Commit 7141394e authored by Felix Kuehling's avatar Felix Kuehling Committed by Alex Deucher
Browse files

drm/amdgpu: Simplify AQL queue mapping 



Do AQL queue double-mapping with a single attach call. That will make it
easier to create per-GPU BOs later, to be shared between the two BO VA
mappings on the same GPU.

Freeing the attachments is not necessary if map_to_gpu fails. These will be
cleaned up when the kdg_mem object is destroyed in
amdgpu_amdkfd_gpuvm_free_memory_of_gpu.

Signed-off-by: default avatarFelix Kuehling <Felix.Kuehling@amd.com>
Acked-by: default avatarOak Zeng <Oak.Zeng@amd.com>
Acked-by: default avatarRamesh Errabolu <Ramesh.Errabolu@amd.com>
Signed-off-by: default avatarAlex Deucher <alexander.deucher@amd.com>
parent 4e94272f

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c

+48 −55
Original line number Diff line number Diff line
@@ -488,70 +488,76 @@ static uint64_t get_pte_flags(struct amdgpu_device *adev, struct kgd_mem *mem) 
 * 4a.  Validate new page tables and directories

 */

static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem,

		struct amdgpu_vm *vm, bool is_aql,

		struct kfd_mem_attachment **p_attachment)

		struct amdgpu_vm *vm, bool is_aql)

{

	unsigned long bo_size = mem->bo->tbo.base.size;

	uint64_t va = mem->va;

	struct kfd_mem_attachment *attachment;

	struct amdgpu_bo *bo;

	int ret;

	struct kfd_mem_attachment *attachment[2] = {NULL, NULL};

	struct amdgpu_bo *bo[2] = {NULL, NULL};

	int i, ret;



	if (!va) {

		pr_err("Invalid VA when adding BO to VM\n");

		return -EINVAL;

	}



	if (is_aql)

		va += bo_size;



	attachment = kzalloc(sizeof(*attachment), GFP_KERNEL);

	if (!attachment)

		return -ENOMEM;

	for (i = 0; i <= is_aql; i++) {

		attachment[i] = kzalloc(sizeof(*attachment[i]), GFP_KERNEL);

		if (unlikely(!attachment[i])) {

			ret = -ENOMEM;

			goto unwind;

		}



		pr_debug("\t add VA 0x%llx - 0x%llx to vm %p\n", va,

			 va + bo_size, vm);



	/* FIXME: For now all attachments use the same BO. This is incorrect

	 * because one BO can only have one DMA mapping for one GPU. We need

	 * one BO per GPU, e.g. a DMABuf import with dynamic attachment. This

	 * will be addressed one BO-type at a time in subsequent patches.

		/* FIXME: For now all attachments use the same BO. This is

		 * incorrect because one BO can only have one DMA mapping

		 * for one GPU. We need one BO per GPU, e.g. a DMABuf

		 * import with dynamic attachment. This will be addressed

		 * one BO-type at a time in subsequent patches.

		 */

	bo = mem->bo;

	drm_gem_object_get(&bo->tbo.base);

		bo[i] = mem->bo;

		drm_gem_object_get(&bo[i]->tbo.base);



		/* Add BO to VM internal data structures */

	attachment->bo_va = amdgpu_vm_bo_add(adev, vm, bo);

	if (!attachment->bo_va) {

		ret = -EINVAL;

		attachment[i]->bo_va = amdgpu_vm_bo_add(adev, vm, bo[i]);

		if (unlikely(!attachment[i]->bo_va)) {

			ret = -ENOMEM;

			pr_err("Failed to add BO object to VM. ret == %d\n",

			       ret);

		goto err_vmadd;

			goto unwind;

		}



	attachment->va = va;

	attachment->pte_flags = get_pte_flags(adev, mem);

	attachment->adev = adev;

	list_add(&attachment->list, &mem->attachments);

		attachment[i]->va = va;

		attachment[i]->pte_flags = get_pte_flags(adev, mem);

		attachment[i]->adev = adev;

		list_add(&attachment[i]->list, &mem->attachments);



	if (p_attachment)

		*p_attachment = attachment;

		va += bo_size;

	}



	/* Allocate validate page tables if needed */

	ret = vm_validate_pt_pd_bos(vm);

	if (unlikely(ret)) {

		pr_err("validate_pt_pd_bos() failed\n");

		goto err_alloc_pts;

		goto unwind;

	}



	return 0;



err_alloc_pts:

	amdgpu_vm_bo_rmv(adev, attachment->bo_va);

	list_del(&attachment->list);

err_vmadd:

	drm_gem_object_put(&bo->tbo.base);

	kfree(attachment);

unwind:

	for (; i >= 0; i--) {

		if (!attachment[i])

			continue;

		if (attachment[i]->bo_va) {

			amdgpu_vm_bo_rmv(adev, attachment[i]->bo_va);

			list_del(&attachment[i]->list);

		}

		if (bo[i])

			drm_gem_object_put(&bo[i]->tbo.base);

		kfree(attachment[i]);

	}

	return ret;

}
@@ -1384,8 +1390,6 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( 
	uint32_t domain;

	struct kfd_mem_attachment *entry;

	struct bo_vm_reservation_context ctx;

	struct kfd_mem_attachment *attachment = NULL;

	struct kfd_mem_attachment *attachment_aql = NULL;

	unsigned long bo_size;

	bool is_invalid_userptr = false;
@@ -1435,15 +1439,9 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( 
		is_invalid_userptr = true;



	if (!kfd_mem_is_attached(avm, mem)) {

		ret = kfd_mem_attach(adev, mem, avm, false, &attachment);

		ret = kfd_mem_attach(adev, mem, avm, mem->aql_queue);

		if (ret)

			goto attach_failed;

		if (mem->aql_queue) {

			ret = kfd_mem_attach(adev, mem, avm, true,

					     &attachment_aql);

			if (ret)

				goto attach_failed_aql;

		}

	} else {

		ret = vm_validate_pt_pd_bos(avm);

		if (unlikely(ret))
@@ -1498,11 +1496,6 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( 
	goto out;



map_bo_to_gpuvm_failed:

	if (attachment_aql)

		kfd_mem_detach(attachment_aql);

attach_failed_aql:

	if (attachment)

		kfd_mem_detach(attachment);

attach_failed:

	unreserve_bo_and_vms(&ctx, false, false);

out: