Skip to contentSkip to conversation

drm/scheduler: Partially revert "drm/scheduler: track GPU active time per entity"

PATCH In review xinhui pan 2 participants 3 messages Aug 17, 2023

Subsystems (3)
lore
Share & tools
lore

xinhui pan

author
#0 lore raw

PATCH 1/1 drm/scheduler: Partially revert "drm/scheduler: track GPU active time per entity"

This patch partially revert commit df622729ddbf ("drm/scheduler: track GPU active time per entity") which touchs entity without any reference.

I notice there is one memory overwritten from gpu scheduler side. The case is like below.
A(drm_sched_main) B(vm fini)
drm_sched_job_begin drm_sched_entity_kill

	//job in pending_list			wait_for_completion
complete_all				...
...					kfree entity
drm_sched_get_cleanup_job
	//fetch job from pending_list
	access job->entity //memory overwitten

As long as we can NOT guarantee entity is alive in this case, lets revert it for now.

---
 drivers/gpu/drm/scheduler/sched_main.c | 6 ------
 1 file changed, 6 deletions(-)
drivers/gpu/drm/scheduler/sched_main.c +0 -6
index 602361c690c9..1b3f1a6a8514 100644
@@ -907,12 +907,6 @@ drm_sched_get_cleanup_job(struct drm_gpu_scheduler *sched)
 
 	spin_unlock(&sched->job_list_lock);
 
-	if (job) {
-		job->entity->elapsed_ns += ktime_to_ns(
-			ktime_sub(job->s_fence->finished.timestamp,
-				  job->s_fence->scheduled.timestamp));
-	}
-
 	return job;
 }
 
Signature · 1 line
2.34.1
#1 lore raw

[AMD Official Use Only - General]

Can we just add kref for entity?
Or just collect such job time usage somewhere else?

-----Original Message-----
From: Pan, Xinhui <[email protected]>
Sent: Thursday, August 17, 2023 1:05 PM
To: [email protected]
Cc: Tuikov, Luben <[email protected]>; [email protected]; [email protected]; [email protected]; Koenig, Christian <[email protected]>; Pan, Xinhui <[email protected]>
Subject: [PATCH] drm/scheduler: Partially revert "drm/scheduler: track GPU active time per entity"

This patch partially revert commit df622729ddbf ("drm/scheduler: track GPU active time per entity") which touchs entity without any reference.

I notice there is one memory overwritten from gpu scheduler side. The case is like below.
A(drm_sched_main) B(vm fini)
drm_sched_job_begin drm_sched_entity_kill

        //job in pending_list                   wait_for_completion
complete_all                            ...
...                                     kfree entity
drm_sched_get_cleanup_job
        //fetch job from pending_list
        access job->entity //memory overwitten

As long as we can NOT guarantee entity is alive in this case, lets revert it for now.

---
 drivers/gpu/drm/scheduler/sched_main.c | 6 ------
 1 file changed, 6 deletions(-)
drivers/gpu/drm/scheduler/sched_main.c +0 -6
index 602361c690c9..1b3f1a6a8514 100644
@@ -907,12 +907,6 @@ drm_sched_get_cleanup_job(struct drm_gpu_scheduler *sched)
 
        spin_unlock(&sched->job_list_lock);
 
-       if (job) {
-               job->entity->elapsed_ns += ktime_to_ns(
-                       ktime_sub(job->s_fence->finished.timestamp,
-                                 job->s_fence->scheduled.timestamp));
-       }
-
        return job;
 }
 

--
2.34.1

#2 lore raw
In reply to xinhui pan

[Public]

Hi Xinhui,

That patch has been reverted on Linux mainline.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/gpu/drm/scheduler/sched_main.c?h=v6.5-rc6&id=baad10973fdb442912af676de3348e80bd8fe602

Regards,
Guchun

Quoted 54 lines

-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of xinhui pan
Sent: Thursday, August 17, 2023 1:05 PM
To: [email protected]
Cc: Pan, Xinhui <[email protected]>; [email protected]; Tuikov, Luben <[email protected]>; [email protected]; Koenig, Christian <[email protected]>; [email protected]
Subject: [PATCH] drm/scheduler: Partially revert "drm/scheduler: track GPU active time per entity"

This patch partially revert commit df622729ddbf ("drm/scheduler: track GPU active time per entity") which touchs entity without any reference.

I notice there is one memory overwritten from gpu scheduler side. The case is like below.
A(drm_sched_main) B(vm fini)
drm_sched_job_begin drm_sched_entity_kill

      //job in pending_list                   wait_for_completion
complete_all                          ...
...                                   kfree entity
drm_sched_get_cleanup_job
      //fetch job from pending_list
      access job->entity //memory overwitten

As long as we can NOT guarantee entity is alive in this case, lets revert it for now.

---
 drivers/gpu/drm/scheduler/sched_main.c | 6 ------
 1 file changed, 6 deletions(-)
drivers/gpu/drm/scheduler/sched_main.c +0 -7
index 602361c690c9..1b3f1a6a8514 100644
@@ -907,12 +907,6 @@ drm_sched_get_cleanup_job(struct
 drm_gpu_scheduler *sched)
 
      spin_unlock(&sched->job_list_lock);
 
-     if (job) {
-             job->entity->elapsed_ns += ktime_to_ns(
-                     ktime_sub(job->s_fence->finished.timestamp,
-                               job->s_fence->scheduled.timestamp));
-     }
-
      return job;
 }
 
--

2.34.1