[Intel-gfx] [PATCH 0/3] HDMI optimization series

[Daniel Vetter]

On Wed, Aug 12, 2015 at 06:04:50PM +0530, Sharma, Shashank wrote:
> Regards
> Shashank
> 
> On 8/12/2015 5:55 PM, Daniel Vetter wrote:
> >On Tue, Aug 11, 2015 at 11:03:54AM +0000, Sharma, Shashank wrote:
> >>HI Daniel,
> >>
> >>Looks like we are not getting an IVB machine here. I think instead of blocking the patch set for the whole series, we can just block it for the platforms we don’t get success for.
> >>We are shipping many VLV/CHV systems with these patches applied, and they are passing all Android test suits and test benches, so we don’t doubt the solution as such.
> >>
> >>We have done testing for following platforms (gen 9 is tested by Imre also):
> >>1. CHV / BSW
> >>2. SKL
> >>3. BXT
> >>4. BDW
> >
> >I know that it works on piles of other platforms too. The problem is
> >figuring out why it doesn't work on some and what we're doing wrong.
> >
> >>Why don’t we go ahead with this solution for gen 8 onwards, and keep the
> >>rest of the code un touched, this will serve both the purpose.
> >
> >I guess we can, but if we get a regression report on bdw because of this
> >I'll be furious.
> >-Daniel
> >
> I completely agree, and understand your concern about adding regression. How
> about this, we perform an extended set of testing on BDW also, and only
> after seeing the test reports, we can take a call on merging these patches.

The problem is that no one seems to have machines with broken hpd, I
stumbled over the broken ivb here purely by accident. The only way to get
testing is to merge the patches and then wait for someone to complain.
Trying to do exhaustive testing beforehand is imo impossible.

The only thing which would alleviate my concerns is fixing the hpd
handling bugs we know about already, since in the past we've just done "oh
doesn't work on genX, try on genX+1 only then" and that never worked
out. And we've tried this for _every_ _single_ platform since gen4. But we
can try once more, simply be aware that if it breaks again I will not
accept a patch to restrict to gen9 but instead will rip it out again.
-Daniel

> >>
> >>Regards
> >>Shashank
> >>-----Original Message-----
> >>From: Intel-gfx [mailto:intel-gfx-bounces at lists.freedesktop.org] On Behalf Of Daniel Vetter
> >>Sent: Tuesday, August 11, 2015 3:12 PM
> >>To: Jindal, Sonika
> >>Cc: intel-gfx at lists.freedesktop.org
> >>Subject: Re: [Intel-gfx] [PATCH 0/3] HDMI optimization series
> >>
> >>On Mon, Aug 10, 2015 at 02:05:47PM +0530, Jindal, Sonika wrote:
> >>>
> >>>
> >>>On 8/10/2015 1:38 PM, Daniel Vetter wrote:
> >>>>On Mon, Aug 10, 2015 at 04:50:37AM +0000, Jindal, Sonika wrote:
> >>>>>Hi Daniel,
> >>>>>
> >>>>>That patch was already merged:
> >>>>>http://lists.freedesktop.org/archives/intel-gfx/2015-July/071142.htm
> >>>>>l
> >>>>>
> >>>>>For SKL, the above patch helped in getting the correct ISR bits set.
> >>>>>One option is to enable the HDMI optimization from VLV onwards.
> >>>>>I don't have an ivb machine to try out the issue.
> >>>>
> >>>>ivb is simple the machine I have here, but when we tried this the
> >>>>last time around we had reports for all platforms (your patch still
> >>>>doesn't cite the relevant sha1 btw). I think there are 3 possible explanations:
> >>>>
> >>>Yes, I don't know which were those patches and how to find them..
> >>
> >>git blame (recursively) and git log on intel_hdmi.c will tell you the story. I require all these extensive git commit messages because I dig them out daily. Also git log --pickaxe (well I use the equivalent in the gitk gui). If these tools are generally unknown in the vpg display team then we absolutely need to do a training session about them, they're extremely powerful and useful to dig out the history of the i915 code.
> >>
> >>>>a) we do something wrong with hpd handling on these platforms. That
> >>>>seems to be the explanation you favour (with the gen >= 7 checks and
> >>>>all that), but I think it's very unlikely: On each platform where we
> >>>>had reports of hpd being broken there was also machines where hpd works perfectly fine.
> >>>>
> >>>Not sure, I will find one ivb system and try on that.
> >>>
> >>>>b) There's broken HDMI (or DVI) sinks out there. If that's the case
> >>>>we can never merge your patch.
> >>>I doubt this because we have tested these patches with many sinks in
> >>>the past with VLV/CHV.
> >>>>
> >>>>c) There's something in vbt or somewhere else that tells the windows
> >>>>driver whether using hpd or not is ok (and the hpd problem is
> >>>>actually an issue with certain OEM machines ...).
> >>>>
> >>>No, nothing like that.
> >>>
> >>>>I hoped that with your hpd handling fix we'd have some indication
> >>>>that our hpd troubles are of type a). But since I tested with your
> >>>>patch that didn't work out.
> >>>>
> >>>>And until we have some evidence that our hpd troubles aren't type b)
> >>>>I really don't want to merge any patch which relies upon hpd bits for hdmi.
> >>>>-Daniel
> >>>>
> >>>I will try on ivb.
> >>
> >>I don't think it's ivb specific, since we do have ivb machines where hpd works perfectly fine. Same for every other platform. The only thing which seems common is that DP+ connectors work, but some of the hdmi-only connectors fail. That's why I wondered whether there's something i915 does different than the windows driver. In case you can get hold of one, the broken laptop I have here is an Asus UX31A with ivb.
> >>-Daniel
> >>--
> >>Daniel Vetter
> >>Software Engineer, Intel Corporation
> >>http://blog.ffwll.ch
> >>_______________________________________________
> >>Intel-gfx mailing list
> >>Intel-gfx at lists.freedesktop.org
> >>http://lists.freedesktop.org/mailman/listinfo/intel-gfx
> >

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch