[PATCH v2 1/9] drm/i915: Mark up clflushes as belonging to an unordered timeline

[PATCH v2 1/9] drm/i915: Mark up clflushes as belonging to an unordered timeline

[Chris Wilson]

2 clflushes on two different objects are not ordered, and so do not
belong to the same timeline (context). Either we use a unique context
for each, or we reserve a special global context to mean unordered.
Ideally, we would reserve 0 to mean unordered (DMA_FENCE_NO_CONTEXT) to
have the same semantics everywhere.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
 drivers/gpu/drm/i915/i915_drv.h         | 2 ++
 drivers/gpu/drm/i915/i915_gem.c         | 2 +-
 drivers/gpu/drm/i915/i915_gem_clflush.c | 8 +-------
 drivers/gpu/drm/i915/i915_gem_clflush.h | 1 -
 4 files changed, 4 insertions(+), 9 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 1af4e6f5410c..9ff399fdf92b 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1511,6 +1511,8 @@ struct i915_gem_mm {
 	/** LRU list of objects with fence regs on them. */
 	struct list_head fence_list;
 
+	u64 unordered_timeline;
+
 	/* the indicator for dispatch video commands on two BSD rings */
 	atomic_t bsd_engine_dispatch_index;
 
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 33fb11cc5acc..68c0a1c3bf77 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -4746,7 +4746,7 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
 
 	mutex_lock(&dev_priv->drm.struct_mutex);
 
-	i915_gem_clflush_init(dev_priv);
+	dev_priv->mm.unordered_timeline = dma_fence_context_alloc(1);
 
 	if (!i915.enable_execlists) {
 		dev_priv->gt.resume = intel_legacy_submission_resume;
diff --git a/drivers/gpu/drm/i915/i915_gem_clflush.c b/drivers/gpu/drm/i915/i915_gem_clflush.c
index ffd01e02fe94..ffac7a1f0caf 100644
--- a/drivers/gpu/drm/i915/i915_gem_clflush.c
+++ b/drivers/gpu/drm/i915/i915_gem_clflush.c
@@ -27,7 +27,6 @@
 #include "i915_gem_clflush.h"
 
 static DEFINE_SPINLOCK(clflush_lock);
-static u64 clflush_context;
 
 struct clflush {
 	struct dma_fence dma; /* Must be first for dma_fence_free() */
@@ -157,7 +156,7 @@ void i915_gem_clflush_object(struct drm_i915_gem_object *obj,
 		dma_fence_init(&clflush->dma,
 			       &i915_clflush_ops,
 			       &clflush_lock,
-			       clflush_context,
+			       to_i915(obj->base.dev)->mm.unordered_timeline,
 			       0);
 		i915_sw_fence_init(&clflush->wait, i915_clflush_notify);
 
@@ -182,8 +181,3 @@ void i915_gem_clflush_object(struct drm_i915_gem_object *obj,
 		GEM_BUG_ON(obj->base.write_domain != I915_GEM_DOMAIN_CPU);
 	}
 }
-
-void i915_gem_clflush_init(struct drm_i915_private *i915)
-{
-	clflush_context = dma_fence_context_alloc(1);
-}
diff --git a/drivers/gpu/drm/i915/i915_gem_clflush.h b/drivers/gpu/drm/i915/i915_gem_clflush.h
index b62d61a2d15f..2455a7820937 100644
--- a/drivers/gpu/drm/i915/i915_gem_clflush.h
+++ b/drivers/gpu/drm/i915/i915_gem_clflush.h
@@ -28,7 +28,6 @@
 struct drm_i915_private;
 struct drm_i915_gem_object;
 
-void i915_gem_clflush_init(struct drm_i915_private *i915);
 void i915_gem_clflush_object(struct drm_i915_gem_object *obj,
 			     unsigned int flags);
 #define I915_CLFLUSH_FORCE BIT(0)
-- 
2.11.0

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH v2 2/9] drm/i915: Lift timeline ordering to await_dma_fence

[Chris Wilson]

Currently we filter out repeated use of the same timeline in the low
level i915_gem_request_await_request(), after having added the
dependency on the old request. However, we can lift this to
i915_gem_request_await_dma_fence() (before the dependency is added)
using the observation that requests along the same timeline are
explicitly ordered via i915_add_request (along with the dependencies).

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
 drivers/gpu/drm/i915/i915_gem_request.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index 313cdff7c6dd..31874a38752e 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -663,6 +663,7 @@ i915_gem_request_await_request(struct drm_i915_gem_request *to,
 	int ret;
 
 	GEM_BUG_ON(to == from);
+	GEM_BUG_ON(to->timeline == from->timeline);
 
 	if (to->engine->schedule) {
 		ret = i915_priotree_add_dependency(to->i915,
@@ -672,9 +673,6 @@ i915_gem_request_await_request(struct drm_i915_gem_request *to,
 			return ret;
 	}
 
-	if (to->timeline == from->timeline)
-		return 0;
-
 	if (to->engine == from->engine) {
 		ret = i915_sw_fence_await_sw_fence_gfp(&to->submit,
 						       &from->submit,
@@ -723,6 +721,13 @@ i915_gem_request_await_dma_fence(struct drm_i915_gem_request *req,
 	if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
 		return 0;
 
+	/* Requests on the same timeline are explicitly ordered, along with
+	 * their dependencies, by i915_add_request() which ensures that requests
+	 * are submitted in-order through each ring.
+	 */
+	if (fence->context == req->fence.context)
+		return 0;
+
 	if (dma_fence_is_i915(fence))
 		return i915_gem_request_await_request(req, to_request(fence));
 
-- 
2.11.0

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH v2 3/9] drm/i915: Make ptr_unpack_bits() more function-like

[Chris Wilson]

ptr_unpack_bits() is a function-like macro, as such it is meant to be
replaceable by a function. In this case, we should be passing in the
out-param as a pointer.

Bizarrely this does affect code generation:

function                                     old     new   delta
i915_gem_object_pin_map                      409     389     -20

An improvement(?) in this case, but one can't help wonder what
strict-aliasing optimisations we are preventing.

The generated code looks identical in using ptr_unpack_bits (no extra
motions to stack, the pointer and bits appear to be kept in registers),
the difference appears to be code ordering and with a reorder it is able
to use smaller forward jumps.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
 drivers/gpu/drm/i915/i915_gem.c   | 2 +-
 drivers/gpu/drm/i915/i915_utils.h | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 68c0a1c3bf77..d993a3cc8f97 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2612,7 +2612,7 @@ void *i915_gem_object_pin_map(struct drm_i915_gem_object *obj,
 	}
 	GEM_BUG_ON(!obj->mm.pages);
 
-	ptr = ptr_unpack_bits(obj->mm.mapping, has_type);
+	ptr = ptr_unpack_bits(obj->mm.mapping, &has_type);
 	if (ptr && has_type != type) {
 		if (pinned) {
 			ret = -EBUSY;
diff --git a/drivers/gpu/drm/i915/i915_utils.h b/drivers/gpu/drm/i915/i915_utils.h
index c5455d36b617..aca11aad5da7 100644
--- a/drivers/gpu/drm/i915/i915_utils.h
+++ b/drivers/gpu/drm/i915/i915_utils.h
@@ -77,7 +77,7 @@
 
 #define ptr_unpack_bits(ptr, bits) ({					\
 	unsigned long __v = (unsigned long)(ptr);			\
-	(bits) = __v & ~PAGE_MASK;					\
+	*(bits) = __v & ~PAGE_MASK;					\
 	(typeof(ptr))(__v & PAGE_MASK);					\
 })
 
-- 
2.11.0

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH v2 4/9] drm/i915: Redefine ptr_pack_bits() and friends

[Chris Wilson]

Rebrand the current (pointer | bits) pack/unpack utility macros as
explicit bit twiddling for PAGE_SIZE so that we can use the more
flexible underlying macros for different bits.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
 drivers/gpu/drm/i915/i915_cmd_parser.c |  2 +-
 drivers/gpu/drm/i915/i915_gem.c        |  6 +++---
 drivers/gpu/drm/i915/i915_utils.h      | 19 +++++++++++++------
 3 files changed, 17 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index 2a1a3347495a..f0cb22cc0dd6 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -1284,7 +1284,7 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 
 		if (*cmd == MI_BATCH_BUFFER_END) {
 			if (needs_clflush_after) {
-				void *ptr = ptr_mask_bits(shadow_batch_obj->mm.mapping);
+				void *ptr = page_mask_bits(shadow_batch_obj->mm.mapping);
 				drm_clflush_virt_range(ptr,
 						       (void *)(cmd + 1) - ptr);
 			}
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index d993a3cc8f97..e5f943ca5216 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2280,7 +2280,7 @@ void __i915_gem_object_put_pages(struct drm_i915_gem_object *obj,
 	if (obj->mm.mapping) {
 		void *ptr;
 
-		ptr = ptr_mask_bits(obj->mm.mapping);
+		ptr = page_mask_bits(obj->mm.mapping);
 		if (is_vmalloc_addr(ptr))
 			vunmap(ptr);
 		else
@@ -2612,7 +2612,7 @@ void *i915_gem_object_pin_map(struct drm_i915_gem_object *obj,
 	}
 	GEM_BUG_ON(!obj->mm.pages);
 
-	ptr = ptr_unpack_bits(obj->mm.mapping, &has_type);
+	ptr = page_unpack_bits(obj->mm.mapping, &has_type);
 	if (ptr && has_type != type) {
 		if (pinned) {
 			ret = -EBUSY;
@@ -2634,7 +2634,7 @@ void *i915_gem_object_pin_map(struct drm_i915_gem_object *obj,
 			goto err_unpin;
 		}
 
-		obj->mm.mapping = ptr_pack_bits(ptr, type);
+		obj->mm.mapping = page_pack_bits(ptr, type);
 	}
 
 out_unlock:
diff --git a/drivers/gpu/drm/i915/i915_utils.h b/drivers/gpu/drm/i915/i915_utils.h
index aca11aad5da7..f0500c65726d 100644
--- a/drivers/gpu/drm/i915/i915_utils.h
+++ b/drivers/gpu/drm/i915/i915_utils.h
@@ -70,20 +70,27 @@
 #define overflows_type(x, T) \
 	(sizeof(x) > sizeof(T) && (x) >> (sizeof(T) * BITS_PER_BYTE))
 
-#define ptr_mask_bits(ptr) ({						\
+#define ptr_mask_bits(ptr, n) ({					\
 	unsigned long __v = (unsigned long)(ptr);			\
-	(typeof(ptr))(__v & PAGE_MASK);					\
+	(typeof(ptr))(__v & -BIT(n));					\
 })
 
-#define ptr_unpack_bits(ptr, bits) ({					\
+#define ptr_unmask_bits(ptr, n) ((unsigned long)(ptr) & (BIT(n) - 1))
+
+#define ptr_unpack_bits(ptr, bits, n) ({				\
 	unsigned long __v = (unsigned long)(ptr);			\
-	*(bits) = __v & ~PAGE_MASK;					\
-	(typeof(ptr))(__v & PAGE_MASK);					\
+	*(bits) = __v & (BIT(n) - 1);					\
+	(typeof(ptr))(__v & -BIT(n));					\
 })
 
-#define ptr_pack_bits(ptr, bits)					\
+#define ptr_pack_bits(ptr, bits, n)					\
 	((typeof(ptr))((unsigned long)(ptr) | (bits)))
 
+#define page_mask_bits(ptr) ptr_mask_bits(ptr, PAGE_SHIFT)
+#define page_unmask_bits(ptr) ptr_unmask_bits(ptr, PAGE_SHIFT)
+#define page_pack_bits(ptr, bits) ptr_pack_bits(ptr, bits, PAGE_SHIFT)
+#define page_unpack_bits(ptr, bits) ptr_unpack_bits(ptr, bits, PAGE_SHIFT)
+
 #define ptr_offset(ptr, member) offsetof(typeof(*(ptr)), member)
 
 #define fetch_and_zero(ptr) ({						\
-- 
2.11.0

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH v2 5/9] drm/i915: Squash repeated awaits on the same fence

[Chris Wilson]

Track the latest fence waited upon on each context, and only add a new
asynchronous wait if the new fence is more recent than the recorded
fence for that context. This requires us to filter out unordered
timelines, which are noted by DMA_FENCE_NO_CONTEXT. However, in the
absence of a universal identifier, we have to use our own
i915->mm.unordered_timeline token.

v2: Throw around the debug crutches
v3: Inline the likely case of the pre-allocation cache being full.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
 drivers/gpu/drm/i915/i915_gem_request.c            |  72 ++++---
 drivers/gpu/drm/i915/i915_gem_timeline.c           | 217 +++++++++++++++++++++
 drivers/gpu/drm/i915/i915_gem_timeline.h           |  34 ++++
 drivers/gpu/drm/i915/selftests/i915_gem_timeline.c | 127 ++++++++++++
 .../gpu/drm/i915/selftests/i915_mock_selftests.h   |   1 +
 5 files changed, 424 insertions(+), 27 deletions(-)
 create mode 100644 drivers/gpu/drm/i915/selftests/i915_gem_timeline.c

diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index 31874a38752e..6e68387aa097 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -714,9 +714,7 @@ int
 i915_gem_request_await_dma_fence(struct drm_i915_gem_request *req,
 				 struct dma_fence *fence)
 {
-	struct dma_fence_array *array;
 	int ret;
-	int i;
 
 	if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
 		return 0;
@@ -728,39 +726,59 @@ i915_gem_request_await_dma_fence(struct drm_i915_gem_request *req,
 	if (fence->context == req->fence.context)
 		return 0;
 
-	if (dma_fence_is_i915(fence))
-		return i915_gem_request_await_request(req, to_request(fence));
+	/* Squash repeated waits to the same timelines, picking the latest */
+	if (fence->context != req->i915->mm.unordered_timeline) {
+		if (intel_timeline_sync_get(req->timeline,
+					    fence->context,
+					    fence->seqno))
+			return 0;
 
-	if (!dma_fence_is_array(fence)) {
+		ret = intel_timeline_sync_reserve(req->timeline);
+		if (unlikely(ret))
+			return ret;
+	}
+
+	if (dma_fence_is_i915(fence)) {
+		ret = i915_gem_request_await_request(req, to_request(fence));
+		if (ret < 0)
+			return ret;
+	} else if (!dma_fence_is_array(fence)) {
 		ret = i915_sw_fence_await_dma_fence(&req->submit,
 						    fence, I915_FENCE_TIMEOUT,
 						    GFP_KERNEL);
-		return ret < 0 ? ret : 0;
-	}
-
-	/* Note that if the fence-array was created in signal-on-any mode,
-	 * we should *not* decompose it into its individual fences. However,
-	 * we don't currently store which mode the fence-array is operating
-	 * in. Fortunately, the only user of signal-on-any is private to
-	 * amdgpu and we should not see any incoming fence-array from
-	 * sync-file being in signal-on-any mode.
-	 */
-
-	array = to_dma_fence_array(fence);
-	for (i = 0; i < array->num_fences; i++) {
-		struct dma_fence *child = array->fences[i];
-
-		if (dma_fence_is_i915(child))
-			ret = i915_gem_request_await_request(req,
-							     to_request(child));
-		else
-			ret = i915_sw_fence_await_dma_fence(&req->submit,
-							    child, I915_FENCE_TIMEOUT,
-							    GFP_KERNEL);
 		if (ret < 0)
 			return ret;
+	} else {
+		struct dma_fence_array *array = to_dma_fence_array(fence);
+		int i;
+
+		/* Note that if the fence-array was created in signal-on-any mode,
+		 * we should *not* decompose it into its individual fences. However,
+		 * we don't currently store which mode the fence-array is operating
+		 * in. Fortunately, the only user of signal-on-any is private to
+		 * amdgpu and we should not see any incoming fence-array from
+		 * sync-file being in signal-on-any mode.
+		 */
+
+		for (i = 0; i < array->num_fences; i++) {
+			struct dma_fence *child = array->fences[i];
+
+			if (dma_fence_is_i915(child))
+				ret = i915_gem_request_await_request(req,
+								     to_request(child));
+			else
+				ret = i915_sw_fence_await_dma_fence(&req->submit,
+								    child, I915_FENCE_TIMEOUT,
+								    GFP_KERNEL);
+			if (ret < 0)
+				return ret;
+		}
 	}
 
+	if (fence->context != req->i915->mm.unordered_timeline)
+		intel_timeline_sync_set(req->timeline,
+					fence->context, fence->seqno);
+
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_gem_timeline.c b/drivers/gpu/drm/i915/i915_gem_timeline.c
index b596ca7ee058..c25edfde4833 100644
--- a/drivers/gpu/drm/i915/i915_gem_timeline.c
+++ b/drivers/gpu/drm/i915/i915_gem_timeline.c
@@ -24,6 +24,217 @@
 
 #include "i915_drv.h"
 
+#define SHIFT ilog2(NSEQMAP)
+#define MASK (NSEQMAP - 1)
+
+static void seqmap_free_layers(struct seqmap_layer *p)
+{
+	unsigned int i;
+
+	if (p->height) {
+		for (; (i = ffs(p->bitmap)); p->bitmap &= ~0u << i)
+			seqmap_free_layers(p->slot[i - 1]);
+	}
+
+	kfree(p);
+}
+
+static void seqmap_free(struct seqmap *seqmap)
+{
+	if (seqmap->top)
+		seqmap_free_layers(seqmap->top);
+
+	while (seqmap->freed) {
+		struct seqmap_layer *p;
+
+		p = ptr_mask_bits(seqmap->freed, SEQMAP_COUNT_BITS);
+		seqmap->freed = p->parent;
+		kfree(p);
+	}
+}
+
+__malloc static struct seqmap_layer *
+seqmap_alloc_layer(struct seqmap *shared)
+{
+	struct seqmap_layer *p;
+
+	GEM_BUG_ON(!shared->freed);
+
+	p = ptr_mask_bits(shared->freed, SEQMAP_COUNT_BITS);
+	shared->freed = p->parent;
+
+	return p;
+}
+
+static int layer_idx(const struct seqmap_layer *p, u64 id)
+{
+	return (id >> p->height) & MASK;
+}
+
+bool intel_timeline_sync_get(struct intel_timeline *tl, u64 id, u32 seqno)
+{
+	struct seqmap *shared = &tl->sync;
+	struct seqmap_layer *p;
+	unsigned int idx;
+
+	p = shared->hint;
+	if (!p)
+		return false;
+
+	if ((id >> SHIFT) == p->prefix)
+		goto found;
+
+	p = shared->top;
+	do {
+		if ((id >> p->height >> SHIFT) != p->prefix)
+			return false;
+
+		if (!p->height)
+			break;
+
+		p = p->slot[layer_idx(p, id)];
+		if (!p)
+			return false;
+	} while (1);
+
+	shared->hint = p;
+found:
+	idx = id & MASK;
+	if (!(p->bitmap & BIT(idx)))
+		return false;
+
+	return i915_seqno_passed((uintptr_t)p->slot[idx], seqno);
+}
+
+void intel_timeline_sync_set(struct intel_timeline *tl, u64 id, u32 seqno)
+{
+	struct seqmap *shared = &tl->sync;
+	struct seqmap_layer *p, *cur;
+	unsigned int idx;
+
+	/* We expect to be called in sequence following a  _get(id), which
+	 * should have preloaded the ->hint for us.
+	 */
+	p = shared->hint;
+	if (likely(p && (id >> SHIFT) == p->prefix))
+		goto found_layer;
+
+	if (!p) {
+		GEM_BUG_ON(shared->top);
+		cur = seqmap_alloc_layer(shared);
+		cur->parent = NULL;
+		shared->top = cur;
+		goto new_layer;
+	}
+
+	/* No shortcut, we have to descend the tree to find the right layer
+	 * containing this fence.
+	 *
+	 * Each layer in the tree holds 16 (NSEQMAP) pointers, either fences
+	 * or lower layers. Leaf nodes (height = 0) contain the fences, all
+	 * other nodes (height > 0) are internal layers that point to a lower
+	 * node. Each internal layer has at least 2 descendents.
+	 *
+	 * Starting at the top, we check whether the current prefix matches. If
+	 * it doesn't, we have gone passed our layer and need to insert a join
+	 * into the tree, and a new leaf node as a descendent as well as the
+	 * original layer.
+	 *
+	 * The matching prefix means we are still following the right branch
+	 * of the tree. If it has height 0, we have found our leaf and just
+	 * need to replace the fence slot with ourselves. If the height is
+	 * not zero, our slot contains the next layer in the tree (unless
+	 * it is empty, in which case we can add ourselves as a new leaf).
+	 * As descend the tree the prefix grows (and height decreases).
+	 */
+	p = shared->top;
+	do {
+		if ((id >> p->height >> SHIFT) != p->prefix) {
+			/* insert a join above the current layer */
+			cur = seqmap_alloc_layer(shared);
+			cur->height = ALIGN(fls64((id >> p->height >> SHIFT) ^ p->prefix),
+					    SHIFT) + p->height;
+			cur->prefix = id >> cur->height >> SHIFT;
+
+			if (p->parent)
+				p->parent->slot[layer_idx(p->parent, id)] = cur;
+			else
+				shared->top = cur;
+			cur->parent = p->parent;
+
+			idx = p->prefix >> (cur->height - p->height - SHIFT) & MASK;
+			cur->slot[idx] = p;
+			cur->bitmap |= BIT(idx);
+			p->parent = cur;
+		} else if (!p->height) {
+			/* matching base layer */
+			shared->hint = p;
+			goto found_layer;
+		} else {
+			/* descend into the next layer */
+			idx = layer_idx(p, id);
+			cur = p->slot[idx];
+			if (unlikely(!cur)) {
+				cur = seqmap_alloc_layer(shared);
+				p->slot[idx] = cur;
+				p->bitmap |= BIT(idx);
+				cur->parent = p;
+				goto new_layer;
+			}
+		}
+
+		p = cur;
+	} while (1);
+
+found_layer:
+	GEM_BUG_ON(p->height);
+	GEM_BUG_ON(p->prefix != id >> SHIFT);
+	idx = id & MASK;
+	p->slot[idx] = (void *)(uintptr_t)seqno;
+	p->bitmap |= BIT(idx);
+	GEM_BUG_ON(shared->hint != p);
+	return;
+
+new_layer:
+	GEM_BUG_ON(cur->height);
+	cur->prefix = id >> SHIFT;
+	idx = id & MASK;
+	cur->slot[idx] = (void *)(uintptr_t)seqno;
+	cur->bitmap = BIT(idx);
+	shared->hint = cur;
+	return;
+}
+
+int __intel_timeline_sync_reserve(struct intel_timeline *tl)
+{
+	struct seqmap *shared = &tl->sync;
+	int count;
+
+	might_sleep();
+
+	/* To guarantee being able to replace a fence in the radixtree,
+	 * we need at most 2 layers: one to create a join in the tree,
+	 * and one to contain the fence. Typically we expect to reuse
+	 * a layer and so avoid any insertions.
+	 *
+	 * We use the low bits of the freed list to track its length
+	 * since we only need a couple of bits.
+	 */
+	count = ptr_unmask_bits(shared->freed, SEQMAP_COUNT_BITS);
+	while (count++ < 2) {
+		struct seqmap_layer *p;
+
+		p = kzalloc(sizeof(*p), GFP_KERNEL);
+		if (unlikely(!p))
+			return -ENOMEM;
+
+		p->parent = shared->freed;
+		shared->freed = ptr_pack_bits(p, count, SEQMAP_COUNT_BITS);
+	}
+
+	return 0;
+}
+
 static int __i915_gem_timeline_init(struct drm_i915_private *i915,
 				    struct i915_gem_timeline *timeline,
 				    const char *name,
@@ -91,8 +302,14 @@ void i915_gem_timeline_fini(struct i915_gem_timeline *timeline)
 		struct intel_timeline *tl = &timeline->engine[i];
 
 		GEM_BUG_ON(!list_empty(&tl->requests));
+
+		seqmap_free(&tl->sync);
 	}
 
 	list_del(&timeline->link);
 	kfree(timeline->name);
 }
+
+#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
+#include "selftests/i915_gem_timeline.c"
+#endif
diff --git a/drivers/gpu/drm/i915/i915_gem_timeline.h b/drivers/gpu/drm/i915/i915_gem_timeline.h
index 6c53e14cab2a..4cb597401803 100644
--- a/drivers/gpu/drm/i915/i915_gem_timeline.h
+++ b/drivers/gpu/drm/i915/i915_gem_timeline.h
@@ -26,11 +26,30 @@
 #define I915_GEM_TIMELINE_H
 
 #include <linux/list.h>
+#include <linux/radix-tree.h>
 
+#include "i915_utils.h"
 #include "i915_gem_request.h"
 
 struct i915_gem_timeline;
 
+#define NSEQMAP 16
+
+struct seqmap_layer {
+	u64 prefix;
+	unsigned int height;
+	unsigned int bitmap;
+	struct seqmap_layer *parent;
+	void *slot[NSEQMAP];
+};
+
+struct seqmap {
+	struct seqmap_layer *hint;
+	struct seqmap_layer *top;
+	struct seqmap_layer *freed;
+#define SEQMAP_COUNT_BITS 2
+};
+
 struct intel_timeline {
 	u64 fence_context;
 	u32 seqno;
@@ -55,6 +74,8 @@ struct intel_timeline {
 	 * struct_mutex.
 	 */
 	struct i915_gem_active last_request;
+
+	struct seqmap sync;
 	u32 sync_seqno[I915_NUM_ENGINES];
 
 	struct i915_gem_timeline *common;
@@ -75,4 +96,17 @@ int i915_gem_timeline_init(struct drm_i915_private *i915,
 int i915_gem_timeline_init__global(struct drm_i915_private *i915);
 void i915_gem_timeline_fini(struct i915_gem_timeline *tl);
 
+bool intel_timeline_sync_get(struct intel_timeline *tl, u64 id, u32 seqno);
+
+int __intel_timeline_sync_reserve(struct intel_timeline *tl);
+static inline int intel_timeline_sync_reserve(struct intel_timeline *tl)
+{
+	if (likely(ptr_unmask_bits(tl->sync.freed, SEQMAP_COUNT_BITS) == 2))
+		return 0;
+
+	return __intel_timeline_sync_reserve(tl);
+}
+
+void intel_timeline_sync_set(struct intel_timeline *tl, u64 id, u32 seqno);
+
 #endif
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_timeline.c b/drivers/gpu/drm/i915/selftests/i915_gem_timeline.c
new file mode 100644
index 000000000000..15e2b2e63081
--- /dev/null
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_timeline.c
@@ -0,0 +1,127 @@
+/*
+ * Copyright © 2017 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#include "../i915_selftest.h"
+#include "mock_gem_device.h"
+
+static int igt_seqmap(void *arg)
+{
+	struct drm_i915_private *i915 = arg;
+	const struct {
+		const char *name;
+		u32 seqno;
+		bool expected;
+		bool set;
+	} pass[] = {
+		{ "unset", 0, false, false },
+		{ "new", 0, false, true },
+		{ "0a", 0, true, true },
+		{ "1a", 1, false, true },
+		{ "1b", 1, true, true },
+		{ "0b", 0, true, false },
+		{ "2a", 2, false, true },
+		{ "4", 4, false, true },
+		{ "INT_MAX", INT_MAX, false, true },
+		{ "INT_MAX-1", INT_MAX-1, true, false },
+		{ "INT_MAX+1", (u32)INT_MAX+1, false, true },
+		{ "INT_MAX", INT_MAX, true, false },
+		{ "UINT_MAX", UINT_MAX, false, true },
+		{ "wrap", 0, false, true },
+		{ "unwrap", UINT_MAX, true, false },
+		{},
+	}, *p;
+	struct intel_timeline *tl;
+	int order, offset;
+	int ret;
+
+	tl = &i915->gt.global_timeline.engine[RCS];
+	for (p = pass; p->name; p++) {
+		for (order = 1; order < 64; order++) {
+			for (offset = -1; offset <= (order > 1); offset++) {
+				u64 ctx = BIT_ULL(order) + offset;
+
+				if (intel_timeline_sync_get(tl,
+							    ctx,
+							    p->seqno) != p->expected) {
+					pr_err("1: %s(ctx=%llu, seqno=%u) expected passed %s but failed\n",
+					       p->name, ctx, p->seqno, yesno(p->expected));
+					return -EINVAL;
+				}
+
+				if (p->set) {
+					ret = intel_timeline_sync_reserve(tl);
+					if (ret)
+						return ret;
+
+					intel_timeline_sync_set(tl, ctx, p->seqno);
+				}
+			}
+		}
+	}
+
+	tl = &i915->gt.global_timeline.engine[BCS];
+	for (order = 1; order < 64; order++) {
+		for (offset = -1; offset <= (order > 1); offset++) {
+			u64 ctx = BIT_ULL(order) + offset;
+
+			for (p = pass; p->name; p++) {
+				if (intel_timeline_sync_get(tl,
+							    ctx,
+							    p->seqno) != p->expected) {
+					pr_err("2: %s(ctx=%llu, seqno=%u) expected passed %s but failed\n",
+					       p->name, ctx, p->seqno, yesno(p->expected));
+					return -EINVAL;
+				}
+
+				if (p->set) {
+					ret = intel_timeline_sync_reserve(tl);
+					if (ret)
+						return ret;
+
+					intel_timeline_sync_set(tl, ctx, p->seqno);
+				}
+			}
+		}
+	}
+
+	return 0;
+}
+
+int i915_gem_timeline_mock_selftests(void)
+{
+	static const struct i915_subtest tests[] = {
+		SUBTEST(igt_seqmap),
+	};
+	struct drm_i915_private *i915;
+	int err;
+
+	i915 = mock_gem_device();
+	if (!i915)
+		return -ENOMEM;
+
+	err = i915_subtests(tests, i915);
+	drm_dev_unref(&i915->drm);
+
+	return err;
+}
diff --git a/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h b/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h
index be9a9ebf5692..8d0f50c25df8 100644
--- a/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h
+++ b/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h
@@ -12,6 +12,7 @@ selftest(sanitycheck, i915_mock_sanitycheck) /* keep first (igt selfcheck) */
 selftest(scatterlist, scatterlist_mock_selftests)
 selftest(uncore, intel_uncore_mock_selftests)
 selftest(breadcrumbs, intel_breadcrumbs_mock_selftests)
+selftest(timelines, i915_gem_timeline_mock_selftests)
 selftest(requests, i915_gem_request_mock_selftests)
 selftest(objects, i915_gem_object_mock_selftests)
 selftest(dmabuf, i915_gem_dmabuf_mock_selftests)
-- 
2.11.0

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH v3] drm/i915: Squash repeated awaits on the same fence

[Chris Wilson]

Track the latest fence waited upon on each context, and only add a new
asynchronous wait if the new fence is more recent than the recorded
fence for that context. This requires us to filter out unordered
timelines, which are noted by DMA_FENCE_NO_CONTEXT. However, in the
absence of a universal identifier, we have to use our own
i915->mm.unordered_timeline token.

v2: Throw around the debug crutches
v3: Inline the likely case of the pre-allocation cache being full.
v4: Drop the pre-allocation support, we can lose the most recent fence
in case of allocation failure -- it just means we may emit more awaits
than strictly necessary but will not break.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
 drivers/gpu/drm/i915/i915_gem_request.c            |  66 ++++---
 drivers/gpu/drm/i915/i915_gem_timeline.c           | 194 +++++++++++++++++++++
 drivers/gpu/drm/i915/i915_gem_timeline.h           |  18 ++
 drivers/gpu/drm/i915/selftests/i915_gem_timeline.c | 123 +++++++++++++
 .../gpu/drm/i915/selftests/i915_mock_selftests.h   |   1 +
 5 files changed, 375 insertions(+), 27 deletions(-)
 create mode 100644 drivers/gpu/drm/i915/selftests/i915_gem_timeline.c

diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index 31874a38752e..f679c16aa5c6 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -714,9 +714,7 @@ int
 i915_gem_request_await_dma_fence(struct drm_i915_gem_request *req,
 				 struct dma_fence *fence)
 {
-	struct dma_fence_array *array;
 	int ret;
-	int i;
 
 	if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
 		return 0;
@@ -728,39 +726,53 @@ i915_gem_request_await_dma_fence(struct drm_i915_gem_request *req,
 	if (fence->context == req->fence.context)
 		return 0;
 
-	if (dma_fence_is_i915(fence))
-		return i915_gem_request_await_request(req, to_request(fence));
+	/* Squash repeated waits to the same timelines, picking the latest */
+	if (fence->context != req->i915->mm.unordered_timeline &&
+	    intel_timeline_sync_get(req->timeline,
+				    fence->context, fence->seqno))
+		return 0;
 
-	if (!dma_fence_is_array(fence)) {
+	if (dma_fence_is_i915(fence)) {
+		ret = i915_gem_request_await_request(req, to_request(fence));
+		if (ret < 0)
+			return ret;
+	} else if (!dma_fence_is_array(fence)) {
 		ret = i915_sw_fence_await_dma_fence(&req->submit,
 						    fence, I915_FENCE_TIMEOUT,
 						    GFP_KERNEL);
-		return ret < 0 ? ret : 0;
-	}
-
-	/* Note that if the fence-array was created in signal-on-any mode,
-	 * we should *not* decompose it into its individual fences. However,
-	 * we don't currently store which mode the fence-array is operating
-	 * in. Fortunately, the only user of signal-on-any is private to
-	 * amdgpu and we should not see any incoming fence-array from
-	 * sync-file being in signal-on-any mode.
-	 */
-
-	array = to_dma_fence_array(fence);
-	for (i = 0; i < array->num_fences; i++) {
-		struct dma_fence *child = array->fences[i];
-
-		if (dma_fence_is_i915(child))
-			ret = i915_gem_request_await_request(req,
-							     to_request(child));
-		else
-			ret = i915_sw_fence_await_dma_fence(&req->submit,
-							    child, I915_FENCE_TIMEOUT,
-							    GFP_KERNEL);
 		if (ret < 0)
 			return ret;
+	} else {
+		struct dma_fence_array *array = to_dma_fence_array(fence);
+		int i;
+
+		/* Note that if the fence-array was created in signal-on-any mode,
+		 * we should *not* decompose it into its individual fences. However,
+		 * we don't currently store which mode the fence-array is operating
+		 * in. Fortunately, the only user of signal-on-any is private to
+		 * amdgpu and we should not see any incoming fence-array from
+		 * sync-file being in signal-on-any mode.
+		 */
+
+		for (i = 0; i < array->num_fences; i++) {
+			struct dma_fence *child = array->fences[i];
+
+			if (dma_fence_is_i915(child))
+				ret = i915_gem_request_await_request(req,
+								     to_request(child));
+			else
+				ret = i915_sw_fence_await_dma_fence(&req->submit,
+								    child, I915_FENCE_TIMEOUT,
+								    GFP_KERNEL);
+			if (ret < 0)
+				return ret;
+		}
 	}
 
+	if (fence->context != req->i915->mm.unordered_timeline)
+		intel_timeline_sync_set(req->timeline,
+					fence->context, fence->seqno);
+
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_gem_timeline.c b/drivers/gpu/drm/i915/i915_gem_timeline.c
index b596ca7ee058..446eefa70e6d 100644
--- a/drivers/gpu/drm/i915/i915_gem_timeline.c
+++ b/drivers/gpu/drm/i915/i915_gem_timeline.c
@@ -24,6 +24,194 @@
 
 #include "i915_drv.h"
 
+#define SHIFT ilog2(NSEQMAP)
+#define MASK (NSEQMAP - 1)
+
+static void __seqmap_free(struct seqmap *p)
+{
+	unsigned int i;
+
+	if (p->height) {
+		for (; (i = ffs(p->bitmap)); p->bitmap &= ~0u << i)
+			__seqmap_free(p->slot[i - 1]);
+	}
+
+	kfree(p);
+}
+
+static void seqmap_free(struct seqmap *seqmap)
+{
+	if (!seqmap)
+		return;
+
+	while (seqmap->parent)
+		seqmap = seqmap->parent;
+
+	__seqmap_free(seqmap);
+}
+
+static int layer_idx(const struct seqmap *p, u64 id)
+{
+	return (id >> p->height) & MASK;
+}
+
+bool intel_timeline_sync_get(struct intel_timeline *tl, u64 id, u32 seqno)
+{
+	struct seqmap *p;
+	unsigned int idx;
+
+	p = tl->sync;
+	if (!p)
+		return false;
+
+	if ((id >> SHIFT) == p->prefix)
+		goto found;
+
+	/* First climb the tree back to a parent branch */
+	do {
+		p = p->parent;
+		if (!p)
+			return false;
+
+		if ((id >> p->height >> SHIFT) == p->prefix)
+			break;
+	} while (1);
+
+	/* And then descend again until we find our leaf */
+	do {
+		if (!p->height)
+			break;
+
+		p = p->slot[layer_idx(p, id)];
+		if (!p)
+			return false;
+
+		if ((id >> p->height >> SHIFT) != p->prefix)
+			return false;
+	} while (1);
+
+	tl->sync = p;
+found:
+	idx = id & MASK;
+	if (!(p->bitmap & BIT(idx)))
+		return false;
+
+	return i915_seqno_passed((uintptr_t)p->slot[idx], seqno);
+}
+
+int intel_timeline_sync_set(struct intel_timeline *tl, u64 id, u32 seqno)
+{
+	struct seqmap *p;
+	unsigned int idx;
+
+	/* We expect to be called in sequence following a  _get(id), which
+	 * should have preloaded the tl->sync hint for us.
+	 */
+	p = tl->sync;
+	if (likely(p && (id >> SHIFT) == p->prefix))
+		goto set;
+
+	if (!p) {
+		p = kzalloc(sizeof(*p), GFP_KERNEL);
+		if (unlikely(!p))
+			return -ENOMEM;
+
+		p->prefix = id >> SHIFT;
+		goto found;
+	}
+
+	/* Climb back up the tree until we find a common prefix */
+	do {
+		if (!p->parent)
+			break;
+
+		p = p->parent;
+
+		if ((id >> p->height >> SHIFT) == p->prefix)
+			break;
+	} while (1);
+
+	/* No shortcut, we have to descend the tree to find the right layer
+	 * containing this fence.
+	 *
+	 * Each layer in the tree holds 16 (NSEQMAP) pointers, either fences
+	 * or lower layers. Leaf nodes (height = 0) contain the fences, all
+	 * other nodes (height > 0) are internal layers that point to a lower
+	 * node. Each internal layer has at least 2 descendents.
+	 *
+	 * Starting at the top, we check whether the current prefix matches. If
+	 * it doesn't, we have gone passed our layer and need to insert a join
+	 * into the tree, and a new leaf node as a descendent as well as the
+	 * original layer.
+	 *
+	 * The matching prefix means we are still following the right branch
+	 * of the tree. If it has height 0, we have found our leaf and just
+	 * need to replace the fence slot with ourselves. If the height is
+	 * not zero, our slot contains the next layer in the tree (unless
+	 * it is empty, in which case we can add ourselves as a new leaf).
+	 * As descend the tree the prefix grows (and height decreases).
+	 */
+	do {
+		struct seqmap *next;
+
+		if ((id >> p->height >> SHIFT) != p->prefix) {
+			/* insert a join above the current layer */
+			next = kzalloc(sizeof(*next), GFP_KERNEL);
+			if (unlikely(!next))
+				return -ENOMEM;
+
+			next->height = ALIGN(fls64((id >> p->height >> SHIFT) ^ p->prefix),
+					    SHIFT) + p->height;
+			next->prefix = id >> next->height >> SHIFT;
+
+			if (p->parent)
+				p->parent->slot[layer_idx(p->parent, id)] = next;
+			next->parent = p->parent;
+
+			idx = p->prefix >> (next->height - p->height - SHIFT) & MASK;
+			next->slot[idx] = p;
+			next->bitmap |= BIT(idx);
+			p->parent = next;
+
+			/* ascend to the join */
+			p = next;
+		} else {
+			if (!p->height)
+				break;
+		}
+
+		/* descend into the next layer */
+		GEM_BUG_ON(!p->height);
+		idx = layer_idx(p, id);
+		next = p->slot[idx];
+		if (unlikely(!next)) {
+			next = kzalloc(sizeof(*next), GFP_KERNEL);
+			if (unlikely(!next))
+				return -ENOMEM;
+
+			p->slot[idx] = next;
+			p->bitmap |= BIT(idx);
+			next->parent = p;
+			next->prefix = id >> SHIFT;
+
+			p = next;
+			break;
+		}
+
+		p = next;
+	} while (1);
+
+found:
+	GEM_BUG_ON(p->height);
+	GEM_BUG_ON(p->prefix != id >> SHIFT);
+	tl->sync = p;
+set:
+	idx = id & MASK;
+	p->slot[idx] = (void *)(uintptr_t)seqno;
+	p->bitmap |= BIT(idx);
+	return 0;
+}
+
 static int __i915_gem_timeline_init(struct drm_i915_private *i915,
 				    struct i915_gem_timeline *timeline,
 				    const char *name,
@@ -91,8 +279,14 @@ void i915_gem_timeline_fini(struct i915_gem_timeline *timeline)
 		struct intel_timeline *tl = &timeline->engine[i];
 
 		GEM_BUG_ON(!list_empty(&tl->requests));
+
+		seqmap_free(tl->sync);
 	}
 
 	list_del(&timeline->link);
 	kfree(timeline->name);
 }
+
+#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
+#include "selftests/i915_gem_timeline.c"
+#endif
diff --git a/drivers/gpu/drm/i915/i915_gem_timeline.h b/drivers/gpu/drm/i915/i915_gem_timeline.h
index 6c53e14cab2a..64bad92320de 100644
--- a/drivers/gpu/drm/i915/i915_gem_timeline.h
+++ b/drivers/gpu/drm/i915/i915_gem_timeline.h
@@ -26,11 +26,23 @@
 #define I915_GEM_TIMELINE_H
 
 #include <linux/list.h>
+#include <linux/radix-tree.h>
 
+#include "i915_utils.h"
 #include "i915_gem_request.h"
 
 struct i915_gem_timeline;
 
+#define NSEQMAP 16
+
+struct seqmap {
+	u64 prefix;
+	unsigned int height;
+	unsigned int bitmap;
+	struct seqmap *parent;
+	void *slot[NSEQMAP];
+};
+
 struct intel_timeline {
 	u64 fence_context;
 	u32 seqno;
@@ -55,6 +67,8 @@ struct intel_timeline {
 	 * struct_mutex.
 	 */
 	struct i915_gem_active last_request;
+
+	struct seqmap *sync;
 	u32 sync_seqno[I915_NUM_ENGINES];
 
 	struct i915_gem_timeline *common;
@@ -75,4 +89,8 @@ int i915_gem_timeline_init(struct drm_i915_private *i915,
 int i915_gem_timeline_init__global(struct drm_i915_private *i915);
 void i915_gem_timeline_fini(struct i915_gem_timeline *tl);
 
+bool intel_timeline_sync_get(struct intel_timeline *tl, u64 id, u32 seqno);
+
+int intel_timeline_sync_set(struct intel_timeline *tl, u64 id, u32 seqno);
+
 #endif
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_timeline.c b/drivers/gpu/drm/i915/selftests/i915_gem_timeline.c
new file mode 100644
index 000000000000..c0bb8ecac93b
--- /dev/null
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_timeline.c
@@ -0,0 +1,123 @@
+/*
+ * Copyright © 2017 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#include "../i915_selftest.h"
+#include "mock_gem_device.h"
+
+static int igt_seqmap(void *arg)
+{
+	struct drm_i915_private *i915 = arg;
+	const struct {
+		const char *name;
+		u32 seqno;
+		bool expected;
+		bool set;
+	} pass[] = {
+		{ "unset", 0, false, false },
+		{ "new", 0, false, true },
+		{ "0a", 0, true, true },
+		{ "1a", 1, false, true },
+		{ "1b", 1, true, true },
+		{ "0b", 0, true, false },
+		{ "2a", 2, false, true },
+		{ "4", 4, false, true },
+		{ "INT_MAX", INT_MAX, false, true },
+		{ "INT_MAX-1", INT_MAX-1, true, false },
+		{ "INT_MAX+1", (u32)INT_MAX+1, false, true },
+		{ "INT_MAX", INT_MAX, true, false },
+		{ "UINT_MAX", UINT_MAX, false, true },
+		{ "wrap", 0, false, true },
+		{ "unwrap", UINT_MAX, true, false },
+		{},
+	}, *p;
+	struct intel_timeline *tl;
+	int order, offset;
+	int ret;
+
+	tl = &i915->gt.global_timeline.engine[RCS];
+	for (p = pass; p->name; p++) {
+		for (order = 1; order < 64; order++) {
+			for (offset = -1; offset <= (order > 1); offset++) {
+				u64 ctx = BIT_ULL(order) + offset;
+
+				if (intel_timeline_sync_get(tl,
+							    ctx,
+							    p->seqno) != p->expected) {
+					pr_err("1: %s(ctx=%llu, seqno=%u) expected passed %s but failed\n",
+					       p->name, ctx, p->seqno, yesno(p->expected));
+					return -EINVAL;
+				}
+
+				if (p->set) {
+					ret = intel_timeline_sync_set(tl, ctx, p->seqno);
+					if (ret)
+						return ret;
+				}
+			}
+		}
+	}
+
+	tl = &i915->gt.global_timeline.engine[BCS];
+	for (order = 1; order < 64; order++) {
+		for (offset = -1; offset <= (order > 1); offset++) {
+			u64 ctx = BIT_ULL(order) + offset;
+
+			for (p = pass; p->name; p++) {
+				if (intel_timeline_sync_get(tl,
+							    ctx,
+							    p->seqno) != p->expected) {
+					pr_err("2: %s(ctx=%llu, seqno=%u) expected passed %s but failed\n",
+					       p->name, ctx, p->seqno, yesno(p->expected));
+					return -EINVAL;
+				}
+
+				if (p->set) {
+					ret = intel_timeline_sync_set(tl, ctx, p->seqno);
+					if (ret)
+						return ret;
+				}
+			}
+		}
+	}
+
+	return 0;
+}
+
+int i915_gem_timeline_mock_selftests(void)
+{
+	static const struct i915_subtest tests[] = {
+		SUBTEST(igt_seqmap),
+	};
+	struct drm_i915_private *i915;
+	int err;
+
+	i915 = mock_gem_device();
+	if (!i915)
+		return -ENOMEM;
+
+	err = i915_subtests(tests, i915);
+	drm_dev_unref(&i915->drm);
+
+	return err;
+}
diff --git a/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h b/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h
index be9a9ebf5692..8d0f50c25df8 100644
--- a/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h
+++ b/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h
@@ -12,6 +12,7 @@ selftest(sanitycheck, i915_mock_sanitycheck) /* keep first (igt selfcheck) */
 selftest(scatterlist, scatterlist_mock_selftests)
 selftest(uncore, intel_uncore_mock_selftests)
 selftest(breadcrumbs, intel_breadcrumbs_mock_selftests)
+selftest(timelines, i915_gem_timeline_mock_selftests)
 selftest(requests, i915_gem_request_mock_selftests)
 selftest(objects, i915_gem_object_mock_selftests)
 selftest(dmabuf, i915_gem_dmabuf_mock_selftests)
-- 
2.11.0

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH v4] drm/i915: Squash repeated awaits on the same fence

[Chris Wilson]

Track the latest fence waited upon on each context, and only add a new
asynchronous wait if the new fence is more recent than the recorded
fence for that context. This requires us to filter out unordered
timelines, which are noted by DMA_FENCE_NO_CONTEXT. However, in the
absence of a universal identifier, we have to use our own
i915->mm.unordered_timeline token.

v2: Throw around the debug crutches
v3: Inline the likely case of the pre-allocation cache being full.
v4: Drop the pre-allocation support, we can lose the most recent fence
in case of allocation failure -- it just means we may emit more awaits
than strictly necessary but will not break.
v5: Trim allocation size for leaf nodes, they only need an array of u32
not pointers.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
 drivers/gpu/drm/i915/i915_gem_request.c            |  66 ++++---
 drivers/gpu/drm/i915/i915_gem_timeline.c           | 215 +++++++++++++++++++++
 drivers/gpu/drm/i915/i915_gem_timeline.h           |   8 +
 drivers/gpu/drm/i915/selftests/i915_gem_timeline.c | 123 ++++++++++++
 .../gpu/drm/i915/selftests/i915_mock_selftests.h   |   1 +
 5 files changed, 386 insertions(+), 27 deletions(-)
 create mode 100644 drivers/gpu/drm/i915/selftests/i915_gem_timeline.c

diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index 31874a38752e..f679c16aa5c6 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -714,9 +714,7 @@ int
 i915_gem_request_await_dma_fence(struct drm_i915_gem_request *req,
 				 struct dma_fence *fence)
 {
-	struct dma_fence_array *array;
 	int ret;
-	int i;
 
 	if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
 		return 0;
@@ -728,39 +726,53 @@ i915_gem_request_await_dma_fence(struct drm_i915_gem_request *req,
 	if (fence->context == req->fence.context)
 		return 0;
 
-	if (dma_fence_is_i915(fence))
-		return i915_gem_request_await_request(req, to_request(fence));
+	/* Squash repeated waits to the same timelines, picking the latest */
+	if (fence->context != req->i915->mm.unordered_timeline &&
+	    intel_timeline_sync_get(req->timeline,
+				    fence->context, fence->seqno))
+		return 0;
 
-	if (!dma_fence_is_array(fence)) {
+	if (dma_fence_is_i915(fence)) {
+		ret = i915_gem_request_await_request(req, to_request(fence));
+		if (ret < 0)
+			return ret;
+	} else if (!dma_fence_is_array(fence)) {
 		ret = i915_sw_fence_await_dma_fence(&req->submit,
 						    fence, I915_FENCE_TIMEOUT,
 						    GFP_KERNEL);
-		return ret < 0 ? ret : 0;
-	}
-
-	/* Note that if the fence-array was created in signal-on-any mode,
-	 * we should *not* decompose it into its individual fences. However,
-	 * we don't currently store which mode the fence-array is operating
-	 * in. Fortunately, the only user of signal-on-any is private to
-	 * amdgpu and we should not see any incoming fence-array from
-	 * sync-file being in signal-on-any mode.
-	 */
-
-	array = to_dma_fence_array(fence);
-	for (i = 0; i < array->num_fences; i++) {
-		struct dma_fence *child = array->fences[i];
-
-		if (dma_fence_is_i915(child))
-			ret = i915_gem_request_await_request(req,
-							     to_request(child));
-		else
-			ret = i915_sw_fence_await_dma_fence(&req->submit,
-							    child, I915_FENCE_TIMEOUT,
-							    GFP_KERNEL);
 		if (ret < 0)
 			return ret;
+	} else {
+		struct dma_fence_array *array = to_dma_fence_array(fence);
+		int i;
+
+		/* Note that if the fence-array was created in signal-on-any mode,
+		 * we should *not* decompose it into its individual fences. However,
+		 * we don't currently store which mode the fence-array is operating
+		 * in. Fortunately, the only user of signal-on-any is private to
+		 * amdgpu and we should not see any incoming fence-array from
+		 * sync-file being in signal-on-any mode.
+		 */
+
+		for (i = 0; i < array->num_fences; i++) {
+			struct dma_fence *child = array->fences[i];
+
+			if (dma_fence_is_i915(child))
+				ret = i915_gem_request_await_request(req,
+								     to_request(child));
+			else
+				ret = i915_sw_fence_await_dma_fence(&req->submit,
+								    child, I915_FENCE_TIMEOUT,
+								    GFP_KERNEL);
+			if (ret < 0)
+				return ret;
+		}
 	}
 
+	if (fence->context != req->i915->mm.unordered_timeline)
+		intel_timeline_sync_set(req->timeline,
+					fence->context, fence->seqno);
+
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_gem_timeline.c b/drivers/gpu/drm/i915/i915_gem_timeline.c
index b596ca7ee058..15ff9f37b6dc 100644
--- a/drivers/gpu/drm/i915/i915_gem_timeline.c
+++ b/drivers/gpu/drm/i915/i915_gem_timeline.c
@@ -24,6 +24,215 @@
 
 #include "i915_drv.h"
 
+#define NSYNC 16
+#define SHIFT ilog2(NSYNC)
+#define MASK (NSYNC - 1)
+
+struct intel_timeline_sync {
+	u64 prefix;
+	unsigned int height;
+	unsigned int bitmap;
+	struct intel_timeline_sync *parent;
+};
+
+static inline u32 *__sync_seqno(struct intel_timeline_sync *p)
+{
+	return (u32 *)(p + 1);
+}
+
+static inline struct intel_timeline_sync **
+__sync_child(struct intel_timeline_sync *p)
+{
+	return (struct intel_timeline_sync **)(p + 1);
+}
+
+static void __sync_free(struct intel_timeline_sync *p)
+{
+	unsigned int i;
+
+	if (p->height) {
+		for (; (i = ffs(p->bitmap)); p->bitmap &= ~0u << i)
+			__sync_free(__sync_child(p)[i - 1]);
+	}
+
+	kfree(p);
+}
+
+static void sync_free(struct intel_timeline_sync *sync)
+{
+	if (!sync)
+		return;
+
+	while (sync->parent)
+		sync = sync->parent;
+
+	__sync_free(sync);
+}
+
+static int layer_idx(const struct intel_timeline_sync *p, u64 id)
+{
+	return (id >> p->height) & MASK;
+}
+
+bool intel_timeline_sync_get(struct intel_timeline *tl, u64 id, u32 seqno)
+{
+	struct intel_timeline_sync *p;
+	unsigned int idx;
+
+	p = tl->sync;
+	if (!p)
+		return false;
+
+	if ((id >> SHIFT) == p->prefix)
+		goto found;
+
+	/* First climb the tree back to a parent branch */
+	do {
+		p = p->parent;
+		if (!p)
+			return false;
+
+		if ((id >> p->height >> SHIFT) == p->prefix)
+			break;
+	} while (1);
+
+	/* And then descend again until we find our leaf */
+	do {
+		if (!p->height)
+			break;
+
+		p = __sync_child(p)[layer_idx(p, id)];
+		if (!p)
+			return false;
+
+		if ((id >> p->height >> SHIFT) != p->prefix)
+			return false;
+	} while (1);
+
+	tl->sync = p;
+found:
+	idx = id & MASK;
+	if (!(p->bitmap & BIT(idx)))
+		return false;
+
+	return i915_seqno_passed(__sync_seqno(p)[idx], seqno);
+}
+
+int intel_timeline_sync_set(struct intel_timeline *tl, u64 id, u32 seqno)
+{
+	struct intel_timeline_sync *p;
+	unsigned int idx;
+
+	/* We expect to be called in sequence following a  _get(id), which
+	 * should have preloaded the tl->sync hint for us.
+	 */
+	p = tl->sync;
+	if (likely(p && (id >> SHIFT) == p->prefix))
+		goto set;
+
+	if (!p) {
+		p = kzalloc(sizeof(*p) + NSYNC * sizeof(seqno), GFP_KERNEL);
+		if (unlikely(!p))
+			return -ENOMEM;
+
+		p->prefix = id >> SHIFT;
+		goto found;
+	}
+
+	/* Climb back up the tree until we find a common prefix */
+	do {
+		if (!p->parent)
+			break;
+
+		p = p->parent;
+
+		if ((id >> p->height >> SHIFT) == p->prefix)
+			break;
+	} while (1);
+
+	/* No shortcut, we have to descend the tree to find the right layer
+	 * containing this fence.
+	 *
+	 * Each layer in the tree holds 16 (NSYNC) pointers, either fences
+	 * or lower layers. Leaf nodes (height = 0) contain the fences, all
+	 * other nodes (height > 0) are internal layers that point to a lower
+	 * node. Each internal layer has at least 2 descendents.
+	 *
+	 * Starting at the top, we check whether the current prefix matches. If
+	 * it doesn't, we have gone passed our layer and need to insert a join
+	 * into the tree, and a new leaf node as a descendent as well as the
+	 * original layer.
+	 *
+	 * The matching prefix means we are still following the right branch
+	 * of the tree. If it has height 0, we have found our leaf and just
+	 * need to replace the fence slot with ourselves. If the height is
+	 * not zero, our slot contains the next layer in the tree (unless
+	 * it is empty, in which case we can add ourselves as a new leaf).
+	 * As descend the tree the prefix grows (and height decreases).
+	 */
+	do {
+		struct intel_timeline_sync *next;
+
+		if ((id >> p->height >> SHIFT) != p->prefix) {
+			/* insert a join above the current layer */
+			next = kzalloc(sizeof(*next) + NSYNC * sizeof(next),
+				       GFP_KERNEL);
+			if (unlikely(!next))
+				return -ENOMEM;
+
+			next->height = ALIGN(fls64((id >> p->height >> SHIFT) ^ p->prefix),
+					    SHIFT) + p->height;
+			next->prefix = id >> next->height >> SHIFT;
+
+			if (p->parent)
+				__sync_child(p->parent)[layer_idx(p->parent, id)] = next;
+			next->parent = p->parent;
+
+			idx = p->prefix >> (next->height - p->height - SHIFT) & MASK;
+			__sync_child(next)[idx] = p;
+			next->bitmap |= BIT(idx);
+			p->parent = next;
+
+			/* ascend to the join */
+			p = next;
+		} else {
+			if (!p->height)
+				break;
+		}
+
+		/* descend into the next layer */
+		GEM_BUG_ON(!p->height);
+		idx = layer_idx(p, id);
+		next = __sync_child(p)[idx];
+		if (unlikely(!next)) {
+			next = kzalloc(sizeof(*next) + NSYNC * sizeof(seqno),
+				       GFP_KERNEL);
+			if (unlikely(!next))
+				return -ENOMEM;
+
+			__sync_child(p)[idx] = next;
+			p->bitmap |= BIT(idx);
+			next->parent = p;
+			next->prefix = id >> SHIFT;
+
+			p = next;
+			break;
+		}
+
+		p = next;
+	} while (1);
+
+found:
+	GEM_BUG_ON(p->height);
+	GEM_BUG_ON(p->prefix != id >> SHIFT);
+	tl->sync = p;
+set:
+	idx = id & MASK;
+	__sync_seqno(p)[idx] = seqno;
+	p->bitmap |= BIT(idx);
+	return 0;
+}
+
 static int __i915_gem_timeline_init(struct drm_i915_private *i915,
 				    struct i915_gem_timeline *timeline,
 				    const char *name,
@@ -91,8 +300,14 @@ void i915_gem_timeline_fini(struct i915_gem_timeline *timeline)
 		struct intel_timeline *tl = &timeline->engine[i];
 
 		GEM_BUG_ON(!list_empty(&tl->requests));
+
+		sync_free(tl->sync);
 	}
 
 	list_del(&timeline->link);
 	kfree(timeline->name);
 }
+
+#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
+#include "selftests/i915_gem_timeline.c"
+#endif
diff --git a/drivers/gpu/drm/i915/i915_gem_timeline.h b/drivers/gpu/drm/i915/i915_gem_timeline.h
index 6c53e14cab2a..22a80daa6efa 100644
--- a/drivers/gpu/drm/i915/i915_gem_timeline.h
+++ b/drivers/gpu/drm/i915/i915_gem_timeline.h
@@ -26,10 +26,13 @@
 #define I915_GEM_TIMELINE_H
 
 #include <linux/list.h>
+#include <linux/radix-tree.h>
 
+#include "i915_utils.h"
 #include "i915_gem_request.h"
 
 struct i915_gem_timeline;
+struct intel_timeline_sync;
 
 struct intel_timeline {
 	u64 fence_context;
@@ -55,6 +58,8 @@ struct intel_timeline {
 	 * struct_mutex.
 	 */
 	struct i915_gem_active last_request;
+
+	struct intel_timeline_sync *sync;
 	u32 sync_seqno[I915_NUM_ENGINES];
 
 	struct i915_gem_timeline *common;
@@ -75,4 +80,7 @@ int i915_gem_timeline_init(struct drm_i915_private *i915,
 int i915_gem_timeline_init__global(struct drm_i915_private *i915);
 void i915_gem_timeline_fini(struct i915_gem_timeline *tl);
 
+bool intel_timeline_sync_get(struct intel_timeline *tl, u64 id, u32 seqno);
+int intel_timeline_sync_set(struct intel_timeline *tl, u64 id, u32 seqno);
+
 #endif
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_timeline.c b/drivers/gpu/drm/i915/selftests/i915_gem_timeline.c
new file mode 100644
index 000000000000..c0bb8ecac93b
--- /dev/null
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_timeline.c
@@ -0,0 +1,123 @@
+/*
+ * Copyright © 2017 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ */
+
+#include "../i915_selftest.h"
+#include "mock_gem_device.h"
+
+static int igt_seqmap(void *arg)
+{
+	struct drm_i915_private *i915 = arg;
+	const struct {
+		const char *name;
+		u32 seqno;
+		bool expected;
+		bool set;
+	} pass[] = {
+		{ "unset", 0, false, false },
+		{ "new", 0, false, true },
+		{ "0a", 0, true, true },
+		{ "1a", 1, false, true },
+		{ "1b", 1, true, true },
+		{ "0b", 0, true, false },
+		{ "2a", 2, false, true },
+		{ "4", 4, false, true },
+		{ "INT_MAX", INT_MAX, false, true },
+		{ "INT_MAX-1", INT_MAX-1, true, false },
+		{ "INT_MAX+1", (u32)INT_MAX+1, false, true },
+		{ "INT_MAX", INT_MAX, true, false },
+		{ "UINT_MAX", UINT_MAX, false, true },
+		{ "wrap", 0, false, true },
+		{ "unwrap", UINT_MAX, true, false },
+		{},
+	}, *p;
+	struct intel_timeline *tl;
+	int order, offset;
+	int ret;
+
+	tl = &i915->gt.global_timeline.engine[RCS];
+	for (p = pass; p->name; p++) {
+		for (order = 1; order < 64; order++) {
+			for (offset = -1; offset <= (order > 1); offset++) {
+				u64 ctx = BIT_ULL(order) + offset;
+
+				if (intel_timeline_sync_get(tl,
+							    ctx,
+							    p->seqno) != p->expected) {
+					pr_err("1: %s(ctx=%llu, seqno=%u) expected passed %s but failed\n",
+					       p->name, ctx, p->seqno, yesno(p->expected));
+					return -EINVAL;
+				}
+
+				if (p->set) {
+					ret = intel_timeline_sync_set(tl, ctx, p->seqno);
+					if (ret)
+						return ret;
+				}
+			}
+		}
+	}
+
+	tl = &i915->gt.global_timeline.engine[BCS];
+	for (order = 1; order < 64; order++) {
+		for (offset = -1; offset <= (order > 1); offset++) {
+			u64 ctx = BIT_ULL(order) + offset;
+
+			for (p = pass; p->name; p++) {
+				if (intel_timeline_sync_get(tl,
+							    ctx,
+							    p->seqno) != p->expected) {
+					pr_err("2: %s(ctx=%llu, seqno=%u) expected passed %s but failed\n",
+					       p->name, ctx, p->seqno, yesno(p->expected));
+					return -EINVAL;
+				}
+
+				if (p->set) {
+					ret = intel_timeline_sync_set(tl, ctx, p->seqno);
+					if (ret)
+						return ret;
+				}
+			}
+		}
+	}
+
+	return 0;
+}
+
+int i915_gem_timeline_mock_selftests(void)
+{
+	static const struct i915_subtest tests[] = {
+		SUBTEST(igt_seqmap),
+	};
+	struct drm_i915_private *i915;
+	int err;
+
+	i915 = mock_gem_device();
+	if (!i915)
+		return -ENOMEM;
+
+	err = i915_subtests(tests, i915);
+	drm_dev_unref(&i915->drm);
+
+	return err;
+}
diff --git a/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h b/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h
index be9a9ebf5692..8d0f50c25df8 100644
--- a/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h
+++ b/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h
@@ -12,6 +12,7 @@ selftest(sanitycheck, i915_mock_sanitycheck) /* keep first (igt selfcheck) */
 selftest(scatterlist, scatterlist_mock_selftests)
 selftest(uncore, intel_uncore_mock_selftests)
 selftest(breadcrumbs, intel_breadcrumbs_mock_selftests)
+selftest(timelines, i915_gem_timeline_mock_selftests)
 selftest(requests, i915_gem_request_mock_selftests)
 selftest(objects, i915_gem_object_mock_selftests)
 selftest(dmabuf, i915_gem_dmabuf_mock_selftests)
-- 
2.11.0

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH v2 6/9] drm/i915: Rename intel_timeline.sync_seqno[] to .global_sync[]

[Chris Wilson]

With the addition of the inter-context intel_time.sync map, having a
very similar sync_seqno[] is confusing. Aide the reader by denoting that
this a pre-allocated array for storing semaphore sync points wrt to the
global seqno.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/i915_gem_request.c  | 8 ++++----
 drivers/gpu/drm/i915/i915_gem_timeline.h | 2 +-
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index 6e68387aa097..87ca3453140a 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -218,8 +218,8 @@ static int reset_all_global_seqno(struct drm_i915_private *i915, u32 seqno)
 		tl->seqno = seqno;
 
 		list_for_each_entry(timeline, &i915->gt.timelines, link)
-			memset(timeline->engine[id].sync_seqno, 0,
-			       sizeof(timeline->engine[id].sync_seqno));
+			memset(timeline->engine[id].global_sync, 0,
+			       sizeof(timeline->engine[id].global_sync));
 	}
 
 	return 0;
@@ -688,7 +688,7 @@ i915_gem_request_await_request(struct drm_i915_gem_request *to,
 		return ret < 0 ? ret : 0;
 	}
 
-	if (seqno <= to->timeline->sync_seqno[from->engine->id])
+	if (seqno <= to->timeline->global_sync[from->engine->id])
 		return 0;
 
 	trace_i915_gem_ring_sync_to(to, from);
@@ -706,7 +706,7 @@ i915_gem_request_await_request(struct drm_i915_gem_request *to,
 			return ret;
 	}
 
-	to->timeline->sync_seqno[from->engine->id] = seqno;
+	to->timeline->global_sync[from->engine->id] = seqno;
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_gem_timeline.h b/drivers/gpu/drm/i915/i915_gem_timeline.h
index 4cb597401803..48437612bb4b 100644
--- a/drivers/gpu/drm/i915/i915_gem_timeline.h
+++ b/drivers/gpu/drm/i915/i915_gem_timeline.h
@@ -76,7 +76,7 @@ struct intel_timeline {
 	struct i915_gem_active last_request;
 
 	struct seqmap sync;
-	u32 sync_seqno[I915_NUM_ENGINES];
+	u32 global_sync[I915_NUM_ENGINES];
 
 	struct i915_gem_timeline *common;
 };
-- 
2.11.0

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH v2 6/9] drm/i915: Rename intel_timeline.sync_seqno[] to .global_sync[]

[Joonas Lahtinen]

On ke, 2017-04-12 at 13:48 +0100, Chris Wilson wrote:
> With the addition of the inter-context intel_time.sync map, having a

intel_timeline.sync?

> very similar sync_seqno[] is confusing. Aide the reader by denoting that
> this a pre-allocated array for storing semaphore sync points wrt to the
> global seqno.
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

Regards, Joonas
-- 
Joonas Lahtinen
Open Source Technology Center
Intel Corporation
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH v2 7/9] drm/i915: Confirm the request is still active before adding it to the await

[Chris Wilson]

Although we do check the completion-status of the request before
actually adding a wait on it (either to its submit fence or its
completion dma-fence), we currently do not check before adding it to the
dependency lists.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/i915_gem_request.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index 87ca3453140a..436693b12bae 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -665,6 +665,9 @@ i915_gem_request_await_request(struct drm_i915_gem_request *to,
 	GEM_BUG_ON(to == from);
 	GEM_BUG_ON(to->timeline == from->timeline);
 
+	if (i915_gem_request_completed(from))
+		return 0;
+
 	if (to->engine->schedule) {
 		ret = i915_priotree_add_dependency(to->i915,
 						   &to->priotree,
-- 
2.11.0

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH v2 7/9] drm/i915: Confirm the request is still active before adding it to the await

[Michał Winiarski]

On Wed, Apr 12, 2017 at 01:48:23PM +0100, Chris Wilson wrote:
> Although we do check the completion-status of the request before
> actually adding a wait on it (either to its submit fence or its
> completion dma-fence), we currently do not check before adding it to the
> dependency lists.
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>

-Michał

> ---
>  drivers/gpu/drm/i915/i915_gem_request.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
> index 87ca3453140a..436693b12bae 100644
> --- a/drivers/gpu/drm/i915/i915_gem_request.c
> +++ b/drivers/gpu/drm/i915/i915_gem_request.c
> @@ -665,6 +665,9 @@ i915_gem_request_await_request(struct drm_i915_gem_request *to,
>  	GEM_BUG_ON(to == from);
>  	GEM_BUG_ON(to->timeline == from->timeline);
>  
> +	if (i915_gem_request_completed(from))
> +		return 0;
> +
>  	if (to->engine->schedule) {
>  		ret = i915_priotree_add_dependency(to->i915,
>  						   &to->priotree,
> -- 
> 2.11.0
> 
> _______________________________________________
> 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

Re: [PATCH v2 7/9] drm/i915: Confirm the request is still active before adding it to the await

[Joonas Lahtinen]

On ke, 2017-04-12 at 13:48 +0100, Chris Wilson wrote:
> Although we do check the completion-status of the request before
> actually adding a wait on it (either to its submit fence or its
> completion dma-fence), we currently do not check before adding it to the
> dependency lists.
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Well, it's not going to hurt.

Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

Regards, Joonas
-- 
Joonas Lahtinen
Open Source Technology Center
Intel Corporation
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH v2 8/9] drm/i915: Do not record a successful syncpoint for a dma-await

[Chris Wilson]

As we may unwind the requests, even though the request we are awaiting
has a global_seqno that seqno may be revoked during the await and so we
can not reliably use it as a barrier for all future awaits on the same
timeline.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michał Winiarski <michal.winiarski@intel.com>
---
 drivers/gpu/drm/i915/i915_gem_request.c | 34 ++++++++++++++++-----------------
 1 file changed, 16 insertions(+), 18 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index 436693b12bae..17feeb68cdff 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -684,33 +684,31 @@ i915_gem_request_await_request(struct drm_i915_gem_request *to,
 	}
 
 	seqno = i915_gem_request_global_seqno(from);
-	if (!seqno) {
-		ret = i915_sw_fence_await_dma_fence(&to->submit,
-						    &from->fence, 0,
-						    GFP_KERNEL);
-		return ret < 0 ? ret : 0;
-	}
+	if (!seqno)
+		goto await_dma_fence;
 
-	if (seqno <= to->timeline->global_sync[from->engine->id])
-		return 0;
-
-	trace_i915_gem_ring_sync_to(to, from);
 	if (!i915.semaphores) {
-		if (!i915_spin_request(from, TASK_INTERRUPTIBLE, 2)) {
-			ret = i915_sw_fence_await_dma_fence(&to->submit,
-							    &from->fence, 0,
-							    GFP_KERNEL);
-			if (ret < 0)
-				return ret;
-		}
+		if (!__i915_spin_request(from, seqno, TASK_INTERRUPTIBLE, 2))
+			goto await_dma_fence;
 	} else {
+		if (seqno <= to->timeline->global_sync[from->engine->id])
+			return 0;
+
+		trace_i915_gem_ring_sync_to(to, from);
 		ret = to->engine->semaphore.sync_to(to, from);
 		if (ret)
 			return ret;
+
+		to->timeline->global_sync[from->engine->id] = seqno;
 	}
 
-	to->timeline->global_sync[from->engine->id] = seqno;
 	return 0;
+
+await_dma_fence:
+	ret = i915_sw_fence_await_dma_fence(&to->submit,
+					    &from->fence, 0,
+					    GFP_KERNEL);
+	return ret < 0 ? ret : 0;
 }
 
 int
-- 
2.11.0

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH v2 8/9] drm/i915: Do not record a successful syncpoint for a dma-await

[Michał Winiarski]

On Wed, Apr 12, 2017 at 01:48:24PM +0100, Chris Wilson wrote:
> As we may unwind the requests, even though the request we are awaiting
> has a global_seqno that seqno may be revoked during the await and so we
> can not reliably use it as a barrier for all future awaits on the same
> timeline.
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Michał Winiarski <michal.winiarski@intel.com>

Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>

-Michał

> ---
>  drivers/gpu/drm/i915/i915_gem_request.c | 34 ++++++++++++++++-----------------
>  1 file changed, 16 insertions(+), 18 deletions(-)
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH v2 9/9] drm/i915: Switch the global i915.semaphores check to a local predicate

[Chris Wilson]

Rather than use a global modparam, we can just check to see if the
engine has semaphores configured upon it.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/i915_gem_request.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index 17feeb68cdff..38307079bd35 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -687,10 +687,12 @@ i915_gem_request_await_request(struct drm_i915_gem_request *to,
 	if (!seqno)
 		goto await_dma_fence;
 
-	if (!i915.semaphores) {
+	if (!to->engine->semaphore.sync_to) {
 		if (!__i915_spin_request(from, seqno, TASK_INTERRUPTIBLE, 2))
 			goto await_dma_fence;
 	} else {
+		GEM_BUG_ON(!from->engine->semaphore.signal);
+
 		if (seqno <= to->timeline->global_sync[from->engine->id])
 			return 0;
 
-- 
2.11.0

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH v2 9/9] drm/i915: Switch the global i915.semaphores check to a local predicate

[Joonas Lahtinen]

On ke, 2017-04-12 at 13:48 +0100, Chris Wilson wrote:
> Rather than use a global modparam, we can just check to see if the
> engine has semaphores configured upon it.
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Should be good.

Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

Regars, Joonas
-- 
Joonas Lahtinen
Open Source Technology Center
Intel Corporation
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

✓ Fi.CI.BAT: success for series starting with [v2,1/9] drm/i915: Mark up clflushes as belonging to an unordered timeline

[Patchwork]

== Series Details ==

Series: series starting with [v2,1/9] drm/i915: Mark up clflushes as belonging to an unordered timeline
URL   : https://patchwork.freedesktop.org/series/22924/
State : success

== Summary ==

Series 22924v1 Series without cover letter
https://patchwork.freedesktop.org/api/1.0/series/22924/revisions/1/mbox/

Test gem_exec_fence:
        Subgroup await-hang-default:
                incomplete -> PASS       (fi-hsw-4770) fdo#99726
Test gem_exec_flush:
        Subgroup basic-batch-kernel-default-uc:
                fail       -> PASS       (fi-snb-2600) fdo#100007

fdo#99726 https://bugs.freedesktop.org/show_bug.cgi?id=99726
fdo#100007 https://bugs.freedesktop.org/show_bug.cgi?id=100007

fi-bdw-5557u     total:278  pass:267  dwarn:0   dfail:0   fail:0   skip:11  time:431s
fi-bdw-gvtdvm    total:278  pass:256  dwarn:8   dfail:0   fail:0   skip:14  time:425s
fi-bsw-n3050     total:278  pass:242  dwarn:0   dfail:0   fail:0   skip:36  time:575s
fi-bxt-j4205     total:278  pass:259  dwarn:0   dfail:0   fail:0   skip:19  time:507s
fi-byt-j1900     total:278  pass:254  dwarn:0   dfail:0   fail:0   skip:24  time:487s
fi-byt-n2820     total:278  pass:250  dwarn:0   dfail:0   fail:0   skip:28  time:487s
fi-hsw-4770      total:278  pass:262  dwarn:0   dfail:0   fail:0   skip:16  time:603s
fi-hsw-4770r     total:278  pass:262  dwarn:0   dfail:0   fail:0   skip:16  time:403s
fi-ilk-650       total:278  pass:228  dwarn:0   dfail:0   fail:0   skip:50  time:424s
fi-ivb-3520m     total:278  pass:260  dwarn:0   dfail:0   fail:0   skip:18  time:490s
fi-ivb-3770      total:278  pass:260  dwarn:0   dfail:0   fail:0   skip:18  time:463s
fi-kbl-7500u     total:278  pass:260  dwarn:0   dfail:0   fail:0   skip:18  time:455s
fi-kbl-7560u     total:278  pass:268  dwarn:0   dfail:0   fail:0   skip:10  time:567s
fi-skl-6260u     total:278  pass:268  dwarn:0   dfail:0   fail:0   skip:10  time:453s
fi-skl-6700hq    total:278  pass:261  dwarn:0   dfail:0   fail:0   skip:17  time:570s
fi-skl-6700k     total:278  pass:256  dwarn:4   dfail:0   fail:0   skip:18  time:457s
fi-skl-6770hq    total:278  pass:268  dwarn:0   dfail:0   fail:0   skip:10  time:496s
fi-skl-gvtdvm    total:278  pass:265  dwarn:0   dfail:0   fail:0   skip:13  time:434s
fi-snb-2520m     total:278  pass:250  dwarn:0   dfail:0   fail:0   skip:28  time:527s
fi-snb-2600      total:278  pass:249  dwarn:0   dfail:0   fail:0   skip:29  time:410s

53b9521a6f113b5fcfd8c6a1f2ba3194e2d21381 drm-tip: 2017y-04m-12d-12h-39m-39s UTC integration manifest
d00ced5 drm/i915: Switch the global i915.semaphores check to a local predicate
ee45627 drm/i915: Do not record a successful syncpoint for a dma-await
a3ec281 drm/i915: Confirm the request is still active before adding it to the await
4132e6e drm/i915: Rename intel_timeline.sync_seqno[] to .global_sync[]
77635c1 drm/i915: Squash repeated awaits on the same fence
037170d drm/i915: Redefine ptr_pack_bits() and friends
a06b39f drm/i915: Make ptr_unpack_bits() more function-like
0ed32ff drm/i915: Lift timeline ordering to await_dma_fence
0547c98 drm/i915: Mark up clflushes as belonging to an unordered timeline

== Logs ==

For more details see: https://intel-gfx-ci.01.org/CI/Patchwork_4492/
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

✗ Fi.CI.BAT: failure for series starting with [v2,1/9] drm/i915: Mark up clflushes as belonging to an unordered timeline (rev2)

[Patchwork]

== Series Details ==

Series: series starting with [v2,1/9] drm/i915: Mark up clflushes as belonging to an unordered timeline (rev2)
URL   : https://patchwork.freedesktop.org/series/22924/
State : failure

== Summary ==

scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/i915_perf.o' failed
make[4]: *** [drivers/gpu/drm/i915/i915_perf.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/opregion.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/opregion.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_dp_aux_backlight.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_dp_aux_backlight.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/dvo_tfp410.o' failed
make[4]: *** [drivers/gpu/drm/i915/dvo_tfp410.o] Error 1
  LD      net/ipv6/built-in.o
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_dsi_pll.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_dsi_pll.o] Error 1
  LD      drivers/video/console/built-in.o
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/sched_policy.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/sched_policy.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_bios.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_bios.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_panel.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_panel.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_dvo.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_dvo.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/display.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/display.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/cmd_parser.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/cmd_parser.o] Error 1
  LD      drivers/video/fbdev/built-in.o
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/firmware.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/firmware.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_cdclk.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_cdclk.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/dvo_ch7xxx.o' failed
make[4]: *** [drivers/gpu/drm/i915/dvo_ch7xxx.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_dsi_vbt.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_dsi_vbt.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_pm.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_pm.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_lspcon.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_lspcon.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_opregion.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_opregion.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_sdvo.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_sdvo.o] Error 1
  LD      drivers/video/built-in.o
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_sprite.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_sprite.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/render.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/render.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/handlers.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/handlers.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_lpe_audio.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_lpe_audio.o] Error 1
  LD      drivers/usb/gadget/udc/udc-core.o
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/mmio.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/mmio.o] Error 1
  LD      drivers/usb/gadget/udc/built-in.o
  LD      drivers/usb/gadget/built-in.o
  LD      drivers/scsi/scsi_mod.o
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_display.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_display.o] Error 1
  LD      drivers/pci/built-in.o
  LD      net/ipv4/built-in.o
scripts/Makefile.build:553: recipe for target 'drivers/gpu/drm/i915' failed
make[3]: *** [drivers/gpu/drm/i915] Error 2
make[3]: *** Waiting for unfinished jobs....
  LD      drivers/tty/serial/8250/8250_base.o
  LD      drivers/tty/serial/8250/built-in.o
  LD      drivers/tty/serial/built-in.o
  LD [M]  drivers/net/ethernet/intel/e1000/e1000.o
scripts/Makefile.build:553: recipe for target 'drivers/gpu/drm' failed
make[2]: *** [drivers/gpu/drm] Error 2
scripts/Makefile.build:553: recipe for target 'drivers/gpu' failed
make[1]: *** [drivers/gpu] Error 2
make[1]: *** Waiting for unfinished jobs....
  LD      fs/btrfs/btrfs.o
  CC      arch/x86/kernel/cpu/capflags.o
  LD      arch/x86/kernel/cpu/built-in.o
  LD      drivers/usb/core/usbcore.o
  LD      drivers/scsi/sd_mod.o
  LD      arch/x86/kernel/built-in.o
  LD      drivers/scsi/built-in.o
  LD      drivers/usb/core/built-in.o
  LD      fs/btrfs/built-in.o
  LD      arch/x86/built-in.o
  LD      drivers/tty/vt/built-in.o
  LD      drivers/tty/built-in.o
  LD [M]  drivers/net/ethernet/intel/igb/igb.o
  LD      drivers/usb/host/xhci-hcd.o
  LD      drivers/md/md-mod.o
  LD      drivers/md/built-in.o
  LD      net/core/built-in.o
  LD      net/built-in.o
  LD      fs/ext4/ext4.o
  LD      fs/ext4/built-in.o
  LD      fs/built-in.o
  LD      drivers/usb/host/built-in.o
  LD      drivers/usb/built-in.o
  LD [M]  drivers/net/ethernet/intel/e1000e/e1000e.o
  LD      drivers/net/ethernet/built-in.o
  LD      drivers/net/built-in.o
Makefile:1002: recipe for target 'drivers' failed
make: *** [drivers] Error 2

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

✗ Fi.CI.BAT: failure for series starting with [v2,1/9] drm/i915: Mark up clflushes as belonging to an unordered timeline (rev3)

[Patchwork]

== Series Details ==

Series: series starting with [v2,1/9] drm/i915: Mark up clflushes as belonging to an unordered timeline (rev3)
URL   : https://patchwork.freedesktop.org/series/22924/
State : failure

== Summary ==

scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/cfg_space.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/cfg_space.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_modes.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_modes.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_cdclk.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_cdclk.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/i915_gem_gtt.o' failed
make[4]: *** [drivers/gpu/drm/i915/i915_gem_gtt.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/dvo_ns2501.o' failed
make[4]: *** [drivers/gpu/drm/i915/dvo_ns2501.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/i915_vgpu.o' failed
make[4]: *** [drivers/gpu/drm/i915/i915_vgpu.o] Error 1
  LD      net/ipv6/ipv6.o
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_pm.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_pm.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/edid.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/edid.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/render.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/render.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_lvds.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_lvds.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_dp.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_dp.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/vgpu.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/vgpu.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/mmio.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/mmio.o] Error 1
  LD [M]  drivers/net/ethernet/intel/igbvf/igbvf.o
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/i915_gem.o' failed
make[4]: *** [drivers/gpu/drm/i915/i915_gem.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_i2c.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_i2c.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/sched_policy.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/sched_policy.o] Error 1
  LD      drivers/usb/gadget/udc/udc-core.o
  LD      net/ipv6/built-in.o
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_dp_mst.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_dp_mst.o] Error 1
  LD      drivers/usb/gadget/udc/built-in.o
  LD      drivers/usb/gadget/built-in.o
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_opregion.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_opregion.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/aperture_gm.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/aperture_gm.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_display.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_display.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/i915_oa_hsw.o' failed
make[4]: *** [drivers/gpu/drm/i915/i915_oa_hsw.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/gtt.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/gtt.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/cmd_parser.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/cmd_parser.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/i915_gpu_error.o' failed
make[4]: *** [drivers/gpu/drm/i915/i915_gpu_error.o] Error 1
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/intel_gvt.o' failed
make[4]: *** [drivers/gpu/drm/i915/intel_gvt.o] Error 1
  LD      drivers/gpu/drm/drm.o
  LD [M]  drivers/net/ethernet/intel/e1000/e1000.o
  LD      drivers/tty/serial/8250/8250_base.o
  LD      drivers/tty/serial/8250/built-in.o
  LD      fs/btrfs/btrfs.o
  LD      drivers/tty/serial/built-in.o
scripts/Makefile.build:294: recipe for target 'drivers/gpu/drm/i915/gvt/handlers.o' failed
make[4]: *** [drivers/gpu/drm/i915/gvt/handlers.o] Error 1
scripts/Makefile.build:553: recipe for target 'drivers/gpu/drm/i915' failed
make[3]: *** [drivers/gpu/drm/i915] Error 2
scripts/Makefile.build:553: recipe for target 'drivers/gpu/drm' failed
make[2]: *** [drivers/gpu/drm] Error 2
scripts/Makefile.build:553: recipe for target 'drivers/gpu' failed
make[1]: *** [drivers/gpu] Error 2
make[1]: *** Waiting for unfinished jobs....
  LD      fs/btrfs/built-in.o
  LD      drivers/tty/vt/built-in.o
  CC      arch/x86/kernel/cpu/capflags.o
  LD      drivers/tty/built-in.o
  LD      arch/x86/kernel/cpu/built-in.o
  LD      drivers/md/md-mod.o
  LD      drivers/md/built-in.o
  LD      arch/x86/kernel/built-in.o
  AR      lib/lib.a
  EXPORTS lib/lib-ksyms.o
  LD      lib/built-in.o
  LD      drivers/usb/core/usbcore.o
  LD      drivers/usb/core/built-in.o
  LD      arch/x86/built-in.o
  LD [M]  drivers/net/ethernet/intel/igb/igb.o
  LD [M]  drivers/net/ethernet/intel/e1000e/e1000e.o
  LD      drivers/usb/host/xhci-hcd.o
  LD      net/ipv4/built-in.o
  LD      drivers/usb/host/built-in.o
  LD      drivers/usb/built-in.o
  LD      fs/ext4/ext4.o
  LD      fs/ext4/built-in.o
  LD      fs/built-in.o
  LD      net/core/built-in.o
  LD      net/built-in.o
  LD      drivers/net/ethernet/built-in.o
  LD      drivers/net/built-in.o
Makefile:1002: recipe for target 'drivers' failed
make: *** [drivers] Error 2

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx