[PATCH v2] generic/648: dmerror must be unmounted

[PATCH v2] generic/648: dmerror must be unmounted

[Zorro Lang]

Sometimes g/648 fail to unmount dmerror with this error:

   umount: /mnt/xfstests/scratch: target is busy.

Even worse, it will cause all later test cases fail as:

  mount: bad usage
  Try 'mount --help' for more information.
  check: failed to mount $SCRATCH_DEV using specified options
  Interrupted!

So we shouldn't _fail directly if dmerror_unmount fails, use a while
loop to try to unmount it enough times.

Signed-off-by: Zorro Lang <zlang@redhat.com>
---

As Darrick's review point, V2 brings the "iteration XXX scratch
unmount failed" _fail back.

Thanks,
Zorro

 tests/generic/648 | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

diff --git a/tests/generic/648 b/tests/generic/648
index 83dd111d..e5c743c5 100755
--- a/tests/generic/648
+++ b/tests/generic/648
@@ -104,7 +104,7 @@ for i in $(seq 1 $((25 * TIME_FACTOR)) ); do
 		wait > /dev/null 2>&1
 		ps -e | grep fsstress > /dev/null 2>&1
 	done
-	for ((i = 0; i < 10; i++)); do
+	for ((j = 0; j < 10; j++)); do
 		test -e "$snap_aliveflag" || break
 		sleep 1
 	done
@@ -112,7 +112,21 @@ for i in $(seq 1 $((25 * TIME_FACTOR)) ); do
 	# Mount again to replay log after loading working table, so we have a
 	# consistent fs after test.
 	$UMOUNT_PROG $loopmnt
-	_dmerror_unmount || _fail "iteration $i scratch unmount failed"
+	is_unmounted=1
+	# We must unmount dmerror at here, or whole later testing will crash.
+	# So try to umount enough times, before we have no choice.
+	for ((j = 0; j < 100; j++)); do
+		sleep 1
+		_dmerror_unmount > $tmp.unmount.err 2>&1
+		if [ $? -eq 0 ];then
+			is_unmounted=0
+			break
+		fi
+	done
+	if [ $is_unmounted -ne 0 ];then
+		cat $tmp.unmount.err
+		_fail "iteration $i scratch unmount failed"
+	fi
 	_dmerror_load_working_table
 	if ! _dmerror_mount; then
 		_metadump_dev $DMERROR_DEV $seqres.scratch.$i.md
-- 
2.31.1

Re: [PATCH v2] generic/648: dmerror must be unmounted

[xuyang2018.jy]

on 2021/10/24 20:05, Zorro Lang wrote:
> Sometimes g/648 fail to unmount dmerror with this error:
> 
>     umount: /mnt/xfstests/scratch: target is busy.
> 
> Even worse, it will cause all later test cases fail as:
> 
>    mount: bad usage
>    Try 'mount --help' for more information.
>    check: failed to mount $SCRATCH_DEV using specified options
>    Interrupted!
> 
> So we shouldn't _fail directly if dmerror_unmount fails, use a while
> loop to try to unmount it enough times.
I also met EBUSY error when testing other cases(I need to filter about
60 cases for this reason and do signle test for them) that using dmerror
or dmflakey.

Can we add a umount_api for dmerror/dmflakey that use a while loop to
umount during enough time in xfstest layer?

Best Regards
Yang Xu
> 
> Signed-off-by: Zorro Lang<zlang@redhat.com>
> ---
> 
> As Darrick's review point, V2 brings the "iteration XXX scratch
> unmount failed" _fail back.
> 
> Thanks,
> Zorro
> 
>   tests/generic/648 | 18 ++++++++++++++++--
>   1 file changed, 16 insertions(+), 2 deletions(-)
> 
> diff --git a/tests/generic/648 b/tests/generic/648
> index 83dd111d..e5c743c5 100755
> --- a/tests/generic/648
> +++ b/tests/generic/648
> @@ -104,7 +104,7 @@ for i in $(seq 1 $((25 * TIME_FACTOR)) ); do
>   		wait>  /dev/null 2>&1
>   		ps -e | grep fsstress>  /dev/null 2>&1
>   	done
> -	for ((i = 0; i<  10; i++)); do
> +	for ((j = 0; j<  10; j++)); do
>   		test -e "$snap_aliveflag" || break
>   		sleep 1
>   	done
> @@ -112,7 +112,21 @@ for i in $(seq 1 $((25 * TIME_FACTOR)) ); do
>   	# Mount again to replay log after loading working table, so we have a
>   	# consistent fs after test.
>   	$UMOUNT_PROG $loopmnt
> -	_dmerror_unmount || _fail "iteration $i scratch unmount failed"
> +	is_unmounted=1
> +	# We must unmount dmerror at here, or whole later testing will crash.
> +	# So try to umount enough times, before we have no choice.
> +	for ((j = 0; j<  100; j++)); do
> +		sleep 1
> +		_dmerror_unmount>  $tmp.unmount.err 2>&1
> +		if [ $? -eq 0 ];then
> +			is_unmounted=0
> +			break
> +		fi
> +	done
> +	if [ $is_unmounted -ne 0 ];then
> +		cat $tmp.unmount.err
> +		_fail "iteration $i scratch unmount failed"
> +	fi
>   	_dmerror_load_working_table
>   	if ! _dmerror_mount; then
>   		_metadump_dev $DMERROR_DEV $seqres.scratch.$i.md

Re: [PATCH v2] generic/648: dmerror must be unmounted

[Zorro Lang]

On Sun, Oct 24, 2021 at 08:05:12PM +0800, Zorro Lang wrote:
> Sometimes g/648 fail to unmount dmerror with this error:
> 
>    umount: /mnt/xfstests/scratch: target is busy.
> 
> Even worse, it will cause all later test cases fail as:
> 
>   mount: bad usage
>   Try 'mount --help' for more information.
>   check: failed to mount $SCRATCH_DEV using specified options
>   Interrupted!
> 
> So we shouldn't _fail directly if dmerror_unmount fails, use a while
> loop to try to unmount it enough times.
> 
> Signed-off-by: Zorro Lang <zlang@redhat.com>
> ---

Ping, it's been 2+ weeks passed, any more review points about this?

Thanks,
Zorro

> 
> As Darrick's review point, V2 brings the "iteration XXX scratch
> unmount failed" _fail back.
> 
> Thanks,
> Zorro
> 
>  tests/generic/648 | 18 ++++++++++++++++--
>  1 file changed, 16 insertions(+), 2 deletions(-)
> 
> diff --git a/tests/generic/648 b/tests/generic/648
> index 83dd111d..e5c743c5 100755
> --- a/tests/generic/648
> +++ b/tests/generic/648
> @@ -104,7 +104,7 @@ for i in $(seq 1 $((25 * TIME_FACTOR)) ); do
>  		wait > /dev/null 2>&1
>  		ps -e | grep fsstress > /dev/null 2>&1
>  	done
> -	for ((i = 0; i < 10; i++)); do
> +	for ((j = 0; j < 10; j++)); do
>  		test -e "$snap_aliveflag" || break
>  		sleep 1
>  	done
> @@ -112,7 +112,21 @@ for i in $(seq 1 $((25 * TIME_FACTOR)) ); do
>  	# Mount again to replay log after loading working table, so we have a
>  	# consistent fs after test.
>  	$UMOUNT_PROG $loopmnt
> -	_dmerror_unmount || _fail "iteration $i scratch unmount failed"
> +	is_unmounted=1
> +	# We must unmount dmerror at here, or whole later testing will crash.
> +	# So try to umount enough times, before we have no choice.
> +	for ((j = 0; j < 100; j++)); do
> +		sleep 1
> +		_dmerror_unmount > $tmp.unmount.err 2>&1
> +		if [ $? -eq 0 ];then
> +			is_unmounted=0
> +			break
> +		fi
> +	done
> +	if [ $is_unmounted -ne 0 ];then
> +		cat $tmp.unmount.err
> +		_fail "iteration $i scratch unmount failed"
> +	fi
>  	_dmerror_load_working_table
>  	if ! _dmerror_mount; then
>  		_metadump_dev $DMERROR_DEV $seqres.scratch.$i.md
> -- 
> 2.31.1
> 

Re: [PATCH v2] generic/648: dmerror must be unmounted

[Darrick J. Wong]

On Sun, Oct 24, 2021 at 08:05:12PM +0800, Zorro Lang wrote:
> Sometimes g/648 fail to unmount dmerror with this error:
> 
>    umount: /mnt/xfstests/scratch: target is busy.
> 
> Even worse, it will cause all later test cases fail as:
> 
>   mount: bad usage
>   Try 'mount --help' for more information.
>   check: failed to mount $SCRATCH_DEV using specified options
>   Interrupted!
> 
> So we shouldn't _fail directly if dmerror_unmount fails, use a while
> loop to try to unmount it enough times.
> 
> Signed-off-by: Zorro Lang <zlang@redhat.com>

Looks good to me now,
Reviewed-by: Darrick J. Wong <djwong@kernel.org>

--D

> ---
> 
> As Darrick's review point, V2 brings the "iteration XXX scratch
> unmount failed" _fail back.
> 
> Thanks,
> Zorro
> 
>  tests/generic/648 | 18 ++++++++++++++++--
>  1 file changed, 16 insertions(+), 2 deletions(-)
> 
> diff --git a/tests/generic/648 b/tests/generic/648
> index 83dd111d..e5c743c5 100755
> --- a/tests/generic/648
> +++ b/tests/generic/648
> @@ -104,7 +104,7 @@ for i in $(seq 1 $((25 * TIME_FACTOR)) ); do
>  		wait > /dev/null 2>&1
>  		ps -e | grep fsstress > /dev/null 2>&1
>  	done
> -	for ((i = 0; i < 10; i++)); do
> +	for ((j = 0; j < 10; j++)); do
>  		test -e "$snap_aliveflag" || break
>  		sleep 1
>  	done
> @@ -112,7 +112,21 @@ for i in $(seq 1 $((25 * TIME_FACTOR)) ); do
>  	# Mount again to replay log after loading working table, so we have a
>  	# consistent fs after test.
>  	$UMOUNT_PROG $loopmnt
> -	_dmerror_unmount || _fail "iteration $i scratch unmount failed"
> +	is_unmounted=1
> +	# We must unmount dmerror at here, or whole later testing will crash.
> +	# So try to umount enough times, before we have no choice.
> +	for ((j = 0; j < 100; j++)); do
> +		sleep 1
> +		_dmerror_unmount > $tmp.unmount.err 2>&1
> +		if [ $? -eq 0 ];then
> +			is_unmounted=0
> +			break
> +		fi
> +	done
> +	if [ $is_unmounted -ne 0 ];then
> +		cat $tmp.unmount.err
> +		_fail "iteration $i scratch unmount failed"
> +	fi
>  	_dmerror_load_working_table
>  	if ! _dmerror_mount; then
>  		_metadump_dev $DMERROR_DEV $seqres.scratch.$i.md
> -- 
> 2.31.1
> 

Re: [PATCH v2] generic/648: dmerror must be unmounted

[Eryu Guan]

On Tue, Nov 09, 2021 at 12:09:13PM +0800, Zorro Lang wrote:
> On Sun, Oct 24, 2021 at 08:05:12PM +0800, Zorro Lang wrote:
> > Sometimes g/648 fail to unmount dmerror with this error:
> > 
> >    umount: /mnt/xfstests/scratch: target is busy.
> > 
> > Even worse, it will cause all later test cases fail as:
> > 
> >   mount: bad usage
> >   Try 'mount --help' for more information.
> >   check: failed to mount $SCRATCH_DEV using specified options
> >   Interrupted!
> > 
> > So we shouldn't _fail directly if dmerror_unmount fails, use a while
> > loop to try to unmount it enough times.
> > 
> > Signed-off-by: Zorro Lang <zlang@redhat.com>
> > ---
> 
> Ping, it's been 2+ weeks passed, any more review points about this?

Sorry, I lost it in my to-review queue.. Applied now.

Thanks,
Eryu