1 files changed, 172 insertions, 37 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
index 4b3000c21ef2..ddb9d3269357 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
@@ -30,6 +30,60 @@
 #include "amdgpu.h"
 #include "amdgpu_trace.h"
 #include "amdgpu_reset.h"
+#include "amdgpu_dev_coredump.h"
+#include "amdgpu_xgmi.h"
+
+static void amdgpu_job_do_core_dump(struct amdgpu_device *adev,
+				    struct amdgpu_job *job)
+{
+	int i;
+
+	dev_info(adev->dev, "Dumping IP State\n");
+	for (i = 0; i < adev->num_ip_blocks; i++)
+		if (adev->ip_blocks[i].version->funcs->dump_ip_state)
+			adev->ip_blocks[i].version->funcs
+				->dump_ip_state((void *)&adev->ip_blocks[i]);
+	dev_info(adev->dev, "Dumping IP State Completed\n");
+
+	amdgpu_coredump(adev, true, false, job);
+}
+
+static void amdgpu_job_core_dump(struct amdgpu_device *adev,
+				 struct amdgpu_job *job)
+{
+	struct list_head device_list, *device_list_handle =  NULL;
+	struct amdgpu_device *tmp_adev = NULL;
+	struct amdgpu_hive_info *hive = NULL;
+
+	if (!amdgpu_sriov_vf(adev))
+		hive = amdgpu_get_xgmi_hive(adev);
+	if (hive)
+		mutex_lock(&hive->hive_lock);
+	/*
+	 * Reuse the logic in amdgpu_device_gpu_recover() to build list of
+	 * devices for code dump
+	 */
+	INIT_LIST_HEAD(&device_list);
+	if (!amdgpu_sriov_vf(adev) && (adev->gmc.xgmi.num_physical_nodes > 1) && hive) {
+		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head)
+			list_add_tail(&tmp_adev->reset_list, &device_list);
+		if (!list_is_first(&adev->reset_list, &device_list))
+			list_rotate_to_front(&adev->reset_list, &device_list);
+		device_list_handle = &device_list;
+	} else {
+		list_add_tail(&adev->reset_list, &device_list);
+		device_list_handle = &device_list;
+	}
+
+	/* Do the coredump for each device */
+	list_for_each_entry(tmp_adev, device_list_handle, reset_list)
+		amdgpu_job_do_core_dump(tmp_adev, job);
+
+	if (hive) {
+		mutex_unlock(&hive->hive_lock);
+		amdgpu_put_xgmi_hive(hive);
+	}
+}
 
 static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 {
@@ -41,34 +95,82 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 	int r;
 
 	if (!drm_dev_enter(adev_to_drm(adev), &idx)) {
-		DRM_INFO("%s - device unplugged skipping recovery on scheduler:%s",
+		dev_info(adev->dev, "%s - device unplugged skipping recovery on scheduler:%s",
 			 __func__, s_job->sched->name);
 
 		/* Effectively the job is aborted as the device is gone */
 		return DRM_GPU_SCHED_STAT_ENODEV;
 	}
 
-
-	adev->job_hang = true;
+	/*
+	 * Do the coredump immediately after a job timeout to get a very
+	 * close dump/snapshot/representation of GPU's current error status
+	 * Skip it for SRIOV, since VF FLR will be triggered by host driver
+	 * before job timeout
+	 */
+	if (!amdgpu_sriov_vf(adev))
+		amdgpu_job_core_dump(adev, job);
 
 	if (amdgpu_gpu_recovery &&
 	    amdgpu_ring_soft_recovery(ring, job->vmid, s_job->s_fence->parent)) {
-		DRM_ERROR("ring %s timeout, but soft recovered\n",
-			  s_job->sched->name);
+		dev_err(adev->dev, "ring %s timeout, but soft recovered\n",
+			s_job->sched->name);
 		goto exit;
 	}
 
-	DRM_ERROR("ring %s timeout, signaled seq=%u, emitted seq=%u\n",
-		   job->base.sched->name, atomic_read(&ring->fence_drv.last_seq),
-		   ring->fence_drv.sync_seq);
+	dev_err(adev->dev, "ring %s timeout, signaled seq=%u, emitted seq=%u\n",
+		job->base.sched->name, atomic_read(&ring->fence_drv.last_seq),
+		ring->fence_drv.sync_seq);
 
 	ti = amdgpu_vm_get_task_info_pasid(ring->adev, job->pasid);
 	if (ti) {
-		DRM_ERROR("Process information: process %s pid %d thread %s pid %d\n",
-			  ti->process_name, ti->tgid, ti->task_name, ti->pid);
+		dev_err(adev->dev,
+			"Process information: process %s pid %d thread %s pid %d\n",
+			ti->process_name, ti->tgid, ti->task_name, ti->pid);
 		amdgpu_vm_put_task_info(ti);
 	}
 
+	/* attempt a per ring reset */
+	if (unlikely(adev->debug_disable_gpu_ring_reset)) {
+		dev_err(adev->dev, "Ring reset disabled by debug mask\n");
+	} else if (amdgpu_gpu_recovery && ring->funcs->reset) {
+		bool is_guilty;
+
+		dev_err(adev->dev, "Starting %s ring reset\n", s_job->sched->name);
+		/* stop the scheduler, but don't mess with the
+		 * bad job yet because if ring reset fails
+		 * we'll fall back to full GPU reset.
+		 */
+		drm_sched_wqueue_stop(&ring->sched);
+
+		/* for engine resets, we need to reset the engine,
+		 * but individual queues may be unaffected.
+		 * check here to make sure the accounting is correct.
+		 */
+		if (ring->funcs->is_guilty)
+			is_guilty = ring->funcs->is_guilty(ring);
+		else
+			is_guilty = true;
+
+		if (is_guilty)
+			dma_fence_set_error(&s_job->s_fence->finished, -ETIME);
+
+		r = amdgpu_ring_reset(ring, job->vmid);
+		if (!r) {
+			if (amdgpu_ring_sched_ready(ring))
+				drm_sched_stop(&ring->sched, s_job);
+			if (is_guilty) {
+				atomic_inc(&ring->adev->gpu_reset_counter);
+				amdgpu_fence_driver_force_completion(ring);
+			}
+			if (amdgpu_ring_sched_ready(ring))
+				drm_sched_start(&ring->sched, 0);
+			dev_err(adev->dev, "Ring %s reset succeeded\n", ring->sched.name);
+			drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE);
+			goto exit;
+		}
+		dev_err(adev->dev, "Ring %s reset failure\n", ring->sched.name);
+	}
 	dma_fence_set_error(&s_job->s_fence->finished, -ETIME);
 
 	if (amdgpu_device_should_recover_gpu(ring->adev)) {
@@ -77,11 +179,18 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 
 		reset_context.method = AMD_RESET_METHOD_NONE;
 		reset_context.reset_req_dev = adev;
+		reset_context.src = AMDGPU_RESET_SRC_JOB;
 		clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
 
+		/*
+		 * To avoid an unnecessary extra coredump, as we have already
+		 * got the very close representation of GPU's error status
+		 */
+		set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags);
+
 		r = amdgpu_device_gpu_recover(ring->adev, job, &reset_context);
 		if (r)
-			DRM_ERROR("GPU Recovery Failed: %d\n", r);
+			dev_err(adev->dev, "GPU Recovery Failed: %d\n", r);
 	} else {
 		drm_sched_suspend_timeout(&ring->sched);
 		if (amdgpu_sriov_vf(adev))
@@ -89,7 +198,6 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 	}
 
 exit:
-	adev->job_hang = false;
 	drm_dev_exit(idx);
 	return DRM_GPU_SCHED_STAT_NOMINAL;
 }
@@ -105,11 +213,6 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	if (!*job)
 		return -ENOMEM;
 
-	/*
-	 * Initialize the scheduler to at least some ring so that we always
-	 * have a pointer to adev.
-	 */
-	(*job)->base.sched = &adev->rings[0]->sched;
 	(*job)->vm = vm;
 
 	amdgpu_sync_create(&(*job)->explicit_sync);
@@ -163,20 +266,19 @@ void amdgpu_job_set_resources(struct amdgpu_job *job, struct amdgpu_bo *gds,
 
 void amdgpu_job_free_resources(struct amdgpu_job *job)
 {
-	struct amdgpu_ring *ring = to_amdgpu_ring(job->base.sched);
 	struct dma_fence *f;
 	unsigned i;
 
 	/* Check if any fences where initialized */
 	if (job->base.s_fence && job->base.s_fence->finished.ops)
 		f = &job->base.s_fence->finished;
-	else if (job->hw_fence.ops)
-		f = &job->hw_fence;
+	else if (job->hw_fence.base.ops)
+		f = &job->hw_fence.base;
 	else
 		f = NULL;
 
 	for (i = 0; i < job->num_ibs; ++i)
-		amdgpu_ib_free(ring->adev, &job->ibs[i], f);
+		amdgpu_ib_free(&job->ibs[i], f);
 }
 
 static void amdgpu_job_free_cb(struct drm_sched_job *s_job)
@@ -188,10 +290,10 @@ static void amdgpu_job_free_cb(struct drm_sched_job *s_job)
 	amdgpu_sync_free(&job->explicit_sync);
 
 	/* only put the hw fence if has embedded fence */
-	if (!job->hw_fence.ops)
+	if (!job->hw_fence.base.ops)
 		kfree(job);
 	else
-		dma_fence_put(&job->hw_fence);
+		dma_fence_put(&job->hw_fence.base);
 }
 
 void amdgpu_job_set_gang_leader(struct amdgpu_job *job,
@@ -220,10 +322,10 @@ void amdgpu_job_free(struct amdgpu_job *job)
 	if (job->gang_submit != &job->base.s_fence->scheduled)
 		dma_fence_put(job->gang_submit);
 
-	if (!job->hw_fence.ops)
+	if (!job->hw_fence.base.ops)
 		kfree(job);
 	else
-		dma_fence_put(&job->hw_fence);
+		dma_fence_put(&job->hw_fence.base);
 }
 
 struct dma_fence *amdgpu_job_submit(struct amdgpu_job *job)
@@ -259,26 +361,40 @@ amdgpu_job_prepare_job(struct drm_sched_job *sched_job,
 {
 	struct amdgpu_ring *ring = to_amdgpu_ring(s_entity->rq->sched);
 	struct amdgpu_job *job = to_amdgpu_job(sched_job);
-	struct dma_fence *fence = NULL;
+	struct dma_fence *fence;
 	int r;
 
-	/* Ignore soft recovered fences here */
 	r = drm_sched_entity_error(s_entity);
-	if (r && r != -ENODATA)
+	if (r)
 		goto error;
 
-	if (!fence && job->gang_submit)
+	if (job->gang_submit) {
 		fence = amdgpu_device_switch_gang(ring->adev, job->gang_submit);
+		if (fence)
+			return fence;
+	}
+
+	fence = amdgpu_device_enforce_isolation(ring->adev, ring, job);
+	if (fence)
+		return fence;
 
-	while (!fence && job->vm && !job->vmid) {
+	if (job->vm && !job->vmid) {
 		r = amdgpu_vmid_grab(job->vm, ring, job, &fence);
 		if (r) {
-			DRM_ERROR("Error getting VM ID (%d)\n", r);
+			dev_err(ring->adev->dev, "Error getting VM ID (%d)\n", r);
 			goto error;
 		}
+		/*
+		 * The VM structure might be released after the VMID is
+		 * assigned, we had multiple problems with people trying to use
+		 * the VM pointer so better set it to NULL.
+		 */
+		if (!fence)
+			job->vm = NULL;
+		return fence;
 	}
 
-	return fence;
+	return NULL;
 
 error:
 	dma_fence_set_error(&job->base.s_fence->finished, r);
@@ -304,12 +420,15 @@ static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job)
 		dma_fence_set_error(finished, -ECANCELED);
 
 	if (finished->error < 0) {
-		DRM_INFO("Skip scheduling IBs!\n");
+		dev_dbg(adev->dev, "Skip scheduling IBs in ring(%s)",
+			ring->name);
 	} else {
 		r = amdgpu_ib_schedule(ring, job->num_ibs, job->ibs, job,
 				       &fence);
 		if (r)
-			DRM_ERROR("Error scheduling IBs (%d)\n", r);
+			dev_err(adev->dev,
+				"Error scheduling IBs (%d) in ring(%s)", r,
+				ring->name);
 	}
 
 	job->job_run_counter++;
@@ -319,8 +438,24 @@ static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job)
 	return fence;
 }
 
-#define to_drm_sched_job(sched_job)		\
-		container_of((sched_job), struct drm_sched_job, queue_node)
+/*
+ * This is a duplicate function from DRM scheduler sched_internal.h.
+ * Plan is to remove it when amdgpu_job_stop_all_jobs_on_sched is removed, due
+ * latter being incorrect and racy.
+ *
+ * See https://lore.kernel.org/amd-gfx/44edde63-7181-44fb-a4f7-94e50514f539@amd.com/
+ */
+static struct drm_sched_job *
+drm_sched_entity_queue_pop(struct drm_sched_entity *entity)
+{
+	struct spsc_node *node;
+
+	node = spsc_queue_pop(&entity->job_queue);
+	if (!node)
+		return NULL;
+
+	return container_of(node, struct drm_sched_job, queue_node);
+}
 
 void amdgpu_job_stop_all_jobs_on_sched(struct drm_gpu_scheduler *sched)
 {
@@ -333,7 +468,7 @@ void amdgpu_job_stop_all_jobs_on_sched(struct drm_gpu_scheduler *sched)
 		struct drm_sched_rq *rq = sched->sched_rq[i];
 		spin_lock(&rq->lock);
 		list_for_each_entry(s_entity, &rq->entities, list) {
-			while ((s_job = to_drm_sched_job(spsc_queue_pop(&s_entity->job_queue)))) {
+			while ((s_job = drm_sched_entity_queue_pop(s_entity))) {
 				struct drm_sched_fence *s_fence = s_job->s_fence;
 
 				dma_fence_signal(&s_fence->scheduled);