On Fri, Aug 18, 2017 at 09:35:25AM +0200, Carlos Maiolino wrote: > Hi Eryu, > > > Thanks a lot for the updated version! It looks much better than v1 :) > > > > Though I hit another issue while testing this patch, please see below. > > > > > + > > > +# agi->freecount is 0 here, corrupt it to show extra free inodes > > > +$XFS_DB_PROG -x -c "agi 0" -c "write freecount 10" $SCRATCH_DEV >> $seqres.full 2>&1 > > > > Minor nit, please use _scratch_xfs_db helper. > > > > NP at all > > > + > > > +_scratch_mount > > > + > > > +# Lock up a buggy kernel > > > +touch $SCRATCH_MNT/lockupfile >> $seqres.full 2>&1 > > > > > > XFS: Assertion failed: freecount == be32_to_cpu(agi->agi_freecount), file: fs/xfs/libxfs/xfs_ialloc.c, line: 246 > > > > Perhaps test should _notrun with CONFIG_XFS_DEBUG on? > > > > The Assert is expected. > > We have debug code to catch the agi->agi_freecount corruption which will explode > in this assert, such code is debug only though because it will search the > whole btree and 'manually' count how many free slots are in the tree, then > compare with agi_freecount. Yeah, I know it's expected, we turn on fatal assert in debug mode. > > Although, I'm not sure if you notice, but even you hit this assert, you will > not be able to unmount the filesystem. > > I'm not sure now if simply disable this test on XFS_DEBUG is the best approach, > hitting the bug in XFS_DEBUG is less catastrophic because it will not lockup, > but will make the FS unmountable. It crashed my host immediately, because I have /proc/sys/kernel/panic_on_oops set to 1. Anyway, an unmountable fs is as bad as a crash, they all prevent subsequent tests from running. The underlying bug is fixed, we expect test to pass, not crash the host or block further testings. > > I'd simply disable it in XFS_DEBUG, once the bug is exactly the same as with non > debug code, but with a different behavior. Not sure though if there are people > running xfstests exclusively on XFS_DEBUG. I believe running xfstests with XFS_DEBUG is pretty common. Thanks, Eryu -- To unsubscribe from this list: send the line "unsubscribe fstests" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
