Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
drm/amdkfd: Handle GPU reset and drain retry fault race
Only check and drain IH1 ring if CAM is not enabled.
If GPU is under reset, don't access IH to drain retry fault.
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
authored by
Philip Yang
and committed by
Alex Deucher
5b57c3c3
e1c94109
+6
-1
+6
-1
drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+6
-1
drivers/gpu/drm/amd/amdkfd/kfd_svm.c
···
33
33
#include "amdgpu_hmm.h"
34
34
#include "amdgpu.h"
35
35
#include "amdgpu_xgmi.h"
36
+
#include "amdgpu_reset.h"
36
37
#include "kfd_priv.h"
37
38
#include "kfd_svm.h"
38
39
#include "kfd_migrate.h"
···
2370
2369
2371
2370
pr_debug("drain retry fault gpu %d svms %p\n", i, svms);
2372
2371
2372
+
if (!down_read_trylock(&pdd->dev->adev->reset_domain->sem))
2373
+
continue;
2374
+
2373
2375
amdgpu_ih_wait_on_checkpoint_process_ts(pdd->dev->adev,
2374
2376
pdd->dev->adev->irq.retry_cam_enabled ?
2375
2377
&pdd->dev->adev->irq.ih :
···
2382
2378
amdgpu_ih_wait_on_checkpoint_process_ts(pdd->dev->adev,
2383
2379
&pdd->dev->adev->irq.ih_soft);
2384
2380
2381
+
up_read(&pdd->dev->adev->reset_domain->sem);
2385
2382
2386
2383
pr_debug("drain retry fault gpu %d svms 0x%p done\n", i, svms);
2387
2384
}
···
2566
2561
adev = pdd->dev->adev;
2567
2562
2568
2563
/* Check and drain ih1 ring if cam not available */
2569
-
if (adev->irq.ih1.ring_size) {
2564
+
if (!adev->irq.retry_cam_enabled && adev->irq.ih1.ring_size) {
2570
2565
ih = &adev->irq.ih1;
2571
2566
checkpoint_wptr = amdgpu_ih_get_wptr(adev, ih);
2572
2567
if (ih->rptr != checkpoint_wptr) {