drm/amdgpu: put job to list before done

the mirror_list will be used for later time out detect
feature.  This is needed to properly detect a GPU
timeout with the scheduler.

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

3 files changed, 27 insertions, 0 deletions

diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
index 8d49ea2e4134..af846f208c67 100644
--- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
+++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
@@ -349,12 +349,15 @@ int amd_sched_job_init(struct amd_sched_job *job,
                                                 struct amd_sched_entity *entity,
                                                 void *owner, struct fence **fence)
 {
+        INIT_LIST_HEAD(&job->node);
         job->sched = sched;
         job->s_entity = entity;
         job->s_fence = amd_sched_fence_create(entity, owner);
         if (!job->s_fence)
                 return -ENOMEM;
 
+        job->s_fence->s_job = job;
+
         if (fence)
                 *fence = &job->s_fence->base;
         return 0;
@@ -408,6 +411,12 @@ static void amd_sched_process_job(struct fence *f, struct fence_cb *cb)
         unsigned long flags;
 
         atomic_dec(&sched->hw_rq_count);
+
+        /* remove job from ring_mirror_list */
+        spin_lock_irqsave(&sched->job_list_lock, flags);
+        list_del_init(&s_fence->s_job->node);
+        spin_unlock_irqrestore(&sched->job_list_lock, flags);
+
         amd_sched_fence_signal(s_fence);
         if (sched->timeout != MAX_SCHEDULE_TIMEOUT) {
                 cancel_delayed_work(&s_fence->dwork);
@@ -480,6 +489,7 @@ static int amd_sched_main(void *param)
                 }
 
                 atomic_inc(&sched->hw_rq_count);
+                amd_sched_job_pre_schedule(sched, sched_job);
                 fence = sched->ops->run_job(sched_job);
                 amd_sched_fence_scheduled(s_fence);
                 if (fence) {
@@ -527,6 +537,8 @@ int amd_sched_init(struct amd_gpu_scheduler *sched,
 
         init_waitqueue_head(&sched->wake_up_worker);
         init_waitqueue_head(&sched->job_scheduled);
+        INIT_LIST_HEAD(&sched->ring_mirror_list);
+        spin_lock_init(&sched->job_list_lock);
         atomic_set(&sched->hw_rq_count, 0);
         if (atomic_inc_return(&sched_fence_slab_ref) == 1) {
                 sched_fence_slab = kmem_cache_create(
diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h
index ee1e8127f863..2e3b8308186c 100644
--- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h
+++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h
@@ -76,6 +76,7 @@ struct amd_sched_fence {
         void                            *owner;
         struct delayed_work             dwork;
         struct list_head                list;
+        struct amd_sched_job    *s_job;
 };
 
 struct amd_sched_job {
@@ -85,6 +86,7 @@ struct amd_sched_job {
         bool    use_sched;      /* true if the job goes to scheduler */
         struct fence_cb                cb_free_job;
         struct work_struct             work_free_job;
+        struct list_head                           node;
 };
 
 extern const struct fence_ops amd_sched_fence_ops;
@@ -128,6 +130,8 @@ struct amd_gpu_scheduler {
         struct list_head                fence_list;
         spinlock_t                      fence_list_lock;
         struct task_struct              *thread;
+        struct list_head        ring_mirror_list;
+        spinlock_t                      job_list_lock;
 };
 
 int amd_sched_init(struct amd_gpu_scheduler *sched,
@@ -151,4 +155,6 @@ int amd_sched_job_init(struct amd_sched_job *job,
                                         struct amd_gpu_scheduler *sched,
                                         struct amd_sched_entity *entity,
                                         void *owner, struct fence **fence);
+void amd_sched_job_pre_schedule(struct amd_gpu_scheduler *sched ,
+                                                                struct amd_sched_job *s_job);
 #endif
diff --git a/drivers/gpu/drm/amd/scheduler/sched_fence.c b/drivers/gpu/drm/amd/scheduler/sched_fence.c
index dc115aea352b..33ddd38185d5 100644
--- a/drivers/gpu/drm/amd/scheduler/sched_fence.c
+++ b/drivers/gpu/drm/amd/scheduler/sched_fence.c
@@ -57,6 +57,15 @@ void amd_sched_fence_signal(struct amd_sched_fence *fence)
                 FENCE_TRACE(&fence->base, "was already signaled\n");
 }
 
+void amd_sched_job_pre_schedule(struct amd_gpu_scheduler *sched ,
+                                struct amd_sched_job *s_job)
+{
+        unsigned long flags;
+        spin_lock_irqsave(&sched->job_list_lock, flags);
+        list_add_tail(&s_job->node, &sched->ring_mirror_list);
+        spin_unlock_irqrestore(&sched->job_list_lock, flags);
+}
+
 void amd_sched_fence_scheduled(struct amd_sched_fence *s_fence)
 {
         struct fence_cb *cur, *tmp;