Re: [PATCH v2] xfs: test inode allocation state missmatch corruption

[Dave Chinner <david@fromorbit.com>]

On Mon, May 21, 2018 at 08:55:30AM -0700, Darrick J. Wong wrote:
> On Mon, May 21, 2018 at 05:04:11PM +0800, Eryu Guan wrote:
> > On Mon, May 21, 2018 at 09:58:57AM +0800, Zorro Lang wrote:
> > > +$XFS_IO_PROG -fc 'pwrite 0 $blksz' -c fsync $SCRATCH_MNT/dir/newfile 2>&1 | \
> > > +	grep -i "Structure needs cleaning" | _filter_scratch
> > > +
> > > +# filter a intentional internal errors
> > > +_check_dmesg _filter_dmesg
> > > +
> > > +# success, all done
> > > +status=0
> 
> Does this test leave the scratch fs mounted but shut down if the kernel
> hasn't been patched?  The _require_scratch_nocheck means that the broken
> fs doesn't get scraped off the system and subsequent tests can go all
> wacky when operations on $SCRATCH_MNT error out...

There's another discussion going on about fix this because xfs/132
triggers the same issue (it's a problem with _check_filesystem() not
cycling the mount after the test when _require_scratch_nocheck() is
used).

I didn't see it here in testing because there was no test run after
the new test I proposed. Same here - its xfs/999, and no test is run
after it so this sort of thing isn't noticed.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com