Bug 21903

Created attachment 26156 [details]
dmesg

Created attachment 26157 [details]
Xorg.0.log after successful startup

Created attachment 26158 [details]
the used xorg.conf

Are you still seeing this with recent bits?

No more black screens with current master and different boot conditions. I will see during the next days if it will happen again, but in 10 (?) tests it looked good so far. As it seems to be fixed, I close the bug now as resolved/fixed. (if it occurs again, I will reopen this bug, but hopefully this will not be necessary)

Sorry. This was too early. After a freeze in fullscreen mode, I made a hard reset and there again came only a black screen when starting X.

Can you attach your current logs & dmesg from when the problem occurs?  Register dumps using intel_reg_dumper from the working and broken cases might also help...

I see a couple of issues in your last dmesg; looks like you're using kernel mode setting but also you've builtin or loaded FB_INTEL.

Also, recent kernels have many fixes relative to 2.6.29, so I'd recommend trying 2.6.30 or something from git... (preferably the drm-intel tree).

(In reply to comment #7)
> Can you attach your current logs & dmesg from when the problem occurs? 
> Register dumps using intel_reg_dumper from the working and broken cases might
> also help...
I will attach Xorg.0.log and the output of dmesg and intel_gpu_dump.

> I see a couple of issues in your last dmesg; looks like you're using kernel
> mode setting but also you've builtin or loaded FB_INTEL.

Now using KMS without intelfb.

> Also, recent kernels have many fixes relative to 2.6.29, so I'd recommend
> trying 2.6.30 or something from git... (preferably the drm-intel tree).

Updated to 2.6.30, mesa 7.5_rc4 and latest master of xf86-video-intel.

Created attachment 27372 [details]
Xorg.0.log after a failed start

Created attachment 27373 [details]
dmesg

Created attachment 27374 [details]
bzip2 compressed dump of intel_gpu_dump right after a failed start of X

Created attachment 28881 [details]
gpu dump from the working case

Sorry, I had forgotten that one.

This is the result of
- cold boot, with automatically running (Gentoo) xdm init script, but failing
- switching to VT1
- login as root
- /etc/init.d/xdm restart
- switching again to VT1
- intel_gpu_dump with intel-gpu-tools 1.0.1

Created attachment 28882 [details]
gpu dump from the broken case with git snapshot

This is the result of
- cold boot, with automatically running (Gentoo) xdm init script, but failing
- switching to VT1
- intel_gpu_dump with intel-gpu-tools 1.0.1

the dumps of the working case above and this one were done with the same bits of xf86-video-intel (git snapshot from yesterday)


Please let me know if any further information is missing.

This one has me stumped...  If X crashed or your session ended early somehow I can imagine how this might happen (there's kernel code to restore the framebuffer console when the last app exits, but it was broken until recently and might have restored an empty console on configs like yours).

Did you happen to capture register dumps as well?  The GPU dumps might be helpful if this was a hang, but it doesn't sound like that's what's happening here.

I'm hoping something else is wrong with your desktop environment that causes X to exit early, but I don't see any evidence of a crash in your logs (maybe your X session logs might have something?).

Any update here?  Register dumps might help, but your logs look good so it seems like something else might be wrong with your environment that leads to X exiting.

Sorry, I did not find the time for further investigations, yet.

Also, the problem did not occur any more since several weeks. Maybe the problem is gone (or maybe it is only a coincidence).

This time I am pretty sure this bug is definitely gone. I do not know what was the cause, but I could not reproduce it for several months now. Probably it has been fixed with one of the hundreds of updates since October.

Thank you for your assistance. Closing again as RESOLVED/FIXED