4 ms·

Note that this is just the kernel modules, not the actual graphics driver. IIRC these sources were already released under a permissive license along with their

Note that this is just the kernel modules, not the actual graphics driver. IIRC these sources were already released under a permissive license along with their driver distribution. This just seems to be them putting those sources on GitHub and being more open to contributions.

amluto 4y ago
Are they? I thought the “OS-agnostic” part was historically only available as a binary. Maybe that changed since last time I looked. On quick inspection, this is a complete, MIT-licensed kernel driver.
- mzs 4y ago
  >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.
  - monocasa 4y ago
    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.
  - amluto 4y ago
    Closed source firmware is acceptable for Linux. See: https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/about/ 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.
    - kaladin-jasnah 4y ago
      It already is for the 510 driver.
tomxor 4y ago
Thanks for clearing that up, for a second I thought nVidia was finally Linux viable.
pluc 4y ago
The same could be said of a lot of Nvidia IP which was leaked a few months ago. But this is different, it's voluntary.
odensc 4y ago
This isn't the same as the source previously provided with the drivers. This is a completely new kernel module.
jakogut 4y ago
The kernel driver was never distributed as source. The driver itself was a giant compiled object file that was linked into a kernel module with some shim code using DKMS. I know because I dug through it trying to fix the RTX 2060 in my G14 not going into D3.

amluto 4y ago

Are they? I thought the “OS-agnostic” part was historically only available as a binary. Maybe that changed since last time I looked. On quick inspection, this is a complete, MIT-licensed kernel driver.

mzs 4y ago
>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.
- monocasa 4y ago
  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.
- amluto 4y ago
  Closed source firmware is acceptable for Linux. See: https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/about/ 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.
  - kaladin-jasnah 4y ago
    It already is for the 510 driver.

tomxor 4y ago

Thanks for clearing that up, for a second I thought nVidia was finally Linux viable.

pluc 4y ago

The same could be said of a lot of Nvidia IP which was leaked a few months ago. But this is different, it's voluntary.

odensc 4y ago

This isn't the same as the source previously provided with the drivers. This is a completely new kernel module.

jakogut 4y ago

The kernel driver was never distributed as source. The driver itself was a giant compiled object file that was linked into a kernel module with some shim code using DKMS. I know because I dug through it trying to fix the RTX 2060 in my G14 not going into D3.