On Wed, May 18, 2022 at 01:34:24PM -0400, Joe Lawrence wrote:
> The test_klp_callbacks_busy module conditionally blocks a future
> livepatch transition by busy waiting inside its workqueue function,
> busymod_work_func(). After scheduling this work, a test livepatch is
> loaded, introducing the transition under test.
>
> Both events are marked in the kernel log for later verification, but
> there is no synchronization to ensure that busymod_work_func() logs its
> function entry message before subsequent selftest commands log their own
> messages. This can lead to a rare test failure due to unexpected
> ordering like:
>
> --- expected
> +++ result
> @@ -1,7 +1,7 @@
> % modprobe test_klp_callbacks_busy block_transition=Y
> test_klp_callbacks_busy: test_klp_callbacks_busy_init
> -test_klp_callbacks_busy: busymod_work_func enter
> % modprobe test_klp_callbacks_demo
> +test_klp_callbacks_busy: busymod_work_func enter
> livepatch: enabling patch 'test_klp_callbacks_demo'
> livepatch: 'test_klp_callbacks_demo': initializing patching transition
> test_klp_callbacks_demo: pre_patch_callback: vmlinux
>
> Force the module init function to wait until busymod_work_func() has
> started (and logged its message), before exiting to the next selftest
> steps.
>
> Fixes: 547840bd5ae5 ("selftests/livepatch: simplify test-klp-callbacks busy target tests")
> Signed-off-by: Joe Lawrence <joe.lawrence@xxxxxxxxxx>
> ---
>
> Hi Petr, I remember you discouraged against a completion variable the
> first time around [1], but is there any better way with the workqueue
> API to ensure our "enter" message gets logged first? Or should we just
> drop the msg altogether to avoid the situation? I don't think it's
> absolutely necessary for the tests.
>
> [1] https://lore.kernel.org/live-patching/20200602081654.GI27273@linux-b0ei/
>
> lib/livepatch/test_klp_callbacks_busy.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/lib/livepatch/test_klp_callbacks_busy.c b/lib/livepatch/test_klp_callbacks_busy.c
> index 7ac845f65be5..eb502b2bb3ef 100644
> --- a/lib/livepatch/test_klp_callbacks_busy.c
> +++ b/lib/livepatch/test_klp_callbacks_busy.c
> @@ -16,10 +16,16 @@ MODULE_PARM_DESC(block_transition, "block_transition (default=false)");
>
> static void busymod_work_func(struct work_struct *work);
> static DECLARE_WORK(work, busymod_work_func);
> +static DECLARE_COMPLETION(busymod_work_started);
>
> static void busymod_work_func(struct work_struct *work)
> {
> + /*
> + * Hold the init function from exiting until we've started and
> + * announced our appearence in the kernel log.
> + */
> pr_info("%s enter\n", __func__);
> + complete(&busymod_work_started);
>
> while (READ_ONCE(block_transition)) {
> /*
> @@ -36,6 +42,7 @@ static int test_klp_callbacks_busy_init(void)
> {
> pr_info("%s\n", __func__);
> schedule_work(&work);
> + wait_for_completion(&busymod_work_started);
>
> if (!block_transition) {
> /*
> --
> 2.26.3
>
[ fixup ycote's email address, sorry Yannick :D ]
-- Joe
