On Thu, Jun 29, 2017 at 05:36:14PM +0800, Eryu Guan wrote:

On Wed, Jun 21, 2017 at 02:57:48PM -0700, Darrick J. Wong wrote:

quoted

From: Darrick J. Wong <redacted>

Test a regression in XFS where we blow out a transaction reservation if
we create a big file, share every other block, and delete the first
file.  There's nothing particularly fs-specific about this stress test,
so put it in generic.

Signed-off-by: Darrick J. Wong <redacted>

This test took me 3019s to finish with v4.12-rc7 kernel, and another
host "hung" at "Delete file1" (it had been more than 1 hour, and I lost
my patience and hard-reboot the host). Are they expected results?

No.  The take-forever-or-crash behavior should be fixed by "xfs: try to
avoid blowing out the transaction reservation when bunmaping a shared
extent" in 4.13.  Feel free to hang on to this one until -rc1. :)

If the bug is still existed in latest upstream kernel, I tend to merge
it after the fix landing in linus tree. If v4.12-rc7 doesn't suffer from
this bug, the test time should be reduced.

<shrug> This is what I saw just now:

FSTYP         -- xfs (debug)
PLATFORM      -- Linux/x86_64 birch-mtr0 4.12.0-rc6-dgc
MKFS_OPTIONS  -- -f -m reflink=1,rmapbt=1, -i sparse=1, /dev/pmem1
MOUNT_OPTIONS -- /dev/pmem1 /opt

generic/931      21s
Ran: generic/931
Passed all 1 tests

Though if it takes forever for everyone else, please kick this one out
of auto/quick.  At least in theory, before the patch the test will
either blow out a transaction reservation and hang the system, or if it
does succeed it'll have done so by scraping long and hard for log space.
That is probably why it takes 3000+ seconds on your test box, unless
you were also testing xfs for-next.

And another minor nit below.

quoted

---
 tests/generic/931     |   94 +++++++++++++++++++++++++++++++++++++++++++++++++
 tests/generic/931.out |    6 +++
 tests/generic/group   |    1 +
 3 files changed, 101 insertions(+)
 create mode 100755 tests/generic/931
 create mode 100644 tests/generic/931.out

diff --git a/tests/generic/931 b/tests/generic/931
new file mode 100755
index 0000000..afadf81
--- /dev/null
+++ b/tests/generic/931

@@ -0,0 +1,94 @@
+#! /bin/bash
+# FS QA Test No. 931
+#
+# See how well we handle deleting a file with a million refcount extents.
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2017, Oracle and/or its affiliates.  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 -rf "$tmp".* $testdir/file1

'rm -rf' looks a bit scary, and we're only deleting regular files not
directories, 'rm -f' should be sufficient.

Yes.  Will you fix it on the way in or should I resend?

--D

Thanks,
Eryu

quoted

+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+. ./common/attr
+. ./common/reflink
+
+# real QA test starts here
+_supported_os Linux
+_require_scratch_reflink
+_require_cp_reflink
+_require_test_program "punch-alternating"
+
+rm -f "$seqres.full"
+
+echo "Format and mount"
+_scratch_mkfs > "$seqres.full" 2>&1
+_scratch_mount >> "$seqres.full" 2>&1
+
+testdir="$SCRATCH_MNT/test-$seq"
+mkdir "$testdir"
+
+# Setup for one million blocks, but we'll accept stress testing down to
+# 2^17 blocks... that should be plenty for anyone.
+fnr=20
+free_blocks=$(stat -f -c '%a' "$testdir")
+blksz=$(_get_block_size "$testdir")
+space_avail=$((free_blocks * blksz))
+calc_space() {
+	blocks_needed=$(( 2 ** (fnr + 1) ))
+	space_needed=$((blocks_needed * blksz * 5 / 4))
+}
+calc_space
+while test $space_needed -gt $space_avail; do
+	fnr=$((fnr - 1))
+	calc_space
+done
+test $fnr -lt 17 && _notrun "Insufficient space for stress test; would only create $blocks_needed extents ($space_needed/$space_avail blocks)."
+
+echo "Create a many-block file"
+echo "creating $blocks_needed blocks..." >> "$seqres.full"
+$XFS_IO_PROG -f -c "pwrite -S 0x61 -b 4194304 0 $((2 ** (fnr + 1) * blksz))" "$testdir/file1" >> "$seqres.full"
+
+echo "Reflinking file"
+_cp_reflink $testdir/file1 $testdir/file2
+
+echo "Punch file2"
+echo "Punching file2..." >> "$seqres.full"
+"$here/src/punch-alternating" "$testdir/file2" >> "$seqres.full"
+echo "...done" >> "$seqres.full"
+_scratch_cycle_mount
+
+echo "Delete file1"
+rm -rf $testdir/file1
+
+# success, all done
+status=0
+exit

diff --git a/tests/generic/931.out b/tests/generic/931.out
new file mode 100644
index 0000000..c7b724e
--- /dev/null
+++ b/tests/generic/931.out

@@ -0,0 +1,6 @@
+QA output created by 931
+Format and mount
+Create a many-block file
+Reflinking file
+Punch file2
+Delete file1

diff --git a/tests/generic/group b/tests/generic/group
index ab1e9d3..b0d1844 100644
--- a/tests/generic/group
+++ b/tests/generic/group

@@ -443,3 +443,4 @@
 438 auto
 439 auto quick punch
 440 auto quick encrypt
+931 auto quick clone

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