aboutsummaryrefslogtreecommitdiffstatshomepage
path: root/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c')
-rw-r--r--drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c60
1 files changed, 35 insertions, 25 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index df58a6a1a67e..8975cf41a91a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -196,7 +196,7 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
return -EINVAL;
vram_size = KFD_XCP_MEMORY_SIZE(adev, xcp_id);
- if (adev->gmc.is_app_apu) {
+ if (adev->gmc.is_app_apu || adev->flags & AMD_IS_APU) {
system_mem_needed = size;
ttm_mem_needed = size;
}
@@ -220,7 +220,7 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
(kfd_mem_limit.ttm_mem_used + ttm_mem_needed >
kfd_mem_limit.max_ttm_mem_limit) ||
(adev && xcp_id >= 0 && adev->kfd.vram_used[xcp_id] + vram_needed >
- vram_size - reserved_for_pt)) {
+ vram_size - reserved_for_pt - atomic64_read(&adev->vram_pin_size))) {
ret = -ENOMEM;
goto release;
}
@@ -232,7 +232,8 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
"adev reference can't be null when vram is used");
if (adev && xcp_id >= 0) {
adev->kfd.vram_used[xcp_id] += vram_needed;
- adev->kfd.vram_used_aligned[xcp_id] += adev->gmc.is_app_apu ?
+ adev->kfd.vram_used_aligned[xcp_id] +=
+ (adev->gmc.is_app_apu || adev->flags & AMD_IS_APU) ?
vram_needed :
ALIGN(vram_needed, VRAM_AVAILABLITY_ALIGN);
}
@@ -260,7 +261,7 @@ void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
if (adev) {
adev->kfd.vram_used[xcp_id] -= size;
- if (adev->gmc.is_app_apu) {
+ if (adev->gmc.is_app_apu || adev->flags & AMD_IS_APU) {
adev->kfd.vram_used_aligned[xcp_id] -= size;
kfd_mem_limit.system_mem_used -= size;
kfd_mem_limit.ttm_mem_used -= size;
@@ -889,7 +890,7 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem,
* if peer device has large BAR. In contrast, access over xGMI is
* allowed for both small and large BAR configurations of peer device
*/
- if ((adev != bo_adev && !adev->gmc.is_app_apu) &&
+ if ((adev != bo_adev && !(adev->gmc.is_app_apu || adev->flags & AMD_IS_APU)) &&
((mem->domain == AMDGPU_GEM_DOMAIN_VRAM) ||
(mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL) ||
(mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP))) {
@@ -1188,7 +1189,8 @@ static int reserve_bo_and_cond_vms(struct kgd_mem *mem,
int ret;
ctx->sync = &mem->sync;
- drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0);
+ drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
+ DRM_EXEC_IGNORE_DUPLICATES, 0);
drm_exec_until_all_locked(&ctx->exec) {
ctx->n_vms = 0;
list_for_each_entry(entry, &mem->attachments, list) {
@@ -1656,7 +1658,7 @@ size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev,
- atomic64_read(&adev->vram_pin_size)
- reserved_for_pt;
- if (adev->gmc.is_app_apu) {
+ if (adev->gmc.is_app_apu || adev->flags & AMD_IS_APU) {
system_mem_available = no_system_mem_limit ?
kfd_mem_limit.max_system_mem_limit :
kfd_mem_limit.max_system_mem_limit -
@@ -1704,7 +1706,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
if (flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) {
domain = alloc_domain = AMDGPU_GEM_DOMAIN_VRAM;
- if (adev->gmc.is_app_apu) {
+ if (adev->gmc.is_app_apu || adev->flags & AMD_IS_APU) {
domain = AMDGPU_GEM_DOMAIN_GTT;
alloc_domain = AMDGPU_GEM_DOMAIN_GTT;
alloc_flags = 0;
@@ -1854,6 +1856,7 @@ err_node_allow:
err_bo_create:
amdgpu_amdkfd_unreserve_mem_limit(adev, aligned_size, flags, xcp_id);
err_reserve_limit:
+ amdgpu_sync_free(&(*mem)->sync);
mutex_destroy(&(*mem)->lock);
if (gobj)
drm_gem_object_put(gobj);
@@ -1950,7 +1953,7 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
if (size) {
if (!is_imported &&
(mem->bo->preferred_domains == AMDGPU_GEM_DOMAIN_VRAM ||
- (adev->gmc.is_app_apu &&
+ ((adev->gmc.is_app_apu || adev->flags & AMD_IS_APU) &&
mem->bo->preferred_domains == AMDGPU_GEM_DOMAIN_GTT)))
*size = bo_size;
else
@@ -2372,8 +2375,9 @@ static int import_obj_create(struct amdgpu_device *adev,
(*mem)->dmabuf = dma_buf;
(*mem)->bo = bo;
(*mem)->va = va;
- (*mem)->domain = (bo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM) && !adev->gmc.is_app_apu ?
- AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT;
+ (*mem)->domain = (bo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM) &&
+ !(adev->gmc.is_app_apu || adev->flags & AMD_IS_APU) ?
+ AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT;
(*mem)->mapped_to_gpu_memory = 0;
(*mem)->process_info = avm->process_info;
@@ -2900,13 +2904,12 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence __rcu *
amdgpu_sync_create(&sync_obj);
- /* Validate BOs and map them to GPUVM (update VM page tables). */
+ /* Validate BOs managed by KFD */
list_for_each_entry(mem, &process_info->kfd_bo_list,
validate_list) {
struct amdgpu_bo *bo = mem->bo;
uint32_t domain = mem->domain;
- struct kfd_mem_attachment *attachment;
struct dma_resv_iter cursor;
struct dma_fence *fence;
@@ -2931,6 +2934,25 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence __rcu *
goto validate_map_fail;
}
}
+ }
+
+ if (failed_size)
+ pr_debug("0x%lx/0x%lx in system\n", failed_size, total_size);
+
+ /* Validate PDs, PTs and evicted DMABuf imports last. Otherwise BO
+ * validations above would invalidate DMABuf imports again.
+ */
+ ret = process_validate_vms(process_info, &exec.ticket);
+ if (ret) {
+ pr_debug("Validating VMs failed, ret: %d\n", ret);
+ goto validate_map_fail;
+ }
+
+ /* Update mappings managed by KFD. */
+ list_for_each_entry(mem, &process_info->kfd_bo_list,
+ validate_list) {
+ struct kfd_mem_attachment *attachment;
+
list_for_each_entry(attachment, &mem->attachments, list) {
if (!attachment->is_mapped)
continue;
@@ -2947,18 +2969,6 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence __rcu *
}
}
- if (failed_size)
- pr_debug("0x%lx/0x%lx in system\n", failed_size, total_size);
-
- /* Validate PDs, PTs and evicted DMABuf imports last. Otherwise BO
- * validations above would invalidate DMABuf imports again.
- */
- ret = process_validate_vms(process_info, &exec.ticket);
- if (ret) {
- pr_debug("Validating VMs failed, ret: %d\n", ret);
- goto validate_map_fail;
- }
-
/* Update mappings not managed by KFD */
list_for_each_entry(peer_vm, &process_info->vm_list_head,
vm_list_node) {
Copyright © 1996 – 2023 Jason A. Donenfeld. All Rights Reverse Engineered.