[Bug 99334] New: [SKL] External screen not recognized on boot, disconnects during operation

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=99334

            Bug ID: 99334
           Summary: [SKL] External screen not recognized on boot,
                    disconnects during operation
           Product: DRI
           Version: unspecified
          Hardware: Other
                OS: All
            Status: NEW
          Severity: normal
          Priority: medium
         Component: DRM/Intel
          Assignee: intel-gfx-bugs at lists.freedesktop.org
          Reporter: post+fdo at ralfj.de
        QA Contact: intel-gfx-bugs at lists.freedesktop.org
                CC: intel-gfx-bugs at lists.freedesktop.org

Created attachment 128835
  --> https://bugs.freedesktop.org/attachment.cgi?id=128835&action=edit
dmesg captured with drm-tip

I am having trouble with an external monitor connected via HDMI.  This is with
an Acer Aspire VN7-592G.  The machine has an HDMI port and a Thunderbolt 3
port; all of this is with a screen directly connected to the HDMI port (and
nothing connected to Thunderbolt).  I am on Debian testing and using KDE.  The
trouble may not be specific to HDMI at all, that's just the only connector I
happen to be able to try right now.

The issue is as follows:  The external screen is not recognized properly when
it is plugged in during boot.  "xrandr" shows no external screen after logging
it.  I have to un-plug and re-plug the screen for it to be detected. 
Sometimes, I have to do this multiple times.  Only then it will show up in
xrandr and be recognized by the screen setup of my DE).  I then configure my
system to use only the external screen, i.e., the internal one is disabled.
  Furthermore, there are some things that may trigger the screen to suddenly
disconnect (so that the screen setup tool automatically re-enables the internal
screen):  Starting a game (tried this with neverball [both in windowed and
full-screen mode] and a proprietary game on Steam [full-screen]), starting
Steam (in the moment the Steam main window pops up), restarting kwin, logging
out.  In each case (potentially after logging back in), I can unplug and replug
the screen (potentially multiple times) and then everything works again.  If I
try this often enough (well, I tried with neverball as that is fastest to
start), at some point, it will stop making trouble -- the external screen will
stay connected with neverball started -- and from then on, starting Steam or
other, proprietary games also generally works.  Somehow, this fix is "sticky"
until I reboot.  Notice that this is not related to modesetting to happen; I
saw the disconnect with neverball in windowed mode or with the steam main
window.
  On drm-tip, when the external screen disconnects, it can also happen that the
X session dies.  I was unable to find any backtrace or anything in the X logs,
they just end with some modesetting stuff.  When I wanted to reproduce this to
capture the logs, it did not happen again.  The X session dying never happened
with the debian 4.8 kernel or with a vanilla 4.9.1 kernel.  
  Needless to say, this gets rather frustrating as it becomes hard to play a
game on the external screen.  The screen missing after boot is less of a
problem, I don't frequently reboot anyway.

I am honestly quite lost about how to even start diagnosing this.  I report
this bug against the Kernel component as I saw this behavior both with the
modesetting Xorg driver and with the intel Xorg driver -- and I can't imagine
this being the fault of OpenGL, but what do I know?

Attached you can find:

* dmesg-4.10:  This is a dmesg (via journalctl as "dmesg" would only output
parts of the logs) from drm-tip (9ea6a075).  I booted the machine, logged in,
unplugged and re-plugged the screen, then started neverball.  This triggered a
disconnect.
* xrandr-strange-verbose:  The output of "xrandr -q --verbose" after neverball
was started and the screen got disconnected on the Debian 4.8 kernel with the
modesetting Xorg driver.  This looks rather strange, there are some additional
modelines being printed after all screens have been listed; these modes seem
not to belong to any screen?  Notice that these are *not* modes of HDMI-2 which
is disconnected.  Furthermore, these modes are shown in "verbose" mode even
without the "--verbose" flag, as the attachment xrandr-strange (output of
"xrandr -q" caught at another occasion) shows.
* lspci: What lspci -vv has to say about my GPU.

Please let me know if you need any further information.  I would be very
grateful for any assistance.

-- 
You are receiving this mail because:
You are the assignee for the bug.
You are the QA Contact for the bug.
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20170109/f115c436/attachment.html>