| Message ID | 20200619103636.11974-5-m.szyprowski@samsung.com |
|---|---|
| State | Accepted |
| Commit | 39913934e67cab214480cc95a3632696aeb8e71a |
| Headers | show |
| Series | DRM: fix struct sg_table nents vs. orig_nents misuse | expand |
Am 19.06.20 um 12:36 schrieb Marek Szyprowski: > The Documentation/DMA-API-HOWTO.txt states that the dma_map_sg() function > returns the number of the created entries in the DMA address space. > However the subsequent calls to the dma_sync_sg_for_{device,cpu}() and > dma_unmap_sg must be called with the original number of the entries > passed to the dma_map_sg(). > > struct sg_table is a common structure used for describing a non-contiguous > memory buffer, used commonly in the DRM and graphics subsystems. It > consists of a scatterlist with memory pages and DMA addresses (sgl entry), > as well as the number of scatterlist entries: CPU pages (orig_nents entry) > and DMA mapped pages (nents entry). > > It turned out that it was a common mistake to misuse nents and orig_nents > entries, calling DMA-mapping functions with a wrong number of entries or > ignoring the number of mapped entries returned by the dma_map_sg() > function. > > To avoid such issues, lets use a common dma-mapping wrappers operating > directly on the struct sg_table objects and use scatterlist page > iterators where possible. This, almost always, hides references to the > nents and orig_nents entries, making the code robust, easier to follow > and copy/paste safe. > > Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com> > Reviewed-by: Christian König <christian.koenig@amd.com> Any objection that we pick this one and the radeon up into our branches for upstreaming? That should about clashes with other driver changes. Thanks, Christian. > --- > drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c | 6 +++--- > drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 9 +++------ > drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c | 8 ++++---- > 3 files changed, 10 insertions(+), 13 deletions(-) > > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c > index 43d8ed7dbd00..519ce4427fce 100644 > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c > @@ -307,8 +307,8 @@ static struct sg_table *amdgpu_dma_buf_map(struct dma_buf_attachment *attach, > if (IS_ERR(sgt)) > return sgt; > > - if (!dma_map_sg_attrs(attach->dev, sgt->sgl, sgt->nents, dir, > - DMA_ATTR_SKIP_CPU_SYNC)) > + if (dma_map_sgtable(attach->dev, sgt, dir, > + DMA_ATTR_SKIP_CPU_SYNC)) > goto error_free; > break; > > @@ -349,7 +349,7 @@ static void amdgpu_dma_buf_unmap(struct dma_buf_attachment *attach, > struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); > > if (sgt->sgl->page_link) { > - dma_unmap_sg(attach->dev, sgt->sgl, sgt->nents, dir); > + dma_unmap_sgtable(attach->dev, sgt, dir, 0); > sg_free_table(sgt); > kfree(sgt); > } else { > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c > index 5129a996e941..97fb73e5a6ae 100644 > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c > @@ -1025,7 +1025,6 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm) > { > struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); > struct amdgpu_ttm_tt *gtt = (void *)ttm; > - unsigned nents; > int r; > > int write = !(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY); > @@ -1040,9 +1039,8 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm) > goto release_sg; > > /* Map SG to device */ > - r = -ENOMEM; > - nents = dma_map_sg(adev->dev, ttm->sg->sgl, ttm->sg->nents, direction); > - if (nents == 0) > + r = dma_map_sgtable(adev->dev, ttm->sg, direction, 0); > + if (r) > goto release_sg; > > /* convert SG to linear array of pages and dma addresses */ > @@ -1073,8 +1071,7 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm) > return; > > /* unmap the pages mapped to the device */ > - dma_unmap_sg(adev->dev, ttm->sg->sgl, ttm->sg->nents, direction); > - > + dma_unmap_sgtable(adev->dev, ttm->sg, direction, 0); > sg_free_table(ttm->sg); > > #if IS_ENABLED(CONFIG_DRM_AMDGPU_USERPTR) > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c > index d399e5893170..c281aa13f5ec 100644 > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c > @@ -477,11 +477,11 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev, > if (r) > goto error_free; > > - for_each_sg((*sgt)->sgl, sg, num_entries, i) > + for_each_sgtable_sg((*sgt), sg, i) > sg->length = 0; > > node = mem->mm_node; > - for_each_sg((*sgt)->sgl, sg, num_entries, i) { > + for_each_sgtable_sg((*sgt), sg, i) { > phys_addr_t phys = (node->start << PAGE_SHIFT) + > adev->gmc.aper_base; > size_t size = node->size << PAGE_SHIFT; > @@ -501,7 +501,7 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev, > return 0; > > error_unmap: > - for_each_sg((*sgt)->sgl, sg, num_entries, i) { > + for_each_sgtable_sg((*sgt), sg, i) { > if (!sg->length) > continue; > > @@ -532,7 +532,7 @@ void amdgpu_vram_mgr_free_sgt(struct amdgpu_device *adev, > struct scatterlist *sg; > int i; > > - for_each_sg(sgt->sgl, sg, sgt->nents, i) > + for_each_sgtable_sg(sgt, sg, i) > dma_unmap_resource(dev, sg->dma_address, > sg->length, dir, > DMA_ATTR_SKIP_CPU_SYNC);
Hi Christian, On 22.06.2020 15:27, Christian König wrote: > Am 19.06.20 um 12:36 schrieb Marek Szyprowski: >> The Documentation/DMA-API-HOWTO.txt states that the dma_map_sg() >> function >> returns the number of the created entries in the DMA address space. >> However the subsequent calls to the dma_sync_sg_for_{device,cpu}() and >> dma_unmap_sg must be called with the original number of the entries >> passed to the dma_map_sg(). >> >> struct sg_table is a common structure used for describing a >> non-contiguous >> memory buffer, used commonly in the DRM and graphics subsystems. It >> consists of a scatterlist with memory pages and DMA addresses (sgl >> entry), >> as well as the number of scatterlist entries: CPU pages (orig_nents >> entry) >> and DMA mapped pages (nents entry). >> >> It turned out that it was a common mistake to misuse nents and >> orig_nents >> entries, calling DMA-mapping functions with a wrong number of entries or >> ignoring the number of mapped entries returned by the dma_map_sg() >> function. >> >> To avoid such issues, lets use a common dma-mapping wrappers operating >> directly on the struct sg_table objects and use scatterlist page >> iterators where possible. This, almost always, hides references to the >> nents and orig_nents entries, making the code robust, easier to follow >> and copy/paste safe. >> >> Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com> >> Reviewed-by: Christian König <christian.koenig@amd.com> > > Any objection that we pick this one and the radeon up into our > branches for upstreaming? > > That should about clashes with other driver changes. I'm fine. This one and radeon doesn't depend on the prime changes, so it should merge fine via your tree. I will try to ask for more review of the remaining patches and then try merging via drm-misc. Best regards
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c index 43d8ed7dbd00..519ce4427fce 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c @@ -307,8 +307,8 @@ static struct sg_table *amdgpu_dma_buf_map(struct dma_buf_attachment *attach, if (IS_ERR(sgt)) return sgt; - if (!dma_map_sg_attrs(attach->dev, sgt->sgl, sgt->nents, dir, - DMA_ATTR_SKIP_CPU_SYNC)) + if (dma_map_sgtable(attach->dev, sgt, dir, + DMA_ATTR_SKIP_CPU_SYNC)) goto error_free; break; @@ -349,7 +349,7 @@ static void amdgpu_dma_buf_unmap(struct dma_buf_attachment *attach, struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); if (sgt->sgl->page_link) { - dma_unmap_sg(attach->dev, sgt->sgl, sgt->nents, dir); + dma_unmap_sgtable(attach->dev, sgt, dir, 0); sg_free_table(sgt); kfree(sgt); } else { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 5129a996e941..97fb73e5a6ae 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -1025,7 +1025,6 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); struct amdgpu_ttm_tt *gtt = (void *)ttm; - unsigned nents; int r; int write = !(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY); @@ -1040,9 +1039,8 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm) goto release_sg; /* Map SG to device */ - r = -ENOMEM; - nents = dma_map_sg(adev->dev, ttm->sg->sgl, ttm->sg->nents, direction); - if (nents == 0) + r = dma_map_sgtable(adev->dev, ttm->sg, direction, 0); + if (r) goto release_sg; /* convert SG to linear array of pages and dma addresses */ @@ -1073,8 +1071,7 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm) return; /* unmap the pages mapped to the device */ - dma_unmap_sg(adev->dev, ttm->sg->sgl, ttm->sg->nents, direction); - + dma_unmap_sgtable(adev->dev, ttm->sg, direction, 0); sg_free_table(ttm->sg); #if IS_ENABLED(CONFIG_DRM_AMDGPU_USERPTR) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c index d399e5893170..c281aa13f5ec 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c @@ -477,11 +477,11 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev, if (r) goto error_free; - for_each_sg((*sgt)->sgl, sg, num_entries, i) + for_each_sgtable_sg((*sgt), sg, i) sg->length = 0; node = mem->mm_node; - for_each_sg((*sgt)->sgl, sg, num_entries, i) { + for_each_sgtable_sg((*sgt), sg, i) { phys_addr_t phys = (node->start << PAGE_SHIFT) + adev->gmc.aper_base; size_t size = node->size << PAGE_SHIFT; @@ -501,7 +501,7 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev, return 0; error_unmap: - for_each_sg((*sgt)->sgl, sg, num_entries, i) { + for_each_sgtable_sg((*sgt), sg, i) { if (!sg->length) continue; @@ -532,7 +532,7 @@ void amdgpu_vram_mgr_free_sgt(struct amdgpu_device *adev, struct scatterlist *sg; int i; - for_each_sg(sgt->sgl, sg, sgt->nents, i) + for_each_sgtable_sg(sgt, sg, i) dma_unmap_resource(dev, sg->dma_address, sg->length, dir, DMA_ATTR_SKIP_CPU_SYNC);
The Documentation/DMA-API-HOWTO.txt states that the dma_map_sg() function returns the number of the created entries in the DMA address space. However the subsequent calls to the dma_sync_sg_for_{device,cpu}() and dma_unmap_sg must be called with the original number of the entries passed to the dma_map_sg(). struct sg_table is a common structure used for describing a non-contiguous memory buffer, used commonly in the DRM and graphics subsystems. It consists of a scatterlist with memory pages and DMA addresses (sgl entry), as well as the number of scatterlist entries: CPU pages (orig_nents entry) and DMA mapped pages (nents entry). It turned out that it was a common mistake to misuse nents and orig_nents entries, calling DMA-mapping functions with a wrong number of entries or ignoring the number of mapped entries returned by the dma_map_sg() function. To avoid such issues, lets use a common dma-mapping wrappers operating directly on the struct sg_table objects and use scatterlist page iterators where possible. This, almost always, hides references to the nents and orig_nents entries, making the code robust, easier to follow and copy/paste safe. Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com> Reviewed-by: Christian König <christian.koenig@amd.com> --- drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c | 6 +++--- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 9 +++------ drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c | 8 ++++---- 3 files changed, 10 insertions(+), 13 deletions(-)