Skip to contentSkip to conversation

drm/amdgpu/gfx8: fix driver reload with KIQ

PATCH In review Alex Deucher 3 participants 4 messages Jul 4, 2017 1 message not in the archive

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

PATCH 1/2 drm/amdgpu/gfx8: fix driver reload with KIQ

Drop the deactivation in KIQ init and drop the KCQ disabling via KIQ. We disable the MEC shortly after anyway, so there is no need to wait for all of this. Doing so seems to leave the MEC in a bad way.

---
 drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 69 +----------------------------------
 1 file changed, 1 insertion(+), 68 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +1 -68
index 1ca642b..4bb12ee 100644
@@ -4636,56 +4636,6 @@ static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev)
 	return r;
 }
 
-static int gfx_v8_0_kiq_kcq_disable(struct amdgpu_device *adev)
-{
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
-	uint32_t scratch, tmp = 0;
-	int r, i;
-
-	r = amdgpu_gfx_scratch_get(adev, &scratch);
-	if (r) {
-		DRM_ERROR("Failed to get scratch reg (%d).\n", r);
-		return r;
-	}
-	WREG32(scratch, 0xCAFEDEAD);
-
-	r = amdgpu_ring_alloc(kiq_ring, 6 + 3);
-	if (r) {
-		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
-		amdgpu_gfx_scratch_free(adev, scratch);
-		return r;
-	}
-	/* unmap queues */
-	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4));
-	amdgpu_ring_write(kiq_ring,
-			  PACKET3_UNMAP_QUEUES_ACTION(1)| /* RESET_QUEUES */
-			  PACKET3_UNMAP_QUEUES_QUEUE_SEL(2)); /* select all queues */
-	amdgpu_ring_write(kiq_ring, 0);
-	amdgpu_ring_write(kiq_ring, 0);
-	amdgpu_ring_write(kiq_ring, 0);
-	amdgpu_ring_write(kiq_ring, 0);
-	/* write to scratch for completion */
-	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
-	amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
-	amdgpu_ring_write(kiq_ring, 0xDEADBEEF);
-	amdgpu_ring_commit(kiq_ring);
-
-	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(scratch);
-		if (tmp == 0xDEADBEEF)
-			break;
-		DRM_UDELAY(1);
-	}
-	if (i >= adev->usec_timeout) {
-		DRM_ERROR("KCQ disabled failed (scratch(0x%04X)=0x%08X)\n",
-			  scratch, tmp);
-		r = -EINVAL;
-	}
-	amdgpu_gfx_scratch_free(adev, scratch);
-
-	return r;
-}
-
 static int gfx_v8_0_deactivate_hqd(struct amdgpu_device *adev, u32 req)
 {
 	int i, r = 0;
@@ -4891,7 +4841,6 @@ int gfx_v8_0_mqd_commit(struct amdgpu_device *adev,
 
 static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
 {
-	int r = 0;
 	struct amdgpu_device *adev = ring->adev;
 	struct vi_mqd *mqd = ring->mqd_ptr;
 	int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS;
@@ -4908,11 +4857,6 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
 		amdgpu_ring_clear_ring(ring);
 		mutex_lock(&adev->srbm_mutex);
 		vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
-		r = gfx_v8_0_deactivate_hqd(adev, 1);
-		if (r) {
-			dev_err(adev->dev, "failed to deactivate ring %s\n", ring->name);
-			goto out_unlock;
-		}
 		gfx_v8_0_mqd_commit(adev, mqd);
 		vi_srbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
@@ -4923,11 +4867,6 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
 		mutex_lock(&adev->srbm_mutex);
 		vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
 		gfx_v8_0_mqd_init(ring);
-		r = gfx_v8_0_deactivate_hqd(adev, 1);
-		if (r) {
-			dev_err(adev->dev, "failed to deactivate ring %s\n", ring->name);
-			goto out_unlock;
-		}
 		gfx_v8_0_mqd_commit(adev, mqd);
 		vi_srbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
@@ -4936,12 +4875,7 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
 			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation));
 	}
 
-	return r;
-
-out_unlock:
-	vi_srbm_select(adev, 0, 0, 0, 0);
-	mutex_unlock(&adev->srbm_mutex);
-	return r;
+	return 0;
 }
 
 static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring)
@@ -5145,7 +5079,6 @@ static int gfx_v8_0_hw_fini(void *handle)
 		pr_debug("For SRIOV client, shouldn't do anything.\n");
 		return 0;
 	}
-	gfx_v8_0_kiq_kcq_disable(adev);
 	gfx_v8_0_cp_enable(adev, false);
 	gfx_v8_0_rlc_stop(adev);
 
Signature · 6 lines
2.5.5

_______________________________________________
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): 1499148001-11497-1-git-send-email-alexander.deucher-5C7GfCeVMHo@public.gmane.org

#2 lore raw

PATCH 2/2 drm/amdgpu/gfx9: fix driver reload with KIQ

Drop the KCQ disabling via KIQ. We disable the MEC shortly after anyway, so there is no need to wait for all of this. Doing so seems to leave the MEC in a bad way.

---
 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 51 -----------------------------------
 1 file changed, 51 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +0 -51
index 2034592..3a0b69b 100644
@@ -2502,56 +2502,6 @@ static int gfx_v9_0_kiq_kcq_enable(struct amdgpu_device *adev)
 	return r;
 }
 
-static int gfx_v9_0_kiq_kcq_disable(struct amdgpu_device *adev)
-{
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
-	uint32_t scratch, tmp = 0;
-	int r, i;
-
-	r = amdgpu_gfx_scratch_get(adev, &scratch);
-	if (r) {
-		DRM_ERROR("Failed to get scratch reg (%d).\n", r);
-		return r;
-	}
-	WREG32(scratch, 0xCAFEDEAD);
-
-	r = amdgpu_ring_alloc(kiq_ring, 6 + 3);
-	if (r) {
-		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
-		amdgpu_gfx_scratch_free(adev, scratch);
-		return r;
-	}
-	/* unmap queues */
-	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4));
-	amdgpu_ring_write(kiq_ring,
-			  PACKET3_UNMAP_QUEUES_ACTION(1)| /* RESET_QUEUES */
-			  PACKET3_UNMAP_QUEUES_QUEUE_SEL(2)); /* select all queues */
-	amdgpu_ring_write(kiq_ring, 0);
-	amdgpu_ring_write(kiq_ring, 0);
-	amdgpu_ring_write(kiq_ring, 0);
-	amdgpu_ring_write(kiq_ring, 0);
-	/* write to scratch for completion */
-	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
-	amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
-	amdgpu_ring_write(kiq_ring, 0xDEADBEEF);
-	amdgpu_ring_commit(kiq_ring);
-
-	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(scratch);
-		if (tmp == 0xDEADBEEF)
-			break;
-		DRM_UDELAY(1);
-	}
-	if (i >= adev->usec_timeout) {
-		DRM_ERROR("KCQ disable failed (scratch(0x%04X)=0x%08X)\n",
-			  scratch, tmp);
-		r = -EINVAL;
-	}
-	amdgpu_gfx_scratch_free(adev, scratch);
-
-	return r;
-}
-
 static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
@@ -2996,7 +2946,6 @@ static int gfx_v9_0_hw_fini(void *handle)
 		pr_debug("For SRIOV client, shouldn't do anything.\n");
 		return 0;
 	}
-	gfx_v9_0_kiq_kcq_disable(adev);
 	gfx_v9_0_cp_enable(adev, false);
 	gfx_v9_0_rlc_stop(adev);
 
Signature · 6 lines
2.5.5

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Tested-by
#3 lore raw
In reply to a message not in the archive
Quoted 6 lines · Alex Deucher

Drop the deactivation in KIQ init and drop the KCQ disabling via KIQ. We disable the MEC shortly after anyway, so there is no need to wait for all of this. Doing so seems to leave the MEC in a bad way.

Signature · 6 lines
Earthling Michel Dänzer               |               http://www.amd.com
Libre software enthusiast             |             Mesa and X developer
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#4 lore raw
In reply to a message not in the archive

Am 04.07.2017 um 08:00 schrieb Alex Deucher:

Quoted 6 lines

Drop the deactivation in KIQ init and drop the KCQ disabling via KIQ. We disable the MEC shortly after anyway, so there is no need to wait for all of this. Doing so seems to leave the MEC in a bad way.

Quoted 112 lines
---
  drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 69 +----------------------------------
  1 file changed, 1 insertion(+), 68 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +1 -67
index 1ca642b..4bb12ee 100644
@@ -4636,56 +4636,6 @@ static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev)
  	return r;
  }
 
-static int gfx_v8_0_kiq_kcq_disable(struct amdgpu_device *adev)
-{
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
-	uint32_t scratch, tmp = 0;
-	int r, i;
-
-	r = amdgpu_gfx_scratch_get(adev, &scratch);
-	if (r) {
-		DRM_ERROR("Failed to get scratch reg (%d).\n", r);
-		return r;
-	}
-	WREG32(scratch, 0xCAFEDEAD);
-
-	r = amdgpu_ring_alloc(kiq_ring, 6 + 3);
-	if (r) {
-		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
-		amdgpu_gfx_scratch_free(adev, scratch);
-		return r;
-	}
-	/* unmap queues */
-	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4));
-	amdgpu_ring_write(kiq_ring,
-			  PACKET3_UNMAP_QUEUES_ACTION(1)| /* RESET_QUEUES */
-			  PACKET3_UNMAP_QUEUES_QUEUE_SEL(2)); /* select all queues */
-	amdgpu_ring_write(kiq_ring, 0);
-	amdgpu_ring_write(kiq_ring, 0);
-	amdgpu_ring_write(kiq_ring, 0);
-	amdgpu_ring_write(kiq_ring, 0);
-	/* write to scratch for completion */
-	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
-	amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
-	amdgpu_ring_write(kiq_ring, 0xDEADBEEF);
-	amdgpu_ring_commit(kiq_ring);
-
-	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(scratch);
-		if (tmp == 0xDEADBEEF)
-			break;
-		DRM_UDELAY(1);
-	}
-	if (i >= adev->usec_timeout) {
-		DRM_ERROR("KCQ disabled failed (scratch(0x%04X)=0x%08X)\n",
-			  scratch, tmp);
-		r = -EINVAL;
-	}
-	amdgpu_gfx_scratch_free(adev, scratch);
-
-	return r;
-}
-
  static int gfx_v8_0_deactivate_hqd(struct amdgpu_device *adev, u32 req)
  {
  	int i, r = 0;
@@ -4891,7 +4841,6 @@ int gfx_v8_0_mqd_commit(struct amdgpu_device *adev,
 
  static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
  {
-	int r = 0;
  	struct amdgpu_device *adev = ring->adev;
  	struct vi_mqd *mqd = ring->mqd_ptr;
  	int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS;
@@ -4908,11 +4857,6 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
  		amdgpu_ring_clear_ring(ring);
  		mutex_lock(&adev->srbm_mutex);
  		vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
-		r = gfx_v8_0_deactivate_hqd(adev, 1);
-		if (r) {
-			dev_err(adev->dev, "failed to deactivate ring %s\n", ring->name);
-			goto out_unlock;
-		}
  		gfx_v8_0_mqd_commit(adev, mqd);
  		vi_srbm_select(adev, 0, 0, 0, 0);
  		mutex_unlock(&adev->srbm_mutex);
@@ -4923,11 +4867,6 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
  		mutex_lock(&adev->srbm_mutex);
  		vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
  		gfx_v8_0_mqd_init(ring);
-		r = gfx_v8_0_deactivate_hqd(adev, 1);
-		if (r) {
-			dev_err(adev->dev, "failed to deactivate ring %s\n", ring->name);
-			goto out_unlock;
-		}
  		gfx_v8_0_mqd_commit(adev, mqd);
  		vi_srbm_select(adev, 0, 0, 0, 0);
  		mutex_unlock(&adev->srbm_mutex);
@@ -4936,12 +4875,7 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
  			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation));
  	}
 
-	return r;
-
-out_unlock:
-	vi_srbm_select(adev, 0, 0, 0, 0);
-	mutex_unlock(&adev->srbm_mutex);
-	return r;
+	return 0;
  }
 
  static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring)

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c static int gfx_v8_0_hw_fini(void *handle) · line 5084

Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +0 -1
@@ -5145,7 +5079,6 @@ static int gfx_v8_0_hw_fini(void *handle)
  		pr_debug("For SRIOV client, shouldn't do anything.\n");
  		return 0;
  	}
-	gfx_v8_0_kiq_kcq_disable(adev);
  	gfx_v8_0_cp_enable(adev, false);
  	gfx_v8_0_rlc_stop(adev);
 

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