Created attachment 24847 [details]
distribution: debian gnu/linux sid/experimental with amd64
kernel version: 2.6.30-rc2
on a Dell Latitude D820 using LVDS (no other screens attached).
xf86-video-intel: git commit 2e3b95ed0197971e81ab7509245c899e96859d5b
xserver: 1.6.1 (debian sid/experimental)
mesa: 7.4-2 (debian sid/experimental)
Easy to reproduce here by just trying to start the xserver with UXA enabled. I
get a black screen and i'm unable to change back to a VT anymore. Just got the
log by syncing/killing/rebooting using sysrq. sysrq+g doesn't work either.
"buscher" in IRC seems to have the same problem on the MacBook 2, with the
difference that it works with EXA there. Unfortunately I get the same result
with UXA. For me it doesn't make any difference if using with and/or without
PAT and Tiling leading always to that black screen.
here is the xorg.conf:
Identifier "Configured Video Device"
Option "AccelMethod" "UXA"
Option "ModeDebug" "yes"
dmesg and Xorg.0.log are attached.
Created attachment 24848 [details]
Created attachment 24971 [details]
register dump before xserver start
Created attachment 24972 [details]
register dump after xserver start
Created attachment 25005 [details]
GPU dump after xserver start
it works when the drm stuff isn't built-in into the kernel and modprobed later.
my first guess was that it may have problems with the agpart module which may
be loaded afther drm, but i checked and that wasn't the case.
I think we went over this on IRC but now I've forgotten. :/
I think the builtin vs. modprobed issue should be gone now, and if not we should fix it, but I'm not sure that's behind this problem.
We recently found that some distros run vbetool at bootup, which was causing interrupts to get disabled, preventing any rendering from completing. So that could be one cause of this bug.
If that's not it, can you get a backtrace of X when it displays the blank screen? It's probably stuck trying to render into the front buffer, but I'd like to get that confirmed. Also, please copy out the contents of the debugfs files and attach them here during the failure.
# mount -t debugfs nodev /sys/kernel/debug
# cd /sys/kernel/debug/dri/0
# mkdir /tmp/gpu-debug; cp * /tmp/gpu-debug
I added myself to the CC list because I had the same issue (read: same symptoms), with a "Mobile Intel GMA 4500MHD" (8086:2a43). Linux 220.127.116.11 (vanilla) & driver 2.6.1 (from Debian)
I had all graphics stuff as modules so I don't think it matters.
KMS works since I upgraded to 18.104.22.168 + 2.7.0. Now, I even have i915 buitin.
Works for me now (tested with 2.6.30-rc5) and thus closed :-)