Re: generic: test statx attribute_mask setting

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

On Thu, Apr 25, 2019 at 04:42:55PM -0500, Steve French wrote:
> I verified that although the "attributes_mask"
>  is not returned by current master branch of xfsprogs (in xfs_io statx -r),
> I do see it with the for-next branch of xfsprogs.
> 
> I replaced xfsprogs with the for-next version and generic/532 runs
> (and tested on smb3 mounts with cifs.ko and the test worked fine)

Oh good!

--D

> On Thu, Apr 25, 2019 at 4:10 PM Darrick J. Wong <darrick.wong@oracle.com> wrote:
> >
> > On Thu, Apr 25, 2019 at 01:39:40PM -0700, Steve French wrote:
> > > I noticed that the patch "generic: test statx attribute_mask setting"
> > > seems to cause a failure on xfstest 532.  I am trying with current
> > > xfsprogs and xfstests from git.kernel.org git trees for those.  The
> > > output is:
> > >
> > > (generic/532    [not run] xfs_io statx command does not support attributes_mask)
> > >
> > > due to this line:
> > >
> > > 0ff18d367 (Darrick J. Wong 2019-03-04 13:19:35 -0800 35) get_attributes_mask() {
> > > 0ff18d367 (Darrick J. Wong 2019-03-04 13:19:35 -0800 36)
> > > $XFS_IO_PROG -r -c "statx -r" $1 | grep 'stat.attributes_mask =' | cut
> > > -d ' ' -f 3
> > >
> > > But attributes_mask is not returned by current xfsprogs (instead it
> > > appears to be "mask"
> >
> > That's stx_mask, which are the STATX_* flags.
> >
> > > not "attributes_mask").   If you grep xfsprogs for "attributes_mask" -
> > > there are no hits.
> > >
> > > Is this a typo in line 36 of tests/generic/532 where it is searching
> > > for "stat.attributes_mask" not "stat.mask"?
> >
> > Nope, stx_attributes_mask is a "new" field that shows which
> > stx_attributes flags (STATX_ATTR_*) are known to the filesystem.  It's
> > not really new (it showed up in 4.11 along with the rest of statx) but
> > I only seemed to notice its existence recently.  Support for this new
> > field will appear in xfsprogs 5.0.
> >
> > --D
> >
> > > --
> > > Thanks,
> > >
> > > Steve
> 
> 
> 
> -- 
> Thanks,
> 
> Steve