|
Bugzilla – Full Text Bug Listing |
| Summary: | nvidia related X server crash | ||
|---|---|---|---|
| Product: | [openSUSE] SUSE LINUX 10.0 | Reporter: | Richard Brown <richardbb> |
| Component: | X11 3rd Party | Assignee: | Stefan Dirsch <sndirsch> |
| Status: | RESOLVED WONTFIX | QA Contact: | E-mail List <qa-bugs> |
| Severity: | Normal | ||
| Priority: | P5 - None | CC: | aritger, eich |
| Version: | Final | ||
| Target Milestone: | --- | ||
| Hardware: | x86 | ||
| OS: | SuSE Linux 10.0 | ||
| Whiteboard: | |||
| Found By: | Other | Services Priority: | |
| Business Priority: | Blocker: | --- | |
| Marketing QA Status: | --- | IT Deployment: | --- |
|
Description
Richard Brown
2005-12-11 14:06:32 UTC
Many of these oopses have krb5_decrypt in them. Are you using NFS with RPCSEC_GSS and kerberos? If so, do these crashes disappear if you disable it? (In reply to comment #1) > Many of these oopses have krb5_decrypt in them. Are you using NFS with > RPCSEC_GSS and kerberos? If so, do these crashes disappear if you disable > it? > Sorry, but I'm not sure what all these are. However: I checked Yast2 NFS configuration, and the NFS Server Do Not Start option is selected. Selecting Kerberos Client in Yast2 results in it telling me that installation of pam_krb5 and krb5-client is required. Is this relevant to what you're asking? The motherboard has a Marvell Gigabit LAN controller chipset, but this is unused. (The machine is not networked). Hope this helps. On perhaps a different tack I am currently using fluxbox instead of KDE, and the crashes are much less frequent. Looking at the oopses more closely I notice you're using the nvidia module. I strongly suspect this is the main culprit. Can you please try running with the stock nv driver and see if the problem persists? Right. I ran the NVIDIA uninstaller program, and it told me that there was no NVIDIA driver installed! Also I found that some software that needs 3D hardware acceleration tells me that it can't find any. What is yast doing?! So, instead I just changed the driver to "nv" in Xorg.conf. At the moment it's working, and so far so good; no crash yet. I think if I try the NVIDIA driver again I'll use its own installer. Incidentally, down amongst the oopses is a mention of ac97. The terratec soundcard doesn't use this. It's only for onboard sound, isn't it? which is disabled. richard@linux:~> dmesg | grep ac97 ALSA sound/pci/ac97/ac97_codec.c:1959: AC'97 0 does not respond - RESET ALSA sound/pci/ac97/ac97_codec.c:1968: AC'97 0 access is not valid [0xffffffff], removing mixer. ice1712: cannot initialize ac97 for consumer, skipped I'll let you know if the nv switch remains crash-free. Richard. Hi again, It's much more stable now running with the nv driver, but I have had one lock-up since the last posting. I was running firefox with multiple tabs, while connected to the net. X died and I ended up at a text page with a login prompt. The display had broken green lines across it. The machine had to be reset. I could find no reports of this in the logs. What's the story with the nvidia driver? What makes you suspect it? Hi Richard, The nvidia driver is known to mess with a lot of things - including the memory subsystem. And since it is closed source, we cannot support configurations with this driver loaded. The crashes you reported just confirmed that it is always a good idea to see if kernel crashes go away by removing all closed source third party drivers. Please report this issue to Nvidia. Concerning the crash you reported in comment #5, this sounds like an X server bug, not a kernel bug. When the X server crashes, it will leave the video hardware in a funky state. I'm re-assigning this report to the X people. Please attach the results of "nvidia-bug-report.sh". Thanks. Sorry, but where is nvidia-bug-report.sh? updatedb followed by locate can't find it. The same day as my last comment, X died. I reinstalled the nvidia driver (and tried to at least get X running again. It didn't work.) and then uninstalled it with its installer, and then removed the glx module from Xorg.conf. Leaving /usr/X11R6/lib/modules/drivers/nvidia_drv.o intact and running Xorg -configure results in a could not load module fatal error. Removing it and running Xorg -configure results in "Could not init font path element" errors, and "fatal server error could not open default font 'fixed'", or "cursor". There is much wailing and gnashing of gears here at the moment. Richard nvidia-bug-report.sh is part of the nvidia driver installation. I'm afraid I can't help you much in your now messed up system. :-( Hopefully at least you'll manage to get at least a nv driver ocnfiguration running again. Finally got X up and running again. The thing that's bugging (sic) me is that the last crash WAS with the nv driver. So there won't be an nvidia-bug-report. Sorry. I wish I'd known about that script. It appears that when X crashed, it corrupted a number of font files and XKB related stuff. So I ripped them out and reinstalled/updated in yast. Is it true that when X crashes it dumps stuff into a cache which it hopes will get written to disk? If the machine is locked up, how does it expect this stuff to get written? The Xserver does nothing special when writing to its logfile. Anyway, since you can't provide the output of nvidia-bug-report.sh me (and NVIDIA) can't investigate this issue. :-( I'm sorry! (--> WONTFIX) |