Bug 99362

Please try the drm-tip branch of the drm-tip repo. That's the bleeding edge. Let's see if we've fixed this already upstream.

If it doesn't, you have the starting points for a git bisect:

git bisect start
git bisect good f8896f5d58e64bfd3c2b5f7c5ba5c3f3967e93c7
git bisect bad ca6e4405779ed56ebac941570615abd667c72c02

and try the kernels git gives you, and tell git what the result was.

Sorry, my fault. I have already tried drm-intel and drm-tip with no success. 
Thanks for your advice to try bisect,  I'll do so.

looks likes connector is hdmi, isn't it?  Can you attached kernel log linked to drm-tip branch use ? 
BTW please follow Jani's advice and keep us posted on your result.

yes, hdmi. and it's a long wire (to the projector). 

current status:

git bisect start
# good: [f8896f5d58e64bfd3c2b5f7c5ba5c3f3967e93c7] drm/i915/skl: Buffer translation improvements
git bisect good f8896f5d58e64bfd3c2b5f7c5ba5c3f3967e93c7
# bad: [ca6e4405779ed56ebac941570615abd667c72c02] Merge tag 'drm-intel-fixes-2015-07-15' into drm-intel-next-queued
git bisect bad ca6e4405779ed56ebac941570615abd667c72c02
# good: [f7b08217c755e88a29b5bd53b4a1d10cd8b3c5f8] Merge branch 'dmi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jdelvare/staging
git bisect good f7b08217c755e88a29b5bd53b4a1d10cd8b3c5f8
# skip: [23908db413eccd77084b09c9b0a4451dfb0524c0] Merge tag 'staging-4.2-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging
git bisect skip 23908db413eccd77084b09c9b0a4451dfb0524c0
# good: [22165fa79814e71e7a5974b3c37a5028ed16c8f9] Merge tag 'dm-4.2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm
git bisect good 22165fa79814e71e7a5974b3c37a5028ed16c8f9
# good: [e22619a29fcdb513b7bc020e84225bb3b5914259] Merge branch 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security
git bisect good e22619a29fcdb513b7bc020e84225bb3b5914259
# skip: [a611fb75d0517fce65f588cde94f80bb4052c6b2] Merge tag 'module-misc-v4.1-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux
git bisect skip a611fb75d0517fce65f588cde94f80bb4052c6b2

... window is shrinking, slowly ...
(skipping nonbootable and crashdumping kernels)

Created attachment 128919 [details]
dmesg: drm-tip

Are you referring dmesg? No problem to attach one. 

(btw. this log was created while trying hints from https://aboutsimon.com/blog/2016/07/20/Ubuntu-16.04-external-monitor-flickering-and-turning-off-on-intel-i915.html plus forced uxa in xorg.conf with no success)

(In reply to Michael Heide from comment #5)
> Created attachment 128919 [details]
> dmesg: drm-tip

We'll need a dmesg with drm.debug=0xe passed to the kernel. Otherwise there's not much useful infromation in the log.

Does the projector have an EDID? If so can you also attach that?

Created attachment 128930 [details]
dmesg with drm.debug=0xe

Created attachment 128931 [details]
hdmi2 monitor edid

Created attachment 128932 [details]
xrandr --verbose (human readable edid / modelines)

Status update: my bisect run has stalled so far... restart needed.

First there were good and bad kernels, besides a few unbootable builds. But later on there were only unbootable builds. First I thought "ok, someone has broken the kernel, find the commit when they fixed this". But then I've tried some commit which I had build successfully once before: now broken. 

It seems my local git database is broken, even cleaning the whole tree (git clean -df) is not enough. 

Stay tuned...

Sorry but I don't get it. 

ca6e4405779ed56ebac941570615abd667c72c02 and newer is bad while ccfb8b2ed4d4e12c3c35de3db5fbbbaa11277736 and olders (in that line) are good. 

But 8f539a83efa7dceb7c2257ca96e2dfc846bd12f6 - which is the other parent of ca6e4405779ed56ebac941570615abd667c72c02 besides ccfb8b2ed4d4e12c3c35de3db5fbbbaa11277736 - and all others I tested so far in that line do not boot at all. Grub loads initrd and kernel, then a black screen and (sometimes) a blinking cursor appears. 

What am I doing wrong?

I compiled all those commits and all of them do not boot into a running system:
97173eaf5f33b1e85efdb06d593d333480b60bf3 
e62925567c7926e78bc8ca976cde5c28ea265a49 
d04df7325ac9def8c4a68b49822c1d0a0c5379c6 
3e6da4a9d9005fbc83420686fb897bd8ad749e91 
eeebeac5e476800991ea1d10827307c41650ba11 
8f539a83efa7dceb7c2257ca96e2dfc846bd12f6 
ce65e47b78789b4f78be1fd7e4c884df74a9f075 346add7834557b5b9628b9bf2387106d42e631d4 
793dfa59bcfde9d642295480674926827e9adcfc 
05712c1561041d102d3cac7d7e79548b1446caa4 
f8896f5d58e64bfd3c2b5f7c5ba5c3f3967e93c7 

Are all those commits are indeed broken or am I missing something?

(btw: f8896f5d58e64bfd3c2b5f7c5ba5c3f3967e93c7 is one of those in the line of non-booting kernels. I don't get it, but either I have mistakenly used this as a "good" example in my initial bug report, or I indeed was able to successfully compile and boot it back then?)

What I do:
git clean -dfX # only if not a clean git clone
git checkout or bisect good/bad/skip
cp /boot/config-$(uname -r) .config
yes "" | make oldconfig
fakeroot make-kpkg -j5 --initrd --append-to-version=-sometag kernel-image
dpkg -i ...
reboot

(In reply to Ville Syrjala from comment #7)
> (In reply to Michael Heide from comment #5)
> > Created attachment 128919 [details]
> > dmesg: drm-tip
> 
> We'll need a dmesg with drm.debug=0xe passed to the kernel. Otherwise
> there's not much useful infromation in the log.
> 
> Does the projector have an EDID? If so can you also attach that?

Ville anything else you can do to help Michael?

Everything looks fine based on the log. We are now using 12bpc though whereas in the past we would have used 8bpc. So that could explain the change in behaviour. If the signal quality is marginal it's possible the problem would only manifest in the modes that require the highest bandwidth, which seems to be the case here.

Are you using a HDMI<->HDMI, DVI<->DVI or HDMI<->DVI cable? Any extra dongles/adaptors? Trying to eliminate extra dongles and whatnot might help, or the cable itself might not be of sufficiently high quality so trying another cable might help. Also if you're using the HDMI connector on the projector it might make sense to try out the DVI connection instead (AFAICS that projector should have both?). In theory it should not report 12bpc capability over DVI which would force us to pick 8bpc.

Still waiting on feedback from Michael about Ville's questions.  It still shows High/Critical so wondering if that priority still makes sense?

Assigning the bug to submitter to answer some questions from comment 14

Reporter, please comment or otherwise bug will be closed as no feedback.

(In reply to Jani Saarinen from comment #17)
> Reporter, please comment or otherwise bug will be closed as no feedback.

Hello, 
I'm changing the bug to CLOSE due to the lack of feedback. If there is any new information relevant in this case, please share it and mark as REOPEN. Thank you.