[PATCH] xfs: log record crc mismatch test

[PATCH] xfs: log record crc mismatch test

[Alexander Tsvetkov]

Added new test case for simulation of V5 filesystem
log crc record corruption using of external log.

Signed-off-by: Alexander Tsvetkov <alexander.tsvetkov@oracle.com>
---
  tests/xfs/005     | 28 ++++++++++++++++++++++++++--
  tests/xfs/005.out |  3 +++
  2 files changed, 29 insertions(+), 2 deletions(-)

diff --git a/tests/xfs/005 b/tests/xfs/005
index fade4bb..ba0dcbb 100755
--- a/tests/xfs/005
+++ b/tests/xfs/005
@@ -1,7 +1,7 @@
  #! /bin/bash
  # FS QA Test No. 005
  #
-# Test that a bad crc on a primary V5 superblock will fail the mount
+# Test that a bad crc on a V5 filesystem will fail the mount
  #
  # 10e6e65 xfs: be more forgiving of a v4 secondary sb w/ junk in v5 fields
  # inadvertently caused primary SB CRC failures to not error out, this
@@ -55,10 +55,12 @@ _supported_os Linux

  _require_scratch_nocheck
  _require_xfs_mkfs_crc
+_require_logdev
+logsize=2560

  rm -f $seqres.full

-_scratch_mkfs_xfs -m crc=1 >> $seqres.full 2>&1 || _fail "mkfs failed"
+_scratch_mkfs_xfs -m crc=1 -lsize=${logsize}b >> $seqres.full 2>&1 || 
_fail "mkfs failed"

  # Zap the crc.  xfs_db updates the CRC post-write, so poke it directly
  $XFS_IO_PROG -c "pwrite 224 4" -c fsync $SCRATCH_DEV | _filter_xfs_io
@@ -66,6 +68,28 @@ $XFS_IO_PROG -c "pwrite 224 4" -c fsync $SCRATCH_DEV 
| _filter_xfs_io
  # should FAIL, the crc is bad; golden output contains mount failure
  _scratch_mount 2>&1 | filter_mount

+_scratch_unmount &> /dev/null
+_scratch_mkfs_xfs -m crc=1 -lsize=${logsize}b  >> $seqres.full 2>&1 || 
_fail "mkfs failed"
+_scratch_mount >> $seqres.full 2>&1 || _fail "mount failed"
+
+# Run filesytem exercising for some time
+_param="-f link=10 -f creat=10 -f mkdir=10 -f truncate=5 -f symlink=10"
+$FSSTRESS_PROG $_param -s 1 $FSSTRESS_AVOID -n 1000 -d $SCRATCH_MNT &> 
/dev/null
+sleep 3
+
+# Shutdown filesystem and unmount
+src/godown -v $SCRATCH_MNT >> $seqres.full
+killall -q fsstress
+wait
+sleep 3
+_scratch_unmount
+
+# Zap log record crc.
+$XFS_IO_PROG -c "pwrite 5000 5" $SCRATCH_LOGDEV | _filter_xfs_io
+
+# should FAIL, log record crc is bad; golden output contains mount failure
+_scratch_mount 2>&1 | filter_mount
+
  # success, all done
  status=0
  exit
diff --git a/tests/xfs/005.out b/tests/xfs/005.out
index 13d0125..8907301 100644
--- a/tests/xfs/005.out
+++ b/tests/xfs/005.out
@@ -2,3 +2,6 @@ QA output created by 005
  wrote 4/4 bytes at offset 224
  XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
  mount: Structure needs cleaning
+wrote 5/5 bytes at offset 5000
+XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+mount: Structure needs cleaning
-- 
1.9.3

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH] xfs: log record crc mismatch test

[Eric Sandeen]

On 4/27/15 10:25 AM, Alexander Tsvetkov wrote:
> Added new test case for simulation of V5 filesystem
> log crc record corruption using of external log.
> 
> Signed-off-by: Alexander Tsvetkov <alexander.tsvetkov@oracle.com>
> ---
>  tests/xfs/005     | 28 ++++++++++++++++++++++++++--
>  tests/xfs/005.out |  3 +++
>  2 files changed, 29 insertions(+), 2 deletions(-)
> 
> diff --git a/tests/xfs/005 b/tests/xfs/005
> index fade4bb..ba0dcbb 100755
> --- a/tests/xfs/005
> +++ b/tests/xfs/005
> @@ -1,7 +1,7 @@
>  #! /bin/bash
>  # FS QA Test No. 005
>  #
> -# Test that a bad crc on a primary V5 superblock will fail the mount
> +# Test that a bad crc on a V5 filesystem will fail the mount

bad crc on what?

>  #
>  # 10e6e65 xfs: be more forgiving of a v4 secondary sb w/ junk in v5 fields
>  # inadvertently caused primary SB CRC failures to not error out, this
> @@ -55,10 +55,12 @@ _supported_os Linux
> 
>  _require_scratch_nocheck
>  _require_xfs_mkfs_crc
> +_require_logdev
> +logsize=2560

No, this means that xfs/05 will no longer work without an external logdev,
right?  So you've stopped testing the internal case, and we've lost that
coverage.

And, we generally don't want to add new testcases to old tests because then
it gets harder to investigate regressions.  It'd probably be best to
create a new test which specifically tests log crc corruption, and make
it handle both internal and external logs automatically.

Thanks,
-Eric

>  rm -f $seqres.full
> 
> -_scratch_mkfs_xfs -m crc=1 >> $seqres.full 2>&1 || _fail "mkfs failed"
> +_scratch_mkfs_xfs -m crc=1 -lsize=${logsize}b >> $seqres.full 2>&1 || _fail "mkfs failed"
> 
>  # Zap the crc.  xfs_db updates the CRC post-write, so poke it directly
>  $XFS_IO_PROG -c "pwrite 224 4" -c fsync $SCRATCH_DEV | _filter_xfs_io
> @@ -66,6 +68,28 @@ $XFS_IO_PROG -c "pwrite 224 4" -c fsync $SCRATCH_DEV | _filter_xfs_io
>  # should FAIL, the crc is bad; golden output contains mount failure
>  _scratch_mount 2>&1 | filter_mount
> 
> +_scratch_unmount &> /dev/null
> +_scratch_mkfs_xfs -m crc=1 -lsize=${logsize}b  >> $seqres.full 2>&1 || _fail "mkfs failed"
> +_scratch_mount >> $seqres.full 2>&1 || _fail "mount failed"
> +
> +# Run filesytem exercising for some time
> +_param="-f link=10 -f creat=10 -f mkdir=10 -f truncate=5 -f symlink=10"
> +$FSSTRESS_PROG $_param -s 1 $FSSTRESS_AVOID -n 1000 -d $SCRATCH_MNT &> /dev/null
> +sleep 3
> +
> +# Shutdown filesystem and unmount
> +src/godown -v $SCRATCH_MNT >> $seqres.full
> +killall -q fsstress
> +wait
> +sleep 3
> +_scratch_unmount
> +
> +# Zap log record crc.
> +$XFS_IO_PROG -c "pwrite 5000 5" $SCRATCH_LOGDEV | _filter_xfs_io
> +
> +# should FAIL, log record crc is bad; golden output contains mount failure
> +_scratch_mount 2>&1 | filter_mount
> +
>  # success, all done
>  status=0
>  exit
> diff --git a/tests/xfs/005.out b/tests/xfs/005.out
> index 13d0125..8907301 100644
> --- a/tests/xfs/005.out
> +++ b/tests/xfs/005.out
> @@ -2,3 +2,6 @@ QA output created by 005
>  wrote 4/4 bytes at offset 224
>  XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
>  mount: Structure needs cleaning
> +wrote 5/5 bytes at offset 5000
> +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> +mount: Structure needs cleaning

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH v2] xfs: log record crc mismatch test

[Alexander Tsvetkov]

Hi Eric,

On 04/27/2015 07:43 PM, Eric Sandeen wrote:
> On 4/27/15 10:25 AM, Alexander Tsvetkov wrote:
>> Added new test case for simulation of V5 filesystem
>> log crc record corruption using of external log.
>>
>> Signed-off-by: Alexander Tsvetkov <alexander.tsvetkov@oracle.com>
>> ---
>>   tests/xfs/005     | 28 ++++++++++++++++++++++++++--
>>   tests/xfs/005.out |  3 +++
>>   2 files changed, 29 insertions(+), 2 deletions(-)
>>
>> diff --git a/tests/xfs/005 b/tests/xfs/005
>> index fade4bb..ba0dcbb 100755
>> --- a/tests/xfs/005
>> +++ b/tests/xfs/005
>> @@ -1,7 +1,7 @@
>>   #! /bin/bash
>>   # FS QA Test No. 005
>>   #
>> -# Test that a bad crc on a primary V5 superblock will fail the mount
>> +# Test that a bad crc on a V5 filesystem will fail the mount
> bad crc on what?
>
>>   #
>>   # 10e6e65 xfs: be more forgiving of a v4 secondary sb w/ junk in v5 fields
>>   # inadvertently caused primary SB CRC failures to not error out, this
>> @@ -55,10 +55,12 @@ _supported_os Linux
>>
>>   _require_scratch_nocheck
>>   _require_xfs_mkfs_crc
>> +_require_logdev
>> +logsize=2560
> No, this means that xfs/05 will no longer work without an external logdev,
> right?  So you've stopped testing the internal case, and we've lost that
> coverage.
>
> And, we generally don't want to add new testcases to old tests because then
> it gets harder to investigate regressions.  It'd probably be best to
> create a new test which specifically tests log crc corruption, and make
> it handle both internal and external logs automatically.
I've prepared the test with both cases, see at the bottom of this mail


> Thanks,
> -Eric
>
>>   rm -f $seqres.full
>>
>> -_scratch_mkfs_xfs -m crc=1 >> $seqres.full 2>&1 || _fail "mkfs failed"
>> +_scratch_mkfs_xfs -m crc=1 -lsize=${logsize}b >> $seqres.full 2>&1 || _fail "mkfs failed"
>>
>>   # Zap the crc.  xfs_db updates the CRC post-write, so poke it directly
>>   $XFS_IO_PROG -c "pwrite 224 4" -c fsync $SCRATCH_DEV | _filter_xfs_io
>> @@ -66,6 +68,28 @@ $XFS_IO_PROG -c "pwrite 224 4" -c fsync $SCRATCH_DEV | _filter_xfs_io
>>   # should FAIL, the crc is bad; golden output contains mount failure
>>   _scratch_mount 2>&1 | filter_mount
>>
>> +_scratch_unmount &> /dev/null
>> +_scratch_mkfs_xfs -m crc=1 -lsize=${logsize}b  >> $seqres.full 2>&1 || _fail "mkfs failed"
>> +_scratch_mount >> $seqres.full 2>&1 || _fail "mount failed"
>> +
>> +# Run filesytem exercising for some time
>> +_param="-f link=10 -f creat=10 -f mkdir=10 -f truncate=5 -f symlink=10"
>> +$FSSTRESS_PROG $_param -s 1 $FSSTRESS_AVOID -n 1000 -d $SCRATCH_MNT &> /dev/null
>> +sleep 3
>> +
>> +# Shutdown filesystem and unmount
>> +src/godown -v $SCRATCH_MNT >> $seqres.full
>> +killall -q fsstress
>> +wait
>> +sleep 3
>> +_scratch_unmount
>> +
>> +# Zap log record crc.
>> +$XFS_IO_PROG -c "pwrite 5000 5" $SCRATCH_LOGDEV | _filter_xfs_io
>> +
>> +# should FAIL, log record crc is bad; golden output contains mount failure
>> +_scratch_mount 2>&1 | filter_mount
>> +
>>   # success, all done
>>   status=0
>>   exit
>> diff --git a/tests/xfs/005.out b/tests/xfs/005.out
>> index 13d0125..8907301 100644
>> --- a/tests/xfs/005.out
>> +++ b/tests/xfs/005.out
>> @@ -2,3 +2,6 @@ QA output created by 005
>>   wrote 4/4 bytes at offset 224
>>   XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
>>   mount: Structure needs cleaning
>> +wrote 5/5 bytes at offset 5000
>> +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
>> +mount: Structure needs cleaning
Added new test for simulation of V5 filesystem
log crc record corruption in internal and external logs

Signed-off-by: Alexander Tsvetkov <alexander.tsvetkov@oracle.com>
---
  tests/xfs/306     | 110 
++++++++++++++++++++++++++++++++++++++++++++++++++++++
  tests/xfs/306.out |   7 ++++
  tests/xfs/group   |   1 +
  3 files changed, 118 insertions(+)
  create mode 100755 tests/xfs/306
  create mode 100644 tests/xfs/306.out

diff --git a/tests/xfs/306 b/tests/xfs/306
new file mode 100755
index 0000000..6e025fe
--- /dev/null
+++ b/tests/xfs/306
@@ -0,0 +1,110 @@
+#! /bin/bash
+# FS QA Test No. 306
+#
+# Test that bad crc on a V5 log record will fail the mount
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2015 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 -f $tmp.*
+}
+
+filter_mount()
+{
+    sed -e "s/mount .* failed: //"
+}
+
+filter_io()
+{
+    sed -e "s/at offset .*/at offset XXX/"
+}
+
+# $1 - log start offset
+# $2 - log device
+do_test()
+{
+    off=$1
+    dev=$2
+
+    _scratch_unmount >/dev/null 2>&1
+    _scratch_mount >> $seqres.full 2>&1 || _fail "mount failed"
+
+    # Run filesytem exercising for some time
+    param="-f link=10 -f creat=10 -f mkdir=10 -f truncate=5 -f symlink=10"
+    $FSSTRESS_PROG $param -s 1 $FSSTRESS_AVOID -n 1000 -d $SCRATCH_MNT 
&> /dev/null
+    sleep 3
+
+    # Shutdown filesystem and unmount
+    src/godown -v $SCRATCH_MNT >> $seqres.full
+    killall -q fsstress
+    wait
+    sleep 3
+    _scratch_unmount
+
+    # Zap log record crc
+    off=$((off + 5000))
+    $XFS_IO_PROG -c "pwrite $off 5" $dev | _filter_xfs_io | filter_io
+
+    # should FAIL, log record crc is bad; golden output contains mount 
failure
+    _scratch_mount 2>&1 | filter_mount
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+
+_supported_fs xfs
+_supported_os Linux
+
+_require_scratch_nocheck
+_require_xfs_mkfs_crc
+_require_logdev
+
+rm -f $seqres.full
+fssize=$((32 * 1024 * 1024))
+MKFS_OPTIONS="-m crc=1 $MKFS_OPTIONS"
+_scratch_mkfs_sized $fssize 2> /dev/null | _filter_mkfs 2> $tmp.mkfs > 
/dev/null
+. $tmp.mkfs
+
+# internal log corruption
+unset USE_EXTERNAL
+_scratch_mkfs_sized $fssize >> $seqres.full 2>&1 || _fail "mkfs failed"
+logstart=$($XFS_DB_PROG -c "sb 0" -c "print" $SCRATCH_DEV | grep 
logstart | cut -f3 -d' ')
+offset=$((logstart * lbsize))
+do_test $offset $SCRATCH_DEV
+
+# external log corruption
+USE_EXTERNAL="yes"
+_scratch_mkfs_sized $fssize >> $seqres.full 2>&1 || _fail "mkfs failed"
+do_test 0 $SCRATCH_LOGDEV
+
+# success, all done
+status=0
+exit
diff --git a/tests/xfs/306.out b/tests/xfs/306.out
new file mode 100644
index 0000000..aef98f2
--- /dev/null
+++ b/tests/xfs/306.out
@@ -0,0 +1,7 @@
+QA output created by 306
+wrote 5/5 bytes at offset XXX
+XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+mount: Structure needs cleaning
+wrote 5/5 bytes at offset XXX
+XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+mount: Structure needs cleaning
diff --git a/tests/xfs/group b/tests/xfs/group
index 58144d2..f108377 100644
--- a/tests/xfs/group
+++ b/tests/xfs/group
@@ -191,3 +191,4 @@
  303 auto quick quota
  304 auto quick quota
  305 auto quota
+306 auto quick
-- 
1.9.3

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH] xfs: log record crc mismatch test

[Dave Chinner]

On Mon, Apr 27, 2015 at 06:25:57PM +0300, Alexander Tsvetkov wrote:
> Added new test case for simulation of V5 filesystem
> log crc record corruption using of external log.
> 
> Signed-off-by: Alexander Tsvetkov <alexander.tsvetkov@oracle.com>

Hi Alexander,

In future, can you please post xfstests patches to
fstests@vger.kernel.org? I've cc'd that list on my reply.

> ---
>  tests/xfs/005     | 28 ++++++++++++++++++++++++++--
>  tests/xfs/005.out |  3 +++
>  2 files changed, 29 insertions(+), 2 deletions(-)
> 
> diff --git a/tests/xfs/005 b/tests/xfs/005
> index fade4bb..ba0dcbb 100755
> --- a/tests/xfs/005
> +++ b/tests/xfs/005
> @@ -1,7 +1,7 @@
>  #! /bin/bash
>  # FS QA Test No. 005
>  #
> -# Test that a bad crc on a primary V5 superblock will fail the mount
> +# Test that a bad crc on a V5 filesystem will fail the mount
>  #
>  # 10e6e65 xfs: be more forgiving of a v4 secondary sb w/ junk in v5 fields
>  # inadvertently caused primary SB CRC failures to not error out, this
> @@ -55,10 +55,12 @@ _supported_os Linux
> 
>  _require_scratch_nocheck
>  _require_xfs_mkfs_crc
> +_require_logdev
> +logsize=2560

As eric pointed out, this fundamentally changes the context of the
test, and so should be written as a new test rather than a change to
an existing test.

Also, you don't need an external log device to do this - you can use
xfs_db to zero parts of the internal log just fine.

$ sudo mkfs.xfs -f -m crc=1,finobt=1 /dev/ram0
meta-data=/dev/ram0              isize=512    agcount=4, agsize=250000 blks
         =                       sectsz=4096  attr=2, projid32bit=1
         =                       crc=1        finobt=1
data     =                       bsize=4096   blocks=1000000, imaxpct=25
         =                       sunit=0      swidth=0 blks
naming   =version 2              bsize=4096   ascii-ci=0 ftype=1
log      =internal log           bsize=4096   blocks=2560, version=2
         =                       sectsz=4096  sunit=1 blks, lazy-count=1
realtime =none                   extsz=4096   blocks=0, rtextents=0
$ sudo xfs_db -x /dev/ram0
xfs_db> sb 0
xfs_db> p logstart
logstart = 524296
xfs_db> a logstart
xfs_db> type data
xfs_db> write fill 0xa5a5a5a5 0 4096
xfs_db> p
000: a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5
020: a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5
.....
fe0: a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5 a5a5a5a5
xfs_db> q
$ sudo mount /dev/ram0 /mnt/test
[346491.179086] XFS (ram0): Mounting V5 Filesystem
[346491.180844] XFS (ram0): Log inconsistent or not a log (last==0, first!=1)
[346491.182946] XFS (ram0): empty log check failed
[346491.184403] XFS (ram0): log mount/recovery failed: error -22
[346491.186276] XFS (ram0): log mount failed
.....

Note that if you have an external log device, logstart will be zero,
and you'll need to use xfs_io to overwrite it so you'll need to
explicitly test for that case before calling xfs_db to modify the
log.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs