Fixes: b3dc91f97317 ("drm/amdkfd: enable subsequent retry fault")
Reported-by: Sashiko <[email protected]>
Closes: https://lore.kernel.org/r/[email protected]
Signed-off-by: Aveline Noir <[email protected]>
---
No testing on AMD hardware. Any test will be appreciated.
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 2 +-
drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 7 ++++---
drivers/gpu/drm/amd/amdkfd/kfd_svm.h | 4 ++--
3 files changed, 7 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index c41e9c0916b7..86ef849097fa 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -3092,7 +3092,7 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev,
u32 pasid,
drm_exec_fini(&exec);
if (!svm_range_restore_pages(adev, pasid, vmid,
- node_id, addr >> PAGE_SHIFT, ts,
write_fault))
+ node_id, addr, ts, write_fault))
return true;
/* Re-acquire the VM lock, could be that the VM was freed in between. */
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
index 79455e23a030..c85d6bb73629 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
@@ -3130,8 +3130,9 @@ svm_fault_allowed(struct vm_area_struct *vma, bool
write_fault)
int
svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
uint32_t vmid, uint32_t node_id,
- uint64_t addr, uint64_t ts, bool write_fault)
+ uint64_t fault_addr, uint64_t ts, bool write_fault)
{
+ uint64_t addr = fault_addr >> PAGE_SHIFT;
unsigned long start, last, size;
struct mm_struct *mm = NULL;
struct svm_range_list *svms;
@@ -3249,7 +3250,7 @@ svm_range_restore_pages(struct amdgpu_device *adev,
unsigned int pasid,
mutex_lock(&prange->migrate_mutex);
if (svm_range_skip_recover(prange)) {
- amdgpu_gmc_filter_faults_remove(node->adev, addr, pasid);
+ amdgpu_gmc_filter_faults_remove(node->adev, fault_addr, pasid);
r = 0;
goto out_unlock_range;
}
@@ -3353,7 +3354,7 @@ svm_range_restore_pages(struct amdgpu_device *adev,
unsigned int pasid,
if (r == -EAGAIN) {
pr_debug("recover vm fault later\n");
- amdgpu_gmc_filter_faults_remove(node->adev, addr, pasid);
+ amdgpu_gmc_filter_faults_remove(node->adev, fault_addr, pasid);
r = 0;
}
return r;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h
b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h
index c7d7adae4476..38744e716f74 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h
@@ -178,7 +178,7 @@ int svm_range_vram_node_new(struct kfd_node *node, struct
svm_range *prange,
bool clear);
void svm_range_vram_node_free(struct svm_range *prange);
int svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
- uint32_t vmid, uint32_t node_id, uint64_t addr,
uint64_t ts,
+ uint32_t vmid, uint32_t node_id, uint64_t
fault_addr, uint64_t ts,
bool write_fault);
int svm_range_evict_svm_bo(struct amdgpu_bo *bo);
@@ -230,7 +230,7 @@ static inline void svm_range_list_fini(struct kfd_process *p)
static inline int svm_range_restore_pages(struct amdgpu_device *adev,
unsigned int pasid,
uint32_t client_id, uint32_t node_id,
- uint64_t addr, uint64_t ts, bool
write_fault)
+ uint64_t fault_addr, uint64_t ts,
bool write_fault)
{
return -EFAULT;
}
---
base-commit: 403b3d68346d7ba25f474f0686ad1d05feb65b7c
change-id: 20261108-kfd-fault-filter-address-20261108-4f096f2b2a1a
Best regards,
--
Aveline Noir <[email protected]>