Re: [PATCH] xfs: Regression test for vulnerable directory integrity check

["Darrick J. Wong" <darrick.wong@oracle.com>]

On Thu, May 24, 2018 at 05:32:12PM +0800, Xiao Yang wrote:
> If a malicious XFS contains a block+ format directory wherein the
> directory inode's core.mode is corrupted, and there are subdirectories
> of the corrupted directory, an attempt to traverse up the directory
> tree by running xfs_scrub will crash the kernel in __xfs_dir3_data_check.
> 
> Signed-off-by: Xiao Yang <yangx.jy@cn.fujitsu.com>
> ---
>  tests/xfs/448     | 90 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
>  tests/xfs/448.out |  2 ++
>  tests/xfs/group   |  1 +
>  3 files changed, 93 insertions(+)
>  create mode 100755 tests/xfs/448
>  create mode 100644 tests/xfs/448.out
> 
> diff --git a/tests/xfs/448 b/tests/xfs/448
> new file mode 100755
> index 0000000..bc151a4
> --- /dev/null
> +++ b/tests/xfs/448
> @@ -0,0 +1,90 @@
> +#! /bin/bash
> +# FS QA Test No. 448
> +#
> +# Regression test for commit:
> +# 46c5973 ("xfs: harden directory integrity checks some more")
> +#
> +# If a malicious XFS contains a block+ format directory wherein
> +# the directory inode's core.mode is corrupted, and there are
> +# subdirectories of the corrupted directory, an attempt to traverse
> +# up the directory tree by running xfs_scrub will crash the
> +# kernel in __xfs_dir3_data_check.
> +#
> +# Notice:
> +# we should have non fatal asserts configured, because assert
> +# failures triggered by the intentional corrupt would crash system.
> +#
> +#-----------------------------------------------------------------------
> +# Copyright (c) 2018 FUJITSU.  All Rights Reserved.
> +# Author: Xiao Yang <yangx.jy@cn.fujitsu.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()
> +{
> +	rm -rf $tmp.*
> +}
> +
> +# get standard environment, filters and checks
> +. ./common/rc
> +. ./common/filter
> +. ./common/populate
> +. ./common/fuzzy
> +
> +# real QA test starts here
> +_supported_os Linux
> +_supported_fs xfs
> +_require_scratch
> +_require_scrub
> +_require_scratch_nocheck
> +# Corrupt XFS on purpose, and skip if assert failures would crash system.
> +_require_no_xfs_bug_on_assert
> +
> +rm -f "$seqres.full"
> +
> +# Format and mount
> +_scratch_mkfs > $seqres.full 2>&1 || _fail "mkfs failed"
> +_scratch_mount
> +
> +# Create a block+(e.g. leaf) format directory
> +dblksz="$(xfs_info "${SCRATCH_MNT}" | grep naming.*bsize | sed -e 's/^.*bsize=//g' -e 's/\([0-9]*\).*$/\1/g')"
> +__populate_create_dir "${SCRATCH_MNT}/dir_leaf" "$((dblksz / 12))"
> +dino=$(stat -c "%i" "${SCRATCH_MNT}/dir_leaf")
> +
> +# Corrupt the directory inode's core.mode
> +_scratch_unmount
> +setmode="0100755"
> +_scratch_xfs_set_metadata_field "core.mode" "$setmode" "inode $dino" >> $seqres.full
> +getmode=$(_scratch_xfs_get_metadata_field "core.mode" "inode $dino")
> +[ "$getmode" != "$setmode" ] && _notrun "failed to set core.mode"

When does the set fail?  And isn't that a _fail()ure?

> +
> +# Check a mounted XFS (online)
> +_scratch_mount
> +$XFS_SCRUB_PROG -d -T -v -n $SCRATCH_MNT >> $seqres.full 2>&1

I don't think you want to rely on xfs_scrub at this point if you can
avoid it--scrub is totally experimental and can be deconfigured from the
kernel.  Can you poke the directory using regular commands (like ls) to
trigger the assert?

--D

> +echo "Silence is golden"
> +
> +# success, all done
> +status=0
> +exit
> diff --git a/tests/xfs/448.out b/tests/xfs/448.out
> new file mode 100644
> index 0000000..b6f0a53
> --- /dev/null
> +++ b/tests/xfs/448.out
> @@ -0,0 +1,2 @@
> +QA output created by 448
> +Silence is golden
> diff --git a/tests/xfs/group b/tests/xfs/group
> index 51326d9..dd39d08 100644
> --- a/tests/xfs/group
> +++ b/tests/xfs/group
> @@ -445,3 +445,4 @@
>  445 auto quick filestreams
>  446 auto quick
>  447 auto mount
> +448 auto quick fuzzers
> -- 
> 1.8.3.1
> 
> 
>