xfs: test for post umount readahead completion panic

XFS has a bug where directory readahead completions can occur after
unmount. This can lead to a crash or panic because metadata read
verification attempts to access core XFS data structures (e.g., the
log) after they have been freed and certain pointers have been
reset.

Add a test that triggers directory readahead, delays the readahead
I/O and immediately unmounts the filesystem. This test is part of
the dangerous group as it will cause kernels affected by the bug to
crash.

[eguan replaced touch with echo to speedup file creation]

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>

tests/xfs/311

@@ -0,0 +1,91 @@
+#! /bin/bash
+# FS QA Test No. 311
+#
+# Test to reproduce an XFS unmount crash due to races with directory readahead.
+# XFS had a bug in which unmount would proceed with a readahead I/O in flight.
+# If the unmount deconstructed the log by the time I/O completion occurs,
+# certain metadata read verifier checks could access invalid memory and cause a
+# panic.
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2016 Red Hat, Inc.  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.*
+	_scratch_unmount > /dev/null 2>&1
+	_cleanup_delay > /dev/null 2>&1
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/dmdelay
+
+# Modify as appropriate.
+_supported_fs xfs
+_supported_os Linux
+
+_require_scratch
+_require_dm_target delay
+
+rm -f $seqres.full
+
+echo "Silence is golden."
+
+_scratch_mkfs_xfs >> $seqres.full 2>&1
+
+_init_delay
+_mount_delay
+
+# insert entries to grow the directory to at least one extent, which is what
+# triggers readahead on dir open
+mkdir $SCRATCH_MNT/dir
+for i in $(seq 0 999); do
+	echo > $SCRATCH_MNT/dir/$i
+done
+
+# remount to clear the buffer cache
+_unmount_delay
+_mount_delay
+
+# introduce a read I/O delay
+_load_delay_table $DELAY_READ
+
+# Map the directory and immediately unmount. This should invoke an asynchronous
+# readahead on the first block of the directory. The readahead is delayed by
+# dm-delay. If the unmount doesn't properly wait for the readahead completion,
+# the read verifier can run after core data structures have been freed and lead
+# to a crash/panic.
+$XFS_IO_PROG -c "bmap -v" $SCRATCH_MNT/dir >> $seqres.full 2>&1
+_unmount_delay
+
+_cleanup_delay
+
+# success, all done
+status=0
+exit

tests/xfs/311.out

@@ -0,0 +1,2 @@
+QA output created by 311
+Silence is golden.

tests/xfs/group

@@ -290,3 +290,4 @@
 308 auto quick clone
 309 auto clone
 310 auto clone rmap
+311 auto dangerous quick