linux-ext4.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH v3 1/2] generic: test data corruption on ext4 caused by written/delayed extent
@ 2015-04-27 14:33 Lukas Czerner
  2015-04-27 14:33 ` [PATCH v3 2/2] generic: test bug when zero range is crossing isize within single block Lukas Czerner
  0 siblings, 1 reply; 2+ messages in thread
From: Lukas Czerner @ 2015-04-27 14:33 UTC (permalink / raw)
  To: fstests; +Cc: linux-ext4, Lukas Czerner

This test exercises the problem with unwritten and delayed extents
in ext4 extent status tree where we might in some cases lose a block
worth of data. Even though this was a ext4 specific problem the
reproducer can be easily tun on any file system so let's do that just
in case.

This tests exercises the problem fixed in kernel with commit
"ext4: Fix data corruption caused by unwritten and delayed extents"

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
---
 tests/generic/082     | 85 +++++++++++++++++++++++++++++++++++++++++++++++++++
 tests/generic/082.out | 14 +++++++++
 tests/generic/group   |  1 +
 3 files changed, 100 insertions(+)
 create mode 100755 tests/generic/082
 create mode 100644 tests/generic/082.out

diff --git a/tests/generic/082 b/tests/generic/082
new file mode 100755
index 0000000..6d3d5df
--- /dev/null
+++ b/tests/generic/082
@@ -0,0 +1,85 @@
+#! /bin/bash
+# FS QA Test 082
+#
+# This test exercises the problem with unwritten and delayed extents
+# in ext4 extent status tree where we might in some cases lose a block
+# worth of data. Even though this was a ext4 specific problem the
+# reproducer can be easily run on any file system so let's do that just
+# in case.
+#
+# This test exercises the problem fixed in kernel with commit
+# "ext4: Fix data corruption caused by unwritten and delayed extents"
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2015 Red Hat, Inc., Lukas Czerner <lczerner@redhat.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"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_supported_os IRIX Linux
+_require_test
+_require_xfs_io_command "falloc"
+
+test_file=${TEST_DIR}/testfile-$seq
+
+rm -f $test_file
+
+# The first write creates a delayed extent, fallocate creates
+# unwritten extent which will be marked as delayed in ext4
+# extent status tree. Second write will convert unwritten/delayed
+# block into written/delayed.
+$XFS_IO_PROG -f -c "pwrite -S 0xaa 4096 2048" \
+		-c "falloc 0 131072" \
+		-c "pwrite -S 0xbb 65536 2048" \
+		-c "fsync"  $test_file > $seqres.full 2>&1
+
+# Drop the caches to evict dirty buffers from memory
+echo 3 > /proc/sys/vm/drop_caches
+
+# Write into the second part of the block with 0xbb write from before
+# will create new empty! buffer because the block is still marked as
+# delayed even though it's already written - resulting in
+# overwriting previous data.
+$XFS_IO_PROG -c "pwrite -S 0xdd 67584 2048" $test_file >> $seqres.full 2>&1
+
+# On a faulty ext4 oxbb data will be missing, overwritten by zeroes.
+hexdump -C $test_file
+
+# success, all done
+status=0
+exit
diff --git a/tests/generic/082.out b/tests/generic/082.out
new file mode 100644
index 0000000..ef28b9a
--- /dev/null
+++ b/tests/generic/082.out
@@ -0,0 +1,14 @@
+QA output created by 082
+00000000  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
+*
+00001000  aa aa aa aa aa aa aa aa  aa aa aa aa aa aa aa aa  |................|
+*
+00001800  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
+*
+00010000  bb bb bb bb bb bb bb bb  bb bb bb bb bb bb bb bb  |................|
+*
+00010800  dd dd dd dd dd dd dd dd  dd dd dd dd dd dd dd dd  |................|
+*
+00011000  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
+*
+00020000
diff --git a/tests/generic/group b/tests/generic/group
index 63b883c..c6ea3d8 100644
--- a/tests/generic/group
+++ b/tests/generic/group
@@ -84,6 +84,7 @@
 079 acl attr ioctl metadata auto quick
 080 auto quick
 081 auto quick
+082 auto prealloc preallocrw quick
 083 rw auto enospc stress
 088 perms auto quick
 089 metadata auto
-- 
1.8.3.1

^ permalink raw reply related	[flat|nested] 2+ messages in thread

* [PATCH v3 2/2] generic: test bug when zero range is crossing isize within single block
  2015-04-27 14:33 [PATCH v3 1/2] generic: test data corruption on ext4 caused by written/delayed extent Lukas Czerner
@ 2015-04-27 14:33 ` Lukas Czerner
  0 siblings, 0 replies; 2+ messages in thread
From: Lukas Czerner @ 2015-04-27 14:33 UTC (permalink / raw)
  To: fstests; +Cc: linux-ext4, Lukas Czerner

Exercise the situation that cause ext4 to BUG_ON() when we use
zero range to zero a range which starts within the isize but ends
past the isize but still in the same block. This particular problem
has only been seen on systems with page_size > block_size.

This tests exercises the problem fixed in kernel with commit
0f2af21aae11972fa924374ddcf52e88347cf5a8
ext4: Allocate entire range in zero range

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
---
 tests/generic/084     | 60 +++++++++++++++++++++++++++++++++++++++++++++++++++
 tests/generic/084.out |  1 +
 tests/generic/group   |  1 +
 3 files changed, 62 insertions(+)
 create mode 100755 tests/generic/084
 create mode 100644 tests/generic/084.out

diff --git a/tests/generic/084 b/tests/generic/084
new file mode 100755
index 0000000..5066b14
--- /dev/null
+++ b/tests/generic/084
@@ -0,0 +1,60 @@
+#! /bin/bash
+# FS QA Test 084
+#
+# Exercise the situation that cause ext4 to BUG_ON() when we use
+# zero range to zero a range which starts within the isize but ends
+# past the isize but still in the same block. This particular problem
+# has only been seen on systems with page_size > block_size.
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2015 Red Hat, Inc., Lukas Czerner <lczerner@redhat.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"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_supported_os IRIX Linux
+_require_test
+_require_xfs_io_command "fzero"
+
+testfile=$TEST_DIR/$seq.$$
+
+$XFS_IO_PROG -f -c "pwrite 4096 512" -c "fzero 4351 512" $testfile > $seqres.full 2>&1
+
+# success, all done
+status=0
+exit
diff --git a/tests/generic/084.out b/tests/generic/084.out
new file mode 100644
index 0000000..d0aa380
--- /dev/null
+++ b/tests/generic/084.out
@@ -0,0 +1 @@
+QA output created by 084
diff --git a/tests/generic/group b/tests/generic/group
index c6ea3d8..6cb7dc3 100644
--- a/tests/generic/group
+++ b/tests/generic/group
@@ -86,6 +86,7 @@
 081 auto quick
 082 auto prealloc preallocrw quick
 083 rw auto enospc stress
+084 auto prealloc quick
 088 perms auto quick
 089 metadata auto
 091 rw auto quick
-- 
1.8.3.1

^ permalink raw reply related	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2015-04-27 14:33 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2015-04-27 14:33 [PATCH v3 1/2] generic: test data corruption on ext4 caused by written/delayed extent Lukas Czerner
2015-04-27 14:33 ` [PATCH v3 2/2] generic: test bug when zero range is crossing isize within single block Lukas Czerner

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).