Luis Chamberlain sent out the modules changes today for the Linux 6.6 merge window. Most notable with the modules update is a change that better builds up the defenses against NVIDIA's proprietary kernel driver from using GPL-only symbols. Or in other words, bits that only true open-source drivers should be utilizing and not proprietary kernel drivers like NVIDIA's default Linux driver in respecting the original kernel code author's intent.
Back in 2020 when the original defense was added, NVIDIA recommended avoiding the Linux 5.9 for the time being. They ended up having a supported driver several weeks later. It will be interesting to see this time how long Linux 6.6+ thwarts their kernel driver.
b) How long before NVIDIA throws up their hands at the whole thing and does their own Linux distro + pushes all their cloud AI customers to use it? (it doesn't seem like they're ever going to be shamed / coerced into actually open-sourcing their driver)
Would having their own distro even help? It seems like working around this would require forking from Linux at a lower level, and even that would only circumvent technical (rather than copyright) barriers.
They can probably just drop some kernel packages in their driver PPAs or whatever. You don't need to fork the whole distro to customize the kernel. But it will still be a huge pain.
They can beef of linux support of freebsd a little and do some other help to the desktop experience there. Freebsd has always been more pragmatic, and for most uses of an os you can't tell a real world difference. (pkg instead of apt, and other such differences are minor)
The userland differences are not too great, but I would assume a kernel module as significant as a modern GPU driver is pretty deeply tied to Linux's kernel internals.
Nvidia maintains a driver for FreeBSD already, same version as Linux and everything. IIRC the closed-source portion is “unified” and they just build the interface for whatever OS.
The “Linux support” piece is on the application side.
There's an interesting discussion about the whole topic on the Phoronix forums about this. Some people claim that removing them and Nvidia's current behavior is a DMCA violation:
The kernel includes IP only licensed under GPLv2.
While a module linked against the kernel isn't necessarily a derived work which in turn would need to be licensed GPLv2 as well, there are specific interfaces that are meant for internal use and by their very nature would make your work derived if using them. These are the interfaces marked EXPORT_GPL_ONLY.
Using these interfaces with a module not licensed GPLv2, you taint the kernel and violate the licensing.
Removing the check, you aren't necessarily yet violating GPLv2, but you're removing a technical protection measure which is a violation of the DMCA.
It also raises the question why you'd remove checks that only prevent a possible GPLv2 violation if you're not violating GPLv2 anyways as Nvidia claims.
you aren't necessarily yet violating GPLv2, but you're removing a technical protection measure which is a violation of the DMCA.
Isn't overcoming a technical limit a violation itself? That's what made DeCSS illegal. They didn't have to prove anyone was actually copying DVDs with it, just that DeCSS could allow you to copy a DVD
B) can't happen because of gpl. Even if it could, not many customers will move to an nvidia distro. ML people need good distros and good drivers.
If a hypothetical nvidia distro would speed up training by 10% but cause drop of productivity of humans of as small as 5%, no many will "buy" it. We can throw more hardware, people are the bottleneck nowadays