| Message ID | 1527154332-13234-1-git-send-email-yangx.jy@cn.fujitsu.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
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 > > > -- To unsubscribe from this list: send the line "unsubscribe fstests" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 2018/05/25 12:37, Darrick J. Wong wrote: > 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" != "$setmode" ]&& _notrun "failed to set core.mode" > When does the set fail? And isn't that a _fail()ure? Hi Darrick, The set never failed on my enviroment, and i just ensure set succeeded. Do you want to remove the check? >> + >> +# Check a mounted XFS (online) >> +_scratch_mount >> +$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? I tried and failed to trigger the assert by some regular commands(ls, stat, etc.). If you have some ideas, could you tell me how to trigger the assert by using regular commands? Thanks, Xiao Yang > --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 >> >> >> > > . > -- To unsubscribe from this list: send the line "unsubscribe fstests" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, May 25, 2018 at 02:33:38PM +0800, Xiao Yang wrote: > On 2018/05/25 12:37, Darrick J. Wong wrote: > > 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" != "$setmode" ]&& _notrun "failed to set core.mode" > > When does the set fail? And isn't that a _fail()ure? > Hi Darrick, > > The set never failed on my enviroment, and i just ensure set succeeded. > Do you want to remove the check? No, it's usually a good idea to make sure that the debugger actually did the thing we told it to. I was simply curious if this occurred on a regular basis. > > > + > > > +# Check a mounted XFS (online) > > > +_scratch_mount > > > +$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? > I tried and failed to trigger the assert by some regular commands(ls, stat, > etc.). > If you have some ideas, could you tell me how to trigger the assert by using > regular commands? Ah, ok. I'd have thought that "find $SCRATCH_MNT -type f -print0 | xargs -0 cat > /dev/null" would've triggered the assert, but I don't have any better suggestions. --D > Thanks, > Xiao Yang > > --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 > > > > > > > > > > > > > . > > > > > > -- > To unsubscribe from this list: send the line "unsubscribe fstests" in > the body of a message to majordomo@vger.kernel.org > 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@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Eryu, Do you have any better ways to trigger the assert? On 2018/05/30 1:53, Darrick J. Wong wrote: > On Fri, May 25, 2018 at 02:33:38PM +0800, Xiao Yang wrote: >> On 2018/05/25 12:37, Darrick J. Wong wrote: >>> 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" != "$setmode" ]&& _notrun "failed to set core.mode" >>> When does the set fail? And isn't that a _fail()ure? >> Hi Darrick, >> >> The set never failed on my enviroment, and i just ensure set succeeded. >> Do you want to remove the check? > No, it's usually a good idea to make sure that the debugger actually did > the thing we told it to. I was simply curious if this occurred on a > regular basis. Hi Darrick, OK, keep it. >>>> + >>>> +# Check a mounted XFS (online) >>>> +_scratch_mount >>>> +$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? >> I tried and failed to trigger the assert by some regular commands(ls, stat, >> etc.). >> If you have some ideas, could you tell me how to trigger the assert by using >> regular commands? > Ah, ok. I'd have thought that "find $SCRATCH_MNT -type f -print0 | xargs > -0 cat> /dev/null" would've triggered the assert, but I don't have any > better suggestions. I tried the find command, but failed to trigger the assert. If nobody has a better way, can we still use xfs_scrub command in the test? Thanks, Xiao Yang > --D > >> Thanks, >> Xiao Yang >>> --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 >>>> >>>> >>>> >>> . >>> >> >> >> -- >> To unsubscribe from this list: send the line "unsubscribe fstests" in >> the body of a message to majordomo@vger.kernel.org >> 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@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, May 30, 2018 at 12:04:42PM +0800, Xiao Yang wrote: > Hi Eryu, > > Do you have any better ways to trigger the assert? > > On 2018/05/30 1:53, Darrick J. Wong wrote: > > On Fri, May 25, 2018 at 02:33:38PM +0800, Xiao Yang wrote: > > > On 2018/05/25 12:37, Darrick J. Wong wrote: > > > > 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" != "$setmode" ]&& _notrun "failed to set core.mode" > > > > When does the set fail? And isn't that a _fail()ure? > > > Hi Darrick, > > > > > > The set never failed on my enviroment, and i just ensure set succeeded. > > > Do you want to remove the check? > > No, it's usually a good idea to make sure that the debugger actually did > > the thing we told it to. I was simply curious if this occurred on a > > regular basis. > Hi Darrick, > > OK, keep it. > > > > > + > > > > > +# Check a mounted XFS (online) > > > > > +_scratch_mount > > > > > +$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? > > > I tried and failed to trigger the assert by some regular commands(ls, stat, > > > etc.). > > > If you have some ideas, could you tell me how to trigger the assert by using > > > regular commands? > > Ah, ok. I'd have thought that "find $SCRATCH_MNT -type f -print0 | xargs > > -0 cat> /dev/null" would've triggered the assert, but I don't have any > > better suggestions. > I tried the find command, but failed to trigger the assert. > If nobody has a better way, can we still use xfs_scrub command in the test? Sure. To be fair, scrub had a silly bug in it that made this easy to trip over + test for. :) Though you might consider the simpler: $XFS_IO_PROG -c 'scrub parent' /path/to/the/child/file since we're not going to change xfs_io unless we absolutely have to. --D > Thanks, > Xiao Yang > > --D > > > > > Thanks, > > > Xiao Yang > > > > --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 > > > > > > > > > > > > > > > > > > > . > > > > > > > > > > > > > -- > > > To unsubscribe from this list: send the line "unsubscribe fstests" in > > > the body of a message to majordomo@vger.kernel.org > > > 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@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, May 30, 2018 at 12:04:42PM +0800, Xiao Yang wrote: > Hi Eryu, > > Do you have any better ways to trigger the assert? Sorry, I've no better ideas here.. > > On 2018/05/30 1:53, Darrick J. Wong wrote: > > On Fri, May 25, 2018 at 02:33:38PM +0800, Xiao Yang wrote: > > > On 2018/05/25 12:37, Darrick J. Wong wrote: > > > > 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") It'd be better to reference the commit with 12-digits commit id. > > > > > +# > > > > > +# 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> Author not needed. > > > > > +# > > > > > +# 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 $tmp.* Please follow the 'new' template in _cleanup(), and add any cleanup steps specific to the test. > > > > > +} > > > > > + > > > > > +# get standard environment, filters and checks > > > > > +. ./common/rc > > > > > +. ./common/filter > > > > > +. ./common/populate > > > > > +. ./common/fuzzy > > > > > + > > > > > +# real QA test starts here > > > > > +_supported_os Linux > > > > > +_supported_fs xfs > > > > > +_require_scratch _require_scratch_nocheck below is sufficient. > > > > > +_require_scrub > > > > > +_require_scratch_nocheck > > > > > +# Corrupt XFS on purpose, and skip if assert failures would crash system. > > > > > +_require_no_xfs_bug_on_assert > > > > > + > > > > > +rm -f "$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')" I think we can update _filter_mkfs to print out bsize for naming section too and get the size by sourcing filter output, e.g. _scratch_mkfs > $seqres.full 2> $tmp.mkfs || _fail "mkfs failed" ... # get directory block size . $tmp.mkfs ... > > > > > +__populate_create_dir "${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" != "$setmode" ]&& _notrun "failed to set core.mode" > > > > When does the set fail? And isn't that a _fail()ure? Yeah, looks like a _fail to me. Thanks, Eryu > > > Hi Darrick, > > > > > > The set never failed on my enviroment, and i just ensure set succeeded. > > > Do you want to remove the check? > > No, it's usually a good idea to make sure that the debugger actually did > > the thing we told it to. I was simply curious if this occurred on a > > regular basis. > Hi Darrick, > > OK, keep it. > > > > > + > > > > > +# Check a mounted XFS (online) > > > > > +_scratch_mount > > > > > +$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? > > > I tried and failed to trigger the assert by some regular commands(ls, stat, > > > etc.). > > > If you have some ideas, could you tell me how to trigger the assert by using > > > regular commands? > > Ah, ok. I'd have thought that "find $SCRATCH_MNT -type f -print0 | xargs > > -0 cat> /dev/null" would've triggered the assert, but I don't have any > > better suggestions. > I tried the find command, but failed to trigger the assert. > If nobody has a better way, can we still use xfs_scrub command in the test? > > Thanks, > Xiao Yang > > --D > > > > > Thanks, > > > Xiao Yang > > > > --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 > > > > > > > > > > > > > > > > > > > . > > > > > > > > > > > > > -- > > > To unsubscribe from this list: send the line "unsubscribe fstests" in > > > the body of a message to majordomo@vger.kernel.org > > > 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@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
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" != "$setmode" ] && _notrun "failed to set core.mode" + +# Check a mounted XFS (online) +_scratch_mount +$XFS_SCRUB_PROG -d -T -v -n $SCRATCH_MNT >> $seqres.full 2>&1 + +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
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