Re: [PATCH] btrfs: test ENOSPC caused by many orphan items

[Eryu Guan <guaneryu@gmail.com>]

On Tue, May 15, 2018 at 07:14:02PM -0700, Omar Sandoval wrote:
> On Wed, May 16, 2018 at 09:48:58AM +0800, Eryu Guan wrote:
> > On Wed, May 09, 2018 at 11:21:55PM -0700, Omar Sandoval wrote:
> > > From: Omar Sandoval <osandov@fb.com>
> > > 
> > > Btrfs has a bug where we can prematurely ENOSPC if we have lots of
> > > orphaned files, i.e., deleted files which are still open. Add a test
> > > which repeatedly creates and deletes a file while keeping all of the
> > > file descriptors open. This should succeed but doesn't on Btrfs without
> > > the fix.
> > > 
> > > Signed-off-by: Omar Sandoval <osandov@fb.com>
> > > ---
> > >  tests/generic/479     |  0
> > >  tests/generic/487     | 65 +++++++++++++++++++++++++++++++++++++++++++
> > >  tests/generic/487.out |  2 ++
> > >  tests/generic/group   |  1 +
> > >  4 files changed, 68 insertions(+)
> > >  mode change 100644 => 100755 tests/generic/479
> > >  create mode 100755 tests/generic/487
> > >  create mode 100644 tests/generic/487.out
> > > 
> > > diff --git a/tests/generic/479 b/tests/generic/479
> > > old mode 100644
> > > new mode 100755
> > > diff --git a/tests/generic/487 b/tests/generic/487
> > > new file mode 100755
> > > index 00000000..66379cf0
> > > --- /dev/null
> > > +++ b/tests/generic/487
> > > @@ -0,0 +1,65 @@
> > > +#! /bin/bash
> > > +# FS QA Test 487
> > > +#
> > > +# Test having many file descriptors referring to deleted files open. Regression
> > > +# test for patch "Btrfs: fix ENOSPC caused by orphan items reservations".
> > > +#
> > > +#-----------------------------------------------------------------------
> > > +# Copyright (c) 2018 Omar Sandoval.  All Rights Reserved.
> > > +#
> > > +# This program is free software; you can redistribute it and/or
> > > +# modify it under the terms of the GNU General Public License as
> > > +# published by the Free Software Foundation.
> > > +#
> > > +# This program is distributed in the hope that it would be useful,
> > > +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> > > +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> > > +# GNU General Public License for more details.
> > > +#
> > > +# You should have received a copy of the GNU General Public License
> > > +# along with this program; if not, write the Free Software Foundation,
> > > +# Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
> > > +#-----------------------------------------------------------------------
> > > +#
> > > +
> > > +seq=`basename $0`
> > > +seqres=$RESULT_DIR/$seq
> > > +echo "QA output created by $seq"
> > > +
> > > +here=`pwd`
> > > +tmp=/tmp/$$
> > > +status=1	# failure is the default!
> > > +trap "_cleanup; exit \$status" 0 1 2 3 15
> > > +
> > > +_cleanup()
> > > +{
> > > +	cd /
> > > +	rm -f $tmp.*
> > > +}
> > > +
> > > +. ./common/rc
> > > +. ./common/filter
> > > +
> > > +rm -f $seqres.full
> > > +
> > > +_supported_fs generic
> > > +_supported_os Linux
> > > +_require_scratch
> > > +
> > > +_scratch_mkfs_sized $((1024 * 1024 * 1024)) >> $seqres.full 2>&1
> > > +_scratch_mount
> > > +
> > > +test_file="$SCRATCH_MNT/$seq"
> > > +
> > > +(
> > > +ulimit -n $((16 * 1024))
> > > +# ~10000 files on a 1 GB filesystem should be no problem.
> > > +for ((i = 1000; i < 10000; i++)); do
> > > +	eval "exec $i<> \"$test_file\"" && rm "$test_file" || break
> > > +done
> > 
> > There's a helper command in src that does exactly this job, e.g.
> > 
> > $here/src/multi_open_unlink -f $SCRATCH_MNT/$seq -n 10000 -s 0
> > 
> > which creates & unlinks 10000 files and keeps them open for 0 second in
> > $SCRATCH_MNT using "$seq" as name prefix. This reduces the test run time
> > from 13s to 1s for me.
> > 
> > It's a straightforward change, I'll just update on commit, please let me
> > know if you have different thoughts.
> > 
> > Thanks,
> > Eryu
> 
> Great, as long as it still reproduces the bug without the fix applied,
> that's perfect. Thanks!

Yes, I've verified that it still reproduces the bug on btrfs and test
passes with the fix(es) applied.

Thanks,
Eryu