On Mon, Oct 21, 2019 at 02:45:39PM +0100, David Howells wrote:
> Nicolas Pitre <nico@xxxxxxxxxxx> wrote:
>
> > From: Maxime Bizon <mbizon@xxxxxxxxxx>
> >
> > When both CONFIG_CRAMFS_MTD and CONFIG_CRAMFS_BLOCKDEV are enabled, if
> > we fail to mount on MTD, we don't try on block device.
> >
> > Fixes: 74f78fc5ef43 ("vfs: Convert cramfs to use the new mount API")
> >
> > Signed-off-by: Maxime Bizon <mbizon@xxxxxxxxxx>
> > Signed-off-by: Nicolas Pitre <nico@xxxxxxxxxxx>
>
> Acked-by: David Howells <dhowells@xxxxxxxxxx>
FWIW, the thing that worries me here is the possibility of
side effects on fs_context in case if fill_super fails really
late... OTOH, cramfs one seems to be safe in that respect.
OK, will apply, but that's fairly brittle and needs to be
documented. If we *ever* grow non-trivial options parsing
there, that'll be a serious landmine. If something gets
transferred from fs_context into a superblock, which
fails later in setup and takes that object with it, the
second part (get_tree_bdev()) would be in trouble.
