Re: [PATCH 1/2] check: try to fix the test device if it gets corrupted

["Darrick J. Wong" <djwong@kernel.org>]

On Fri, Apr 14, 2023 at 03:11:32PM -0700, Leah Rumancik wrote:
> From: Theodore Ts'o <tytso@mit.edu>
> 
> If the test device gets corrupted all subsequent tests will fail.  To
> prevent this from causing all subsequent tests to be useless, try
> repair the file system on TEST_DEV if possible.  We don't need to do
> this with the scratch device since that file system gets recreated
> each time anyway.
> 
> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
> ---
>  check      |  7 ++++++-
>  common/rc  | 41 +++++++++++++++++++++++++++++++++++++++++
>  common/xfs | 12 ++++++++++++
>  3 files changed, 59 insertions(+), 1 deletion(-)
> 
> diff --git a/check b/check
> index 1a58a2b2..befbf465 100755
> --- a/check
> +++ b/check
> @@ -536,7 +536,12 @@ _check_filesystems()
>  	local ret=0
>  
>  	if [ -f ${RESULT_DIR}/require_test ]; then
> -		_check_test_fs || ret=1
> +		if ! _check_test_fs ; then
> +			ret=1
> +			echo "Trying to repair broken TEST_DEV file system"
> +			_repair_test_fs
> +			_test_mount
> +		fi
>  		rm -f ${RESULT_DIR}/require_test*
>  	else
>  		_test_unmount 2> /dev/null
> diff --git a/common/rc b/common/rc
> index 90749343..e8fc7e86 100644
> --- a/common/rc
> +++ b/common/rc
> @@ -1199,6 +1199,47 @@ _repair_scratch_fs()
>      esac
>  }
>  
> +_repair_test_fs()
> +{
> +	case $FSTYP in
> +	xfs)
> +		_repair_xfs_test_fs "$@" >$tmp.repair 2>&1
> +		res=$?
> +		if [ "$res" -ne 0 ]; then
> +			echo "xfs_repair returns $res; replay log?" >>$tmp.repair
> +			_test_mount
> +			res=$?
> +			if [ $res -gt 0 ]; then
> +				echo "mount returns $res; zap log?" >>$tmp.repair
> +				_xfs_repair_test_fs -L >>$tmp.repair 2>&1
> +				echo "log zap returns $?" >> $tmp.repair
> +			else
> +				umount "$TEST_DEV"
> +			fi
> +			_xfs_repair_test_fs "$@" >>$tmp.repair 2>&1
> +			res=$?
> +		fi
> +		;;
> +	*)
> +		# Let's hope fsck -y suffices...
> +		fsck -t $FSTYP -fy $TEST_DEV >$tmp.repair 2>&1
> +		res=$?
> +		if test "$res" -lt 4 ; then
> +			res=0
> +		fi
> +		;;
> +	esac
> +	if [ $res -ne 0 ]; then
> +		_log_err "_repair_test_fs: failed, err=$res"
> +		echo "*** fsck.$FSTYP output ***"	>>$seqres.full
> +		cat $tmp.repair				>>$seqres.full
> +		echo "*** end fsck.$FSTYP output"	>>$seqres.full
> +
> +	fi
> +	rm -f $tmp.repair
> +	return $res
> +}
> +
>  _get_pids_by_name()
>  {
>      if [ $# -ne 1 ]
> diff --git a/common/xfs b/common/xfs
> index e8e4832c..4a130493 100644
> --- a/common/xfs
> +++ b/common/xfs
> @@ -988,6 +988,18 @@ _check_xfs_test_fs()
>  	return $?
>  }
>  
> +# modeled after _scratch_xfs_repair
> +_repair_xfs_test_fs()

Dumb bikeshed: Can this be named _test_xfs_repair? :D

With that changed,
Reviewed-by: Darrick J. Wong <djwong@kernel.org>

--D

> +{
> +	TEST_OPTIONS=""
> +	[ "$USE_EXTERNAL" = yes -a ! -z "$TEST_LOGDEV" ] && \
> +		TEST_OPTIONS="-l$TEST_LOGDEV"
> +	[ "$USE_EXTERNAL" = yes -a ! -z "$TEST_RTDEV" ] && \
> +		TEST_OPTIONS=$TEST_OPTIONS" -r$TEST_RTDEV"
> +	[ "$LARGE_TEST_DEV" = yes ] && TEST_OPTIONS=$TEST_OPTIONS" -t"
> +	$XFS_REPAIR_PROG $TEST_OPTIONS $* $TEST_DEV
> +}
> +
>  _require_xfs_test_rmapbt()
>  {
>  	_require_test
> -- 
> 2.40.0.634.g4ca3ef3211-goog
>