We’ve tried to install two different drivers on a machine running Red Hat 7.9 with the latest patch set. We’ve followed the instructions and blacklisted the Nouveau driver in GRUB. The .run file prompted us about having a key (we need to us to boot as UEFI and with secure boot) and we answered the question which then allowed us to proceed with what appears to be a successful install.
After rebooting, we no longer get the graphical login.
Digging around the /var/log area it mentions that nvidia.ko couldn’t be loaded and it has something to do with Secure Boot, but it also says it will proceed to boot up.
We’ve done some searches in google and the general advice is to disable secure boot - that is not an option for us.
Does NVIDIA have a version of the driver that is signed to avoid this issue? This will happen each time there is a kernel update, how can this be avoided if patching is automated?
The model of the graphics card is a NVS 510 GK107.
I don’t know of a signed driver for RHEL 7, I guess you’ll have to create your own certs, enroll them and then use dkms to auto-sign the modules on kernel/driver updates.
@johnsoohoo
@Cliff_Woolley
Hi team,
I have copied source code for nvidia drivers in to my centos, and generated nvidia.ko,nvidia-uvm.ko,nvidia-drm.ko files in the centos…
These .ko modules are copied to yacto based SDU OS, While installing these modules in to kernel space of ram using insmod i am facing ERROR AS BELOW…
unknown sbol in nvidia_register_module
Unknow symbol in nvidia_get_rm_ops
We have done signature for .ko files, but issue is still present
Please help in this to resolve module dependency issue
You can’t use modules compiled for a different kernel.
1 Like
If i use same kernel version in both operating systems,
Can I succeed in this, Or i will face compiler issues too??
You’ll not only need the same kernel version but the same kernel image. You should rather install the kernel headers of the target on your local system, make sure that compiler linker are the same and then you can build foreign modules.
Thank you everyone…😊 i will check this
Hi @generix ,
While running nvidia-smi command
Failed to initialize nvml : driver/library version mismatch…
Could you please check this
Kernel module is 515.48.07 but userland is 510.47.03 so nothing will work.
1 Like
@generix
Can you help to share the mknod commands to create device files for the following modules
nvidia.ko,
nvidia-uvm.ko,
nvidia-modeset.ko,
nvidia-drm.ko
nvidia-peermem.ko
I don’t know, the dev nodes should be created on module load.
