Using the standard clear turned out to be to inflexible.
First of all it is executed on the system queue, together with buffer
moves instead on the per VM queue.
And second we need to fill in the page tables with more than just zero.
Signed-off-by: Christian König <christian.koenig at amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 116 ++++++++++++++++++++++++++-------
1 file changed, 92 insertions(+), 24 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index 0df52cb1765b..5cdd8d9c3311 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -257,6 +257,74 @@ bool amdgpu_vm_ready(struct amdgpu_vm *vm)
return ready;
}
+/**
+ * amdgpu_vm_clear_bo - initially clear the PDs/PTs
+ *
+ * @adev: amdgpu_device pointer
+ * @bo: BO to clear
+ * @level: level this BO is at
+ *
+ * Root PD needs to be reserved when calling this.
+ */
+static int amdgpu_vm_clear_bo(struct amdgpu_device *adev,
+ struct amdgpu_vm *vm,
+ struct amdgpu_bo *bo,
+ unsigned level)
+{
+ struct ttm_operation_ctx ctx = { true, false };
+ struct dma_fence *fence = NULL;
+ uint64_t addr, init_value;
+ struct amdgpu_ring *ring;
+ struct amdgpu_job *job;
+ unsigned entries;
+ int r;
+
+ if (vm->pte_support_ats) {
+ init_value = AMDGPU_PTE_DEFAULT_ATC;
+ if (level != AMDGPU_VM_PTB)
+ init_value |= AMDGPU_PDE_PTE;
+ } else {
+ init_value = 0;
+ }
+
+ ring = container_of(vm->entity.sched, struct amdgpu_ring, sched);
+
+ r = reservation_object_reserve_shared(bo->tbo.resv);
+ if (r)
+ return r;
+
+ r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+ if (r)
+ goto error;
+
+ addr = amdgpu_bo_gpu_offset(bo);
+ entries = amdgpu_bo_size(bo) / 8;
+
+ r = amdgpu_job_alloc_with_ib(adev, 64, &job);
+ if (r)
+ goto error;
+
+ amdgpu_vm_set_pte_pde(adev, &job->ibs[0], addr, 0,
+ entries, 0, init_value);
+ amdgpu_ring_pad_ib(ring, &job->ibs[0]);
+
+ WARN_ON(job->ibs[0].length_dw > 64);
+ r = amdgpu_job_submit(job, ring, &vm->entity,
+ AMDGPU_FENCE_OWNER_UNDEFINED, &fence);
+ if (r)
+ goto error_free;
+
+ amdgpu_bo_fence(bo, fence, true);
+ dma_fence_put(fence);
+ return 0;
+
+error_free:
+ amdgpu_job_free(job);
+
+error:
+ return r;
+}
+
/**
* amdgpu_vm_alloc_levels - allocate the PD/PT levels
*
@@ -275,9 +343,8 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev,
{
unsigned shift = amdgpu_vm_level_shift(adev, level);
unsigned pt_idx, from, to;
- int r;
u64 flags;
- uint64_t init_value = 0;
+ int r;
if (!parent->entries) {
unsigned num_entries = amdgpu_vm_num_entries(adev, level);
@@ -300,21 +367,13 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev,
saddr = saddr & ((1 << shift) - 1);
eaddr = eaddr & ((1 << shift) - 1);
- flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS |
- AMDGPU_GEM_CREATE_VRAM_CLEARED;
+ flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
if (vm->use_cpu_for_update)
flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
else
flags |= (AMDGPU_GEM_CREATE_NO_CPU_ACCESS |
AMDGPU_GEM_CREATE_SHADOW);
- if (vm->pte_support_ats) {
- init_value = AMDGPU_PTE_DEFAULT_ATC;
- if (level != AMDGPU_VM_PTB)
- init_value |= AMDGPU_PDE_PTE;
-
- }
-
/* walk over the address space and allocate the page tables */
for (pt_idx = from; pt_idx <= to; ++pt_idx) {
struct reservation_object *resv = vm->root.base.bo->tbo.resv;
@@ -325,12 +384,17 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev,
r = amdgpu_bo_create(adev,
amdgpu_vm_bo_size(adev, level),
AMDGPU_GPU_PAGE_SIZE, true,
- AMDGPU_GEM_DOMAIN_VRAM,
- flags,
- NULL, resv, init_value, &pt);
+ AMDGPU_GEM_DOMAIN_VRAM, flags,
+ NULL, resv, 0, &pt);
if (r)
return r;
+ r = amdgpu_vm_clear_bo(adev, vm, pt, level);
+ if (r) {
+ amdgpu_bo_unref(&pt);
+ return r;
+ }
+
if (vm->use_cpu_for_update) {
r = amdgpu_bo_kmap(pt, NULL);
if (r) {
@@ -2241,11 +2305,11 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
{
const unsigned align = min(AMDGPU_VM_PTB_ALIGN_SIZE,
AMDGPU_VM_PTE_COUNT(adev) * 8);
- uint64_t init_pde_value = 0, flags;
unsigned ring_instance;
struct amdgpu_ring *ring;
struct drm_sched_rq *rq;
unsigned long size;
+ uint64_t flags;
int r, i;
vm->va = RB_ROOT_CACHED;
@@ -2274,23 +2338,19 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode &
AMDGPU_VM_USE_CPU_FOR_COMPUTE);
- if (adev->asic_type == CHIP_RAVEN) {
+ if (adev->asic_type == CHIP_RAVEN)
vm->pte_support_ats = true;
- init_pde_value = AMDGPU_PTE_DEFAULT_ATC
- | AMDGPU_PDE_PTE;
-
- }
- } else
+ } else {
vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode &
AMDGPU_VM_USE_CPU_FOR_GFX);
+ }
DRM_DEBUG_DRIVER("VM update mode is %s\n",
vm->use_cpu_for_update ? "CPU" : "SDMA");
WARN_ONCE((vm->use_cpu_for_update & !amdgpu_vm_is_large_bar(adev)),
"CPU update of VM recommended only for large BAR system\n");
vm->last_update = NULL;
- flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS |
- AMDGPU_GEM_CREATE_VRAM_CLEARED;
+ flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
if (vm->use_cpu_for_update)
flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
else
@@ -2299,7 +2359,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
size = amdgpu_vm_bo_size(adev, adev->vm_manager.root_level);
r = amdgpu_bo_create(adev, size, align, true, AMDGPU_GEM_DOMAIN_VRAM,
- flags, NULL, NULL, init_pde_value,
+ flags, NULL, NULL, 0,
&vm->root.base.bo);
if (r)
goto error_free_sched_entity;
@@ -2308,6 +2368,11 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
if (r)
goto error_free_root;
+ r = amdgpu_vm_clear_bo(adev, vm, vm->root.base.bo,
+ adev->vm_manager.root_level);
+ if (r)
+ goto error_unreserve;
+
vm->root.base.vm = vm;
list_add_tail(&vm->root.base.bo_list, &vm->root.base.bo->va);
list_add_tail(&vm->root.base.vm_status, &vm->evicted);
@@ -2331,6 +2396,9 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
return 0;
+error_unreserve:
+ amdgpu_bo_unreserve(vm->root.base.bo);
+
error_free_root:
amdgpu_bo_unref(&vm->root.base.bo->shadow);
amdgpu_bo_unref(&vm->root.base.bo);
--
2.14.1
