[igt-dev] [PATCH i-g-t 3/4] lib: Don't leak children in igt_waitchildren_timeout

[Daniel Vetter]

On Thu, Feb 07, 2019 at 03:20:37PM +0000, Chris Wilson wrote:
> Quoting Daniel Vetter (2019-02-07 14:57:06)
> > Instead of cleaning up the mess in igt_exit make sure we don't even
> > let it out of the container. See also
> > 
> > commit 754876378d6c9b2775e8c07b4d16f9878c55949f
> > Author: Chris Wilson <chris at chris-wilson.co.uk>
> > Date:   Fri Feb 26 22:11:10 2016 +0000
> > 
> >     igt/gem_sync: Enforce a timeout of 20s
> > 
> > which added this helper.
> > 
> > To make sure that everyone follows the rules, add an assert.
> > 
> > We're keeping the cleanup code as a failsafe, and because it speeds
> > up the testcase I'm following up with.
> > 
> > Cc: Chris Wilson <chris at chris-wilson.co.uk>
> > Signed-off-by: Daniel Vetter <daniel.vetter at intel.com>
> > ---
> >  lib/igt_core.c | 6 ++++++
> >  1 file changed, 6 insertions(+)
> > 
> > diff --git a/lib/igt_core.c b/lib/igt_core.c
> > index 49fbf70deb06..a7105f0591fc 100644
> > --- a/lib/igt_core.c
> > +++ b/lib/igt_core.c
> > @@ -1525,6 +1525,7 @@ void igt_exit(void)
> >  
> >         for (int c = 0; c < num_test_children; c++)
> >                 kill(test_children[c], SIGKILL);
> > +       assert(!num_test_children);
> >  
> >         if (!test_with_subtests) {
> >                 struct timespec now;
> > @@ -1842,6 +1843,11 @@ void igt_waitchildren_timeout(int seconds, const char *reason)
> >         igt_set_timeout(seconds, reason);
> >         igt_waitchildren();
> >         igt_reset_timeout();
> > +
> > +       for (int c = 0; c < num_test_children; c++)
> > +               kill(test_children[c], SIGKILL);
> > +
> > +       __igt_waitchildren();
> 
> How did we escape?
> 
> If the SIGALRM fired, we hit igt_fail() in the parent with children in
> tow, otherwise we completed igt_waitchildren() successfully and cleaned
> everything up.

Hm right, this does nothing.

The motivation for this is that I noticed that without an explicit
igt_waitchildren() the igt_assert forwarding doesn't work. So I've tried
to catch that kind of leaking.

The trouble that does happen is if you run multiple tests, then the
children do leak into the next subtest (since igt_exit is only called at
the very end). That's the usual trouble with our exit handlers not
properly stacking. In the end I wanted to put a wait() into igt_exit and
check that it gives us ECHILD, to make sure everything is cleaned up.

So not sure what to do ... any ideas?
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch