[igt-dev] [PATCH i-g-t] tests/kms_content_protection: CP cleanup exit handler

[Petri Latvala]

On Tue, May 19, 2020 at 03:57:13PM +0530, Anshuman Gupta wrote:
> On 2020-05-19 at 11:42:26 +0300, Petri Latvala wrote:
> > On Tue, May 19, 2020 at 11:40:28AM +0530, Anshuman Gupta wrote:
> > > Add HDCP content protection cleanup igt exit handler,
> > > earlier it was done through igt_fixture which doesn't
> > > trigger hdcp cleanup on igt abortion due to any signal.
> > > This should avoid any HDCP content protection leak.
> > > 
> > > Cc: Ramalingam C <ramalingam.c at intel.com>
> > > Signed-off-by: Anshuman Gupta <anshuman.gupta at intel.com>
> > > ---
> > >  tests/kms_content_protection.c | 14 +++++++++-----
> > >  1 file changed, 9 insertions(+), 5 deletions(-)
> > > 
> > > diff --git a/tests/kms_content_protection.c b/tests/kms_content_protection.c
> > > index 3b9cedcb..f3101b8a 100644
> > > --- a/tests/kms_content_protection.c
> > > +++ b/tests/kms_content_protection.c
> > > @@ -651,12 +651,21 @@ static void test_content_protection_cleanup(void)
> > >  	}
> > >  }
> > >  
> > > +static void test_content_protection_exit_handler(int sig)
> > > +{
> > > +	igt_fixture {
> > > +		test_content_protection_cleanup();
> > > +		igt_display_fini(&data.display);
> > > +	}
> >
> Thanks Petri for review. 
> > Don't put it in an igt_fixture. If a test fails, you're still inside a
> > subtest and entering an igt_fixture will fail.
> test_content_protection_cleanup is using for_each_connected_output() which
> of using assert on igt_can_fail(), because of that i had to used
> igt_fixture(), could you please suggest to mitigate above.

Ohhh bugger, you're right =(

Well, the lib is in need of a good cleanup regarding exit handlers, as
this has now revealed. We need to fix the context tracking to properly
support exit handlers and have igt_can_fail work in those.

I suppose this code should work for now, the comment included:


/*
 * FIXME: igt_can_fail() is sometimes incorrect in exit handlers. Wrap
 * cleanup in igt_fixture if igt_core thinks we're still in a subtest.
 */
if (igt_subtest_name()) {
  test_content_protection_cleanup();
  igt_display_fini(&data.display);
 } else {
  igt_fixture {
    test_content_protection_cleanup();
    igt_display_fini(&data.display);
  }
 }


-- 
Petri Latvala