[Intel-gfx] drm/i915 X crashes on 835GM (IBM R31)

[Thomas Richter]

Hi Chris, hi Daniel,
> A full drm.debug=6 dmesg, Xorg.0.log and xrandr --verbose would be
> useful starting point to distangling what is going on. Is this an 830gm
> or an 845g? google says 830gm, so there should be 2 crtcs available for
> you to drive the two outputs independently, and we need to find out what
> is going wrong.
> -Chris
>
You are right, this is a 830GM as far as I know. I'm attaching the 
dmesg.out with drm.debug=255 (hopefully all the data) with an external
monitor attached. I'm also attaching the stderr and stdout of X - but it 
doesn't do much except crashing. If I'm attaching the external monitor
later when X is running, I do get a working X, but no picture on the 
external monitor. The xrandr output is also attached.

I also checked the bios: The bios is right now configured to create only 
a display on the external monitor. Consequently, if I boot with the 
external monitor connected, the boot screen and grub appear there, but 
during bootstrap, it seems like the drm code grabs the display pipe for 
its internal display.
X is "flickering" when panning in this mode.

Things are working better if I switch the bios to "both" such that 
external and internal monitor are activated. I then get a flickering 
display on the external monitor, and a working display on the internal 
during bootstrap in the console with drm loaded, and a working display 
with X. I *also* get a working panning without flicker on the internal 
display, but flicker when panning on the VGA output.

Thus, apparently:

*) There is at least one bug related to X which let it crash if it 
cannot find the pipe for the internal display or only one pipe is available.
*) Something is wrong with the initial drm setup when an external 
monitor and an internal monitor is connected. The external display is 
not usable.
*) Panning through the internal LVDS interface seems to be ok and 
working, but the external VGA pipe does not work well for panning. 
Interestingly,
if you compare that with the results from the Fujitsu Lifebook (using 
essentially the same graphics chipset, but the internal display is 
connected via
the DVO and not via LVDS) something's wrong with non-LVDS output somehow.
*) It looks as if the bios would disable the internal pipe (Pipe B I 
assume) and the drm kernel module would probably then grab pipe A for 
the internal
display, causing the problem with panning. X seems to be very confused 
about this setup.

HTHH,
     Thomas

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: dmesg.out
URL: <http://lists.freedesktop.org/archives/intel-gfx/attachments/20131022/5f2397eb/attachment.ksh>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: x.log
Type: text/x-log
Size: 1769 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/intel-gfx/attachments/20131022/5f2397eb/attachment.bin>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: xrandr.verbose
URL: <http://lists.freedesktop.org/archives/intel-gfx/attachments/20131022/5f2397eb/attachment-0001.ksh>