Re: [PATCH] xfstests: create a test for xfs log grant head leak detection

[Brian Foster <bfoster@redhat.com>]

On Wed, Jun 11, 2014 at 09:31:06AM +1000, Dave Chinner wrote:
> On Tue, Jun 10, 2014 at 07:17:50AM -0400, Brian Foster wrote:
> > On Tue, Jun 10, 2014 at 11:21:49AM +1000, Dave Chinner wrote:
> > > On Fri, Jun 06, 2014 at 09:14:43AM -0400, Brian Foster wrote:
> > > > +# real QA test starts here
> > > > +_supported_fs xfs
> > > > +_supported_os Linux
> > > > +
> > > > +_require_scratch
> > > > +_require_freeze
> > > > +
> > > > +if [ ! -e /sys/fs/xfs ]
> > > > +then
> > > > +	_notrun "no kernel support for XFS sysfs attributes"
> > > > +fi
> > > 
> > > _requires_xfs_sysfs
> > > 
> > 
> > I was mulling this over as I think we'll probably end up in a situation
> > where a test that depends on sysfs bits will need to check for a
> > specific attribute file. E.g., some new test comes along using a new
> > attribute file. Checking for /sys/fs/xfs is not sufficient for that test
> > once we release a version that so far only exports the log bits.
> > 
> > I think we could handle that by supporting a parameter to
> > _requires_xfs_sysfs that specifies the sub-attribute that must exist
> > (similar to what we have for xfs_io commands). We don't need that at the
> > moment, but that's good enough for me to create the requires func.
> 
> Yup, passing the name and/or sub-path of the paramter set required
> sounds fine to me. it would become:
> 
> _requires_xfs_sysfs log
> 
> in this case, because the presence of the /sys/fs/xfs/<dev>/log
> directory would be suficinet to indicate the test should run, yes?
> 

Yep, that works. Should any new log attributes come along as a test
dependency, something like the following would be required:

_requires_xfs_sysfs log/logattr

So _requires_xfs_sysfs can basically just test for the existence of the
parameter under the xfs/<dev> path for TEST_DEV. I'll go with that.
Thanks.

Brian

> Cheers,
> 
> Dave.
> -- 
> Dave Chinner
> david@fromorbit.com