On Mon, Aug 10, 2020 at 06:48:19PM -0600, Jens Axboe wrote:
> With the change allowing read-ahead for IOCB_NOWAIT, we changed the
> RWF_NOWAIT semantics of only doing cached reads. Since we know have
> IOCB_NOIO to manage that specific side of it, just make RWF_NOWAIT
> imply IOCB_NOIO as well to restore the previous behavior.
> 
> Fixes: 2e85abf053b9 ("mm: allow read-ahead with IOCB_NOWAIT set")
> Reported-by: Dave Chinner <da...@fromorbit.com>
> Signed-off-by: Jens Axboe <ax...@kernel.dk>
> 
> ---
> 
> This was a known change with the buffered async read change, but we
> didn't have IOCB_NOIO until late in 5.8. Now that bases are synced,
> make the change to make RWF_NOWAIT behave like past kernels.
> 
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index bd7ec3eaeed0..f1cca4bfdd7b 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -3293,7 +3293,7 @@ static inline int kiocb_set_rw_flags(struct kiocb *ki, 
> rwf_t flags)
>       if (flags & RWF_NOWAIT) {
>               if (!(ki->ki_filp->f_mode & FMODE_NOWAIT))
>                       return -EOPNOTSUPP;
> -             kiocb_flags |= IOCB_NOWAIT;
> +             kiocb_flags |= IOCB_NOWAIT | IOCB_NOIO;
>       }
>       if (flags & RWF_HIPRI)
>               kiocb_flags |= IOCB_HIPRI;

Looks good.

Reviewed-by: Dave Chinner <dchin...@redhat.com>

-- 
Dave Chinner
da...@fromorbit.com

Reply via email to