Created attachment 20247 [details]
Xorg.0.log piped through uniq (it was 22MB before)
The external vga monitor often stays blank after resume from suspend to ram (s3). Sometimes its filled with a single color, like plain blue. I have to switch back and forth between the console and X several times until it works again.
While the display is blank this is flooded in the Xorg log file:
(EE) intel(0): underrun on pipe A!
I will attach the complete log and my config file plus lspci.
The problem exists since at least Ubuntu 8.04 with much older drivers. It also exists with the intel driver coming with 8.10.
xorg version: 7.4~5ubuntu3
intel driver: 1.5.0
Created attachment 20248 [details]
Created attachment 20249 [details]
Created attachment 20250 [details]
btw. The kernel is 2.6.27 vanilla and has intelfb compiled-in (but not active)
Syslog just told me:
Nov 12 16:57:28 xxx acpid: client has disconnected
after the second display goes on again.
Created attachment 20334 [details]
Xorg.0.log with intel 2.5.1 (uncut)
The problem has for the first time now come up on the first display init with version 2.5.1.
[ 37.499292] [drm:i915_getparam] *ERROR* Unknown parameter 5
Created attachment 20335 [details]
dmesg with intel 2.5.1
I've just confirmed there's a regression between 2.5.0 and 2.5.1. My external display does not work anymore on first start, but only after switching (Ctl-Alt-F1, Alt-F7) three times. Kernel 2.6.27, dmesg with intel 2.5.1 attached.
Created attachment 20336 [details]
dmesg with 2.6.28-rc4 and 2.5.1
With the combination of kernel 2.6.28-rc4 and driver 2.5.1 the display *is* initialized properly the first time (but there's a warning with ioremap (dmesg attached).
Can you attach the output of intel_reg_dumper from when you see one of the displays fail to come up? The pipe underrun errors in some of your logs indicate a serious bug in the way we program the outputs on your system...
The ioremap warning should be harmless iirc, I think it'll be fixed for the final .28 release.
Sorry I didn't update for a while. The problems disappeared some time before 2.6.28rc7. Do you want me to check out rc4 and then post the output of intel_reg_dumper? As long as you don't ask for a bisect...
Created attachment 21286 [details]
kernel warning with 2.5.1 and 2.6.28-rc7
Well, to be honest there's still a warning when using the combination of 2.6.28-rc7 and 2.5.1. (see attachment)(not with 2.5.0). But as you said that shouldn't be a problem...
Ok, great. Thanks for confirming.
By the way, the oops is still there in 2.6.28-rc9...
see http://bugzilla.kernel.org/show_bug.cgi?id=12267 for the kernel bug.
Correction: The warning actually first appeared after 2.6.28-rc7 in the combination with intel driver 2.5.0. It used to be only with 2.5.1 and 2.6.28-rcX.