Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20200421023959.20879-1-harshadshirwadkar@gmail.com>
In-Reply-To: <20200421023959.20879-1-harshadshirwadkar@gmail.com>
From:   harshad shirwadkar <harshadshirwadkar@gmail.com>
Date:   Mon, 20 Apr 2020 20:02:03 -0700
Message-ID: <CAD+ocbzDoOkvtM4KKhmhv9Tke=D9B8q_PfGezfd4H+H_auDJ5w@mail.gmail.com>
Subject: Re: [PATCH] ext4: don't ignore return values from ext4_ext_dirty()
To:     Ext4 Developers List <linux-ext4@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-ext4-owner@vger.kernel.org
Precedence: bulk

On Mon, Apr 20, 2020 at 7:40 PM Harshad Shirwadkar
<harshadshirwadkar@gmail.com> wrote:
>
> Don't ignore return values from ext4_ext_dirty, since the errors
> indicate valid failures below Ext4.  In all of the other instances of
> ext4_ext_dirty calls, the error return value is handled in some
> way. This patch makes those remaining couple of places to handle
> ext4_ext_dirty errors as well. In the longer run, we probably should
> make sure that errors from other mark_dirty routines are handled as
> well.
>
> Ran gce-xfstests smoke tests and verified that there were no
> regressions.
>
> Signed-off-by: Harshad Shirwadkar <harshadshirwadkar@gmail.com>
> ---
>  fs/ext4/extents.c | 5 ++---
>  1 file changed, 2 insertions(+), 3 deletions(-)
>
> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> index f2b577b315a0..f62f55a16fe3 100644
> --- a/fs/ext4/extents.c
> +++ b/fs/ext4/extents.c
> @@ -3244,8 +3244,7 @@ static int ext4_split_extent_at(handle_t *handle,
>
>  fix_extent_len:
>         ex->ee_len = orig_ex.ee_len;
> -       ext4_ext_dirty(handle, inode, path + path->p_depth);
> -       return err;
> +       return ext4_ext_dirty(handle, inode, path + path->p_depth);
>  }

I realized that this is not correct. That's because fix_extent_len is
an error path and this change would make ext4_split_extent_at() return
success if ext4_ext_dirty succeeds in this path. I think instead I
should be adding a comment here explaining why we are not handling
ext4_ext_dirty(). Sorry for that.

- Harshad

>
>  /*
> @@ -3503,7 +3502,7 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
>         }
>         if (allocated) {
>                 /* Mark the block containing both extents as dirty */
> -               ext4_ext_dirty(handle, inode, path + depth);
> +               err = ext4_ext_dirty(handle, inode, path + depth);
>
>                 /* Update path to point to the right extent */
>                 path[depth].p_ext = abut_ex;
> --
> 2.26.1.301.g55bc3eb7cb9-goog
>