Opt-in Linux 7.0 Kernel for Proxmox VE 9 available

[bossanova808]

I'm a relative novice with this stuff, so please excuse this noise if I am wrong, but that ^ smells a lot like: https://forum.proxmox.com/threads/a...-with-kernel-7-and-multiple-nic-types.183574/

 

[Blackclaws]

I'm a relative novice with this stuff, so please excuse this noise if I am wrong, but that ^ smells a lot like: https://forum.proxmox.com/threads/a...-with-kernel-7-and-multiple-nic-types.183574/

Could be related but tailscale is specifically UDP and not TCP so similar problems with offloading but a different datapath. Could just be that the whole off the offloading featureset is somewhat broken in some NIC combinations.

 

[tomten]

Leaving a link to this here as well so others looking for problems can see it: https://bugzilla.proxmox.com/show_bug.cgi?id=7627

TCP checksum offloading for virtio is broken for at least some NIC types on latest kernel + qemu for later linux guests / windows guests.

I've seen similar issues but with a realtek nic, I just figured it was realtek being realtek and turned off checksum offloading.

 

[Blackclaws]

I've seen similar issues but with a realtek nic, I just figured it was realtek being realtek and turned off checksum offloading.

See the bug, I can reproduce this on Broadcom and Intel NICs as well so I think its a wider problem. The problem exists when newer linux kernel features in 6.16+ interact with new Qemu versions 11.0+ which introduce new offloading scenarios. I don't know why these break if its a problem in Qemu or the host Kernel though.

 

[chrcoluk]

Does choosing a lower machine version prevent these issues? 11.0 was only released late april, it seems a bit too new.

 

[Pcom]

Hi,

with QEMU 10.2, there was a switch to using io_uring for the IO thread event loops and the IO pressure/wait accounting is set via the io_uring subsystem now. It's a different kernel subsystem from before, so it's not unexpected if it's different.

So, if I understood correctly, this is just a different way the graph is calculated after the update, and it does not necessarily mean that performance is affected, right?

 

[p-v-a]

see here - https://bugzilla.kernel.org/show_bug.cgi?id=220693. in my case I had to regress to 6.14 for FPDMA errors go away.

 

[fiona]

So, if I understood correctly, this is just a different way the graph is calculated after the update, and it does not necessarily mean that performance is affected, right?

Yes. To be precise: a different way the IO wait metric is calculated.

 

[kromberg]

I am trying to install the NVidia host grid drivers on 7.0.2-7-pve and I am getting this error:

fatal error: os-interface.h: No such file or directory

I have these installed:

proxmox-headers-6.17.13-12-pve
proxmox-headers-7.0.2-7-pve

What am I missing?

 

[daanw]

I am trying to install the NVidia host grid drivers on 7.0.2-7-pve and I am getting this error:

fatal error: os-interface.h: No such file or directory

I have these installed:

proxmox-headers-6.17.13-12-pve
proxmox-headers-7.0.2-7-pve

What am I missing?

Which driver version? This could indicate incompatibility between driver and kernel version.

 

[kromberg]

550.144.02 looks like the latest version a P100 and V100 supports.

 

[daanw]

550.144.02 looks like the latest version a P100 and V100 supports.

Linux 6.15 or newer has no support for the EXTRA_CFLAGS variable in out-of-tree module Kbuild files, needed for 550.144.02.

You can try to manually replace EXTRA_CFLAGS with ccflags-y in the driver's Kbuild files before installation, but no guarantee:
https://gist.github.com/mrdemonbkgit/9ba6962002829121486a0c6f4e8d4ac3

 

[kromberg]

Thanks for that, it helped some but still not working. Looks like I will need to reinstall and configure proxmox back to version 8.