7
Jan

Troubleshooting Nvidia TwinView on Ubuntu Linux

  • Any X experts out there? I need help setting up Nvidia TwinView on an Ubuntu Linux machine. PLEASE NOTE: In order for me to pay the big bucks, my machine must reliably boot from a Linux distribution with TwinView working!!! That is, a single desktop across two monitors. My motherboard is an ASUS A8N32-SLI, my video card is an ASUS EN7800 GTX. I have 64-bit Ubuntu Breezy installed. Here's the fun part - the Nvidia 1.0.8756 driver installs successfully. I then let nvidia-configx run. No errors. Then if I do startx, X starts with TwinView working! Great! Then when I reboot, X fails. I've edited the xorg.conf files many times to try to eliminate errors, but I must still be missing something. Since TwinView works after the driver is installed, the hardware should be OK. My current xorg.conf is here: http://levelofdetail.com/errs/xorg.conf My current xorg.0.log is here: http://levelofdetail.com/errs/Xorg.0.log These are only the current versions. I've tried many other versions before these. I'll take whatever advice you give and report back results until something works. I assume this will take a few exchanges back and forth. This is a new system install, so there is no data to lose. I'd like to stick with Ubuntu or at least some other Debian-based distro. My hope is that I'm just missing some small thing in the xorg.conf file. But I'm completely lost as to how this could work on installation and then not work after rebooting. (I'm out for the rest of today. Will check back first thing tomorrow.) Thanks, nevada2244-ga


  • Hello Nevada2244, Did you declare success? If so, I would like to prepare a proper answer. If not, please describe the current situation so I can help you correct it. --Maniac


  • Hello Nevada2224, It is good that progress has been made. Let me comment first on the new questions / comments and then I'll follow up with a few more things to check. [1] Loading NV-GLX, etc. This appears to be a side effect of the NVIDIA module loading. The documentation I looked at on Nvidia's site talks about loading of other modules, etc. As to why it still says it is using this capability is beyond me [may be a bug...]. [2] More than one... The configuration file refers to Device "NVIDIA Corporation NVIDIA Default Card" twice (one for each screen). It is an info message so its MAY be OK. Searching for that phrase on various mailing lists doesn't bring up much of a hint on why you would get that. Since you are using Twinview, it might not be necessary to actually specify both screens - see http://download.nvidia.com/XFree86/Linux-x86_64/1.0-8762/README/appendix-g.html for details. See also http://download.nvidia.com/XFree86/Linux-x86_64/1.0-8762/README/appendix-p.html which describes the set up of two X displays on the single card (not using Twinview). [3] Other comments. The error messages [those marked with (EE)] indicate GLX is not loaded and the driver considers this an error. I suggest checking that all files are the Nvidia ones as described at http://www.nvidia.com/object/linux_readme_install.html (scroll down to appendix C) or more directly at http://download.nvidia.com/XFree86/Linux-x86_64/1.0-8762/README/appendix-c.html which describes the files, links and specific versions and locations. If you have to fix something, try enabling GLX again to see if it is completely fixed. [4] Xinerama messages... Again, based on the Twinview section of the README http://download.nvidia.com/XFree86/Linux-x86_64/1.0-8762/README/appendix-g.html (search for Xinerama) it mentions that the Xinerama interface is used to allow the window manager (and other applications) find out the edges of the two displays. It looks like this is "normal behavior". --- I suggest you briefly view http://www.linuxforums.org/forum/linux-desktop-x-windows/34147-nvidia-driver-wont-work-glx-xorg-gentoo.html where another user has a similar problem / uses the "disable GLX" solution to make it work. This was over a year ago and I find it odd that you have the same problem with the updated driver. As a side note - in addition to the reboot, I suggest doing it twice: - once with power off / on - once without power off / on to make sure the configuration works on a completely uninitialized card (after power off) as well as the indeterminate state the card is in when you reboot. --Maniac


  • Hello Nevada2224, Hmm. I noticed you removed the old error log and configuration files. I wanted to do a comparison but if they are not available, we may be able to make progress without them. [I checked my browser cache as well - they were already removed] I noticed something that may be significant in the new log file. At the top, it indicates X Protocol Version 11, Revision 0, Release 7.0 which I believe refers to X11R7 (a relatively new release). The instructions on the NVidia site refers to X11R6 directories. Please check your system & move the files /usr/X11R6/lib/modules/drivers/nvidia_drv.o /usr/X11R6/lib/modules/extensions/libglx.so.x.y.z /usr/X11R6/lib/modules/extensions/libglx.so -> libglx.so.x.y.z to /usr/X11R7/lib/modules/drivers/nvidia_drv.o /usr/X11R7/lib/modules/extensions/libglx.so.x.y.z /usr/X11R7/lib/modules/extensions/libglx.so -> libglx.so.x.y.z (or wherever the new X library files are stored - may be in /usr/lib/xorg/modules as referenced in the log file) and try restarting the X server. According to the comments at http://lwn.net/Articles/165407/ the module library moved and you should move the files to the "right" location. --Maniac


  • Hello Nevada2244, Not a problem. You now seem to be on the right track now. Let me know next week how things work out and I'll look for some better references to describe the overall X set up / configuration, etc. --Maniac


  • Hello Maniac, Yes, we're successfully done. Thanks for your patience and expertise. I learned a lot and look forward to your wrap up. Will gladly try Google Answers again. Nevada2244


  • Hi Maniac, I did an apt-get upgrade, X failed, I reinstalled the driver without letting it rewrite the config file, and X with TwinView has again started fine since then. I'd like to do one more thing before finally declaring victory and letting you go. (I really don't want to let my researcher go!) That is a "dist-upgrade" to Ubuntu's Dapper Drake just released. I'll do this by about Wednesday of next week. Then I'll try again. I suspect I'll have to reinstall the driver yet again and it will work again with the same config file I'm using now. For your final response, I'd just like some good general resources on X and the X config file. When I first ran into this problem, I posted questions on several discussion groups and I got a wide variety of wrong answers. If you could point me to a couple good resources that I can study, that would help. Thanks for your help and patience. Nevada2244


  • Hello Maniac, I recursively copied the /usr/X11R6 directory to /usr/X11R7. Still would not start. I saw in the log file: (==) ModulePath set to "/usr/lib/xorg/modules" I copied nvidia_drv.o and nvidia_drv.so there and X STARTED with TwinView working! I rebooted a couple times and it still worked. I restarted from full power down and it still worked. Sorry you didn't see the old log files. I moved ISPs and just didn't think to move those directories. Very sorry if that held you up. I put the new log file here: http://levelofdetail.com/errs5/Xorg.0.log and moved my previous directory err to errs4 so the directories are in sequence. You don't need (or want!) these, but in case you are interested these are now on levelofdetail.com: directory: errs2 (bad conf file) Xorg.0.log.afterInstallReboot.txt Xorg.0.log.broke.txt Xorg.0.log.worked.txt nvidia-installer.log xorg.conf.nevada.txt xorg.conf.nvidia.txt errs3 Xorg.0.log xorg.conf errs4 (Dapper install - could not find driver) Xorg.0.log nvidia-installer.log xorg.conf errs5 (working again) Xorg.0.log If I can try to summarize what happened: With Dapper an updated version of X was installed. X11R7 looks for a hardware driver in /usr/lib/xorg/modules But the nvidia installer still puts it in the old location /usr/X11R6. Is this right? Does this mean everyone who uses X11R7 and an nvidia installer will have this same problem? I'll reboot again tomorrow and see how it goes. But we should finally have it. Will let you go soon! I would still like a couple of good X resources that you could recommend. Nevada2244


  • Hello Nevada2224, I am glad that it has worked out so far. Let me summarize the problems and current solution and then provide some further resources on the X window system. The initial problem was a failure to start the X server. The symptom was messages in the log file that appeared to run OK until the following messages appeared: Fatal server error: Caught signal 11. Server aborting After some initial diagnosis, we determined that the 3D acceleration, otherwise known as "GLX" was causing the problem. GLX is described at http://en.wikipedia.org/wiki/Xgl and several other sites - search for glx (or xgl) x windows to find more resources Disabling this was the initial fix for your system. When you upgraded your system from "breezy" to "dapper", that update introduced a newer version of the X server which broke the X server in a different manner. In this case, the update relocated only SOME of the files necessary to run the X server. We resolved that by moving the nvidia module / driver files to /usr/lib/xorg/modules and restarting the server. At this point, your display works "OK", though there may be some additional changes needed to fully enable the 3D acceleration (GLX). Make a clarification request if you want to follow up on this or continue to use the system as is. Additional references on the X window system and how it works follows: [1] A general overview of the system, historical descriptions, and a number of links at http://en.wikipedia.org/wiki/X_Window_System [2] Another high level overview with some specifics to Linux http://www.faqs.org/docs/Linux-HOWTO/XWindow-Overview-HOWTO.html [3] Another high level description of the X window system http://www.togaware.com/linux/survivor/X_Window.html Some of the more detailed pages include nice illustrations comparing the displays with no window manager, a minimal window manager, and Gnome. [4] A single page describing the architecture (with some nice diagrams) of the X window system http://maguolo.web.cs.unibo.it/xwindow/ [5] A section out of "The Art of Unix Usability" by Eric Raymond is on line at http://www.catb.org/~esr/writings/taouu/html/ch02s06.html which describes some of the characteristics of the X window system, primarily from a user's perspective. The main document indexed at http://www.catb.org/~esr/writings/taouu/html/index.html has a lot of related material you may find interesting as well. [6] If you wonder why applications don't seem to respond as you expect - an example at http://www.jwz.org/doc/x-cut-and-paste.html describes the different methods "Cut" and "Paste" are implemented by an X window client (application). Its somewhat unfortunate that X has more than one way to implement some capabilities, and not all applications implement them properly. [7] A generally sarcastic view of the X window system is at http://www.art.net/~hopkins/Don/unix-haters/x-windows/disaster.html which is a chapter of the "Unix Haters Handbook" by Garfinkel, Heise, and Strassmann. [8] As an alternative to the method you used to install the Nvidia drivers, your distribution has a "How To" at https://wiki.ubuntu.com/BinaryDriverHowto/Nvidia for the Nvidia drivers. You may wish to change to this method so others debug the installation (e.g., the difference between X11 version 6 and 7) and get everything set up properly. [9] The main X.org site at http://x.org/ Primarily of interest to those preparing distributions such as Ubuntu. There is a bug list and mailing lists. I suggest you review the mailing list archives prior to signing up to a mailing list to make sure the discussion has the kind of information you are looking for. However, there is documentation embedded in the distribution as described at http://tronche.com/gui/x/xlib/getting-the-manual-in-postscript.html which describes how to get a copy of the documentation (for viewing or printing). [10] O'Reilly sells a comprehensive series of books on the X window system. For example http://www.oreilly.com/catalog/v1/ is a programming manual for Xlib (the lowest level of X programming). Other volumes in the series describe network messages, higher level packages (e.g., Motif), and related materials. [11] Searching with a phrase such as x window linux faq brings up sites such as http://necrotic.deadbeast.net/xsf/XFree86/trunk/debian/local/FAQ.xhtml which has a variety of answers to "Frequently Asked Questions" related to the X window system or http://www.rahul.net/kenton/xsites.html which has a huge list of links organized in broad categories. I hope you find these resources helpful in your further work with your systems. Please make a clarification request if this list is incomplete or unclear. Good luck. --Maniac


  • Hi maniac, I did the dist-upgrade from Ubuntu Breezy to Dapper with no errors reported by apt-get. It did ask me if I wanted to overwrite /etc/gdm/gdm.conf. I replied no. X would not start so I reinstalled the driver. The installation appeared to go smoothly. However, X still will not start. It looks like it is a different problem than what I had before. The installer log file is here: http://levelofdetail.com/err/nvidia-installer.log As you can see, it looks good. The log file is here: http://levelofdetail.com/err/Xorg.0.log It's not finding the driver. On looking around, I found this: http://www.linuxquestions.org/questions/showthread.php?t=444218 I might have the same problem, but it doesn't say how it was fixed. I have no idea how the system finds its modules. "nvidia" is listed in the file /etc/modules. In the dmesg log file it does say: module license 'NVIDIA' taints kernel I'm wondering about that gdm.conf file. When looking at the directory I did see an "XKeepsCrashing" file. I tried it and it didn't report or change anything. What do you think?


  • Hello Maniac, I checked the files from the page you cited. They were all there except nvidia.o. Instead there was nvidia.ko and then in an nvidia subdirectory the file nvidiafb.ko. So the files I found were: /lib/modules/2.6.15-23-amd64-generic/kernel/drivers/video/nvidia.ko /lib/modules/2.6.15-23-amd64-generic/kernel/drivers/video/nvidia/nvidiafb.ko I could NOT find nvidia.o anywhere in the file system. I put my conf file here: http://levelofdetail.com/err/xorg.conf I did not let xconfig rewrite the file, so it should be the same as before. If this doesn't give a good idea of what to do next, I'm open to doing a fresh install of Dapper and then trying the nvidia install again from there. Nevada2244


  • maniac-ga, I tried turning off "Load glx" in the conf file and it looks like it's working. (It's the installer that put that line in there, not me!) My current log (on successful booting) is here: http://www.levelofdetail.com/errs3/Xorg.0.log My current conf file is here: http://www.levelofdetail.com/errs3/xorg.conf This is my edited "best try" conf file after my own study, not nvidia's xconfig. A few questions: 1. In the current log file I see these lines: (II) Loading extension NV-GLX (II) NVIDIA(0): NVIDIA 3D Acceleration Architecture Initialized (II) NVIDIA(0): Using the NVIDIA 2D acceleration architecture What? GLX IS being loaded, even though it's commented out in the conf file? Any guess what was happening? The driver can't figure this out? 2. Also in the log file, there's still a warning "More than one matching Device section for instances" Is it because I specify a device for each screen? How can I eliminate this warning? 3. Any other comment regarding the xorg.conf file? 4. Also in the log file I still see, as you did: (II) Loading extension XINERAMA I'm curious if you found out anything about this. I don't think it's important. Thanks for taking on this question. I'll see what happens when I reboot tomorrow morning before declaring final victory, but we might have it!


  • Hello Nevada2244, I agree that the build looks OK (I don't like all those warnings - but there are no errors and the kernel module was built). I agree with your assessment on the log file. On the location of driver issue, if you haven't seen this before, please review http://download.nvidia.com/XFree86/Linux-x86_64/1.0-8762/README/appendix-c.html which indicates the locations / versions of the files for the Nvidia install (both the kernel modules and X modules). Ensuring everything matches the files / locations listed here should restore proper operation. If not, I'll need to look at the X config file again to see if anything changed from the previous copy. > On looking around, I found this: > > http://www.linuxquestions.org/questions/showthread.php?t=444218 > > I might have the same problem, but it doesn't say how it was fixed. [see above for where Nvidia says where the files SHOULD be] > "nvidia" is listed in the file /etc/modules. That's the kernel module for the low level interface to the card (not the higher level module being loaded by the X server). > In the dmesg log file it does say: > > module license 'NVIDIA' taints kernel > Yes - it will do that. The concept of "kernel tainting" is briefly described at http://www.redhat.com/archives/linux-lvm/2001-October/msg00412.html and in the kernel FAQ at http://www.tux.org/lkml/#s1-18 and refers to the use of non GPL modules in Linux (and the extreme reluctance of kernel developers to debug tainted kernels). The kernel can still work OK - and usually does - but to get help for crashes, you have to go to the author of the kernel module that tainted the kernel. > I'm wondering about that gdm.conf file. That's the configuration file for the GNOME display manager (which runs on X). We're not getting far enough to start the display manager at this point (or if it starts - it notices that X is not running and gives up). The "its crashing" file basically tells you what you already know - X is not starting properly. --Maniac


  • Looks like the same problem here: http://www.linuxquestions.org/questions/showthread.php?t=452126


  • Hello Nevada2244, How did the tests go today? Can we declare success at this point? If so, I can post a proper answer with the overview of the solution as well as the additional X references / materials. If not, please let me know what still needs to be fixed and I can do some more research solving this problem. --Maniac


  • Hello Nevada2244, Hmm. It sounds like the configuration tool (nvidia-configx) is setting up the card differently than a cold boot. A few things to try. [0] I suggest adding something like... export NOWIS=`date` mv Xorg.0.log "Xorg.0.log.$NOWIS" to capture each log file with a unique name prior to starting X. You could leave this in permanently if you rm the old files (say - after a week). [1] Are there any differences in Xorg.0.log between the successful and unsuccessful runs? The diff output may be quite illuminating. [2] Since you don't get either (WW) or (EE) lines prior to the SIGSEGV (signal 11) - the X server code isn't finding the real problem. Try trapping the error by running gdb on the X server. There are at least two ways to do this: (a) debug using the core dumps. See http://www.redhat.com/archives/rhl-devel-list/2005-November/msg00531.html for a pretty good explanation of the steps involved (b) debug on line (suggest you use a serial interface or network login for this method). You can attach while the X server is starting, or start the X server by hand with gdb. The same reference as above describes the attach method. Starting the X server by hand requires some voodoo magic, and I'll have to do some further research to get the exact steps. If you have symbols loaded, you should get a good indication of the cause of the failure (at least the function / module it occurs in) and can report that back. If you don't have symbols - let me know and I'll see if I can find some version that does for you to test with. [3] Try disabling some features. It looks like you disabled the second monitor in your test, but it seems to be incomplete (e.g., TwinView is still enabled). Disabling glx (hardware acceleration) support might help as well. It also appears odd that the log file also refers to Xinerama (multiple monitors act as a single desktop) as a loadable module and built-in. Something is fishy there but I'll have to do some research on that. See also this note http://www.linuxquestions.org/questions/showthread.php?t=444218 Another user reported that the nvidia software wasn't all in the right locations. You may check your system to see if this is your problem as well. --Maniac







    • #If you have any other info about this subject , Please add it free.#
    Your name:
    E-mail:
    Telphone:

    Your comments:


    If you have any other info about Troubleshooting Nvidia TwinView on Ubuntu Linux , Please add it free.
    Posted by jack under xn--6kr096akncb2j.com
    edit