Media (Intel® Video Processing Library, Intel Media SDK)
Access community support with transcoding, decoding, and encoding in applications using media tools like Intel® oneAPI Video Processing Library and Intel® Media SDK
Announcements
The Intel Media SDK project is no longer active. For continued support and access to new features, Intel Media SDK users are encouraged to read the transition guide on upgrading from Intel® Media SDK to Intel® Video Processing Library (VPL), and to move to VPL as soon as possible.
For more information, see the VPL website.

Installing linux NVIDIA drivers after SDK

Ask__Erik
Beginner
505 Views

Hi

I'm having troubles installing the NVIDIA proprietary drivers after installing the media SDK with patched DKM.

system: ubuntu 12.04, kernel 3.8.0-23-generic

mediaSDK: 2014 linux servers

Install of media sdk works fine, i followed the instructions including removing libdrm. After reboot the patched modules load correctly, but the nvidia module doesn't load and can't be modprobed, attempting to reinstall results in error, among stated possible reasons is building against an incorrect or badly configured kernel. The last entries from the install log are

-> Kernel module load error: insmod: error inserting './kernel/nvidia.ko': -1 Invalid parameters
-> Kernel messages:
[   94.471413] nvidia: Unknown symbol drm_gem_prime_export (err -22)
[   94.481574] nvidia: disagrees about version of symbol drm_pci_init
[   94.481577] nvidia: Unknown symbol drm_pci_init (err -22)
[   94.481617] nvidia: disagrees about version of symbol drm_gem_prime_handle_to_fd
[   94.481618] nvidia: Unknown symbol drm_gem_prime_handle_to_fd (err -22)
[   94.481634] nvidia: disagrees about version of symbol drm_gem_private_object_init
[   94.481635] nvidia: Unknown symbol drm_gem_private_object_init (err -22)
[   94.481687] nvidia: disagrees about version of symbol drm_gem_handle_create
[   94.481688] nvidia: Unknown symbol drm_gem_handle_create (err -22)
[   94.481726] nvidia: disagrees about version of symbol drm_pci_exit
[   94.481727] nvidia: Unknown symbol drm_pci_exit (err -22)
[   94.481745] nvidia: disagrees about version of symbol drm_gem_prime_export
[   94.481746] nvidia: Unknown symbol drm_gem_prime_export (err -22)
[  108.071591] nvidia: disagrees about version of symbol drm_pci_init
[  108.071595] nvidia: Unknown symbol drm_pci_init (err -22)
[  108.071628] nvidia: disagrees about version of symbol drm_gem_prime_handle_to_fd
[  108.071629] nvidia: Unknown symbol drm_gem_prime_handle_to_fd (err -22)
[  108.071642] nvidia: disagrees about version of symbol drm_gem_private_object_init
[  108.071643] nvidia: Unknown symbol drm_gem_private_object_init (err -22)
[  108.071687] nvidia: disagrees about version of symbol drm_gem_handle_create
[  108.071688] nvidia: Unknown symbol drm_gem_handle_create (err -22)
[  108.071718] nvidia: disagrees about version of symbol drm_pci_exit
[  108.071719] nvidia: Unknown symbol drm_pci_exit (err -22)
[  108.071734] nvidia: disagrees about version of symbol drm_gem_prime_export
[  108.071735] nvidia: Unknown symbol drm_gem_prime_export (err -22)

I have tried to manually move the patched sources to the corresponding locations in /usr/src/linux-headers-3.8.0-23-generic/ but there was no difference.

Reinstalling the generic kernel from repositories again I could install nvidia-drivers without problems, but now the incorrect drm_kms_helper (among others) modules is loaded.

Have anyone had similar issues? are there a complete patched sources package one could download? I'd really rather not rebuild the kernel manually, but if necessary are there any more steps than moving the files from the patched sources directories in the sdk to the above directory structure?

 

 

 

0 Kudos
3 Replies
Jeffrey_M_Intel1
Employee
505 Views

Thank you for bringing up this scenario so we can start considering potential solutions.  For now this isn't a supported configuration.  It isn't ideal, but for the current release it will be easiest to stick with either MSDK or NVENC/CUDA but not both.  Manual patching for a case like this is still a work in progress, and not something we're prepared to support at this time.

Regards,

Jeff

 

0 Kudos
Ask__Erik
Beginner
505 Views

Hi

Are the patches included with the SDK added later kernels?

Since NVENC is yet to be unlocked for linux (windows drivers got NVENC support in march ths year), and we require CUDA, dropping either isn't an option. Since we have to run an unsupported configuration, I'd rather run one that actually has the patched modules.

 

0 Kudos
Jeffrey_M_Intel1
Employee
505 Views

Most of the i915 patches are added to the kernel trunk and available in 3.13 or newer.  However, there are still a few minor patches which are not.  We're evaluating ways to make an "advanced" kernel available with the full set of updates for cases like this.  I'm still tracking down exactly which patches aren't committed.  In the meantime, there is a very good chance that you'll be able to proceed with a newer kernel.

0 Kudos
Reply