github-mirror/linux
1
0
Fork 0
mirror of https://github.com/torvalds/linux.git synced 2026-10-12 05:17:15 +02:00
Code Issues Actions Packages Projects Releases Wiki Activity Graph

drm/amdgpu: Remove faulty queue before resume

Browse Source 
When driver already knows a bad queue but MES suspend_all is successful
and MES hung queue detection doesn't detect it, remove this queue refore
resume_all.

Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This commit is contained in:
Amber Lin 2026-05-29 15:36:52 -04:00 committed by Alex Deucher
parent 913c0d83be
commit f401a2633e
6 changed files with 23 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
View File

@ -2244,7 +2244,8 @@ int amdgpu_gfx_reset_mes_compute(struct amdgpu_device *adev,
struct amdgpu_ring *ring, struct amdgpu_ring *ring,
struct amdgpu_fence *guilty_fence, struct amdgpu_fence *guilty_fence,
struct amdgpu_usermode_queue *uq, struct amdgpu_usermode_queue *uq,
unsigned int *hung_queue_count) unsigned int *hung_queue_count,
void *faulty_queue_input)
{ {
struct amdgpu_mes_hung_queue_hqd_info *hqd_info = struct amdgpu_mes_hung_queue_hqd_info *hqd_info =
(struct amdgpu_mes_hung_queue_hqd_info *) (struct amdgpu_mes_hung_queue_hqd_info *)
@ -2252,6 +2253,7 @@ int amdgpu_gfx_reset_mes_compute(struct amdgpu_device *adev,
int i, r, pipe, queue, queue_type; int i, r, pipe, queue, queue_type;
unsigned int num_hung = 0; unsigned int num_hung = 0;
bool use_mmio = adev->gfx.mec.use_mmio_for_reset; bool use_mmio = adev->gfx.mec.use_mmio_for_reset;
struct mes_remove_queue_input *queue_input = (struct mes_remove_queue_input *)faulty_queue_input;
guard(mutex)(&adev->gfx.mec.reset_mutex); guard(mutex)(&adev->gfx.mec.reset_mutex);
/* stop the drm schedulers for all compute queues */ /* stop the drm schedulers for all compute queues */
@ -2306,6 +2308,20 @@ int amdgpu_gfx_reset_mes_compute(struct amdgpu_device *adev,
if (r) if (r)
goto out; goto out;
} }
/* MES doesn't detect any hung queue but we have a known bad queue
* and it is not KCQ
*/
if (!num_hung && queue_input && !ring) {
/* MES suspend_all is successful means this bad queue is
* preempted successfuly. Remove it before resume all so it
* doesn't get mapped back
*/
amdgpu_mes_lock(&adev->mes);
r = adev->mes.funcs->remove_hw_queue(&adev->mes, queue_input);
amdgpu_mes_unlock(&adev->mes);
}
out: out:
/* resume all will enable the non-hung queues */ /* resume all will enable the non-hung queues */
amdgpu_mes_resume(adev, 0); amdgpu_mes_resume(adev, 0);

3
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
View File

@ -651,7 +651,8 @@ int amdgpu_gfx_reset_mes_compute(struct amdgpu_device *adev,
struct amdgpu_ring *ring, struct amdgpu_ring *ring,
struct amdgpu_fence *guilty_fence, struct amdgpu_fence *guilty_fence,
struct amdgpu_usermode_queue *uq, struct amdgpu_usermode_queue *uq,
unsigned int *hung_queue_count); unsigned int *hung_queue_count,
void *faulty_queue_input);
void amdgpu_gfx_ras_error_func(struct amdgpu_device *adev, void amdgpu_gfx_ras_error_func(struct amdgpu_device *adev,
void *ras_error_status, void *ras_error_status,
void (*func)(struct amdgpu_device *adev, void *ras_error_status, void (*func)(struct amdgpu_device *adev, void *ras_error_status,

2
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
View File

@ -141,7 +141,7 @@ static void amdgpu_userq_hang_detect_work(struct work_struct *work)
int r; int r;
if (queue->queue_type == AMDGPU_HW_IP_COMPUTE) if (queue->queue_type == AMDGPU_HW_IP_COMPUTE)
r = amdgpu_gfx_reset_mes_compute(adev, NULL, NULL, NULL, NULL); r = amdgpu_gfx_reset_mes_compute(adev, NULL, NULL, NULL, NULL, NULL);
else else
r = userq_funcs->reset(queue); r = userq_funcs->reset(queue);
if (r) if (r)

2
drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
View File

@ -6835,7 +6835,7 @@ static int gfx_v11_0_reset_kcq(struct amdgpu_ring *ring,
{ {
struct amdgpu_device *adev = ring->adev; struct amdgpu_device *adev = ring->adev;
return amdgpu_gfx_reset_mes_compute(adev, ring, timedout_fence, NULL, NULL); return amdgpu_gfx_reset_mes_compute(adev, ring, timedout_fence, NULL, NULL, NULL);
} }
static void gfx_v11_ip_print(struct amdgpu_ip_block *ip_block, struct drm_printer *p) static void gfx_v11_ip_print(struct amdgpu_ip_block *ip_block, struct drm_printer *p)

2
drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c
View File

@ -5261,7 +5261,7 @@ static int gfx_v12_0_reset_kcq(struct amdgpu_ring *ring,
{ {
struct amdgpu_device *adev = ring->adev; struct amdgpu_device *adev = ring->adev;
return amdgpu_gfx_reset_mes_compute(adev, ring, timedout_fence, NULL, NULL); return amdgpu_gfx_reset_mes_compute(adev, ring, timedout_fence, NULL, NULL, NULL);
} }
static void gfx_v12_0_ring_begin_use(struct amdgpu_ring *ring) static void gfx_v12_0_ring_begin_use(struct amdgpu_ring *ring)

2
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
View File

@ -464,7 +464,7 @@ static int reset_queues_mes(struct device_queue_manager *dqm)
goto fail; goto fail;
} }
r = amdgpu_gfx_reset_mes_compute(adev, NULL, NULL, NULL, &num_hung); r = amdgpu_gfx_reset_mes_compute(adev, NULL, NULL, NULL, &num_hung, NULL);
if (r) if (r)
goto fail; goto fail;