General :: Nvidia Driver Is Not Working After Using A Self Compiled Kernel (2.6.38)?
Mar 19, 2011
About a month ago, I decided to go deeper in my Linux knowledge. I've been reading a lot and found out that Arch linux would be my learning distro. As I was installing Arch, it was a pain at first but I really learned a lot which I would never ever learn with Mint. Now I decided to take a step further is which "compiling your own kernel".
MY PROBLEM:
Everything was smooth in my Arch for a 2 weeks until I decided to compile my own kernel. I currently have a Nvidia GTX 460 1GB card.
BTW, the method I used to install nvidia with 2.6.37 was
Code:
pacman -S nvidia nvidia-utils
and that worked with all of the .37 kernels (-1 -2 -3 -4) which was download from kernel.org.
When I compiled the .38 (using the same .config of .37), it just boots up to the terminal (not loading gdm). However, I could still login by typing my username and password. So I've checked dmesg of both .37 and .38 and noticed that the Nvidia driver is not loaded in .38.
I tried reinstall it by running pacman and it doesn't do anything. It was thinking of uninstalling nvidia and nvidia-utils but there are so many dependencies conflicts (like screensaver, compiz, etc).
Since, I didn't want to mess up my .37 install, I just grabbed an old HDD and installed from scratch again. This time, I compiled the .38 kernel first (without gui) and then installed nvidia nvidia-utils. It was the same problem.
With this observation, I'm concluding that the nvidia and nvidia-utils from pacman is not compatible with .38.
I've read that I have to wait for nvidia to release a driver that will be compatible with the .38 kernel. Is that true? Does it mean I have to wait for nvidia/nvidia-utils to be updated from pacman? How would I know when it is updated?
I've also read about nouveau, but I guess that is not for me because it doesn't support 3D.
Is there a work around for me to use nvidia/nvidia-utils with .38?
Is nvidia and nvidia-utils proprietary drivers? What is the difference with these two and the one you download directly from nvidia?
First of all, I apologize for the bombardment of questions. As you can tell, I'm so clueless on how nvidia drivers work on linux in general (since it was spoon fed by mint) and I really would love to learn about this is a deeper level. Could someone please explain to me (LAYMANS terms) how nvidia works (and possibly a solution to my issue).
EDIT: Additional info - I have a netbook that also runs arch. It uses an Intel GMA integrated video chip which I used "xf86-video-intel" from pacman and I believe since it is open source, it works with .38 fine. So does that mean if you use an open source driver, it will work with all other kernels?
I am running the newest Nvidia driver from nvidia.com, version 260.19.44. The built in proprietary Nvidia driver does not work properly.
Recently ubuntu was updated to a newer kernel 2.6.35.27 from 2.6.35.25. Under 2.6.35.25 the driver is working. Under 2.6.35.27 the desktop is not starting, and I can login to a 'console'. After login i try to run startx, and gets some errors. See attached Xorg.0.log.
I compiled the 2.6.31.6 kernel and <insert drumroll> it boots!(my first kernel roll, I'm kind of shocked actually) That's the good news. The bad news is that my NVIDIA drivers are gone in the wind. That's not entirely true as I can still boot into the old kernel and startx. Is there a way to download the driver using the command line for reinstall?
my problem is on installing nvidia driver on fc12 32bit but, first of all, as i understood the pae kernel requires more than 4gb of ram,i have a 2.2 ghz cpu with 2 gb ram,but when i run command:uname -r it answers: 2.6.31.5-127.PAE [i have fc12 32 bit] when we try to download linux we have a 32bit edition or 64bit edition,do we have an edition which is only for pae? or when we install for example the 32bit edition on a computer with more than 4gb of ram then the kernel automatically will change to be a pae kernel??
I installed the latest kernel liquorix (2.6.35) but when i want to install the Nvidia driver downloaded on the Nvidia website (256.53), i have an error message because Nvidia doesn't found the kernel source tree.
I install linux-image-2.6.35-6.dmz.2-liquorix-686_2.6.35-16_i386.deb, linux-headers-2.6.35-6.dmz.2-liquorix-686_2.6.35-16_i386.deb and build-essential. I don't understand why the installation doesn't works.
I compiled my own kernel with tun/tap and bridge support. Both modules load fine at boot time (I could read that in the dmesg output). Now I want to use it, and the /dev/net/tun node is not there, so my application gives that error. I'm trying to bridge openvpn connections. Is it possible that udevd is not doing his work?
PS: I'm on a WD MyBook World Ed NAS device. It's ARM, so I cross compiled the kernel from my debian linux machine. I also installed debian on the NAS.)
I installed it on a Dell Inspiron 531 with the GeForce 6150SE nForce 430 built in video card. From the recommend driver list I installed �NVIDIA accelerated graphics driver (version current) [Recommended
Well, turns out it should not have been recommended. I had restarted and all I got was a low res ubuntu logo and a boot right into a full screen terminal. Tried startx and got a no screens found, I look online for about an hour last night and decided to just reinstall, which takes a couple hours when installing all the packages.
I have done more research today and found to install the latest linux x64 driver from nVidias website, which I did, but it does not run. I followed some more instructions and it said to do a �sudo chmod +x <file>� which I did, and it starts to open and I get �You appear to be running an X server; please exit X before installing.�
I�m starting to get a little frustrated here, guess I�m just used to a lot of the ease of windows and assumed that something like installing a graphics card driver would be easy.
This is a followup/secondary question from another post earlier this morning.[URL].. Info: Lucid 10.04, 4GB ram, Nvidia video, home built pc. I'm fairly certain that when I enable the pae kernel this time, I'll get usage of the full 4GB of memory that I have installed.
The problem is, that I did enable the pae kernel earlier this week to test something out, but ran into a problem. It came up in "low graphics mode" (800x600). As you can image, I wasn't real thrilled about that.
I went to System->Administration->Hardware Drivers to get the video driver straightened out. I figured it just needed one compiled for the pae kernel... Well, it didn't find one. Is there one? Is there a procedure to get one installed/compiled, if there isn't one?
I don't know if I want to mess with trying to get the Nvidia drivers working for pae, when I only stand to gain another ~700MB of memory. Currently showing 3.2GB. I'm not sure if its worth my trouble..
Having just updated various files including the kernel using Package Manager I no longer seem to have the correct version of the Nvidia graphics driver. On previous updates this has been done automatically by the "kmod Nvidia" Metapackage. My last kernel was 2.6.32.19-163 fc12.i686.PAE and the Nvidia driver for that did get downloaded correctly. Looking on Yumex I cannot see a driver for this latest kernel listed.
I'm encountering a problem running X and Gnome from a Xen-enabled Kernel with NVIDIA Binary driver compiled with IGNORE_XEN_PRESENCE=y on debian squeeze
Hardware: NVIDIA NVS 5100M Kernel: Debian Squeeze : 2.6.32-5-xen-amd64 NVIDIA Kernel from the official package
Boot and module loading are successful, but when X starts, I only get a black screen. I attached here my Xorg.0.log, however it doesn't seems to have any problem.
i ran yum update which updated kernel to 2.6.31.9.174.fc12.i686.PAE. Now after logging in i get a blank white screen. With previous kernel updates i have had no such problem. Anyway, the boot messages are following:
Quote:
checking for module nvidia.ko [FAILED] nvidia.ko for kernel 2.6.31.9-174.fc12.i686.PAE was not found [WARNING] The nvidia driver will not be enabled until one is found [WARNING] Driver already disabled
I just responded to a (packagekit?) prompt to update packages, which included new kernel 2.6.35.12-88.fc14.x86_64. I use nVidia on my notebook and usually the new driver is installed automatically. This time, the computer would not boot to the stage that the nVidia logo appears, indicating the driver is not present. I edited grub.conf to take me back a version and I am now running under the previous kernel 2.6.35.11-83.fc14.x86_64. I tried a yum search for kmod-nvidia-2.6.35.12-88.fc14.x86_64 and nothing was found. I have enabled these repos:
I'm running Slack64 13.1 and I'm in good shape with the 33.4 kernel, but if I upgrade the kernel to 35.x what do I need to do to get it working with my Nvidia 7300GT? I have the nvidia installer, NVIDIA-Linux-x86_64-260.19.12.run. Do I just do from the stable repo
Code: slackpkg upgrade then boot to run level 3 and run NVIDIA-Linux-x86_64-260.19.12.run? Thanks for your help.
I have an Nvidia graphics card,... actually I manage several workstations that run centos and have an nvidia video card. I also have a personal computer with ubuntu and an nvidia network card.
I would like to do a regular automatic update of those Centos workstations. (With a pilot group to test and then a full roll-out). Until oktober 2009 no major difference in automatic updating ubuntu and centos (apart from the differences between apt and yum):
After a kernel upgrade, the systems can not boot into it's Xorg gui, because the nvidia driver must be rebuild (=not recompiled, because this is partially object code, the driver is not opensource).
But from ubuntu 9.10 onwards, the kernel update process checks for the presence of propietary drivers like those of nvidia and does a rebuild on the reboot, so that the system can succesfully boot into the xorg GUI (and gdm or kdm) My question is: Are ther any plans for Centos to do the same, this would relief me from some upgrade hassle for the Centos workstations that I manage. Or does anyone know about a (good) automagic workaround?
I compiled my kernel, compiled scsi support into kernel, used the new kernel and initrd, the boot failed.Then ,i unzip my initrd, found that sd_mod.ko can't be insert, i added it manual, and reboot OK.so, why! in the kernel configure , the sd_mod.ko is set to <M> , but why it can't be found in initrd?
I've posted here before (viewtopic.php?f=5&t=66322) about problems arising from my attempts to get an nVidia driver to work with my custom kernel. Now those problems are all fixed, and I'm back to where I was: the built kernel boots fine, but the nVidia driver fails.
The custom kernel is as near to the stock one as I can make it, I'm just trying to find a working build process at present, before trying to build a later-version kernel.
I used sgfxi with "-! 40" to build the nVidia driver for my custom kernel; it reported that everything was fine.
With stock kernel - 2.6.32-5-amd64 ...
Extracts from /var/log/Xorg.0.log: X.Org X Server 1.7.7 Release Date: 2010-05-04 X Protocol Version 11, Revision 0 code....
I'm running Fedora 13 x64 and updated the kernel to the latest version (2.6.34.7-56 to 2.6.34.7-61). I use the proprietary nvidia drivers (currently 260.19.12) so I had to compile the kernel module against the new kernel sources and reinstall the driver. The process was apparently successful, but when I try to start X nothing happens, it's as if the computer had been suspended, my monitor acts like it isn't receiving any input. I have full runlevel 3 access, and the system seems pretty much fine up to that point. Nvidia's own sanity tests which are built into the installer reported no problems with my driver.
I used to have FC11 installed with the nvidia supplied binary driver. Worked fine, very stable. Trying FC12 and FC13 live CDs caused a hang, so I stuck with FC11. FC14 live CD did boot so I backed up my own files and did a clean FC14 install. Upgraded to get kernel 2.6.35.6-48, installed the nvidia driver from RPMFusion (260.19.21) and had a working system! Now, after a cold boot, the kernel always hangs during boot. Tried a number of kernel boot option (apci=off, noapic and a few others too). The hang is not always at the same place. The earliest it has hung is after:Disabling lock debugging due to kernel taint
Sometimes its a few lines later at:NVRM: loading NVIDIA UNIX x86_64 Kernel Module 260.19.12 ...(which would seem to imply that the hang can even occur a step or two before the actual driver is loaded!)
Sometimes it keeps going for a few more seconds and randomly hangs during some other driver load.
Without a specific nvidia driver the system always boots ok, so I'm pretty sure its the nvidia driver but I just can't seem to get a handle on this in a reproducible way. Sometimes, when I have make some other change, it will boot and work ok, but a reboot locks it up again (for example, in case it was SELinux, I disabled it and on the next reboot it booted and worked ok. Then, next boot, it wasn't working again). When I make changes, I reboot using the previous kernel, make the changes and reboot, with an occasional success. Once I start rebooting using the latest kernel, it always hangs.
I'd love to get a solution to this, but even some advice on debugging this would be appreciated!
I'm using 8600GT, 256Mb, PCIe card in a MA69VM motherboard with 3Gb of RAM.
I would like to install above mentioned driver. I need to prepare kernel for this instalation, but this information I didnt find on openSUSE pages. I add openSUSE 11.3 KDE
I've been trying to install the latest driver NVIDIA-Linux-x86_64-256.35.run. I do ctrl+alt+f1 and login, then sudo sh ./NVIDIA-Linux-x86_64-256.35.run and it gives me the license agreement, but after I accept and right after it gets to 100% it says it can't install the kernel or access or whatever. Has anyone done this installation and had the same problem but figured it out??
I've just installed Lucid Lynx on both my machines in the interest of sitting and waiting for the Unity/Gnome 3 storm to blow over. On the HP (see below), everything works great, and I've followed instructions online on how to upgrade to LibreOffice, upgrade the kernel to 2.6.38 using the kernel PPA, etc.
However, on the IBM, I'm using Nvidia proprietary graphics drivers. These work well on the stock kernel that 10.04 installed (2.6.32-32), but installing 2.6.38 seems to break the driver. If I install the driver first, and then the kernel on top, X stops working and I have to revert to the default, generic driver to get back in. Once there, I cannot install the driver again. The Additional Drivers dialog goes through the motions, but then drops a "systemerror: installarchives() failed" error message.
So, is there a different version of the driver I should be trying to install? I should clarify at this point I tried all three options the Additional Drivers dialog provided me, all gave the same result (version 96, version 173, and version current).
Or should I leave the kernel at 2.6.32? Is there any downside to leaving it?
After upgrading kernel package to 2.6.32-5 NVIDIA installation gave me ERROR:Unable to load the kernel module 'nvidia.ko'. This happens most frequently when this kernel module was built against the wrong or improperly configured kernel sources, with a version of gcc that differs from the one used to build the target kernel, or if a driver such as rivafb/nvidiafb is present and prevents the NVIDIA kernel module from obtaining ownership of the NVIDIA graphics device(s), or NVIDIA GPU installed in this system is not supported by this NVIDIA Linux graphics driver release.
I have made a full install of Slack 13.37.0. When I try to run the installer of the NVIDIA GF 8400GS card (NVIDIA-Linux-x86-180.29-pkg1.run) downloaded from [URL], I receive an error message. It says that the kernel source cannot be indetified/found. /lib/modules/2.6.37.6-smp/source and .../build links to /usr/src/linux-2.6.37.6-smp, which contains the full kernel source (can be compiled), including the header .h files in include/linux. The same NVIDIA installer can be run successfully on my previous Slack 11.
I've built a new kernel (2.6.34) on our workstation at work. It boots and runs beautifully, but there is one minor problem. I created the kernel as a Debian package along with the kernel headers. Upon installing both and attempting to build the nVidia driver for said kernel, the installer tells me that it cannot determine the version and quits. This happens even if I manually specify the path to the headers. What's going on here, did I miss something during my compilation of the new kernel?
In order to configure a dual display in a fresh 11.4 installation, I am attempting to install the nVidia driver according to the instructions in SDB:NVIDIA the hard way - openSUSE. I have not previously compiled or configured a kernel so I largely followed the instructions from OpenSUSE 11.2 - How to compile a Kernel for Newbies.The currently installed kernel is 2.6.37.1-1.2-desktop. The graphics card is an ASUS EN9600GSO (512MB).
In Yast, I installed gcc, make, kernel-devel (v. 2.6.37.1-1.2) and kernel-desktop-devel (v. 2.6.37.1-1.2). Per the "hard way" instructions, I did not install kernel-source.
I am fairly new to Linux. My machine is running Debian 5.0.5 with gcc version 4.3. When I try to install the Nvidia QuadroFX 3450 driver I got from the Nvidia web site I get an error saying: "The compiler used to compile the kernel (gcc 4.1) does not exactly match the current compiler (gcc 4.3)....." Does this mean my gcc version is too new for the driver? if so, how do I roll back to a older version?
I'm using Slackware 13.0 on rather old PC with old Riva TNT2 video card. Default driver is "nv". Everything work fine, but without hardware 2D acceleration under X.
After studying various manuals I 1. downloaded Nvidia binary package suitable for my video card. 2. Recompile kernel without Riva framebuffer support. 3. Start Nvidia script.
Script said: "Error: unable to find the kernel source tree for the currently running kernel. Please make sure you have installed the kernel source files for your kernel and that they are properly configured...""
Kernel sources are in /usr/src/linux-2.6.29.6 I have all kernel packages installed. I was trying various switches for script in order to show the right path - nothing! 8 Some people say that Nvidia script don't like 4th digit in a kernel's name and get it from "uname -r" output. Can I change it somehow?
I had to have the Nvidia Driver build a module to my kernel. It goes through, says it was successful, but I an unable to start X and get this log:
Quote:
X.Org X Server 1.6.1.901 (1.6.2 RC 1) Release Date: 2009-5-8 X Protocol Version 11, Revision 0 Build Operating System: Linux 2.6.18-128.1.6.el5 i686
[code]....
Its an Nvidia 8200 integrated graphics card... Not sure if there is a wrong driver or the driver is in a different location then where Fedora is looking.
