[Intel-gfx] [PATCH 09/12] tests/kms_psr_sink_crc: Fix all testcases.

[Daniel Vetter]

On Thu, Sep 04, 2014 at 05:55:24PM -0700, Rodrigo Vivi wrote:
> adding suspend_autoresume on primary tests like this:
> @ -470,6 +472,8 @@ igt_main
>                         data.test_plane = PRIMARY;
>                         data.op = op;
>                         run_test(&data);
> +                       igt_system_suspend_autoresume();
> +                       run_test(&data);
> 
> on BDW I got these results:
> 
> 
> vivijim rdvivi-seattle tests$ sudo ./kms_psr_sink_crc
> IGT-Version: 1.7-gd4b43f0 (x86_64) (Linux: 3.17.0-rc2+ x86_64)
> rtcwake: wakeup from "mem" using /dev/rtc0 at Fri Sep  5 00:44:03 2014
> Subtest primary_page_flip: SUCCESS
> rtcwake: wakeup from "mem" using /dev/rtc0 at Fri Sep  5 00:44:40 2014
> Subtest primary_mmap_gtt: SUCCESS
> Waiting 10s...
> rtcwake: wakeup from "mem" using /dev/rtc0 at Fri Sep  5 00:45:27 2014
> Waiting 10s...
> Subtest primary_mmap_gtt_waiting: SUCCESS
> rtcwake: wakeup from "mem" using /dev/rtc0 at Fri Sep  5 00:46:13 2014
> Subtest primary_mmap_cpu: SUCCESS
> rtcwake: wakeup from "mem" using /dev/rtc0 at Fri Sep  5 00:46:50 2014
> Subtest primary_blt: SUCCESS
> rtcwake: wakeup from "mem" using /dev/rtc0 at Fri Sep  5 00:47:27 2014
> Subtest primary_render: SUCCESS
> 
> on HSW I couldn't test because suspend/resume breaks even with psr disabled.
> I'm going to check more tomorrow..
> 
> But regarding the suspend resume test, how do you suggest to organize it?
> Extra loops for all current cases?
> suspend_{primary, sprite, cursor}_{page_flip, mmap_gtt, etc}? I believe the
> test will take so long to finish on this case what is bad for qa alghouth
> it is the complete one. What do you think?

I think we don't need the full set of tests also with system suspend. I
think just one test which catches the current bug is good enough, after
all if psr is set up correctly it should work the same at runtime than
over s/r. And since this is a test I'd just copypaste the relevant subtest
(if it doesn't integrate quickly into the existing code), not worth at all
to make a big fuzz.

And we have lots of resume tests already, they "only" take about
30 second. Only important to have "suspend" somewhere in the subtest name
so that all system suspend tests can easily be filtered out/selected.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch