The branch main has been updated by manu:
URL:
https://cgit.FreeBSD.org/src/commit/?id=4085bde9fa2ea7ebe861a2e4a07312aac00a26fc
commit 4085bde9fa2ea7ebe861a2e4a07312aac00a26fc
Author: Austin Shafer <asha...@badland.io>
AuthorDate: 2022-12-06 15:25:53 +0000
Commit: Emmanuel Vadot <m...@freebsd.org>
CommitDate: 2022-12-06 15:25:53 +0000
linuxkpi: Fix return value of dma_map_sgtable
dma_map_sgtable internally uses the dma_map_sg_attrs helper. The problem is
that dma_map_sg_attrs returns the number of entries mapped, whereas
dma_map_sgtable returns nonzero on failure. This leads to dma_map_sgtable
returning non-zero-but-positive values which tricks other areas of the stack
into thinking nents is a valid pointer.
This checks if nents is valid and returns zero if so, updating the nents
field
in sgt. This fixes PRIME render offload with nvidia-drm.
Fixes: 9202c95f47c2 ("linuxkpi: Add dma_{un,}map_sgtable")
---
sys/compat/linuxkpi/common/include/linux/dma-mapping.h | 9 ++++++++-
1 file changed, 8 insertions(+), 1 deletion(-)
diff --git a/sys/compat/linuxkpi/common/include/linux/dma-mapping.h
b/sys/compat/linuxkpi/common/include/linux/dma-mapping.h
index d3d25fcce857..7f2579787b4b 100644
--- a/sys/compat/linuxkpi/common/include/linux/dma-mapping.h
+++ b/sys/compat/linuxkpi/common/include/linux/dma-mapping.h
@@ -362,7 +362,14 @@ dma_map_sgtable(struct device *dev, struct sg_table *sgt,
unsigned long attrs)
{
- return (dma_map_sg_attrs(dev, sgt->sgl, sgt->nents, dir, attrs));
+ int nents = dma_map_sg_attrs(dev, sgt->sgl, sgt->nents, dir, attrs);
+
+ if (nents < 0) {
+ return nents;
+ } else {
+ sgt->nents = nents;
+ return 0;
+ }
}
static inline void
