Skip to contentSkip to conversation

drm/amdgpu: recover vram bo from shadow after gpu reset V2

PATCH In review Chunming Zhou 2 participants 3 messages Aug 18, 2016 – Aug 19, 2016 2 messages not in the archive

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 7/7 drm/amdgpu: recover vram bo from shadow after gpu reset V2

V2:
1. don't directly submit to many jobs at the same time.
2. delete unrelated printk.

Change-Id I963598ba6eb44bc8620d70e026c0175d1a1de120
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 58 +++++++++++++++++++++++++++++-
 1 file changed, 57 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +57 -1
index c9b15c0..0bedb26 100644
@@ -2120,6 +2120,35 @@ bool amdgpu_need_backup(struct amdgpu_device *adev)
 	return amdgpu_lockup_timeout > 0 ? true : false;
 }
 
+static int amdgpu_recover_vram_from_shadow(struct amdgpu_device *adev,
+					   struct amdgpu_ring *ring,
+					   struct amdgpu_bo *bo,
+					   struct fence **fence)
+{
+	uint32_t domain;
+	int r;
+
+       if (!bo->shadow)
+               return 0;
+
+       r = amdgpu_bo_reserve(bo, false);
+       if (r)
+               return r;
+       domain = amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type);
+       /* if bo has been evicted, then no need to recover */
+       if (domain == AMDGPU_GEM_DOMAIN_VRAM) {
+               r = amdgpu_bo_restore_from_shadow(adev, ring, bo,
+						 NULL, fence, true);
+               if (r) {
+                       DRM_ERROR("recover page table failed!\n");
+                       goto err;
+               }
+       }
+err:
+       amdgpu_bo_unreserve(bo);
+       return r;
+}
+
 /**
  * amdgpu_gpu_reset - reset the asic
  *
@@ -2202,13 +2231,40 @@ retry:
 		if (r) {
 			dev_err(adev->dev, "ib ring test failed (%d).\n", r);
 			r = amdgpu_suspend(adev);
+			need_full_reset = true;
 			goto retry;
 		}
-
+		/**
+		 * recovery vm page tables, since we cannot depend on VRAM is
+		 * consistent after gpu full reset.
+		 */
+		if (need_full_reset && amdgpu_need_backup(adev)) {
+			struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
+			struct amdgpu_bo *bo, *tmp;
+			struct fence *fence = NULL, *next = NULL;
+
+			DRM_INFO("recover vram bo from shadow\n");
+			mutex_lock(&adev->shadow_list_lock);
+			list_for_each_entry_safe(bo, tmp, &adev->shadow_list, shadow_list) {
+				amdgpu_recover_vram_from_shadow(adev, ring, bo, &next);
+				if (fence)
+					fence_wait(fence, false);
+				fence_put(fence);
+				fence = next;
+			}
+			mutex_unlock(&adev->shadow_list_lock);
+			if (fence) {
+				r = fence_wait(fence, false);
+				if (r)
+					WARN(r, "recovery from shadow isn't comleted\n");
+			}
+			fence_put(fence);
+		}
 		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 			struct amdgpu_ring *ring = adev->rings[i];
 			if (!ring)
 				continue;
+
 			amd_sched_job_recovery(&ring->sched);
 			kthread_unpark(ring->sched.thread);
 		}
Signature · 6 lines
1.9.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): 1471519071-30545-1-git-send-email-David1.Zhou-5C7GfCeVMHo@public.gmane.org

#2 lore raw

Am 18.08.2016 um 13:17 schrieb Chunming Zhou:

Quoted 63 lines

V2:
1. don't directly submit to many jobs at the same time.
2. delete unrelated printk.

Change-Id I963598ba6eb44bc8620d70e026c0175d1a1de120
---
  drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 58 +++++++++++++++++++++++++++++-
  1 file changed, 57 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +35 -1
index c9b15c0..0bedb26 100644
@@ -2120,6 +2120,35 @@ bool amdgpu_need_backup(struct amdgpu_device *adev)
  	return amdgpu_lockup_timeout > 0 ? true : false;
  }
 
+static int amdgpu_recover_vram_from_shadow(struct amdgpu_device *adev,
+					   struct amdgpu_ring *ring,
+					   struct amdgpu_bo *bo,
+					   struct fence **fence)
+{
+	uint32_t domain;
+	int r;
+
+       if (!bo->shadow)
+               return 0;
+
+       r = amdgpu_bo_reserve(bo, false);
+       if (r)
+               return r;
+       domain = amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type);
+       /* if bo has been evicted, then no need to recover */
+       if (domain == AMDGPU_GEM_DOMAIN_VRAM) {
+               r = amdgpu_bo_restore_from_shadow(adev, ring, bo,
+						 NULL, fence, true);
+               if (r) {
+                       DRM_ERROR("recover page table failed!\n");
+                       goto err;
+               }
+       }
+err:
+       amdgpu_bo_unreserve(bo);
+       return r;
+}
+
  /**
   * amdgpu_gpu_reset - reset the asic
   *
@@ -2202,13 +2231,40 @@ retry:
  		if (r) {
  			dev_err(adev->dev, "ib ring test failed (%d).\n", r);
  			r = amdgpu_suspend(adev);
+			need_full_reset = true;
  			goto retry;
  		}
-
+		/**
+		 * recovery vm page tables, since we cannot depend on VRAM is
+		 * consistent after gpu full reset.
+		 */
+		if (need_full_reset && amdgpu_need_backup(adev)) {

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c retry: · line 2251

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +10 -0
+			struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
+			struct amdgpu_bo *bo, *tmp;
+			struct fence *fence = NULL, *next = NULL;
+
+			DRM_INFO("recover vram bo from shadow\n");
+			mutex_lock(&adev->shadow_list_lock);
+			list_for_each_entry_safe(bo, tmp, &adev->shadow_list, shadow_list) {
+				amdgpu_recover_vram_from_shadow(adev, ring, bo, &next);
+				if (fence)
+					fence_wait(fence, false);

You should check the return code here as well and abort the loop if anything fails.

With that fixed the patch is Reviewed-by: Christian König <[email protected]> as well.

Finally done with that!

Cheers,
Christian.

Quoted 9 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +9 -0
+				fence_put(fence);
+				fence = next;
+			}
+			mutex_unlock(&adev->shadow_list_lock);
+			if (fence) {
+				r = fence_wait(fence, false);
+				if (r)
+					WARN(r, "recovery from shadow isn't comleted\n");
+			}

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c retry: · line 2270

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +3 -0
+			fence_put(fence);
+		}
  		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
  			struct amdgpu_ring *ring = adev->rings[i];
  			if (!ring)
  				continue;
+
  			amd_sched_job_recovery(&ring->sched);
  			kthread_unpark(ring->sched.thread);
  		}

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): 78fefefd-b088-3923-e8a1-5a02bcb2a406-ANTagKRnAhcb1SvskN2V4Q@public.gmane.org

author
#4 lore raw
Quoted 80 lines · Christian König

Am 18.08.2016 um 13:17 schrieb Chunming Zhou:

Quoted 66 lines

V2:
1. don't directly submit to many jobs at the same time.
2. delete unrelated printk.

Change-Id I963598ba6eb44bc8620d70e026c0175d1a1de120
---
  drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 58 
+++++++++++++++++++++++++++++-
  1 file changed, 57 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +35 -1
index c9b15c0..0bedb26 100644
@@ -2120,6 +2120,35 @@ bool amdgpu_need_backup(struct amdgpu_device 
 *adev)
      return amdgpu_lockup_timeout > 0 ? true : false;
  }
  +static int amdgpu_recover_vram_from_shadow(struct amdgpu_device 
 *adev,
+                       struct amdgpu_ring *ring,
+                       struct amdgpu_bo *bo,
+                       struct fence **fence)
+{
+    uint32_t domain;
+    int r;
+
+       if (!bo->shadow)
+               return 0;
+
+       r = amdgpu_bo_reserve(bo, false);
+       if (r)
+               return r;
+       domain = amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type);
+       /* if bo has been evicted, then no need to recover */
+       if (domain == AMDGPU_GEM_DOMAIN_VRAM) {
+               r = amdgpu_bo_restore_from_shadow(adev, ring, bo,
+                         NULL, fence, true);
+               if (r) {
+                       DRM_ERROR("recover page table failed!\n");
+                       goto err;
+               }
+       }
+err:
+       amdgpu_bo_unreserve(bo);
+       return r;
+}
+
  /**
   * amdgpu_gpu_reset - reset the asic
   *
@@ -2202,13 +2231,40 @@ retry:
          if (r) {
              dev_err(adev->dev, "ib ring test failed (%d).\n", r);
              r = amdgpu_suspend(adev);
+            need_full_reset = true;
              goto retry;
          }
-
+        /**
+         * recovery vm page tables, since we cannot depend on VRAM is
+         * consistent after gpu full reset.
+         */
+        if (need_full_reset && amdgpu_need_backup(adev)) {
+            struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c retry: · line 2252

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +9 -0
+            struct amdgpu_bo *bo, *tmp;
+            struct fence *fence = NULL, *next = NULL;
+
+            DRM_INFO("recover vram bo from shadow\n");
+            mutex_lock(&adev->shadow_list_lock);
+            list_for_each_entry_safe(bo, tmp, &adev->shadow_list, 
 shadow_list) {
+                amdgpu_recover_vram_from_shadow(adev, ring, bo, &next);
+                if (fence)
+                    fence_wait(fence, false);

You should check the return code here as well and abort the loop if anything fails.

Quoted 3 lines

With that fixed the patch is Reviewed-by: Christian König <[email protected]> as well.

Finally done with that!

Agree, Thanks very much for review. Next will be 'recover gart table':)

Cheers,
David Zhou

Quoted 25 lines

Cheers,
Christian.

Quoted 18 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +12 -0
+                fence_put(fence);
+                fence = next;
+            }
+            mutex_unlock(&adev->shadow_list_lock);
+            if (fence) {
+                r = fence_wait(fence, false);
+                if (r)
+                    WARN(r, "recovery from shadow isn't comleted\n");
+            }
+            fence_put(fence);
+        }
          for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
              struct amdgpu_ring *ring = adev->rings[i];
              if (!ring)
                  continue;
+
              amd_sched_job_recovery(&ring->sched);
              kthread_unpark(ring->sched.thread);
Quoted 1 line
          }
Quoted 4 lines

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx