[PATCH 7/9] drm/amdgpu: Add DMA mapping of GTT BOs

[Felix Kuehling]

Use DMABufs with dynamic attachment to DMA-map GTT BOs on other GPUs.

Signed-off-by: Felix Kuehling <Felix.Kuehling at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h    |  2 +
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c  | 74 ++++++++++++++++++-
 2 files changed, 75 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index fc3514ed1b74..3ea51982b720 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -41,6 +41,7 @@ struct amdgpu_device;
 enum kfd_mem_attachment_type {
 	KFD_MEM_ATT_SHARED,	/* Share kgd_mem->bo or another attachment's */
 	KFD_MEM_ATT_USERPTR,	/* SG bo to DMA map pages from a userptr bo */
+	KFD_MEM_ATT_DMABUF,	/* DMAbuf to DMA map TTM BOs */
 };
 
 struct kfd_mem_attachment {
@@ -56,6 +57,7 @@ struct kfd_mem_attachment {
 struct kgd_mem {
 	struct mutex lock;
 	struct amdgpu_bo *bo;
+	struct dma_buf *dmabuf;
 	struct list_head attachments;
 	/* protected by amdkfd_process_info.lock */
 	struct ttm_validate_buffer validate_list;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index 1416f3c03f1d..bb3a96ab8f20 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -522,6 +522,16 @@ kfd_mem_dmamap_userptr(struct kgd_mem *mem,
 	return ret;
 }
 
+static int
+kfd_mem_dmamap_dmabuf(struct kfd_mem_attachment *attachment)
+{
+	struct ttm_operation_ctx ctx = {.interruptible = true};
+	struct amdgpu_bo *bo = attachment->bo_va->base.bo;
+
+	amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT);
+	return ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+}
+
 static int
 kfd_mem_dmamap_attachment(struct kgd_mem *mem,
 			  struct kfd_mem_attachment *attachment)
@@ -531,6 +541,8 @@ kfd_mem_dmamap_attachment(struct kgd_mem *mem,
 		return 0;
 	case KFD_MEM_ATT_USERPTR:
 		return kfd_mem_dmamap_userptr(mem, attachment);
+	case KFD_MEM_ATT_DMABUF:
+		return kfd_mem_dmamap_dmabuf(attachment);
 	default:
 		WARN_ON_ONCE(1);
 	}
@@ -560,6 +572,19 @@ kfd_mem_dmaunmap_userptr(struct kgd_mem *mem,
 	ttm->sg = NULL;
 }
 
+static void
+kfd_mem_dmaunmap_dmabuf(struct kfd_mem_attachment *attachment)
+{
+	struct ttm_operation_ctx ctx = {.interruptible = true};
+	struct amdgpu_bo *bo = attachment->bo_va->base.bo;
+
+	amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU);
+	ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+	/* FIXME: This does not guarantee that amdgpu_ttm_tt_unpopulate is
+	 * called
+	 */
+}
+
 static void
 kfd_mem_dmaunmap_attachment(struct kgd_mem *mem,
 			    struct kfd_mem_attachment *attachment)
@@ -570,6 +595,9 @@ kfd_mem_dmaunmap_attachment(struct kgd_mem *mem,
 	case KFD_MEM_ATT_USERPTR:
 		kfd_mem_dmaunmap_userptr(mem, attachment);
 		break;
+	case KFD_MEM_ATT_DMABUF:
+		kfd_mem_dmaunmap_dmabuf(attachment);
+		break;
 	default:
 		WARN_ON_ONCE(1);
 	}
@@ -601,6 +629,36 @@ kfd_mem_attach_userptr(struct amdgpu_device *adev, struct kgd_mem *mem,
 	return 0;
 }
 
+static int
+kfd_mem_attach_dmabuf(struct amdgpu_device *adev, struct kgd_mem *mem,
+		      struct amdgpu_bo **bo)
+{
+	struct drm_gem_object *gobj;
+
+	if (!mem->dmabuf) {
+		mem->dmabuf = amdgpu_gem_prime_export(&mem->bo->tbo.base,
+			mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_WRITABLE ?
+				DRM_RDWR : 0);
+		if (IS_ERR(mem->dmabuf)) {
+			mem->dmabuf = NULL;
+			return PTR_ERR(mem->dmabuf);
+		}
+	}
+
+	gobj = amdgpu_gem_prime_import(&adev->ddev, mem->dmabuf);
+	if (IS_ERR(gobj))
+		return PTR_ERR(gobj);
+
+	/* Import takes an extra reference on the dmabuf. Drop it now to
+	 * avoid leaking it. We only need the one reference in
+	 * kgd_mem->dmabuf.
+	 */
+	dma_buf_put(mem->dmabuf);
+
+	*bo = gem_to_amdgpu_bo(gobj);
+	return 0;
+}
+
 /* kfd_mem_attach - Add a BO to a VM
  *
  * Everything that needs to bo done only once when a BO is first added
@@ -658,8 +716,20 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem,
 			ret = kfd_mem_attach_userptr(adev, mem, &bo[i]);
 			if (ret)
 				goto unwind;
+		} else if (mem->domain == AMDGPU_GEM_DOMAIN_GTT &&
+			   mem->bo->tbo.type != ttm_bo_type_sg) {
+			/* GTT BOs use DMA-mapping ability of dynamic-attach
+			 * DMA bufs. TODO: The same should work for VRAM on
+			 * large-BAR GPUs.
+			 */
+			attachment[i]->type = KFD_MEM_ATT_DMABUF;
+			ret = kfd_mem_attach_dmabuf(adev, mem, &bo[i]);
+			if (ret)
+				goto unwind;
 		} else {
-			/* FIXME: Need to DMA-map other BO types */
+			/* FIXME: Need to DMA-map other BO types:
+			 * large-BAR VRAM, doorbells, MMIO remap
+			 */
 			attachment[i]->type = KFD_MEM_ATT_SHARED;
 			bo[i] = mem->bo;
 			drm_gem_object_get(&bo[i]->tbo.base);
@@ -1558,6 +1628,8 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
 	}
 
 	/* Free the BO*/
+	if (mem->dmabuf)
+		dma_buf_put(mem->dmabuf);
 	drm_gem_object_put(&mem->bo->tbo.base);
 	mutex_destroy(&mem->lock);
 	kfree(mem);
-- 
2.31.1