Re: [PATCH] fstests: _fail the tests if _scratch_mount failed to avoid fully filling root fs

[Eryu Guan <eguan@redhat.com>]

On Wed, Mar 23, 2016 at 11:23:29AM +0800, Eryu Guan wrote:
> On Wed, Mar 23, 2016 at 11:08:56AM +1100, Dave Chinner wrote:
> > On Mon, Mar 21, 2016 at 03:23:41PM +0800, Eryu Guan wrote:
> > > btrfs failed to mount small fs on ppc64 host with error ENOSPC, even
> > > creating such small fs succeeded, then generic/027 consumed all free
> > > space on root fs not on SCRATCH_DEV and test harness cannot create tmp
> > > files and continue other tests.
> > > 
> > > Though I think it's a btrfs bug, it's still worth preventing this
> > > situation from happening in the harness, as such tests usually aim to
> > > exercise fs on ENOSPC conditions, there's no point to continue if the
> > > small fs is not mounted.
> > 
> > I think the btrfs bug should be fixed. At minimum, the workaround to
> > see if the filesytem can be mounted should be in btrfs's
> > implementation of scratch_mkfs_sized....
> 
> OK, I'll add workaround in _scratch_mkfs_sized. Thanks for reviewing!

I think about it another time, I agree that what really should be fixed
is the btrfs bug, not fstests. I'll drop this patch and report bug to
btrfs list.

Thanks,
Eryu