From c0dd8a9253fadfb8e5357217d085f1989da4ef0a Mon Sep 17 00:00:00 2001 From: Matthew Auld Date: Mon, 7 Apr 2025 15:18:25 +0100 Subject: drm/amdgpu/dma_buf: fix page_link check MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The page_link lower bits of the first sg could contain something like SG_END, if we are mapping a single VRAM page or contiguous blob which fits into one sg entry. Rather pull out the struct page, and use that in our check to know if we mapped struct pages vs VRAM. Fixes: f44ffd677fb3 ("drm/amdgpu: add support for exporting VRAM using DMA-buf v3") Signed-off-by: Matthew Auld Cc: Christian König Cc: amd-gfx@lists.freedesktop.org Cc: # v5.8+ Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c index 9f627caedc3f..c9842a0e2a1c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c @@ -184,7 +184,7 @@ static void amdgpu_dma_buf_unmap(struct dma_buf_attachment *attach, struct sg_table *sgt, enum dma_data_direction dir) { - if (sgt->sgl->page_link) { + if (sg_page(sgt->sgl)) { dma_unmap_sgtable(attach->dev, sgt, dir, 0); sg_free_table(sgt); kfree(sgt); -- cgit v1.2.3 From f5e7fabd1f5c65b2e077efcdb118cfa67eae7311 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Thu, 9 Jan 2025 11:57:56 -0500 Subject: drm/amdgpu: allow pinning DMA-bufs into VRAM if all importers can do P2P MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Try pinning into VRAM to allow P2P with RDMA NICs without ODP support if all attachments can do P2P. If any attachment can't do P2P just pin into GTT instead. Acked-by: Simona Vetter Signed-off-by: Christian König Signed-off-by: Felix Kuehling Reviewed-by: Felix Kuehling Tested-by: Pak Nin Lui Cc: Simona Vetter Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c | 25 ++++++++++++++++++------- 1 file changed, 18 insertions(+), 7 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c index c9842a0e2a1c..667080cc9ae1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c @@ -75,11 +75,25 @@ static int amdgpu_dma_buf_attach(struct dma_buf *dmabuf, */ static int amdgpu_dma_buf_pin(struct dma_buf_attachment *attach) { - struct drm_gem_object *obj = attach->dmabuf->priv; - struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj); + struct dma_buf *dmabuf = attach->dmabuf; + struct amdgpu_bo *bo = gem_to_amdgpu_bo(dmabuf->priv); + u32 domains = bo->preferred_domains; - /* pin buffer into GTT */ - return amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT); + dma_resv_assert_held(dmabuf->resv); + + /* + * Try pinning into VRAM to allow P2P with RDMA NICs without ODP + * support if all attachments can do P2P. If any attachment can't do + * P2P just pin into GTT instead. + */ + list_for_each_entry(attach, &dmabuf->attachments, node) + if (!attach->peer2peer) + domains &= ~AMDGPU_GEM_DOMAIN_VRAM; + + if (domains & AMDGPU_GEM_DOMAIN_VRAM) + bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; + + return amdgpu_bo_pin(bo, domains); } /** @@ -134,9 +148,6 @@ static struct sg_table *amdgpu_dma_buf_map(struct dma_buf_attachment *attach, r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (r) return ERR_PTR(r); - - } else if (bo->tbo.resource->mem_type != TTM_PL_TT) { - return ERR_PTR(-EBUSY); } switch (bo->tbo.resource->mem_type) { -- cgit v1.2.3 From 5cf3c602df88b471178a5717b17e529d09acad84 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Thu, 17 Apr 2025 10:23:15 -0400 Subject: drm/amdgpu: Use allowed_domains for pinning dmabufs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When determining the domains for pinning DMABufs, filter allowed_domains and fail with a warning if VRAM is forbidden and GTT is not an allowed domain. Fixes: f5e7fabd1f5c ("drm/amdgpu: allow pinning DMA-bufs into VRAM if all importers can do P2P") Suggested-by: Christian König Signed-off-by: Felix Kuehling Reviewed-by: Christian König Signed-off-by: Alex Deucher (cherry picked from commit 3940796a6eefa555fec688a4adee5659ef9fa431) --- drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c index 667080cc9ae1..0446586bd5a7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c @@ -77,7 +77,7 @@ static int amdgpu_dma_buf_pin(struct dma_buf_attachment *attach) { struct dma_buf *dmabuf = attach->dmabuf; struct amdgpu_bo *bo = gem_to_amdgpu_bo(dmabuf->priv); - u32 domains = bo->preferred_domains; + u32 domains = bo->allowed_domains; dma_resv_assert_held(dmabuf->resv); @@ -93,6 +93,9 @@ static int amdgpu_dma_buf_pin(struct dma_buf_attachment *attach) if (domains & AMDGPU_GEM_DOMAIN_VRAM) bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; + if (WARN_ON(!domains)) + return -EINVAL; + return amdgpu_bo_pin(bo, domains); } -- cgit v1.2.3 From 5e56935b519b2fbbca1cafa0cef3c7c3d062f62d Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Tue, 15 Apr 2025 23:58:28 -0400 Subject: drm/amdgpu: Don't pin VRAM without DMABUF_MOVE_NOTIFY MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Pinning of VRAM is for peer devices that don't support dynamic attachment and move notifiers. But it requires that all such peer devices are able to access VRAM via PCIe P2P. Any device without P2P access requires migration to GTT, which fails if the memory is already pinned for another peer device. Sharing between GPUs should not require pinning in VRAM. However, if DMABUF_MOVE_NOTIFY is disabled in the kernel build, even DMABufs shared between GPUs must be pinned, which can lead to failures and functional regressions on systems where some peer GPUs are not P2P accessible. Disable VRAM pinning if move notifiers are disabled in the kernel build to fix regressions when sharing BOs between GPUs. Signed-off-by: Felix Kuehling Tested-by: Hao (Claire) Zhou Reviewed-by: Christian König Signed-off-by: Alex Deucher (cherry picked from commit 05185812ae3695fe049c14847ce3cbeccff1bf2e) --- drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c | 17 ++++++++++++----- 1 file changed, 12 insertions(+), 5 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c index 0446586bd5a7..5740e8d1a522 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c @@ -81,14 +81,21 @@ static int amdgpu_dma_buf_pin(struct dma_buf_attachment *attach) dma_resv_assert_held(dmabuf->resv); - /* - * Try pinning into VRAM to allow P2P with RDMA NICs without ODP + /* Try pinning into VRAM to allow P2P with RDMA NICs without ODP * support if all attachments can do P2P. If any attachment can't do * P2P just pin into GTT instead. + * + * To avoid with conflicting pinnings between GPUs and RDMA when move + * notifiers are disabled, only allow pinning in VRAM when move + * notiers are enabled. */ - list_for_each_entry(attach, &dmabuf->attachments, node) - if (!attach->peer2peer) - domains &= ~AMDGPU_GEM_DOMAIN_VRAM; + if (!IS_ENABLED(CONFIG_DMABUF_MOVE_NOTIFY)) { + domains &= ~AMDGPU_GEM_DOMAIN_VRAM; + } else { + list_for_each_entry(attach, &dmabuf->attachments, node) + if (!attach->peer2peer) + domains &= ~AMDGPU_GEM_DOMAIN_VRAM; + } if (domains & AMDGPU_GEM_DOMAIN_VRAM) bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; -- cgit v1.2.3 From a92741e72f91b904c1d8c3d409ed8dbe9c1f2b26 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Wed, 16 Apr 2025 00:19:13 -0400 Subject: drm/amdgpu: Allow P2P access through XGMI MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit If peer memory is accessible through XGMI, allow leaving it in VRAM rather than forcing its migration to GTT on DMABuf attachment. Signed-off-by: Felix Kuehling Tested-by: Hao (Claire) Zhou Reviewed-by: Christian König Signed-off-by: Alex Deucher (cherry picked from commit 372c8d72c3680fdea3fbb2d6b089f76b4a6d596a) --- drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c | 30 ++++++++++++++++++++++++++++- 1 file changed, 29 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c index 5740e8d1a522..e6913fcf2c7b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c @@ -43,6 +43,29 @@ #include #include +static const struct dma_buf_attach_ops amdgpu_dma_buf_attach_ops; + +/** + * dma_buf_attach_adev - Helper to get adev of an attachment + * + * @attach: attachment + * + * Returns: + * A struct amdgpu_device * if the attaching device is an amdgpu device or + * partition, NULL otherwise. + */ +static struct amdgpu_device *dma_buf_attach_adev(struct dma_buf_attachment *attach) +{ + if (attach->importer_ops == &amdgpu_dma_buf_attach_ops) { + struct drm_gem_object *obj = attach->importer_priv; + struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj); + + return amdgpu_ttm_adev(bo->tbo.bdev); + } + + return NULL; +} + /** * amdgpu_dma_buf_attach - &dma_buf_ops.attach implementation * @@ -54,11 +77,13 @@ static int amdgpu_dma_buf_attach(struct dma_buf *dmabuf, struct dma_buf_attachment *attach) { + struct amdgpu_device *attach_adev = dma_buf_attach_adev(attach); struct drm_gem_object *obj = dmabuf->priv; struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj); struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - if (pci_p2pdma_distance(adev->pdev, attach->dev, false) < 0) + if (!amdgpu_dmabuf_is_xgmi_accessible(attach_adev, bo) && + pci_p2pdma_distance(adev->pdev, attach->dev, false) < 0) attach->peer2peer = false; amdgpu_vm_bo_update_shared(bo); @@ -480,6 +505,9 @@ bool amdgpu_dmabuf_is_xgmi_accessible(struct amdgpu_device *adev, struct drm_gem_object *obj = &bo->tbo.base; struct drm_gem_object *gobj; + if (!adev) + return false; + if (obj->import_attach) { struct dma_buf *dma_buf = obj->import_attach->dmabuf; -- cgit v1.2.3