Network configuration issue after kernel 6.8 upgrade

[MMartinez]

Hi,

I've experimented a problem after upgrading kernel to 6.8 on Proxmox 8.2 on a Dell C6420 node.

After renaming interfaces as described in Kernel 6.8 - "Known Issues & Breaking Changes" (https://pve.proxmox.com/wiki/Roadmap#Proxmox_VE_8.2), I've been able to use almost every network defined in the node, but there is one network defined as a VLAN on the host with Traditional Linux Bridge that doesn't work.

I've changed the configuration to use VLAN aware Linux Bridge and it is working again, but I prefer to notify the issue.

Following the configuration recommended on https://pve.proxmox.com/wiki/Network_Configuration, my NOT working configuration is:

iface enp24s0f0np0 inet manual
iface enp24s0f0np0.106 inet manual


auto vmbr0v106
iface vmbr0v106 inet static
        address 192.168.106.21/24
        bridge-ports enp24s0f0np0.106
        bridge-stp off
        bridge-fd 0


auto vmbr0
iface vmbr0 inet static
        address 192.168.105.21/24
        gateway 192.168.105.1
        bridge-ports enp24s0f0np0
        bridge-stp off
        bridge-fd 0

With this configuration network 192.168.106.0/24 does not work. I get an UNKNOWN state when checking with "ip a" and ping doesn't work.

Next configuration using VLAN aware works fine:

iface enp24s0f0np0 inet manual
iface enp24s0f0np0.106 inet manual

auto vmbr0.106
iface vmbr0.106 inet static
        address 192.168.106.21/24

auto vmbr0
iface vmbr0 inet static
        address 192.168.105.21/24
        gateway 192.168.105.1
        bridge-ports enp24s0f0np0
        bridge-stp off
        bridge-fd 0
        bridge-vlan-aware yes
        bridge-vids 2-4094

Hope it helps.

Regards,

Manuel Martínez

 

[cheiss]

Hi,

iface enp24s0f0np0.106 inet manual

please note interface names can be at most 15 bytes long - as Linux defines the maximum interface name length IFNAMSIZ. This includes to the trailing NUL-byte, so you get 15 bytes at most.

enp24s0f0np0.106 is 16 characters long, so - on a hunch - that is probably why it's not working.

 

[leesteken]

Following the configuration recommended on https://pve.proxmox.com/wiki/Network_Configuration, my NOT working configuration is:

iface enp24s0f0np0 inet manual
iface enp24s0f0np0.106 inet manual


auto vmbr0v106
iface vmbr0v106 inet static
        address 192.168.106.21/24
        bridge-ports enp24s0f0np0.106
        bridge-stp off
        bridge-fd 0


auto vmbr0
iface vmbr0 inet static
        address 192.168.105.21/24
        gateway 192.168.105.1
        bridge-ports enp24s0f0np0
        bridge-stp off
        bridge-fd 0

I think that iface enp24s0f0np0.106 inet manual is unnecessary here.

 

[MMartinez]

I think that iface enp24s0f0np0.106 inet manual is unnecessary here.

Probably is not needed. I'ts been working without this line with kernel 6.5 and previous.

 

[MMartinez]

Hi,


please note interface names can be at most 15 bytes long - as Linux defines the maximum interface name length IFNAMSIZ. This includes to the trailing NUL-byte, so you get 15 bytes at most.

enp24s0f0np0.106 is 16 characters long, so - on a hunch - that is probably why it's not working.

Thanks. That could be the reason. But the network interfaces naming scheme is the cause of the problem. Any way to avoid it? I've seen that I can pin the network name (https://pve.proxmox.com/pve-docs/pve-admin-guide.html#network_override_device_names) but, does it make sense?

Regards,

Manuel

 

[cheiss]

I've seen that I can pin the network name (https://pve.proxmox.com/pve-docs/pve-admin-guide.html#network_override_device_names) but, does it make sense?

Yeah, that would work and IMO make sense here.
Overriding the network interface name does not have any adverse effects.

It can even be beneficial if e.g. the firmware does not have stable, port-based names and possibly change them when adding e.g. new PCIe devices. Which is not so uncommon, at least with consumer hardware.

 

[MMartinez]

Yeah, that would work and IMO make sense here.
Overriding the network interface name does not have any adverse effects.

It can even be beneficial if e.g. the firmware does not have stable, port-based names and possibly change them when adding e.g. new PCIe devices. Which is not so uncommon, at least with consumer hardware.

I'll do it then. Thank's for your help!!

Kind regards,

Manuel Martínez