On Thu, Oct 16, 2014 at 09:42:20AM +0800, Qu Wenruo wrote:
> The following kernel commit introduced the bug:
> 51f395ad btrfs: Use right extent length when inserting overlap extent map.
>
> When btrfs commit race with btrfs_get_extent(), merge_extent_mapping()
> may build up a new extent which length overflows and cause extent map
> insert fail, causing the caller get a -EEXIST error.
>
> This regression is fixed by the following patches:
> btrfs: Fix and enhance merge_extent_mapping() to insert best fitted extent map
> btrfs: Fix the wrong condition judgment about subset extent map
>
> Cc: Filipe Manana <fdmanana@xxxxxxxx>
> Signed-off-by: Qu Wenruo <quwenruo@xxxxxxxxxxxxxx>
> ---
> tests/btrfs/078 | 128 ++++++++++++++++++++++++++++++++++++++++++++++++++++
> tests/btrfs/078.out | 2 +
> tests/btrfs/group | 1 +
> 3 files changed, 131 insertions(+)
> create mode 100755 tests/btrfs/078
> create mode 100644 tests/btrfs/078.out
>
> diff --git a/tests/btrfs/078 b/tests/btrfs/078
> new file mode 100755
> index 0000000..bdc20af
> --- /dev/null
> +++ b/tests/btrfs/078
> @@ -0,0 +1,128 @@
> +#! /bin/bash
> +# FS QA Test No. btrfs/078
> +#
> +# Do write along with fiemap ioctl.
> +# Regression test for the kernel comit:
> +# 51f395ad btrfs: Use right extent length when inserting overlap extent map.
> +#
> +# When calling fiemap(without SYNC flag) and btrfs fs is commiting,
> +# it will cause race condition and cause btrfs to generate a wrong extent
> +# whose len is overflow and fail to insert into the extent map tree,
> +# returning -EEXIST.
> +#
> +# Fixed by the following patches (not merged in mainline yet):
> +# btrfs: Fix and enhance merge_extent_mapping() to insert best fitted extent map
> +# btrfs: Fix the wrong condition judgment about subset extent map
> +#
> +#-----------------------------------------------------------------------
> +# Copyright (C) 2014 Fujitsu All Rights Reserved.
> +# Author: Qu Wenruo <quwenruo@xxxxxxxxxxxxxx>
> +#
> +# 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=`mktemp -d`
I'm a bit concerned about this. Some functions in common/rc use
$tmp.some_sufix as temp file. If $tmp defined as a dir here, I'm not
sure if it will cause any trouble.
> +status=1 # failure is the default!
> +trap "_cleanup; exit \$status" 0 1 2 3 15
> +
> +_cleanup()
> +{
> + kill $dd_pid &> /dev/null
> + kill $fiemap_pid &> /dev/null
> + wait
> + rm -fr $test_file
> + rm -fr $tmp
> +}
> +
> +# get standard environment, filters and checks
> +. ./common/rc
> +. ./common/filter
> +
> +# real QA test starts here
> +_supported_fs btrfs
> +_supported_os Linux
> +_require_scratch
> +_need_to_be_root
> +# Since xfs_io's fiemap always use SYNC flag and can't be unset,
> +# we must use filefrag to call fiemap without SYNC flag.
> +_require_command "/usr/sbin/filefrag"
> +_require_xfs_io_command "falloc"
> +
> +test_file=$SCRATCH_MNT/testfile
> +filesize=$((10 * 1024 * 1024 * 1024)) #10G size
> +buffersize=$((1024 * 1024)) # 1M bs for dd
> +count=$(($filesize / $buffersize))
> +testfile=$SCRATCH_MNT/testfile
test_file and testfile are both set to $SCRATCH_MNT/testfile and you
use both $test_file and $testfile through the test, seems you can use
only one?
> +
> +rm -f $seqres.full
> +rm -f $test_file
> +
> +_scratch_mkfs >>$seqres.full 2>&1
> +_scratch_mount
> +$XFS_IO_PROG -f -c "falloc 0 $filesize" $testfile || \
> + _fail 'need device larger than 10G'
You can use _require_fs_space() helper here.
Thanks,
Eryu
> +
> +dd_work() {
> + out=$1
> + dd if=/dev/zero of=$out bs=$buffersize count=$count \
> + conv=notrunc &> /dev/null
> +}
> +
> +# There is a bug for e2fsprogs, at least in version 1.42.9, filefrag will
> +# leak the return value, so we can't judge return value only,
> +# but also to filter the output
> +_filter_error() {
> + # when filefrag fails FIEMAP ioctl, it will fall back to FIBMAP,
> + # which is not supported by btrfs and will report "FIBMAP: strerr()"
> + # However due to other btrfs commits, kernel will return -ENOTTY
> + # or -EINVAL, so only grep for "FIBMAP" for max compatibility.
> + grep "FIBMAP"
> +}
> +
> +fiemap_work() {
> + filename=$1
> + while true; do
> + filefrag $filename 2> $tmp/output 1> /dev/null
> + ret=$?
> + err=`cat $tmp/output | _filter_error`
> + if [ $ret -ne 0 -o -n "$err" ]; then
> + kill $dd_pid
> + return 1
> + fi
> + done
> +}
> +
> +dd_work $testfile &
> +dd_pid=$!
> +fiemap_work $testfile &
> +fiemap_pid=$!
> +wait $dd_pid
> +ddret=$?
> +kill $fiemap_pid &> /dev/null
> +wait $fiemap_pid
> +
> +if [ $ddret -ne 0 ]; then
> + echo "Extent merge bug detected"
> + status=1
> + exit
> +else
> + echo "Silence is golden"
> + status=0
> + exit
> +fi
> diff --git a/tests/btrfs/078.out b/tests/btrfs/078.out
> new file mode 100644
> index 0000000..b8acea8
> --- /dev/null
> +++ b/tests/btrfs/078.out
> @@ -0,0 +1,2 @@
> +QA output created by 078
> +Silence is golden
> diff --git a/tests/btrfs/group b/tests/btrfs/group
> index 9adf862..40e7430 100644
> --- a/tests/btrfs/group
> +++ b/tests/btrfs/group
> @@ -80,3 +80,4 @@
> 075 auto quick subvol
> 076 auto quick
> 077 auto quick
> +078 auto
> --
> 2.1.2
>
> --
> To unsubscribe from this list: send the line "unsubscribe fstests" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe fstests" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
