All of lore.kernel.org
 help / color / mirror / Atom feed
From: Daniel Vetter <daniel.vetter@ffwll.ch>
To: DRI Development <dri-devel@lists.freedesktop.org>
Cc: "Daniel Vetter" <daniel.vetter@ffwll.ch>,
	"Intel Graphics Development" <intel-gfx@lists.freedesktop.org>,
	"Steven Price" <steven.price@arm.com>,
	"Boris Brezillon" <boris.brezillon@collabora.com>,
	"Daniel Vetter" <daniel.vetter@intel.com>,
	"Lee Jones" <lee.jones@linaro.org>,
	"Christian König" <christian.koenig@amd.com>
Subject: [PATCH v4 02/18] drm/sched: Barriers are needed for entity->last_scheduled
Date: Mon, 12 Jul 2021 19:53:36 +0200	[thread overview]
Message-ID: <20210712175352.802687-3-daniel.vetter@ffwll.ch> (raw)
In-Reply-To: <20210712175352.802687-1-daniel.vetter@ffwll.ch>

It might be good enough on x86 with just READ_ONCE, but the write side
should then at least be WRITE_ONCE because x86 has total store order.

It's definitely not enough on arm.

Fix this proplery, which means
- explain the need for the barrier in both places
- point at the other side in each comment

Also pull out the !sched_list case as the first check, so that the
code flow is clearer.

While at it sprinkle some comments around because it was very
non-obvious to me what's actually going on here and why.

Note that we really need full barriers here, at first I thought
store-release and load-acquire on ->last_scheduled would be enough,
but we actually requiring ordering between that and the queue state.

v2: Put smp_rmp() in the right place and fix up comment (Andrey)

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Cc: "Christian König" <christian.koenig@amd.com>
Cc: Steven Price <steven.price@arm.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Cc: Lee Jones <lee.jones@linaro.org>
Cc: Boris Brezillon <boris.brezillon@collabora.com>
---
 drivers/gpu/drm/scheduler/sched_entity.c | 27 ++++++++++++++++++++++--
 1 file changed, 25 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
index f7347c284886..89e3f6eaf519 100644
--- a/drivers/gpu/drm/scheduler/sched_entity.c
+++ b/drivers/gpu/drm/scheduler/sched_entity.c
@@ -439,8 +439,16 @@ struct drm_sched_job *drm_sched_entity_pop_job(struct drm_sched_entity *entity)
 		dma_fence_set_error(&sched_job->s_fence->finished, -ECANCELED);
 
 	dma_fence_put(entity->last_scheduled);
+
 	entity->last_scheduled = dma_fence_get(&sched_job->s_fence->finished);
 
+	/*
+	 * If the queue is empty we allow drm_sched_entity_select_rq() to
+	 * locklessly access ->last_scheduled. This only works if we set the
+	 * pointer before we dequeue and if we a write barrier here.
+	 */
+	smp_wmb();
+
 	spsc_queue_pop(&entity->job_queue);
 	return sched_job;
 }
@@ -459,10 +467,25 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
 	struct drm_gpu_scheduler *sched;
 	struct drm_sched_rq *rq;
 
-	if (spsc_queue_count(&entity->job_queue) || !entity->sched_list)
+	/* single possible engine and already selected */
+	if (!entity->sched_list)
+		return;
+
+	/* queue non-empty, stay on the same engine */
+	if (spsc_queue_count(&entity->job_queue))
 		return;
 
-	fence = READ_ONCE(entity->last_scheduled);
+	/*
+	 * Only when the queue is empty are we guaranteed that the scheduler
+	 * thread cannot change ->last_scheduled. To enforce ordering we need
+	 * a read barrier here. See drm_sched_entity_pop_job() for the other
+	 * side.
+	 */
+	smp_rmb();
+
+	fence = entity->last_scheduled;
+
+	/* stay on the same engine if the previous job hasn't finished */
 	if (fence && !dma_fence_is_signaled(fence))
 		return;
 
-- 
2.32.0


WARNING: multiple messages have this Message-ID
From: Daniel Vetter <daniel.vetter@ffwll.ch>
To: DRI Development <dri-devel@lists.freedesktop.org>
Cc: "Andrey Grodzovsky" <andrey.grodzovsky@amd.com>,
	"Daniel Vetter" <daniel.vetter@ffwll.ch>,
	"Intel Graphics Development" <intel-gfx@lists.freedesktop.org>,
	"Steven Price" <steven.price@arm.com>,
	"Daniel Vetter" <daniel.vetter@intel.com>,
	"Lee Jones" <lee.jones@linaro.org>,
	"Christian König" <christian.koenig@amd.com>
Subject: [Intel-gfx] [PATCH v4 02/18] drm/sched: Barriers are needed for entity->last_scheduled
Date: Mon, 12 Jul 2021 19:53:36 +0200	[thread overview]
Message-ID: <20210712175352.802687-3-daniel.vetter@ffwll.ch> (raw)
In-Reply-To: <20210712175352.802687-1-daniel.vetter@ffwll.ch>

It might be good enough on x86 with just READ_ONCE, but the write side
should then at least be WRITE_ONCE because x86 has total store order.

It's definitely not enough on arm.

Fix this proplery, which means
- explain the need for the barrier in both places
- point at the other side in each comment

Also pull out the !sched_list case as the first check, so that the
code flow is clearer.

While at it sprinkle some comments around because it was very
non-obvious to me what's actually going on here and why.

Note that we really need full barriers here, at first I thought
store-release and load-acquire on ->last_scheduled would be enough,
but we actually requiring ordering between that and the queue state.

v2: Put smp_rmp() in the right place and fix up comment (Andrey)

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Cc: "Christian König" <christian.koenig@amd.com>
Cc: Steven Price <steven.price@arm.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Cc: Lee Jones <lee.jones@linaro.org>
Cc: Boris Brezillon <boris.brezillon@collabora.com>
---
 drivers/gpu/drm/scheduler/sched_entity.c | 27 ++++++++++++++++++++++--
 1 file changed, 25 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
index f7347c284886..89e3f6eaf519 100644
--- a/drivers/gpu/drm/scheduler/sched_entity.c
+++ b/drivers/gpu/drm/scheduler/sched_entity.c
@@ -439,8 +439,16 @@ struct drm_sched_job *drm_sched_entity_pop_job(struct drm_sched_entity *entity)
 		dma_fence_set_error(&sched_job->s_fence->finished, -ECANCELED);
 
 	dma_fence_put(entity->last_scheduled);
+
 	entity->last_scheduled = dma_fence_get(&sched_job->s_fence->finished);
 
+	/*
+	 * If the queue is empty we allow drm_sched_entity_select_rq() to
+	 * locklessly access ->last_scheduled. This only works if we set the
+	 * pointer before we dequeue and if we a write barrier here.
+	 */
+	smp_wmb();
+
 	spsc_queue_pop(&entity->job_queue);
 	return sched_job;
 }
@@ -459,10 +467,25 @@ void drm_sched_entity_select_rq(struct drm_sched_entity *entity)
 	struct drm_gpu_scheduler *sched;
 	struct drm_sched_rq *rq;
 
-	if (spsc_queue_count(&entity->job_queue) || !entity->sched_list)
+	/* single possible engine and already selected */
+	if (!entity->sched_list)
+		return;
+
+	/* queue non-empty, stay on the same engine */
+	if (spsc_queue_count(&entity->job_queue))
 		return;
 
-	fence = READ_ONCE(entity->last_scheduled);
+	/*
+	 * Only when the queue is empty are we guaranteed that the scheduler
+	 * thread cannot change ->last_scheduled. To enforce ordering we need
+	 * a read barrier here. See drm_sched_entity_pop_job() for the other
+	 * side.
+	 */
+	smp_rmb();
+
+	fence = entity->last_scheduled;
+
+	/* stay on the same engine if the previous job hasn't finished */
 	if (fence && !dma_fence_is_signaled(fence))
 		return;
 
-- 
2.32.0

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

  parent reply	other threads:[~2021-07-12 20:02 UTC|newest]

Thread overview: 112+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-07-12 17:53 [PATCH v4 00/18] drm/sched dependency tracking and dma-resv fixes Daniel Vetter
2021-07-12 17:53 ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53 ` [PATCH v4 01/18] drm/sched: Split drm_sched_job_init Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53   ` Daniel Vetter
2021-07-12 20:22   ` Emma Anholt
2021-07-12 20:22     ` [Intel-gfx] " Emma Anholt
2021-07-12 20:22     ` Emma Anholt
2021-07-13  6:40   ` Christian König
2021-07-13  6:40     ` [Intel-gfx] " Christian König
2021-07-13  6:40     ` Christian König
2021-07-13  6:53     ` Daniel Vetter
2021-07-13  6:53       ` [Intel-gfx] " Daniel Vetter
2021-07-13  6:53       ` Daniel Vetter
2021-07-12 17:53 ` Daniel Vetter [this message]
2021-07-12 17:53   ` [Intel-gfx] [PATCH v4 02/18] drm/sched: Barriers are needed for entity->last_scheduled Daniel Vetter
2021-07-13  6:35   ` Christian König
2021-07-13  6:35     ` [Intel-gfx] " Christian König
2021-07-13  6:50     ` Daniel Vetter
2021-07-13  6:50       ` [Intel-gfx] " Daniel Vetter
2021-07-13  7:25       ` Christian König
2021-07-13  7:25         ` [Intel-gfx] " Christian König
2021-07-13  9:10         ` Daniel Vetter
2021-07-13  9:10           ` [Intel-gfx] " Daniel Vetter
2021-07-13 11:20           ` Christian König
2021-07-13 11:20             ` [Intel-gfx] " Christian König
2021-07-13 16:11           ` Andrey Grodzovsky
2021-07-13 16:11             ` [Intel-gfx] " Andrey Grodzovsky
2021-07-13 16:45             ` Daniel Vetter
2021-07-13 16:45               ` [Intel-gfx] " Daniel Vetter
2021-07-14 22:12               ` Andrey Grodzovsky
2021-07-14 22:12                 ` [Intel-gfx] " Andrey Grodzovsky
2021-07-15 10:16                 ` Daniel Vetter
2021-07-15 10:16                   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53 ` [PATCH v4 03/18] drm/sched: Add dependency tracking Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53   ` Daniel Vetter
2021-07-27 11:09   ` Daniel Vetter
2021-07-27 11:09     ` [Intel-gfx] " Daniel Vetter
2021-07-27 11:09     ` Daniel Vetter
2021-07-28 11:28     ` [Linaro-mm-sig] " Christian König
2021-07-28 11:28       ` [Intel-gfx] " Christian König
2021-07-28 11:28       ` Christian König
2021-07-28 12:09       ` Daniel Vetter
2021-07-28 12:09         ` [Intel-gfx] " Daniel Vetter
2021-07-28 12:09         ` Daniel Vetter
2021-07-28 12:46         ` Christian König
2021-07-28 12:46           ` [Intel-gfx] " Christian König
2021-07-28 12:46           ` Christian König
2021-07-28 15:20         ` Melissa Wen
2021-07-28 15:20           ` [Intel-gfx] " Melissa Wen
2021-07-28 15:20           ` Melissa Wen
2021-07-12 17:53 ` [PATCH v4 04/18] drm/sched: drop entity parameter from drm_sched_push_job Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53   ` Daniel Vetter
2021-07-12 17:53 ` [PATCH v4 05/18] drm/sched: improve docs around drm_sched_entity Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53 ` [PATCH v4 06/18] drm/panfrost: use scheduler dependency tracking Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53   ` Daniel Vetter
2021-07-12 17:53 ` [PATCH v4 07/18] drm/lima: " Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53   ` Daniel Vetter
2021-07-12 17:53 ` [PATCH v4 08/18] drm/v3d: Move drm_sched_job_init to v3d_job_init Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-14  9:34   ` Melissa Wen
2021-07-14  9:34     ` [Intel-gfx] " Melissa Wen
2021-07-12 17:53 ` [PATCH v4 09/18] drm/v3d: Use scheduler dependency handling Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-14  9:37   ` Melissa Wen
2021-07-14  9:37     ` [Intel-gfx] " Melissa Wen
2021-07-12 17:53 ` [PATCH v4 10/18] drm/etnaviv: " Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53   ` Daniel Vetter
2021-07-12 17:53 ` [PATCH v4 11/18] drm/gem: Delete gem array fencing helpers Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53   ` Daniel Vetter
2021-07-12 17:53 ` [PATCH v4 12/18] drm/sched: Don't store self-dependencies Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53 ` [PATCH v4 13/18] drm/sched: Check locking in drm_sched_job_await_implicit Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53 ` [PATCH v4 14/18] drm/msm: Don't break exclusive fence ordering Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53   ` Daniel Vetter
2021-07-13 16:55   ` Rob Clark
2021-07-13 16:55     ` [Intel-gfx] " Rob Clark
2021-07-13 16:55     ` Rob Clark
2021-07-13 16:58     ` Daniel Vetter
2021-07-13 16:58       ` [Intel-gfx] " Daniel Vetter
2021-07-13 16:58       ` Daniel Vetter
2021-07-13 17:46       ` Rob Clark
2021-07-13 17:46         ` [Intel-gfx] " Rob Clark
2021-07-13 17:46         ` Rob Clark
2021-07-13 17:45         ` Daniel Vetter
2021-07-13 17:45           ` [Intel-gfx] " Daniel Vetter
2021-07-13 17:45           ` Daniel Vetter
2021-07-12 17:53 ` [PATCH v4 15/18] drm/etnaviv: " Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53 ` [PATCH v4 16/18] drm/i915: delete exclude argument from i915_sw_fence_await_reservation Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53 ` [PATCH v4 17/18] drm/i915: Don't break exclusive fence ordering Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53 ` [PATCH v4 18/18] dma-resv: Give the docs a do-over Daniel Vetter
2021-07-12 17:53   ` [Intel-gfx] " Daniel Vetter
2021-07-12 17:53   ` Daniel Vetter
2021-07-12 20:47 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for drm/sched dependency tracking and dma-resv fixes (rev3) Patchwork
2021-07-12 21:13 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork
2021-07-12 23:43 ` [Intel-gfx] ✗ Fi.CI.IGT: failure " Patchwork
2021-07-27 11:51 ` [PATCH v4 00/18] drm/sched dependency tracking and dma-resv fixes Boris Brezillon
2021-07-27 11:51   ` [Intel-gfx] " Boris Brezillon
2021-07-27 14:47 ` Melissa Wen
2021-07-27 14:47   ` [Intel-gfx] " Melissa Wen

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210712175352.802687-3-daniel.vetter@ffwll.ch \
    --to=daniel.vetter@ffwll.ch \
    --cc=boris.brezillon@collabora.com \
    --cc=christian.koenig@amd.com \
    --cc=daniel.vetter@intel.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=intel-gfx@lists.freedesktop.org \
    --cc=lee.jones@linaro.org \
    --cc=steven.price@arm.com \
    --subject='Re: [PATCH v4 02/18] drm/sched: Barriers are needed for entity->last_scheduled' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.