Re: [PATCH 1/2] btrfs: do not SetPageError on a read error for extent buffers

[Filipe Manana <fdmanana@kernel.org>]

On Fri, Feb 11, 2022 at 11:24:38AM -0500, Josef Bacik wrote:
> For reads the page is marked !uptodate and that indicates that there was
> a problem.  We should only be using SetPageError for write errors, and
> in fact do that everywhere except for extent buffer reads.  

We are also doing that for data page reads, at end_page_read() - we clear
the uptodate bit from the page and then set the error bit as well. So it
needs to be fixed as well.

> Fix this so
> we maintain that PageError == write error.
> 
> Signed-off-by: Josef Bacik <josef@toxicpanda.com>
> ---
>  fs/btrfs/extent_io.c | 1 -
>  1 file changed, 1 deletion(-)
> 
> diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
> index 16b6820c913d..bb3c29984fcd 100644
> --- a/fs/btrfs/extent_io.c
> +++ b/fs/btrfs/extent_io.c
> @@ -6671,7 +6671,6 @@ int read_extent_buffer_pages(struct extent_buffer *eb, int wait, int mirror_num)
>  				 * i.e unlock page/set error bit.
>  				 */
>  				ret = err;
> -				SetPageError(page);

The comment above also needs to be upodated, and it mentions we
must set the error bit on the page.

Thanks.

>  				unlock_page(page);
>  				atomic_dec(&eb->io_pages);
>  			}
> -- 
> 2.26.3
>