nVidia version mismatch on sidux-box
effo
Status: New User - Welcome
Joined: 30 Aug 2009
Posts: 3
Reply Quote
Yesterday I did a du using apt, then I rebooted and re-installed the nVidia drivers using sgfxi.

I rebooted once again, then X refused to start. In the log I see:

:: Code ::
Aug 30 12:44:07 siduxbox kernel: NVRM: API mismatch: the client has the version 185.18.36, but
Aug 30 12:44:07 siduxbox kernel: NVRM: this kernel module has the version 185.18.31.  Please
Aug 30 12:44:07 siduxbox kernel: NVRM: make sure that this kernel module and all NVIDIA driver
Aug 30 12:44:07 siduxbox kernel: NVRM: components have the same version.
Aug 30 12:46:25 siduxbox kernel: NVRM: loading NVIDIA UNIX x86 Kernel Module  185.18.36  Fri Aug 14 17:18:04 PDT 2009
A


X.org.log says:
:: Code ::
(EE) NVIDIA(0): Failed to initialize the NVIDIA kernel module. Please see the
(EE) NVIDIA(0):     system's kernel log for additional error messages and
(EE) NVIDIA(0):     consult the NVIDIA README for details.
(EE) NVIDIA(0):  *** Aborting ***


I run sgfxi once again, and after that, I could start X with 'init 5'. However, after another reboot, X doesn't start, with the same errors in logs. If I run sgfxi, then I can start X.

I have absolutely no clue on how to solve this. If more information is needed, please say so.
Back to top
techAdmin
Status: Site Admin
Joined: 26 Sep 2003
Posts: 4127
Location: East Coast, West Coast? I know it's one of them.
Reply Quote
Are you using that dkams cr#p? That's my guess as to the source of the problem. You need to purge all that junk and then just use straight nvidia install.
Back to top
effo
Status: New User - Welcome
Joined: 30 Aug 2009
Posts: 3
Reply Quote
Thanks for the reply.

Yes, the dmakms was installed (as suggested in the sidux manual), in an attempt to manage nvidia drivers on my sidux install without using sgfxi. Well, that failed miserably...

After purging dmakms, I also had to remove the latest kernel, and re-install it from apt. Now my sidux is running fine again, thanks a lot.
Back to top
techAdmin
Status: Site Admin
Joined: 26 Sep 2003
Posts: 4127
Location: East Coast, West Coast? I know it's one of them.
Reply Quote
Can you please post the exact steps required to removing the dmakms stuff to get a working system again?

I debated briefly adding in an option to do that in sgfxi, or at least to alerting the user that it's running, but since I won't install it on my systems, it's hard to test or debug.

It is however virtually guaranteed to fail with a rolling release system like sidux, although for a stable kernel release like ubuntu or debian stable it should in theory work ok.

Also, can you be more precise about what failed? and how?

The idea, which is nice in theory, of this auto kernel module rebuilder stuff is that when a new kernel is installed, the old driver modules get rebuilt to new kernel, automatically.

Sgfxi doesn't have anything to do with this method, and probably won't for quite some time, since the failure points with it are so clear and obvious if you follow non free driver development, almost comically obvious in fact.

So what made you decide to not use it, and to use sgfxi instead, what failed, that is? Clearly you can't swap between the two, it has to be one or the other, personally I prefer the method tested and proved reliable over years by 100s of thousands of real world rolling release driver installs, but that's just me, I'm strange about that.
Back to top
effo
Status: New User - Welcome
Joined: 30 Aug 2009
Posts: 3
Reply Quote
Problem description
-------------------------
After doing a DU and bringing in a new kernel with apt (not smxi), I rebooted and rebuild the nVidia kernel module using sgfxi. Then I could start X without problem. The next day I started the machine, X didn't run with the errors shown above in the log.

To work around this issue, I rebuild the nVidia module, and started X. However, the next day X didn't start with the same errors. Now I realized that I had problems...

Solution
-----------
Please notice, I am no pro, and all these steps might not be necessary, but this is how I did to get my system running.

First, undo everything that was specified in the sidux manual: manual.sidux.com/en/hw-dev-hw-dri-en.htm#non-free-3d

I rebooted, but this didn't help, I guess the module was still loaded in some way. I rebooted once again and booted with the previous kernel, then I removed the latest installed kernel. This could not be done using the kernel-remover, so I had to use apt for this.

Just to be sure, I rebooted again, then did a DU to bring in the newest kernel again. I rebooted and used sgfxi to build the nVidia module, and started X.

As a final step I rebooted again, and now I could boot the latest kernel with the newly built nVidia module.

Again, I bet there are better ways to do this, but this is what I did.
Back to top
techAdmin
Status: Site Admin
Joined: 26 Sep 2003
Posts: 4127
Location: East Coast, West Coast? I know it's one of them.
Reply Quote
That manual entry is so utterly painful to read.... Talk about talking yourself and your users into a corner....

Very depressing, no wonder their user numbers are plummetting.

But this is why I left that project.

My main interest in that method is if I can make an easy dmakms remover option in sgfxi to help users who followed that manual advice, which like many sidux generated methods, is incomplete, half thought out, and user unfriendly to the max.

The sidux kernel-remover script I had to give up on and do a full rewrite of in smxi in order to get bugs and issues handled, by the way, that's why smxi has it natively now. But I don't know about dmakms stuff re kernels, nor did I really want to know about it.
Back to top
Display posts from previous:   

All times are GMT - 8 Hours