[igt-dev] [PATCH i-g-t] igt/kms_frontbuffer_tracking: Skip tests if FBC was disabled

[Chris Wilson]

Quoting Lofstedt, Marta (2018-03-06 11:32:56)
> 
> 
> > -----Original Message-----
> > From: Chris Wilson [mailto:chris at chris-wilson.co.uk]
> > Sent: Tuesday, March 6, 2018 1:03 PM
> > To: Nikula, Jani <jani.nikula at intel.com>; Lofstedt, Marta
> > <marta.lofstedt at intel.com>; igt-dev at lists.freedesktop.org
> > Cc: Zanoni, Paulo R <paulo.r.zanoni at intel.com>; Lankhorst, Maarten
> > <maarten.lankhorst at intel.com>
> > Subject: RE: [igt-dev] [PATCH i-g-t] igt/kms_frontbuffer_tracking: Skip tests if
> > FBC was disabled
> > 
> > Quoting Jani Nikula (2018-03-06 10:55:02)
> > > On Tue, 06 Mar 2018, Chris Wilson <chris at chris-wilson.co.uk> wrote:
> > > > Quoting Lofstedt, Marta (2018-03-06 10:39:04)
> > > >>
> > > >>
> > > >> > -----Original Message-----
> > > >> > From: Chris Wilson [mailto:chris at chris-wilson.co.uk]
> > > >> > Sent: Tuesday, March 6, 2018 11:32 AM
> > > >> > To: Lofstedt, Marta <marta.lofstedt at intel.com>; igt-
> > > >> > dev at lists.freedesktop.org
> > > >> > Subject: Re: [igt-dev] [PATCH i-g-t]
> > > >> > igt/kms_frontbuffer_tracking: Skip tests if FBC was disabled
> > > >> >
> > > >> > Quoting Marta Lofstedt (2018-03-06 09:24:57)
> > > >> > > If FBC has been disabled in the system due to previous
> > > >> > > malfuction, we would save time on CI if we bail out early.
> > > >> > >
> > > >> > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=105359
> > > >> > > Signed-off-by: Marta Lofstedt <marta.lofstedt at intel.com>
> > > >> >
> > > >> > Hmm, FBC starts inactive on a system, so we would need to ensure
> > > >> > that we enabled FBC and did a proper modeset first and wait for
> > > >> > FBC to be enabled first.
> > > >>
> > > >> On the SKL NUCi5 system I have been running this on, the FBC disabled
> > sysfs, isn't re-set after a modeset. Maybe this is a driver error that should not
> > be hidden by me suggesting to skip the tests.
> > > >
> > > > It may just be that we see the same error. Some errors like
> > > > underruns permanently disable FBC (afaik).
> 
> Is, one underrun and forever no FBC is how we want it in the driver?

More or less, yes. Underruns == screen blinking == irate users. (Or
worse, dead machine.) FBC is a usual suspect in these matters.

Now, I would much rather put this under control of userspace; signal the
loss of FBC and let it decide whether to reprogram with a different mode
or give up. Just as for PSR, let userspace detect if FBC/PSR/whatever is
supported at this mode, and if not and userspace is power conscious, let
it try a reduced mode. (One of the purposes of the atomic modesetting API
was to allow userspace to make such global decisions.)
-Chris