summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd/kfd_svm.c')
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_svm.c119
1 files changed, 55 insertions, 64 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
index bd3e20d981e0..a0f22ea6d15a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
@@ -563,7 +563,8 @@ svm_range_vram_node_new(struct kfd_node *node, struct svm_range *prange,
int r;
p = container_of(prange->svms, struct kfd_process, svms);
- pr_debug("pasid: %x svms 0x%p [0x%lx 0x%lx]\n", p->pasid, prange->svms,
+ pr_debug("process pid: %d svms 0x%p [0x%lx 0x%lx]\n",
+ p->lead_thread->pid, prange->svms,
prange->start, prange->last);
if (svm_range_validate_svm_bo(node, prange))
@@ -1170,13 +1171,12 @@ svm_range_split_head(struct svm_range *prange, uint64_t new_start,
}
static void
-svm_range_add_child(struct svm_range *prange, struct mm_struct *mm,
- struct svm_range *pchild, enum svm_work_list_ops op)
+svm_range_add_child(struct svm_range *prange, struct svm_range *pchild, enum svm_work_list_ops op)
{
pr_debug("add child 0x%p [0x%lx 0x%lx] to prange 0x%p child list %d\n",
pchild, pchild->start, pchild->last, prange, op);
- pchild->work_item.mm = mm;
+ pchild->work_item.mm = NULL;
pchild->work_item.op = op;
list_add_tail(&pchild->child_list, &prange->child_list);
}
@@ -1244,8 +1244,7 @@ svm_range_get_pte_flags(struct kfd_node *node,
case IP_VERSION(9, 4, 4):
case IP_VERSION(9, 5, 0):
if (ext_coherent)
- mtype_local = (gc_ip_version < IP_VERSION(9, 5, 0) && !node->adev->rev_id) ?
- AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_CC;
+ mtype_local = AMDGPU_VM_MTYPE_CC;
else
mtype_local = amdgpu_mtype_local == 1 ? AMDGPU_VM_MTYPE_NC :
amdgpu_mtype_local == 2 ? AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW;
@@ -1278,7 +1277,7 @@ svm_range_get_pte_flags(struct kfd_node *node,
mapping_flags |= ext_coherent ? AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
/* system memory accessed by the dGPU */
} else {
- if (gc_ip_version < IP_VERSION(9, 5, 0))
+ if (gc_ip_version < IP_VERSION(9, 5, 0) || ext_coherent)
mapping_flags |= AMDGPU_VM_MTYPE_UC;
else
mapping_flags |= AMDGPU_VM_MTYPE_NC;
@@ -1286,13 +1285,7 @@ svm_range_get_pte_flags(struct kfd_node *node,
break;
case IP_VERSION(12, 0, 0):
case IP_VERSION(12, 0, 1):
- if (domain == SVM_RANGE_VRAM_DOMAIN) {
- if (bo_node != node)
- mapping_flags |= AMDGPU_VM_MTYPE_NC;
- } else {
- mapping_flags |= coherent ?
- AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
- }
+ mapping_flags |= AMDGPU_VM_MTYPE_NC;
break;
default:
mapping_flags |= coherent ?
@@ -2400,15 +2393,17 @@ svm_range_add_list_work(struct svm_range_list *svms, struct svm_range *prange,
prange->work_item.op != SVM_OP_UNMAP_RANGE)
prange->work_item.op = op;
} else {
- prange->work_item.op = op;
-
- /* Pairs with mmput in deferred_list_work */
- mmget(mm);
- prange->work_item.mm = mm;
- list_add_tail(&prange->deferred_list,
- &prange->svms->deferred_range_list);
- pr_debug("add prange 0x%p [0x%lx 0x%lx] to work list op %d\n",
- prange, prange->start, prange->last, op);
+ /* Pairs with mmput in deferred_list_work.
+ * If process is exiting and mm is gone, don't update mmu notifier.
+ */
+ if (mmget_not_zero(mm)) {
+ prange->work_item.mm = mm;
+ prange->work_item.op = op;
+ list_add_tail(&prange->deferred_list,
+ &prange->svms->deferred_range_list);
+ pr_debug("add prange 0x%p [0x%lx 0x%lx] to work list op %d\n",
+ prange, prange->start, prange->last, op);
+ }
}
spin_unlock(&svms->deferred_list_lock);
}
@@ -2422,8 +2417,7 @@ void schedule_deferred_list_work(struct svm_range_list *svms)
}
static void
-svm_range_unmap_split(struct mm_struct *mm, struct svm_range *parent,
- struct svm_range *prange, unsigned long start,
+svm_range_unmap_split(struct svm_range *parent, struct svm_range *prange, unsigned long start,
unsigned long last)
{
struct svm_range *head;
@@ -2444,12 +2438,12 @@ svm_range_unmap_split(struct mm_struct *mm, struct svm_range *parent,
svm_range_split(tail, last + 1, tail->last, &head);
if (head != prange && tail != prange) {
- svm_range_add_child(parent, mm, head, SVM_OP_UNMAP_RANGE);
- svm_range_add_child(parent, mm, tail, SVM_OP_ADD_RANGE);
+ svm_range_add_child(parent, head, SVM_OP_UNMAP_RANGE);
+ svm_range_add_child(parent, tail, SVM_OP_ADD_RANGE);
} else if (tail != prange) {
- svm_range_add_child(parent, mm, tail, SVM_OP_UNMAP_RANGE);
+ svm_range_add_child(parent, tail, SVM_OP_UNMAP_RANGE);
} else if (head != prange) {
- svm_range_add_child(parent, mm, head, SVM_OP_UNMAP_RANGE);
+ svm_range_add_child(parent, head, SVM_OP_UNMAP_RANGE);
} else if (parent != prange) {
prange->work_item.op = SVM_OP_UNMAP_RANGE;
}
@@ -2526,14 +2520,14 @@ svm_range_unmap_from_cpu(struct mm_struct *mm, struct svm_range *prange,
l = min(last, pchild->last);
if (l >= s)
svm_range_unmap_from_gpus(pchild, s, l, trigger);
- svm_range_unmap_split(mm, prange, pchild, start, last);
+ svm_range_unmap_split(prange, pchild, start, last);
mutex_unlock(&pchild->lock);
}
s = max(start, prange->start);
l = min(last, prange->last);
if (l >= s)
svm_range_unmap_from_gpus(prange, s, l, trigger);
- svm_range_unmap_split(mm, prange, prange, start, last);
+ svm_range_unmap_split(prange, prange, start, last);
if (unmap_parent)
svm_range_add_list_work(svms, prange, mm, SVM_OP_UNMAP_RANGE);
@@ -2576,8 +2570,6 @@ svm_range_cpu_invalidate_pagetables(struct mmu_interval_notifier *mni,
if (range->event == MMU_NOTIFY_RELEASE)
return true;
- if (!mmget_not_zero(mni->mm))
- return true;
start = mni->interval_tree.start;
last = mni->interval_tree.last;
@@ -2604,7 +2596,6 @@ svm_range_cpu_invalidate_pagetables(struct mmu_interval_notifier *mni,
}
svm_range_unlock(prange);
- mmput(mni->mm);
return true;
}
@@ -2691,7 +2682,7 @@ svm_range_best_restore_location(struct svm_range *prange,
return -1;
}
- if (node->adev->flags & AMD_IS_APU)
+ if (node->adev->apu_prefer_gtt)
return 0;
if (prange->preferred_loc == gpuid ||
@@ -2979,7 +2970,7 @@ svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
return -EFAULT;
}
- p = kfd_lookup_process_by_pasid(pasid);
+ p = kfd_lookup_process_by_pasid(pasid, NULL);
if (!p) {
pr_debug("kfd process not founded pasid 0x%x\n", pasid);
return 0;
@@ -3008,19 +2999,6 @@ svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
goto out;
}
- /* check if this page fault time stamp is before svms->checkpoint_ts */
- if (svms->checkpoint_ts[gpuidx] != 0) {
- if (amdgpu_ih_ts_after(ts, svms->checkpoint_ts[gpuidx])) {
- pr_debug("draining retry fault, drop fault 0x%llx\n", addr);
- r = 0;
- goto out;
- } else
- /* ts is after svms->checkpoint_ts now, reset svms->checkpoint_ts
- * to zero to avoid following ts wrap around give wrong comparing
- */
- svms->checkpoint_ts[gpuidx] = 0;
- }
-
if (!p->xnack_enabled) {
pr_debug("XNACK not enabled for pasid 0x%x\n", pasid);
r = -EFAULT;
@@ -3040,6 +3018,21 @@ svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
mmap_read_lock(mm);
retry_write_locked:
mutex_lock(&svms->lock);
+
+ /* check if this page fault time stamp is before svms->checkpoint_ts */
+ if (svms->checkpoint_ts[gpuidx] != 0) {
+ if (amdgpu_ih_ts_after_or_equal(ts, svms->checkpoint_ts[gpuidx])) {
+ pr_debug("draining retry fault, drop fault 0x%llx\n", addr);
+ r = -EAGAIN;
+ goto out_unlock_svms;
+ } else {
+ /* ts is after svms->checkpoint_ts now, reset svms->checkpoint_ts
+ * to zero to avoid following ts wrap around give wrong comparing
+ */
+ svms->checkpoint_ts[gpuidx] = 0;
+ }
+ }
+
prange = svm_range_from_addr(svms, addr, NULL);
if (!prange) {
pr_debug("failed to find prange svms 0x%p address [0x%llx]\n",
@@ -3165,7 +3158,8 @@ out_unlock_svms:
mutex_unlock(&svms->lock);
mmap_read_unlock(mm);
- svm_range_count_fault(node, p, gpuidx);
+ if (r != -EAGAIN)
+ svm_range_count_fault(node, p, gpuidx);
mmput(mm);
out:
@@ -3242,7 +3236,8 @@ void svm_range_list_fini(struct kfd_process *p)
struct svm_range *prange;
struct svm_range *next;
- pr_debug("pasid 0x%x svms 0x%p\n", p->pasid, &p->svms);
+ pr_debug("process pid %d svms 0x%p\n", p->lead_thread->pid,
+ &p->svms);
cancel_delayed_work_sync(&p->svms.restore_work);
@@ -3265,7 +3260,8 @@ void svm_range_list_fini(struct kfd_process *p)
mutex_destroy(&p->svms.lock);
- pr_debug("pasid 0x%x svms 0x%p done\n", p->pasid, &p->svms);
+ pr_debug("process pid %d svms 0x%p done\n",
+ p->lead_thread->pid, &p->svms);
}
int svm_range_list_init(struct kfd_process *p)
@@ -3438,7 +3434,7 @@ svm_range_best_prefetch_location(struct svm_range *prange)
goto out;
}
- if (bo_node->adev->flags & AMD_IS_APU) {
+ if (bo_node->adev->apu_prefer_gtt) {
best_loc = 0;
goto out;
}
@@ -3628,8 +3624,8 @@ svm_range_set_attr(struct kfd_process *p, struct mm_struct *mm,
bool flush_tlb;
int r, ret = 0;
- pr_debug("pasid 0x%x svms 0x%p [0x%llx 0x%llx] pages 0x%llx\n",
- p->pasid, &p->svms, start, start + size - 1, size);
+ pr_debug("process pid %d svms 0x%p [0x%llx 0x%llx] pages 0x%llx\n",
+ p->lead_thread->pid, &p->svms, start, start + size - 1, size);
r = svm_range_check_attr(p, nattr, attrs);
if (r)
@@ -3737,8 +3733,8 @@ out_unlock_range:
out:
mutex_unlock(&process_info->lock);
- pr_debug("pasid 0x%x svms 0x%p [0x%llx 0x%llx] done, r=%d\n", p->pasid,
- &p->svms, start, start + size - 1, r);
+ pr_debug("process pid %d svms 0x%p [0x%llx 0x%llx] done, r=%d\n",
+ p->lead_thread->pid, &p->svms, start, start + size - 1, r);
return ret ? ret : r;
}
@@ -4076,8 +4072,8 @@ exit:
return ret;
}
-int svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges,
- uint64_t *svm_priv_data_size)
+void svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges,
+ uint64_t *svm_priv_data_size)
{
uint64_t total_size, accessibility_size, common_attr_size;
int nattr_common = 4, nattr_accessibility = 1;
@@ -4089,8 +4085,6 @@ int svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges,
*svm_priv_data_size = 0;
svms = &p->svms;
- if (!svms)
- return -EINVAL;
mutex_lock(&svms->lock);
list_for_each_entry(prange, &svms->list, list) {
@@ -4132,7 +4126,6 @@ int svm_range_get_info(struct kfd_process *p, uint32_t *num_svm_ranges,
pr_debug("num_svm_ranges %u total_priv_size %llu\n", *num_svm_ranges,
*svm_priv_data_size);
- return 0;
}
int kfd_criu_checkpoint_svm(struct kfd_process *p,
@@ -4149,8 +4142,6 @@ int kfd_criu_checkpoint_svm(struct kfd_process *p,
struct mm_struct *mm;
svms = &p->svms;
- if (!svms)
- return -EINVAL;
mm = get_task_mm(p->lead_thread);
if (!mm) {
generated by cgit v1.2.3 (git 2.39.0) at 2025-08-28 23:25:41 +0300