Re: [PATCH] ext4: regression test for "tune2fs -l" after ext4 shutdown

[Zorro Lang <zlang@redhat.com>]

On Wed, Aug 18, 2021 at 04:26:01PM +0200, Jan Kara wrote:
> On Wed 18-08-21 21:20:44, Boyang Xue wrote:
> > > > +
> > > > +# real QA test starts here
> > > > +_supported_fs ext4
> > >
> > > I'm wondering if this case can be a generic case, there's nothing
> > > ext4 specified operations, except this line:
> > >
> > > "$TUNE2FS_PROG -l $SCRATCH_DEV"
> > >
> > > Hmm... if we can change this line to something likes _get_fs_super(),
> > > it might help to make this test to be a generic test.
> > 
> > I think this bug is heavily related to "tune2fs", ext4 only. So I
> > guess an ext4 only test is enough?
> 
> FWIW I agree with Boyang here. For this test to make sense for any other
> filesystem other the filesystem would need to read fs metadata through
> buffer cache in _get_fs_super(). Furthermore it is somewhat ext2/3/4
> specific (due to historical reasons) that reading superblock from the
> buffer cache of a mounted filesystem is expected to result in something
> sensible. Usually this is plain unsupported use...

Thanks for this explanation:) I didn't ask for extending this test to be
a generic test, just checking others ideas:) Due to although tune2fs is
special, but the test steps are common:
1) mkfs
2) mount
3) write io
4) shutdown fs
5) umount && mount
6) read sb from a mounted fs (make sure using tune2fs for ext4)

Anyway, keep this test as ext4 only is fine for me :)

Thanks,
Zorro

> 
> 								Honza
> -- 
> Jan Kara <jack@suse.com>
> SUSE Labs, CR
>