Re: [PATCH] xfs: remove "no-allocation" reservations for file creations

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

On Thu, Dec 07, 2017 at 01:02:59AM +0100, Christoph Hellwig wrote:
> On Tue, Dec 05, 2017 at 10:59:36AM -0800, Darrick J. Wong wrote:
> > I fixed most of the auto group regressions with small fixes to xfstests,
> > but xfs/013 is a problem because once we run out of space to create new
> > dirX we simply get nothing.  Will have to study the behavior of that
> > test in more depth...
> 
> With what kind of config do you see xfs/013 failures?

MKFS_OPTIONS='-m rmapbt=1,reflink=1 -i sparse=1'

I /think/ the fix is that the cp -Rl can sometimes fail to create
dir$((i+1)) because we happen to hit exactly the point where the cleaner
hasn't caught up with fsstress and so the dir$((i+1)) creation
transaction fails with ENOSPC.  AFAICT if I change that to:

while ! test -d $SCRATCH_MNT/dir$((i+1)); do
	cp -Rl ...
done

then xfs/013 exercises the finobt just fine and doesn't fail.

--D

> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html