On Fri, Nov 04, 2016 at 11:54:34AM +1100, Dave Chinner wrote:
> On Thu, Nov 03, 2016 at 05:52:00PM +0800, Eryu Guan wrote:
> > When sb_inopblock is corrupted (out of boundary in this case), XFS
> > should not crash and refuse to mount.
> >
> > Kernel commit 392c6de98af1 ("xfs: sanitize sb_inopblock in
> > xfs_mount_validate_sb") addresses this issue.
>
> Ok, seems like something addressed by the fuzzing tests, but we can
> ignore that for now.
>
> > +_scratch_mkfs >>$seqres.full
> > +
> > +# corrupt sb_inopblock
> > +_scratch_xfs_db -x -c "sb 0" -c "write -c inopblock 512" >>$seqres.full
> > +
> > +# mount corrupted fs
> > +_scratch_mount >>$seqres.full 2>&1
> > +
> > +# no crash, and SCRATCH_DEV should not be mounted either
> > +_is_mounted $SCRATCH_DEV
>
> Rather than test that we caught /one/ corrupt field, why not
> loop here corrupting each superblock field in turn and checking
> that the corruption is caught properly?
>
> i.e. shouldn't we be proactively testing that we're catching all the
> obvious corruptions, rather than just testing the out-of-bounds
> issues that we've already found and fixed?
I've been working on xfstests that do exactly that for a while now and
am getting ready to do a big patch dump of online scrub/repair + more
dangerous_fuzzers tests to see how well the recovery tools work.
In the meantime I doubt it hurts to have this regression test while I
get my act together. :)
--D
>
> Cheers,
>
> Dave.
> --
> Dave Chinner
> david@xxxxxxxxxxxxx
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
