diff options
Diffstat (limited to 'meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.14.71/2055-drm-amdgpu-cleanup-job-reset-routine-v2.patch')
| -rw-r--r-- | meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.14.71/2055-drm-amdgpu-cleanup-job-reset-routine-v2.patch | 114 |
1 files changed, 114 insertions, 0 deletions
diff --git a/meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.14.71/2055-drm-amdgpu-cleanup-job-reset-routine-v2.patch b/meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.14.71/2055-drm-amdgpu-cleanup-job-reset-routine-v2.patch new file mode 100644 index 00000000..dbe748ce --- /dev/null +++ b/meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.14.71/2055-drm-amdgpu-cleanup-job-reset-routine-v2.patch @@ -0,0 +1,114 @@ +From 22c6aecb8e3ec8a1b6c1fb43076e86a7de4c3780 Mon Sep 17 00:00:00 2001 +From: Monk Liu <Monk.Liu@amd.com> +Date: Mon, 16 Oct 2017 19:46:43 +0800 +Subject: [PATCH 2055/4131] drm/amdgpu:cleanup job reset routine(v2) + +merge the setting guilty on context into this function +to avoid implement extra routine. + +v2: +go through entity list and compare the fence_ctx +before operate on the entity, otherwise the entity +may be just a wild pointer + +Change-Id: I7a0063464fdc85d5ac9080046380e745565ff540 +Signed-off-by: Monk Liu <Monk.Liu@amd.com> +Reviewed-by: Chunming Zhou <David1.Zhou@amd.com> + + Conflicts: + drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +--- + drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 ++-- + drivers/gpu/drm/amd/scheduler/gpu_scheduler.c | 31 ++++++++++++++++++++++++++- + drivers/gpu/drm/amd/scheduler/gpu_scheduler.h | 2 +- + 3 files changed, 33 insertions(+), 4 deletions(-) + +diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +index d0aaed4..29349bf 100644 +--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c ++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +@@ -2896,7 +2896,7 @@ int amdgpu_sriov_gpu_reset(struct amdgpu_device *adev, struct amdgpu_job *job) + amd_sched_job_kickout(&job->base); + + /* only do job_reset on the hang ring if @job not NULL */ +- amd_sched_hw_job_reset(&ring->sched); ++ amd_sched_hw_job_reset(&ring->sched, NULL); + + /* after all hw jobs are reset, hw fence is meaningless, so force_completion */ + amdgpu_fence_driver_force_completion(ring); +@@ -3017,7 +3017,7 @@ int amdgpu_gpu_reset(struct amdgpu_device *adev) + if (!ring || !ring->sched.thread) + continue; + kthread_park(ring->sched.thread); +- amd_sched_hw_job_reset(&ring->sched); ++ amd_sched_hw_job_reset(&ring->sched, NULL); + /* after all hw jobs are reset, hw fence is meaningless, so force_completion */ + amdgpu_fence_driver_force_completion(ring); + } +diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c +index 041510e..2634846 100644 +--- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c ++++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c +@@ -443,9 +443,18 @@ static void amd_sched_job_timedout(struct work_struct *work) + job->sched->ops->timedout_job(job); + } + +-void amd_sched_hw_job_reset(struct amd_gpu_scheduler *sched) ++static void amd_sched_set_guilty(struct amd_sched_job *s_job) ++{ ++ if (atomic_inc_return(&s_job->karma) > s_job->sched->hang_limit) ++ if (s_job->s_entity->guilty) ++ atomic_set(s_job->s_entity->guilty, 1); ++} ++ ++void amd_sched_hw_job_reset(struct amd_gpu_scheduler *sched, struct amd_sched_job *bad) + { + struct amd_sched_job *s_job; ++ struct amd_sched_entity *entity, *tmp; ++ int i;; + + spin_lock(&sched->job_list_lock); + list_for_each_entry_reverse(s_job, &sched->ring_mirror_list, node) { +@@ -458,6 +467,26 @@ void amd_sched_hw_job_reset(struct amd_gpu_scheduler *sched) + } + } + spin_unlock(&sched->job_list_lock); ++ ++ if (bad) { ++ bool found = false; ++ ++ for (i = AMD_SCHED_PRIORITY_MIN; i < AMD_SCHED_PRIORITY_MAX; i++ ) { ++ struct amd_sched_rq *rq = &sched->sched_rq[i]; ++ ++ spin_lock(&rq->lock); ++ list_for_each_entry_safe(entity, tmp, &rq->entities, list) { ++ if (bad->s_fence->scheduled.context == entity->fence_context) { ++ found = true; ++ amd_sched_set_guilty(bad); ++ break; ++ } ++ } ++ spin_unlock(&rq->lock); ++ if (found) ++ break; ++ } ++ } + } + + void amd_sched_job_kickout(struct amd_sched_job *s_job) +diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h +index 7342763..64c887f 100644 +--- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h ++++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h +@@ -173,7 +173,7 @@ int amd_sched_job_init(struct amd_sched_job *job, + struct amd_gpu_scheduler *sched, + struct amd_sched_entity *entity, + void *owner); +-void amd_sched_hw_job_reset(struct amd_gpu_scheduler *sched); ++void amd_sched_hw_job_reset(struct amd_gpu_scheduler *sched, struct amd_sched_job *job); + void amd_sched_job_recovery(struct amd_gpu_scheduler *sched); + bool amd_sched_dependency_optimized(struct dma_fence* fence, + struct amd_sched_entity *entity); +-- +2.7.4 + |