diff mbox series

drm/nouveau/mmu: fix reuse of nvkm_umem

Message ID 20220207161443.1843660-1-kherbst@redhat.com (mailing list archive)
State New, archived
Headers show
Series drm/nouveau/mmu: fix reuse of nvkm_umem | expand

Commit Message

Karol Herbst Feb. 7, 2022, 4:14 p.m. UTC 
I am not entirely sure if this fixes anything, but the code standed out
while investigated problematic calls to vunmap.

nvkm_umem.io is only ever set for the NVKM_OBJECT_MAP_IO case in
nvkm_umem_map, but never for the NVKM_OBJECT_MAP_VA one, which could lead
to taking the wrong patch inside nvkm_umem_unmap.

I just don't know if this is a real issue or not, but the code doesn't
look correct this way.

Fixes: c83c4097eba8 ("drm/nouveau/mmu: define user interfaces to mmu memory allocation")
Cc: <stable@vger.kernel.org> # v4.15+
---
 drivers/gpu/drm/nouveau/nvkm/subdev/mmu/umem.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
diff mbox series

Patch

diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/umem.c b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/umem.c
index e530bb8b3b17..2608e0796066 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/umem.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/mmu/umem.c
@@ -102,6 +102,7 @@  nvkm_umem_map(struct nvkm_object *object, void *argv, u32 argc,
 		*handle = (unsigned long)(void *)umem->map;
 		*length = nvkm_memory_size(umem->memory);
 		*type = NVKM_OBJECT_MAP_VA;
+		umem->io = false;
 		return 0;
 	} else
 	if ((umem->type & NVKM_MEM_VRAM) ||
@@ -112,12 +113,11 @@  nvkm_umem_map(struct nvkm_object *object, void *argv, u32 argc,
 			return ret;
 
 		*type = NVKM_OBJECT_MAP_IO;
-	} else {
-		return -EINVAL;
+		umem->io = true;
+		return 0;
 	}
 
-	umem->io = (*type == NVKM_OBJECT_MAP_IO);
-	return 0;
+	return -EINVAL;
 }
 
 static void *