[PATCH blktests 1/2] loop/009: only skip on a genuinely unrecognized --ping option

[PATCH blktests 1/2] loop/009: only skip on a genuinely unrecognized --ping option

[Sebastian Chlad]

udevadm control --ping can also fail when udevd is busy or unresponsive
not only when the option is genuinely unsupported. Treating any failure
as "unsupported" risked masking a real udevd problem behind a
misleading skip reason instead of letting the test run and report it.

Signed-off-by: Sebastian Chlad <sebastian.chlad@suse.com>
---
 tests/loop/009 | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/loop/009 b/tests/loop/009
index d3c7991..944f53b 100755
--- a/tests/loop/009
+++ b/tests/loop/009
@@ -11,7 +11,7 @@ DESCRIPTION="check that LOOP_CONFIGURE sends uevents for partitions"
 QUICK=1
 
 requires() {
-	if ! udevadm control --ping >& /dev/null; then
+	if ! udevadm control --help 2>&1 | grep -q -- '--ping'; then
 		SKIP_REASONS+=("udevadm control does not support --ping option")
 	fi
 }
-- 
2.51.0

[PATCH blktests 2/2] loop/006: settle udev before completing the test

[Sebastian Chlad]

Ensures udev has processed the burst of attach/detach/LOOP_CHANGE_FD
activity generated by this test before control is handed back, so it
does not leave a backlog for the next test to contend with.

Signed-off-by: Sebastian Chlad <sebastian.chlad@suse.com>
---
 tests/loop/006 | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/tests/loop/006 b/tests/loop/006
index 549c112..701d44c 100755
--- a/tests/loop/006
+++ b/tests/loop/006
@@ -69,5 +69,10 @@ test() {
 	fi
 	losetup -d "$loop_dev"
 
+	# Let udev catch up on the attach/detach/change-fd before
+	# handing control back, so it doesn't leave a backlog for
+	# whatever test runs next.
+	udevadm settle
+
 	echo "Test complete"
 }
-- 
2.51.0

Re: [PATCH blktests 1/2] loop/009: only skip on a genuinely unrecognized --ping option

[Shin'ichiro Kawasaki]

On Jul 21, 2026 / 18:04, Sebastian Chlad wrote:
> udevadm control --ping can also fail when udevd is busy or unresponsive
> not only when the option is genuinely unsupported. Treating any failure
> as "unsupported" risked masking a real udevd problem behind a
> misleading skip reason instead of letting the test run and report it.
> 
> Signed-off-by: Sebastian Chlad <sebastian.chlad@suse.com>
> ---
>  tests/loop/009 | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tests/loop/009 b/tests/loop/009
> index d3c7991..944f53b 100755
> --- a/tests/loop/009
> +++ b/tests/loop/009
> @@ -11,7 +11,7 @@ DESCRIPTION="check that LOOP_CONFIGURE sends uevents for partitions"
>  QUICK=1
>  
>  requires() {
> -	if ! udevadm control --ping >& /dev/null; then
> +	if ! udevadm control --help 2>&1 | grep -q -- '--ping'; then

I applied this patch. FYI, I took the liberty to replace the short option -q
above with the long option --quiet. Thanks!

Re: [PATCH blktests 2/2] loop/006: settle udev before completing the test

[Shin'ichiro Kawasaki]

On Jul 21, 2026 / 18:04, Sebastian Chlad wrote:
> Ensures udev has processed the burst of attach/detach/LOOP_CHANGE_FD
> activity generated by this test before control is handed back, so it
> does not leave a backlog for the next test to contend with.
> 
> Signed-off-by: Sebastian Chlad <sebastian.chlad@suse.com>

I applied this patch also. Thanks!