[igt-dev] [PATCH i-g-t] tests/kms_cursor_legacy: increase timeout for nonblocking flip wait

[Ville Syrjälä]

On Fri, Feb 21, 2020 at 01:16:25PM +0000, Chris Wilson wrote:
> Quoting Juha-Pekka Heikkila (2020-02-21 13:12:46)
> > On 21.2.2020 14.23, Chris Wilson wrote:
> > > Quoting Juha-Pekka Heikkila (2020-02-21 12:08:07)
> > >> When run on more than one monitor there may be modeset in queue
> > >> which will block flip. Kernel will timeout with modeset at
> > >> ten seconds so lets wait in test also that same ten seconds at
> > >> maximum.
> > > 
> > > 10s! You are joking, right? For a modeset that should not be there?
> > 
> > 10s sounded quite big number to me too but that is from Ville. This test 
> > is now failing because modeset + flip take more than one second, on my 
> > ICL box with eDP + HDMI connected it becomes 1.19s. Arek is talking 
> > about test box where it is more than two seconds. Anyway if 10s is what 
> > kernel is waiting in the worst case I don't think test should be more 
> > strict.
> 
> I think the kernel is broken. From a UX perspective that needs fixing.

10s is the flip_done timeout. We shouldn't need quite that long.
IIRC the max panel power cycle delay we can program is something
like 5s.

I had a quick look at the test case but couldn't quite figure out
what specifically it was trying to say with the different timeouts.

-- 
Ville Syrjälä
Intel