On Tue, 2026-07-07 at 09:00 +0200, Nam Cao wrote:

Gabriele Monaco [off-list ref] writes:

quoted

Monitor initialisation also called during KUnit tests may register some
tracepoints, this can lead to issues since we don't expect real monitor
events running during KUnit tests.

Prevent tracepoint registration if an RV KUnit test is running.

Reviewed-by: Nam Cao <redacted>
Signed-off-by: Gabriele Monaco <gmonaco@redhat.com>

Somehow I have no recollection of seeing this patch.

Well, I'm fairly sure you reviewed it ;)

But since we already hold rv_interface_lock during the test, is this
really necessary?

This is for things like ltl_monitor_init() that attaches the newtask probe.
The function is also doing a few other things we may not want to do, to be fair,
but calling the monitor_init() function as-is was the easiest solution I could
think of. Also ha_monitor_init() may attach a probe.

I suppose I could also do some more targeted initialisation/destruction, in fact
right now, the LTL initialisation does a bunch of stuff tests don't need and
doesn't do what they do need (initialise the dummy tasks, which aren't part of
the task list).

I should probably rethink this a little.

Thanks,
Gabriele