On 2020-08-26 07:32, Marek Szyprowski wrote:
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
drivers/gpu/drm/armada/armada_gem.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/armada/armada_gem.c b/drivers/gpu/drm/armada/armada_gem.c index 8005614d2e6b..bedd8937d8a1 100644 --- a/drivers/gpu/drm/armada/armada_gem.c +++ b/drivers/gpu/drm/armada/armada_gem.c @@ -395,7 +395,7 @@ armada_gem_prime_map_dma_buf(struct dma_buf_attachment *attach, mapping = dobj->obj.filp->f_mapping;
for_each_sg(sgt->sgl, sg, count, i) {
for_each_sgtable_sg(sgt, sg, i) { struct page *page;
page = shmem_read_mapping_page(mapping, i); @@ -407,8 +407,8 @@ armada_gem_prime_map_dma_buf(struct dma_buf_attachment *attach, sg_set_page(sg, page, PAGE_SIZE, 0); }
if (dma_map_sg(attach->dev, sgt->sgl, sgt->nents, dir) == 0) {
num = sgt->nents;
if (dma_map_sgtable(attach->dev, sgt, dir, 0)) {
num = count;
I think it might be even nicer to get rid of "num" entirely and convert the cleanup path to for_each_sgtable_sg() for completeness - AFAICS it should only need an extra "if (sg_page(sg))..." check in that loop. Then "count" could possibly be squashed into its one remaining use as well, but maybe it's worth keeping for readability.
Robin.
goto release; }
} else if (dobj->page) { @@ -418,7 +418,7 @@ armada_gem_prime_map_dma_buf(struct dma_buf_attachment *attach, sg_set_page(sgt->sgl, dobj->page, dobj->obj.size, 0);
if (dma_map_sg(attach->dev, sgt->sgl, sgt->nents, dir) == 0)
} else if (dobj->linear) { /* Single contiguous physical region - no struct page */if (dma_map_sgtable(attach->dev, sgt, dir, 0)) goto free_table;
@@ -449,11 +449,11 @@ static void armada_gem_prime_unmap_dma_buf(struct dma_buf_attachment *attach, int i; if (!dobj->linear)
dma_unmap_sg(attach->dev, sgt->sgl, sgt->nents, dir);
dma_unmap_sgtable(attach->dev, sgt, dir, 0);
if (dobj->obj.filp) { struct scatterlist *sg;
for_each_sg(sgt->sgl, sg, sgt->nents, i)
}for_each_sgtable_sg(sgt, sg, i) put_page(sg_page(sg));