On Tue, Dec 05, 2017 at 09:29:19AM +0200, Nikolay Borisov wrote:
> This function was introduced by 247e743cbe6e ("Btrfs: Use async helpers to deal
> with pages that have been improperly dirtied") and it didn't do any error
> handling then. This function might very well fail in ENOMEM situation, yet
> it's not handled, this could lead to inconsistent state. So let's handle the
> failure by setting the mapping error bit.
In current code the ENOMEM cannot happen (returned by __set_extent_bit),
but EEXIST could, so there's a possibility of an unhandled error.
> Signed-off-by: Nikolay Borisov <nborisov@xxxxxxxx>
> Cc: stable@xxxxxxxxxxxxxxx
> ---
> fs/btrfs/inode.c | 11 +++++++++--
> 1 file changed, 9 insertions(+), 2 deletions(-)
>
> diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
> index 993061f83067..7a5a46fefdb4 100644
> --- a/fs/btrfs/inode.c
> +++ b/fs/btrfs/inode.c
> @@ -2098,8 +2098,15 @@ static void btrfs_writepage_fixup_worker(struct btrfs_work *work)
> goto out;
> }
>
> - btrfs_set_extent_delalloc(inode, page_start, page_end, 0, &cached_state,
> - 0);
> + ret = btrfs_set_extent_delalloc(inode, page_start, page_end, 0,
> + &cached_state, 0);
> + if (ret) {
> + mapping_set_error(page->mapping, ret);
> + end_extent_writepage(page, ret, page_start, page_end);
> + ClearPageChecked(page);
So this repeats the cleanup code after the preceding call to
btrfs_delalloc_reserve_space fails, I don't see a better way how to get
out of that so it's probably ok.