On Wed, Oct 14, 2015 at 11:23:44AM -0700, Christoph Hellwig wrote:
> On Wed, Oct 14, 2015 at 11:20:13AM -0700, Darrick J. Wong wrote:
> > > As is this makes xfstests rather unhappy by failing tests. I'm not
> > > sure if that's issues in the blockget / xfs_check functionality or
> > > because it actually finds bugs in the kernel code.
> >
> > Which tests, specifically? I ran all the tests in the 'quick' group and of
> > the relatively few errors I saw, I couldn't trace any of them back to
> > blockget.
> >
> > Earlier patch editions caused such problems, but afaict I've fixed them
> > all. :)
>
> shared/006 xfs/076 xfs/206 xfs/250
>
> although xfs/206 also fails just due to reflink output from mkfs, but
> I'll send a patch for that soon.
Current failures here with v4.3-rcX on both sides and a current
xfstests:
Failures: generic/042 xfs/076 xfs/078 xfs/079
042 is expected to fail, xfs/076 is testing sparse inodes (check
doesn't understand them), xfs/078 is failing because I'm testing on
4k sector devices and it wants to use a 2k block size.
xfs/079 is failing because repair is reporting:
found illegal null character in symlink inode 16811448
problem with symbolic link in inode 16811448
would have cleared inode 16811448
but the check is clean. So that may be a new regression. I haven't
looked into it more deeply than that.
Cheers,
Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
|