From: =?UTF-8?q?Christian=20K=C3=B6nig?= <christian.koenig@amd.com>
Date: Thu, 26 Jul 2018 13:43:49 +0200
Subject: drm/scheduler: only kill entity if last user is killed v2
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Git-commit: 43bce41cf48eb51eab5ad9e0d40ed382a7bb61d7
Patch-mainline: v4.19-rc1
References: FATE#326289 FATE#326079 FATE#326049 FATE#322398 FATE#326166

Note which task is using the entity and only kill it if the last user of
the entity is killed. This should prevent problems when entities are leaked to
child processes.

v2: add missing kernel doc

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Acked-by: Nayan Deshmukh <nayan26deshmukh@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Petr Tesarik <ptesarik@suse.com>
---
 drivers/gpu/drm/scheduler/gpu_scheduler.c |    6 +++++-
 include/drm/gpu_scheduler.h               |    2 ++
 2 files changed, 7 insertions(+), 1 deletion(-)

--- a/drivers/gpu/drm/scheduler/gpu_scheduler.c
+++ b/drivers/gpu/drm/scheduler/gpu_scheduler.c
@@ -275,6 +275,7 @@ static void drm_sched_entity_kill_jobs_c
 long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout)
 {
 	struct drm_gpu_scheduler *sched;
+	struct task_struct *last_user;
 	long ret = timeout;
 
 	sched = entity->rq->sched;
@@ -295,7 +296,9 @@ long drm_sched_entity_flush(struct drm_s
 
 
 	/* For killed process disable any more IBs enqueue right now */
-	if ((current->flags & PF_EXITING) && (current->exit_code == SIGKILL))
+	last_user = cmpxchg(&entity->last_user, current->group_leader, NULL);
+	if ((!last_user || last_user == current->group_leader) &&
+	    (current->flags & PF_EXITING) && (current->exit_code == SIGKILL))
 		drm_sched_entity_set_rq(entity, NULL);
 
 	return ret;
@@ -541,6 +544,7 @@ void drm_sched_entity_push_job(struct dr
 
 	trace_drm_sched_job(sched_job, entity);
 
+	WRITE_ONCE(entity->last_user, current->group_leader);
 	first = spsc_queue_push(&entity->job_queue, &sched_job->queue_node);
 
 	/* first job wakes up scheduler */
--- a/include/drm/gpu_scheduler.h
+++ b/include/drm/gpu_scheduler.h
@@ -66,6 +66,7 @@ enum drm_sched_priority {
  * @guilty: points to ctx's guilty.
  * @fini_status: contains the exit status in case the process was signalled.
  * @last_scheduled: points to the finished fence of the last scheduled job.
+ * @last_user: last group leader pushing a job into the entity.
  *
  * Entities will emit jobs in order to their corresponding hardware
  * ring, and the scheduler will alternate between entities based on
@@ -85,6 +86,7 @@ struct drm_sched_entity {
 	struct dma_fence_cb		cb;
 	atomic_t			*guilty;
 	struct dma_fence                *last_scheduled;
+	struct task_struct		*last_user;
 };
 
 /**