Re: [PATCH 0/3] tracing: Fix some selftest issues

[Google]

On Tue, 4 Jun 2024 09:57:46 -0400
Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:

> On Fri, 31 May 2024 23:20:47 +0900
> Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx> wrote:
> 
> > The major conflict happens when the boot-time test cleans up the kprobe
> > events by
> > 
> >   dyn_events_release_all(&trace_kprobe_ops);
> > 
> > And I removed it by [3/3] patch in this series :) because it does not
> > needed and not confirmed there is no other kprobe events when the test
> > starts. Also the warning message are redundant so I removed it by [2/3].
> > 
> > So without this [1/3], if we apply [2/3] and [3/3], the problem will be
> > mitigated, but I think the root cause is that these modules are built-in.
> 
> I'm OK with making them module only, but I don't see any selftests for
> sythetic events. I think they should have a boot up test as well. If we
> remove them, let's add something to test them at boot up. Then the boot up
> code could clean it up.
> 
> Or change the test module to be a boot up test that cleans itself up if it
> is compiled in as not a module?

Yeah, I think we may need another test code for synthetic events, which
also triggering the synthetic events.

BTW, some these bootup tests can be ported on KUnit. Do you have a plan to
use KUnit?

Thank you,

> 
> -- Steve
> 


-- 
Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx>