Open Linux Kernel Modules Installed by Default Starting in the release 560 series, it will be recommended to use the open flavor of NVIDIA Linux Kernel Modules wherever possible (Turing or later GPUs, or Ada or later when using GPU virtualization). If installing from the .run file, installation will detect what GPUs are present and default to installing the open kernel modules if all NVIDIA GPUs in the system can be driven by the open kernel modules. Distribution-specific repackaging of the...
It’s the part that can legally be distributed with Linux distributions (including in-car OS) due the kernel’s license. The actual functionality is in the proprietary user space driver
I’m not sure. Didn’t they just move the code that was previously executed in the proprietary kernel module to the new also proprietary userspace driver that’s just connected to the hardware by this new and open source wrapper module? And the other half into firmware? It’s still arbitrary and closed code that gets forwarded to the hardware. And running there it has access to all the memory, screen content etc… I’m not sure if this is a win concerning security. I think it’s pretty much unchanged.
But there are several big advantages. Now the kernel probably won’t get tainted any longer and we can have signed kernels and activate secure boot easily. And that’s maybe a big plus for security. And I hope we’ll get the convenience, too. In the past I had the NVidia driver crap out on me while debugging stuff with recent kernel versions or release candidates. And NVidia was lagging behind, leaving me with a console instead of the desktop environment…
Didn’t they just move the code that was previously executed in the proprietary kernel module to the new also proprietary userspace driver
Probably. And that is exactly what was expected from them since the beginning of their Linux drivers. Kernel is not a place for such big and proprietary piece of code. So this is the important change.
Yes, the driver is still proprietary, but it does not break the kernel any more the way it did.
But this is the part where being open source is most important. For security, maintainability and convenience reasons
One could even argue if the usespace part, the OpenGl or Vulkan implementation, is still ‘a driver’. (I think it is, at least partially)
It’s the part that can legally be distributed with Linux distributions (including in-car OS) due the kernel’s license. The actual functionality is in the proprietary user space driver
I’m not sure. Didn’t they just move the code that was previously executed in the proprietary kernel module to the new also proprietary userspace driver that’s just connected to the hardware by this new and open source wrapper module? And the other half into firmware? It’s still arbitrary and closed code that gets forwarded to the hardware. And running there it has access to all the memory, screen content etc… I’m not sure if this is a win concerning security. I think it’s pretty much unchanged.
But there are several big advantages. Now the kernel probably won’t get tainted any longer and we can have signed kernels and activate secure boot easily. And that’s maybe a big plus for security. And I hope we’ll get the convenience, too. In the past I had the NVidia driver crap out on me while debugging stuff with recent kernel versions or release candidates. And NVidia was lagging behind, leaving me with a console instead of the desktop environment…
Probably. And that is exactly what was expected from them since the beginning of their Linux drivers. Kernel is not a place for such big and proprietary piece of code. So this is the important change.
Yes, the driver is still proprietary, but it does not break the kernel any more the way it did.