WORKAROUND - Nvidia 260.19.29 + Xorg-Server 1.9.2 Crash X11

[luispa]

Hi, after it became stable I upgrade both nvidia and xorg-server

[luispa]

Marking it as WORKAROUND,

Please share if you hit this one and find a solution.

Thanks
Luis

[michaelhallquist]

Hi Luis,

I have virtually the same problem, but with the ATI drivers. I have a Radeon HD 2400 XT (rv610 board). I was also running xorg-server 1.7.7 happily and when I upgraded to xorg-server 1.9.2 in late January, the same issue occurred: a couple of minutes after VMWare Workstation 7.1.3 (Win 7 guest) started, all of X11 crashed. I wasn't able to find any evidence of a backtrace in my Xorg, gdm, or messages log files. Just this error showed up when gdm restarted moments later:

Feb 17 09:02:35 macgyver polkitd(authority=local): Unregistered Authentication Agent for session /org/freedesktop/ConsoleKit/Session5 (system bus name :1.33, object path /org/gnome/PolicyKit1/AuthenticationAgent, locale en_US.UTF- (disconnected from bus)
Feb 17 09:02:36 macgyver kernel: [drm:drm_mode_getfb] *ERROR* invalid framebuffer id

Relevant packages under the failing installation include:

x11-drivers/radeon-ucode-20110106
x11-drivers/xf86-input-evdev-2.6.0
x11-drivers/xf86-video-ati-6.14.0
x11-base/xorg-drivers-1.9
x11-base/xorg-server-1.9.4
media-libs/mesa-7.9.1

I came across this Gentoo bug concerning gallium vs. classic drivers:

https://bugs.gentoo.org/show_bug.cgi?id=350280

But setting my r300 or r600 driver to gallium or classic has no effect. As you described, Luis, the X11 crash also occurs independent of VMWare, but much more sporadically (e.g., once per hour, as opposed to within a few minutes of starting VMWare). My sense was that Firefox might have increased the likelihood of crash, but that might just be superstitious.

My workaround is the same as yours: I downgraded to xorg-server-1.8.2 (or 1.7.7, which continues to work, but I wanted to do away with HAL on my system). Everything works properly with xorg-server-1.8.2 (and xorg-drivers-1.. No other package downgrades are necessary (e.g., mesa or the ati drivers).

Does anyone have suggestions about how to debug this problem? I haven't succeeded in getting backtraces or core dumps from X even after enabling splitdebug in make.conf.

Thanks,
Michael

[bpkent]

[luispa]

Thanks "michaelhallquist" for detecting exactly the same with ATI, so looks like not a driver problem.

Today I upgrade serveral packages and didn't realized xorg-server got upgraded up to x11-base/xorg-server-1.9.4 (had masked 1.9.2), so the problem appeared again. This time is even before than expected. After login in, two minutes later, whole X crash by itself.

My problem is that xorg-server 1.7.7 is masked, so can't go back, have to stay at 1.9.x. And it's unusable...

[chris...]

Im having periodic pausing with nvidia260 and xorg1.9
It gets very annoying at times

Using an older nvidia driver makes it more user friendly

[luispa]

I guess I've been able to solve it finally. Don't know where the problem was, though by doing the following two things now X doesn't crash eventually, at least it's been working for 6 hours without problems.

Now, for your infomration, I've tried multiple things without any succeed, just for your information:
- Recompiled the whole system world
- Tried Gnome KDE
- Tried Nvidia and Nouveau
- Tried compiz / no-compiz in all kind of combinations with above
None of the above solved the problem

What I did:

1) First I'm using this new file with xorg-server 1.9.x, I'm not sure if it has anything to do, anyway, was one of my changes.

[michaelhallquist]

Hi Luis,

I'm frustrated to say that the fixes that worked on your machine had no effect on mine (xorg-server 1.9.4)! I applied the changes you suggested separately: evdev for keyboard, evdev for mouse, mv $HOME/.config directory. No single change fixed the problem. I then applied the changes sequentially, culiminating in all three being present at once. Unfortunately, the X server crashed as usual in VMWare (and elsewhere if I wait long enough). I also tried tweaking the system evdev config file: /usr/share/X11/xorg.conf.d/10-evdev.conf. This seemed to contain redundant evdev trapping for keyboard and mouse relative to the keyboard config file you mentioned. Also, in your config file, you had used /dev/ipput/evdev* as the match for mouse, but at least on my machine, it is /dev/input/event*.

The fact that your fixes seem to depend on evdev is interesting... perhaps it's not a video problem, but has to do with the keyboard/mouse devices. Would you mind posting the Xorg.log from your working configuration so that I can see if your machine is handling things differently? Also, is there any chance one of the other things you tried along the way is necessary for this fix to work?

Thanks,
Michael

[luispa]

Hi, this is my xorg.conf

[krinn]

i don't think your title is more accurate like that: using vmware is the main problem i saw here, and now your title suggest xorg 1.9.2 doesn't work with nvidia-drivers-260.19.29

[luispa]

Thanks for the input, however my main original problem was a X crash and so I renamed the thread.

The problem is not due to VMWare, but xorg-server 1.9, being the main root cause still something unknown, but very well identified under one or more of the files under $HOME/.config.

VMWare Workstation is an application that was accelerating the crash of xorg, but no the root cause and so I got confused. After researching I was able to reproduce the crash very easely without even starting VMWare.

Notice, though, that we may have hitted two different problems, as "michaelhallquist" seems to keep crashing on VMWare, and doesn't seem to crash X when vmware is not running, or at least that's what I understood.

luis

[michaelhallquist]

Hi Luis,

I'm happy to report that my problem with Vmware workstation (and other applications like Firefox) crashing X is finally resolved! I believe the problems on both of our machines were identical. X (xorg-server 1.9 or higher) would crash with little evidence in the log file particularly when Vmware was running, but this also occurred for other X apps. The problem was much more prevalent with Vmware. I was able to trace the issue specifically to the xorg-server package. Simply downgrading from xorg-server 1.9 to xorg-server 1.8.2 stopped the problem from occurring. But because 1.8.2 has been deprecated and removed from the Gentoo tree, I had to download the ebuilds from the Gentoo archives and use a local overlay (plus masking 1.9 packages) to get this working. (not optimal)

As for how I fixed the problem, it's somewhat difficult to say given the number of changes. Deleting the $HOME/.config file alone was insufficient to correct my problem. Here are the steps I performed that jointly seem to have corrected the issue (although I remain uncertain which step is critical).

1) Upgraded my kernel to 2.6.38-r4 with kernel modesetting (KMS), radeon r600 firmware blobs compiled into the kernel, and with X direct rendering manager (DRM) turned on.
2) Removed my old $HOME/.config directory
3) Installed Vmware workstation 7.1.4 (vmware-modules are version 238.4)
4) Upgraded to xorg-server-1.9.5
5) X11 driver packages are:
x11-drivers/radeon-ucode-20110302
x11-drivers/xf86-input-evdev-2.6.0
x11-drivers/xf86-video-ati-6.14.1

The combination of these changes has fixed the crashing issue. X, Vmware workstation, and other previously problematic applications including Firefox have been running for 72 hours with no crashes!

Anyway, thanks for your leads, Luis. And I'm glad we've both resolved this issue!

Best,
Michael