Scenario:
Moved 12.3 HDD from failed hardware with nVidia GPU to new hardware with Intel HD 4600 GPU.
No logs suggested why Intel GPU driver wasn’t loading
Upgraded to 13.1
Now, Xorg log says that nVidia GPU kernel module is loading but no screens to attach to(of course).
Inspected installed packages, found full listing of packages installed for both nVidia/nouveau/nv and Intel(although Intel tools not installed).
Attempt
Tried to uninstall all the nVidia/nouveau/nv packages, but this results in practically all applications wanting to be uninstalled as dependencies. If I pursue this route, I would have to manually mark every app, library and plenty more to “keep” and run the dependency checker for every item and that’s a BIG job.
Suggestion?
Possible to blacklist a GPU kernel module to prevent it from loading, how would this work?
Ugh.
If as you describe I’ll literally be nearly installing an entire OS selecting one package at a time manually, and because the point of upgrading vs a new install is to preserve apps and settings it’s much bigger than just a new default install.
>
> Scenario:
> Moved 12.3 HDD from failed hardware with nVidia GPU to new hardware with
> Intel HD 4600 GPU.
> No logs suggested why Intel GPU driver wasn’t loading
> Upgraded to 13.1
> Now, Xorg log says that nVidia GPU kernel module is loading but no
> screens to attach to(of course).
> Inspected installed packages, found full listing of packages installed
> for both nVidia/nouveau/nv and Intel(although Intel tools not
> installed).
>
> Attempt
> Tried to uninstall all the nVidia/nouveau/nv packages, but this results
> in practically all applications wanting to be uninstalled as
> dependencies. If I pursue this route, I would have to manually mark
> every app, library and plenty more to “keep” and run the dependency
> checker for every item and that’s a BIG job.
>
> Suggestion?
> Possible to blacklist a GPU kernel module to prevent it from loading,
> how would this work?
>
> TIA,
> TSU
>
>
Check if you have a /etc/X11/xorg.conf that is referencing the nvidia
driver. If so, get rid of it.
Thx for replying.
One of the first things I did was verify no xorg.conf exists.
I came up with a solution for removing the nVidia/nouveau/nv packages. I removed them <breaking dependencies>, then did another offline install (equivalent to a zypper dup). Rebooted. Now the error isn’t loading the wrong GPU driver, is saying the driver is missing. So, progress but not the root solution.
Have been researching how the kernel loads the kernel module for the appropriate display, obviously the udev system doesn’t do that, with numerous reboots I’ve provided multiple and ample opportunity to automatically detect and load new display hardware. Apparently the new display hardware is detected, but still the wrong drivers are loaded.
A short while ago, I finally think I found what is required. In the following location are a number of files similar to init.d files and this specific file still specifies the nouveau driver
/etc/X11/xorg.conf.d/50-device.conf
I’m pretty sure if I modify this file to something like “Intel” it might fix my problem once and for all but am still looking around for the exact string to enter.
I haven’t found much documenting what I’m feeling my way through…
On 2013-11-20 21:56, tsu2 wrote:
>
> robin_listas;2599431 Wrote:
>> On 2013-11-20 21:06, tsu2 wrote:
>>
>>> Attempt
>>> Tried to uninstall all the nVidia/nouveau/nv packages, but this
>> results
>>
>> No. Only uninstall nvidia packages.
> If as you describe I’ll literally be nearly installing an entire OS
> selecting one package at a time manually, and because the point of
> upgrading vs a new install is to preserve apps and settings it’s much
> bigger than just a new default install.
I said “nvidia”. Not nouveau nor nv.
I mean “nvidia” as in nvidia proprietary driver.
–
Cheers / Saludos,
Carlos E. R.
(from 12.3 x86_64 “Dartmouth” at Telcontar)
That was the issue, when these HDD were in the old hardware running 12.3, the nouveau driver was “hard coded” into the file I specified above. Apparently no upgrade or update routine exists which checks the contents of the file.
I then ran the KDE LiveCD to see how this file is configured and found it empty(all lines commented out). So, that was what I did since I couldn’t be sure the exact correct driver to specify.
I also found this after-the-fact SDB which although out of date (verified to work only up to 11.4) but contains still relevant information including my “discovery” SDB:Configuring graphics cards - openSUSE
Thx.
But in this situation they are all a family of drivers that apply to the same chipset so the problem and eventually solution applies to using any of those drivers.
As it turns out, the issues I encountered and the specific problem I fixed can apply to any hardware change from one GPU to another. Or, for that matter might apply to any change in xorg related hardware like monitor, mouse, keyboard, more.
> Thx.
> But in this situation they are all a family of drivers that apply to the
> same chipset so the problem and eventually solution applies to using any
> of those drivers.
What I meant was that the only “driver” you can remove is the
proprietary driver; all the rest (nouveau) are integrated parts of the
kernel and xorg, and attempting to remove them will cause destruction
and havoc. In fact, it would be removing the nvidia rpm or binary
(soft/hard way), and then reapply a certain xorg package for mesa
libraries I always forget the name.
A different thing is changing back whatever configuration files the
nvidia install changed.
50-nvidia.conf
–
Cheers / Saludos,
Carlos E. R.
(from 12.3 x86_64 “Dartmouth” at Telcontar)
Another thing to consider when migrating a system in this way is which modules are included in initrd, one will usually have to generate a new one.
Check Yast>sysem>etcsysconfigeditor>system>kernel>initrd_modules, if nvidia is listed remove it.