X crashes and system is only responsive via sys-req on keyboard

[Liam O'Reilly]

I have tried to report this bug at https://bugs.launchpad.net/ubuntu/+source/linux/+bug/341363 where I was told to report a separate bug.

I am running Jaunty (fully up-to-date as of 13th July 2009) 64 bit version. My kern.log file shows alot of
[drm:i915_getparam] *ERROR* Unknown parameter 6
messages just before my system crashes.

My system is crashing in this way (on average) at least once a day at the moment. I believe these messages in the kern.log file are the last messages produced before such a crash.

When such a crash happens the graphics system is totally unresponsive, alt-ctr-backspace twice does not restart x (dont-zap disabled). When I try to get to a virtual terminal say Ctr-F1 - nothing happens. I can use sys-req to restart the system. I can ssh in to the machine and sound etc still works. Trying to restart gdm via /etc/init.d/gdm restart fails. In the end I restart using sys-req.

This bug is making Ubuntu totally unusable right now. I am pretty sure it due to the Intel graphic driver. I will attach logs of various command in the next posts.

[Liam O'Reilly (aliam13-2)]

I will attach a report of dmesg after a reboot of the crash (that I guess will happen later today).

This crash seems to happen more the more graphic intensive applications I run, i.e., compiz / OpenGL seem to cause more crashes. Also the crash seems to happen a lot when opening Firefox, but that may just be because I open Firefox a lot. I have no way of forcing a crash.

[ZelinskiyIS (ivze)]

All the symptoms, described by Liam O'Reilly, are the same with my netbook, except the crashes are more rare. The crashes are less frequent when using Ubuntu "Standart" instead of "Extra" desktop effects. The device, affected, is Acer Aspire One ZG5.

# uname -a
Linux ubuntu-laptop 2.6.28-14-generic #47-Ubuntu SMP Sat Jul 25 00:28:35 UTC 2009 i686 GNU/Linux
# cat /proc/version_signature
Ubuntu 2.6.28-14.47-generic

[ZelinskiyIS (ivze)]

...and to clarify, Jaunty 32bit is running on the netbook. (This could, probably, be discovered by reading the previous post, but this way the whole looks more clear).

[MrAuer (mr-auer)]

I am also now waiting for my Eee 901 to crash again, now I have X server and display driver debug symbols installed and Apport running. I hope it crashes as often as yesterday now that I can hope to trace it :)

On this machine, the crashes happen more often with XFCE than Gnome (I have tried bot Xubuntu and regular Ubuntu on it) - even when I do not have XFCEs own compositing, nor Compiz, on. Under Gnome using Compiz perhaps made the crashes more common - and under Gnome, I often get a hard lock instead of just X crashing and restarting, like under XFCE.

I will report here when I get more data.

[Geir Ove Myhr (gomyhr)]

Liam, could you check if this bug also is present in Karmic and if it is post the log files from when it crashes? I understand that you may have to run Karmic for a while to see if it is still there, but we need to know it is a problem with a recent version of the linux kernel and the driver to get upstream's attention and we also would like the ubuntu developers to spend their time on fixing bugs that will anyway be gone in Karmic.

You other guys, from your comments I understand that this is a relatively common problem. We would rather that you don't post your logs here, though. This is because some of the details in your problem could be slightly different and looking at the log files together makes it harder to keep issues apart and report upstream. What you can do is file a new bug report, preferably with the ubuntu-bug command on Karmic, and indicate that you think your bug may be the same as this. If one of your reports turn out have better log information we can mark the other bugs as duplicates, and if something turns out to be different we can remove the duplicate mark later. Either way, we keep the log files from each person/computer on a separate bug report and that is necessary to keep things manageable for the ubuntu developers and to be able to upstream the issue efficiently.

I know Launchpad during the bug reporting really discourages filing new bug reports if you can find a similar issue. This is unfortunate for xorg-related packages where we much prefer separate bug reports, but is probably a fact we have to live with.

[Liam O'Reilly (aliam13-2)]

Here is the dmesg report that I promised above (a little late, but here now). I found when trying to cause this problem that, that I simply cannot reproduce the problem. It happens at least once a day, but I have no idea what triggers it. It seems to happen a lot when I interface with Firefox, but this could be coincidence as I use Firefox often.

I will now try to do my daily work on a Ubuntu 9.10 64 bit Alpha 4 live cd and see if the bug is triggered, If so I will try to get a backtrace etc (using instructions at https://wiki.ubuntu.com/X/Backtracing).

[MrAuer (mr-auer)]

OK..
I have now tried running apport, but so far it has not generated any reports, even thou X has crashed twice. First time hard lock, had to power down. Second time, dropped to login prompt again, but still no report..

Completely random still, like with Liam O'Reilly. I have found a few other people (on IRC) who have the same crash behaviour also on Eee pc 901, with the intel 945g chipset - one of these people used Debian (testing I think) instead of Ubuntu. They have had that present also at least during Jaunty cycle, I had these crashes happen also on Intrepid (reason I upgraded).

I downloaded Karmic alpha 4, I will put it on a USB stick today and try it out. Alpha 3 crashed constantly and had issues with display going dark for a couple seconds at time, I will see if those have been resolved.

[Bryce Harrington (bryce)]

(Setting to incomplete while waiting on a full backtrace)

[Liam O'Reilly (aliam13-2)]

Please note that unlike MrAuer I did not have this problem on Intrepid. The problems only started to happen on Jaunty for me (although I don't seem to remember them happening when I first upgraded and the few weeks following).

[Liam O'Reilly (aliam13-2)]

I noticed a vital clue which I missed before.

My mouse is responsive, but no clicks are registered (as if I never clicked), I can move the mouse around the screen. The keyboard is unresponsive apart from sys-req. I did not think to check the caps lock led. I will do this next time the crash happens.

Here is what I think is a full backtrace (I don't really know what one of these should look like). I followed the instructions at https://wiki.ubuntu.com/X/Backtracing). I found Xorg's pid then started gdb, attached to xorg's pid and run the gdb command "backtrace full".

I hope this helps. If not then please let me know how I can help more.

[Liam O'Reilly (aliam13-2)]

As I have provided what I think is a full backtrace

[Liam O'Reilly (aliam13-2)]

I just remembered that I have a modified xorg.conf file. I have changed my AccelMethod to UXA (Ubuntu Juanty 64 bit). This was because under the default setting, the terminals in gnome were unusable for me as the scrolling speed was too slow. Using UXA the terminals are normal and much faster.

[Bryce Harrington (bryce)]

Ahhh yes, we specifically did not ship Jaunty with UXA enabled due to people finding lots of crash and freeze bugs (see https://wiki.ubuntu.com/X/UxaTesting). We provided UXA as an option for testers but do not support it on Jaunty.

As mentioned before, if you can reproduce it on Karmic (where UXA works much more solidly, and where we do support it) feel free to reopen.

[MrAuer (mr-auer)]

:(
Does this mean I need to open a new bug report again?
I do not have UXA enabled, and I am still getting these crashes and hard locks on Jaunty.
I dont know why Apport does not generate a report upon crash, though - except when it hard locks - I guess nothing gets done at that point as even the kernel is unresponsive to SysRq key combos.

I am still trying to get a usable backtrace somehow (I am running Apport in the background all the time) but haven't yet had the time to setup ssh on this machine and wait for it to crash. In the hard lock case I don't know if I would even be able to log in after the fact.

[Liam O'Reilly (aliam13-2)]

Dear MrAuer, unfortunately, you should open a new bug report. Since I have disabled UXA then my Jaunty system seems to be fine - that is there have been no more crashes yet. I have been using the alpha 4 of Karmic at the moment to see if my bug is present there - as I feel finding it there is more important that finding if UXA was the problem on Jaunty. So far Karmic seems to be fine, with my bug not appearing. I think this bug can be marked as invalid, unless I find an occurrence of this on Jaunty with no UXA or Karmic.

[MrAuer (mr-auer)]

I should get Karmic alpha 4 installed on my memory card as well..
I wasnt able to get it on a USB stick using unetbootin, I guess it was because Karmic used GRUB2.
Ill have to look into it and try running Karmic (I guess you could run it off a CD and then make a bootable USB via Karmic itself). I wouldnt mind upgrading to that if it just works without crashing.

[Syj (jsaintyv)]

We meet the same issue on Asus EEE Top. We upgrade 8.04.1, which worked fine, to 9.04.

Since the X freeze after GDM authentication. All the system is freeze. No more answer to ping and ssh connection. No numlock activity.

I verify UXA, it was not enable.

X work one time with twm but it freeze when i close the session to go back console.

[Geir Ove Myhr (gomyhr)]

Syj, if you read the comments above, you will see that we are interested in freeze reports from Karmic, and that it is too late to fix anything in Jaunty now. If you still have your problem on an updated Karmic, please report it as a new bug. This bug has been closed, since the original reporter (Liam) does not experience this on Karmic.

[bomanizer (janne-matolaatikko)]

I suspect this bug on my T61 running Linux Mint 7 (Gloria). Is there any chance that this will be fixed for Jaunty?

[bomanizer (janne-matolaatikko)]

I don't know if this is of interest anymore, but I'm confident that I can 100% reproduce the crash.