Bug 67311 - Display on Displayport does not work if plugged in at boot time
Summary: Display on Displayport does not work if plugged in at boot time
Status: RESOLVED CODE_FIX
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(DRI - Intel) (show other bugs)
Hardware: x86-64 Linux
: P1 normal
Assignee: Jesse Barnes
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-12-18 17:58 UTC by Philip Langdale
Modified: 2014-04-11 14:42 UTC (History)
3 users (show)

See Also:
Kernel Version: 3.12.5
Subsystem:
Regression: No
Bisected commit-id:


Attachments
Failure when turning on machine with U3011 plugged in (133.15 KB, text/plain)
2013-12-19 17:46 UTC, Philip Langdale
Details
Failure when turning on machine with U3011 not plugged in and plugged in after X starts (153.11 KB, text/plain)
2013-12-19 17:47 UTC, Philip Langdale
Details
Success when rebooting with U3011 off and turning it on after X starts (227.79 KB, text/plain)
2013-12-19 17:48 UTC, Philip Langdale
Details
Xorg log when things don't work (53.57 KB, text/plain)
2014-01-06 19:21 UTC, Philip Langdale
Details
Xorg log when things do work (68.12 KB, text/plain)
2014-01-06 19:22 UTC, Philip Langdale
Details
dmesg output with drm-intel-nightly for initial boot with no display (122.62 KB, text/plain)
2014-01-08 19:28 UTC, Philip Langdale
Details
dmesg output with drm-intel-nightly for initial boot with no display after vt switch back and forth (131.29 KB, text/plain)
2014-01-08 19:28 UTC, Philip Langdale
Details

Description Philip Langdale 2013-12-18 17:58:22 UTC
I'm running on a new Intel NUC d54250wyk.

If I boot the system with my Dell U3011 plugged in, I see a blank screen instead of X. Switching to a console VT looks fine, and then trying to switch back to X results in me seeing the desktop for a second, then the picture starts flickering and goes blank.

The following scenarios work fine:

1) I plug in my U2410 over HDMI and boot with that
2) I boot with no display attached and then plug in the U3011
3) I run with the U2410, then suspend, then resume with the U3011 plugged in.

So it seems very specific to this case of booting with the display present.
Comment 1 Jani Nikula 2013-12-19 09:15:49 UTC
(In reply to Philip Langdale from comment #0)
> I'm running on a new Intel NUC d54250wyk.
> 
> If I boot the system with my Dell U3011 plugged in, I see a blank screen
> instead of X. Switching to a console VT looks fine, and then trying to
> switch back to X results in me seeing the desktop for a second, then the
> picture starts flickering and goes blank.
> 
> The following scenarios work fine:
> 
> 1) I plug in my U2410 over HDMI and boot with that
> 2) I boot with no display attached and then plug in the U3011
> 3) I run with the U2410, then suspend, then resume with the U3011 plugged in.
> 
> So it seems very specific to this case of booting with the display present.

Please attach dmesgs from early boot, with drm.debug=0xe module parameter set, for both the reported problem *and* the working case #2. Thanks.
Comment 2 Jani Nikula 2013-12-19 09:20:41 UTC
Oh, which kernel version are you using? Please try at least 3.12 or later.
Comment 3 Philip Langdale 2013-12-19 16:19:20 UTC
Kernel 3.12.5 and also tested with 3.13.rc3. No difference.
Comment 4 Philip Langdale 2013-12-19 17:46:30 UTC
Created attachment 119051 [details]
Failure when turning on machine with U3011 plugged in
Comment 5 Philip Langdale 2013-12-19 17:47:17 UTC
Created attachment 119061 [details]
Failure when turning on machine with U3011 not plugged in and plugged in after X starts

Turns out that from a cold start, plugging in the display later doesn't help. I have to reboot with the display off and then turn it on after X starts for it to work.
Comment 6 Philip Langdale 2013-12-19 17:48:05 UTC
Created attachment 119071 [details]
Success when rebooting with U3011 off and turning it on after X starts
Comment 7 Jani Nikula 2013-12-20 06:38:12 UTC
Philip, please attach your Xorg.0.log.

Chris, please have a look at this. I fail to see anything wrong in the dmesg; I suspect userspace, but I suck at debugging there...
Comment 8 Philip Langdale 2013-12-20 18:11:51 UTC
The Xorg.0.log at default logging level trivially doesn't show any errors or warnings, so it'll be a waste of time for me to attach that. What debug level do you need?
Comment 9 Jani Nikula 2013-12-30 09:15:15 UTC
Please just attach the Xorg.0.log. Thanks.
Comment 10 Philip Langdale 2014-01-06 19:21:46 UTC
Created attachment 121111 [details]
Xorg log when things don't work
Comment 11 Philip Langdale 2014-01-06 19:22:05 UTC
Created attachment 121121 [details]
Xorg log when things do work
Comment 12 Philip Langdale 2014-01-06 19:22:15 UTC
Done
Comment 13 Daniel Vetter 2014-01-08 16:49:12 UTC
Can you please also test latest drm-intel-nightly from

http://cgit.freedesktop.org/~danvet/drm-intel/

Might be that we've improved the situation a bit meanwhile - your descriptions sounds an awful lot like there's some underrun.

Also, when you restore the display momentarily with a vt switch and the go back to X, are there any underruns reported in dmesg (dmesg | grep underrun, you need drm.debug=0xe enabled)? There's one at boot for the failing case ...
Comment 14 Philip Langdale 2014-01-08 19:28:02 UTC
Created attachment 121301 [details]
dmesg output with drm-intel-nightly for initial boot with no display
Comment 15 Philip Langdale 2014-01-08 19:28:23 UTC
Created attachment 121311 [details]
dmesg output with drm-intel-nightly for initial boot with no display after vt switch back and forth
Comment 16 Philip Langdale 2014-01-08 19:30:42 UTC
I tried running with drm-intel-nightly and attached dmesg logs for it. No change in the display behaviour.

You can see from the logs that there is no additional underrun reported after the VT switch with the brief flickering X display visible.

Also, drm-intel-nightly locked up all the time. When I tried to boot it normally with the hdmi display, it locked up after a few seconds at the desktop. In the displayport case it locked up when I tried to reboot - apparentling in the fb panning code. I have panic logs but that's a separate problem and perhaps you know what's going on there already.

Do you guys have access to one of these machines? It is an Intel product. :-)
Comment 17 Philip Langdale 2014-03-31 16:57:26 UTC
Appears to be fixed in 3.14 final, but I couldn't tell you why.
Comment 18 Daniel Vetter 2014-04-11 14:42:27 UTC
Hm, I guess we'll take that. Thanks for your report and please reopen if the issue resurfaces.

Note You need to log in before you can comment on or make changes to this bug.