Linux 5.13 Yanks A NVIDIA NVLink Driver For Lack Of Open-Source User

[phoronix]

Phoronix: Linux 5.13 Yanks A NVIDIA NVLink Driver For Lack Of Open-Source User

The VFIO changes for the Linux 5.13 kernel aren't particularly exciting this cycle but one of the changes does raise some eyebrows with the VFIO NVIDIA NVLink2 driver being removed. This driver is being removed as it shouldn't have been even added in the first place for lack of an open-source client/user exercising it...

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.13-Yanks-VFIO-NVLink2

[Alexmitter]

Of course it has to be kicked, it can not be tested after all.

[Mitch]

Forgive my ignorance but as I understand it in Linux, the number 1 rule is "we do not break userspace". Does this not break that rule?

I'm not making a statement or anything, just asking. I'm actually glad maintainers stick to their guns on this and remove this on the basis that it never should have been there.

[microcode]

Forgive my ignorance but as I understand it in Linux, the number 1 rule is "we do not break userspace". Does this not break that rule?

I'm not making a statement or anything, just asking. I'm actually glad maintainers stick to their guns on this and remove this on the basis that it never should have been there.

I think in this case, the retort is "what userspace?".

[bug77]

Forgive my ignorance but as I understand it in Linux, the number 1 rule is "we do not break userspace". Does this not break that rule?

I'm not making a statement or anything, just asking. I'm actually glad maintainers stick to their guns on this and remove this on the basis that it never should have been there.

Rule #0 is it doesn't get into the kernel without an open source client that exercises it
And when there's no client, there's nothing to break.

[milkylainen]

Pissing contest. All this is just needs to be called for what it is.

Does this mean that we invalidate all drivers for hardware that cannot be bought yet?
But this was NvLink2. And it is available in Power9?
Or what the hell does "open source user/client" in the hardware space mean?

Also. The kernel community does not have to hide anything or being coy about it.
This should be called. "We like to take a dump on NVidia because we can."

(No, I do not like NVidia.)

[bug77]

Pissing contest. All this is just needs to be called for what it is.

Does this mean that we invalidate all drivers for hardware that cannot be bought yet?
Fine. No more drivers until we see actual hardware first?
Or what the hell does "open source user/client" in the hardware space mean?

Also. The kernel community does not have to hide anything or being coy about it.
This should be called. "We like to take a dump on NVidia because we can."

(No, I do not like NVidia.)

Well, I like Nvidia, but I disagree with what you said.

Inclusion into the Linux kernel requires an open source client, because that's the only way to actually test what's being included. It's sensible and straightforward, no pissing contest here.
Unreleased hardware can be tested, that's what engineering samples are for.

[dh04000]

So say for a user that had hardware that depends on that code. Did that permeantly break that hardware for future kernels? Or can that be readded on the userside when say I stalling the Nvidia driver?

[dh04000]

When installing* the

[chuckatkins]

The issue in this case I suppose is what defines "client". HPC and AI are the biggest market for NVLink I believe and most of the widely used HPC codes using NVIDIA GPUs are Open Source. So if the "client" is the application using GPUs across NVLink then the assertion is simply wrong. However, if they are instead viewing the GPU driver as the client then the assertion holds.