Re: xfs_check vs. xfs_repair vs. the world^W^Wfstests

["Darrick J. Wong" <darrick.wong@oracle.com>]

On Thu, Apr 09, 2020 at 12:40:28AM -0700, Christoph Hellwig wrote:
> On Wed, Apr 08, 2020 at 10:23:21AM +0530, Chandan Rajendra wrote:
> > > Not sure what to do about quota in repair -- we could build in the
> > > ability to do quota counts since we scan the whole inode table and
> > > directory tree anyway.  From there it's not so hard to rebuild the quota
> > > inodes too.
> > >
> > 
> > I will take up this work and get it completed.
> > 
> > Since I have other higher priority tasks at work place, I will have this as my
> > secondary focus. Meanwhile, until it gets done, can we disable running these
> > tests on block size > 4k i.e. https://patchwork.kernel.org/patch/11454399/.
> 
> I still think even with the above outstanding issues we should not
> run check by default.  We can still opt into it, but check extends
> the run time of xfstests for not very good reasons.

So do I, and when I proposed it back in 2018[1], Dave indicated that we
ought to get to the point where we don't need to xfs_check as a
cross-check for xfs_repair.  I'm working on fixing up the smaller things
that repair doesn't catch, so once that series lands maybe we can just
drop xfs_check?

(Since the only big discrepancy at that point will be the lack of quota
checking...)

--D

[1] https://lore.kernel.org/fstests/20180214212228.GG7000@dastard/