[V2] drm/sched: fix fence reference count leak

Commit Message

Qianyi Liu Feb. 25, 2025, 9:41 a.m. UTC

From: qianyi liu <[e-mail]>

We leaked last_scheduled fences when the entity was being killed and the
fence callback add fails.

To fix this, we should decrement the reference count of prev when
dma_fence_add_callback() fails, ensuring proper balance.

v2:
 * Make commit message more clearly. (Philipp and Matt)
 * Add "Fixes: " tag and put the stable kernel on Cc. (Philipp)
 * Correct subject line from "drm/scheduler" to "drm/sched". (Philipp)

Cc: [e-mail]
Fixes: 2fdb8a8f07c2 ("drm/scheduler: rework entity flush, kill and fini")

Signed-off-by: qianyi liu <[e-mail]>
---
 drivers/gpu/drm/scheduler/sched_entity.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

Comments

Markus Elfring Feb. 25, 2025, 2:17 p.m. UTC | #1

…
> fence callback add fails.

                     failed?


> To fix this, we should decrement the reference count of prev when
…

See also:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst?h=v6.14-rc4#n94


> v2:

Patch version descriptions may be specified behind the marker line.
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst?h=v6.14-rc4#n763

Regards,
Markus

Qianyi Liu Feb. 26, 2025, 12:10 a.m. UTC | #2

Hello Markus,

>…
>> fence callback add fails.

>                     failed?


>> To fix this, we should decrement the reference count of prev when
>…

> See also:
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst?h=v6.14-rc4#n94


>> v2:

> Patch version descriptions may be specified behind the marker line.
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst?h=v6.14-rc4#n763

Thanks for your comments, I will update these in V3.

> Regards,
> Markus

Best Regards
Qianyi

Patch

diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
index 69bcf0e99d57..1c0c14bcf726 100644
--- a/drivers/gpu/drm/scheduler/sched_entity.c
+++ b/drivers/gpu/drm/scheduler/sched_entity.c
@@ -259,9 +259,12 @@  static void drm_sched_entity_kill(struct drm_sched_entity *entity)
 		struct drm_sched_fence *s_fence = job->s_fence;
 
 		dma_fence_get(&s_fence->finished);
-		if (!prev || dma_fence_add_callback(prev, &job->finish_cb,
-					   drm_sched_entity_kill_jobs_cb))
+		if (!prev ||
+		    dma_fence_add_callback(prev, &job->finish_cb,
+					   drm_sched_entity_kill_jobs_cb)) {
+			dma_fence_put(prev);
 			drm_sched_entity_kill_jobs_cb(NULL, &job->finish_cb);
+		}
 
 		prev = &s_fence->finished;
 	}