[Intel-gfx] [PATCH] drm/i915/crt: Do not rely upon the HPD presence pin

[Chris Wilson]

On Sun, 10 Jun 2012 19:04:10 +0200, Daniel Vetter <daniel at ffwll.ch> wrote:
> On Fri, Jun 08, 2012 at 11:23:10PM +0100, Chris Wilson wrote:
> > On Sat, 9 Jun 2012 00:22:12 +0200, Daniel Vetter <daniel at ffwll.ch> wrote:
> > > On Thu, May 31, 2012 at 01:08:53PM +0100, Chris Wilson wrote:
> > > > Whilst most monitors do wire up the HPD presence pin, it seems quite a
> > > > few KVM do not. Therefore if we simply rely on the HPD pin being
> > > > asserted to indicate a connected monitor we fail miserable, so fall back
> > > > to performing a DCC query for the EDID.
> > > > 
> > > > Reported-and-tested-by: Matthieu LAVIE <boiteamadmax at hotmail.com>
> > > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=50501
> > > > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > > 
> > > Ok, this blew up ... Can you please resend, with Dave's suggestion for a
> > > rectified commit message & comment and with a check added such that we
> > > don't try to do load_detect on HAS_HOTPLUG machines - I guess it doesn't
> > > work too well.
> > 
> > I disagree, if we cannot trust the hw autodetection, then we know that
> > there are monitors/kvm that do not report an EDID and so we need to do
> > the whole shebang. Which will continue to annoy Linus since his machine
> > is behaving as expected given the circumstances.
> 
> Well, I don't disagree on doing the whole shebang. The proplem is that the
> load-detect code as-is is gen3 only (and maybe gen4, haven't checked
> that) - it surely can't work on pch split platforms if half the registers
> we use in there are gone.
> 
> Until that is fixed and properly tested on all relevant platforms, we
> should be able to help the bug reporters by simply using the edid
> detection, but bailing on the load detect stuff for all HAS_HOTPLUG
> platforms (as we do now already). I'll whip up a patch.
> 
> For actual load-detect stuff is imo -next material, and I think we should
> dodge that bullet until we have an actual bug reporter wanting it ...

I can send them one of my monitors that fails to report an EDID to them
so that they can put it behind their KVM that breaks autodetection...

Coming up with a solution to handling unknown connection status is indeed
-next material, so I'm not too concerned if we punt the entire thing so
that we can do a thorough job. If we could handle unknown cleanly, it
would have prevented a lot of misery over the years with spurious TV
detection and the like.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre