Re: [LTP] [PATCH] tst_supported_fs: Implement checking paths against skiplist

[Petr Vorel <pvorel@suse.cz>]

> Hi!
> > Should I understand this that you don't trust /proc/self/mountinfo output?
> > (show_mountinfo() in fs/proc_namespace.c in kernel git)?

> Not much as I wouldn't trust it, it's just that the ext2/ext3/ext4 names
> are just short names for a subset of additional functionality.

> Consider this:

> $ dd if=/dev/zero of=fsimage.img bs=1024 count=32768
> $ mkfs.ext3 fsimage.img
> $ tune2fs -O ^has_journal fsimage.img
> $ mount -o loop fsimage.img /mntpoint
> $ cat /proc/self/mountinfo
> 43 19 7:0 / /mntpoint rw,relatime - ext3 /dev/loop0 rw

> Now we have mounted ext3 without journaling, is that really ext3?

> $ umount /mntpoint
> $ mount -t ext2 -o loop fsimage.img /mntpoint
> $ cat /proc/self/mountinfo
> 43 19 7:0 / /mntpoint rw,relatime - ext2 /dev/loop0 rw

> Now we mounted the same filesystem as ext2.

> And in both cases this is handled by the ext4 kernel driver anyways:

> $ dmesg
> ...
> EXT4-fs (loop0): mounting ext3 file system using the ext4 subsystem
> ...
> EXT4-fs (loop0): mounting ext2 file system using the ext4 subsystem
> ...

Thanks a lot for a nice example to document the problem.

> So I would argue that these are not three different filesystems, but
> just one that can turn on and off various features. And if you look at
> 'man ext4' you will see that the set of supported features for any extX
> variant does change with kernel version.

Makes sense.

> The question is if we really need to be able to distinguish them or if
> we can keep calling them by a common name.

As I noted before, using Martin's patch require to use ext2/ext3/ext4 for
.skip_filesystems and $TST_SKIP_FILESYSTEMS. I'm OK with it, although it's
a bit misleading.

Kind regards,
Petr

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp