Re: [PATCH] overlay: Plumb through flush method

[Amir Goldstein <amir73il@gmail.com>]

On Wed, Jan 20, 2021 at 5:24 PM Miklos Szeredi <miklos@szeredi.hu> wrote:
>
> On Thu, Dec 3, 2020 at 7:32 PM Sargun Dhillon <sargun@sargun.me> wrote:
> >
> > On Thu, Dec 3, 2020 at 2:32 AM Amir Goldstein <amir73il@gmail.com> wrote:
> > >
> > > On Thu, Dec 3, 2020 at 12:16 PM Sargun Dhillon <sargun@sargun.me> wrote:
> > > >
> > > > On Sun, Nov 29, 2020 at 7:00 PM Sargun Dhillon <sargun@sargun.me> wrote:
> > > > >
> > > > > Filesystems can implement their own flush method that release
> > > > > resources, or manipulate caches. Currently if one of these
> > > > > filesystems is used with overlayfs, the flush method is not called.
> > > > >
> > > > > Signed-off-by: Sargun Dhillon <sargun@sargun.me>
> > > > > Cc: linux-fsdevel@vger.kernel.org
> > > > > Cc: linux-unionfs@vger.kernel.org
> > > > > Cc: Miklos Szeredi <miklos@szeredi.hu>
> > > > > Cc: Amir Goldstein <amir73il@gmail.com>
> > > > > ---
> > > > >  fs/overlayfs/file.c | 11 +++++++++++
> > > > >  1 file changed, 11 insertions(+)
> > > > >
> > > > > diff --git a/fs/overlayfs/file.c b/fs/overlayfs/file.c
> > > > > index efccb7c1f9bc..802259f33c28 100644
> > > > > --- a/fs/overlayfs/file.c
> > > > > +++ b/fs/overlayfs/file.c
> > > > > @@ -787,6 +787,16 @@ static loff_t ovl_remap_file_range(struct file *file_in, loff_t pos_in,
> > > > >                             remap_flags, op);
> > > > >  }
> > > > >
> > > > > +static int ovl_flush(struct file *file, fl_owner_t id)
> > > > > +{
> > > > > +       struct file *realfile = file->private_data;
> > > > > +
> > > > > +       if (realfile->f_op->flush)
> > > > > +               return realfile->f_op->flush(realfile, id);
> > > > > +
> > > > > +       return 0;
> > > > > +}
> > > > > +
> > > > >  const struct file_operations ovl_file_operations = {
> > > > >         .open           = ovl_open,
> > > > >         .release        = ovl_release,
> > > > > @@ -798,6 +808,7 @@ const struct file_operations ovl_file_operations = {
> > > > >         .fallocate      = ovl_fallocate,
> > > > >         .fadvise        = ovl_fadvise,
> > > > >         .unlocked_ioctl = ovl_ioctl,
> > > > > +       .flush          = ovl_flush,
> > > > >  #ifdef CONFIG_COMPAT
> > > > >         .compat_ioctl   = ovl_compat_ioctl,
> > > > >  #endif
> > > > > --
> > > > > 2.25.1
> > > > >
> > > >
> > > > Amir, Miklos,
> > > > Is this acceptable? I discovered this being a problem when we had the discussion
> > > > of whether the volatile fs should return an error on close on dirty files.
> > >
> > > Yes, looks ok.
> > > Maybe we want to check if the realfile is upper although
> > > maybe flush can release resources also on read only fs?
> > >
> > > >
> > > > It seems like it would be useful if anyone uses NFS, or CIFS as an upperdir.
> > >
> > > They are not supported as upperdir. only FUSE is.
> > >
> > > Thanks,
> > > Amir.
> >
> > VFS does it on read-only files / mounts, so we should probably do the
> > same thing.
>
> Right, but it should handle files copied up after the oipen (i.e. call
> ovl_real_fdget() to get the real file).
>

Applied patch is missing fdput() xfstests fail.
Following tested fix.

Thanks,
Amir.

--- a/fs/overlayfs/file.c
+++ b/fs/overlayfs/file.c
@@ -693,12 +693,17 @@ static int ovl_flush(struct file *file, fl_owner_t id)
        int err;

        err = ovl_real_fdget(file, &real);
-       if (!err && real.file->f_op->flush) {
+       if (err)
+               return err;
+
+       if (real.file->f_op->flush) {
                old_cred = ovl_override_creds(file_inode(file)->i_sb);
                err = real.file->f_op->flush(real.file, id);
                revert_creds(old_cred);
        }

+       fdput(real);
+
        return err;
 }