From: Felix Kuehling <Felix.Kuehling@amd.com> To: amd-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org Cc: christian.koenig@amd.com Subject: [PATCH 7/9] drm/amdgpu: Add DMA mapping of GTT BOs Date: Wed, 14 Apr 2021 02:48:02 -0400 [thread overview] Message-ID: <20210414064804.29356-8-Felix.Kuehling@amd.com> (raw) In-Reply-To: <20210414064804.29356-1-Felix.Kuehling@amd.com> Use DMABufs with dynamic attachment to DMA-map GTT BOs on other GPUs. Signed-off-by: Felix Kuehling <Felix.Kuehling@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 _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
WARNING: multiple messages have this Message-ID (diff)
From: Felix Kuehling <Felix.Kuehling@amd.com> To: amd-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org Cc: christian.koenig@amd.com Subject: [PATCH 7/9] drm/amdgpu: Add DMA mapping of GTT BOs Date: Wed, 14 Apr 2021 02:48:02 -0400 [thread overview] Message-ID: <20210414064804.29356-8-Felix.Kuehling@amd.com> (raw) In-Reply-To: <20210414064804.29356-1-Felix.Kuehling@amd.com> Use DMABufs with dynamic attachment to DMA-map GTT BOs on other GPUs. Signed-off-by: Felix Kuehling <Felix.Kuehling@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 _______________________________________________ amd-gfx mailing list amd-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/amd-gfx
next prev parent reply other threads:[~2021-04-14 6:48 UTC|newest] Thread overview: 41+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-04-14 6:47 [PATCH 0/9] Implement multi-GPU DMA mappings for KFD Felix Kuehling 2021-04-14 6:47 ` Felix Kuehling 2021-04-14 6:47 ` [PATCH 1/9] drm/amdgpu: Rename kfd_bo_va_list to kfd_mem_attachment Felix Kuehling 2021-04-14 6:47 ` Felix Kuehling 2021-04-14 6:47 ` [PATCH 2/9] drm/amdgpu: Keep a bo-reference per-attachment Felix Kuehling 2021-04-14 6:47 ` Felix Kuehling 2021-04-14 6:47 ` [PATCH 3/9] drm/amdgpu: Simplify AQL queue mapping Felix Kuehling 2021-04-14 6:47 ` Felix Kuehling 2021-04-14 6:47 ` [PATCH 4/9] drm/amdgpu: Add multi-GPU DMA mapping helpers Felix Kuehling 2021-04-14 6:47 ` Felix Kuehling 2021-04-14 6:48 ` [PATCH 5/9] drm/amdgpu: DMA map/unmap when updating GPU mappings Felix Kuehling 2021-04-14 6:48 ` Felix Kuehling 2021-04-14 6:48 ` [PATCH 6/9] drm/amdgpu: Move kfd_mem_attach outside reservation Felix Kuehling 2021-04-14 6:48 ` Felix Kuehling 2021-04-14 6:48 ` Felix Kuehling [this message] 2021-04-14 6:48 ` [PATCH 7/9] drm/amdgpu: Add DMA mapping of GTT BOs Felix Kuehling 2021-04-14 6:48 ` [PATCH 8/9] drm/ttm: Don't count pages in SG BOs against pages_limit Felix Kuehling 2021-04-14 6:48 ` Felix Kuehling 2021-04-14 6:51 ` Christian König 2021-04-14 6:51 ` Christian König 2021-04-14 9:15 ` Daniel Vetter 2021-04-14 9:15 ` Daniel Vetter 2021-04-14 9:19 ` Christian König 2021-04-14 9:19 ` Christian König 2021-04-14 10:26 ` Daniel Vetter 2021-04-14 10:26 ` Daniel Vetter 2021-04-14 10:49 ` Christian König 2021-04-14 10:49 ` Christian König 2021-04-14 12:25 ` Daniel Vetter 2021-04-14 12:25 ` Daniel Vetter 2021-04-14 12:43 ` Christian König 2021-04-14 12:43 ` Christian König 2021-04-14 12:47 ` Daniel Vetter 2021-04-14 12:47 ` Daniel Vetter 2021-04-14 12:49 ` Christian König 2021-04-14 12:49 ` Christian König 2021-04-14 14:41 ` Felix Kuehling 2021-04-14 14:41 ` Felix Kuehling 2021-04-14 6:48 ` [PATCH 9/9] drm/amdgpu: Lock the attached dmabuf in unpopulate Felix Kuehling 2021-04-14 6:48 ` Felix Kuehling -- strict thread matches above, loose matches on Subject: below -- 2021-04-14 6:46 [PATCH 0/9] Implement multi-GPU DMA mappings for KFD Felix Kuehling 2021-04-14 6:46 ` [PATCH 7/9] drm/amdgpu: Add DMA mapping of GTT BOs Felix Kuehling
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=20210414064804.29356-8-Felix.Kuehling@amd.com \ --to=felix.kuehling@amd.com \ --cc=amd-gfx@lists.freedesktop.org \ --cc=christian.koenig@amd.com \ --cc=dri-devel@lists.freedesktop.org \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.