btrfs: exercise transaction abort on device with discard support

Test that btrfs' transaction abortion does not corrupt a filesystem
mounted with -o discard nor allows a subsequent fstrim to corrupt the
filesystem (regardless of being mounted with or without -o discard).

This issue was fixed by the following linux kernel patch:

    Btrfs: fix fs corruption on transaction abort if device supports discard
    (commit 678886bdc6378c1cbd5072da2c5a3035000214e3)

Without the corresponding btrfs fix the fs becomes unmountable and fails
like this:

  $ ./check btrfs/089
  FSTYP         -- btrfs
  PLATFORM      -- Linux/x86_64 debian3 3.19.0-btrfs-next-7+
  MKFS_OPTIONS  -- /dev/sdc
  MOUNT_OPTIONS -- /dev/sdc /home/fdmanana/btrfs-tests/scratch_1

  btrfs/089 2s ... - output mismatch (see /home/fdmanana/git/hub/xfstests/results//btrfs/089.out.bad)
      --- tests/btrfs/089.out	2015-04-03 19:29:42.969594083 +0100
      +++ /home/fdmanana/git/hub/xfstests/results//btrfs/089.out.bad	2015-04-03 19:42:37.419181019 +0100
      @@ -3,7 +3,10 @@
       XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
       wrote 1048576/1048576 bytes at offset 524288
       XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
      +mount: wrong fs type, bad option, bad superblock on /dev/sdc,
      +       missing codepage or helper program, or other error
      +       In some cases useful info is found in syslog - try
      +       dmesg | tail  or so
      ...
      (Run 'diff -u tests/btrfs/089.out /home/fdmanana/git/hub/xfstests/results//btrfs/089.out.bad'  to see the entire diff)
  _check_btrfs_filesystem: filesystem on /dev/sdc is inconsistent (see /home/fdmanana/git/hub/xfstests/results//btrfs/089.full)
  Ran: btrfs/089
  Failures: btrfs/089
  Failed 1 of 1 tests

  $ cat /home/fdmanana/git/hub/xfstests/results//btrfs/089.full
  Performing full device TRIM (100.00GiB) ...
  touch: cannot touch '/home/fdmanana/btrfs-tests/scratch_1/abc': Read-only file system
  _check_btrfs_filesystem: filesystem on /dev/sdc is inconsistent
  *** fsck.btrfs output ***
  Check tree block failed, want=29589504, have=0
  Check tree block failed, want=29589504, have=0
  Check tree block failed, want=29589504, have=0
  Check tree block failed, want=29589504, have=0
  Check tree block failed, want=29589504, have=0
  read block failed check_tree_block
  Couldn't read tree root
  Couldn't open file system
  (...)

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

tests/btrfs/088

@@ -0,0 +1,133 @@
+#! /bin/bash
+# FS QA Test No. btrfs/088
+#
+# Test that btrfs' transaction abortion does not corrupt a filesystem mounted
+# with -o discard nor allows a subsequent fstrim to corrupt the filesystem
+# (regardless of being mounted with or without -o discard).
+#
+# This issue was fixed by the following linux kernel patch:
+#
+#    Btrfs: fix fs corruption on transaction abort if device supports discard
+#    (commit 678886bdc6378c1cbd5072da2c5a3035000214e3)
+#
+#-----------------------------------------------------------------------
+# Copyright (C) 2015 SUSE Linux Products GmbH. All Rights Reserved.
+# Author: Filipe Manana <fdmanana@suse.com>
+#
+# 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"
+
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	rm -f $tmp.*
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+
+# real QA test starts here
+_supported_fs btrfs
+_supported_os Linux
+_require_scratch
+_require_fail_make_request
+_need_to_be_root
+
+SCRATCH_BDEV=`_short_dev $SCRATCH_DEV`
+
+enable_io_failure()
+{
+	echo 100 > $DEBUGFS_MNT/fail_make_request/probability
+	echo 1000 > $DEBUGFS_MNT/fail_make_request/times
+	echo 0 > $DEBUGFS_MNT/fail_make_request/verbose
+	echo 1 > /sys/block/$SCRATCH_BDEV/make-it-fail
+}
+
+disable_io_failure()
+{
+	echo 0 > /sys/block/$SCRATCH_BDEV/make-it-fail
+	echo 0 > $DEBUGFS_MNT/fail_make_request/probability
+	echo 0 > $DEBUGFS_MNT/fail_make_request/times
+}
+
+rm -f $seqres.full
+
+# We will abort a btrfs transaction later, which always produces a warning in
+# dmesg. We do not want the test to fail because of this.
+_disable_dmesg_check
+
+_scratch_mkfs >>$seqres.full 2>&1
+_scratch_mount "-o discard"
+_require_batched_discard $SCRATCH_MNT
+
+# Create a file and call sync to commit our first transaction.
+$XFS_IO_PROG -f -c "pwrite -S 0xaa 0 1M" $SCRATCH_MNT/foo | _filter_xfs_io
+sync
+
+# Create some other file, which forces a COW operation of the fs root, adding
+# the old root location to the pinned extents list, and opens a new btrfs
+# transaction.
+touch $SCRATCH_MNT/bar
+
+# Write to the first file to verify later that the original data extent was not
+# a victim of a discard operation.
+$XFS_IO_PROG -c "pwrite -S 0xbb 512K 1M" $SCRATCH_MNT/foo | _filter_xfs_io
+
+# Now make sure the next transaction commit will abort and turn the fs readonly,
+# unmount the fs, mount it again and verify we can open file foo and read its
+# content, which should be what it had when the first transaction was committed
+# (first call to sync), since btrfs is a COW filesystem and foo was not fsynced.
+# Btrfs used to issue a discard operation on the extents in the pinned extents
+# list, resulting in corruption of metadata and data, and used too to return the
+# pinned extents to the free space caches, allowing future fstrim operations to
+# perform a discard operation against the pinned exents. This made the fs
+# unmountable because the btree roots that the superblock points at were written
+# in place (by the discard operations).
+enable_io_failure
+
+# This sync will trigger a commit of the current transaction, which will be
+# aborted because IO will fail for metadata extents (btree nodes/leafs).
+sync
+disable_io_failure
+
+touch $SCRATCH_MNT/abc >>$seqres.full 2>&1 && \
+	echo "Transaction was not aborted, filesystem is not in readonly mode"
+
+# This fstrim operation should not cause discard operations to be performed
+# against extents that were COWed, otherwise the next mount will fail since
+# the btree roots that the superblock points at have their physical areas
+# on disk full of zeroes.
+$FSTRIM_PROG $SCRATCH_MNT
+
+# We expect to be able to mount the fs again and have available all metadata and
+# data that got persisted in the first transaction.
+_scratch_remount
+
+# We now expect file's foo content to match what it had when the first
+# transaction was committed because the second transaction was aborted and we
+# did not fsync foo.
+echo "File foo content after transaction abort + remount:"
+od -t x1 $SCRATCH_MNT/foo
+
+status=0
+exit

tests/btrfs/088.out

@@ -0,0 +1,9 @@
+QA output created by 088
+wrote 1048576/1048576 bytes at offset 0
+XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+wrote 1048576/1048576 bytes at offset 524288
+XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+File foo content after transaction abort + remount:
+0000000 aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa
+*
+4000000

tests/btrfs/group

@@ -90,3 +90,4 @@
 085 auto quick metadata subvol
 086 auto quick clone
 087 auto quick send
+088 auto quick metadata