Fedora 11: X Windows configuration gets messed up after reboot (X server crashing)

[maverick280857]

Is it advisable to remove the PAE kernel altogether, installing something else, and then re-installing the NVIDIA drivers?

Why can't I use the drivers (in the pkg.run file) given by nVidia themselves? They don't seem to be able to compile. I think the problem is that they are not designed for PAE kernels, and so do not detect the kernel sources for the PAE kernel.

I think I will need to use nVidia's proprietary drivers to use CUDA. A side question: does anyone here have experience of using an nVidia driver on a PAE kernel to run CUDA?

[Hlingler]

Something is obviously very wrong here. I seem to recall reading something about this version of the driver, but I'll have to verify that....

Have you ever previously installed the NVidia driver binaries on this OS ? You should not install the driver, even the same version, from more than one source (binary, RPM), else you'll basically be "double-installing" the same (or conflicting) files in different (conflicting) locations. Installing different versions of the driver, from any source, is usually fatal.

Why can't I use the drivers (in the pkg.run file) given by nVidia themselves? They don't seem to be able to compile.

You've answered your own question: if the kmods don't compile, what good is trying the binary ?? At this point, I don't know why it won't compile.

V

EDIT: Here's the post I was talking about: http://forums.fedoraforum.org/showpo...&postcount=492

I have not yet found where to get the other driver RPMs....

EDIT #2: Here's the new DKMS-based beta driver: http://forums.fedoraforum.org/showpo...0&postcount=67
http://forums.fedoraforum.org/showthread.php?p=1251632

[maverick280857]

I went through the links you gave me, but how I can use them is not obvious to me. They seem to be for the non-PAE kernel.

Even I think now that something is wrong. I suppose I should reinstall Fedora 11 (from scratch). But when I do that, the PAE kernel gets installed by default. Shall I go ahead with the PAE kernel and re-try

yum install kmod-nvidia-PAE

or should I explicitly remove the PAE kernel first, then install the drivers?

Is it still possible to fix this non-invasively (i.e. without a re-install)?

[Hlingler]

I went through the links you gave me, but how I can use them is not obvious to me. They seem to be for the non-PAE kernel.

The PAE kernels are, for all practical purposes, simply i686 kernels, so the 32-bit DKMS driver RPM(s) should work. Use this post as a guide: http://forums.fedoraforum.org/showpo...91&postcount=4

V

[maverick280857]

I'm reinstalling Fedora 11 and will follow the steps to install the dkms-based driver, and post my experiences here in a short while. This is the fourth time I've had to reinstall Fedora 11 in 3 days :-/

[maverick280857]

Ok, I did a complete reinstall of Fedora 11, set up the network, ran

yum update yum

and then I followed the instructions given on http://forums.fedoraforum.org/showpo...91&postcount=4. The last step gve me an error, which is really weird. I don't know why this is happening to me despite repeated reinstalls. Please have a look (scroll down)

[root@farpoint Download]# yum --nogpgcheck install nvidia-x11-drv-185.18.31-1.fc11.i586.rpm
Loaded plugins: refresh-packagekit
Setting up Install Process
Examining nvidia-x11-drv-185.18.31-1.fc11.i586.rpm: nvidia-x11-drv-185.18.31-1.fc11.i586
Marking nvidia-x11-drv-185.18.31-1.fc11.i586.rpm to be installed
Resolving Dependencies
--> Running transaction check
---> Package nvidia-x11-drv.i586 0:185.18.31-1.fc11 set to be updated
--> Finished Dependency Resolution

Dependencies Resolved

================================================================================
 Package
     Arch   Version               Repository                               Size
================================================================================
Installing:
 nvidia-x11-drv
     i586   185.18.31-1.fc11      /nvidia-x11-drv-185.18.31-1.fc11.i586    40 M

Transaction Summary
================================================================================
Install      1 Package(s)         
Update       0 Package(s)         
Remove       0 Package(s)         

Total size: 40 M
Is this ok [y/N]: y
Downloading Packages:
Running rpm_check_debug
Running Transaction Test
Finished Transaction Test
Transaction Test Succeeded
Running Transaction
  Installing     : nvidia-x11-drv-185.18.31-1.fc11.i586                     1/1 
Error: Cannot open X config file (missing or malformed)
Error: Cannot open X config file (missing or malformed)

Installed:
  nvidia-x11-drv.i586 0:185.18.31-1.fc11                                        

Complete!
[root@farpoint Download]#

Anyway, then I rebooted the system, and since nvidia-settings told me the driver wasn't being used, I did

[root@farpoint vivek]# nvidia-xconfig

WARNING: Unable to locate/open X configuration file.

New X configuration file written to '/etc/X11/xorg.conf'

[root@farpoint vivek]# nvidia-xconfig

Using X configuration file: "/etc/X11/xorg.conf".
Backed up file '/etc/X11/xorg.conf' as '/etc/X11/xorg.conf.backup'
New X configuration file written to '/etc/X11/xorg.conf'

[root@farpoint vivek]#

Then, I rebooted the system. The X server did not start. So, I am still having the same problem.

I have tried dkms, kmod-nvidia and nvidia's own proprietary drivers. None of them is working. I can only use "vesa" and "nv" drivers. I have the default PAE kernel that FC11 automatically installs. What do you suggest I should do next?

[Hlingler]

First: there is no point in repeated re-installs of the OS - this will fix nothing.

The X server did not start.

Same exact results ? I.E.: the Xorg.0.log shows the same thing:

(EE) NVIDIA: Failed to load the NVIDIA kernel module. Please check your
(EE) NVIDIA:     system's kernel log for additional error messages.
(II) UnloadModule: "nvidia"
(II) Unloading /usr/lib/xorg/modules/drivers//nvidia_drv.so
(EE) Failed to load module "nvidia" (module-specific error, 0)
(EE) No drivers available.

What do you suggest I should do next?

Try the DKMS-based beta driver as previously suggested: http://forums.fedoraforum.org/showpo...0&postcount=67

Download it, 'yum --nogpgcheck localupdate ...', re-boot, see what happens. If that fails, we can try a kernel downgrade.

V

[maverick280857]

Ok, I did what you said, but X still does not start. I've attached the new Xorg.log.0 file with this message.

[Hlingler]

Well, progress: this time a different error:

(EE) Aug 14 13:56:56 NVIDIA(0): Failed to initialize the NVIDIA graphics device PCI:5:0:0. 
(EE) Aug 14 13:56:56 NVIDIA(0):     Please check your system's kernel log for additional error
(EE) Aug 14 13:56:56 NVIDIA(0):     messages and refer to Chapter 8: Common Problems in the
(EE) Aug 14 13:56:56 NVIDIA(0):     README for additional information.
(EE) Aug 14 13:56:56 NVIDIA(0): Failed to initialize the NVIDIA graphics device!
(II) UnloadModule: "nvidia"
(II) UnloadModule: "wfb"
(II) UnloadModule: "fb"
(EE) Screen(s) found, but none have a usable configuration.

Fatal server error:
no screens found

Try: go to Koji and fetch the last stable i686/PAE kernel released, plus the matching kernel-devel: http://koji.fedoraproject.org/koji/b...buildID=113377

Install these two packages, and re-boot into this kernel. See what happens.... You may need to edit (as root user or sudo) /boot/grub/grub.conf to allow you time to choose/switch which kernel to boot:

default=0
timeout=8
splashimage=(hd0,4)/grub/splash.xpm.gz
#hiddenmenu

I'm beginning to wonder if this chipset is supported (repeated checks at NVidia's site show it listed as supported), or if there may be a hardware problem.

V

[maverick280857]

I'll try this and get back in a while.

Before installing FC 11, I was successfully using FC 10 on this computer -- both kmod-nvidia and nvidia's own (pkg.run) drivers worked and so did Cuda. So this is a disturbing aberration. I doubt if a hardware problem is the cause.

[fanjinfei]

After install CUDA 2.3 driver
cd /lib/modules/<your running version>
cp kernel/drivers/video/nvidia.ko extra/nvidia/nvidia.ko

reboot

[maverick280857]

Hi fanjinfei, thanks for your reply. I tried your method, but it didn't work (could not find nvidia.ko, so the installation could not continue).

Anyway, Hlingler, it seems things are working now. Thanks! This is what I did:

Step 1: On a clean install of Fedora 11, run (as superuser)

yum update yum

yum install presto fastestmirrors

Step 2: Install dkms drivers (taken from http://forums.fedoraforum.org/showpo...91&postcount=4)

su
yum erase *nvidia* livna-config-display
yum install dkms kernel-devel kernel-headers gcc wget
wget http://www.dfm.uninsubria.it/compiz/fusion-testing/nvidia/i386/nvidia-x11-drv-185.18.31-1.fc11.i586.rpm
yum --nogpgcheck install nvidia-x11-drv-185.18.31-1.fc11.i586.rpm

then I got the nvidia driver from http://fedoraforum.org/leigh123linux....fc11.i586.rpm and did

yum --nogpgcheck update nvidia-x11-drv-190.18-1.fc11.i586.rpm

There was some problem at this stage, but then you suggested that I download the correct sources from http://koji.fedoraproject.org/koji/b...buildID=113377. So I downloaded the following files:

kernel-2.6.29.6-213.fc11.src.rpm
kernel-debuginfo-common-2.6.29.6-213.fc11.i686.rpm
kernel-PAE-2.6.29.6-213.fc11.i686.rpm
kernel-PAEdebug-2.6.29.6-213.fc11.i686.rpm
kernel-PAEdebug-debuginfo-2.6.29.6-213.fc11.i686.rpm
kernel-PAEdebug-devel-2.6.29.6-213.fc11.i686.rpm
kernel-PAE-debuginfo-2.6.29.6-213.fc11.i686.rpm
kernel-PAE-devel-2.6.29.6-213.fc11.i686.rpm

(humongous download!)

Step 3: Install the above

yum --nogpgcheck localupdate kernel*

This seems to have been the crucial step: dkms automatically detected the nvidia rpm file in the same directory as the kernel files (correct me if I'm wrong), and updated the driver. I rebooted the system and everything seems to be normal now.

I'd appreciate if you could go over these steps and fix them or convert them into a proper procedure which will be useful to everyone in future. I am not very happy with the haphazard way in which I installed the driver, and had it not been for your help, I would be running the vesa or the nv drivers Thanks a ton!

PS -- if you can fix these steps and make them more general and useful, they could be put up as some sticky here, and I can mirror it on my webpage.

[Hlingler]

(1) You did not need most of those kernel-* packages, only the ones in GREEN; RED are extraneous, and may be removed without issue:

kernel-2.6.29.6-213.fc11.src.rpm
kernel-debuginfo-common-2.6.29.6-213.fc11.i686.rpm
kernel-PAE-2.6.29.6-213.fc11.i686.rpm
kernel-PAEdebug-2.6.29.6-213.fc11.i686.rpm
kernel-PAEdebug-debuginfo-2.6.29.6-213.fc11.i686.rpm
kernel-PAEdebug-devel-2.6.29.6-213.fc11.i686.rpm
kernel-PAE-debuginfo-2.6.29.6-213.fc11.i686.rpm
kernel-PAE-devel-2.6.29.6-213.fc11.i686.rpm

(2) The nvidia driver version (185.18.31 versus 190.18) may or may not be an issue here - I don't know for sure, but 185.18.31 seems problematic.
(3) I also don't know whether the kernel downgrade helped, because I don't know which kernel you're booted into, which works. But I think the newer kernel-2.6.29.6-217.2.3 is a bad kernel (at least for NVidia driver). Please post for working driver:
uname -r

V

[maverick280857]

[vivek@farpoint ~]$ uname -r
2.6.29.6-213.fc11.i686.PAE
[vivek@farpoint ~]$

[Hlingler]

So: it's the kernel-2.6.29.6-217.2.3.fc11, in all likelyhood; probably (but not certainly) the NVidia driver version is not important.

Because of the timing (and repeated re-installs) of the Fedora OS, you missed the working kernel, and got the current/stable "questionable" kernel with a 'yum update'. Therefore, downgrading the kernel had to be done manually - as you did - from Koji.

On a clean install, I would recommend:
(1) install either stable (preferred) or beta NVidia driver (DKMS-based)
(2) 'yum update', but exclude kernel-*2.6.29.6-217.2.3.fc11
(3) Manually update to stable kernel-*2.6.29.6-213.fc11

V