[PATCH v2] media: videobuf2: Fix IS_ERR checking in vb2_vmalloc_put_userptr()

[PATCH v2] media: videobuf2: Fix IS_ERR checking in vb2_vmalloc_put_userptr()

[Ma Ke]

In order to avoid error pointers from frame_vector_pages(), we could
use IS_ERR() to check the return value to fix this. This checking
operation could make sure that vector contains pages.

Signed-off-by: Ma Ke <make_ruc2021@163.com>
---
 drivers/media/common/videobuf2/videobuf2-vmalloc.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/drivers/media/common/videobuf2/videobuf2-vmalloc.c b/drivers/media/common/videobuf2/videobuf2-vmalloc.c
index 7c635e292106..5aa66305546d 100644
--- a/drivers/media/common/videobuf2/videobuf2-vmalloc.c
+++ b/drivers/media/common/videobuf2/videobuf2-vmalloc.c
@@ -133,13 +133,15 @@ static void vb2_vmalloc_put_userptr(void *buf_priv)
 
 	if (!buf->vec->is_pfns) {
 		n_pages = frame_vector_count(buf->vec);
-		pages = frame_vector_pages(buf->vec);
 		if (vaddr)
 			vm_unmap_ram((void *)vaddr, n_pages);
 		if (buf->dma_dir == DMA_FROM_DEVICE ||
-		    buf->dma_dir == DMA_BIDIRECTIONAL)
-			for (i = 0; i < n_pages; i++)
-				set_page_dirty_lock(pages[i]);
+		    buf->dma_dir == DMA_BIDIRECTIONAL){
+			pages = frame_vector_pages(buf->vec);
+			if (!WARN_ON_ONCE(IS_ERR(pages)))
+				for (i = 0; i < n_pages; i++)
+					set_page_dirty_lock(pages[i]);
+		}
 	} else {
 		iounmap((__force void __iomem *)buf->vaddr);
 	}
-- 
2.37.2

Re: [PATCH v2] media: videobuf2: Fix IS_ERR checking in vb2_vmalloc_put_userptr()

[Tomasz Figa]

On Tue, Oct 3, 2023 at 6:13 PM Ma Ke <make_ruc2021@163.com> wrote:
>
> In order to avoid error pointers from frame_vector_pages(), we could
> use IS_ERR() to check the return value to fix this. This checking
> operation could make sure that vector contains pages.
>
> Signed-off-by: Ma Ke <make_ruc2021@163.com>
> ---
>  drivers/media/common/videobuf2/videobuf2-vmalloc.c | 10 ++++++----
>  1 file changed, 6 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/media/common/videobuf2/videobuf2-vmalloc.c b/drivers/media/common/videobuf2/videobuf2-vmalloc.c
> index 7c635e292106..5aa66305546d 100644
> --- a/drivers/media/common/videobuf2/videobuf2-vmalloc.c
> +++ b/drivers/media/common/videobuf2/videobuf2-vmalloc.c
> @@ -133,13 +133,15 @@ static void vb2_vmalloc_put_userptr(void *buf_priv)
>
>         if (!buf->vec->is_pfns) {
>                 n_pages = frame_vector_count(buf->vec);
> -               pages = frame_vector_pages(buf->vec);
>                 if (vaddr)
>                         vm_unmap_ram((void *)vaddr, n_pages);
>                 if (buf->dma_dir == DMA_FROM_DEVICE ||
> -                   buf->dma_dir == DMA_BIDIRECTIONAL)
> -                       for (i = 0; i < n_pages; i++)
> -                               set_page_dirty_lock(pages[i]);
> +                   buf->dma_dir == DMA_BIDIRECTIONAL){

Missing space between ) and { .

> +                       pages = frame_vector_pages(buf->vec);
> +                       if (!WARN_ON_ONCE(IS_ERR(pages)))

I still think this cannot happen, but the frame vector code is a mess,
so being a bit more defensive here is probably fine, until someone can
clean up that code (probably by removing the PFN-only code path
completely).

> +                               for (i = 0; i < n_pages; i++)
> +                                       set_page_dirty_lock(pages[i]);
> +               }
>         } else {
>                 iounmap((__force void __iomem *)buf->vaddr);
>         }
> --
> 2.37.2
>

With that one formatting issue fixed: (Mauro, Hans, maybe it could be
fixed up when applying?)

Acked-by: Tomasz Figa <tfiga@chromium.org>

Best regards,
Tomasz