Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

>Note that the kernel modules built here must be used with gsp.bin firmware and user-space NVIDIA GPU driver components from a corresponding 515.43.04 driver release. This can be achieved by installing the NVIDIA GPU driver from the .run file using the --no-kernel-modules option.


Closed source firmware is acceptable for Linux. See:

https://git.kernel.org/pub/scm/linux/kernel/git/firmware/lin...

It would not be much of a stretch at all for nvidia to include their firmware there.

The userspace stack is an issue. For upstream Linux, there’s a fairly hard rule that graphics driver interfaces must be usable with open source userspace. But I don’t think the nvidia graphics user parts are particularly secret, and someone could write such a thing.


It already is for the 510 driver.


That's orthogonal to the parent's question.

My understanding as well was that the bulk of the kernel space driver was previously only available as a binary blob, with only a small shim layer open sourced that loaded the rest of the kernel module and translated all of the internal kernel calls. I heard a rumor this was actually core to their internal legal theory about why they could have a binary blob kernel driver.

Required firmware blobs and the user space libraries are ultimately different components in the stack.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: