ASUS M3N78-EM

[colinjones]

Hmm, I've never seen that link and have no idea why that is thought necessary. Plus the article is none-to-specific about which libraries may need updating....

I would suggest just running the driver update normally, using the main process and see what happens.... my understanding is that it should work just fine. Yes, the info you have is for the 64 bit driver, and it really ought not be because the vast majority of people are using i386. You just need to replace the URL with the 32bit one.... go to the nVidia downloads site, choose the universal GPU driver for Linux 32 bit, and hit OK. It will then present you with a download link. You can right click that and copy it, then in the process type wget <paste> and you will get the right one.

[rocketlynx]

Colin, at the bottom of the wiki page (http://wiki.linuxmce.org/index.php/Upgrading_the_Kernel), Upgrading the Kernel, it goes as follows:

wget http://us.download.nvidia.com/XFree86/Linux-x86_64/173.14.12/NVIDIA-Linux-x86_64-173.14.12-pkg2.run
sh NVIDIA-Linux-x86_64-173.14.12-pkg2.run --keep --add-this-kernel
./stripheader.sh NVIDIA-Linux-x86_64-173.14.12-pkg2/usr/src/nv/precompiled/precompiled-nv-linux.o* NVIDIA-Linux-x86_64-173.14.12-pkg2/usr/src/nv/nv-linux.o
ld -d -r -o nvidia.ko NVIDIA-Linux-x86_64-173.14.12-pkg2/usr/src/nv/nv-linux.o NVIDIA-Linux-x86_64-173.14.12-pkg2/usr/src/nv/nv-kernel.o
for md in /usr/pluto/diskless/*; do
cp -r NVIDIA-Linux-x86_64-173.14.12-pkg2/usr/* $md/usr/
cp nvidia.ko $md/lib/modules/2.6.26.5/kernel/drivers/video/
done

I'll go to the nvidia download and find a 32 bit sub.
Thanks

[rednaxelan]

rocketlynx / anyone did you ever get this working as a diskless MD? I just read through the entire thread hoping to find the solution to all my M3N78-EM woes but there isn't a solution that worked for me in here.

My onboard NIC is disabled and I have a PCI NIC in its place that works fine for net booting from the core.
The MD was set as an i386 MD and image rebuilt to overcome the initial boot issues.
I got to the point that the screen is cycling but AVWizard never shows, I updated the nvidia driver to the latest from their site (180.60) on the core and again rebuilt the MD image but it still just sits there with the screen cycling. I can't hit any combo of keys to escape from it and am totally stuck, would appreciate anyones suggestions.

[Zaerc]

rocketlynx / anyone did you ever get this working as a diskless MD? I just read through the entire thread hoping to find the solution to all my M3N78-EM woes but there isn't a solution that worked for me in here.

My onboard NIC is disabled and I have a PCI NIC in its place that works fine for net booting from the core.
The MD was set as an i386 MD and image rebuilt to overcome the initial boot issues.
I got to the point that the screen is cycling but AVWizard never shows, I updated the nvidia driver to the latest from their site (180.60) on the core and again rebuilt the MD image but it still just sits there with the screen cycling. I can't hit any combo of keys to escape from it and am totally stuck, would appreciate anyones suggestions.

You probably should not have rebuilt the MD image after installing the nvidia driver.

[rednaxelan]

Thanks for the reply, but why should I not have rebuilt the image?

[colinjones]

Because rebuilding dumps and completely recreates the diskless image, thus blowing away 1) your new driver and 2) any changes you made to xorg.conf...

If you need to rebuild the diskless image to change architecture from AMD64 to i386, or something like that, you should do that first. Only then upgrade the drivers, modify config files, etc...

[rednaxelan]

  I am a complete newb here so please treat me as such, I am trying my best to be a Linux convert but years of windoze use has addled my brain.

I didn't update the drivers on the MD as I have never been able to get to a prompt of any kind on the MD, instead I updated the drivers on my core, with the thought (maybe the wrong thought) that when I rebuilt the MD image it would push those drivers out to the new image. Am I completely off track?

[qball4]

Rebuilding the image basically wipes the old data from /usr/pluto/diskless/XX and replaces it with a clean version from the /usr/pluto/install/Pluto-MD-i386.tar.bz2 archive. In order to have the updated drivers automatically pushed you need to rebuild the archive per the wiki. Here's a syntax-corrected version:

(Notes - Replace XXX with your kernel version. If you are using amd64 for your MD's, replace i386 with amd64.)

Code: [Select]

cd /tmp
mkdir moonfs_x86
cd moonfs_x86
tar -jxvf /usr/pluto/install/PlutoMD-i386.tar.bz2
cp /boot/vmlinuz-XXX boot
cp /boot/initrd.img-XXX boot
cp -r /lib/modules/XXX/ lib/modules/
tar -jcvpf /usr/pluto/install/PlutoMD-i386.tar.bz2 *
cd ..
rm -rf moonfs_x86

:Matt

[rednaxelan]

Well we are definitely getting somewhere. I followed those steps and they worked perfectly (thanks Matt). I booted the MD and finally got into and completed the AVWizard. I fully expected it then to finish up and go into lmce but no instead I get the following:

*Reloading /etc/samba/smb.conf
kdeinit:Can't connect to the X server
kdeinit: Might not terminate at the end of the session
DCOP aborting call from 'anonymous-9258' to 'kded': ERROR communication problem with kded, it probably crashed

Then it repeats the following message continuously:
lmce_launch_manager: cannot connect to X server :0

I can shift between tty1, tty2 etc with no problem but it definitely will not connect or start X. Is this possibly something to do with the PCI NIC I am using?

[colinjones]

Not likely to be anything to do with your NIC, as the errors are about X Windows failing to start up.

Rather than attempting to update the diskless image archive on your core. Why not just directly install the nVidia driver on the MD ... ie just update the bit you are interested in, rather than wholesale refresh all the system modules.

Delete the MD device and start again.... It will completely delete the /usr/pluto/diskless/<dev#> folder. Now boot the MD again to start the process. Once you have gotten to the "cycling" part again (make sure it is definitely complete in the setup).

From the core, type

sudo -s
(and your linuxmce password, then)
ssh moon##
(where ## is the device number of the MD just created)

This will leave you at a root prompt for the MD, now you can install the nVidia driver. If it complains about X running or trying to run whilst you are installing the driver, you can always type at the MD -

telinit 1

To kill off all the higher level processes.

Once done, reboot. If there are still issues, then you need to review your /var/log/Xorg.0.log file for errors whilst X is starting up. You can also do this through the ssh shell, or directly on your core, here /usr/pluto/diskless/##/var/log/Xorg.0.log

[rednaxelan]

Thanks Colin, but unfortunately it gave the same results: got through the AVWizard with no problem but started cycling the same error message over and over: I did however notice a section of the error that I had not seem before because the screen scrolled through too fast. It said:

DCOPClient :: attachinternal. Attach Failed Could not open network socket.

The rest of the errors then came after that. Any help would be appreciated.