire);
+ i915_active_init(&cl->active, __cacheline_active, __cacheline_retire);
return cl;
}
@@ -255,7 +254,7 @@ int intel_timeline_init(struct intel_timeline *timeline,
mutex_init(&timeline->mutex);
- INIT_ACTIVE_REQUEST(&timeline->last_request, &timeline->mutex);
+ INIT_ACTIVE_FENCE(&timeline->last_request, &timeline->mutex);
INIT_LIST_HEAD(&timeline->requests);
i915_syncmap_init(&timeline->sync);
@@ -443,7 +442,7 @@ __intel_timeline_get_seqno(struct intel_timeline *tl,
* free it after the current request is retired, which ensures that
* all writes into the cacheline from previous requests are complete.
*/
- err = i915_active_ref(&tl->hwsp_cacheline->active, tl, rq);
+ err = i915_active_ref(&tl->hwsp_cacheline->active, tl, &rq->fence);
if (err)
goto err_cacheline;
diff --git a/drivers/gpu/drm/i915/gt/intel_timeline_types.h b/drivers/gpu/drm/i915/gt/intel_timeline_types.h
index c668c4c50e75..7faf81f94355 100644
--- a/drivers/gpu/drm/i915/gt/intel_timeline_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_timeline_types.h
@@ -63,7 +63,7 @@ struct intel_timeline {
* the request using i915_active_request_get_request_rcu(), or hold the
* struct_mutex.
*/
- struct i915_active_request last_request;
+ struct i915_active_fence last_request;
/**
* We track the most recent seqno that we wait on in every context so
diff --git a/drivers/gpu/drm/i915/gt/selftest_context.c b/drivers/gpu/drm/i915/gt/selftest_context.c
index 4ce1e25433d2..e6bcbe7ab5e1 100644
--- a/drivers/gpu/drm/i915/gt/selftest_context.c
+++ b/drivers/gpu/drm/i915/gt/selftest_context.c
@@ -47,24 +47,20 @@ static int context_sync(struct intel_context *ce)
mutex_lock(&tl->mutex);
do {
- struct i915_request *rq;
+ struct dma_fence *fence;
long timeout;
- rcu_read_lock();
- rq = rcu_dereference(tl->last_request.request);
- if (rq)
- rq = i915_request_get_rcu(rq);
- rcu_read_unlock();
- if (!rq)
+ fence = i915_active_fence_get(&tl->last_request);
+ if (!fence)
break;
- timeout = i915_request_wait(rq, 0, HZ / 10);
+ timeout = dma_fence_wait_timeout(fence, false, HZ / 10);
if (timeout < 0)
err = timeout;
else
- i915_request_retire_upto(rq);
+ i915_request_retire_upto(to_request(fence));
- i915_request_put(rq);
+ dma_fence_put(fence);
} while (!err);
mutex_unlock(&tl->mutex);
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index 22ea2e747064..8eec4ead12a4 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -994,9 +994,13 @@ static struct i915_request *dummy_request(struct intel_engine_cs *engine)
if (!rq)
return NULL;
- INIT_LIST_HEAD(&rq->active_list);
rq->engine = engine;
+ spin_lock_init(&rq->lock);
+ INIT_LIST_HEAD(&rq->fence.cb_list);
+ rq->fence.lock = &rq->lock;
+ rq->fence.ops = &i915_fence_ops;
+
i915_sched_node_init(&rq->sched);
/* mark this request as permanently incomplete */
@@ -1089,8 +1093,8 @@ static int live_suppress_wait_preempt(void *arg)
}
/* Disable NEWCLIENT promotion */
- __i915_active_request_set(&i915_request_timeline(rq[i])->last_request,
- dummy);
+ __i915_active_fence_set(&i915_request_timeline(rq[i])->last_request,
+ &dummy->fence);
i915_request_add(rq[i]);
}
diff --git a/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c b/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c
index 598170efcaf6..2a77c051f36a 100644
--- a/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c
+++ b/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c
@@ -15,7 +15,7 @@ void mock_timeline_init(struct intel_timeline *timeline, u64 context)
mutex_init(&timeline->mutex);
- INIT_ACTIVE_REQUEST(&timeline->last_request, &timeline->mutex);
+ INIT_ACTIVE_FENCE(&timeline->last_request, &timeline->mutex);
INIT_LIST_HEAD(&timeline->requests);
i915_syncmap_init(&timeline->sync);
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index 6c79d16b381e..03f567084548 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -385,11 +385,8 @@ intel_gvt_workload_req_alloc(struct intel_vgpu_workload *workload)
{
struct intel_vgpu *vgpu = workload->vgpu;
struct intel_vgpu_submission *s = &vgpu->submission;
- struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
struct i915_request *rq;
- lockdep_assert_held(&dev_priv->drm.struct_mutex);
-
if (workload->req)
return 0;
diff --git a/drivers/gpu/drm/i915/i915_active.c b/drivers/gpu/drm/i915/i915_active.c
index 7ca066688b98..f23590ec9e68 100644
--- a/drivers/gpu/drm/i915/i915_active.c
+++ b/drivers/gpu/drm/i915/i915_active.c
@@ -12,8 +12,6 @@
#include "i915_active.h"
#include "i915_globals.h"
-#define BKL(ref) (&(ref)->i915->drm.struct_mutex)
-
/*
* Active refs memory management
*
@@ -27,35 +25,35 @@ static struct i915_global_active {
} global;
struct active_node {
- struct i915_active_request base;
+ struct i915_active_fence base;
struct i915_active *ref;
struct rb_node node;
u64 timeline;
};
static inline struct active_node *
-node_from_active(struct i915_active_request *active)
+node_from_active(struct i915_active_fence *active)
{
return container_of(active, struct active_node, base);
}
#define take_preallocated_barriers(x) llist_del_all(&(x)->preallocated_barriers)
-static inline bool is_barrier(const struct i915_active_request *active)
+static inline bool is_barrier(const struct i915_active_fence *active)
{
- return IS_ERR(rcu_access_pointer(active->request));
+ return IS_ERR(rcu_access_pointer(active->fence));
}
static inline struct llist_node *barrier_to_ll(struct active_node *node)
{
GEM_BUG_ON(!is_barrier(&node->base));
- return (struct llist_node *)&node->base.link;
+ return (struct llist_node *)&node->base.cb.node;
}
static inline struct intel_engine_cs *
__barrier_to_engine(struct active_node *node)
{
- return (struct intel_engine_cs *)READ_ONCE(node->base.link.prev);
+ return (struct intel_engine_cs *)READ_ONCE(node->base.cb.node.prev);
}
static inline struct intel_engine_cs *
@@ -68,7 +66,7 @@ barrier_to_engine(struct active_node *node)
static inline struct active_node *barrier_from_ll(struct llist_node *x)
{
return container_of((struct list_head *)x,
- struct active_node, base.link);
+ struct active_node, base.cb.node);
}
#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM) && IS_ENABLED(CONFIG_DEBUG_OBJECTS)
@@ -147,15 +145,18 @@ __active_retire(struct i915_active *ref)
if (!retire)
return;
- GEM_BUG_ON(rcu_access_pointer(ref->excl));
+ GEM_BUG_ON(rcu_access_pointer(ref->excl.fence));
rbtree_postorder_for_each_entry_safe(it, n, &root, node) {
- GEM_BUG_ON(i915_active_request_isset(&it->base));
+ GEM_BUG_ON(i915_active_fence_isset(&it->base));
kmem_cache_free(global.slab_cache, it);
}
/* After the final retire, the entire struct may be freed */
if (ref->retire)
ref->retire(ref);
+
+ /* ... except if you wait on it, you must manage your own references! */
+ wake_up_var(ref);
}
static void
@@ -189,12 +190,20 @@ active_retire(struct i915_active *ref)
}
static void
-node_retire(struct i915_active_request *base, struct i915_request *rq)
+node_retire(struct dma_fence *fence, struct dma_fence_cb *cb)
{
- active_retire(node_from_active(base)->ref);
+ i915_active_fence_cb(fence, cb);
+ active_retire(container_of(cb, struct active_node, base.cb)->ref);
}
-static struct i915_active_request *
+static void
+excl_retire(struct dma_fence *fence, struct dma_fence_cb *cb)
+{
+ i915_active_fence_cb(fence, cb);
+ active_retire(container_of(cb, struct i915_active, excl.cb));
+}
+
+static struct i915_active_fence *
active_instance(struct i915_active *ref, struct intel_timeline *tl)
{
struct active_node *node, *prealloc;
@@ -238,7 +247,7 @@ active_instance(struct i915_active *ref, struct intel_timeline *tl)
}
node = prealloc;
- i915_active_request_init(&node->base, &tl->mutex, NULL, node_retire);
+ __i915_active_fence_init(&node->base, &tl->mutex, NULL, node_retire);
node->ref = ref;
node->timeline = idx;
@@ -253,8 +262,7 @@ active_instance(struct i915_active *ref, struct intel_timeline *tl)
return &node->base;
}
-void __i915_active_init(struct drm_i915_private *i915,
- struct i915_active *ref,
+void __i915_active_init(struct i915_active *ref,
int (*active)(struct i915_active *ref),
void (*retire)(struct i915_active *ref),
struct lock_class_key *key)
@@ -263,19 +271,18 @@ void __i915_active_init(struct drm_i915_private *i915,
debug_active_init(ref);
- ref->i915 = i915;
ref->flags = 0;
ref->active = active;
ref->retire = ptr_unpack_bits(retire, &bits, 2);
if (bits & I915_ACTIVE_MAY_SLEEP)
ref->flags |= I915_ACTIVE_RETIRE_SLEEPS;
- ref->excl = NULL;
ref->tree = RB_ROOT;
ref->cache = NULL;
init_llist_head(&ref->preallocated_barriers);
atomic_set(&ref->count, 0);
__mutex_init(&ref->mutex, "i915_active", key);
+ __i915_active_fence_init(&ref->excl, &ref->mutex, NULL, excl_retire);
INIT_WORK(&ref->work, active_work);
}
@@ -329,9 +336,9 @@ __active_del_barrier(struct i915_active *ref, struct active_node *node)
int i915_active_ref(struct i915_active *ref,
struct intel_timeline *tl,
- struct i915_request *rq)
+ struct dma_fence *fence)
{
- struct i915_active_request *active;
+ struct i915_active_fence *active;
int err;
lockdep_assert_held(&tl->mutex);
@@ -354,66 +361,39 @@ int i915_active_ref(struct i915_active *ref,
* request that we want to emit on the kernel_context.
*/
__active_del_barrier(ref, node_from_active(active));
- RCU_INIT_POINTER(active->request, NULL);
- INIT_LIST_HEAD(&active->link);
- } else {
- if (!i915_active_request_isset(active))
- atomic_inc(&ref->count);
+ RCU_INIT_POINTER(active->fence, NULL);
+ atomic_dec(&ref->count);
}
- GEM_BUG_ON(!atomic_read(&ref->count));
- __i915_active_request_set(active, rq);
+ if (!__i915_active_fence_set(active, fence))
+ atomic_inc(&ref->count);
out:
i915_active_release(ref);
return err;
}
-static void excl_cb(struct dma_fence *f, struct dma_fence_cb *cb)
-{
- struct i915_active *ref = container_of(cb, typeof(*ref), excl_cb);
-
- RCU_INIT_POINTER(ref->excl, NULL);
- dma_fence_put(f);
-
- active_retire(ref);
-}
-
void i915_active_set_exclusive(struct i915_active *ref, struct dma_fence *f)
{
/* We expect the caller to manage the exclusive timeline ordering */
GEM_BUG_ON(i915_active_is_idle(ref));
- dma_fence_get(f);
-
- rcu_read_lock();
- if (rcu_access_pointer(ref->excl)) {
- struct dma_fence *old;
-
- old = dma_fence_get_rcu_safe(&ref->excl);
- if (old) {
- if (dma_fence_remove_callback(old, &ref->excl_cb))
- atomic_dec(&ref->count);
- dma_fence_put(old);
- }
- }
- rcu_read_unlock();
-
- atomic_inc(&ref->count);
- rcu_assign_pointer(ref->excl, f);
+ mutex_acquire(&ref->mutex.dep_map, 0, 0, _THIS_IP_);
+ if (!__i915_active_fence_set(&ref->excl, f))
+ atomic_inc(&ref->count);
+ mutex_release(&ref->mutex.dep_map, 0, _THIS_IP_);
+}
- if (dma_fence_add_callback(f, &ref->excl_cb, excl_cb)) {
- RCU_INIT_POINTER(ref->excl, NULL);
- atomic_dec(&ref->count);
- dma_fence_put(f);
- }
+bool i915_active_acquire_if_busy(struct i915_active *ref)
+{
+ debug_active_assert(ref);
+ return atomic_add_unless(&ref->count, 1, 0);
}
int i915_active_acquire(struct i915_active *ref)
{
int err;
- debug_active_assert(ref);
- if (atomic_add_unless(&ref->count, 1, 0))
+ if (i915_active_acquire_if_busy(ref))
return 0;
err = mutex_lock_interruptible(&ref->mutex);
@@ -438,121 +418,57 @@ void i915_active_release(struct i915_active *ref)
active_retire(ref);
}
-static void __active_ungrab(struct i915_active *ref)
-{
- clear_and_wake_up_bit(I915_ACTIVE_GRAB_BIT, &ref->flags);
-}
-
-bool i915_active_trygrab(struct i915_active *ref)
+static void enable_signaling(struct i915_active_fence *active)
{
- debug_active_assert(ref);
-
- if (test_and_set_bit(I915_ACTIVE_GRAB_BIT, &ref->flags))
- return false;
-
- if (!atomic_add_unless(&ref->count, 1, 0)) {
- __active_ungrab(ref);
- return false;
- }
-
- return true;
-}
+ struct dma_fence *fence;
-void i915_active_ungrab(struct i915_active *ref)
-{
- GEM_BUG_ON(!test_bit(I915_ACTIVE_GRAB_BIT, &ref->flags));
-
- active_retire(ref);
- __active_ungrab(ref);
-}
-
-static int excl_wait(struct i915_active *ref)
-{
- struct dma_fence *old;
- int err = 0;
-
- if (!rcu_access_pointer(ref->excl))
- return 0;
-
- rcu_read_lock();
- old = dma_fence_get_rcu_safe(&ref->excl);
- rcu_read_unlock();
- if (old) {
- err = dma_fence_wait(old, true);
- dma_fence_put(old);
- }
+ fence = i915_active_fence_get(active);
+ if (!fence)
+ return;
- return err;
+ dma_fence_enable_sw_signaling(fence);
+ dma_fence_put(fence);
}
int i915_active_wait(struct i915_active *ref)
{
struct active_node *it, *n;
- int err;
+ int err = 0;
might_sleep();
- might_lock(&ref->mutex);
- if (i915_active_is_idle(ref))
+ if (!i915_active_acquire_if_busy(ref))
return 0;
- err = mutex_lock_interruptible(&ref->mutex);
- if (err)
- return err;
-
- if (!atomic_add_unless(&ref->count, 1, 0)) {
- mutex_unlock(&ref->mutex);
- return 0;
- }
-
- err = excl_wait(ref);
- if (err)
- goto out;
-
+ /* Flush lazy signals */
+ enable_signaling(&ref->excl);
rbtree_postorder_for_each_entry_safe(it, n, &ref->tree, node) {
- if (is_barrier(&it->base)) { /* unconnected idle-barrier */
- err = -EBUSY;
- break;
- }
+ if (is_barrier(&it->base)) /* unconnected idle barrier */
+ continue;
- err = i915_active_request_retire(&it->base, BKL(ref));
- if (err)
- break;
+ enable_signaling(&it->base);
}
+ /* Any fence added after the wait begins will not be auto-signaled */
-out:
- __active_retire(ref);
+ i915_active_release(ref);
if (err)
return err;
- if (wait_on_bit(&ref->flags, I915_ACTIVE_GRAB_BIT, TASK_KILLABLE))
+ if (wait_var_event_interruptible(ref, i915_active_is_idle(ref)))
return -EINTR;
- flush_work(&ref->work);
- if (!i915_active_is_idle(ref))
- return -EBUSY;
-
return 0;
}
-int i915_request_await_active_request(struct i915_request *rq,
- struct i915_active_request *active)
-{
- struct i915_request *barrier =
- i915_active_request_raw(active, &rq->i915->drm.struct_mutex);
-
- return barrier ? i915_request_await_dma_fence(rq, &barrier->fence) : 0;
-}
-
int i915_request_await_active(struct i915_request *rq, struct i915_active *ref)
{
int err = 0;
- if (rcu_access_pointer(ref->excl)) {
+ if (rcu_access_pointer(ref->excl.fence)) {
struct dma_fence *fence;
rcu_read_lock();
- fence = dma_fence_get_rcu_safe(&ref->excl);
+ fence = dma_fence_get_rcu_safe(&ref->excl.fence);
rcu_read_unlock();
if (fence) {
err = i915_request_await_dma_fence(rq, fence);
@@ -578,7 +494,7 @@ void i915_active_fini(struct i915_active *ref)
static inline bool is_idle_barrier(struct active_node *node, u64 idx)
{
- return node->timeline == idx && !i915_active_request_isset(&node->base);
+ return node->timeline == idx && !i915_active_fence_isset(&node->base);
}
static struct active_node *reuse_idle_barrier(struct i915_active *ref, u64 idx)
@@ -698,13 +614,13 @@ int i915_active_acquire_preallocate_barrier(struct i915_active *ref,
node->base.lock =
&engine->kernel_context->timeline->mutex;
#endif
- RCU_INIT_POINTER(node->base.request, NULL);
- node->base.retire = node_retire;
+ RCU_INIT_POINTER(node->base.fence, NULL);
+ node->base.cb.func = node_retire;
node->timeline = idx;
node->ref = ref;
}
- if (!i915_active_request_isset(&node->base)) {
+ if (!i915_active_fence_isset(&node->base)) {
/*
* Mark this as being *our* unconnected proto-node.
*
@@ -714,8 +630,8 @@ int i915_active_acquire_preallocate_barrier(struct i915_active *ref,
* and then we can use the rb_node and list pointers
* for our tracking of the pending barrier.
*/
- RCU_INIT_POINTER(node->base.request, ERR_PTR(-EAGAIN));
- node->base.link.prev = (void *)engine;
+ RCU_INIT_POINTER(node->base.fence, ERR_PTR(-EAGAIN));
+ node->base.cb.node.prev = (void *)engine;
atomic_inc(&ref->count);
}
@@ -782,25 +698,65 @@ void i915_request_add_active_barriers(struct i915_request *rq)
{
struct intel_engine_cs *engine = rq->engine;
struct llist_node *node, *next;
+ unsigned long flags;
GEM_BUG_ON(intel_engine_is_virtual(engine));
GEM_BUG_ON(i915_request_timeline(rq) != engine->kernel_context->timeline);
+ node = llist_del_all(&engine->barrier_tasks);
+ if (!node)
+ return;
/*
* Attach the list of proto-fences to the in-flight request such
* that the parent i915_active will be released when this request
* is retired.
*/
- llist_for_each_safe(node, next, llist_del_all(&engine->barrier_tasks)) {
- RCU_INIT_POINTER(barrier_from_ll(node)->base.request, rq);
+ spin_lock_irqsave(&rq->lock, flags);
+ llist_for_each_safe(node, next, node) {
+ RCU_INIT_POINTER(barrier_from_ll(node)->base.fence, &rq->fence);
smp_wmb(); /* serialise with reuse_idle_barrier */
- list_add_tail((struct list_head *)node, &rq->active_list);
+ list_add_tail((struct list_head *)node, &rq->fence.cb_list);
+ }
+ spin_unlock_irqrestore(&rq->lock, flags);
+}
+
+#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
+#define active_is_held(active) lockdep_is_held((active)->lock)
+#else
+#define active_is_held(active) true
+#endif
+
+struct dma_fence *
+__i915_active_fence_set(struct i915_active_fence *active,
+ struct dma_fence *fence)
+{
+ struct dma_fence *prev;
+ unsigned long flags;
+
+ /* NB: updates must be serialised by an outer timeline mutex */
+ spin_lock_irqsave(fence->lock, flags);
+ GEM_BUG_ON(test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags));
+
+ prev = rcu_dereference_protected(active->fence, active_is_held(active));
+ if (prev) {
+ spin_lock_nested(prev->lock, SINGLE_DEPTH_NESTING);
+ __list_del_entry(&active->cb.node);
+ spin_unlock(prev->lock); /* serialise with prev->cb_list */
+ prev = rcu_access_pointer(active->fence);
}
+
+ rcu_assign_pointer(active->fence, fence);
+ list_add_tail(&active->cb.node, &fence->cb_list);
+
+ spin_unlock_irqrestore(fence->lock, flags);
+
+ return prev;
}
-int i915_active_request_set(struct i915_active_request *active,
- struct i915_request *rq)
+int i915_active_fence_set(struct i915_active_fence *active,
+ struct i915_request *rq)
{
+ struct dma_fence *fence;
int err;
#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
@@ -808,18 +764,25 @@ int i915_active_request_set(struct i915_active_request *active,
#endif
/* Must maintain ordering wrt previous active requests */
- err = i915_request_await_active_request(rq, active);
- if (err)
- return err;
+ rcu_read_lock();
+ fence = __i915_active_fence_set(active, &rq->fence);
+ if (fence)
+ fence = dma_fence_get_rcu(fence);
+ rcu_read_unlock();
+
+ if (fence) {
+ err = i915_request_await_dma_fence(rq, fence);
+ dma_fence_put(fence);
+ if (err)
+ return err;
+ }
- __i915_active_request_set(active, rq);
return 0;
}
-void i915_active_retire_noop(struct i915_active_request *active,
- struct i915_request *request)
+void i915_active_noop(struct dma_fence *fence, struct dma_fence_cb *cb)
{
- /* Space left intentionally blank */
+ i915_active_fence_cb(fence, cb);
}
#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
diff --git a/drivers/gpu/drm/i915/i915_active.h b/drivers/gpu/drm/i915/i915_active.h
index 90034f61b7c2..4f52fe6146d2 100644
--- a/drivers/gpu/drm/i915/i915_active.h
+++ b/drivers/gpu/drm/i915/i915_active.h
@@ -12,6 +12,10 @@
#include "i915_active_types.h"
#include "i915_request.h"
+struct i915_request;
+struct intel_engine_cs;
+struct intel_timeline;
+
/*
* We treat requests as fences. This is not be to confused with our
* "fence registers" but pipeline synchronisation objects ala GL_ARB_sync.
@@ -28,308 +32,108 @@
* write access so that we can perform concurrent read operations between
* the CPU and GPU engines, as well as waiting for all rendering to
* complete, or waiting for the last GPU user of a "fence register". The
- * object then embeds a #i915_active_request to track the most recent (in
+ * object then embeds a #i915_active_fence to track the most recent (in
* retirement order) request relevant for the desired mode of access.
- * The #i915_active_request is updated with i915_active_request_set() to
+ * The #i915_active_fence is updated with i915_active_fence_set() to
* track the most recent fence request, typically this is done as part of
* i915_vma_move_to_active().
*
- * When the #i915_active_request completes (is retired), it will
+ * When the #i915_active_fence completes (is retired), it will
* signal its completion to the owner through a callback as well as mark
- * itself as idle (i915_active_request.request == NULL). The owner
+ * itself as idle (i915_active_fence.request == NULL). The owner
* can then perform any action, such as delayed freeing of an active
* resource including itself.
*/
-void i915_active_retire_noop(struct i915_active_request *active,
- struct i915_request *request);
+void i915_active_noop(struct dma_fence *fence, struct dma_fence_cb *cb);
/**
- * i915_active_request_init - prepares the activity tracker for use
+ * __i915_active_fence_init - prepares the activity tracker for use
* @active - the active tracker
- * @rq - initial request to track, can be NULL
+ * @fence - initial fence to track, can be NULL
* @func - a callback when then the tracker is retired (becomes idle),
* can be NULL
*
- * i915_active_request_init() prepares the embedded @active struct for use as
- * an activity tracker, that is for tracking the last known active request
- * associated with it. When the last request becomes idle, when it is retired
+ * i915_active_fence_init() prepares the embedded @active struct for use as
+ * an activity tracker, that is for tracking the last known active fence
+ * associated with it. When the last fence becomes idle, when it is retired
* after completion, the optional callback @func is invoked.
*/
static inline void
-i915_active_request_init(struct i915_active_request *active,
+__i915_active_fence_init(struct i915_active_fence *active,
struct mutex *lock,
- struct i915_request *rq,
- i915_active_retire_fn retire)
+ void *fence,
+ dma_fence_func_t fn)
{
- RCU_INIT_POINTER(active->request, rq);
- INIT_LIST_HEAD(&active->link);
- active->retire = retire ?: i915_active_retire_noop;
+ RCU_INIT_POINTER(active->fence, fence);
+ active->cb.func = fn ?: i915_active_noop;
#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
active->lock = lock;
#endif
}
-#define INIT_ACTIVE_REQUEST(name, lock) \
- i915_active_request_init((name), (lock), NULL, NULL)
-
-/**
- * i915_active_request_set - updates the tracker to watch the current request
- * @active - the active tracker
- * @request - the request to watch
- *
- * __i915_active_request_set() watches the given @request for completion. Whilst
- * that @request is busy, the @active reports busy. When that @request is
- * retired, the @active tracker is updated to report idle.
- */
-static inline void
-__i915_active_request_set(struct i915_active_request *active,
- struct i915_request *request)
-{
-#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
- lockdep_assert_held(active->lock);
-#endif
- list_move(&active->link, &request->active_list);
- rcu_assign_pointer(active->request, request);
-}
+#define INIT_ACTIVE_FENCE(A, LOCK) \
+ __i915_active_fence_init((A), (LOCK), NULL, NULL)
-int __must_check
-i915_active_request_set(struct i915_active_request *active,
- struct i915_request *rq);
+struct dma_fence *
+__i915_active_fence_set(struct i915_active_fence *active,
+ struct dma_fence *fence);
/**
- * i915_active_request_raw - return the active request
+ * i915_active_fence_set - updates the tracker to watch the current fence
* @active - the active tracker
+ * @rq - the request to watch
*
- * i915_active_request_raw() returns the current request being tracked, or NULL.
- * It does not obtain a reference on the request for the caller, so the caller
- * must hold struct_mutex.
+ * i915_active_fence_set() watches the given @rq for completion. While
+ * that @rq is busy, the @active reports busy. When that @rq is signaled
+ * (or else retired) the @active tracker is updated to report idle.
*/
-static inline struct i915_request *
-i915_active_request_raw(const struct i915_active_request *active,
- struct mutex *mutex)
-{
- return rcu_dereference_protected(active->request,
- lockdep_is_held(mutex));
-}
-
-/**
- * i915_active_request_peek - report the active request being monitored
- * @active - the active tracker
- *
- * i915_active_request_peek() returns the current request being tracked if
- * still active, or NULL. It does not obtain a reference on the request
- * for the caller, so the caller must hold struct_mutex.
- */
-static inline struct i915_request *
-i915_active_request_peek(const struct i915_active_request *active,
- struct mutex *mutex)
-{
- struct i915_request *request;
-
- request = i915_active_request_raw(active, mutex);
- if (!request || i915_request_completed(request))
- return NULL;
-
- return request;
-}
-
-/**
- * i915_active_request_get - return a reference to the active request
- * @active - the active tracker
- *
- * i915_active_request_get() returns a reference to the active request, or NULL
- * if the active tracker is idle. The caller must hold struct_mutex.
- */
-static inline struct i915_request *
-i915_active_request_get(const struct i915_active_request *active,
- struct mutex *mutex)
-{
- return i915_request_get(i915_active_request_peek(active, mutex));
-}
-
-/**
- * __i915_active_request_get_rcu - return a reference to the active request
- * @active - the active tracker
- *
- * __i915_active_request_get() returns a reference to the active request,
- * or NULL if the active tracker is idle. The caller must hold the RCU read
- * lock, but the returned pointer is safe to use outside of RCU.
- */
-static inline struct i915_request *
-__i915_active_request_get_rcu(const struct i915_active_request *active)
-{
- /*
- * Performing a lockless retrieval of the active request is super
- * tricky. SLAB_TYPESAFE_BY_RCU merely guarantees that the backing
- * slab of request objects will not be freed whilst we hold the
- * RCU read lock. It does not guarantee that the request itself
- * will not be freed and then *reused*. Viz,
- *
- * Thread A Thread B
- *
- * rq = active.request
- * retire(rq) -> free(rq);
- * (rq is now first on the slab freelist)
- * active.request = NULL
- *
- * rq = new submission on a new object
- * ref(rq)
- *
- * To prevent the request from being reused whilst the caller
- * uses it, we take a reference like normal. Whilst acquiring
- * the reference we check that it is not in a destroyed state
- * (refcnt == 0). That prevents the request being reallocated
- * whilst the caller holds on to it. To check that the request
- * was not reallocated as we acquired the reference we have to
- * check that our request remains the active request across
- * the lookup, in the same manner as a seqlock. The visibility
- * of the pointer versus the reference counting is controlled
- * by using RCU barriers (rcu_dereference and rcu_assign_pointer).
- *
- * In the middle of all that, we inspect whether the request is
- * complete. Retiring is lazy so the request may be completed long
- * before the active tracker is updated. Querying whether the
- * request is complete is far cheaper (as it involves no locked
- * instructions setting cachelines to exclusive) than acquiring
- * the reference, so we do it first. The RCU read lock ensures the
- * pointer dereference is valid, but does not ensure that the
- * seqno nor HWS is the right one! However, if the request was
- * reallocated, that means the active tracker's request was complete.
- * If the new request is also complete, then both are and we can
- * just report the active tracker is idle. If the new request is
- * incomplete, then we acquire a reference on it and check that
- * it remained the active request.
- *
- * It is then imperative that we do not zero the request on
- * reallocation, so that we can chase the dangling pointers!
- * See i915_request_alloc().
- */
- do {
- struct i915_request *request;
-
- request = rcu_dereference(active->request);
- if (!request || i915_request_completed(request))
- return NULL;
-
- /*
- * An especially silly compiler could decide to recompute the
- * result of i915_request_completed, more specifically
- * re-emit the load for request->fence.seqno. A race would catch
- * a later seqno value, which could flip the result from true to
- * false. Which means part of the instructions below might not
- * be executed, while later on instructions are executed. Due to
- * barriers within the refcounting the inconsistency can't reach
- * past the call to i915_request_get_rcu, but not executing
- * that while still executing i915_request_put() creates
- * havoc enough. Prevent this with a compiler barrier.
- */
- barrier();
-
- request = i915_request_get_rcu(request);
-
- /*
- * What stops the following rcu_access_pointer() from occurring
- * before the above i915_request_get_rcu()? If we were
- * to read the value before pausing to get the reference to
- * the request, we may not notice a change in the active
- * tracker.
- *
- * The rcu_access_pointer() is a mere compiler barrier, which
- * means both the CPU and compiler are free to perform the
- * memory read without constraint. The compiler only has to
- * ensure that any operations after the rcu_access_pointer()
- * occur afterwards in program order. This means the read may
- * be performed earlier by an out-of-order CPU, or adventurous
- * compiler.
- *
- * The atomic operation at the heart of
- * i915_request_get_rcu(), see dma_fence_get_rcu(), is
- * atomic_inc_not_zero() which is only a full memory barrier
- * when successful. That is, if i915_request_get_rcu()
- * returns the request (and so with the reference counted
- * incremented) then the following read for rcu_access_pointer()
- * must occur after the atomic operation and so confirm
- * that this request is the one currently being tracked.
- *
- * The corresponding write barrier is part of
- * rcu_assign_pointer().
- */
- if (!request || request == rcu_access_pointer(active->request))
- return rcu_pointer_handoff(request);
-
- i915_request_put(request);
- } while (1);
-}
-
+int __must_check
+i915_active_fence_set(struct i915_active_fence *active,
+ struct i915_request *rq);
/**
- * i915_active_request_get_unlocked - return a reference to the active request
+ * i915_active_fence_get - return a reference to the active fence
* @active - the active tracker
*
- * i915_active_request_get_unlocked() returns a reference to the active request,
+ * i915_active_fence_get() returns a reference to the active fence,
* or NULL if the active tracker is idle. The reference is obtained under RCU,
* so no locking is required by the caller.
*
- * The reference should be freed with i915_request_put().
+ * The reference should be freed with dma_fence_put().
*/
-static inline struct i915_request *
-i915_active_request_get_unlocked(const struct i915_active_request *active)
+static inline struct dma_fence *
+i915_active_fence_get(struct i915_active_fence *active)
{
- struct i915_request *request;
+ struct dma_fence *fence;
rcu_read_lock();
- request = __i915_active_request_get_rcu(active);
+ fence = dma_fence_get_rcu_safe(&active->fence);
rcu_read_unlock();
- return request;
+ return fence;
}
/**
- * i915_active_request_isset - report whether the active tracker is assigned
+ * i915_active_fence_isset - report whether the active tracker is assigned
* @active - the active tracker
*
- * i915_active_request_isset() returns true if the active tracker is currently
- * assigned to a request. Due to the lazy retiring, that request may be idle
+ * i915_active_fence_isset() returns true if the active tracker is currently
+ * assigned to a fence. Due to the lazy retiring, that fence may be idle
* and this may report stale information.
*/
static inline bool
-i915_active_request_isset(const struct i915_active_request *active)
+i915_active_fence_isset(const struct i915_active_fence *active)
{
- return rcu_access_pointer(active->request);
+ return rcu_access_pointer(active->fence);
}
-/**
- * i915_active_request_retire - waits until the request is retired
- * @active - the active request on which to wait
- *
- * i915_active_request_retire() waits until the request is completed,
- * and then ensures that at least the retirement handler for this
- * @active tracker is called before returning. If the @active
- * tracker is idle, the function returns immediately.
- */
-static inline int __must_check
-i915_active_request_retire(struct i915_active_request *active,
- struct mutex *mutex)
+static inline void
+i915_active_fence_cb(struct dma_fence *fence, struct dma_fence_cb *cb)
{
- struct i915_request *request;
- long ret;
-
- request = i915_active_request_raw(active, mutex);
- if (!request)
- return 0;
-
- ret = i915_request_wait(request,
- I915_WAIT_INTERRUPTIBLE,
- MAX_SCHEDULE_TIMEOUT);
- if (ret < 0)
- return ret;
+ struct i915_active_fence *active =
+ container_of(cb, typeof(*active), cb);
- list_del_init(&active->link);
- RCU_INIT_POINTER(active->request, NULL);
-
- active->retire(active, request);
-
- return 0;
+ RCU_INIT_POINTER(active->fence, NULL);
}
/*
@@ -358,47 +162,40 @@ i915_active_request_retire(struct i915_active_request *active,
* synchronisation.
*/
-void __i915_active_init(struct drm_i915_private *i915,
- struct i915_active *ref,
+void __i915_active_init(struct i915_active *ref,
int (*active)(struct i915_active *ref),
void (*retire)(struct i915_active *ref),
struct lock_class_key *key);
-#define i915_active_init(i915, ref, active, retire) do { \
+#define i915_active_init(ref, active, retire) do { \
static struct lock_class_key __key; \
\
- __i915_active_init(i915, ref, active, retire, &__key); \
+ __i915_active_init(ref, active, retire, &__key); \
} while (0)
int i915_active_ref(struct i915_active *ref,
struct intel_timeline *tl,
- struct i915_request *rq);
+ struct dma_fence *fence);
static inline int
i915_active_add_request(struct i915_active *ref, struct i915_request *rq)
{
- return i915_active_ref(ref, i915_request_timeline(rq), rq);
+ return i915_active_ref(ref, i915_request_timeline(rq), &rq->fence);
}
void i915_active_set_exclusive(struct i915_active *ref, struct dma_fence *f);
static inline bool i915_active_has_exclusive(struct i915_active *ref)
{
- return rcu_access_pointer(ref->excl);
+ return rcu_access_pointer(ref->excl.fence);
}
int i915_active_wait(struct i915_active *ref);
-int i915_request_await_active(struct i915_request *rq,
- struct i915_active *ref);
-int i915_request_await_active_request(struct i915_request *rq,
- struct i915_active_request *active);
+int i915_request_await_active(struct i915_request *rq, struct i915_active *ref);
int i915_active_acquire(struct i915_active *ref);
+bool i915_active_acquire_if_busy(struct i915_active *ref);
void i915_active_release(struct i915_active *ref);
-void __i915_active_release_nested(struct i915_active *ref, int subclass);
-
-bool i915_active_trygrab(struct i915_active *ref);
-void i915_active_ungrab(struct i915_active *ref);
static inline bool
i915_active_is_idle(const struct i915_active *ref)
diff --git a/drivers/gpu/drm/i915/i915_active_types.h b/drivers/gpu/drm/i915/i915_active_types.h
index 021167f0004d..d89a74c142c6 100644
--- a/drivers/gpu/drm/i915/i915_active_types.h
+++ b/drivers/gpu/drm/i915/i915_active_types.h
@@ -17,17 +17,9 @@
#include "i915_utils.h"
-struct drm_i915_private;
-struct i915_active_request;
-struct i915_request;
-
-typedef void (*i915_active_retire_fn)(struct i915_active_request *,
- struct i915_request *);
-
-struct i915_active_request {
- struct i915_request __rcu *request;
- struct list_head link;
- i915_active_retire_fn retire;
+struct i915_active_fence {
+ struct dma_fence __rcu *fence;
+ struct dma_fence_cb cb;
#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
/*
* Incorporeal!
@@ -53,20 +45,17 @@ struct active_node;
#define i915_active_may_sleep(fn) ptr_pack_bits(&(fn), I915_ACTIVE_MAY_SLEEP, 2)
struct i915_active {
- struct drm_i915_private *i915;
+ atomic_t count;
+ struct mutex mutex;
struct active_node *cache;
struct rb_root tree;
- struct mutex mutex;
- atomic_t count;
/* Preallocated "exclusive" node */
- struct dma_fence __rcu *excl;
- struct dma_fence_cb excl_cb;
+ struct i915_active_fence excl;
unsigned long flags;
#define I915_ACTIVE_RETIRE_SLEEPS BIT(0)
-#define I915_ACTIVE_GRAB_BIT 1
int (*active)(struct i915_active *ref);
void (*retire)(struct i915_active *ref);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 934ea776a46a..a88849544c8b 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -892,28 +892,38 @@ wait_for_timelines(struct intel_gt *gt, unsigned int wait, long timeout)
spin_lock_irqsave(&timelines->lock, flags);
list_for_each_entry(tl, &timelines->active_list, link) {
- struct i915_request *rq;
+ struct dma_fence *fence;
- rq = i915_active_request_get_unlocked(&tl->last_request);
- if (!rq)
+ fence = i915_active_fence_get(&tl->last_request);
+ if (!fence)
continue;
spin_unlock_irqrestore(&timelines->lock, flags);
- /*
- * "Race-to-idle".
- *
- * Switching to the kernel context is often used a synchronous
- * step prior to idling, e.g. in suspend for flushing all
- * current operations to memory before sleeping. These we
- * want to complete as quickly as possible to avoid prolonged
- * stalls, so allow the gpu to boost to maximum clocks.
- */
- if (wait & I915_WAIT_FOR_IDLE_BOOST)
- gen6_rps_boost(rq);
+ if (!dma_fence_is_i915(fence)) {
+ timeout = dma_fence_wait_timeout(fence,
+ flags & I915_WAIT_INTERRUPTIBLE,
+ timeout);
+ } else {
+ struct i915_request *rq = to_request(fence);
+
+ /*
+ * "Race-to-idle".
+ *
+ * Switching to the kernel context is often used as
+ * a synchronous step prior to idling, e.g. in suspend
+ * for flushing all current operations to memory before
+ * sleeping. These we want to complete as quickly as
+ * possible to avoid prolonged stalls, so allow the gpu
+ * to boost to maximum clocks.
+ */
+ if (flags & I915_WAIT_FOR_IDLE_BOOST)
+ gen6_rps_boost(rq);
+
+ timeout = i915_request_wait(rq, flags, timeout);
+ }
- timeout = i915_request_wait(rq, wait, timeout);
- i915_request_put(rq);
+ dma_fence_put(fence);
if (timeout < 0)
return timeout;
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index a351b98afe03..bf57c41fa1fb 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -1861,7 +1861,6 @@ static const struct i915_vma_ops pd_vma_ops = {
static struct i915_vma *pd_vma_create(struct gen6_ppgtt *ppgtt, int size)
{
- struct drm_i915_private *i915 = ppgtt->base.vm.i915;
struct i915_ggtt *ggtt = ppgtt->base.vm.gt->ggtt;
struct i915_vma *vma;
@@ -1872,7 +1871,7 @@ static struct i915_vma *pd_vma_create(struct gen6_ppgtt *ppgtt, int size)
if (!vma)
return ERR_PTR(-ENOMEM);
- i915_active_init(i915, &vma->active, NULL, NULL);
+ i915_active_init(&vma->active, NULL, NULL);
mutex_init(&vma->pages_mutex);
vma->vm = i915_vm_get(&ggtt->vm);
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index 6384a06aa5bf..a28ee754b7b4 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -1299,7 +1299,7 @@ capture_vma(struct capture_vma *next,
if (!c)
return next;
- if (!i915_active_trygrab(&vma->active)) {
+ if (!i915_active_acquire_if_busy(&vma->active)) {
kfree(c);
return next;
}
@@ -1439,7 +1439,7 @@ gem_record_rings(struct i915_gpu_state *error, struct compress *compress)
*this->slot =
i915_error_object_create(i915, vma, compress);
- i915_active_ungrab(&vma->active);
+ i915_active_release(&vma->active);
i915_vma_put(vma);
capture = this->next;
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index a8916412759b..4ffe62a42186 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -218,8 +218,6 @@ static void remove_from_engine(struct i915_request *rq)
static bool i915_request_retire(struct i915_request *rq)
{
- struct i915_active_request *active, *next;
-
if (!i915_request_completed(rq))
return false;
@@ -244,35 +242,6 @@ static bool i915_request_retire(struct i915_request *rq)
&i915_request_timeline(rq)->requests));
rq->ring->head = rq->postfix;
- /*
- * Walk through the active list, calling retire on each. This allows
- * objects to track their GPU activity and mark themselves as idle
- * when their *last* active request is completed (updating state
- * tracking lists for eviction, active references for GEM, etc).
- *
- * As the ->retire() may free the node, we decouple it first and
- * pass along the auxiliary information (to avoid dereferencing
- * the node after the callback).
- */
- 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
- * the retirement of requests and processing their callbacks.
- * Of which, this loop itself is particularly hot due to the
- * cache misses when jumping around the list of
- * i915_active_request. So we try to keep this loop as
- * streamlined as possible and also prefetch the next
- * i915_active_request to try and hide the likely cache miss.
- */
- prefetchw(next);
-
- INIT_LIST_HEAD(&active->link);
- RCU_INIT_POINTER(active->request, NULL);
-
- active->retire(active, rq);
- }
-
local_irq_disable();
/*
@@ -704,7 +673,6 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp)
rq->flags = 0;
rq->execution_mask = ALL_ENGINES;
- INIT_LIST_HEAD(&rq->active_list);
INIT_LIST_HEAD(&rq->execute_cb);
/*
@@ -743,7 +711,6 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp)
ce->ring->emit = rq->head;
/* Make sure we didn't add ourselves to external state before freeing */
- GEM_BUG_ON(!list_empty(&rq->active_list));
GEM_BUG_ON(!list_empty(&rq->sched.signalers_list));
GEM_BUG_ON(!list_empty(&rq->sched.waiters_list));
@@ -1174,8 +1141,8 @@ __i915_request_add_to_timeline(struct i915_request *rq)
* precludes optimising to use semaphores serialisation of a single
* timeline across engines.
*/
- prev = rcu_dereference_protected(timeline->last_request.request,
- lockdep_is_held(&timeline->mutex));
+ prev = to_request(__i915_active_fence_set(&timeline->last_request,
+ &rq->fence));
if (prev && !i915_request_completed(prev)) {
if (is_power_of_2(prev->engine->mask | rq->engine->mask))
i915_sw_fence_await_sw_fence(&rq->submit,
@@ -1200,7 +1167,6 @@ __i915_request_add_to_timeline(struct i915_request *rq)
* us, the timeline will hold its seqno which is later than ours.
*/
GEM_BUG_ON(timeline->seqno != rq->fence.seqno);
- __i915_active_request_set(&timeline->last_request, rq);
return prev;
}
diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
index ec5bb4c2e5ae..91a885c36c6b 100644
--- a/drivers/gpu/drm/i915/i915_request.h
+++ b/drivers/gpu/drm/i915/i915_request.h
@@ -211,7 +211,6 @@ struct i915_request {
* on the active_list (of their final request).
*/
struct i915_capture_list *capture_list;
- struct list_head active_list;
/** Time at which this request was emitted, in jiffies. */
unsigned long emitted_jiffies;
diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
index 2717441a9f3a..647bb96a6220 100644
--- a/drivers/gpu/drm/i915/i915_vma.c
+++ b/drivers/gpu/drm/i915/i915_vma.c
@@ -120,8 +120,7 @@ vma_create(struct drm_i915_gem_object *obj,
vma->size = obj->base.size;
vma->display_alignment = I915_GTT_MIN_ALIGNMENT;
- i915_active_init(vm->i915, &vma->active,
- __i915_vma_active, __i915_vma_retire);
+ i915_active_init(&vma->active, __i915_vma_active, __i915_vma_retire);
/* Declare ourselves safe for use inside shrinkers */
if (IS_ENABLED(CONFIG_LOCKDEP)) {
@@ -1148,6 +1147,7 @@ int __i915_vma_unbind(struct i915_vma *vma)
if (ret)
return ret;
+ GEM_BUG_ON(i915_vma_is_active(vma));
if (i915_vma_is_pinned(vma)) {
vma_print_allocator(vma, "is pinned");
return -EBUSY;
diff --git a/drivers/gpu/drm/i915/selftests/i915_active.c b/drivers/gpu/drm/i915/selftests/i915_active.c
index a41785822ed9..2cc71bcf884f 100644
--- a/drivers/gpu/drm/i915/selftests/i915_active.c
+++ b/drivers/gpu/drm/i915/selftests/i915_active.c
@@ -68,7 +68,7 @@ static struct live_active *__live_alloc(struct drm_i915_private *i915)
return NULL;
kref_init(&active->ref);
- i915_active_init(i915, &active->base, __live_active, __live_retire);
+ i915_active_init(&active->base, __live_active, __live_retire);
return active;
}
@@ -146,19 +146,13 @@ static int live_active_wait(void *arg)
{
struct drm_i915_private *i915 = arg;
struct live_active *active;
- intel_wakeref_t wakeref;
int err = 0;
/* Check that we get a callback when requests retire upon waiting */
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
active = __live_active_setup(i915);
- if (IS_ERR(active)) {
- err = PTR_ERR(active);
- goto err;
- }
+ if (IS_ERR(active))
+ return PTR_ERR(active);
i915_active_wait(&active->base);
if (!READ_ONCE(active->retired)) {
@@ -168,11 +162,9 @@ static int live_active_wait(void *arg)
__live_put(active);
+ mutex_lock(&i915->drm.struct_mutex);
if (igt_flush_test(i915, I915_WAIT_LOCKED))
err = -EIO;
-
-err:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
mutex_unlock(&i915->drm.struct_mutex);
return err;
@@ -182,23 +174,19 @@ static int live_active_retire(void *arg)
{
struct drm_i915_private *i915 = arg;
struct live_active *active;
- intel_wakeref_t wakeref;
int err = 0;
/* Check that we get a callback when requests are indirectly retired */
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
active = __live_active_setup(i915);
- if (IS_ERR(active)) {
- err = PTR_ERR(active);
- goto err;
- }
+ if (IS_ERR(active))
+ return PTR_ERR(active);
/* waits for & retires all requests */
+ mutex_lock(&i915->drm.struct_mutex);
if (igt_flush_test(i915, I915_WAIT_LOCKED))
err = -EIO;
+ mutex_unlock(&i915->drm.struct_mutex);
if (!READ_ONCE(active->retired)) {
pr_err("i915_active not retired after flushing!\n");
@@ -207,10 +195,6 @@ static int live_active_retire(void *arg)
__live_put(active);
-err:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
-
return err;
}
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* Re: [PATCH 07/27] drm/i915: Coordinate i915_active with its own mutex
2019-09-25 10:01 ` [PATCH 07/27] drm/i915: Coordinate i915_active with its own mutex Chris Wilson
@ 2019-09-27 11:10 ` Tvrtko Ursulin
2019-09-27 11:25 ` Chris Wilson
0 siblings, 1 reply; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-27 11:10 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 25/09/2019 11:01, Chris Wilson wrote:
> Forgo the struct_mutex serialisation for i915_active, and interpose its
> own mutex handling for active/retire.
>
> This is a multi-layered sleight-of-hand. First, we had to ensure that no
> active/retire callbacks accidentally inverted the mutex ordering rules,
> nor assumed that they were themselves serialised by struct_mutex. More
> challenging though, is the rule over updating elements of the active
> rbtree. Instead of the whole i915_active now being serialised by
> struct_mutex, allocations/rotations of the tree are serialised by the
> i915_active.mutex and individual nodes are serialised by the caller
> using the i915_timeline.mutex (we need to use nested spinlocks to
> interact with the dma_fence callback lists).
>
> The pain point here is that instead of a single mutex around execbuf, we
> now have to take a mutex for active tracker (one for each vma, context,
> etc) and a couple of spinlocks for each fence update. The improvement in
> fine grained locking allowing for multiple concurrent clients
> (eventually!) should be worth it in typical loads.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> .../gpu/drm/i915/display/intel_frontbuffer.c | 2 +-
> drivers/gpu/drm/i915/display/intel_overlay.c | 3 +-
> drivers/gpu/drm/i915/gem/i915_gem_context.c | 4 +-
> .../gpu/drm/i915/gem/i915_gem_object_types.h | 1 +
> drivers/gpu/drm/i915/gem/i915_gem_pm.c | 9 +-
> drivers/gpu/drm/i915/gt/intel_context.c | 4 +-
> drivers/gpu/drm/i915/gt/intel_engine_pool.c | 2 +-
> drivers/gpu/drm/i915/gt/intel_reset.c | 10 +-
> drivers/gpu/drm/i915/gt/intel_timeline.c | 7 +-
> .../gpu/drm/i915/gt/intel_timeline_types.h | 2 +-
> drivers/gpu/drm/i915/gt/selftest_context.c | 16 +-
> drivers/gpu/drm/i915/gt/selftest_lrc.c | 10 +-
> .../gpu/drm/i915/gt/selftests/mock_timeline.c | 2 +-
> drivers/gpu/drm/i915/gvt/scheduler.c | 3 -
> drivers/gpu/drm/i915/i915_active.c | 289 +++++++---------
> drivers/gpu/drm/i915/i915_active.h | 319 ++++--------------
> drivers/gpu/drm/i915/i915_active_types.h | 23 +-
> drivers/gpu/drm/i915/i915_gem.c | 42 ++-
> drivers/gpu/drm/i915/i915_gem_gtt.c | 3 +-
> drivers/gpu/drm/i915/i915_gpu_error.c | 4 +-
> drivers/gpu/drm/i915/i915_request.c | 38 +--
> drivers/gpu/drm/i915/i915_request.h | 1 -
> drivers/gpu/drm/i915/i915_vma.c | 4 +-
> drivers/gpu/drm/i915/selftests/i915_active.c | 32 +-
> 24 files changed, 264 insertions(+), 566 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/display/intel_frontbuffer.c b/drivers/gpu/drm/i915/display/intel_frontbuffer.c
> index 6428b8dd70d3..84b164f31895 100644
> --- a/drivers/gpu/drm/i915/display/intel_frontbuffer.c
> +++ b/drivers/gpu/drm/i915/display/intel_frontbuffer.c
> @@ -257,7 +257,7 @@ intel_frontbuffer_get(struct drm_i915_gem_object *obj)
> front->obj = obj;
> kref_init(&front->ref);
> atomic_set(&front->bits, 0);
> - i915_active_init(i915, &front->write,
> + i915_active_init(&front->write,
> frontbuffer_active,
> i915_active_may_sleep(frontbuffer_retire));
>
> diff --git a/drivers/gpu/drm/i915/display/intel_overlay.c b/drivers/gpu/drm/i915/display/intel_overlay.c
> index 3f4ac1ee7668..e12e1a753af0 100644
> --- a/drivers/gpu/drm/i915/display/intel_overlay.c
> +++ b/drivers/gpu/drm/i915/display/intel_overlay.c
> @@ -1360,8 +1360,7 @@ void intel_overlay_setup(struct drm_i915_private *dev_priv)
> overlay->contrast = 75;
> overlay->saturation = 146;
>
> - i915_active_init(dev_priv,
> - &overlay->last_flip,
> + i915_active_init(&overlay->last_flip,
> NULL, intel_overlay_last_flip_retire);
>
> ret = get_registers(overlay, OVERLAY_NEEDS_PHYSICAL(dev_priv));
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
> index 4cd7d2ecf1d5..9d85aab68d34 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
> @@ -868,20 +868,18 @@ static int context_barrier_task(struct i915_gem_context *ctx,
> void (*task)(void *data),
> void *data)
> {
> - struct drm_i915_private *i915 = ctx->i915;
> struct context_barrier_task *cb;
> struct i915_gem_engines_iter it;
> struct intel_context *ce;
> int err = 0;
>
> - lockdep_assert_held(&i915->drm.struct_mutex);
> GEM_BUG_ON(!task);
>
> cb = kmalloc(sizeof(*cb), GFP_KERNEL);
> if (!cb)
> return -ENOMEM;
>
> - i915_active_init(i915, &cb->base, NULL, cb_retire);
> + i915_active_init(&cb->base, NULL, cb_retire);
> err = i915_active_acquire(&cb->base);
> if (err) {
> kfree(cb);
> 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 d695f187b790..e291ddfeb7dd 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
> @@ -8,6 +8,7 @@
> #define __I915_GEM_OBJECT_TYPES_H__
>
> #include <drm/drm_gem.h>
> +#include <uapi/drm/i915_drm.h>
>
> #include "i915_active.h"
> #include "i915_selftest.h"
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
> index a11ad4d914ca..3a47e1190ce1 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
> @@ -16,14 +16,11 @@ 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 =
> + struct dma_fence_cb *cb =
> container_of((struct list_head *)node,
> - typeof(*active), link);
> + typeof(*cb), node);
>
> - INIT_LIST_HEAD(&active->link);
> - RCU_INIT_POINTER(active->request, NULL);
> -
> - active->retire(active, NULL);
> + cb->func(NULL, cb);
> }
> }
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_context.c b/drivers/gpu/drm/i915/gt/intel_context.c
> index 06fabdf205cf..35a40c2820a2 100644
> --- a/drivers/gpu/drm/i915/gt/intel_context.c
> +++ b/drivers/gpu/drm/i915/gt/intel_context.c
> @@ -240,7 +240,7 @@ intel_context_init(struct intel_context *ce,
>
> mutex_init(&ce->pin_mutex);
>
> - i915_active_init(ctx->i915, &ce->active,
> + i915_active_init(&ce->active,
> __intel_context_active, __intel_context_retire);
> }
>
> @@ -307,7 +307,7 @@ int intel_context_prepare_remote_request(struct intel_context *ce,
> return err;
>
> /* Queue this switch after current activity by this context. */
> - err = i915_active_request_set(&tl->last_request, rq);
> + err = i915_active_fence_set(&tl->last_request, rq);
> mutex_unlock(&tl->mutex);
> if (err)
> return err;
> diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pool.c b/drivers/gpu/drm/i915/gt/intel_engine_pool.c
> index 81fab101fdb4..3cdbd5f8b5be 100644
> --- a/drivers/gpu/drm/i915/gt/intel_engine_pool.c
> +++ b/drivers/gpu/drm/i915/gt/intel_engine_pool.c
> @@ -95,7 +95,7 @@ node_create(struct intel_engine_pool *pool, size_t sz)
> return ERR_PTR(-ENOMEM);
>
> node->pool = pool;
> - i915_active_init(engine->i915, &node->active, pool_active, pool_retire);
> + i915_active_init(&node->active, pool_active, pool_retire);
>
> obj = i915_gem_object_create_internal(engine->i915, sz);
> if (IS_ERR(obj)) {
> diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c
> index ae68c3786f63..e8c8d0041e8f 100644
> --- a/drivers/gpu/drm/i915/gt/intel_reset.c
> +++ b/drivers/gpu/drm/i915/gt/intel_reset.c
> @@ -828,10 +828,10 @@ static bool __intel_gt_unset_wedged(struct intel_gt *gt)
> */
> spin_lock_irqsave(&timelines->lock, flags);
> list_for_each_entry(tl, &timelines->active_list, link) {
> - struct i915_request *rq;
> + struct dma_fence *fence;
>
> - rq = i915_active_request_get_unlocked(&tl->last_request);
> - if (!rq)
> + fence = i915_active_fence_get(&tl->last_request);
> + if (!fence)
> continue;
>
> spin_unlock_irqrestore(&timelines->lock, flags);
> @@ -843,8 +843,8 @@ static bool __intel_gt_unset_wedged(struct intel_gt *gt)
> * (I915_FENCE_TIMEOUT) so this wait should not be unbounded
> * in the worst case.
> */
> - dma_fence_default_wait(&rq->fence, false, MAX_SCHEDULE_TIMEOUT);
> - i915_request_put(rq);
> + dma_fence_default_wait(fence, false, MAX_SCHEDULE_TIMEOUT);
> + dma_fence_put(fence);
>
> /* Restart iteration after droping lock */
> spin_lock_irqsave(&timelines->lock, flags);
> diff --git a/drivers/gpu/drm/i915/gt/intel_timeline.c b/drivers/gpu/drm/i915/gt/intel_timeline.c
> index 653f60e78392..0f959694303c 100644
> --- a/drivers/gpu/drm/i915/gt/intel_timeline.c
> +++ b/drivers/gpu/drm/i915/gt/intel_timeline.c
> @@ -178,8 +178,7 @@ cacheline_alloc(struct intel_timeline_hwsp *hwsp, unsigned int cacheline)
> cl->hwsp = hwsp;
> cl->vaddr = page_pack_bits(vaddr, cacheline);
>
> - i915_active_init(hwsp->gt->i915, &cl->active,
> - __cacheline_active, __cacheline_retire);
> + i915_active_init(&cl->active, __cacheline_active, __cacheline_retire);
>
> return cl;
> }
> @@ -255,7 +254,7 @@ int intel_timeline_init(struct intel_timeline *timeline,
>
> mutex_init(&timeline->mutex);
>
> - INIT_ACTIVE_REQUEST(&timeline->last_request, &timeline->mutex);
> + INIT_ACTIVE_FENCE(&timeline->last_request, &timeline->mutex);
> INIT_LIST_HEAD(&timeline->requests);
>
> i915_syncmap_init(&timeline->sync);
> @@ -443,7 +442,7 @@ __intel_timeline_get_seqno(struct intel_timeline *tl,
> * free it after the current request is retired, which ensures that
> * all writes into the cacheline from previous requests are complete.
> */
> - err = i915_active_ref(&tl->hwsp_cacheline->active, tl, rq);
> + err = i915_active_ref(&tl->hwsp_cacheline->active, tl, &rq->fence);
> if (err)
> goto err_cacheline;
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_timeline_types.h b/drivers/gpu/drm/i915/gt/intel_timeline_types.h
> index c668c4c50e75..7faf81f94355 100644
> --- a/drivers/gpu/drm/i915/gt/intel_timeline_types.h
> +++ b/drivers/gpu/drm/i915/gt/intel_timeline_types.h
> @@ -63,7 +63,7 @@ struct intel_timeline {
> * the request using i915_active_request_get_request_rcu(), or hold
This was probably referring to __i915_active_request_get_rcu which this
patch removes in which case the comment should be updated.
the
> * struct_mutex.
> */
> - struct i915_active_request last_request;
> + struct i915_active_fence last_request;
>
> /**
> * We track the most recent seqno that we wait on in every context so
> diff --git a/drivers/gpu/drm/i915/gt/selftest_context.c b/drivers/gpu/drm/i915/gt/selftest_context.c
> index 4ce1e25433d2..e6bcbe7ab5e1 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_context.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_context.c
> @@ -47,24 +47,20 @@ static int context_sync(struct intel_context *ce)
>
> mutex_lock(&tl->mutex);
> do {
> - struct i915_request *rq;
> + struct dma_fence *fence;
> long timeout;
>
> - rcu_read_lock();
> - rq = rcu_dereference(tl->last_request.request);
> - if (rq)
> - rq = i915_request_get_rcu(rq);
> - rcu_read_unlock();
> - if (!rq)
> + fence = i915_active_fence_get(&tl->last_request);
> + if (!fence)
> break;
>
> - timeout = i915_request_wait(rq, 0, HZ / 10);
> + timeout = dma_fence_wait_timeout(fence, false, HZ / 10);
> if (timeout < 0)
> err = timeout;
> else
> - i915_request_retire_upto(rq);
> + i915_request_retire_upto(to_request(fence));
>
> - i915_request_put(rq);
> + dma_fence_put(fence);
> } while (!err);
> mutex_unlock(&tl->mutex);
>
> diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
> index 22ea2e747064..8eec4ead12a4 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
> @@ -994,9 +994,13 @@ static struct i915_request *dummy_request(struct intel_engine_cs *engine)
> if (!rq)
> return NULL;
>
> - INIT_LIST_HEAD(&rq->active_list);
> rq->engine = engine;
>
> + spin_lock_init(&rq->lock);
> + INIT_LIST_HEAD(&rq->fence.cb_list);
> + rq->fence.lock = &rq->lock;
> + rq->fence.ops = &i915_fence_ops;
> +
> i915_sched_node_init(&rq->sched);
>
> /* mark this request as permanently incomplete */
> @@ -1089,8 +1093,8 @@ static int live_suppress_wait_preempt(void *arg)
> }
>
> /* Disable NEWCLIENT promotion */
> - __i915_active_request_set(&i915_request_timeline(rq[i])->last_request,
> - dummy);
> + __i915_active_fence_set(&i915_request_timeline(rq[i])->last_request,
> + &dummy->fence);
> i915_request_add(rq[i]);
> }
>
> diff --git a/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c b/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c
> index 598170efcaf6..2a77c051f36a 100644
> --- a/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c
> +++ b/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c
> @@ -15,7 +15,7 @@ void mock_timeline_init(struct intel_timeline *timeline, u64 context)
>
> mutex_init(&timeline->mutex);
>
> - INIT_ACTIVE_REQUEST(&timeline->last_request, &timeline->mutex);
> + INIT_ACTIVE_FENCE(&timeline->last_request, &timeline->mutex);
> INIT_LIST_HEAD(&timeline->requests);
>
> i915_syncmap_init(&timeline->sync);
> diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
> index 6c79d16b381e..03f567084548 100644
> --- a/drivers/gpu/drm/i915/gvt/scheduler.c
> +++ b/drivers/gpu/drm/i915/gvt/scheduler.c
> @@ -385,11 +385,8 @@ intel_gvt_workload_req_alloc(struct intel_vgpu_workload *workload)
> {
> struct intel_vgpu *vgpu = workload->vgpu;
> struct intel_vgpu_submission *s = &vgpu->submission;
> - struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
> struct i915_request *rq;
>
> - lockdep_assert_held(&dev_priv->drm.struct_mutex);
> -
> if (workload->req)
> return 0;
>
> diff --git a/drivers/gpu/drm/i915/i915_active.c b/drivers/gpu/drm/i915/i915_active.c
> index 7ca066688b98..f23590ec9e68 100644
> --- a/drivers/gpu/drm/i915/i915_active.c
> +++ b/drivers/gpu/drm/i915/i915_active.c
> @@ -12,8 +12,6 @@
> #include "i915_active.h"
> #include "i915_globals.h"
>
> -#define BKL(ref) (&(ref)->i915->drm.struct_mutex)
> -
> /*
> * Active refs memory management
> *
> @@ -27,35 +25,35 @@ static struct i915_global_active {
> } global;
>
> struct active_node {
> - struct i915_active_request base;
> + struct i915_active_fence base;
> struct i915_active *ref;
> struct rb_node node;
> u64 timeline;
> };
>
> static inline struct active_node *
> -node_from_active(struct i915_active_request *active)
> +node_from_active(struct i915_active_fence *active)
> {
> return container_of(active, struct active_node, base);
> }
>
> #define take_preallocated_barriers(x) llist_del_all(&(x)->preallocated_barriers)
>
> -static inline bool is_barrier(const struct i915_active_request *active)
> +static inline bool is_barrier(const struct i915_active_fence *active)
> {
> - return IS_ERR(rcu_access_pointer(active->request));
> + return IS_ERR(rcu_access_pointer(active->fence));
> }
>
> static inline struct llist_node *barrier_to_ll(struct active_node *node)
> {
> GEM_BUG_ON(!is_barrier(&node->base));
> - return (struct llist_node *)&node->base.link;
> + return (struct llist_node *)&node->base.cb.node;
> }
>
> static inline struct intel_engine_cs *
> __barrier_to_engine(struct active_node *node)
> {
> - return (struct intel_engine_cs *)READ_ONCE(node->base.link.prev);
> + return (struct intel_engine_cs *)READ_ONCE(node->base.cb.node.prev);
> }
>
> static inline struct intel_engine_cs *
> @@ -68,7 +66,7 @@ barrier_to_engine(struct active_node *node)
> static inline struct active_node *barrier_from_ll(struct llist_node *x)
> {
> return container_of((struct list_head *)x,
> - struct active_node, base.link);
> + struct active_node, base.cb.node);
> }
>
> #if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM) && IS_ENABLED(CONFIG_DEBUG_OBJECTS)
> @@ -147,15 +145,18 @@ __active_retire(struct i915_active *ref)
> if (!retire)
> return;
>
> - GEM_BUG_ON(rcu_access_pointer(ref->excl));
> + GEM_BUG_ON(rcu_access_pointer(ref->excl.fence));
> rbtree_postorder_for_each_entry_safe(it, n, &root, node) {
> - GEM_BUG_ON(i915_active_request_isset(&it->base));
> + GEM_BUG_ON(i915_active_fence_isset(&it->base));
> kmem_cache_free(global.slab_cache, it);
> }
>
> /* After the final retire, the entire struct may be freed */
> if (ref->retire)
> ref->retire(ref);
> +
> + /* ... except if you wait on it, you must manage your own references! */
> + wake_up_var(ref);
> }
>
> static void
> @@ -189,12 +190,20 @@ active_retire(struct i915_active *ref)
> }
>
> static void
> -node_retire(struct i915_active_request *base, struct i915_request *rq)
> +node_retire(struct dma_fence *fence, struct dma_fence_cb *cb)
> {
> - active_retire(node_from_active(base)->ref);
> + i915_active_fence_cb(fence, cb);
> + active_retire(container_of(cb, struct active_node, base.cb)->ref);
> }
>
> -static struct i915_active_request *
> +static void
> +excl_retire(struct dma_fence *fence, struct dma_fence_cb *cb)
> +{
> + i915_active_fence_cb(fence, cb);
> + active_retire(container_of(cb, struct i915_active, excl.cb));
> +}
> +
> +static struct i915_active_fence *
> active_instance(struct i915_active *ref, struct intel_timeline *tl)
> {
> struct active_node *node, *prealloc;
> @@ -238,7 +247,7 @@ active_instance(struct i915_active *ref, struct intel_timeline *tl)
> }
>
> node = prealloc;
> - i915_active_request_init(&node->base, &tl->mutex, NULL, node_retire);
> + __i915_active_fence_init(&node->base, &tl->mutex, NULL, node_retire);
> node->ref = ref;
> node->timeline = idx;
>
> @@ -253,8 +262,7 @@ active_instance(struct i915_active *ref, struct intel_timeline *tl)
> return &node->base;
> }
>
> -void __i915_active_init(struct drm_i915_private *i915,
> - struct i915_active *ref,
> +void __i915_active_init(struct i915_active *ref,
> int (*active)(struct i915_active *ref),
> void (*retire)(struct i915_active *ref),
> struct lock_class_key *key)
> @@ -263,19 +271,18 @@ void __i915_active_init(struct drm_i915_private *i915,
>
> debug_active_init(ref);
>
> - ref->i915 = i915;
> ref->flags = 0;
> ref->active = active;
> ref->retire = ptr_unpack_bits(retire, &bits, 2);
> if (bits & I915_ACTIVE_MAY_SLEEP)
> ref->flags |= I915_ACTIVE_RETIRE_SLEEPS;
>
> - ref->excl = NULL;
> ref->tree = RB_ROOT;
> ref->cache = NULL;
> init_llist_head(&ref->preallocated_barriers);
> atomic_set(&ref->count, 0);
> __mutex_init(&ref->mutex, "i915_active", key);
> + __i915_active_fence_init(&ref->excl, &ref->mutex, NULL, excl_retire);
> INIT_WORK(&ref->work, active_work);
> }
>
> @@ -329,9 +336,9 @@ __active_del_barrier(struct i915_active *ref, struct active_node *node)
>
> int i915_active_ref(struct i915_active *ref,
> struct intel_timeline *tl,
> - struct i915_request *rq)
> + struct dma_fence *fence)
> {
> - struct i915_active_request *active;
> + struct i915_active_fence *active;
> int err;
>
> lockdep_assert_held(&tl->mutex);
> @@ -354,66 +361,39 @@ int i915_active_ref(struct i915_active *ref,
> * request that we want to emit on the kernel_context.
> */
> __active_del_barrier(ref, node_from_active(active));
> - RCU_INIT_POINTER(active->request, NULL);
> - INIT_LIST_HEAD(&active->link);
> - } else {
> - if (!i915_active_request_isset(active))
> - atomic_inc(&ref->count);
> + RCU_INIT_POINTER(active->fence, NULL);
> + atomic_dec(&ref->count);
> }
> - GEM_BUG_ON(!atomic_read(&ref->count));
> - __i915_active_request_set(active, rq);
> + if (!__i915_active_fence_set(active, fence))
> + atomic_inc(&ref->count);
>
> out:
> i915_active_release(ref);
> return err;
> }
>
> -static void excl_cb(struct dma_fence *f, struct dma_fence_cb *cb)
> -{
> - struct i915_active *ref = container_of(cb, typeof(*ref), excl_cb);
> -
> - RCU_INIT_POINTER(ref->excl, NULL);
> - dma_fence_put(f);
> -
> - active_retire(ref);
> -}
> -
> void i915_active_set_exclusive(struct i915_active *ref, struct dma_fence *f)
> {
> /* We expect the caller to manage the exclusive timeline ordering */
> GEM_BUG_ON(i915_active_is_idle(ref));
>
> - dma_fence_get(f);
> -
> - rcu_read_lock();
> - if (rcu_access_pointer(ref->excl)) {
> - struct dma_fence *old;
> -
> - old = dma_fence_get_rcu_safe(&ref->excl);
> - if (old) {
> - if (dma_fence_remove_callback(old, &ref->excl_cb))
> - atomic_dec(&ref->count);
> - dma_fence_put(old);
> - }
> - }
> - rcu_read_unlock();
> -
> - atomic_inc(&ref->count);
> - rcu_assign_pointer(ref->excl, f);
> + mutex_acquire(&ref->mutex.dep_map, 0, 0, _THIS_IP_);
> + if (!__i915_active_fence_set(&ref->excl, f))
> + atomic_inc(&ref->count);
> + mutex_release(&ref->mutex.dep_map, 0, _THIS_IP_);
Comment for this block please to explain what's going on.
> +}
>
> - if (dma_fence_add_callback(f, &ref->excl_cb, excl_cb)) {
> - RCU_INIT_POINTER(ref->excl, NULL);
> - atomic_dec(&ref->count);
> - dma_fence_put(f);
> - }
> +bool i915_active_acquire_if_busy(struct i915_active *ref)
> +{
> + debug_active_assert(ref);
> + return atomic_add_unless(&ref->count, 1, 0);
> }
>
> int i915_active_acquire(struct i915_active *ref)
> {
> int err;
>
> - debug_active_assert(ref);
> - if (atomic_add_unless(&ref->count, 1, 0))
> + if (i915_active_acquire_if_busy(ref))
> return 0;
>
> err = mutex_lock_interruptible(&ref->mutex);
> @@ -438,121 +418,57 @@ void i915_active_release(struct i915_active *ref)
> active_retire(ref);
> }
>
> -static void __active_ungrab(struct i915_active *ref)
> -{
> - clear_and_wake_up_bit(I915_ACTIVE_GRAB_BIT, &ref->flags);
> -}
> -
> -bool i915_active_trygrab(struct i915_active *ref)
> +static void enable_signaling(struct i915_active_fence *active)
> {
> - debug_active_assert(ref);
> -
> - if (test_and_set_bit(I915_ACTIVE_GRAB_BIT, &ref->flags))
> - return false;
> -
> - if (!atomic_add_unless(&ref->count, 1, 0)) {
> - __active_ungrab(ref);
> - return false;
> - }
> -
> - return true;
> -}
> + struct dma_fence *fence;
>
> -void i915_active_ungrab(struct i915_active *ref)
> -{
> - GEM_BUG_ON(!test_bit(I915_ACTIVE_GRAB_BIT, &ref->flags));
> -
> - active_retire(ref);
> - __active_ungrab(ref);
> -}
> -
> -static int excl_wait(struct i915_active *ref)
> -{
> - struct dma_fence *old;
> - int err = 0;
> -
> - if (!rcu_access_pointer(ref->excl))
> - return 0;
> -
> - rcu_read_lock();
> - old = dma_fence_get_rcu_safe(&ref->excl);
> - rcu_read_unlock();
> - if (old) {
> - err = dma_fence_wait(old, true);
> - dma_fence_put(old);
> - }
> + fence = i915_active_fence_get(active);
> + if (!fence)
> + return;
>
> - return err;
> + dma_fence_enable_sw_signaling(fence);
> + dma_fence_put(fence);
> }
>
> int i915_active_wait(struct i915_active *ref)
> {
> struct active_node *it, *n;
> - int err;
> + int err = 0;
>
> might_sleep();
> - might_lock(&ref->mutex);
>
> - if (i915_active_is_idle(ref))
> + if (!i915_active_acquire_if_busy(ref))
> return 0;
>
> - err = mutex_lock_interruptible(&ref->mutex);
> - if (err)
> - return err;
> -
> - if (!atomic_add_unless(&ref->count, 1, 0)) {
> - mutex_unlock(&ref->mutex);
> - return 0;
> - }
> -
> - err = excl_wait(ref);
> - if (err)
> - goto out;
> -
> + /* Flush lazy signals */
> + enable_signaling(&ref->excl);
> rbtree_postorder_for_each_entry_safe(it, n, &ref->tree, node) {
> - if (is_barrier(&it->base)) { /* unconnected idle-barrier */
> - err = -EBUSY;
> - break;
> - }
> + if (is_barrier(&it->base)) /* unconnected idle barrier */
> + continue;
>
> - err = i915_active_request_retire(&it->base, BKL(ref));
> - if (err)
> - break;
> + enable_signaling(&it->base);
> }
> + /* Any fence added after the wait begins will not be auto-signaled */
>
> -out:
> - __active_retire(ref);
> + i915_active_release(ref);
> if (err)
> return err;
>
> - if (wait_on_bit(&ref->flags, I915_ACTIVE_GRAB_BIT, TASK_KILLABLE))
> + if (wait_var_event_interruptible(ref, i915_active_is_idle(ref)))
> return -EINTR;
>
> - flush_work(&ref->work);
> - if (!i915_active_is_idle(ref))
> - return -EBUSY;
> -
> return 0;
> }
>
> -int i915_request_await_active_request(struct i915_request *rq,
> - struct i915_active_request *active)
> -{
> - struct i915_request *barrier =
> - i915_active_request_raw(active, &rq->i915->drm.struct_mutex);
> -
> - return barrier ? i915_request_await_dma_fence(rq, &barrier->fence) : 0;
> -}
> -
> int i915_request_await_active(struct i915_request *rq, struct i915_active *ref)
> {
> int err = 0;
>
> - if (rcu_access_pointer(ref->excl)) {
> + if (rcu_access_pointer(ref->excl.fence)) {
> struct dma_fence *fence;
>
> rcu_read_lock();
> - fence = dma_fence_get_rcu_safe(&ref->excl);
> + fence = dma_fence_get_rcu_safe(&ref->excl.fence);
> rcu_read_unlock();
> if (fence) {
> err = i915_request_await_dma_fence(rq, fence);
> @@ -578,7 +494,7 @@ void i915_active_fini(struct i915_active *ref)
>
> static inline bool is_idle_barrier(struct active_node *node, u64 idx)
> {
> - return node->timeline == idx && !i915_active_request_isset(&node->base);
> + return node->timeline == idx && !i915_active_fence_isset(&node->base);
> }
>
> static struct active_node *reuse_idle_barrier(struct i915_active *ref, u64 idx)
> @@ -698,13 +614,13 @@ int i915_active_acquire_preallocate_barrier(struct i915_active *ref,
> node->base.lock =
> &engine->kernel_context->timeline->mutex;
> #endif
> - RCU_INIT_POINTER(node->base.request, NULL);
> - node->base.retire = node_retire;
> + RCU_INIT_POINTER(node->base.fence, NULL);
> + node->base.cb.func = node_retire;
> node->timeline = idx;
> node->ref = ref;
> }
>
> - if (!i915_active_request_isset(&node->base)) {
> + if (!i915_active_fence_isset(&node->base)) {
> /*
> * Mark this as being *our* unconnected proto-node.
> *
> @@ -714,8 +630,8 @@ int i915_active_acquire_preallocate_barrier(struct i915_active *ref,
> * and then we can use the rb_node and list pointers
> * for our tracking of the pending barrier.
> */
> - RCU_INIT_POINTER(node->base.request, ERR_PTR(-EAGAIN));
> - node->base.link.prev = (void *)engine;
> + RCU_INIT_POINTER(node->base.fence, ERR_PTR(-EAGAIN));
> + node->base.cb.node.prev = (void *)engine;
> atomic_inc(&ref->count);
> }
>
> @@ -782,25 +698,65 @@ void i915_request_add_active_barriers(struct i915_request *rq)
> {
> struct intel_engine_cs *engine = rq->engine;
> struct llist_node *node, *next;
> + unsigned long flags;
>
> GEM_BUG_ON(intel_engine_is_virtual(engine));
> GEM_BUG_ON(i915_request_timeline(rq) != engine->kernel_context->timeline);
>
> + node = llist_del_all(&engine->barrier_tasks);
> + if (!node)
> + return;
> /*
> * Attach the list of proto-fences to the in-flight request such
> * that the parent i915_active will be released when this request
> * is retired.
> */
> - llist_for_each_safe(node, next, llist_del_all(&engine->barrier_tasks)) {
> - RCU_INIT_POINTER(barrier_from_ll(node)->base.request, rq);
> + spin_lock_irqsave(&rq->lock, flags);
> + llist_for_each_safe(node, next, node) {
> + RCU_INIT_POINTER(barrier_from_ll(node)->base.fence, &rq->fence);
> smp_wmb(); /* serialise with reuse_idle_barrier */
> - list_add_tail((struct list_head *)node, &rq->active_list);
> + list_add_tail((struct list_head *)node, &rq->fence.cb_list);
> + }
> + spin_unlock_irqrestore(&rq->lock, flags);
> +}
> +
> +#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
> +#define active_is_held(active) lockdep_is_held((active)->lock)
> +#else
> +#define active_is_held(active) true
> +#endif
> +
> +struct dma_fence *
> +__i915_active_fence_set(struct i915_active_fence *active,
> + struct dma_fence *fence)
Can you add a short comment for this function explaining the racyness
and so why it returns prev and what should the callers do with it?
> +{
> + struct dma_fence *prev;
> + unsigned long flags;
> +
> + /* NB: updates must be serialised by an outer timeline mutex */
Can't check here?
> + spin_lock_irqsave(fence->lock, flags);
> + GEM_BUG_ON(test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags));
> +
> + prev = rcu_dereference_protected(active->fence, active_is_held(active));
> + if (prev) {
> + spin_lock_nested(prev->lock, SINGLE_DEPTH_NESTING);
> + __list_del_entry(&active->cb.node);
> + spin_unlock(prev->lock); /* serialise with prev->cb_list */
> + prev = rcu_access_pointer(active->fence);
> }
> +
> + rcu_assign_pointer(active->fence, fence);
> + list_add_tail(&active->cb.node, &fence->cb_list);
> +
> + spin_unlock_irqrestore(fence->lock, flags);
> +
> + return prev;
> }
>
> -int i915_active_request_set(struct i915_active_request *active,
> - struct i915_request *rq)
> +int i915_active_fence_set(struct i915_active_fence *active,
> + struct i915_request *rq)
> {
> + struct dma_fence *fence;
> int err;
>
> #if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
Here not in diff we actually have:
#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
lockdep_assert_held(active->lock);
So why only under GEM debug and how does it related to the NB comment in
__i915_active_fence_set?
> @@ -808,18 +764,25 @@ int i915_active_request_set(struct i915_active_request *active,
> #endif
>
> /* Must maintain ordering wrt previous active requests */
> - err = i915_request_await_active_request(rq, active);
> - if (err)
> - return err;
> + rcu_read_lock();
> + fence = __i915_active_fence_set(active, &rq->fence);
> + if (fence)
> + fence = dma_fence_get_rcu(fence);
> + rcu_read_unlock();
> +
> + if (fence) {
> + err = i915_request_await_dma_fence(rq, fence);
> + dma_fence_put(fence);
> + if (err)
> + return err;
> + }
>
> - __i915_active_request_set(active, rq);
> return 0;
> }
>
> -void i915_active_retire_noop(struct i915_active_request *active,
> - struct i915_request *request)
> +void i915_active_noop(struct dma_fence *fence, struct dma_fence_cb *cb)
> {
> - /* Space left intentionally blank */
> + i915_active_fence_cb(fence, cb);
> }
>
> #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
> diff --git a/drivers/gpu/drm/i915/i915_active.h b/drivers/gpu/drm/i915/i915_active.h
> index 90034f61b7c2..4f52fe6146d2 100644
> --- a/drivers/gpu/drm/i915/i915_active.h
> +++ b/drivers/gpu/drm/i915/i915_active.h
> @@ -12,6 +12,10 @@
> #include "i915_active_types.h"
> #include "i915_request.h"
>
> +struct i915_request;
> +struct intel_engine_cs;
> +struct intel_timeline;
> +
> /*
> * We treat requests as fences. This is not be to confused with our
> * "fence registers" but pipeline synchronisation objects ala GL_ARB_sync.
> @@ -28,308 +32,108 @@
> * write access so that we can perform concurrent read operations between
> * the CPU and GPU engines, as well as waiting for all rendering to
> * complete, or waiting for the last GPU user of a "fence register". The
> - * object then embeds a #i915_active_request to track the most recent (in
> + * object then embeds a #i915_active_fence to track the most recent (in
> * retirement order) request relevant for the desired mode of access.
> - * The #i915_active_request is updated with i915_active_request_set() to
> + * The #i915_active_fence is updated with i915_active_fence_set() to
> * track the most recent fence request, typically this is done as part of
> * i915_vma_move_to_active().
> *
> - * When the #i915_active_request completes (is retired), it will
> + * When the #i915_active_fence completes (is retired), it will
> * signal its completion to the owner through a callback as well as mark
> - * itself as idle (i915_active_request.request == NULL). The owner
> + * itself as idle (i915_active_fence.request == NULL). The owner
> * can then perform any action, such as delayed freeing of an active
> * resource including itself.
> */
>
> -void i915_active_retire_noop(struct i915_active_request *active,
> - struct i915_request *request);
> +void i915_active_noop(struct dma_fence *fence, struct dma_fence_cb *cb);
>
> /**
> - * i915_active_request_init - prepares the activity tracker for use
> + * __i915_active_fence_init - prepares the activity tracker for use
> * @active - the active tracker
> - * @rq - initial request to track, can be NULL
> + * @fence - initial fence to track, can be NULL
> * @func - a callback when then the tracker is retired (becomes idle),
> * can be NULL
> *
> - * i915_active_request_init() prepares the embedded @active struct for use as
> - * an activity tracker, that is for tracking the last known active request
> - * associated with it. When the last request becomes idle, when it is retired
> + * i915_active_fence_init() prepares the embedded @active struct for use as
> + * an activity tracker, that is for tracking the last known active fence
> + * associated with it. When the last fence becomes idle, when it is retired
> * after completion, the optional callback @func is invoked.
> */
> static inline void
> -i915_active_request_init(struct i915_active_request *active,
> +__i915_active_fence_init(struct i915_active_fence *active,
> struct mutex *lock,
> - struct i915_request *rq,
> - i915_active_retire_fn retire)
> + void *fence,
> + dma_fence_func_t fn)
> {
> - RCU_INIT_POINTER(active->request, rq);
> - INIT_LIST_HEAD(&active->link);
> - active->retire = retire ?: i915_active_retire_noop;
> + RCU_INIT_POINTER(active->fence, fence);
> + active->cb.func = fn ?: i915_active_noop;
> #if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
> active->lock = lock;
> #endif
> }
>
> -#define INIT_ACTIVE_REQUEST(name, lock) \
> - i915_active_request_init((name), (lock), NULL, NULL)
> -
> -/**
> - * i915_active_request_set - updates the tracker to watch the current request
> - * @active - the active tracker
> - * @request - the request to watch
> - *
> - * __i915_active_request_set() watches the given @request for completion. Whilst
> - * that @request is busy, the @active reports busy. When that @request is
> - * retired, the @active tracker is updated to report idle.
> - */
> -static inline void
> -__i915_active_request_set(struct i915_active_request *active,
> - struct i915_request *request)
> -{
> -#if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
> - lockdep_assert_held(active->lock);
> -#endif
> - list_move(&active->link, &request->active_list);
> - rcu_assign_pointer(active->request, request);
> -}
> +#define INIT_ACTIVE_FENCE(A, LOCK) \
> + __i915_active_fence_init((A), (LOCK), NULL, NULL)
>
> -int __must_check
> -i915_active_request_set(struct i915_active_request *active,
> - struct i915_request *rq);
> +struct dma_fence *
> +__i915_active_fence_set(struct i915_active_fence *active,
> + struct dma_fence *fence);
>
> /**
> - * i915_active_request_raw - return the active request
> + * i915_active_fence_set - updates the tracker to watch the current fence
> * @active - the active tracker
> + * @rq - the request to watch
> *
> - * i915_active_request_raw() returns the current request being tracked, or NULL.
> - * It does not obtain a reference on the request for the caller, so the caller
> - * must hold struct_mutex.
> + * i915_active_fence_set() watches the given @rq for completion. While
> + * that @rq is busy, the @active reports busy. When that @rq is signaled
> + * (or else retired) the @active tracker is updated to report idle.
> */
> -static inline struct i915_request *
> -i915_active_request_raw(const struct i915_active_request *active,
> - struct mutex *mutex)
> -{
> - return rcu_dereference_protected(active->request,
> - lockdep_is_held(mutex));
> -}
> -
> -/**
> - * i915_active_request_peek - report the active request being monitored
> - * @active - the active tracker
> - *
> - * i915_active_request_peek() returns the current request being tracked if
> - * still active, or NULL. It does not obtain a reference on the request
> - * for the caller, so the caller must hold struct_mutex.
> - */
> -static inline struct i915_request *
> -i915_active_request_peek(const struct i915_active_request *active,
> - struct mutex *mutex)
> -{
> - struct i915_request *request;
> -
> - request = i915_active_request_raw(active, mutex);
> - if (!request || i915_request_completed(request))
> - return NULL;
> -
> - return request;
> -}
> -
> -/**
> - * i915_active_request_get - return a reference to the active request
> - * @active - the active tracker
> - *
> - * i915_active_request_get() returns a reference to the active request, or NULL
> - * if the active tracker is idle. The caller must hold struct_mutex.
> - */
> -static inline struct i915_request *
> -i915_active_request_get(const struct i915_active_request *active,
> - struct mutex *mutex)
> -{
> - return i915_request_get(i915_active_request_peek(active, mutex));
> -}
> -
> -/**
> - * __i915_active_request_get_rcu - return a reference to the active request
> - * @active - the active tracker
> - *
> - * __i915_active_request_get() returns a reference to the active request,
> - * or NULL if the active tracker is idle. The caller must hold the RCU read
> - * lock, but the returned pointer is safe to use outside of RCU.
> - */
> -static inline struct i915_request *
> -__i915_active_request_get_rcu(const struct i915_active_request *active)
> -{
> - /*
> - * Performing a lockless retrieval of the active request is super
> - * tricky. SLAB_TYPESAFE_BY_RCU merely guarantees that the backing
> - * slab of request objects will not be freed whilst we hold the
> - * RCU read lock. It does not guarantee that the request itself
> - * will not be freed and then *reused*. Viz,
> - *
> - * Thread A Thread B
> - *
> - * rq = active.request
> - * retire(rq) -> free(rq);
> - * (rq is now first on the slab freelist)
> - * active.request = NULL
> - *
> - * rq = new submission on a new object
> - * ref(rq)
> - *
> - * To prevent the request from being reused whilst the caller
> - * uses it, we take a reference like normal. Whilst acquiring
> - * the reference we check that it is not in a destroyed state
> - * (refcnt == 0). That prevents the request being reallocated
> - * whilst the caller holds on to it. To check that the request
> - * was not reallocated as we acquired the reference we have to
> - * check that our request remains the active request across
> - * the lookup, in the same manner as a seqlock. The visibility
> - * of the pointer versus the reference counting is controlled
> - * by using RCU barriers (rcu_dereference and rcu_assign_pointer).
> - *
> - * In the middle of all that, we inspect whether the request is
> - * complete. Retiring is lazy so the request may be completed long
> - * before the active tracker is updated. Querying whether the
> - * request is complete is far cheaper (as it involves no locked
> - * instructions setting cachelines to exclusive) than acquiring
> - * the reference, so we do it first. The RCU read lock ensures the
> - * pointer dereference is valid, but does not ensure that the
> - * seqno nor HWS is the right one! However, if the request was
> - * reallocated, that means the active tracker's request was complete.
> - * If the new request is also complete, then both are and we can
> - * just report the active tracker is idle. If the new request is
> - * incomplete, then we acquire a reference on it and check that
> - * it remained the active request.
> - *
> - * It is then imperative that we do not zero the request on
> - * reallocation, so that we can chase the dangling pointers!
> - * See i915_request_alloc().
> - */
> - do {
> - struct i915_request *request;
> -
> - request = rcu_dereference(active->request);
> - if (!request || i915_request_completed(request))
> - return NULL;
> -
> - /*
> - * An especially silly compiler could decide to recompute the
> - * result of i915_request_completed, more specifically
> - * re-emit the load for request->fence.seqno. A race would catch
> - * a later seqno value, which could flip the result from true to
> - * false. Which means part of the instructions below might not
> - * be executed, while later on instructions are executed. Due to
> - * barriers within the refcounting the inconsistency can't reach
> - * past the call to i915_request_get_rcu, but not executing
> - * that while still executing i915_request_put() creates
> - * havoc enough. Prevent this with a compiler barrier.
> - */
> - barrier();
> -
> - request = i915_request_get_rcu(request);
> -
> - /*
> - * What stops the following rcu_access_pointer() from occurring
> - * before the above i915_request_get_rcu()? If we were
> - * to read the value before pausing to get the reference to
> - * the request, we may not notice a change in the active
> - * tracker.
> - *
> - * The rcu_access_pointer() is a mere compiler barrier, which
> - * means both the CPU and compiler are free to perform the
> - * memory read without constraint. The compiler only has to
> - * ensure that any operations after the rcu_access_pointer()
> - * occur afterwards in program order. This means the read may
> - * be performed earlier by an out-of-order CPU, or adventurous
> - * compiler.
> - *
> - * The atomic operation at the heart of
> - * i915_request_get_rcu(), see dma_fence_get_rcu(), is
> - * atomic_inc_not_zero() which is only a full memory barrier
> - * when successful. That is, if i915_request_get_rcu()
> - * returns the request (and so with the reference counted
> - * incremented) then the following read for rcu_access_pointer()
> - * must occur after the atomic operation and so confirm
> - * that this request is the one currently being tracked.
> - *
> - * The corresponding write barrier is part of
> - * rcu_assign_pointer().
> - */
> - if (!request || request == rcu_access_pointer(active->request))
> - return rcu_pointer_handoff(request);
> -
> - i915_request_put(request);
> - } while (1);
> -}
> -
> +int __must_check
> +i915_active_fence_set(struct i915_active_fence *active,
> + struct i915_request *rq);
> /**
> - * i915_active_request_get_unlocked - return a reference to the active request
> + * i915_active_fence_get - return a reference to the active fence
> * @active - the active tracker
> *
> - * i915_active_request_get_unlocked() returns a reference to the active request,
> + * i915_active_fence_get() returns a reference to the active fence,
> * or NULL if the active tracker is idle. The reference is obtained under RCU,
> * so no locking is required by the caller.
> *
> - * The reference should be freed with i915_request_put().
> + * The reference should be freed with dma_fence_put().
> */
> -static inline struct i915_request *
> -i915_active_request_get_unlocked(const struct i915_active_request *active)
> +static inline struct dma_fence *
> +i915_active_fence_get(struct i915_active_fence *active)
> {
> - struct i915_request *request;
> + struct dma_fence *fence;
>
> rcu_read_lock();
> - request = __i915_active_request_get_rcu(active);
> + fence = dma_fence_get_rcu_safe(&active->fence);
> rcu_read_unlock();
>
> - return request;
> + return fence;
> }
>
> /**
> - * i915_active_request_isset - report whether the active tracker is assigned
> + * i915_active_fence_isset - report whether the active tracker is assigned
> * @active - the active tracker
> *
> - * i915_active_request_isset() returns true if the active tracker is currently
> - * assigned to a request. Due to the lazy retiring, that request may be idle
> + * i915_active_fence_isset() returns true if the active tracker is currently
> + * assigned to a fence. Due to the lazy retiring, that fence may be idle
> * and this may report stale information.
> */
> static inline bool
> -i915_active_request_isset(const struct i915_active_request *active)
> +i915_active_fence_isset(const struct i915_active_fence *active)
> {
> - return rcu_access_pointer(active->request);
> + return rcu_access_pointer(active->fence);
> }
>
> -/**
> - * i915_active_request_retire - waits until the request is retired
> - * @active - the active request on which to wait
> - *
> - * i915_active_request_retire() waits until the request is completed,
> - * and then ensures that at least the retirement handler for this
> - * @active tracker is called before returning. If the @active
> - * tracker is idle, the function returns immediately.
> - */
> -static inline int __must_check
> -i915_active_request_retire(struct i915_active_request *active,
> - struct mutex *mutex)
> +static inline void
> +i915_active_fence_cb(struct dma_fence *fence, struct dma_fence_cb *cb)
> {
> - struct i915_request *request;
> - long ret;
> -
> - request = i915_active_request_raw(active, mutex);
> - if (!request)
> - return 0;
> -
> - ret = i915_request_wait(request,
> - I915_WAIT_INTERRUPTIBLE,
> - MAX_SCHEDULE_TIMEOUT);
> - if (ret < 0)
> - return ret;
> + struct i915_active_fence *active =
> + container_of(cb, typeof(*active), cb);
>
> - list_del_init(&active->link);
> - RCU_INIT_POINTER(active->request, NULL);
> -
> - active->retire(active, request);
> -
> - return 0;
> + RCU_INIT_POINTER(active->fence, NULL);
> }
>
> /*
> @@ -358,47 +162,40 @@ i915_active_request_retire(struct i915_active_request *active,
> * synchronisation.
> */
>
> -void __i915_active_init(struct drm_i915_private *i915,
> - struct i915_active *ref,
> +void __i915_active_init(struct i915_active *ref,
> int (*active)(struct i915_active *ref),
> void (*retire)(struct i915_active *ref),
> struct lock_class_key *key);
> -#define i915_active_init(i915, ref, active, retire) do { \
> +#define i915_active_init(ref, active, retire) do { \
> static struct lock_class_key __key; \
> \
> - __i915_active_init(i915, ref, active, retire, &__key); \
> + __i915_active_init(ref, active, retire, &__key); \
> } while (0)
>
> int i915_active_ref(struct i915_active *ref,
> struct intel_timeline *tl,
> - struct i915_request *rq);
> + struct dma_fence *fence);
>
> static inline int
> i915_active_add_request(struct i915_active *ref, struct i915_request *rq)
> {
> - return i915_active_ref(ref, i915_request_timeline(rq), rq);
> + return i915_active_ref(ref, i915_request_timeline(rq), &rq->fence);
> }
>
> void i915_active_set_exclusive(struct i915_active *ref, struct dma_fence *f);
>
> static inline bool i915_active_has_exclusive(struct i915_active *ref)
> {
> - return rcu_access_pointer(ref->excl);
> + return rcu_access_pointer(ref->excl.fence);
> }
>
> int i915_active_wait(struct i915_active *ref);
>
> -int i915_request_await_active(struct i915_request *rq,
> - struct i915_active *ref);
> -int i915_request_await_active_request(struct i915_request *rq,
> - struct i915_active_request *active);
> +int i915_request_await_active(struct i915_request *rq, struct i915_active *ref);
>
> int i915_active_acquire(struct i915_active *ref);
> +bool i915_active_acquire_if_busy(struct i915_active *ref);
> void i915_active_release(struct i915_active *ref);
> -void __i915_active_release_nested(struct i915_active *ref, int subclass);
> -
> -bool i915_active_trygrab(struct i915_active *ref);
> -void i915_active_ungrab(struct i915_active *ref);
>
> static inline bool
> i915_active_is_idle(const struct i915_active *ref)
> diff --git a/drivers/gpu/drm/i915/i915_active_types.h b/drivers/gpu/drm/i915/i915_active_types.h
> index 021167f0004d..d89a74c142c6 100644
> --- a/drivers/gpu/drm/i915/i915_active_types.h
> +++ b/drivers/gpu/drm/i915/i915_active_types.h
> @@ -17,17 +17,9 @@
>
> #include "i915_utils.h"
>
> -struct drm_i915_private;
> -struct i915_active_request;
> -struct i915_request;
> -
> -typedef void (*i915_active_retire_fn)(struct i915_active_request *,
> - struct i915_request *);
> -
> -struct i915_active_request {
> - struct i915_request __rcu *request;
> - struct list_head link;
> - i915_active_retire_fn retire;
> +struct i915_active_fence {
> + struct dma_fence __rcu *fence;
> + struct dma_fence_cb cb;
> #if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
> /*
> * Incorporeal!
> @@ -53,20 +45,17 @@ struct active_node;
> #define i915_active_may_sleep(fn) ptr_pack_bits(&(fn), I915_ACTIVE_MAY_SLEEP, 2)
>
> struct i915_active {
> - struct drm_i915_private *i915;
> + atomic_t count;
> + struct mutex mutex;
>
> struct active_node *cache;
> struct rb_root tree;
> - struct mutex mutex;
> - atomic_t count;
>
> /* Preallocated "exclusive" node */
> - struct dma_fence __rcu *excl;
> - struct dma_fence_cb excl_cb;
> + struct i915_active_fence excl;
>
> unsigned long flags;
> #define I915_ACTIVE_RETIRE_SLEEPS BIT(0)
> -#define I915_ACTIVE_GRAB_BIT 1
>
> int (*active)(struct i915_active *ref);
> void (*retire)(struct i915_active *ref);
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index 934ea776a46a..a88849544c8b 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -892,28 +892,38 @@ wait_for_timelines(struct intel_gt *gt, unsigned int wait, long timeout)
>
> spin_lock_irqsave(&timelines->lock, flags);
> list_for_each_entry(tl, &timelines->active_list, link) {
> - struct i915_request *rq;
> + struct dma_fence *fence;
>
> - rq = i915_active_request_get_unlocked(&tl->last_request);
> - if (!rq)
> + fence = i915_active_fence_get(&tl->last_request);
> + if (!fence)
> continue;
>
> spin_unlock_irqrestore(&timelines->lock, flags);
>
> - /*
> - * "Race-to-idle".
> - *
> - * Switching to the kernel context is often used a synchronous
> - * step prior to idling, e.g. in suspend for flushing all
> - * current operations to memory before sleeping. These we
> - * want to complete as quickly as possible to avoid prolonged
> - * stalls, so allow the gpu to boost to maximum clocks.
> - */
> - if (wait & I915_WAIT_FOR_IDLE_BOOST)
> - gen6_rps_boost(rq);
> + if (!dma_fence_is_i915(fence)) {
> + timeout = dma_fence_wait_timeout(fence,
> + flags & I915_WAIT_INTERRUPTIBLE,
> + timeout);
> + } else {
> + struct i915_request *rq = to_request(fence);
> +
> + /*
> + * "Race-to-idle".
> + *
> + * Switching to the kernel context is often used as
> + * a synchronous step prior to idling, e.g. in suspend
> + * for flushing all current operations to memory before
> + * sleeping. These we want to complete as quickly as
> + * possible to avoid prolonged stalls, so allow the gpu
> + * to boost to maximum clocks.
> + */
> + if (flags & I915_WAIT_FOR_IDLE_BOOST)
> + gen6_rps_boost(rq);
> +
> + timeout = i915_request_wait(rq, flags, timeout);
> + }
>
> - timeout = i915_request_wait(rq, wait, timeout);
> - i915_request_put(rq);
> + dma_fence_put(fence);
> if (timeout < 0)
> return timeout;
>
> diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
> index a351b98afe03..bf57c41fa1fb 100644
> --- a/drivers/gpu/drm/i915/i915_gem_gtt.c
> +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
> @@ -1861,7 +1861,6 @@ static const struct i915_vma_ops pd_vma_ops = {
>
> static struct i915_vma *pd_vma_create(struct gen6_ppgtt *ppgtt, int size)
> {
> - struct drm_i915_private *i915 = ppgtt->base.vm.i915;
> struct i915_ggtt *ggtt = ppgtt->base.vm.gt->ggtt;
> struct i915_vma *vma;
>
> @@ -1872,7 +1871,7 @@ static struct i915_vma *pd_vma_create(struct gen6_ppgtt *ppgtt, int size)
> if (!vma)
> return ERR_PTR(-ENOMEM);
>
> - i915_active_init(i915, &vma->active, NULL, NULL);
> + i915_active_init(&vma->active, NULL, NULL);
>
> mutex_init(&vma->pages_mutex);
> vma->vm = i915_vm_get(&ggtt->vm);
> diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
> index 6384a06aa5bf..a28ee754b7b4 100644
> --- a/drivers/gpu/drm/i915/i915_gpu_error.c
> +++ b/drivers/gpu/drm/i915/i915_gpu_error.c
> @@ -1299,7 +1299,7 @@ capture_vma(struct capture_vma *next,
> if (!c)
> return next;
>
> - if (!i915_active_trygrab(&vma->active)) {
> + if (!i915_active_acquire_if_busy(&vma->active)) {
> kfree(c);
> return next;
> }
> @@ -1439,7 +1439,7 @@ gem_record_rings(struct i915_gpu_state *error, struct compress *compress)
> *this->slot =
> i915_error_object_create(i915, vma, compress);
>
> - i915_active_ungrab(&vma->active);
> + i915_active_release(&vma->active);
> i915_vma_put(vma);
>
> capture = this->next;
> diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
> index a8916412759b..4ffe62a42186 100644
> --- a/drivers/gpu/drm/i915/i915_request.c
> +++ b/drivers/gpu/drm/i915/i915_request.c
> @@ -218,8 +218,6 @@ static void remove_from_engine(struct i915_request *rq)
>
> static bool i915_request_retire(struct i915_request *rq)
> {
> - struct i915_active_request *active, *next;
> -
> if (!i915_request_completed(rq))
> return false;
>
> @@ -244,35 +242,6 @@ static bool i915_request_retire(struct i915_request *rq)
> &i915_request_timeline(rq)->requests));
> rq->ring->head = rq->postfix;
>
> - /*
> - * Walk through the active list, calling retire on each. This allows
> - * objects to track their GPU activity and mark themselves as idle
> - * when their *last* active request is completed (updating state
> - * tracking lists for eviction, active references for GEM, etc).
> - *
> - * As the ->retire() may free the node, we decouple it first and
> - * pass along the auxiliary information (to avoid dereferencing
> - * the node after the callback).
> - */
> - 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
> - * the retirement of requests and processing their callbacks.
> - * Of which, this loop itself is particularly hot due to the
> - * cache misses when jumping around the list of
> - * i915_active_request. So we try to keep this loop as
> - * streamlined as possible and also prefetch the next
> - * i915_active_request to try and hide the likely cache miss.
> - */
> - prefetchw(next);
> -
> - INIT_LIST_HEAD(&active->link);
> - RCU_INIT_POINTER(active->request, NULL);
> -
> - active->retire(active, rq);
> - }
> -
> local_irq_disable();
>
> /*
> @@ -704,7 +673,6 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp)
> rq->flags = 0;
> rq->execution_mask = ALL_ENGINES;
>
> - INIT_LIST_HEAD(&rq->active_list);
> INIT_LIST_HEAD(&rq->execute_cb);
>
> /*
> @@ -743,7 +711,6 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp)
> ce->ring->emit = rq->head;
>
> /* Make sure we didn't add ourselves to external state before freeing */
> - GEM_BUG_ON(!list_empty(&rq->active_list));
> GEM_BUG_ON(!list_empty(&rq->sched.signalers_list));
> GEM_BUG_ON(!list_empty(&rq->sched.waiters_list));
>
> @@ -1174,8 +1141,8 @@ __i915_request_add_to_timeline(struct i915_request *rq)
> * precludes optimising to use semaphores serialisation of a single
> * timeline across engines.
> */
> - prev = rcu_dereference_protected(timeline->last_request.request,
> - lockdep_is_held(&timeline->mutex));
> + prev = to_request(__i915_active_fence_set(&timeline->last_request,
> + &rq->fence));
> if (prev && !i915_request_completed(prev)) {
> if (is_power_of_2(prev->engine->mask | rq->engine->mask))
> i915_sw_fence_await_sw_fence(&rq->submit,
> @@ -1200,7 +1167,6 @@ __i915_request_add_to_timeline(struct i915_request *rq)
> * us, the timeline will hold its seqno which is later than ours.
> */
> GEM_BUG_ON(timeline->seqno != rq->fence.seqno);
> - __i915_active_request_set(&timeline->last_request, rq);
>
> return prev;
> }
> diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
> index ec5bb4c2e5ae..91a885c36c6b 100644
> --- a/drivers/gpu/drm/i915/i915_request.h
> +++ b/drivers/gpu/drm/i915/i915_request.h
> @@ -211,7 +211,6 @@ struct i915_request {
> * on the active_list (of their final request).
> */
> struct i915_capture_list *capture_list;
> - struct list_head active_list;
>
> /** Time at which this request was emitted, in jiffies. */
> unsigned long emitted_jiffies;
> diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
> index 2717441a9f3a..647bb96a6220 100644
> --- a/drivers/gpu/drm/i915/i915_vma.c
> +++ b/drivers/gpu/drm/i915/i915_vma.c
> @@ -120,8 +120,7 @@ vma_create(struct drm_i915_gem_object *obj,
> vma->size = obj->base.size;
> vma->display_alignment = I915_GTT_MIN_ALIGNMENT;
>
> - i915_active_init(vm->i915, &vma->active,
> - __i915_vma_active, __i915_vma_retire);
> + i915_active_init(&vma->active, __i915_vma_active, __i915_vma_retire);
>
> /* Declare ourselves safe for use inside shrinkers */
> if (IS_ENABLED(CONFIG_LOCKDEP)) {
> @@ -1148,6 +1147,7 @@ int __i915_vma_unbind(struct i915_vma *vma)
> if (ret)
> return ret;
>
> + GEM_BUG_ON(i915_vma_is_active(vma));
> if (i915_vma_is_pinned(vma)) {
> vma_print_allocator(vma, "is pinned");
> return -EBUSY;
> diff --git a/drivers/gpu/drm/i915/selftests/i915_active.c b/drivers/gpu/drm/i915/selftests/i915_active.c
> index a41785822ed9..2cc71bcf884f 100644
> --- a/drivers/gpu/drm/i915/selftests/i915_active.c
> +++ b/drivers/gpu/drm/i915/selftests/i915_active.c
> @@ -68,7 +68,7 @@ static struct live_active *__live_alloc(struct drm_i915_private *i915)
> return NULL;
>
> kref_init(&active->ref);
> - i915_active_init(i915, &active->base, __live_active, __live_retire);
> + i915_active_init(&active->base, __live_active, __live_retire);
>
> return active;
> }
> @@ -146,19 +146,13 @@ static int live_active_wait(void *arg)
> {
> struct drm_i915_private *i915 = arg;
> struct live_active *active;
> - intel_wakeref_t wakeref;
> int err = 0;
>
> /* Check that we get a callback when requests retire upon waiting */
>
> - mutex_lock(&i915->drm.struct_mutex);
> - wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> -
> active = __live_active_setup(i915);
> - if (IS_ERR(active)) {
> - err = PTR_ERR(active);
> - goto err;
> - }
> + if (IS_ERR(active))
> + return PTR_ERR(active);
>
> i915_active_wait(&active->base);
> if (!READ_ONCE(active->retired)) {
> @@ -168,11 +162,9 @@ static int live_active_wait(void *arg)
>
> __live_put(active);
>
> + mutex_lock(&i915->drm.struct_mutex);
> if (igt_flush_test(i915, I915_WAIT_LOCKED))
> err = -EIO;
> -
> -err:
> - intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> mutex_unlock(&i915->drm.struct_mutex);
>
> return err;
> @@ -182,23 +174,19 @@ static int live_active_retire(void *arg)
> {
> struct drm_i915_private *i915 = arg;
> struct live_active *active;
> - intel_wakeref_t wakeref;
> int err = 0;
>
> /* Check that we get a callback when requests are indirectly retired */
>
> - mutex_lock(&i915->drm.struct_mutex);
> - wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> -
> active = __live_active_setup(i915);
> - if (IS_ERR(active)) {
> - err = PTR_ERR(active);
> - goto err;
> - }
> + if (IS_ERR(active))
> + return PTR_ERR(active);
>
> /* waits for & retires all requests */
> + mutex_lock(&i915->drm.struct_mutex);
> if (igt_flush_test(i915, I915_WAIT_LOCKED))
> err = -EIO;
> + mutex_unlock(&i915->drm.struct_mutex);
>
> if (!READ_ONCE(active->retired)) {
> pr_err("i915_active not retired after flushing!\n");
> @@ -207,10 +195,6 @@ static int live_active_retire(void *arg)
>
> __live_put(active);
>
> -err:
> - intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> return err;
> }
>
>
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* Re: [PATCH 07/27] drm/i915: Coordinate i915_active with its own mutex
2019-09-27 11:10 ` Tvrtko Ursulin
@ 2019-09-27 11:25 ` Chris Wilson
2019-09-27 12:08 ` Tvrtko Ursulin
0 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-27 11:25 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Tvrtko Ursulin (2019-09-27 12:10:29)
>
> On 25/09/2019 11:01, Chris Wilson wrote:
> > void i915_active_set_exclusive(struct i915_active *ref, struct dma_fence *f)
> > {
> > /* We expect the caller to manage the exclusive timeline ordering */
> > GEM_BUG_ON(i915_active_is_idle(ref));
> >
> > - dma_fence_get(f);
> > -
> > - rcu_read_lock();
> > - if (rcu_access_pointer(ref->excl)) {
> > - struct dma_fence *old;
> > -
> > - old = dma_fence_get_rcu_safe(&ref->excl);
> > - if (old) {
> > - if (dma_fence_remove_callback(old, &ref->excl_cb))
> > - atomic_dec(&ref->count);
> > - dma_fence_put(old);
> > - }
> > - }
> > - rcu_read_unlock();
> > -
> > - atomic_inc(&ref->count);
> > - rcu_assign_pointer(ref->excl, f);
> > + mutex_acquire(&ref->mutex.dep_map, 0, 0, _THIS_IP_);
> > + if (!__i915_active_fence_set(&ref->excl, f))
> > + atomic_inc(&ref->count);
> > + mutex_release(&ref->mutex.dep_map, 0, _THIS_IP_);
>
> Comment for this block please to explain what's going on.
This was meant to be clued in by the opening comment that the caller is
expected to manage the exclusive timeline. But since the
i915_active_fence debug API requires us to declare what mutex guards
each timeline, we had to fake one.
> > +struct dma_fence *
> > +__i915_active_fence_set(struct i915_active_fence *active,
> > + struct dma_fence *fence)
>
> Can you add a short comment for this function explaining the racyness
> and so why it returns prev and what should the callers do with it?
Before using this function, we had the callers declare what mutex guards
this timeline and we check here that is held.
> > +{
> > + struct dma_fence *prev;
> > + unsigned long flags;
> > +
> > + /* NB: updates must be serialised by an outer timeline mutex */
>
> Can't check here?
We do, see active_is_held()
> > + spin_lock_irqsave(fence->lock, flags);
> > + GEM_BUG_ON(test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags));
> > +
> > + prev = rcu_dereference_protected(active->fence, active_is_held(active));
> > + if (prev) {
> > + spin_lock_nested(prev->lock, SINGLE_DEPTH_NESTING);
> > + __list_del_entry(&active->cb.node);
> > + spin_unlock(prev->lock); /* serialise with prev->cb_list */
> > + prev = rcu_access_pointer(active->fence);
> > }
> > +
> > + rcu_assign_pointer(active->fence, fence);
> > + list_add_tail(&active->cb.node, &fence->cb_list);
> > +
> > + spin_unlock_irqrestore(fence->lock, flags);
> > +
> > + return prev;
> > }
> >
> > -int i915_active_request_set(struct i915_active_request *active,
> > - struct i915_request *rq)
> > +int i915_active_fence_set(struct i915_active_fence *active,
> > + struct i915_request *rq)
> > {
> > + struct dma_fence *fence;
> > int err;
> >
> > #if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
>
> Here not in diff we actually have:
>
> #if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
> lockdep_assert_held(active->lock);
>
> So why only under GEM debug and how does it related to the NB comment in
> __i915_active_fence_set?
We only expand the struct when we want to for debugging.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* Re: [PATCH 07/27] drm/i915: Coordinate i915_active with its own mutex
2019-09-27 11:25 ` Chris Wilson
@ 2019-09-27 12:08 ` Tvrtko Ursulin
2019-09-27 12:16 ` Chris Wilson
0 siblings, 1 reply; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-27 12:08 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 27/09/2019 12:25, Chris Wilson wrote:
> Quoting Tvrtko Ursulin (2019-09-27 12:10:29)
>>
>> On 25/09/2019 11:01, Chris Wilson wrote:
>>> void i915_active_set_exclusive(struct i915_active *ref, struct dma_fence *f)
>>> {
>>> /* We expect the caller to manage the exclusive timeline ordering */
>>> GEM_BUG_ON(i915_active_is_idle(ref));
>>>
>>> - dma_fence_get(f);
>>> -
>>> - rcu_read_lock();
>>> - if (rcu_access_pointer(ref->excl)) {
>>> - struct dma_fence *old;
>>> -
>>> - old = dma_fence_get_rcu_safe(&ref->excl);
>>> - if (old) {
>>> - if (dma_fence_remove_callback(old, &ref->excl_cb))
>>> - atomic_dec(&ref->count);
>>> - dma_fence_put(old);
>>> - }
>>> - }
>>> - rcu_read_unlock();
>>> -
>>> - atomic_inc(&ref->count);
>>> - rcu_assign_pointer(ref->excl, f);
>>> + mutex_acquire(&ref->mutex.dep_map, 0, 0, _THIS_IP_);
>>> + if (!__i915_active_fence_set(&ref->excl, f))
>>> + atomic_inc(&ref->count);
>>> + mutex_release(&ref->mutex.dep_map, 0, _THIS_IP_);
>>
>> Comment for this block please to explain what's going on.
>
> This was meant to be clued in by the opening comment that the caller is
> expected to manage the exclusive timeline. But since the
> i915_active_fence debug API requires us to declare what mutex guards
> each timeline, we had to fake one.
I think the commentary about duality of which mutex guards things is
only present in the commit message at the moment and it would be really
good to have that somewhere in the code as well. Because after a few
months of refactoring, going from git blame to commits can stop working
so well, and then in code commentary becomes very valuable.
>>> +struct dma_fence *
>>> +__i915_active_fence_set(struct i915_active_fence *active,
>>> + struct dma_fence *fence)
>>
>> Can you add a short comment for this function explaining the racyness
>> and so why it returns prev and what should the callers do with it?
>
> Before using this function, we had the callers declare what mutex guards
> this timeline and we check here that is held.
No, I mean because it has to reload prev and return it to the caller,
which implies that is to handle some designed-in racy-ness which I think
should be mentioned.
>
>>> +{
>>> + struct dma_fence *prev;
>>> + unsigned long flags;
>>> +
>>> + /* NB: updates must be serialised by an outer timeline mutex */
>>
>> Can't check here?
>
> We do, see active_is_held()
Ack!
>>> + spin_lock_irqsave(fence->lock, flags);
>>> + GEM_BUG_ON(test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags));
>>> +
>>> + prev = rcu_dereference_protected(active->fence, active_is_held(active));
>>> + if (prev) {
>>> + spin_lock_nested(prev->lock, SINGLE_DEPTH_NESTING);
>>> + __list_del_entry(&active->cb.node);
>>> + spin_unlock(prev->lock); /* serialise with prev->cb_list */
>>> + prev = rcu_access_pointer(active->fence);
>>> }
>>> +
>>> + rcu_assign_pointer(active->fence, fence);
>>> + list_add_tail(&active->cb.node, &fence->cb_list);
>>> +
>>> + spin_unlock_irqrestore(fence->lock, flags);
>>> +
>>> + return prev;
>>> }
>>>
>>> -int i915_active_request_set(struct i915_active_request *active,
>>> - struct i915_request *rq)
>>> +int i915_active_fence_set(struct i915_active_fence *active,
>>> + struct i915_request *rq)
>>> {
>>> + struct dma_fence *fence;
>>> int err;
>>>
>>> #if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
>>
>> Here not in diff we actually have:
>>
>> #if IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)
>> lockdep_assert_held(active->lock);
>>
>> So why only under GEM debug and how does it related to the NB comment in
>> __i915_active_fence_set?
>
> We only expand the struct when we want to for debugging.
Missed that detail, ok.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* Re: [PATCH 07/27] drm/i915: Coordinate i915_active with its own mutex
2019-09-27 12:08 ` Tvrtko Ursulin
@ 2019-09-27 12:16 ` Chris Wilson
2019-09-27 12:25 ` Tvrtko Ursulin
0 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-27 12:16 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Tvrtko Ursulin (2019-09-27 13:08:51)
>
> On 27/09/2019 12:25, Chris Wilson wrote:
> > Quoting Tvrtko Ursulin (2019-09-27 12:10:29)
> >>
> >> On 25/09/2019 11:01, Chris Wilson wrote:
> >>> +struct dma_fence *
> >>> +__i915_active_fence_set(struct i915_active_fence *active,
> >>> + struct dma_fence *fence)
> >>
> >> Can you add a short comment for this function explaining the racyness
> >> and so why it returns prev and what should the callers do with it?
> >
> > Before using this function, we had the callers declare what mutex guards
> > this timeline and we check here that is held.
>
> No, I mean because it has to reload prev and return it to the caller,
> which implies that is to handle some designed-in racy-ness which I think
> should be mentioned.
But that's not racing with the caller, that just racing with the
callback from the interrupt handler and the reason why we have to check
after serialising is called out. /* serialise with prev->cb_list */ ?
The caller is responsible for ensuring that prev is executed before
fence to keep the timeline in the same order as recorded.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* Re: [PATCH 07/27] drm/i915: Coordinate i915_active with its own mutex
2019-09-27 12:16 ` Chris Wilson
@ 2019-09-27 12:25 ` Tvrtko Ursulin
2019-09-27 12:32 ` Chris Wilson
0 siblings, 1 reply; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-27 12:25 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 27/09/2019 13:16, Chris Wilson wrote:
> Quoting Tvrtko Ursulin (2019-09-27 13:08:51)
>>
>> On 27/09/2019 12:25, Chris Wilson wrote:
>>> Quoting Tvrtko Ursulin (2019-09-27 12:10:29)
>>>>
>>>> On 25/09/2019 11:01, Chris Wilson wrote:
>>>>> +struct dma_fence *
>>>>> +__i915_active_fence_set(struct i915_active_fence *active,
>>>>> + struct dma_fence *fence)
>>>>
>>>> Can you add a short comment for this function explaining the racyness
>>>> and so why it returns prev and what should the callers do with it?
>>>
>>> Before using this function, we had the callers declare what mutex guards
>>> this timeline and we check here that is held.
>>
>> No, I mean because it has to reload prev and return it to the caller,
>> which implies that is to handle some designed-in racy-ness which I think
>> should be mentioned.
>
> But that's not racing with the caller, that just racing with the
> callback from the interrupt handler and the reason why we have to check
> after serialising is called out. /* serialise with prev->cb_list */ ?
>
> The caller is responsible for ensuring that prev is executed before
> fence to keep the timeline in the same order as recorded.
I did not say it is racing with the caller, but that the caller has to
handle a race.
"Serialise with prev->cb_list" is too low level for me. Trust me, I
always struggle with the active tracking code since there is so many
indirections and relationships. So in the absence of a visual diagram a
higher level comment would be beneficial for the future self. Just
expanding on what you replied here talking about how interrupts
interacts with new stuff entering tracking would do it.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* Re: [PATCH 07/27] drm/i915: Coordinate i915_active with its own mutex
2019-09-27 12:25 ` Tvrtko Ursulin
@ 2019-09-27 12:32 ` Chris Wilson
2019-09-27 13:58 ` Tvrtko Ursulin
0 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-27 12:32 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Tvrtko Ursulin (2019-09-27 13:25:23)
>
> On 27/09/2019 13:16, Chris Wilson wrote:
> > Quoting Tvrtko Ursulin (2019-09-27 13:08:51)
> >>
> >> On 27/09/2019 12:25, Chris Wilson wrote:
> >>> Quoting Tvrtko Ursulin (2019-09-27 12:10:29)
> >>>>
> >>>> On 25/09/2019 11:01, Chris Wilson wrote:
> >>>>> +struct dma_fence *
> >>>>> +__i915_active_fence_set(struct i915_active_fence *active,
> >>>>> + struct dma_fence *fence)
> >>>>
> >>>> Can you add a short comment for this function explaining the racyness
> >>>> and so why it returns prev and what should the callers do with it?
> >>>
> >>> Before using this function, we had the callers declare what mutex guards
> >>> this timeline and we check here that is held.
> >>
> >> No, I mean because it has to reload prev and return it to the caller,
> >> which implies that is to handle some designed-in racy-ness which I think
> >> should be mentioned.
> >
> > But that's not racing with the caller, that just racing with the
> > callback from the interrupt handler and the reason why we have to check
> > after serialising is called out. /* serialise with prev->cb_list */ ?
> >
> > The caller is responsible for ensuring that prev is executed before
> > fence to keep the timeline in the same order as recorded.
>
> I did not say it is racing with the caller, but that the caller has to
> handle a race.
But the caller only has to care about "was there already an active fence
on this timeline? If so, I must make sure it executes before me and take
that into consideration for the flow along the timeline"
I'm not seeing what race the caller has to be concerned with here. If
they replace the last request in the timeline, they have it returned.
If there was no request previously in the timeline, they have NULL.
(That just seems straightforward.)
> "Serialise with prev->cb_list" is too low level for me. Trust me, I
> always struggle with the active tracking code since there is so many
> indirections and relationships. So in the absence of a visual diagram a
> higher level comment would be beneficial for the future self. Just
> expanding on what you replied here talking about how interrupts
> interacts with new stuff entering tracking would do it.
It's just about that the callback may be executing and so we need to
serialise the list manipulation under the lock; having performed that
list manipulation, we then know whether or not we were successful in
capturing the previous fence.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* Re: [PATCH 07/27] drm/i915: Coordinate i915_active with its own mutex
2019-09-27 12:32 ` Chris Wilson
@ 2019-09-27 13:58 ` Tvrtko Ursulin
2019-09-27 14:10 ` Chris Wilson
0 siblings, 1 reply; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-27 13:58 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 27/09/2019 13:32, Chris Wilson wrote:
> Quoting Tvrtko Ursulin (2019-09-27 13:25:23)
>>
>> On 27/09/2019 13:16, Chris Wilson wrote:
>>> Quoting Tvrtko Ursulin (2019-09-27 13:08:51)
>>>>
>>>> On 27/09/2019 12:25, Chris Wilson wrote:
>>>>> Quoting Tvrtko Ursulin (2019-09-27 12:10:29)
>>>>>>
>>>>>> On 25/09/2019 11:01, Chris Wilson wrote:
>>>>>>> +struct dma_fence *
>>>>>>> +__i915_active_fence_set(struct i915_active_fence *active,
>>>>>>> + struct dma_fence *fence)
>>>>>>
>>>>>> Can you add a short comment for this function explaining the racyness
>>>>>> and so why it returns prev and what should the callers do with it?
>>>>>
>>>>> Before using this function, we had the callers declare what mutex guards
>>>>> this timeline and we check here that is held.
>>>>
>>>> No, I mean because it has to reload prev and return it to the caller,
>>>> which implies that is to handle some designed-in racy-ness which I think
>>>> should be mentioned.
>>>
>>> But that's not racing with the caller, that just racing with the
>>> callback from the interrupt handler and the reason why we have to check
>>> after serialising is called out. /* serialise with prev->cb_list */ ?
>>>
>>> The caller is responsible for ensuring that prev is executed before
>>> fence to keep the timeline in the same order as recorded.
>>
>> I did not say it is racing with the caller, but that the caller has to
>> handle a race.
>
> But the caller only has to care about "was there already an active fence
> on this timeline? If so, I must make sure it executes before me and take
> that into consideration for the flow along the timeline"
>
> I'm not seeing what race the caller has to be concerned with here. If
> they replace the last request in the timeline, they have it returned.
> If there was no request previously in the timeline, they have NULL.
> (That just seems straightforward.)
>
>> "Serialise with prev->cb_list" is too low level for me. Trust me, I
>> always struggle with the active tracking code since there is so many
>> indirections and relationships. So in the absence of a visual diagram a
>> higher level comment would be beneficial for the future self. Just
>> expanding on what you replied here talking about how interrupts
>> interacts with new stuff entering tracking would do it.
>
> It's just about that the callback may be executing and so we need to
> serialise the list manipulation under the lock; having performed that
> list manipulation, we then know whether or not we were successful in
> capturing the previous fence.
Ok ok :), just expand the comment next to re-fetch of prev =
active->fence to that effect, that's all I'm asking. :)
It will make it easier for future reader to understand why it is
required and under what circumstances could active->fence have became
NULL in there.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* Re: [PATCH 07/27] drm/i915: Coordinate i915_active with its own mutex
2019-09-27 13:58 ` Tvrtko Ursulin
@ 2019-09-27 14:10 ` Chris Wilson
0 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-27 14:10 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Tvrtko Ursulin (2019-09-27 14:58:24)
>
> On 27/09/2019 13:32, Chris Wilson wrote:
> > Quoting Tvrtko Ursulin (2019-09-27 13:25:23)
> >>
> >> On 27/09/2019 13:16, Chris Wilson wrote:
> >>> Quoting Tvrtko Ursulin (2019-09-27 13:08:51)
> >>>>
> >>>> On 27/09/2019 12:25, Chris Wilson wrote:
> >>>>> Quoting Tvrtko Ursulin (2019-09-27 12:10:29)
> >>>>>>
> >>>>>> On 25/09/2019 11:01, Chris Wilson wrote:
> >>>>>>> +struct dma_fence *
> >>>>>>> +__i915_active_fence_set(struct i915_active_fence *active,
> >>>>>>> + struct dma_fence *fence)
> >>>>>>
> >>>>>> Can you add a short comment for this function explaining the racyness
> >>>>>> and so why it returns prev and what should the callers do with it?
> >>>>>
> >>>>> Before using this function, we had the callers declare what mutex guards
> >>>>> this timeline and we check here that is held.
> >>>>
> >>>> No, I mean because it has to reload prev and return it to the caller,
> >>>> which implies that is to handle some designed-in racy-ness which I think
> >>>> should be mentioned.
> >>>
> >>> But that's not racing with the caller, that just racing with the
> >>> callback from the interrupt handler and the reason why we have to check
> >>> after serialising is called out. /* serialise with prev->cb_list */ ?
> >>>
> >>> The caller is responsible for ensuring that prev is executed before
> >>> fence to keep the timeline in the same order as recorded.
> >>
> >> I did not say it is racing with the caller, but that the caller has to
> >> handle a race.
> >
> > But the caller only has to care about "was there already an active fence
> > on this timeline? If so, I must make sure it executes before me and take
> > that into consideration for the flow along the timeline"
> >
> > I'm not seeing what race the caller has to be concerned with here. If
> > they replace the last request in the timeline, they have it returned.
> > If there was no request previously in the timeline, they have NULL.
> > (That just seems straightforward.)
> >
> >> "Serialise with prev->cb_list" is too low level for me. Trust me, I
> >> always struggle with the active tracking code since there is so many
> >> indirections and relationships. So in the absence of a visual diagram a
> >> higher level comment would be beneficial for the future self. Just
> >> expanding on what you replied here talking about how interrupts
> >> interacts with new stuff entering tracking would do it.
> >
> > It's just about that the callback may be executing and so we need to
> > serialise the list manipulation under the lock; having performed that
> > list manipulation, we then know whether or not we were successful in
> > capturing the previous fence.
>
> Ok ok :), just expand the comment next to re-fetch of prev =
> active->fence to that effect, that's all I'm asking. :)
>
> It will make it easier for future reader to understand why it is
> required and under what circumstances could active->fence have became
> NULL in there.
spin_lock_nested(prev->lock, SINGLE_DEPTH_NESTING);
__list_del_entry(&active->cb.node);
spin_unlock(prev->lock); /* serialise with prev->cb_list */
+
+ /*
+ * active->fence is reset by the callback from inside
+ * interrupt context. We need to serialise our list
+ * manipulation with the fence->lock to prevent the prev
+ * being lost inside an interrupt (it can't be replaced as
+ * no other caller is allowed to enter __i915_active_fence_set
+ * as we hold the timeline lock). After serialising with
+ * the callback, we need to double check which ran first,
+ * our list_del() or the callback...
+ */
prev = rcu_access_pointer(active->fence);
}
Feels a little too tautological, but it is Friday afternoon.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* [PATCH 08/27] drm/i915: Move idle barrier cleanup into engine-pm
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (6 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 07/27] drm/i915: Coordinate i915_active with its own mutex Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:01 ` [PATCH 09/27] drm/i915: Drop struct_mutex from around i915_retire_requests() Chris Wilson
` (22 subsequent siblings)
30 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
Now that we now longer need to guarantee that the active callback is
under the struct_mutex, we can lift it out of the i915_gem_park() and
into the engine parking itself.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/gem/i915_gem_pm.c | 19 -------------------
drivers/gpu/drm/i915/gt/intel_engine_pm.c | 15 +++++++++++++++
drivers/gpu/drm/i915/i915_active.c | 1 +
3 files changed, 16 insertions(+), 19 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
index 3a47e1190ce1..42a6021d2df8 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
@@ -11,29 +11,10 @@
#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 dma_fence_cb *cb =
- container_of((struct list_head *)node,
- typeof(*cb), node);
-
- cb->func(NULL, cb);
- }
-}
-
static void i915_gem_park(struct drm_i915_private *i915)
{
- struct intel_engine_cs *engine;
- enum intel_engine_id id;
-
lockdep_assert_held(&i915->drm.struct_mutex);
- for_each_engine(engine, i915, id)
- call_idle_barriers(engine); /* cleanup after wedging */
-
i915_vma_parked(i915);
i915_globals_park();
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pm.c b/drivers/gpu/drm/i915/gt/intel_engine_pm.c
index ce61c83d68e9..360fd5cb07ab 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_pm.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_pm.c
@@ -123,6 +123,19 @@ static bool switch_to_kernel_context(struct intel_engine_cs *engine)
return result;
}
+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 dma_fence_cb *cb =
+ container_of((struct list_head *)node,
+ typeof(*cb), node);
+
+ cb->func(NULL, cb);
+ }
+}
+
static int __engine_park(struct intel_wakeref *wf)
{
struct intel_engine_cs *engine =
@@ -142,6 +155,8 @@ static int __engine_park(struct intel_wakeref *wf)
GEM_TRACE("%s\n", engine->name);
+ call_idle_barriers(engine); /* cleanup after wedging */
+
intel_engine_disarm_breadcrumbs(engine);
intel_engine_pool_park(&engine->pool);
diff --git a/drivers/gpu/drm/i915/i915_active.c b/drivers/gpu/drm/i915/i915_active.c
index f23590ec9e68..b5a817888a25 100644
--- a/drivers/gpu/drm/i915/i915_active.c
+++ b/drivers/gpu/drm/i915/i915_active.c
@@ -688,6 +688,7 @@ void i915_active_acquire_barrier(struct i915_active *ref)
rb_link_node(&node->node, parent, p);
rb_insert_color(&node->node, &ref->tree);
+ GEM_BUG_ON(!intel_engine_pm_is_awake(engine));
llist_add(barrier_to_ll(node), &engine->barrier_tasks);
intel_engine_pm_put(engine);
}
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* [PATCH 09/27] drm/i915: Drop struct_mutex from around i915_retire_requests()
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (7 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 08/27] drm/i915: Move idle barrier cleanup into engine-pm Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:01 ` [PATCH 10/27] drm/i915: Remove the GEM idle worker Chris Wilson
` (21 subsequent siblings)
30 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
We don't need to hold struct_mutex now for retiring requests, so drop it
from i915_retire_requests() and i915_gem_wait_for_idle(), finally
removing I915_WAIT_LOCKED for good.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
.../gpu/drm/i915/gem/i915_gem_client_blt.c | 7 +-
drivers/gpu/drm/i915/gem/i915_gem_context.c | 20 +--
drivers/gpu/drm/i915/gem/i915_gem_pm.c | 45 +++---
.../i915/gem/selftests/i915_gem_coherency.c | 40 +++---
.../drm/i915/gem/selftests/i915_gem_context.c | 4 +-
.../drm/i915/gem/selftests/i915_gem_mman.c | 6 +-
.../i915/gem/selftests/i915_gem_object_blt.c | 4 -
drivers/gpu/drm/i915/gt/selftest_context.c | 4 +-
drivers/gpu/drm/i915/gt/selftest_hangcheck.c | 89 ++----------
drivers/gpu/drm/i915/gt/selftest_lrc.c | 23 ++-
drivers/gpu/drm/i915/gt/selftest_timeline.c | 91 +++++-------
.../gpu/drm/i915/gt/selftest_workarounds.c | 6 +-
drivers/gpu/drm/i915/i915_debugfs.c | 42 ++----
drivers/gpu/drm/i915/i915_gem.c | 19 ++-
drivers/gpu/drm/i915/i915_request.h | 7 +-
drivers/gpu/drm/i915/selftests/i915_active.c | 8 +-
.../gpu/drm/i915/selftests/i915_gem_evict.c | 2 +-
drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 6 +-
drivers/gpu/drm/i915/selftests/i915_request.c | 134 ++++++------------
.../gpu/drm/i915/selftests/i915_selftest.c | 8 +-
drivers/gpu/drm/i915/selftests/i915_vma.c | 4 -
.../gpu/drm/i915/selftests/igt_flush_test.c | 30 ++--
.../gpu/drm/i915/selftests/igt_flush_test.h | 2 +-
.../gpu/drm/i915/selftests/igt_live_test.c | 9 +-
.../gpu/drm/i915/selftests/mock_gem_device.c | 4 -
25 files changed, 201 insertions(+), 413 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c b/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c
index c1fca5728e6e..81366aa4812b 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_client_blt.c
@@ -155,7 +155,6 @@ static void clear_pages_dma_fence_cb(struct dma_fence *fence,
static void clear_pages_worker(struct work_struct *work)
{
struct clear_pages_work *w = container_of(work, typeof(*w), work);
- struct drm_i915_private *i915 = w->ce->engine->i915;
struct drm_i915_gem_object *obj = w->sleeve->vma->obj;
struct i915_vma *vma = w->sleeve->vma;
struct i915_request *rq;
@@ -173,11 +172,9 @@ static void clear_pages_worker(struct work_struct *work)
obj->read_domains = I915_GEM_GPU_DOMAINS;
obj->write_domain = 0;
- /* XXX: we need to kill this */
- mutex_lock(&i915->drm.struct_mutex);
err = i915_vma_pin(vma, 0, 0, PIN_USER);
if (unlikely(err))
- goto out_unlock;
+ goto out_signal;
batch = intel_emit_vma_fill_blt(w->ce, vma, w->value);
if (IS_ERR(batch)) {
@@ -229,8 +226,6 @@ static void clear_pages_worker(struct work_struct *work)
intel_emit_vma_release(w->ce, batch);
out_unpin:
i915_vma_unpin(vma);
-out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
out_signal:
if (unlikely(err)) {
dma_fence_set_error(&w->dma, err);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index 9d85aab68d34..0ab416887fc2 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -1159,8 +1159,7 @@ gen8_modify_rpcs(struct intel_context *ce, struct intel_sseu sseu)
}
static int
-__intel_context_reconfigure_sseu(struct intel_context *ce,
- struct intel_sseu sseu)
+intel_context_reconfigure_sseu(struct intel_context *ce, struct intel_sseu sseu)
{
int ret;
@@ -1183,23 +1182,6 @@ __intel_context_reconfigure_sseu(struct intel_context *ce,
return ret;
}
-static int
-intel_context_reconfigure_sseu(struct intel_context *ce, struct intel_sseu sseu)
-{
- struct drm_i915_private *i915 = ce->engine->i915;
- int ret;
-
- ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
- if (ret)
- return ret;
-
- ret = __intel_context_reconfigure_sseu(ce, sseu);
-
- mutex_unlock(&i915->drm.struct_mutex);
-
- return ret;
-}
-
static int
user_to_context_sseu(struct drm_i915_private *i915,
const struct drm_i915_gem_context_param_sseu *user,
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
index 42a6021d2df8..535f2f8a23aa 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
@@ -48,11 +48,7 @@ static void retire_work_handler(struct work_struct *work)
struct drm_i915_private *i915 =
container_of(work, typeof(*i915), gem.retire_work.work);
- /* Come back later if the device is busy... */
- if (mutex_trylock(&i915->drm.struct_mutex)) {
- i915_retire_requests(i915);
- mutex_unlock(&i915->drm.struct_mutex);
- }
+ i915_retire_requests(i915);
queue_delayed_work(i915->wq,
&i915->gem.retire_work,
@@ -86,26 +82,23 @@ static bool switch_to_kernel_context_sync(struct intel_gt *gt)
{
bool result = !intel_gt_is_wedged(gt);
- do {
- if (i915_gem_wait_for_idle(gt->i915,
- I915_WAIT_LOCKED |
- I915_WAIT_FOR_IDLE_BOOST,
- I915_GEM_IDLE_TIMEOUT) == -ETIME) {
- /* XXX hide warning from gem_eio */
- if (i915_modparams.reset) {
- dev_err(gt->i915->drm.dev,
- "Failed to idle engines, declaring wedged!\n");
- GEM_TRACE_DUMP();
- }
-
- /*
- * Forcibly cancel outstanding work and leave
- * the gpu quiet.
- */
- intel_gt_set_wedged(gt);
- result = false;
+ if (i915_gem_wait_for_idle(gt->i915,
+ I915_WAIT_FOR_IDLE_BOOST,
+ I915_GEM_IDLE_TIMEOUT) == -ETIME) {
+ /* XXX hide warning from gem_eio */
+ if (i915_modparams.reset) {
+ dev_err(gt->i915->drm.dev,
+ "Failed to idle engines, declaring wedged!\n");
+ GEM_TRACE_DUMP();
}
- } while (i915_retire_requests(gt->i915) && result);
+
+ /*
+ * Forcibly cancel outstanding work and leave
+ * the gpu quiet.
+ */
+ intel_gt_set_wedged(gt);
+ result = false;
+ }
if (intel_gt_pm_wait_for_idle(gt))
result = false;
@@ -146,8 +139,6 @@ void i915_gem_suspend(struct drm_i915_private *i915)
user_forcewake(&i915->gt, true);
- mutex_lock(&i915->drm.struct_mutex);
-
/*
* We have to flush all the executing contexts to main memory so
* that they can saved in the hibernation image. To ensure the last
@@ -159,8 +150,6 @@ void i915_gem_suspend(struct drm_i915_private *i915)
*/
switch_to_kernel_context_sync(&i915->gt);
- mutex_unlock(&i915->drm.struct_mutex);
-
cancel_delayed_work_sync(&i915->gt.hangcheck.work);
i915_gem_drain_freed_objects(i915);
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 0ff7a89aadca..549810f70aeb 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
@@ -7,6 +7,7 @@
#include <linux/prime_numbers.h>
#include "gt/intel_gt.h"
+#include "gt/intel_gt_pm.h"
#include "i915_selftest.h"
#include "selftests/i915_random.h"
@@ -78,7 +79,7 @@ static int gtt_set(struct drm_i915_gem_object *obj,
{
struct i915_vma *vma;
u32 __iomem *map;
- int err;
+ int err = 0;
i915_gem_object_lock(obj);
err = i915_gem_object_set_to_gtt_domain(obj, true);
@@ -90,15 +91,21 @@ static int gtt_set(struct drm_i915_gem_object *obj,
if (IS_ERR(vma))
return PTR_ERR(vma);
+ intel_gt_pm_get(vma->vm->gt);
+
map = i915_vma_pin_iomap(vma);
i915_vma_unpin(vma);
- if (IS_ERR(map))
- return PTR_ERR(map);
+ if (IS_ERR(map)) {
+ err = PTR_ERR(map);
+ goto out_rpm;
+ }
iowrite32(v, &map[offset / sizeof(*map)]);
i915_vma_unpin_iomap(vma);
- return 0;
+out_rpm:
+ intel_gt_pm_put(vma->vm->gt);
+ return err;
}
static int gtt_get(struct drm_i915_gem_object *obj,
@@ -107,7 +114,7 @@ static int gtt_get(struct drm_i915_gem_object *obj,
{
struct i915_vma *vma;
u32 __iomem *map;
- int err;
+ int err = 0;
i915_gem_object_lock(obj);
err = i915_gem_object_set_to_gtt_domain(obj, false);
@@ -119,15 +126,21 @@ static int gtt_get(struct drm_i915_gem_object *obj,
if (IS_ERR(vma))
return PTR_ERR(vma);
+ intel_gt_pm_get(vma->vm->gt);
+
map = i915_vma_pin_iomap(vma);
i915_vma_unpin(vma);
- if (IS_ERR(map))
- return PTR_ERR(map);
+ if (IS_ERR(map)) {
+ err = PTR_ERR(map);
+ goto out_rpm;
+ }
*v = ioread32(&map[offset / sizeof(*map)]);
i915_vma_unpin_iomap(vma);
- return 0;
+out_rpm:
+ intel_gt_pm_put(vma->vm->gt);
+ return err;
}
static int wc_set(struct drm_i915_gem_object *obj,
@@ -280,7 +293,6 @@ static int igt_gem_coherency(void *arg)
struct drm_i915_private *i915 = arg;
const struct igt_coherency_mode *read, *write, *over;
struct drm_i915_gem_object *obj;
- intel_wakeref_t wakeref;
unsigned long count, n;
u32 *offsets, *values;
int err = 0;
@@ -299,8 +311,6 @@ static int igt_gem_coherency(void *arg)
values = offsets + ncachelines;
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
for (over = igt_coherency_mode; over->name; over++) {
if (!over->set)
continue;
@@ -326,7 +336,7 @@ static int igt_gem_coherency(void *arg)
obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
if (IS_ERR(obj)) {
err = PTR_ERR(obj);
- goto unlock;
+ goto free;
}
i915_random_reorder(offsets, ncachelines, &prng);
@@ -377,15 +387,13 @@ static int igt_gem_coherency(void *arg)
}
}
}
-unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
+free:
kfree(offsets);
return err;
put_object:
i915_gem_object_put(obj);
- goto unlock;
+ goto free;
}
int i915_gem_coherency_live_selftests(struct drm_i915_private *i915)
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 6ef3d745dc8d..82f54e819eb0 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
@@ -949,7 +949,7 @@ __sseu_test(const char *name,
if (ret)
return ret;
- ret = __intel_context_reconfigure_sseu(ce, sseu);
+ ret = intel_context_reconfigure_sseu(ce, sseu);
if (ret)
goto out_spin;
@@ -1053,7 +1053,7 @@ __igt_ctx_sseu(struct drm_i915_private *i915,
goto out_fail;
out_fail:
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
ret = -EIO;
intel_context_unpin(ce);
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 36aca1c172e7..856b8e467ee8 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
@@ -581,12 +581,8 @@ static void disable_retire_worker(struct drm_i915_private *i915)
static void restore_retire_worker(struct drm_i915_private *i915)
{
+ igt_flush_test(i915);
intel_gt_pm_put(&i915->gt);
-
- mutex_lock(&i915->drm.struct_mutex);
- igt_flush_test(i915, I915_WAIT_LOCKED);
- mutex_unlock(&i915->drm.struct_mutex);
-
i915_gem_driver_register__shrinker(i915);
}
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_object_blt.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_object_blt.c
index c21d747e7d05..9ec55b3a3815 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_object_blt.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_object_blt.c
@@ -65,9 +65,7 @@ static int igt_fill_blt(void *arg)
if (!(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE))
obj->cache_dirty = true;
- mutex_lock(&i915->drm.struct_mutex);
err = i915_gem_object_fill_blt(obj, ce, val);
- mutex_unlock(&i915->drm.struct_mutex);
if (err)
goto err_unpin;
@@ -166,9 +164,7 @@ static int igt_copy_blt(void *arg)
if (!(dst->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE))
dst->cache_dirty = true;
- mutex_lock(&i915->drm.struct_mutex);
err = i915_gem_object_copy_blt(src, dst, ce);
- mutex_unlock(&i915->drm.struct_mutex);
if (err)
goto err_unpin;
diff --git a/drivers/gpu/drm/i915/gt/selftest_context.c b/drivers/gpu/drm/i915/gt/selftest_context.c
index e6bcbe7ab5e1..86cffbb0a9cb 100644
--- a/drivers/gpu/drm/i915/gt/selftest_context.c
+++ b/drivers/gpu/drm/i915/gt/selftest_context.c
@@ -318,7 +318,7 @@ static int live_active_context(void *arg)
if (err)
break;
- err = igt_flush_test(gt->i915, I915_WAIT_LOCKED);
+ err = igt_flush_test(gt->i915);
if (err)
break;
}
@@ -431,7 +431,7 @@ static int live_remote_context(void *arg)
if (err)
break;
- err = igt_flush_test(gt->i915, I915_WAIT_LOCKED);
+ err = igt_flush_test(gt->i915);
if (err)
break;
}
diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
index e53eea1050f8..fc4d02406536 100644
--- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
+++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
@@ -58,7 +58,9 @@ static int hang_init(struct hang *h, struct intel_gt *gt)
memset(h, 0, sizeof(*h));
h->gt = gt;
+ mutex_lock(>->i915->drm.struct_mutex);
h->ctx = kernel_context(gt->i915);
+ mutex_unlock(>->i915->drm.struct_mutex);
if (IS_ERR(h->ctx))
return PTR_ERR(h->ctx);
@@ -285,7 +287,7 @@ static void hang_fini(struct hang *h)
kernel_context_close(h->ctx);
- igt_flush_test(h->gt->i915, I915_WAIT_LOCKED);
+ igt_flush_test(h->gt->i915);
}
static bool wait_until_running(struct hang *h, struct i915_request *rq)
@@ -309,10 +311,9 @@ static int igt_hang_sanitycheck(void *arg)
/* Basic check that we can execute our hanging batch */
- mutex_lock(>->i915->drm.struct_mutex);
err = hang_init(&h, gt);
if (err)
- goto unlock;
+ return err;
for_each_engine(engine, gt->i915, id) {
struct intel_wedge_me w;
@@ -355,8 +356,6 @@ static int igt_hang_sanitycheck(void *arg)
fini:
hang_fini(&h);
-unlock:
- mutex_unlock(>->i915->drm.struct_mutex);
return err;
}
@@ -395,8 +394,6 @@ static int igt_reset_nop(void *arg)
reset_count = i915_reset_count(global);
count = 0;
do {
- mutex_lock(>->i915->drm.struct_mutex);
-
for_each_engine(engine, gt->i915, id) {
int i;
@@ -417,7 +414,6 @@ static int igt_reset_nop(void *arg)
intel_gt_reset(gt, ALL_ENGINES, NULL);
igt_global_reset_unlock(gt);
- mutex_unlock(>->i915->drm.struct_mutex);
if (intel_gt_is_wedged(gt)) {
err = -EIO;
break;
@@ -429,16 +425,13 @@ static int igt_reset_nop(void *arg)
break;
}
- err = igt_flush_test(gt->i915, 0);
+ err = igt_flush_test(gt->i915);
if (err)
break;
} while (time_before(jiffies, end_time));
pr_info("%s: %d resets\n", __func__, count);
- mutex_lock(>->i915->drm.struct_mutex);
- err = igt_flush_test(gt->i915, I915_WAIT_LOCKED);
- mutex_unlock(>->i915->drm.struct_mutex);
-
+ err = igt_flush_test(gt->i915);
out:
mock_file_free(gt->i915, file);
if (intel_gt_is_wedged(gt))
@@ -494,7 +487,6 @@ static int igt_reset_nop_engine(void *arg)
break;
}
- mutex_lock(>->i915->drm.struct_mutex);
for (i = 0; i < 16; i++) {
struct i915_request *rq;
@@ -507,7 +499,6 @@ static int igt_reset_nop_engine(void *arg)
i915_request_add(rq);
}
err = intel_engine_reset(engine, NULL);
- mutex_unlock(>->i915->drm.struct_mutex);
if (err) {
pr_err("i915_reset_engine failed\n");
break;
@@ -533,15 +524,12 @@ static int igt_reset_nop_engine(void *arg)
if (err)
break;
- err = igt_flush_test(gt->i915, 0);
+ err = igt_flush_test(gt->i915);
if (err)
break;
}
- mutex_lock(>->i915->drm.struct_mutex);
- err = igt_flush_test(gt->i915, I915_WAIT_LOCKED);
- mutex_unlock(>->i915->drm.struct_mutex);
-
+ err = igt_flush_test(gt->i915);
out:
mock_file_free(gt->i915, file);
if (intel_gt_is_wedged(gt))
@@ -563,9 +551,7 @@ static int __igt_reset_engine(struct intel_gt *gt, bool active)
return 0;
if (active) {
- mutex_lock(>->i915->drm.struct_mutex);
err = hang_init(&h, gt);
- mutex_unlock(>->i915->drm.struct_mutex);
if (err)
return err;
}
@@ -593,17 +579,14 @@ static int __igt_reset_engine(struct intel_gt *gt, bool active)
if (active) {
struct i915_request *rq;
- mutex_lock(>->i915->drm.struct_mutex);
rq = hang_create_request(&h, engine);
if (IS_ERR(rq)) {
err = PTR_ERR(rq);
- mutex_unlock(>->i915->drm.struct_mutex);
break;
}
i915_request_get(rq);
i915_request_add(rq);
- mutex_unlock(>->i915->drm.struct_mutex);
if (!wait_until_running(&h, rq)) {
struct drm_printer p = drm_info_printer(gt->i915->drm.dev);
@@ -647,7 +630,7 @@ static int __igt_reset_engine(struct intel_gt *gt, bool active)
if (err)
break;
- err = igt_flush_test(gt->i915, 0);
+ err = igt_flush_test(gt->i915);
if (err)
break;
}
@@ -655,11 +638,8 @@ static int __igt_reset_engine(struct intel_gt *gt, bool active)
if (intel_gt_is_wedged(gt))
err = -EIO;
- if (active) {
- mutex_lock(>->i915->drm.struct_mutex);
+ if (active)
hang_fini(&h);
- mutex_unlock(>->i915->drm.struct_mutex);
- }
return err;
}
@@ -741,10 +721,8 @@ static int active_engine(void *data)
struct i915_request *old = rq[idx];
struct i915_request *new;
- mutex_lock(&engine->i915->drm.struct_mutex);
new = igt_request_alloc(ctx[idx], engine);
if (IS_ERR(new)) {
- mutex_unlock(&engine->i915->drm.struct_mutex);
err = PTR_ERR(new);
break;
}
@@ -755,7 +733,6 @@ static int active_engine(void *data)
rq[idx] = i915_request_get(new);
i915_request_add(new);
- mutex_unlock(&engine->i915->drm.struct_mutex);
err = active_request_put(old);
if (err)
@@ -795,9 +772,7 @@ static int __igt_reset_engines(struct intel_gt *gt,
return 0;
if (flags & TEST_ACTIVE) {
- mutex_lock(>->i915->drm.struct_mutex);
err = hang_init(&h, gt);
- mutex_unlock(>->i915->drm.struct_mutex);
if (err)
return err;
@@ -855,17 +830,14 @@ static int __igt_reset_engines(struct intel_gt *gt,
struct i915_request *rq = NULL;
if (flags & TEST_ACTIVE) {
- mutex_lock(>->i915->drm.struct_mutex);
rq = hang_create_request(&h, engine);
if (IS_ERR(rq)) {
err = PTR_ERR(rq);
- mutex_unlock(>->i915->drm.struct_mutex);
break;
}
i915_request_get(rq);
i915_request_add(rq);
- mutex_unlock(>->i915->drm.struct_mutex);
if (!wait_until_running(&h, rq)) {
struct drm_printer p = drm_info_printer(gt->i915->drm.dev);
@@ -977,9 +949,7 @@ static int __igt_reset_engines(struct intel_gt *gt,
if (err)
break;
- mutex_lock(>->i915->drm.struct_mutex);
- err = igt_flush_test(gt->i915, I915_WAIT_LOCKED);
- mutex_unlock(>->i915->drm.struct_mutex);
+ err = igt_flush_test(gt->i915);
if (err)
break;
}
@@ -987,11 +957,8 @@ static int __igt_reset_engines(struct intel_gt *gt,
if (intel_gt_is_wedged(gt))
err = -EIO;
- if (flags & TEST_ACTIVE) {
- mutex_lock(>->i915->drm.struct_mutex);
+ if (flags & TEST_ACTIVE)
hang_fini(&h);
- mutex_unlock(>->i915->drm.struct_mutex);
- }
return err;
}
@@ -1061,7 +1028,6 @@ static int igt_reset_wait(void *arg)
igt_global_reset_lock(gt);
- mutex_lock(>->i915->drm.struct_mutex);
err = hang_init(&h, gt);
if (err)
goto unlock;
@@ -1109,7 +1075,6 @@ static int igt_reset_wait(void *arg)
fini:
hang_fini(&h);
unlock:
- mutex_unlock(>->i915->drm.struct_mutex);
igt_global_reset_unlock(gt);
if (intel_gt_is_wedged(gt))
@@ -1189,10 +1154,9 @@ static int __igt_reset_evict_vma(struct intel_gt *gt,
/* Check that we can recover an unbind stuck on a hanging request */
- mutex_lock(>->i915->drm.struct_mutex);
err = hang_init(&h, gt);
if (err)
- goto unlock;
+ return err;
obj = i915_gem_object_create_internal(gt->i915, SZ_1M);
if (IS_ERR(obj)) {
@@ -1255,8 +1219,6 @@ static int __igt_reset_evict_vma(struct intel_gt *gt,
if (err)
goto out_rq;
- mutex_unlock(>->i915->drm.struct_mutex);
-
if (!wait_until_running(&h, rq)) {
struct drm_printer p = drm_info_printer(gt->i915->drm.dev);
@@ -1305,16 +1267,12 @@ static int __igt_reset_evict_vma(struct intel_gt *gt,
put_task_struct(tsk);
}
- mutex_lock(>->i915->drm.struct_mutex);
out_rq:
i915_request_put(rq);
out_obj:
i915_gem_object_put(obj);
fini:
hang_fini(&h);
-unlock:
- mutex_unlock(>->i915->drm.struct_mutex);
-
if (intel_gt_is_wedged(gt))
return -EIO;
@@ -1396,7 +1354,6 @@ static int igt_reset_queue(void *arg)
igt_global_reset_lock(gt);
- mutex_lock(>->i915->drm.struct_mutex);
err = hang_init(&h, gt);
if (err)
goto unlock;
@@ -1511,7 +1468,7 @@ static int igt_reset_queue(void *arg)
i915_request_put(prev);
- err = igt_flush_test(gt->i915, I915_WAIT_LOCKED);
+ err = igt_flush_test(gt->i915);
if (err)
break;
}
@@ -1519,7 +1476,6 @@ static int igt_reset_queue(void *arg)
fini:
hang_fini(&h);
unlock:
- mutex_unlock(>->i915->drm.struct_mutex);
igt_global_reset_unlock(gt);
if (intel_gt_is_wedged(gt))
@@ -1546,11 +1502,9 @@ static int igt_handle_error(void *arg)
if (!engine || !intel_engine_can_store_dword(engine))
return 0;
- mutex_lock(>->i915->drm.struct_mutex);
-
err = hang_init(&h, gt);
if (err)
- goto err_unlock;
+ return err;
rq = hang_create_request(&h, engine);
if (IS_ERR(rq)) {
@@ -1574,8 +1528,6 @@ static int igt_handle_error(void *arg)
goto err_request;
}
- mutex_unlock(>->i915->drm.struct_mutex);
-
/* Temporarily disable error capture */
error = xchg(&global->first_error, (void *)-1);
@@ -1583,8 +1535,6 @@ static int igt_handle_error(void *arg)
xchg(&global->first_error, error);
- mutex_lock(>->i915->drm.struct_mutex);
-
if (rq->fence.error != -EIO) {
pr_err("Guilty request not identified!\n");
err = -EINVAL;
@@ -1595,8 +1545,6 @@ static int igt_handle_error(void *arg)
i915_request_put(rq);
err_fini:
hang_fini(&h);
-err_unlock:
- mutex_unlock(>->i915->drm.struct_mutex);
return err;
}
@@ -1689,7 +1637,6 @@ static int igt_reset_engines_atomic(void *arg)
return 0;
igt_global_reset_lock(gt);
- mutex_lock(>->i915->drm.struct_mutex);
/* Flush any requests before we get started and check basics */
if (!igt_force_reset(gt))
@@ -1709,9 +1656,7 @@ static int igt_reset_engines_atomic(void *arg)
out:
/* As we poke around the guts, do a full reset before continuing. */
igt_force_reset(gt);
-
unlock:
- mutex_unlock(>->i915->drm.struct_mutex);
igt_global_reset_unlock(gt);
return err;
@@ -1751,10 +1696,6 @@ int intel_hangcheck_live_selftests(struct drm_i915_private *i915)
err = intel_gt_live_subtests(tests, gt);
- mutex_lock(>->i915->drm.struct_mutex);
- igt_flush_test(gt->i915, I915_WAIT_LOCKED);
- mutex_unlock(>->i915->drm.struct_mutex);
-
i915_modparams.enable_hangcheck = saved_hangcheck;
intel_runtime_pm_put(>->i915->runtime_pm, wakeref);
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index 8eec4ead12a4..3300acd9199d 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -61,7 +61,7 @@ static int live_sanitycheck(void *arg)
}
igt_spinner_end(&spin);
- if (igt_flush_test(i915, I915_WAIT_LOCKED)) {
+ if (igt_flush_test(i915)) {
err = -EIO;
goto err_ctx;
}
@@ -206,8 +206,7 @@ slice_semaphore_queue(struct intel_engine_cs *outer,
if (err)
goto out;
- if (i915_request_wait(head,
- I915_WAIT_LOCKED,
+ if (i915_request_wait(head, 0,
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);
@@ -279,7 +278,7 @@ static int live_timeslice_preempt(void *arg)
if (err)
goto err_pin;
- if (igt_flush_test(i915, I915_WAIT_LOCKED)) {
+ if (igt_flush_test(i915)) {
err = -EIO;
goto err_pin;
}
@@ -832,7 +831,7 @@ static int live_nopreempt(void *arg)
goto err_wedged;
}
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
goto err_wedged;
}
@@ -897,7 +896,7 @@ static int live_suppress_self_preempt(void *arg)
if (!intel_engine_has_preemption(engine))
continue;
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
goto err_wedged;
intel_engine_pm_get(engine);
@@ -958,7 +957,7 @@ static int live_suppress_self_preempt(void *arg)
}
intel_engine_pm_put(engine);
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
goto err_wedged;
}
@@ -1119,7 +1118,7 @@ static int live_suppress_wait_preempt(void *arg)
for (i = 0; i < ARRAY_SIZE(client); i++)
igt_spinner_end(&client[i].spin);
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
goto err_wedged;
if (engine->execlists.preempt_hang.count) {
@@ -1398,7 +1397,7 @@ static int live_preempt_hang(void *arg)
igt_spinner_end(&spin_hi);
igt_spinner_end(&spin_lo);
- if (igt_flush_test(i915, I915_WAIT_LOCKED)) {
+ if (igt_flush_test(i915)) {
err = -EIO;
goto err_ctx_lo;
}
@@ -1795,7 +1794,7 @@ static int nop_virtual_engine(struct drm_i915_private *i915,
prime, div64_u64(ktime_to_ns(times[1]), prime));
out:
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
err = -EIO;
for (nc = 0; nc < nctx; nc++) {
@@ -1940,7 +1939,7 @@ static int mask_virtual_engine(struct drm_i915_private *i915,
goto out;
out:
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
err = -EIO;
for (n = 0; n < nsibling; n++)
@@ -2118,7 +2117,7 @@ static int bond_virtual_engine(struct drm_i915_private *i915,
out:
for (n = 0; !IS_ERR(rq[n]); n++)
i915_request_put(rq[n]);
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
err = -EIO;
kernel_context_close(ctx);
diff --git a/drivers/gpu/drm/i915/gt/selftest_timeline.c b/drivers/gpu/drm/i915/gt/selftest_timeline.c
index 321481403165..16abfabf08c7 100644
--- a/drivers/gpu/drm/i915/gt/selftest_timeline.c
+++ b/drivers/gpu/drm/i915/gt/selftest_timeline.c
@@ -6,7 +6,7 @@
#include <linux/prime_numbers.h>
-#include "gem/i915_gem_pm.h"
+#include "intel_engine_pm.h"
#include "intel_gt.h"
#include "../selftests/i915_random.h"
@@ -136,7 +136,6 @@ static int mock_hwsp_freelist(void *arg)
goto err_put;
}
- mutex_lock(&state.i915->drm.struct_mutex);
for (p = phases; p->name; p++) {
pr_debug("%s(%s)\n", __func__, p->name);
for_each_prime_number_from(na, 1, 2 * CACHELINES_PER_PAGE) {
@@ -149,7 +148,6 @@ static int mock_hwsp_freelist(void *arg)
out:
for (na = 0; na < state.max; na++)
__mock_hwsp_record(&state, na, NULL);
- mutex_unlock(&state.i915->drm.struct_mutex);
kfree(state.history);
err_put:
drm_dev_put(&state.i915->drm);
@@ -449,8 +447,6 @@ tl_write(struct intel_timeline *tl, struct intel_engine_cs *engine, u32 value)
struct i915_request *rq;
int err;
- lockdep_assert_held(&tl->gt->i915->drm.struct_mutex); /* lazy rq refs */
-
err = intel_timeline_pin(tl);
if (err) {
rq = ERR_PTR(err);
@@ -461,10 +457,14 @@ tl_write(struct intel_timeline *tl, struct intel_engine_cs *engine, u32 value)
if (IS_ERR(rq))
goto out_unpin;
+ i915_request_get(rq);
+
err = emit_ggtt_store_dw(rq, tl->hwsp_offset, value);
i915_request_add(rq);
- if (err)
+ if (err) {
+ i915_request_put(rq);
rq = ERR_PTR(err);
+ }
out_unpin:
intel_timeline_unpin(tl);
@@ -500,7 +500,6 @@ static int live_hwsp_engine(void *arg)
struct intel_timeline **timelines;
struct intel_engine_cs *engine;
enum intel_engine_id id;
- intel_wakeref_t wakeref;
unsigned long count, n;
int err = 0;
@@ -515,14 +514,13 @@ static int live_hwsp_engine(void *arg)
if (!timelines)
return -ENOMEM;
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
count = 0;
for_each_engine(engine, i915, id) {
if (!intel_engine_can_store_dword(engine))
continue;
+ intel_engine_pm_get(engine);
+
for (n = 0; n < NUM_TIMELINES; n++) {
struct intel_timeline *tl;
struct i915_request *rq;
@@ -530,22 +528,26 @@ static int live_hwsp_engine(void *arg)
tl = checked_intel_timeline_create(i915);
if (IS_ERR(tl)) {
err = PTR_ERR(tl);
- goto out;
+ break;
}
rq = tl_write(tl, engine, count);
if (IS_ERR(rq)) {
intel_timeline_put(tl);
err = PTR_ERR(rq);
- goto out;
+ break;
}
timelines[count++] = tl;
+ i915_request_put(rq);
}
+
+ intel_engine_pm_put(engine);
+ if (err)
+ break;
}
-out:
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
err = -EIO;
for (n = 0; n < count; n++) {
@@ -559,11 +561,7 @@ static int live_hwsp_engine(void *arg)
intel_timeline_put(tl);
}
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
-
kvfree(timelines);
-
return err;
#undef NUM_TIMELINES
}
@@ -575,7 +573,6 @@ static int live_hwsp_alternate(void *arg)
struct intel_timeline **timelines;
struct intel_engine_cs *engine;
enum intel_engine_id id;
- intel_wakeref_t wakeref;
unsigned long count, n;
int err = 0;
@@ -591,9 +588,6 @@ static int live_hwsp_alternate(void *arg)
if (!timelines)
return -ENOMEM;
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
count = 0;
for (n = 0; n < NUM_TIMELINES; n++) {
for_each_engine(engine, i915, id) {
@@ -605,11 +599,14 @@ static int live_hwsp_alternate(void *arg)
tl = checked_intel_timeline_create(i915);
if (IS_ERR(tl)) {
+ intel_engine_pm_put(engine);
err = PTR_ERR(tl);
goto out;
}
+ intel_engine_pm_get(engine);
rq = tl_write(tl, engine, count);
+ intel_engine_pm_put(engine);
if (IS_ERR(rq)) {
intel_timeline_put(tl);
err = PTR_ERR(rq);
@@ -617,11 +614,12 @@ static int live_hwsp_alternate(void *arg)
}
timelines[count++] = tl;
+ i915_request_put(rq);
}
}
out:
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
err = -EIO;
for (n = 0; n < count; n++) {
@@ -635,11 +633,7 @@ static int live_hwsp_alternate(void *arg)
intel_timeline_put(tl);
}
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
-
kvfree(timelines);
-
return err;
#undef NUM_TIMELINES
}
@@ -650,7 +644,6 @@ static int live_hwsp_wrap(void *arg)
struct intel_engine_cs *engine;
struct intel_timeline *tl;
enum intel_engine_id id;
- intel_wakeref_t wakeref;
int err = 0;
/*
@@ -658,14 +651,10 @@ static int live_hwsp_wrap(void *arg)
* foreign GPU references.
*/
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
tl = intel_timeline_create(&i915->gt, NULL);
- if (IS_ERR(tl)) {
- err = PTR_ERR(tl);
- goto out_rpm;
- }
+ if (IS_ERR(tl))
+ return PTR_ERR(tl);
+
if (!tl->has_initial_breadcrumb || !tl->hwsp_cacheline)
goto out_free;
@@ -681,7 +670,9 @@ static int live_hwsp_wrap(void *arg)
if (!intel_engine_can_store_dword(engine))
continue;
+ intel_engine_pm_get(engine);
rq = i915_request_create(engine->kernel_context);
+ intel_engine_pm_put(engine);
if (IS_ERR(rq)) {
err = PTR_ERR(rq);
goto out;
@@ -747,16 +738,12 @@ static int live_hwsp_wrap(void *arg)
}
out:
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
err = -EIO;
intel_timeline_unpin(tl);
out_free:
intel_timeline_put(tl);
-out_rpm:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
-
return err;
}
@@ -765,7 +752,6 @@ static int live_hwsp_recycle(void *arg)
struct drm_i915_private *i915 = arg;
struct intel_engine_cs *engine;
enum intel_engine_id id;
- intel_wakeref_t wakeref;
unsigned long count;
int err = 0;
@@ -775,9 +761,6 @@ static int live_hwsp_recycle(void *arg)
* want to confuse ourselves or the GPU.
*/
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
count = 0;
for_each_engine(engine, i915, id) {
IGT_TIMEOUT(end_time);
@@ -785,6 +768,8 @@ static int live_hwsp_recycle(void *arg)
if (!intel_engine_can_store_dword(engine))
continue;
+ intel_engine_pm_get(engine);
+
do {
struct intel_timeline *tl;
struct i915_request *rq;
@@ -792,21 +777,22 @@ static int live_hwsp_recycle(void *arg)
tl = checked_intel_timeline_create(i915);
if (IS_ERR(tl)) {
err = PTR_ERR(tl);
- goto out;
+ break;
}
rq = tl_write(tl, engine, count);
if (IS_ERR(rq)) {
intel_timeline_put(tl);
err = PTR_ERR(rq);
- goto out;
+ break;
}
if (i915_request_wait(rq, 0, HZ / 5) < 0) {
pr_err("Wait for timeline writes timed out!\n");
+ i915_request_put(rq);
intel_timeline_put(tl);
err = -EIO;
- goto out;
+ break;
}
if (*tl->hwsp_seqno != count) {
@@ -815,17 +801,18 @@ static int live_hwsp_recycle(void *arg)
err = -EINVAL;
}
+ i915_request_put(rq);
intel_timeline_put(tl);
count++;
if (err)
- goto out;
+ break;
} while (!__igt_timeout(end_time, NULL));
- }
-out:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
+ intel_engine_pm_put(engine);
+ if (err)
+ break;
+ }
return err;
}
diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
index 999a98f00494..06351fefbbf3 100644
--- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
@@ -676,7 +676,7 @@ static int check_dirty_whitelist(struct i915_gem_context *ctx,
break;
}
- if (igt_flush_test(ctx->i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(ctx->i915))
err = -EIO;
out_batch:
i915_vma_unpin_and_release(&batch, 0);
@@ -1090,7 +1090,7 @@ static int live_isolated_whitelist(void *arg)
kernel_context_close(client[i].ctx);
}
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
err = -EIO;
return err;
@@ -1248,7 +1248,7 @@ live_engine_reset_workarounds(void *arg)
igt_global_reset_unlock(&i915->gt);
kernel_context_close(ctx);
- igt_flush_test(i915, I915_WAIT_LOCKED);
+ igt_flush_test(i915);
return ret;
}
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index b5b449a88cf1..2047f170fbe7 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -3618,6 +3618,7 @@ static int
i915_drop_caches_set(void *data, u64 val)
{
struct drm_i915_private *i915 = data;
+ int ret;
DRM_DEBUG("Dropping caches: 0x%08llx [0x%08llx]\n",
val, val & DROP_ALL);
@@ -3627,40 +3628,21 @@ i915_drop_caches_set(void *data, u64 val)
I915_IDLE_ENGINES_TIMEOUT))
intel_gt_set_wedged(&i915->gt);
- /* No need to check and wait for gpu resets, only libdrm auto-restarts
- * on ioctls on -EAGAIN. */
- if (val & (DROP_ACTIVE | DROP_IDLE | DROP_RETIRE | DROP_RESET_SEQNO)) {
- int ret;
+ if (val & DROP_RETIRE)
+ i915_retire_requests(i915);
- ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
+ if (val & (DROP_IDLE | DROP_ACTIVE)) {
+ ret = i915_gem_wait_for_idle(i915,
+ I915_WAIT_INTERRUPTIBLE,
+ MAX_SCHEDULE_TIMEOUT);
if (ret)
return ret;
+ }
- /*
- * To finish the flush of the idle_worker, we must complete
- * the switch-to-kernel-context, which requires a double
- * pass through wait_for_idle: first queues the switch,
- * second waits for the switch.
- */
- if (ret == 0 && val & (DROP_IDLE | DROP_ACTIVE))
- ret = i915_gem_wait_for_idle(i915,
- I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED,
- MAX_SCHEDULE_TIMEOUT);
-
- if (ret == 0 && val & DROP_IDLE)
- ret = i915_gem_wait_for_idle(i915,
- I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED,
- MAX_SCHEDULE_TIMEOUT);
-
- if (val & DROP_RETIRE)
- i915_retire_requests(i915);
-
- mutex_unlock(&i915->drm.struct_mutex);
-
- if (ret == 0 && val & DROP_IDLE)
- ret = intel_gt_pm_wait_for_idle(&i915->gt);
+ if (val & DROP_IDLE) {
+ ret = intel_gt_pm_wait_for_idle(&i915->gt);
+ if (ret)
+ return ret;
}
if (val & DROP_RESET_ACTIVE && intel_gt_terminally_wedged(&i915->gt))
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index a88849544c8b..47c53bb9eb89 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -945,19 +945,16 @@ int i915_gem_wait_for_idle(struct drm_i915_private *i915,
if (!intel_gt_pm_is_awake(gt))
return 0;
- GEM_TRACE("flags=%x (%s), timeout=%ld%s\n",
- flags, flags & I915_WAIT_LOCKED ? "locked" : "unlocked",
- timeout, timeout == MAX_SCHEDULE_TIMEOUT ? " (forever)" : "");
-
- timeout = wait_for_timelines(gt, flags, timeout);
- if (timeout < 0)
- return timeout;
+ do {
+ timeout = wait_for_timelines(gt, flags, timeout);
+ if (timeout < 0)
+ return timeout;
- if (flags & I915_WAIT_LOCKED) {
- lockdep_assert_held(&i915->drm.struct_mutex);
+ cond_resched();
+ if (signal_pending(current))
+ return -EINTR;
- i915_retire_requests(i915);
- }
+ } while (i915_retire_requests(i915));
return 0;
}
diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
index 91a885c36c6b..621fb33cda30 100644
--- a/drivers/gpu/drm/i915/i915_request.h
+++ b/drivers/gpu/drm/i915/i915_request.h
@@ -308,10 +308,9 @@ long i915_request_wait(struct i915_request *rq,
long timeout)
__attribute__((nonnull(1)));
#define I915_WAIT_INTERRUPTIBLE BIT(0)
-#define I915_WAIT_LOCKED BIT(1) /* struct_mutex held, handle GPU reset */
-#define I915_WAIT_PRIORITY BIT(2) /* small priority bump for the request */
-#define I915_WAIT_ALL BIT(3) /* used by i915_gem_object_wait() */
-#define I915_WAIT_FOR_IDLE_BOOST BIT(4)
+#define I915_WAIT_PRIORITY BIT(1) /* small priority bump for the request */
+#define I915_WAIT_ALL BIT(2) /* used by i915_gem_object_wait() */
+#define I915_WAIT_FOR_IDLE_BOOST BIT(3)
static inline bool i915_request_signaled(const struct i915_request *rq)
{
diff --git a/drivers/gpu/drm/i915/selftests/i915_active.c b/drivers/gpu/drm/i915/selftests/i915_active.c
index 2cc71bcf884f..268192b5613b 100644
--- a/drivers/gpu/drm/i915/selftests/i915_active.c
+++ b/drivers/gpu/drm/i915/selftests/i915_active.c
@@ -162,10 +162,8 @@ static int live_active_wait(void *arg)
__live_put(active);
- mutex_lock(&i915->drm.struct_mutex);
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
err = -EIO;
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -183,10 +181,8 @@ static int live_active_retire(void *arg)
return PTR_ERR(active);
/* waits for & retires all requests */
- mutex_lock(&i915->drm.struct_mutex);
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
err = -EIO;
- mutex_unlock(&i915->drm.struct_mutex);
if (!READ_ONCE(active->retired)) {
pr_err("i915_active not retired after flushing!\n");
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
index 75a4695b82bb..52d2df843148 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
@@ -523,7 +523,7 @@ static int igt_evict_contexts(void *arg)
mutex_lock(&i915->ggtt.vm.mutex);
out_locked:
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
err = -EIO;
while (reserved) {
struct reserved *next = reserved->next;
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
index 373fbd58a17a..9939dd40c6a8 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
@@ -1703,12 +1703,8 @@ int i915_gem_gtt_mock_selftests(void)
err = i915_subtests(tests, ggtt);
- mutex_lock(&i915->drm.struct_mutex);
mock_device_flush(i915);
- mutex_unlock(&i915->drm.struct_mutex);
-
i915_gem_drain_freed_objects(i915);
-
mock_fini_ggtt(ggtt);
kfree(ggtt);
out_put:
@@ -2002,7 +1998,7 @@ static int igt_cs_tlb(void *arg)
}
}
end:
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
err = -EIO;
i915_gem_context_unlock_engines(ctx);
i915_gem_object_unpin_map(out);
diff --git a/drivers/gpu/drm/i915/selftests/i915_request.c b/drivers/gpu/drm/i915/selftests/i915_request.c
index 6b5b03081880..d1ee384a3260 100644
--- a/drivers/gpu/drm/i915/selftests/i915_request.c
+++ b/drivers/gpu/drm/i915/selftests/i915_request.c
@@ -41,21 +41,16 @@ static int igt_add_request(void *arg)
{
struct drm_i915_private *i915 = arg;
struct i915_request *request;
- int err = -ENOMEM;
/* Basic preliminary test to create a request and let it loose! */
- mutex_lock(&i915->drm.struct_mutex);
request = mock_request(i915->engine[RCS0]->kernel_context, HZ / 10);
if (!request)
- goto out_unlock;
+ return -ENOMEM;
i915_request_add(request);
- err = 0;
-out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
- return err;
+ return 0;
}
static int igt_wait_request(void *arg)
@@ -67,12 +62,10 @@ static int igt_wait_request(void *arg)
/* Submit a request, then wait upon it */
- mutex_lock(&i915->drm.struct_mutex);
request = mock_request(i915->engine[RCS0]->kernel_context, T);
- if (!request) {
- err = -ENOMEM;
- goto out_unlock;
- }
+ if (!request)
+ return -ENOMEM;
+
i915_request_get(request);
if (i915_request_wait(request, 0, 0) != -ETIME) {
@@ -125,9 +118,7 @@ static int igt_wait_request(void *arg)
err = 0;
out_request:
i915_request_put(request);
-out_unlock:
mock_device_flush(i915);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -140,52 +131,45 @@ static int igt_fence_wait(void *arg)
/* Submit a request, treat it as a fence and wait upon it */
- mutex_lock(&i915->drm.struct_mutex);
request = mock_request(i915->engine[RCS0]->kernel_context, T);
- if (!request) {
- err = -ENOMEM;
- goto out_locked;
- }
+ if (!request)
+ return -ENOMEM;
if (dma_fence_wait_timeout(&request->fence, false, T) != -ETIME) {
pr_err("fence wait success before submit (expected timeout)!\n");
- goto out_locked;
+ goto out;
}
i915_request_add(request);
- mutex_unlock(&i915->drm.struct_mutex);
if (dma_fence_is_signaled(&request->fence)) {
pr_err("fence signaled immediately!\n");
- goto out_device;
+ goto out;
}
if (dma_fence_wait_timeout(&request->fence, false, T / 2) != -ETIME) {
pr_err("fence wait success after submit (expected timeout)!\n");
- goto out_device;
+ goto out;
}
if (dma_fence_wait_timeout(&request->fence, false, T) <= 0) {
pr_err("fence wait timed out (expected success)!\n");
- goto out_device;
+ goto out;
}
if (!dma_fence_is_signaled(&request->fence)) {
pr_err("fence unsignaled after waiting!\n");
- goto out_device;
+ goto out;
}
if (dma_fence_wait_timeout(&request->fence, false, T) <= 0) {
pr_err("fence wait timed out when complete (expected success)!\n");
- goto out_device;
+ goto out;
}
err = 0;
-out_device:
- mutex_lock(&i915->drm.struct_mutex);
-out_locked:
+out:
mock_device_flush(i915);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -199,6 +183,8 @@ static int igt_request_rewind(void *arg)
mutex_lock(&i915->drm.struct_mutex);
ctx[0] = mock_context(i915, "A");
+ mutex_unlock(&i915->drm.struct_mutex);
+
ce = i915_gem_context_get_engine(ctx[0], RCS0);
GEM_BUG_ON(IS_ERR(ce));
request = mock_request(ce, 2 * HZ);
@@ -211,7 +197,10 @@ static int igt_request_rewind(void *arg)
i915_request_get(request);
i915_request_add(request);
+ mutex_lock(&i915->drm.struct_mutex);
ctx[1] = mock_context(i915, "B");
+ mutex_unlock(&i915->drm.struct_mutex);
+
ce = i915_gem_context_get_engine(ctx[1], RCS0);
GEM_BUG_ON(IS_ERR(ce));
vip = mock_request(ce, 0);
@@ -233,7 +222,6 @@ static int igt_request_rewind(void *arg)
request->engine->submit_request(request);
rcu_read_unlock();
- mutex_unlock(&i915->drm.struct_mutex);
if (i915_request_wait(vip, 0, HZ) == -ETIME) {
pr_err("timed out waiting for high priority request\n");
@@ -248,14 +236,12 @@ static int igt_request_rewind(void *arg)
err = 0;
err:
i915_request_put(vip);
- mutex_lock(&i915->drm.struct_mutex);
err_context_1:
mock_context_close(ctx[1]);
i915_request_put(request);
err_context_0:
mock_context_close(ctx[0]);
mock_device_flush(i915);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -282,7 +268,6 @@ __live_request_alloc(struct intel_context *ce)
static int __igt_breadcrumbs_smoketest(void *arg)
{
struct smoketest *t = arg;
- struct mutex * const BKL = &t->engine->i915->drm.struct_mutex;
const unsigned int max_batch = min(t->ncontexts, t->max_batch) - 1;
const unsigned int total = 4 * t->ncontexts + 1;
unsigned int num_waits = 0, num_fences = 0;
@@ -337,14 +322,11 @@ static int __igt_breadcrumbs_smoketest(void *arg)
struct i915_request *rq;
struct intel_context *ce;
- mutex_lock(BKL);
-
ce = i915_gem_context_get_engine(ctx, t->engine->legacy_idx);
GEM_BUG_ON(IS_ERR(ce));
rq = t->request_alloc(ce);
intel_context_put(ce);
if (IS_ERR(rq)) {
- mutex_unlock(BKL);
err = PTR_ERR(rq);
count = n;
break;
@@ -357,8 +339,6 @@ static int __igt_breadcrumbs_smoketest(void *arg)
requests[n] = i915_request_get(rq);
i915_request_add(rq);
- mutex_unlock(BKL);
-
if (err >= 0)
err = i915_sw_fence_await_dma_fence(wait,
&rq->fence,
@@ -457,15 +437,15 @@ static int mock_breadcrumbs_smoketest(void *arg)
goto out_threads;
}
- mutex_lock(&t.engine->i915->drm.struct_mutex);
for (n = 0; n < t.ncontexts; n++) {
+ mutex_lock(&t.engine->i915->drm.struct_mutex);
t.contexts[n] = mock_context(t.engine->i915, "mock");
+ mutex_unlock(&t.engine->i915->drm.struct_mutex);
if (!t.contexts[n]) {
ret = -ENOMEM;
goto out_contexts;
}
}
- mutex_unlock(&t.engine->i915->drm.struct_mutex);
for (n = 0; n < ncpus; n++) {
threads[n] = kthread_run(__igt_breadcrumbs_smoketest,
@@ -495,18 +475,15 @@ static int mock_breadcrumbs_smoketest(void *arg)
atomic_long_read(&t.num_fences),
ncpus);
- mutex_lock(&t.engine->i915->drm.struct_mutex);
out_contexts:
for (n = 0; n < t.ncontexts; n++) {
if (!t.contexts[n])
break;
mock_context_close(t.contexts[n]);
}
- mutex_unlock(&t.engine->i915->drm.struct_mutex);
kfree(t.contexts);
out_threads:
kfree(threads);
-
return ret;
}
@@ -539,7 +516,6 @@ static int live_nop_request(void *arg)
{
struct drm_i915_private *i915 = arg;
struct intel_engine_cs *engine;
- intel_wakeref_t wakeref;
struct igt_live_test t;
unsigned int id;
int err = -ENODEV;
@@ -549,28 +525,25 @@ static int live_nop_request(void *arg)
* the overhead of submitting requests to the hardware.
*/
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
for_each_engine(engine, i915, id) {
- struct i915_request *request = NULL;
unsigned long n, prime;
IGT_TIMEOUT(end_time);
ktime_t times[2] = {};
err = igt_live_test_begin(&t, i915, __func__, engine->name);
if (err)
- goto out_unlock;
+ return err;
for_each_prime_number_from(prime, 1, 8192) {
+ struct i915_request *request = NULL;
+
times[1] = ktime_get_raw();
for (n = 0; n < prime; n++) {
+ i915_request_put(request);
request = i915_request_create(engine->kernel_context);
- if (IS_ERR(request)) {
- err = PTR_ERR(request);
- goto out_unlock;
- }
+ if (IS_ERR(request))
+ return PTR_ERR(request);
/* This space is left intentionally blank.
*
@@ -585,9 +558,11 @@ static int live_nop_request(void *arg)
* for latency.
*/
+ i915_request_get(request);
i915_request_add(request);
}
i915_request_wait(request, 0, MAX_SCHEDULE_TIMEOUT);
+ i915_request_put(request);
times[1] = ktime_sub(ktime_get_raw(), times[1]);
if (prime == 1)
@@ -599,7 +574,7 @@ static int live_nop_request(void *arg)
err = igt_live_test_end(&t);
if (err)
- goto out_unlock;
+ return err;
pr_info("Request latencies on %s: 1 = %lluns, %lu = %lluns\n",
engine->name,
@@ -607,9 +582,6 @@ static int live_nop_request(void *arg)
prime, div64_u64(ktime_to_ns(times[1]), prime));
}
-out_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -679,6 +651,7 @@ empty_request(struct intel_engine_cs *engine,
if (err)
goto out_request;
+ i915_request_get(request);
out_request:
i915_request_add(request);
return err ? ERR_PTR(err) : request;
@@ -688,7 +661,6 @@ static int live_empty_request(void *arg)
{
struct drm_i915_private *i915 = arg;
struct intel_engine_cs *engine;
- intel_wakeref_t wakeref;
struct igt_live_test t;
struct i915_vma *batch;
unsigned int id;
@@ -699,14 +671,9 @@ static int live_empty_request(void *arg)
* the overhead of submitting requests to the hardware.
*/
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
batch = empty_batch(i915);
- if (IS_ERR(batch)) {
- err = PTR_ERR(batch);
- goto out_unlock;
- }
+ if (IS_ERR(batch))
+ return PTR_ERR(batch);
for_each_engine(engine, i915, id) {
IGT_TIMEOUT(end_time);
@@ -730,6 +697,7 @@ static int live_empty_request(void *arg)
times[1] = ktime_get_raw();
for (n = 0; n < prime; n++) {
+ i915_request_put(request);
request = empty_request(engine, batch);
if (IS_ERR(request)) {
err = PTR_ERR(request);
@@ -745,6 +713,7 @@ static int live_empty_request(void *arg)
if (__igt_timeout(end_time, NULL))
break;
}
+ i915_request_put(request);
err = igt_live_test_end(&t);
if (err)
@@ -759,9 +728,6 @@ static int live_empty_request(void *arg)
out_batch:
i915_vma_unpin(batch);
i915_vma_put(batch);
-out_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -841,7 +807,6 @@ static int live_all_engines(void *arg)
struct drm_i915_private *i915 = arg;
struct intel_engine_cs *engine;
struct i915_request *request[I915_NUM_ENGINES];
- intel_wakeref_t wakeref;
struct igt_live_test t;
struct i915_vma *batch;
unsigned int id;
@@ -852,18 +817,15 @@ static int live_all_engines(void *arg)
* block doing so, and that they don't complete too soon.
*/
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
err = igt_live_test_begin(&t, i915, __func__, "");
if (err)
- goto out_unlock;
+ return err;
batch = recursive_batch(i915);
if (IS_ERR(batch)) {
err = PTR_ERR(batch);
pr_err("%s: Unable to create batch, err=%d\n", __func__, err);
- goto out_unlock;
+ return err;
}
for_each_engine(engine, i915, id) {
@@ -933,9 +895,6 @@ static int live_all_engines(void *arg)
i915_request_put(request[id]);
i915_vma_unpin(batch);
i915_vma_put(batch);
-out_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -945,7 +904,6 @@ static int live_sequential_engines(void *arg)
struct i915_request *request[I915_NUM_ENGINES] = {};
struct i915_request *prev = NULL;
struct intel_engine_cs *engine;
- intel_wakeref_t wakeref;
struct igt_live_test t;
unsigned int id;
int err;
@@ -956,12 +914,9 @@ static int live_sequential_engines(void *arg)
* they are running on independent engines.
*/
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
err = igt_live_test_begin(&t, i915, __func__, "");
if (err)
- goto out_unlock;
+ return err;
for_each_engine(engine, i915, id) {
struct i915_vma *batch;
@@ -971,7 +926,7 @@ static int live_sequential_engines(void *arg)
err = PTR_ERR(batch);
pr_err("%s: Unable to create batch for %s, err=%d\n",
__func__, engine->name, err);
- goto out_unlock;
+ return err;
}
request[id] = i915_request_create(engine->kernel_context);
@@ -1063,9 +1018,6 @@ static int live_sequential_engines(void *arg)
i915_vma_put(request[id]->batch);
i915_request_put(request[id]);
}
-out_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -1156,9 +1108,10 @@ static int live_breadcrumbs_smoketest(void *arg)
goto out_threads;
}
- mutex_lock(&i915->drm.struct_mutex);
for (n = 0; n < t[0].ncontexts; n++) {
+ mutex_lock(&i915->drm.struct_mutex);
t[0].contexts[n] = live_context(i915, file);
+ mutex_unlock(&i915->drm.struct_mutex);
if (!t[0].contexts[n]) {
ret = -ENOMEM;
goto out_contexts;
@@ -1175,7 +1128,6 @@ static int live_breadcrumbs_smoketest(void *arg)
t[id].max_batch = max_batches(t[0].contexts[0], engine);
if (t[id].max_batch < 0) {
ret = t[id].max_batch;
- mutex_unlock(&i915->drm.struct_mutex);
goto out_flush;
}
/* One ring interleaved between requests from all cpus */
@@ -1190,7 +1142,6 @@ static int live_breadcrumbs_smoketest(void *arg)
&t[id], "igt/%d.%d", id, n);
if (IS_ERR(tsk)) {
ret = PTR_ERR(tsk);
- mutex_unlock(&i915->drm.struct_mutex);
goto out_flush;
}
@@ -1198,7 +1149,6 @@ static int live_breadcrumbs_smoketest(void *arg)
threads[id * ncpus + n] = tsk;
}
}
- mutex_unlock(&i915->drm.struct_mutex);
msleep(jiffies_to_msecs(i915_selftest.timeout_jiffies));
@@ -1226,10 +1176,8 @@ static int live_breadcrumbs_smoketest(void *arg)
pr_info("Completed %lu waits for %lu fences across %d engines and %d cpus\n",
num_waits, num_fences, RUNTIME_INFO(i915)->num_engines, ncpus);
- mutex_lock(&i915->drm.struct_mutex);
ret = igt_live_test_end(&live) ?: ret;
out_contexts:
- mutex_unlock(&i915->drm.struct_mutex);
kfree(t[0].contexts);
out_threads:
kfree(threads);
diff --git a/drivers/gpu/drm/i915/selftests/i915_selftest.c b/drivers/gpu/drm/i915/selftests/i915_selftest.c
index 438ea0eaa416..825a8286cbe8 100644
--- a/drivers/gpu/drm/i915/selftests/i915_selftest.c
+++ b/drivers/gpu/drm/i915/selftests/i915_selftest.c
@@ -263,10 +263,8 @@ int __i915_live_teardown(int err, void *data)
{
struct drm_i915_private *i915 = data;
- mutex_lock(&i915->drm.struct_mutex);
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
err = -EIO;
- mutex_unlock(&i915->drm.struct_mutex);
i915_gem_drain_freed_objects(i915);
@@ -284,10 +282,8 @@ int __intel_gt_live_teardown(int err, void *data)
{
struct intel_gt *gt = data;
- mutex_lock(>->i915->drm.struct_mutex);
- if (igt_flush_test(gt->i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(gt->i915))
err = -EIO;
- mutex_unlock(>->i915->drm.struct_mutex);
i915_gem_drain_freed_objects(gt->i915);
diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c
index 0e4f66312b39..1c9db08f7c28 100644
--- a/drivers/gpu/drm/i915/selftests/i915_vma.c
+++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
@@ -833,12 +833,8 @@ int i915_vma_mock_selftests(void)
err = i915_subtests(tests, ggtt);
- mutex_lock(&i915->drm.struct_mutex);
mock_device_flush(i915);
- mutex_unlock(&i915->drm.struct_mutex);
-
i915_gem_drain_freed_objects(i915);
-
mock_fini_ggtt(ggtt);
kfree(ggtt);
out_put:
diff --git a/drivers/gpu/drm/i915/selftests/igt_flush_test.c b/drivers/gpu/drm/i915/selftests/igt_flush_test.c
index d3b5eb402d33..2a5fbe46ea9f 100644
--- a/drivers/gpu/drm/i915/selftests/igt_flush_test.c
+++ b/drivers/gpu/drm/i915/selftests/igt_flush_test.c
@@ -12,31 +12,25 @@
#include "igt_flush_test.h"
-int igt_flush_test(struct drm_i915_private *i915, unsigned int flags)
+int igt_flush_test(struct drm_i915_private *i915)
{
int ret = intel_gt_is_wedged(&i915->gt) ? -EIO : 0;
- int repeat = !!(flags & I915_WAIT_LOCKED);
cond_resched();
- do {
- if (i915_gem_wait_for_idle(i915, flags, HZ / 5) == -ETIME) {
- pr_err("%pS timed out, cancelling all further testing.\n",
- __builtin_return_address(0));
+ i915_retire_requests(i915);
+ if (i915_gem_wait_for_idle(i915, 0, HZ / 5) == -ETIME) {
+ pr_err("%pS timed out, cancelling all further testing.\n",
+ __builtin_return_address(0));
- GEM_TRACE("%pS timed out.\n",
- __builtin_return_address(0));
- GEM_TRACE_DUMP();
+ GEM_TRACE("%pS timed out.\n",
+ __builtin_return_address(0));
+ GEM_TRACE_DUMP();
- intel_gt_set_wedged(&i915->gt);
- repeat = 0;
- ret = -EIO;
- }
-
- /* Ensure we also flush after wedging. */
- if (flags & I915_WAIT_LOCKED)
- i915_retire_requests(i915);
- } while (repeat--);
+ intel_gt_set_wedged(&i915->gt);
+ ret = -EIO;
+ }
+ i915_retire_requests(i915);
return ret;
}
diff --git a/drivers/gpu/drm/i915/selftests/igt_flush_test.h b/drivers/gpu/drm/i915/selftests/igt_flush_test.h
index 63e009927c43..7541fa74e641 100644
--- a/drivers/gpu/drm/i915/selftests/igt_flush_test.h
+++ b/drivers/gpu/drm/i915/selftests/igt_flush_test.h
@@ -9,6 +9,6 @@
struct drm_i915_private;
-int igt_flush_test(struct drm_i915_private *i915, unsigned int flags);
+int igt_flush_test(struct drm_i915_private *i915);
#endif /* IGT_FLUSH_TEST_H */
diff --git a/drivers/gpu/drm/i915/selftests/igt_live_test.c b/drivers/gpu/drm/i915/selftests/igt_live_test.c
index 3e902761cd16..04a6f88fdf64 100644
--- a/drivers/gpu/drm/i915/selftests/igt_live_test.c
+++ b/drivers/gpu/drm/i915/selftests/igt_live_test.c
@@ -19,15 +19,12 @@ int igt_live_test_begin(struct igt_live_test *t,
enum intel_engine_id id;
int err;
- lockdep_assert_held(&i915->drm.struct_mutex);
-
t->i915 = i915;
t->func = func;
t->name = name;
err = i915_gem_wait_for_idle(i915,
- I915_WAIT_INTERRUPTIBLE |
- I915_WAIT_LOCKED,
+ I915_WAIT_INTERRUPTIBLE,
MAX_SCHEDULE_TIMEOUT);
if (err) {
pr_err("%s(%s): failed to idle before, with err=%d!",
@@ -50,9 +47,7 @@ int igt_live_test_end(struct igt_live_test *t)
struct intel_engine_cs *engine;
enum intel_engine_id id;
- lockdep_assert_held(&i915->drm.struct_mutex);
-
- if (igt_flush_test(i915, I915_WAIT_LOCKED))
+ if (igt_flush_test(i915))
return -EIO;
if (t->reset_global != i915_reset_count(&i915->gpu_error)) {
diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
index 91f15fa728cd..94e61a1c7c65 100644
--- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
+++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
@@ -41,8 +41,6 @@ void mock_device_flush(struct drm_i915_private *i915)
struct intel_engine_cs *engine;
enum intel_engine_id id;
- lockdep_assert_held(&i915->drm.struct_mutex);
-
do {
for_each_engine(engine, i915, id)
mock_engine_flush(engine);
@@ -55,9 +53,7 @@ static void mock_device_release(struct drm_device *dev)
struct intel_engine_cs *engine;
enum intel_engine_id id;
- mutex_lock(&i915->drm.struct_mutex);
mock_device_flush(i915);
- mutex_unlock(&i915->drm.struct_mutex);
flush_work(&i915->gem.idle_work);
i915_gem_drain_workqueue(i915);
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* [PATCH 10/27] drm/i915: Remove the GEM idle worker
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (8 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 09/27] drm/i915: Drop struct_mutex from around i915_retire_requests() Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:01 ` [PATCH 11/27] drm/i915: Merge wait_for_timelines with retire_request Chris Wilson
` (20 subsequent siblings)
30 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
Nothing inside the idle worker now requires struct_mutex, so we can
remove the indirection of using our own worker.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/gem/i915_gem_pm.c | 28 ++-----------------
.../drm/i915/gem/selftests/i915_gem_mman.c | 3 --
drivers/gpu/drm/i915/i915_debugfs.c | 5 ----
drivers/gpu/drm/i915/i915_drv.h | 9 ------
.../gpu/drm/i915/selftests/mock_gem_device.c | 6 ----
5 files changed, 2 insertions(+), 49 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
index 535f2f8a23aa..c9c9e49391fa 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
@@ -13,36 +13,13 @@
static void i915_gem_park(struct drm_i915_private *i915)
{
- lockdep_assert_held(&i915->drm.struct_mutex);
+ cancel_delayed_work(&i915->gem.retire_work);
i915_vma_parked(i915);
i915_globals_park();
}
-static void idle_work_handler(struct work_struct *work)
-{
- struct drm_i915_private *i915 =
- container_of(work, typeof(*i915), gem.idle_work);
- bool park;
-
- cancel_delayed_work_sync(&i915->gem.retire_work);
- mutex_lock(&i915->drm.struct_mutex);
-
- intel_wakeref_lock(&i915->gt.wakeref);
- park = (!intel_wakeref_is_active(&i915->gt.wakeref) &&
- !work_pending(work));
- intel_wakeref_unlock(&i915->gt.wakeref);
- if (park)
- i915_gem_park(i915);
- else
- queue_delayed_work(i915->wq,
- &i915->gem.retire_work,
- round_jiffies_up_relative(HZ));
-
- mutex_unlock(&i915->drm.struct_mutex);
-}
-
static void retire_work_handler(struct work_struct *work)
{
struct drm_i915_private *i915 =
@@ -71,7 +48,7 @@ static int pm_notifier(struct notifier_block *nb,
break;
case INTEL_GT_PARK:
- queue_work(i915->wq, &i915->gem.idle_work);
+ i915_gem_park(i915);
break;
}
@@ -264,7 +241,6 @@ void i915_gem_resume(struct drm_i915_private *i915)
void i915_gem_init__pm(struct drm_i915_private *i915)
{
- INIT_WORK(&i915->gem.idle_work, idle_work_handler);
INIT_DELAYED_WORK(&i915->gem.retire_work, retire_work_handler);
i915->gem.pm_notifier.notifier_call = pm_notifier;
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 856b8e467ee8..4ba6ed5c8313 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
@@ -572,11 +572,8 @@ static bool assert_mmap_offset(struct drm_i915_private *i915,
static void disable_retire_worker(struct drm_i915_private *i915)
{
i915_gem_driver_unregister__shrinker(i915);
-
intel_gt_pm_get(&i915->gt);
-
cancel_delayed_work_sync(&i915->gem.retire_work);
- flush_work(&i915->gem.idle_work);
}
static void restore_retire_worker(struct drm_i915_private *i915)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 2047f170fbe7..7bbc9eda8ebe 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -3659,11 +3659,6 @@ i915_drop_caches_set(void *data, u64 val)
i915_gem_shrink_all(i915);
fs_reclaim_release(GFP_KERNEL);
- if (val & DROP_IDLE) {
- flush_delayed_work(&i915->gem.retire_work);
- flush_work(&i915->gem.idle_work);
- }
-
if (val & DROP_FREED)
i915_gem_drain_freed_objects(i915);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index fcf7423075ef..3611fd5446f3 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1726,15 +1726,6 @@ struct drm_i915_private {
* fires, go retire requests.
*/
struct delayed_work retire_work;
-
- /**
- * When we detect an idle GPU, we want to turn on
- * powersaving features. So once we see that there
- * are no more requests outstanding and no more
- * arrive within a small period of time, we fire
- * off the idle_work.
- */
- struct work_struct idle_work;
} gem;
/* For i945gm vblank irq vs. C3 workaround */
diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
index 94e61a1c7c65..91f6253939ad 100644
--- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
+++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
@@ -55,7 +55,6 @@ static void mock_device_release(struct drm_device *dev)
mock_device_flush(i915);
- flush_work(&i915->gem.idle_work);
i915_gem_drain_workqueue(i915);
mutex_lock(&i915->drm.struct_mutex);
@@ -103,10 +102,6 @@ static void mock_retire_work_handler(struct work_struct *work)
{
}
-static void mock_idle_work_handler(struct work_struct *work)
-{
-}
-
static int pm_domain_resume(struct device *dev)
{
return pm_generic_runtime_resume(dev);
@@ -186,7 +181,6 @@ struct drm_i915_private *mock_gem_device(void)
mock_init_contexts(i915);
INIT_DELAYED_WORK(&i915->gem.retire_work, mock_retire_work_handler);
- INIT_WORK(&i915->gem.idle_work, mock_idle_work_handler);
i915->gt.awake = -1;
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* [PATCH 11/27] drm/i915: Merge wait_for_timelines with retire_request
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (9 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 10/27] drm/i915: Remove the GEM idle worker Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:47 ` Tvrtko Ursulin
2019-09-25 10:01 ` [PATCH 12/27] drm/i915: Move request runtime management onto gt Chris Wilson
` (19 subsequent siblings)
30 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
wait_for_timelines is essentially the same loop as retiring requests
(with an extra timeout), so merge the two into one routine.
v2: i915_retire_requests_timeout and keep VT'd w/a as !interruptible
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/gem/i915_gem_mman.c | 4 +-
drivers/gpu/drm/i915/gem/i915_gem_pm.c | 4 +-
.../drm/i915/gem/selftests/i915_gem_context.c | 2 +-
drivers/gpu/drm/i915/i915_debugfs.c | 4 +-
drivers/gpu/drm/i915/i915_drv.h | 3 +-
drivers/gpu/drm/i915/i915_gem.c | 67 ++-----------------
drivers/gpu/drm/i915/i915_gem_evict.c | 12 ++--
drivers/gpu/drm/i915/i915_gem_gtt.c | 4 +-
drivers/gpu/drm/i915/i915_request.c | 26 ++++++-
drivers/gpu/drm/i915/i915_request.h | 7 +-
.../gpu/drm/i915/selftests/igt_flush_test.c | 4 +-
.../gpu/drm/i915/selftests/igt_live_test.c | 4 +-
.../gpu/drm/i915/selftests/mock_gem_device.c | 2 +-
13 files changed, 49 insertions(+), 94 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
index c44614895487..9cb59fe1a736 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_mman.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
@@ -431,9 +431,7 @@ static int create_mmap_offset(struct drm_i915_gem_object *obj)
/* Attempt to reap some mmap space from dead objects */
do {
- err = i915_gem_wait_for_idle(i915,
- I915_WAIT_INTERRUPTIBLE,
- MAX_SCHEDULE_TIMEOUT);
+ err = i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
if (err)
break;
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
index c9c9e49391fa..04a9a27d659d 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
@@ -59,9 +59,7 @@ static bool switch_to_kernel_context_sync(struct intel_gt *gt)
{
bool result = !intel_gt_is_wedged(gt);
- if (i915_gem_wait_for_idle(gt->i915,
- I915_WAIT_FOR_IDLE_BOOST,
- I915_GEM_IDLE_TIMEOUT) == -ETIME) {
+ if (i915_gem_wait_for_idle(gt->i915, I915_GEM_IDLE_TIMEOUT) == -ETIME) {
/* XXX hide warning from gem_eio */
if (i915_modparams.reset) {
dev_err(gt->i915->drm.dev,
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 82f54e819eb0..b26f550f8430 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
@@ -923,7 +923,7 @@ __sseu_finish(const char *name,
if ((flags & TEST_IDLE) && ret == 0) {
ret = i915_gem_wait_for_idle(ce->engine->i915,
- 0, MAX_SCHEDULE_TIMEOUT);
+ MAX_SCHEDULE_TIMEOUT);
if (ret)
return ret;
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 7bbc9eda8ebe..033764326601 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -3632,9 +3632,7 @@ i915_drop_caches_set(void *data, u64 val)
i915_retire_requests(i915);
if (val & (DROP_IDLE | DROP_ACTIVE)) {
- ret = i915_gem_wait_for_idle(i915,
- I915_WAIT_INTERRUPTIBLE,
- MAX_SCHEDULE_TIMEOUT);
+ ret = i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
if (ret)
return ret;
}
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 3611fd5446f3..6e5d5e4ca91d 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2328,8 +2328,7 @@ void i915_gem_driver_register(struct drm_i915_private *i915);
void i915_gem_driver_unregister(struct drm_i915_private *i915);
void i915_gem_driver_remove(struct drm_i915_private *dev_priv);
void i915_gem_driver_release(struct drm_i915_private *dev_priv);
-int i915_gem_wait_for_idle(struct drm_i915_private *dev_priv,
- unsigned int flags, long timeout);
+int i915_gem_wait_for_idle(struct drm_i915_private *dev_priv, long timeout);
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);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 47c53bb9eb89..ab23944571fa 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -883,61 +883,7 @@ void i915_gem_runtime_suspend(struct drm_i915_private *i915)
}
}
-static long
-wait_for_timelines(struct intel_gt *gt, unsigned int wait, long timeout)
-{
- struct intel_gt_timelines *timelines = >->timelines;
- struct intel_timeline *tl;
- unsigned long flags;
-
- spin_lock_irqsave(&timelines->lock, flags);
- list_for_each_entry(tl, &timelines->active_list, link) {
- struct dma_fence *fence;
-
- fence = i915_active_fence_get(&tl->last_request);
- if (!fence)
- continue;
-
- spin_unlock_irqrestore(&timelines->lock, flags);
-
- if (!dma_fence_is_i915(fence)) {
- timeout = dma_fence_wait_timeout(fence,
- flags & I915_WAIT_INTERRUPTIBLE,
- timeout);
- } else {
- struct i915_request *rq = to_request(fence);
-
- /*
- * "Race-to-idle".
- *
- * Switching to the kernel context is often used as
- * a synchronous step prior to idling, e.g. in suspend
- * for flushing all current operations to memory before
- * sleeping. These we want to complete as quickly as
- * possible to avoid prolonged stalls, so allow the gpu
- * to boost to maximum clocks.
- */
- if (flags & I915_WAIT_FOR_IDLE_BOOST)
- gen6_rps_boost(rq);
-
- timeout = i915_request_wait(rq, flags, timeout);
- }
-
- dma_fence_put(fence);
- if (timeout < 0)
- return timeout;
-
- /* restart after reacquiring the lock */
- spin_lock_irqsave(&timelines->lock, flags);
- tl = list_entry(&timelines->active_list, typeof(*tl), link);
- }
- spin_unlock_irqrestore(&timelines->lock, flags);
-
- return timeout;
-}
-
-int i915_gem_wait_for_idle(struct drm_i915_private *i915,
- unsigned int flags, long timeout)
+int i915_gem_wait_for_idle(struct drm_i915_private *i915, long timeout)
{
struct intel_gt *gt = &i915->gt;
@@ -945,18 +891,13 @@ int i915_gem_wait_for_idle(struct drm_i915_private *i915,
if (!intel_gt_pm_is_awake(gt))
return 0;
- do {
- timeout = wait_for_timelines(gt, flags, timeout);
- if (timeout < 0)
- return timeout;
-
+ while ((timeout = i915_retire_requests_timeout(i915, timeout)) > 0) {
cond_resched();
if (signal_pending(current))
return -EINTR;
+ }
- } while (i915_retire_requests(i915));
-
- return 0;
+ return timeout;
}
struct i915_vma *
diff --git a/drivers/gpu/drm/i915/i915_gem_evict.c b/drivers/gpu/drm/i915/i915_gem_evict.c
index 0552bf93eea3..0a412f6d01d7 100644
--- a/drivers/gpu/drm/i915/i915_gem_evict.c
+++ b/drivers/gpu/drm/i915/i915_gem_evict.c
@@ -46,9 +46,7 @@ static int ggtt_flush(struct drm_i915_private *i915)
* the hopes that we can then remove contexts and the like only
* bound by their active reference.
*/
- return i915_gem_wait_for_idle(i915,
- I915_WAIT_INTERRUPTIBLE,
- MAX_SCHEDULE_TIMEOUT);
+ return i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
}
static bool
@@ -126,6 +124,8 @@ i915_gem_evict_something(struct i915_address_space *vm,
min_size, alignment, color,
start, end, mode);
+ i915_retire_requests(vm->i915);
+
search_again:
active = NULL;
INIT_LIST_HEAD(&eviction_list);
@@ -264,13 +264,13 @@ int i915_gem_evict_for_node(struct i915_address_space *vm,
trace_i915_gem_evict_node(vm, target, flags);
- /* Retire before we search the active list. Although we have
+ /*
+ * Retire before we search the active list. Although we have
* reasonable accuracy in our retirement lists, we may have
* a stray pin (preventing eviction) that can only be resolved by
* retiring.
*/
- if (!(flags & PIN_NONBLOCK))
- i915_retire_requests(vm->i915);
+ i915_retire_requests(vm->i915);
if (i915_vm_has_cache_coloring(vm)) {
/* Expand search to cover neighbouring guard pages (or lack!) */
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index bf57c41fa1fb..30ee016ac2ba 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -2528,7 +2528,9 @@ void i915_gem_gtt_finish_pages(struct drm_i915_gem_object *obj,
struct i915_ggtt *ggtt = &dev_priv->ggtt;
if (unlikely(ggtt->do_idle_maps)) {
- if (i915_gem_wait_for_idle(dev_priv, 0, MAX_SCHEDULE_TIMEOUT)) {
+ /* XXX This does not prevent more requests being submitted! */
+ if (i915_retire_requests_timeout(dev_priv,
+ -MAX_SCHEDULE_TIMEOUT)) {
DRM_ERROR("Failed to wait for idle; VT'd may hang.\n");
/* Wait a bit, in hopes it avoids the hang */
udelay(10);
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 4ffe62a42186..52f7c4e5b644 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -1508,13 +1508,19 @@ long i915_request_wait(struct i915_request *rq,
return timeout;
}
-bool i915_retire_requests(struct drm_i915_private *i915)
+long i915_retire_requests_timeout(struct drm_i915_private *i915, long timeout)
{
struct intel_gt_timelines *timelines = &i915->gt.timelines;
struct intel_timeline *tl, *tn;
+ unsigned long active_count = 0;
unsigned long flags;
+ bool interruptible;
LIST_HEAD(free);
+ interruptible = true;
+ if (timeout < 0)
+ timeout = -timeout, interruptible = false;
+
spin_lock_irqsave(&timelines->lock, flags);
list_for_each_entry_safe(tl, tn, &timelines->active_list, link) {
if (!mutex_trylock(&tl->mutex))
@@ -1525,13 +1531,27 @@ bool i915_retire_requests(struct drm_i915_private *i915)
tl->active_count++; /* pin the list element */
spin_unlock_irqrestore(&timelines->lock, flags);
+ if (timeout > 0) {
+ struct dma_fence *fence;
+
+ fence = i915_active_fence_get(&tl->last_request);
+ if (fence) {
+ timeout = dma_fence_wait_timeout(fence,
+ interruptible,
+ timeout);
+ dma_fence_put(fence);
+ }
+ }
+
retire_requests(tl);
spin_lock_irqsave(&timelines->lock, flags);
/* Resume iteration after dropping lock */
list_safe_reset_next(tl, tn, link);
- if (!--tl->active_count)
+ if (--tl->active_count)
+ active_count += !!rcu_access_pointer(tl->last_request.fence);
+ else
list_del(&tl->link);
mutex_unlock(&tl->mutex);
@@ -1547,7 +1567,7 @@ bool i915_retire_requests(struct drm_i915_private *i915)
list_for_each_entry_safe(tl, tn, &free, link)
__intel_timeline_free(&tl->kref);
- return !list_empty(&timelines->active_list);
+ return active_count ? timeout : 0;
}
#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
index 621fb33cda30..256b0715180f 100644
--- a/drivers/gpu/drm/i915/i915_request.h
+++ b/drivers/gpu/drm/i915/i915_request.h
@@ -310,7 +310,6 @@ long i915_request_wait(struct i915_request *rq,
#define I915_WAIT_INTERRUPTIBLE BIT(0)
#define I915_WAIT_PRIORITY BIT(1) /* small priority bump for the request */
#define I915_WAIT_ALL BIT(2) /* used by i915_gem_object_wait() */
-#define I915_WAIT_FOR_IDLE_BOOST BIT(3)
static inline bool i915_request_signaled(const struct i915_request *rq)
{
@@ -460,6 +459,10 @@ i915_request_active_timeline(struct i915_request *rq)
lockdep_is_held(&rq->engine->active.lock));
}
-bool i915_retire_requests(struct drm_i915_private *i915);
+long i915_retire_requests_timeout(struct drm_i915_private *i915, long timeout);
+static inline void i915_retire_requests(struct drm_i915_private *i915)
+{
+ i915_retire_requests_timeout(i915, 0);
+}
#endif /* I915_REQUEST_H */
diff --git a/drivers/gpu/drm/i915/selftests/igt_flush_test.c b/drivers/gpu/drm/i915/selftests/igt_flush_test.c
index 2a5fbe46ea9f..ed496bd6d84f 100644
--- a/drivers/gpu/drm/i915/selftests/igt_flush_test.c
+++ b/drivers/gpu/drm/i915/selftests/igt_flush_test.c
@@ -18,8 +18,7 @@ int igt_flush_test(struct drm_i915_private *i915)
cond_resched();
- i915_retire_requests(i915);
- if (i915_gem_wait_for_idle(i915, 0, HZ / 5) == -ETIME) {
+ if (i915_gem_wait_for_idle(i915, HZ / 5) == -ETIME) {
pr_err("%pS timed out, cancelling all further testing.\n",
__builtin_return_address(0));
@@ -30,7 +29,6 @@ int igt_flush_test(struct drm_i915_private *i915)
intel_gt_set_wedged(&i915->gt);
ret = -EIO;
}
- i915_retire_requests(i915);
return ret;
}
diff --git a/drivers/gpu/drm/i915/selftests/igt_live_test.c b/drivers/gpu/drm/i915/selftests/igt_live_test.c
index 04a6f88fdf64..eae90f97df6c 100644
--- a/drivers/gpu/drm/i915/selftests/igt_live_test.c
+++ b/drivers/gpu/drm/i915/selftests/igt_live_test.c
@@ -23,9 +23,7 @@ int igt_live_test_begin(struct igt_live_test *t,
t->func = func;
t->name = name;
- err = i915_gem_wait_for_idle(i915,
- I915_WAIT_INTERRUPTIBLE,
- MAX_SCHEDULE_TIMEOUT);
+ err = i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
if (err) {
pr_err("%s(%s): failed to idle before, with err=%d!",
func, name, err);
diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
index 91f6253939ad..87d34167bf89 100644
--- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
+++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
@@ -44,7 +44,7 @@ void mock_device_flush(struct drm_i915_private *i915)
do {
for_each_engine(engine, i915, id)
mock_engine_flush(engine);
- } while (i915_retire_requests(i915));
+ } while (i915_retire_requests_timeout(i915, MAX_SCHEDULE_TIMEOUT));
}
static void mock_device_release(struct drm_device *dev)
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* Re: [PATCH 11/27] drm/i915: Merge wait_for_timelines with retire_request
2019-09-25 10:01 ` [PATCH 11/27] drm/i915: Merge wait_for_timelines with retire_request Chris Wilson
@ 2019-09-25 10:47 ` Tvrtko Ursulin
2019-09-25 10:54 ` Chris Wilson
0 siblings, 1 reply; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-25 10:47 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 25/09/2019 11:01, Chris Wilson wrote:
> wait_for_timelines is essentially the same loop as retiring requests
> (with an extra timeout), so merge the two into one routine.
>
> v2: i915_retire_requests_timeout and keep VT'd w/a as !interruptible
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/gem/i915_gem_mman.c | 4 +-
> drivers/gpu/drm/i915/gem/i915_gem_pm.c | 4 +-
> .../drm/i915/gem/selftests/i915_gem_context.c | 2 +-
> drivers/gpu/drm/i915/i915_debugfs.c | 4 +-
> drivers/gpu/drm/i915/i915_drv.h | 3 +-
> drivers/gpu/drm/i915/i915_gem.c | 67 ++-----------------
> drivers/gpu/drm/i915/i915_gem_evict.c | 12 ++--
> drivers/gpu/drm/i915/i915_gem_gtt.c | 4 +-
> drivers/gpu/drm/i915/i915_request.c | 26 ++++++-
> drivers/gpu/drm/i915/i915_request.h | 7 +-
> .../gpu/drm/i915/selftests/igt_flush_test.c | 4 +-
> .../gpu/drm/i915/selftests/igt_live_test.c | 4 +-
> .../gpu/drm/i915/selftests/mock_gem_device.c | 2 +-
> 13 files changed, 49 insertions(+), 94 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
> index c44614895487..9cb59fe1a736 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_mman.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
> @@ -431,9 +431,7 @@ static int create_mmap_offset(struct drm_i915_gem_object *obj)
>
> /* Attempt to reap some mmap space from dead objects */
> do {
> - err = i915_gem_wait_for_idle(i915,
> - I915_WAIT_INTERRUPTIBLE,
> - MAX_SCHEDULE_TIMEOUT);
> + err = i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
> if (err)
> break;
>
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
> index c9c9e49391fa..04a9a27d659d 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
> @@ -59,9 +59,7 @@ static bool switch_to_kernel_context_sync(struct intel_gt *gt)
> {
> bool result = !intel_gt_is_wedged(gt);
>
> - if (i915_gem_wait_for_idle(gt->i915,
> - I915_WAIT_FOR_IDLE_BOOST,
> - I915_GEM_IDLE_TIMEOUT) == -ETIME) {
> + if (i915_gem_wait_for_idle(gt->i915, I915_GEM_IDLE_TIMEOUT) == -ETIME) {
> /* XXX hide warning from gem_eio */
> if (i915_modparams.reset) {
> dev_err(gt->i915->drm.dev,
> 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 82f54e819eb0..b26f550f8430 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
> @@ -923,7 +923,7 @@ __sseu_finish(const char *name,
>
> if ((flags & TEST_IDLE) && ret == 0) {
> ret = i915_gem_wait_for_idle(ce->engine->i915,
> - 0, MAX_SCHEDULE_TIMEOUT);
> + MAX_SCHEDULE_TIMEOUT);
> if (ret)
> return ret;
>
> diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
> index 7bbc9eda8ebe..033764326601 100644
> --- a/drivers/gpu/drm/i915/i915_debugfs.c
> +++ b/drivers/gpu/drm/i915/i915_debugfs.c
> @@ -3632,9 +3632,7 @@ i915_drop_caches_set(void *data, u64 val)
> i915_retire_requests(i915);
>
> if (val & (DROP_IDLE | DROP_ACTIVE)) {
> - ret = i915_gem_wait_for_idle(i915,
> - I915_WAIT_INTERRUPTIBLE,
> - MAX_SCHEDULE_TIMEOUT);
> + ret = i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
> if (ret)
> return ret;
> }
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index 3611fd5446f3..6e5d5e4ca91d 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -2328,8 +2328,7 @@ void i915_gem_driver_register(struct drm_i915_private *i915);
> void i915_gem_driver_unregister(struct drm_i915_private *i915);
> void i915_gem_driver_remove(struct drm_i915_private *dev_priv);
> void i915_gem_driver_release(struct drm_i915_private *dev_priv);
> -int i915_gem_wait_for_idle(struct drm_i915_private *dev_priv,
> - unsigned int flags, long timeout);
> +int i915_gem_wait_for_idle(struct drm_i915_private *dev_priv, long timeout);
> 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);
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index 47c53bb9eb89..ab23944571fa 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -883,61 +883,7 @@ void i915_gem_runtime_suspend(struct drm_i915_private *i915)
> }
> }
>
> -static long
> -wait_for_timelines(struct intel_gt *gt, unsigned int wait, long timeout)
> -{
> - struct intel_gt_timelines *timelines = >->timelines;
> - struct intel_timeline *tl;
> - unsigned long flags;
> -
> - spin_lock_irqsave(&timelines->lock, flags);
> - list_for_each_entry(tl, &timelines->active_list, link) {
> - struct dma_fence *fence;
> -
> - fence = i915_active_fence_get(&tl->last_request);
> - if (!fence)
> - continue;
> -
> - spin_unlock_irqrestore(&timelines->lock, flags);
> -
> - if (!dma_fence_is_i915(fence)) {
> - timeout = dma_fence_wait_timeout(fence,
> - flags & I915_WAIT_INTERRUPTIBLE,
> - timeout);
> - } else {
> - struct i915_request *rq = to_request(fence);
> -
> - /*
> - * "Race-to-idle".
> - *
> - * Switching to the kernel context is often used as
> - * a synchronous step prior to idling, e.g. in suspend
> - * for flushing all current operations to memory before
> - * sleeping. These we want to complete as quickly as
> - * possible to avoid prolonged stalls, so allow the gpu
> - * to boost to maximum clocks.
> - */
> - if (flags & I915_WAIT_FOR_IDLE_BOOST)
> - gen6_rps_boost(rq);
> -
> - timeout = i915_request_wait(rq, flags, timeout);
> - }
> -
> - dma_fence_put(fence);
> - if (timeout < 0)
> - return timeout;
> -
> - /* restart after reacquiring the lock */
> - spin_lock_irqsave(&timelines->lock, flags);
> - tl = list_entry(&timelines->active_list, typeof(*tl), link);
> - }
> - spin_unlock_irqrestore(&timelines->lock, flags);
> -
> - return timeout;
> -}
> -
> -int i915_gem_wait_for_idle(struct drm_i915_private *i915,
> - unsigned int flags, long timeout)
> +int i915_gem_wait_for_idle(struct drm_i915_private *i915, long timeout)
> {
> struct intel_gt *gt = &i915->gt;
>
> @@ -945,18 +891,13 @@ int i915_gem_wait_for_idle(struct drm_i915_private *i915,
> if (!intel_gt_pm_is_awake(gt))
> return 0;
>
> - do {
> - timeout = wait_for_timelines(gt, flags, timeout);
> - if (timeout < 0)
> - return timeout;
> -
> + while ((timeout = i915_retire_requests_timeout(i915, timeout)) > 0) {
> cond_resched();
> if (signal_pending(current))
> return -EINTR;
> + }
>
> - } while (i915_retire_requests(i915));
> -
> - return 0;
> + return timeout;
> }
>
> struct i915_vma *
> diff --git a/drivers/gpu/drm/i915/i915_gem_evict.c b/drivers/gpu/drm/i915/i915_gem_evict.c
> index 0552bf93eea3..0a412f6d01d7 100644
> --- a/drivers/gpu/drm/i915/i915_gem_evict.c
> +++ b/drivers/gpu/drm/i915/i915_gem_evict.c
> @@ -46,9 +46,7 @@ static int ggtt_flush(struct drm_i915_private *i915)
> * the hopes that we can then remove contexts and the like only
> * bound by their active reference.
> */
> - return i915_gem_wait_for_idle(i915,
> - I915_WAIT_INTERRUPTIBLE,
> - MAX_SCHEDULE_TIMEOUT);
> + return i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
> }
>
> static bool
> @@ -126,6 +124,8 @@ i915_gem_evict_something(struct i915_address_space *vm,
> min_size, alignment, color,
> start, end, mode);
>
> + i915_retire_requests(vm->i915);
> +
> search_again:
> active = NULL;
> INIT_LIST_HEAD(&eviction_list);
> @@ -264,13 +264,13 @@ int i915_gem_evict_for_node(struct i915_address_space *vm,
>
> trace_i915_gem_evict_node(vm, target, flags);
>
> - /* Retire before we search the active list. Although we have
> + /*
> + * Retire before we search the active list. Although we have
> * reasonable accuracy in our retirement lists, we may have
> * a stray pin (preventing eviction) that can only be resolved by
> * retiring.
> */
> - if (!(flags & PIN_NONBLOCK))
> - i915_retire_requests(vm->i915);
> + i915_retire_requests(vm->i915);
>
> if (i915_vm_has_cache_coloring(vm)) {
> /* Expand search to cover neighbouring guard pages (or lack!) */
> diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
> index bf57c41fa1fb..30ee016ac2ba 100644
> --- a/drivers/gpu/drm/i915/i915_gem_gtt.c
> +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
> @@ -2528,7 +2528,9 @@ void i915_gem_gtt_finish_pages(struct drm_i915_gem_object *obj,
> struct i915_ggtt *ggtt = &dev_priv->ggtt;
>
> if (unlikely(ggtt->do_idle_maps)) {
> - if (i915_gem_wait_for_idle(dev_priv, 0, MAX_SCHEDULE_TIMEOUT)) {
> + /* XXX This does not prevent more requests being submitted! */
> + if (i915_retire_requests_timeout(dev_priv,
> + -MAX_SCHEDULE_TIMEOUT)) {
> DRM_ERROR("Failed to wait for idle; VT'd may hang.\n");
> /* Wait a bit, in hopes it avoids the hang */
> udelay(10);
> diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
> index 4ffe62a42186..52f7c4e5b644 100644
> --- a/drivers/gpu/drm/i915/i915_request.c
> +++ b/drivers/gpu/drm/i915/i915_request.c
> @@ -1508,13 +1508,19 @@ long i915_request_wait(struct i915_request *rq,
> return timeout;
> }
>
> -bool i915_retire_requests(struct drm_i915_private *i915)
> +long i915_retire_requests_timeout(struct drm_i915_private *i915, long timeout)
> {
> struct intel_gt_timelines *timelines = &i915->gt.timelines;
> struct intel_timeline *tl, *tn;
> + unsigned long active_count = 0;
> unsigned long flags;
> + bool interruptible;
> LIST_HEAD(free);
>
> + interruptible = true;
> + if (timeout < 0)
> + timeout = -timeout, interruptible = false;
Slightly uneasy about the negative trick but alternatives which came to
mind of or-ing the negative bit or wrapper for passing in the
(non-)interruptible(timeout) do not sound so hot either.
> +
> spin_lock_irqsave(&timelines->lock, flags);
> list_for_each_entry_safe(tl, tn, &timelines->active_list, link) {
> if (!mutex_trylock(&tl->mutex))
> @@ -1525,13 +1531,27 @@ bool i915_retire_requests(struct drm_i915_private *i915)
> tl->active_count++; /* pin the list element */
> spin_unlock_irqrestore(&timelines->lock, flags);
>
> + if (timeout > 0) {
> + struct dma_fence *fence;
> +
> + fence = i915_active_fence_get(&tl->last_request);
> + if (fence) {
> + timeout = dma_fence_wait_timeout(fence,
> + interruptible,
> + timeout);
> + dma_fence_put(fence);
> + }
> + }
> +
> retire_requests(tl);
>
> spin_lock_irqsave(&timelines->lock, flags);
>
> /* Resume iteration after dropping lock */
> list_safe_reset_next(tl, tn, link);
> - if (!--tl->active_count)
> + if (--tl->active_count)
> + active_count += !!rcu_access_pointer(tl->last_request.fence);
> + else
> list_del(&tl->link);
>
> mutex_unlock(&tl->mutex);
> @@ -1547,7 +1567,7 @@ bool i915_retire_requests(struct drm_i915_private *i915)
> list_for_each_entry_safe(tl, tn, &free, link)
> __intel_timeline_free(&tl->kref);
>
> - return !list_empty(&timelines->active_list);
> + return active_count ? timeout : 0;
> }
>
> #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
> diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
> index 621fb33cda30..256b0715180f 100644
> --- a/drivers/gpu/drm/i915/i915_request.h
> +++ b/drivers/gpu/drm/i915/i915_request.h
> @@ -310,7 +310,6 @@ long i915_request_wait(struct i915_request *rq,
> #define I915_WAIT_INTERRUPTIBLE BIT(0)
> #define I915_WAIT_PRIORITY BIT(1) /* small priority bump for the request */
> #define I915_WAIT_ALL BIT(2) /* used by i915_gem_object_wait() */
> -#define I915_WAIT_FOR_IDLE_BOOST BIT(3)
>
> static inline bool i915_request_signaled(const struct i915_request *rq)
> {
> @@ -460,6 +459,10 @@ i915_request_active_timeline(struct i915_request *rq)
> lockdep_is_held(&rq->engine->active.lock));
> }
>
> -bool i915_retire_requests(struct drm_i915_private *i915);
> +long i915_retire_requests_timeout(struct drm_i915_private *i915, long timeout);
> +static inline void i915_retire_requests(struct drm_i915_private *i915)
> +{
> + i915_retire_requests_timeout(i915, 0);
> +}
>
> #endif /* I915_REQUEST_H */
> diff --git a/drivers/gpu/drm/i915/selftests/igt_flush_test.c b/drivers/gpu/drm/i915/selftests/igt_flush_test.c
> index 2a5fbe46ea9f..ed496bd6d84f 100644
> --- a/drivers/gpu/drm/i915/selftests/igt_flush_test.c
> +++ b/drivers/gpu/drm/i915/selftests/igt_flush_test.c
> @@ -18,8 +18,7 @@ int igt_flush_test(struct drm_i915_private *i915)
>
> cond_resched();
>
> - i915_retire_requests(i915);
> - if (i915_gem_wait_for_idle(i915, 0, HZ / 5) == -ETIME) {
> + if (i915_gem_wait_for_idle(i915, HZ / 5) == -ETIME) {
> pr_err("%pS timed out, cancelling all further testing.\n",
> __builtin_return_address(0));
>
> @@ -30,7 +29,6 @@ int igt_flush_test(struct drm_i915_private *i915)
> intel_gt_set_wedged(&i915->gt);
> ret = -EIO;
> }
> - i915_retire_requests(i915);
>
> return ret;
> }
> diff --git a/drivers/gpu/drm/i915/selftests/igt_live_test.c b/drivers/gpu/drm/i915/selftests/igt_live_test.c
> index 04a6f88fdf64..eae90f97df6c 100644
> --- a/drivers/gpu/drm/i915/selftests/igt_live_test.c
> +++ b/drivers/gpu/drm/i915/selftests/igt_live_test.c
> @@ -23,9 +23,7 @@ int igt_live_test_begin(struct igt_live_test *t,
> t->func = func;
> t->name = name;
>
> - err = i915_gem_wait_for_idle(i915,
> - I915_WAIT_INTERRUPTIBLE,
> - MAX_SCHEDULE_TIMEOUT);
> + err = i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
> if (err) {
> pr_err("%s(%s): failed to idle before, with err=%d!",
> func, name, err);
> diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
> index 91f6253939ad..87d34167bf89 100644
> --- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
> +++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
> @@ -44,7 +44,7 @@ void mock_device_flush(struct drm_i915_private *i915)
> do {
> for_each_engine(engine, i915, id)
> mock_engine_flush(engine);
> - } while (i915_retire_requests(i915));
> + } while (i915_retire_requests_timeout(i915, MAX_SCHEDULE_TIMEOUT));
> }
>
> static void mock_device_release(struct drm_device *dev)
>
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] 70+ messages in thread* Re: [PATCH 11/27] drm/i915: Merge wait_for_timelines with retire_request
2019-09-25 10:47 ` Tvrtko Ursulin
@ 2019-09-25 10:54 ` Chris Wilson
0 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:54 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Tvrtko Ursulin (2019-09-25 11:47:35)
>
> On 25/09/2019 11:01, Chris Wilson wrote:
> > -bool i915_retire_requests(struct drm_i915_private *i915)
> > +long i915_retire_requests_timeout(struct drm_i915_private *i915, long timeout)
> > {
> > struct intel_gt_timelines *timelines = &i915->gt.timelines;
> > struct intel_timeline *tl, *tn;
> > + unsigned long active_count = 0;
> > unsigned long flags;
> > + bool interruptible;
> > LIST_HEAD(free);
> >
> > + interruptible = true;
> > + if (timeout < 0)
> > + timeout = -timeout, interruptible = false;
>
> Slightly uneasy about the negative trick but alternatives which came to
> mind of or-ing the negative bit or wrapper for passing in the
> (non-)interruptible(timeout) do not sound so hot either.
Queasy feeling concurred. Our mission is not to have any
non-interruptible sleeps, but if we do we'll have to add unsigned flags
back. :|
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* [PATCH 12/27] drm/i915: Move request runtime management onto gt
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (10 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 11/27] drm/i915: Merge wait_for_timelines with retire_request Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:57 ` Tvrtko Ursulin
2019-09-25 10:01 ` [PATCH 13/27] drm/i915: Move global activity tracking from GEM to GT Chris Wilson
` (18 subsequent siblings)
30 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
Requests are run from the gt and are tided into the gt runtime power
management, so pull the runtime request management under gt/
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/Makefile | 1 +
drivers/gpu/drm/i915/gem/i915_gem_mman.c | 6 +-
drivers/gpu/drm/i915/gem/i915_gem_pm.c | 22 +---
.../drm/i915/gem/selftests/i915_gem_context.c | 5 +-
.../drm/i915/gem/selftests/i915_gem_mman.c | 2 +-
drivers/gpu/drm/i915/gt/intel_gt.c | 2 +
drivers/gpu/drm/i915/gt/intel_gt_pm.c | 3 +
drivers/gpu/drm/i915/gt/intel_gt_requests.c | 123 ++++++++++++++++++
drivers/gpu/drm/i915/gt/intel_gt_requests.h | 24 ++++
drivers/gpu/drm/i915/gt/intel_gt_types.h | 11 ++
drivers/gpu/drm/i915/gt/selftest_timeline.c | 8 +-
drivers/gpu/drm/i915/i915_debugfs.c | 17 +--
drivers/gpu/drm/i915/i915_drv.h | 10 --
drivers/gpu/drm/i915/i915_gem.c | 17 ---
drivers/gpu/drm/i915/i915_gem_evict.c | 14 +-
drivers/gpu/drm/i915/i915_gem_gtt.c | 5 +-
drivers/gpu/drm/i915/i915_request.c | 64 +--------
drivers/gpu/drm/i915/i915_request.h | 7 +-
.../gpu/drm/i915/selftests/igt_flush_test.c | 9 +-
.../gpu/drm/i915/selftests/igt_live_test.c | 5 +-
.../gpu/drm/i915/selftests/mock_gem_device.c | 10 +-
21 files changed, 211 insertions(+), 154 deletions(-)
create mode 100644 drivers/gpu/drm/i915/gt/intel_gt_requests.c
create mode 100644 drivers/gpu/drm/i915/gt/intel_gt_requests.h
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index 6313e7b4bd78..cba19470feb5 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -83,6 +83,7 @@ gt-y += \
gt/intel_gt_irq.o \
gt/intel_gt_pm.o \
gt/intel_gt_pm_irq.o \
+ gt/intel_gt_requests.o \
gt/intel_hangcheck.o \
gt/intel_lrc.o \
gt/intel_renderstate.o \
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
index 9cb59fe1a736..96c09d40bb7a 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_mman.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
@@ -8,6 +8,7 @@
#include <linux/sizes.h>
#include "gt/intel_gt.h"
+#include "gt/intel_gt_requests.h"
#include "i915_drv.h"
#include "i915_gem_gtt.h"
@@ -423,6 +424,7 @@ void i915_gem_object_release_mmap(struct drm_i915_gem_object *obj)
static int create_mmap_offset(struct drm_i915_gem_object *obj)
{
struct drm_i915_private *i915 = to_i915(obj->base.dev);
+ struct intel_gt *gt = &i915->gt;
int err;
err = drm_gem_create_mmap_offset(&obj->base);
@@ -431,7 +433,7 @@ static int create_mmap_offset(struct drm_i915_gem_object *obj)
/* Attempt to reap some mmap space from dead objects */
do {
- err = i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
+ err = intel_gt_wait_for_idle(gt, MAX_SCHEDULE_TIMEOUT);
if (err)
break;
@@ -440,7 +442,7 @@ static int create_mmap_offset(struct drm_i915_gem_object *obj)
if (!err)
break;
- } while (flush_delayed_work(&i915->gem.retire_work));
+ } while (flush_delayed_work(>->requests.retire_work));
return err;
}
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
index 04a9a27d659d..1e18045ee982 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
@@ -7,31 +7,18 @@
#include "gem/i915_gem_pm.h"
#include "gt/intel_gt.h"
#include "gt/intel_gt_pm.h"
+#include "gt/intel_gt_requests.h"
#include "i915_drv.h"
#include "i915_globals.h"
static void i915_gem_park(struct drm_i915_private *i915)
{
- cancel_delayed_work(&i915->gem.retire_work);
-
i915_vma_parked(i915);
i915_globals_park();
}
-static void retire_work_handler(struct work_struct *work)
-{
- struct drm_i915_private *i915 =
- container_of(work, typeof(*i915), gem.retire_work.work);
-
- i915_retire_requests(i915);
-
- queue_delayed_work(i915->wq,
- &i915->gem.retire_work,
- round_jiffies_up_relative(HZ));
-}
-
static int pm_notifier(struct notifier_block *nb,
unsigned long action,
void *data)
@@ -42,9 +29,6 @@ static int pm_notifier(struct notifier_block *nb,
switch (action) {
case INTEL_GT_UNPARK:
i915_globals_unpark();
- queue_delayed_work(i915->wq,
- &i915->gem.retire_work,
- round_jiffies_up_relative(HZ));
break;
case INTEL_GT_PARK:
@@ -59,7 +43,7 @@ static bool switch_to_kernel_context_sync(struct intel_gt *gt)
{
bool result = !intel_gt_is_wedged(gt);
- if (i915_gem_wait_for_idle(gt->i915, I915_GEM_IDLE_TIMEOUT) == -ETIME) {
+ if (intel_gt_wait_for_idle(gt, I915_GEM_IDLE_TIMEOUT) == -ETIME) {
/* XXX hide warning from gem_eio */
if (i915_modparams.reset) {
dev_err(gt->i915->drm.dev,
@@ -239,8 +223,6 @@ void i915_gem_resume(struct drm_i915_private *i915)
void i915_gem_init__pm(struct drm_i915_private *i915)
{
- INIT_DELAYED_WORK(&i915->gem.retire_work, retire_work_handler);
-
i915->gem.pm_notifier.notifier_call = pm_notifier;
blocking_notifier_chain_register(&i915->gt.pm_notifications,
&i915->gem.pm_notifier);
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 b26f550f8430..b39224ed3817 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
@@ -8,6 +8,7 @@
#include "gem/i915_gem_pm.h"
#include "gt/intel_gt.h"
+#include "gt/intel_gt_requests.h"
#include "gt/intel_reset.h"
#include "i915_selftest.h"
@@ -304,7 +305,7 @@ create_test_object(struct i915_address_space *vm,
int err;
/* Keep in GEM's good graces */
- i915_retire_requests(vm->i915);
+ intel_gt_retire_requests(vm->gt);
size = min(vm->total / 2, 1024ull * DW_PER_PAGE * PAGE_SIZE);
size = round_down(size, DW_PER_PAGE * PAGE_SIZE);
@@ -922,7 +923,7 @@ __sseu_finish(const char *name,
igt_spinner_end(spin);
if ((flags & TEST_IDLE) && ret == 0) {
- ret = i915_gem_wait_for_idle(ce->engine->i915,
+ ret = intel_gt_wait_for_idle(ce->engine->gt,
MAX_SCHEDULE_TIMEOUT);
if (ret)
return ret;
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 4ba6ed5c8313..1cd25cfd0246 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
@@ -573,7 +573,7 @@ static void disable_retire_worker(struct drm_i915_private *i915)
{
i915_gem_driver_unregister__shrinker(i915);
intel_gt_pm_get(&i915->gt);
- cancel_delayed_work_sync(&i915->gem.retire_work);
+ cancel_delayed_work_sync(&i915->gt.requests.retire_work);
}
static void restore_retire_worker(struct drm_i915_private *i915)
diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c b/drivers/gpu/drm/i915/gt/intel_gt.c
index 95e67db52668..39fb69609be4 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt.c
+++ b/drivers/gpu/drm/i915/gt/intel_gt.c
@@ -6,6 +6,7 @@
#include "i915_drv.h"
#include "intel_gt.h"
#include "intel_gt_pm.h"
+#include "intel_gt_requests.h"
#include "intel_mocs.h"
#include "intel_uncore.h"
#include "intel_pm.h"
@@ -22,6 +23,7 @@ void intel_gt_init_early(struct intel_gt *gt, struct drm_i915_private *i915)
intel_gt_init_hangcheck(gt);
intel_gt_init_reset(gt);
+ intel_gt_init_requests(gt);
intel_gt_pm_init_early(gt);
intel_uc_init_early(>->uc);
}
diff --git a/drivers/gpu/drm/i915/gt/intel_gt_pm.c b/drivers/gpu/drm/i915/gt/intel_gt_pm.c
index 2ccf8cacaa0a..7b12e9dc6c2f 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt_pm.c
+++ b/drivers/gpu/drm/i915/gt/intel_gt_pm.c
@@ -10,6 +10,7 @@
#include "intel_engine_pm.h"
#include "intel_gt.h"
#include "intel_gt_pm.h"
+#include "intel_gt_requests.h"
#include "intel_pm.h"
#include "intel_wakeref.h"
@@ -48,6 +49,7 @@ static int __gt_unpark(struct intel_wakeref *wf)
i915_pmu_gt_unparked(i915);
intel_gt_queue_hangcheck(gt);
+ intel_gt_unpark_requests(gt);
pm_notify(gt, INTEL_GT_UNPARK);
@@ -63,6 +65,7 @@ static int __gt_park(struct intel_wakeref *wf)
GEM_TRACE("\n");
pm_notify(gt, INTEL_GT_PARK);
+ intel_gt_park_requests(gt);
i915_pmu_gt_parked(i915);
if (INTEL_GEN(i915) >= 6)
diff --git a/drivers/gpu/drm/i915/gt/intel_gt_requests.c b/drivers/gpu/drm/i915/gt/intel_gt_requests.c
new file mode 100644
index 000000000000..8aed89fd2cdc
--- /dev/null
+++ b/drivers/gpu/drm/i915/gt/intel_gt_requests.c
@@ -0,0 +1,123 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2019 Intel Corporation
+ */
+
+#include "i915_request.h"
+#include "intel_gt.h"
+#include "intel_gt_pm.h"
+#include "intel_gt_requests.h"
+#include "intel_timeline.h"
+
+static void retire_requests(struct intel_timeline *tl)
+{
+ struct i915_request *rq, *rn;
+
+ list_for_each_entry_safe(rq, rn, &tl->requests, link)
+ if (!i915_request_retire(rq))
+ break;
+}
+
+long intel_gt_retire_requests_timeout(struct intel_gt *gt, long timeout)
+{
+ struct intel_gt_timelines *timelines = >->timelines;
+ struct intel_timeline *tl, *tn;
+ unsigned long active_count = 0;
+ unsigned long flags;
+ bool interruptible;
+ LIST_HEAD(free);
+
+ interruptible = true;
+ if (unlikely(timeout < 0))
+ timeout = -timeout, interruptible = false;
+
+ spin_lock_irqsave(&timelines->lock, flags);
+ list_for_each_entry_safe(tl, tn, &timelines->active_list, link) {
+ if (!mutex_trylock(&tl->mutex))
+ continue;
+
+ intel_timeline_get(tl);
+ GEM_BUG_ON(!tl->active_count);
+ tl->active_count++; /* pin the list element */
+ spin_unlock_irqrestore(&timelines->lock, flags);
+
+ if (timeout > 0) {
+ struct dma_fence *fence;
+
+ fence = i915_active_fence_get(&tl->last_request);
+ if (fence) {
+ timeout = dma_fence_wait_timeout(fence,
+ true,
+ timeout);
+ dma_fence_put(fence);
+ }
+ }
+
+ retire_requests(tl);
+
+ spin_lock_irqsave(&timelines->lock, flags);
+
+ /* Resume iteration after dropping lock */
+ list_safe_reset_next(tl, tn, link);
+ if (--tl->active_count)
+ active_count += !!rcu_access_pointer(tl->last_request.fence);
+ else
+ list_del(&tl->link);
+
+ mutex_unlock(&tl->mutex);
+
+ /* Defer the final release to after the spinlock */
+ if (refcount_dec_and_test(&tl->kref.refcount)) {
+ GEM_BUG_ON(tl->active_count);
+ list_add(&tl->link, &free);
+ }
+ }
+ spin_unlock_irqrestore(&timelines->lock, flags);
+
+ list_for_each_entry_safe(tl, tn, &free, link)
+ __intel_timeline_free(&tl->kref);
+
+ return active_count ? timeout : 0;
+}
+
+int intel_gt_wait_for_idle(struct intel_gt *gt, long timeout)
+{
+ /* If the device is asleep, we have no requests outstanding */
+ if (!intel_gt_pm_is_awake(gt))
+ return 0;
+
+ while ((timeout = intel_gt_retire_requests_timeout(gt, timeout)) > 0) {
+ cond_resched();
+ if (signal_pending(current))
+ return -EINTR;
+ }
+
+ return timeout;
+}
+
+static void retire_work_handler(struct work_struct *work)
+{
+ struct intel_gt *gt =
+ container_of(work, typeof(*gt), requests.retire_work.work);
+
+ intel_gt_retire_requests(gt);
+ schedule_delayed_work(>->requests.retire_work,
+ round_jiffies_up_relative(HZ));
+}
+
+void intel_gt_init_requests(struct intel_gt *gt)
+{
+ INIT_DELAYED_WORK(>->requests.retire_work, retire_work_handler);
+}
+
+void intel_gt_park_requests(struct intel_gt *gt)
+{
+ cancel_delayed_work(>->requests.retire_work);
+}
+
+void intel_gt_unpark_requests(struct intel_gt *gt)
+{
+ schedule_delayed_work(>->requests.retire_work,
+ round_jiffies_up_relative(HZ));
+}
diff --git a/drivers/gpu/drm/i915/gt/intel_gt_requests.h b/drivers/gpu/drm/i915/gt/intel_gt_requests.h
new file mode 100644
index 000000000000..bd31cbce47e0
--- /dev/null
+++ b/drivers/gpu/drm/i915/gt/intel_gt_requests.h
@@ -0,0 +1,24 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2019 Intel Corporation
+ */
+
+#ifndef INTEL_GT_REQUESTS_H
+#define INTEL_GT_REQUESTS_H
+
+struct intel_gt;
+
+long intel_gt_retire_requests_timeout(struct intel_gt *gt, long timeout);
+static inline void intel_gt_retire_requests(struct intel_gt *gt)
+{
+ intel_gt_retire_requests_timeout(gt, 0);
+}
+
+int intel_gt_wait_for_idle(struct intel_gt *gt, long timeout);
+
+void intel_gt_init_requests(struct intel_gt *gt);
+void intel_gt_park_requests(struct intel_gt *gt);
+void intel_gt_unpark_requests(struct intel_gt *gt);
+
+#endif /* INTEL_GT_REQUESTS_H */
diff --git a/drivers/gpu/drm/i915/gt/intel_gt_types.h b/drivers/gpu/drm/i915/gt/intel_gt_types.h
index 3039cef64b11..f524942c82aa 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_gt_types.h
@@ -49,6 +49,17 @@ struct intel_gt {
struct list_head hwsp_free_list;
} timelines;
+ struct intel_gt_requests {
+ /**
+ * We leave the user IRQ off as much as possible,
+ * but this means that requests will finish and never
+ * be retired once the system goes idle. Set a timer to
+ * fire periodically while the ring is running. When it
+ * fires, go retire requests.
+ */
+ struct delayed_work retire_work;
+ } requests;
+
struct intel_wakeref wakeref;
atomic_t user_wakeref;
diff --git a/drivers/gpu/drm/i915/gt/selftest_timeline.c b/drivers/gpu/drm/i915/gt/selftest_timeline.c
index 16abfabf08c7..d6df40cdc8a6 100644
--- a/drivers/gpu/drm/i915/gt/selftest_timeline.c
+++ b/drivers/gpu/drm/i915/gt/selftest_timeline.c
@@ -8,6 +8,7 @@
#include "intel_engine_pm.h"
#include "intel_gt.h"
+#include "intel_gt_requests.h"
#include "../selftests/i915_random.h"
#include "../i915_selftest.h"
@@ -641,6 +642,7 @@ static int live_hwsp_alternate(void *arg)
static int live_hwsp_wrap(void *arg)
{
struct drm_i915_private *i915 = arg;
+ struct intel_gt *gt = &i915->gt;
struct intel_engine_cs *engine;
struct intel_timeline *tl;
enum intel_engine_id id;
@@ -651,7 +653,7 @@ static int live_hwsp_wrap(void *arg)
* foreign GPU references.
*/
- tl = intel_timeline_create(&i915->gt, NULL);
+ tl = intel_timeline_create(gt, NULL);
if (IS_ERR(tl))
return PTR_ERR(tl);
@@ -662,7 +664,7 @@ static int live_hwsp_wrap(void *arg)
if (err)
goto out_free;
- for_each_engine(engine, i915, id) {
+ for_each_engine(engine, gt->i915, id) {
const u32 *hwsp_seqno[2];
struct i915_request *rq;
u32 seqno[2];
@@ -734,7 +736,7 @@ static int live_hwsp_wrap(void *arg)
goto out;
}
- i915_retire_requests(i915); /* recycle HWSP */
+ intel_gt_retire_requests(gt); /* recycle HWSP */
}
out:
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 033764326601..5f9703ca3104 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -41,6 +41,7 @@
#include "gem/i915_gem_context.h"
#include "gt/intel_gt_pm.h"
+#include "gt/intel_gt_requests.h"
#include "gt/intel_reset.h"
#include "gt/uc/intel_guc_submission.h"
@@ -3618,33 +3619,33 @@ static int
i915_drop_caches_set(void *data, u64 val)
{
struct drm_i915_private *i915 = data;
+ struct intel_gt *gt = &i915->gt;
int ret;
DRM_DEBUG("Dropping caches: 0x%08llx [0x%08llx]\n",
val, val & DROP_ALL);
if (val & DROP_RESET_ACTIVE &&
- wait_for(intel_engines_are_idle(&i915->gt),
- I915_IDLE_ENGINES_TIMEOUT))
- intel_gt_set_wedged(&i915->gt);
+ wait_for(intel_engines_are_idle(gt), I915_IDLE_ENGINES_TIMEOUT))
+ intel_gt_set_wedged(gt);
if (val & DROP_RETIRE)
- i915_retire_requests(i915);
+ intel_gt_retire_requests(gt);
if (val & (DROP_IDLE | DROP_ACTIVE)) {
- ret = i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
+ ret = intel_gt_wait_for_idle(gt, MAX_SCHEDULE_TIMEOUT);
if (ret)
return ret;
}
if (val & DROP_IDLE) {
- ret = intel_gt_pm_wait_for_idle(&i915->gt);
+ ret = intel_gt_pm_wait_for_idle(gt);
if (ret)
return ret;
}
- if (val & DROP_RESET_ACTIVE && intel_gt_terminally_wedged(&i915->gt))
- intel_gt_handle_error(&i915->gt, ALL_ENGINES, 0, NULL);
+ if (val & DROP_RESET_ACTIVE && intel_gt_terminally_wedged(gt))
+ intel_gt_handle_error(gt, ALL_ENGINES, 0, NULL);
fs_reclaim_acquire(GFP_KERNEL);
if (val & DROP_BOUND)
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 6e5d5e4ca91d..4e0c4ba2d257 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1717,15 +1717,6 @@ struct drm_i915_private {
struct {
struct notifier_block pm_notifier;
-
- /**
- * We leave the user IRQ off as much as possible,
- * but this means that requests will finish and never
- * be retired once the system goes idle. Set a timer to
- * fire periodically while the ring is running. When it
- * fires, go retire requests.
- */
- struct delayed_work retire_work;
} gem;
/* For i945gm vblank irq vs. C3 workaround */
@@ -2328,7 +2319,6 @@ void i915_gem_driver_register(struct drm_i915_private *i915);
void i915_gem_driver_unregister(struct drm_i915_private *i915);
void i915_gem_driver_remove(struct drm_i915_private *dev_priv);
void i915_gem_driver_release(struct drm_i915_private *dev_priv);
-int i915_gem_wait_for_idle(struct drm_i915_private *dev_priv, long timeout);
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);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index ab23944571fa..9c2877441540 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -883,23 +883,6 @@ void i915_gem_runtime_suspend(struct drm_i915_private *i915)
}
}
-int i915_gem_wait_for_idle(struct drm_i915_private *i915, long timeout)
-{
- struct intel_gt *gt = &i915->gt;
-
- /* If the device is asleep, we have no requests outstanding */
- if (!intel_gt_pm_is_awake(gt))
- return 0;
-
- while ((timeout = i915_retire_requests_timeout(i915, timeout)) > 0) {
- cond_resched();
- if (signal_pending(current))
- return -EINTR;
- }
-
- return timeout;
-}
-
struct i915_vma *
i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj,
const struct i915_ggtt_view *view,
diff --git a/drivers/gpu/drm/i915/i915_gem_evict.c b/drivers/gpu/drm/i915/i915_gem_evict.c
index 0a412f6d01d7..7e62c310290f 100644
--- a/drivers/gpu/drm/i915/i915_gem_evict.c
+++ b/drivers/gpu/drm/i915/i915_gem_evict.c
@@ -29,6 +29,7 @@
#include <drm/i915_drm.h>
#include "gem/i915_gem_context.h"
+#include "gt/intel_gt_requests.h"
#include "i915_drv.h"
#include "i915_trace.h"
@@ -37,7 +38,7 @@ I915_SELFTEST_DECLARE(static struct igt_evict_ctl {
bool fail_if_busy:1;
} igt_evict_ctl;)
-static int ggtt_flush(struct drm_i915_private *i915)
+static int ggtt_flush(struct intel_gt *gt)
{
/*
* Not everything in the GGTT is tracked via vma (otherwise we
@@ -46,7 +47,7 @@ static int ggtt_flush(struct drm_i915_private *i915)
* the hopes that we can then remove contexts and the like only
* bound by their active reference.
*/
- return i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
+ return intel_gt_wait_for_idle(gt, MAX_SCHEDULE_TIMEOUT);
}
static bool
@@ -92,7 +93,6 @@ i915_gem_evict_something(struct i915_address_space *vm,
u64 start, u64 end,
unsigned flags)
{
- struct drm_i915_private *dev_priv = vm->i915;
struct drm_mm_scan scan;
struct list_head eviction_list;
struct i915_vma *vma, *next;
@@ -124,7 +124,7 @@ i915_gem_evict_something(struct i915_address_space *vm,
min_size, alignment, color,
start, end, mode);
- i915_retire_requests(vm->i915);
+ intel_gt_retire_requests(vm->gt);
search_again:
active = NULL;
@@ -197,7 +197,7 @@ i915_gem_evict_something(struct i915_address_space *vm,
if (I915_SELFTEST_ONLY(igt_evict_ctl.fail_if_busy))
return -EBUSY;
- ret = ggtt_flush(dev_priv);
+ ret = ggtt_flush(vm->gt);
if (ret)
return ret;
@@ -270,7 +270,7 @@ int i915_gem_evict_for_node(struct i915_address_space *vm,
* a stray pin (preventing eviction) that can only be resolved by
* retiring.
*/
- i915_retire_requests(vm->i915);
+ intel_gt_retire_requests(vm->gt);
if (i915_vm_has_cache_coloring(vm)) {
/* Expand search to cover neighbouring guard pages (or lack!) */
@@ -372,7 +372,7 @@ int i915_gem_evict_vm(struct i915_address_space *vm)
* switch otherwise is ineffective.
*/
if (i915_is_ggtt(vm)) {
- ret = ggtt_flush(vm->i915);
+ ret = ggtt_flush(vm->gt);
if (ret)
return ret;
}
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 30ee016ac2ba..0d040517787b 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -38,6 +38,7 @@
#include "display/intel_frontbuffer.h"
#include "gt/intel_gt.h"
+#include "gt/intel_gt_requests.h"
#include "i915_drv.h"
#include "i915_scatterlist.h"
@@ -2529,8 +2530,8 @@ void i915_gem_gtt_finish_pages(struct drm_i915_gem_object *obj,
if (unlikely(ggtt->do_idle_maps)) {
/* XXX This does not prevent more requests being submitted! */
- if (i915_retire_requests_timeout(dev_priv,
- -MAX_SCHEDULE_TIMEOUT)) {
+ if (intel_gt_retire_requests_timeout(ggtt->vm.gt,
+ -MAX_SCHEDULE_TIMEOUT)) {
DRM_ERROR("Failed to wait for idle; VT'd may hang.\n");
/* Wait a bit, in hopes it avoids the hang */
udelay(10);
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 52f7c4e5b644..437f9fc6282e 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -216,7 +216,7 @@ static void remove_from_engine(struct i915_request *rq)
spin_unlock(&locked->active.lock);
}
-static bool i915_request_retire(struct i915_request *rq)
+bool i915_request_retire(struct i915_request *rq)
{
if (!i915_request_completed(rq))
return false;
@@ -1508,68 +1508,6 @@ long i915_request_wait(struct i915_request *rq,
return timeout;
}
-long i915_retire_requests_timeout(struct drm_i915_private *i915, long timeout)
-{
- struct intel_gt_timelines *timelines = &i915->gt.timelines;
- struct intel_timeline *tl, *tn;
- unsigned long active_count = 0;
- unsigned long flags;
- bool interruptible;
- LIST_HEAD(free);
-
- interruptible = true;
- if (timeout < 0)
- timeout = -timeout, interruptible = false;
-
- spin_lock_irqsave(&timelines->lock, flags);
- list_for_each_entry_safe(tl, tn, &timelines->active_list, link) {
- if (!mutex_trylock(&tl->mutex))
- continue;
-
- intel_timeline_get(tl);
- GEM_BUG_ON(!tl->active_count);
- tl->active_count++; /* pin the list element */
- spin_unlock_irqrestore(&timelines->lock, flags);
-
- if (timeout > 0) {
- struct dma_fence *fence;
-
- fence = i915_active_fence_get(&tl->last_request);
- if (fence) {
- timeout = dma_fence_wait_timeout(fence,
- interruptible,
- timeout);
- dma_fence_put(fence);
- }
- }
-
- retire_requests(tl);
-
- spin_lock_irqsave(&timelines->lock, flags);
-
- /* Resume iteration after dropping lock */
- list_safe_reset_next(tl, tn, link);
- if (--tl->active_count)
- active_count += !!rcu_access_pointer(tl->last_request.fence);
- else
- list_del(&tl->link);
-
- mutex_unlock(&tl->mutex);
-
- /* Defer the final release to after the spinlock */
- if (refcount_dec_and_test(&tl->kref.refcount)) {
- GEM_BUG_ON(tl->active_count);
- list_add(&tl->link, &free);
- }
- }
- spin_unlock_irqrestore(&timelines->lock, flags);
-
- list_for_each_entry_safe(tl, tn, &free, link)
- __intel_timeline_free(&tl->kref);
-
- return active_count ? timeout : 0;
-}
-
#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
#include "selftests/mock_request.c"
#include "selftests/i915_request.c"
diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
index 256b0715180f..6a95242b280d 100644
--- a/drivers/gpu/drm/i915/i915_request.h
+++ b/drivers/gpu/drm/i915/i915_request.h
@@ -250,6 +250,7 @@ struct i915_request *__i915_request_commit(struct i915_request *request);
void __i915_request_queue(struct i915_request *rq,
const struct i915_sched_attr *attr);
+bool i915_request_retire(struct i915_request *rq);
void i915_request_retire_upto(struct i915_request *rq);
static inline struct i915_request *
@@ -459,10 +460,4 @@ i915_request_active_timeline(struct i915_request *rq)
lockdep_is_held(&rq->engine->active.lock));
}
-long i915_retire_requests_timeout(struct drm_i915_private *i915, long timeout);
-static inline void i915_retire_requests(struct drm_i915_private *i915)
-{
- i915_retire_requests_timeout(i915, 0);
-}
-
#endif /* I915_REQUEST_H */
diff --git a/drivers/gpu/drm/i915/selftests/igt_flush_test.c b/drivers/gpu/drm/i915/selftests/igt_flush_test.c
index ed496bd6d84f..7b0939e3f007 100644
--- a/drivers/gpu/drm/i915/selftests/igt_flush_test.c
+++ b/drivers/gpu/drm/i915/selftests/igt_flush_test.c
@@ -4,8 +4,8 @@
* Copyright © 2018 Intel Corporation
*/
-#include "gem/i915_gem_context.h"
#include "gt/intel_gt.h"
+#include "gt/intel_gt_requests.h"
#include "i915_drv.h"
#include "i915_selftest.h"
@@ -14,11 +14,12 @@
int igt_flush_test(struct drm_i915_private *i915)
{
- int ret = intel_gt_is_wedged(&i915->gt) ? -EIO : 0;
+ struct intel_gt *gt = &i915->gt;
+ int ret = intel_gt_is_wedged(gt) ? -EIO : 0;
cond_resched();
- if (i915_gem_wait_for_idle(i915, HZ / 5) == -ETIME) {
+ if (intel_gt_wait_for_idle(gt, HZ / 5) == -ETIME) {
pr_err("%pS timed out, cancelling all further testing.\n",
__builtin_return_address(0));
@@ -26,7 +27,7 @@ int igt_flush_test(struct drm_i915_private *i915)
__builtin_return_address(0));
GEM_TRACE_DUMP();
- intel_gt_set_wedged(&i915->gt);
+ intel_gt_set_wedged(gt);
ret = -EIO;
}
diff --git a/drivers/gpu/drm/i915/selftests/igt_live_test.c b/drivers/gpu/drm/i915/selftests/igt_live_test.c
index eae90f97df6c..810b60100c2c 100644
--- a/drivers/gpu/drm/i915/selftests/igt_live_test.c
+++ b/drivers/gpu/drm/i915/selftests/igt_live_test.c
@@ -4,7 +4,8 @@
* Copyright © 2018 Intel Corporation
*/
-#include "../i915_drv.h"
+#include "i915_drv.h"
+#include "gt/intel_gt_requests.h"
#include "../i915_selftest.h"
#include "igt_flush_test.h"
@@ -23,7 +24,7 @@ int igt_live_test_begin(struct igt_live_test *t,
t->func = func;
t->name = name;
- err = i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
+ err = intel_gt_wait_for_idle(&i915->gt, MAX_SCHEDULE_TIMEOUT);
if (err) {
pr_err("%s(%s): failed to idle before, with err=%d!",
func, name, err);
diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
index 87d34167bf89..efe40c29fcf2 100644
--- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
+++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
@@ -26,6 +26,7 @@
#include <linux/pm_runtime.h>
#include "gt/intel_gt.h"
+#include "gt/intel_gt_requests.h"
#include "gt/mock_engine.h"
#include "mock_request.h"
@@ -44,7 +45,8 @@ void mock_device_flush(struct drm_i915_private *i915)
do {
for_each_engine(engine, i915, id)
mock_engine_flush(engine);
- } while (i915_retire_requests_timeout(i915, MAX_SCHEDULE_TIMEOUT));
+ } while (intel_gt_retire_requests_timeout(&i915->gt,
+ MAX_SCHEDULE_TIMEOUT));
}
static void mock_device_release(struct drm_device *dev)
@@ -98,10 +100,6 @@ static void release_dev(struct device *dev)
kfree(pdev);
}
-static void mock_retire_work_handler(struct work_struct *work)
-{
-}
-
static int pm_domain_resume(struct device *dev)
{
return pm_generic_runtime_resume(dev);
@@ -180,8 +178,6 @@ struct drm_i915_private *mock_gem_device(void)
mock_init_contexts(i915);
- INIT_DELAYED_WORK(&i915->gem.retire_work, mock_retire_work_handler);
-
i915->gt.awake = -1;
intel_timelines_init(i915);
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* Re: [PATCH 12/27] drm/i915: Move request runtime management onto gt
2019-09-25 10:01 ` [PATCH 12/27] drm/i915: Move request runtime management onto gt Chris Wilson
@ 2019-09-25 10:57 ` Tvrtko Ursulin
2019-09-25 11:17 ` Chris Wilson
0 siblings, 1 reply; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-25 10:57 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 25/09/2019 11:01, Chris Wilson wrote:
> Requests are run from the gt and are tided into the gt runtime power
> management, so pull the runtime request management under gt/
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/Makefile | 1 +
> drivers/gpu/drm/i915/gem/i915_gem_mman.c | 6 +-
> drivers/gpu/drm/i915/gem/i915_gem_pm.c | 22 +---
> .../drm/i915/gem/selftests/i915_gem_context.c | 5 +-
> .../drm/i915/gem/selftests/i915_gem_mman.c | 2 +-
> drivers/gpu/drm/i915/gt/intel_gt.c | 2 +
> drivers/gpu/drm/i915/gt/intel_gt_pm.c | 3 +
> drivers/gpu/drm/i915/gt/intel_gt_requests.c | 123 ++++++++++++++++++
> drivers/gpu/drm/i915/gt/intel_gt_requests.h | 24 ++++
> drivers/gpu/drm/i915/gt/intel_gt_types.h | 11 ++
> drivers/gpu/drm/i915/gt/selftest_timeline.c | 8 +-
> drivers/gpu/drm/i915/i915_debugfs.c | 17 +--
> drivers/gpu/drm/i915/i915_drv.h | 10 --
> drivers/gpu/drm/i915/i915_gem.c | 17 ---
> drivers/gpu/drm/i915/i915_gem_evict.c | 14 +-
> drivers/gpu/drm/i915/i915_gem_gtt.c | 5 +-
> drivers/gpu/drm/i915/i915_request.c | 64 +--------
> drivers/gpu/drm/i915/i915_request.h | 7 +-
> .../gpu/drm/i915/selftests/igt_flush_test.c | 9 +-
> .../gpu/drm/i915/selftests/igt_live_test.c | 5 +-
> .../gpu/drm/i915/selftests/mock_gem_device.c | 10 +-
> 21 files changed, 211 insertions(+), 154 deletions(-)
> create mode 100644 drivers/gpu/drm/i915/gt/intel_gt_requests.c
> create mode 100644 drivers/gpu/drm/i915/gt/intel_gt_requests.h
>
> diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
> index 6313e7b4bd78..cba19470feb5 100644
> --- a/drivers/gpu/drm/i915/Makefile
> +++ b/drivers/gpu/drm/i915/Makefile
> @@ -83,6 +83,7 @@ gt-y += \
> gt/intel_gt_irq.o \
> gt/intel_gt_pm.o \
> gt/intel_gt_pm_irq.o \
> + gt/intel_gt_requests.o \
> gt/intel_hangcheck.o \
> gt/intel_lrc.o \
> gt/intel_renderstate.o \
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
> index 9cb59fe1a736..96c09d40bb7a 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_mman.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
> @@ -8,6 +8,7 @@
> #include <linux/sizes.h>
>
> #include "gt/intel_gt.h"
> +#include "gt/intel_gt_requests.h"
>
> #include "i915_drv.h"
> #include "i915_gem_gtt.h"
> @@ -423,6 +424,7 @@ void i915_gem_object_release_mmap(struct drm_i915_gem_object *obj)
> static int create_mmap_offset(struct drm_i915_gem_object *obj)
> {
> struct drm_i915_private *i915 = to_i915(obj->base.dev);
> + struct intel_gt *gt = &i915->gt;
> int err;
>
> err = drm_gem_create_mmap_offset(&obj->base);
> @@ -431,7 +433,7 @@ static int create_mmap_offset(struct drm_i915_gem_object *obj)
>
> /* Attempt to reap some mmap space from dead objects */
> do {
> - err = i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
> + err = intel_gt_wait_for_idle(gt, MAX_SCHEDULE_TIMEOUT);
It may be useful to keep the i915_gem_wait_for_idle as trivial wrapper
which can then call intel_gt_wait_for_idle. Keeps the separation of GEM
and GT better, but also..
> if (err)
> break;
>
> @@ -440,7 +442,7 @@ static int create_mmap_offset(struct drm_i915_gem_object *obj)
> if (!err)
> break;
>
> - } while (flush_delayed_work(&i915->gem.retire_work));
> + } while (flush_delayed_work(>->requests.retire_work));
.. here could be considered a layering violation that GEM is peeking
into GT internals like this. Shall we have another wrapper like
i915_gem_retire_requests_sync which would call intel_gt_retire_requests
and new intel_gt_flush_retire, or something?
>
> return err;
> }
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
> index 04a9a27d659d..1e18045ee982 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
> @@ -7,31 +7,18 @@
> #include "gem/i915_gem_pm.h"
> #include "gt/intel_gt.h"
> #include "gt/intel_gt_pm.h"
> +#include "gt/intel_gt_requests.h"
>
> #include "i915_drv.h"
> #include "i915_globals.h"
>
> static void i915_gem_park(struct drm_i915_private *i915)
> {
> - cancel_delayed_work(&i915->gem.retire_work);
> -
> i915_vma_parked(i915);
>
> i915_globals_park();
> }
>
> -static void retire_work_handler(struct work_struct *work)
> -{
> - struct drm_i915_private *i915 =
> - container_of(work, typeof(*i915), gem.retire_work.work);
> -
> - i915_retire_requests(i915);
> -
> - queue_delayed_work(i915->wq,
> - &i915->gem.retire_work,
> - round_jiffies_up_relative(HZ));
> -}
> -
> static int pm_notifier(struct notifier_block *nb,
> unsigned long action,
> void *data)
> @@ -42,9 +29,6 @@ static int pm_notifier(struct notifier_block *nb,
> switch (action) {
> case INTEL_GT_UNPARK:
> i915_globals_unpark();
> - queue_delayed_work(i915->wq,
> - &i915->gem.retire_work,
> - round_jiffies_up_relative(HZ));
> break;
>
> case INTEL_GT_PARK:
> @@ -59,7 +43,7 @@ static bool switch_to_kernel_context_sync(struct intel_gt *gt)
> {
> bool result = !intel_gt_is_wedged(gt);
>
> - if (i915_gem_wait_for_idle(gt->i915, I915_GEM_IDLE_TIMEOUT) == -ETIME) {
> + if (intel_gt_wait_for_idle(gt, I915_GEM_IDLE_TIMEOUT) == -ETIME) {
> /* XXX hide warning from gem_eio */
> if (i915_modparams.reset) {
> dev_err(gt->i915->drm.dev,
> @@ -239,8 +223,6 @@ void i915_gem_resume(struct drm_i915_private *i915)
>
> void i915_gem_init__pm(struct drm_i915_private *i915)
> {
> - INIT_DELAYED_WORK(&i915->gem.retire_work, retire_work_handler);
> -
> i915->gem.pm_notifier.notifier_call = pm_notifier;
> blocking_notifier_chain_register(&i915->gt.pm_notifications,
> &i915->gem.pm_notifier);
> 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 b26f550f8430..b39224ed3817 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
> @@ -8,6 +8,7 @@
>
> #include "gem/i915_gem_pm.h"
> #include "gt/intel_gt.h"
> +#include "gt/intel_gt_requests.h"
> #include "gt/intel_reset.h"
> #include "i915_selftest.h"
>
> @@ -304,7 +305,7 @@ create_test_object(struct i915_address_space *vm,
> int err;
>
> /* Keep in GEM's good graces */
> - i915_retire_requests(vm->i915);
> + intel_gt_retire_requests(vm->gt);
>
> size = min(vm->total / 2, 1024ull * DW_PER_PAGE * PAGE_SIZE);
> size = round_down(size, DW_PER_PAGE * PAGE_SIZE);
> @@ -922,7 +923,7 @@ __sseu_finish(const char *name,
> igt_spinner_end(spin);
>
> if ((flags & TEST_IDLE) && ret == 0) {
> - ret = i915_gem_wait_for_idle(ce->engine->i915,
> + ret = intel_gt_wait_for_idle(ce->engine->gt,
> MAX_SCHEDULE_TIMEOUT);
> if (ret)
> return ret;
> 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 4ba6ed5c8313..1cd25cfd0246 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
> @@ -573,7 +573,7 @@ static void disable_retire_worker(struct drm_i915_private *i915)
> {
> i915_gem_driver_unregister__shrinker(i915);
> intel_gt_pm_get(&i915->gt);
> - cancel_delayed_work_sync(&i915->gem.retire_work);
> + cancel_delayed_work_sync(&i915->gt.requests.retire_work);
Same pattern of GEM helper to call a new GT export?
> }
>
> static void restore_retire_worker(struct drm_i915_private *i915)
> diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c b/drivers/gpu/drm/i915/gt/intel_gt.c
> index 95e67db52668..39fb69609be4 100644
> --- a/drivers/gpu/drm/i915/gt/intel_gt.c
> +++ b/drivers/gpu/drm/i915/gt/intel_gt.c
> @@ -6,6 +6,7 @@
> #include "i915_drv.h"
> #include "intel_gt.h"
> #include "intel_gt_pm.h"
> +#include "intel_gt_requests.h"
> #include "intel_mocs.h"
> #include "intel_uncore.h"
> #include "intel_pm.h"
> @@ -22,6 +23,7 @@ void intel_gt_init_early(struct intel_gt *gt, struct drm_i915_private *i915)
>
> intel_gt_init_hangcheck(gt);
> intel_gt_init_reset(gt);
> + intel_gt_init_requests(gt);
> intel_gt_pm_init_early(gt);
> intel_uc_init_early(>->uc);
> }
> diff --git a/drivers/gpu/drm/i915/gt/intel_gt_pm.c b/drivers/gpu/drm/i915/gt/intel_gt_pm.c
> index 2ccf8cacaa0a..7b12e9dc6c2f 100644
> --- a/drivers/gpu/drm/i915/gt/intel_gt_pm.c
> +++ b/drivers/gpu/drm/i915/gt/intel_gt_pm.c
> @@ -10,6 +10,7 @@
> #include "intel_engine_pm.h"
> #include "intel_gt.h"
> #include "intel_gt_pm.h"
> +#include "intel_gt_requests.h"
> #include "intel_pm.h"
> #include "intel_wakeref.h"
>
> @@ -48,6 +49,7 @@ static int __gt_unpark(struct intel_wakeref *wf)
> i915_pmu_gt_unparked(i915);
>
> intel_gt_queue_hangcheck(gt);
> + intel_gt_unpark_requests(gt);
>
> pm_notify(gt, INTEL_GT_UNPARK);
>
> @@ -63,6 +65,7 @@ static int __gt_park(struct intel_wakeref *wf)
> GEM_TRACE("\n");
>
> pm_notify(gt, INTEL_GT_PARK);
> + intel_gt_park_requests(gt);
>
> i915_pmu_gt_parked(i915);
> if (INTEL_GEN(i915) >= 6)
> diff --git a/drivers/gpu/drm/i915/gt/intel_gt_requests.c b/drivers/gpu/drm/i915/gt/intel_gt_requests.c
> new file mode 100644
> index 000000000000..8aed89fd2cdc
> --- /dev/null
> +++ b/drivers/gpu/drm/i915/gt/intel_gt_requests.c
> @@ -0,0 +1,123 @@
> +/*
> + * SPDX-License-Identifier: MIT
> + *
> + * Copyright © 2019 Intel Corporation
> + */
> +
> +#include "i915_request.h"
> +#include "intel_gt.h"
> +#include "intel_gt_pm.h"
> +#include "intel_gt_requests.h"
> +#include "intel_timeline.h"
> +
> +static void retire_requests(struct intel_timeline *tl)
> +{
> + struct i915_request *rq, *rn;
> +
> + list_for_each_entry_safe(rq, rn, &tl->requests, link)
> + if (!i915_request_retire(rq))
> + break;
> +}
> +
> +long intel_gt_retire_requests_timeout(struct intel_gt *gt, long timeout)
> +{
> + struct intel_gt_timelines *timelines = >->timelines;
> + struct intel_timeline *tl, *tn;
> + unsigned long active_count = 0;
> + unsigned long flags;
> + bool interruptible;
> + LIST_HEAD(free);
> +
> + interruptible = true;
> + if (unlikely(timeout < 0))
> + timeout = -timeout, interruptible = false;
> +
> + spin_lock_irqsave(&timelines->lock, flags);
> + list_for_each_entry_safe(tl, tn, &timelines->active_list, link) {
> + if (!mutex_trylock(&tl->mutex))
> + continue;
> +
> + intel_timeline_get(tl);
> + GEM_BUG_ON(!tl->active_count);
> + tl->active_count++; /* pin the list element */
> + spin_unlock_irqrestore(&timelines->lock, flags);
> +
> + if (timeout > 0) {
> + struct dma_fence *fence;
> +
> + fence = i915_active_fence_get(&tl->last_request);
> + if (fence) {
> + timeout = dma_fence_wait_timeout(fence,
> + true,
> + timeout);
> + dma_fence_put(fence);
> + }
> + }
> +
> + retire_requests(tl);
> +
> + spin_lock_irqsave(&timelines->lock, flags);
> +
> + /* Resume iteration after dropping lock */
> + list_safe_reset_next(tl, tn, link);
> + if (--tl->active_count)
> + active_count += !!rcu_access_pointer(tl->last_request.fence);
> + else
> + list_del(&tl->link);
> +
> + mutex_unlock(&tl->mutex);
> +
> + /* Defer the final release to after the spinlock */
> + if (refcount_dec_and_test(&tl->kref.refcount)) {
> + GEM_BUG_ON(tl->active_count);
> + list_add(&tl->link, &free);
> + }
> + }
> + spin_unlock_irqrestore(&timelines->lock, flags);
> +
> + list_for_each_entry_safe(tl, tn, &free, link)
> + __intel_timeline_free(&tl->kref);
> +
> + return active_count ? timeout : 0;
> +}
> +
> +int intel_gt_wait_for_idle(struct intel_gt *gt, long timeout)
> +{
> + /* If the device is asleep, we have no requests outstanding */
> + if (!intel_gt_pm_is_awake(gt))
> + return 0;
> +
> + while ((timeout = intel_gt_retire_requests_timeout(gt, timeout)) > 0) {
> + cond_resched();
> + if (signal_pending(current))
> + return -EINTR;
> + }
> +
> + return timeout;
> +}
> +
> +static void retire_work_handler(struct work_struct *work)
> +{
> + struct intel_gt *gt =
> + container_of(work, typeof(*gt), requests.retire_work.work);
> +
> + intel_gt_retire_requests(gt);
> + schedule_delayed_work(>->requests.retire_work,
> + round_jiffies_up_relative(HZ));
> +}
> +
> +void intel_gt_init_requests(struct intel_gt *gt)
> +{
> + INIT_DELAYED_WORK(>->requests.retire_work, retire_work_handler);
> +}
> +
> +void intel_gt_park_requests(struct intel_gt *gt)
> +{
> + cancel_delayed_work(>->requests.retire_work);
> +}
> +
> +void intel_gt_unpark_requests(struct intel_gt *gt)
> +{
> + schedule_delayed_work(>->requests.retire_work,
> + round_jiffies_up_relative(HZ));
> +}
> diff --git a/drivers/gpu/drm/i915/gt/intel_gt_requests.h b/drivers/gpu/drm/i915/gt/intel_gt_requests.h
> new file mode 100644
> index 000000000000..bd31cbce47e0
> --- /dev/null
> +++ b/drivers/gpu/drm/i915/gt/intel_gt_requests.h
> @@ -0,0 +1,24 @@
> +/*
> + * SPDX-License-Identifier: MIT
> + *
> + * Copyright © 2019 Intel Corporation
> + */
> +
> +#ifndef INTEL_GT_REQUESTS_H
> +#define INTEL_GT_REQUESTS_H
> +
> +struct intel_gt;
> +
> +long intel_gt_retire_requests_timeout(struct intel_gt *gt, long timeout);
> +static inline void intel_gt_retire_requests(struct intel_gt *gt)
> +{
> + intel_gt_retire_requests_timeout(gt, 0);
> +}
> +
> +int intel_gt_wait_for_idle(struct intel_gt *gt, long timeout);
> +
> +void intel_gt_init_requests(struct intel_gt *gt);
> +void intel_gt_park_requests(struct intel_gt *gt);
> +void intel_gt_unpark_requests(struct intel_gt *gt);
> +
> +#endif /* INTEL_GT_REQUESTS_H */
> diff --git a/drivers/gpu/drm/i915/gt/intel_gt_types.h b/drivers/gpu/drm/i915/gt/intel_gt_types.h
> index 3039cef64b11..f524942c82aa 100644
> --- a/drivers/gpu/drm/i915/gt/intel_gt_types.h
> +++ b/drivers/gpu/drm/i915/gt/intel_gt_types.h
> @@ -49,6 +49,17 @@ struct intel_gt {
> struct list_head hwsp_free_list;
> } timelines;
>
> + struct intel_gt_requests {
> + /**
> + * We leave the user IRQ off as much as possible,
> + * but this means that requests will finish and never
> + * be retired once the system goes idle. Set a timer to
> + * fire periodically while the ring is running. When it
> + * fires, go retire requests.
> + */
> + struct delayed_work retire_work;
> + } requests;
> +
> struct intel_wakeref wakeref;
> atomic_t user_wakeref;
>
> diff --git a/drivers/gpu/drm/i915/gt/selftest_timeline.c b/drivers/gpu/drm/i915/gt/selftest_timeline.c
> index 16abfabf08c7..d6df40cdc8a6 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_timeline.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_timeline.c
> @@ -8,6 +8,7 @@
>
> #include "intel_engine_pm.h"
> #include "intel_gt.h"
> +#include "intel_gt_requests.h"
>
> #include "../selftests/i915_random.h"
> #include "../i915_selftest.h"
> @@ -641,6 +642,7 @@ static int live_hwsp_alternate(void *arg)
> static int live_hwsp_wrap(void *arg)
> {
> struct drm_i915_private *i915 = arg;
> + struct intel_gt *gt = &i915->gt;
> struct intel_engine_cs *engine;
> struct intel_timeline *tl;
> enum intel_engine_id id;
> @@ -651,7 +653,7 @@ static int live_hwsp_wrap(void *arg)
> * foreign GPU references.
> */
>
> - tl = intel_timeline_create(&i915->gt, NULL);
> + tl = intel_timeline_create(gt, NULL);
> if (IS_ERR(tl))
> return PTR_ERR(tl);
>
> @@ -662,7 +664,7 @@ static int live_hwsp_wrap(void *arg)
> if (err)
> goto out_free;
>
> - for_each_engine(engine, i915, id) {
> + for_each_engine(engine, gt->i915, id) {
> const u32 *hwsp_seqno[2];
> struct i915_request *rq;
> u32 seqno[2];
> @@ -734,7 +736,7 @@ static int live_hwsp_wrap(void *arg)
> goto out;
> }
>
> - i915_retire_requests(i915); /* recycle HWSP */
> + intel_gt_retire_requests(gt); /* recycle HWSP */
> }
>
> out:
> diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
> index 033764326601..5f9703ca3104 100644
> --- a/drivers/gpu/drm/i915/i915_debugfs.c
> +++ b/drivers/gpu/drm/i915/i915_debugfs.c
> @@ -41,6 +41,7 @@
>
> #include "gem/i915_gem_context.h"
> #include "gt/intel_gt_pm.h"
> +#include "gt/intel_gt_requests.h"
> #include "gt/intel_reset.h"
> #include "gt/uc/intel_guc_submission.h"
>
> @@ -3618,33 +3619,33 @@ static int
> i915_drop_caches_set(void *data, u64 val)
> {
> struct drm_i915_private *i915 = data;
> + struct intel_gt *gt = &i915->gt;
> int ret;
>
> DRM_DEBUG("Dropping caches: 0x%08llx [0x%08llx]\n",
> val, val & DROP_ALL);
>
> if (val & DROP_RESET_ACTIVE &&
> - wait_for(intel_engines_are_idle(&i915->gt),
> - I915_IDLE_ENGINES_TIMEOUT))
> - intel_gt_set_wedged(&i915->gt);
> + wait_for(intel_engines_are_idle(gt), I915_IDLE_ENGINES_TIMEOUT))
> + intel_gt_set_wedged(gt);
>
> if (val & DROP_RETIRE)
> - i915_retire_requests(i915);
> + intel_gt_retire_requests(gt);
>
> if (val & (DROP_IDLE | DROP_ACTIVE)) {
> - ret = i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
> + ret = intel_gt_wait_for_idle(gt, MAX_SCHEDULE_TIMEOUT);
> if (ret)
> return ret;
> }
>
> if (val & DROP_IDLE) {
> - ret = intel_gt_pm_wait_for_idle(&i915->gt);
> + ret = intel_gt_pm_wait_for_idle(gt);
> if (ret)
> return ret;
> }
>
> - if (val & DROP_RESET_ACTIVE && intel_gt_terminally_wedged(&i915->gt))
> - intel_gt_handle_error(&i915->gt, ALL_ENGINES, 0, NULL);
> + if (val & DROP_RESET_ACTIVE && intel_gt_terminally_wedged(gt))
> + intel_gt_handle_error(gt, ALL_ENGINES, 0, NULL);
>
> fs_reclaim_acquire(GFP_KERNEL);
> if (val & DROP_BOUND)
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index 6e5d5e4ca91d..4e0c4ba2d257 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -1717,15 +1717,6 @@ struct drm_i915_private {
>
> struct {
> struct notifier_block pm_notifier;
> -
> - /**
> - * We leave the user IRQ off as much as possible,
> - * but this means that requests will finish and never
> - * be retired once the system goes idle. Set a timer to
> - * fire periodically while the ring is running. When it
> - * fires, go retire requests.
> - */
> - struct delayed_work retire_work;
> } gem;
>
> /* For i945gm vblank irq vs. C3 workaround */
> @@ -2328,7 +2319,6 @@ void i915_gem_driver_register(struct drm_i915_private *i915);
> void i915_gem_driver_unregister(struct drm_i915_private *i915);
> void i915_gem_driver_remove(struct drm_i915_private *dev_priv);
> void i915_gem_driver_release(struct drm_i915_private *dev_priv);
> -int i915_gem_wait_for_idle(struct drm_i915_private *dev_priv, long timeout);
> 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);
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index ab23944571fa..9c2877441540 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -883,23 +883,6 @@ void i915_gem_runtime_suspend(struct drm_i915_private *i915)
> }
> }
>
> -int i915_gem_wait_for_idle(struct drm_i915_private *i915, long timeout)
> -{
> - struct intel_gt *gt = &i915->gt;
> -
> - /* If the device is asleep, we have no requests outstanding */
> - if (!intel_gt_pm_is_awake(gt))
> - return 0;
> -
> - while ((timeout = i915_retire_requests_timeout(i915, timeout)) > 0) {
> - cond_resched();
> - if (signal_pending(current))
> - return -EINTR;
> - }
> -
> - return timeout;
> -}
> -
> struct i915_vma *
> i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj,
> const struct i915_ggtt_view *view,
> diff --git a/drivers/gpu/drm/i915/i915_gem_evict.c b/drivers/gpu/drm/i915/i915_gem_evict.c
> index 0a412f6d01d7..7e62c310290f 100644
> --- a/drivers/gpu/drm/i915/i915_gem_evict.c
> +++ b/drivers/gpu/drm/i915/i915_gem_evict.c
> @@ -29,6 +29,7 @@
> #include <drm/i915_drm.h>
>
> #include "gem/i915_gem_context.h"
> +#include "gt/intel_gt_requests.h"
>
> #include "i915_drv.h"
> #include "i915_trace.h"
> @@ -37,7 +38,7 @@ I915_SELFTEST_DECLARE(static struct igt_evict_ctl {
> bool fail_if_busy:1;
> } igt_evict_ctl;)
>
> -static int ggtt_flush(struct drm_i915_private *i915)
> +static int ggtt_flush(struct intel_gt *gt)
> {
> /*
> * Not everything in the GGTT is tracked via vma (otherwise we
> @@ -46,7 +47,7 @@ static int ggtt_flush(struct drm_i915_private *i915)
> * the hopes that we can then remove contexts and the like only
> * bound by their active reference.
> */
> - return i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
> + return intel_gt_wait_for_idle(gt, MAX_SCHEDULE_TIMEOUT);
> }
>
> static bool
> @@ -92,7 +93,6 @@ i915_gem_evict_something(struct i915_address_space *vm,
> u64 start, u64 end,
> unsigned flags)
> {
> - struct drm_i915_private *dev_priv = vm->i915;
> struct drm_mm_scan scan;
> struct list_head eviction_list;
> struct i915_vma *vma, *next;
> @@ -124,7 +124,7 @@ i915_gem_evict_something(struct i915_address_space *vm,
> min_size, alignment, color,
> start, end, mode);
>
> - i915_retire_requests(vm->i915);
> + intel_gt_retire_requests(vm->gt);
>
> search_again:
> active = NULL;
> @@ -197,7 +197,7 @@ i915_gem_evict_something(struct i915_address_space *vm,
> if (I915_SELFTEST_ONLY(igt_evict_ctl.fail_if_busy))
> return -EBUSY;
>
> - ret = ggtt_flush(dev_priv);
> + ret = ggtt_flush(vm->gt);
> if (ret)
> return ret;
>
> @@ -270,7 +270,7 @@ int i915_gem_evict_for_node(struct i915_address_space *vm,
> * a stray pin (preventing eviction) that can only be resolved by
> * retiring.
> */
> - i915_retire_requests(vm->i915);
> + intel_gt_retire_requests(vm->gt);
>
> if (i915_vm_has_cache_coloring(vm)) {
> /* Expand search to cover neighbouring guard pages (or lack!) */
> @@ -372,7 +372,7 @@ int i915_gem_evict_vm(struct i915_address_space *vm)
> * switch otherwise is ineffective.
> */
> if (i915_is_ggtt(vm)) {
> - ret = ggtt_flush(vm->i915);
> + ret = ggtt_flush(vm->gt);
> if (ret)
> return ret;
> }
> diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
> index 30ee016ac2ba..0d040517787b 100644
> --- a/drivers/gpu/drm/i915/i915_gem_gtt.c
> +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
> @@ -38,6 +38,7 @@
>
> #include "display/intel_frontbuffer.h"
> #include "gt/intel_gt.h"
> +#include "gt/intel_gt_requests.h"
>
> #include "i915_drv.h"
> #include "i915_scatterlist.h"
> @@ -2529,8 +2530,8 @@ void i915_gem_gtt_finish_pages(struct drm_i915_gem_object *obj,
>
> if (unlikely(ggtt->do_idle_maps)) {
> /* XXX This does not prevent more requests being submitted! */
> - if (i915_retire_requests_timeout(dev_priv,
> - -MAX_SCHEDULE_TIMEOUT)) {
> + if (intel_gt_retire_requests_timeout(ggtt->vm.gt,
> + -MAX_SCHEDULE_TIMEOUT)) {
> DRM_ERROR("Failed to wait for idle; VT'd may hang.\n");
> /* Wait a bit, in hopes it avoids the hang */
> udelay(10);
> diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
> index 52f7c4e5b644..437f9fc6282e 100644
> --- a/drivers/gpu/drm/i915/i915_request.c
> +++ b/drivers/gpu/drm/i915/i915_request.c
> @@ -216,7 +216,7 @@ static void remove_from_engine(struct i915_request *rq)
> spin_unlock(&locked->active.lock);
> }
>
> -static bool i915_request_retire(struct i915_request *rq)
> +bool i915_request_retire(struct i915_request *rq)
> {
> if (!i915_request_completed(rq))
> return false;
> @@ -1508,68 +1508,6 @@ long i915_request_wait(struct i915_request *rq,
> return timeout;
> }
>
> -long i915_retire_requests_timeout(struct drm_i915_private *i915, long timeout)
> -{
> - struct intel_gt_timelines *timelines = &i915->gt.timelines;
> - struct intel_timeline *tl, *tn;
> - unsigned long active_count = 0;
> - unsigned long flags;
> - bool interruptible;
> - LIST_HEAD(free);
> -
> - interruptible = true;
> - if (timeout < 0)
> - timeout = -timeout, interruptible = false;
> -
> - spin_lock_irqsave(&timelines->lock, flags);
> - list_for_each_entry_safe(tl, tn, &timelines->active_list, link) {
> - if (!mutex_trylock(&tl->mutex))
> - continue;
> -
> - intel_timeline_get(tl);
> - GEM_BUG_ON(!tl->active_count);
> - tl->active_count++; /* pin the list element */
> - spin_unlock_irqrestore(&timelines->lock, flags);
> -
> - if (timeout > 0) {
> - struct dma_fence *fence;
> -
> - fence = i915_active_fence_get(&tl->last_request);
> - if (fence) {
> - timeout = dma_fence_wait_timeout(fence,
> - interruptible,
> - timeout);
> - dma_fence_put(fence);
> - }
> - }
> -
> - retire_requests(tl);
> -
> - spin_lock_irqsave(&timelines->lock, flags);
> -
> - /* Resume iteration after dropping lock */
> - list_safe_reset_next(tl, tn, link);
> - if (--tl->active_count)
> - active_count += !!rcu_access_pointer(tl->last_request.fence);
> - else
> - list_del(&tl->link);
> -
> - mutex_unlock(&tl->mutex);
> -
> - /* Defer the final release to after the spinlock */
> - if (refcount_dec_and_test(&tl->kref.refcount)) {
> - GEM_BUG_ON(tl->active_count);
> - list_add(&tl->link, &free);
> - }
> - }
> - spin_unlock_irqrestore(&timelines->lock, flags);
> -
> - list_for_each_entry_safe(tl, tn, &free, link)
> - __intel_timeline_free(&tl->kref);
> -
> - return active_count ? timeout : 0;
> -}
> -
> #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
> #include "selftests/mock_request.c"
> #include "selftests/i915_request.c"
> diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
> index 256b0715180f..6a95242b280d 100644
> --- a/drivers/gpu/drm/i915/i915_request.h
> +++ b/drivers/gpu/drm/i915/i915_request.h
> @@ -250,6 +250,7 @@ struct i915_request *__i915_request_commit(struct i915_request *request);
> void __i915_request_queue(struct i915_request *rq,
> const struct i915_sched_attr *attr);
>
> +bool i915_request_retire(struct i915_request *rq);
> void i915_request_retire_upto(struct i915_request *rq);
>
> static inline struct i915_request *
> @@ -459,10 +460,4 @@ i915_request_active_timeline(struct i915_request *rq)
> lockdep_is_held(&rq->engine->active.lock));
> }
>
> -long i915_retire_requests_timeout(struct drm_i915_private *i915, long timeout);
> -static inline void i915_retire_requests(struct drm_i915_private *i915)
> -{
> - i915_retire_requests_timeout(i915, 0);
> -}
> -
> #endif /* I915_REQUEST_H */
> diff --git a/drivers/gpu/drm/i915/selftests/igt_flush_test.c b/drivers/gpu/drm/i915/selftests/igt_flush_test.c
> index ed496bd6d84f..7b0939e3f007 100644
> --- a/drivers/gpu/drm/i915/selftests/igt_flush_test.c
> +++ b/drivers/gpu/drm/i915/selftests/igt_flush_test.c
> @@ -4,8 +4,8 @@
> * Copyright © 2018 Intel Corporation
> */
>
> -#include "gem/i915_gem_context.h"
> #include "gt/intel_gt.h"
> +#include "gt/intel_gt_requests.h"
>
> #include "i915_drv.h"
> #include "i915_selftest.h"
> @@ -14,11 +14,12 @@
>
> int igt_flush_test(struct drm_i915_private *i915)
> {
> - int ret = intel_gt_is_wedged(&i915->gt) ? -EIO : 0;
> + struct intel_gt *gt = &i915->gt;
> + int ret = intel_gt_is_wedged(gt) ? -EIO : 0;
>
> cond_resched();
>
> - if (i915_gem_wait_for_idle(i915, HZ / 5) == -ETIME) {
> + if (intel_gt_wait_for_idle(gt, HZ / 5) == -ETIME) {
> pr_err("%pS timed out, cancelling all further testing.\n",
> __builtin_return_address(0));
>
> @@ -26,7 +27,7 @@ int igt_flush_test(struct drm_i915_private *i915)
> __builtin_return_address(0));
> GEM_TRACE_DUMP();
>
> - intel_gt_set_wedged(&i915->gt);
> + intel_gt_set_wedged(gt);
> ret = -EIO;
> }
>
> diff --git a/drivers/gpu/drm/i915/selftests/igt_live_test.c b/drivers/gpu/drm/i915/selftests/igt_live_test.c
> index eae90f97df6c..810b60100c2c 100644
> --- a/drivers/gpu/drm/i915/selftests/igt_live_test.c
> +++ b/drivers/gpu/drm/i915/selftests/igt_live_test.c
> @@ -4,7 +4,8 @@
> * Copyright © 2018 Intel Corporation
> */
>
> -#include "../i915_drv.h"
> +#include "i915_drv.h"
> +#include "gt/intel_gt_requests.h"
>
> #include "../i915_selftest.h"
> #include "igt_flush_test.h"
> @@ -23,7 +24,7 @@ int igt_live_test_begin(struct igt_live_test *t,
> t->func = func;
> t->name = name;
>
> - err = i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
> + err = intel_gt_wait_for_idle(&i915->gt, MAX_SCHEDULE_TIMEOUT);
> if (err) {
> pr_err("%s(%s): failed to idle before, with err=%d!",
> func, name, err);
> diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
> index 87d34167bf89..efe40c29fcf2 100644
> --- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
> +++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
> @@ -26,6 +26,7 @@
> #include <linux/pm_runtime.h>
>
> #include "gt/intel_gt.h"
> +#include "gt/intel_gt_requests.h"
> #include "gt/mock_engine.h"
>
> #include "mock_request.h"
> @@ -44,7 +45,8 @@ void mock_device_flush(struct drm_i915_private *i915)
> do {
> for_each_engine(engine, i915, id)
> mock_engine_flush(engine);
> - } while (i915_retire_requests_timeout(i915, MAX_SCHEDULE_TIMEOUT));
> + } while (intel_gt_retire_requests_timeout(&i915->gt,
> + MAX_SCHEDULE_TIMEOUT));
> }
>
> static void mock_device_release(struct drm_device *dev)
> @@ -98,10 +100,6 @@ static void release_dev(struct device *dev)
> kfree(pdev);
> }
>
> -static void mock_retire_work_handler(struct work_struct *work)
> -{
> -}
> -
> static int pm_domain_resume(struct device *dev)
> {
> return pm_generic_runtime_resume(dev);
> @@ -180,8 +178,6 @@ struct drm_i915_private *mock_gem_device(void)
>
> mock_init_contexts(i915);
>
> - INIT_DELAYED_WORK(&i915->gem.retire_work, mock_retire_work_handler);
> -
> i915->gt.awake = -1;
>
> intel_timelines_init(i915);
>
I am still slightly uneasy about having requests, which I see as a GEM
concept, be retired from GT, but okay, it's not the most important issue
at the moment.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* Re: [PATCH 12/27] drm/i915: Move request runtime management onto gt
2019-09-25 10:57 ` Tvrtko Ursulin
@ 2019-09-25 11:17 ` Chris Wilson
2019-09-25 11:24 ` Chris Wilson
0 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 11:17 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Tvrtko Ursulin (2019-09-25 11:57:53)
>
> On 25/09/2019 11:01, Chris Wilson wrote:
> > @@ -423,6 +424,7 @@ void i915_gem_object_release_mmap(struct drm_i915_gem_object *obj)
> > static int create_mmap_offset(struct drm_i915_gem_object *obj)
> > {
> > struct drm_i915_private *i915 = to_i915(obj->base.dev);
> > + struct intel_gt *gt = &i915->gt;
> > int err;
> >
> > err = drm_gem_create_mmap_offset(&obj->base);
> > @@ -431,7 +433,7 @@ static int create_mmap_offset(struct drm_i915_gem_object *obj)
> >
> > /* Attempt to reap some mmap space from dead objects */
> > do {
> > - err = i915_gem_wait_for_idle(i915, MAX_SCHEDULE_TIMEOUT);
> > + err = intel_gt_wait_for_idle(gt, MAX_SCHEDULE_TIMEOUT);
>
> It may be useful to keep the i915_gem_wait_for_idle as trivial wrapper
> which can then call intel_gt_wait_for_idle. Keeps the separation of GEM
> and GT better, but also..
> > if (err)
> > break;
> >
> > @@ -440,7 +442,7 @@ static int create_mmap_offset(struct drm_i915_gem_object *obj)
> > if (!err)
> > break;
> >
> > - } while (flush_delayed_work(&i915->gem.retire_work));
> > + } while (flush_delayed_work(>->requests.retire_work));
>
> .. here could be considered a layering violation that GEM is peeking
> into GT internals like this. Shall we have another wrapper like
> i915_gem_retire_requests_sync which would call intel_gt_retire_requests
> and new intel_gt_flush_retire, or something?
This is the most complicated case because we're managing a global
resource. Actually this loop can be eliminated, as it was only trying to
avoid taking struct_mutex here (using the worker instead). Now that
we can call retire directly, it can be reduced.
After which I think the point is moot as the path forward is clearer.
> > @@ -180,8 +178,6 @@ struct drm_i915_private *mock_gem_device(void)
> >
> > mock_init_contexts(i915);
> >
> > - INIT_DELAYED_WORK(&i915->gem.retire_work, mock_retire_work_handler);
> > -
> > i915->gt.awake = -1;
> >
> > intel_timelines_init(i915);
> >
>
> I am still slightly uneasy about having requests, which I see as a GEM
> concept, be retired from GT, but okay, it's not the most important issue
> at the moment.
requests are not a GEM concept, they belong in sched/ :)
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* Re: [PATCH 12/27] drm/i915: Move request runtime management onto gt
2019-09-25 11:17 ` Chris Wilson
@ 2019-09-25 11:24 ` Chris Wilson
2019-09-25 11:29 ` Chris Wilson
0 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 11:24 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Chris Wilson (2019-09-25 12:17:14)
> Quoting Tvrtko Ursulin (2019-09-25 11:57:53)
> >
> > I am still slightly uneasy about having requests, which I see as a GEM
> > concept, be retired from GT, but okay, it's not the most important issue
> > at the moment.
>
> requests are not a GEM concept, they belong in sched/ :)
Overall though, the compartmentalisation is intel_gt based. Our
timelines and their powermanagement is associated with the gt, so that's
where it feels necessary to provide the background cleanup. If we drop
the timeline->gt, then I think it's time for a new home?
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* Re: [PATCH 12/27] drm/i915: Move request runtime management onto gt
2019-09-25 11:24 ` Chris Wilson
@ 2019-09-25 11:29 ` Chris Wilson
2019-09-25 11:33 ` Chris Wilson
0 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 11:29 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Chris Wilson (2019-09-25 12:24:08)
> Quoting Chris Wilson (2019-09-25 12:17:14)
> > Quoting Tvrtko Ursulin (2019-09-25 11:57:53)
> > >
> > > I am still slightly uneasy about having requests, which I see as a GEM
> > > concept, be retired from GT, but okay, it's not the most important issue
> > > at the moment.
> >
> > requests are not a GEM concept, they belong in sched/ :)
>
> Overall though, the compartmentalisation is intel_gt based. Our
> timelines and their powermanagement is associated with the gt, so that's
> where it feels necessary to provide the background cleanup. If we drop
> the timeline->gt, then I think it's time for a new home?
Actually would a simple s//intel_gt_retire_timelines()/ make it more
palatable?
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* Re: [PATCH 12/27] drm/i915: Move request runtime management onto gt
2019-09-25 11:29 ` Chris Wilson
@ 2019-09-25 11:33 ` Chris Wilson
2019-09-25 15:17 ` Tvrtko Ursulin
0 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 11:33 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Chris Wilson (2019-09-25 12:29:40)
> Quoting Chris Wilson (2019-09-25 12:24:08)
> > Quoting Chris Wilson (2019-09-25 12:17:14)
> > > Quoting Tvrtko Ursulin (2019-09-25 11:57:53)
> > > >
> > > > I am still slightly uneasy about having requests, which I see as a GEM
> > > > concept, be retired from GT, but okay, it's not the most important issue
> > > > at the moment.
> > >
> > > requests are not a GEM concept, they belong in sched/ :)
> >
> > Overall though, the compartmentalisation is intel_gt based. Our
> > timelines and their powermanagement is associated with the gt, so that's
> > where it feels necessary to provide the background cleanup. If we drop
> > the timeline->gt, then I think it's time for a new home?
>
> Actually would a simple s//intel_gt_retire_timelines()/ make it more
> palatable?
Or even just intel_gt_retire()
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* Re: [PATCH 12/27] drm/i915: Move request runtime management onto gt
2019-09-25 11:33 ` Chris Wilson
@ 2019-09-25 15:17 ` Tvrtko Ursulin
0 siblings, 0 replies; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-25 15:17 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 25/09/2019 12:33, Chris Wilson wrote:
> Quoting Chris Wilson (2019-09-25 12:29:40)
>> Quoting Chris Wilson (2019-09-25 12:24:08)
>>> Quoting Chris Wilson (2019-09-25 12:17:14)
>>>> Quoting Tvrtko Ursulin (2019-09-25 11:57:53)
>>>>>
>>>>> I am still slightly uneasy about having requests, which I see as a GEM
>>>>> concept, be retired from GT, but okay, it's not the most important issue
>>>>> at the moment.
>>>>
>>>> requests are not a GEM concept, they belong in sched/ :)
>>>
>>> Overall though, the compartmentalisation is intel_gt based. Our
>>> timelines and their powermanagement is associated with the gt, so that's
>>> where it feels necessary to provide the background cleanup. If we drop
>>> the timeline->gt, then I think it's time for a new home?
>>
>> Actually would a simple s//intel_gt_retire_timelines()/ make it more
>> palatable?
>
> Or even just intel_gt_retire()
This bit is fine as it is. As we have both pointed out there is natural
and unavoidable intertwining between GEM and GT and that's completely
natural. As long as we have clear APIs between the two (so those helpers
I was suggesting) I am happy.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* [PATCH 13/27] drm/i915: Move global activity tracking from GEM to GT
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (11 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 12/27] drm/i915: Move request runtime management onto gt Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:01 ` [PATCH 14/27] drm/i915: Expose engine properties via sysfs Chris Wilson
` (17 subsequent siblings)
30 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
As our global unpark/park keep track of the number of active users, we
can simply move the accounting from the GEM layer to the base GT layer.
It was placed originally inside GEM to benefit from the 100ms extra
delay on idleness, but that has been eliminated and now there is no
substantive difference between the layers. In moving it, we move another
piece of the puzzle out from underneath struct_mutex.
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/gem/i915_gem_pm.c | 11 +----------
drivers/gpu/drm/i915/gt/intel_gt_pm.c | 5 +++++
2 files changed, 6 insertions(+), 10 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
index 1e18045ee982..9b52aca70eee 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
@@ -10,14 +10,6 @@
#include "gt/intel_gt_requests.h"
#include "i915_drv.h"
-#include "i915_globals.h"
-
-static void i915_gem_park(struct drm_i915_private *i915)
-{
- i915_vma_parked(i915);
-
- i915_globals_park();
-}
static int pm_notifier(struct notifier_block *nb,
unsigned long action,
@@ -28,11 +20,10 @@ static int pm_notifier(struct notifier_block *nb,
switch (action) {
case INTEL_GT_UNPARK:
- i915_globals_unpark();
break;
case INTEL_GT_PARK:
- i915_gem_park(i915);
+ i915_vma_parked(i915);
break;
}
diff --git a/drivers/gpu/drm/i915/gt/intel_gt_pm.c b/drivers/gpu/drm/i915/gt/intel_gt_pm.c
index 7b12e9dc6c2f..660857a665e7 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt_pm.c
+++ b/drivers/gpu/drm/i915/gt/intel_gt_pm.c
@@ -5,6 +5,7 @@
*/
#include "i915_drv.h"
+#include "i915_globals.h"
#include "i915_params.h"
#include "intel_context.h"
#include "intel_engine_pm.h"
@@ -26,6 +27,8 @@ static int __gt_unpark(struct intel_wakeref *wf)
GEM_TRACE("\n");
+ i915_globals_unpark();
+
/*
* It seems that the DMC likes to transition between the DC states a lot
* when there are no connected displays (no active power domains) during
@@ -77,6 +80,8 @@ static int __gt_park(struct intel_wakeref *wf)
GEM_BUG_ON(!wakeref);
intel_display_power_put(i915, POWER_DOMAIN_GT_IRQ, wakeref);
+ i915_globals_park();
+
return 0;
}
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* [PATCH 14/27] drm/i915: Expose engine properties via sysfs
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (12 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 13/27] drm/i915: Move global activity tracking from GEM to GT Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-27 20:48 ` Rodrigo Vivi
2019-09-25 10:01 ` [PATCH 15/27] drm/i915/execlists: Force preemption Chris Wilson
` (16 subsequent siblings)
30 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
Preliminary stub to add engines underneath /sys/class/drm/cardN/, so
that we can expose properties on each engine to the sysadmin.
To start with we have basic analogues of the i915_query ioctl so that we
can pretty print engine discovery from the shell, and flesh out the
directory structure. Later we will add writeable sysadmin properties such
as per-engine timeout controls.
An example tree of the engine properties on Braswell:
/sys/class/drm/card0
└── engine
├── bcs0
│ ├── class
│ ├── heartbeat_interval_ms
│ ├── instance
│ ├── mmio_base
│ └── name
├── rcs0
│ ├── class
│ ├── heartbeat_interval_ms
│ ├── instance
│ ├── mmio_base
│ └── name
├── vcs0
│ ├── class
│ ├── heartbeat_interval_ms
│ ├── instance
│ ├── mmio_base
│ └── name
└── vecs0
├── class
├── heartbeat_interval_ms
├── instance
├── mmio_base
└── name
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
---
drivers/gpu/drm/i915/Makefile | 3 +-
drivers/gpu/drm/i915/gt/intel_engine_sysfs.c | 119 +++++++++++++++++++
drivers/gpu/drm/i915/gt/intel_engine_sysfs.h | 14 +++
drivers/gpu/drm/i915/i915_sysfs.c | 4 +
4 files changed, 139 insertions(+), 1 deletion(-)
create mode 100644 drivers/gpu/drm/i915/gt/intel_engine_sysfs.c
create mode 100644 drivers/gpu/drm/i915/gt/intel_engine_sysfs.h
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index cba19470feb5..ba98a2067433 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -76,8 +76,9 @@ gt-y += \
gt/intel_breadcrumbs.o \
gt/intel_context.o \
gt/intel_engine_cs.o \
- gt/intel_engine_pool.o \
gt/intel_engine_pm.o \
+ gt/intel_engine_pool.o \
+ gt/intel_engine_sysfs.o \
gt/intel_engine_user.o \
gt/intel_gt.o \
gt/intel_gt_irq.o \
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_sysfs.c b/drivers/gpu/drm/i915/gt/intel_engine_sysfs.c
new file mode 100644
index 000000000000..cbe9ec59beeb
--- /dev/null
+++ b/drivers/gpu/drm/i915/gt/intel_engine_sysfs.c
@@ -0,0 +1,119 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2019 Intel Corporation
+ */
+
+#include <linux/kobject.h>
+#include <linux/sysfs.h>
+
+#include "i915_drv.h"
+#include "intel_engine.h"
+#include "intel_engine_sysfs.h"
+
+struct kobj_engine {
+ struct kobject base;
+ struct intel_engine_cs *engine;
+};
+
+static struct intel_engine_cs *kobj_to_engine(struct kobject *kobj)
+{
+ return container_of(kobj, struct kobj_engine, base)->engine;
+}
+
+static ssize_t
+name_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
+{
+ return sprintf(buf, "%s\n", kobj_to_engine(kobj)->name);
+}
+
+static ssize_t
+class_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
+{
+ return sprintf(buf, "%d\n", kobj_to_engine(kobj)->uabi_class);
+}
+
+static ssize_t
+inst_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
+{
+ return sprintf(buf, "%d\n", kobj_to_engine(kobj)->uabi_instance);
+}
+
+static ssize_t
+mmio_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
+{
+ return sprintf(buf, "0x%x\n", kobj_to_engine(kobj)->mmio_base);
+}
+
+static struct kobj_attribute name_attr = __ATTR(name, 0444, name_show, NULL);
+static struct kobj_attribute class_attr = __ATTR(class, 0444, class_show, NULL);
+static struct kobj_attribute inst_attr = __ATTR(instance, 0444, inst_show, NULL);
+static struct kobj_attribute mmio_attr = __ATTR(mmio_base, 0444, mmio_show, NULL);
+
+static void kobj_engine_release(struct kobject *kobj)
+{
+ kfree(kobj);
+}
+
+static struct kobj_type kobj_engine_type = {
+ .release = kobj_engine_release,
+ .sysfs_ops = &kobj_sysfs_ops
+};
+
+static struct kobject *
+kobj_engine(struct kobject *dir, struct intel_engine_cs *engine)
+{
+ struct kobj_engine *ke;
+
+ ke = kzalloc(sizeof(*ke), GFP_KERNEL);
+ if (!ke)
+ return NULL;
+
+ kobject_init(&ke->base, &kobj_engine_type);
+ ke->engine = engine;
+
+ if (kobject_add(&ke->base, dir, "%s", engine->name)) {
+ kobject_put(&ke->base);
+ return NULL;
+ }
+
+ /* xfer ownership to sysfs tree */
+ return &ke->base;
+}
+
+void intel_engines_add_sysfs(struct drm_i915_private *i915)
+{
+ static const struct attribute *files[] = {
+ &name_attr.attr,
+ &class_attr.attr,
+ &inst_attr.attr,
+ &mmio_attr.attr,
+ NULL
+ };
+
+ struct device *kdev = i915->drm.primary->kdev;
+ struct intel_engine_cs *engine;
+ struct kobject *dir;
+
+ dir = kobject_create_and_add("engine", &kdev->kobj);
+ if (!dir)
+ return;
+
+ for_each_uabi_engine(engine, i915) {
+ struct kobject *kobj;
+
+ kobj = kobj_engine(dir, engine);
+ if (!kobj)
+ goto err_engine;
+
+ if (sysfs_create_files(kobj, files))
+ goto err_engine;
+
+ if (0) {
+err_engine:
+ dev_err(kdev, "Failed to add sysfs engine '%s'\n",
+ engine->name);
+ break;
+ }
+ }
+}
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_sysfs.h b/drivers/gpu/drm/i915/gt/intel_engine_sysfs.h
new file mode 100644
index 000000000000..ef44a745b70a
--- /dev/null
+++ b/drivers/gpu/drm/i915/gt/intel_engine_sysfs.h
@@ -0,0 +1,14 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2019 Intel Corporation
+ */
+
+#ifndef INTEL_ENGINE_SYSFS_H
+#define INTEL_ENGINE_SYSFS_H
+
+struct drm_i915_private;
+
+void intel_engines_add_sysfs(struct drm_i915_private *i915);
+
+#endif /* INTEL_ENGINE_SYSFS_H */
diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c
index d8a3b180c084..6b88d934927a 100644
--- a/drivers/gpu/drm/i915/i915_sysfs.c
+++ b/drivers/gpu/drm/i915/i915_sysfs.c
@@ -30,6 +30,8 @@
#include <linux/stat.h>
#include <linux/sysfs.h>
+#include "gt/intel_engine_sysfs.h"
+
#include "i915_drv.h"
#include "i915_sysfs.h"
#include "intel_pm.h"
@@ -618,6 +620,8 @@ void i915_setup_sysfs(struct drm_i915_private *dev_priv)
DRM_ERROR("RPS sysfs setup failed\n");
i915_setup_error_capture(kdev);
+
+ intel_engines_add_sysfs(dev_priv);
}
void i915_teardown_sysfs(struct drm_i915_private *dev_priv)
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* Re: [PATCH 14/27] drm/i915: Expose engine properties via sysfs
2019-09-25 10:01 ` [PATCH 14/27] drm/i915: Expose engine properties via sysfs Chris Wilson
@ 2019-09-27 20:48 ` Rodrigo Vivi
0 siblings, 0 replies; 70+ messages in thread
From: Rodrigo Vivi @ 2019-09-27 20:48 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
On Wed, Sep 25, 2019 at 11:01:24AM +0100, Chris Wilson wrote:
> Preliminary stub to add engines underneath /sys/class/drm/cardN/, so
> that we can expose properties on each engine to the sysadmin.
>
> To start with we have basic analogues of the i915_query ioctl so that we
> can pretty print engine discovery from the shell, and flesh out the
> directory structure. Later we will add writeable sysadmin properties such
> as per-engine timeout controls.
>
> An example tree of the engine properties on Braswell:
> /sys/class/drm/card0
> └── engine
> ├── bcs0
> │ ├── class
> │ ├── heartbeat_interval_ms
> │ ├── instance
> │ ├── mmio_base
> │ └── name
> ├── rcs0
> │ ├── class
> │ ├── heartbeat_interval_ms
> │ ├── instance
> │ ├── mmio_base
> │ └── name
> ├── vcs0
> │ ├── class
> │ ├── heartbeat_interval_ms
> │ ├── instance
> │ ├── mmio_base
> │ └── name
> └── vecs0
> ├── class
> ├── heartbeat_interval_ms
> ├── instance
> ├── mmio_base
> └── name
>
Acked-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
> ---
> drivers/gpu/drm/i915/Makefile | 3 +-
> drivers/gpu/drm/i915/gt/intel_engine_sysfs.c | 119 +++++++++++++++++++
> drivers/gpu/drm/i915/gt/intel_engine_sysfs.h | 14 +++
> drivers/gpu/drm/i915/i915_sysfs.c | 4 +
> 4 files changed, 139 insertions(+), 1 deletion(-)
> create mode 100644 drivers/gpu/drm/i915/gt/intel_engine_sysfs.c
> create mode 100644 drivers/gpu/drm/i915/gt/intel_engine_sysfs.h
>
> diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
> index cba19470feb5..ba98a2067433 100644
> --- a/drivers/gpu/drm/i915/Makefile
> +++ b/drivers/gpu/drm/i915/Makefile
> @@ -76,8 +76,9 @@ gt-y += \
> gt/intel_breadcrumbs.o \
> gt/intel_context.o \
> gt/intel_engine_cs.o \
> - gt/intel_engine_pool.o \
> gt/intel_engine_pm.o \
> + gt/intel_engine_pool.o \
> + gt/intel_engine_sysfs.o \
> gt/intel_engine_user.o \
> gt/intel_gt.o \
> gt/intel_gt_irq.o \
> diff --git a/drivers/gpu/drm/i915/gt/intel_engine_sysfs.c b/drivers/gpu/drm/i915/gt/intel_engine_sysfs.c
> new file mode 100644
> index 000000000000..cbe9ec59beeb
> --- /dev/null
> +++ b/drivers/gpu/drm/i915/gt/intel_engine_sysfs.c
> @@ -0,0 +1,119 @@
> +/*
> + * SPDX-License-Identifier: MIT
> + *
> + * Copyright © 2019 Intel Corporation
> + */
> +
> +#include <linux/kobject.h>
> +#include <linux/sysfs.h>
> +
> +#include "i915_drv.h"
> +#include "intel_engine.h"
> +#include "intel_engine_sysfs.h"
> +
> +struct kobj_engine {
> + struct kobject base;
> + struct intel_engine_cs *engine;
> +};
> +
> +static struct intel_engine_cs *kobj_to_engine(struct kobject *kobj)
> +{
> + return container_of(kobj, struct kobj_engine, base)->engine;
> +}
> +
> +static ssize_t
> +name_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
> +{
> + return sprintf(buf, "%s\n", kobj_to_engine(kobj)->name);
> +}
> +
> +static ssize_t
> +class_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
> +{
> + return sprintf(buf, "%d\n", kobj_to_engine(kobj)->uabi_class);
> +}
> +
> +static ssize_t
> +inst_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
> +{
> + return sprintf(buf, "%d\n", kobj_to_engine(kobj)->uabi_instance);
> +}
> +
> +static ssize_t
> +mmio_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
> +{
> + return sprintf(buf, "0x%x\n", kobj_to_engine(kobj)->mmio_base);
> +}
> +
> +static struct kobj_attribute name_attr = __ATTR(name, 0444, name_show, NULL);
> +static struct kobj_attribute class_attr = __ATTR(class, 0444, class_show, NULL);
> +static struct kobj_attribute inst_attr = __ATTR(instance, 0444, inst_show, NULL);
> +static struct kobj_attribute mmio_attr = __ATTR(mmio_base, 0444, mmio_show, NULL);
> +
> +static void kobj_engine_release(struct kobject *kobj)
> +{
> + kfree(kobj);
> +}
> +
> +static struct kobj_type kobj_engine_type = {
> + .release = kobj_engine_release,
> + .sysfs_ops = &kobj_sysfs_ops
> +};
> +
> +static struct kobject *
> +kobj_engine(struct kobject *dir, struct intel_engine_cs *engine)
> +{
> + struct kobj_engine *ke;
> +
> + ke = kzalloc(sizeof(*ke), GFP_KERNEL);
> + if (!ke)
> + return NULL;
> +
> + kobject_init(&ke->base, &kobj_engine_type);
> + ke->engine = engine;
> +
> + if (kobject_add(&ke->base, dir, "%s", engine->name)) {
> + kobject_put(&ke->base);
> + return NULL;
> + }
> +
> + /* xfer ownership to sysfs tree */
> + return &ke->base;
> +}
> +
> +void intel_engines_add_sysfs(struct drm_i915_private *i915)
> +{
> + static const struct attribute *files[] = {
> + &name_attr.attr,
> + &class_attr.attr,
> + &inst_attr.attr,
> + &mmio_attr.attr,
> + NULL
> + };
> +
> + struct device *kdev = i915->drm.primary->kdev;
> + struct intel_engine_cs *engine;
> + struct kobject *dir;
> +
> + dir = kobject_create_and_add("engine", &kdev->kobj);
> + if (!dir)
> + return;
> +
> + for_each_uabi_engine(engine, i915) {
> + struct kobject *kobj;
> +
> + kobj = kobj_engine(dir, engine);
> + if (!kobj)
> + goto err_engine;
> +
> + if (sysfs_create_files(kobj, files))
> + goto err_engine;
> +
> + if (0) {
> +err_engine:
> + dev_err(kdev, "Failed to add sysfs engine '%s'\n",
> + engine->name);
> + break;
> + }
> + }
> +}
> diff --git a/drivers/gpu/drm/i915/gt/intel_engine_sysfs.h b/drivers/gpu/drm/i915/gt/intel_engine_sysfs.h
> new file mode 100644
> index 000000000000..ef44a745b70a
> --- /dev/null
> +++ b/drivers/gpu/drm/i915/gt/intel_engine_sysfs.h
> @@ -0,0 +1,14 @@
> +/*
> + * SPDX-License-Identifier: MIT
> + *
> + * Copyright © 2019 Intel Corporation
> + */
> +
> +#ifndef INTEL_ENGINE_SYSFS_H
> +#define INTEL_ENGINE_SYSFS_H
> +
> +struct drm_i915_private;
> +
> +void intel_engines_add_sysfs(struct drm_i915_private *i915);
> +
> +#endif /* INTEL_ENGINE_SYSFS_H */
> diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c
> index d8a3b180c084..6b88d934927a 100644
> --- a/drivers/gpu/drm/i915/i915_sysfs.c
> +++ b/drivers/gpu/drm/i915/i915_sysfs.c
> @@ -30,6 +30,8 @@
> #include <linux/stat.h>
> #include <linux/sysfs.h>
>
> +#include "gt/intel_engine_sysfs.h"
> +
> #include "i915_drv.h"
> #include "i915_sysfs.h"
> #include "intel_pm.h"
> @@ -618,6 +620,8 @@ void i915_setup_sysfs(struct drm_i915_private *dev_priv)
> DRM_ERROR("RPS sysfs setup failed\n");
>
> i915_setup_error_capture(kdev);
> +
> + intel_engines_add_sysfs(dev_priv);
> }
>
> void i915_teardown_sysfs(struct drm_i915_private *dev_priv)
> --
> 2.23.0
>
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* [PATCH 15/27] drm/i915/execlists: Force preemption
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (13 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 14/27] drm/i915: Expose engine properties via sysfs Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:01 ` [PATCH 16/27] drm/i915: Mark up "sentinel" requests Chris Wilson
` (15 subsequent siblings)
30 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
If the preempted context takes too long to relinquish control, e.g. it
is stuck inside a shader with arbitration disabled, evict that context
with an engine reset. This ensures that preemptions are reasonably
responsive, providing a tighter QoS for the more important context at
the cost of flagging unresponsive contexts more frequently (i.e. instead
of using an ~10s hangcheck, we now evict at ~100ms). The challenge of
lies in picking a timeout that can be reasonably serviced by HW for
typical workloads, balancing the existing clients against the needs for
responsiveness.
Note that coupled with timeslicing, this will lead to rapid GPU "hang"
detection with multiple active contexts vying for GPU time.
v2: Couple in sysfs control of preemption timeout
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
---
drivers/gpu/drm/i915/Kconfig.profile | 12 ++++
drivers/gpu/drm/i915/gt/intel_engine_cs.c | 2 +
drivers/gpu/drm/i915/gt/intel_engine_sysfs.c | 31 +++++++++++
drivers/gpu/drm/i915/gt/intel_engine_types.h | 4 ++
drivers/gpu/drm/i915/gt/intel_lrc.c | 58 ++++++++++++++++++++
drivers/gpu/drm/i915/i915_params.h | 2 +-
6 files changed, 108 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/i915/Kconfig.profile b/drivers/gpu/drm/i915/Kconfig.profile
index 48df8889a88a..3184e8491333 100644
--- a/drivers/gpu/drm/i915/Kconfig.profile
+++ b/drivers/gpu/drm/i915/Kconfig.profile
@@ -25,3 +25,15 @@ config DRM_I915_SPIN_REQUEST
May be 0 to disable the initial spin. In practice, we estimate
the cost of enabling the interrupt (if currently disabled) to be
a few microseconds.
+
+config DRM_I915_PREEMPT_TIMEOUT
+ int "Preempt timeout (ms)"
+ default 100 # milliseconds
+ help
+ How long to wait (in milliseconds) for a preemption event to occur
+ when submitting a new context via execlists. If the current context
+ does not hit an arbitration point and yield to HW before the timer
+ expires, the HW will be reset to allow the more important context
+ to execute.
+
+ May be 0 to disable the timeout.
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
index f451d5076bde..12855947bc9b 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
@@ -304,6 +304,8 @@ static int intel_engine_setup(struct intel_gt *gt, enum intel_engine_id id)
engine->instance = info->instance;
__sprint_engine_name(engine);
+ engine->props.preempt_timeout = CONFIG_DRM_I915_PREEMPT_TIMEOUT;
+
/*
* To be overridden by the backend on setup. However to facilitate
* cleanup on error during setup, we always provide the destroy vfunc.
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_sysfs.c b/drivers/gpu/drm/i915/gt/intel_engine_sysfs.c
index cbe9ec59beeb..5aa3b67bd6cd 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_sysfs.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_sysfs.c
@@ -45,10 +45,36 @@ mmio_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
return sprintf(buf, "0x%x\n", kobj_to_engine(kobj)->mmio_base);
}
+static ssize_t
+preempt_timeout_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
+{
+ struct intel_engine_cs *engine = kobj_to_engine(kobj);
+
+ return sprintf(buf, "%lu\n", engine->props.preempt_timeout);
+}
+
+static ssize_t
+preempt_timeout_store(struct kobject *kobj, struct kobj_attribute *attr,
+ const char *buf, size_t count)
+{
+ struct intel_engine_cs *engine = kobj_to_engine(kobj);
+ unsigned long timeout;
+ int err;
+
+ err = kstrtoul(buf, 0, &timeout);
+ if (err)
+ return err;
+
+ engine->props.preempt_timeout = timeout;
+ return count;
+}
+
static struct kobj_attribute name_attr = __ATTR(name, 0444, name_show, NULL);
static struct kobj_attribute class_attr = __ATTR(class, 0444, class_show, NULL);
static struct kobj_attribute inst_attr = __ATTR(instance, 0444, inst_show, NULL);
static struct kobj_attribute mmio_attr = __ATTR(mmio_base, 0444, mmio_show, NULL);
+static struct kobj_attribute preempt_timeout_attr =
+__ATTR(preempt_timeout_ms, 0600, preempt_timeout_show, preempt_timeout_store);
static void kobj_engine_release(struct kobject *kobj)
{
@@ -109,6 +135,11 @@ void intel_engines_add_sysfs(struct drm_i915_private *i915)
if (sysfs_create_files(kobj, files))
goto err_engine;
+ if (CONFIG_DRM_I915_PREEMPT_TIMEOUT &&
+ intel_engine_has_preemption(engine) &&
+ sysfs_create_file(kobj, &preempt_timeout_attr.attr))
+ goto err_engine;
+
if (0) {
err_engine:
dev_err(kdev, "Failed to add sysfs engine '%s'\n",
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
index 943f0663837e..b144704d2f21 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
@@ -539,6 +539,10 @@ struct intel_engine_cs {
*/
ktime_t total;
} stats;
+
+ struct {
+ unsigned long preempt_timeout;
+ } props;
};
static inline bool
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index caac091b3eb9..3f79970aeadd 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -1384,6 +1384,29 @@ static void record_preemption(struct intel_engine_execlists *execlists)
(void)I915_SELFTEST_ONLY(execlists->preempt_hang.count++);
}
+static void set_preempt_timeout(struct intel_engine_cs *engine)
+{
+ unsigned long timeout;
+
+ if (!CONFIG_DRM_I915_PREEMPT_TIMEOUT)
+ return;
+
+ timeout = READ_ONCE(engine->props.preempt_timeout);
+ if (!timeout)
+ return;
+
+ timeout = msecs_to_jiffies_timeout(timeout);
+ /*
+ * Paranoia to make sure the compiler computes the timeout before
+ * loading 'jiffies' as jiffies is volatile and may be updated in
+ * the background by a timer tick. All to reduce the complexity
+ * of the addition and reduce the risk of losing a jiffie.
+ */
+ barrier();
+
+ mod_timer(&engine->execlists.timer, jiffies + timeout);
+}
+
static void execlists_dequeue(struct intel_engine_cs *engine)
{
struct intel_engine_execlists * const execlists = &engine->execlists;
@@ -1733,6 +1756,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
execlists->switch_priority_hint =
switch_prio(engine, *execlists->pending);
execlists_submit_ports(engine);
+ set_preempt_timeout(engine);
} else {
ring_set_paused(engine, 0);
}
@@ -1962,6 +1986,37 @@ static void __execlists_submission_tasklet(struct intel_engine_cs *const engine)
execlists_dequeue(engine);
}
+static noinline void preempt_reset(struct intel_engine_cs *engine)
+{
+ const unsigned int bit = I915_RESET_ENGINE + engine->id;
+ unsigned long *lock = &engine->gt->reset.flags;
+
+ if (i915_modparams.reset < 3)
+ return;
+
+ if (test_and_set_bit(bit, lock))
+ return;
+
+ /* Mark this tasklet as disabled to avoid waiting for it to complete */
+ tasklet_disable_nosync(&engine->execlists.tasklet);
+
+ intel_engine_reset(engine, "preemption time out");
+
+ tasklet_enable(&engine->execlists.tasklet);
+ clear_and_wake_up_bit(bit, lock);
+}
+
+static bool preempt_timeout(struct intel_engine_cs *const engine)
+{
+ if (!CONFIG_DRM_I915_PREEMPT_TIMEOUT)
+ return false;
+
+ if (!intel_engine_has_preemption(engine))
+ return false;
+
+ return !timer_pending(&engine->execlists.timer);
+}
+
/*
* Check the unread Context Status Buffers and manage the submission of new
* contexts to the ELSP accordingly.
@@ -1976,6 +2031,9 @@ static void execlists_submission_tasklet(unsigned long data)
spin_lock_irqsave(&engine->active.lock, flags);
__execlists_submission_tasklet(engine);
spin_unlock_irqrestore(&engine->active.lock, flags);
+ } else {
+ if (preempt_timeout(engine))
+ preempt_reset(engine);
}
}
diff --git a/drivers/gpu/drm/i915/i915_params.h b/drivers/gpu/drm/i915/i915_params.h
index d29ade3b7de6..56058978bb27 100644
--- a/drivers/gpu/drm/i915/i915_params.h
+++ b/drivers/gpu/drm/i915/i915_params.h
@@ -61,7 +61,7 @@ struct drm_printer;
param(char *, dmc_firmware_path, NULL) \
param(int, mmio_debug, -IS_ENABLED(CONFIG_DRM_I915_DEBUG_MMIO)) \
param(int, edp_vswing, 0) \
- param(int, reset, 2) \
+ param(int, reset, 3) \
param(unsigned int, inject_load_failure, 0) \
param(int, fastboot, -1) \
param(int, enable_dpcd_backlight, 0) \
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* [PATCH 16/27] drm/i915: Mark up "sentinel" requests
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (14 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 15/27] drm/i915/execlists: Force preemption Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:01 ` [PATCH 17/27] drm/i915/execlists: Cancel banned contexts on schedule-out Chris Wilson
` (14 subsequent siblings)
30 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
Sometimes we want to emit a terminator request, a request that flushes
the pipeline and allows no request to come after it. This can be used
for a "preempt-to-idle" to ensure that upon processing the
context-switch to that request, all other active contexts have been
flushed.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/gt/intel_lrc.c | 6 ++++++
drivers/gpu/drm/i915/i915_request.h | 10 ++++++++--
2 files changed, 14 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 3f79970aeadd..0518828129aa 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -1223,6 +1223,9 @@ static bool can_merge_rq(const struct i915_request *prev,
if (i915_request_completed(next))
return true;
+ if (i915_request_has_sentinel(prev))
+ return false;
+
if (!can_merge_ctx(prev->hw_context, next->hw_context))
return false;
@@ -1694,6 +1697,9 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
if (last->hw_context == rq->hw_context)
goto done;
+ if (i915_request_has_sentinel(last))
+ goto done;
+
/*
* If GVT overrides us we only ever submit
* port[0], leaving port[1] empty. Note that we
diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
index 6a95242b280d..96991d64759c 100644
--- a/drivers/gpu/drm/i915/i915_request.h
+++ b/drivers/gpu/drm/i915/i915_request.h
@@ -216,8 +216,9 @@ struct i915_request {
unsigned long emitted_jiffies;
unsigned long flags;
-#define I915_REQUEST_WAITBOOST BIT(0)
-#define I915_REQUEST_NOPREEMPT BIT(1)
+#define I915_REQUEST_WAITBOOST BIT(0)
+#define I915_REQUEST_NOPREEMPT BIT(1)
+#define I915_REQUEST_SENTINEL BIT(2)
/** timeline->request entry for this request */
struct list_head link;
@@ -440,6 +441,11 @@ static inline bool i915_request_has_nopreempt(const struct i915_request *rq)
return unlikely(rq->flags & I915_REQUEST_NOPREEMPT);
}
+static inline bool i915_request_has_sentinel(const struct i915_request *rq)
+{
+ return unlikely(rq->flags & I915_REQUEST_SENTINEL);
+}
+
static inline struct intel_timeline *
i915_request_timeline(struct i915_request *rq)
{
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* [PATCH 17/27] drm/i915/execlists: Cancel banned contexts on schedule-out
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (15 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 16/27] drm/i915: Mark up "sentinel" requests Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:01 ` [PATCH 18/27] drm/i915: Cancel non-persistent contexts on close Chris Wilson
` (13 subsequent siblings)
30 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
On completion of a banned context, scrub the context image so that we do
not replay the active payload. The intent is that we skip banned
payloads on request submission so that the timeline advancement
continues on in the background. However, if we are returning to a
preempted request, i915_request_skip() is ineffective and instead we
need to patch up the context image so that it continues from the start
of the next request.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/gt/intel_lrc.c | 78 +++++++++++++++++++++++++++--
1 file changed, 73 insertions(+), 5 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 0518828129aa..0ea2f4d7dc43 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -234,6 +234,9 @@ static void execlists_init_reg_state(u32 *reg_state,
const struct intel_engine_cs *engine,
const struct intel_ring *ring,
bool close);
+static void
+__execlists_update_reg_state(const struct intel_context *ce,
+ const struct intel_engine_cs *engine);
static void mark_eio(struct i915_request *rq)
{
@@ -1012,6 +1015,69 @@ static void kick_siblings(struct i915_request *rq, struct intel_context *ce)
tasklet_schedule(&ve->base.execlists.tasklet);
}
+static void
+mark_complete(struct i915_request *rq, struct intel_engine_cs *engine)
+{
+ const struct intel_timeline * const tl = rcu_dereference(rq->timeline);
+
+ *(u32 *)tl->hwsp_seqno = rq->fence.seqno;
+ GEM_BUG_ON(!i915_request_completed(rq));
+
+ list_for_each_entry_from_reverse(rq, &tl->requests, link) {
+ if (i915_request_signaled(rq))
+ break;
+
+ mark_eio(rq);
+ }
+
+ intel_engine_queue_breadcrumbs(engine);
+}
+
+static void __context_pin_acquire(struct intel_context *ce)
+{
+ GEM_BUG_ON(!intel_context_is_pinned(ce));
+ mutex_acquire(&ce->pin_mutex.dep_map, 2, 0, _RET_IP_);
+}
+
+static void __context_pin_release(struct intel_context *ce)
+{
+ mutex_release(&ce->pin_mutex.dep_map, 0, _RET_IP_);
+}
+
+static void cancel_active(struct i915_request *rq,
+ struct intel_engine_cs *engine)
+{
+ struct intel_context * const ce = rq->hw_context;
+ u32 *regs = ce->lrc_reg_state;
+
+ if (i915_request_completed(rq))
+ return;
+
+ GEM_TRACE("%s(%s): { rq=%llx:%lld }\n",
+ __func__, engine->name, rq->fence.context, rq->fence.seqno);
+ __context_pin_acquire(ce);
+
+ /* Scrub the context image to prevent replaying the previous batch */
+ memcpy(regs, /* skip restoring the vanilla PPHWSP */
+ engine->pinned_default_state + LRC_STATE_PN * PAGE_SIZE,
+ engine->context_size - PAGE_SIZE);
+ execlists_init_reg_state(regs, ce, engine, ce->ring, false);
+
+ /* Ring will be advanced on retire; here we need to reset the context */
+ ce->ring->head = intel_ring_wrap(ce->ring, rq->wa_tail);
+ __execlists_update_reg_state(ce, engine);
+
+ /* We've switched away, so this should be a no-op, but intent matters */
+ ce->lrc_desc |= CTX_DESC_FORCE_RESTORE;
+
+ /* Let everyone know that the request may now be retired */
+ rcu_read_lock();
+ mark_complete(rq, engine);
+ rcu_read_unlock();
+
+ __context_pin_release(ce);
+}
+
static inline void
__execlists_schedule_out(struct i915_request *rq,
struct intel_engine_cs * const engine)
@@ -1022,6 +1088,9 @@ __execlists_schedule_out(struct i915_request *rq,
execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_OUT);
intel_gt_pm_put(engine->gt);
+ if (unlikely(i915_gem_context_is_banned(ce->gem_context)))
+ cancel_active(rq, engine);
+
/*
* If this is part of a virtual engine, its next request may
* have been blocked waiting for access to the active context.
@@ -2833,7 +2902,7 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
GEM_BUG_ON(!i915_vma_is_pinned(ce->state));
/* Proclaim we have exclusive access to the context image! */
- mutex_acquire(&ce->pin_mutex.dep_map, 2, 0, _THIS_IP_);
+ __context_pin_acquire(ce);
rq = active_request(rq);
if (!rq) {
@@ -2897,7 +2966,7 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled)
__execlists_reset_reg_state(ce, engine);
__execlists_update_reg_state(ce, engine);
ce->lrc_desc |= CTX_DESC_FORCE_RESTORE; /* paranoid: GPU was reset! */
- mutex_release(&ce->pin_mutex.dep_map, 0, _THIS_IP_);
+ __context_pin_release(ce);
unwind:
/* Push back any incomplete requests for replay after the reset. */
@@ -4522,8 +4591,7 @@ void intel_lr_context_reset(struct intel_engine_cs *engine,
u32 head,
bool scrub)
{
- GEM_BUG_ON(!intel_context_is_pinned(ce));
- mutex_acquire(&ce->pin_mutex.dep_map, 2, 0, _THIS_IP_);
+ __context_pin_acquire(ce);
/*
* We want a simple context + ring to execute the breadcrumb update.
@@ -4549,7 +4617,7 @@ void intel_lr_context_reset(struct intel_engine_cs *engine,
intel_ring_update_space(ce->ring);
__execlists_update_reg_state(ce, engine);
- mutex_release(&ce->pin_mutex.dep_map, 0, _THIS_IP_);
+ __context_pin_release(ce);
}
#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* [PATCH 18/27] drm/i915: Cancel non-persistent contexts on close
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (16 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 17/27] drm/i915/execlists: Cancel banned contexts on schedule-out Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 10:01 ` [PATCH 19/27] drm/i915: Replace hangcheck by heartbeats Chris Wilson
` (12 subsequent siblings)
30 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
Normally, we rely on our hangcheck to prevent persistent batches from
hogging the GPU. However, if the user disables hangcheck, this mechanism
breaks down. Despite our insistence that this is unsafe, the users are
equally insistent that they want to use endless batches and will disable
the hangcheck mechanism. We are looking at perhaps replacing hangcheck
with a softer mechanism, that sends a pulse down the engine to check if
it is well. We can use the same preemptive pulse to flush an active
persistent context off the GPU upon context close, preventing resources
being lost and unkillable requests remaining on the GPU after process
termination. To avoid changing the ABI and accidentally breaking
existing userspace, we make the persistence of a context explicit and
enable it by default (matching current ABI). Userspace can opt out of
persistent mode (forcing requests to be cancelled when the context is
closed by process termination or explicitly) by a context parameter. To
facilitate existing use-cases of disabling hangcheck, if the modparam is
disabled (i915.enable_hangcheck=0), we disable persistence mode by
default. (Note, one of the outcomes for supporting endless mode will be
the removal of hangchecking, at which point opting into persistent mode
will be mandatory, or maybe the default perhaps controlled by cgroups.)
Testcase: igt/gem_ctx_persistence
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Cc: Jon Bloomfield <jon.bloomfield@intel.com>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
---
drivers/gpu/drm/i915/Makefile | 1 +
drivers/gpu/drm/i915/gem/i915_gem_context.c | 122 ++++++++++++++++++
drivers/gpu/drm/i915/gem/i915_gem_context.h | 15 +++
.../gpu/drm/i915/gem/i915_gem_context_types.h | 1 +
.../gpu/drm/i915/gem/selftests/mock_context.c | 2 +
.../gpu/drm/i915/gt/intel_engine_heartbeat.c | 56 ++++++++
.../gpu/drm/i915/gt/intel_engine_heartbeat.h | 14 ++
drivers/gpu/drm/i915/gt/intel_engine_pm.c | 2 +-
drivers/gpu/drm/i915/i915_priolist_types.h | 1 +
include/uapi/drm/i915_drm.h | 15 +++
10 files changed, 228 insertions(+), 1 deletion(-)
create mode 100644 drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c
create mode 100644 drivers/gpu/drm/i915/gt/intel_engine_heartbeat.h
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index ba98a2067433..82088f748409 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -76,6 +76,7 @@ gt-y += \
gt/intel_breadcrumbs.o \
gt/intel_context.o \
gt/intel_engine_cs.o \
+ gt/intel_engine_heartbeat.o \
gt/intel_engine_pm.o \
gt/intel_engine_pool.o \
gt/intel_engine_sysfs.o \
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index 0ab416887fc2..e832238a72e5 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -70,6 +70,7 @@
#include <drm/i915_drm.h>
#include "gt/intel_lrc_reg.h"
+#include "gt/intel_engine_heartbeat.h"
#include "gt/intel_engine_user.h"
#include "i915_gem_context.h"
@@ -375,6 +376,78 @@ void i915_gem_context_release(struct kref *ref)
queue_work(i915->wq, &i915->contexts.free_work);
}
+static inline struct i915_gem_engines *
+__context_engines_static(struct i915_gem_context *ctx)
+{
+ return rcu_dereference_protected(ctx->engines, true);
+}
+
+static void kill_context(struct i915_gem_context *ctx)
+{
+ intel_engine_mask_t tmp, active, reset;
+ struct intel_gt *gt = &ctx->i915->gt;
+ struct i915_gem_engines_iter it;
+ struct intel_engine_cs *engine;
+ struct intel_context *ce;
+
+ /*
+ * If we are already banned, it was due to a guilty request causing
+ * a reset and the entire context being evicted from the GPU.
+ */
+ if (i915_gem_context_is_banned(ctx))
+ return;
+
+ i915_gem_context_set_banned(ctx);
+
+ /*
+ * Map the user's engine back to the actual engines; one virtual
+ * engine will be mapped to multiple engines, and using ctx->engine[]
+ * the same engine may be have multiple instances in the user's map.
+ * However, we only care about pending requests, so only include
+ * engines on which there are incomplete requests.
+ */
+ active = 0;
+ for_each_gem_engine(ce, __context_engines_static(ctx), it) {
+ struct dma_fence *fence;
+
+ if (!ce->timeline)
+ continue;
+
+ fence = i915_active_fence_get(&ce->timeline->last_request);
+ if (!fence)
+ continue;
+
+ engine = to_request(fence)->engine;
+ if (HAS_EXECLISTS(gt->i915))
+ engine = intel_context_inflight(ce);
+ if (engine)
+ active |= engine->mask;
+
+ dma_fence_put(fence);
+ }
+
+ /*
+ * Send a "high priority pulse" down the engine to cause the
+ * current request to be momentarily preempted. (If it fails to
+ * be preempted, it will be reset). As we have marked our context
+ * as banned, any incomplete request, including any running, will
+ * be skipped following the preemption.
+ */
+ reset = 0;
+ for_each_engine_masked(engine, gt->i915, active, tmp)
+ if (intel_engine_pulse(engine))
+ reset |= engine->mask;
+
+ /*
+ * If we are unable to send a preemptive pulse to bump
+ * the context from the GPU, we have to resort to a full
+ * reset. We hope the collateral damage is worth it.
+ */
+ if (reset)
+ intel_gt_handle_error(gt, reset, 0,
+ "context closure in %s", ctx->name);
+}
+
static void context_close(struct i915_gem_context *ctx)
{
i915_gem_context_set_closed(ctx);
@@ -400,6 +473,10 @@ static void context_close(struct i915_gem_context *ctx)
lut_close(ctx);
mutex_unlock(&ctx->mutex);
+
+ if (!i915_gem_context_is_persistent(ctx))
+ kill_context(ctx);
+
i915_gem_context_put(ctx);
}
@@ -440,6 +517,21 @@ __create_context(struct drm_i915_private *i915)
i915_gem_context_set_bannable(ctx);
i915_gem_context_set_recoverable(ctx);
+ /*
+ * If the user has disabled hangchecking, we can not be sure that
+ * the batches will ever complete after the context is closed,
+ * keep the context and all resources pinned forever. So in this
+ * case we opt to forcibly kill off all remaining requests on
+ * context close.
+ *
+ * Note that the user may chance the value of the modparam between
+ * context creation and close, we choose to ignore this for the
+ * sake of determinism and expect the user to set the parameter
+ * on module load and never touch it again.
+ */
+ if (i915_modparams.enable_hangcheck) /* cgroup hook? */
+ i915_gem_context_set_persistence(ctx);
+
for (i = 0; i < ARRAY_SIZE(ctx->hang_timestamp); i++)
ctx->hang_timestamp[i] = jiffies - CONTEXT_FAST_HANG_JIFFIES;
@@ -597,6 +689,7 @@ i915_gem_context_create_kernel(struct drm_i915_private *i915, int prio)
}
i915_gem_context_clear_bannable(ctx);
+ i915_gem_context_set_persistence(ctx);
ctx->sched.priority = I915_USER_PRIORITY(prio);
GEM_BUG_ON(!i915_gem_context_is_kernel(ctx));
@@ -1724,6 +1817,26 @@ get_engines(struct i915_gem_context *ctx,
return err;
}
+static int
+set_persistence(struct i915_gem_context *ctx,
+ const struct drm_i915_gem_context_param *args)
+{
+ if (args->size)
+ return -EINVAL;
+
+ if (args->value) {
+ i915_gem_context_set_persistence(ctx);
+ return 0;
+ }
+
+ /* To cancel a context we use "preempt-to-idle" */
+ if (!(ctx->i915->caps.scheduler & I915_SCHEDULER_CAP_PREEMPTION))
+ return -ENODEV;
+
+ i915_gem_context_clear_persistence(ctx);
+ return 0;
+}
+
static int ctx_setparam(struct drm_i915_file_private *fpriv,
struct i915_gem_context *ctx,
struct drm_i915_gem_context_param *args)
@@ -1801,6 +1914,10 @@ static int ctx_setparam(struct drm_i915_file_private *fpriv,
ret = set_engines(ctx, args);
break;
+ case I915_CONTEXT_PARAM_PERSISTENCE:
+ ret = set_persistence(ctx, args);
+ break;
+
case I915_CONTEXT_PARAM_BAN_PERIOD:
default:
ret = -EINVAL;
@@ -2250,6 +2367,11 @@ int i915_gem_context_getparam_ioctl(struct drm_device *dev, void *data,
ret = get_engines(ctx, args);
break;
+ case I915_CONTEXT_PARAM_PERSISTENCE:
+ args->size = 0;
+ args->value = i915_gem_context_is_persistent(ctx);
+ break;
+
case I915_CONTEXT_PARAM_BAN_PERIOD:
default:
ret = -EINVAL;
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.h b/drivers/gpu/drm/i915/gem/i915_gem_context.h
index 176978608b6f..e0f5b6c6a331 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.h
@@ -74,6 +74,21 @@ static inline void i915_gem_context_clear_recoverable(struct i915_gem_context *c
clear_bit(UCONTEXT_RECOVERABLE, &ctx->user_flags);
}
+static inline bool i915_gem_context_is_persistent(const struct i915_gem_context *ctx)
+{
+ return test_bit(UCONTEXT_PERSISTENCE, &ctx->user_flags);
+}
+
+static inline void i915_gem_context_set_persistence(struct i915_gem_context *ctx)
+{
+ set_bit(UCONTEXT_PERSISTENCE, &ctx->user_flags);
+}
+
+static inline void i915_gem_context_clear_persistence(struct i915_gem_context *ctx)
+{
+ clear_bit(UCONTEXT_PERSISTENCE, &ctx->user_flags);
+}
+
static inline bool i915_gem_context_is_banned(const struct i915_gem_context *ctx)
{
return test_bit(CONTEXT_BANNED, &ctx->flags);
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 260d59cc3de8..daf1ea5075a6 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
@@ -137,6 +137,7 @@ struct i915_gem_context {
#define UCONTEXT_NO_ERROR_CAPTURE 1
#define UCONTEXT_BANNABLE 2
#define UCONTEXT_RECOVERABLE 3
+#define UCONTEXT_PERSISTENCE 4
/**
* @flags: small set of booleans
diff --git a/drivers/gpu/drm/i915/gem/selftests/mock_context.c b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
index be8974ccff24..0d1a275ec316 100644
--- a/drivers/gpu/drm/i915/gem/selftests/mock_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
@@ -23,6 +23,8 @@ mock_context(struct drm_i915_private *i915,
INIT_LIST_HEAD(&ctx->link);
ctx->i915 = i915;
+ i915_gem_context_set_persistence(ctx);
+
mutex_init(&ctx->engines_mutex);
e = default_engines(ctx);
if (IS_ERR(e))
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c b/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c
new file mode 100644
index 000000000000..2fc413f9d506
--- /dev/null
+++ b/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c
@@ -0,0 +1,56 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2019 Intel Corporation
+ */
+
+#include "i915_request.h"
+
+#include "intel_context.h"
+#include "intel_engine_heartbeat.h"
+#include "intel_engine_pm.h"
+#include "intel_engine.h"
+#include "intel_gt.h"
+
+static void idle_pulse(struct intel_engine_cs *engine, struct i915_request *rq)
+{
+ engine->wakeref_serial = READ_ONCE(engine->serial) + 1;
+ i915_request_add_active_barriers(rq);
+}
+
+int intel_engine_pulse(struct intel_engine_cs *engine)
+{
+ struct i915_sched_attr attr = { .priority = I915_PRIORITY_BARRIER };
+ struct intel_context *ce = engine->kernel_context;
+ struct i915_request *rq;
+ int err = 0;
+
+ if (!intel_engine_has_preemption(engine))
+ return -ENODEV;
+
+ if (!intel_engine_pm_get_if_awake(engine))
+ return 0;
+
+ if (mutex_lock_interruptible(&ce->timeline->mutex))
+ goto out_rpm;
+
+ intel_context_enter(ce);
+ rq = __i915_request_create(ce, GFP_NOWAIT | __GFP_NOWARN);
+ intel_context_exit(ce);
+ if (IS_ERR(rq)) {
+ err = PTR_ERR(rq);
+ goto out_unlock;
+ }
+
+ rq->flags |= I915_REQUEST_SENTINEL;
+ idle_pulse(engine, rq);
+
+ __i915_request_commit(rq);
+ __i915_request_queue(rq, &attr);
+
+out_unlock:
+ mutex_unlock(&ce->timeline->mutex);
+out_rpm:
+ intel_engine_pm_put(engine);
+ return err;
+}
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.h b/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.h
new file mode 100644
index 000000000000..b950451b5998
--- /dev/null
+++ b/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.h
@@ -0,0 +1,14 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2019 Intel Corporation
+ */
+
+#ifndef INTEL_ENGINE_HEARTBEAT_H
+#define INTEL_ENGINE_HEARTBEAT_H
+
+struct intel_engine_cs;
+
+int intel_engine_pulse(struct intel_engine_cs *engine);
+
+#endif /* INTEL_ENGINE_HEARTBEAT_H */
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pm.c b/drivers/gpu/drm/i915/gt/intel_engine_pm.c
index 360fd5cb07ab..d11107d92651 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_pm.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_pm.c
@@ -110,7 +110,7 @@ static bool switch_to_kernel_context(struct intel_engine_cs *engine)
i915_request_add_active_barriers(rq);
/* Install ourselves as a preemption barrier */
- rq->sched.attr.priority = I915_PRIORITY_UNPREEMPTABLE;
+ rq->sched.attr.priority = I915_PRIORITY_BARRIER;
__i915_request_commit(rq);
/* Release our exclusive hold on the engine */
diff --git a/drivers/gpu/drm/i915/i915_priolist_types.h b/drivers/gpu/drm/i915/i915_priolist_types.h
index 21037a2e2038..ae8bb3cb627e 100644
--- a/drivers/gpu/drm/i915/i915_priolist_types.h
+++ b/drivers/gpu/drm/i915/i915_priolist_types.h
@@ -39,6 +39,7 @@ enum {
* active request.
*/
#define I915_PRIORITY_UNPREEMPTABLE INT_MAX
+#define I915_PRIORITY_BARRIER INT_MAX
#define __NO_PREEMPTION (I915_PRIORITY_WAIT)
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index 30c542144016..eb9e704d717a 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -1565,6 +1565,21 @@ struct drm_i915_gem_context_param {
* i915_context_engines_bond (I915_CONTEXT_ENGINES_EXT_BOND)
*/
#define I915_CONTEXT_PARAM_ENGINES 0xa
+
+/*
+ * I915_CONTEXT_PARAM_PERSISTENCE:
+ *
+ * Allow the context and active rendering to survive the process until
+ * completion. Persistence allows fire-and-forget clients to queue up a
+ * bunch of work, hand the output over to a display server and the quit.
+ * If the context is not marked as persistent, upon closing (either via
+ * an explicit DRM_I915_GEM_CONTEXT_DESTROY or implicitly from file closure
+ * or process termination), the context and any outstanding requests will be
+ * cancelled (and exported fences for cancelled requests marked as -EIO).
+ *
+ * By default, new contexts allow persistence.
+ */
+#define I915_CONTEXT_PARAM_PERSISTENCE 0xb
/* Must be kept compact -- no holes and well documented */
__u64 value;
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* [PATCH 19/27] drm/i915: Replace hangcheck by heartbeats
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (17 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 18/27] drm/i915: Cancel non-persistent contexts on close Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-27 8:26 ` Joonas Lahtinen
2019-09-25 10:01 ` [PATCH 20/27] drm/i915: Remove logical HW ID Chris Wilson
` (11 subsequent siblings)
30 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
Replace sampling the engine state every so often with a periodic
heartbeat request to measure the health of an engine. This is coupled
with the forced-preemption to allow long running requests to survive so
long as they do not block other users.
v2: Couple in sysfs controls
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Jon Bloomfield <jon.bloomfield@intel.com>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
---
drivers/gpu/drm/i915/Kconfig.profile | 11 +
drivers/gpu/drm/i915/Makefile | 1 -
drivers/gpu/drm/i915/display/intel_display.c | 2 +-
drivers/gpu/drm/i915/gem/i915_gem_object.h | 1 -
drivers/gpu/drm/i915/gem/i915_gem_pm.c | 2 -
drivers/gpu/drm/i915/gt/intel_engine.h | 32 --
drivers/gpu/drm/i915/gt/intel_engine_cs.c | 11 +-
.../gpu/drm/i915/gt/intel_engine_heartbeat.c | 115 ++++++
.../gpu/drm/i915/gt/intel_engine_heartbeat.h | 5 +
drivers/gpu/drm/i915/gt/intel_engine_pm.c | 5 +-
drivers/gpu/drm/i915/gt/intel_engine_sysfs.c | 29 ++
drivers/gpu/drm/i915/gt/intel_engine_types.h | 17 +-
drivers/gpu/drm/i915/gt/intel_gt.c | 1 -
drivers/gpu/drm/i915/gt/intel_gt.h | 4 -
drivers/gpu/drm/i915/gt/intel_gt_pm.c | 1 -
drivers/gpu/drm/i915/gt/intel_gt_types.h | 9 -
drivers/gpu/drm/i915/gt/intel_hangcheck.c | 361 ------------------
drivers/gpu/drm/i915/gt/intel_reset.c | 3 +-
drivers/gpu/drm/i915/gt/selftest_hangcheck.c | 4 -
drivers/gpu/drm/i915/i915_debugfs.c | 87 -----
drivers/gpu/drm/i915/i915_drv.c | 3 -
drivers/gpu/drm/i915/i915_drv.h | 1 -
drivers/gpu/drm/i915/i915_gpu_error.c | 33 +-
drivers/gpu/drm/i915/i915_gpu_error.h | 2 -
drivers/gpu/drm/i915/i915_priolist_types.h | 6 +
25 files changed, 191 insertions(+), 555 deletions(-)
delete mode 100644 drivers/gpu/drm/i915/gt/intel_hangcheck.c
diff --git a/drivers/gpu/drm/i915/Kconfig.profile b/drivers/gpu/drm/i915/Kconfig.profile
index 3184e8491333..aafb57f84169 100644
--- a/drivers/gpu/drm/i915/Kconfig.profile
+++ b/drivers/gpu/drm/i915/Kconfig.profile
@@ -37,3 +37,14 @@ config DRM_I915_PREEMPT_TIMEOUT
to execute.
May be 0 to disable the timeout.
+
+config DRM_I915_HEARTBEAT_INTERVAL
+ int "Interval between heartbeat pulses (ms)"
+ default 2500 # microseconds
+ help
+ While active the driver uses a periodic request, a heartbeat, to
+ check the wellness of the GPU and to regularly flush state changes
+ (idle barriers).
+
+ May be 0 to disable heartbeats and therefore disable automatic GPU
+ hang detection.
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index 82088f748409..f1f150344541 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -86,7 +86,6 @@ gt-y += \
gt/intel_gt_pm.o \
gt/intel_gt_pm_irq.o \
gt/intel_gt_requests.o \
- gt/intel_hangcheck.o \
gt/intel_lrc.o \
gt/intel_renderstate.o \
gt/intel_reset.o \
diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
index 12abcd1e471b..99ddbb7bedbc 100644
--- a/drivers/gpu/drm/i915/display/intel_display.c
+++ b/drivers/gpu/drm/i915/display/intel_display.c
@@ -14345,7 +14345,7 @@ static void intel_plane_unpin_fb(struct intel_plane_state *old_plane_state)
static void fb_obj_bump_render_priority(struct drm_i915_gem_object *obj)
{
struct i915_sched_attr attr = {
- .priority = I915_PRIORITY_DISPLAY,
+ .priority = I915_USER_PRIORITY(I915_PRIORITY_DISPLAY),
};
i915_gem_object_wait_priority(obj, 0, &attr);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index a78af25dce36..967c30737dc5 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -429,6 +429,5 @@ int i915_gem_object_wait(struct drm_i915_gem_object *obj,
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_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
index 9b52aca70eee..e04e4ff7b52e 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
@@ -100,8 +100,6 @@ void i915_gem_suspend(struct drm_i915_private *i915)
*/
switch_to_kernel_context_sync(&i915->gt);
- cancel_delayed_work_sync(&i915->gt.hangcheck.work);
-
i915_gem_drain_freed_objects(i915);
intel_uc_suspend(&i915->gt.uc);
diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h
index d3c6993f4f46..da09eea0fb86 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine.h
@@ -89,38 +89,6 @@ struct drm_printer;
/* seqno size is actually only a uint32, but since we plan to use MI_FLUSH_DW to
* do the writes, and that must have qw aligned offsets, simply pretend it's 8b.
*/
-enum intel_engine_hangcheck_action {
- ENGINE_IDLE = 0,
- ENGINE_WAIT,
- ENGINE_ACTIVE_SEQNO,
- ENGINE_ACTIVE_HEAD,
- ENGINE_ACTIVE_SUBUNITS,
- ENGINE_WAIT_KICK,
- ENGINE_DEAD,
-};
-
-static inline const char *
-hangcheck_action_to_str(const enum intel_engine_hangcheck_action a)
-{
- switch (a) {
- case ENGINE_IDLE:
- return "idle";
- case ENGINE_WAIT:
- return "wait";
- case ENGINE_ACTIVE_SEQNO:
- return "active seqno";
- case ENGINE_ACTIVE_HEAD:
- return "active head";
- case ENGINE_ACTIVE_SUBUNITS:
- return "active subunits";
- case ENGINE_WAIT_KICK:
- return "wait kick";
- case ENGINE_DEAD:
- return "dead";
- }
-
- return "unknown";
-}
static inline unsigned int
execlists_num_ports(const struct intel_engine_execlists * const execlists)
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
index 12855947bc9b..358333de409c 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
@@ -305,6 +305,7 @@ static int intel_engine_setup(struct intel_gt *gt, enum intel_engine_id id)
__sprint_engine_name(engine);
engine->props.preempt_timeout = CONFIG_DRM_I915_PREEMPT_TIMEOUT;
+ engine->props.heartbeat_interval = CONFIG_DRM_I915_HEARTBEAT_INTERVAL;
/*
* To be overridden by the backend on setup. However to facilitate
@@ -599,7 +600,6 @@ static int intel_engine_setup_common(struct intel_engine_cs *engine)
intel_engine_init_active(engine, ENGINE_PHYSICAL);
intel_engine_init_breadcrumbs(engine);
intel_engine_init_execlists(engine);
- intel_engine_init_hangcheck(engine);
intel_engine_init_cmd_parser(engine);
intel_engine_init__pm(engine);
@@ -1416,8 +1416,13 @@ 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: %d ms ago\n",
- jiffies_to_msecs(jiffies - engine->hangcheck.action_timestamp));
+
+ rcu_read_lock();
+ rq = READ_ONCE(engine->heartbeat.systole);
+ if (rq)
+ drm_printf(m, "\tHeartbeat: %d ms ago\n",
+ jiffies_to_msecs(jiffies - rq->emitted_jiffies));
+ rcu_read_unlock();
drm_printf(m, "\tReset count: %d (global %d)\n",
i915_reset_engine_count(error, engine),
i915_reset_count(error));
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c b/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c
index 2fc413f9d506..f68acf9118f3 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c
@@ -11,6 +11,27 @@
#include "intel_engine_pm.h"
#include "intel_engine.h"
#include "intel_gt.h"
+#include "intel_reset.h"
+
+/*
+ * While the engine is active, we send a periodic pulse along the engine
+ * to check on its health and to flush any idle-barriers. If that request
+ * is stuck, and we fail to preempt it, we declare the engine hung and
+ * issue a reset -- in the hope that restores progress.
+ */
+
+static void next_heartbeat(struct intel_engine_cs *engine)
+{
+ long delay;
+
+ delay = READ_ONCE(engine->props.heartbeat_interval);
+ if (!delay)
+ return;
+
+ delay = msecs_to_jiffies_timeout(delay);
+ schedule_delayed_work(&engine->heartbeat.work,
+ round_jiffies_up_relative(delay));
+}
static void idle_pulse(struct intel_engine_cs *engine, struct i915_request *rq)
{
@@ -18,6 +39,100 @@ static void idle_pulse(struct intel_engine_cs *engine, struct i915_request *rq)
i915_request_add_active_barriers(rq);
}
+static void heartbeat(struct work_struct *wrk)
+{
+ struct i915_sched_attr attr = {
+ .priority = I915_USER_PRIORITY(I915_PRIORITY_MIN),
+ };
+ struct intel_engine_cs *engine =
+ container_of(wrk, typeof(*engine), heartbeat.work.work);
+ struct intel_context *ce = engine->kernel_context;
+ struct i915_request *rq;
+
+ if (!intel_engine_pm_get_if_awake(engine))
+ return;
+
+ rq = engine->heartbeat.systole;
+ if (rq && i915_request_completed(rq)) {
+ i915_request_put(rq);
+ engine->heartbeat.systole = NULL;
+ }
+
+ if (intel_gt_is_wedged(engine->gt))
+ goto out;
+
+ if (engine->heartbeat.systole) {
+ if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) {
+ struct drm_printer p = drm_debug_printer(__func__);
+
+ intel_engine_dump(engine, &p,
+ "%s heartbeat not ticking\n",
+ engine->name);
+ }
+
+ if (engine->schedule &&
+ rq->sched.attr.priority < I915_PRIORITY_BARRIER) {
+ attr.priority =
+ I915_USER_PRIORITY(I915_PRIORITY_HEARTBEAT);
+ if (rq->sched.attr.priority >= attr.priority)
+ attr.priority = I915_PRIORITY_BARRIER;
+
+ local_bh_disable();
+ engine->schedule(rq, &attr);
+ local_bh_enable();
+ } else {
+ intel_gt_handle_error(engine->gt, engine->mask,
+ I915_ERROR_CAPTURE,
+ "stopped heartbeat on %s",
+ engine->name);
+ }
+ goto out;
+ }
+
+ if (engine->wakeref_serial == engine->serial)
+ goto out;
+
+ mutex_lock(&ce->timeline->mutex);
+
+ intel_context_enter(ce);
+ rq = __i915_request_create(ce, GFP_NOWAIT | __GFP_NOWARN);
+ intel_context_exit(ce);
+ if (IS_ERR(rq))
+ goto unlock;
+
+ idle_pulse(engine, rq);
+ if (i915_modparams.enable_hangcheck)
+ engine->heartbeat.systole = i915_request_get(rq);
+
+ __i915_request_commit(rq);
+ __i915_request_queue(rq, &attr);
+
+unlock:
+ mutex_unlock(&ce->timeline->mutex);
+out:
+ next_heartbeat(engine);
+ intel_engine_pm_put(engine);
+}
+
+void intel_engine_unpark_heartbeat(struct intel_engine_cs *engine)
+{
+ if (!CONFIG_DRM_I915_HEARTBEAT_INTERVAL)
+ return;
+
+ next_heartbeat(engine);
+}
+
+void intel_engine_park_heartbeat(struct intel_engine_cs *engine)
+{
+ cancel_delayed_work(&engine->heartbeat.work);
+ i915_request_put(fetch_and_zero(&engine->heartbeat.systole));
+}
+
+void intel_engine_init_heartbeat(struct intel_engine_cs *engine)
+{
+ INIT_DELAYED_WORK(&engine->heartbeat.work, heartbeat);
+}
+
int intel_engine_pulse(struct intel_engine_cs *engine)
{
struct i915_sched_attr attr = { .priority = I915_PRIORITY_BARRIER };
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.h b/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.h
index b950451b5998..39391004554d 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine_heartbeat.h
@@ -9,6 +9,11 @@
struct intel_engine_cs;
+void intel_engine_init_heartbeat(struct intel_engine_cs *engine);
+
+void intel_engine_park_heartbeat(struct intel_engine_cs *engine);
+void intel_engine_unpark_heartbeat(struct intel_engine_cs *engine);
+
int intel_engine_pulse(struct intel_engine_cs *engine);
#endif /* INTEL_ENGINE_HEARTBEAT_H */
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pm.c b/drivers/gpu/drm/i915/gt/intel_engine_pm.c
index d11107d92651..f2ec976ca1dd 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_pm.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_pm.c
@@ -7,6 +7,7 @@
#include "i915_drv.h"
#include "intel_engine.h"
+#include "intel_engine_heartbeat.h"
#include "intel_engine_pm.h"
#include "intel_engine_pool.h"
#include "intel_gt.h"
@@ -33,7 +34,7 @@ static int __engine_unpark(struct intel_wakeref *wf)
if (engine->unpark)
engine->unpark(engine);
- intel_engine_init_hangcheck(engine);
+ intel_engine_unpark_heartbeat(engine);
return 0;
}
@@ -157,6 +158,7 @@ static int __engine_park(struct intel_wakeref *wf)
call_idle_barriers(engine); /* cleanup after wedging */
+ intel_engine_park_heartbeat(engine);
intel_engine_disarm_breadcrumbs(engine);
intel_engine_pool_park(&engine->pool);
@@ -187,6 +189,7 @@ void intel_engine_init__pm(struct intel_engine_cs *engine)
struct intel_runtime_pm *rpm = &engine->i915->runtime_pm;
intel_wakeref_init(&engine->wakeref, rpm, &wf_ops);
+ intel_engine_init_heartbeat(engine);
}
#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_sysfs.c b/drivers/gpu/drm/i915/gt/intel_engine_sysfs.c
index 5aa3b67bd6cd..be588eba1156 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_sysfs.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_sysfs.c
@@ -69,12 +69,38 @@ preempt_timeout_store(struct kobject *kobj, struct kobj_attribute *attr,
return count;
}
+static ssize_t
+heartbeat_interval_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
+{
+ struct intel_engine_cs *engine = kobj_to_engine(kobj);
+
+ return sprintf(buf, "%lu\n", engine->props.heartbeat_interval);
+}
+
+static ssize_t
+heartbeat_interval_store(struct kobject *kobj, struct kobj_attribute *attr,
+ const char *buf, size_t count)
+{
+ struct intel_engine_cs *engine = kobj_to_engine(kobj);
+ unsigned long delay;
+ int err;
+
+ err = kstrtoul(buf, 0, &delay);
+ if (err)
+ return err;
+
+ engine->props.heartbeat_interval = delay;
+ return count;
+}
+
static struct kobj_attribute name_attr = __ATTR(name, 0444, name_show, NULL);
static struct kobj_attribute class_attr = __ATTR(class, 0444, class_show, NULL);
static struct kobj_attribute inst_attr = __ATTR(instance, 0444, inst_show, NULL);
static struct kobj_attribute mmio_attr = __ATTR(mmio_base, 0444, mmio_show, NULL);
static struct kobj_attribute preempt_timeout_attr =
__ATTR(preempt_timeout_ms, 0600, preempt_timeout_show, preempt_timeout_store);
+static struct kobj_attribute heartbeat_interval_attr =
+__ATTR(heartbeat_interval_ms, 0600, heartbeat_interval_show, heartbeat_interval_store);
static void kobj_engine_release(struct kobject *kobj)
{
@@ -114,6 +140,9 @@ void intel_engines_add_sysfs(struct drm_i915_private *i915)
&class_attr.attr,
&inst_attr.attr,
&mmio_attr.attr,
+#if CONFIG_DRM_I915_HEARTBEAT_INTERVAL
+ &heartbeat_interval_attr.attr,
+#endif
NULL
};
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
index b144704d2f21..da4b45aa84b1 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
@@ -15,6 +15,7 @@
#include <linux/rbtree.h>
#include <linux/timer.h>
#include <linux/types.h>
+#include <linux/workqueue.h>
#include "i915_gem.h"
#include "i915_pmu.h"
@@ -76,14 +77,6 @@ struct intel_instdone {
u32 row[I915_MAX_SLICES][I915_MAX_SUBSLICES];
};
-struct intel_engine_hangcheck {
- u64 acthd;
- u32 last_ring;
- u32 last_head;
- unsigned long action_timestamp;
- struct intel_instdone instdone;
-};
-
struct intel_ring {
struct kref ref;
struct i915_vma *vma;
@@ -323,6 +316,11 @@ struct intel_engine_cs {
intel_engine_mask_t saturated; /* submitting semaphores too late? */
+ struct {
+ struct delayed_work work;
+ struct i915_request *systole;
+ } heartbeat;
+
unsigned long serial;
unsigned long wakeref_serial;
@@ -473,8 +471,6 @@ struct intel_engine_cs {
/* status_notifier: list of callbacks for context-switch changes */
struct atomic_notifier_head context_status_notifier;
- struct intel_engine_hangcheck hangcheck;
-
#define I915_ENGINE_NEEDS_CMD_PARSER BIT(0)
#define I915_ENGINE_SUPPORTS_STATS BIT(1)
#define I915_ENGINE_HAS_PREEMPTION BIT(2)
@@ -542,6 +538,7 @@ struct intel_engine_cs {
struct {
unsigned long preempt_timeout;
+ unsigned long heartbeat_interval;
} props;
};
diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c b/drivers/gpu/drm/i915/gt/intel_gt.c
index 39fb69609be4..700be75a1177 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt.c
+++ b/drivers/gpu/drm/i915/gt/intel_gt.c
@@ -21,7 +21,6 @@ void intel_gt_init_early(struct intel_gt *gt, struct drm_i915_private *i915)
INIT_LIST_HEAD(>->closed_vma);
spin_lock_init(>->closed_lock);
- intel_gt_init_hangcheck(gt);
intel_gt_init_reset(gt);
intel_gt_init_requests(gt);
intel_gt_pm_init_early(gt);
diff --git a/drivers/gpu/drm/i915/gt/intel_gt.h b/drivers/gpu/drm/i915/gt/intel_gt.h
index e6ab0bff0efb..5b6effed3713 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt.h
+++ b/drivers/gpu/drm/i915/gt/intel_gt.h
@@ -46,8 +46,6 @@ void intel_gt_clear_error_registers(struct intel_gt *gt,
void intel_gt_flush_ggtt_writes(struct intel_gt *gt);
void intel_gt_chipset_flush(struct intel_gt *gt);
-void intel_gt_init_hangcheck(struct intel_gt *gt);
-
static inline u32 intel_gt_scratch_offset(const struct intel_gt *gt,
enum intel_gt_scratch_field field)
{
@@ -59,6 +57,4 @@ static inline bool intel_gt_is_wedged(struct intel_gt *gt)
return __intel_reset_failed(>->reset);
}
-void intel_gt_queue_hangcheck(struct intel_gt *gt);
-
#endif /* __INTEL_GT_H__ */
diff --git a/drivers/gpu/drm/i915/gt/intel_gt_pm.c b/drivers/gpu/drm/i915/gt/intel_gt_pm.c
index 660857a665e7..475083f9680b 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt_pm.c
+++ b/drivers/gpu/drm/i915/gt/intel_gt_pm.c
@@ -51,7 +51,6 @@ static int __gt_unpark(struct intel_wakeref *wf)
i915_pmu_gt_unparked(i915);
- intel_gt_queue_hangcheck(gt);
intel_gt_unpark_requests(gt);
pm_notify(gt, INTEL_GT_UNPARK);
diff --git a/drivers/gpu/drm/i915/gt/intel_gt_types.h b/drivers/gpu/drm/i915/gt/intel_gt_types.h
index f524942c82aa..e045354e78e2 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_gt_types.h
@@ -25,14 +25,6 @@ struct i915_ggtt;
struct intel_engine_cs;
struct intel_uncore;
-struct intel_hangcheck {
- /* For hangcheck timer */
-#define DRM_I915_HANGCHECK_PERIOD 1500 /* in ms */
-#define DRM_I915_HANGCHECK_JIFFIES msecs_to_jiffies(DRM_I915_HANGCHECK_PERIOD)
-
- struct delayed_work work;
-};
-
struct intel_gt {
struct drm_i915_private *i915;
struct intel_uncore *uncore;
@@ -66,7 +58,6 @@ struct intel_gt {
struct list_head closed_vma;
spinlock_t closed_lock; /* guards the list of closed_vma */
- struct intel_hangcheck hangcheck;
struct intel_reset reset;
/**
diff --git a/drivers/gpu/drm/i915/gt/intel_hangcheck.c b/drivers/gpu/drm/i915/gt/intel_hangcheck.c
deleted file mode 100644
index 40f62f780be5..000000000000
--- a/drivers/gpu/drm/i915/gt/intel_hangcheck.c
+++ /dev/null
@@ -1,361 +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.
- *
- */
-
-#include "i915_drv.h"
-#include "intel_engine.h"
-#include "intel_gt.h"
-#include "intel_reset.h"
-
-struct hangcheck {
- u64 acthd;
- u32 ring;
- u32 head;
- enum intel_engine_hangcheck_action action;
- unsigned long action_timestamp;
- int deadlock;
- struct intel_instdone instdone;
- bool wedged:1;
- bool stalled:1;
-};
-
-static bool instdone_unchanged(u32 current_instdone, u32 *old_instdone)
-{
- u32 tmp = current_instdone | *old_instdone;
- bool unchanged;
-
- unchanged = tmp == *old_instdone;
- *old_instdone |= tmp;
-
- return unchanged;
-}
-
-static bool subunits_stuck(struct intel_engine_cs *engine)
-{
- struct drm_i915_private *dev_priv = engine->i915;
- const struct sseu_dev_info *sseu = &RUNTIME_INFO(dev_priv)->sseu;
- struct intel_instdone instdone;
- struct intel_instdone *accu_instdone = &engine->hangcheck.instdone;
- bool stuck;
- int slice;
- int subslice;
-
- intel_engine_get_instdone(engine, &instdone);
-
- /* There might be unstable subunit states even when
- * actual head is not moving. Filter out the unstable ones by
- * accumulating the undone -> done transitions and only
- * consider those as progress.
- */
- stuck = instdone_unchanged(instdone.instdone,
- &accu_instdone->instdone);
- stuck &= instdone_unchanged(instdone.slice_common,
- &accu_instdone->slice_common);
-
- for_each_instdone_slice_subslice(dev_priv, sseu, slice, subslice) {
- stuck &= instdone_unchanged(instdone.sampler[slice][subslice],
- &accu_instdone->sampler[slice][subslice]);
- stuck &= instdone_unchanged(instdone.row[slice][subslice],
- &accu_instdone->row[slice][subslice]);
- }
-
- return stuck;
-}
-
-static enum intel_engine_hangcheck_action
-head_stuck(struct intel_engine_cs *engine, u64 acthd)
-{
- if (acthd != engine->hangcheck.acthd) {
-
- /* Clear subunit states on head movement */
- memset(&engine->hangcheck.instdone, 0,
- sizeof(engine->hangcheck.instdone));
-
- return ENGINE_ACTIVE_HEAD;
- }
-
- if (!subunits_stuck(engine))
- return ENGINE_ACTIVE_SUBUNITS;
-
- return ENGINE_DEAD;
-}
-
-static enum intel_engine_hangcheck_action
-engine_stuck(struct intel_engine_cs *engine, u64 acthd)
-{
- enum intel_engine_hangcheck_action ha;
- u32 tmp;
-
- ha = head_stuck(engine, acthd);
- if (ha != ENGINE_DEAD)
- return ha;
-
- if (IS_GEN(engine->i915, 2))
- return ENGINE_DEAD;
-
- /* Is the chip hanging on a WAIT_FOR_EVENT?
- * If so we can simply poke the RB_WAIT bit
- * and break the hang. This should work on
- * all but the second generation chipsets.
- */
- tmp = ENGINE_READ(engine, RING_CTL);
- if (tmp & RING_WAIT) {
- intel_gt_handle_error(engine->gt, engine->mask, 0,
- "stuck wait on %s", engine->name);
- ENGINE_WRITE(engine, RING_CTL, tmp);
- return ENGINE_WAIT_KICK;
- }
-
- return ENGINE_DEAD;
-}
-
-static void hangcheck_load_sample(struct intel_engine_cs *engine,
- struct hangcheck *hc)
-{
- hc->acthd = intel_engine_get_active_head(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_ring = hc->ring;
- engine->hangcheck.last_head = hc->head;
-}
-
-static enum intel_engine_hangcheck_action
-hangcheck_get_action(struct intel_engine_cs *engine,
- const struct hangcheck *hc)
-{
- if (intel_engine_is_idle(engine))
- return ENGINE_IDLE;
-
- if (engine->hangcheck.last_ring != hc->ring)
- return ENGINE_ACTIVE_SEQNO;
-
- if (engine->hangcheck.last_head != hc->head)
- return ENGINE_ACTIVE_SEQNO;
-
- return engine_stuck(engine, hc->acthd);
-}
-
-static void hangcheck_accumulate_sample(struct intel_engine_cs *engine,
- struct hangcheck *hc)
-{
- unsigned long timeout = I915_ENGINE_DEAD_TIMEOUT;
-
- hc->action = hangcheck_get_action(engine, hc);
-
- /* We always increment the progress
- * if the engine is busy and still processing
- * the same request, so that no single request
- * can run indefinitely (such as a chain of
- * batches). The only time we do not increment
- * the hangcheck score on this ring, if this
- * engine is in a legitimate wait for another
- * engine. In that case the waiting engine is a
- * victim and we want to be sure we catch the
- * right culprit. Then every time we do kick
- * the ring, make it as a progress as the seqno
- * advancement might ensure and if not, it
- * will catch the hanging engine.
- */
-
- switch (hc->action) {
- case ENGINE_IDLE:
- case ENGINE_ACTIVE_SEQNO:
- /* Clear head and subunit states on seqno movement */
- hc->acthd = 0;
-
- memset(&engine->hangcheck.instdone, 0,
- sizeof(engine->hangcheck.instdone));
-
- /* Intentional fall through */
- case ENGINE_WAIT_KICK:
- case ENGINE_WAIT:
- engine->hangcheck.action_timestamp = jiffies;
- break;
-
- case ENGINE_ACTIVE_HEAD:
- case ENGINE_ACTIVE_SUBUNITS:
- /*
- * Seqno stuck with still active engine gets leeway,
- * in hopes that it is just a long shader.
- */
- timeout = I915_SEQNO_DEAD_TIMEOUT;
- break;
-
- case ENGINE_DEAD:
- break;
-
- default:
- MISSING_CASE(hc->action);
- }
-
- hc->stalled = time_after(jiffies,
- engine->hangcheck.action_timestamp + timeout);
- hc->wedged = time_after(jiffies,
- engine->hangcheck.action_timestamp +
- I915_ENGINE_WEDGED_TIMEOUT);
-}
-
-static void hangcheck_declare_hang(struct intel_gt *gt,
- intel_engine_mask_t hung,
- intel_engine_mask_t stuck)
-{
- struct intel_engine_cs *engine;
- intel_engine_mask_t tmp;
- char msg[80];
- int len;
-
- /* If some rings hung but others were still busy, only
- * blame the hanging rings in the synopsis.
- */
- if (stuck != hung)
- hung &= ~stuck;
- len = scnprintf(msg, sizeof(msg),
- "%s on ", stuck == hung ? "no progress" : "hang");
- for_each_engine_masked(engine, gt->i915, hung, tmp)
- len += scnprintf(msg + len, sizeof(msg) - len,
- "%s, ", engine->name);
- msg[len-2] = '\0';
-
- return intel_gt_handle_error(gt, hung, I915_ERROR_CAPTURE, "%s", msg);
-}
-
-/*
- * This is called when the chip hasn't reported back with completed
- * batchbuffers in a long time. We keep track per ring seqno progress and
- * if there are no progress, hangcheck score for that ring is increased.
- * Further, acthd is inspected to see if the ring is stuck. On stuck case
- * we kick the ring. If we see no progress on three subsequent calls
- * we assume chip is wedged and try to fix it by resetting the chip.
- */
-static void hangcheck_elapsed(struct work_struct *work)
-{
- struct intel_gt *gt =
- container_of(work, typeof(*gt), hangcheck.work.work);
- intel_engine_mask_t hung = 0, stuck = 0, wedged = 0;
- struct intel_engine_cs *engine;
- enum intel_engine_id id;
- intel_wakeref_t wakeref;
-
- if (!i915_modparams.enable_hangcheck)
- return;
-
- if (!READ_ONCE(gt->awake))
- return;
-
- if (intel_gt_is_wedged(gt))
- return;
-
- wakeref = intel_runtime_pm_get_if_in_use(>->i915->runtime_pm);
- if (!wakeref)
- return;
-
- /* As enabling the GPU requires fairly extensive mmio access,
- * periodically arm the mmio checker to see if we are triggering
- * any invalid access.
- */
- intel_uncore_arm_unclaimed_mmio_detection(gt->uncore);
-
- for_each_engine(engine, gt->i915, id) {
- struct hangcheck hc;
-
- intel_engine_signal_breadcrumbs(engine);
-
- hangcheck_load_sample(engine, &hc);
- hangcheck_accumulate_sample(engine, &hc);
- hangcheck_store_sample(engine, &hc);
-
- if (hc.stalled) {
- hung |= engine->mask;
- if (hc.action != ENGINE_DEAD)
- stuck |= engine->mask;
- }
-
- if (hc.wedged)
- wedged |= engine->mask;
- }
-
- if (GEM_SHOW_DEBUG() && (hung | stuck)) {
- struct drm_printer p = drm_debug_printer("hangcheck");
-
- for_each_engine(engine, gt->i915, id) {
- if (intel_engine_is_idle(engine))
- continue;
-
- intel_engine_dump(engine, &p, "%s\n", engine->name);
- }
- }
-
- if (wedged) {
- dev_err(gt->i915->drm.dev,
- "GPU recovery timed out,"
- " cancelling all in-flight rendering.\n");
- GEM_TRACE_DUMP();
- intel_gt_set_wedged(gt);
- }
-
- if (hung)
- hangcheck_declare_hang(gt, hung, stuck);
-
- intel_runtime_pm_put(>->i915->runtime_pm, wakeref);
-
- /* Reset timer in case GPU hangs without another request being added */
- intel_gt_queue_hangcheck(gt);
-}
-
-void intel_gt_queue_hangcheck(struct intel_gt *gt)
-{
- unsigned long delay;
-
- if (unlikely(!i915_modparams.enable_hangcheck))
- return;
-
- /*
- * Don't continually defer the hangcheck so that it is always run at
- * least once after work has been scheduled on any ring. Otherwise,
- * we will ignore a hung ring if a second ring is kept busy.
- */
-
- delay = round_jiffies_up_relative(DRM_I915_HANGCHECK_JIFFIES);
- queue_delayed_work(system_long_wq, >->hangcheck.work, delay);
-}
-
-void intel_engine_init_hangcheck(struct intel_engine_cs *engine)
-{
- memset(&engine->hangcheck, 0, sizeof(engine->hangcheck));
- engine->hangcheck.action_timestamp = jiffies;
-}
-
-void intel_gt_init_hangcheck(struct intel_gt *gt)
-{
- INIT_DELAYED_WORK(>->hangcheck.work, hangcheck_elapsed);
-}
-
-#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
-#include "selftest_hangcheck.c"
-#endif
diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c
index e8c8d0041e8f..2685c0103e90 100644
--- a/drivers/gpu/drm/i915/gt/intel_reset.c
+++ b/drivers/gpu/drm/i915/gt/intel_reset.c
@@ -997,8 +997,6 @@ void intel_gt_reset(struct intel_gt *gt,
if (ret)
goto taint;
- intel_gt_queue_hangcheck(gt);
-
finish:
reset_finish(gt, awake);
unlock:
@@ -1322,4 +1320,5 @@ void __intel_fini_wedge(struct intel_wedge_me *w)
#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
#include "selftest_reset.c"
+#include "selftest_hangcheck.c"
#endif
diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
index fc4d02406536..f2b5ab1e5a2e 100644
--- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
+++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
@@ -1681,7 +1681,6 @@ int intel_hangcheck_live_selftests(struct drm_i915_private *i915)
};
struct intel_gt *gt = &i915->gt;
intel_wakeref_t wakeref;
- bool saved_hangcheck;
int err;
if (!intel_has_gpu_reset(gt->i915))
@@ -1691,12 +1690,9 @@ int intel_hangcheck_live_selftests(struct drm_i915_private *i915)
return -EIO; /* we're long past hope of a successful reset */
wakeref = intel_runtime_pm_get(>->i915->runtime_pm);
- saved_hangcheck = fetch_and_zero(&i915_modparams.enable_hangcheck);
- drain_delayed_work(>->hangcheck.work); /* flush param */
err = intel_gt_live_subtests(tests, gt);
- i915_modparams.enable_hangcheck = saved_hangcheck;
intel_runtime_pm_put(>->i915->runtime_pm, wakeref);
return err;
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 5f9703ca3104..fd41505d52ec 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -1003,92 +1003,6 @@ static int i915_frequency_info(struct seq_file *m, void *unused)
return ret;
}
-static void i915_instdone_info(struct drm_i915_private *dev_priv,
- struct seq_file *m,
- struct intel_instdone *instdone)
-{
- const struct sseu_dev_info *sseu = &RUNTIME_INFO(dev_priv)->sseu;
- int slice;
- int subslice;
-
- seq_printf(m, "\t\tINSTDONE: 0x%08x\n",
- instdone->instdone);
-
- if (INTEL_GEN(dev_priv) <= 3)
- return;
-
- seq_printf(m, "\t\tSC_INSTDONE: 0x%08x\n",
- instdone->slice_common);
-
- if (INTEL_GEN(dev_priv) <= 6)
- return;
-
- for_each_instdone_slice_subslice(dev_priv, sseu, slice, subslice)
- seq_printf(m, "\t\tSAMPLER_INSTDONE[%d][%d]: 0x%08x\n",
- slice, subslice, instdone->sampler[slice][subslice]);
-
- for_each_instdone_slice_subslice(dev_priv, sseu, slice, subslice)
- seq_printf(m, "\t\tROW_INSTDONE[%d][%d]: 0x%08x\n",
- slice, subslice, instdone->row[slice][subslice]);
-}
-
-static int i915_hangcheck_info(struct seq_file *m, void *unused)
-{
- struct drm_i915_private *i915 = node_to_i915(m->private);
- struct intel_gt *gt = &i915->gt;
- struct intel_engine_cs *engine;
- intel_wakeref_t wakeref;
- enum intel_engine_id id;
-
- seq_printf(m, "Reset flags: %lx\n", gt->reset.flags);
- if (test_bit(I915_WEDGED, >->reset.flags))
- seq_puts(m, "\tWedged\n");
- if (test_bit(I915_RESET_BACKOFF, >->reset.flags))
- seq_puts(m, "\tDevice (global) reset in progress\n");
-
- if (!i915_modparams.enable_hangcheck) {
- seq_puts(m, "Hangcheck disabled\n");
- return 0;
- }
-
- if (timer_pending(>->hangcheck.work.timer))
- seq_printf(m, "Hangcheck active, timer fires in %dms\n",
- jiffies_to_msecs(gt->hangcheck.work.timer.expires -
- jiffies));
- else if (delayed_work_pending(>->hangcheck.work))
- seq_puts(m, "Hangcheck active, work pending\n");
- else
- seq_puts(m, "Hangcheck inactive\n");
-
- seq_printf(m, "GT active? %s\n", yesno(gt->awake));
-
- with_intel_runtime_pm(&i915->runtime_pm, wakeref) {
- for_each_engine(engine, i915, id) {
- struct intel_instdone instdone;
-
- seq_printf(m, "%s: %d ms ago\n",
- engine->name,
- jiffies_to_msecs(jiffies -
- engine->hangcheck.action_timestamp));
-
- seq_printf(m, "\tACTHD = 0x%08llx [current 0x%08llx]\n",
- (long long)engine->hangcheck.acthd,
- intel_engine_get_active_head(engine));
-
- intel_engine_get_instdone(engine, &instdone);
-
- seq_puts(m, "\tinstdone read =\n");
- i915_instdone_info(i915, m, &instdone);
-
- seq_puts(m, "\tinstdone accu =\n");
- i915_instdone_info(i915, m,
- &engine->hangcheck.instdone);
- }
- }
-
- return 0;
-}
-
static int ironlake_drpc_info(struct seq_file *m)
{
struct drm_i915_private *i915 = node_to_i915(m->private);
@@ -4302,7 +4216,6 @@ static const struct drm_info_list i915_debugfs_list[] = {
{"i915_guc_stage_pool", i915_guc_stage_pool, 0},
{"i915_huc_load_status", i915_huc_load_status_info, 0},
{"i915_frequency_info", i915_frequency_info, 0},
- {"i915_hangcheck_info", i915_hangcheck_info, 0},
{"i915_drpc_info", i915_drpc_info, 0},
{"i915_ring_freq_table", i915_ring_freq_table, 0},
{"i915_frontbuffer_tracking", i915_frontbuffer_tracking, 0},
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index ab00ba6f695b..8cac6124d16a 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -1621,10 +1621,7 @@ void i915_driver_remove(struct drm_i915_private *i915)
i915_driver_modeset_remove(i915);
- /* Free error state after interrupts are fully disabled. */
- cancel_delayed_work_sync(&i915->gt.hangcheck.work);
i915_reset_error_state(i915);
-
i915_gem_driver_remove(i915);
intel_power_domains_driver_remove(i915);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 4e0c4ba2d257..ad7ed7f72da8 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2212,7 +2212,6 @@ extern const struct dev_pm_ops i915_pm_ops;
int i915_driver_probe(struct pci_dev *pdev, const struct pci_device_id *ent);
void i915_driver_remove(struct drm_i915_private *i915);
-void intel_engine_init_hangcheck(struct intel_engine_cs *engine);
int vlv_force_gfx_clock(struct drm_i915_private *dev_priv, bool on);
static inline bool intel_gvt_active(struct drm_i915_private *dev_priv)
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index a28ee754b7b4..a86d28bda6dd 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -534,10 +534,6 @@ static void error_print_engine(struct drm_i915_error_state_buf *m,
}
err_printf(m, " ring->head: 0x%08x\n", ee->cpu_ring_head);
err_printf(m, " ring->tail: 0x%08x\n", ee->cpu_ring_tail);
- err_printf(m, " hangcheck timestamp: %dms (%lu%s)\n",
- jiffies_to_msecs(ee->hangcheck_timestamp - epoch),
- ee->hangcheck_timestamp,
- ee->hangcheck_timestamp == epoch ? "; epoch" : "");
err_printf(m, " engine reset count: %u\n", ee->reset_count);
for (n = 0; n < ee->num_ports; n++) {
@@ -679,11 +675,8 @@ static void __err_print_to_sgl(struct drm_i915_error_state_buf *m,
ts = ktime_to_timespec64(error->uptime);
err_printf(m, "Uptime: %lld s %ld us\n",
(s64)ts.tv_sec, ts.tv_nsec / NSEC_PER_USEC);
- err_printf(m, "Epoch: %lu jiffies (%u HZ)\n", error->epoch, HZ);
- err_printf(m, "Capture: %lu jiffies; %d ms ago, %d ms after epoch\n",
- error->capture,
- jiffies_to_msecs(jiffies - error->capture),
- jiffies_to_msecs(error->capture - error->epoch));
+ err_printf(m, "Capture: %lu jiffies; %d ms ago\n",
+ error->capture, jiffies_to_msecs(jiffies - error->capture));
for (ee = error->engine; ee; ee = ee->next)
err_printf(m, "Active process (on ring %s): %s [%d]\n",
@@ -742,7 +735,7 @@ static void __err_print_to_sgl(struct drm_i915_error_state_buf *m,
err_printf(m, "GTT_CACHE_EN: 0x%08x\n", error->gtt_cache);
for (ee = error->engine; ee; ee = ee->next)
- error_print_engine(m, ee, error->epoch);
+ error_print_engine(m, ee, error->capture);
for (ee = error->engine; ee; ee = ee->next) {
const struct drm_i915_error_object *obj;
@@ -770,7 +763,7 @@ static void __err_print_to_sgl(struct drm_i915_error_state_buf *m,
for (j = 0; j < ee->num_requests; j++)
error_print_request(m, " ",
&ee->requests[j],
- error->epoch);
+ error->capture);
}
print_error_obj(m, ee->engine, "ringbuffer", ee->ringbuffer);
@@ -1144,8 +1137,6 @@ static void error_record_engine_registers(struct i915_gpu_state *error,
}
ee->idle = intel_engine_is_idle(engine);
- if (!ee->idle)
- ee->hangcheck_timestamp = engine->hangcheck.action_timestamp;
ee->reset_count = i915_reset_engine_count(&dev_priv->gpu_error,
engine);
@@ -1658,20 +1649,6 @@ static void capture_params(struct i915_gpu_state *error)
i915_params_copy(&error->params, &i915_modparams);
}
-static unsigned long capture_find_epoch(const struct i915_gpu_state *error)
-{
- const struct drm_i915_error_engine *ee;
- unsigned long epoch = error->capture;
-
- for (ee = error->engine; ee; ee = ee->next) {
- if (ee->hangcheck_timestamp &&
- time_before(ee->hangcheck_timestamp, epoch))
- epoch = ee->hangcheck_timestamp;
- }
-
- return epoch;
-}
-
static void capture_finish(struct i915_gpu_state *error)
{
struct i915_ggtt *ggtt = &error->i915->ggtt;
@@ -1723,8 +1700,6 @@ i915_capture_gpu_state(struct drm_i915_private *i915)
error->overlay = intel_overlay_capture_error_state(i915);
error->display = intel_display_capture_error_state(i915);
- error->epoch = capture_find_epoch(error);
-
capture_finish(error);
compress_fini(&compress);
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.h b/drivers/gpu/drm/i915/i915_gpu_error.h
index 63cf387411e0..caaed5093d95 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.h
+++ b/drivers/gpu/drm/i915/i915_gpu_error.h
@@ -34,7 +34,6 @@ struct i915_gpu_state {
ktime_t boottime;
ktime_t uptime;
unsigned long capture;
- unsigned long epoch;
struct drm_i915_private *i915;
@@ -86,7 +85,6 @@ struct i915_gpu_state {
/* Software tracked state */
bool idle;
- unsigned long hangcheck_timestamp;
int num_requests;
u32 reset_count;
diff --git a/drivers/gpu/drm/i915/i915_priolist_types.h b/drivers/gpu/drm/i915/i915_priolist_types.h
index ae8bb3cb627e..732aad148881 100644
--- a/drivers/gpu/drm/i915/i915_priolist_types.h
+++ b/drivers/gpu/drm/i915/i915_priolist_types.h
@@ -16,6 +16,12 @@ enum {
I915_PRIORITY_MIN = I915_CONTEXT_MIN_USER_PRIORITY - 1,
I915_PRIORITY_NORMAL = I915_CONTEXT_DEFAULT_PRIORITY,
I915_PRIORITY_MAX = I915_CONTEXT_MAX_USER_PRIORITY + 1,
+
+ /* A preemptive pulse used to monitor the health of each engine */
+ I915_PRIORITY_HEARTBEAT,
+
+ /* Interactive workload, scheduled for immediate pageflipping */
+ I915_PRIORITY_DISPLAY,
};
#define I915_USER_PRIORITY_SHIFT 2
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* Re: [PATCH 19/27] drm/i915: Replace hangcheck by heartbeats
2019-09-25 10:01 ` [PATCH 19/27] drm/i915: Replace hangcheck by heartbeats Chris Wilson
@ 2019-09-27 8:26 ` Joonas Lahtinen
2019-09-27 9:18 ` Chris Wilson
0 siblings, 1 reply; 70+ messages in thread
From: Joonas Lahtinen @ 2019-09-27 8:26 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
Quoting Chris Wilson (2019-09-25 13:01:29)
> Replace sampling the engine state every so often with a periodic
> heartbeat request to measure the health of an engine. This is coupled
> with the forced-preemption to allow long running requests to survive so
> long as they do not block other users.
>
> v2: Couple in sysfs controls
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> Cc: Jon Bloomfield <jon.bloomfield@intel.com>
> Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
<SNIP>
> +++ b/drivers/gpu/drm/i915/Kconfig.profile
> @@ -37,3 +37,14 @@ config DRM_I915_PREEMPT_TIMEOUT
> to execute.
>
> May be 0 to disable the timeout.
> +
> +config DRM_I915_HEARTBEAT_INTERVAL
> + int "Interval between heartbeat pulses (ms)"
> + default 2500 # microseconds
"ms" or "us", pick one?
> + help
> + While active the driver uses a periodic request, a heartbeat, to
> + check the wellness of the GPU and to regularly flush state changes
> + (idle barriers).
> +
> + May be 0 to disable heartbeats and therefore disable automatic GPU
> + hang detection.
Worth to mention this can be overridden from sysfs.
> +static void heartbeat(struct work_struct *wrk)
> +{
<SNIP>
> + if (i915_modparams.enable_hangcheck)
> + engine->heartbeat.systole = i915_request_get(rq);
I'd be more inclined to the userspace opt-in for running indefinitely and
getting rid of the modparam completely.
The long workloads might even not pre-empt at desired granularity.
Regards, Joonas
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* Re: [PATCH 19/27] drm/i915: Replace hangcheck by heartbeats
2019-09-27 8:26 ` Joonas Lahtinen
@ 2019-09-27 9:18 ` Chris Wilson
0 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-27 9:18 UTC (permalink / raw)
To: Joonas Lahtinen, intel-gfx
Quoting Joonas Lahtinen (2019-09-27 09:26:52)
> Quoting Chris Wilson (2019-09-25 13:01:29)
> > Replace sampling the engine state every so often with a periodic
> > heartbeat request to measure the health of an engine. This is coupled
> > with the forced-preemption to allow long running requests to survive so
> > long as they do not block other users.
> >
> > v2: Couple in sysfs controls
> >
> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> > Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
> > Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> > Cc: Jon Bloomfield <jon.bloomfield@intel.com>
> > Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
>
> <SNIP>
>
> > +++ b/drivers/gpu/drm/i915/Kconfig.profile
> > @@ -37,3 +37,14 @@ config DRM_I915_PREEMPT_TIMEOUT
> > to execute.
> >
> > May be 0 to disable the timeout.
> > +
> > +config DRM_I915_HEARTBEAT_INTERVAL
> > + int "Interval between heartbeat pulses (ms)"
> > + default 2500 # microseconds
>
> "ms" or "us", pick one?
It began with 'm', close enough when copy'n'paste :)
> > + help
> > + While active the driver uses a periodic request, a heartbeat, to
> > + check the wellness of the GPU and to regularly flush state changes
> > + (idle barriers).
> > +
> > + May be 0 to disable heartbeats and therefore disable automatic GPU
> > + hang detection.
>
> Worth to mention this can be overridden from sysfs.
The idea of setting 0 here is to disable it at compilation and DCE.
But any other value could be overridden...
> > +static void heartbeat(struct work_struct *wrk)
> > +{
>
> <SNIP>
>
> > + if (i915_modparams.enable_hangcheck)
> > + engine->heartbeat.systole = i915_request_get(rq);
>
> I'd be more inclined to the userspace opt-in for running indefinitely and
> getting rid of the modparam completely.
That's a separate challenge. :)
> The long workloads might even not pre-empt at desired granularity.
Indeed, but as that is a qos impact on another user, I would similarly
suggest that opting out of this is akin to setting yourself to be
non-preemptable, ergo a restricted operation.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* [PATCH 20/27] drm/i915: Remove logical HW ID
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (18 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 19/27] drm/i915: Replace hangcheck by heartbeats Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 12:41 ` Tvrtko Ursulin
2019-09-25 10:01 ` [PATCH 21/27] drm/i915: Move context management under GEM Chris Wilson
` (10 subsequent siblings)
30 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
With the introduction of ctx->engines[] we allow multiple logical
contexts to be used on the same engine (e.g. with virtual engines). Each
logical context requires a unique tag in order for context-switching to
occur correctly between them.
We only need to keep a unique tag for the active lifetime of the
context, and for as long as we need to identify that context. The HW
uses the tag to determine if it should use a lite-restore (why not the
LRCA?) and passes the tag back for various status identifies. The only
status we need to track is for OA, so when using perf, we assign the
specific context a unique tag.
Fixes: 976b55f0e1db ("drm/i915: Allow a context to define its set of engines")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/gem/i915_gem_context.c | 147 ------------------
drivers/gpu/drm/i915/gem/i915_gem_context.h | 15 --
.../gpu/drm/i915/gem/i915_gem_context_types.h | 18 ---
.../drm/i915/gem/selftests/i915_gem_context.c | 13 +-
.../gpu/drm/i915/gem/selftests/mock_context.c | 8 -
drivers/gpu/drm/i915/gt/intel_context_types.h | 1 +
drivers/gpu/drm/i915/gt/intel_engine_types.h | 4 +-
drivers/gpu/drm/i915/gt/intel_lrc.c | 29 ++--
drivers/gpu/drm/i915/gvt/kvmgt.c | 17 --
drivers/gpu/drm/i915/i915_debugfs.c | 3 -
drivers/gpu/drm/i915/i915_gpu_error.c | 7 +-
drivers/gpu/drm/i915/i915_gpu_error.h | 1 -
drivers/gpu/drm/i915/i915_perf.c | 25 ++-
drivers/gpu/drm/i915/i915_trace.h | 38 ++---
.../gpu/drm/i915/selftests/i915_gem_evict.c | 4 +-
drivers/gpu/drm/i915/selftests/i915_vma.c | 2 +-
16 files changed, 48 insertions(+), 284 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index e832238a72e5..4e59b809d901 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -168,97 +168,6 @@ lookup_user_engine(struct i915_gem_context *ctx,
return i915_gem_context_get_engine(ctx, idx);
}
-static inline int new_hw_id(struct drm_i915_private *i915, gfp_t gfp)
-{
- unsigned int max;
-
- lockdep_assert_held(&i915->contexts.mutex);
-
- if (INTEL_GEN(i915) >= 12)
- max = GEN12_MAX_CONTEXT_HW_ID;
- else if (INTEL_GEN(i915) >= 11)
- max = GEN11_MAX_CONTEXT_HW_ID;
- else if (USES_GUC_SUBMISSION(i915))
- /*
- * When using GuC in proxy submission, GuC consumes the
- * highest bit in the context id to indicate proxy submission.
- */
- max = MAX_GUC_CONTEXT_HW_ID;
- else
- max = MAX_CONTEXT_HW_ID;
-
- return ida_simple_get(&i915->contexts.hw_ida, 0, max, gfp);
-}
-
-static int steal_hw_id(struct drm_i915_private *i915)
-{
- struct i915_gem_context *ctx, *cn;
- LIST_HEAD(pinned);
- int id = -ENOSPC;
-
- lockdep_assert_held(&i915->contexts.mutex);
-
- list_for_each_entry_safe(ctx, cn,
- &i915->contexts.hw_id_list, hw_id_link) {
- if (atomic_read(&ctx->hw_id_pin_count)) {
- list_move_tail(&ctx->hw_id_link, &pinned);
- continue;
- }
-
- GEM_BUG_ON(!ctx->hw_id); /* perma-pinned kernel context */
- list_del_init(&ctx->hw_id_link);
- id = ctx->hw_id;
- break;
- }
-
- /*
- * Remember how far we got up on the last repossesion scan, so the
- * list is kept in a "least recently scanned" order.
- */
- list_splice_tail(&pinned, &i915->contexts.hw_id_list);
- return id;
-}
-
-static int assign_hw_id(struct drm_i915_private *i915, unsigned int *out)
-{
- int ret;
-
- lockdep_assert_held(&i915->contexts.mutex);
-
- /*
- * We prefer to steal/stall ourselves and our users over that of the
- * entire system. That may be a little unfair to our users, and
- * even hurt high priority clients. The choice is whether to oomkill
- * something else, or steal a context id.
- */
- ret = new_hw_id(i915, GFP_KERNEL | __GFP_RETRY_MAYFAIL | __GFP_NOWARN);
- if (unlikely(ret < 0)) {
- ret = steal_hw_id(i915);
- if (ret < 0) /* once again for the correct errno code */
- ret = new_hw_id(i915, GFP_KERNEL);
- if (ret < 0)
- return ret;
- }
-
- *out = ret;
- return 0;
-}
-
-static void release_hw_id(struct i915_gem_context *ctx)
-{
- struct drm_i915_private *i915 = ctx->i915;
-
- if (list_empty(&ctx->hw_id_link))
- return;
-
- mutex_lock(&i915->contexts.mutex);
- if (!list_empty(&ctx->hw_id_link)) {
- ida_simple_remove(&i915->contexts.hw_ida, ctx->hw_id);
- list_del_init(&ctx->hw_id_link);
- }
- mutex_unlock(&i915->contexts.mutex);
-}
-
static void __free_engines(struct i915_gem_engines *e, unsigned int count)
{
while (count--) {
@@ -313,8 +222,6 @@ static void i915_gem_context_free(struct i915_gem_context *ctx)
lockdep_assert_held(&ctx->i915->drm.struct_mutex);
GEM_BUG_ON(!i915_gem_context_is_closed(ctx));
- release_hw_id(ctx);
-
free_engines(rcu_access_pointer(ctx->engines));
mutex_destroy(&ctx->engines_mutex);
@@ -459,12 +366,6 @@ static void context_close(struct i915_gem_context *ctx)
ctx->file_priv = ERR_PTR(-EBADF);
- /*
- * This context will never again be assinged to HW, so we can
- * reuse its ID for the next context.
- */
- release_hw_id(ctx);
-
/*
* The LUT uses the VMA as a backpointer to unref the object,
* so we need to clear the LUT before we close all the VMA (inside
@@ -507,7 +408,6 @@ __create_context(struct drm_i915_private *i915)
RCU_INIT_POINTER(ctx->engines, e);
INIT_RADIX_TREE(&ctx->handles_vma, GFP_KERNEL);
- INIT_LIST_HEAD(&ctx->hw_id_link);
/* NB: Mark all slices as needing a remap so that when the context first
* loads it will restore whatever remap state already exists. If there
@@ -676,18 +576,11 @@ struct i915_gem_context *
i915_gem_context_create_kernel(struct drm_i915_private *i915, int prio)
{
struct i915_gem_context *ctx;
- int err;
ctx = i915_gem_create_context(i915, 0);
if (IS_ERR(ctx))
return ctx;
- err = i915_gem_context_pin_hw_id(ctx);
- if (err) {
- destroy_kernel_context(&ctx);
- return ERR_PTR(err);
- }
-
i915_gem_context_clear_bannable(ctx);
i915_gem_context_set_persistence(ctx);
ctx->sched.priority = I915_USER_PRIORITY(prio);
@@ -727,15 +620,6 @@ int i915_gem_contexts_init(struct drm_i915_private *dev_priv)
DRM_ERROR("Failed to create default global context\n");
return PTR_ERR(ctx);
}
- /*
- * For easy recognisablity, we want the kernel context to be 0 and then
- * all user contexts will have non-zero hw_id. Kernel contexts are
- * permanently pinned, so that we never suffer a stall and can
- * use them from any allocation context (e.g. for evicting other
- * contexts and from inside the shrinker).
- */
- GEM_BUG_ON(ctx->hw_id);
- GEM_BUG_ON(!atomic_read(&ctx->hw_id_pin_count));
dev_priv->kernel_context = ctx;
DRM_DEBUG_DRIVER("%s context support initialized\n",
@@ -749,10 +633,6 @@ void i915_gem_contexts_fini(struct drm_i915_private *i915)
lockdep_assert_held(&i915->drm.struct_mutex);
destroy_kernel_context(&i915->kernel_context);
-
- /* Must free all deferred contexts (via flush_workqueue) first */
- GEM_BUG_ON(!list_empty(&i915->contexts.hw_id_list));
- ida_destroy(&i915->contexts.hw_ida);
}
static int context_idr_cleanup(int id, void *p, void *data)
@@ -2438,33 +2318,6 @@ int i915_gem_context_reset_stats_ioctl(struct drm_device *dev,
return ret;
}
-int __i915_gem_context_pin_hw_id(struct i915_gem_context *ctx)
-{
- struct drm_i915_private *i915 = ctx->i915;
- int err = 0;
-
- mutex_lock(&i915->contexts.mutex);
-
- GEM_BUG_ON(i915_gem_context_is_closed(ctx));
-
- if (list_empty(&ctx->hw_id_link)) {
- GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count));
-
- err = assign_hw_id(i915, &ctx->hw_id);
- if (err)
- goto out_unlock;
-
- list_add_tail(&ctx->hw_id_link, &i915->contexts.hw_id_list);
- }
-
- GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count) == ~0u);
- atomic_inc(&ctx->hw_id_pin_count);
-
-out_unlock:
- mutex_unlock(&i915->contexts.mutex);
- return err;
-}
-
/* GEM context-engines iterator: for_each_gem_engine() */
struct intel_context *
i915_gem_engines_iter_next(struct i915_gem_engines_iter *it)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.h b/drivers/gpu/drm/i915/gem/i915_gem_context.h
index e0f5b6c6a331..1b0df53436cf 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.h
@@ -127,21 +127,6 @@ 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)
-{
- if (atomic_inc_not_zero(&ctx->hw_id_pin_count))
- return 0;
-
- return __i915_gem_context_pin_hw_id(ctx);
-}
-
-static inline void i915_gem_context_unpin_hw_id(struct i915_gem_context *ctx)
-{
- GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count) == 0u);
- atomic_dec(&ctx->hw_id_pin_count);
-}
-
static inline bool i915_gem_context_is_kernel(struct i915_gem_context *ctx)
{
return !ctx->file_priv;
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 daf1ea5075a6..6419da7c9f90 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
@@ -148,24 +148,6 @@ struct i915_gem_context {
#define CONTEXT_FORCE_SINGLE_SUBMISSION 2
#define CONTEXT_USER_ENGINES 3
- /**
- * @hw_id: - unique identifier for the context
- *
- * The hardware needs to uniquely identify the context for a few
- * functions like fault reporting, PASID, scheduling. The
- * &drm_i915_private.context_hw_ida is used to assign a unqiue
- * id for the lifetime of the context.
- *
- * @hw_id_pin_count: - number of times this context had been pinned
- * for use (should be, at most, once per engine).
- *
- * @hw_id_link: - all contexts with an assigned id are tracked
- * for possible repossession.
- */
- unsigned int hw_id;
- atomic_t hw_id_pin_count;
- struct list_head hw_id_link;
-
struct mutex mutex;
struct i915_sched_attr sched;
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 b39224ed3817..a8a9293f4ac6 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
@@ -446,9 +446,9 @@ static int igt_ctx_exec(void *arg)
err = gpu_fill(ce, obj, dw);
if (err) {
- pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n",
+ pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
ndwords, dw, max_dwords(obj),
- engine->name, ctx->hw_id,
+ engine->name,
yesno(!!ctx->vm), err);
intel_context_put(ce);
kernel_context_close(ctx);
@@ -584,9 +584,9 @@ static int igt_shared_ctx_exec(void *arg)
err = gpu_fill(ce, obj, dw);
if (err) {
- pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n",
+ pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
ndwords, dw, max_dwords(obj),
- engine->name, ctx->hw_id,
+ engine->name,
yesno(!!ctx->vm), err);
intel_context_put(ce);
kernel_context_close(ctx);
@@ -1168,10 +1168,9 @@ static int igt_ctx_readonly(void *arg)
err = gpu_fill(ce, obj, dw);
if (err) {
- pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n",
+ pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
ndwords, dw, max_dwords(obj),
- ce->engine->name, ctx->hw_id,
- yesno(!!ctx->vm), err);
+ ce->engine->name, yesno(!!ctx->vm), err);
i915_gem_context_unlock_engines(ctx);
goto out_unlock;
}
diff --git a/drivers/gpu/drm/i915/gem/selftests/mock_context.c b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
index 0d1a275ec316..ebc46f098561 100644
--- a/drivers/gpu/drm/i915/gem/selftests/mock_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
@@ -13,7 +13,6 @@ mock_context(struct drm_i915_private *i915,
{
struct i915_gem_context *ctx;
struct i915_gem_engines *e;
- int ret;
ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
if (!ctx)
@@ -32,13 +31,8 @@ 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->hw_id_link);
mutex_init(&ctx->mutex);
- ret = i915_gem_context_pin_hw_id(ctx);
- if (ret < 0)
- goto err_engines;
-
if (name) {
struct i915_ppgtt *ppgtt;
@@ -56,8 +50,6 @@ mock_context(struct drm_i915_private *i915,
return ctx;
-err_engines:
- free_engines(rcu_access_pointer(ctx->engines));
err_free:
kfree(ctx);
return NULL;
diff --git a/drivers/gpu/drm/i915/gt/intel_context_types.h b/drivers/gpu/drm/i915/gt/intel_context_types.h
index bf9cedfccbf0..6959b05ae5f8 100644
--- a/drivers/gpu/drm/i915/gt/intel_context_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_context_types.h
@@ -58,6 +58,7 @@ struct intel_context {
u32 *lrc_reg_state;
u64 lrc_desc;
+ u32 tag; /* cookie passed to HW to track this context on submission */
unsigned int active_count; /* protected by timeline->mutex */
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
index da4b45aa84b1..7bce176e696c 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
@@ -296,10 +296,12 @@ struct intel_engine_cs {
u8 uabi_class;
u8 uabi_instance;
+ u32 uabi_capabilities;
u32 context_size;
u32 mmio_base;
- u32 uabi_capabilities;
+ unsigned int context_tag;
+#define NUM_CONTEXT_TAG (256)
struct rb_node uabi_node;
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 0ea2f4d7dc43..af79b8d70800 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -433,12 +433,8 @@ assert_priority_queue(const struct i915_request *prev,
static u64
lrc_descriptor(struct intel_context *ce, struct intel_engine_cs *engine)
{
- struct i915_gem_context *ctx = ce->gem_context;
u64 desc;
- BUILD_BUG_ON(MAX_CONTEXT_HW_ID > (BIT(GEN8_CTX_ID_WIDTH)));
- BUILD_BUG_ON(GEN11_MAX_CONTEXT_HW_ID > (BIT(GEN11_SW_CTX_ID_WIDTH)));
-
desc = INTEL_LEGACY_32B_CONTEXT;
if (i915_vm_is_4lvl(ce->vm))
desc = INTEL_LEGACY_64B_CONTEXT;
@@ -456,20 +452,11 @@ lrc_descriptor(struct intel_context *ce, struct intel_engine_cs *engine)
* anything below.
*/
if (INTEL_GEN(engine->i915) >= 11) {
- GEM_BUG_ON(ctx->hw_id >= BIT(GEN11_SW_CTX_ID_WIDTH));
- desc |= (u64)ctx->hw_id << GEN11_SW_CTX_ID_SHIFT;
- /* bits 37-47 */
-
desc |= (u64)engine->instance << GEN11_ENGINE_INSTANCE_SHIFT;
/* bits 48-53 */
- /* TODO: decide what to do with SW counter (bits 55-60) */
-
desc |= (u64)engine->class << GEN11_ENGINE_CLASS_SHIFT;
/* bits 61-63 */
- } else {
- GEM_BUG_ON(ctx->hw_id >= BIT(GEN8_CTX_ID_WIDTH));
- desc |= (u64)ctx->hw_id << GEN8_CTX_ID_SHIFT; /* bits 32-52 */
}
return desc;
@@ -978,6 +965,15 @@ __execlists_schedule_in(struct i915_request *rq)
intel_context_get(ce);
+ if (ce->tag) {
+ ce->lrc_desc |= (u64)ce->tag << 32;
+ } else {
+ ce->lrc_desc &= ~GENMASK_ULL(47, 37);
+ ce->lrc_desc |=
+ (u64)(engine->context_tag++ % NUM_CONTEXT_TAG) <<
+ GEN11_SW_CTX_ID_SHIFT;
+ }
+
intel_gt_pm_get(engine->gt);
execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN);
intel_engine_context_in(engine);
@@ -2224,7 +2220,6 @@ static void execlists_context_unpin(struct intel_context *ce)
check_redzone((void *)ce->lrc_reg_state - LRC_STATE_PN * PAGE_SIZE,
ce->engine);
- i915_gem_context_unpin_hw_id(ce->gem_context);
i915_gem_object_unpin_map(ce->state->obj);
intel_ring_reset(ce->ring, ce->ring->tail);
}
@@ -2274,18 +2269,12 @@ __execlists_context_pin(struct intel_context *ce,
goto unpin_active;
}
- ret = i915_gem_context_pin_hw_id(ce->gem_context);
- if (ret)
- goto unpin_map;
-
ce->lrc_desc = lrc_descriptor(ce, engine);
ce->lrc_reg_state = vaddr + LRC_STATE_PN * PAGE_SIZE;
__execlists_update_reg_state(ce, engine);
return 0;
-unpin_map:
- i915_gem_object_unpin_map(ce->state->obj);
unpin_active:
intel_context_active_release(ce);
err:
diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
index 343d79c1cb7e..04a5a0d90823 100644
--- a/drivers/gpu/drm/i915/gvt/kvmgt.c
+++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
@@ -1564,27 +1564,10 @@ vgpu_id_show(struct device *dev, struct device_attribute *attr,
return sprintf(buf, "\n");
}
-static ssize_t
-hw_id_show(struct device *dev, struct device_attribute *attr,
- char *buf)
-{
- struct mdev_device *mdev = mdev_from_dev(dev);
-
- if (mdev) {
- struct intel_vgpu *vgpu = (struct intel_vgpu *)
- mdev_get_drvdata(mdev);
- return sprintf(buf, "%u\n",
- vgpu->submission.shadow[0]->gem_context->hw_id);
- }
- return sprintf(buf, "\n");
-}
-
static DEVICE_ATTR_RO(vgpu_id);
-static DEVICE_ATTR_RO(hw_id);
static struct attribute *intel_vgpu_attrs[] = {
&dev_attr_vgpu_id.attr,
- &dev_attr_hw_id.attr,
NULL
};
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index fd41505d52ec..8caaa446490f 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -1504,9 +1504,6 @@ static int i915_context_status(struct seq_file *m, void *unused)
struct intel_context *ce;
seq_puts(m, "HW context ");
- if (!list_empty(&ctx->hw_id_link))
- seq_printf(m, "%x [pin %u]", ctx->hw_id,
- atomic_read(&ctx->hw_id_pin_count));
if (ctx->pid) {
struct task_struct *task;
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index a86d28bda6dd..47239df653f2 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -471,9 +471,9 @@ static void error_print_context(struct drm_i915_error_state_buf *m,
const char *header,
const struct drm_i915_error_context *ctx)
{
- err_printf(m, "%s%s[%d] hw_id %d, prio %d, guilty %d active %d\n",
- header, ctx->comm, ctx->pid, ctx->hw_id,
- ctx->sched_attr.priority, ctx->guilty, ctx->active);
+ err_printf(m, "%s%s[%d] prio %d, guilty %d active %d\n",
+ header, ctx->comm, ctx->pid, ctx->sched_attr.priority,
+ ctx->guilty, ctx->active);
}
static void error_print_engine(struct drm_i915_error_state_buf *m,
@@ -1262,7 +1262,6 @@ static bool record_context(struct drm_i915_error_context *e,
rcu_read_unlock();
}
- e->hw_id = ctx->hw_id;
e->sched_attr = ctx->sched;
e->guilty = atomic_read(&ctx->guilty_count);
e->active = atomic_read(&ctx->active_count);
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.h b/drivers/gpu/drm/i915/i915_gpu_error.h
index caaed5093d95..4dc36d6ee3a2 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.h
+++ b/drivers/gpu/drm/i915/i915_gpu_error.h
@@ -117,7 +117,6 @@ struct i915_gpu_state {
struct drm_i915_error_context {
char comm[TASK_COMM_LEN];
pid_t pid;
- u32 hw_id;
int active;
int guilty;
struct i915_sched_attr sched_attr;
diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index 80055501eccb..d36ba248d438 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -1283,22 +1283,15 @@ static int oa_get_render_ctx_id(struct i915_perf_stream *stream)
} else {
stream->specific_ctx_id_mask =
(1U << GEN8_CTX_ID_WIDTH) - 1;
- stream->specific_ctx_id =
- upper_32_bits(ce->lrc_desc);
- stream->specific_ctx_id &=
- stream->specific_ctx_id_mask;
+ stream->specific_ctx_id = stream->specific_ctx_id_mask;
}
break;
case 11:
case 12: {
stream->specific_ctx_id_mask =
- ((1U << GEN11_SW_CTX_ID_WIDTH) - 1) << (GEN11_SW_CTX_ID_SHIFT - 32) |
- ((1U << GEN11_ENGINE_INSTANCE_WIDTH) - 1) << (GEN11_ENGINE_INSTANCE_SHIFT - 32) |
- ((1 << GEN11_ENGINE_CLASS_WIDTH) - 1) << (GEN11_ENGINE_CLASS_SHIFT - 32);
- stream->specific_ctx_id = upper_32_bits(ce->lrc_desc);
- stream->specific_ctx_id &=
- stream->specific_ctx_id_mask;
+ ((1U << GEN11_SW_CTX_ID_WIDTH) - 1) << (GEN11_SW_CTX_ID_SHIFT - 32);
+ stream->specific_ctx_id = stream->specific_ctx_id_mask;
break;
}
@@ -1306,6 +1299,8 @@ static int oa_get_render_ctx_id(struct i915_perf_stream *stream)
MISSING_CASE(INTEL_GEN(i915));
}
+ ce->tag = stream->specific_ctx_id_mask;
+
DRM_DEBUG_DRIVER("filtering on ctx_id=0x%x ctx_id_mask=0x%x\n",
stream->specific_ctx_id,
stream->specific_ctx_id_mask);
@@ -1324,12 +1319,14 @@ static void oa_put_render_ctx_id(struct i915_perf_stream *stream)
{
struct intel_context *ce;
- stream->specific_ctx_id = INVALID_CTX_ID;
- stream->specific_ctx_id_mask = 0;
-
ce = fetch_and_zero(&stream->pinned_ctx);
- if (ce)
+ if (ce) {
+ ce->tag = 0;
intel_context_unpin(ce);
+ }
+
+ stream->specific_ctx_id = INVALID_CTX_ID;
+ stream->specific_ctx_id_mask = 0;
}
static void
diff --git a/drivers/gpu/drm/i915/i915_trace.h b/drivers/gpu/drm/i915/i915_trace.h
index 24f2944da09d..1f2cf6cfafb5 100644
--- a/drivers/gpu/drm/i915/i915_trace.h
+++ b/drivers/gpu/drm/i915/i915_trace.h
@@ -665,7 +665,6 @@ TRACE_EVENT(i915_request_queue,
TP_STRUCT__entry(
__field(u32, dev)
- __field(u32, hw_id)
__field(u64, ctx)
__field(u16, class)
__field(u16, instance)
@@ -675,7 +674,6 @@ TRACE_EVENT(i915_request_queue,
TP_fast_assign(
__entry->dev = rq->i915->drm.primary->index;
- __entry->hw_id = rq->gem_context->hw_id;
__entry->class = rq->engine->uabi_class;
__entry->instance = rq->engine->uabi_instance;
__entry->ctx = rq->fence.context;
@@ -683,10 +681,9 @@ TRACE_EVENT(i915_request_queue,
__entry->flags = flags;
),
- TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u, flags=0x%x",
+ TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, flags=0x%x",
__entry->dev, __entry->class, __entry->instance,
- __entry->hw_id, __entry->ctx, __entry->seqno,
- __entry->flags)
+ __entry->ctx, __entry->seqno, __entry->flags)
);
DECLARE_EVENT_CLASS(i915_request,
@@ -695,7 +692,6 @@ DECLARE_EVENT_CLASS(i915_request,
TP_STRUCT__entry(
__field(u32, dev)
- __field(u32, hw_id)
__field(u64, ctx)
__field(u16, class)
__field(u16, instance)
@@ -704,16 +700,15 @@ DECLARE_EVENT_CLASS(i915_request,
TP_fast_assign(
__entry->dev = rq->i915->drm.primary->index;
- __entry->hw_id = rq->gem_context->hw_id;
__entry->class = rq->engine->uabi_class;
__entry->instance = rq->engine->uabi_instance;
__entry->ctx = rq->fence.context;
__entry->seqno = rq->fence.seqno;
),
- TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u",
+ TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u",
__entry->dev, __entry->class, __entry->instance,
- __entry->hw_id, __entry->ctx, __entry->seqno)
+ __entry->ctx, __entry->seqno)
);
DEFINE_EVENT(i915_request, i915_request_add,
@@ -738,7 +733,6 @@ TRACE_EVENT(i915_request_in,
TP_STRUCT__entry(
__field(u32, dev)
- __field(u32, hw_id)
__field(u64, ctx)
__field(u16, class)
__field(u16, instance)
@@ -749,7 +743,6 @@ TRACE_EVENT(i915_request_in,
TP_fast_assign(
__entry->dev = rq->i915->drm.primary->index;
- __entry->hw_id = rq->gem_context->hw_id;
__entry->class = rq->engine->uabi_class;
__entry->instance = rq->engine->uabi_instance;
__entry->ctx = rq->fence.context;
@@ -758,9 +751,9 @@ TRACE_EVENT(i915_request_in,
__entry->port = port;
),
- TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u, prio=%u, port=%u",
+ TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, prio=%u, port=%u",
__entry->dev, __entry->class, __entry->instance,
- __entry->hw_id, __entry->ctx, __entry->seqno,
+ __entry->ctx, __entry->seqno,
__entry->prio, __entry->port)
);
@@ -770,7 +763,6 @@ TRACE_EVENT(i915_request_out,
TP_STRUCT__entry(
__field(u32, dev)
- __field(u32, hw_id)
__field(u64, ctx)
__field(u16, class)
__field(u16, instance)
@@ -780,7 +772,6 @@ TRACE_EVENT(i915_request_out,
TP_fast_assign(
__entry->dev = rq->i915->drm.primary->index;
- __entry->hw_id = rq->gem_context->hw_id;
__entry->class = rq->engine->uabi_class;
__entry->instance = rq->engine->uabi_instance;
__entry->ctx = rq->fence.context;
@@ -788,10 +779,9 @@ TRACE_EVENT(i915_request_out,
__entry->completed = i915_request_completed(rq);
),
- TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u, completed?=%u",
+ TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, completed?=%u",
__entry->dev, __entry->class, __entry->instance,
- __entry->hw_id, __entry->ctx, __entry->seqno,
- __entry->completed)
+ __entry->ctx, __entry->seqno, __entry->completed)
);
#else
@@ -829,7 +819,6 @@ TRACE_EVENT(i915_request_wait_begin,
TP_STRUCT__entry(
__field(u32, dev)
- __field(u32, hw_id)
__field(u64, ctx)
__field(u16, class)
__field(u16, instance)
@@ -845,7 +834,6 @@ TRACE_EVENT(i915_request_wait_begin,
*/
TP_fast_assign(
__entry->dev = rq->i915->drm.primary->index;
- __entry->hw_id = rq->gem_context->hw_id;
__entry->class = rq->engine->uabi_class;
__entry->instance = rq->engine->uabi_instance;
__entry->ctx = rq->fence.context;
@@ -853,9 +841,9 @@ TRACE_EVENT(i915_request_wait_begin,
__entry->flags = flags;
),
- TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u, flags=0x%x",
+ TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, flags=0x%x",
__entry->dev, __entry->class, __entry->instance,
- __entry->hw_id, __entry->ctx, __entry->seqno,
+ __entry->ctx, __entry->seqno,
__entry->flags)
);
@@ -958,19 +946,17 @@ DECLARE_EVENT_CLASS(i915_context,
TP_STRUCT__entry(
__field(u32, dev)
__field(struct i915_gem_context *, ctx)
- __field(u32, hw_id)
__field(struct i915_address_space *, vm)
),
TP_fast_assign(
__entry->dev = ctx->i915->drm.primary->index;
__entry->ctx = ctx;
- __entry->hw_id = ctx->hw_id;
__entry->vm = ctx->vm;
),
- TP_printk("dev=%u, ctx=%p, ctx_vm=%p, hw_id=%u",
- __entry->dev, __entry->ctx, __entry->vm, __entry->hw_id)
+ TP_printk("dev=%u, ctx=%p, ctx_vm=%p",
+ __entry->dev, __entry->ctx, __entry->vm)
)
DEFINE_EVENT(i915_context, i915_context_create,
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
index 52d2df843148..f39f0282e78c 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
@@ -491,8 +491,8 @@ static int igt_evict_contexts(void *arg)
if (IS_ERR(rq)) {
/* When full, fail_if_busy will trigger EBUSY */
if (PTR_ERR(rq) != -EBUSY) {
- pr_err("Unexpected error from request alloc (ctx hw id %u, on %s): %d\n",
- ctx->hw_id, engine->name,
+ pr_err("Unexpected error from request alloc (on %s): %d\n",
+ engine->name,
(int)PTR_ERR(rq));
err = PTR_ERR(rq);
}
diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c
index 1c9db08f7c28..ac1ff558eb90 100644
--- a/drivers/gpu/drm/i915/selftests/i915_vma.c
+++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
@@ -170,7 +170,7 @@ static int igt_vma_create(void *arg)
}
nc = 0;
- for_each_prime_number(num_ctx, MAX_CONTEXT_HW_ID) {
+ for_each_prime_number(num_ctx, 2 * NUM_CONTEXT_TAG) {
for (; nc < num_ctx; nc++) {
ctx = mock_context(i915, "mock");
if (!ctx)
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* Re: [PATCH 20/27] drm/i915: Remove logical HW ID
2019-09-25 10:01 ` [PATCH 20/27] drm/i915: Remove logical HW ID Chris Wilson
@ 2019-09-25 12:41 ` Tvrtko Ursulin
2019-09-25 12:51 ` Chris Wilson
2019-09-25 17:59 ` Daniele Ceraolo Spurio
0 siblings, 2 replies; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-25 12:41 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
[+ Daniele, I think he might want to have a look at this.]
On 25/09/2019 11:01, Chris Wilson wrote:
> With the introduction of ctx->engines[] we allow multiple logical
> contexts to be used on the same engine (e.g. with virtual engines). Each
> logical context requires a unique tag in order for context-switching to
> occur correctly between them.
>
> We only need to keep a unique tag for the active lifetime of the
> context, and for as long as we need to identify that context. The HW
> uses the tag to determine if it should use a lite-restore (why not the
> LRCA?) and passes the tag back for various status identifies. The only
> status we need to track is for OA, so when using perf, we assign the
> specific context a unique tag.
>
> Fixes: 976b55f0e1db ("drm/i915: Allow a context to define its set of engines")
Fixes? Why?
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/gem/i915_gem_context.c | 147 ------------------
> drivers/gpu/drm/i915/gem/i915_gem_context.h | 15 --
> .../gpu/drm/i915/gem/i915_gem_context_types.h | 18 ---
> .../drm/i915/gem/selftests/i915_gem_context.c | 13 +-
> .../gpu/drm/i915/gem/selftests/mock_context.c | 8 -
> drivers/gpu/drm/i915/gt/intel_context_types.h | 1 +
> drivers/gpu/drm/i915/gt/intel_engine_types.h | 4 +-
> drivers/gpu/drm/i915/gt/intel_lrc.c | 29 ++--
> drivers/gpu/drm/i915/gvt/kvmgt.c | 17 --
> drivers/gpu/drm/i915/i915_debugfs.c | 3 -
> drivers/gpu/drm/i915/i915_gpu_error.c | 7 +-
> drivers/gpu/drm/i915/i915_gpu_error.h | 1 -
> drivers/gpu/drm/i915/i915_perf.c | 25 ++-
> drivers/gpu/drm/i915/i915_trace.h | 38 ++---
> .../gpu/drm/i915/selftests/i915_gem_evict.c | 4 +-
> drivers/gpu/drm/i915/selftests/i915_vma.c | 2 +-
> 16 files changed, 48 insertions(+), 284 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
> index e832238a72e5..4e59b809d901 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
> @@ -168,97 +168,6 @@ lookup_user_engine(struct i915_gem_context *ctx,
> return i915_gem_context_get_engine(ctx, idx);
> }
>
> -static inline int new_hw_id(struct drm_i915_private *i915, gfp_t gfp)
> -{
> - unsigned int max;
> -
> - lockdep_assert_held(&i915->contexts.mutex);
> -
> - if (INTEL_GEN(i915) >= 12)
> - max = GEN12_MAX_CONTEXT_HW_ID;
> - else if (INTEL_GEN(i915) >= 11)
> - max = GEN11_MAX_CONTEXT_HW_ID;
> - else if (USES_GUC_SUBMISSION(i915))
> - /*
> - * When using GuC in proxy submission, GuC consumes the
> - * highest bit in the context id to indicate proxy submission.
> - */
> - max = MAX_GUC_CONTEXT_HW_ID;
> - else
> - max = MAX_CONTEXT_HW_ID;
> -
> - return ida_simple_get(&i915->contexts.hw_ida, 0, max, gfp);
> -}
> -
> -static int steal_hw_id(struct drm_i915_private *i915)
> -{
> - struct i915_gem_context *ctx, *cn;
> - LIST_HEAD(pinned);
> - int id = -ENOSPC;
> -
> - lockdep_assert_held(&i915->contexts.mutex);
> -
> - list_for_each_entry_safe(ctx, cn,
> - &i915->contexts.hw_id_list, hw_id_link) {
> - if (atomic_read(&ctx->hw_id_pin_count)) {
> - list_move_tail(&ctx->hw_id_link, &pinned);
> - continue;
> - }
> -
> - GEM_BUG_ON(!ctx->hw_id); /* perma-pinned kernel context */
> - list_del_init(&ctx->hw_id_link);
> - id = ctx->hw_id;
> - break;
> - }
> -
> - /*
> - * Remember how far we got up on the last repossesion scan, so the
> - * list is kept in a "least recently scanned" order.
> - */
> - list_splice_tail(&pinned, &i915->contexts.hw_id_list);
> - return id;
> -}
> -
> -static int assign_hw_id(struct drm_i915_private *i915, unsigned int *out)
> -{
> - int ret;
> -
> - lockdep_assert_held(&i915->contexts.mutex);
> -
> - /*
> - * We prefer to steal/stall ourselves and our users over that of the
> - * entire system. That may be a little unfair to our users, and
> - * even hurt high priority clients. The choice is whether to oomkill
> - * something else, or steal a context id.
> - */
> - ret = new_hw_id(i915, GFP_KERNEL | __GFP_RETRY_MAYFAIL | __GFP_NOWARN);
> - if (unlikely(ret < 0)) {
> - ret = steal_hw_id(i915);
> - if (ret < 0) /* once again for the correct errno code */
> - ret = new_hw_id(i915, GFP_KERNEL);
> - if (ret < 0)
> - return ret;
> - }
> -
> - *out = ret;
> - return 0;
> -}
> -
> -static void release_hw_id(struct i915_gem_context *ctx)
> -{
> - struct drm_i915_private *i915 = ctx->i915;
> -
> - if (list_empty(&ctx->hw_id_link))
> - return;
> -
> - mutex_lock(&i915->contexts.mutex);
> - if (!list_empty(&ctx->hw_id_link)) {
> - ida_simple_remove(&i915->contexts.hw_ida, ctx->hw_id);
> - list_del_init(&ctx->hw_id_link);
> - }
> - mutex_unlock(&i915->contexts.mutex);
> -}
> -
> static void __free_engines(struct i915_gem_engines *e, unsigned int count)
> {
> while (count--) {
> @@ -313,8 +222,6 @@ static void i915_gem_context_free(struct i915_gem_context *ctx)
> lockdep_assert_held(&ctx->i915->drm.struct_mutex);
> GEM_BUG_ON(!i915_gem_context_is_closed(ctx));
>
> - release_hw_id(ctx);
> -
> free_engines(rcu_access_pointer(ctx->engines));
> mutex_destroy(&ctx->engines_mutex);
>
> @@ -459,12 +366,6 @@ static void context_close(struct i915_gem_context *ctx)
>
> ctx->file_priv = ERR_PTR(-EBADF);
>
> - /*
> - * This context will never again be assinged to HW, so we can
> - * reuse its ID for the next context.
> - */
> - release_hw_id(ctx);
> -
> /*
> * The LUT uses the VMA as a backpointer to unref the object,
> * so we need to clear the LUT before we close all the VMA (inside
> @@ -507,7 +408,6 @@ __create_context(struct drm_i915_private *i915)
> RCU_INIT_POINTER(ctx->engines, e);
>
> INIT_RADIX_TREE(&ctx->handles_vma, GFP_KERNEL);
> - INIT_LIST_HEAD(&ctx->hw_id_link);
>
> /* NB: Mark all slices as needing a remap so that when the context first
> * loads it will restore whatever remap state already exists. If there
> @@ -676,18 +576,11 @@ struct i915_gem_context *
> i915_gem_context_create_kernel(struct drm_i915_private *i915, int prio)
> {
> struct i915_gem_context *ctx;
> - int err;
>
> ctx = i915_gem_create_context(i915, 0);
> if (IS_ERR(ctx))
> return ctx;
>
> - err = i915_gem_context_pin_hw_id(ctx);
> - if (err) {
> - destroy_kernel_context(&ctx);
> - return ERR_PTR(err);
> - }
> -
> i915_gem_context_clear_bannable(ctx);
> i915_gem_context_set_persistence(ctx);
> ctx->sched.priority = I915_USER_PRIORITY(prio);
> @@ -727,15 +620,6 @@ int i915_gem_contexts_init(struct drm_i915_private *dev_priv)
> DRM_ERROR("Failed to create default global context\n");
> return PTR_ERR(ctx);
> }
> - /*
> - * For easy recognisablity, we want the kernel context to be 0 and then
> - * all user contexts will have non-zero hw_id. Kernel contexts are
> - * permanently pinned, so that we never suffer a stall and can
> - * use them from any allocation context (e.g. for evicting other
> - * contexts and from inside the shrinker).
> - */
> - GEM_BUG_ON(ctx->hw_id);
> - GEM_BUG_ON(!atomic_read(&ctx->hw_id_pin_count));
> dev_priv->kernel_context = ctx;
>
> DRM_DEBUG_DRIVER("%s context support initialized\n",
> @@ -749,10 +633,6 @@ void i915_gem_contexts_fini(struct drm_i915_private *i915)
> lockdep_assert_held(&i915->drm.struct_mutex);
>
> destroy_kernel_context(&i915->kernel_context);
> -
> - /* Must free all deferred contexts (via flush_workqueue) first */
> - GEM_BUG_ON(!list_empty(&i915->contexts.hw_id_list));
> - ida_destroy(&i915->contexts.hw_ida);
> }
>
> static int context_idr_cleanup(int id, void *p, void *data)
> @@ -2438,33 +2318,6 @@ int i915_gem_context_reset_stats_ioctl(struct drm_device *dev,
> return ret;
> }
>
> -int __i915_gem_context_pin_hw_id(struct i915_gem_context *ctx)
> -{
> - struct drm_i915_private *i915 = ctx->i915;
> - int err = 0;
> -
> - mutex_lock(&i915->contexts.mutex);
> -
> - GEM_BUG_ON(i915_gem_context_is_closed(ctx));
> -
> - if (list_empty(&ctx->hw_id_link)) {
> - GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count));
> -
> - err = assign_hw_id(i915, &ctx->hw_id);
> - if (err)
> - goto out_unlock;
> -
> - list_add_tail(&ctx->hw_id_link, &i915->contexts.hw_id_list);
> - }
> -
> - GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count) == ~0u);
> - atomic_inc(&ctx->hw_id_pin_count);
> -
> -out_unlock:
> - mutex_unlock(&i915->contexts.mutex);
> - return err;
> -}
> -
> /* GEM context-engines iterator: for_each_gem_engine() */
> struct intel_context *
> i915_gem_engines_iter_next(struct i915_gem_engines_iter *it)
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.h b/drivers/gpu/drm/i915/gem/i915_gem_context.h
> index e0f5b6c6a331..1b0df53436cf 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_context.h
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.h
> @@ -127,21 +127,6 @@ 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)
> -{
> - if (atomic_inc_not_zero(&ctx->hw_id_pin_count))
> - return 0;
> -
> - return __i915_gem_context_pin_hw_id(ctx);
> -}
> -
> -static inline void i915_gem_context_unpin_hw_id(struct i915_gem_context *ctx)
> -{
> - GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count) == 0u);
> - atomic_dec(&ctx->hw_id_pin_count);
> -}
> -
> static inline bool i915_gem_context_is_kernel(struct i915_gem_context *ctx)
> {
> return !ctx->file_priv;
> 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 daf1ea5075a6..6419da7c9f90 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
> @@ -148,24 +148,6 @@ struct i915_gem_context {
> #define CONTEXT_FORCE_SINGLE_SUBMISSION 2
> #define CONTEXT_USER_ENGINES 3
>
> - /**
> - * @hw_id: - unique identifier for the context
> - *
> - * The hardware needs to uniquely identify the context for a few
> - * functions like fault reporting, PASID, scheduling. The
> - * &drm_i915_private.context_hw_ida is used to assign a unqiue
> - * id for the lifetime of the context.
> - *
> - * @hw_id_pin_count: - number of times this context had been pinned
> - * for use (should be, at most, once per engine).
> - *
> - * @hw_id_link: - all contexts with an assigned id are tracked
> - * for possible repossession.
> - */
> - unsigned int hw_id;
> - atomic_t hw_id_pin_count;
> - struct list_head hw_id_link;
> -
> struct mutex mutex;
>
> struct i915_sched_attr sched;
> 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 b39224ed3817..a8a9293f4ac6 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
> @@ -446,9 +446,9 @@ static int igt_ctx_exec(void *arg)
>
> err = gpu_fill(ce, obj, dw);
> if (err) {
> - pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n",
> + pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
> ndwords, dw, max_dwords(obj),
> - engine->name, ctx->hw_id,
> + engine->name,
> yesno(!!ctx->vm), err);
> intel_context_put(ce);
> kernel_context_close(ctx);
> @@ -584,9 +584,9 @@ static int igt_shared_ctx_exec(void *arg)
>
> err = gpu_fill(ce, obj, dw);
> if (err) {
> - pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n",
> + pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
> ndwords, dw, max_dwords(obj),
> - engine->name, ctx->hw_id,
> + engine->name,
> yesno(!!ctx->vm), err);
> intel_context_put(ce);
> kernel_context_close(ctx);
> @@ -1168,10 +1168,9 @@ static int igt_ctx_readonly(void *arg)
>
> err = gpu_fill(ce, obj, dw);
> if (err) {
> - pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n",
> + pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
> ndwords, dw, max_dwords(obj),
> - ce->engine->name, ctx->hw_id,
> - yesno(!!ctx->vm), err);
> + ce->engine->name, yesno(!!ctx->vm), err);
> i915_gem_context_unlock_engines(ctx);
> goto out_unlock;
> }
> diff --git a/drivers/gpu/drm/i915/gem/selftests/mock_context.c b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
> index 0d1a275ec316..ebc46f098561 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/mock_context.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
> @@ -13,7 +13,6 @@ mock_context(struct drm_i915_private *i915,
> {
> struct i915_gem_context *ctx;
> struct i915_gem_engines *e;
> - int ret;
>
> ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
> if (!ctx)
> @@ -32,13 +31,8 @@ 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->hw_id_link);
> mutex_init(&ctx->mutex);
>
> - ret = i915_gem_context_pin_hw_id(ctx);
> - if (ret < 0)
> - goto err_engines;
> -
> if (name) {
> struct i915_ppgtt *ppgtt;
>
> @@ -56,8 +50,6 @@ mock_context(struct drm_i915_private *i915,
>
> return ctx;
>
> -err_engines:
> - free_engines(rcu_access_pointer(ctx->engines));
> err_free:
> kfree(ctx);
> return NULL;
> diff --git a/drivers/gpu/drm/i915/gt/intel_context_types.h b/drivers/gpu/drm/i915/gt/intel_context_types.h
> index bf9cedfccbf0..6959b05ae5f8 100644
> --- a/drivers/gpu/drm/i915/gt/intel_context_types.h
> +++ b/drivers/gpu/drm/i915/gt/intel_context_types.h
> @@ -58,6 +58,7 @@ struct intel_context {
>
> u32 *lrc_reg_state;
> u64 lrc_desc;
> + u32 tag; /* cookie passed to HW to track this context on submission */
>
> unsigned int active_count; /* protected by timeline->mutex */
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
> index da4b45aa84b1..7bce176e696c 100644
> --- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
> +++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
> @@ -296,10 +296,12 @@ struct intel_engine_cs {
> u8 uabi_class;
> u8 uabi_instance;
>
> + u32 uabi_capabilities;
> u32 context_size;
> u32 mmio_base;
>
> - u32 uabi_capabilities;
> + unsigned int context_tag;
> +#define NUM_CONTEXT_TAG (256)
AFAICT this define now defines how deep ELSP we can have before we start
getting hard to debug problems. GuC angle needs considering here.
Can we add some asserts somewhere to express the relationship?
And a comment here explaining this, or with the assert.
>
> struct rb_node uabi_node;
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
> index 0ea2f4d7dc43..af79b8d70800 100644
> --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
> +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
> @@ -433,12 +433,8 @@ assert_priority_queue(const struct i915_request *prev,
> static u64
> lrc_descriptor(struct intel_context *ce, struct intel_engine_cs *engine)
> {
> - struct i915_gem_context *ctx = ce->gem_context;
> u64 desc;
>
> - BUILD_BUG_ON(MAX_CONTEXT_HW_ID > (BIT(GEN8_CTX_ID_WIDTH)));
> - BUILD_BUG_ON(GEN11_MAX_CONTEXT_HW_ID > (BIT(GEN11_SW_CTX_ID_WIDTH)));
> -
> desc = INTEL_LEGACY_32B_CONTEXT;
> if (i915_vm_is_4lvl(ce->vm))
> desc = INTEL_LEGACY_64B_CONTEXT;
> @@ -456,20 +452,11 @@ lrc_descriptor(struct intel_context *ce, struct intel_engine_cs *engine)
> * anything below.
> */
> if (INTEL_GEN(engine->i915) >= 11) {
> - GEM_BUG_ON(ctx->hw_id >= BIT(GEN11_SW_CTX_ID_WIDTH));
> - desc |= (u64)ctx->hw_id << GEN11_SW_CTX_ID_SHIFT;
> - /* bits 37-47 */
> -
> desc |= (u64)engine->instance << GEN11_ENGINE_INSTANCE_SHIFT;
> /* bits 48-53 */
>
> - /* TODO: decide what to do with SW counter (bits 55-60) */
> -
> desc |= (u64)engine->class << GEN11_ENGINE_CLASS_SHIFT;
> /* bits 61-63 */
> - } else {
> - GEM_BUG_ON(ctx->hw_id >= BIT(GEN8_CTX_ID_WIDTH));
> - desc |= (u64)ctx->hw_id << GEN8_CTX_ID_SHIFT; /* bits 32-52 */
> }
>
> return desc;
> @@ -978,6 +965,15 @@ __execlists_schedule_in(struct i915_request *rq)
>
> intel_context_get(ce);
>
> + if (ce->tag) {
> + ce->lrc_desc |= (u64)ce->tag << 32;
> + } else {
> + ce->lrc_desc &= ~GENMASK_ULL(47, 37);
> + ce->lrc_desc |=
> + (u64)(engine->context_tag++ % NUM_CONTEXT_TAG) <<
> + GEN11_SW_CTX_ID_SHIFT;
> + }
So hw id is valid only while context is in ELSP and it changes with
every submission except in the OA case?
Shifts and masks look dodgy. For >= gen11 the current code has the hw_id
in [37, 42], otherwise [32, 52]. This patch does not seem to handle the
differences between gens.
> +
> intel_gt_pm_get(engine->gt);
> execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN);
> intel_engine_context_in(engine);
> @@ -2224,7 +2220,6 @@ static void execlists_context_unpin(struct intel_context *ce)
> check_redzone((void *)ce->lrc_reg_state - LRC_STATE_PN * PAGE_SIZE,
> ce->engine);
>
> - i915_gem_context_unpin_hw_id(ce->gem_context);
> i915_gem_object_unpin_map(ce->state->obj);
> intel_ring_reset(ce->ring, ce->ring->tail);
> }
> @@ -2274,18 +2269,12 @@ __execlists_context_pin(struct intel_context *ce,
> goto unpin_active;
> }
>
> - ret = i915_gem_context_pin_hw_id(ce->gem_context);
> - if (ret)
> - goto unpin_map;
> -
> ce->lrc_desc = lrc_descriptor(ce, engine);
> ce->lrc_reg_state = vaddr + LRC_STATE_PN * PAGE_SIZE;
> __execlists_update_reg_state(ce, engine);
>
> return 0;
>
> -unpin_map:
> - i915_gem_object_unpin_map(ce->state->obj);
> unpin_active:
> intel_context_active_release(ce);
> err:
> diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
> index 343d79c1cb7e..04a5a0d90823 100644
> --- a/drivers/gpu/drm/i915/gvt/kvmgt.c
> +++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
> @@ -1564,27 +1564,10 @@ vgpu_id_show(struct device *dev, struct device_attribute *attr,
> return sprintf(buf, "\n");
> }
>
> -static ssize_t
> -hw_id_show(struct device *dev, struct device_attribute *attr,
> - char *buf)
> -{
> - struct mdev_device *mdev = mdev_from_dev(dev);
> -
> - if (mdev) {
> - struct intel_vgpu *vgpu = (struct intel_vgpu *)
> - mdev_get_drvdata(mdev);
> - return sprintf(buf, "%u\n",
> - vgpu->submission.shadow[0]->gem_context->hw_id);
> - }
> - return sprintf(buf, "\n");
> -}
> -
> static DEVICE_ATTR_RO(vgpu_id);
> -static DEVICE_ATTR_RO(hw_id);
>
> static struct attribute *intel_vgpu_attrs[] = {
> &dev_attr_vgpu_id.attr,
> - &dev_attr_hw_id.attr,
> NULL
> };
>
> diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
> index fd41505d52ec..8caaa446490f 100644
> --- a/drivers/gpu/drm/i915/i915_debugfs.c
> +++ b/drivers/gpu/drm/i915/i915_debugfs.c
> @@ -1504,9 +1504,6 @@ static int i915_context_status(struct seq_file *m, void *unused)
> struct intel_context *ce;
>
> seq_puts(m, "HW context ");
> - if (!list_empty(&ctx->hw_id_link))
> - seq_printf(m, "%x [pin %u]", ctx->hw_id,
> - atomic_read(&ctx->hw_id_pin_count));
> if (ctx->pid) {
> struct task_struct *task;
>
> diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
> index a86d28bda6dd..47239df653f2 100644
> --- a/drivers/gpu/drm/i915/i915_gpu_error.c
> +++ b/drivers/gpu/drm/i915/i915_gpu_error.c
> @@ -471,9 +471,9 @@ static void error_print_context(struct drm_i915_error_state_buf *m,
> const char *header,
> const struct drm_i915_error_context *ctx)
> {
> - err_printf(m, "%s%s[%d] hw_id %d, prio %d, guilty %d active %d\n",
> - header, ctx->comm, ctx->pid, ctx->hw_id,
> - ctx->sched_attr.priority, ctx->guilty, ctx->active);
> + err_printf(m, "%s%s[%d] prio %d, guilty %d active %d\n",
> + header, ctx->comm, ctx->pid, ctx->sched_attr.priority,
> + ctx->guilty, ctx->active);
> }
>
> static void error_print_engine(struct drm_i915_error_state_buf *m,
> @@ -1262,7 +1262,6 @@ static bool record_context(struct drm_i915_error_context *e,
> rcu_read_unlock();
> }
>
> - e->hw_id = ctx->hw_id;
> e->sched_attr = ctx->sched;
> e->guilty = atomic_read(&ctx->guilty_count);
> e->active = atomic_read(&ctx->active_count);
> diff --git a/drivers/gpu/drm/i915/i915_gpu_error.h b/drivers/gpu/drm/i915/i915_gpu_error.h
> index caaed5093d95..4dc36d6ee3a2 100644
> --- a/drivers/gpu/drm/i915/i915_gpu_error.h
> +++ b/drivers/gpu/drm/i915/i915_gpu_error.h
> @@ -117,7 +117,6 @@ struct i915_gpu_state {
> struct drm_i915_error_context {
> char comm[TASK_COMM_LEN];
> pid_t pid;
> - u32 hw_id;
> int active;
> int guilty;
> struct i915_sched_attr sched_attr;
> diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
> index 80055501eccb..d36ba248d438 100644
> --- a/drivers/gpu/drm/i915/i915_perf.c
> +++ b/drivers/gpu/drm/i915/i915_perf.c
> @@ -1283,22 +1283,15 @@ static int oa_get_render_ctx_id(struct i915_perf_stream *stream)
> } else {
> stream->specific_ctx_id_mask =
> (1U << GEN8_CTX_ID_WIDTH) - 1;
> - stream->specific_ctx_id =
> - upper_32_bits(ce->lrc_desc);
> - stream->specific_ctx_id &=
> - stream->specific_ctx_id_mask;
> + stream->specific_ctx_id = stream->specific_ctx_id_mask;
> }
> break;
>
> case 11:
> case 12: {
> stream->specific_ctx_id_mask =
> - ((1U << GEN11_SW_CTX_ID_WIDTH) - 1) << (GEN11_SW_CTX_ID_SHIFT - 32) |
> - ((1U << GEN11_ENGINE_INSTANCE_WIDTH) - 1) << (GEN11_ENGINE_INSTANCE_SHIFT - 32) |
> - ((1 << GEN11_ENGINE_CLASS_WIDTH) - 1) << (GEN11_ENGINE_CLASS_SHIFT - 32);
> - stream->specific_ctx_id = upper_32_bits(ce->lrc_desc);
> - stream->specific_ctx_id &=
> - stream->specific_ctx_id_mask;
> + ((1U << GEN11_SW_CTX_ID_WIDTH) - 1) << (GEN11_SW_CTX_ID_SHIFT - 32);
> + stream->specific_ctx_id = stream->specific_ctx_id_mask;
> break;
> }
>
> @@ -1306,6 +1299,8 @@ static int oa_get_render_ctx_id(struct i915_perf_stream *stream)
> MISSING_CASE(INTEL_GEN(i915));
> }
>
> + ce->tag = stream->specific_ctx_id_mask;
> +
> DRM_DEBUG_DRIVER("filtering on ctx_id=0x%x ctx_id_mask=0x%x\n",
> stream->specific_ctx_id,
> stream->specific_ctx_id_mask);
> @@ -1324,12 +1319,14 @@ static void oa_put_render_ctx_id(struct i915_perf_stream *stream)
> {
> struct intel_context *ce;
>
> - stream->specific_ctx_id = INVALID_CTX_ID;
> - stream->specific_ctx_id_mask = 0;
> -
> ce = fetch_and_zero(&stream->pinned_ctx);
> - if (ce)
> + if (ce) {
> + ce->tag = 0;
> intel_context_unpin(ce);
> + }
> +
> + stream->specific_ctx_id = INVALID_CTX_ID;
> + stream->specific_ctx_id_mask = 0;
OA hunks looks dodgy as well. ce->tag is set to
stream->specific_ctx_id_mask while I think it should be id. Furthermore
id is assigned from mask which is fixed and not unique for different
contexts.
> }
>
> static void
> diff --git a/drivers/gpu/drm/i915/i915_trace.h b/drivers/gpu/drm/i915/i915_trace.h
> index 24f2944da09d..1f2cf6cfafb5 100644
> --- a/drivers/gpu/drm/i915/i915_trace.h
> +++ b/drivers/gpu/drm/i915/i915_trace.h
> @@ -665,7 +665,6 @@ TRACE_EVENT(i915_request_queue,
>
> TP_STRUCT__entry(
> __field(u32, dev)
> - __field(u32, hw_id)
> __field(u64, ctx)
> __field(u16, class)
> __field(u16, instance)
> @@ -675,7 +674,6 @@ TRACE_EVENT(i915_request_queue,
>
> TP_fast_assign(
> __entry->dev = rq->i915->drm.primary->index;
> - __entry->hw_id = rq->gem_context->hw_id;
> __entry->class = rq->engine->uabi_class;
> __entry->instance = rq->engine->uabi_instance;
> __entry->ctx = rq->fence.context;
> @@ -683,10 +681,9 @@ TRACE_EVENT(i915_request_queue,
> __entry->flags = flags;
> ),
>
> - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u, flags=0x%x",
> + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, flags=0x%x",
> __entry->dev, __entry->class, __entry->instance,
> - __entry->hw_id, __entry->ctx, __entry->seqno,
> - __entry->flags)
> + __entry->ctx, __entry->seqno, __entry->flags)
> );
>
> DECLARE_EVENT_CLASS(i915_request,
> @@ -695,7 +692,6 @@ DECLARE_EVENT_CLASS(i915_request,
>
> TP_STRUCT__entry(
> __field(u32, dev)
> - __field(u32, hw_id)
> __field(u64, ctx)
> __field(u16, class)
> __field(u16, instance)
> @@ -704,16 +700,15 @@ DECLARE_EVENT_CLASS(i915_request,
>
> TP_fast_assign(
> __entry->dev = rq->i915->drm.primary->index;
> - __entry->hw_id = rq->gem_context->hw_id;
> __entry->class = rq->engine->uabi_class;
> __entry->instance = rq->engine->uabi_instance;
> __entry->ctx = rq->fence.context;
> __entry->seqno = rq->fence.seqno;
> ),
>
> - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u",
> + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u",
> __entry->dev, __entry->class, __entry->instance,
> - __entry->hw_id, __entry->ctx, __entry->seqno)
> + __entry->ctx, __entry->seqno)
> );
>
> DEFINE_EVENT(i915_request, i915_request_add,
> @@ -738,7 +733,6 @@ TRACE_EVENT(i915_request_in,
>
> TP_STRUCT__entry(
> __field(u32, dev)
> - __field(u32, hw_id)
> __field(u64, ctx)
> __field(u16, class)
> __field(u16, instance)
> @@ -749,7 +743,6 @@ TRACE_EVENT(i915_request_in,
>
> TP_fast_assign(
> __entry->dev = rq->i915->drm.primary->index;
> - __entry->hw_id = rq->gem_context->hw_id;
> __entry->class = rq->engine->uabi_class;
> __entry->instance = rq->engine->uabi_instance;
> __entry->ctx = rq->fence.context;
> @@ -758,9 +751,9 @@ TRACE_EVENT(i915_request_in,
> __entry->port = port;
> ),
>
> - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u, prio=%u, port=%u",
> + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, prio=%u, port=%u",
> __entry->dev, __entry->class, __entry->instance,
> - __entry->hw_id, __entry->ctx, __entry->seqno,
> + __entry->ctx, __entry->seqno,
> __entry->prio, __entry->port)
> );
>
> @@ -770,7 +763,6 @@ TRACE_EVENT(i915_request_out,
>
> TP_STRUCT__entry(
> __field(u32, dev)
> - __field(u32, hw_id)
> __field(u64, ctx)
> __field(u16, class)
> __field(u16, instance)
> @@ -780,7 +772,6 @@ TRACE_EVENT(i915_request_out,
>
> TP_fast_assign(
> __entry->dev = rq->i915->drm.primary->index;
> - __entry->hw_id = rq->gem_context->hw_id;
> __entry->class = rq->engine->uabi_class;
> __entry->instance = rq->engine->uabi_instance;
> __entry->ctx = rq->fence.context;
> @@ -788,10 +779,9 @@ TRACE_EVENT(i915_request_out,
> __entry->completed = i915_request_completed(rq);
> ),
>
> - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u, completed?=%u",
> + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, completed?=%u",
> __entry->dev, __entry->class, __entry->instance,
> - __entry->hw_id, __entry->ctx, __entry->seqno,
> - __entry->completed)
> + __entry->ctx, __entry->seqno, __entry->completed)
> );
>
> #else
> @@ -829,7 +819,6 @@ TRACE_EVENT(i915_request_wait_begin,
>
> TP_STRUCT__entry(
> __field(u32, dev)
> - __field(u32, hw_id)
> __field(u64, ctx)
> __field(u16, class)
> __field(u16, instance)
> @@ -845,7 +834,6 @@ TRACE_EVENT(i915_request_wait_begin,
> */
> TP_fast_assign(
> __entry->dev = rq->i915->drm.primary->index;
> - __entry->hw_id = rq->gem_context->hw_id;
> __entry->class = rq->engine->uabi_class;
> __entry->instance = rq->engine->uabi_instance;
> __entry->ctx = rq->fence.context;
> @@ -853,9 +841,9 @@ TRACE_EVENT(i915_request_wait_begin,
> __entry->flags = flags;
> ),
>
> - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u, flags=0x%x",
> + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, flags=0x%x",
> __entry->dev, __entry->class, __entry->instance,
> - __entry->hw_id, __entry->ctx, __entry->seqno,
> + __entry->ctx, __entry->seqno,
> __entry->flags)
> );
>
> @@ -958,19 +946,17 @@ DECLARE_EVENT_CLASS(i915_context,
> TP_STRUCT__entry(
> __field(u32, dev)
> __field(struct i915_gem_context *, ctx)
> - __field(u32, hw_id)
> __field(struct i915_address_space *, vm)
> ),
>
> TP_fast_assign(
> __entry->dev = ctx->i915->drm.primary->index;
> __entry->ctx = ctx;
> - __entry->hw_id = ctx->hw_id;
> __entry->vm = ctx->vm;
> ),
>
> - TP_printk("dev=%u, ctx=%p, ctx_vm=%p, hw_id=%u",
> - __entry->dev, __entry->ctx, __entry->vm, __entry->hw_id)
> + TP_printk("dev=%u, ctx=%p, ctx_vm=%p",
> + __entry->dev, __entry->ctx, __entry->vm)
> )
>
> DEFINE_EVENT(i915_context, i915_context_create,
> diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
> index 52d2df843148..f39f0282e78c 100644
> --- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
> +++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
> @@ -491,8 +491,8 @@ static int igt_evict_contexts(void *arg)
> if (IS_ERR(rq)) {
> /* When full, fail_if_busy will trigger EBUSY */
> if (PTR_ERR(rq) != -EBUSY) {
> - pr_err("Unexpected error from request alloc (ctx hw id %u, on %s): %d\n",
> - ctx->hw_id, engine->name,
> + pr_err("Unexpected error from request alloc (on %s): %d\n",
> + engine->name,
> (int)PTR_ERR(rq));
> err = PTR_ERR(rq);
> }
> diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c
> index 1c9db08f7c28..ac1ff558eb90 100644
> --- a/drivers/gpu/drm/i915/selftests/i915_vma.c
> +++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
> @@ -170,7 +170,7 @@ static int igt_vma_create(void *arg)
> }
>
> nc = 0;
> - for_each_prime_number(num_ctx, MAX_CONTEXT_HW_ID) {
> + for_each_prime_number(num_ctx, 2 * NUM_CONTEXT_TAG) {
> for (; nc < num_ctx; nc++) {
> ctx = mock_context(i915, "mock");
> if (!ctx)
>
Unless I am missing something I see this patch as simplification and not
a fix. If we can get away with it in the GuC world then it's quite a big
simplification so it's fine by me.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* Re: [PATCH 20/27] drm/i915: Remove logical HW ID
2019-09-25 12:41 ` Tvrtko Ursulin
@ 2019-09-25 12:51 ` Chris Wilson
2019-09-25 14:38 ` Tvrtko Ursulin
2019-09-25 17:59 ` Daniele Ceraolo Spurio
1 sibling, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 12:51 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Tvrtko Ursulin (2019-09-25 13:41:10)
> [+ Daniele, I think he might want to have a look at this.]
>
> On 25/09/2019 11:01, Chris Wilson wrote:
> > With the introduction of ctx->engines[] we allow multiple logical
> > contexts to be used on the same engine (e.g. with virtual engines). Each
> > logical context requires a unique tag in order for context-switching to
> > occur correctly between them.
> >
> > We only need to keep a unique tag for the active lifetime of the
> > context, and for as long as we need to identify that context. The HW
> > uses the tag to determine if it should use a lite-restore (why not the
> > LRCA?) and passes the tag back for various status identifies. The only
> > status we need to track is for OA, so when using perf, we assign the
> > specific context a unique tag.
> >
> > Fixes: 976b55f0e1db ("drm/i915: Allow a context to define its set of engines")
>
> Fixes? Why?
The above paragraph explains that as we give two distinct contexts the
same tag then it will only perform (according to bspec) a lite-restore.
-Chris
> > diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
> > index da4b45aa84b1..7bce176e696c 100644
> > --- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
> > +++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
> > @@ -296,10 +296,12 @@ struct intel_engine_cs {
> > u8 uabi_class;
> > u8 uabi_instance;
> >
> > + u32 uabi_capabilities;
> > u32 context_size;
> > u32 mmio_base;
> >
> > - u32 uabi_capabilities;
> > + unsigned int context_tag;
> > +#define NUM_CONTEXT_TAG (256)
>
> AFAICT this define now defines how deep ELSP we can have before we start
> getting hard to debug problems. GuC angle needs considering here.
The guc is immaterial here, since they use their own mechanism. It just
needs to be a value larger than 2 * EXECLIST_MAX_PORTS and less then
BIT(10).
> > @@ -978,6 +965,15 @@ __execlists_schedule_in(struct i915_request *rq)
> >
> > intel_context_get(ce);
> >
> > + if (ce->tag) {
> > + ce->lrc_desc |= (u64)ce->tag << 32;
> > + } else {
> > + ce->lrc_desc &= ~GENMASK_ULL(47, 37);
> > + ce->lrc_desc |=
> > + (u64)(engine->context_tag++ % NUM_CONTEXT_TAG) <<
> > + GEN11_SW_CTX_ID_SHIFT;
> > + }
>
> So hw id is valid only while context is in ELSP and it changes with
> every submission except in the OA case?
Aye. OA being a pita.
> Shifts and masks look dodgy. For >= gen11 the current code has the hw_id
> in [37, 42], otherwise [32, 52]. This patch does not seem to handle the
> differences between gens.
Because the values we supply do not matter (they are cookies), just the
lifetime.
> > +
> > intel_gt_pm_get(engine->gt);
> > execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN);
> > intel_engine_context_in(engine);
> > @@ -2224,7 +2220,6 @@ static void execlists_context_unpin(struct intel_context *ce)
> > check_redzone((void *)ce->lrc_reg_state - LRC_STATE_PN * PAGE_SIZE,
> > ce->engine);
> >
> > - i915_gem_context_unpin_hw_id(ce->gem_context);
> > i915_gem_object_unpin_map(ce->state->obj);
> > intel_ring_reset(ce->ring, ce->ring->tail);
> > }
> > @@ -2274,18 +2269,12 @@ __execlists_context_pin(struct intel_context *ce,
> > goto unpin_active;
> > }
> >
> > - ret = i915_gem_context_pin_hw_id(ce->gem_context);
> > - if (ret)
> > - goto unpin_map;
> > -
> > ce->lrc_desc = lrc_descriptor(ce, engine);
> > ce->lrc_reg_state = vaddr + LRC_STATE_PN * PAGE_SIZE;
> > __execlists_update_reg_state(ce, engine);
> >
> > return 0;
> >
> > -unpin_map:
> > - i915_gem_object_unpin_map(ce->state->obj);
> > unpin_active:
> > intel_context_active_release(ce);
> > err:
> > diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
> > index 343d79c1cb7e..04a5a0d90823 100644
> > --- a/drivers/gpu/drm/i915/gvt/kvmgt.c
> > +++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
> > @@ -1564,27 +1564,10 @@ vgpu_id_show(struct device *dev, struct device_attribute *attr,
> > return sprintf(buf, "\n");
> > }
> >
> > -static ssize_t
> > -hw_id_show(struct device *dev, struct device_attribute *attr,
> > - char *buf)
> > -{
> > - struct mdev_device *mdev = mdev_from_dev(dev);
> > -
> > - if (mdev) {
> > - struct intel_vgpu *vgpu = (struct intel_vgpu *)
> > - mdev_get_drvdata(mdev);
> > - return sprintf(buf, "%u\n",
> > - vgpu->submission.shadow[0]->gem_context->hw_id);
> > - }
> > - return sprintf(buf, "\n");
> > -}
> > -
> > static DEVICE_ATTR_RO(vgpu_id);
> > -static DEVICE_ATTR_RO(hw_id);
> >
> > static struct attribute *intel_vgpu_attrs[] = {
> > &dev_attr_vgpu_id.attr,
> > - &dev_attr_hw_id.attr,
> > NULL
> > };
> >
> > diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
> > index fd41505d52ec..8caaa446490f 100644
> > --- a/drivers/gpu/drm/i915/i915_debugfs.c
> > +++ b/drivers/gpu/drm/i915/i915_debugfs.c
> > @@ -1504,9 +1504,6 @@ static int i915_context_status(struct seq_file *m, void *unused)
> > struct intel_context *ce;
> >
> > seq_puts(m, "HW context ");
> > - if (!list_empty(&ctx->hw_id_link))
> > - seq_printf(m, "%x [pin %u]", ctx->hw_id,
> > - atomic_read(&ctx->hw_id_pin_count));
> > if (ctx->pid) {
> > struct task_struct *task;
> >
> > diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
> > index a86d28bda6dd..47239df653f2 100644
> > --- a/drivers/gpu/drm/i915/i915_gpu_error.c
> > +++ b/drivers/gpu/drm/i915/i915_gpu_error.c
> > @@ -471,9 +471,9 @@ static void error_print_context(struct drm_i915_error_state_buf *m,
> > const char *header,
> > const struct drm_i915_error_context *ctx)
> > {
> > - err_printf(m, "%s%s[%d] hw_id %d, prio %d, guilty %d active %d\n",
> > - header, ctx->comm, ctx->pid, ctx->hw_id,
> > - ctx->sched_attr.priority, ctx->guilty, ctx->active);
> > + err_printf(m, "%s%s[%d] prio %d, guilty %d active %d\n",
> > + header, ctx->comm, ctx->pid, ctx->sched_attr.priority,
> > + ctx->guilty, ctx->active);
> > }
> >
> > static void error_print_engine(struct drm_i915_error_state_buf *m,
> > @@ -1262,7 +1262,6 @@ static bool record_context(struct drm_i915_error_context *e,
> > rcu_read_unlock();
> > }
> >
> > - e->hw_id = ctx->hw_id;
> > e->sched_attr = ctx->sched;
> > e->guilty = atomic_read(&ctx->guilty_count);
> > e->active = atomic_read(&ctx->active_count);
> > diff --git a/drivers/gpu/drm/i915/i915_gpu_error.h b/drivers/gpu/drm/i915/i915_gpu_error.h
> > index caaed5093d95..4dc36d6ee3a2 100644
> > --- a/drivers/gpu/drm/i915/i915_gpu_error.h
> > +++ b/drivers/gpu/drm/i915/i915_gpu_error.h
> > @@ -117,7 +117,6 @@ struct i915_gpu_state {
> > struct drm_i915_error_context {
> > char comm[TASK_COMM_LEN];
> > pid_t pid;
> > - u32 hw_id;
> > int active;
> > int guilty;
> > struct i915_sched_attr sched_attr;
> > diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
> > index 80055501eccb..d36ba248d438 100644
> > --- a/drivers/gpu/drm/i915/i915_perf.c
> > +++ b/drivers/gpu/drm/i915/i915_perf.c
> > @@ -1283,22 +1283,15 @@ static int oa_get_render_ctx_id(struct i915_perf_stream *stream)
> > } else {
> > stream->specific_ctx_id_mask =
> > (1U << GEN8_CTX_ID_WIDTH) - 1;
> > - stream->specific_ctx_id =
> > - upper_32_bits(ce->lrc_desc);
> > - stream->specific_ctx_id &=
> > - stream->specific_ctx_id_mask;
> > + stream->specific_ctx_id = stream->specific_ctx_id_mask;
> > }
> > break;
> >
> > case 11:
> > case 12: {
> > stream->specific_ctx_id_mask =
> > - ((1U << GEN11_SW_CTX_ID_WIDTH) - 1) << (GEN11_SW_CTX_ID_SHIFT - 32) |
> > - ((1U << GEN11_ENGINE_INSTANCE_WIDTH) - 1) << (GEN11_ENGINE_INSTANCE_SHIFT - 32) |
> > - ((1 << GEN11_ENGINE_CLASS_WIDTH) - 1) << (GEN11_ENGINE_CLASS_SHIFT - 32);
> > - stream->specific_ctx_id = upper_32_bits(ce->lrc_desc);
> > - stream->specific_ctx_id &=
> > - stream->specific_ctx_id_mask;
> > + ((1U << GEN11_SW_CTX_ID_WIDTH) - 1) << (GEN11_SW_CTX_ID_SHIFT - 32);
> > + stream->specific_ctx_id = stream->specific_ctx_id_mask;
> > break;
> > }
> >
> > @@ -1306,6 +1299,8 @@ static int oa_get_render_ctx_id(struct i915_perf_stream *stream)
> > MISSING_CASE(INTEL_GEN(i915));
> > }
> >
> > + ce->tag = stream->specific_ctx_id_mask;
> > +
> > DRM_DEBUG_DRIVER("filtering on ctx_id=0x%x ctx_id_mask=0x%x\n",
> > stream->specific_ctx_id,
> > stream->specific_ctx_id_mask);
> > @@ -1324,12 +1319,14 @@ static void oa_put_render_ctx_id(struct i915_perf_stream *stream)
> > {
> > struct intel_context *ce;
> >
> > - stream->specific_ctx_id = INVALID_CTX_ID;
> > - stream->specific_ctx_id_mask = 0;
> > -
> > ce = fetch_and_zero(&stream->pinned_ctx);
> > - if (ce)
> > + if (ce) {
> > + ce->tag = 0;
> > intel_context_unpin(ce);
> > + }
> > +
> > + stream->specific_ctx_id = INVALID_CTX_ID;
> > + stream->specific_ctx_id_mask = 0;
>
> OA hunks looks dodgy as well. ce->tag is set to
> stream->specific_ctx_id_mask while I think it should be id. Furthermore
> id is assigned from mask which is fixed and not unique for different
> contexts.
It is singular in the entire system. The mask is a guaranteed unique
value.
> > diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c
> > index 1c9db08f7c28..ac1ff558eb90 100644
> > --- a/drivers/gpu/drm/i915/selftests/i915_vma.c
> > +++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
> > @@ -170,7 +170,7 @@ static int igt_vma_create(void *arg)
> > }
> >
> > nc = 0;
> > - for_each_prime_number(num_ctx, MAX_CONTEXT_HW_ID) {
> > + for_each_prime_number(num_ctx, 2 * NUM_CONTEXT_TAG) {
> > for (; nc < num_ctx; nc++) {
> > ctx = mock_context(i915, "mock");
> > if (!ctx)
> >
>
> Unless I am missing something I see this patch as simplification and not
> a fix. If we can get away with it in the GuC world then it's quite a big
> simplification so it's fine by me.
Bspec says that for lrc_desc with matching tags (at least for gen8-gen11
I believe), it performs a lite restore (regardless of lrca). Ergo it is
quite a major fix.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* Re: [PATCH 20/27] drm/i915: Remove logical HW ID
2019-09-25 12:51 ` Chris Wilson
@ 2019-09-25 14:38 ` Tvrtko Ursulin
0 siblings, 0 replies; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-25 14:38 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 25/09/2019 13:51, Chris Wilson wrote:
> Quoting Tvrtko Ursulin (2019-09-25 13:41:10)
>> [+ Daniele, I think he might want to have a look at this.]
>>
>> On 25/09/2019 11:01, Chris Wilson wrote:
>>> With the introduction of ctx->engines[] we allow multiple logical
>>> contexts to be used on the same engine (e.g. with virtual engines). Each
>>> logical context requires a unique tag in order for context-switching to
>>> occur correctly between them.
>>>
>>> We only need to keep a unique tag for the active lifetime of the
>>> context, and for as long as we need to identify that context. The HW
>>> uses the tag to determine if it should use a lite-restore (why not the
>>> LRCA?) and passes the tag back for various status identifies. The only
>>> status we need to track is for OA, so when using perf, we assign the
>>> specific context a unique tag.
>>>
>>> Fixes: 976b55f0e1db ("drm/i915: Allow a context to define its set of engines")
>>
>> Fixes? Why?
>
> The above paragraph explains that as we give two distinct contexts the
> same tag then it will only perform (according to bspec) a lite-restore.
Right I see. But not with virtual engines per se, but even if someone
created an engine map with two same engines in it. Virtual engines are
then a super set of this problem.
>>> diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
>>> index da4b45aa84b1..7bce176e696c 100644
>>> --- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
>>> +++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
>>> @@ -296,10 +296,12 @@ struct intel_engine_cs {
>>> u8 uabi_class;
>>> u8 uabi_instance;
>>>
>>> + u32 uabi_capabilities;
>>> u32 context_size;
>>> u32 mmio_base;
>>>
>>> - u32 uabi_capabilities;
>>> + unsigned int context_tag;
>>> +#define NUM_CONTEXT_TAG (256)
>>
>> AFAICT this define now defines how deep ELSP we can have before we start
>> getting hard to debug problems. GuC angle needs considering here.
>
> The guc is immaterial here, since they use their own mechanism. It just
> needs to be a value larger than 2 * EXECLIST_MAX_PORTS and less then
> BIT(10).
I don't know how GuC (will) work so hopefully Daniele can confirm or
deny this won't be a problem there.
>
>>> @@ -978,6 +965,15 @@ __execlists_schedule_in(struct i915_request *rq)
>>>
>>> intel_context_get(ce);
>>>
>>> + if (ce->tag) {
>>> + ce->lrc_desc |= (u64)ce->tag << 32;
>>> + } else {
>>> + ce->lrc_desc &= ~GENMASK_ULL(47, 37);
>>> + ce->lrc_desc |=
>>> + (u64)(engine->context_tag++ % NUM_CONTEXT_TAG) <<
>>> + GEN11_SW_CTX_ID_SHIFT;
>>> + }
>>
>> So hw id is valid only while context is in ELSP and it changes with
>> every submission except in the OA case?
>
> Aye. OA being a pita.
>
>> Shifts and masks look dodgy. For >= gen11 the current code has the hw_id
>> in [37, 42], otherwise [32, 52]. This patch does not seem to handle the
>> differences between gens.
>
> Because the values we supply do not matter (they are cookies), just the
> lifetime.
Okay so you are choosing to not use some of the available bits since we
don't need all 20 (pre gen11), or even 10 (gen11+). This definitely
deserves a comment. And another assert with NUM_CONTEXT_TAG relating to
available bits.
>
>>> +
>>> intel_gt_pm_get(engine->gt);
>>> execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN);
>>> intel_engine_context_in(engine);
>>> @@ -2224,7 +2220,6 @@ static void execlists_context_unpin(struct intel_context *ce)
>>> check_redzone((void *)ce->lrc_reg_state - LRC_STATE_PN * PAGE_SIZE,
>>> ce->engine);
>>>
>>> - i915_gem_context_unpin_hw_id(ce->gem_context);
>>> i915_gem_object_unpin_map(ce->state->obj);
>>> intel_ring_reset(ce->ring, ce->ring->tail);
>>> }
>>> @@ -2274,18 +2269,12 @@ __execlists_context_pin(struct intel_context *ce,
>>> goto unpin_active;
>>> }
>>>
>>> - ret = i915_gem_context_pin_hw_id(ce->gem_context);
>>> - if (ret)
>>> - goto unpin_map;
>>> -
>>> ce->lrc_desc = lrc_descriptor(ce, engine);
>>> ce->lrc_reg_state = vaddr + LRC_STATE_PN * PAGE_SIZE;
>>> __execlists_update_reg_state(ce, engine);
>>>
>>> return 0;
>>>
>>> -unpin_map:
>>> - i915_gem_object_unpin_map(ce->state->obj);
>>> unpin_active:
>>> intel_context_active_release(ce);
>>> err:
>>> diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
>>> index 343d79c1cb7e..04a5a0d90823 100644
>>> --- a/drivers/gpu/drm/i915/gvt/kvmgt.c
>>> +++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
>>> @@ -1564,27 +1564,10 @@ vgpu_id_show(struct device *dev, struct device_attribute *attr,
>>> return sprintf(buf, "\n");
>>> }
>>>
>>> -static ssize_t
>>> -hw_id_show(struct device *dev, struct device_attribute *attr,
>>> - char *buf)
>>> -{
>>> - struct mdev_device *mdev = mdev_from_dev(dev);
>>> -
>>> - if (mdev) {
>>> - struct intel_vgpu *vgpu = (struct intel_vgpu *)
>>> - mdev_get_drvdata(mdev);
>>> - return sprintf(buf, "%u\n",
>>> - vgpu->submission.shadow[0]->gem_context->hw_id);
>>> - }
>>> - return sprintf(buf, "\n");
>>> -}
>>> -
>>> static DEVICE_ATTR_RO(vgpu_id);
>>> -static DEVICE_ATTR_RO(hw_id);
>>>
>>> static struct attribute *intel_vgpu_attrs[] = {
>>> &dev_attr_vgpu_id.attr,
>>> - &dev_attr_hw_id.attr,
>>> NULL
>>> };
>>>
>>> diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
>>> index fd41505d52ec..8caaa446490f 100644
>>> --- a/drivers/gpu/drm/i915/i915_debugfs.c
>>> +++ b/drivers/gpu/drm/i915/i915_debugfs.c
>>> @@ -1504,9 +1504,6 @@ static int i915_context_status(struct seq_file *m, void *unused)
>>> struct intel_context *ce;
>>>
>>> seq_puts(m, "HW context ");
>>> - if (!list_empty(&ctx->hw_id_link))
>>> - seq_printf(m, "%x [pin %u]", ctx->hw_id,
>>> - atomic_read(&ctx->hw_id_pin_count));
>>> if (ctx->pid) {
>>> struct task_struct *task;
>>>
>>> diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
>>> index a86d28bda6dd..47239df653f2 100644
>>> --- a/drivers/gpu/drm/i915/i915_gpu_error.c
>>> +++ b/drivers/gpu/drm/i915/i915_gpu_error.c
>>> @@ -471,9 +471,9 @@ static void error_print_context(struct drm_i915_error_state_buf *m,
>>> const char *header,
>>> const struct drm_i915_error_context *ctx)
>>> {
>>> - err_printf(m, "%s%s[%d] hw_id %d, prio %d, guilty %d active %d\n",
>>> - header, ctx->comm, ctx->pid, ctx->hw_id,
>>> - ctx->sched_attr.priority, ctx->guilty, ctx->active);
>>> + err_printf(m, "%s%s[%d] prio %d, guilty %d active %d\n",
>>> + header, ctx->comm, ctx->pid, ctx->sched_attr.priority,
>>> + ctx->guilty, ctx->active);
>>> }
>>>
>>> static void error_print_engine(struct drm_i915_error_state_buf *m,
>>> @@ -1262,7 +1262,6 @@ static bool record_context(struct drm_i915_error_context *e,
>>> rcu_read_unlock();
>>> }
>>>
>>> - e->hw_id = ctx->hw_id;
>>> e->sched_attr = ctx->sched;
>>> e->guilty = atomic_read(&ctx->guilty_count);
>>> e->active = atomic_read(&ctx->active_count);
>>> diff --git a/drivers/gpu/drm/i915/i915_gpu_error.h b/drivers/gpu/drm/i915/i915_gpu_error.h
>>> index caaed5093d95..4dc36d6ee3a2 100644
>>> --- a/drivers/gpu/drm/i915/i915_gpu_error.h
>>> +++ b/drivers/gpu/drm/i915/i915_gpu_error.h
>>> @@ -117,7 +117,6 @@ struct i915_gpu_state {
>>> struct drm_i915_error_context {
>>> char comm[TASK_COMM_LEN];
>>> pid_t pid;
>>> - u32 hw_id;
>>> int active;
>>> int guilty;
>>> struct i915_sched_attr sched_attr;
>>> diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
>>> index 80055501eccb..d36ba248d438 100644
>>> --- a/drivers/gpu/drm/i915/i915_perf.c
>>> +++ b/drivers/gpu/drm/i915/i915_perf.c
>>> @@ -1283,22 +1283,15 @@ static int oa_get_render_ctx_id(struct i915_perf_stream *stream)
>>> } else {
>>> stream->specific_ctx_id_mask =
>>> (1U << GEN8_CTX_ID_WIDTH) - 1;
>>> - stream->specific_ctx_id =
>>> - upper_32_bits(ce->lrc_desc);
>>> - stream->specific_ctx_id &=
>>> - stream->specific_ctx_id_mask;
>>> + stream->specific_ctx_id = stream->specific_ctx_id_mask;
>>> }
>>> break;
>>>
>>> case 11:
>>> case 12: {
>>> stream->specific_ctx_id_mask =
>>> - ((1U << GEN11_SW_CTX_ID_WIDTH) - 1) << (GEN11_SW_CTX_ID_SHIFT - 32) |
>>> - ((1U << GEN11_ENGINE_INSTANCE_WIDTH) - 1) << (GEN11_ENGINE_INSTANCE_SHIFT - 32) |
>>> - ((1 << GEN11_ENGINE_CLASS_WIDTH) - 1) << (GEN11_ENGINE_CLASS_SHIFT - 32);
>>> - stream->specific_ctx_id = upper_32_bits(ce->lrc_desc);
>>> - stream->specific_ctx_id &=
>>> - stream->specific_ctx_id_mask;
>>> + ((1U << GEN11_SW_CTX_ID_WIDTH) - 1) << (GEN11_SW_CTX_ID_SHIFT - 32);
>>> + stream->specific_ctx_id = stream->specific_ctx_id_mask;
>>> break;
>>> }
>>>
>>> @@ -1306,6 +1299,8 @@ static int oa_get_render_ctx_id(struct i915_perf_stream *stream)
>>> MISSING_CASE(INTEL_GEN(i915));
>>> }
>>>
>>> + ce->tag = stream->specific_ctx_id_mask;
>>> +
>>> DRM_DEBUG_DRIVER("filtering on ctx_id=0x%x ctx_id_mask=0x%x\n",
>>> stream->specific_ctx_id,
>>> stream->specific_ctx_id_mask);
>>> @@ -1324,12 +1319,14 @@ static void oa_put_render_ctx_id(struct i915_perf_stream *stream)
>>> {
>>> struct intel_context *ce;
>>>
>>> - stream->specific_ctx_id = INVALID_CTX_ID;
>>> - stream->specific_ctx_id_mask = 0;
>>> -
>>> ce = fetch_and_zero(&stream->pinned_ctx);
>>> - if (ce)
>>> + if (ce) {
>>> + ce->tag = 0;
>>> intel_context_unpin(ce);
>>> + }
>>> +
>>> + stream->specific_ctx_id = INVALID_CTX_ID;
>>> + stream->specific_ctx_id_mask = 0;
>>
>> OA hunks looks dodgy as well. ce->tag is set to
>> stream->specific_ctx_id_mask while I think it should be id. Furthermore
>> id is assigned from mask which is fixed and not unique for different
>> contexts.
>
> It is singular in the entire system. The mask is a guaranteed unique
> value.
What is singular? The OA context and it's id? I assumed what needs to be
done is fix hw_id for _all_ contexts while OA is sampling. But on a more
detailed look this assumption does not seem to hold in code. Okay, I
declare to not know how OA works well enough.
>
>>> diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c
>>> index 1c9db08f7c28..ac1ff558eb90 100644
>>> --- a/drivers/gpu/drm/i915/selftests/i915_vma.c
>>> +++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
>>> @@ -170,7 +170,7 @@ static int igt_vma_create(void *arg)
>>> }
>>>
>>> nc = 0;
>>> - for_each_prime_number(num_ctx, MAX_CONTEXT_HW_ID) {
>>> + for_each_prime_number(num_ctx, 2 * NUM_CONTEXT_TAG) {
>>> for (; nc < num_ctx; nc++) {
>>> ctx = mock_context(i915, "mock");
>>> if (!ctx)
>>>
>>
>> Unless I am missing something I see this patch as simplification and not
>> a fix. If we can get away with it in the GuC world then it's quite a big
>> simplification so it's fine by me.
>
> Bspec says that for lrc_desc with matching tags (at least for gen8-gen11
> I believe), it performs a lite restore (regardless of lrca). Ergo it is
> quite a major fix.
Yes I see that now, my bad. This one will not lend itself well to
backporting.. But I don't think there is a lighter-weight solution.
Moving hw_id to intel_context would also end up large-ish I think.
Because of the locking requirement etc.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* Re: [PATCH 20/27] drm/i915: Remove logical HW ID
2019-09-25 12:41 ` Tvrtko Ursulin
2019-09-25 12:51 ` Chris Wilson
@ 2019-09-25 17:59 ` Daniele Ceraolo Spurio
2019-09-25 18:23 ` Matthew Brost
1 sibling, 1 reply; 70+ messages in thread
From: Daniele Ceraolo Spurio @ 2019-09-25 17:59 UTC (permalink / raw)
To: Tvrtko Ursulin, Chris Wilson, intel-gfx, Matthew Brost
+ Matt
On 9/25/19 5:41 AM, Tvrtko Ursulin wrote:
> [+ Daniele, I think he might want to have a look at this.]
>
> On 25/09/2019 11:01, Chris Wilson wrote:
>> With the introduction of ctx->engines[] we allow multiple logical
>> contexts to be used on the same engine (e.g. with virtual engines). Each
>> logical context requires a unique tag in order for context-switching to
>> occur correctly between them.
>>
>> We only need to keep a unique tag for the active lifetime of the
>> context, and for as long as we need to identify that context. The HW
>> uses the tag to determine if it should use a lite-restore (why not the
>> LRCA?) and passes the tag back for various status identifies. The only
>> status we need to track is for OA, so when using perf, we assign the
>> specific context a unique tag.
>>
>> Fixes: 976b55f0e1db ("drm/i915: Allow a context to define its set of
>> engines")
>
> Fixes? Why?
>
>> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
>> ---
>> drivers/gpu/drm/i915/gem/i915_gem_context.c | 147 ------------------
>> drivers/gpu/drm/i915/gem/i915_gem_context.h | 15 --
>> .../gpu/drm/i915/gem/i915_gem_context_types.h | 18 ---
>> .../drm/i915/gem/selftests/i915_gem_context.c | 13 +-
>> .../gpu/drm/i915/gem/selftests/mock_context.c | 8 -
>> drivers/gpu/drm/i915/gt/intel_context_types.h | 1 +
>> drivers/gpu/drm/i915/gt/intel_engine_types.h | 4 +-
>> drivers/gpu/drm/i915/gt/intel_lrc.c | 29 ++--
>> drivers/gpu/drm/i915/gvt/kvmgt.c | 17 --
>> drivers/gpu/drm/i915/i915_debugfs.c | 3 -
>> drivers/gpu/drm/i915/i915_gpu_error.c | 7 +-
>> drivers/gpu/drm/i915/i915_gpu_error.h | 1 -
>> drivers/gpu/drm/i915/i915_perf.c | 25 ++-
>> drivers/gpu/drm/i915/i915_trace.h | 38 ++---
>> .../gpu/drm/i915/selftests/i915_gem_evict.c | 4 +-
>> drivers/gpu/drm/i915/selftests/i915_vma.c | 2 +-
>> 16 files changed, 48 insertions(+), 284 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c
>> b/drivers/gpu/drm/i915/gem/i915_gem_context.c
>> index e832238a72e5..4e59b809d901 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
>> @@ -168,97 +168,6 @@ lookup_user_engine(struct i915_gem_context *ctx,
>> return i915_gem_context_get_engine(ctx, idx);
>> }
>> -static inline int new_hw_id(struct drm_i915_private *i915, gfp_t gfp)
>> -{
>> - unsigned int max;
>> -
>> - lockdep_assert_held(&i915->contexts.mutex);
>> -
>> - if (INTEL_GEN(i915) >= 12)
>> - max = GEN12_MAX_CONTEXT_HW_ID;
>> - else if (INTEL_GEN(i915) >= 11)
>> - max = GEN11_MAX_CONTEXT_HW_ID;
>> - else if (USES_GUC_SUBMISSION(i915))
>> - /*
>> - * When using GuC in proxy submission, GuC consumes the
>> - * highest bit in the context id to indicate proxy submission.
>> - */
>> - max = MAX_GUC_CONTEXT_HW_ID;
>> - else
>> - max = MAX_CONTEXT_HW_ID;
>> -
>> - return ida_simple_get(&i915->contexts.hw_ida, 0, max, gfp);
>> -}
>> -
>> -static int steal_hw_id(struct drm_i915_private *i915)
>> -{
>> - struct i915_gem_context *ctx, *cn;
>> - LIST_HEAD(pinned);
>> - int id = -ENOSPC;
>> -
>> - lockdep_assert_held(&i915->contexts.mutex);
>> -
>> - list_for_each_entry_safe(ctx, cn,
>> - &i915->contexts.hw_id_list, hw_id_link) {
>> - if (atomic_read(&ctx->hw_id_pin_count)) {
>> - list_move_tail(&ctx->hw_id_link, &pinned);
>> - continue;
>> - }
>> -
>> - GEM_BUG_ON(!ctx->hw_id); /* perma-pinned kernel context */
>> - list_del_init(&ctx->hw_id_link);
>> - id = ctx->hw_id;
>> - break;
>> - }
>> -
>> - /*
>> - * Remember how far we got up on the last repossesion scan, so the
>> - * list is kept in a "least recently scanned" order.
>> - */
>> - list_splice_tail(&pinned, &i915->contexts.hw_id_list);
>> - return id;
>> -}
>> -
>> -static int assign_hw_id(struct drm_i915_private *i915, unsigned int
>> *out)
>> -{
>> - int ret;
>> -
>> - lockdep_assert_held(&i915->contexts.mutex);
>> -
>> - /*
>> - * We prefer to steal/stall ourselves and our users over that of the
>> - * entire system. That may be a little unfair to our users, and
>> - * even hurt high priority clients. The choice is whether to oomkill
>> - * something else, or steal a context id.
>> - */
>> - ret = new_hw_id(i915, GFP_KERNEL | __GFP_RETRY_MAYFAIL |
>> __GFP_NOWARN);
>> - if (unlikely(ret < 0)) {
>> - ret = steal_hw_id(i915);
>> - if (ret < 0) /* once again for the correct errno code */
>> - ret = new_hw_id(i915, GFP_KERNEL);
>> - if (ret < 0)
>> - return ret;
>> - }
>> -
>> - *out = ret;
>> - return 0;
>> -}
>> -
>> -static void release_hw_id(struct i915_gem_context *ctx)
>> -{
>> - struct drm_i915_private *i915 = ctx->i915;
>> -
>> - if (list_empty(&ctx->hw_id_link))
>> - return;
>> -
>> - mutex_lock(&i915->contexts.mutex);
>> - if (!list_empty(&ctx->hw_id_link)) {
>> - ida_simple_remove(&i915->contexts.hw_ida, ctx->hw_id);
>> - list_del_init(&ctx->hw_id_link);
>> - }
>> - mutex_unlock(&i915->contexts.mutex);
>> -}
>> -
>> static void __free_engines(struct i915_gem_engines *e, unsigned int
>> count)
>> {
>> while (count--) {
>> @@ -313,8 +222,6 @@ static void i915_gem_context_free(struct
>> i915_gem_context *ctx)
>> lockdep_assert_held(&ctx->i915->drm.struct_mutex);
>> GEM_BUG_ON(!i915_gem_context_is_closed(ctx));
>> - release_hw_id(ctx);
>> -
>> free_engines(rcu_access_pointer(ctx->engines));
>> mutex_destroy(&ctx->engines_mutex);
>> @@ -459,12 +366,6 @@ static void context_close(struct i915_gem_context
>> *ctx)
>> ctx->file_priv = ERR_PTR(-EBADF);
>> - /*
>> - * This context will never again be assinged to HW, so we can
>> - * reuse its ID for the next context.
>> - */
>> - release_hw_id(ctx);
>> -
>> /*
>> * The LUT uses the VMA as a backpointer to unref the object,
>> * so we need to clear the LUT before we close all the VMA (inside
>> @@ -507,7 +408,6 @@ __create_context(struct drm_i915_private *i915)
>> RCU_INIT_POINTER(ctx->engines, e);
>> INIT_RADIX_TREE(&ctx->handles_vma, GFP_KERNEL);
>> - INIT_LIST_HEAD(&ctx->hw_id_link);
>> /* NB: Mark all slices as needing a remap so that when the
>> context first
>> * loads it will restore whatever remap state already exists. If
>> there
>> @@ -676,18 +576,11 @@ struct i915_gem_context *
>> i915_gem_context_create_kernel(struct drm_i915_private *i915, int prio)
>> {
>> struct i915_gem_context *ctx;
>> - int err;
>> ctx = i915_gem_create_context(i915, 0);
>> if (IS_ERR(ctx))
>> return ctx;
>> - err = i915_gem_context_pin_hw_id(ctx);
>> - if (err) {
>> - destroy_kernel_context(&ctx);
>> - return ERR_PTR(err);
>> - }
>> -
>> i915_gem_context_clear_bannable(ctx);
>> i915_gem_context_set_persistence(ctx);
>> ctx->sched.priority = I915_USER_PRIORITY(prio);
>> @@ -727,15 +620,6 @@ int i915_gem_contexts_init(struct
>> drm_i915_private *dev_priv)
>> DRM_ERROR("Failed to create default global context\n");
>> return PTR_ERR(ctx);
>> }
>> - /*
>> - * For easy recognisablity, we want the kernel context to be 0
>> and then
>> - * all user contexts will have non-zero hw_id. Kernel contexts are
>> - * permanently pinned, so that we never suffer a stall and can
>> - * use them from any allocation context (e.g. for evicting other
>> - * contexts and from inside the shrinker).
>> - */
>> - GEM_BUG_ON(ctx->hw_id);
>> - GEM_BUG_ON(!atomic_read(&ctx->hw_id_pin_count));
>> dev_priv->kernel_context = ctx;
>> DRM_DEBUG_DRIVER("%s context support initialized\n",
>> @@ -749,10 +633,6 @@ void i915_gem_contexts_fini(struct
>> drm_i915_private *i915)
>> lockdep_assert_held(&i915->drm.struct_mutex);
>> destroy_kernel_context(&i915->kernel_context);
>> -
>> - /* Must free all deferred contexts (via flush_workqueue) first */
>> - GEM_BUG_ON(!list_empty(&i915->contexts.hw_id_list));
>> - ida_destroy(&i915->contexts.hw_ida);
>> }
>> static int context_idr_cleanup(int id, void *p, void *data)
>> @@ -2438,33 +2318,6 @@ int i915_gem_context_reset_stats_ioctl(struct
>> drm_device *dev,
>> return ret;
>> }
>> -int __i915_gem_context_pin_hw_id(struct i915_gem_context *ctx)
>> -{
>> - struct drm_i915_private *i915 = ctx->i915;
>> - int err = 0;
>> -
>> - mutex_lock(&i915->contexts.mutex);
>> -
>> - GEM_BUG_ON(i915_gem_context_is_closed(ctx));
>> -
>> - if (list_empty(&ctx->hw_id_link)) {
>> - GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count));
>> -
>> - err = assign_hw_id(i915, &ctx->hw_id);
>> - if (err)
>> - goto out_unlock;
>> -
>> - list_add_tail(&ctx->hw_id_link, &i915->contexts.hw_id_list);
>> - }
>> -
>> - GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count) == ~0u);
>> - atomic_inc(&ctx->hw_id_pin_count);
>> -
>> -out_unlock:
>> - mutex_unlock(&i915->contexts.mutex);
>> - return err;
>> -}
>> -
>> /* GEM context-engines iterator: for_each_gem_engine() */
>> struct intel_context *
>> i915_gem_engines_iter_next(struct i915_gem_engines_iter *it)
>> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.h
>> b/drivers/gpu/drm/i915/gem/i915_gem_context.h
>> index e0f5b6c6a331..1b0df53436cf 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_context.h
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.h
>> @@ -127,21 +127,6 @@ 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)
>> -{
>> - if (atomic_inc_not_zero(&ctx->hw_id_pin_count))
>> - return 0;
>> -
>> - return __i915_gem_context_pin_hw_id(ctx);
>> -}
>> -
>> -static inline void i915_gem_context_unpin_hw_id(struct
>> i915_gem_context *ctx)
>> -{
>> - GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count) == 0u);
>> - atomic_dec(&ctx->hw_id_pin_count);
>> -}
>> -
>> static inline bool i915_gem_context_is_kernel(struct
>> i915_gem_context *ctx)
>> {
>> return !ctx->file_priv;
>> 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 daf1ea5075a6..6419da7c9f90 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
>> @@ -148,24 +148,6 @@ struct i915_gem_context {
>> #define CONTEXT_FORCE_SINGLE_SUBMISSION 2
>> #define CONTEXT_USER_ENGINES 3
>> - /**
>> - * @hw_id: - unique identifier for the context
>> - *
>> - * The hardware needs to uniquely identify the context for a few
>> - * functions like fault reporting, PASID, scheduling. The
>> - * &drm_i915_private.context_hw_ida is used to assign a unqiue
>> - * id for the lifetime of the context.
>> - *
>> - * @hw_id_pin_count: - number of times this context had been pinned
>> - * for use (should be, at most, once per engine).
>> - *
>> - * @hw_id_link: - all contexts with an assigned id are tracked
>> - * for possible repossession.
>> - */
>> - unsigned int hw_id;
>> - atomic_t hw_id_pin_count;
>> - struct list_head hw_id_link;
>> -
>> struct mutex mutex;
>> struct i915_sched_attr sched;
>> 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 b39224ed3817..a8a9293f4ac6 100644
>> --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
>> +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
>> @@ -446,9 +446,9 @@ static int igt_ctx_exec(void *arg)
>> err = gpu_fill(ce, obj, dw);
>> if (err) {
>> - pr_err("Failed to fill dword %lu [%lu/%lu] with gpu
>> (%s) in ctx %u [full-ppgtt? %s], err=%d\n",
>> + pr_err("Failed to fill dword %lu [%lu/%lu] with gpu
>> (%s) [full-ppgtt? %s], err=%d\n",
>> ndwords, dw, max_dwords(obj),
>> - engine->name, ctx->hw_id,
>> + engine->name,
>> yesno(!!ctx->vm), err);
>> intel_context_put(ce);
>> kernel_context_close(ctx);
>> @@ -584,9 +584,9 @@ static int igt_shared_ctx_exec(void *arg)
>> err = gpu_fill(ce, obj, dw);
>> if (err) {
>> - pr_err("Failed to fill dword %lu [%lu/%lu] with gpu
>> (%s) in ctx %u [full-ppgtt? %s], err=%d\n",
>> + pr_err("Failed to fill dword %lu [%lu/%lu] with gpu
>> (%s) [full-ppgtt? %s], err=%d\n",
>> ndwords, dw, max_dwords(obj),
>> - engine->name, ctx->hw_id,
>> + engine->name,
>> yesno(!!ctx->vm), err);
>> intel_context_put(ce);
>> kernel_context_close(ctx);
>> @@ -1168,10 +1168,9 @@ static int igt_ctx_readonly(void *arg)
>> err = gpu_fill(ce, obj, dw);
>> if (err) {
>> - pr_err("Failed to fill dword %lu [%lu/%lu] with gpu
>> (%s) in ctx %u [full-ppgtt? %s], err=%d\n",
>> + pr_err("Failed to fill dword %lu [%lu/%lu] with gpu
>> (%s) [full-ppgtt? %s], err=%d\n",
>> ndwords, dw, max_dwords(obj),
>> - ce->engine->name, ctx->hw_id,
>> - yesno(!!ctx->vm), err);
>> + ce->engine->name, yesno(!!ctx->vm), err);
>> i915_gem_context_unlock_engines(ctx);
>> goto out_unlock;
>> }
>> diff --git a/drivers/gpu/drm/i915/gem/selftests/mock_context.c
>> b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
>> index 0d1a275ec316..ebc46f098561 100644
>> --- a/drivers/gpu/drm/i915/gem/selftests/mock_context.c
>> +++ b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
>> @@ -13,7 +13,6 @@ mock_context(struct drm_i915_private *i915,
>> {
>> struct i915_gem_context *ctx;
>> struct i915_gem_engines *e;
>> - int ret;
>> ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
>> if (!ctx)
>> @@ -32,13 +31,8 @@ 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->hw_id_link);
>> mutex_init(&ctx->mutex);
>> - ret = i915_gem_context_pin_hw_id(ctx);
>> - if (ret < 0)
>> - goto err_engines;
>> -
>> if (name) {
>> struct i915_ppgtt *ppgtt;
>> @@ -56,8 +50,6 @@ mock_context(struct drm_i915_private *i915,
>> return ctx;
>> -err_engines:
>> - free_engines(rcu_access_pointer(ctx->engines));
>> err_free:
>> kfree(ctx);
>> return NULL;
>> diff --git a/drivers/gpu/drm/i915/gt/intel_context_types.h
>> b/drivers/gpu/drm/i915/gt/intel_context_types.h
>> index bf9cedfccbf0..6959b05ae5f8 100644
>> --- a/drivers/gpu/drm/i915/gt/intel_context_types.h
>> +++ b/drivers/gpu/drm/i915/gt/intel_context_types.h
>> @@ -58,6 +58,7 @@ struct intel_context {
>> u32 *lrc_reg_state;
>> u64 lrc_desc;
>> + u32 tag; /* cookie passed to HW to track this context on
>> submission */
>> unsigned int active_count; /* protected by timeline->mutex */
>> diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h
>> b/drivers/gpu/drm/i915/gt/intel_engine_types.h
>> index da4b45aa84b1..7bce176e696c 100644
>> --- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
>> +++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
>> @@ -296,10 +296,12 @@ struct intel_engine_cs {
>> u8 uabi_class;
>> u8 uabi_instance;
>> + u32 uabi_capabilities;
>> u32 context_size;
>> u32 mmio_base;
>> - u32 uabi_capabilities;
>> + unsigned int context_tag;
>> +#define NUM_CONTEXT_TAG (256)
>
> AFAICT this define now defines how deep ELSP we can have before we start
> getting hard to debug problems. GuC angle needs considering here.
>
> Can we add some asserts somewhere to express the relationship?
>
> And a comment here explaining this, or with the assert.
>
>> struct rb_node uabi_node;
>> diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c
>> b/drivers/gpu/drm/i915/gt/intel_lrc.c
>> index 0ea2f4d7dc43..af79b8d70800 100644
>> --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
>> +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
>> @@ -433,12 +433,8 @@ assert_priority_queue(const struct i915_request
>> *prev,
>> static u64
>> lrc_descriptor(struct intel_context *ce, struct intel_engine_cs
>> *engine)
>> {
>> - struct i915_gem_context *ctx = ce->gem_context;
>> u64 desc;
>> - BUILD_BUG_ON(MAX_CONTEXT_HW_ID > (BIT(GEN8_CTX_ID_WIDTH)));
>> - BUILD_BUG_ON(GEN11_MAX_CONTEXT_HW_ID >
>> (BIT(GEN11_SW_CTX_ID_WIDTH)));
>> -
>> desc = INTEL_LEGACY_32B_CONTEXT;
>> if (i915_vm_is_4lvl(ce->vm))
>> desc = INTEL_LEGACY_64B_CONTEXT;
>> @@ -456,20 +452,11 @@ lrc_descriptor(struct intel_context *ce, struct
>> intel_engine_cs *engine)
>> * anything below.
>> */
>> if (INTEL_GEN(engine->i915) >= 11) {
>> - GEM_BUG_ON(ctx->hw_id >= BIT(GEN11_SW_CTX_ID_WIDTH));
>> - desc |= (u64)ctx->hw_id << GEN11_SW_CTX_ID_SHIFT;
>> - /* bits 37-47 */
>> -
>> desc |= (u64)engine->instance << GEN11_ENGINE_INSTANCE_SHIFT;
>> /* bits 48-53 */
>> - /* TODO: decide what to do with SW counter (bits 55-60) */
>> -
>> desc |= (u64)engine->class << GEN11_ENGINE_CLASS_SHIFT;
>> /* bits 61-63 */
>> - } else {
>> - GEM_BUG_ON(ctx->hw_id >= BIT(GEN8_CTX_ID_WIDTH));
>> - desc |= (u64)ctx->hw_id << GEN8_CTX_ID_SHIFT; /* bits
>> 32-52 */
>> }
>> return desc;
>> @@ -978,6 +965,15 @@ __execlists_schedule_in(struct i915_request *rq)
>> intel_context_get(ce);
>> + if (ce->tag) {
>> + ce->lrc_desc |= (u64)ce->tag << 32;
>> + } else {
>> + ce->lrc_desc &= ~GENMASK_ULL(47, 37);
>> + ce->lrc_desc |=
>> + (u64)(engine->context_tag++ % NUM_CONTEXT_TAG) <<
>> + GEN11_SW_CTX_ID_SHIFT;
>> + }
>
> So hw id is valid only while context is in ELSP and it changes with
> every submission except in the OA case?
>
> Shifts and masks look dodgy. For >= gen11 the current code has the hw_id
> in [37, 42], otherwise [32, 52]. This patch does not seem to handle the
> differences between gens.
>
>> +
>> intel_gt_pm_get(engine->gt);
>> execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN);
>> intel_engine_context_in(engine);
>> @@ -2224,7 +2220,6 @@ static void execlists_context_unpin(struct
>> intel_context *ce)
>> check_redzone((void *)ce->lrc_reg_state - LRC_STATE_PN * PAGE_SIZE,
>> ce->engine);
>> - i915_gem_context_unpin_hw_id(ce->gem_context);
>> i915_gem_object_unpin_map(ce->state->obj);
>> intel_ring_reset(ce->ring, ce->ring->tail);
>> }
>> @@ -2274,18 +2269,12 @@ __execlists_context_pin(struct intel_context *ce,
>> goto unpin_active;
>> }
>> - ret = i915_gem_context_pin_hw_id(ce->gem_context);
>> - if (ret)
>> - goto unpin_map;
>> -
>> ce->lrc_desc = lrc_descriptor(ce, engine);
>> ce->lrc_reg_state = vaddr + LRC_STATE_PN * PAGE_SIZE;
>> __execlists_update_reg_state(ce, engine);
>> return 0;
>> -unpin_map:
>> - i915_gem_object_unpin_map(ce->state->obj);
>> unpin_active:
>> intel_context_active_release(ce);
>> err:
>> diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c
>> b/drivers/gpu/drm/i915/gvt/kvmgt.c
>> index 343d79c1cb7e..04a5a0d90823 100644
>> --- a/drivers/gpu/drm/i915/gvt/kvmgt.c
>> +++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
>> @@ -1564,27 +1564,10 @@ vgpu_id_show(struct device *dev, struct
>> device_attribute *attr,
>> return sprintf(buf, "\n");
>> }
>> -static ssize_t
>> -hw_id_show(struct device *dev, struct device_attribute *attr,
>> - char *buf)
>> -{
>> - struct mdev_device *mdev = mdev_from_dev(dev);
>> -
>> - if (mdev) {
>> - struct intel_vgpu *vgpu = (struct intel_vgpu *)
>> - mdev_get_drvdata(mdev);
>> - return sprintf(buf, "%u\n",
>> - vgpu->submission.shadow[0]->gem_context->hw_id);
>> - }
>> - return sprintf(buf, "\n");
>> -}
>> -
>> static DEVICE_ATTR_RO(vgpu_id);
>> -static DEVICE_ATTR_RO(hw_id);
>> static struct attribute *intel_vgpu_attrs[] = {
>> &dev_attr_vgpu_id.attr,
>> - &dev_attr_hw_id.attr,
>> NULL
>> };
>> diff --git a/drivers/gpu/drm/i915/i915_debugfs.c
>> b/drivers/gpu/drm/i915/i915_debugfs.c
>> index fd41505d52ec..8caaa446490f 100644
>> --- a/drivers/gpu/drm/i915/i915_debugfs.c
>> +++ b/drivers/gpu/drm/i915/i915_debugfs.c
>> @@ -1504,9 +1504,6 @@ static int i915_context_status(struct seq_file
>> *m, void *unused)
>> struct intel_context *ce;
>> seq_puts(m, "HW context ");
>> - if (!list_empty(&ctx->hw_id_link))
>> - seq_printf(m, "%x [pin %u]", ctx->hw_id,
>> - atomic_read(&ctx->hw_id_pin_count));
>> if (ctx->pid) {
>> struct task_struct *task;
>> diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c
>> b/drivers/gpu/drm/i915/i915_gpu_error.c
>> index a86d28bda6dd..47239df653f2 100644
>> --- a/drivers/gpu/drm/i915/i915_gpu_error.c
>> +++ b/drivers/gpu/drm/i915/i915_gpu_error.c
>> @@ -471,9 +471,9 @@ static void error_print_context(struct
>> drm_i915_error_state_buf *m,
>> const char *header,
>> const struct drm_i915_error_context *ctx)
>> {
>> - err_printf(m, "%s%s[%d] hw_id %d, prio %d, guilty %d active %d\n",
>> - header, ctx->comm, ctx->pid, ctx->hw_id,
>> - ctx->sched_attr.priority, ctx->guilty, ctx->active);
>> + err_printf(m, "%s%s[%d] prio %d, guilty %d active %d\n",
>> + header, ctx->comm, ctx->pid, ctx->sched_attr.priority,
>> + ctx->guilty, ctx->active);
>> }
>> static void error_print_engine(struct drm_i915_error_state_buf *m,
>> @@ -1262,7 +1262,6 @@ static bool record_context(struct
>> drm_i915_error_context *e,
>> rcu_read_unlock();
>> }
>> - e->hw_id = ctx->hw_id;
>> e->sched_attr = ctx->sched;
>> e->guilty = atomic_read(&ctx->guilty_count);
>> e->active = atomic_read(&ctx->active_count);
>> diff --git a/drivers/gpu/drm/i915/i915_gpu_error.h
>> b/drivers/gpu/drm/i915/i915_gpu_error.h
>> index caaed5093d95..4dc36d6ee3a2 100644
>> --- a/drivers/gpu/drm/i915/i915_gpu_error.h
>> +++ b/drivers/gpu/drm/i915/i915_gpu_error.h
>> @@ -117,7 +117,6 @@ struct i915_gpu_state {
>> struct drm_i915_error_context {
>> char comm[TASK_COMM_LEN];
>> pid_t pid;
>> - u32 hw_id;
>> int active;
>> int guilty;
>> struct i915_sched_attr sched_attr;
>> diff --git a/drivers/gpu/drm/i915/i915_perf.c
>> b/drivers/gpu/drm/i915/i915_perf.c
>> index 80055501eccb..d36ba248d438 100644
>> --- a/drivers/gpu/drm/i915/i915_perf.c
>> +++ b/drivers/gpu/drm/i915/i915_perf.c
>> @@ -1283,22 +1283,15 @@ static int oa_get_render_ctx_id(struct
>> i915_perf_stream *stream)
>> } else {
>> stream->specific_ctx_id_mask =
>> (1U << GEN8_CTX_ID_WIDTH) - 1;
>> - stream->specific_ctx_id =
>> - upper_32_bits(ce->lrc_desc);
>> - stream->specific_ctx_id &=
>> - stream->specific_ctx_id_mask;
>> + stream->specific_ctx_id = stream->specific_ctx_id_mask;
>> }
>> break;
>> case 11:
>> case 12: {
>> stream->specific_ctx_id_mask =
>> - ((1U << GEN11_SW_CTX_ID_WIDTH) - 1) <<
>> (GEN11_SW_CTX_ID_SHIFT - 32) |
>> - ((1U << GEN11_ENGINE_INSTANCE_WIDTH) - 1) <<
>> (GEN11_ENGINE_INSTANCE_SHIFT - 32) |
>> - ((1 << GEN11_ENGINE_CLASS_WIDTH) - 1) <<
>> (GEN11_ENGINE_CLASS_SHIFT - 32);
>> - stream->specific_ctx_id = upper_32_bits(ce->lrc_desc);
>> - stream->specific_ctx_id &=
>> - stream->specific_ctx_id_mask;
>> + ((1U << GEN11_SW_CTX_ID_WIDTH) - 1) <<
>> (GEN11_SW_CTX_ID_SHIFT - 32);
>> + stream->specific_ctx_id = stream->specific_ctx_id_mask;
>> break;
>> }
>> @@ -1306,6 +1299,8 @@ static int oa_get_render_ctx_id(struct
>> i915_perf_stream *stream)
>> MISSING_CASE(INTEL_GEN(i915));
>> }
>> + ce->tag = stream->specific_ctx_id_mask;
>> +
>> DRM_DEBUG_DRIVER("filtering on ctx_id=0x%x ctx_id_mask=0x%x\n",
>> stream->specific_ctx_id,
>> stream->specific_ctx_id_mask);
>> @@ -1324,12 +1319,14 @@ static void oa_put_render_ctx_id(struct
>> i915_perf_stream *stream)
>> {
>> struct intel_context *ce;
>> - stream->specific_ctx_id = INVALID_CTX_ID;
>> - stream->specific_ctx_id_mask = 0;
>> -
>> ce = fetch_and_zero(&stream->pinned_ctx);
>> - if (ce)
>> + if (ce) {
>> + ce->tag = 0;
>> intel_context_unpin(ce);
>> + }
>> +
>> + stream->specific_ctx_id = INVALID_CTX_ID;
>> + stream->specific_ctx_id_mask = 0;
>
> OA hunks looks dodgy as well. ce->tag is set to
> stream->specific_ctx_id_mask while I think it should be id. Furthermore
> id is assigned from mask which is fixed and not unique for different
> contexts.
>
>> }
>> static void
>> diff --git a/drivers/gpu/drm/i915/i915_trace.h
>> b/drivers/gpu/drm/i915/i915_trace.h
>> index 24f2944da09d..1f2cf6cfafb5 100644
>> --- a/drivers/gpu/drm/i915/i915_trace.h
>> +++ b/drivers/gpu/drm/i915/i915_trace.h
>> @@ -665,7 +665,6 @@ TRACE_EVENT(i915_request_queue,
>> TP_STRUCT__entry(
>> __field(u32, dev)
>> - __field(u32, hw_id)
>> __field(u64, ctx)
>> __field(u16, class)
>> __field(u16, instance)
>> @@ -675,7 +674,6 @@ TRACE_EVENT(i915_request_queue,
>> TP_fast_assign(
>> __entry->dev = rq->i915->drm.primary->index;
>> - __entry->hw_id = rq->gem_context->hw_id;
>> __entry->class = rq->engine->uabi_class;
>> __entry->instance = rq->engine->uabi_instance;
>> __entry->ctx = rq->fence.context;
>> @@ -683,10 +681,9 @@ TRACE_EVENT(i915_request_queue,
>> __entry->flags = flags;
>> ),
>> - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu,
>> seqno=%u, flags=0x%x",
>> + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u,
>> flags=0x%x",
>> __entry->dev, __entry->class, __entry->instance,
>> - __entry->hw_id, __entry->ctx, __entry->seqno,
>> - __entry->flags)
>> + __entry->ctx, __entry->seqno, __entry->flags)
>> );
>> DECLARE_EVENT_CLASS(i915_request,
>> @@ -695,7 +692,6 @@ DECLARE_EVENT_CLASS(i915_request,
>> TP_STRUCT__entry(
>> __field(u32, dev)
>> - __field(u32, hw_id)
>> __field(u64, ctx)
>> __field(u16, class)
>> __field(u16, instance)
>> @@ -704,16 +700,15 @@ DECLARE_EVENT_CLASS(i915_request,
>> TP_fast_assign(
>> __entry->dev = rq->i915->drm.primary->index;
>> - __entry->hw_id = rq->gem_context->hw_id;
>> __entry->class = rq->engine->uabi_class;
>> __entry->instance = rq->engine->uabi_instance;
>> __entry->ctx = rq->fence.context;
>> __entry->seqno = rq->fence.seqno;
>> ),
>> - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u",
>> + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u",
>> __entry->dev, __entry->class, __entry->instance,
>> - __entry->hw_id, __entry->ctx, __entry->seqno)
>> + __entry->ctx, __entry->seqno)
>> );
>> DEFINE_EVENT(i915_request, i915_request_add,
>> @@ -738,7 +733,6 @@ TRACE_EVENT(i915_request_in,
>> TP_STRUCT__entry(
>> __field(u32, dev)
>> - __field(u32, hw_id)
>> __field(u64, ctx)
>> __field(u16, class)
>> __field(u16, instance)
>> @@ -749,7 +743,6 @@ TRACE_EVENT(i915_request_in,
>> TP_fast_assign(
>> __entry->dev = rq->i915->drm.primary->index;
>> - __entry->hw_id = rq->gem_context->hw_id;
>> __entry->class = rq->engine->uabi_class;
>> __entry->instance = rq->engine->uabi_instance;
>> __entry->ctx = rq->fence.context;
>> @@ -758,9 +751,9 @@ TRACE_EVENT(i915_request_in,
>> __entry->port = port;
>> ),
>> - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu,
>> seqno=%u, prio=%u, port=%u",
>> + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, prio=%u,
>> port=%u",
>> __entry->dev, __entry->class, __entry->instance,
>> - __entry->hw_id, __entry->ctx, __entry->seqno,
>> + __entry->ctx, __entry->seqno,
>> __entry->prio, __entry->port)
>> );
>> @@ -770,7 +763,6 @@ TRACE_EVENT(i915_request_out,
>> TP_STRUCT__entry(
>> __field(u32, dev)
>> - __field(u32, hw_id)
>> __field(u64, ctx)
>> __field(u16, class)
>> __field(u16, instance)
>> @@ -780,7 +772,6 @@ TRACE_EVENT(i915_request_out,
>> TP_fast_assign(
>> __entry->dev = rq->i915->drm.primary->index;
>> - __entry->hw_id = rq->gem_context->hw_id;
>> __entry->class = rq->engine->uabi_class;
>> __entry->instance = rq->engine->uabi_instance;
>> __entry->ctx = rq->fence.context;
>> @@ -788,10 +779,9 @@ TRACE_EVENT(i915_request_out,
>> __entry->completed = i915_request_completed(rq);
>> ),
>> - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu,
>> seqno=%u, completed?=%u",
>> + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u,
>> completed?=%u",
>> __entry->dev, __entry->class, __entry->instance,
>> - __entry->hw_id, __entry->ctx, __entry->seqno,
>> - __entry->completed)
>> + __entry->ctx, __entry->seqno, __entry->completed)
>> );
>> #else
>> @@ -829,7 +819,6 @@ TRACE_EVENT(i915_request_wait_begin,
>> TP_STRUCT__entry(
>> __field(u32, dev)
>> - __field(u32, hw_id)
>> __field(u64, ctx)
>> __field(u16, class)
>> __field(u16, instance)
>> @@ -845,7 +834,6 @@ TRACE_EVENT(i915_request_wait_begin,
>> */
>> TP_fast_assign(
>> __entry->dev = rq->i915->drm.primary->index;
>> - __entry->hw_id = rq->gem_context->hw_id;
>> __entry->class = rq->engine->uabi_class;
>> __entry->instance = rq->engine->uabi_instance;
>> __entry->ctx = rq->fence.context;
>> @@ -853,9 +841,9 @@ TRACE_EVENT(i915_request_wait_begin,
>> __entry->flags = flags;
>> ),
>> - TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu,
>> seqno=%u, flags=0x%x",
>> + TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u,
>> flags=0x%x",
>> __entry->dev, __entry->class, __entry->instance,
>> - __entry->hw_id, __entry->ctx, __entry->seqno,
>> + __entry->ctx, __entry->seqno,
>> __entry->flags)
>> );
>> @@ -958,19 +946,17 @@ DECLARE_EVENT_CLASS(i915_context,
>> TP_STRUCT__entry(
>> __field(u32, dev)
>> __field(struct i915_gem_context *, ctx)
>> - __field(u32, hw_id)
>> __field(struct i915_address_space *, vm)
>> ),
>> TP_fast_assign(
>> __entry->dev = ctx->i915->drm.primary->index;
>> __entry->ctx = ctx;
>> - __entry->hw_id = ctx->hw_id;
>> __entry->vm = ctx->vm;
>> ),
>> - TP_printk("dev=%u, ctx=%p, ctx_vm=%p, hw_id=%u",
>> - __entry->dev, __entry->ctx, __entry->vm, __entry->hw_id)
>> + TP_printk("dev=%u, ctx=%p, ctx_vm=%p",
>> + __entry->dev, __entry->ctx, __entry->vm)
>> )
>> DEFINE_EVENT(i915_context, i915_context_create,
>> diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
>> b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
>> index 52d2df843148..f39f0282e78c 100644
>> --- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
>> +++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
>> @@ -491,8 +491,8 @@ static int igt_evict_contexts(void *arg)
>> if (IS_ERR(rq)) {
>> /* When full, fail_if_busy will trigger EBUSY */
>> if (PTR_ERR(rq) != -EBUSY) {
>> - pr_err("Unexpected error from request alloc (ctx
>> hw id %u, on %s): %d\n",
>> - ctx->hw_id, engine->name,
>> + pr_err("Unexpected error from request alloc (on
>> %s): %d\n",
>> + engine->name,
>> (int)PTR_ERR(rq));
>> err = PTR_ERR(rq);
>> }
>> diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c
>> b/drivers/gpu/drm/i915/selftests/i915_vma.c
>> index 1c9db08f7c28..ac1ff558eb90 100644
>> --- a/drivers/gpu/drm/i915/selftests/i915_vma.c
>> +++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
>> @@ -170,7 +170,7 @@ static int igt_vma_create(void *arg)
>> }
>> nc = 0;
>> - for_each_prime_number(num_ctx, MAX_CONTEXT_HW_ID) {
>> + for_each_prime_number(num_ctx, 2 * NUM_CONTEXT_TAG) {
>> for (; nc < num_ctx; nc++) {
>> ctx = mock_context(i915, "mock");
>> if (!ctx)
>>
>
> Unless I am missing something I see this patch as simplification and not
> a fix. If we can get away with it in the GuC world then it's quite a big
> simplification so it's fine by me.
The GuC will require a fixed ID per intel_context for the whole time the
context is in use/pinned. The GUC IDs are a global list (i.e. all
engines share it) with 64k entries, so the logic to assign them is going
to be a bit different and most likely confined to the guc
context_pin/unpin callbacks, but we might bring back some of the
low-level ID management logic that has been removed in this patch.
An ID at the gem_context level definitely doesn't help on the guc side, so:
Acked-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Daniele
>
> Regards,
>
> Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* Re: [PATCH 20/27] drm/i915: Remove logical HW ID
2019-09-25 17:59 ` Daniele Ceraolo Spurio
@ 2019-09-25 18:23 ` Matthew Brost
0 siblings, 0 replies; 70+ messages in thread
From: Matthew Brost @ 2019-09-25 18:23 UTC (permalink / raw)
To: Daniele Ceraolo Spurio; +Cc: intel-gfx
On Wed, Sep 25, 2019 at 10:59:32AM -0700, Daniele Ceraolo Spurio wrote:
>+ Matt
>
I've reviewed this patch and this affects some of the work I'm doing on the new
GuC interface. For the new GuC interface I have two patches that rework the HW
ID assignment.
The first patch moves ownership of the HW ID from i915_gem_context to
intel_context as the new GuC has 1 to 1 mapping betweens HW IDs and LRCs.
The second patch changes the allocation algorithm to use a bitmap rather than
ida as contiguous HW IDs are required for multi-lrc to work with the new GuC.
All of this being said, I don't think anything in this patch will badly break
what I have for the new GuC. At most it will be minor refactor once the new GuC
patches are rebased on top of these patches.
-Matt
>On 9/25/19 5:41 AM, Tvrtko Ursulin wrote:
>>[+ Daniele, I think he might want to have a look at this.]
>>
>>On 25/09/2019 11:01, Chris Wilson wrote:
>>>With the introduction of ctx->engines[] we allow multiple logical
>>>contexts to be used on the same engine (e.g. with virtual engines). Each
>>>logical context requires a unique tag in order for context-switching to
>>>occur correctly between them.
>>>
>>>We only need to keep a unique tag for the active lifetime of the
>>>context, and for as long as we need to identify that context. The HW
>>>uses the tag to determine if it should use a lite-restore (why not the
>>>LRCA?) and passes the tag back for various status identifies. The only
>>>status we need to track is for OA, so when using perf, we assign the
>>>specific context a unique tag.
>>>
>>>Fixes: 976b55f0e1db ("drm/i915: Allow a context to define its set
>>>of engines")
>>
>>Fixes? Why?
>>
>>>Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
>>>---
>>> drivers/gpu/drm/i915/gem/i915_gem_context.c | 147 ------------------
>>> drivers/gpu/drm/i915/gem/i915_gem_context.h | 15 --
>>> .../gpu/drm/i915/gem/i915_gem_context_types.h | 18 ---
>>> .../drm/i915/gem/selftests/i915_gem_context.c | 13 +-
>>> .../gpu/drm/i915/gem/selftests/mock_context.c | 8 -
>>> drivers/gpu/drm/i915/gt/intel_context_types.h | 1 +
>>> drivers/gpu/drm/i915/gt/intel_engine_types.h | 4 +-
>>> drivers/gpu/drm/i915/gt/intel_lrc.c | 29 ++--
>>> drivers/gpu/drm/i915/gvt/kvmgt.c | 17 --
>>> drivers/gpu/drm/i915/i915_debugfs.c | 3 -
>>> drivers/gpu/drm/i915/i915_gpu_error.c | 7 +-
>>> drivers/gpu/drm/i915/i915_gpu_error.h | 1 -
>>> drivers/gpu/drm/i915/i915_perf.c | 25 ++-
>>> drivers/gpu/drm/i915/i915_trace.h | 38 ++---
>>> .../gpu/drm/i915/selftests/i915_gem_evict.c | 4 +-
>>> drivers/gpu/drm/i915/selftests/i915_vma.c | 2 +-
>>> 16 files changed, 48 insertions(+), 284 deletions(-)
>>>
>>>diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c
>>>b/drivers/gpu/drm/i915/gem/i915_gem_context.c
>>>index e832238a72e5..4e59b809d901 100644
>>>--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
>>>+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
>>>@@ -168,97 +168,6 @@ lookup_user_engine(struct i915_gem_context *ctx,
>>> return i915_gem_context_get_engine(ctx, idx);
>>> }
>>>-static inline int new_hw_id(struct drm_i915_private *i915, gfp_t gfp)
>>>-{
>>>- unsigned int max;
>>>-
>>>- lockdep_assert_held(&i915->contexts.mutex);
>>>-
>>>- if (INTEL_GEN(i915) >= 12)
>>>- max = GEN12_MAX_CONTEXT_HW_ID;
>>>- else if (INTEL_GEN(i915) >= 11)
>>>- max = GEN11_MAX_CONTEXT_HW_ID;
>>>- else if (USES_GUC_SUBMISSION(i915))
>>>- /*
>>>- * When using GuC in proxy submission, GuC consumes the
>>>- * highest bit in the context id to indicate proxy submission.
>>>- */
>>>- max = MAX_GUC_CONTEXT_HW_ID;
>>>- else
>>>- max = MAX_CONTEXT_HW_ID;
>>>-
>>>- return ida_simple_get(&i915->contexts.hw_ida, 0, max, gfp);
>>>-}
>>>-
>>>-static int steal_hw_id(struct drm_i915_private *i915)
>>>-{
>>>- struct i915_gem_context *ctx, *cn;
>>>- LIST_HEAD(pinned);
>>>- int id = -ENOSPC;
>>>-
>>>- lockdep_assert_held(&i915->contexts.mutex);
>>>-
>>>- list_for_each_entry_safe(ctx, cn,
>>>- &i915->contexts.hw_id_list, hw_id_link) {
>>>- if (atomic_read(&ctx->hw_id_pin_count)) {
>>>- list_move_tail(&ctx->hw_id_link, &pinned);
>>>- continue;
>>>- }
>>>-
>>>- GEM_BUG_ON(!ctx->hw_id); /* perma-pinned kernel context */
>>>- list_del_init(&ctx->hw_id_link);
>>>- id = ctx->hw_id;
>>>- break;
>>>- }
>>>-
>>>- /*
>>>- * Remember how far we got up on the last repossesion scan, so the
>>>- * list is kept in a "least recently scanned" order.
>>>- */
>>>- list_splice_tail(&pinned, &i915->contexts.hw_id_list);
>>>- return id;
>>>-}
>>>-
>>>-static int assign_hw_id(struct drm_i915_private *i915, unsigned
>>>int *out)
>>>-{
>>>- int ret;
>>>-
>>>- lockdep_assert_held(&i915->contexts.mutex);
>>>-
>>>- /*
>>>- * We prefer to steal/stall ourselves and our users over that of the
>>>- * entire system. That may be a little unfair to our users, and
>>>- * even hurt high priority clients. The choice is whether to oomkill
>>>- * something else, or steal a context id.
>>>- */
>>>- ret = new_hw_id(i915, GFP_KERNEL | __GFP_RETRY_MAYFAIL |
>>>__GFP_NOWARN);
>>>- if (unlikely(ret < 0)) {
>>>- ret = steal_hw_id(i915);
>>>- if (ret < 0) /* once again for the correct errno code */
>>>- ret = new_hw_id(i915, GFP_KERNEL);
>>>- if (ret < 0)
>>>- return ret;
>>>- }
>>>-
>>>- *out = ret;
>>>- return 0;
>>>-}
>>>-
>>>-static void release_hw_id(struct i915_gem_context *ctx)
>>>-{
>>>- struct drm_i915_private *i915 = ctx->i915;
>>>-
>>>- if (list_empty(&ctx->hw_id_link))
>>>- return;
>>>-
>>>- mutex_lock(&i915->contexts.mutex);
>>>- if (!list_empty(&ctx->hw_id_link)) {
>>>- ida_simple_remove(&i915->contexts.hw_ida, ctx->hw_id);
>>>- list_del_init(&ctx->hw_id_link);
>>>- }
>>>- mutex_unlock(&i915->contexts.mutex);
>>>-}
>>>-
>>> static void __free_engines(struct i915_gem_engines *e, unsigned
>>>int count)
>>> {
>>> while (count--) {
>>>@@ -313,8 +222,6 @@ static void i915_gem_context_free(struct
>>>i915_gem_context *ctx)
>>> lockdep_assert_held(&ctx->i915->drm.struct_mutex);
>>> GEM_BUG_ON(!i915_gem_context_is_closed(ctx));
>>>- release_hw_id(ctx);
>>>-
>>> free_engines(rcu_access_pointer(ctx->engines));
>>> mutex_destroy(&ctx->engines_mutex);
>>>@@ -459,12 +366,6 @@ static void context_close(struct
>>>i915_gem_context *ctx)
>>> ctx->file_priv = ERR_PTR(-EBADF);
>>>- /*
>>>- * This context will never again be assinged to HW, so we can
>>>- * reuse its ID for the next context.
>>>- */
>>>- release_hw_id(ctx);
>>>-
>>> /*
>>> * The LUT uses the VMA as a backpointer to unref the object,
>>> * so we need to clear the LUT before we close all the VMA (inside
>>>@@ -507,7 +408,6 @@ __create_context(struct drm_i915_private *i915)
>>> RCU_INIT_POINTER(ctx->engines, e);
>>> INIT_RADIX_TREE(&ctx->handles_vma, GFP_KERNEL);
>>>- INIT_LIST_HEAD(&ctx->hw_id_link);
>>> /* NB: Mark all slices as needing a remap so that when the
>>>context first
>>> * loads it will restore whatever remap state already
>>>exists. If there
>>>@@ -676,18 +576,11 @@ struct i915_gem_context *
>>> i915_gem_context_create_kernel(struct drm_i915_private *i915, int prio)
>>> {
>>> struct i915_gem_context *ctx;
>>>- int err;
>>> ctx = i915_gem_create_context(i915, 0);
>>> if (IS_ERR(ctx))
>>> return ctx;
>>>- err = i915_gem_context_pin_hw_id(ctx);
>>>- if (err) {
>>>- destroy_kernel_context(&ctx);
>>>- return ERR_PTR(err);
>>>- }
>>>-
>>> i915_gem_context_clear_bannable(ctx);
>>> i915_gem_context_set_persistence(ctx);
>>> ctx->sched.priority = I915_USER_PRIORITY(prio);
>>>@@ -727,15 +620,6 @@ int i915_gem_contexts_init(struct
>>>drm_i915_private *dev_priv)
>>> DRM_ERROR("Failed to create default global context\n");
>>> return PTR_ERR(ctx);
>>> }
>>>- /*
>>>- * For easy recognisablity, we want the kernel context to be
>>>0 and then
>>>- * all user contexts will have non-zero hw_id. Kernel contexts are
>>>- * permanently pinned, so that we never suffer a stall and can
>>>- * use them from any allocation context (e.g. for evicting other
>>>- * contexts and from inside the shrinker).
>>>- */
>>>- GEM_BUG_ON(ctx->hw_id);
>>>- GEM_BUG_ON(!atomic_read(&ctx->hw_id_pin_count));
>>> dev_priv->kernel_context = ctx;
>>> DRM_DEBUG_DRIVER("%s context support initialized\n",
>>>@@ -749,10 +633,6 @@ void i915_gem_contexts_fini(struct
>>>drm_i915_private *i915)
>>> lockdep_assert_held(&i915->drm.struct_mutex);
>>> destroy_kernel_context(&i915->kernel_context);
>>>-
>>>- /* Must free all deferred contexts (via flush_workqueue) first */
>>>- GEM_BUG_ON(!list_empty(&i915->contexts.hw_id_list));
>>>- ida_destroy(&i915->contexts.hw_ida);
>>> }
>>> static int context_idr_cleanup(int id, void *p, void *data)
>>>@@ -2438,33 +2318,6 @@ int
>>>i915_gem_context_reset_stats_ioctl(struct drm_device *dev,
>>> return ret;
>>> }
>>>-int __i915_gem_context_pin_hw_id(struct i915_gem_context *ctx)
>>>-{
>>>- struct drm_i915_private *i915 = ctx->i915;
>>>- int err = 0;
>>>-
>>>- mutex_lock(&i915->contexts.mutex);
>>>-
>>>- GEM_BUG_ON(i915_gem_context_is_closed(ctx));
>>>-
>>>- if (list_empty(&ctx->hw_id_link)) {
>>>- GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count));
>>>-
>>>- err = assign_hw_id(i915, &ctx->hw_id);
>>>- if (err)
>>>- goto out_unlock;
>>>-
>>>- list_add_tail(&ctx->hw_id_link, &i915->contexts.hw_id_list);
>>>- }
>>>-
>>>- GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count) == ~0u);
>>>- atomic_inc(&ctx->hw_id_pin_count);
>>>-
>>>-out_unlock:
>>>- mutex_unlock(&i915->contexts.mutex);
>>>- return err;
>>>-}
>>>-
>>> /* GEM context-engines iterator: for_each_gem_engine() */
>>> struct intel_context *
>>> i915_gem_engines_iter_next(struct i915_gem_engines_iter *it)
>>>diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.h
>>>b/drivers/gpu/drm/i915/gem/i915_gem_context.h
>>>index e0f5b6c6a331..1b0df53436cf 100644
>>>--- a/drivers/gpu/drm/i915/gem/i915_gem_context.h
>>>+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.h
>>>@@ -127,21 +127,6 @@ 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)
>>>-{
>>>- if (atomic_inc_not_zero(&ctx->hw_id_pin_count))
>>>- return 0;
>>>-
>>>- return __i915_gem_context_pin_hw_id(ctx);
>>>-}
>>>-
>>>-static inline void i915_gem_context_unpin_hw_id(struct
>>>i915_gem_context *ctx)
>>>-{
>>>- GEM_BUG_ON(atomic_read(&ctx->hw_id_pin_count) == 0u);
>>>- atomic_dec(&ctx->hw_id_pin_count);
>>>-}
>>>-
>>> static inline bool i915_gem_context_is_kernel(struct
>>>i915_gem_context *ctx)
>>> {
>>> return !ctx->file_priv;
>>>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 daf1ea5075a6..6419da7c9f90 100644
>>>--- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
>>>+++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
>>>@@ -148,24 +148,6 @@ struct i915_gem_context {
>>> #define CONTEXT_FORCE_SINGLE_SUBMISSION 2
>>> #define CONTEXT_USER_ENGINES 3
>>>- /**
>>>- * @hw_id: - unique identifier for the context
>>>- *
>>>- * The hardware needs to uniquely identify the context for a few
>>>- * functions like fault reporting, PASID, scheduling. The
>>>- * &drm_i915_private.context_hw_ida is used to assign a unqiue
>>>- * id for the lifetime of the context.
>>>- *
>>>- * @hw_id_pin_count: - number of times this context had been pinned
>>>- * for use (should be, at most, once per engine).
>>>- *
>>>- * @hw_id_link: - all contexts with an assigned id are tracked
>>>- * for possible repossession.
>>>- */
>>>- unsigned int hw_id;
>>>- atomic_t hw_id_pin_count;
>>>- struct list_head hw_id_link;
>>>-
>>> struct mutex mutex;
>>> struct i915_sched_attr sched;
>>>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 b39224ed3817..a8a9293f4ac6 100644
>>>--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
>>>+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
>>>@@ -446,9 +446,9 @@ static int igt_ctx_exec(void *arg)
>>> err = gpu_fill(ce, obj, dw);
>>> if (err) {
>>>- pr_err("Failed to fill dword %lu [%lu/%lu] with
>>>gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n",
>>>+ pr_err("Failed to fill dword %lu [%lu/%lu] with
>>>gpu (%s) [full-ppgtt? %s], err=%d\n",
>>> ndwords, dw, max_dwords(obj),
>>>- engine->name, ctx->hw_id,
>>>+ engine->name,
>>> yesno(!!ctx->vm), err);
>>> intel_context_put(ce);
>>> kernel_context_close(ctx);
>>>@@ -584,9 +584,9 @@ static int igt_shared_ctx_exec(void *arg)
>>> err = gpu_fill(ce, obj, dw);
>>> if (err) {
>>>- pr_err("Failed to fill dword %lu [%lu/%lu] with
>>>gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n",
>>>+ pr_err("Failed to fill dword %lu [%lu/%lu] with
>>>gpu (%s) [full-ppgtt? %s], err=%d\n",
>>> ndwords, dw, max_dwords(obj),
>>>- engine->name, ctx->hw_id,
>>>+ engine->name,
>>> yesno(!!ctx->vm), err);
>>> intel_context_put(ce);
>>> kernel_context_close(ctx);
>>>@@ -1168,10 +1168,9 @@ static int igt_ctx_readonly(void *arg)
>>> err = gpu_fill(ce, obj, dw);
>>> if (err) {
>>>- pr_err("Failed to fill dword %lu [%lu/%lu] with
>>>gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n",
>>>+ pr_err("Failed to fill dword %lu [%lu/%lu] with
>>>gpu (%s) [full-ppgtt? %s], err=%d\n",
>>> ndwords, dw, max_dwords(obj),
>>>- ce->engine->name, ctx->hw_id,
>>>- yesno(!!ctx->vm), err);
>>>+ ce->engine->name, yesno(!!ctx->vm), err);
>>> i915_gem_context_unlock_engines(ctx);
>>> goto out_unlock;
>>> }
>>>diff --git a/drivers/gpu/drm/i915/gem/selftests/mock_context.c
>>>b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
>>>index 0d1a275ec316..ebc46f098561 100644
>>>--- a/drivers/gpu/drm/i915/gem/selftests/mock_context.c
>>>+++ b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
>>>@@ -13,7 +13,6 @@ mock_context(struct drm_i915_private *i915,
>>> {
>>> struct i915_gem_context *ctx;
>>> struct i915_gem_engines *e;
>>>- int ret;
>>> ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
>>> if (!ctx)
>>>@@ -32,13 +31,8 @@ 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->hw_id_link);
>>> mutex_init(&ctx->mutex);
>>>- ret = i915_gem_context_pin_hw_id(ctx);
>>>- if (ret < 0)
>>>- goto err_engines;
>>>-
>>> if (name) {
>>> struct i915_ppgtt *ppgtt;
>>>@@ -56,8 +50,6 @@ mock_context(struct drm_i915_private *i915,
>>> return ctx;
>>>-err_engines:
>>>- free_engines(rcu_access_pointer(ctx->engines));
>>> err_free:
>>> kfree(ctx);
>>> return NULL;
>>>diff --git a/drivers/gpu/drm/i915/gt/intel_context_types.h
>>>b/drivers/gpu/drm/i915/gt/intel_context_types.h
>>>index bf9cedfccbf0..6959b05ae5f8 100644
>>>--- a/drivers/gpu/drm/i915/gt/intel_context_types.h
>>>+++ b/drivers/gpu/drm/i915/gt/intel_context_types.h
>>>@@ -58,6 +58,7 @@ struct intel_context {
>>> u32 *lrc_reg_state;
>>> u64 lrc_desc;
>>>+ u32 tag; /* cookie passed to HW to track this context on
>>>submission */
>>> unsigned int active_count; /* protected by timeline->mutex */
>>>diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h
>>>b/drivers/gpu/drm/i915/gt/intel_engine_types.h
>>>index da4b45aa84b1..7bce176e696c 100644
>>>--- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
>>>+++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
>>>@@ -296,10 +296,12 @@ struct intel_engine_cs {
>>> u8 uabi_class;
>>> u8 uabi_instance;
>>>+ u32 uabi_capabilities;
>>> u32 context_size;
>>> u32 mmio_base;
>>>- u32 uabi_capabilities;
>>>+ unsigned int context_tag;
>>>+#define NUM_CONTEXT_TAG (256)
>>
>>AFAICT this define now defines how deep ELSP we can have before we
>>start getting hard to debug problems. GuC angle needs considering
>>here.
>>
>>Can we add some asserts somewhere to express the relationship?
>>
>>And a comment here explaining this, or with the assert.
>>
>>> struct rb_node uabi_node;
>>>diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c
>>>b/drivers/gpu/drm/i915/gt/intel_lrc.c
>>>index 0ea2f4d7dc43..af79b8d70800 100644
>>>--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
>>>+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
>>>@@ -433,12 +433,8 @@ assert_priority_queue(const struct
>>>i915_request *prev,
>>> static u64
>>> lrc_descriptor(struct intel_context *ce, struct intel_engine_cs
>>>*engine)
>>> {
>>>- struct i915_gem_context *ctx = ce->gem_context;
>>> u64 desc;
>>>- BUILD_BUG_ON(MAX_CONTEXT_HW_ID > (BIT(GEN8_CTX_ID_WIDTH)));
>>>- BUILD_BUG_ON(GEN11_MAX_CONTEXT_HW_ID >
>>>(BIT(GEN11_SW_CTX_ID_WIDTH)));
>>>-
>>> desc = INTEL_LEGACY_32B_CONTEXT;
>>> if (i915_vm_is_4lvl(ce->vm))
>>> desc = INTEL_LEGACY_64B_CONTEXT;
>>>@@ -456,20 +452,11 @@ lrc_descriptor(struct intel_context *ce,
>>>struct intel_engine_cs *engine)
>>> * anything below.
>>> */
>>> if (INTEL_GEN(engine->i915) >= 11) {
>>>- GEM_BUG_ON(ctx->hw_id >= BIT(GEN11_SW_CTX_ID_WIDTH));
>>>- desc |= (u64)ctx->hw_id << GEN11_SW_CTX_ID_SHIFT;
>>>- /* bits 37-47 */
>>>-
>>> desc |= (u64)engine->instance << GEN11_ENGINE_INSTANCE_SHIFT;
>>> /* bits 48-53 */
>>>- /* TODO: decide what to do with SW counter (bits 55-60) */
>>>-
>>> desc |= (u64)engine->class << GEN11_ENGINE_CLASS_SHIFT;
>>> /* bits 61-63 */
>>>- } else {
>>>- GEM_BUG_ON(ctx->hw_id >= BIT(GEN8_CTX_ID_WIDTH));
>>>- desc |= (u64)ctx->hw_id << GEN8_CTX_ID_SHIFT; /* bits
>>>32-52 */
>>> }
>>> return desc;
>>>@@ -978,6 +965,15 @@ __execlists_schedule_in(struct i915_request *rq)
>>> intel_context_get(ce);
>>>+ if (ce->tag) {
>>>+ ce->lrc_desc |= (u64)ce->tag << 32;
>>>+ } else {
>>>+ ce->lrc_desc &= ~GENMASK_ULL(47, 37);
>>>+ ce->lrc_desc |=
>>>+ (u64)(engine->context_tag++ % NUM_CONTEXT_TAG) <<
>>>+ GEN11_SW_CTX_ID_SHIFT;
>>>+ }
>>
>>So hw id is valid only while context is in ELSP and it changes with
>>every submission except in the OA case?
>>
>>Shifts and masks look dodgy. For >= gen11 the current code has the
>>hw_id in [37, 42], otherwise [32, 52]. This patch does not seem to
>>handle the differences between gens.
>>
>>>+
>>> intel_gt_pm_get(engine->gt);
>>> execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN);
>>> intel_engine_context_in(engine);
>>>@@ -2224,7 +2220,6 @@ static void execlists_context_unpin(struct
>>>intel_context *ce)
>>> check_redzone((void *)ce->lrc_reg_state - LRC_STATE_PN * PAGE_SIZE,
>>> ce->engine);
>>>- i915_gem_context_unpin_hw_id(ce->gem_context);
>>> i915_gem_object_unpin_map(ce->state->obj);
>>> intel_ring_reset(ce->ring, ce->ring->tail);
>>> }
>>>@@ -2274,18 +2269,12 @@ __execlists_context_pin(struct intel_context *ce,
>>> goto unpin_active;
>>> }
>>>- ret = i915_gem_context_pin_hw_id(ce->gem_context);
>>>- if (ret)
>>>- goto unpin_map;
>>>-
>>> ce->lrc_desc = lrc_descriptor(ce, engine);
>>> ce->lrc_reg_state = vaddr + LRC_STATE_PN * PAGE_SIZE;
>>> __execlists_update_reg_state(ce, engine);
>>> return 0;
>>>-unpin_map:
>>>- i915_gem_object_unpin_map(ce->state->obj);
>>> unpin_active:
>>> intel_context_active_release(ce);
>>> err:
>>>diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c
>>>b/drivers/gpu/drm/i915/gvt/kvmgt.c
>>>index 343d79c1cb7e..04a5a0d90823 100644
>>>--- a/drivers/gpu/drm/i915/gvt/kvmgt.c
>>>+++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
>>>@@ -1564,27 +1564,10 @@ vgpu_id_show(struct device *dev, struct
>>>device_attribute *attr,
>>> return sprintf(buf, "\n");
>>> }
>>>-static ssize_t
>>>-hw_id_show(struct device *dev, struct device_attribute *attr,
>>>- char *buf)
>>>-{
>>>- struct mdev_device *mdev = mdev_from_dev(dev);
>>>-
>>>- if (mdev) {
>>>- struct intel_vgpu *vgpu = (struct intel_vgpu *)
>>>- mdev_get_drvdata(mdev);
>>>- return sprintf(buf, "%u\n",
>>>- vgpu->submission.shadow[0]->gem_context->hw_id);
>>>- }
>>>- return sprintf(buf, "\n");
>>>-}
>>>-
>>> static DEVICE_ATTR_RO(vgpu_id);
>>>-static DEVICE_ATTR_RO(hw_id);
>>> static struct attribute *intel_vgpu_attrs[] = {
>>> &dev_attr_vgpu_id.attr,
>>>- &dev_attr_hw_id.attr,
>>> NULL
>>> };
>>>diff --git a/drivers/gpu/drm/i915/i915_debugfs.c
>>>b/drivers/gpu/drm/i915/i915_debugfs.c
>>>index fd41505d52ec..8caaa446490f 100644
>>>--- a/drivers/gpu/drm/i915/i915_debugfs.c
>>>+++ b/drivers/gpu/drm/i915/i915_debugfs.c
>>>@@ -1504,9 +1504,6 @@ static int i915_context_status(struct
>>>seq_file *m, void *unused)
>>> struct intel_context *ce;
>>> seq_puts(m, "HW context ");
>>>- if (!list_empty(&ctx->hw_id_link))
>>>- seq_printf(m, "%x [pin %u]", ctx->hw_id,
>>>- atomic_read(&ctx->hw_id_pin_count));
>>> if (ctx->pid) {
>>> struct task_struct *task;
>>>diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c
>>>b/drivers/gpu/drm/i915/i915_gpu_error.c
>>>index a86d28bda6dd..47239df653f2 100644
>>>--- a/drivers/gpu/drm/i915/i915_gpu_error.c
>>>+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
>>>@@ -471,9 +471,9 @@ static void error_print_context(struct
>>>drm_i915_error_state_buf *m,
>>> const char *header,
>>> const struct drm_i915_error_context *ctx)
>>> {
>>>- err_printf(m, "%s%s[%d] hw_id %d, prio %d, guilty %d active %d\n",
>>>- header, ctx->comm, ctx->pid, ctx->hw_id,
>>>- ctx->sched_attr.priority, ctx->guilty, ctx->active);
>>>+ err_printf(m, "%s%s[%d] prio %d, guilty %d active %d\n",
>>>+ header, ctx->comm, ctx->pid, ctx->sched_attr.priority,
>>>+ ctx->guilty, ctx->active);
>>> }
>>> static void error_print_engine(struct drm_i915_error_state_buf *m,
>>>@@ -1262,7 +1262,6 @@ static bool record_context(struct
>>>drm_i915_error_context *e,
>>> rcu_read_unlock();
>>> }
>>>- e->hw_id = ctx->hw_id;
>>> e->sched_attr = ctx->sched;
>>> e->guilty = atomic_read(&ctx->guilty_count);
>>> e->active = atomic_read(&ctx->active_count);
>>>diff --git a/drivers/gpu/drm/i915/i915_gpu_error.h
>>>b/drivers/gpu/drm/i915/i915_gpu_error.h
>>>index caaed5093d95..4dc36d6ee3a2 100644
>>>--- a/drivers/gpu/drm/i915/i915_gpu_error.h
>>>+++ b/drivers/gpu/drm/i915/i915_gpu_error.h
>>>@@ -117,7 +117,6 @@ struct i915_gpu_state {
>>> struct drm_i915_error_context {
>>> char comm[TASK_COMM_LEN];
>>> pid_t pid;
>>>- u32 hw_id;
>>> int active;
>>> int guilty;
>>> struct i915_sched_attr sched_attr;
>>>diff --git a/drivers/gpu/drm/i915/i915_perf.c
>>>b/drivers/gpu/drm/i915/i915_perf.c
>>>index 80055501eccb..d36ba248d438 100644
>>>--- a/drivers/gpu/drm/i915/i915_perf.c
>>>+++ b/drivers/gpu/drm/i915/i915_perf.c
>>>@@ -1283,22 +1283,15 @@ static int oa_get_render_ctx_id(struct
>>>i915_perf_stream *stream)
>>> } else {
>>> stream->specific_ctx_id_mask =
>>> (1U << GEN8_CTX_ID_WIDTH) - 1;
>>>- stream->specific_ctx_id =
>>>- upper_32_bits(ce->lrc_desc);
>>>- stream->specific_ctx_id &=
>>>- stream->specific_ctx_id_mask;
>>>+ stream->specific_ctx_id = stream->specific_ctx_id_mask;
>>> }
>>> break;
>>> case 11:
>>> case 12: {
>>> stream->specific_ctx_id_mask =
>>>- ((1U << GEN11_SW_CTX_ID_WIDTH) - 1) <<
>>>(GEN11_SW_CTX_ID_SHIFT - 32) |
>>>- ((1U << GEN11_ENGINE_INSTANCE_WIDTH) - 1) <<
>>>(GEN11_ENGINE_INSTANCE_SHIFT - 32) |
>>>- ((1 << GEN11_ENGINE_CLASS_WIDTH) - 1) <<
>>>(GEN11_ENGINE_CLASS_SHIFT - 32);
>>>- stream->specific_ctx_id = upper_32_bits(ce->lrc_desc);
>>>- stream->specific_ctx_id &=
>>>- stream->specific_ctx_id_mask;
>>>+ ((1U << GEN11_SW_CTX_ID_WIDTH) - 1) <<
>>>(GEN11_SW_CTX_ID_SHIFT - 32);
>>>+ stream->specific_ctx_id = stream->specific_ctx_id_mask;
>>> break;
>>> }
>>>@@ -1306,6 +1299,8 @@ static int oa_get_render_ctx_id(struct
>>>i915_perf_stream *stream)
>>> MISSING_CASE(INTEL_GEN(i915));
>>> }
>>>+ ce->tag = stream->specific_ctx_id_mask;
>>>+
>>> DRM_DEBUG_DRIVER("filtering on ctx_id=0x%x ctx_id_mask=0x%x\n",
>>> stream->specific_ctx_id,
>>> stream->specific_ctx_id_mask);
>>>@@ -1324,12 +1319,14 @@ static void oa_put_render_ctx_id(struct
>>>i915_perf_stream *stream)
>>> {
>>> struct intel_context *ce;
>>>- stream->specific_ctx_id = INVALID_CTX_ID;
>>>- stream->specific_ctx_id_mask = 0;
>>>-
>>> ce = fetch_and_zero(&stream->pinned_ctx);
>>>- if (ce)
>>>+ if (ce) {
>>>+ ce->tag = 0;
>>> intel_context_unpin(ce);
>>>+ }
>>>+
>>>+ stream->specific_ctx_id = INVALID_CTX_ID;
>>>+ stream->specific_ctx_id_mask = 0;
>>
>>OA hunks looks dodgy as well. ce->tag is set to
>>stream->specific_ctx_id_mask while I think it should be id.
>>Furthermore id is assigned from mask which is fixed and not unique
>>for different contexts.
>>
>>> }
>>> static void
>>>diff --git a/drivers/gpu/drm/i915/i915_trace.h
>>>b/drivers/gpu/drm/i915/i915_trace.h
>>>index 24f2944da09d..1f2cf6cfafb5 100644
>>>--- a/drivers/gpu/drm/i915/i915_trace.h
>>>+++ b/drivers/gpu/drm/i915/i915_trace.h
>>>@@ -665,7 +665,6 @@ TRACE_EVENT(i915_request_queue,
>>> TP_STRUCT__entry(
>>> __field(u32, dev)
>>>- __field(u32, hw_id)
>>> __field(u64, ctx)
>>> __field(u16, class)
>>> __field(u16, instance)
>>>@@ -675,7 +674,6 @@ TRACE_EVENT(i915_request_queue,
>>> TP_fast_assign(
>>> __entry->dev = rq->i915->drm.primary->index;
>>>- __entry->hw_id = rq->gem_context->hw_id;
>>> __entry->class = rq->engine->uabi_class;
>>> __entry->instance = rq->engine->uabi_instance;
>>> __entry->ctx = rq->fence.context;
>>>@@ -683,10 +681,9 @@ TRACE_EVENT(i915_request_queue,
>>> __entry->flags = flags;
>>> ),
>>>- TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu,
>>>seqno=%u, flags=0x%x",
>>>+ TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u,
>>>flags=0x%x",
>>> __entry->dev, __entry->class, __entry->instance,
>>>- __entry->hw_id, __entry->ctx, __entry->seqno,
>>>- __entry->flags)
>>>+ __entry->ctx, __entry->seqno, __entry->flags)
>>> );
>>> DECLARE_EVENT_CLASS(i915_request,
>>>@@ -695,7 +692,6 @@ DECLARE_EVENT_CLASS(i915_request,
>>> TP_STRUCT__entry(
>>> __field(u32, dev)
>>>- __field(u32, hw_id)
>>> __field(u64, ctx)
>>> __field(u16, class)
>>> __field(u16, instance)
>>>@@ -704,16 +700,15 @@ DECLARE_EVENT_CLASS(i915_request,
>>> TP_fast_assign(
>>> __entry->dev = rq->i915->drm.primary->index;
>>>- __entry->hw_id = rq->gem_context->hw_id;
>>> __entry->class = rq->engine->uabi_class;
>>> __entry->instance = rq->engine->uabi_instance;
>>> __entry->ctx = rq->fence.context;
>>> __entry->seqno = rq->fence.seqno;
>>> ),
>>>- TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu, seqno=%u",
>>>+ TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u",
>>> __entry->dev, __entry->class, __entry->instance,
>>>- __entry->hw_id, __entry->ctx, __entry->seqno)
>>>+ __entry->ctx, __entry->seqno)
>>> );
>>> DEFINE_EVENT(i915_request, i915_request_add,
>>>@@ -738,7 +733,6 @@ TRACE_EVENT(i915_request_in,
>>> TP_STRUCT__entry(
>>> __field(u32, dev)
>>>- __field(u32, hw_id)
>>> __field(u64, ctx)
>>> __field(u16, class)
>>> __field(u16, instance)
>>>@@ -749,7 +743,6 @@ TRACE_EVENT(i915_request_in,
>>> TP_fast_assign(
>>> __entry->dev = rq->i915->drm.primary->index;
>>>- __entry->hw_id = rq->gem_context->hw_id;
>>> __entry->class = rq->engine->uabi_class;
>>> __entry->instance = rq->engine->uabi_instance;
>>> __entry->ctx = rq->fence.context;
>>>@@ -758,9 +751,9 @@ TRACE_EVENT(i915_request_in,
>>> __entry->port = port;
>>> ),
>>>- TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu,
>>>seqno=%u, prio=%u, port=%u",
>>>+ TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u,
>>>prio=%u, port=%u",
>>> __entry->dev, __entry->class, __entry->instance,
>>>- __entry->hw_id, __entry->ctx, __entry->seqno,
>>>+ __entry->ctx, __entry->seqno,
>>> __entry->prio, __entry->port)
>>> );
>>>@@ -770,7 +763,6 @@ TRACE_EVENT(i915_request_out,
>>> TP_STRUCT__entry(
>>> __field(u32, dev)
>>>- __field(u32, hw_id)
>>> __field(u64, ctx)
>>> __field(u16, class)
>>> __field(u16, instance)
>>>@@ -780,7 +772,6 @@ TRACE_EVENT(i915_request_out,
>>> TP_fast_assign(
>>> __entry->dev = rq->i915->drm.primary->index;
>>>- __entry->hw_id = rq->gem_context->hw_id;
>>> __entry->class = rq->engine->uabi_class;
>>> __entry->instance = rq->engine->uabi_instance;
>>> __entry->ctx = rq->fence.context;
>>>@@ -788,10 +779,9 @@ TRACE_EVENT(i915_request_out,
>>> __entry->completed = i915_request_completed(rq);
>>> ),
>>>- TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu,
>>>seqno=%u, completed?=%u",
>>>+ TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u,
>>>completed?=%u",
>>> __entry->dev, __entry->class, __entry->instance,
>>>- __entry->hw_id, __entry->ctx, __entry->seqno,
>>>- __entry->completed)
>>>+ __entry->ctx, __entry->seqno, __entry->completed)
>>> );
>>> #else
>>>@@ -829,7 +819,6 @@ TRACE_EVENT(i915_request_wait_begin,
>>> TP_STRUCT__entry(
>>> __field(u32, dev)
>>>- __field(u32, hw_id)
>>> __field(u64, ctx)
>>> __field(u16, class)
>>> __field(u16, instance)
>>>@@ -845,7 +834,6 @@ TRACE_EVENT(i915_request_wait_begin,
>>> */
>>> TP_fast_assign(
>>> __entry->dev = rq->i915->drm.primary->index;
>>>- __entry->hw_id = rq->gem_context->hw_id;
>>> __entry->class = rq->engine->uabi_class;
>>> __entry->instance = rq->engine->uabi_instance;
>>> __entry->ctx = rq->fence.context;
>>>@@ -853,9 +841,9 @@ TRACE_EVENT(i915_request_wait_begin,
>>> __entry->flags = flags;
>>> ),
>>>- TP_printk("dev=%u, engine=%u:%u, hw_id=%u, ctx=%llu,
>>>seqno=%u, flags=0x%x",
>>>+ TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u,
>>>flags=0x%x",
>>> __entry->dev, __entry->class, __entry->instance,
>>>- __entry->hw_id, __entry->ctx, __entry->seqno,
>>>+ __entry->ctx, __entry->seqno,
>>> __entry->flags)
>>> );
>>>@@ -958,19 +946,17 @@ DECLARE_EVENT_CLASS(i915_context,
>>> TP_STRUCT__entry(
>>> __field(u32, dev)
>>> __field(struct i915_gem_context *, ctx)
>>>- __field(u32, hw_id)
>>> __field(struct i915_address_space *, vm)
>>> ),
>>> TP_fast_assign(
>>> __entry->dev = ctx->i915->drm.primary->index;
>>> __entry->ctx = ctx;
>>>- __entry->hw_id = ctx->hw_id;
>>> __entry->vm = ctx->vm;
>>> ),
>>>- TP_printk("dev=%u, ctx=%p, ctx_vm=%p, hw_id=%u",
>>>- __entry->dev, __entry->ctx, __entry->vm, __entry->hw_id)
>>>+ TP_printk("dev=%u, ctx=%p, ctx_vm=%p",
>>>+ __entry->dev, __entry->ctx, __entry->vm)
>>> )
>>> DEFINE_EVENT(i915_context, i915_context_create,
>>>diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
>>>b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
>>>index 52d2df843148..f39f0282e78c 100644
>>>--- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
>>>+++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
>>>@@ -491,8 +491,8 @@ static int igt_evict_contexts(void *arg)
>>> if (IS_ERR(rq)) {
>>> /* When full, fail_if_busy will trigger EBUSY */
>>> if (PTR_ERR(rq) != -EBUSY) {
>>>- pr_err("Unexpected error from request alloc
>>>(ctx hw id %u, on %s): %d\n",
>>>- ctx->hw_id, engine->name,
>>>+ pr_err("Unexpected error from request alloc
>>>(on %s): %d\n",
>>>+ engine->name,
>>> (int)PTR_ERR(rq));
>>> err = PTR_ERR(rq);
>>> }
>>>diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c
>>>b/drivers/gpu/drm/i915/selftests/i915_vma.c
>>>index 1c9db08f7c28..ac1ff558eb90 100644
>>>--- a/drivers/gpu/drm/i915/selftests/i915_vma.c
>>>+++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
>>>@@ -170,7 +170,7 @@ static int igt_vma_create(void *arg)
>>> }
>>> nc = 0;
>>>- for_each_prime_number(num_ctx, MAX_CONTEXT_HW_ID) {
>>>+ for_each_prime_number(num_ctx, 2 * NUM_CONTEXT_TAG) {
>>> for (; nc < num_ctx; nc++) {
>>> ctx = mock_context(i915, "mock");
>>> if (!ctx)
>>>
>>
>>Unless I am missing something I see this patch as simplification and
>>not a fix. If we can get away with it in the GuC world then it's
>>quite a big simplification so it's fine by me.
>
>The GuC will require a fixed ID per intel_context for the whole time
>the context is in use/pinned. The GUC IDs are a global list (i.e. all
>engines share it) with 64k entries, so the logic to assign them is
>going to be a bit different and most likely confined to the guc
>context_pin/unpin callbacks, but we might bring back some of the
>low-level ID management logic that has been removed in this patch.
>
>An ID at the gem_context level definitely doesn't help on the guc side, so:
>
>Acked-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
>
>Daniele
>
>>
>>Regards,
>>
>>Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* [PATCH 21/27] drm/i915: Move context management under GEM
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (19 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 20/27] drm/i915: Remove logical HW ID Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-26 13:57 ` Tvrtko Ursulin
2019-09-25 10:01 ` [PATCH 22/27] drm/i915/overlay: Drop struct_mutex guard Chris Wilson
` (9 subsequent siblings)
30 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
Keep track of the GEM contexts underneath i915->gem.contexts and assign
them their own lock for the purposes of list management.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/gem/i915_gem_context.c | 155 +++++++-----------
drivers/gpu/drm/i915/gem/i915_gem_context.h | 4 +-
.../gpu/drm/i915/gem/i915_gem_context_types.h | 2 +-
.../gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 +-
drivers/gpu/drm/i915/gem/i915_gem_userptr.c | 2 +-
.../gpu/drm/i915/gem/selftests/huge_pages.c | 12 +-
.../drm/i915/gem/selftests/i915_gem_context.c | 122 ++++++--------
.../gpu/drm/i915/gem/selftests/mock_context.c | 4 +-
drivers/gpu/drm/i915/gt/intel_context.c | 8 +-
drivers/gpu/drm/i915/gt/selftest_context.c | 24 +--
drivers/gpu/drm/i915/gt/selftest_hangcheck.c | 19 +--
drivers/gpu/drm/i915/gt/selftest_lrc.c | 4 +-
.../gpu/drm/i915/gt/selftest_workarounds.c | 10 +-
drivers/gpu/drm/i915/gvt/scheduler.c | 20 +--
drivers/gpu/drm/i915/i915_debugfs.c | 50 +++---
drivers/gpu/drm/i915/i915_drv.c | 2 -
drivers/gpu/drm/i915/i915_drv.h | 27 +--
drivers/gpu/drm/i915/i915_gem.c | 10 +-
drivers/gpu/drm/i915/i915_gem_gtt.c | 4 +-
drivers/gpu/drm/i915/i915_perf.c | 24 ++-
drivers/gpu/drm/i915/i915_sysfs.c | 77 ++++-----
drivers/gpu/drm/i915/i915_trace.h | 2 +-
drivers/gpu/drm/i915/selftests/i915_gem.c | 8 -
.../gpu/drm/i915/selftests/i915_gem_evict.c | 3 -
drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 4 +-
drivers/gpu/drm/i915/selftests/i915_request.c | 11 +-
drivers/gpu/drm/i915/selftests/i915_vma.c | 5 +-
.../gpu/drm/i915/selftests/mock_gem_device.c | 6 +-
28 files changed, 272 insertions(+), 349 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index 4e59b809d901..a77f439358d7 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -219,9 +219,12 @@ static struct i915_gem_engines *default_engines(struct i915_gem_context *ctx)
static void i915_gem_context_free(struct i915_gem_context *ctx)
{
- lockdep_assert_held(&ctx->i915->drm.struct_mutex);
GEM_BUG_ON(!i915_gem_context_is_closed(ctx));
+ spin_lock(&ctx->i915->gem.contexts.lock);
+ list_del(&ctx->link);
+ spin_unlock(&ctx->i915->gem.contexts.lock);
+
free_engines(rcu_access_pointer(ctx->engines));
mutex_destroy(&ctx->engines_mutex);
@@ -231,56 +234,40 @@ static void i915_gem_context_free(struct i915_gem_context *ctx)
kfree(ctx->name);
put_pid(ctx->pid);
- list_del(&ctx->link);
mutex_destroy(&ctx->mutex);
kfree_rcu(ctx, rcu);
}
-static void contexts_free(struct drm_i915_private *i915)
+static void contexts_free_all(struct llist_node *list)
{
- struct llist_node *freed = llist_del_all(&i915->contexts.free_list);
struct i915_gem_context *ctx, *cn;
- lockdep_assert_held(&i915->drm.struct_mutex);
-
- llist_for_each_entry_safe(ctx, cn, freed, free_link)
+ llist_for_each_entry_safe(ctx, cn, list, free_link)
i915_gem_context_free(ctx);
}
-static void contexts_free_first(struct drm_i915_private *i915)
+static void contexts_flush_free(struct i915_gem_contexts *gc)
{
- struct i915_gem_context *ctx;
- struct llist_node *freed;
-
- lockdep_assert_held(&i915->drm.struct_mutex);
-
- freed = llist_del_first(&i915->contexts.free_list);
- if (!freed)
- return;
-
- ctx = container_of(freed, typeof(*ctx), free_link);
- i915_gem_context_free(ctx);
+ contexts_free_all(llist_del_all(&gc->free_list));
}
static void contexts_free_worker(struct work_struct *work)
{
- struct drm_i915_private *i915 =
- container_of(work, typeof(*i915), contexts.free_work);
+ struct i915_gem_contexts *gc =
+ container_of(work, typeof(*gc), free_work);
- mutex_lock(&i915->drm.struct_mutex);
- contexts_free(i915);
- mutex_unlock(&i915->drm.struct_mutex);
+ contexts_flush_free(gc);
}
void i915_gem_context_release(struct kref *ref)
{
struct i915_gem_context *ctx = container_of(ref, typeof(*ctx), ref);
- struct drm_i915_private *i915 = ctx->i915;
+ struct i915_gem_contexts *gc = &ctx->i915->gem.contexts;
trace_i915_context_free(ctx);
- if (llist_add(&ctx->free_link, &i915->contexts.free_list))
- queue_work(i915->wq, &i915->contexts.free_work);
+ if (llist_add(&ctx->free_link, &gc->free_list))
+ queue_work(ctx->i915->wq, &gc->free_work);
}
static inline struct i915_gem_engines *
@@ -359,8 +346,8 @@ static void context_close(struct i915_gem_context *ctx)
{
i915_gem_context_set_closed(ctx);
- if (ctx->vm)
- i915_vm_close(ctx->vm);
+ if (rcu_access_pointer(ctx->vm))
+ i915_vm_close(rcu_dereference_protected(ctx->vm, true));
mutex_lock(&ctx->mutex);
@@ -394,7 +381,6 @@ __create_context(struct drm_i915_private *i915)
return ERR_PTR(-ENOMEM);
kref_init(&ctx->ref);
- list_add_tail(&ctx->link, &i915->contexts.list);
ctx->i915 = i915;
ctx->sched.priority = I915_USER_PRIORITY(I915_PRIORITY_NORMAL);
mutex_init(&ctx->mutex);
@@ -435,6 +421,10 @@ __create_context(struct drm_i915_private *i915)
for (i = 0; i < ARRAY_SIZE(ctx->hang_timestamp); i++)
ctx->hang_timestamp[i] = jiffies - CONTEXT_FAST_HANG_JIFFIES;
+ spin_lock(&i915->gem.contexts.lock);
+ list_add_tail(&ctx->link, &i915->gem.contexts.list);
+ spin_unlock(&i915->gem.contexts.lock);
+
return ctx;
err_free:
@@ -464,11 +454,11 @@ static void __apply_ppgtt(struct intel_context *ce, void *vm)
static struct i915_address_space *
__set_ppgtt(struct i915_gem_context *ctx, struct i915_address_space *vm)
{
- struct i915_address_space *old = ctx->vm;
+ struct i915_address_space *old = rcu_dereference_protected(ctx->vm, 1);
GEM_BUG_ON(old && i915_vm_is_4lvl(vm) != i915_vm_is_4lvl(old));
- ctx->vm = i915_vm_open(vm);
+ rcu_assign_pointer(ctx->vm, i915_vm_open(vm));
context_apply_all(ctx, __apply_ppgtt, vm);
return old;
@@ -477,7 +467,7 @@ __set_ppgtt(struct i915_gem_context *ctx, struct i915_address_space *vm)
static void __assign_ppgtt(struct i915_gem_context *ctx,
struct i915_address_space *vm)
{
- if (vm == ctx->vm)
+ if (vm == rcu_access_pointer(ctx->vm))
return;
vm = __set_ppgtt(ctx, vm);
@@ -509,27 +499,25 @@ static void __assign_timeline(struct i915_gem_context *ctx,
}
static struct i915_gem_context *
-i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags)
+i915_gem_create_context(struct drm_i915_private *i915, unsigned int flags)
{
struct i915_gem_context *ctx;
- lockdep_assert_held(&dev_priv->drm.struct_mutex);
-
if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE &&
- !HAS_EXECLISTS(dev_priv))
+ !HAS_EXECLISTS(i915))
return ERR_PTR(-EINVAL);
- /* Reap the most stale context */
- contexts_free_first(dev_priv);
+ /* Reap the stale contexts */
+ contexts_flush_free(&i915->gem.contexts);
- ctx = __create_context(dev_priv);
+ ctx = __create_context(i915);
if (IS_ERR(ctx))
return ctx;
- if (HAS_FULL_PPGTT(dev_priv)) {
+ if (HAS_FULL_PPGTT(i915)) {
struct i915_ppgtt *ppgtt;
- ppgtt = i915_ppgtt_create(dev_priv);
+ ppgtt = i915_ppgtt_create(i915);
if (IS_ERR(ppgtt)) {
DRM_DEBUG_DRIVER("PPGTT setup failed (%ld)\n",
PTR_ERR(ppgtt));
@@ -544,7 +532,7 @@ i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags)
if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE) {
struct intel_timeline *timeline;
- timeline = intel_timeline_create(&dev_priv->gt, NULL);
+ timeline = intel_timeline_create(&i915->gt, NULL);
if (IS_ERR(timeline)) {
context_close(ctx);
return ERR_CAST(timeline);
@@ -590,48 +578,40 @@ i915_gem_context_create_kernel(struct drm_i915_private *i915, int prio)
return ctx;
}
-static void init_contexts(struct drm_i915_private *i915)
+static void init_contexts(struct i915_gem_contexts *gc)
{
- mutex_init(&i915->contexts.mutex);
- INIT_LIST_HEAD(&i915->contexts.list);
-
- /* Using the simple ida interface, the max is limited by sizeof(int) */
- BUILD_BUG_ON(MAX_CONTEXT_HW_ID > INT_MAX);
- BUILD_BUG_ON(GEN11_MAX_CONTEXT_HW_ID > INT_MAX);
- ida_init(&i915->contexts.hw_ida);
- INIT_LIST_HEAD(&i915->contexts.hw_id_list);
+ spin_lock_init(&gc->lock);
+ INIT_LIST_HEAD(&gc->list);
- INIT_WORK(&i915->contexts.free_work, contexts_free_worker);
- init_llist_head(&i915->contexts.free_list);
+ INIT_WORK(&gc->free_work, contexts_free_worker);
+ init_llist_head(&gc->free_list);
}
-int i915_gem_contexts_init(struct drm_i915_private *dev_priv)
+int i915_gem_init_contexts(struct drm_i915_private *i915)
{
struct i915_gem_context *ctx;
/* Reassure ourselves we are only called once */
- GEM_BUG_ON(dev_priv->kernel_context);
+ GEM_BUG_ON(i915->kernel_context);
- init_contexts(dev_priv);
+ init_contexts(&i915->gem.contexts);
/* lowest priority; idle task */
- ctx = i915_gem_context_create_kernel(dev_priv, I915_PRIORITY_MIN);
+ ctx = i915_gem_context_create_kernel(i915, I915_PRIORITY_MIN);
if (IS_ERR(ctx)) {
DRM_ERROR("Failed to create default global context\n");
return PTR_ERR(ctx);
}
- dev_priv->kernel_context = ctx;
+ i915->kernel_context = ctx;
DRM_DEBUG_DRIVER("%s context support initialized\n",
- DRIVER_CAPS(dev_priv)->has_logical_contexts ?
+ DRIVER_CAPS(i915)->has_logical_contexts ?
"logical" : "fake");
return 0;
}
-void i915_gem_contexts_fini(struct drm_i915_private *i915)
+void i915_gem_driver_release__contexts(struct drm_i915_private *i915)
{
- lockdep_assert_held(&i915->drm.struct_mutex);
-
destroy_kernel_context(&i915->kernel_context);
}
@@ -653,8 +633,8 @@ static int gem_context_register(struct i915_gem_context *ctx,
int ret;
ctx->file_priv = fpriv;
- if (ctx->vm)
- ctx->vm->file = fpriv;
+ if (rcu_access_pointer(ctx->vm))
+ rcu_dereference_protected(ctx->vm, true)->file = fpriv;
ctx->pid = get_task_pid(current, PIDTYPE_PID);
ctx->name = kasprintf(GFP_KERNEL, "%s[%d]",
@@ -691,9 +671,7 @@ int i915_gem_context_open(struct drm_i915_private *i915,
idr_init(&file_priv->context_idr);
idr_init_base(&file_priv->vm_idr, 1);
- mutex_lock(&i915->drm.struct_mutex);
ctx = i915_gem_create_context(i915, 0);
- mutex_unlock(&i915->drm.struct_mutex);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
goto err;
@@ -721,6 +699,7 @@ int i915_gem_context_open(struct drm_i915_private *i915,
void i915_gem_context_close(struct drm_file *file)
{
struct drm_i915_file_private *file_priv = file->driver_priv;
+ struct drm_i915_private *i915 = file_priv->dev_priv;
idr_for_each(&file_priv->context_idr, context_idr_cleanup, NULL);
idr_destroy(&file_priv->context_idr);
@@ -729,6 +708,8 @@ void i915_gem_context_close(struct drm_file *file)
idr_for_each(&file_priv->vm_idr, vm_idr_cleanup, NULL);
idr_destroy(&file_priv->vm_idr);
mutex_destroy(&file_priv->vm_idr_lock);
+
+ contexts_flush_free(&i915->gem.contexts);
}
int i915_gem_vm_create_ioctl(struct drm_device *dev, void *data,
@@ -907,16 +888,12 @@ static int get_ppgtt(struct drm_i915_file_private *file_priv,
struct i915_address_space *vm;
int ret;
- if (!ctx->vm)
+ if (!rcu_access_pointer(ctx->vm))
return -ENODEV;
- /* XXX rcu acquire? */
- ret = mutex_lock_interruptible(&ctx->i915->drm.struct_mutex);
- if (ret)
- return ret;
-
+ rcu_read_lock();
vm = i915_vm_get(ctx->vm);
- mutex_unlock(&ctx->i915->drm.struct_mutex);
+ rcu_read_unlock();
ret = mutex_lock_interruptible(&file_priv->vm_idr_lock);
if (ret)
@@ -1025,7 +1002,7 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv,
if (args->size)
return -EINVAL;
- if (!ctx->vm)
+ if (!rcu_access_pointer(ctx->vm))
return -ENODEV;
if (upper_32_bits(args->value))
@@ -1039,17 +1016,15 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv,
if (!vm)
return -ENOENT;
- err = mutex_lock_interruptible(&ctx->i915->drm.struct_mutex);
+ err = mutex_lock_interruptible(&ctx->mutex);
if (err)
goto out;
- if (vm == ctx->vm)
+ if (vm == rcu_access_pointer(ctx->vm))
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, vm);
@@ -1069,8 +1044,7 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv,
}
unlock:
- mutex_unlock(&ctx->i915->drm.struct_mutex);
-
+ mutex_unlock(&ctx->mutex);
out:
i915_vm_put(vm);
return err;
@@ -1953,7 +1927,7 @@ static int clone_vm(struct i915_gem_context *dst,
rcu_read_lock();
do {
- vm = READ_ONCE(src->vm);
+ vm = rcu_dereference(src->vm);
if (!vm)
break;
@@ -1975,7 +1949,7 @@ static int clone_vm(struct i915_gem_context *dst,
* it cannot be reallocated elsewhere.
*/
- if (vm == READ_ONCE(src->vm))
+ if (vm == rcu_access_pointer(src->vm))
break;
i915_vm_put(vm);
@@ -2077,12 +2051,7 @@ int i915_gem_context_create_ioctl(struct drm_device *dev, void *data,
return -EIO;
}
- ret = i915_mutex_lock_interruptible(dev);
- if (ret)
- return ret;
-
ext_data.ctx = i915_gem_create_context(i915, args->flags);
- mutex_unlock(&dev->struct_mutex);
if (IS_ERR(ext_data.ctx))
return PTR_ERR(ext_data.ctx);
@@ -2209,10 +2178,12 @@ int i915_gem_context_getparam_ioctl(struct drm_device *dev, void *data,
case I915_CONTEXT_PARAM_GTT_SIZE:
args->size = 0;
- if (ctx->vm)
- args->value = ctx->vm->total;
+ rcu_read_lock();
+ if (rcu_access_pointer(ctx->vm))
+ args->value = rcu_dereference(ctx->vm)->total;
else
args->value = to_i915(dev)->ggtt.vm.total;
+ rcu_read_unlock();
break;
case I915_CONTEXT_PARAM_NO_ERROR_CAPTURE:
@@ -2283,7 +2254,7 @@ int i915_gem_context_setparam_ioctl(struct drm_device *dev, void *data,
int i915_gem_context_reset_stats_ioctl(struct drm_device *dev,
void *data, struct drm_file *file)
{
- struct drm_i915_private *dev_priv = to_i915(dev);
+ struct drm_i915_private *i915 = to_i915(dev);
struct drm_i915_reset_stats *args = data;
struct i915_gem_context *ctx;
int ret;
@@ -2305,7 +2276,7 @@ int i915_gem_context_reset_stats_ioctl(struct drm_device *dev,
*/
if (capable(CAP_SYS_ADMIN))
- args->reset_count = i915_reset_count(&dev_priv->gpu_error);
+ args->reset_count = i915_reset_count(&i915->gpu_error);
else
args->reset_count = 0;
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.h b/drivers/gpu/drm/i915/gem/i915_gem_context.h
index 1b0df53436cf..4ee5dfc5794e 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.h
@@ -133,8 +133,8 @@ 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_fini(struct drm_i915_private *dev_priv);
+int __must_check i915_gem_init_contexts(struct drm_i915_private *i915);
+void i915_gem_driver_release__contexts(struct drm_i915_private *i915);
int i915_gem_context_open(struct drm_i915_private *i915,
struct drm_file *file);
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 6419da7c9f90..a3ecd19f2303 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
@@ -88,7 +88,7 @@ struct i915_gem_context {
* In other modes, this is a NULL pointer with the expectation that
* the caller uses the shared global GTT.
*/
- struct i915_address_space *vm;
+ struct i915_address_space __rcu *vm;
/**
* @pid: process id of creator
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 88a881be12ec..98816c35ffc3 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -728,7 +728,7 @@ static int eb_select_context(struct i915_execbuffer *eb)
return -ENOENT;
eb->gem_context = ctx;
- if (ctx->vm)
+ if (rcu_access_pointer(ctx->vm))
eb->invalid_flags |= EXEC_OBJECT_NEEDS_GTT;
eb->context_flags = 0;
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
index 27573ef903b4..7c225e1d2a11 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
@@ -757,7 +757,7 @@ i915_gem_userptr_ioctl(struct drm_device *dev,
* On almost all of the older hw, we cannot tell the GPU that
* a page is readonly.
*/
- vm = dev_priv->kernel_context->vm;
+ vm = rcu_dereference_protected(dev_priv->kernel_context->vm, true);
if (!vm || !vm->has_read_only)
return -ENODEV;
}
diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
index 98b2a6ccfcc1..e204e653b459 100644
--- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
+++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
@@ -1322,7 +1322,7 @@ static int igt_ppgtt_pin_update(void *arg)
struct i915_gem_context *ctx = arg;
struct drm_i915_private *dev_priv = ctx->i915;
unsigned long supported = INTEL_INFO(dev_priv)->page_sizes;
- struct i915_address_space *vm = ctx->vm;
+ struct i915_address_space *vm = rcu_dereference_protected(ctx->vm, 1);
struct drm_i915_gem_object *obj;
struct i915_gem_engines_iter it;
struct intel_context *ce;
@@ -1460,7 +1460,8 @@ static int igt_tmpfs_fallback(void *arg)
struct i915_gem_context *ctx = arg;
struct drm_i915_private *i915 = ctx->i915;
struct vfsmount *gemfs = i915->mm.gemfs;
- struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm;
+ struct i915_address_space *vm =
+ rcu_dereference_protected(ctx->vm, true) ?: &i915->ggtt.vm;
struct drm_i915_gem_object *obj;
struct i915_vma *vma;
u32 *vaddr;
@@ -1517,7 +1518,8 @@ static int igt_shrink_thp(void *arg)
{
struct i915_gem_context *ctx = arg;
struct drm_i915_private *i915 = ctx->i915;
- struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm;
+ struct i915_address_space *vm =
+ rcu_dereference_protected(ctx->vm, true) ?: &i915->ggtt.vm;
struct drm_i915_gem_object *obj;
struct i915_gem_engines_iter it;
struct intel_context *ce;
@@ -1699,8 +1701,8 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *i915)
goto out_unlock;
}
- if (ctx->vm)
- ctx->vm->scrub_64K = true;
+ if (rcu_access_pointer(ctx->vm))
+ rcu_dereference_protected(ctx->vm, true)->scrub_64K = true;
err = i915_subtests(tests, ctx);
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 a8a9293f4ac6..7de1c0078d6f 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
@@ -53,19 +53,17 @@ static int live_nop_switch(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(&i915->drm.struct_mutex);
-
ctx = kcalloc(nctx, sizeof(*ctx), GFP_KERNEL);
if (!ctx) {
err = -ENOMEM;
- goto out_unlock;
+ goto out_file;
}
for (n = 0; n < nctx; n++) {
ctx[n] = live_context(i915, file);
if (IS_ERR(ctx[n])) {
err = PTR_ERR(ctx[n]);
- goto out_unlock;
+ goto out_file;
}
}
@@ -79,7 +77,7 @@ static int live_nop_switch(void *arg)
rq = igt_request_alloc(ctx[n], engine);
if (IS_ERR(rq)) {
err = PTR_ERR(rq);
- goto out_unlock;
+ goto out_file;
}
i915_request_add(rq);
}
@@ -87,7 +85,7 @@ static int live_nop_switch(void *arg)
pr_err("Failed to populated %d contexts\n", nctx);
intel_gt_set_wedged(&i915->gt);
err = -EIO;
- goto out_unlock;
+ goto out_file;
}
times[1] = ktime_get_raw();
@@ -97,7 +95,7 @@ static int live_nop_switch(void *arg)
err = igt_live_test_begin(&t, i915, __func__, engine->name);
if (err)
- goto out_unlock;
+ goto out_file;
end_time = jiffies + i915_selftest.timeout_jiffies;
for_each_prime_number_from(prime, 2, 8192) {
@@ -107,7 +105,7 @@ static int live_nop_switch(void *arg)
rq = igt_request_alloc(ctx[n % nctx], engine);
if (IS_ERR(rq)) {
err = PTR_ERR(rq);
- goto out_unlock;
+ goto out_file;
}
/*
@@ -143,7 +141,7 @@ static int live_nop_switch(void *arg)
err = igt_live_test_end(&t);
if (err)
- goto out_unlock;
+ goto out_file;
pr_info("Switch latencies on %s: 1 = %lluns, %lu = %lluns\n",
engine->name,
@@ -151,8 +149,7 @@ static int live_nop_switch(void *arg)
prime - 1, div64_u64(ktime_to_ns(times[1]), prime - 1));
}
-out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
+out_file:
mock_file_free(i915, file);
return err;
}
@@ -412,11 +409,9 @@ static int igt_ctx_exec(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(&i915->drm.struct_mutex);
-
err = igt_live_test_begin(&t, i915, __func__, engine->name);
if (err)
- goto out_unlock;
+ goto out_file;
ncontexts = 0;
ndwords = 0;
@@ -428,7 +423,7 @@ static int igt_ctx_exec(void *arg)
ctx = kernel_context(i915);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
- goto out_unlock;
+ goto out_file;
}
ce = i915_gem_context_get_engine(ctx, engine->legacy_idx);
@@ -440,7 +435,7 @@ static int igt_ctx_exec(void *arg)
err = PTR_ERR(obj);
intel_context_put(ce);
kernel_context_close(ctx);
- goto out_unlock;
+ goto out_file;
}
}
@@ -449,17 +444,18 @@ static int igt_ctx_exec(void *arg)
pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
ndwords, dw, max_dwords(obj),
engine->name,
- yesno(!!ctx->vm), err);
+ yesno(!!rcu_access_pointer(ctx->vm)),
+ err);
intel_context_put(ce);
kernel_context_close(ctx);
- goto out_unlock;
+ goto out_file;
}
err = throttle(ce, tq, ARRAY_SIZE(tq));
if (err) {
intel_context_put(ce);
kernel_context_close(ctx);
- goto out_unlock;
+ goto out_file;
}
if (++dw == max_dwords(obj)) {
@@ -489,11 +485,10 @@ static int igt_ctx_exec(void *arg)
dw += rem;
}
-out_unlock:
+out_file:
throttle_release(tq, ARRAY_SIZE(tq));
if (igt_live_test_end(&t))
err = -EIO;
- mutex_unlock(&i915->drm.struct_mutex);
mock_file_free(i915, file);
if (err)
@@ -528,22 +523,20 @@ static int igt_shared_ctx_exec(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(&i915->drm.struct_mutex);
-
parent = live_context(i915, file);
if (IS_ERR(parent)) {
err = PTR_ERR(parent);
- goto out_unlock;
+ goto out_file;
}
if (!parent->vm) { /* not full-ppgtt; nothing to share */
err = 0;
- goto out_unlock;
+ goto out_file;
}
err = igt_live_test_begin(&t, i915, __func__, "");
if (err)
- goto out_unlock;
+ goto out_file;
for_each_engine(engine, i915, id) {
unsigned long ncontexts, ndwords, dw;
@@ -587,7 +580,8 @@ static int igt_shared_ctx_exec(void *arg)
pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
ndwords, dw, max_dwords(obj),
engine->name,
- yesno(!!ctx->vm), err);
+ yesno(!!rcu_access_pointer(ctx->vm)),
+ err);
intel_context_put(ce);
kernel_context_close(ctx);
goto out_test;
@@ -626,17 +620,13 @@ static int igt_shared_ctx_exec(void *arg)
dw += rem;
}
- mutex_unlock(&i915->drm.struct_mutex);
i915_gem_drain_freed_objects(i915);
- mutex_lock(&i915->drm.struct_mutex);
}
out_test:
throttle_release(tq, ARRAY_SIZE(tq));
if (igt_live_test_end(&t))
err = -EIO;
-out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
-
+out_file:
mock_file_free(i915, file);
return err;
}
@@ -1008,8 +998,6 @@ __igt_ctx_sseu(struct drm_i915_private *i915,
if (flags & TEST_RESET)
igt_global_reset_lock(&i915->gt);
- mutex_lock(&i915->drm.struct_mutex);
-
ctx = live_context(i915, file);
if (IS_ERR(ctx)) {
ret = PTR_ERR(ctx);
@@ -1064,8 +1052,6 @@ __igt_ctx_sseu(struct drm_i915_private *i915,
i915_gem_object_put(obj);
out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
-
if (flags & TEST_RESET)
igt_global_reset_unlock(&i915->gt);
@@ -1125,23 +1111,24 @@ static int igt_ctx_readonly(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(&i915->drm.struct_mutex);
-
err = igt_live_test_begin(&t, i915, __func__, "");
if (err)
- goto out_unlock;
+ goto out_file;
ctx = live_context(i915, file);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
- goto out_unlock;
+ goto out_file;
}
- vm = ctx->vm ?: &i915->ggtt.alias->vm;
+ rcu_read_lock();
+ vm = rcu_dereference(ctx->vm) ?: &i915->ggtt.alias->vm;
if (!vm || !vm->has_read_only) {
+ rcu_read_unlock();
err = 0;
- goto out_unlock;
+ goto out_file;
}
+ rcu_read_unlock();
ndwords = 0;
dw = 0;
@@ -1159,7 +1146,7 @@ static int igt_ctx_readonly(void *arg)
if (IS_ERR(obj)) {
err = PTR_ERR(obj);
i915_gem_context_unlock_engines(ctx);
- goto out_unlock;
+ goto out_file;
}
if (prandom_u32_state(&prng) & 1)
@@ -1170,15 +1157,17 @@ static int igt_ctx_readonly(void *arg)
if (err) {
pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
ndwords, dw, max_dwords(obj),
- ce->engine->name, yesno(!!ctx->vm), err);
+ ce->engine->name,
+ yesno(!!rcu_access_pointer(ctx->vm)),
+ err);
i915_gem_context_unlock_engines(ctx);
- goto out_unlock;
+ goto out_file;
}
err = throttle(ce, tq, ARRAY_SIZE(tq));
if (err) {
i915_gem_context_unlock_engines(ctx);
- goto out_unlock;
+ goto out_file;
}
if (++dw == max_dwords(obj)) {
@@ -1210,11 +1199,10 @@ static int igt_ctx_readonly(void *arg)
dw += rem;
}
-out_unlock:
+out_file:
throttle_release(tq, ARRAY_SIZE(tq));
if (igt_live_test_end(&t))
err = -EIO;
- mutex_unlock(&i915->drm.struct_mutex);
mock_file_free(i915, file);
return err;
@@ -1223,7 +1211,7 @@ static int igt_ctx_readonly(void *arg)
static int check_scratch(struct i915_gem_context *ctx, u64 offset)
{
struct drm_mm_node *node =
- __drm_mm_interval_first(&ctx->vm->mm,
+ __drm_mm_interval_first(&rcu_dereference_protected(ctx->vm, true)->mm,
offset, offset + sizeof(u32) - 1);
if (!node || node->start > offset)
return 0;
@@ -1273,7 +1261,9 @@ static int write_to_scratch(struct i915_gem_context *ctx,
intel_gt_chipset_flush(engine->gt);
- vma = i915_vma_instance(obj, ctx->vm, NULL);
+ vma = i915_vma_instance(obj,
+ rcu_dereference_protected(ctx->vm, true),
+ NULL);
if (IS_ERR(vma)) {
err = PTR_ERR(vma);
goto err;
@@ -1372,7 +1362,9 @@ static int read_from_scratch(struct i915_gem_context *ctx,
intel_gt_chipset_flush(engine->gt);
- vma = i915_vma_instance(obj, ctx->vm, NULL);
+ vma = i915_vma_instance(obj,
+ rcu_dereference_protected(ctx->vm, true),
+ NULL);
if (IS_ERR(vma)) {
err = PTR_ERR(vma);
goto err;
@@ -1463,27 +1455,25 @@ static int igt_vm_isolation(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(&i915->drm.struct_mutex);
-
err = igt_live_test_begin(&t, i915, __func__, "");
if (err)
- goto out_unlock;
+ goto out_file;
ctx_a = live_context(i915, file);
if (IS_ERR(ctx_a)) {
err = PTR_ERR(ctx_a);
- goto out_unlock;
+ goto out_file;
}
ctx_b = live_context(i915, file);
if (IS_ERR(ctx_b)) {
err = PTR_ERR(ctx_b);
- goto out_unlock;
+ goto out_file;
}
/* We can only test vm isolation, if the vm are distinct */
if (ctx_a->vm == ctx_b->vm)
- goto out_unlock;
+ goto out_file;
vm_total = ctx_a->vm->total;
GEM_BUG_ON(ctx_b->vm->total != vm_total);
@@ -1512,7 +1502,7 @@ static int igt_vm_isolation(void *arg)
err = read_from_scratch(ctx_b, engine,
offset, &value);
if (err)
- goto out_unlock;
+ goto out_file;
if (value) {
pr_err("%s: Read %08x from scratch (offset 0x%08x_%08x), after %lu reads!\n",
@@ -1521,7 +1511,7 @@ static int igt_vm_isolation(void *arg)
lower_32_bits(offset),
this);
err = -EINVAL;
- goto out_unlock;
+ goto out_file;
}
this++;
@@ -1531,11 +1521,9 @@ static int igt_vm_isolation(void *arg)
pr_info("Checked %lu scratch offsets across %d engines\n",
count, RUNTIME_INFO(i915)->num_engines);
-out_unlock:
+out_file:
if (igt_live_test_end(&t))
err = -EIO;
- mutex_unlock(&i915->drm.struct_mutex);
-
mock_file_free(i915, file);
return err;
}
@@ -1567,13 +1555,9 @@ static int mock_context_barrier(void *arg)
* a request; useful for retiring old state after loading new.
*/
- mutex_lock(&i915->drm.struct_mutex);
-
ctx = mock_context(i915, "mock");
- if (!ctx) {
- err = -ENOMEM;
- goto unlock;
- }
+ if (!ctx)
+ return -ENOMEM;
counter = 0;
err = context_barrier_task(ctx, 0,
@@ -1646,8 +1630,6 @@ static int mock_context_barrier(void *arg)
out:
mock_context_close(ctx);
-unlock:
- mutex_unlock(&i915->drm.struct_mutex);
return err;
#undef pr_fmt
#define pr_fmt(x) x
diff --git a/drivers/gpu/drm/i915/gem/selftests/mock_context.c b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
index ebc46f098561..6ef86e7923a7 100644
--- a/drivers/gpu/drm/i915/gem/selftests/mock_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
@@ -67,7 +67,7 @@ void mock_context_close(struct i915_gem_context *ctx)
void mock_init_contexts(struct drm_i915_private *i915)
{
- init_contexts(i915);
+ init_contexts(&i915->gem.contexts);
}
struct i915_gem_context *
@@ -76,8 +76,6 @@ live_context(struct drm_i915_private *i915, struct drm_file *file)
struct i915_gem_context *ctx;
int err;
- lockdep_assert_held(&i915->drm.struct_mutex);
-
ctx = i915_gem_create_context(i915, 0);
if (IS_ERR(ctx))
return ctx;
diff --git a/drivers/gpu/drm/i915/gt/intel_context.c b/drivers/gpu/drm/i915/gt/intel_context.c
index 35a40c2820a2..107eb03d6dd4 100644
--- a/drivers/gpu/drm/i915/gt/intel_context.c
+++ b/drivers/gpu/drm/i915/gt/intel_context.c
@@ -226,7 +226,13 @@ intel_context_init(struct intel_context *ce,
kref_init(&ce->ref);
ce->gem_context = ctx;
- ce->vm = i915_vm_get(ctx->vm ?: &engine->gt->ggtt->vm);
+ rcu_read_lock();
+ ce->vm = rcu_dereference(ctx->vm);
+ if (ce->vm && !kref_get_unless_zero(&ce->vm->ref))
+ ce->vm = NULL;
+ if (!ce->vm)
+ ce->vm = i915_vm_get(&engine->gt->ggtt->vm);
+ rcu_read_unlock();
if (ctx->timeline)
ce->timeline = intel_timeline_get(ctx->timeline);
diff --git a/drivers/gpu/drm/i915/gt/selftest_context.c b/drivers/gpu/drm/i915/gt/selftest_context.c
index 86cffbb0a9cb..7c838a57e174 100644
--- a/drivers/gpu/drm/i915/gt/selftest_context.c
+++ b/drivers/gpu/drm/i915/gt/selftest_context.c
@@ -155,13 +155,9 @@ static int live_context_size(void *arg)
* HW tries to write past the end of one.
*/
- mutex_lock(>->i915->drm.struct_mutex);
-
fixme = kernel_context(gt->i915);
- if (IS_ERR(fixme)) {
- err = PTR_ERR(fixme);
- goto unlock;
- }
+ if (IS_ERR(fixme))
+ return PTR_ERR(fixme);
for_each_engine(engine, gt->i915, id) {
struct {
@@ -201,8 +197,6 @@ static int live_context_size(void *arg)
}
kernel_context_close(fixme);
-unlock:
- mutex_unlock(>->i915->drm.struct_mutex);
return err;
}
@@ -305,12 +299,10 @@ static int live_active_context(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(>->i915->drm.struct_mutex);
-
fixme = live_context(gt->i915, file);
if (IS_ERR(fixme)) {
err = PTR_ERR(fixme);
- goto unlock;
+ goto out_file;
}
for_each_engine(engine, gt->i915, id) {
@@ -323,8 +315,7 @@ static int live_active_context(void *arg)
break;
}
-unlock:
- mutex_unlock(>->i915->drm.struct_mutex);
+out_file:
mock_file_free(gt->i915, file);
return err;
}
@@ -418,12 +409,10 @@ static int live_remote_context(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(>->i915->drm.struct_mutex);
-
fixme = live_context(gt->i915, file);
if (IS_ERR(fixme)) {
err = PTR_ERR(fixme);
- goto unlock;
+ goto out_file;
}
for_each_engine(engine, gt->i915, id) {
@@ -436,8 +425,7 @@ static int live_remote_context(void *arg)
break;
}
-unlock:
- mutex_unlock(>->i915->drm.struct_mutex);
+out_file:
mock_file_free(gt->i915, file);
return err;
}
diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
index f2b5ab1e5a2e..1b836ff0a756 100644
--- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
+++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
@@ -58,9 +58,7 @@ static int hang_init(struct hang *h, struct intel_gt *gt)
memset(h, 0, sizeof(*h));
h->gt = gt;
- mutex_lock(>->i915->drm.struct_mutex);
h->ctx = kernel_context(gt->i915);
- mutex_unlock(>->i915->drm.struct_mutex);
if (IS_ERR(h->ctx))
return PTR_ERR(h->ctx);
@@ -133,7 +131,8 @@ static struct i915_request *
hang_create_request(struct hang *h, struct intel_engine_cs *engine)
{
struct intel_gt *gt = h->gt;
- struct i915_address_space *vm = h->ctx->vm ?: &engine->gt->ggtt->vm;
+ struct i915_address_space *vm =
+ rcu_dereference_protected(h->ctx->vm, true) ?: &engine->gt->ggtt->vm;
struct drm_i915_gem_object *obj;
struct i915_request *rq = NULL;
struct i915_vma *hws, *vma;
@@ -382,9 +381,7 @@ static int igt_reset_nop(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(>->i915->drm.struct_mutex);
ctx = live_context(gt->i915, file);
- mutex_unlock(>->i915->drm.struct_mutex);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
goto out;
@@ -458,9 +455,7 @@ static int igt_reset_nop_engine(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(>->i915->drm.struct_mutex);
ctx = live_context(gt->i915, file);
- mutex_unlock(>->i915->drm.struct_mutex);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
goto out;
@@ -705,9 +700,7 @@ static int active_engine(void *data)
return PTR_ERR(file);
for (count = 0; count < ARRAY_SIZE(ctx); count++) {
- mutex_lock(&engine->i915->drm.struct_mutex);
ctx[count] = live_context(engine->i915, file);
- mutex_unlock(&engine->i915->drm.struct_mutex);
if (IS_ERR(ctx[count])) {
err = PTR_ERR(ctx[count]);
while (--count)
@@ -1298,18 +1291,18 @@ static int igt_reset_evict_ppgtt(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(>->i915->drm.struct_mutex);
ctx = live_context(gt->i915, file);
- mutex_unlock(>->i915->drm.struct_mutex);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
goto out;
}
err = 0;
- if (ctx->vm) /* aliasing == global gtt locking, covered above */
- err = __igt_reset_evict_vma(gt, ctx->vm,
+ if (rcu_access_pointer(ctx->vm)) {
+ /* aliasing == global gtt locking, covered above */
+ err = __igt_reset_evict_vma(gt, rcu_dereference_protected(ctx->vm, true),
evict_vma, EXEC_OBJECT_WRITE);
+ }
out:
mock_file_free(gt->i915, file);
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index 3300acd9199d..e33c2f66c683 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -1453,7 +1453,9 @@ static int smoke_submit(struct preempt_smoke *smoke,
int err = 0;
if (batch) {
- vma = i915_vma_instance(batch, ctx->vm, NULL);
+ vma = i915_vma_instance(batch,
+ rcu_dereference_protected(ctx->vm, true),
+ NULL);
if (IS_ERR(vma))
return PTR_ERR(vma);
diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
index 06351fefbbf3..ed44333d9e20 100644
--- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
@@ -33,6 +33,8 @@ struct wa_lists {
} engine[I915_NUM_ENGINES];
};
+#define ctx_vm(ctx) rcu_dereference_protected((ctx)->vm, true)
+
static void
reference_lists_init(struct drm_i915_private *i915, struct wa_lists *lists)
{
@@ -362,7 +364,7 @@ static struct i915_vma *create_batch(struct i915_gem_context *ctx)
if (IS_ERR(obj))
return ERR_CAST(obj);
- vma = i915_vma_instance(obj, ctx->vm, NULL);
+ vma = i915_vma_instance(obj, ctx_vm(ctx), NULL);
if (IS_ERR(vma)) {
err = PTR_ERR(vma);
goto err_obj;
@@ -468,7 +470,7 @@ static int check_dirty_whitelist(struct i915_gem_context *ctx,
int err = 0, i, v;
u32 *cs, *results;
- scratch = create_scratch(ctx->vm, 2 * ARRAY_SIZE(values) + 1);
+ scratch = create_scratch(ctx_vm(ctx), 2 * ARRAY_SIZE(values) + 1);
if (IS_ERR(scratch))
return PTR_ERR(scratch);
@@ -1018,14 +1020,14 @@ static int live_isolated_whitelist(void *arg)
goto err;
}
- client[i].scratch[0] = create_scratch(c->vm, 1024);
+ client[i].scratch[0] = create_scratch(ctx_vm(c), 1024);
if (IS_ERR(client[i].scratch[0])) {
err = PTR_ERR(client[i].scratch[0]);
kernel_context_close(c);
goto err;
}
- client[i].scratch[1] = create_scratch(c->vm, 1024);
+ client[i].scratch[1] = create_scratch(ctx_vm(c), 1024);
if (IS_ERR(client[i].scratch[1])) {
err = PTR_ERR(client[i].scratch[1]);
i915_vma_unpin_and_release(&client[i].scratch[0], 0);
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index 03f567084548..ce5cff01f44c 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -365,7 +365,8 @@ static void set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload,
struct i915_gem_context *ctx)
{
struct intel_vgpu_mm *mm = workload->shadow_mm;
- struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(ctx->vm);
+ struct i915_ppgtt *ppgtt =
+ i915_vm_to_ppgtt(rcu_dereference_protected(ctx->vm, true));
int i = 0;
if (mm->ppgtt_mm.root_entry_type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY) {
@@ -1230,20 +1231,18 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
struct intel_vgpu_submission *s = &vgpu->submission;
struct intel_engine_cs *engine;
struct i915_gem_context *ctx;
+ struct i915_ppgtt *ppgtt;
enum intel_engine_id i;
int ret;
- mutex_lock(&i915->drm.struct_mutex);
-
ctx = i915_gem_context_create_kernel(i915, I915_PRIORITY_MAX);
- if (IS_ERR(ctx)) {
- ret = PTR_ERR(ctx);
- goto out_unlock;
- }
+ if (IS_ERR(ctx))
+ return PTR_ERR(ctx);
i915_gem_context_set_force_single_submission(ctx);
- i915_context_ppgtt_root_save(s, i915_vm_to_ppgtt(ctx->vm));
+ ppgtt = i915_vm_to_ppgtt(rcu_dereference_protected(ctx->vm, true));
+ i915_context_ppgtt_root_save(s, ppgtt);
for_each_engine(engine, i915, i) {
struct intel_context *ce;
@@ -1289,11 +1288,10 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
bitmap_zero(s->tlb_handle_pending, I915_NUM_ENGINES);
i915_gem_context_put(ctx);
- mutex_unlock(&i915->drm.struct_mutex);
return 0;
out_shadow_ctx:
- i915_context_ppgtt_root_restore(s, i915_vm_to_ppgtt(ctx->vm));
+ i915_context_ppgtt_root_restore(s, ppgtt);
for_each_engine(engine, i915, i) {
if (IS_ERR(s->shadow[i]))
break;
@@ -1302,8 +1300,6 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
intel_context_put(s->shadow[i]);
}
i915_gem_context_put(ctx);
-out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
return ret;
}
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 8caaa446490f..d4a7cda60679 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -315,12 +315,18 @@ static void print_context_stats(struct seq_file *m,
struct drm_i915_private *i915)
{
struct file_stats kstats = {};
- struct i915_gem_context *ctx;
+ struct i915_gem_context *ctx, *cn;
- list_for_each_entry(ctx, &i915->contexts.list, link) {
+ spin_lock(&i915->gem.contexts.lock);
+ list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) {
struct i915_gem_engines_iter it;
struct intel_context *ce;
+ if (!kref_get_unless_zero(&ctx->ref))
+ continue;
+
+ spin_unlock(&i915->gem.contexts.lock);
+
for_each_gem_engine(ce,
i915_gem_context_lock_engines(ctx), it) {
intel_context_lock_pinned(ce);
@@ -337,7 +343,9 @@ static void print_context_stats(struct seq_file *m,
i915_gem_context_unlock_engines(ctx);
if (!IS_ERR_OR_NULL(ctx->file_priv)) {
- struct file_stats stats = { .vm = ctx->vm, };
+ struct file_stats stats = {
+ .vm = rcu_access_pointer(ctx->vm),
+ };
struct drm_file *file = ctx->file_priv->file;
struct task_struct *task;
char name[80];
@@ -354,7 +362,12 @@ static void print_context_stats(struct seq_file *m,
print_file_stats(m, name, stats);
}
+
+ spin_lock(&i915->gem.contexts.lock);
+ list_safe_reset_next(ctx, cn, link);
+ i915_gem_context_put(ctx);
}
+ spin_unlock(&i915->gem.contexts.lock);
print_file_stats(m, "[k]contexts", kstats);
}
@@ -362,7 +375,6 @@ static void print_context_stats(struct seq_file *m,
static int i915_gem_object_info(struct seq_file *m, void *data)
{
struct drm_i915_private *i915 = node_to_i915(m->private);
- int ret;
seq_printf(m, "%u shrinkable [%u free] objects, %llu bytes\n",
i915->mm.shrink_count,
@@ -371,12 +383,7 @@ static int i915_gem_object_info(struct seq_file *m, void *data)
seq_putc(m, '\n');
- ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
- if (ret)
- return ret;
-
print_context_stats(m, i915);
- mutex_unlock(&i915->drm.struct_mutex);
return 0;
}
@@ -1490,19 +1497,19 @@ static void describe_ctx_ring(struct seq_file *m, struct intel_ring *ring)
static int i915_context_status(struct seq_file *m, void *unused)
{
- struct drm_i915_private *dev_priv = node_to_i915(m->private);
- struct drm_device *dev = &dev_priv->drm;
- struct i915_gem_context *ctx;
- int ret;
-
- ret = mutex_lock_interruptible(&dev->struct_mutex);
- if (ret)
- return ret;
+ struct drm_i915_private *i915 = node_to_i915(m->private);
+ struct i915_gem_context *ctx, *cn;
- list_for_each_entry(ctx, &dev_priv->contexts.list, link) {
+ spin_lock(&i915->gem.contexts.lock);
+ list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) {
struct i915_gem_engines_iter it;
struct intel_context *ce;
+ if (!kref_get_unless_zero(&ctx->ref))
+ continue;
+
+ spin_unlock(&i915->gem.contexts.lock);
+
seq_puts(m, "HW context ");
if (ctx->pid) {
struct task_struct *task;
@@ -1537,9 +1544,12 @@ static int i915_context_status(struct seq_file *m, void *unused)
i915_gem_context_unlock_engines(ctx);
seq_putc(m, '\n');
- }
- mutex_unlock(&dev->struct_mutex);
+ spin_lock(&i915->gem.contexts.lock);
+ list_safe_reset_next(ctx, cn, link);
+ i915_gem_context_put(ctx);
+ }
+ spin_unlock(&i915->gem.contexts.lock);
return 0;
}
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 8cac6124d16a..a8ddac22d633 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -1685,10 +1685,8 @@ static void i915_driver_postclose(struct drm_device *dev, struct drm_file *file)
{
struct drm_i915_file_private *file_priv = file->driver_priv;
- mutex_lock(&dev->struct_mutex);
i915_gem_context_close(file);
i915_gem_release(dev, file);
- mutex_unlock(&dev->struct_mutex);
kfree_rcu(file_priv, rcu);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index ad7ed7f72da8..23f311172e75 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1543,25 +1543,6 @@ struct drm_i915_private {
int audio_power_refcount;
u32 audio_freq_cntrl;
- struct {
- struct mutex mutex;
- struct list_head list;
- struct llist_head free_list;
- struct work_struct free_work;
-
- /* The hw wants to have a stable context identifier for the
- * lifetime of the context (for OA, PASID, faults, etc).
- * This is limited in execlists to 21 bits.
- */
- struct ida hw_ida;
-#define MAX_CONTEXT_HW_ID (1<<21) /* exclusive */
-#define MAX_GUC_CONTEXT_HW_ID (1 << 20) /* exclusive */
-#define GEN11_MAX_CONTEXT_HW_ID (1<<11) /* exclusive */
-/* in Gen12 ID 0x7FF is reserved to indicate idle */
-#define GEN12_MAX_CONTEXT_HW_ID (GEN11_MAX_CONTEXT_HW_ID - 1)
- struct list_head hw_id_list;
- } contexts;
-
u32 fdi_rx_config;
/* Shadow for DISPLAY_PHY_CONTROL which can't be safely read */
@@ -1717,6 +1698,14 @@ struct drm_i915_private {
struct {
struct notifier_block pm_notifier;
+
+ struct i915_gem_contexts {
+ spinlock_t lock; /* locks list */
+ struct list_head list;
+
+ struct llist_head free_list;
+ struct work_struct free_work;
+ } contexts;
} gem;
/* For i945gm vblank irq vs. C3 workaround */
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 9c2877441540..8814db7023db 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1263,7 +1263,7 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
goto err_unlock;
}
- ret = i915_gem_contexts_init(dev_priv);
+ ret = i915_gem_init_contexts(dev_priv);
if (ret) {
GEM_BUG_ON(ret == -EIO);
goto err_scratch;
@@ -1345,7 +1345,7 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
}
err_context:
if (ret != -EIO)
- i915_gem_contexts_fini(dev_priv);
+ i915_gem_driver_release__contexts(dev_priv);
err_scratch:
intel_gt_driver_release(&dev_priv->gt);
err_unlock:
@@ -1413,11 +1413,9 @@ void i915_gem_driver_remove(struct drm_i915_private *dev_priv)
void i915_gem_driver_release(struct drm_i915_private *dev_priv)
{
- mutex_lock(&dev_priv->drm.struct_mutex);
intel_engines_cleanup(dev_priv);
- i915_gem_contexts_fini(dev_priv);
+ i915_gem_driver_release__contexts(dev_priv);
intel_gt_driver_release(&dev_priv->gt);
- mutex_unlock(&dev_priv->drm.struct_mutex);
intel_wa_list_free(&dev_priv->gt_wa_list);
@@ -1427,7 +1425,7 @@ void i915_gem_driver_release(struct drm_i915_private *dev_priv)
i915_gem_drain_freed_objects(dev_priv);
- WARN_ON(!list_empty(&dev_priv->contexts.list));
+ WARN_ON(!list_empty(&dev_priv->gem.contexts.list));
}
void i915_gem_init_mmio(struct drm_i915_private *i915)
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 0d040517787b..db32e0fe78b1 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -1366,7 +1366,9 @@ static int gen8_init_scratch(struct i915_address_space *vm)
if (vm->has_read_only &&
vm->i915->kernel_context &&
vm->i915->kernel_context->vm) {
- struct i915_address_space *clone = vm->i915->kernel_context->vm;
+ struct i915_address_space *clone =
+ rcu_dereference_protected(vm->i915->kernel_context->vm,
+ true);
GEM_BUG_ON(!clone->has_read_only);
diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index d36ba248d438..bfe6e9b6d71d 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -1853,8 +1853,8 @@ static int gen8_configure_all_contexts(struct i915_perf_stream *stream,
};
#undef ctx_flexeuN
struct intel_engine_cs *engine;
- struct i915_gem_context *ctx;
- int i;
+ struct i915_gem_context *ctx, *cn;
+ int i, err;
for (i = 2; i < ARRAY_SIZE(regs); i++)
regs[i].value = oa_config_flex_reg(oa_config, regs[i].reg);
@@ -1877,16 +1877,27 @@ static int gen8_configure_all_contexts(struct i915_perf_stream *stream,
* context. Contexts idle at the time of reconfiguration are not
* trapped behind the barrier.
*/
- list_for_each_entry(ctx, &i915->contexts.list, link) {
- int err;
-
+ spin_lock(&i915->gem.contexts.lock);
+ list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) {
if (ctx == i915->kernel_context)
continue;
+ if (!kref_get_unless_zero(&ctx->ref))
+ continue;
+
+ spin_unlock(&i915->gem.contexts.lock);
+
err = gen8_configure_context(ctx, regs, ARRAY_SIZE(regs));
- if (err)
+ if (err) {
+ i915_gem_context_put(ctx);
return err;
+ }
+
+ spin_lock(&i915->gem.contexts.lock);
+ list_safe_reset_next(ctx, cn, link);
+ i915_gem_context_put(ctx);
}
+ spin_unlock(&i915->gem.contexts.lock);
/*
* After updating all other contexts, we need to modify ourselves.
@@ -1895,7 +1906,6 @@ static int gen8_configure_all_contexts(struct i915_perf_stream *stream,
*/
for_each_uabi_engine(engine, i915) {
struct intel_context *ce = engine->kernel_context;
- int err;
if (engine->class != RENDER_CLASS)
continue;
diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c
index 6b88d934927a..585808c2ee84 100644
--- a/drivers/gpu/drm/i915/i915_sysfs.c
+++ b/drivers/gpu/drm/i915/i915_sysfs.c
@@ -144,9 +144,9 @@ static const struct attribute_group media_rc6_attr_group = {
};
#endif
-static int l3_access_valid(struct drm_i915_private *dev_priv, loff_t offset)
+static int l3_access_valid(struct drm_i915_private *i915, loff_t offset)
{
- if (!HAS_L3_DPF(dev_priv))
+ if (!HAS_L3_DPF(i915))
return -EPERM;
if (offset % 4 != 0)
@@ -164,31 +164,24 @@ i915_l3_read(struct file *filp, struct kobject *kobj,
loff_t offset, size_t count)
{
struct device *kdev = kobj_to_dev(kobj);
- struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev);
- struct drm_device *dev = &dev_priv->drm;
+ struct drm_i915_private *i915 = kdev_minor_to_i915(kdev);
int slice = (int)(uintptr_t)attr->private;
int ret;
- count = round_down(count, 4);
-
- ret = l3_access_valid(dev_priv, offset);
+ ret = l3_access_valid(i915, offset);
if (ret)
return ret;
+ count = round_down(count, 4);
count = min_t(size_t, GEN7_L3LOG_SIZE - offset, count);
+ memset(buf, 0, count);
- ret = i915_mutex_lock_interruptible(dev);
- if (ret)
- return ret;
-
- if (dev_priv->l3_parity.remap_info[slice])
+ spin_lock(&i915->gem.contexts.lock);
+ if (i915->l3_parity.remap_info[slice])
memcpy(buf,
- dev_priv->l3_parity.remap_info[slice] + (offset/4),
+ i915->l3_parity.remap_info[slice] + offset / 4,
count);
- else
- memset(buf, 0, count);
-
- mutex_unlock(&dev->struct_mutex);
+ spin_unlock(&i915->gem.contexts.lock);
return count;
}
@@ -199,46 +192,46 @@ i915_l3_write(struct file *filp, struct kobject *kobj,
loff_t offset, size_t count)
{
struct device *kdev = kobj_to_dev(kobj);
- struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev);
- struct drm_device *dev = &dev_priv->drm;
+ struct drm_i915_private *i915 = kdev_minor_to_i915(kdev);
struct i915_gem_context *ctx;
int slice = (int)(uintptr_t)attr->private;
- u32 **remap_info;
+ u32 *remap_info, *freeme = NULL;
int ret;
- ret = l3_access_valid(dev_priv, offset);
+ ret = l3_access_valid(i915, offset);
if (ret)
return ret;
- ret = i915_mutex_lock_interruptible(dev);
- if (ret)
- return ret;
+ remap_info = kzalloc(GEN7_L3LOG_SIZE, GFP_KERNEL);
+ if (!remap_info)
+ return -ENOMEM;
- remap_info = &dev_priv->l3_parity.remap_info[slice];
- if (!*remap_info) {
- *remap_info = kzalloc(GEN7_L3LOG_SIZE, GFP_KERNEL);
- if (!*remap_info) {
- ret = -ENOMEM;
- goto out;
- }
+ spin_lock(&i915->gem.contexts.lock);
+
+ if (i915->l3_parity.remap_info[slice]) {
+ freeme = remap_info;
+ remap_info = i915->l3_parity.remap_info[slice];
+ } else {
+ i915->l3_parity.remap_info[slice] = remap_info;
}
- /* TODO: Ideally we really want a GPU reset here to make sure errors
- * aren't propagated. Since I cannot find a stable way to reset the GPU
- * at this point it is left as a TODO.
- */
- memcpy(*remap_info + (offset/4), buf, count);
+ count = round_down(count, 4);
+ memcpy(remap_info + offset / 4, buf, count);
/* NB: We defer the remapping until we switch to the context */
- list_for_each_entry(ctx, &dev_priv->contexts.list, link)
- ctx->remap_slice |= (1<<slice);
+ list_for_each_entry(ctx, &i915->gem.contexts.list, link)
+ ctx->remap_slice |= BIT(slice);
- ret = count;
+ spin_unlock(&i915->gem.contexts.lock);
+ kfree(freeme);
-out:
- mutex_unlock(&dev->struct_mutex);
+ /*
+ * TODO: Ideally we really want a GPU reset here to make sure errors
+ * aren't propagated. Since I cannot find a stable way to reset the GPU
+ * at this point it is left as a TODO.
+ */
- return ret;
+ return count;
}
static const struct bin_attribute dpf_attrs = {
diff --git a/drivers/gpu/drm/i915/i915_trace.h b/drivers/gpu/drm/i915/i915_trace.h
index 1f2cf6cfafb5..7ef7a1e1664c 100644
--- a/drivers/gpu/drm/i915/i915_trace.h
+++ b/drivers/gpu/drm/i915/i915_trace.h
@@ -952,7 +952,7 @@ DECLARE_EVENT_CLASS(i915_context,
TP_fast_assign(
__entry->dev = ctx->i915->drm.primary->index;
__entry->ctx = ctx;
- __entry->vm = ctx->vm;
+ __entry->vm = rcu_access_pointer(ctx->vm);
),
TP_printk("dev=%u, ctx=%p, ctx_vm=%p",
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem.c b/drivers/gpu/drm/i915/selftests/i915_gem.c
index 0346c3e5b6b6..bfa40a5b6d98 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem.c
@@ -138,11 +138,9 @@ static int igt_gem_suspend(void *arg)
return PTR_ERR(file);
err = -ENOMEM;
- mutex_lock(&i915->drm.struct_mutex);
ctx = live_context(i915, file);
if (!IS_ERR(ctx))
err = switch_to_context(i915, ctx);
- mutex_unlock(&i915->drm.struct_mutex);
if (err)
goto out;
@@ -157,9 +155,7 @@ static int igt_gem_suspend(void *arg)
pm_resume(i915);
- mutex_lock(&i915->drm.struct_mutex);
err = switch_to_context(i915, ctx);
- mutex_unlock(&i915->drm.struct_mutex);
out:
mock_file_free(i915, file);
return err;
@@ -177,11 +173,9 @@ static int igt_gem_hibernate(void *arg)
return PTR_ERR(file);
err = -ENOMEM;
- mutex_lock(&i915->drm.struct_mutex);
ctx = live_context(i915, file);
if (!IS_ERR(ctx))
err = switch_to_context(i915, ctx);
- mutex_unlock(&i915->drm.struct_mutex);
if (err)
goto out;
@@ -196,9 +190,7 @@ static int igt_gem_hibernate(void *arg)
pm_resume(i915);
- mutex_lock(&i915->drm.struct_mutex);
err = switch_to_context(i915, ctx);
- mutex_unlock(&i915->drm.struct_mutex);
out:
mock_file_free(i915, file);
return err;
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
index f39f0282e78c..0af9a58d011d 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
@@ -473,7 +473,6 @@ static int igt_evict_contexts(void *arg)
}
count = 0;
- mutex_lock(&i915->drm.struct_mutex);
onstack_fence_init(&fence);
do {
struct i915_request *rq;
@@ -510,8 +509,6 @@ static int igt_evict_contexts(void *arg)
count++;
err = 0;
} while(1);
- mutex_unlock(&i915->drm.struct_mutex);
-
onstack_fence_fini(&fence);
pr_info("Submitted %lu contexts/requests on %s\n",
count, engine->name);
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
index 9939dd40c6a8..6777a335bd44 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
@@ -1246,6 +1246,7 @@ static int exercise_mock(struct drm_i915_private *i915,
unsigned long end_time))
{
const u64 limit = totalram_pages() << PAGE_SHIFT;
+ struct i915_address_space *vm;
struct i915_gem_context *ctx;
IGT_TIMEOUT(end_time);
int err;
@@ -1254,7 +1255,8 @@ static int exercise_mock(struct drm_i915_private *i915,
if (!ctx)
return -ENOMEM;
- err = func(i915, ctx->vm, 0, min(ctx->vm->total, limit), end_time);
+ vm = rcu_dereference_protected(ctx->vm, true);
+ err = func(i915, vm, 0, min(vm->total, limit), end_time);
mock_context_close(ctx);
return err;
diff --git a/drivers/gpu/drm/i915/selftests/i915_request.c b/drivers/gpu/drm/i915/selftests/i915_request.c
index d1ee384a3260..4abf52733520 100644
--- a/drivers/gpu/drm/i915/selftests/i915_request.c
+++ b/drivers/gpu/drm/i915/selftests/i915_request.c
@@ -181,9 +181,7 @@ static int igt_request_rewind(void *arg)
struct intel_context *ce;
int err = -EINVAL;
- mutex_lock(&i915->drm.struct_mutex);
ctx[0] = mock_context(i915, "A");
- mutex_unlock(&i915->drm.struct_mutex);
ce = i915_gem_context_get_engine(ctx[0], RCS0);
GEM_BUG_ON(IS_ERR(ce));
@@ -197,9 +195,7 @@ static int igt_request_rewind(void *arg)
i915_request_get(request);
i915_request_add(request);
- mutex_lock(&i915->drm.struct_mutex);
ctx[1] = mock_context(i915, "B");
- mutex_unlock(&i915->drm.struct_mutex);
ce = i915_gem_context_get_engine(ctx[1], RCS0);
GEM_BUG_ON(IS_ERR(ce));
@@ -438,9 +434,7 @@ static int mock_breadcrumbs_smoketest(void *arg)
}
for (n = 0; n < t.ncontexts; n++) {
- mutex_lock(&t.engine->i915->drm.struct_mutex);
t.contexts[n] = mock_context(t.engine->i915, "mock");
- mutex_unlock(&t.engine->i915->drm.struct_mutex);
if (!t.contexts[n]) {
ret = -ENOMEM;
goto out_contexts;
@@ -734,7 +728,8 @@ static int live_empty_request(void *arg)
static struct i915_vma *recursive_batch(struct drm_i915_private *i915)
{
struct i915_gem_context *ctx = i915->kernel_context;
- struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm;
+ struct i915_address_space *vm =
+ rcu_dereference_protected(ctx->vm, true) ?: &i915->ggtt.vm;
struct drm_i915_gem_object *obj;
const int gen = INTEL_GEN(i915);
struct i915_vma *vma;
@@ -1109,9 +1104,7 @@ static int live_breadcrumbs_smoketest(void *arg)
}
for (n = 0; n < t[0].ncontexts; n++) {
- mutex_lock(&i915->drm.struct_mutex);
t[0].contexts[n] = live_context(i915, file);
- mutex_unlock(&i915->drm.struct_mutex);
if (!t[0].contexts[n]) {
ret = -ENOMEM;
goto out_contexts;
diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c
index ac1ff558eb90..f6e6b0aae38c 100644
--- a/drivers/gpu/drm/i915/selftests/i915_vma.c
+++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
@@ -38,7 +38,7 @@ static bool assert_vma(struct i915_vma *vma,
{
bool ok = true;
- if (vma->vm != ctx->vm) {
+ if (vma->vm != rcu_access_pointer(ctx->vm)) {
pr_err("VMA created with wrong VM\n");
ok = false;
}
@@ -113,7 +113,8 @@ static int create_vmas(struct drm_i915_private *i915,
list_for_each_entry(obj, objects, st_link) {
for (pinned = 0; pinned <= 1; pinned++) {
list_for_each_entry(ctx, contexts, link) {
- struct i915_address_space *vm = ctx->vm;
+ struct i915_address_space *vm =
+ rcu_dereference_protected(ctx->vm, true);
struct i915_vma *vma;
int err;
diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
index efe40c29fcf2..0ed21a7b8682 100644
--- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
+++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
@@ -59,11 +59,9 @@ static void mock_device_release(struct drm_device *dev)
i915_gem_drain_workqueue(i915);
- mutex_lock(&i915->drm.struct_mutex);
for_each_engine(engine, i915, id)
mock_engine_free(engine);
- i915_gem_contexts_fini(i915);
- mutex_unlock(&i915->drm.struct_mutex);
+ i915_gem_driver_release__contexts(i915);
intel_timelines_fini(i915);
@@ -207,7 +205,7 @@ struct drm_i915_private *mock_gem_device(void)
return i915;
err_context:
- i915_gem_contexts_fini(i915);
+ i915_gem_driver_release__contexts(i915);
err_engine:
mock_engine_free(i915->engine[RCS0]);
err_unlock:
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* Re: [PATCH 21/27] drm/i915: Move context management under GEM
2019-09-25 10:01 ` [PATCH 21/27] drm/i915: Move context management under GEM Chris Wilson
@ 2019-09-26 13:57 ` Tvrtko Ursulin
2019-10-02 16:09 ` Tvrtko Ursulin
2019-10-03 7:35 ` Chris Wilson
0 siblings, 2 replies; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-26 13:57 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 25/09/2019 11:01, Chris Wilson wrote:
> Keep track of the GEM contexts underneath i915->gem.contexts and assign
> them their own lock for the purposes of list management.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> ---
> drivers/gpu/drm/i915/gem/i915_gem_context.c | 155 +++++++-----------
> drivers/gpu/drm/i915/gem/i915_gem_context.h | 4 +-
> .../gpu/drm/i915/gem/i915_gem_context_types.h | 2 +-
> .../gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 +-
> drivers/gpu/drm/i915/gem/i915_gem_userptr.c | 2 +-
> .../gpu/drm/i915/gem/selftests/huge_pages.c | 12 +-
> .../drm/i915/gem/selftests/i915_gem_context.c | 122 ++++++--------
> .../gpu/drm/i915/gem/selftests/mock_context.c | 4 +-
> drivers/gpu/drm/i915/gt/intel_context.c | 8 +-
> drivers/gpu/drm/i915/gt/selftest_context.c | 24 +--
> drivers/gpu/drm/i915/gt/selftest_hangcheck.c | 19 +--
> drivers/gpu/drm/i915/gt/selftest_lrc.c | 4 +-
> .../gpu/drm/i915/gt/selftest_workarounds.c | 10 +-
> drivers/gpu/drm/i915/gvt/scheduler.c | 20 +--
> drivers/gpu/drm/i915/i915_debugfs.c | 50 +++---
> drivers/gpu/drm/i915/i915_drv.c | 2 -
> drivers/gpu/drm/i915/i915_drv.h | 27 +--
> drivers/gpu/drm/i915/i915_gem.c | 10 +-
> drivers/gpu/drm/i915/i915_gem_gtt.c | 4 +-
> drivers/gpu/drm/i915/i915_perf.c | 24 ++-
> drivers/gpu/drm/i915/i915_sysfs.c | 77 ++++-----
> drivers/gpu/drm/i915/i915_trace.h | 2 +-
> drivers/gpu/drm/i915/selftests/i915_gem.c | 8 -
> .../gpu/drm/i915/selftests/i915_gem_evict.c | 3 -
> drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 4 +-
> drivers/gpu/drm/i915/selftests/i915_request.c | 11 +-
> drivers/gpu/drm/i915/selftests/i915_vma.c | 5 +-
> .../gpu/drm/i915/selftests/mock_gem_device.c | 6 +-
> 28 files changed, 272 insertions(+), 349 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
> index 4e59b809d901..a77f439358d7 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
> @@ -219,9 +219,12 @@ static struct i915_gem_engines *default_engines(struct i915_gem_context *ctx)
>
> static void i915_gem_context_free(struct i915_gem_context *ctx)
> {
> - lockdep_assert_held(&ctx->i915->drm.struct_mutex);
> GEM_BUG_ON(!i915_gem_context_is_closed(ctx));
>
> + spin_lock(&ctx->i915->gem.contexts.lock);
> + list_del(&ctx->link);
> + spin_unlock(&ctx->i915->gem.contexts.lock);
> +
> free_engines(rcu_access_pointer(ctx->engines));
> mutex_destroy(&ctx->engines_mutex);
>
> @@ -231,56 +234,40 @@ static void i915_gem_context_free(struct i915_gem_context *ctx)
> kfree(ctx->name);
> put_pid(ctx->pid);
>
> - list_del(&ctx->link);
> mutex_destroy(&ctx->mutex);
>
> kfree_rcu(ctx, rcu);
> }
>
> -static void contexts_free(struct drm_i915_private *i915)
> +static void contexts_free_all(struct llist_node *list)
> {
> - struct llist_node *freed = llist_del_all(&i915->contexts.free_list);
> struct i915_gem_context *ctx, *cn;
>
> - lockdep_assert_held(&i915->drm.struct_mutex);
> -
> - llist_for_each_entry_safe(ctx, cn, freed, free_link)
> + llist_for_each_entry_safe(ctx, cn, list, free_link)
> i915_gem_context_free(ctx);
> }
>
> -static void contexts_free_first(struct drm_i915_private *i915)
> +static void contexts_flush_free(struct i915_gem_contexts *gc)
> {
> - struct i915_gem_context *ctx;
> - struct llist_node *freed;
> -
> - lockdep_assert_held(&i915->drm.struct_mutex);
> -
> - freed = llist_del_first(&i915->contexts.free_list);
> - if (!freed)
> - return;
> -
> - ctx = container_of(freed, typeof(*ctx), free_link);
> - i915_gem_context_free(ctx);
> + contexts_free_all(llist_del_all(&gc->free_list));
> }
>
> static void contexts_free_worker(struct work_struct *work)
> {
> - struct drm_i915_private *i915 =
> - container_of(work, typeof(*i915), contexts.free_work);
> + struct i915_gem_contexts *gc =
> + container_of(work, typeof(*gc), free_work);
>
> - mutex_lock(&i915->drm.struct_mutex);
> - contexts_free(i915);
> - mutex_unlock(&i915->drm.struct_mutex);
> + contexts_flush_free(gc);
> }
>
> void i915_gem_context_release(struct kref *ref)
> {
> struct i915_gem_context *ctx = container_of(ref, typeof(*ctx), ref);
> - struct drm_i915_private *i915 = ctx->i915;
> + struct i915_gem_contexts *gc = &ctx->i915->gem.contexts;
>
> trace_i915_context_free(ctx);
> - if (llist_add(&ctx->free_link, &i915->contexts.free_list))
> - queue_work(i915->wq, &i915->contexts.free_work);
> + if (llist_add(&ctx->free_link, &gc->free_list))
> + queue_work(ctx->i915->wq, &gc->free_work);
At first I thought gc was some sort of garbage collection list. But it
is a GEM contexts list. :) This hunk looks completely avoidable, I don't
see it brings much benefit on balance since in turn it adds ctx->i915
twice . But as you wish..
> }
>
> static inline struct i915_gem_engines *
> @@ -359,8 +346,8 @@ static void context_close(struct i915_gem_context *ctx)
> {
> i915_gem_context_set_closed(ctx);
>
> - if (ctx->vm)
> - i915_vm_close(ctx->vm);
> + if (rcu_access_pointer(ctx->vm))
> + i915_vm_close(rcu_dereference_protected(ctx->vm, true));
>
> mutex_lock(&ctx->mutex);
>
> @@ -394,7 +381,6 @@ __create_context(struct drm_i915_private *i915)
> return ERR_PTR(-ENOMEM);
>
> kref_init(&ctx->ref);
> - list_add_tail(&ctx->link, &i915->contexts.list);
> ctx->i915 = i915;
> ctx->sched.priority = I915_USER_PRIORITY(I915_PRIORITY_NORMAL);
> mutex_init(&ctx->mutex);
> @@ -435,6 +421,10 @@ __create_context(struct drm_i915_private *i915)
> for (i = 0; i < ARRAY_SIZE(ctx->hang_timestamp); i++)
> ctx->hang_timestamp[i] = jiffies - CONTEXT_FAST_HANG_JIFFIES;
>
> + spin_lock(&i915->gem.contexts.lock);
> + list_add_tail(&ctx->link, &i915->gem.contexts.list);
> + spin_unlock(&i915->gem.contexts.lock);
> +
> return ctx;
>
> err_free:
> @@ -464,11 +454,11 @@ static void __apply_ppgtt(struct intel_context *ce, void *vm)
> static struct i915_address_space *
> __set_ppgtt(struct i915_gem_context *ctx, struct i915_address_space *vm)
> {
> - struct i915_address_space *old = ctx->vm;
> + struct i915_address_space *old = rcu_dereference_protected(ctx->vm, 1);
>
> GEM_BUG_ON(old && i915_vm_is_4lvl(vm) != i915_vm_is_4lvl(old));
>
> - ctx->vm = i915_vm_open(vm);
> + rcu_assign_pointer(ctx->vm, i915_vm_open(vm));
Are updaters here serialized? Lost track if struct mutex is still held
in set_ppgtt at this point in the series or not.. Thinking of "old =
rcu_dereference_protected.." and here.
> context_apply_all(ctx, __apply_ppgtt, vm);
>
> return old;
> @@ -477,7 +467,7 @@ __set_ppgtt(struct i915_gem_context *ctx, struct i915_address_space *vm)
> static void __assign_ppgtt(struct i915_gem_context *ctx,
> struct i915_address_space *vm)
> {
> - if (vm == ctx->vm)
> + if (vm == rcu_access_pointer(ctx->vm))
> return;
>
> vm = __set_ppgtt(ctx, vm);
> @@ -509,27 +499,25 @@ static void __assign_timeline(struct i915_gem_context *ctx,
> }
>
> static struct i915_gem_context *
> -i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags)
> +i915_gem_create_context(struct drm_i915_private *i915, unsigned int flags)
> {
> struct i915_gem_context *ctx;
>
> - lockdep_assert_held(&dev_priv->drm.struct_mutex);
> -
> if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE &&
> - !HAS_EXECLISTS(dev_priv))
> + !HAS_EXECLISTS(i915))
> return ERR_PTR(-EINVAL);
>
> - /* Reap the most stale context */
> - contexts_free_first(dev_priv);
> + /* Reap the stale contexts */
> + contexts_flush_free(&i915->gem.contexts);
>
> - ctx = __create_context(dev_priv);
> + ctx = __create_context(i915);
> if (IS_ERR(ctx))
> return ctx;
>
> - if (HAS_FULL_PPGTT(dev_priv)) {
> + if (HAS_FULL_PPGTT(i915)) {
> struct i915_ppgtt *ppgtt;
>
> - ppgtt = i915_ppgtt_create(dev_priv);
> + ppgtt = i915_ppgtt_create(i915);
> if (IS_ERR(ppgtt)) {
> DRM_DEBUG_DRIVER("PPGTT setup failed (%ld)\n",
> PTR_ERR(ppgtt));
> @@ -544,7 +532,7 @@ i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags)
> if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE) {
> struct intel_timeline *timeline;
>
> - timeline = intel_timeline_create(&dev_priv->gt, NULL);
> + timeline = intel_timeline_create(&i915->gt, NULL);
> if (IS_ERR(timeline)) {
> context_close(ctx);
> return ERR_CAST(timeline);
> @@ -590,48 +578,40 @@ i915_gem_context_create_kernel(struct drm_i915_private *i915, int prio)
> return ctx;
> }
>
> -static void init_contexts(struct drm_i915_private *i915)
> +static void init_contexts(struct i915_gem_contexts *gc)
> {
> - mutex_init(&i915->contexts.mutex);
> - INIT_LIST_HEAD(&i915->contexts.list);
> -
> - /* Using the simple ida interface, the max is limited by sizeof(int) */
> - BUILD_BUG_ON(MAX_CONTEXT_HW_ID > INT_MAX);
> - BUILD_BUG_ON(GEN11_MAX_CONTEXT_HW_ID > INT_MAX);
> - ida_init(&i915->contexts.hw_ida);
> - INIT_LIST_HEAD(&i915->contexts.hw_id_list);
This (and some more bits) belong to the hw_id removal patch.
> + spin_lock_init(&gc->lock);
> + INIT_LIST_HEAD(&gc->list);
>
> - INIT_WORK(&i915->contexts.free_work, contexts_free_worker);
> - init_llist_head(&i915->contexts.free_list);
> + INIT_WORK(&gc->free_work, contexts_free_worker);
> + init_llist_head(&gc->free_list);
> }
>
> -int i915_gem_contexts_init(struct drm_i915_private *dev_priv)
> +int i915_gem_init_contexts(struct drm_i915_private *i915)
> {
> struct i915_gem_context *ctx;
>
> /* Reassure ourselves we are only called once */
> - GEM_BUG_ON(dev_priv->kernel_context);
> + GEM_BUG_ON(i915->kernel_context);
>
> - init_contexts(dev_priv);
> + init_contexts(&i915->gem.contexts);
>
> /* lowest priority; idle task */
> - ctx = i915_gem_context_create_kernel(dev_priv, I915_PRIORITY_MIN);
> + ctx = i915_gem_context_create_kernel(i915, I915_PRIORITY_MIN);
> if (IS_ERR(ctx)) {
> DRM_ERROR("Failed to create default global context\n");
> return PTR_ERR(ctx);
> }
> - dev_priv->kernel_context = ctx;
> + i915->kernel_context = ctx;
>
> DRM_DEBUG_DRIVER("%s context support initialized\n",
> - DRIVER_CAPS(dev_priv)->has_logical_contexts ?
> + DRIVER_CAPS(i915)->has_logical_contexts ?
> "logical" : "fake");
> return 0;
> }
>
> -void i915_gem_contexts_fini(struct drm_i915_private *i915)
> +void i915_gem_driver_release__contexts(struct drm_i915_private *i915)
> {
> - lockdep_assert_held(&i915->drm.struct_mutex);
> -
> destroy_kernel_context(&i915->kernel_context);
> }
>
> @@ -653,8 +633,8 @@ static int gem_context_register(struct i915_gem_context *ctx,
> int ret;
>
> ctx->file_priv = fpriv;
> - if (ctx->vm)
> - ctx->vm->file = fpriv;
> + if (rcu_access_pointer(ctx->vm))
> + rcu_dereference_protected(ctx->vm, true)->file = fpriv;
Here rcu accesses are just to satisfy sparse?
>
> ctx->pid = get_task_pid(current, PIDTYPE_PID);
> ctx->name = kasprintf(GFP_KERNEL, "%s[%d]",
> @@ -691,9 +671,7 @@ int i915_gem_context_open(struct drm_i915_private *i915,
> idr_init(&file_priv->context_idr);
> idr_init_base(&file_priv->vm_idr, 1);
>
> - mutex_lock(&i915->drm.struct_mutex);
> ctx = i915_gem_create_context(i915, 0);
> - mutex_unlock(&i915->drm.struct_mutex);
> if (IS_ERR(ctx)) {
> err = PTR_ERR(ctx);
> goto err;
> @@ -721,6 +699,7 @@ int i915_gem_context_open(struct drm_i915_private *i915,
> void i915_gem_context_close(struct drm_file *file)
> {
> struct drm_i915_file_private *file_priv = file->driver_priv;
> + struct drm_i915_private *i915 = file_priv->dev_priv;
>
> idr_for_each(&file_priv->context_idr, context_idr_cleanup, NULL);
> idr_destroy(&file_priv->context_idr);
> @@ -729,6 +708,8 @@ void i915_gem_context_close(struct drm_file *file)
> idr_for_each(&file_priv->vm_idr, vm_idr_cleanup, NULL);
> idr_destroy(&file_priv->vm_idr);
> mutex_destroy(&file_priv->vm_idr_lock);
> +
> + contexts_flush_free(&i915->gem.contexts);
> }
>
> int i915_gem_vm_create_ioctl(struct drm_device *dev, void *data,
> @@ -907,16 +888,12 @@ static int get_ppgtt(struct drm_i915_file_private *file_priv,
> struct i915_address_space *vm;
> int ret;
>
> - if (!ctx->vm)
> + if (!rcu_access_pointer(ctx->vm))
> return -ENODEV;
>
> - /* XXX rcu acquire? */
> - ret = mutex_lock_interruptible(&ctx->i915->drm.struct_mutex);
> - if (ret)
> - return ret;
> -
> + rcu_read_lock();
> vm = i915_vm_get(ctx->vm);
> - mutex_unlock(&ctx->i915->drm.struct_mutex);
> + rcu_read_unlock();
>
> ret = mutex_lock_interruptible(&file_priv->vm_idr_lock);
> if (ret)
> @@ -1025,7 +1002,7 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv,
> if (args->size)
> return -EINVAL;
>
> - if (!ctx->vm)
> + if (!rcu_access_pointer(ctx->vm))
> return -ENODEV;
>
> if (upper_32_bits(args->value))
> @@ -1039,17 +1016,15 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv,
> if (!vm)
> return -ENOENT;
>
> - err = mutex_lock_interruptible(&ctx->i915->drm.struct_mutex);
> + err = mutex_lock_interruptible(&ctx->mutex);
> if (err)
> goto out;
>
> - if (vm == ctx->vm)
> + if (vm == rcu_access_pointer(ctx->vm))
> 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, vm);
>
> @@ -1069,8 +1044,7 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv,
> }
>
> unlock:
> - mutex_unlock(&ctx->i915->drm.struct_mutex);
> -
> + mutex_unlock(&ctx->mutex);
> out:
> i915_vm_put(vm);
> return err;
> @@ -1953,7 +1927,7 @@ static int clone_vm(struct i915_gem_context *dst,
>
> rcu_read_lock();
> do {
> - vm = READ_ONCE(src->vm);
> + vm = rcu_dereference(src->vm);
> if (!vm)
> break;
>
> @@ -1975,7 +1949,7 @@ static int clone_vm(struct i915_gem_context *dst,
> * it cannot be reallocated elsewhere.
> */
>
> - if (vm == READ_ONCE(src->vm))
> + if (vm == rcu_access_pointer(src->vm))
> break;
>
> i915_vm_put(vm);
> @@ -2077,12 +2051,7 @@ int i915_gem_context_create_ioctl(struct drm_device *dev, void *data,
> return -EIO;
> }
>
> - ret = i915_mutex_lock_interruptible(dev);
> - if (ret)
> - return ret;
> -
> ext_data.ctx = i915_gem_create_context(i915, args->flags);
> - mutex_unlock(&dev->struct_mutex);
> if (IS_ERR(ext_data.ctx))
> return PTR_ERR(ext_data.ctx);
>
> @@ -2209,10 +2178,12 @@ int i915_gem_context_getparam_ioctl(struct drm_device *dev, void *data,
>
> case I915_CONTEXT_PARAM_GTT_SIZE:
> args->size = 0;
> - if (ctx->vm)
> - args->value = ctx->vm->total;
> + rcu_read_lock();
> + if (rcu_access_pointer(ctx->vm))
> + args->value = rcu_dereference(ctx->vm)->total;
> else
> args->value = to_i915(dev)->ggtt.vm.total;
> + rcu_read_unlock();
> break;
>
> case I915_CONTEXT_PARAM_NO_ERROR_CAPTURE:
> @@ -2283,7 +2254,7 @@ int i915_gem_context_setparam_ioctl(struct drm_device *dev, void *data,
> int i915_gem_context_reset_stats_ioctl(struct drm_device *dev,
> void *data, struct drm_file *file)
> {
> - struct drm_i915_private *dev_priv = to_i915(dev);
> + struct drm_i915_private *i915 = to_i915(dev);
> struct drm_i915_reset_stats *args = data;
> struct i915_gem_context *ctx;
> int ret;
> @@ -2305,7 +2276,7 @@ int i915_gem_context_reset_stats_ioctl(struct drm_device *dev,
> */
>
> if (capable(CAP_SYS_ADMIN))
> - args->reset_count = i915_reset_count(&dev_priv->gpu_error);
> + args->reset_count = i915_reset_count(&i915->gpu_error);
> else
> args->reset_count = 0;
>
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.h b/drivers/gpu/drm/i915/gem/i915_gem_context.h
> index 1b0df53436cf..4ee5dfc5794e 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_context.h
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.h
> @@ -133,8 +133,8 @@ 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_fini(struct drm_i915_private *dev_priv);
> +int __must_check i915_gem_init_contexts(struct drm_i915_private *i915);
> +void i915_gem_driver_release__contexts(struct drm_i915_private *i915);
>
> int i915_gem_context_open(struct drm_i915_private *i915,
> struct drm_file *file);
> 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 6419da7c9f90..a3ecd19f2303 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
> @@ -88,7 +88,7 @@ struct i915_gem_context {
> * In other modes, this is a NULL pointer with the expectation that
> * the caller uses the shared global GTT.
> */
> - struct i915_address_space *vm;
> + struct i915_address_space __rcu *vm;
>
> /**
> * @pid: process id of creator
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
> index 88a881be12ec..98816c35ffc3 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
> @@ -728,7 +728,7 @@ static int eb_select_context(struct i915_execbuffer *eb)
> return -ENOENT;
>
> eb->gem_context = ctx;
> - if (ctx->vm)
> + if (rcu_access_pointer(ctx->vm))
> eb->invalid_flags |= EXEC_OBJECT_NEEDS_GTT;
>
> eb->context_flags = 0;
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
> index 27573ef903b4..7c225e1d2a11 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
> @@ -757,7 +757,7 @@ i915_gem_userptr_ioctl(struct drm_device *dev,
> * On almost all of the older hw, we cannot tell the GPU that
> * a page is readonly.
> */
> - vm = dev_priv->kernel_context->vm;
> + vm = rcu_dereference_protected(dev_priv->kernel_context->vm, true);
Hm, should it not be rcu_dereference here?
> if (!vm || !vm->has_read_only)
> return -ENODEV;
> }
> diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
> index 98b2a6ccfcc1..e204e653b459 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
> @@ -1322,7 +1322,7 @@ static int igt_ppgtt_pin_update(void *arg)
> struct i915_gem_context *ctx = arg;
> struct drm_i915_private *dev_priv = ctx->i915;
> unsigned long supported = INTEL_INFO(dev_priv)->page_sizes;
> - struct i915_address_space *vm = ctx->vm;
> + struct i915_address_space *vm = rcu_dereference_protected(ctx->vm, 1);
> struct drm_i915_gem_object *obj;
> struct i915_gem_engines_iter it;
> struct intel_context *ce;
> @@ -1460,7 +1460,8 @@ static int igt_tmpfs_fallback(void *arg)
> struct i915_gem_context *ctx = arg;
> struct drm_i915_private *i915 = ctx->i915;
> struct vfsmount *gemfs = i915->mm.gemfs;
> - struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm;
> + struct i915_address_space *vm =
> + rcu_dereference_protected(ctx->vm, true) ?: &i915->ggtt.vm;
> struct drm_i915_gem_object *obj;
> struct i915_vma *vma;
> u32 *vaddr;
> @@ -1517,7 +1518,8 @@ static int igt_shrink_thp(void *arg)
> {
> struct i915_gem_context *ctx = arg;
> struct drm_i915_private *i915 = ctx->i915;
> - struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm;
> + struct i915_address_space *vm =
> + rcu_dereference_protected(ctx->vm, true) ?: &i915->ggtt.vm;
> struct drm_i915_gem_object *obj;
> struct i915_gem_engines_iter it;
> struct intel_context *ce;
> @@ -1699,8 +1701,8 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *i915)
> goto out_unlock;
> }
>
> - if (ctx->vm)
> - ctx->vm->scrub_64K = true;
> + if (rcu_access_pointer(ctx->vm))
> + rcu_dereference_protected(ctx->vm, true)->scrub_64K = true;
>
> err = i915_subtests(tests, ctx);
>
> 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 a8a9293f4ac6..7de1c0078d6f 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
> @@ -53,19 +53,17 @@ static int live_nop_switch(void *arg)
> if (IS_ERR(file))
> return PTR_ERR(file);
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> ctx = kcalloc(nctx, sizeof(*ctx), GFP_KERNEL);
> if (!ctx) {
> err = -ENOMEM;
> - goto out_unlock;
> + goto out_file;
> }
>
> for (n = 0; n < nctx; n++) {
> ctx[n] = live_context(i915, file);
> if (IS_ERR(ctx[n])) {
> err = PTR_ERR(ctx[n]);
> - goto out_unlock;
> + goto out_file;
> }
> }
>
> @@ -79,7 +77,7 @@ static int live_nop_switch(void *arg)
> rq = igt_request_alloc(ctx[n], engine);
> if (IS_ERR(rq)) {
> err = PTR_ERR(rq);
> - goto out_unlock;
> + goto out_file;
> }
> i915_request_add(rq);
> }
> @@ -87,7 +85,7 @@ static int live_nop_switch(void *arg)
> pr_err("Failed to populated %d contexts\n", nctx);
> intel_gt_set_wedged(&i915->gt);
> err = -EIO;
> - goto out_unlock;
> + goto out_file;
> }
>
> times[1] = ktime_get_raw();
> @@ -97,7 +95,7 @@ static int live_nop_switch(void *arg)
>
> err = igt_live_test_begin(&t, i915, __func__, engine->name);
> if (err)
> - goto out_unlock;
> + goto out_file;
>
> end_time = jiffies + i915_selftest.timeout_jiffies;
> for_each_prime_number_from(prime, 2, 8192) {
> @@ -107,7 +105,7 @@ static int live_nop_switch(void *arg)
> rq = igt_request_alloc(ctx[n % nctx], engine);
> if (IS_ERR(rq)) {
> err = PTR_ERR(rq);
> - goto out_unlock;
> + goto out_file;
> }
>
> /*
> @@ -143,7 +141,7 @@ static int live_nop_switch(void *arg)
>
> err = igt_live_test_end(&t);
> if (err)
> - goto out_unlock;
> + goto out_file;
>
> pr_info("Switch latencies on %s: 1 = %lluns, %lu = %lluns\n",
> engine->name,
> @@ -151,8 +149,7 @@ static int live_nop_switch(void *arg)
> prime - 1, div64_u64(ktime_to_ns(times[1]), prime - 1));
> }
>
> -out_unlock:
> - mutex_unlock(&i915->drm.struct_mutex);
> +out_file:
> mock_file_free(i915, file);
> return err;
> }
> @@ -412,11 +409,9 @@ static int igt_ctx_exec(void *arg)
> if (IS_ERR(file))
> return PTR_ERR(file);
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> err = igt_live_test_begin(&t, i915, __func__, engine->name);
> if (err)
> - goto out_unlock;
> + goto out_file;
>
> ncontexts = 0;
> ndwords = 0;
> @@ -428,7 +423,7 @@ static int igt_ctx_exec(void *arg)
> ctx = kernel_context(i915);
> if (IS_ERR(ctx)) {
> err = PTR_ERR(ctx);
> - goto out_unlock;
> + goto out_file;
> }
>
> ce = i915_gem_context_get_engine(ctx, engine->legacy_idx);
> @@ -440,7 +435,7 @@ static int igt_ctx_exec(void *arg)
> err = PTR_ERR(obj);
> intel_context_put(ce);
> kernel_context_close(ctx);
> - goto out_unlock;
> + goto out_file;
> }
> }
>
> @@ -449,17 +444,18 @@ static int igt_ctx_exec(void *arg)
> pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
> ndwords, dw, max_dwords(obj),
> engine->name,
> - yesno(!!ctx->vm), err);
> + yesno(!!rcu_access_pointer(ctx->vm)),
> + err);
> intel_context_put(ce);
> kernel_context_close(ctx);
> - goto out_unlock;
> + goto out_file;
> }
>
> err = throttle(ce, tq, ARRAY_SIZE(tq));
> if (err) {
> intel_context_put(ce);
> kernel_context_close(ctx);
> - goto out_unlock;
> + goto out_file;
> }
>
> if (++dw == max_dwords(obj)) {
> @@ -489,11 +485,10 @@ static int igt_ctx_exec(void *arg)
> dw += rem;
> }
>
> -out_unlock:
> +out_file:
> throttle_release(tq, ARRAY_SIZE(tq));
> if (igt_live_test_end(&t))
> err = -EIO;
> - mutex_unlock(&i915->drm.struct_mutex);
>
> mock_file_free(i915, file);
> if (err)
> @@ -528,22 +523,20 @@ static int igt_shared_ctx_exec(void *arg)
> if (IS_ERR(file))
> return PTR_ERR(file);
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> parent = live_context(i915, file);
> if (IS_ERR(parent)) {
> err = PTR_ERR(parent);
> - goto out_unlock;
> + goto out_file;
> }
>
> if (!parent->vm) { /* not full-ppgtt; nothing to share */
> err = 0;
> - goto out_unlock;
> + goto out_file;
> }
>
> err = igt_live_test_begin(&t, i915, __func__, "");
> if (err)
> - goto out_unlock;
> + goto out_file;
>
> for_each_engine(engine, i915, id) {
> unsigned long ncontexts, ndwords, dw;
> @@ -587,7 +580,8 @@ static int igt_shared_ctx_exec(void *arg)
> pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
> ndwords, dw, max_dwords(obj),
> engine->name,
> - yesno(!!ctx->vm), err);
> + yesno(!!rcu_access_pointer(ctx->vm)),
> + err);
> intel_context_put(ce);
> kernel_context_close(ctx);
> goto out_test;
> @@ -626,17 +620,13 @@ static int igt_shared_ctx_exec(void *arg)
> dw += rem;
> }
>
> - mutex_unlock(&i915->drm.struct_mutex);
> i915_gem_drain_freed_objects(i915);
> - mutex_lock(&i915->drm.struct_mutex);
> }
> out_test:
> throttle_release(tq, ARRAY_SIZE(tq));
> if (igt_live_test_end(&t))
> err = -EIO;
> -out_unlock:
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> +out_file:
> mock_file_free(i915, file);
> return err;
> }
> @@ -1008,8 +998,6 @@ __igt_ctx_sseu(struct drm_i915_private *i915,
> if (flags & TEST_RESET)
> igt_global_reset_lock(&i915->gt);
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> ctx = live_context(i915, file);
> if (IS_ERR(ctx)) {
> ret = PTR_ERR(ctx);
> @@ -1064,8 +1052,6 @@ __igt_ctx_sseu(struct drm_i915_private *i915,
> i915_gem_object_put(obj);
>
> out_unlock:
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> if (flags & TEST_RESET)
> igt_global_reset_unlock(&i915->gt);
>
> @@ -1125,23 +1111,24 @@ static int igt_ctx_readonly(void *arg)
> if (IS_ERR(file))
> return PTR_ERR(file);
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> err = igt_live_test_begin(&t, i915, __func__, "");
> if (err)
> - goto out_unlock;
> + goto out_file;
>
> ctx = live_context(i915, file);
> if (IS_ERR(ctx)) {
> err = PTR_ERR(ctx);
> - goto out_unlock;
> + goto out_file;
> }
>
> - vm = ctx->vm ?: &i915->ggtt.alias->vm;
> + rcu_read_lock();
> + vm = rcu_dereference(ctx->vm) ?: &i915->ggtt.alias->vm;
> if (!vm || !vm->has_read_only) {
> + rcu_read_unlock();
> err = 0;
> - goto out_unlock;
> + goto out_file;
> }
> + rcu_read_unlock();
>
> ndwords = 0;
> dw = 0;
> @@ -1159,7 +1146,7 @@ static int igt_ctx_readonly(void *arg)
> if (IS_ERR(obj)) {
> err = PTR_ERR(obj);
> i915_gem_context_unlock_engines(ctx);
> - goto out_unlock;
> + goto out_file;
> }
>
> if (prandom_u32_state(&prng) & 1)
> @@ -1170,15 +1157,17 @@ static int igt_ctx_readonly(void *arg)
> if (err) {
> pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
> ndwords, dw, max_dwords(obj),
> - ce->engine->name, yesno(!!ctx->vm), err);
> + ce->engine->name,
> + yesno(!!rcu_access_pointer(ctx->vm)),
> + err);
> i915_gem_context_unlock_engines(ctx);
> - goto out_unlock;
> + goto out_file;
> }
>
> err = throttle(ce, tq, ARRAY_SIZE(tq));
> if (err) {
> i915_gem_context_unlock_engines(ctx);
> - goto out_unlock;
> + goto out_file;
> }
>
> if (++dw == max_dwords(obj)) {
> @@ -1210,11 +1199,10 @@ static int igt_ctx_readonly(void *arg)
> dw += rem;
> }
>
> -out_unlock:
> +out_file:
> throttle_release(tq, ARRAY_SIZE(tq));
> if (igt_live_test_end(&t))
> err = -EIO;
> - mutex_unlock(&i915->drm.struct_mutex);
>
> mock_file_free(i915, file);
> return err;
> @@ -1223,7 +1211,7 @@ static int igt_ctx_readonly(void *arg)
> static int check_scratch(struct i915_gem_context *ctx, u64 offset)
> {
> struct drm_mm_node *node =
> - __drm_mm_interval_first(&ctx->vm->mm,
> + __drm_mm_interval_first(&rcu_dereference_protected(ctx->vm, true)->mm,
> offset, offset + sizeof(u32) - 1);
> if (!node || node->start > offset)
> return 0;
> @@ -1273,7 +1261,9 @@ static int write_to_scratch(struct i915_gem_context *ctx,
>
> intel_gt_chipset_flush(engine->gt);
>
> - vma = i915_vma_instance(obj, ctx->vm, NULL);
> + vma = i915_vma_instance(obj,
> + rcu_dereference_protected(ctx->vm, true),
> + NULL);
> if (IS_ERR(vma)) {
> err = PTR_ERR(vma);
> goto err;
> @@ -1372,7 +1362,9 @@ static int read_from_scratch(struct i915_gem_context *ctx,
>
> intel_gt_chipset_flush(engine->gt);
>
> - vma = i915_vma_instance(obj, ctx->vm, NULL);
> + vma = i915_vma_instance(obj,
> + rcu_dereference_protected(ctx->vm, true),
> + NULL);
> if (IS_ERR(vma)) {
> err = PTR_ERR(vma);
> goto err;
> @@ -1463,27 +1455,25 @@ static int igt_vm_isolation(void *arg)
> if (IS_ERR(file))
> return PTR_ERR(file);
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> err = igt_live_test_begin(&t, i915, __func__, "");
> if (err)
> - goto out_unlock;
> + goto out_file;
>
> ctx_a = live_context(i915, file);
> if (IS_ERR(ctx_a)) {
> err = PTR_ERR(ctx_a);
> - goto out_unlock;
> + goto out_file;
> }
>
> ctx_b = live_context(i915, file);
> if (IS_ERR(ctx_b)) {
> err = PTR_ERR(ctx_b);
> - goto out_unlock;
> + goto out_file;
> }
>
> /* We can only test vm isolation, if the vm are distinct */
> if (ctx_a->vm == ctx_b->vm)
> - goto out_unlock;
> + goto out_file;
>
> vm_total = ctx_a->vm->total;
> GEM_BUG_ON(ctx_b->vm->total != vm_total);
> @@ -1512,7 +1502,7 @@ static int igt_vm_isolation(void *arg)
> err = read_from_scratch(ctx_b, engine,
> offset, &value);
> if (err)
> - goto out_unlock;
> + goto out_file;
>
> if (value) {
> pr_err("%s: Read %08x from scratch (offset 0x%08x_%08x), after %lu reads!\n",
> @@ -1521,7 +1511,7 @@ static int igt_vm_isolation(void *arg)
> lower_32_bits(offset),
> this);
> err = -EINVAL;
> - goto out_unlock;
> + goto out_file;
> }
>
> this++;
> @@ -1531,11 +1521,9 @@ static int igt_vm_isolation(void *arg)
> pr_info("Checked %lu scratch offsets across %d engines\n",
> count, RUNTIME_INFO(i915)->num_engines);
>
> -out_unlock:
> +out_file:
> if (igt_live_test_end(&t))
> err = -EIO;
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> mock_file_free(i915, file);
> return err;
> }
> @@ -1567,13 +1555,9 @@ static int mock_context_barrier(void *arg)
> * a request; useful for retiring old state after loading new.
> */
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> ctx = mock_context(i915, "mock");
> - if (!ctx) {
> - err = -ENOMEM;
> - goto unlock;
> - }
> + if (!ctx)
> + return -ENOMEM;
>
> counter = 0;
> err = context_barrier_task(ctx, 0,
> @@ -1646,8 +1630,6 @@ static int mock_context_barrier(void *arg)
>
> out:
> mock_context_close(ctx);
> -unlock:
> - mutex_unlock(&i915->drm.struct_mutex);
> return err;
> #undef pr_fmt
> #define pr_fmt(x) x
> diff --git a/drivers/gpu/drm/i915/gem/selftests/mock_context.c b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
> index ebc46f098561..6ef86e7923a7 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/mock_context.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
> @@ -67,7 +67,7 @@ void mock_context_close(struct i915_gem_context *ctx)
>
> void mock_init_contexts(struct drm_i915_private *i915)
> {
> - init_contexts(i915);
> + init_contexts(&i915->gem.contexts);
> }
>
> struct i915_gem_context *
> @@ -76,8 +76,6 @@ live_context(struct drm_i915_private *i915, struct drm_file *file)
> struct i915_gem_context *ctx;
> int err;
>
> - lockdep_assert_held(&i915->drm.struct_mutex);
> -
> ctx = i915_gem_create_context(i915, 0);
> if (IS_ERR(ctx))
> return ctx;
> diff --git a/drivers/gpu/drm/i915/gt/intel_context.c b/drivers/gpu/drm/i915/gt/intel_context.c
> index 35a40c2820a2..107eb03d6dd4 100644
> --- a/drivers/gpu/drm/i915/gt/intel_context.c
> +++ b/drivers/gpu/drm/i915/gt/intel_context.c
> @@ -226,7 +226,13 @@ intel_context_init(struct intel_context *ce,
> kref_init(&ce->ref);
>
> ce->gem_context = ctx;
> - ce->vm = i915_vm_get(ctx->vm ?: &engine->gt->ggtt->vm);
> + rcu_read_lock();
> + ce->vm = rcu_dereference(ctx->vm);
> + if (ce->vm && !kref_get_unless_zero(&ce->vm->ref))
> + ce->vm = NULL;
> + if (!ce->vm)
> + ce->vm = i915_vm_get(&engine->gt->ggtt->vm);
> + rcu_read_unlock();
> if (ctx->timeline)
> ce->timeline = intel_timeline_get(ctx->timeline);
>
> diff --git a/drivers/gpu/drm/i915/gt/selftest_context.c b/drivers/gpu/drm/i915/gt/selftest_context.c
> index 86cffbb0a9cb..7c838a57e174 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_context.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_context.c
> @@ -155,13 +155,9 @@ static int live_context_size(void *arg)
> * HW tries to write past the end of one.
> */
>
> - mutex_lock(>->i915->drm.struct_mutex);
> -
> fixme = kernel_context(gt->i915);
> - if (IS_ERR(fixme)) {
> - err = PTR_ERR(fixme);
> - goto unlock;
> - }
> + if (IS_ERR(fixme))
> + return PTR_ERR(fixme);
>
> for_each_engine(engine, gt->i915, id) {
> struct {
> @@ -201,8 +197,6 @@ static int live_context_size(void *arg)
> }
>
> kernel_context_close(fixme);
> -unlock:
> - mutex_unlock(>->i915->drm.struct_mutex);
> return err;
> }
>
> @@ -305,12 +299,10 @@ static int live_active_context(void *arg)
> if (IS_ERR(file))
> return PTR_ERR(file);
>
> - mutex_lock(>->i915->drm.struct_mutex);
> -
> fixme = live_context(gt->i915, file);
> if (IS_ERR(fixme)) {
> err = PTR_ERR(fixme);
> - goto unlock;
> + goto out_file;
> }
>
> for_each_engine(engine, gt->i915, id) {
> @@ -323,8 +315,7 @@ static int live_active_context(void *arg)
> break;
> }
>
> -unlock:
> - mutex_unlock(>->i915->drm.struct_mutex);
> +out_file:
> mock_file_free(gt->i915, file);
> return err;
> }
> @@ -418,12 +409,10 @@ static int live_remote_context(void *arg)
> if (IS_ERR(file))
> return PTR_ERR(file);
>
> - mutex_lock(>->i915->drm.struct_mutex);
> -
> fixme = live_context(gt->i915, file);
> if (IS_ERR(fixme)) {
> err = PTR_ERR(fixme);
> - goto unlock;
> + goto out_file;
> }
>
> for_each_engine(engine, gt->i915, id) {
> @@ -436,8 +425,7 @@ static int live_remote_context(void *arg)
> break;
> }
>
> -unlock:
> - mutex_unlock(>->i915->drm.struct_mutex);
> +out_file:
> mock_file_free(gt->i915, file);
> return err;
> }
> diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
> index f2b5ab1e5a2e..1b836ff0a756 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
> @@ -58,9 +58,7 @@ static int hang_init(struct hang *h, struct intel_gt *gt)
> memset(h, 0, sizeof(*h));
> h->gt = gt;
>
> - mutex_lock(>->i915->drm.struct_mutex);
> h->ctx = kernel_context(gt->i915);
> - mutex_unlock(>->i915->drm.struct_mutex);
> if (IS_ERR(h->ctx))
> return PTR_ERR(h->ctx);
>
> @@ -133,7 +131,8 @@ static struct i915_request *
> hang_create_request(struct hang *h, struct intel_engine_cs *engine)
> {
> struct intel_gt *gt = h->gt;
> - struct i915_address_space *vm = h->ctx->vm ?: &engine->gt->ggtt->vm;
> + struct i915_address_space *vm =
> + rcu_dereference_protected(h->ctx->vm, true) ?: &engine->gt->ggtt->vm;
> struct drm_i915_gem_object *obj;
> struct i915_request *rq = NULL;
> struct i915_vma *hws, *vma;
> @@ -382,9 +381,7 @@ static int igt_reset_nop(void *arg)
> if (IS_ERR(file))
> return PTR_ERR(file);
>
> - mutex_lock(>->i915->drm.struct_mutex);
> ctx = live_context(gt->i915, file);
> - mutex_unlock(>->i915->drm.struct_mutex);
> if (IS_ERR(ctx)) {
> err = PTR_ERR(ctx);
> goto out;
> @@ -458,9 +455,7 @@ static int igt_reset_nop_engine(void *arg)
> if (IS_ERR(file))
> return PTR_ERR(file);
>
> - mutex_lock(>->i915->drm.struct_mutex);
> ctx = live_context(gt->i915, file);
> - mutex_unlock(>->i915->drm.struct_mutex);
> if (IS_ERR(ctx)) {
> err = PTR_ERR(ctx);
> goto out;
> @@ -705,9 +700,7 @@ static int active_engine(void *data)
> return PTR_ERR(file);
>
> for (count = 0; count < ARRAY_SIZE(ctx); count++) {
> - mutex_lock(&engine->i915->drm.struct_mutex);
> ctx[count] = live_context(engine->i915, file);
> - mutex_unlock(&engine->i915->drm.struct_mutex);
> if (IS_ERR(ctx[count])) {
> err = PTR_ERR(ctx[count]);
> while (--count)
> @@ -1298,18 +1291,18 @@ static int igt_reset_evict_ppgtt(void *arg)
> if (IS_ERR(file))
> return PTR_ERR(file);
>
> - mutex_lock(>->i915->drm.struct_mutex);
> ctx = live_context(gt->i915, file);
> - mutex_unlock(>->i915->drm.struct_mutex);
> if (IS_ERR(ctx)) {
> err = PTR_ERR(ctx);
> goto out;
> }
>
> err = 0;
> - if (ctx->vm) /* aliasing == global gtt locking, covered above */
> - err = __igt_reset_evict_vma(gt, ctx->vm,
> + if (rcu_access_pointer(ctx->vm)) {
> + /* aliasing == global gtt locking, covered above */
> + err = __igt_reset_evict_vma(gt, rcu_dereference_protected(ctx->vm, true),
> evict_vma, EXEC_OBJECT_WRITE);
> + }
>
> out:
> mock_file_free(gt->i915, file);
> diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
> index 3300acd9199d..e33c2f66c683 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
> @@ -1453,7 +1453,9 @@ static int smoke_submit(struct preempt_smoke *smoke,
> int err = 0;
>
> if (batch) {
> - vma = i915_vma_instance(batch, ctx->vm, NULL);
> + vma = i915_vma_instance(batch,
> + rcu_dereference_protected(ctx->vm, true),
> + NULL);
> if (IS_ERR(vma))
> return PTR_ERR(vma);
>
> diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
> index 06351fefbbf3..ed44333d9e20 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
> @@ -33,6 +33,8 @@ struct wa_lists {
> } engine[I915_NUM_ENGINES];
> };
>
> +#define ctx_vm(ctx) rcu_dereference_protected((ctx)->vm, true)
> +
> static void
> reference_lists_init(struct drm_i915_private *i915, struct wa_lists *lists)
> {
> @@ -362,7 +364,7 @@ static struct i915_vma *create_batch(struct i915_gem_context *ctx)
> if (IS_ERR(obj))
> return ERR_CAST(obj);
>
> - vma = i915_vma_instance(obj, ctx->vm, NULL);
> + vma = i915_vma_instance(obj, ctx_vm(ctx), NULL);
> if (IS_ERR(vma)) {
> err = PTR_ERR(vma);
> goto err_obj;
> @@ -468,7 +470,7 @@ static int check_dirty_whitelist(struct i915_gem_context *ctx,
> int err = 0, i, v;
> u32 *cs, *results;
>
> - scratch = create_scratch(ctx->vm, 2 * ARRAY_SIZE(values) + 1);
> + scratch = create_scratch(ctx_vm(ctx), 2 * ARRAY_SIZE(values) + 1);
> if (IS_ERR(scratch))
> return PTR_ERR(scratch);
>
> @@ -1018,14 +1020,14 @@ static int live_isolated_whitelist(void *arg)
> goto err;
> }
>
> - client[i].scratch[0] = create_scratch(c->vm, 1024);
> + client[i].scratch[0] = create_scratch(ctx_vm(c), 1024);
> if (IS_ERR(client[i].scratch[0])) {
> err = PTR_ERR(client[i].scratch[0]);
> kernel_context_close(c);
> goto err;
> }
>
> - client[i].scratch[1] = create_scratch(c->vm, 1024);
> + client[i].scratch[1] = create_scratch(ctx_vm(c), 1024);
> if (IS_ERR(client[i].scratch[1])) {
> err = PTR_ERR(client[i].scratch[1]);
> i915_vma_unpin_and_release(&client[i].scratch[0], 0);
> diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
> index 03f567084548..ce5cff01f44c 100644
> --- a/drivers/gpu/drm/i915/gvt/scheduler.c
> +++ b/drivers/gpu/drm/i915/gvt/scheduler.c
> @@ -365,7 +365,8 @@ static void set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload,
> struct i915_gem_context *ctx)
> {
> struct intel_vgpu_mm *mm = workload->shadow_mm;
> - struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(ctx->vm);
> + struct i915_ppgtt *ppgtt =
> + i915_vm_to_ppgtt(rcu_dereference_protected(ctx->vm, true));
> int i = 0;
>
> if (mm->ppgtt_mm.root_entry_type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY) {
> @@ -1230,20 +1231,18 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
> struct intel_vgpu_submission *s = &vgpu->submission;
> struct intel_engine_cs *engine;
> struct i915_gem_context *ctx;
> + struct i915_ppgtt *ppgtt;
> enum intel_engine_id i;
> int ret;
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> ctx = i915_gem_context_create_kernel(i915, I915_PRIORITY_MAX);
> - if (IS_ERR(ctx)) {
> - ret = PTR_ERR(ctx);
> - goto out_unlock;
> - }
> + if (IS_ERR(ctx))
> + return PTR_ERR(ctx);
>
> i915_gem_context_set_force_single_submission(ctx);
>
> - i915_context_ppgtt_root_save(s, i915_vm_to_ppgtt(ctx->vm));
> + ppgtt = i915_vm_to_ppgtt(rcu_dereference_protected(ctx->vm, true));
> + i915_context_ppgtt_root_save(s, ppgtt);
>
> for_each_engine(engine, i915, i) {
> struct intel_context *ce;
> @@ -1289,11 +1288,10 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
> bitmap_zero(s->tlb_handle_pending, I915_NUM_ENGINES);
>
> i915_gem_context_put(ctx);
> - mutex_unlock(&i915->drm.struct_mutex);
> return 0;
>
> out_shadow_ctx:
> - i915_context_ppgtt_root_restore(s, i915_vm_to_ppgtt(ctx->vm));
> + i915_context_ppgtt_root_restore(s, ppgtt);
> for_each_engine(engine, i915, i) {
> if (IS_ERR(s->shadow[i]))
> break;
> @@ -1302,8 +1300,6 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
> intel_context_put(s->shadow[i]);
> }
> i915_gem_context_put(ctx);
> -out_unlock:
> - mutex_unlock(&i915->drm.struct_mutex);
> return ret;
> }
>
> diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
> index 8caaa446490f..d4a7cda60679 100644
> --- a/drivers/gpu/drm/i915/i915_debugfs.c
> +++ b/drivers/gpu/drm/i915/i915_debugfs.c
> @@ -315,12 +315,18 @@ static void print_context_stats(struct seq_file *m,
> struct drm_i915_private *i915)
> {
> struct file_stats kstats = {};
> - struct i915_gem_context *ctx;
> + struct i915_gem_context *ctx, *cn;
>
> - list_for_each_entry(ctx, &i915->contexts.list, link) {
> + spin_lock(&i915->gem.contexts.lock);
> + list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) {
> struct i915_gem_engines_iter it;
> struct intel_context *ce;
>
> + if (!kref_get_unless_zero(&ctx->ref))
> + continue;
> +
> + spin_unlock(&i915->gem.contexts.lock);
> +
> for_each_gem_engine(ce,
> i915_gem_context_lock_engines(ctx), it) {
> intel_context_lock_pinned(ce);
> @@ -337,7 +343,9 @@ static void print_context_stats(struct seq_file *m,
> i915_gem_context_unlock_engines(ctx);
>
> if (!IS_ERR_OR_NULL(ctx->file_priv)) {
> - struct file_stats stats = { .vm = ctx->vm, };
> + struct file_stats stats = {
> + .vm = rcu_access_pointer(ctx->vm),
> + };
> struct drm_file *file = ctx->file_priv->file;
> struct task_struct *task;
> char name[80];
> @@ -354,7 +362,12 @@ static void print_context_stats(struct seq_file *m,
>
> print_file_stats(m, name, stats);
> }
> +
> + spin_lock(&i915->gem.contexts.lock);
> + list_safe_reset_next(ctx, cn, link);
> + i915_gem_context_put(ctx);
> }
> + spin_unlock(&i915->gem.contexts.lock);
>
> print_file_stats(m, "[k]contexts", kstats);
> }
> @@ -362,7 +375,6 @@ static void print_context_stats(struct seq_file *m,
> static int i915_gem_object_info(struct seq_file *m, void *data)
> {
> struct drm_i915_private *i915 = node_to_i915(m->private);
> - int ret;
>
> seq_printf(m, "%u shrinkable [%u free] objects, %llu bytes\n",
> i915->mm.shrink_count,
> @@ -371,12 +383,7 @@ static int i915_gem_object_info(struct seq_file *m, void *data)
>
> seq_putc(m, '\n');
>
> - ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
> - if (ret)
> - return ret;
> -
> print_context_stats(m, i915);
> - mutex_unlock(&i915->drm.struct_mutex);
>
> return 0;
> }
> @@ -1490,19 +1497,19 @@ static void describe_ctx_ring(struct seq_file *m, struct intel_ring *ring)
>
> static int i915_context_status(struct seq_file *m, void *unused)
> {
> - struct drm_i915_private *dev_priv = node_to_i915(m->private);
> - struct drm_device *dev = &dev_priv->drm;
> - struct i915_gem_context *ctx;
> - int ret;
> -
> - ret = mutex_lock_interruptible(&dev->struct_mutex);
> - if (ret)
> - return ret;
> + struct drm_i915_private *i915 = node_to_i915(m->private);
> + struct i915_gem_context *ctx, *cn;
>
> - list_for_each_entry(ctx, &dev_priv->contexts.list, link) {
> + spin_lock(&i915->gem.contexts.lock);
> + list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) {
> struct i915_gem_engines_iter it;
> struct intel_context *ce;
>
> + if (!kref_get_unless_zero(&ctx->ref))
> + continue;
> +
> + spin_unlock(&i915->gem.contexts.lock);
> +
> seq_puts(m, "HW context ");
> if (ctx->pid) {
> struct task_struct *task;
> @@ -1537,9 +1544,12 @@ static int i915_context_status(struct seq_file *m, void *unused)
> i915_gem_context_unlock_engines(ctx);
>
> seq_putc(m, '\n');
> - }
>
> - mutex_unlock(&dev->struct_mutex);
> + spin_lock(&i915->gem.contexts.lock);
> + list_safe_reset_next(ctx, cn, link);
> + i915_gem_context_put(ctx);
> + }
> + spin_unlock(&i915->gem.contexts.lock);
>
> return 0;
> }
> diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
> index 8cac6124d16a..a8ddac22d633 100644
> --- a/drivers/gpu/drm/i915/i915_drv.c
> +++ b/drivers/gpu/drm/i915/i915_drv.c
> @@ -1685,10 +1685,8 @@ static void i915_driver_postclose(struct drm_device *dev, struct drm_file *file)
> {
> struct drm_i915_file_private *file_priv = file->driver_priv;
>
> - mutex_lock(&dev->struct_mutex);
> i915_gem_context_close(file);
> i915_gem_release(dev, file);
> - mutex_unlock(&dev->struct_mutex);
>
> kfree_rcu(file_priv, rcu);
>
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index ad7ed7f72da8..23f311172e75 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -1543,25 +1543,6 @@ struct drm_i915_private {
> int audio_power_refcount;
> u32 audio_freq_cntrl;
>
> - struct {
> - struct mutex mutex;
> - struct list_head list;
> - struct llist_head free_list;
> - struct work_struct free_work;
> -
> - /* The hw wants to have a stable context identifier for the
> - * lifetime of the context (for OA, PASID, faults, etc).
> - * This is limited in execlists to 21 bits.
> - */
> - struct ida hw_ida;
> -#define MAX_CONTEXT_HW_ID (1<<21) /* exclusive */
> -#define MAX_GUC_CONTEXT_HW_ID (1 << 20) /* exclusive */
> -#define GEN11_MAX_CONTEXT_HW_ID (1<<11) /* exclusive */
> -/* in Gen12 ID 0x7FF is reserved to indicate idle */
> -#define GEN12_MAX_CONTEXT_HW_ID (GEN11_MAX_CONTEXT_HW_ID - 1)
> - struct list_head hw_id_list;
These bits.
> - } contexts;
> -
> u32 fdi_rx_config;
>
> /* Shadow for DISPLAY_PHY_CONTROL which can't be safely read */
> @@ -1717,6 +1698,14 @@ struct drm_i915_private {
>
> struct {
> struct notifier_block pm_notifier;
> +
> + struct i915_gem_contexts {
> + spinlock_t lock; /* locks list */
> + struct list_head list;
> +
> + struct llist_head free_list;
> + struct work_struct free_work;
> + } contexts;
> } gem;
>
> /* For i945gm vblank irq vs. C3 workaround */
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index 9c2877441540..8814db7023db 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -1263,7 +1263,7 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
> goto err_unlock;
> }
>
> - ret = i915_gem_contexts_init(dev_priv);
> + ret = i915_gem_init_contexts(dev_priv);
> if (ret) {
> GEM_BUG_ON(ret == -EIO);
> goto err_scratch;
> @@ -1345,7 +1345,7 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
> }
> err_context:
> if (ret != -EIO)
> - i915_gem_contexts_fini(dev_priv);
> + i915_gem_driver_release__contexts(dev_priv);
> err_scratch:
> intel_gt_driver_release(&dev_priv->gt);
> err_unlock:
> @@ -1413,11 +1413,9 @@ void i915_gem_driver_remove(struct drm_i915_private *dev_priv)
>
> void i915_gem_driver_release(struct drm_i915_private *dev_priv)
> {
> - mutex_lock(&dev_priv->drm.struct_mutex);
> intel_engines_cleanup(dev_priv);
> - i915_gem_contexts_fini(dev_priv);
> + i915_gem_driver_release__contexts(dev_priv);
> intel_gt_driver_release(&dev_priv->gt);
> - mutex_unlock(&dev_priv->drm.struct_mutex);
>
> intel_wa_list_free(&dev_priv->gt_wa_list);
>
> @@ -1427,7 +1425,7 @@ void i915_gem_driver_release(struct drm_i915_private *dev_priv)
>
> i915_gem_drain_freed_objects(dev_priv);
>
> - WARN_ON(!list_empty(&dev_priv->contexts.list));
> + WARN_ON(!list_empty(&dev_priv->gem.contexts.list));
> }
>
> void i915_gem_init_mmio(struct drm_i915_private *i915)
> diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
> index 0d040517787b..db32e0fe78b1 100644
> --- a/drivers/gpu/drm/i915/i915_gem_gtt.c
> +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
> @@ -1366,7 +1366,9 @@ static int gen8_init_scratch(struct i915_address_space *vm)
> if (vm->has_read_only &&
> vm->i915->kernel_context &&
> vm->i915->kernel_context->vm) {
> - struct i915_address_space *clone = vm->i915->kernel_context->vm;
> + struct i915_address_space *clone =
> + rcu_dereference_protected(vm->i915->kernel_context->vm,
> + true);
What is the protection here?
>
> GEM_BUG_ON(!clone->has_read_only);
>
> diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
> index d36ba248d438..bfe6e9b6d71d 100644
> --- a/drivers/gpu/drm/i915/i915_perf.c
> +++ b/drivers/gpu/drm/i915/i915_perf.c
> @@ -1853,8 +1853,8 @@ static int gen8_configure_all_contexts(struct i915_perf_stream *stream,
> };
> #undef ctx_flexeuN
> struct intel_engine_cs *engine;
> - struct i915_gem_context *ctx;
> - int i;
> + struct i915_gem_context *ctx, *cn;
> + int i, err;
>
> for (i = 2; i < ARRAY_SIZE(regs); i++)
> regs[i].value = oa_config_flex_reg(oa_config, regs[i].reg);
> @@ -1877,16 +1877,27 @@ static int gen8_configure_all_contexts(struct i915_perf_stream *stream,
> * context. Contexts idle at the time of reconfiguration are not
> * trapped behind the barrier.
> */
> - list_for_each_entry(ctx, &i915->contexts.list, link) {
> - int err;
> -
> + spin_lock(&i915->gem.contexts.lock);
> + list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) {
> if (ctx == i915->kernel_context)
> continue;
>
> + if (!kref_get_unless_zero(&ctx->ref))
> + continue;
> +
> + spin_unlock(&i915->gem.contexts.lock);
> +
> err = gen8_configure_context(ctx, regs, ARRAY_SIZE(regs));
> - if (err)
> + if (err) {
> + i915_gem_context_put(ctx);
> return err;
> + }
> +
> + spin_lock(&i915->gem.contexts.lock);
> + list_safe_reset_next(ctx, cn, link);
> + i915_gem_context_put(ctx);
> }
> + spin_unlock(&i915->gem.contexts.lock);
>
> /*
> * After updating all other contexts, we need to modify ourselves.
> @@ -1895,7 +1906,6 @@ static int gen8_configure_all_contexts(struct i915_perf_stream *stream,
> */
> for_each_uabi_engine(engine, i915) {
> struct intel_context *ce = engine->kernel_context;
> - int err;
>
> if (engine->class != RENDER_CLASS)
> continue;
> diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c
> index 6b88d934927a..585808c2ee84 100644
> --- a/drivers/gpu/drm/i915/i915_sysfs.c
> +++ b/drivers/gpu/drm/i915/i915_sysfs.c
> @@ -144,9 +144,9 @@ static const struct attribute_group media_rc6_attr_group = {
> };
> #endif
>
> -static int l3_access_valid(struct drm_i915_private *dev_priv, loff_t offset)
> +static int l3_access_valid(struct drm_i915_private *i915, loff_t offset)
> {
> - if (!HAS_L3_DPF(dev_priv))
> + if (!HAS_L3_DPF(i915))
> return -EPERM;
>
> if (offset % 4 != 0)
> @@ -164,31 +164,24 @@ i915_l3_read(struct file *filp, struct kobject *kobj,
> loff_t offset, size_t count)
> {
> struct device *kdev = kobj_to_dev(kobj);
> - struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev);
> - struct drm_device *dev = &dev_priv->drm;
> + struct drm_i915_private *i915 = kdev_minor_to_i915(kdev);
> int slice = (int)(uintptr_t)attr->private;
> int ret;
>
> - count = round_down(count, 4);
> -
> - ret = l3_access_valid(dev_priv, offset);
> + ret = l3_access_valid(i915, offset);
> if (ret)
> return ret;
>
> + count = round_down(count, 4);
> count = min_t(size_t, GEN7_L3LOG_SIZE - offset, count);
> + memset(buf, 0, count);
>
> - ret = i915_mutex_lock_interruptible(dev);
> - if (ret)
> - return ret;
> -
> - if (dev_priv->l3_parity.remap_info[slice])
> + spin_lock(&i915->gem.contexts.lock);
> + if (i915->l3_parity.remap_info[slice])
> memcpy(buf,
> - dev_priv->l3_parity.remap_info[slice] + (offset/4),
> + i915->l3_parity.remap_info[slice] + offset / 4,
> count);
> - else
> - memset(buf, 0, count);
> -
> - mutex_unlock(&dev->struct_mutex);
> + spin_unlock(&i915->gem.contexts.lock);
>
> return count;
> }
> @@ -199,46 +192,46 @@ i915_l3_write(struct file *filp, struct kobject *kobj,
> loff_t offset, size_t count)
> {
> struct device *kdev = kobj_to_dev(kobj);
> - struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev);
> - struct drm_device *dev = &dev_priv->drm;
> + struct drm_i915_private *i915 = kdev_minor_to_i915(kdev);
> struct i915_gem_context *ctx;
> int slice = (int)(uintptr_t)attr->private;
> - u32 **remap_info;
> + u32 *remap_info, *freeme = NULL;
> int ret;
>
> - ret = l3_access_valid(dev_priv, offset);
> + ret = l3_access_valid(i915, offset);
> if (ret)
> return ret;
>
> - ret = i915_mutex_lock_interruptible(dev);
> - if (ret)
> - return ret;
> + remap_info = kzalloc(GEN7_L3LOG_SIZE, GFP_KERNEL);
> + if (!remap_info)
> + return -ENOMEM;
>
> - remap_info = &dev_priv->l3_parity.remap_info[slice];
> - if (!*remap_info) {
> - *remap_info = kzalloc(GEN7_L3LOG_SIZE, GFP_KERNEL);
> - if (!*remap_info) {
> - ret = -ENOMEM;
> - goto out;
> - }
> + spin_lock(&i915->gem.contexts.lock);
> +
> + if (i915->l3_parity.remap_info[slice]) {
> + freeme = remap_info;
> + remap_info = i915->l3_parity.remap_info[slice];
> + } else {
> + i915->l3_parity.remap_info[slice] = remap_info;
> }
>
> - /* TODO: Ideally we really want a GPU reset here to make sure errors
> - * aren't propagated. Since I cannot find a stable way to reset the GPU
> - * at this point it is left as a TODO.
> - */
> - memcpy(*remap_info + (offset/4), buf, count);
> + count = round_down(count, 4);
Where was this round_down before, can't see it!
> + memcpy(remap_info + offset / 4, buf, count);
>
> /* NB: We defer the remapping until we switch to the context */
> - list_for_each_entry(ctx, &dev_priv->contexts.list, link)
> - ctx->remap_slice |= (1<<slice);
> + list_for_each_entry(ctx, &i915->gem.contexts.list, link)
> + ctx->remap_slice |= BIT(slice);
>
> - ret = count;
> + spin_unlock(&i915->gem.contexts.lock);
> + kfree(freeme);
>
> -out:
> - mutex_unlock(&dev->struct_mutex);
> + /*
> + * TODO: Ideally we really want a GPU reset here to make sure errors
> + * aren't propagated. Since I cannot find a stable way to reset the GPU
> + * at this point it is left as a TODO.
> + */
>
> - return ret;
> + return count;
> }
>
> static const struct bin_attribute dpf_attrs = {
> diff --git a/drivers/gpu/drm/i915/i915_trace.h b/drivers/gpu/drm/i915/i915_trace.h
> index 1f2cf6cfafb5..7ef7a1e1664c 100644
> --- a/drivers/gpu/drm/i915/i915_trace.h
> +++ b/drivers/gpu/drm/i915/i915_trace.h
> @@ -952,7 +952,7 @@ DECLARE_EVENT_CLASS(i915_context,
> TP_fast_assign(
> __entry->dev = ctx->i915->drm.primary->index;
> __entry->ctx = ctx;
> - __entry->vm = ctx->vm;
> + __entry->vm = rcu_access_pointer(ctx->vm);
> ),
>
> TP_printk("dev=%u, ctx=%p, ctx_vm=%p",
> diff --git a/drivers/gpu/drm/i915/selftests/i915_gem.c b/drivers/gpu/drm/i915/selftests/i915_gem.c
> index 0346c3e5b6b6..bfa40a5b6d98 100644
> --- a/drivers/gpu/drm/i915/selftests/i915_gem.c
> +++ b/drivers/gpu/drm/i915/selftests/i915_gem.c
> @@ -138,11 +138,9 @@ static int igt_gem_suspend(void *arg)
> return PTR_ERR(file);
>
> err = -ENOMEM;
> - mutex_lock(&i915->drm.struct_mutex);
> ctx = live_context(i915, file);
> if (!IS_ERR(ctx))
> err = switch_to_context(i915, ctx);
> - mutex_unlock(&i915->drm.struct_mutex);
> if (err)
> goto out;
>
> @@ -157,9 +155,7 @@ static int igt_gem_suspend(void *arg)
>
> pm_resume(i915);
>
> - mutex_lock(&i915->drm.struct_mutex);
> err = switch_to_context(i915, ctx);
> - mutex_unlock(&i915->drm.struct_mutex);
> out:
> mock_file_free(i915, file);
> return err;
> @@ -177,11 +173,9 @@ static int igt_gem_hibernate(void *arg)
> return PTR_ERR(file);
>
> err = -ENOMEM;
> - mutex_lock(&i915->drm.struct_mutex);
> ctx = live_context(i915, file);
> if (!IS_ERR(ctx))
> err = switch_to_context(i915, ctx);
> - mutex_unlock(&i915->drm.struct_mutex);
> if (err)
> goto out;
>
> @@ -196,9 +190,7 @@ static int igt_gem_hibernate(void *arg)
>
> pm_resume(i915);
>
> - mutex_lock(&i915->drm.struct_mutex);
> err = switch_to_context(i915, ctx);
> - mutex_unlock(&i915->drm.struct_mutex);
> out:
> mock_file_free(i915, file);
> return err;
> diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
> index f39f0282e78c..0af9a58d011d 100644
> --- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
> +++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
> @@ -473,7 +473,6 @@ static int igt_evict_contexts(void *arg)
> }
>
> count = 0;
> - mutex_lock(&i915->drm.struct_mutex);
> onstack_fence_init(&fence);
> do {
> struct i915_request *rq;
> @@ -510,8 +509,6 @@ static int igt_evict_contexts(void *arg)
> count++;
> err = 0;
> } while(1);
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> onstack_fence_fini(&fence);
> pr_info("Submitted %lu contexts/requests on %s\n",
> count, engine->name);
> diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
> index 9939dd40c6a8..6777a335bd44 100644
> --- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
> +++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
> @@ -1246,6 +1246,7 @@ static int exercise_mock(struct drm_i915_private *i915,
> unsigned long end_time))
> {
> const u64 limit = totalram_pages() << PAGE_SHIFT;
> + struct i915_address_space *vm;
> struct i915_gem_context *ctx;
> IGT_TIMEOUT(end_time);
> int err;
> @@ -1254,7 +1255,8 @@ static int exercise_mock(struct drm_i915_private *i915,
> if (!ctx)
> return -ENOMEM;
>
> - err = func(i915, ctx->vm, 0, min(ctx->vm->total, limit), end_time);
> + vm = rcu_dereference_protected(ctx->vm, true);
> + err = func(i915, vm, 0, min(vm->total, limit), end_time);
>
> mock_context_close(ctx);
> return err;
> diff --git a/drivers/gpu/drm/i915/selftests/i915_request.c b/drivers/gpu/drm/i915/selftests/i915_request.c
> index d1ee384a3260..4abf52733520 100644
> --- a/drivers/gpu/drm/i915/selftests/i915_request.c
> +++ b/drivers/gpu/drm/i915/selftests/i915_request.c
> @@ -181,9 +181,7 @@ static int igt_request_rewind(void *arg)
> struct intel_context *ce;
> int err = -EINVAL;
>
> - mutex_lock(&i915->drm.struct_mutex);
> ctx[0] = mock_context(i915, "A");
> - mutex_unlock(&i915->drm.struct_mutex);
>
> ce = i915_gem_context_get_engine(ctx[0], RCS0);
> GEM_BUG_ON(IS_ERR(ce));
> @@ -197,9 +195,7 @@ static int igt_request_rewind(void *arg)
> i915_request_get(request);
> i915_request_add(request);
>
> - mutex_lock(&i915->drm.struct_mutex);
> ctx[1] = mock_context(i915, "B");
> - mutex_unlock(&i915->drm.struct_mutex);
>
> ce = i915_gem_context_get_engine(ctx[1], RCS0);
> GEM_BUG_ON(IS_ERR(ce));
> @@ -438,9 +434,7 @@ static int mock_breadcrumbs_smoketest(void *arg)
> }
>
> for (n = 0; n < t.ncontexts; n++) {
> - mutex_lock(&t.engine->i915->drm.struct_mutex);
> t.contexts[n] = mock_context(t.engine->i915, "mock");
> - mutex_unlock(&t.engine->i915->drm.struct_mutex);
> if (!t.contexts[n]) {
> ret = -ENOMEM;
> goto out_contexts;
> @@ -734,7 +728,8 @@ static int live_empty_request(void *arg)
> static struct i915_vma *recursive_batch(struct drm_i915_private *i915)
> {
> struct i915_gem_context *ctx = i915->kernel_context;
> - struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm;
> + struct i915_address_space *vm =
> + rcu_dereference_protected(ctx->vm, true) ?: &i915->ggtt.vm;
> struct drm_i915_gem_object *obj;
> const int gen = INTEL_GEN(i915);
> struct i915_vma *vma;
> @@ -1109,9 +1104,7 @@ static int live_breadcrumbs_smoketest(void *arg)
> }
>
> for (n = 0; n < t[0].ncontexts; n++) {
> - mutex_lock(&i915->drm.struct_mutex);
> t[0].contexts[n] = live_context(i915, file);
> - mutex_unlock(&i915->drm.struct_mutex);
> if (!t[0].contexts[n]) {
> ret = -ENOMEM;
> goto out_contexts;
> diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c
> index ac1ff558eb90..f6e6b0aae38c 100644
> --- a/drivers/gpu/drm/i915/selftests/i915_vma.c
> +++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
> @@ -38,7 +38,7 @@ static bool assert_vma(struct i915_vma *vma,
> {
> bool ok = true;
>
> - if (vma->vm != ctx->vm) {
> + if (vma->vm != rcu_access_pointer(ctx->vm)) {
> pr_err("VMA created with wrong VM\n");
> ok = false;
> }
> @@ -113,7 +113,8 @@ static int create_vmas(struct drm_i915_private *i915,
> list_for_each_entry(obj, objects, st_link) {
> for (pinned = 0; pinned <= 1; pinned++) {
> list_for_each_entry(ctx, contexts, link) {
> - struct i915_address_space *vm = ctx->vm;
> + struct i915_address_space *vm =
> + rcu_dereference_protected(ctx->vm, true);
> struct i915_vma *vma;
> int err;
>
> diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
> index efe40c29fcf2..0ed21a7b8682 100644
> --- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
> +++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
> @@ -59,11 +59,9 @@ static void mock_device_release(struct drm_device *dev)
>
> i915_gem_drain_workqueue(i915);
>
> - mutex_lock(&i915->drm.struct_mutex);
> for_each_engine(engine, i915, id)
> mock_engine_free(engine);
> - i915_gem_contexts_fini(i915);
> - mutex_unlock(&i915->drm.struct_mutex);
> + i915_gem_driver_release__contexts(i915);
>
> intel_timelines_fini(i915);
>
> @@ -207,7 +205,7 @@ struct drm_i915_private *mock_gem_device(void)
> return i915;
>
> err_context:
> - i915_gem_contexts_fini(i915);
> + i915_gem_driver_release__contexts(i915);
> err_engine:
> mock_engine_free(i915->engine[RCS0]);
> err_unlock:
>
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* Re: [PATCH 21/27] drm/i915: Move context management under GEM
2019-09-26 13:57 ` Tvrtko Ursulin
@ 2019-10-02 16:09 ` Tvrtko Ursulin
2019-10-03 7:35 ` Chris Wilson
1 sibling, 0 replies; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-10-02 16:09 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
Ping on some comments from the previous round:
On 26/09/2019 14:57, Tvrtko Ursulin wrote:
>
> On 25/09/2019 11:01, Chris Wilson wrote:
>> Keep track of the GEM contexts underneath i915->gem.contexts and assign
>> them their own lock for the purposes of list management.
>>
>> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
>> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>> ---
>> drivers/gpu/drm/i915/gem/i915_gem_context.c | 155 +++++++-----------
>> drivers/gpu/drm/i915/gem/i915_gem_context.h | 4 +-
>> .../gpu/drm/i915/gem/i915_gem_context_types.h | 2 +-
>> .../gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 +-
>> drivers/gpu/drm/i915/gem/i915_gem_userptr.c | 2 +-
>> .../gpu/drm/i915/gem/selftests/huge_pages.c | 12 +-
>> .../drm/i915/gem/selftests/i915_gem_context.c | 122 ++++++--------
>> .../gpu/drm/i915/gem/selftests/mock_context.c | 4 +-
>> drivers/gpu/drm/i915/gt/intel_context.c | 8 +-
>> drivers/gpu/drm/i915/gt/selftest_context.c | 24 +--
>> drivers/gpu/drm/i915/gt/selftest_hangcheck.c | 19 +--
>> drivers/gpu/drm/i915/gt/selftest_lrc.c | 4 +-
>> .../gpu/drm/i915/gt/selftest_workarounds.c | 10 +-
>> drivers/gpu/drm/i915/gvt/scheduler.c | 20 +--
>> drivers/gpu/drm/i915/i915_debugfs.c | 50 +++---
>> drivers/gpu/drm/i915/i915_drv.c | 2 -
>> drivers/gpu/drm/i915/i915_drv.h | 27 +--
>> drivers/gpu/drm/i915/i915_gem.c | 10 +-
>> drivers/gpu/drm/i915/i915_gem_gtt.c | 4 +-
>> drivers/gpu/drm/i915/i915_perf.c | 24 ++-
>> drivers/gpu/drm/i915/i915_sysfs.c | 77 ++++-----
>> drivers/gpu/drm/i915/i915_trace.h | 2 +-
>> drivers/gpu/drm/i915/selftests/i915_gem.c | 8 -
>> .../gpu/drm/i915/selftests/i915_gem_evict.c | 3 -
>> drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 4 +-
>> drivers/gpu/drm/i915/selftests/i915_request.c | 11 +-
>> drivers/gpu/drm/i915/selftests/i915_vma.c | 5 +-
>> .../gpu/drm/i915/selftests/mock_gem_device.c | 6 +-
>> 28 files changed, 272 insertions(+), 349 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c
>> b/drivers/gpu/drm/i915/gem/i915_gem_context.c
>> index 4e59b809d901..a77f439358d7 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
>> @@ -219,9 +219,12 @@ static struct i915_gem_engines
>> *default_engines(struct i915_gem_context *ctx)
>> static void i915_gem_context_free(struct i915_gem_context *ctx)
>> {
>> - lockdep_assert_held(&ctx->i915->drm.struct_mutex);
>> GEM_BUG_ON(!i915_gem_context_is_closed(ctx));
>> + spin_lock(&ctx->i915->gem.contexts.lock);
>> + list_del(&ctx->link);
>> + spin_unlock(&ctx->i915->gem.contexts.lock);
>> +
>> free_engines(rcu_access_pointer(ctx->engines));
>> mutex_destroy(&ctx->engines_mutex);
>> @@ -231,56 +234,40 @@ static void i915_gem_context_free(struct
>> i915_gem_context *ctx)
>> kfree(ctx->name);
>> put_pid(ctx->pid);
>> - list_del(&ctx->link);
>> mutex_destroy(&ctx->mutex);
>> kfree_rcu(ctx, rcu);
>> }
>> -static void contexts_free(struct drm_i915_private *i915)
>> +static void contexts_free_all(struct llist_node *list)
>> {
>> - struct llist_node *freed = llist_del_all(&i915->contexts.free_list);
>> struct i915_gem_context *ctx, *cn;
>> - lockdep_assert_held(&i915->drm.struct_mutex);
>> -
>> - llist_for_each_entry_safe(ctx, cn, freed, free_link)
>> + llist_for_each_entry_safe(ctx, cn, list, free_link)
>> i915_gem_context_free(ctx);
>> }
>> -static void contexts_free_first(struct drm_i915_private *i915)
>> +static void contexts_flush_free(struct i915_gem_contexts *gc)
>> {
>> - struct i915_gem_context *ctx;
>> - struct llist_node *freed;
>> -
>> - lockdep_assert_held(&i915->drm.struct_mutex);
>> -
>> - freed = llist_del_first(&i915->contexts.free_list);
>> - if (!freed)
>> - return;
>> -
>> - ctx = container_of(freed, typeof(*ctx), free_link);
>> - i915_gem_context_free(ctx);
>> + contexts_free_all(llist_del_all(&gc->free_list));
>> }
>> static void contexts_free_worker(struct work_struct *work)
>> {
>> - struct drm_i915_private *i915 =
>> - container_of(work, typeof(*i915), contexts.free_work);
>> + struct i915_gem_contexts *gc =
>> + container_of(work, typeof(*gc), free_work);
>> - mutex_lock(&i915->drm.struct_mutex);
>> - contexts_free(i915);
>> - mutex_unlock(&i915->drm.struct_mutex);
>> + contexts_flush_free(gc);
>> }
>> void i915_gem_context_release(struct kref *ref)
>> {
>> struct i915_gem_context *ctx = container_of(ref, typeof(*ctx),
>> ref);
>> - struct drm_i915_private *i915 = ctx->i915;
>> + struct i915_gem_contexts *gc = &ctx->i915->gem.contexts;
>> trace_i915_context_free(ctx);
>> - if (llist_add(&ctx->free_link, &i915->contexts.free_list))
>> - queue_work(i915->wq, &i915->contexts.free_work);
>> + if (llist_add(&ctx->free_link, &gc->free_list))
>> + queue_work(ctx->i915->wq, &gc->free_work);
>
> At first I thought gc was some sort of garbage collection list. But it
> is a GEM contexts list. :) This hunk looks completely avoidable, I don't
> see it brings much benefit on balance since in turn it adds ctx->i915
> twice . But as you wish..
>
>> }
>> static inline struct i915_gem_engines *
>> @@ -359,8 +346,8 @@ static void context_close(struct i915_gem_context
>> *ctx)
>> {
>> i915_gem_context_set_closed(ctx);
>> - if (ctx->vm)
>> - i915_vm_close(ctx->vm);
>> + if (rcu_access_pointer(ctx->vm))
>> + i915_vm_close(rcu_dereference_protected(ctx->vm, true));
>> mutex_lock(&ctx->mutex);
>> @@ -394,7 +381,6 @@ __create_context(struct drm_i915_private *i915)
>> return ERR_PTR(-ENOMEM);
>> kref_init(&ctx->ref);
>> - list_add_tail(&ctx->link, &i915->contexts.list);
>> ctx->i915 = i915;
>> ctx->sched.priority = I915_USER_PRIORITY(I915_PRIORITY_NORMAL);
>> mutex_init(&ctx->mutex);
>> @@ -435,6 +421,10 @@ __create_context(struct drm_i915_private *i915)
>> for (i = 0; i < ARRAY_SIZE(ctx->hang_timestamp); i++)
>> ctx->hang_timestamp[i] = jiffies - CONTEXT_FAST_HANG_JIFFIES;
>> + spin_lock(&i915->gem.contexts.lock);
>> + list_add_tail(&ctx->link, &i915->gem.contexts.list);
>> + spin_unlock(&i915->gem.contexts.lock);
>> +
>> return ctx;
>> err_free:
>> @@ -464,11 +454,11 @@ static void __apply_ppgtt(struct intel_context
>> *ce, void *vm)
>> static struct i915_address_space *
>> __set_ppgtt(struct i915_gem_context *ctx, struct i915_address_space
>> *vm)
>> {
>> - struct i915_address_space *old = ctx->vm;
>> + struct i915_address_space *old =
>> rcu_dereference_protected(ctx->vm, 1);
>> GEM_BUG_ON(old && i915_vm_is_4lvl(vm) != i915_vm_is_4lvl(old));
>> - ctx->vm = i915_vm_open(vm);
>> + rcu_assign_pointer(ctx->vm, i915_vm_open(vm));
>
> Are updaters here serialized? Lost track if struct mutex is still held
> in set_ppgtt at this point in the series or not.. Thinking of "old =
> rcu_dereference_protected.." and here.
>
>> context_apply_all(ctx, __apply_ppgtt, vm);
>> return old;
>> @@ -477,7 +467,7 @@ __set_ppgtt(struct i915_gem_context *ctx, struct
>> i915_address_space *vm)
>> static void __assign_ppgtt(struct i915_gem_context *ctx,
>> struct i915_address_space *vm)
>> {
>> - if (vm == ctx->vm)
>> + if (vm == rcu_access_pointer(ctx->vm))
>> return;
>> vm = __set_ppgtt(ctx, vm);
>> @@ -509,27 +499,25 @@ static void __assign_timeline(struct
>> i915_gem_context *ctx,
>> }
>> static struct i915_gem_context *
>> -i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned
>> int flags)
>> +i915_gem_create_context(struct drm_i915_private *i915, unsigned int
>> flags)
>> {
>> struct i915_gem_context *ctx;
>> - lockdep_assert_held(&dev_priv->drm.struct_mutex);
>> -
>> if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE &&
>> - !HAS_EXECLISTS(dev_priv))
>> + !HAS_EXECLISTS(i915))
>> return ERR_PTR(-EINVAL);
>> - /* Reap the most stale context */
>> - contexts_free_first(dev_priv);
>> + /* Reap the stale contexts */
>> + contexts_flush_free(&i915->gem.contexts);
>> - ctx = __create_context(dev_priv);
>> + ctx = __create_context(i915);
>> if (IS_ERR(ctx))
>> return ctx;
>> - if (HAS_FULL_PPGTT(dev_priv)) {
>> + if (HAS_FULL_PPGTT(i915)) {
>> struct i915_ppgtt *ppgtt;
>> - ppgtt = i915_ppgtt_create(dev_priv);
>> + ppgtt = i915_ppgtt_create(i915);
>> if (IS_ERR(ppgtt)) {
>> DRM_DEBUG_DRIVER("PPGTT setup failed (%ld)\n",
>> PTR_ERR(ppgtt));
>> @@ -544,7 +532,7 @@ i915_gem_create_context(struct drm_i915_private
>> *dev_priv, unsigned int flags)
>> if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE) {
>> struct intel_timeline *timeline;
>> - timeline = intel_timeline_create(&dev_priv->gt, NULL);
>> + timeline = intel_timeline_create(&i915->gt, NULL);
>> if (IS_ERR(timeline)) {
>> context_close(ctx);
>> return ERR_CAST(timeline);
>> @@ -590,48 +578,40 @@ i915_gem_context_create_kernel(struct
>> drm_i915_private *i915, int prio)
>> return ctx;
>> }
>> -static void init_contexts(struct drm_i915_private *i915)
>> +static void init_contexts(struct i915_gem_contexts *gc)
>> {
>> - mutex_init(&i915->contexts.mutex);
>> - INIT_LIST_HEAD(&i915->contexts.list);
>> -
>> - /* Using the simple ida interface, the max is limited by
>> sizeof(int) */
>> - BUILD_BUG_ON(MAX_CONTEXT_HW_ID > INT_MAX);
>> - BUILD_BUG_ON(GEN11_MAX_CONTEXT_HW_ID > INT_MAX);
>> - ida_init(&i915->contexts.hw_ida);
>> - INIT_LIST_HEAD(&i915->contexts.hw_id_list);
>
> This (and some more bits) belong to the hw_id removal patch.
>
>> + spin_lock_init(&gc->lock);
>> + INIT_LIST_HEAD(&gc->list);
>> - INIT_WORK(&i915->contexts.free_work, contexts_free_worker);
>> - init_llist_head(&i915->contexts.free_list);
>> + INIT_WORK(&gc->free_work, contexts_free_worker);
>> + init_llist_head(&gc->free_list);
>> }
>> -int i915_gem_contexts_init(struct drm_i915_private *dev_priv)
>> +int i915_gem_init_contexts(struct drm_i915_private *i915)
>> {
>> struct i915_gem_context *ctx;
>> /* Reassure ourselves we are only called once */
>> - GEM_BUG_ON(dev_priv->kernel_context);
>> + GEM_BUG_ON(i915->kernel_context);
>> - init_contexts(dev_priv);
>> + init_contexts(&i915->gem.contexts);
>> /* lowest priority; idle task */
>> - ctx = i915_gem_context_create_kernel(dev_priv, I915_PRIORITY_MIN);
>> + ctx = i915_gem_context_create_kernel(i915, I915_PRIORITY_MIN);
>> if (IS_ERR(ctx)) {
>> DRM_ERROR("Failed to create default global context\n");
>> return PTR_ERR(ctx);
>> }
>> - dev_priv->kernel_context = ctx;
>> + i915->kernel_context = ctx;
>> DRM_DEBUG_DRIVER("%s context support initialized\n",
>> - DRIVER_CAPS(dev_priv)->has_logical_contexts ?
>> + DRIVER_CAPS(i915)->has_logical_contexts ?
>> "logical" : "fake");
>> return 0;
>> }
>> -void i915_gem_contexts_fini(struct drm_i915_private *i915)
>> +void i915_gem_driver_release__contexts(struct drm_i915_private *i915)
>> {
>> - lockdep_assert_held(&i915->drm.struct_mutex);
>> -
>> destroy_kernel_context(&i915->kernel_context);
>> }
>> @@ -653,8 +633,8 @@ static int gem_context_register(struct
>> i915_gem_context *ctx,
>> int ret;
>> ctx->file_priv = fpriv;
>> - if (ctx->vm)
>> - ctx->vm->file = fpriv;
>> + if (rcu_access_pointer(ctx->vm))
>> + rcu_dereference_protected(ctx->vm, true)->file = fpriv;
>
> Here rcu accesses are just to satisfy sparse?
>
>> ctx->pid = get_task_pid(current, PIDTYPE_PID);
>> ctx->name = kasprintf(GFP_KERNEL, "%s[%d]",
>> @@ -691,9 +671,7 @@ int i915_gem_context_open(struct drm_i915_private
>> *i915,
>> idr_init(&file_priv->context_idr);
>> idr_init_base(&file_priv->vm_idr, 1);
>> - mutex_lock(&i915->drm.struct_mutex);
>> ctx = i915_gem_create_context(i915, 0);
>> - mutex_unlock(&i915->drm.struct_mutex);
>> if (IS_ERR(ctx)) {
>> err = PTR_ERR(ctx);
>> goto err;
>> @@ -721,6 +699,7 @@ int i915_gem_context_open(struct drm_i915_private
>> *i915,
>> void i915_gem_context_close(struct drm_file *file)
>> {
>> struct drm_i915_file_private *file_priv = file->driver_priv;
>> + struct drm_i915_private *i915 = file_priv->dev_priv;
>> idr_for_each(&file_priv->context_idr, context_idr_cleanup, NULL);
>> idr_destroy(&file_priv->context_idr);
>> @@ -729,6 +708,8 @@ void i915_gem_context_close(struct drm_file *file)
>> idr_for_each(&file_priv->vm_idr, vm_idr_cleanup, NULL);
>> idr_destroy(&file_priv->vm_idr);
>> mutex_destroy(&file_priv->vm_idr_lock);
>> +
>> + contexts_flush_free(&i915->gem.contexts);
>> }
>> int i915_gem_vm_create_ioctl(struct drm_device *dev, void *data,
>> @@ -907,16 +888,12 @@ static int get_ppgtt(struct
>> drm_i915_file_private *file_priv,
>> struct i915_address_space *vm;
>> int ret;
>> - if (!ctx->vm)
>> + if (!rcu_access_pointer(ctx->vm))
>> return -ENODEV;
>> - /* XXX rcu acquire? */
>> - ret = mutex_lock_interruptible(&ctx->i915->drm.struct_mutex);
>> - if (ret)
>> - return ret;
>> -
>> + rcu_read_lock();
>> vm = i915_vm_get(ctx->vm);
>> - mutex_unlock(&ctx->i915->drm.struct_mutex);
>> + rcu_read_unlock();
>> ret = mutex_lock_interruptible(&file_priv->vm_idr_lock);
>> if (ret)
>> @@ -1025,7 +1002,7 @@ static int set_ppgtt(struct
>> drm_i915_file_private *file_priv,
>> if (args->size)
>> return -EINVAL;
>> - if (!ctx->vm)
>> + if (!rcu_access_pointer(ctx->vm))
>> return -ENODEV;
>> if (upper_32_bits(args->value))
>> @@ -1039,17 +1016,15 @@ static int set_ppgtt(struct
>> drm_i915_file_private *file_priv,
>> if (!vm)
>> return -ENOENT;
>> - err = mutex_lock_interruptible(&ctx->i915->drm.struct_mutex);
>> + err = mutex_lock_interruptible(&ctx->mutex);
>> if (err)
>> goto out;
>> - if (vm == ctx->vm)
>> + if (vm == rcu_access_pointer(ctx->vm))
>> 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, vm);
>> @@ -1069,8 +1044,7 @@ static int set_ppgtt(struct
>> drm_i915_file_private *file_priv,
>> }
>> unlock:
>> - mutex_unlock(&ctx->i915->drm.struct_mutex);
>> -
>> + mutex_unlock(&ctx->mutex);
>> out:
>> i915_vm_put(vm);
>> return err;
>> @@ -1953,7 +1927,7 @@ static int clone_vm(struct i915_gem_context *dst,
>> rcu_read_lock();
>> do {
>> - vm = READ_ONCE(src->vm);
>> + vm = rcu_dereference(src->vm);
>> if (!vm)
>> break;
>> @@ -1975,7 +1949,7 @@ static int clone_vm(struct i915_gem_context *dst,
>> * it cannot be reallocated elsewhere.
>> */
>> - if (vm == READ_ONCE(src->vm))
>> + if (vm == rcu_access_pointer(src->vm))
>> break;
>> i915_vm_put(vm);
>> @@ -2077,12 +2051,7 @@ int i915_gem_context_create_ioctl(struct
>> drm_device *dev, void *data,
>> return -EIO;
>> }
>> - ret = i915_mutex_lock_interruptible(dev);
>> - if (ret)
>> - return ret;
>> -
>> ext_data.ctx = i915_gem_create_context(i915, args->flags);
>> - mutex_unlock(&dev->struct_mutex);
>> if (IS_ERR(ext_data.ctx))
>> return PTR_ERR(ext_data.ctx);
>> @@ -2209,10 +2178,12 @@ int i915_gem_context_getparam_ioctl(struct
>> drm_device *dev, void *data,
>> case I915_CONTEXT_PARAM_GTT_SIZE:
>> args->size = 0;
>> - if (ctx->vm)
>> - args->value = ctx->vm->total;
>> + rcu_read_lock();
>> + if (rcu_access_pointer(ctx->vm))
>> + args->value = rcu_dereference(ctx->vm)->total;
>> else
>> args->value = to_i915(dev)->ggtt.vm.total;
>> + rcu_read_unlock();
>> break;
>> case I915_CONTEXT_PARAM_NO_ERROR_CAPTURE:
>> @@ -2283,7 +2254,7 @@ int i915_gem_context_setparam_ioctl(struct
>> drm_device *dev, void *data,
>> int i915_gem_context_reset_stats_ioctl(struct drm_device *dev,
>> void *data, struct drm_file *file)
>> {
>> - struct drm_i915_private *dev_priv = to_i915(dev);
>> + struct drm_i915_private *i915 = to_i915(dev);
>> struct drm_i915_reset_stats *args = data;
>> struct i915_gem_context *ctx;
>> int ret;
>> @@ -2305,7 +2276,7 @@ int i915_gem_context_reset_stats_ioctl(struct
>> drm_device *dev,
>> */
>> if (capable(CAP_SYS_ADMIN))
>> - args->reset_count = i915_reset_count(&dev_priv->gpu_error);
>> + args->reset_count = i915_reset_count(&i915->gpu_error);
>> else
>> args->reset_count = 0;
>> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.h
>> b/drivers/gpu/drm/i915/gem/i915_gem_context.h
>> index 1b0df53436cf..4ee5dfc5794e 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_context.h
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.h
>> @@ -133,8 +133,8 @@ 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_fini(struct drm_i915_private *dev_priv);
>> +int __must_check i915_gem_init_contexts(struct drm_i915_private *i915);
>> +void i915_gem_driver_release__contexts(struct drm_i915_private *i915);
>> int i915_gem_context_open(struct drm_i915_private *i915,
>> struct drm_file *file);
>> 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 6419da7c9f90..a3ecd19f2303 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
>> @@ -88,7 +88,7 @@ struct i915_gem_context {
>> * In other modes, this is a NULL pointer with the expectation that
>> * the caller uses the shared global GTT.
>> */
>> - struct i915_address_space *vm;
>> + struct i915_address_space __rcu *vm;
>> /**
>> * @pid: process id of creator
>> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
>> b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
>> index 88a881be12ec..98816c35ffc3 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
>> @@ -728,7 +728,7 @@ static int eb_select_context(struct
>> i915_execbuffer *eb)
>> return -ENOENT;
>> eb->gem_context = ctx;
>> - if (ctx->vm)
>> + if (rcu_access_pointer(ctx->vm))
>> eb->invalid_flags |= EXEC_OBJECT_NEEDS_GTT;
>> eb->context_flags = 0;
>> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
>> b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
>> index 27573ef903b4..7c225e1d2a11 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
>> @@ -757,7 +757,7 @@ i915_gem_userptr_ioctl(struct drm_device *dev,
>> * On almost all of the older hw, we cannot tell the GPU that
>> * a page is readonly.
>> */
>> - vm = dev_priv->kernel_context->vm;
>> + vm = rcu_dereference_protected(dev_priv->kernel_context->vm,
>> true);
>
> Hm, should it not be rcu_dereference here?
>
>> if (!vm || !vm->has_read_only)
>> return -ENODEV;
>> }
>> diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
>> b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
>> index 98b2a6ccfcc1..e204e653b459 100644
>> --- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
>> +++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
>> @@ -1322,7 +1322,7 @@ static int igt_ppgtt_pin_update(void *arg)
>> struct i915_gem_context *ctx = arg;
>> struct drm_i915_private *dev_priv = ctx->i915;
>> unsigned long supported = INTEL_INFO(dev_priv)->page_sizes;
>> - struct i915_address_space *vm = ctx->vm;
>> + struct i915_address_space *vm =
>> rcu_dereference_protected(ctx->vm, 1);
>> struct drm_i915_gem_object *obj;
>> struct i915_gem_engines_iter it;
>> struct intel_context *ce;
>> @@ -1460,7 +1460,8 @@ static int igt_tmpfs_fallback(void *arg)
>> struct i915_gem_context *ctx = arg;
>> struct drm_i915_private *i915 = ctx->i915;
>> struct vfsmount *gemfs = i915->mm.gemfs;
>> - struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm;
>> + struct i915_address_space *vm =
>> + rcu_dereference_protected(ctx->vm, true) ?: &i915->ggtt.vm;
>> struct drm_i915_gem_object *obj;
>> struct i915_vma *vma;
>> u32 *vaddr;
>> @@ -1517,7 +1518,8 @@ static int igt_shrink_thp(void *arg)
>> {
>> struct i915_gem_context *ctx = arg;
>> struct drm_i915_private *i915 = ctx->i915;
>> - struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm;
>> + struct i915_address_space *vm =
>> + rcu_dereference_protected(ctx->vm, true) ?: &i915->ggtt.vm;
>> struct drm_i915_gem_object *obj;
>> struct i915_gem_engines_iter it;
>> struct intel_context *ce;
>> @@ -1699,8 +1701,8 @@ int i915_gem_huge_page_live_selftests(struct
>> drm_i915_private *i915)
>> goto out_unlock;
>> }
>> - if (ctx->vm)
>> - ctx->vm->scrub_64K = true;
>> + if (rcu_access_pointer(ctx->vm))
>> + rcu_dereference_protected(ctx->vm, true)->scrub_64K = true;
>> err = i915_subtests(tests, ctx);
>> 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 a8a9293f4ac6..7de1c0078d6f 100644
>> --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
>> +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
>> @@ -53,19 +53,17 @@ static int live_nop_switch(void *arg)
>> if (IS_ERR(file))
>> return PTR_ERR(file);
>> - mutex_lock(&i915->drm.struct_mutex);
>> -
>> ctx = kcalloc(nctx, sizeof(*ctx), GFP_KERNEL);
>> if (!ctx) {
>> err = -ENOMEM;
>> - goto out_unlock;
>> + goto out_file;
>> }
>> for (n = 0; n < nctx; n++) {
>> ctx[n] = live_context(i915, file);
>> if (IS_ERR(ctx[n])) {
>> err = PTR_ERR(ctx[n]);
>> - goto out_unlock;
>> + goto out_file;
>> }
>> }
>> @@ -79,7 +77,7 @@ static int live_nop_switch(void *arg)
>> rq = igt_request_alloc(ctx[n], engine);
>> if (IS_ERR(rq)) {
>> err = PTR_ERR(rq);
>> - goto out_unlock;
>> + goto out_file;
>> }
>> i915_request_add(rq);
>> }
>> @@ -87,7 +85,7 @@ static int live_nop_switch(void *arg)
>> pr_err("Failed to populated %d contexts\n", nctx);
>> intel_gt_set_wedged(&i915->gt);
>> err = -EIO;
>> - goto out_unlock;
>> + goto out_file;
>> }
>> times[1] = ktime_get_raw();
>> @@ -97,7 +95,7 @@ static int live_nop_switch(void *arg)
>> err = igt_live_test_begin(&t, i915, __func__, engine->name);
>> if (err)
>> - goto out_unlock;
>> + goto out_file;
>> end_time = jiffies + i915_selftest.timeout_jiffies;
>> for_each_prime_number_from(prime, 2, 8192) {
>> @@ -107,7 +105,7 @@ static int live_nop_switch(void *arg)
>> rq = igt_request_alloc(ctx[n % nctx], engine);
>> if (IS_ERR(rq)) {
>> err = PTR_ERR(rq);
>> - goto out_unlock;
>> + goto out_file;
>> }
>> /*
>> @@ -143,7 +141,7 @@ static int live_nop_switch(void *arg)
>> err = igt_live_test_end(&t);
>> if (err)
>> - goto out_unlock;
>> + goto out_file;
>> pr_info("Switch latencies on %s: 1 = %lluns, %lu = %lluns\n",
>> engine->name,
>> @@ -151,8 +149,7 @@ static int live_nop_switch(void *arg)
>> prime - 1, div64_u64(ktime_to_ns(times[1]), prime - 1));
>> }
>> -out_unlock:
>> - mutex_unlock(&i915->drm.struct_mutex);
>> +out_file:
>> mock_file_free(i915, file);
>> return err;
>> }
>> @@ -412,11 +409,9 @@ static int igt_ctx_exec(void *arg)
>> if (IS_ERR(file))
>> return PTR_ERR(file);
>> - mutex_lock(&i915->drm.struct_mutex);
>> -
>> err = igt_live_test_begin(&t, i915, __func__, engine->name);
>> if (err)
>> - goto out_unlock;
>> + goto out_file;
>> ncontexts = 0;
>> ndwords = 0;
>> @@ -428,7 +423,7 @@ static int igt_ctx_exec(void *arg)
>> ctx = kernel_context(i915);
>> if (IS_ERR(ctx)) {
>> err = PTR_ERR(ctx);
>> - goto out_unlock;
>> + goto out_file;
>> }
>> ce = i915_gem_context_get_engine(ctx, engine->legacy_idx);
>> @@ -440,7 +435,7 @@ static int igt_ctx_exec(void *arg)
>> err = PTR_ERR(obj);
>> intel_context_put(ce);
>> kernel_context_close(ctx);
>> - goto out_unlock;
>> + goto out_file;
>> }
>> }
>> @@ -449,17 +444,18 @@ static int igt_ctx_exec(void *arg)
>> pr_err("Failed to fill dword %lu [%lu/%lu] with gpu
>> (%s) [full-ppgtt? %s], err=%d\n",
>> ndwords, dw, max_dwords(obj),
>> engine->name,
>> - yesno(!!ctx->vm), err);
>> + yesno(!!rcu_access_pointer(ctx->vm)),
>> + err);
>> intel_context_put(ce);
>> kernel_context_close(ctx);
>> - goto out_unlock;
>> + goto out_file;
>> }
>> err = throttle(ce, tq, ARRAY_SIZE(tq));
>> if (err) {
>> intel_context_put(ce);
>> kernel_context_close(ctx);
>> - goto out_unlock;
>> + goto out_file;
>> }
>> if (++dw == max_dwords(obj)) {
>> @@ -489,11 +485,10 @@ static int igt_ctx_exec(void *arg)
>> dw += rem;
>> }
>> -out_unlock:
>> +out_file:
>> throttle_release(tq, ARRAY_SIZE(tq));
>> if (igt_live_test_end(&t))
>> err = -EIO;
>> - mutex_unlock(&i915->drm.struct_mutex);
>> mock_file_free(i915, file);
>> if (err)
>> @@ -528,22 +523,20 @@ static int igt_shared_ctx_exec(void *arg)
>> if (IS_ERR(file))
>> return PTR_ERR(file);
>> - mutex_lock(&i915->drm.struct_mutex);
>> -
>> parent = live_context(i915, file);
>> if (IS_ERR(parent)) {
>> err = PTR_ERR(parent);
>> - goto out_unlock;
>> + goto out_file;
>> }
>> if (!parent->vm) { /* not full-ppgtt; nothing to share */
>> err = 0;
>> - goto out_unlock;
>> + goto out_file;
>> }
>> err = igt_live_test_begin(&t, i915, __func__, "");
>> if (err)
>> - goto out_unlock;
>> + goto out_file;
>> for_each_engine(engine, i915, id) {
>> unsigned long ncontexts, ndwords, dw;
>> @@ -587,7 +580,8 @@ static int igt_shared_ctx_exec(void *arg)
>> pr_err("Failed to fill dword %lu [%lu/%lu] with gpu
>> (%s) [full-ppgtt? %s], err=%d\n",
>> ndwords, dw, max_dwords(obj),
>> engine->name,
>> - yesno(!!ctx->vm), err);
>> + yesno(!!rcu_access_pointer(ctx->vm)),
>> + err);
>> intel_context_put(ce);
>> kernel_context_close(ctx);
>> goto out_test;
>> @@ -626,17 +620,13 @@ static int igt_shared_ctx_exec(void *arg)
>> dw += rem;
>> }
>> - mutex_unlock(&i915->drm.struct_mutex);
>> i915_gem_drain_freed_objects(i915);
>> - mutex_lock(&i915->drm.struct_mutex);
>> }
>> out_test:
>> throttle_release(tq, ARRAY_SIZE(tq));
>> if (igt_live_test_end(&t))
>> err = -EIO;
>> -out_unlock:
>> - mutex_unlock(&i915->drm.struct_mutex);
>> -
>> +out_file:
>> mock_file_free(i915, file);
>> return err;
>> }
>> @@ -1008,8 +998,6 @@ __igt_ctx_sseu(struct drm_i915_private *i915,
>> if (flags & TEST_RESET)
>> igt_global_reset_lock(&i915->gt);
>> - mutex_lock(&i915->drm.struct_mutex);
>> -
>> ctx = live_context(i915, file);
>> if (IS_ERR(ctx)) {
>> ret = PTR_ERR(ctx);
>> @@ -1064,8 +1052,6 @@ __igt_ctx_sseu(struct drm_i915_private *i915,
>> i915_gem_object_put(obj);
>> out_unlock:
>> - mutex_unlock(&i915->drm.struct_mutex);
>> -
>> if (flags & TEST_RESET)
>> igt_global_reset_unlock(&i915->gt);
>> @@ -1125,23 +1111,24 @@ static int igt_ctx_readonly(void *arg)
>> if (IS_ERR(file))
>> return PTR_ERR(file);
>> - mutex_lock(&i915->drm.struct_mutex);
>> -
>> err = igt_live_test_begin(&t, i915, __func__, "");
>> if (err)
>> - goto out_unlock;
>> + goto out_file;
>> ctx = live_context(i915, file);
>> if (IS_ERR(ctx)) {
>> err = PTR_ERR(ctx);
>> - goto out_unlock;
>> + goto out_file;
>> }
>> - vm = ctx->vm ?: &i915->ggtt.alias->vm;
>> + rcu_read_lock();
>> + vm = rcu_dereference(ctx->vm) ?: &i915->ggtt.alias->vm;
>> if (!vm || !vm->has_read_only) {
>> + rcu_read_unlock();
>> err = 0;
>> - goto out_unlock;
>> + goto out_file;
>> }
>> + rcu_read_unlock();
>> ndwords = 0;
>> dw = 0;
>> @@ -1159,7 +1146,7 @@ static int igt_ctx_readonly(void *arg)
>> if (IS_ERR(obj)) {
>> err = PTR_ERR(obj);
>> i915_gem_context_unlock_engines(ctx);
>> - goto out_unlock;
>> + goto out_file;
>> }
>> if (prandom_u32_state(&prng) & 1)
>> @@ -1170,15 +1157,17 @@ static int igt_ctx_readonly(void *arg)
>> if (err) {
>> pr_err("Failed to fill dword %lu [%lu/%lu] with gpu
>> (%s) [full-ppgtt? %s], err=%d\n",
>> ndwords, dw, max_dwords(obj),
>> - ce->engine->name, yesno(!!ctx->vm), err);
>> + ce->engine->name,
>> + yesno(!!rcu_access_pointer(ctx->vm)),
>> + err);
>> i915_gem_context_unlock_engines(ctx);
>> - goto out_unlock;
>> + goto out_file;
>> }
>> err = throttle(ce, tq, ARRAY_SIZE(tq));
>> if (err) {
>> i915_gem_context_unlock_engines(ctx);
>> - goto out_unlock;
>> + goto out_file;
>> }
>> if (++dw == max_dwords(obj)) {
>> @@ -1210,11 +1199,10 @@ static int igt_ctx_readonly(void *arg)
>> dw += rem;
>> }
>> -out_unlock:
>> +out_file:
>> throttle_release(tq, ARRAY_SIZE(tq));
>> if (igt_live_test_end(&t))
>> err = -EIO;
>> - mutex_unlock(&i915->drm.struct_mutex);
>> mock_file_free(i915, file);
>> return err;
>> @@ -1223,7 +1211,7 @@ static int igt_ctx_readonly(void *arg)
>> static int check_scratch(struct i915_gem_context *ctx, u64 offset)
>> {
>> struct drm_mm_node *node =
>> - __drm_mm_interval_first(&ctx->vm->mm,
>> + __drm_mm_interval_first(&rcu_dereference_protected(ctx->vm,
>> true)->mm,
>> offset, offset + sizeof(u32) - 1);
>> if (!node || node->start > offset)
>> return 0;
>> @@ -1273,7 +1261,9 @@ static int write_to_scratch(struct
>> i915_gem_context *ctx,
>> intel_gt_chipset_flush(engine->gt);
>> - vma = i915_vma_instance(obj, ctx->vm, NULL);
>> + vma = i915_vma_instance(obj,
>> + rcu_dereference_protected(ctx->vm, true),
>> + NULL);
>> if (IS_ERR(vma)) {
>> err = PTR_ERR(vma);
>> goto err;
>> @@ -1372,7 +1362,9 @@ static int read_from_scratch(struct
>> i915_gem_context *ctx,
>> intel_gt_chipset_flush(engine->gt);
>> - vma = i915_vma_instance(obj, ctx->vm, NULL);
>> + vma = i915_vma_instance(obj,
>> + rcu_dereference_protected(ctx->vm, true),
>> + NULL);
>> if (IS_ERR(vma)) {
>> err = PTR_ERR(vma);
>> goto err;
>> @@ -1463,27 +1455,25 @@ static int igt_vm_isolation(void *arg)
>> if (IS_ERR(file))
>> return PTR_ERR(file);
>> - mutex_lock(&i915->drm.struct_mutex);
>> -
>> err = igt_live_test_begin(&t, i915, __func__, "");
>> if (err)
>> - goto out_unlock;
>> + goto out_file;
>> ctx_a = live_context(i915, file);
>> if (IS_ERR(ctx_a)) {
>> err = PTR_ERR(ctx_a);
>> - goto out_unlock;
>> + goto out_file;
>> }
>> ctx_b = live_context(i915, file);
>> if (IS_ERR(ctx_b)) {
>> err = PTR_ERR(ctx_b);
>> - goto out_unlock;
>> + goto out_file;
>> }
>> /* We can only test vm isolation, if the vm are distinct */
>> if (ctx_a->vm == ctx_b->vm)
>> - goto out_unlock;
>> + goto out_file;
>> vm_total = ctx_a->vm->total;
>> GEM_BUG_ON(ctx_b->vm->total != vm_total);
>> @@ -1512,7 +1502,7 @@ static int igt_vm_isolation(void *arg)
>> err = read_from_scratch(ctx_b, engine,
>> offset, &value);
>> if (err)
>> - goto out_unlock;
>> + goto out_file;
>> if (value) {
>> pr_err("%s: Read %08x from scratch (offset
>> 0x%08x_%08x), after %lu reads!\n",
>> @@ -1521,7 +1511,7 @@ static int igt_vm_isolation(void *arg)
>> lower_32_bits(offset),
>> this);
>> err = -EINVAL;
>> - goto out_unlock;
>> + goto out_file;
>> }
>> this++;
>> @@ -1531,11 +1521,9 @@ static int igt_vm_isolation(void *arg)
>> pr_info("Checked %lu scratch offsets across %d engines\n",
>> count, RUNTIME_INFO(i915)->num_engines);
>> -out_unlock:
>> +out_file:
>> if (igt_live_test_end(&t))
>> err = -EIO;
>> - mutex_unlock(&i915->drm.struct_mutex);
>> -
>> mock_file_free(i915, file);
>> return err;
>> }
>> @@ -1567,13 +1555,9 @@ static int mock_context_barrier(void *arg)
>> * a request; useful for retiring old state after loading new.
>> */
>> - mutex_lock(&i915->drm.struct_mutex);
>> -
>> ctx = mock_context(i915, "mock");
>> - if (!ctx) {
>> - err = -ENOMEM;
>> - goto unlock;
>> - }
>> + if (!ctx)
>> + return -ENOMEM;
>> counter = 0;
>> err = context_barrier_task(ctx, 0,
>> @@ -1646,8 +1630,6 @@ static int mock_context_barrier(void *arg)
>> out:
>> mock_context_close(ctx);
>> -unlock:
>> - mutex_unlock(&i915->drm.struct_mutex);
>> return err;
>> #undef pr_fmt
>> #define pr_fmt(x) x
>> diff --git a/drivers/gpu/drm/i915/gem/selftests/mock_context.c
>> b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
>> index ebc46f098561..6ef86e7923a7 100644
>> --- a/drivers/gpu/drm/i915/gem/selftests/mock_context.c
>> +++ b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
>> @@ -67,7 +67,7 @@ void mock_context_close(struct i915_gem_context *ctx)
>> void mock_init_contexts(struct drm_i915_private *i915)
>> {
>> - init_contexts(i915);
>> + init_contexts(&i915->gem.contexts);
>> }
>> struct i915_gem_context *
>> @@ -76,8 +76,6 @@ live_context(struct drm_i915_private *i915, struct
>> drm_file *file)
>> struct i915_gem_context *ctx;
>> int err;
>> - lockdep_assert_held(&i915->drm.struct_mutex);
>> -
>> ctx = i915_gem_create_context(i915, 0);
>> if (IS_ERR(ctx))
>> return ctx;
>> diff --git a/drivers/gpu/drm/i915/gt/intel_context.c
>> b/drivers/gpu/drm/i915/gt/intel_context.c
>> index 35a40c2820a2..107eb03d6dd4 100644
>> --- a/drivers/gpu/drm/i915/gt/intel_context.c
>> +++ b/drivers/gpu/drm/i915/gt/intel_context.c
>> @@ -226,7 +226,13 @@ intel_context_init(struct intel_context *ce,
>> kref_init(&ce->ref);
>> ce->gem_context = ctx;
>> - ce->vm = i915_vm_get(ctx->vm ?: &engine->gt->ggtt->vm);
>> + rcu_read_lock();
>> + ce->vm = rcu_dereference(ctx->vm);
>> + if (ce->vm && !kref_get_unless_zero(&ce->vm->ref))
>> + ce->vm = NULL;
>> + if (!ce->vm)
>> + ce->vm = i915_vm_get(&engine->gt->ggtt->vm);
>> + rcu_read_unlock();
>> if (ctx->timeline)
>> ce->timeline = intel_timeline_get(ctx->timeline);
>> diff --git a/drivers/gpu/drm/i915/gt/selftest_context.c
>> b/drivers/gpu/drm/i915/gt/selftest_context.c
>> index 86cffbb0a9cb..7c838a57e174 100644
>> --- a/drivers/gpu/drm/i915/gt/selftest_context.c
>> +++ b/drivers/gpu/drm/i915/gt/selftest_context.c
>> @@ -155,13 +155,9 @@ static int live_context_size(void *arg)
>> * HW tries to write past the end of one.
>> */
>> - mutex_lock(>->i915->drm.struct_mutex);
>> -
>> fixme = kernel_context(gt->i915);
>> - if (IS_ERR(fixme)) {
>> - err = PTR_ERR(fixme);
>> - goto unlock;
>> - }
>> + if (IS_ERR(fixme))
>> + return PTR_ERR(fixme);
>> for_each_engine(engine, gt->i915, id) {
>> struct {
>> @@ -201,8 +197,6 @@ static int live_context_size(void *arg)
>> }
>> kernel_context_close(fixme);
>> -unlock:
>> - mutex_unlock(>->i915->drm.struct_mutex);
>> return err;
>> }
>> @@ -305,12 +299,10 @@ static int live_active_context(void *arg)
>> if (IS_ERR(file))
>> return PTR_ERR(file);
>> - mutex_lock(>->i915->drm.struct_mutex);
>> -
>> fixme = live_context(gt->i915, file);
>> if (IS_ERR(fixme)) {
>> err = PTR_ERR(fixme);
>> - goto unlock;
>> + goto out_file;
>> }
>> for_each_engine(engine, gt->i915, id) {
>> @@ -323,8 +315,7 @@ static int live_active_context(void *arg)
>> break;
>> }
>> -unlock:
>> - mutex_unlock(>->i915->drm.struct_mutex);
>> +out_file:
>> mock_file_free(gt->i915, file);
>> return err;
>> }
>> @@ -418,12 +409,10 @@ static int live_remote_context(void *arg)
>> if (IS_ERR(file))
>> return PTR_ERR(file);
>> - mutex_lock(>->i915->drm.struct_mutex);
>> -
>> fixme = live_context(gt->i915, file);
>> if (IS_ERR(fixme)) {
>> err = PTR_ERR(fixme);
>> - goto unlock;
>> + goto out_file;
>> }
>> for_each_engine(engine, gt->i915, id) {
>> @@ -436,8 +425,7 @@ static int live_remote_context(void *arg)
>> break;
>> }
>> -unlock:
>> - mutex_unlock(>->i915->drm.struct_mutex);
>> +out_file:
>> mock_file_free(gt->i915, file);
>> return err;
>> }
>> diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
>> b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
>> index f2b5ab1e5a2e..1b836ff0a756 100644
>> --- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
>> +++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
>> @@ -58,9 +58,7 @@ static int hang_init(struct hang *h, struct intel_gt
>> *gt)
>> memset(h, 0, sizeof(*h));
>> h->gt = gt;
>> - mutex_lock(>->i915->drm.struct_mutex);
>> h->ctx = kernel_context(gt->i915);
>> - mutex_unlock(>->i915->drm.struct_mutex);
>> if (IS_ERR(h->ctx))
>> return PTR_ERR(h->ctx);
>> @@ -133,7 +131,8 @@ static struct i915_request *
>> hang_create_request(struct hang *h, struct intel_engine_cs *engine)
>> {
>> struct intel_gt *gt = h->gt;
>> - struct i915_address_space *vm = h->ctx->vm ?: &engine->gt->ggtt->vm;
>> + struct i915_address_space *vm =
>> + rcu_dereference_protected(h->ctx->vm, true) ?:
>> &engine->gt->ggtt->vm;
>> struct drm_i915_gem_object *obj;
>> struct i915_request *rq = NULL;
>> struct i915_vma *hws, *vma;
>> @@ -382,9 +381,7 @@ static int igt_reset_nop(void *arg)
>> if (IS_ERR(file))
>> return PTR_ERR(file);
>> - mutex_lock(>->i915->drm.struct_mutex);
>> ctx = live_context(gt->i915, file);
>> - mutex_unlock(>->i915->drm.struct_mutex);
>> if (IS_ERR(ctx)) {
>> err = PTR_ERR(ctx);
>> goto out;
>> @@ -458,9 +455,7 @@ static int igt_reset_nop_engine(void *arg)
>> if (IS_ERR(file))
>> return PTR_ERR(file);
>> - mutex_lock(>->i915->drm.struct_mutex);
>> ctx = live_context(gt->i915, file);
>> - mutex_unlock(>->i915->drm.struct_mutex);
>> if (IS_ERR(ctx)) {
>> err = PTR_ERR(ctx);
>> goto out;
>> @@ -705,9 +700,7 @@ static int active_engine(void *data)
>> return PTR_ERR(file);
>> for (count = 0; count < ARRAY_SIZE(ctx); count++) {
>> - mutex_lock(&engine->i915->drm.struct_mutex);
>> ctx[count] = live_context(engine->i915, file);
>> - mutex_unlock(&engine->i915->drm.struct_mutex);
>> if (IS_ERR(ctx[count])) {
>> err = PTR_ERR(ctx[count]);
>> while (--count)
>> @@ -1298,18 +1291,18 @@ static int igt_reset_evict_ppgtt(void *arg)
>> if (IS_ERR(file))
>> return PTR_ERR(file);
>> - mutex_lock(>->i915->drm.struct_mutex);
>> ctx = live_context(gt->i915, file);
>> - mutex_unlock(>->i915->drm.struct_mutex);
>> if (IS_ERR(ctx)) {
>> err = PTR_ERR(ctx);
>> goto out;
>> }
>> err = 0;
>> - if (ctx->vm) /* aliasing == global gtt locking, covered above */
>> - err = __igt_reset_evict_vma(gt, ctx->vm,
>> + if (rcu_access_pointer(ctx->vm)) {
>> + /* aliasing == global gtt locking, covered above */
>> + err = __igt_reset_evict_vma(gt,
>> rcu_dereference_protected(ctx->vm, true),
>> evict_vma, EXEC_OBJECT_WRITE);
>> + }
>> out:
>> mock_file_free(gt->i915, file);
>> diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c
>> b/drivers/gpu/drm/i915/gt/selftest_lrc.c
>> index 3300acd9199d..e33c2f66c683 100644
>> --- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
>> +++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
>> @@ -1453,7 +1453,9 @@ static int smoke_submit(struct preempt_smoke
>> *smoke,
>> int err = 0;
>> if (batch) {
>> - vma = i915_vma_instance(batch, ctx->vm, NULL);
>> + vma = i915_vma_instance(batch,
>> + rcu_dereference_protected(ctx->vm, true),
>> + NULL);
>> if (IS_ERR(vma))
>> return PTR_ERR(vma);
>> diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c
>> b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
>> index 06351fefbbf3..ed44333d9e20 100644
>> --- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c
>> +++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
>> @@ -33,6 +33,8 @@ struct wa_lists {
>> } engine[I915_NUM_ENGINES];
>> };
>> +#define ctx_vm(ctx) rcu_dereference_protected((ctx)->vm, true)
>> +
>> static void
>> reference_lists_init(struct drm_i915_private *i915, struct wa_lists
>> *lists)
>> {
>> @@ -362,7 +364,7 @@ static struct i915_vma *create_batch(struct
>> i915_gem_context *ctx)
>> if (IS_ERR(obj))
>> return ERR_CAST(obj);
>> - vma = i915_vma_instance(obj, ctx->vm, NULL);
>> + vma = i915_vma_instance(obj, ctx_vm(ctx), NULL);
>> if (IS_ERR(vma)) {
>> err = PTR_ERR(vma);
>> goto err_obj;
>> @@ -468,7 +470,7 @@ static int check_dirty_whitelist(struct
>> i915_gem_context *ctx,
>> int err = 0, i, v;
>> u32 *cs, *results;
>> - scratch = create_scratch(ctx->vm, 2 * ARRAY_SIZE(values) + 1);
>> + scratch = create_scratch(ctx_vm(ctx), 2 * ARRAY_SIZE(values) + 1);
>> if (IS_ERR(scratch))
>> return PTR_ERR(scratch);
>> @@ -1018,14 +1020,14 @@ static int live_isolated_whitelist(void *arg)
>> goto err;
>> }
>> - client[i].scratch[0] = create_scratch(c->vm, 1024);
>> + client[i].scratch[0] = create_scratch(ctx_vm(c), 1024);
>> if (IS_ERR(client[i].scratch[0])) {
>> err = PTR_ERR(client[i].scratch[0]);
>> kernel_context_close(c);
>> goto err;
>> }
>> - client[i].scratch[1] = create_scratch(c->vm, 1024);
>> + client[i].scratch[1] = create_scratch(ctx_vm(c), 1024);
>> if (IS_ERR(client[i].scratch[1])) {
>> err = PTR_ERR(client[i].scratch[1]);
>> i915_vma_unpin_and_release(&client[i].scratch[0], 0);
>> diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c
>> b/drivers/gpu/drm/i915/gvt/scheduler.c
>> index 03f567084548..ce5cff01f44c 100644
>> --- a/drivers/gpu/drm/i915/gvt/scheduler.c
>> +++ b/drivers/gpu/drm/i915/gvt/scheduler.c
>> @@ -365,7 +365,8 @@ static void set_context_ppgtt_from_shadow(struct
>> intel_vgpu_workload *workload,
>> struct i915_gem_context *ctx)
>> {
>> struct intel_vgpu_mm *mm = workload->shadow_mm;
>> - struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(ctx->vm);
>> + struct i915_ppgtt *ppgtt =
>> + i915_vm_to_ppgtt(rcu_dereference_protected(ctx->vm, true));
>> int i = 0;
>> if (mm->ppgtt_mm.root_entry_type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY) {
>> @@ -1230,20 +1231,18 @@ int intel_vgpu_setup_submission(struct
>> intel_vgpu *vgpu)
>> struct intel_vgpu_submission *s = &vgpu->submission;
>> struct intel_engine_cs *engine;
>> struct i915_gem_context *ctx;
>> + struct i915_ppgtt *ppgtt;
>> enum intel_engine_id i;
>> int ret;
>> - mutex_lock(&i915->drm.struct_mutex);
>> -
>> ctx = i915_gem_context_create_kernel(i915, I915_PRIORITY_MAX);
>> - if (IS_ERR(ctx)) {
>> - ret = PTR_ERR(ctx);
>> - goto out_unlock;
>> - }
>> + if (IS_ERR(ctx))
>> + return PTR_ERR(ctx);
>> i915_gem_context_set_force_single_submission(ctx);
>> - i915_context_ppgtt_root_save(s, i915_vm_to_ppgtt(ctx->vm));
>> + ppgtt = i915_vm_to_ppgtt(rcu_dereference_protected(ctx->vm, true));
>> + i915_context_ppgtt_root_save(s, ppgtt);
>> for_each_engine(engine, i915, i) {
>> struct intel_context *ce;
>> @@ -1289,11 +1288,10 @@ int intel_vgpu_setup_submission(struct
>> intel_vgpu *vgpu)
>> bitmap_zero(s->tlb_handle_pending, I915_NUM_ENGINES);
>> i915_gem_context_put(ctx);
>> - mutex_unlock(&i915->drm.struct_mutex);
>> return 0;
>> out_shadow_ctx:
>> - i915_context_ppgtt_root_restore(s, i915_vm_to_ppgtt(ctx->vm));
>> + i915_context_ppgtt_root_restore(s, ppgtt);
>> for_each_engine(engine, i915, i) {
>> if (IS_ERR(s->shadow[i]))
>> break;
>> @@ -1302,8 +1300,6 @@ int intel_vgpu_setup_submission(struct
>> intel_vgpu *vgpu)
>> intel_context_put(s->shadow[i]);
>> }
>> i915_gem_context_put(ctx);
>> -out_unlock:
>> - mutex_unlock(&i915->drm.struct_mutex);
>> return ret;
>> }
>> diff --git a/drivers/gpu/drm/i915/i915_debugfs.c
>> b/drivers/gpu/drm/i915/i915_debugfs.c
>> index 8caaa446490f..d4a7cda60679 100644
>> --- a/drivers/gpu/drm/i915/i915_debugfs.c
>> +++ b/drivers/gpu/drm/i915/i915_debugfs.c
>> @@ -315,12 +315,18 @@ static void print_context_stats(struct seq_file *m,
>> struct drm_i915_private *i915)
>> {
>> struct file_stats kstats = {};
>> - struct i915_gem_context *ctx;
>> + struct i915_gem_context *ctx, *cn;
>> - list_for_each_entry(ctx, &i915->contexts.list, link) {
>> + spin_lock(&i915->gem.contexts.lock);
>> + list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) {
>> struct i915_gem_engines_iter it;
>> struct intel_context *ce;
>> + if (!kref_get_unless_zero(&ctx->ref))
>> + continue;
>> +
>> + spin_unlock(&i915->gem.contexts.lock);
>> +
>> for_each_gem_engine(ce,
>> i915_gem_context_lock_engines(ctx), it) {
>> intel_context_lock_pinned(ce);
>> @@ -337,7 +343,9 @@ static void print_context_stats(struct seq_file *m,
>> i915_gem_context_unlock_engines(ctx);
>> if (!IS_ERR_OR_NULL(ctx->file_priv)) {
>> - struct file_stats stats = { .vm = ctx->vm, };
>> + struct file_stats stats = {
>> + .vm = rcu_access_pointer(ctx->vm),
>> + };
>> struct drm_file *file = ctx->file_priv->file;
>> struct task_struct *task;
>> char name[80];
>> @@ -354,7 +362,12 @@ static void print_context_stats(struct seq_file *m,
>> print_file_stats(m, name, stats);
>> }
>> +
>> + spin_lock(&i915->gem.contexts.lock);
>> + list_safe_reset_next(ctx, cn, link);
>> + i915_gem_context_put(ctx);
>> }
>> + spin_unlock(&i915->gem.contexts.lock);
>> print_file_stats(m, "[k]contexts", kstats);
>> }
>> @@ -362,7 +375,6 @@ static void print_context_stats(struct seq_file *m,
>> static int i915_gem_object_info(struct seq_file *m, void *data)
>> {
>> struct drm_i915_private *i915 = node_to_i915(m->private);
>> - int ret;
>> seq_printf(m, "%u shrinkable [%u free] objects, %llu bytes\n",
>> i915->mm.shrink_count,
>> @@ -371,12 +383,7 @@ static int i915_gem_object_info(struct seq_file
>> *m, void *data)
>> seq_putc(m, '\n');
>> - ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
>> - if (ret)
>> - return ret;
>> -
>> print_context_stats(m, i915);
>> - mutex_unlock(&i915->drm.struct_mutex);
>> return 0;
>> }
>> @@ -1490,19 +1497,19 @@ static void describe_ctx_ring(struct seq_file
>> *m, struct intel_ring *ring)
>> static int i915_context_status(struct seq_file *m, void *unused)
>> {
>> - struct drm_i915_private *dev_priv = node_to_i915(m->private);
>> - struct drm_device *dev = &dev_priv->drm;
>> - struct i915_gem_context *ctx;
>> - int ret;
>> -
>> - ret = mutex_lock_interruptible(&dev->struct_mutex);
>> - if (ret)
>> - return ret;
>> + struct drm_i915_private *i915 = node_to_i915(m->private);
>> + struct i915_gem_context *ctx, *cn;
>> - list_for_each_entry(ctx, &dev_priv->contexts.list, link) {
>> + spin_lock(&i915->gem.contexts.lock);
>> + list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) {
>> struct i915_gem_engines_iter it;
>> struct intel_context *ce;
>> + if (!kref_get_unless_zero(&ctx->ref))
>> + continue;
>> +
>> + spin_unlock(&i915->gem.contexts.lock);
>> +
>> seq_puts(m, "HW context ");
>> if (ctx->pid) {
>> struct task_struct *task;
>> @@ -1537,9 +1544,12 @@ static int i915_context_status(struct seq_file
>> *m, void *unused)
>> i915_gem_context_unlock_engines(ctx);
>> seq_putc(m, '\n');
>> - }
>> - mutex_unlock(&dev->struct_mutex);
>> + spin_lock(&i915->gem.contexts.lock);
>> + list_safe_reset_next(ctx, cn, link);
>> + i915_gem_context_put(ctx);
>> + }
>> + spin_unlock(&i915->gem.contexts.lock);
>> return 0;
>> }
>> diff --git a/drivers/gpu/drm/i915/i915_drv.c
>> b/drivers/gpu/drm/i915/i915_drv.c
>> index 8cac6124d16a..a8ddac22d633 100644
>> --- a/drivers/gpu/drm/i915/i915_drv.c
>> +++ b/drivers/gpu/drm/i915/i915_drv.c
>> @@ -1685,10 +1685,8 @@ static void i915_driver_postclose(struct
>> drm_device *dev, struct drm_file *file)
>> {
>> struct drm_i915_file_private *file_priv = file->driver_priv;
>> - mutex_lock(&dev->struct_mutex);
>> i915_gem_context_close(file);
>> i915_gem_release(dev, file);
>> - mutex_unlock(&dev->struct_mutex);
>> kfree_rcu(file_priv, rcu);
>> diff --git a/drivers/gpu/drm/i915/i915_drv.h
>> b/drivers/gpu/drm/i915/i915_drv.h
>> index ad7ed7f72da8..23f311172e75 100644
>> --- a/drivers/gpu/drm/i915/i915_drv.h
>> +++ b/drivers/gpu/drm/i915/i915_drv.h
>> @@ -1543,25 +1543,6 @@ struct drm_i915_private {
>> int audio_power_refcount;
>> u32 audio_freq_cntrl;
>> - struct {
>> - struct mutex mutex;
>> - struct list_head list;
>> - struct llist_head free_list;
>> - struct work_struct free_work;
>> -
>> - /* The hw wants to have a stable context identifier for the
>> - * lifetime of the context (for OA, PASID, faults, etc).
>> - * This is limited in execlists to 21 bits.
>> - */
>> - struct ida hw_ida;
>> -#define MAX_CONTEXT_HW_ID (1<<21) /* exclusive */
>> -#define MAX_GUC_CONTEXT_HW_ID (1 << 20) /* exclusive */
>> -#define GEN11_MAX_CONTEXT_HW_ID (1<<11) /* exclusive */
>> -/* in Gen12 ID 0x7FF is reserved to indicate idle */
>> -#define GEN12_MAX_CONTEXT_HW_ID (GEN11_MAX_CONTEXT_HW_ID - 1)
>> - struct list_head hw_id_list;
>
> These bits.
>
>> - } contexts;
>> -
>> u32 fdi_rx_config;
>> /* Shadow for DISPLAY_PHY_CONTROL which can't be safely read */
>> @@ -1717,6 +1698,14 @@ struct drm_i915_private {
>> struct {
>> struct notifier_block pm_notifier;
>> +
>> + struct i915_gem_contexts {
>> + spinlock_t lock; /* locks list */
>> + struct list_head list;
>> +
>> + struct llist_head free_list;
>> + struct work_struct free_work;
>> + } contexts;
>> } gem;
>> /* For i945gm vblank irq vs. C3 workaround */
>> diff --git a/drivers/gpu/drm/i915/i915_gem.c
>> b/drivers/gpu/drm/i915/i915_gem.c
>> index 9c2877441540..8814db7023db 100644
>> --- a/drivers/gpu/drm/i915/i915_gem.c
>> +++ b/drivers/gpu/drm/i915/i915_gem.c
>> @@ -1263,7 +1263,7 @@ int i915_gem_init(struct drm_i915_private
>> *dev_priv)
>> goto err_unlock;
>> }
>> - ret = i915_gem_contexts_init(dev_priv);
>> + ret = i915_gem_init_contexts(dev_priv);
>> if (ret) {
>> GEM_BUG_ON(ret == -EIO);
>> goto err_scratch;
>> @@ -1345,7 +1345,7 @@ int i915_gem_init(struct drm_i915_private
>> *dev_priv)
>> }
>> err_context:
>> if (ret != -EIO)
>> - i915_gem_contexts_fini(dev_priv);
>> + i915_gem_driver_release__contexts(dev_priv);
>> err_scratch:
>> intel_gt_driver_release(&dev_priv->gt);
>> err_unlock:
>> @@ -1413,11 +1413,9 @@ void i915_gem_driver_remove(struct
>> drm_i915_private *dev_priv)
>> void i915_gem_driver_release(struct drm_i915_private *dev_priv)
>> {
>> - mutex_lock(&dev_priv->drm.struct_mutex);
>> intel_engines_cleanup(dev_priv);
>> - i915_gem_contexts_fini(dev_priv);
>> + i915_gem_driver_release__contexts(dev_priv);
>> intel_gt_driver_release(&dev_priv->gt);
>> - mutex_unlock(&dev_priv->drm.struct_mutex);
>> intel_wa_list_free(&dev_priv->gt_wa_list);
>> @@ -1427,7 +1425,7 @@ void i915_gem_driver_release(struct
>> drm_i915_private *dev_priv)
>> i915_gem_drain_freed_objects(dev_priv);
>> - WARN_ON(!list_empty(&dev_priv->contexts.list));
>> + WARN_ON(!list_empty(&dev_priv->gem.contexts.list));
>> }
>> void i915_gem_init_mmio(struct drm_i915_private *i915)
>> diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c
>> b/drivers/gpu/drm/i915/i915_gem_gtt.c
>> index 0d040517787b..db32e0fe78b1 100644
>> --- a/drivers/gpu/drm/i915/i915_gem_gtt.c
>> +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
>> @@ -1366,7 +1366,9 @@ static int gen8_init_scratch(struct
>> i915_address_space *vm)
>> if (vm->has_read_only &&
>> vm->i915->kernel_context &&
>> vm->i915->kernel_context->vm) {
>> - struct i915_address_space *clone = vm->i915->kernel_context->vm;
>> + struct i915_address_space *clone =
>> + rcu_dereference_protected(vm->i915->kernel_context->vm,
>> + true);
>
> What is the protection here?
>
>> GEM_BUG_ON(!clone->has_read_only);
>> diff --git a/drivers/gpu/drm/i915/i915_perf.c
>> b/drivers/gpu/drm/i915/i915_perf.c
>> index d36ba248d438..bfe6e9b6d71d 100644
>> --- a/drivers/gpu/drm/i915/i915_perf.c
>> +++ b/drivers/gpu/drm/i915/i915_perf.c
>> @@ -1853,8 +1853,8 @@ static int gen8_configure_all_contexts(struct
>> i915_perf_stream *stream,
>> };
>> #undef ctx_flexeuN
>> struct intel_engine_cs *engine;
>> - struct i915_gem_context *ctx;
>> - int i;
>> + struct i915_gem_context *ctx, *cn;
>> + int i, err;
>> for (i = 2; i < ARRAY_SIZE(regs); i++)
>> regs[i].value = oa_config_flex_reg(oa_config, regs[i].reg);
>> @@ -1877,16 +1877,27 @@ static int gen8_configure_all_contexts(struct
>> i915_perf_stream *stream,
>> * context. Contexts idle at the time of reconfiguration are not
>> * trapped behind the barrier.
>> */
>> - list_for_each_entry(ctx, &i915->contexts.list, link) {
>> - int err;
>> -
>> + spin_lock(&i915->gem.contexts.lock);
>> + list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) {
>> if (ctx == i915->kernel_context)
>> continue;
>> + if (!kref_get_unless_zero(&ctx->ref))
>> + continue;
>> +
>> + spin_unlock(&i915->gem.contexts.lock);
>> +
>> err = gen8_configure_context(ctx, regs, ARRAY_SIZE(regs));
>> - if (err)
>> + if (err) {
>> + i915_gem_context_put(ctx);
>> return err;
>> + }
>> +
>> + spin_lock(&i915->gem.contexts.lock);
>> + list_safe_reset_next(ctx, cn, link);
>> + i915_gem_context_put(ctx);
>> }
>> + spin_unlock(&i915->gem.contexts.lock);
>> /*
>> * After updating all other contexts, we need to modify ourselves.
>> @@ -1895,7 +1906,6 @@ static int gen8_configure_all_contexts(struct
>> i915_perf_stream *stream,
>> */
>> for_each_uabi_engine(engine, i915) {
>> struct intel_context *ce = engine->kernel_context;
>> - int err;
>> if (engine->class != RENDER_CLASS)
>> continue;
>> diff --git a/drivers/gpu/drm/i915/i915_sysfs.c
>> b/drivers/gpu/drm/i915/i915_sysfs.c
>> index 6b88d934927a..585808c2ee84 100644
>> --- a/drivers/gpu/drm/i915/i915_sysfs.c
>> +++ b/drivers/gpu/drm/i915/i915_sysfs.c
>> @@ -144,9 +144,9 @@ static const struct attribute_group
>> media_rc6_attr_group = {
>> };
>> #endif
>> -static int l3_access_valid(struct drm_i915_private *dev_priv, loff_t
>> offset)
>> +static int l3_access_valid(struct drm_i915_private *i915, loff_t offset)
>> {
>> - if (!HAS_L3_DPF(dev_priv))
>> + if (!HAS_L3_DPF(i915))
>> return -EPERM;
>> if (offset % 4 != 0)
>> @@ -164,31 +164,24 @@ i915_l3_read(struct file *filp, struct kobject
>> *kobj,
>> loff_t offset, size_t count)
>> {
>> struct device *kdev = kobj_to_dev(kobj);
>> - struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev);
>> - struct drm_device *dev = &dev_priv->drm;
>> + struct drm_i915_private *i915 = kdev_minor_to_i915(kdev);
>> int slice = (int)(uintptr_t)attr->private;
>> int ret;
>> - count = round_down(count, 4);
>> -
>> - ret = l3_access_valid(dev_priv, offset);
>> + ret = l3_access_valid(i915, offset);
>> if (ret)
>> return ret;
>> + count = round_down(count, 4);
>> count = min_t(size_t, GEN7_L3LOG_SIZE - offset, count);
>> + memset(buf, 0, count);
>> - ret = i915_mutex_lock_interruptible(dev);
>> - if (ret)
>> - return ret;
>> -
>> - if (dev_priv->l3_parity.remap_info[slice])
>> + spin_lock(&i915->gem.contexts.lock);
>> + if (i915->l3_parity.remap_info[slice])
>> memcpy(buf,
>> - dev_priv->l3_parity.remap_info[slice] + (offset/4),
>> + i915->l3_parity.remap_info[slice] + offset / 4,
>> count);
>> - else
>> - memset(buf, 0, count);
>> -
>> - mutex_unlock(&dev->struct_mutex);
>> + spin_unlock(&i915->gem.contexts.lock);
>> return count;
>> }
>> @@ -199,46 +192,46 @@ i915_l3_write(struct file *filp, struct kobject
>> *kobj,
>> loff_t offset, size_t count)
>> {
>> struct device *kdev = kobj_to_dev(kobj);
>> - struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev);
>> - struct drm_device *dev = &dev_priv->drm;
>> + struct drm_i915_private *i915 = kdev_minor_to_i915(kdev);
>> struct i915_gem_context *ctx;
>> int slice = (int)(uintptr_t)attr->private;
>> - u32 **remap_info;
>> + u32 *remap_info, *freeme = NULL;
>> int ret;
>> - ret = l3_access_valid(dev_priv, offset);
>> + ret = l3_access_valid(i915, offset);
>> if (ret)
>> return ret;
>> - ret = i915_mutex_lock_interruptible(dev);
>> - if (ret)
>> - return ret;
>> + remap_info = kzalloc(GEN7_L3LOG_SIZE, GFP_KERNEL);
>> + if (!remap_info)
>> + return -ENOMEM;
>> - remap_info = &dev_priv->l3_parity.remap_info[slice];
>> - if (!*remap_info) {
>> - *remap_info = kzalloc(GEN7_L3LOG_SIZE, GFP_KERNEL);
>> - if (!*remap_info) {
>> - ret = -ENOMEM;
>> - goto out;
>> - }
>> + spin_lock(&i915->gem.contexts.lock);
>> +
>> + if (i915->l3_parity.remap_info[slice]) {
>> + freeme = remap_info;
>> + remap_info = i915->l3_parity.remap_info[slice];
>> + } else {
>> + i915->l3_parity.remap_info[slice] = remap_info;
>> }
>> - /* TODO: Ideally we really want a GPU reset here to make sure errors
>> - * aren't propagated. Since I cannot find a stable way to reset
>> the GPU
>> - * at this point it is left as a TODO.
>> - */
>> - memcpy(*remap_info + (offset/4), buf, count);
>> + count = round_down(count, 4);
>
> Where was this round_down before, can't see it!
>
>> + memcpy(remap_info + offset / 4, buf, count);
>> /* NB: We defer the remapping until we switch to the context */
>> - list_for_each_entry(ctx, &dev_priv->contexts.list, link)
>> - ctx->remap_slice |= (1<<slice);
>> + list_for_each_entry(ctx, &i915->gem.contexts.list, link)
>> + ctx->remap_slice |= BIT(slice);
>> - ret = count;
>> + spin_unlock(&i915->gem.contexts.lock);
>> + kfree(freeme);
>> -out:
>> - mutex_unlock(&dev->struct_mutex);
>> + /*
>> + * TODO: Ideally we really want a GPU reset here to make sure errors
>> + * aren't propagated. Since I cannot find a stable way to reset
>> the GPU
>> + * at this point it is left as a TODO.
>> + */
>> - return ret;
>> + return count;
>> }
>> static const struct bin_attribute dpf_attrs = {
>> diff --git a/drivers/gpu/drm/i915/i915_trace.h
>> b/drivers/gpu/drm/i915/i915_trace.h
>> index 1f2cf6cfafb5..7ef7a1e1664c 100644
>> --- a/drivers/gpu/drm/i915/i915_trace.h
>> +++ b/drivers/gpu/drm/i915/i915_trace.h
>> @@ -952,7 +952,7 @@ DECLARE_EVENT_CLASS(i915_context,
>> TP_fast_assign(
>> __entry->dev = ctx->i915->drm.primary->index;
>> __entry->ctx = ctx;
>> - __entry->vm = ctx->vm;
>> + __entry->vm = rcu_access_pointer(ctx->vm);
>> ),
>> TP_printk("dev=%u, ctx=%p, ctx_vm=%p",
>> diff --git a/drivers/gpu/drm/i915/selftests/i915_gem.c
>> b/drivers/gpu/drm/i915/selftests/i915_gem.c
>> index 0346c3e5b6b6..bfa40a5b6d98 100644
>> --- a/drivers/gpu/drm/i915/selftests/i915_gem.c
>> +++ b/drivers/gpu/drm/i915/selftests/i915_gem.c
>> @@ -138,11 +138,9 @@ static int igt_gem_suspend(void *arg)
>> return PTR_ERR(file);
>> err = -ENOMEM;
>> - mutex_lock(&i915->drm.struct_mutex);
>> ctx = live_context(i915, file);
>> if (!IS_ERR(ctx))
>> err = switch_to_context(i915, ctx);
>> - mutex_unlock(&i915->drm.struct_mutex);
>> if (err)
>> goto out;
>> @@ -157,9 +155,7 @@ static int igt_gem_suspend(void *arg)
>> pm_resume(i915);
>> - mutex_lock(&i915->drm.struct_mutex);
>> err = switch_to_context(i915, ctx);
>> - mutex_unlock(&i915->drm.struct_mutex);
>> out:
>> mock_file_free(i915, file);
>> return err;
>> @@ -177,11 +173,9 @@ static int igt_gem_hibernate(void *arg)
>> return PTR_ERR(file);
>> err = -ENOMEM;
>> - mutex_lock(&i915->drm.struct_mutex);
>> ctx = live_context(i915, file);
>> if (!IS_ERR(ctx))
>> err = switch_to_context(i915, ctx);
>> - mutex_unlock(&i915->drm.struct_mutex);
>> if (err)
>> goto out;
>> @@ -196,9 +190,7 @@ static int igt_gem_hibernate(void *arg)
>> pm_resume(i915);
>> - mutex_lock(&i915->drm.struct_mutex);
>> err = switch_to_context(i915, ctx);
>> - mutex_unlock(&i915->drm.struct_mutex);
>> out:
>> mock_file_free(i915, file);
>> return err;
>> diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
>> b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
>> index f39f0282e78c..0af9a58d011d 100644
>> --- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
>> +++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
>> @@ -473,7 +473,6 @@ static int igt_evict_contexts(void *arg)
>> }
>> count = 0;
>> - mutex_lock(&i915->drm.struct_mutex);
>> onstack_fence_init(&fence);
>> do {
>> struct i915_request *rq;
>> @@ -510,8 +509,6 @@ static int igt_evict_contexts(void *arg)
>> count++;
>> err = 0;
>> } while(1);
>> - mutex_unlock(&i915->drm.struct_mutex);
>> -
>> onstack_fence_fini(&fence);
>> pr_info("Submitted %lu contexts/requests on %s\n",
>> count, engine->name);
>> diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
>> b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
>> index 9939dd40c6a8..6777a335bd44 100644
>> --- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
>> +++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
>> @@ -1246,6 +1246,7 @@ static int exercise_mock(struct drm_i915_private
>> *i915,
>> unsigned long end_time))
>> {
>> const u64 limit = totalram_pages() << PAGE_SHIFT;
>> + struct i915_address_space *vm;
>> struct i915_gem_context *ctx;
>> IGT_TIMEOUT(end_time);
>> int err;
>> @@ -1254,7 +1255,8 @@ static int exercise_mock(struct drm_i915_private
>> *i915,
>> if (!ctx)
>> return -ENOMEM;
>> - err = func(i915, ctx->vm, 0, min(ctx->vm->total, limit), end_time);
>> + vm = rcu_dereference_protected(ctx->vm, true);
>> + err = func(i915, vm, 0, min(vm->total, limit), end_time);
>> mock_context_close(ctx);
>> return err;
>> diff --git a/drivers/gpu/drm/i915/selftests/i915_request.c
>> b/drivers/gpu/drm/i915/selftests/i915_request.c
>> index d1ee384a3260..4abf52733520 100644
>> --- a/drivers/gpu/drm/i915/selftests/i915_request.c
>> +++ b/drivers/gpu/drm/i915/selftests/i915_request.c
>> @@ -181,9 +181,7 @@ static int igt_request_rewind(void *arg)
>> struct intel_context *ce;
>> int err = -EINVAL;
>> - mutex_lock(&i915->drm.struct_mutex);
>> ctx[0] = mock_context(i915, "A");
>> - mutex_unlock(&i915->drm.struct_mutex);
>> ce = i915_gem_context_get_engine(ctx[0], RCS0);
>> GEM_BUG_ON(IS_ERR(ce));
>> @@ -197,9 +195,7 @@ static int igt_request_rewind(void *arg)
>> i915_request_get(request);
>> i915_request_add(request);
>> - mutex_lock(&i915->drm.struct_mutex);
>> ctx[1] = mock_context(i915, "B");
>> - mutex_unlock(&i915->drm.struct_mutex);
>> ce = i915_gem_context_get_engine(ctx[1], RCS0);
>> GEM_BUG_ON(IS_ERR(ce));
>> @@ -438,9 +434,7 @@ static int mock_breadcrumbs_smoketest(void *arg)
>> }
>> for (n = 0; n < t.ncontexts; n++) {
>> - mutex_lock(&t.engine->i915->drm.struct_mutex);
>> t.contexts[n] = mock_context(t.engine->i915, "mock");
>> - mutex_unlock(&t.engine->i915->drm.struct_mutex);
>> if (!t.contexts[n]) {
>> ret = -ENOMEM;
>> goto out_contexts;
>> @@ -734,7 +728,8 @@ static int live_empty_request(void *arg)
>> static struct i915_vma *recursive_batch(struct drm_i915_private *i915)
>> {
>> struct i915_gem_context *ctx = i915->kernel_context;
>> - struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm;
>> + struct i915_address_space *vm =
>> + rcu_dereference_protected(ctx->vm, true) ?: &i915->ggtt.vm;
>> struct drm_i915_gem_object *obj;
>> const int gen = INTEL_GEN(i915);
>> struct i915_vma *vma;
>> @@ -1109,9 +1104,7 @@ static int live_breadcrumbs_smoketest(void *arg)
>> }
>> for (n = 0; n < t[0].ncontexts; n++) {
>> - mutex_lock(&i915->drm.struct_mutex);
>> t[0].contexts[n] = live_context(i915, file);
>> - mutex_unlock(&i915->drm.struct_mutex);
>> if (!t[0].contexts[n]) {
>> ret = -ENOMEM;
>> goto out_contexts;
>> diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c
>> b/drivers/gpu/drm/i915/selftests/i915_vma.c
>> index ac1ff558eb90..f6e6b0aae38c 100644
>> --- a/drivers/gpu/drm/i915/selftests/i915_vma.c
>> +++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
>> @@ -38,7 +38,7 @@ static bool assert_vma(struct i915_vma *vma,
>> {
>> bool ok = true;
>> - if (vma->vm != ctx->vm) {
>> + if (vma->vm != rcu_access_pointer(ctx->vm)) {
>> pr_err("VMA created with wrong VM\n");
>> ok = false;
>> }
>> @@ -113,7 +113,8 @@ static int create_vmas(struct drm_i915_private *i915,
>> list_for_each_entry(obj, objects, st_link) {
>> for (pinned = 0; pinned <= 1; pinned++) {
>> list_for_each_entry(ctx, contexts, link) {
>> - struct i915_address_space *vm = ctx->vm;
>> + struct i915_address_space *vm =
>> + rcu_dereference_protected(ctx->vm, true);
>> struct i915_vma *vma;
>> int err;
>> diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
>> b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
>> index efe40c29fcf2..0ed21a7b8682 100644
>> --- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
>> +++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
>> @@ -59,11 +59,9 @@ static void mock_device_release(struct drm_device
>> *dev)
>> i915_gem_drain_workqueue(i915);
>> - mutex_lock(&i915->drm.struct_mutex);
>> for_each_engine(engine, i915, id)
>> mock_engine_free(engine);
>> - i915_gem_contexts_fini(i915);
>> - mutex_unlock(&i915->drm.struct_mutex);
>> + i915_gem_driver_release__contexts(i915);
>> intel_timelines_fini(i915);
>> @@ -207,7 +205,7 @@ struct drm_i915_private *mock_gem_device(void)
>> return i915;
>> err_context:
>> - i915_gem_contexts_fini(i915);
>> + i915_gem_driver_release__contexts(i915);
>> err_engine:
>> mock_engine_free(i915->engine[RCS0]);
>> err_unlock:
>>
>
> Regards,
>
> Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* Re: [PATCH 21/27] drm/i915: Move context management under GEM
2019-09-26 13:57 ` Tvrtko Ursulin
2019-10-02 16:09 ` Tvrtko Ursulin
@ 2019-10-03 7:35 ` Chris Wilson
1 sibling, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-10-03 7:35 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Tvrtko Ursulin (2019-09-26 14:57:24)
>
> On 25/09/2019 11:01, Chris Wilson wrote:
> > void i915_gem_context_release(struct kref *ref)
> > {
> > struct i915_gem_context *ctx = container_of(ref, typeof(*ctx), ref);
> > - struct drm_i915_private *i915 = ctx->i915;
> > + struct i915_gem_contexts *gc = &ctx->i915->gem.contexts;
> >
> > trace_i915_context_free(ctx);
> > - if (llist_add(&ctx->free_link, &i915->contexts.free_list))
> > - queue_work(i915->wq, &i915->contexts.free_work);
> > + if (llist_add(&ctx->free_link, &gc->free_list))
> > + queue_work(ctx->i915->wq, &gc->free_work);
>
> At first I thought gc was some sort of garbage collection list. But it
> is a GEM contexts list. :) This hunk looks completely avoidable, I don't
> see it brings much benefit on balance since in turn it adds ctx->i915
> twice . But as you wish..
Since it's not taking struct_mutex it doesn't need i915->wq any more,
good point.
> > @@ -464,11 +454,11 @@ static void __apply_ppgtt(struct intel_context *ce, void *vm)
> > static struct i915_address_space *
> > __set_ppgtt(struct i915_gem_context *ctx, struct i915_address_space *vm)
> > {
> > - struct i915_address_space *old = ctx->vm;
> > + struct i915_address_space *old = rcu_dereference_protected(ctx->vm, 1);
> >
> > GEM_BUG_ON(old && i915_vm_is_4lvl(vm) != i915_vm_is_4lvl(old));
> >
> > - ctx->vm = i915_vm_open(vm);
> > + rcu_assign_pointer(ctx->vm, i915_vm_open(vm));
>
> Are updaters here serialized? Lost track if struct mutex is still held
> in set_ppgtt at this point in the series or not.. Thinking of "old =
> rcu_dereference_protected.." and here.
Yes, ctx->mutex; I was being very lazy in not marking it up; especially
in the selftests where it was so much more dontcare.
> > @@ -653,8 +633,8 @@ static int gem_context_register(struct i915_gem_context *ctx,
> > int ret;
> >
> > ctx->file_priv = fpriv;
> > - if (ctx->vm)
> > - ctx->vm->file = fpriv;
> > + if (rcu_access_pointer(ctx->vm))
> > + rcu_dereference_protected(ctx->vm, true)->file = fpriv;
>
> Here rcu accesses are just to satisfy sparse?
Total eyesore, yup. I'm thinking I just declare ctx->vm to be protected
by ctx->mutex and make it so.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* [PATCH 22/27] drm/i915/overlay: Drop struct_mutex guard
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (20 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 21/27] drm/i915: Move context management under GEM Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 12:43 ` Tvrtko Ursulin
2019-09-25 10:01 ` [PATCH 23/27] drm/i915: Drop struct_mutex guard from debugfs/framebuffer_info Chris Wilson
` (8 subsequent siblings)
30 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
The overlay uses the modeset mutex to control itself and only required
the struct_mutex for requests, which is now obsolete.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/display/intel_display.c | 7 +------
drivers/gpu/drm/i915/display/intel_overlay.c | 13 -------------
drivers/gpu/drm/i915/gt/intel_reset.c | 4 ----
3 files changed, 1 insertion(+), 23 deletions(-)
diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
index 99ddbb7bedbc..038f14d9b5c9 100644
--- a/drivers/gpu/drm/i915/display/intel_display.c
+++ b/drivers/gpu/drm/i915/display/intel_display.c
@@ -5738,13 +5738,8 @@ void hsw_disable_ips(const struct intel_crtc_state *crtc_state)
static void intel_crtc_dpms_overlay_disable(struct intel_crtc *intel_crtc)
{
- if (intel_crtc->overlay) {
- struct drm_device *dev = intel_crtc->base.dev;
-
- mutex_lock(&dev->struct_mutex);
+ if (intel_crtc->overlay)
(void) intel_overlay_switch_off(intel_crtc->overlay);
- mutex_unlock(&dev->struct_mutex);
- }
/* Let userspace switch the overlay on again. In most cases userspace
* has to recompute where to put it anyway.
diff --git a/drivers/gpu/drm/i915/display/intel_overlay.c b/drivers/gpu/drm/i915/display/intel_overlay.c
index e12e1a753af0..daea112cbb87 100644
--- a/drivers/gpu/drm/i915/display/intel_overlay.c
+++ b/drivers/gpu/drm/i915/display/intel_overlay.c
@@ -439,8 +439,6 @@ static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
struct i915_request *rq;
u32 *cs;
- lockdep_assert_held(&dev_priv->drm.struct_mutex);
-
/*
* Only wait if there is actually an old frame to release to
* guarantee forward progress.
@@ -751,7 +749,6 @@ static int intel_overlay_do_put_image(struct intel_overlay *overlay,
struct i915_vma *vma;
int ret, tmp_width;
- lockdep_assert_held(&dev_priv->drm.struct_mutex);
WARN_ON(!drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex));
ret = intel_overlay_release_old_vid(overlay);
@@ -852,7 +849,6 @@ int intel_overlay_switch_off(struct intel_overlay *overlay)
struct drm_i915_private *dev_priv = overlay->i915;
int ret;
- lockdep_assert_held(&dev_priv->drm.struct_mutex);
WARN_ON(!drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex));
ret = intel_overlay_recover_from_interrupt(overlay);
@@ -1068,11 +1064,7 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
if (!(params->flags & I915_OVERLAY_ENABLE)) {
drm_modeset_lock_all(dev);
- mutex_lock(&dev->struct_mutex);
-
ret = intel_overlay_switch_off(overlay);
-
- mutex_unlock(&dev->struct_mutex);
drm_modeset_unlock_all(dev);
return ret;
@@ -1088,7 +1080,6 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
return -ENOENT;
drm_modeset_lock_all(dev);
- mutex_lock(&dev->struct_mutex);
if (i915_gem_object_is_tiled(new_bo)) {
DRM_DEBUG_KMS("buffer used for overlay image can not be tiled\n");
@@ -1152,14 +1143,12 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
if (ret != 0)
goto out_unlock;
- mutex_unlock(&dev->struct_mutex);
drm_modeset_unlock_all(dev);
i915_gem_object_put(new_bo);
return 0;
out_unlock:
- mutex_unlock(&dev->struct_mutex);
drm_modeset_unlock_all(dev);
i915_gem_object_put(new_bo);
@@ -1233,7 +1222,6 @@ int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data,
}
drm_modeset_lock_all(dev);
- mutex_lock(&dev->struct_mutex);
ret = -EINVAL;
if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
@@ -1290,7 +1278,6 @@ int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data,
ret = 0;
out_unlock:
- mutex_unlock(&dev->struct_mutex);
drm_modeset_unlock_all(dev);
return ret;
diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c
index 2685c0103e90..ee0c5c24282c 100644
--- a/drivers/gpu/drm/i915/gt/intel_reset.c
+++ b/drivers/gpu/drm/i915/gt/intel_reset.c
@@ -1265,10 +1265,6 @@ int intel_gt_terminally_wedged(struct intel_gt *gt)
if (!test_bit(I915_RESET_BACKOFF, >->reset.flags))
return -EIO;
- /* XXX intel_reset_finish() still takes struct_mutex!!! */
- if (mutex_is_locked(>->i915->drm.struct_mutex))
- return -EAGAIN;
-
if (wait_event_interruptible(gt->reset.queue,
!test_bit(I915_RESET_BACKOFF,
>->reset.flags)))
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* Re: [PATCH 22/27] drm/i915/overlay: Drop struct_mutex guard
2019-09-25 10:01 ` [PATCH 22/27] drm/i915/overlay: Drop struct_mutex guard Chris Wilson
@ 2019-09-25 12:43 ` Tvrtko Ursulin
2019-09-25 12:53 ` Chris Wilson
0 siblings, 1 reply; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-25 12:43 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 25/09/2019 11:01, Chris Wilson wrote:
> The overlay uses the modeset mutex to control itself and only required
> the struct_mutex for requests, which is now obsolete.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/display/intel_display.c | 7 +------
> drivers/gpu/drm/i915/display/intel_overlay.c | 13 -------------
> drivers/gpu/drm/i915/gt/intel_reset.c | 4 ----
> 3 files changed, 1 insertion(+), 23 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
> index 99ddbb7bedbc..038f14d9b5c9 100644
> --- a/drivers/gpu/drm/i915/display/intel_display.c
> +++ b/drivers/gpu/drm/i915/display/intel_display.c
> @@ -5738,13 +5738,8 @@ void hsw_disable_ips(const struct intel_crtc_state *crtc_state)
>
> static void intel_crtc_dpms_overlay_disable(struct intel_crtc *intel_crtc)
> {
> - if (intel_crtc->overlay) {
> - struct drm_device *dev = intel_crtc->base.dev;
> -
> - mutex_lock(&dev->struct_mutex);
> + if (intel_crtc->overlay)
> (void) intel_overlay_switch_off(intel_crtc->overlay);
> - mutex_unlock(&dev->struct_mutex);
> - }
>
> /* Let userspace switch the overlay on again. In most cases userspace
> * has to recompute where to put it anyway.
> diff --git a/drivers/gpu/drm/i915/display/intel_overlay.c b/drivers/gpu/drm/i915/display/intel_overlay.c
> index e12e1a753af0..daea112cbb87 100644
> --- a/drivers/gpu/drm/i915/display/intel_overlay.c
> +++ b/drivers/gpu/drm/i915/display/intel_overlay.c
> @@ -439,8 +439,6 @@ static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
> struct i915_request *rq;
> u32 *cs;
>
> - lockdep_assert_held(&dev_priv->drm.struct_mutex);
> -
> /*
> * Only wait if there is actually an old frame to release to
> * guarantee forward progress.
> @@ -751,7 +749,6 @@ static int intel_overlay_do_put_image(struct intel_overlay *overlay,
> struct i915_vma *vma;
> int ret, tmp_width;
>
> - lockdep_assert_held(&dev_priv->drm.struct_mutex);
> WARN_ON(!drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex));
>
> ret = intel_overlay_release_old_vid(overlay);
> @@ -852,7 +849,6 @@ int intel_overlay_switch_off(struct intel_overlay *overlay)
> struct drm_i915_private *dev_priv = overlay->i915;
> int ret;
>
> - lockdep_assert_held(&dev_priv->drm.struct_mutex);
> WARN_ON(!drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex));
>
> ret = intel_overlay_recover_from_interrupt(overlay);
> @@ -1068,11 +1064,7 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
>
> if (!(params->flags & I915_OVERLAY_ENABLE)) {
> drm_modeset_lock_all(dev);
> - mutex_lock(&dev->struct_mutex);
> -
> ret = intel_overlay_switch_off(overlay);
> -
> - mutex_unlock(&dev->struct_mutex);
> drm_modeset_unlock_all(dev);
>
> return ret;
> @@ -1088,7 +1080,6 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
> return -ENOENT;
>
> drm_modeset_lock_all(dev);
> - mutex_lock(&dev->struct_mutex);
>
> if (i915_gem_object_is_tiled(new_bo)) {
> DRM_DEBUG_KMS("buffer used for overlay image can not be tiled\n");
> @@ -1152,14 +1143,12 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
> if (ret != 0)
> goto out_unlock;
>
> - mutex_unlock(&dev->struct_mutex);
> drm_modeset_unlock_all(dev);
> i915_gem_object_put(new_bo);
>
> return 0;
>
> out_unlock:
> - mutex_unlock(&dev->struct_mutex);
> drm_modeset_unlock_all(dev);
> i915_gem_object_put(new_bo);
>
> @@ -1233,7 +1222,6 @@ int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data,
> }
>
> drm_modeset_lock_all(dev);
> - mutex_lock(&dev->struct_mutex);
>
> ret = -EINVAL;
> if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
> @@ -1290,7 +1278,6 @@ int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data,
>
> ret = 0;
> out_unlock:
> - mutex_unlock(&dev->struct_mutex);
> drm_modeset_unlock_all(dev);
>
> return ret;
> diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c
> index 2685c0103e90..ee0c5c24282c 100644
> --- a/drivers/gpu/drm/i915/gt/intel_reset.c
> +++ b/drivers/gpu/drm/i915/gt/intel_reset.c
> @@ -1265,10 +1265,6 @@ int intel_gt_terminally_wedged(struct intel_gt *gt)
> if (!test_bit(I915_RESET_BACKOFF, >->reset.flags))
> return -EIO;
>
> - /* XXX intel_reset_finish() still takes struct_mutex!!! */
> - if (mutex_is_locked(>->i915->drm.struct_mutex))
> - return -EAGAIN;
> -
What is this hunk doing here?
Regards,
Tvrtko
> if (wait_event_interruptible(gt->reset.queue,
> !test_bit(I915_RESET_BACKOFF,
> >->reset.flags)))
>
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* Re: [PATCH 22/27] drm/i915/overlay: Drop struct_mutex guard
2019-09-25 12:43 ` Tvrtko Ursulin
@ 2019-09-25 12:53 ` Chris Wilson
2019-09-25 13:01 ` Tvrtko Ursulin
0 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 12:53 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Tvrtko Ursulin (2019-09-25 13:43:46)
>
> On 25/09/2019 11:01, Chris Wilson wrote:
> > The overlay uses the modeset mutex to control itself and only required
> > the struct_mutex for requests, which is now obsolete.
> >
> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> > ---
> > drivers/gpu/drm/i915/display/intel_display.c | 7 +------
> > drivers/gpu/drm/i915/display/intel_overlay.c | 13 -------------
> > drivers/gpu/drm/i915/gt/intel_reset.c | 4 ----
> > 3 files changed, 1 insertion(+), 23 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
> > index 99ddbb7bedbc..038f14d9b5c9 100644
> > --- a/drivers/gpu/drm/i915/display/intel_display.c
> > +++ b/drivers/gpu/drm/i915/display/intel_display.c
> > @@ -5738,13 +5738,8 @@ void hsw_disable_ips(const struct intel_crtc_state *crtc_state)
> >
> > static void intel_crtc_dpms_overlay_disable(struct intel_crtc *intel_crtc)
> > {
> > - if (intel_crtc->overlay) {
> > - struct drm_device *dev = intel_crtc->base.dev;
> > -
> > - mutex_lock(&dev->struct_mutex);
> > + if (intel_crtc->overlay)
> > (void) intel_overlay_switch_off(intel_crtc->overlay);
> > - mutex_unlock(&dev->struct_mutex);
> > - }
> >
> > /* Let userspace switch the overlay on again. In most cases userspace
> > * has to recompute where to put it anyway.
> > diff --git a/drivers/gpu/drm/i915/display/intel_overlay.c b/drivers/gpu/drm/i915/display/intel_overlay.c
> > index e12e1a753af0..daea112cbb87 100644
> > --- a/drivers/gpu/drm/i915/display/intel_overlay.c
> > +++ b/drivers/gpu/drm/i915/display/intel_overlay.c
> > @@ -439,8 +439,6 @@ static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
> > struct i915_request *rq;
> > u32 *cs;
> >
> > - lockdep_assert_held(&dev_priv->drm.struct_mutex);
> > -
> > /*
> > * Only wait if there is actually an old frame to release to
> > * guarantee forward progress.
> > @@ -751,7 +749,6 @@ static int intel_overlay_do_put_image(struct intel_overlay *overlay,
> > struct i915_vma *vma;
> > int ret, tmp_width;
> >
> > - lockdep_assert_held(&dev_priv->drm.struct_mutex);
> > WARN_ON(!drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex));
> >
> > ret = intel_overlay_release_old_vid(overlay);
> > @@ -852,7 +849,6 @@ int intel_overlay_switch_off(struct intel_overlay *overlay)
> > struct drm_i915_private *dev_priv = overlay->i915;
> > int ret;
> >
> > - lockdep_assert_held(&dev_priv->drm.struct_mutex);
> > WARN_ON(!drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex));
> >
> > ret = intel_overlay_recover_from_interrupt(overlay);
> > @@ -1068,11 +1064,7 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
> >
> > if (!(params->flags & I915_OVERLAY_ENABLE)) {
> > drm_modeset_lock_all(dev);
> > - mutex_lock(&dev->struct_mutex);
> > -
> > ret = intel_overlay_switch_off(overlay);
> > -
> > - mutex_unlock(&dev->struct_mutex);
> > drm_modeset_unlock_all(dev);
> >
> > return ret;
> > @@ -1088,7 +1080,6 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
> > return -ENOENT;
> >
> > drm_modeset_lock_all(dev);
> > - mutex_lock(&dev->struct_mutex);
> >
> > if (i915_gem_object_is_tiled(new_bo)) {
> > DRM_DEBUG_KMS("buffer used for overlay image can not be tiled\n");
> > @@ -1152,14 +1143,12 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
> > if (ret != 0)
> > goto out_unlock;
> >
> > - mutex_unlock(&dev->struct_mutex);
> > drm_modeset_unlock_all(dev);
> > i915_gem_object_put(new_bo);
> >
> > return 0;
> >
> > out_unlock:
> > - mutex_unlock(&dev->struct_mutex);
> > drm_modeset_unlock_all(dev);
> > i915_gem_object_put(new_bo);
> >
> > @@ -1233,7 +1222,6 @@ int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data,
> > }
> >
> > drm_modeset_lock_all(dev);
> > - mutex_lock(&dev->struct_mutex);
> >
> > ret = -EINVAL;
> > if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
> > @@ -1290,7 +1278,6 @@ int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data,
> >
> > ret = 0;
> > out_unlock:
> > - mutex_unlock(&dev->struct_mutex);
> > drm_modeset_unlock_all(dev);
> >
> > return ret;
> > diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c
> > index 2685c0103e90..ee0c5c24282c 100644
> > --- a/drivers/gpu/drm/i915/gt/intel_reset.c
> > +++ b/drivers/gpu/drm/i915/gt/intel_reset.c
> > @@ -1265,10 +1265,6 @@ int intel_gt_terminally_wedged(struct intel_gt *gt)
> > if (!test_bit(I915_RESET_BACKOFF, >->reset.flags))
> > return -EIO;
> >
> > - /* XXX intel_reset_finish() still takes struct_mutex!!! */
> > - if (mutex_is_locked(>->i915->drm.struct_mutex))
> > - return -EAGAIN;
> > -
>
> What is this hunk doing here?
Because the overlay used struct_mutex, now it does not.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* Re: [PATCH 22/27] drm/i915/overlay: Drop struct_mutex guard
2019-09-25 12:53 ` Chris Wilson
@ 2019-09-25 13:01 ` Tvrtko Ursulin
2019-09-25 13:11 ` Chris Wilson
0 siblings, 1 reply; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-25 13:01 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 25/09/2019 13:53, Chris Wilson wrote:
> Quoting Tvrtko Ursulin (2019-09-25 13:43:46)
>>
>> On 25/09/2019 11:01, Chris Wilson wrote:
>>> The overlay uses the modeset mutex to control itself and only required
>>> the struct_mutex for requests, which is now obsolete.
>>>
>>> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
>>> ---
>>> drivers/gpu/drm/i915/display/intel_display.c | 7 +------
>>> drivers/gpu/drm/i915/display/intel_overlay.c | 13 -------------
>>> drivers/gpu/drm/i915/gt/intel_reset.c | 4 ----
>>> 3 files changed, 1 insertion(+), 23 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
>>> index 99ddbb7bedbc..038f14d9b5c9 100644
>>> --- a/drivers/gpu/drm/i915/display/intel_display.c
>>> +++ b/drivers/gpu/drm/i915/display/intel_display.c
>>> @@ -5738,13 +5738,8 @@ void hsw_disable_ips(const struct intel_crtc_state *crtc_state)
>>>
>>> static void intel_crtc_dpms_overlay_disable(struct intel_crtc *intel_crtc)
>>> {
>>> - if (intel_crtc->overlay) {
>>> - struct drm_device *dev = intel_crtc->base.dev;
>>> -
>>> - mutex_lock(&dev->struct_mutex);
>>> + if (intel_crtc->overlay)
>>> (void) intel_overlay_switch_off(intel_crtc->overlay);
>>> - mutex_unlock(&dev->struct_mutex);
>>> - }
>>>
>>> /* Let userspace switch the overlay on again. In most cases userspace
>>> * has to recompute where to put it anyway.
>>> diff --git a/drivers/gpu/drm/i915/display/intel_overlay.c b/drivers/gpu/drm/i915/display/intel_overlay.c
>>> index e12e1a753af0..daea112cbb87 100644
>>> --- a/drivers/gpu/drm/i915/display/intel_overlay.c
>>> +++ b/drivers/gpu/drm/i915/display/intel_overlay.c
>>> @@ -439,8 +439,6 @@ static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
>>> struct i915_request *rq;
>>> u32 *cs;
>>>
>>> - lockdep_assert_held(&dev_priv->drm.struct_mutex);
>>> -
>>> /*
>>> * Only wait if there is actually an old frame to release to
>>> * guarantee forward progress.
>>> @@ -751,7 +749,6 @@ static int intel_overlay_do_put_image(struct intel_overlay *overlay,
>>> struct i915_vma *vma;
>>> int ret, tmp_width;
>>>
>>> - lockdep_assert_held(&dev_priv->drm.struct_mutex);
>>> WARN_ON(!drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex));
>>>
>>> ret = intel_overlay_release_old_vid(overlay);
>>> @@ -852,7 +849,6 @@ int intel_overlay_switch_off(struct intel_overlay *overlay)
>>> struct drm_i915_private *dev_priv = overlay->i915;
>>> int ret;
>>>
>>> - lockdep_assert_held(&dev_priv->drm.struct_mutex);
>>> WARN_ON(!drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex));
>>>
>>> ret = intel_overlay_recover_from_interrupt(overlay);
>>> @@ -1068,11 +1064,7 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
>>>
>>> if (!(params->flags & I915_OVERLAY_ENABLE)) {
>>> drm_modeset_lock_all(dev);
>>> - mutex_lock(&dev->struct_mutex);
>>> -
>>> ret = intel_overlay_switch_off(overlay);
>>> -
>>> - mutex_unlock(&dev->struct_mutex);
>>> drm_modeset_unlock_all(dev);
>>>
>>> return ret;
>>> @@ -1088,7 +1080,6 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
>>> return -ENOENT;
>>>
>>> drm_modeset_lock_all(dev);
>>> - mutex_lock(&dev->struct_mutex);
>>>
>>> if (i915_gem_object_is_tiled(new_bo)) {
>>> DRM_DEBUG_KMS("buffer used for overlay image can not be tiled\n");
>>> @@ -1152,14 +1143,12 @@ int intel_overlay_put_image_ioctl(struct drm_device *dev, void *data,
>>> if (ret != 0)
>>> goto out_unlock;
>>>
>>> - mutex_unlock(&dev->struct_mutex);
>>> drm_modeset_unlock_all(dev);
>>> i915_gem_object_put(new_bo);
>>>
>>> return 0;
>>>
>>> out_unlock:
>>> - mutex_unlock(&dev->struct_mutex);
>>> drm_modeset_unlock_all(dev);
>>> i915_gem_object_put(new_bo);
>>>
>>> @@ -1233,7 +1222,6 @@ int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data,
>>> }
>>>
>>> drm_modeset_lock_all(dev);
>>> - mutex_lock(&dev->struct_mutex);
>>>
>>> ret = -EINVAL;
>>> if (!(attrs->flags & I915_OVERLAY_UPDATE_ATTRS)) {
>>> @@ -1290,7 +1278,6 @@ int intel_overlay_attrs_ioctl(struct drm_device *dev, void *data,
>>>
>>> ret = 0;
>>> out_unlock:
>>> - mutex_unlock(&dev->struct_mutex);
>>> drm_modeset_unlock_all(dev);
>>>
>>> return ret;
>>> diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c
>>> index 2685c0103e90..ee0c5c24282c 100644
>>> --- a/drivers/gpu/drm/i915/gt/intel_reset.c
>>> +++ b/drivers/gpu/drm/i915/gt/intel_reset.c
>>> @@ -1265,10 +1265,6 @@ int intel_gt_terminally_wedged(struct intel_gt *gt)
>>> if (!test_bit(I915_RESET_BACKOFF, >->reset.flags))
>>> return -EIO;
>>>
>>> - /* XXX intel_reset_finish() still takes struct_mutex!!! */
>>> - if (mutex_is_locked(>->i915->drm.struct_mutex))
>>> - return -EAGAIN;
>>> -
>>
>> What is this hunk doing here?
>
> Because the overlay used struct_mutex, now it does not.
Now you see why I insist on comments and even better if they are up to
date! :) (intel_reset_finish?!)
If you are confident overlay is the only place which can interfere here
then:
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
If not we'll find out sooner or later.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* Re: [PATCH 22/27] drm/i915/overlay: Drop struct_mutex guard
2019-09-25 13:01 ` Tvrtko Ursulin
@ 2019-09-25 13:11 ` Chris Wilson
0 siblings, 0 replies; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 13:11 UTC (permalink / raw)
To: Tvrtko Ursulin, intel-gfx
Quoting Tvrtko Ursulin (2019-09-25 14:01:51)
>
> On 25/09/2019 13:53, Chris Wilson wrote:
> > Quoting Tvrtko Ursulin (2019-09-25 13:43:46)
> >>
> >> On 25/09/2019 11:01, Chris Wilson wrote:
> >>> diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c
> >>> index 2685c0103e90..ee0c5c24282c 100644
> >>> --- a/drivers/gpu/drm/i915/gt/intel_reset.c
> >>> +++ b/drivers/gpu/drm/i915/gt/intel_reset.c
> >>> @@ -1265,10 +1265,6 @@ int intel_gt_terminally_wedged(struct intel_gt *gt)
> >>> if (!test_bit(I915_RESET_BACKOFF, >->reset.flags))
> >>> return -EIO;
> >>>
> >>> - /* XXX intel_reset_finish() still takes struct_mutex!!! */
> >>> - if (mutex_is_locked(>->i915->drm.struct_mutex))
> >>> - return -EAGAIN;
> >>> -
> >>
> >> What is this hunk doing here?
> >
> > Because the overlay used struct_mutex, now it does not.
>
> Now you see why I insist on comments and even better if they are up to
> date! :) (intel_reset_finish?!)
In this case, the last remaining use was actually intel_reset_prepare,
as we have to disable the overlay before doing a GPU reset. It's just
that we could have mitigated the prepare, but had been stymied by the
intel_reset_finish struct_mutex recursion.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread
* [PATCH 23/27] drm/i915: Drop struct_mutex guard from debugfs/framebuffer_info
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (21 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 22/27] drm/i915/overlay: Drop struct_mutex guard Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 12:45 ` Tvrtko Ursulin
2019-09-25 10:01 ` [PATCH 24/27] drm/i915: Remove struct_mutex guard for debugfs/opregion Chris Wilson
` (7 subsequent siblings)
30 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
It protects nothing being accessed for the intel_framebuffer, so it's
own locking had better be sufficient.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_debugfs.c | 6 ------
1 file changed, 6 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index d4a7cda60679..5e984ba27bef 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -1445,11 +1445,6 @@ static int i915_gem_framebuffer_info(struct seq_file *m, void *data)
struct drm_device *dev = &dev_priv->drm;
struct intel_framebuffer *fbdev_fb = NULL;
struct drm_framebuffer *drm_fb;
- int ret;
-
- ret = mutex_lock_interruptible(&dev->struct_mutex);
- if (ret)
- return ret;
#ifdef CONFIG_DRM_FBDEV_EMULATION
if (dev_priv->fbdev && dev_priv->fbdev->helper.fb) {
@@ -1484,7 +1479,6 @@ static int i915_gem_framebuffer_info(struct seq_file *m, void *data)
seq_putc(m, '\n');
}
mutex_unlock(&dev->mode_config.fb_lock);
- mutex_unlock(&dev->struct_mutex);
return 0;
}
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* Re: [PATCH 23/27] drm/i915: Drop struct_mutex guard from debugfs/framebuffer_info
2019-09-25 10:01 ` [PATCH 23/27] drm/i915: Drop struct_mutex guard from debugfs/framebuffer_info Chris Wilson
@ 2019-09-25 12:45 ` Tvrtko Ursulin
0 siblings, 0 replies; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-25 12:45 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 25/09/2019 11:01, Chris Wilson wrote:
> It protects nothing being accessed for the intel_framebuffer, so it's
> own locking had better be sufficient.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/i915_debugfs.c | 6 ------
> 1 file changed, 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
> index d4a7cda60679..5e984ba27bef 100644
> --- a/drivers/gpu/drm/i915/i915_debugfs.c
> +++ b/drivers/gpu/drm/i915/i915_debugfs.c
> @@ -1445,11 +1445,6 @@ static int i915_gem_framebuffer_info(struct seq_file *m, void *data)
> struct drm_device *dev = &dev_priv->drm;
> struct intel_framebuffer *fbdev_fb = NULL;
> struct drm_framebuffer *drm_fb;
> - int ret;
> -
> - ret = mutex_lock_interruptible(&dev->struct_mutex);
> - if (ret)
> - return ret;
>
> #ifdef CONFIG_DRM_FBDEV_EMULATION
> if (dev_priv->fbdev && dev_priv->fbdev->helper.fb) {
> @@ -1484,7 +1479,6 @@ static int i915_gem_framebuffer_info(struct seq_file *m, void *data)
> seq_putc(m, '\n');
> }
> mutex_unlock(&dev->mode_config.fb_lock);
> - mutex_unlock(&dev->struct_mutex);
>
> return 0;
> }
>
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] 70+ messages in thread
* [PATCH 24/27] drm/i915: Remove struct_mutex guard for debugfs/opregion
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (22 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 23/27] drm/i915: Drop struct_mutex guard from debugfs/framebuffer_info Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 12:51 ` Tvrtko Ursulin
2019-09-25 10:01 ` [PATCH 25/27] drm/i915: Drop struct_mutex from suspend state save/restore Chris Wilson
` (6 subsequent siblings)
30 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
Having a struct_mutex around the read of a BIOS blob serves no purpose.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_debugfs.c | 12 +-----------
1 file changed, 1 insertion(+), 11 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 5e984ba27bef..8f48f2543b7c 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -1411,21 +1411,11 @@ static int i915_ring_freq_table(struct seq_file *m, void *unused)
static int i915_opregion(struct seq_file *m, void *unused)
{
- struct drm_i915_private *dev_priv = node_to_i915(m->private);
- struct drm_device *dev = &dev_priv->drm;
- struct intel_opregion *opregion = &dev_priv->opregion;
- int ret;
-
- ret = mutex_lock_interruptible(&dev->struct_mutex);
- if (ret)
- goto out;
+ struct intel_opregion *opregion = &node_to_i915(m->private)->opregion;
if (opregion->header)
seq_write(m, opregion->header, OPREGION_SIZE);
- mutex_unlock(&dev->struct_mutex);
-
-out:
return 0;
}
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* Re: [PATCH 24/27] drm/i915: Remove struct_mutex guard for debugfs/opregion
2019-09-25 10:01 ` [PATCH 24/27] drm/i915: Remove struct_mutex guard for debugfs/opregion Chris Wilson
@ 2019-09-25 12:51 ` Tvrtko Ursulin
0 siblings, 0 replies; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-25 12:51 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 25/09/2019 11:01, Chris Wilson wrote:
> Having a struct_mutex around the read of a BIOS blob serves no purpose.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/i915_debugfs.c | 12 +-----------
> 1 file changed, 1 insertion(+), 11 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
> index 5e984ba27bef..8f48f2543b7c 100644
> --- a/drivers/gpu/drm/i915/i915_debugfs.c
> +++ b/drivers/gpu/drm/i915/i915_debugfs.c
> @@ -1411,21 +1411,11 @@ static int i915_ring_freq_table(struct seq_file *m, void *unused)
>
> static int i915_opregion(struct seq_file *m, void *unused)
> {
> - struct drm_i915_private *dev_priv = node_to_i915(m->private);
> - struct drm_device *dev = &dev_priv->drm;
> - struct intel_opregion *opregion = &dev_priv->opregion;
> - int ret;
> -
> - ret = mutex_lock_interruptible(&dev->struct_mutex);
> - if (ret)
> - goto out;
> + struct intel_opregion *opregion = &node_to_i915(m->private)->opregion;
>
> if (opregion->header)
> seq_write(m, opregion->header, OPREGION_SIZE);
>
> - mutex_unlock(&dev->struct_mutex);
> -
> -out:
> return 0;
> }
>
>
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] 70+ messages in thread
* [PATCH 25/27] drm/i915: Drop struct_mutex from suspend state save/restore
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (23 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 24/27] drm/i915: Remove struct_mutex guard for debugfs/opregion Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 12:52 ` Tvrtko Ursulin
2019-09-25 10:01 ` [PATCH 26/27] drm/i915/selftests: Drop vestigal struct_mutex guards Chris Wilson
` (5 subsequent siblings)
30 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
struct_mutex provides no serialisation of the registers and data
structures being saved and restored across suspend/resume. It is
completely superfluous here.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_suspend.c | 8 --------
1 file changed, 8 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_suspend.c b/drivers/gpu/drm/i915/i915_suspend.c
index 8508a01ad8b9..1bbfd8fd5124 100644
--- a/drivers/gpu/drm/i915/i915_suspend.c
+++ b/drivers/gpu/drm/i915/i915_suspend.c
@@ -65,8 +65,6 @@ int i915_save_state(struct drm_i915_private *dev_priv)
struct pci_dev *pdev = dev_priv->drm.pdev;
int i;
- mutex_lock(&dev_priv->drm.struct_mutex);
-
i915_save_display(dev_priv);
if (IS_GEN(dev_priv, 4))
@@ -100,8 +98,6 @@ int i915_save_state(struct drm_i915_private *dev_priv)
dev_priv->regfile.saveSWF3[i] = I915_READ(SWF3(i));
}
- mutex_unlock(&dev_priv->drm.struct_mutex);
-
return 0;
}
@@ -110,8 +106,6 @@ int i915_restore_state(struct drm_i915_private *dev_priv)
struct pci_dev *pdev = dev_priv->drm.pdev;
int i;
- mutex_lock(&dev_priv->drm.struct_mutex);
-
if (IS_GEN(dev_priv, 4))
pci_write_config_word(pdev, GCDGMBUS,
dev_priv->regfile.saveGCDGMBUS);
@@ -145,8 +139,6 @@ int i915_restore_state(struct drm_i915_private *dev_priv)
I915_WRITE(SWF3(i), dev_priv->regfile.saveSWF3[i]);
}
- mutex_unlock(&dev_priv->drm.struct_mutex);
-
intel_gmbus_reset(dev_priv);
return 0;
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* Re: [PATCH 25/27] drm/i915: Drop struct_mutex from suspend state save/restore
2019-09-25 10:01 ` [PATCH 25/27] drm/i915: Drop struct_mutex from suspend state save/restore Chris Wilson
@ 2019-09-25 12:52 ` Tvrtko Ursulin
0 siblings, 0 replies; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-25 12:52 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 25/09/2019 11:01, Chris Wilson wrote:
> struct_mutex provides no serialisation of the registers and data
> structures being saved and restored across suspend/resume. It is
> completely superfluous here.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/i915_suspend.c | 8 --------
> 1 file changed, 8 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_suspend.c b/drivers/gpu/drm/i915/i915_suspend.c
> index 8508a01ad8b9..1bbfd8fd5124 100644
> --- a/drivers/gpu/drm/i915/i915_suspend.c
> +++ b/drivers/gpu/drm/i915/i915_suspend.c
> @@ -65,8 +65,6 @@ int i915_save_state(struct drm_i915_private *dev_priv)
> struct pci_dev *pdev = dev_priv->drm.pdev;
> int i;
>
> - mutex_lock(&dev_priv->drm.struct_mutex);
> -
> i915_save_display(dev_priv);
>
> if (IS_GEN(dev_priv, 4))
> @@ -100,8 +98,6 @@ int i915_save_state(struct drm_i915_private *dev_priv)
> dev_priv->regfile.saveSWF3[i] = I915_READ(SWF3(i));
> }
>
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> -
> return 0;
> }
>
> @@ -110,8 +106,6 @@ int i915_restore_state(struct drm_i915_private *dev_priv)
> struct pci_dev *pdev = dev_priv->drm.pdev;
> int i;
>
> - mutex_lock(&dev_priv->drm.struct_mutex);
> -
> if (IS_GEN(dev_priv, 4))
> pci_write_config_word(pdev, GCDGMBUS,
> dev_priv->regfile.saveGCDGMBUS);
> @@ -145,8 +139,6 @@ int i915_restore_state(struct drm_i915_private *dev_priv)
> I915_WRITE(SWF3(i), dev_priv->regfile.saveSWF3[i]);
> }
>
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> -
> intel_gmbus_reset(dev_priv);
>
> return 0;
>
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] 70+ messages in thread
* [PATCH 26/27] drm/i915/selftests: Drop vestigal struct_mutex guards
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (24 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 25/27] drm/i915: Drop struct_mutex from suspend state save/restore Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 12:55 ` Tvrtko Ursulin
2019-09-25 10:01 ` [PATCH 27/27] drm/i915: Drop struct_mutex from around GEM initialisation Chris Wilson
` (4 subsequent siblings)
30 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
We no longer need struct_mutex to serialise request emission, so remove
it from the gt selftests.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
.../gpu/drm/i915/gem/selftests/huge_pages.c | 15 +-
.../drm/i915/gem/selftests/i915_gem_mman.c | 2 -
.../drm/i915/gem/selftests/i915_gem_phys.c | 2 -
drivers/gpu/drm/i915/gt/selftest_lrc.c | 148 +++---------------
.../gpu/drm/i915/gt/selftest_workarounds.c | 11 +-
drivers/gpu/drm/i915/gt/uc/selftest_guc.c | 4 -
6 files changed, 27 insertions(+), 155 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
index e204e653b459..75e717f7bb5a 100644
--- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
+++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
@@ -1633,7 +1633,6 @@ int i915_gem_huge_page_mock_selftests(void)
mkwrite_device_info(dev_priv)->ppgtt_type = INTEL_PPGTT_FULL;
mkwrite_device_info(dev_priv)->ppgtt_size = 48;
- mutex_lock(&dev_priv->drm.struct_mutex);
ppgtt = i915_ppgtt_create(dev_priv);
if (IS_ERR(ppgtt)) {
err = PTR_ERR(ppgtt);
@@ -1659,9 +1658,7 @@ int i915_gem_huge_page_mock_selftests(void)
i915_vm_put(&ppgtt->vm);
out_unlock:
- mutex_unlock(&dev_priv->drm.struct_mutex);
drm_dev_put(&dev_priv->drm);
-
return err;
}
@@ -1677,7 +1674,6 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *i915)
};
struct drm_file *file;
struct i915_gem_context *ctx;
- intel_wakeref_t wakeref;
int err;
if (!HAS_PPGTT(i915)) {
@@ -1692,13 +1688,10 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *i915)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
ctx = live_context(i915, file);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
- goto out_unlock;
+ goto out_file;
}
if (rcu_access_pointer(ctx->vm))
@@ -1706,11 +1699,7 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *i915)
err = i915_subtests(tests, ctx);
-out_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
-
+out_file:
mock_file_free(i915, file);
-
return 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 1cd25cfd0246..cfa52c525691 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
@@ -669,9 +669,7 @@ static int igt_mmap_offset_exhaustion(void *arg)
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;
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 94a15e3f6db8..34932871b3a5 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c
@@ -25,9 +25,7 @@ static int mock_phys_object(void *arg)
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;
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index e33c2f66c683..89affc958d01 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -26,17 +26,13 @@ static int live_sanitycheck(void *arg)
struct i915_gem_context *ctx;
struct intel_context *ce;
struct igt_spinner spin;
- intel_wakeref_t wakeref;
int err = -ENOMEM;
if (!HAS_LOGICAL_RING_CONTEXTS(i915))
return 0;
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
if (igt_spinner_init(&spin, &i915->gt))
- goto err_unlock;
+ return -ENOMEM;
ctx = kernel_context(i915);
if (!ctx)
@@ -73,9 +69,6 @@ static int live_sanitycheck(void *arg)
kernel_context_close(ctx);
err_spin:
igt_spinner_fini(&spin);
-err_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -224,7 +217,6 @@ 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;
@@ -239,14 +231,9 @@ static int live_timeslice_preempt(void *arg)
* ready task.
*/
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
- if (IS_ERR(obj)) {
- err = PTR_ERR(obj);
- goto err_unlock;
- }
+ if (IS_ERR(obj))
+ return PTR_ERR(obj);
vma = i915_vma_instance(obj, &i915->ggtt.vm, NULL);
if (IS_ERR(vma)) {
@@ -291,10 +278,6 @@ static int live_timeslice_preempt(void *arg)
i915_gem_object_unpin_map(obj);
err_obj:
i915_gem_object_put(obj);
-err_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
-
return err;
}
@@ -306,7 +289,6 @@ static int live_busywait_preempt(void *arg)
struct drm_i915_gem_object *obj;
struct i915_vma *vma;
enum intel_engine_id id;
- intel_wakeref_t wakeref;
int err = -ENOMEM;
u32 *map;
@@ -315,12 +297,9 @@ static int live_busywait_preempt(void *arg)
* preempt the busywaits used to synchronise between rings.
*/
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
ctx_hi = kernel_context(i915);
if (!ctx_hi)
- goto err_unlock;
+ return -ENOMEM;
ctx_hi->sched.priority =
I915_USER_PRIORITY(I915_CONTEXT_MAX_USER_PRIORITY);
@@ -474,9 +453,6 @@ static int live_busywait_preempt(void *arg)
kernel_context_close(ctx_lo);
err_ctx_hi:
kernel_context_close(ctx_hi);
-err_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -505,7 +481,6 @@ static int live_preempt(void *arg)
struct igt_spinner spin_hi, spin_lo;
struct intel_engine_cs *engine;
enum intel_engine_id id;
- intel_wakeref_t wakeref;
int err = -ENOMEM;
if (!HAS_LOGICAL_RING_PREEMPTION(i915))
@@ -514,11 +489,8 @@ static int live_preempt(void *arg)
if (!(i915->caps.scheduler & I915_SCHEDULER_CAP_PREEMPTION))
pr_err("Logical preemption supported, but not exposed\n");
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
if (igt_spinner_init(&spin_hi, &i915->gt))
- goto err_unlock;
+ return -ENOMEM;
if (igt_spinner_init(&spin_lo, &i915->gt))
goto err_spin_hi;
@@ -598,9 +570,6 @@ static int live_preempt(void *arg)
igt_spinner_fini(&spin_lo);
err_spin_hi:
igt_spinner_fini(&spin_hi);
-err_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -612,17 +581,13 @@ static int live_late_preempt(void *arg)
struct intel_engine_cs *engine;
struct i915_sched_attr attr = {};
enum intel_engine_id id;
- intel_wakeref_t wakeref;
int err = -ENOMEM;
if (!HAS_LOGICAL_RING_PREEMPTION(i915))
return 0;
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
if (igt_spinner_init(&spin_hi, &i915->gt))
- goto err_unlock;
+ return -ENOMEM;
if (igt_spinner_init(&spin_lo, &i915->gt))
goto err_spin_hi;
@@ -704,9 +669,6 @@ static int live_late_preempt(void *arg)
igt_spinner_fini(&spin_lo);
err_spin_hi:
igt_spinner_fini(&spin_hi);
-err_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
err_wedged:
@@ -751,7 +713,6 @@ static int live_nopreempt(void *arg)
struct intel_engine_cs *engine;
struct preempt_client a, b;
enum intel_engine_id id;
- intel_wakeref_t wakeref;
int err = -ENOMEM;
/*
@@ -762,11 +723,8 @@ static int live_nopreempt(void *arg)
if (!HAS_LOGICAL_RING_PREEMPTION(i915))
return 0;
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
if (preempt_client_init(i915, &a))
- goto err_unlock;
+ return -ENOMEM;
if (preempt_client_init(i915, &b))
goto err_client_a;
b.ctx->sched.priority = I915_USER_PRIORITY(I915_PRIORITY_MAX);
@@ -840,9 +798,6 @@ static int live_nopreempt(void *arg)
preempt_client_fini(&b);
err_client_a:
preempt_client_fini(&a);
-err_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
err_wedged:
@@ -862,7 +817,6 @@ static int live_suppress_self_preempt(void *arg)
};
struct preempt_client a, b;
enum intel_engine_id id;
- intel_wakeref_t wakeref;
int err = -ENOMEM;
/*
@@ -881,11 +835,8 @@ static int live_suppress_self_preempt(void *arg)
if (intel_vgpu_active(i915))
return 0; /* GVT forces single port & request submission */
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
if (preempt_client_init(i915, &a))
- goto err_unlock;
+ return -ENOMEM;
if (preempt_client_init(i915, &b))
goto err_client_a;
@@ -966,9 +917,6 @@ static int live_suppress_self_preempt(void *arg)
preempt_client_fini(&b);
err_client_a:
preempt_client_fini(&a);
-err_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
err_wedged:
@@ -1038,7 +986,6 @@ static int live_suppress_wait_preempt(void *arg)
struct preempt_client client[4];
struct intel_engine_cs *engine;
enum intel_engine_id id;
- intel_wakeref_t wakeref;
int err = -ENOMEM;
int i;
@@ -1051,11 +998,8 @@ static int live_suppress_wait_preempt(void *arg)
if (!HAS_LOGICAL_RING_PREEMPTION(i915))
return 0;
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
if (preempt_client_init(i915, &client[0])) /* ELSP[0] */
- goto err_unlock;
+ return -ENOMEM;
if (preempt_client_init(i915, &client[1])) /* ELSP[1] */
goto err_client_0;
if (preempt_client_init(i915, &client[2])) /* head of queue */
@@ -1141,9 +1085,6 @@ static int live_suppress_wait_preempt(void *arg)
preempt_client_fini(&client[1]);
err_client_0:
preempt_client_fini(&client[0]);
-err_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
err_wedged:
@@ -1160,7 +1101,6 @@ static int live_chain_preempt(void *arg)
struct intel_engine_cs *engine;
struct preempt_client hi, lo;
enum intel_engine_id id;
- intel_wakeref_t wakeref;
int err = -ENOMEM;
/*
@@ -1172,11 +1112,8 @@ static int live_chain_preempt(void *arg)
if (!HAS_LOGICAL_RING_PREEMPTION(i915))
return 0;
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
if (preempt_client_init(i915, &hi))
- goto err_unlock;
+ return -ENOMEM;
if (preempt_client_init(i915, &lo))
goto err_client_hi;
@@ -1287,9 +1224,6 @@ static int live_chain_preempt(void *arg)
preempt_client_fini(&lo);
err_client_hi:
preempt_client_fini(&hi);
-err_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
err_wedged:
@@ -1307,7 +1241,6 @@ static int live_preempt_hang(void *arg)
struct igt_spinner spin_hi, spin_lo;
struct intel_engine_cs *engine;
enum intel_engine_id id;
- intel_wakeref_t wakeref;
int err = -ENOMEM;
if (!HAS_LOGICAL_RING_PREEMPTION(i915))
@@ -1316,11 +1249,8 @@ static int live_preempt_hang(void *arg)
if (!intel_has_reset_engine(i915))
return 0;
- mutex_lock(&i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&i915->runtime_pm);
-
if (igt_spinner_init(&spin_hi, &i915->gt))
- goto err_unlock;
+ return -ENOMEM;
if (igt_spinner_init(&spin_lo, &i915->gt))
goto err_spin_hi;
@@ -1412,9 +1342,6 @@ static int live_preempt_hang(void *arg)
igt_spinner_fini(&spin_lo);
err_spin_hi:
igt_spinner_fini(&spin_hi);
-err_unlock:
- intel_runtime_pm_put(&i915->runtime_pm, wakeref);
- mutex_unlock(&i915->drm.struct_mutex);
return err;
}
@@ -1504,11 +1431,9 @@ static int smoke_crescendo_thread(void *arg)
struct i915_gem_context *ctx = smoke_context(smoke);
int err;
- mutex_lock(&smoke->i915->drm.struct_mutex);
err = smoke_submit(smoke,
ctx, count % I915_PRIORITY_MAX,
smoke->batch);
- mutex_unlock(&smoke->i915->drm.struct_mutex);
if (err)
return err;
@@ -1529,8 +1454,6 @@ static int smoke_crescendo(struct preempt_smoke *smoke, unsigned int flags)
unsigned long count;
int err = 0;
- mutex_unlock(&smoke->i915->drm.struct_mutex);
-
for_each_engine(engine, smoke->i915, id) {
arg[id] = *smoke;
arg[id].engine = engine;
@@ -1563,8 +1486,6 @@ static int smoke_crescendo(struct preempt_smoke *smoke, unsigned int flags)
put_task_struct(tsk[id]);
}
- mutex_lock(&smoke->i915->drm.struct_mutex);
-
pr_info("Submitted %lu crescendo:%x requests across %d engines and %d contexts\n",
count, flags,
RUNTIME_INFO(smoke->i915)->num_engines, smoke->ncontext);
@@ -1607,7 +1528,6 @@ static int live_preempt_smoke(void *arg)
.ncontext = 1024,
};
const unsigned int phase[] = { 0, BATCH };
- intel_wakeref_t wakeref;
struct igt_live_test t;
int err = -ENOMEM;
u32 *cs;
@@ -1622,13 +1542,10 @@ static int live_preempt_smoke(void *arg)
if (!smoke.contexts)
return -ENOMEM;
- mutex_lock(&smoke.i915->drm.struct_mutex);
- wakeref = intel_runtime_pm_get(&smoke.i915->runtime_pm);
-
smoke.batch = i915_gem_object_create_internal(smoke.i915, PAGE_SIZE);
if (IS_ERR(smoke.batch)) {
err = PTR_ERR(smoke.batch);
- goto err_unlock;
+ goto err_free;
}
cs = i915_gem_object_pin_map(smoke.batch, I915_MAP_WB);
@@ -1675,9 +1592,7 @@ static int live_preempt_smoke(void *arg)
err_batch:
i915_gem_object_put(smoke.batch);
-err_unlock:
- intel_runtime_pm_put(&smoke.i915->runtime_pm, wakeref);
- mutex_unlock(&smoke.i915->drm.struct_mutex);
+err_free:
kfree(smoke.contexts);
return err;
@@ -1815,19 +1730,17 @@ static int live_virtual_engine(void *arg)
struct intel_gt *gt = &i915->gt;
enum intel_engine_id id;
unsigned int class, inst;
- int err = -ENODEV;
+ int err;
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;
+ return err;
}
}
@@ -1848,17 +1761,15 @@ static int live_virtual_engine(void *arg)
err = nop_virtual_engine(i915, siblings, nsibling,
n, 0);
if (err)
- goto out_unlock;
+ return err;
}
err = nop_virtual_engine(i915, siblings, nsibling, n, CHAIN);
if (err)
- goto out_unlock;
+ return err;
}
-out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
- return err;
+ return 0;
}
static int mask_virtual_engine(struct drm_i915_private *i915,
@@ -1937,9 +1848,6 @@ static int mask_virtual_engine(struct drm_i915_private *i915,
}
err = igt_live_test_end(&t);
- if (err)
- goto out;
-
out:
if (igt_flush_test(i915))
err = -EIO;
@@ -1962,13 +1870,11 @@ static int live_virtual_mask(void *arg)
struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1];
struct intel_gt *gt = &i915->gt;
unsigned int class, inst;
- int err = 0;
+ int err;
if (USES_GUC_SUBMISSION(i915))
return 0;
- mutex_lock(&i915->drm.struct_mutex);
-
for (class = 0; class <= MAX_ENGINE_CLASS; class++) {
unsigned int nsibling;
@@ -1984,12 +1890,10 @@ static int live_virtual_mask(void *arg)
err = mask_virtual_engine(i915, siblings, nsibling);
if (err)
- goto out_unlock;
+ return err;
}
-out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
- return err;
+ return 0;
}
static int bond_virtual_engine(struct drm_i915_private *i915,
@@ -2140,13 +2044,11 @@ static int live_virtual_bond(void *arg)
struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1];
struct intel_gt *gt = &i915->gt;
unsigned int class, inst;
- int err = 0;
+ int err;
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;
@@ -2169,14 +2071,12 @@ static int live_virtual_bond(void *arg)
if (err) {
pr_err("%s(%s): failed class=%d, nsibling=%d, err=%d\n",
__func__, p->name, class, nsibling, err);
- goto out_unlock;
+ return err;
}
}
}
-out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
- return err;
+ return 0;
}
int intel_execlists_live_selftests(struct drm_i915_private *i915)
diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
index ed44333d9e20..a10990b1a72c 100644
--- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
@@ -704,9 +704,7 @@ static int live_dirty_whitelist(void *arg)
wakeref = intel_runtime_pm_get(&i915->runtime_pm);
- mutex_unlock(&i915->drm.struct_mutex);
file = mock_file(i915);
- mutex_lock(&i915->drm.struct_mutex);
if (IS_ERR(file)) {
err = PTR_ERR(file);
goto out_rpm;
@@ -728,9 +726,7 @@ static int live_dirty_whitelist(void *arg)
}
out_file:
- mutex_unlock(&i915->drm.struct_mutex);
mock_file_free(i915, file);
- mutex_lock(&i915->drm.struct_mutex);
out_rpm:
intel_runtime_pm_put(&i915->runtime_pm, wakeref);
return err;
@@ -1264,14 +1260,9 @@ int intel_workarounds_live_selftests(struct drm_i915_private *i915)
SUBTEST(live_gpu_reset_workarounds),
SUBTEST(live_engine_reset_workarounds),
};
- int err;
if (intel_gt_is_wedged(&i915->gt))
return 0;
- mutex_lock(&i915->drm.struct_mutex);
- err = i915_subtests(tests, i915);
- mutex_unlock(&i915->drm.struct_mutex);
-
- return err;
+ return i915_subtests(tests, i915);
}
diff --git a/drivers/gpu/drm/i915/gt/uc/selftest_guc.c b/drivers/gpu/drm/i915/gt/uc/selftest_guc.c
index bba0eafe1cdb..f927f851aadf 100644
--- a/drivers/gpu/drm/i915/gt/uc/selftest_guc.c
+++ b/drivers/gpu/drm/i915/gt/uc/selftest_guc.c
@@ -116,7 +116,6 @@ static int igt_guc_clients(void *args)
int err = 0;
GEM_BUG_ON(!HAS_GT_UC(dev_priv));
- mutex_lock(&dev_priv->drm.struct_mutex);
wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm);
guc = &dev_priv->gt.uc.guc;
@@ -190,7 +189,6 @@ static int igt_guc_clients(void *args)
guc_clients_enable(guc);
unlock:
intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref);
- mutex_unlock(&dev_priv->drm.struct_mutex);
return err;
}
@@ -208,7 +206,6 @@ static int igt_guc_doorbells(void *arg)
u16 db_id;
GEM_BUG_ON(!HAS_GT_UC(dev_priv));
- mutex_lock(&dev_priv->drm.struct_mutex);
wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm);
guc = &dev_priv->gt.uc.guc;
@@ -299,7 +296,6 @@ static int igt_guc_doorbells(void *arg)
}
unlock:
intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref);
- mutex_unlock(&dev_priv->drm.struct_mutex);
return err;
}
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* Re: [PATCH 26/27] drm/i915/selftests: Drop vestigal struct_mutex guards
2019-09-25 10:01 ` [PATCH 26/27] drm/i915/selftests: Drop vestigal struct_mutex guards Chris Wilson
@ 2019-09-25 12:55 ` Tvrtko Ursulin
0 siblings, 0 replies; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-25 12:55 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 25/09/2019 11:01, Chris Wilson wrote:
> We no longer need struct_mutex to serialise request emission, so remove
> it from the gt selftests.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> .../gpu/drm/i915/gem/selftests/huge_pages.c | 15 +-
> .../drm/i915/gem/selftests/i915_gem_mman.c | 2 -
> .../drm/i915/gem/selftests/i915_gem_phys.c | 2 -
> drivers/gpu/drm/i915/gt/selftest_lrc.c | 148 +++---------------
> .../gpu/drm/i915/gt/selftest_workarounds.c | 11 +-
> drivers/gpu/drm/i915/gt/uc/selftest_guc.c | 4 -
> 6 files changed, 27 insertions(+), 155 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
> index e204e653b459..75e717f7bb5a 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
> @@ -1633,7 +1633,6 @@ int i915_gem_huge_page_mock_selftests(void)
> mkwrite_device_info(dev_priv)->ppgtt_type = INTEL_PPGTT_FULL;
> mkwrite_device_info(dev_priv)->ppgtt_size = 48;
>
> - mutex_lock(&dev_priv->drm.struct_mutex);
> ppgtt = i915_ppgtt_create(dev_priv);
> if (IS_ERR(ppgtt)) {
> err = PTR_ERR(ppgtt);
> @@ -1659,9 +1658,7 @@ int i915_gem_huge_page_mock_selftests(void)
> i915_vm_put(&ppgtt->vm);
>
> out_unlock:
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> drm_dev_put(&dev_priv->drm);
> -
> return err;
> }
>
> @@ -1677,7 +1674,6 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *i915)
> };
> struct drm_file *file;
> struct i915_gem_context *ctx;
> - intel_wakeref_t wakeref;
> int err;
>
> if (!HAS_PPGTT(i915)) {
> @@ -1692,13 +1688,10 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *i915)
> if (IS_ERR(file))
> return PTR_ERR(file);
>
> - mutex_lock(&i915->drm.struct_mutex);
> - wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> -
> ctx = live_context(i915, file);
> if (IS_ERR(ctx)) {
> err = PTR_ERR(ctx);
> - goto out_unlock;
> + goto out_file;
> }
>
> if (rcu_access_pointer(ctx->vm))
> @@ -1706,11 +1699,7 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *i915)
>
> err = i915_subtests(tests, ctx);
>
> -out_unlock:
> - intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> +out_file:
> mock_file_free(i915, file);
> -
> return 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 1cd25cfd0246..cfa52c525691 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
> @@ -669,9 +669,7 @@ static int igt_mmap_offset_exhaustion(void *arg)
> 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;
> 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 94a15e3f6db8..34932871b3a5 100644
> --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c
> +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c
> @@ -25,9 +25,7 @@ static int mock_phys_object(void *arg)
> 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;
> diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
> index e33c2f66c683..89affc958d01 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
> @@ -26,17 +26,13 @@ static int live_sanitycheck(void *arg)
> struct i915_gem_context *ctx;
> struct intel_context *ce;
> struct igt_spinner spin;
> - intel_wakeref_t wakeref;
> int err = -ENOMEM;
>
> if (!HAS_LOGICAL_RING_CONTEXTS(i915))
> return 0;
>
> - mutex_lock(&i915->drm.struct_mutex);
> - wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> -
> if (igt_spinner_init(&spin, &i915->gt))
> - goto err_unlock;
> + return -ENOMEM;
>
> ctx = kernel_context(i915);
> if (!ctx)
> @@ -73,9 +69,6 @@ static int live_sanitycheck(void *arg)
> kernel_context_close(ctx);
> err_spin:
> igt_spinner_fini(&spin);
> -err_unlock:
> - intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> return err;
> }
>
> @@ -224,7 +217,6 @@ 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;
> @@ -239,14 +231,9 @@ static int live_timeslice_preempt(void *arg)
> * ready task.
> */
>
> - mutex_lock(&i915->drm.struct_mutex);
> - wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> -
> obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
> - if (IS_ERR(obj)) {
> - err = PTR_ERR(obj);
> - goto err_unlock;
> - }
> + if (IS_ERR(obj))
> + return PTR_ERR(obj);
>
> vma = i915_vma_instance(obj, &i915->ggtt.vm, NULL);
> if (IS_ERR(vma)) {
> @@ -291,10 +278,6 @@ static int live_timeslice_preempt(void *arg)
> i915_gem_object_unpin_map(obj);
> err_obj:
> i915_gem_object_put(obj);
> -err_unlock:
> - intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> return err;
> }
>
> @@ -306,7 +289,6 @@ static int live_busywait_preempt(void *arg)
> struct drm_i915_gem_object *obj;
> struct i915_vma *vma;
> enum intel_engine_id id;
> - intel_wakeref_t wakeref;
> int err = -ENOMEM;
> u32 *map;
>
> @@ -315,12 +297,9 @@ static int live_busywait_preempt(void *arg)
> * preempt the busywaits used to synchronise between rings.
> */
>
> - mutex_lock(&i915->drm.struct_mutex);
> - wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> -
> ctx_hi = kernel_context(i915);
> if (!ctx_hi)
> - goto err_unlock;
> + return -ENOMEM;
> ctx_hi->sched.priority =
> I915_USER_PRIORITY(I915_CONTEXT_MAX_USER_PRIORITY);
>
> @@ -474,9 +453,6 @@ static int live_busywait_preempt(void *arg)
> kernel_context_close(ctx_lo);
> err_ctx_hi:
> kernel_context_close(ctx_hi);
> -err_unlock:
> - intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> return err;
> }
>
> @@ -505,7 +481,6 @@ static int live_preempt(void *arg)
> struct igt_spinner spin_hi, spin_lo;
> struct intel_engine_cs *engine;
> enum intel_engine_id id;
> - intel_wakeref_t wakeref;
> int err = -ENOMEM;
>
> if (!HAS_LOGICAL_RING_PREEMPTION(i915))
> @@ -514,11 +489,8 @@ static int live_preempt(void *arg)
> if (!(i915->caps.scheduler & I915_SCHEDULER_CAP_PREEMPTION))
> pr_err("Logical preemption supported, but not exposed\n");
>
> - mutex_lock(&i915->drm.struct_mutex);
> - wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> -
> if (igt_spinner_init(&spin_hi, &i915->gt))
> - goto err_unlock;
> + return -ENOMEM;
>
> if (igt_spinner_init(&spin_lo, &i915->gt))
> goto err_spin_hi;
> @@ -598,9 +570,6 @@ static int live_preempt(void *arg)
> igt_spinner_fini(&spin_lo);
> err_spin_hi:
> igt_spinner_fini(&spin_hi);
> -err_unlock:
> - intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> return err;
> }
>
> @@ -612,17 +581,13 @@ static int live_late_preempt(void *arg)
> struct intel_engine_cs *engine;
> struct i915_sched_attr attr = {};
> enum intel_engine_id id;
> - intel_wakeref_t wakeref;
> int err = -ENOMEM;
>
> if (!HAS_LOGICAL_RING_PREEMPTION(i915))
> return 0;
>
> - mutex_lock(&i915->drm.struct_mutex);
> - wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> -
> if (igt_spinner_init(&spin_hi, &i915->gt))
> - goto err_unlock;
> + return -ENOMEM;
>
> if (igt_spinner_init(&spin_lo, &i915->gt))
> goto err_spin_hi;
> @@ -704,9 +669,6 @@ static int live_late_preempt(void *arg)
> igt_spinner_fini(&spin_lo);
> err_spin_hi:
> igt_spinner_fini(&spin_hi);
> -err_unlock:
> - intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> return err;
>
> err_wedged:
> @@ -751,7 +713,6 @@ static int live_nopreempt(void *arg)
> struct intel_engine_cs *engine;
> struct preempt_client a, b;
> enum intel_engine_id id;
> - intel_wakeref_t wakeref;
> int err = -ENOMEM;
>
> /*
> @@ -762,11 +723,8 @@ static int live_nopreempt(void *arg)
> if (!HAS_LOGICAL_RING_PREEMPTION(i915))
> return 0;
>
> - mutex_lock(&i915->drm.struct_mutex);
> - wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> -
> if (preempt_client_init(i915, &a))
> - goto err_unlock;
> + return -ENOMEM;
> if (preempt_client_init(i915, &b))
> goto err_client_a;
> b.ctx->sched.priority = I915_USER_PRIORITY(I915_PRIORITY_MAX);
> @@ -840,9 +798,6 @@ static int live_nopreempt(void *arg)
> preempt_client_fini(&b);
> err_client_a:
> preempt_client_fini(&a);
> -err_unlock:
> - intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> return err;
>
> err_wedged:
> @@ -862,7 +817,6 @@ static int live_suppress_self_preempt(void *arg)
> };
> struct preempt_client a, b;
> enum intel_engine_id id;
> - intel_wakeref_t wakeref;
> int err = -ENOMEM;
>
> /*
> @@ -881,11 +835,8 @@ static int live_suppress_self_preempt(void *arg)
> if (intel_vgpu_active(i915))
> return 0; /* GVT forces single port & request submission */
>
> - mutex_lock(&i915->drm.struct_mutex);
> - wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> -
> if (preempt_client_init(i915, &a))
> - goto err_unlock;
> + return -ENOMEM;
> if (preempt_client_init(i915, &b))
> goto err_client_a;
>
> @@ -966,9 +917,6 @@ static int live_suppress_self_preempt(void *arg)
> preempt_client_fini(&b);
> err_client_a:
> preempt_client_fini(&a);
> -err_unlock:
> - intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> return err;
>
> err_wedged:
> @@ -1038,7 +986,6 @@ static int live_suppress_wait_preempt(void *arg)
> struct preempt_client client[4];
> struct intel_engine_cs *engine;
> enum intel_engine_id id;
> - intel_wakeref_t wakeref;
> int err = -ENOMEM;
> int i;
>
> @@ -1051,11 +998,8 @@ static int live_suppress_wait_preempt(void *arg)
> if (!HAS_LOGICAL_RING_PREEMPTION(i915))
> return 0;
>
> - mutex_lock(&i915->drm.struct_mutex);
> - wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> -
> if (preempt_client_init(i915, &client[0])) /* ELSP[0] */
> - goto err_unlock;
> + return -ENOMEM;
> if (preempt_client_init(i915, &client[1])) /* ELSP[1] */
> goto err_client_0;
> if (preempt_client_init(i915, &client[2])) /* head of queue */
> @@ -1141,9 +1085,6 @@ static int live_suppress_wait_preempt(void *arg)
> preempt_client_fini(&client[1]);
> err_client_0:
> preempt_client_fini(&client[0]);
> -err_unlock:
> - intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> return err;
>
> err_wedged:
> @@ -1160,7 +1101,6 @@ static int live_chain_preempt(void *arg)
> struct intel_engine_cs *engine;
> struct preempt_client hi, lo;
> enum intel_engine_id id;
> - intel_wakeref_t wakeref;
> int err = -ENOMEM;
>
> /*
> @@ -1172,11 +1112,8 @@ static int live_chain_preempt(void *arg)
> if (!HAS_LOGICAL_RING_PREEMPTION(i915))
> return 0;
>
> - mutex_lock(&i915->drm.struct_mutex);
> - wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> -
> if (preempt_client_init(i915, &hi))
> - goto err_unlock;
> + return -ENOMEM;
>
> if (preempt_client_init(i915, &lo))
> goto err_client_hi;
> @@ -1287,9 +1224,6 @@ static int live_chain_preempt(void *arg)
> preempt_client_fini(&lo);
> err_client_hi:
> preempt_client_fini(&hi);
> -err_unlock:
> - intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> return err;
>
> err_wedged:
> @@ -1307,7 +1241,6 @@ static int live_preempt_hang(void *arg)
> struct igt_spinner spin_hi, spin_lo;
> struct intel_engine_cs *engine;
> enum intel_engine_id id;
> - intel_wakeref_t wakeref;
> int err = -ENOMEM;
>
> if (!HAS_LOGICAL_RING_PREEMPTION(i915))
> @@ -1316,11 +1249,8 @@ static int live_preempt_hang(void *arg)
> if (!intel_has_reset_engine(i915))
> return 0;
>
> - mutex_lock(&i915->drm.struct_mutex);
> - wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> -
> if (igt_spinner_init(&spin_hi, &i915->gt))
> - goto err_unlock;
> + return -ENOMEM;
>
> if (igt_spinner_init(&spin_lo, &i915->gt))
> goto err_spin_hi;
> @@ -1412,9 +1342,6 @@ static int live_preempt_hang(void *arg)
> igt_spinner_fini(&spin_lo);
> err_spin_hi:
> igt_spinner_fini(&spin_hi);
> -err_unlock:
> - intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> - mutex_unlock(&i915->drm.struct_mutex);
> return err;
> }
>
> @@ -1504,11 +1431,9 @@ static int smoke_crescendo_thread(void *arg)
> struct i915_gem_context *ctx = smoke_context(smoke);
> int err;
>
> - mutex_lock(&smoke->i915->drm.struct_mutex);
> err = smoke_submit(smoke,
> ctx, count % I915_PRIORITY_MAX,
> smoke->batch);
> - mutex_unlock(&smoke->i915->drm.struct_mutex);
> if (err)
> return err;
>
> @@ -1529,8 +1454,6 @@ static int smoke_crescendo(struct preempt_smoke *smoke, unsigned int flags)
> unsigned long count;
> int err = 0;
>
> - mutex_unlock(&smoke->i915->drm.struct_mutex);
> -
> for_each_engine(engine, smoke->i915, id) {
> arg[id] = *smoke;
> arg[id].engine = engine;
> @@ -1563,8 +1486,6 @@ static int smoke_crescendo(struct preempt_smoke *smoke, unsigned int flags)
> put_task_struct(tsk[id]);
> }
>
> - mutex_lock(&smoke->i915->drm.struct_mutex);
> -
> pr_info("Submitted %lu crescendo:%x requests across %d engines and %d contexts\n",
> count, flags,
> RUNTIME_INFO(smoke->i915)->num_engines, smoke->ncontext);
> @@ -1607,7 +1528,6 @@ static int live_preempt_smoke(void *arg)
> .ncontext = 1024,
> };
> const unsigned int phase[] = { 0, BATCH };
> - intel_wakeref_t wakeref;
> struct igt_live_test t;
> int err = -ENOMEM;
> u32 *cs;
> @@ -1622,13 +1542,10 @@ static int live_preempt_smoke(void *arg)
> if (!smoke.contexts)
> return -ENOMEM;
>
> - mutex_lock(&smoke.i915->drm.struct_mutex);
> - wakeref = intel_runtime_pm_get(&smoke.i915->runtime_pm);
> -
> smoke.batch = i915_gem_object_create_internal(smoke.i915, PAGE_SIZE);
> if (IS_ERR(smoke.batch)) {
> err = PTR_ERR(smoke.batch);
> - goto err_unlock;
> + goto err_free;
> }
>
> cs = i915_gem_object_pin_map(smoke.batch, I915_MAP_WB);
> @@ -1675,9 +1592,7 @@ static int live_preempt_smoke(void *arg)
>
> err_batch:
> i915_gem_object_put(smoke.batch);
> -err_unlock:
> - intel_runtime_pm_put(&smoke.i915->runtime_pm, wakeref);
> - mutex_unlock(&smoke.i915->drm.struct_mutex);
> +err_free:
> kfree(smoke.contexts);
>
> return err;
> @@ -1815,19 +1730,17 @@ static int live_virtual_engine(void *arg)
> struct intel_gt *gt = &i915->gt;
> enum intel_engine_id id;
> unsigned int class, inst;
> - int err = -ENODEV;
> + int err;
>
> 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;
> + return err;
> }
> }
>
> @@ -1848,17 +1761,15 @@ static int live_virtual_engine(void *arg)
> err = nop_virtual_engine(i915, siblings, nsibling,
> n, 0);
> if (err)
> - goto out_unlock;
> + return err;
> }
>
> err = nop_virtual_engine(i915, siblings, nsibling, n, CHAIN);
> if (err)
> - goto out_unlock;
> + return err;
> }
>
> -out_unlock:
> - mutex_unlock(&i915->drm.struct_mutex);
> - return err;
> + return 0;
> }
>
> static int mask_virtual_engine(struct drm_i915_private *i915,
> @@ -1937,9 +1848,6 @@ static int mask_virtual_engine(struct drm_i915_private *i915,
> }
>
> err = igt_live_test_end(&t);
> - if (err)
> - goto out;
> -
> out:
> if (igt_flush_test(i915))
> err = -EIO;
> @@ -1962,13 +1870,11 @@ static int live_virtual_mask(void *arg)
> struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1];
> struct intel_gt *gt = &i915->gt;
> unsigned int class, inst;
> - int err = 0;
> + int err;
>
> if (USES_GUC_SUBMISSION(i915))
> return 0;
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> for (class = 0; class <= MAX_ENGINE_CLASS; class++) {
> unsigned int nsibling;
>
> @@ -1984,12 +1890,10 @@ static int live_virtual_mask(void *arg)
>
> err = mask_virtual_engine(i915, siblings, nsibling);
> if (err)
> - goto out_unlock;
> + return err;
> }
>
> -out_unlock:
> - mutex_unlock(&i915->drm.struct_mutex);
> - return err;
> + return 0;
> }
>
> static int bond_virtual_engine(struct drm_i915_private *i915,
> @@ -2140,13 +2044,11 @@ static int live_virtual_bond(void *arg)
> struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1];
> struct intel_gt *gt = &i915->gt;
> unsigned int class, inst;
> - int err = 0;
> + int err;
>
> 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;
> @@ -2169,14 +2071,12 @@ static int live_virtual_bond(void *arg)
> if (err) {
> pr_err("%s(%s): failed class=%d, nsibling=%d, err=%d\n",
> __func__, p->name, class, nsibling, err);
> - goto out_unlock;
> + return err;
> }
> }
> }
>
> -out_unlock:
> - mutex_unlock(&i915->drm.struct_mutex);
> - return err;
> + return 0;
> }
>
> int intel_execlists_live_selftests(struct drm_i915_private *i915)
> diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
> index ed44333d9e20..a10990b1a72c 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
> @@ -704,9 +704,7 @@ static int live_dirty_whitelist(void *arg)
>
> wakeref = intel_runtime_pm_get(&i915->runtime_pm);
>
> - mutex_unlock(&i915->drm.struct_mutex);
> file = mock_file(i915);
> - mutex_lock(&i915->drm.struct_mutex);
> if (IS_ERR(file)) {
> err = PTR_ERR(file);
> goto out_rpm;
> @@ -728,9 +726,7 @@ static int live_dirty_whitelist(void *arg)
> }
>
> out_file:
> - mutex_unlock(&i915->drm.struct_mutex);
> mock_file_free(i915, file);
> - mutex_lock(&i915->drm.struct_mutex);
> out_rpm:
> intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> return err;
> @@ -1264,14 +1260,9 @@ int intel_workarounds_live_selftests(struct drm_i915_private *i915)
> SUBTEST(live_gpu_reset_workarounds),
> SUBTEST(live_engine_reset_workarounds),
> };
> - int err;
>
> if (intel_gt_is_wedged(&i915->gt))
> return 0;
>
> - mutex_lock(&i915->drm.struct_mutex);
> - err = i915_subtests(tests, i915);
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> - return err;
> + return i915_subtests(tests, i915);
> }
> diff --git a/drivers/gpu/drm/i915/gt/uc/selftest_guc.c b/drivers/gpu/drm/i915/gt/uc/selftest_guc.c
> index bba0eafe1cdb..f927f851aadf 100644
> --- a/drivers/gpu/drm/i915/gt/uc/selftest_guc.c
> +++ b/drivers/gpu/drm/i915/gt/uc/selftest_guc.c
> @@ -116,7 +116,6 @@ static int igt_guc_clients(void *args)
> int err = 0;
>
> GEM_BUG_ON(!HAS_GT_UC(dev_priv));
> - mutex_lock(&dev_priv->drm.struct_mutex);
> wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm);
>
> guc = &dev_priv->gt.uc.guc;
> @@ -190,7 +189,6 @@ static int igt_guc_clients(void *args)
> guc_clients_enable(guc);
> unlock:
> intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref);
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> return err;
> }
>
> @@ -208,7 +206,6 @@ static int igt_guc_doorbells(void *arg)
> u16 db_id;
>
> GEM_BUG_ON(!HAS_GT_UC(dev_priv));
> - mutex_lock(&dev_priv->drm.struct_mutex);
> wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm);
>
> guc = &dev_priv->gt.uc.guc;
> @@ -299,7 +296,6 @@ static int igt_guc_doorbells(void *arg)
> }
> unlock:
> intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref);
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> return err;
> }
>
>
Too boring to read in detail. Eyeballs okay and in asserts, compilers
and static analysers we trust. :)
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] 70+ messages in thread
* [PATCH 27/27] drm/i915: Drop struct_mutex from around GEM initialisation
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (25 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 26/27] drm/i915/selftests: Drop vestigal struct_mutex guards Chris Wilson
@ 2019-09-25 10:01 ` Chris Wilson
2019-09-25 12:56 ` Tvrtko Ursulin
2019-09-25 10:21 ` ✗ Fi.CI.CHECKPATCH: warning for series starting with [01/27] dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling) Patchwork
` (3 subsequent siblings)
30 siblings, 1 reply; 70+ messages in thread
From: Chris Wilson @ 2019-09-25 10:01 UTC (permalink / raw)
To: intel-gfx
We no longer need to placate lockdep by holding struct_mutex for our
initialisation, so don't.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/gem/i915_gem_pm.c | 2 --
drivers/gpu/drm/i915/i915_gem.c | 9 ---------
drivers/gpu/drm/i915/selftests/mock_gem_device.c | 7 -------
3 files changed, 18 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
index e04e4ff7b52e..d1ff854ee7ff 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
@@ -174,7 +174,6 @@ void i915_gem_resume(struct drm_i915_private *i915)
{
GEM_TRACE("\n");
- mutex_lock(&i915->drm.struct_mutex);
intel_uncore_forcewake_get(&i915->uncore, FORCEWAKE_ALL);
if (intel_gt_init_hw(&i915->gt))
@@ -198,7 +197,6 @@ void i915_gem_resume(struct drm_i915_private *i915)
out_unlock:
intel_uncore_forcewake_put(&i915->uncore, FORCEWAKE_ALL);
- mutex_unlock(&i915->drm.struct_mutex);
return;
err_wedged:
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 8814db7023db..a5ef25b5b577 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1246,7 +1246,6 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
* we hold the forcewake during initialisation these problems
* just magically go away.
*/
- mutex_lock(&dev_priv->drm.struct_mutex);
intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL);
ret = i915_init_ggtt(dev_priv);
@@ -1316,7 +1315,6 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
goto err_gt;
intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL);
- mutex_unlock(&dev_priv->drm.struct_mutex);
return 0;
@@ -1327,15 +1325,11 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
* driver doesn't explode during runtime.
*/
err_gt:
- mutex_unlock(&dev_priv->drm.struct_mutex);
-
intel_gt_set_wedged(&dev_priv->gt);
i915_gem_suspend(dev_priv);
i915_gem_suspend_late(dev_priv);
i915_gem_drain_workqueue(dev_priv);
-
- mutex_lock(&dev_priv->drm.struct_mutex);
err_init_hw:
intel_uc_fini_hw(&dev_priv->gt.uc);
err_uc_init:
@@ -1350,7 +1344,6 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
intel_gt_driver_release(&dev_priv->gt);
err_unlock:
intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL);
- mutex_unlock(&dev_priv->drm.struct_mutex);
if (ret != -EIO) {
intel_uc_cleanup_firmwares(&dev_priv->gt.uc);
@@ -1403,10 +1396,8 @@ void i915_gem_driver_remove(struct drm_i915_private *dev_priv)
/* Flush any outstanding unpin_work. */
i915_gem_drain_workqueue(dev_priv);
- mutex_lock(&dev_priv->drm.struct_mutex);
intel_uc_fini_hw(&dev_priv->gt.uc);
intel_uc_fini(&dev_priv->gt.uc);
- mutex_unlock(&dev_priv->drm.struct_mutex);
i915_gem_drain_freed_objects(dev_priv);
}
diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
index 0ed21a7b8682..7950eb9825ed 100644
--- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
+++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
@@ -68,10 +68,7 @@ static void mock_device_release(struct drm_device *dev)
drain_workqueue(i915->wq);
i915_gem_drain_freed_objects(i915);
- mutex_lock(&i915->drm.struct_mutex);
mock_fini_ggtt(&i915->ggtt);
- mutex_unlock(&i915->drm.struct_mutex);
-
destroy_workqueue(i915->wq);
i915_gemfs_fini(i915);
@@ -180,8 +177,6 @@ struct drm_i915_private *mock_gem_device(void)
intel_timelines_init(i915);
- mutex_lock(&i915->drm.struct_mutex);
-
mock_init_ggtt(i915, &i915->ggtt);
mkwrite_device_info(i915)->engine_mask = BIT(0);
@@ -198,7 +193,6 @@ struct drm_i915_private *mock_gem_device(void)
goto err_context;
intel_engines_driver_register(i915);
- mutex_unlock(&i915->drm.struct_mutex);
WARN_ON(i915_gemfs_init(i915));
@@ -209,7 +203,6 @@ struct drm_i915_private *mock_gem_device(void)
err_engine:
mock_engine_free(i915->engine[RCS0]);
err_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
intel_timelines_fini(i915);
destroy_workqueue(i915->wq);
err_drv:
--
2.23.0
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 70+ messages in thread* Re: [PATCH 27/27] drm/i915: Drop struct_mutex from around GEM initialisation
2019-09-25 10:01 ` [PATCH 27/27] drm/i915: Drop struct_mutex from around GEM initialisation Chris Wilson
@ 2019-09-25 12:56 ` Tvrtko Ursulin
0 siblings, 0 replies; 70+ messages in thread
From: Tvrtko Ursulin @ 2019-09-25 12:56 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On 25/09/2019 11:01, Chris Wilson wrote:
> We no longer need to placate lockdep by holding struct_mutex for our
> initialisation, so don't.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/gem/i915_gem_pm.c | 2 --
> drivers/gpu/drm/i915/i915_gem.c | 9 ---------
> drivers/gpu/drm/i915/selftests/mock_gem_device.c | 7 -------
> 3 files changed, 18 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
> index e04e4ff7b52e..d1ff854ee7ff 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
> @@ -174,7 +174,6 @@ void i915_gem_resume(struct drm_i915_private *i915)
> {
> GEM_TRACE("\n");
>
> - mutex_lock(&i915->drm.struct_mutex);
> intel_uncore_forcewake_get(&i915->uncore, FORCEWAKE_ALL);
>
> if (intel_gt_init_hw(&i915->gt))
> @@ -198,7 +197,6 @@ void i915_gem_resume(struct drm_i915_private *i915)
>
> out_unlock:
> intel_uncore_forcewake_put(&i915->uncore, FORCEWAKE_ALL);
> - mutex_unlock(&i915->drm.struct_mutex);
> return;
>
> err_wedged:
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index 8814db7023db..a5ef25b5b577 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -1246,7 +1246,6 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
> * we hold the forcewake during initialisation these problems
> * just magically go away.
> */
> - mutex_lock(&dev_priv->drm.struct_mutex);
> intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL);
>
> ret = i915_init_ggtt(dev_priv);
> @@ -1316,7 +1315,6 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
> goto err_gt;
>
> intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL);
> - mutex_unlock(&dev_priv->drm.struct_mutex);
>
> return 0;
>
> @@ -1327,15 +1325,11 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
> * driver doesn't explode during runtime.
> */
> err_gt:
> - mutex_unlock(&dev_priv->drm.struct_mutex);
> -
> intel_gt_set_wedged(&dev_priv->gt);
> i915_gem_suspend(dev_priv);
> i915_gem_suspend_late(dev_priv);
>
> i915_gem_drain_workqueue(dev_priv);
> -
> - mutex_lock(&dev_priv->drm.struct_mutex);
> err_init_hw:
> intel_uc_fini_hw(&dev_priv->gt.uc);
> err_uc_init:
> @@ -1350,7 +1344,6 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
> intel_gt_driver_release(&dev_priv->gt);
> err_unlock:
> intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL);
> - mutex_unlock(&dev_priv->drm.struct_mutex);
>
> if (ret != -EIO) {
> intel_uc_cleanup_firmwares(&dev_priv->gt.uc);
> @@ -1403,10 +1396,8 @@ void i915_gem_driver_remove(struct drm_i915_private *dev_priv)
> /* Flush any outstanding unpin_work. */
> i915_gem_drain_workqueue(dev_priv);
>
> - mutex_lock(&dev_priv->drm.struct_mutex);
> intel_uc_fini_hw(&dev_priv->gt.uc);
> intel_uc_fini(&dev_priv->gt.uc);
> - mutex_unlock(&dev_priv->drm.struct_mutex);
>
> i915_gem_drain_freed_objects(dev_priv);
> }
> diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
> index 0ed21a7b8682..7950eb9825ed 100644
> --- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
> +++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
> @@ -68,10 +68,7 @@ static void mock_device_release(struct drm_device *dev)
> drain_workqueue(i915->wq);
> i915_gem_drain_freed_objects(i915);
>
> - mutex_lock(&i915->drm.struct_mutex);
> mock_fini_ggtt(&i915->ggtt);
> - mutex_unlock(&i915->drm.struct_mutex);
> -
> destroy_workqueue(i915->wq);
>
> i915_gemfs_fini(i915);
> @@ -180,8 +177,6 @@ struct drm_i915_private *mock_gem_device(void)
>
> intel_timelines_init(i915);
>
> - mutex_lock(&i915->drm.struct_mutex);
> -
> mock_init_ggtt(i915, &i915->ggtt);
>
> mkwrite_device_info(i915)->engine_mask = BIT(0);
> @@ -198,7 +193,6 @@ struct drm_i915_private *mock_gem_device(void)
> goto err_context;
>
> intel_engines_driver_register(i915);
> - mutex_unlock(&i915->drm.struct_mutex);
>
> WARN_ON(i915_gemfs_init(i915));
>
> @@ -209,7 +203,6 @@ struct drm_i915_private *mock_gem_device(void)
> err_engine:
> mock_engine_free(i915->engine[RCS0]);
> err_unlock:
> - mutex_unlock(&i915->drm.struct_mutex);
> intel_timelines_fini(i915);
> destroy_workqueue(i915->wq);
> err_drv:
>
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] 70+ messages in thread
* ✗ Fi.CI.CHECKPATCH: warning for series starting with [01/27] dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling)
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (26 preceding siblings ...)
2019-09-25 10:01 ` [PATCH 27/27] drm/i915: Drop struct_mutex from around GEM initialisation Chris Wilson
@ 2019-09-25 10:21 ` Patchwork
2019-09-25 10:34 ` ✗ Fi.CI.SPARSE: " Patchwork
` (2 subsequent siblings)
30 siblings, 0 replies; 70+ messages in thread
From: Patchwork @ 2019-09-25 10:21 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: series starting with [01/27] dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling)
URL : https://patchwork.freedesktop.org/series/67206/
State : warning
== Summary ==
$ dim checkpatch origin/drm-tip
00c4ac0a46d7 dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling)
-:14: ERROR:GIT_COMMIT_ID: Please use git commit description style 'commit <12+ chars of sha1> ("<title line>")' - ie: 'commit 0fc89b6802ba ("dma-fence: Simply wrap dma_fence_signal_locked with dma_fence_signal")'
#14:
See also 0fc89b6802ba ("dma-fence: Simply wrap dma_fence_signal_locked
total: 1 errors, 0 warnings, 0 checks, 24 lines checked
0a05fd8c35a2 drm/mm: Pack allocated/scanned boolean into a bitfield
a3a715619f6e drm/i915: Only track bound elements of the GTT
11b60deee8dd drm/i915: Mark up address spaces that may need to allocate
250d2e27b808 drm/i915: Pull i915_vma_pin under the vm->mutex
6a9dabdbb8e1 drm/i915: Push the i915_active.retire into a worker
6c93d293a214 drm/i915: Coordinate i915_active with its own mutex
-:1298: CHECK:UNCOMMENTED_DEFINITION: struct mutex definition without comment
#1298: FILE: drivers/gpu/drm/i915/i915_active_types.h:49:
+ struct mutex mutex;
total: 0 errors, 0 warnings, 1 checks, 1441 lines checked
65fe69d37a45 drm/i915: Move idle barrier cleanup into engine-pm
dd53487769e5 drm/i915: Drop struct_mutex from around i915_retire_requests()
6c295f5e8411 drm/i915: Remove the GEM idle worker
75a7caab211c drm/i915: Merge wait_for_timelines with retire_request
fe4bd075906e drm/i915: Move request runtime management onto gt
-:218: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#218:
new file mode 100644
-:223: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#223: FILE: drivers/gpu/drm/i915/gt/intel_gt_requests.c:1:
+/*
-:224: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#224: FILE: drivers/gpu/drm/i915/gt/intel_gt_requests.c:2:
+ * SPDX-License-Identifier: MIT
-:352: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#352: FILE: drivers/gpu/drm/i915/gt/intel_gt_requests.h:1:
+/*
-:353: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#353: FILE: drivers/gpu/drm/i915/gt/intel_gt_requests.h:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 5 warnings, 0 checks, 709 lines checked
492045d0f8c4 drm/i915: Move global activity tracking from GEM to GT
fecb6b0749fd drm/i915: Expose engine properties via sysfs
-:67: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#67:
new file mode 100644
-:72: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#72: FILE: drivers/gpu/drm/i915/gt/intel_engine_sysfs.c:1:
+/*
-:73: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#73: FILE: drivers/gpu/drm/i915/gt/intel_engine_sysfs.c:2:
+ * SPDX-License-Identifier: MIT
-:197: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#197: FILE: drivers/gpu/drm/i915/gt/intel_engine_sysfs.h:1:
+/*
-:198: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#198: FILE: drivers/gpu/drm/i915/gt/intel_engine_sysfs.h:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 5 warnings, 0 checks, 159 lines checked
35cb1337abaf drm/i915/execlists: Force preemption
4e74c6ea7848 drm/i915: Mark up "sentinel" requests
0efddfe0da7b drm/i915/execlists: Cancel banned contexts on schedule-out
58ab43d6168b drm/i915: Cancel non-persistent contexts on close
-:282: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#282:
new file mode 100644
-:287: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#287: FILE: drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c:1:
+/*
-:288: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#288: FILE: drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c:2:
+ * SPDX-License-Identifier: MIT
-:349: WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#349: FILE: drivers/gpu/drm/i915/gt/intel_engine_heartbeat.h:1:
+/*
-:350: WARNING:SPDX_LICENSE_TAG: Misplaced SPDX-License-Identifier tag - use line 1 instead
#350: FILE: drivers/gpu/drm/i915/gt/intel_engine_heartbeat.h:2:
+ * SPDX-License-Identifier: MIT
total: 0 errors, 5 warnings, 0 checks, 319 lines checked
437c1e8cf922 drm/i915: Replace hangcheck by heartbeats
-:229: WARNING:EMBEDDED_FUNCTION_NAME: Prefer using '"%s...", __func__' to using 'heartbeat', this function's name, in a string
#229: FILE: drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c:69:
+ "%s heartbeat not ticking\n",
-:246: WARNING:EMBEDDED_FUNCTION_NAME: Prefer using '"%s...", __func__' to using 'heartbeat', this function's name, in a string
#246: FILE: drivers/gpu/drm/i915/gt/intel_engine_heartbeat.c:86:
+ "stopped heartbeat on %s",
-:533: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#533:
deleted file mode 100644
total: 0 errors, 3 warnings, 0 checks, 669 lines checked
c311dc62779d drm/i915: Remove logical HW ID
a699c736bb88 drm/i915: Move context management under GEM
f42f4f34e168 drm/i915/overlay: Drop struct_mutex guard
d8a7071ed5a6 drm/i915: Drop struct_mutex guard from debugfs/framebuffer_info
5da4a6f1dd1a drm/i915: Remove struct_mutex guard for debugfs/opregion
2c47782ba04b drm/i915: Drop struct_mutex from suspend state save/restore
bf1708691cd3 drm/i915/selftests: Drop vestigal struct_mutex guards
a13c77413805 drm/i915: Drop struct_mutex from around GEM initialisation
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* ✗ Fi.CI.SPARSE: warning for series starting with [01/27] dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling)
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (27 preceding siblings ...)
2019-09-25 10:21 ` ✗ Fi.CI.CHECKPATCH: warning for series starting with [01/27] dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling) Patchwork
@ 2019-09-25 10:34 ` Patchwork
2019-09-25 10:51 ` ✓ Fi.CI.BAT: success " Patchwork
2019-09-26 0:56 ` ✗ Fi.CI.IGT: failure " Patchwork
30 siblings, 0 replies; 70+ messages in thread
From: Patchwork @ 2019-09-25 10:34 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: series starting with [01/27] dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling)
URL : https://patchwork.freedesktop.org/series/67206/
State : warning
== Summary ==
$ dim sparse origin/drm-tip
Sparse version: v0.6.0
Commit: dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling)
Okay!
Commit: drm/mm: Pack allocated/scanned boolean into a bitfield
Okay!
Commit: drm/i915: Only track bound elements of the GTT
Okay!
Commit: drm/i915: Mark up address spaces that may need to allocate
Okay!
Commit: drm/i915: Pull i915_vma_pin under the vm->mutex
Okay!
Commit: drm/i915: Push the i915_active.retire into a worker
Okay!
Commit: drm/i915: Coordinate i915_active with its own mutex
Okay!
Commit: drm/i915: Move idle barrier cleanup into engine-pm
Okay!
Commit: drm/i915: Drop struct_mutex from around i915_retire_requests()
Okay!
Commit: drm/i915: Remove the GEM idle worker
Okay!
Commit: drm/i915: Merge wait_for_timelines with retire_request
Okay!
Commit: drm/i915: Move request runtime management onto gt
Okay!
Commit: drm/i915: Move global activity tracking from GEM to GT
Okay!
Commit: drm/i915: Expose engine properties via sysfs
Okay!
Commit: drm/i915/execlists: Force preemption
Okay!
Commit: drm/i915: Mark up "sentinel" requests
Okay!
Commit: drm/i915/execlists: Cancel banned contexts on schedule-out
Okay!
Commit: drm/i915: Cancel non-persistent contexts on close
Okay!
Commit: drm/i915: Replace hangcheck by heartbeats
Okay!
Commit: drm/i915: Remove logical HW ID
Okay!
Commit: drm/i915: Move context management under GEM
+drivers/gpu/drm/i915/selftests/i915_gem_gtt.c:1803:12: expected struct i915_address_space *vm
+drivers/gpu/drm/i915/selftests/i915_gem_gtt.c:1803:12: got struct i915_address_space [noderef] <asn:4> *vm
+drivers/gpu/drm/i915/selftests/i915_gem_gtt.c:1803:12: warning: incorrect type in assignment (different address spaces)
Commit: drm/i915/overlay: Drop struct_mutex guard
Okay!
Commit: drm/i915: Drop struct_mutex guard from debugfs/framebuffer_info
Okay!
Commit: drm/i915: Remove struct_mutex guard for debugfs/opregion
Okay!
Commit: drm/i915: Drop struct_mutex from suspend state save/restore
Okay!
Commit: drm/i915/selftests: Drop vestigal struct_mutex guards
Okay!
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* ✓ Fi.CI.BAT: success for series starting with [01/27] dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling)
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (28 preceding siblings ...)
2019-09-25 10:34 ` ✗ Fi.CI.SPARSE: " Patchwork
@ 2019-09-25 10:51 ` Patchwork
2019-09-26 0:56 ` ✗ Fi.CI.IGT: failure " Patchwork
30 siblings, 0 replies; 70+ messages in thread
From: Patchwork @ 2019-09-25 10:51 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: series starting with [01/27] dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling)
URL : https://patchwork.freedesktop.org/series/67206/
State : success
== Summary ==
CI Bug Log - changes from CI_DRM_6954 -> Patchwork_14527
====================================================
Summary
-------
**SUCCESS**
No regressions found.
External URL: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/
Known issues
------------
Here are the changes found in Patchwork_14527 that come from known issues:
### IGT changes ###
#### Issues hit ####
* igt@i915_module_load@reload:
- fi-blb-e6850: [PASS][1] -> [INCOMPLETE][2] ([fdo#107718])
[1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/fi-blb-e6850/igt@i915_module_load@reload.html
[2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/fi-blb-e6850/igt@i915_module_load@reload.html
#### Possible fixes ####
* igt@gem_exec_gttfill@basic:
- fi-apl-guc: [DMESG-WARN][3] ([fdo#109385] / [fdo#111652 ]) -> [PASS][4]
[3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/fi-apl-guc/igt@gem_exec_gttfill@basic.html
[4]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/fi-apl-guc/igt@gem_exec_gttfill@basic.html
* igt@kms_frontbuffer_tracking@basic:
- fi-hsw-peppy: [DMESG-WARN][5] ([fdo#102614]) -> [PASS][6]
[5]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/fi-hsw-peppy/igt@kms_frontbuffer_tracking@basic.html
[6]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/fi-hsw-peppy/igt@kms_frontbuffer_tracking@basic.html
{name}: This element is suppressed. This means it is ignored when computing
the status of the difference (SUCCESS, WARNING, or FAILURE).
[fdo#102614]: https://bugs.freedesktop.org/show_bug.cgi?id=102614
[fdo#107713]: https://bugs.freedesktop.org/show_bug.cgi?id=107713
[fdo#107718]: https://bugs.freedesktop.org/show_bug.cgi?id=107718
[fdo#109385]: https://bugs.freedesktop.org/show_bug.cgi?id=109385
[fdo#111600]: https://bugs.freedesktop.org/show_bug.cgi?id=111600
[fdo#111652 ]: https://bugs.freedesktop.org/show_bug.cgi?id=111652
Participating hosts (51 -> 44)
------------------------------
Additional (1): fi-tgl-u
Missing (8): fi-ilk-m540 fi-hsw-4200u fi-byt-j1900 fi-byt-squawks fi-bsw-cyan fi-icl-y fi-byt-clapper fi-bdw-samus
Build changes
-------------
* CI: CI-20190529 -> None
* Linux: CI_DRM_6954 -> Patchwork_14527
CI-20190529: 20190529
CI_DRM_6954: 80fa0e042cdb3da59d2c738db61b8da78cf29e0d @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_5202: 3499c5eb17054e2abd88023fe962768140d24302 @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
Patchwork_14527: a13c77413805d15cafb23dc576cd445a2a6f50ba @ git://anongit.freedesktop.org/gfx-ci/linux
== Linux commits ==
a13c77413805 drm/i915: Drop struct_mutex from around GEM initialisation
bf1708691cd3 drm/i915/selftests: Drop vestigal struct_mutex guards
2c47782ba04b drm/i915: Drop struct_mutex from suspend state save/restore
5da4a6f1dd1a drm/i915: Remove struct_mutex guard for debugfs/opregion
d8a7071ed5a6 drm/i915: Drop struct_mutex guard from debugfs/framebuffer_info
f42f4f34e168 drm/i915/overlay: Drop struct_mutex guard
a699c736bb88 drm/i915: Move context management under GEM
c311dc62779d drm/i915: Remove logical HW ID
437c1e8cf922 drm/i915: Replace hangcheck by heartbeats
58ab43d6168b drm/i915: Cancel non-persistent contexts on close
0efddfe0da7b drm/i915/execlists: Cancel banned contexts on schedule-out
4e74c6ea7848 drm/i915: Mark up "sentinel" requests
35cb1337abaf drm/i915/execlists: Force preemption
fecb6b0749fd drm/i915: Expose engine properties via sysfs
492045d0f8c4 drm/i915: Move global activity tracking from GEM to GT
fe4bd075906e drm/i915: Move request runtime management onto gt
75a7caab211c drm/i915: Merge wait_for_timelines with retire_request
6c295f5e8411 drm/i915: Remove the GEM idle worker
dd53487769e5 drm/i915: Drop struct_mutex from around i915_retire_requests()
65fe69d37a45 drm/i915: Move idle barrier cleanup into engine-pm
6c93d293a214 drm/i915: Coordinate i915_active with its own mutex
6a9dabdbb8e1 drm/i915: Push the i915_active.retire into a worker
250d2e27b808 drm/i915: Pull i915_vma_pin under the vm->mutex
11b60deee8dd drm/i915: Mark up address spaces that may need to allocate
a3a715619f6e drm/i915: Only track bound elements of the GTT
0a05fd8c35a2 drm/mm: Pack allocated/scanned boolean into a bitfield
00c4ac0a46d7 dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling)
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/index.html
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread* ✗ Fi.CI.IGT: failure for series starting with [01/27] dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling)
2019-09-25 10:01 struct_mutex is over the hill and far away Chris Wilson
` (29 preceding siblings ...)
2019-09-25 10:51 ` ✓ Fi.CI.BAT: success " Patchwork
@ 2019-09-26 0:56 ` Patchwork
30 siblings, 0 replies; 70+ messages in thread
From: Patchwork @ 2019-09-26 0:56 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: series starting with [01/27] dma-fence: Serialise signal enabling (dma_fence_enable_sw_signaling)
URL : https://patchwork.freedesktop.org/series/67206/
State : failure
== Summary ==
CI Bug Log - changes from CI_DRM_6954_full -> Patchwork_14527_full
====================================================
Summary
-------
**FAILURE**
Serious unknown changes coming with Patchwork_14527_full absolutely need to be
verified manually.
If you think the reported changes have nothing to do with the changes
introduced in Patchwork_14527_full, please notify your bug team to allow them
to document this new failure mode, which will reduce false positives in CI.
Possible new issues
-------------------
Here are the unknown changes that may have been introduced in Patchwork_14527_full:
### IGT changes ###
#### Possible regressions ####
* igt@gem_softpin@softpin:
- shard-skl: [PASS][1] -> [TIMEOUT][2]
[1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-skl10/igt@gem_softpin@softpin.html
[2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-skl10/igt@gem_softpin@softpin.html
* igt@gem_tiled_swapping@non-threaded:
- shard-hsw: [PASS][3] -> [DMESG-FAIL][4]
[3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-hsw6/igt@gem_tiled_swapping@non-threaded.html
[4]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-hsw6/igt@gem_tiled_swapping@non-threaded.html
* igt@i915_hangman@error-state-capture-bcs0:
- shard-apl: [PASS][5] -> [FAIL][6] +2 similar issues
[5]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-apl4/igt@i915_hangman@error-state-capture-bcs0.html
[6]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-apl7/igt@i915_hangman@error-state-capture-bcs0.html
- shard-iclb: [PASS][7] -> [FAIL][8] +2 similar issues
[7]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-iclb1/igt@i915_hangman@error-state-capture-bcs0.html
[8]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-iclb6/igt@i915_hangman@error-state-capture-bcs0.html
* igt@i915_hangman@error-state-capture-rcs0:
- shard-skl: [PASS][9] -> [FAIL][10] +3 similar issues
[9]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-skl7/igt@i915_hangman@error-state-capture-rcs0.html
[10]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-skl3/igt@i915_hangman@error-state-capture-rcs0.html
- shard-glk: [PASS][11] -> [FAIL][12] +3 similar issues
[11]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-glk5/igt@i915_hangman@error-state-capture-rcs0.html
[12]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-glk5/igt@i915_hangman@error-state-capture-rcs0.html
- shard-iclb: NOTRUN -> [FAIL][13]
[13]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-iclb1/igt@i915_hangman@error-state-capture-rcs0.html
* igt@i915_hangman@error-state-capture-vcs0:
- shard-kbl: [PASS][14] -> [FAIL][15] +4 similar issues
[14]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-kbl6/igt@i915_hangman@error-state-capture-vcs0.html
[15]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-kbl7/igt@i915_hangman@error-state-capture-vcs0.html
* igt@i915_selftest@live_workarounds:
- shard-kbl: [PASS][16] -> [DMESG-FAIL][17]
[16]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-kbl6/igt@i915_selftest@live_workarounds.html
[17]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-kbl1/igt@i915_selftest@live_workarounds.html
Known issues
------------
Here are the changes found in Patchwork_14527_full that come from known issues:
### IGT changes ###
#### Issues hit ####
* igt@gem_ctx_isolation@rcs0-s3:
- shard-kbl: [PASS][18] -> [DMESG-WARN][19] ([fdo#108566]) +1 similar issue
[18]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-kbl4/igt@gem_ctx_isolation@rcs0-s3.html
[19]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-kbl3/igt@gem_ctx_isolation@rcs0-s3.html
* igt@gem_eio@in-flight-contexts-1us:
- shard-hsw: [PASS][20] -> [FAIL][21] ([fdo#105957]) +1 similar issue
[20]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-hsw8/igt@gem_eio@in-flight-contexts-1us.html
[21]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-hsw4/igt@gem_eio@in-flight-contexts-1us.html
* igt@gem_eio@in-flight-contexts-immediate:
- shard-snb: [PASS][22] -> [FAIL][23] ([fdo#105957])
[22]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-snb5/igt@gem_eio@in-flight-contexts-immediate.html
[23]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-snb4/igt@gem_eio@in-flight-contexts-immediate.html
* igt@gem_exec_parallel@bcs0-fds:
- shard-apl: [PASS][24] -> [INCOMPLETE][25] ([fdo#103927]) +2 similar issues
[24]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-apl1/igt@gem_exec_parallel@bcs0-fds.html
[25]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-apl6/igt@gem_exec_parallel@bcs0-fds.html
* igt@gem_exec_schedule@reorder-wide-bsd:
- shard-iclb: [PASS][26] -> [SKIP][27] ([fdo#111325]) +7 similar issues
[26]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-iclb5/igt@gem_exec_schedule@reorder-wide-bsd.html
[27]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-iclb2/igt@gem_exec_schedule@reorder-wide-bsd.html
* igt@gem_softpin@noreloc-s3:
- shard-apl: [PASS][28] -> [DMESG-WARN][29] ([fdo#108566]) +2 similar issues
[28]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-apl8/igt@gem_softpin@noreloc-s3.html
[29]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-apl2/igt@gem_softpin@noreloc-s3.html
* igt@i915_selftest@live_gem_contexts:
- shard-hsw: [PASS][30] -> [DMESG-FAIL][31] ([fdo#111692])
[30]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-hsw6/igt@i915_selftest@live_gem_contexts.html
[31]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-hsw6/igt@i915_selftest@live_gem_contexts.html
* igt@kms_flip@plain-flip-fb-recreate:
- shard-skl: [PASS][32] -> [FAIL][33] ([fdo#100368])
[32]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-skl4/igt@kms_flip@plain-flip-fb-recreate.html
[33]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-skl3/igt@kms_flip@plain-flip-fb-recreate.html
* igt@kms_frontbuffer_tracking@fbcpsr-1p-primscrn-pri-indfb-draw-pwrite:
- shard-iclb: [PASS][34] -> [FAIL][35] ([fdo#103167]) +2 similar issues
[34]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-iclb3/igt@kms_frontbuffer_tracking@fbcpsr-1p-primscrn-pri-indfb-draw-pwrite.html
[35]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-iclb8/igt@kms_frontbuffer_tracking@fbcpsr-1p-primscrn-pri-indfb-draw-pwrite.html
* igt@kms_plane_alpha_blend@pipe-a-constant-alpha-min:
- shard-skl: [PASS][36] -> [FAIL][37] ([fdo#108145])
[36]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-skl7/igt@kms_plane_alpha_blend@pipe-a-constant-alpha-min.html
[37]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-skl3/igt@kms_plane_alpha_blend@pipe-a-constant-alpha-min.html
* igt@kms_psr@psr2_cursor_render:
- shard-iclb: [PASS][38] -> [SKIP][39] ([fdo#109441]) +2 similar issues
[38]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-iclb2/igt@kms_psr@psr2_cursor_render.html
[39]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-iclb4/igt@kms_psr@psr2_cursor_render.html
* igt@perf@blocking:
- shard-skl: [PASS][40] -> [FAIL][41] ([fdo#110728])
[40]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-skl9/igt@perf@blocking.html
[41]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-skl5/igt@perf@blocking.html
* igt@prime_busy@hang-bsd2:
- shard-iclb: [PASS][42] -> [SKIP][43] ([fdo#109276]) +27 similar issues
[42]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-iclb2/igt@prime_busy@hang-bsd2.html
[43]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-iclb6/igt@prime_busy@hang-bsd2.html
#### Possible fixes ####
* igt@gem_ctx_shared@exec-single-timeline-bsd:
- shard-iclb: [SKIP][44] ([fdo#110841]) -> [PASS][45]
[44]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-iclb1/igt@gem_ctx_shared@exec-single-timeline-bsd.html
[45]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-iclb8/igt@gem_ctx_shared@exec-single-timeline-bsd.html
* igt@gem_exec_balancer@smoke:
- shard-iclb: [SKIP][46] ([fdo#110854]) -> [PASS][47]
[46]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-iclb7/igt@gem_exec_balancer@smoke.html
[47]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-iclb4/igt@gem_exec_balancer@smoke.html
* igt@gem_exec_parallel@contexts:
- shard-skl: [DMESG-WARN][48] ([fdo#106107]) -> [PASS][49]
[48]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-skl10/igt@gem_exec_parallel@contexts.html
[49]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-skl10/igt@gem_exec_parallel@contexts.html
* igt@gem_exec_schedule@preempt-queue-bsd:
- shard-iclb: [SKIP][50] ([fdo#111325]) -> [PASS][51] +4 similar issues
[50]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-iclb1/igt@gem_exec_schedule@preempt-queue-bsd.html
[51]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-iclb6/igt@gem_exec_schedule@preempt-queue-bsd.html
* igt@gem_exec_schedule@preempt-queue-bsd1:
- shard-iclb: [SKIP][52] ([fdo#109276]) -> [PASS][53] +11 similar issues
[52]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-iclb8/igt@gem_exec_schedule@preempt-queue-bsd1.html
[53]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-iclb1/igt@gem_exec_schedule@preempt-queue-bsd1.html
* igt@kms_cursor_edge_walk@pipe-c-256x256-top-edge:
- shard-apl: [TIMEOUT][54] ([fdo#111800]) -> [PASS][55]
[54]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-apl7/igt@kms_cursor_edge_walk@pipe-c-256x256-top-edge.html
[55]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-apl5/igt@kms_cursor_edge_walk@pipe-c-256x256-top-edge.html
* igt@kms_flip@2x-flip-vs-expired-vblank-interruptible:
- shard-glk: [FAIL][56] ([fdo#105363]) -> [PASS][57]
[56]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-glk1/igt@kms_flip@2x-flip-vs-expired-vblank-interruptible.html
[57]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-glk1/igt@kms_flip@2x-flip-vs-expired-vblank-interruptible.html
* igt@kms_flip@flip-vs-suspend-interruptible:
- shard-skl: [INCOMPLETE][58] ([fdo#109507]) -> [PASS][59]
[58]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-skl9/igt@kms_flip@flip-vs-suspend-interruptible.html
[59]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-skl5/igt@kms_flip@flip-vs-suspend-interruptible.html
* igt@kms_frontbuffer_tracking@fbc-badstride:
- shard-iclb: [FAIL][60] ([fdo#103167]) -> [PASS][61] +5 similar issues
[60]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-iclb5/igt@kms_frontbuffer_tracking@fbc-badstride.html
[61]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-iclb2/igt@kms_frontbuffer_tracking@fbc-badstride.html
* igt@kms_pipe_crc_basic@suspend-read-crc-pipe-a:
- shard-skl: [INCOMPLETE][62] ([fdo#104108]) -> [PASS][63] +1 similar issue
[62]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-skl6/igt@kms_pipe_crc_basic@suspend-read-crc-pipe-a.html
[63]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-skl2/igt@kms_pipe_crc_basic@suspend-read-crc-pipe-a.html
- shard-iclb: [DMESG-WARN][64] ([fdo#111764]) -> [PASS][65]
[64]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-iclb1/igt@kms_pipe_crc_basic@suspend-read-crc-pipe-a.html
[65]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-iclb7/igt@kms_pipe_crc_basic@suspend-read-crc-pipe-a.html
* igt@kms_plane_alpha_blend@pipe-b-coverage-7efc:
- shard-skl: [FAIL][66] ([fdo#108145] / [fdo#110403]) -> [PASS][67]
[66]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-skl10/igt@kms_plane_alpha_blend@pipe-b-coverage-7efc.html
[67]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-skl6/igt@kms_plane_alpha_blend@pipe-b-coverage-7efc.html
* igt@kms_psr@psr2_sprite_plane_move:
- shard-iclb: [SKIP][68] ([fdo#109441]) -> [PASS][69] +3 similar issues
[68]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-iclb6/igt@kms_psr@psr2_sprite_plane_move.html
[69]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-iclb2/igt@kms_psr@psr2_sprite_plane_move.html
* igt@kms_vblank@pipe-b-ts-continuation-suspend:
- shard-apl: [DMESG-WARN][70] ([fdo#108566]) -> [PASS][71] +1 similar issue
[70]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-apl1/igt@kms_vblank@pipe-b-ts-continuation-suspend.html
[71]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-apl1/igt@kms_vblank@pipe-b-ts-continuation-suspend.html
* igt@perf@gen8-unprivileged-single-ctx-counters:
- shard-skl: [DMESG-FAIL][72] -> [PASS][73]
[72]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-skl10/igt@perf@gen8-unprivileged-single-ctx-counters.html
[73]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-skl6/igt@perf@gen8-unprivileged-single-ctx-counters.html
* igt@perf_pmu@cpu-hotplug:
- shard-apl: [TIMEOUT][74] ([fdo#111546] / [fdo#111800]) -> [PASS][75]
[74]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-apl7/igt@perf_pmu@cpu-hotplug.html
[75]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-apl5/igt@perf_pmu@cpu-hotplug.html
#### Warnings ####
* igt@kms_content_protection@srm:
- shard-apl: [FAIL][76] ([fdo#110321]) -> [INCOMPLETE][77] ([fdo#103927])
[76]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-apl6/igt@kms_content_protection@srm.html
[77]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-apl7/igt@kms_content_protection@srm.html
* igt@kms_flip@2x-plain-flip-ts-check-interruptible:
- shard-apl: [SKIP][78] ([fdo#109271]) -> [INCOMPLETE][79] ([fdo#103927])
[78]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6954/shard-apl7/igt@kms_flip@2x-plain-flip-ts-check-interruptible.html
[79]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/shard-apl8/igt@kms_flip@2x-plain-flip-ts-check-interruptible.html
{name}: This element is suppressed. This means it is ignored when computing
the status of the difference (SUCCESS, WARNING, or FAILURE).
[fdo#100368]: https://bugs.freedesktop.org/show_bug.cgi?id=100368
[fdo#103167]: https://bugs.freedesktop.org/show_bug.cgi?id=103167
[fdo#103927]: https://bugs.freedesktop.org/show_bug.cgi?id=103927
[fdo#104108]: https://bugs.freedesktop.org/show_bug.cgi?id=104108
[fdo#105363]: https://bugs.freedesktop.org/show_bug.cgi?id=105363
[fdo#105957]: https://bugs.freedesktop.org/show_bug.cgi?id=105957
[fdo#106107]: https://bugs.freedesktop.org/show_bug.cgi?id=106107
[fdo#108145]: https://bugs.freedesktop.org/show_bug.cgi?id=108145
[fdo#108566]: https://bugs.freedesktop.org/show_bug.cgi?id=108566
[fdo#109271]: https://bugs.freedesktop.org/show_bug.cgi?id=109271
[fdo#109276]: https://bugs.freedesktop.org/show_bug.cgi?id=109276
[fdo#109441]: https://bugs.freedesktop.org/show_bug.cgi?id=109441
[fdo#109507]: https://bugs.freedesktop.org/show_bug.cgi?id=109507
[fdo#110321]: https://bugs.freedesktop.org/show_bug.cgi?id=110321
[fdo#110403]: https://bugs.freedesktop.org/show_bug.cgi?id=110403
[fdo#110548]: https://bugs.freedesktop.org/show_bug.cgi?id=110548
[fdo#110728]: https://bugs.freedesktop.org/show_bug.cgi?id=110728
[fdo#110841]: https://bugs.freedesktop.org/show_bug.cgi?id=110841
[fdo#110854]: https://bugs.freedesktop.org/show_bug.cgi?id=110854
[fdo#111325]: https://bugs.freedesktop.org/show_bug.cgi?id=111325
[fdo#111546]: https://bugs.freedesktop.org/show_bug.cgi?id=111546
[fdo#111692]: https://bugs.freedesktop.org/show_bug.cgi?id=111692
[fdo#111764]: https://bugs.freedesktop.org/show_bug.cgi?id=111764
[fdo#111800]: https://bugs.freedesktop.org/show_bug.cgi?id=111800
Participating hosts (9 -> 9)
------------------------------
No changes in participating hosts
Build changes
-------------
* CI: CI-20190529 -> None
* Linux: CI_DRM_6954 -> Patchwork_14527
CI-20190529: 20190529
CI_DRM_6954: 80fa0e042cdb3da59d2c738db61b8da78cf29e0d @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_5202: 3499c5eb17054e2abd88023fe962768140d24302 @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
Patchwork_14527: a13c77413805d15cafb23dc576cd445a2a6f50ba @ git://anongit.freedesktop.org/gfx-ci/linux
piglit_4509: fdc5a4ca11124ab8413c7988896eec4c97336694 @ git://anongit.freedesktop.org/piglit
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_14527/
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 70+ messages in thread