Thunderbolt Networking?

[ualex]

I am trying to get thunderbolt network working on Proxmox 8 with 2 Intel NUCs ... I only have partial success, with thunderbolt0 interface showing up.

Does anybody know how to make the interface visible in the GUI and/or automatic IP configuration? Currently nothing works for me.

 

[l-koeln]

Hi, i kind of got it to work....

https://forum.proxmox.com/threads/intel-nuc-13-pro-thunderbolt-ring-network-ceph-cluster.131107/

 

[ualex]

Hi, i kind of got it to work....

https://forum.proxmox.com/threads/intel-nuc-13-pro-thunderbolt-ring-network-ceph-cluster.131107/

Thanks for sharing the link with the information, and I will keep my eye on it ! Hopefully somebody has a solution for your problem.

 

[blebo]

I have this in /etc/network/interfaces on my NUC to QNAP link via thunderbolt, which might help:

iface thunderbolt0 inet manual

auto vmbr8
iface vmbr8 inet static
        address w.x.y.z/24
        bridge-ports thunderbolt0
        bridge-stp off
        bridge-fd 0
        bridge-vlan-aware yes
        bridge-vids 2-4094
#Thunderbolt3

Note: I think if you use the thunderbolt# interfaces names, the kernel auto loads the thunderbolt-net module. At least it has since PVE v7+.

 

[ualex]

I have this in /etc/network/interfaces on my NUC to QNAP link via thunderbolt, which might help:

iface thunderbolt0 inet manual

auto vmbr8
iface vmbr8 inet static
        address w.x.y.z/24
        bridge-ports thunderbolt0
        bridge-stp off
        bridge-fd 0
        bridge-vlan-aware yes
        bridge-vids 2-4094
#Thunderbolt3

Note: I think if you use the thunderbolt# interfaces names, the kernel auto loads the thunderbolt-net module. At least it has since PVE v7+.

I tried it, but the thunderbolt0 link does. not come "UP" and no IP address automatically assigned.

 

[ualex]

Other annoying behavior that sometimes the Thunderbolt link does not come up on my NUC8, and I only see in dmesg:

[    2.420057] ACPI: bus type thunderbolt registered
[    2.420293] thunderbolt 0000:04:00.0: error -EINVAL: cannot obtain PCI resources, aborting
[    2.420301] thunderbolt: probe of 0000:04:00.0 failed with error -22

A reboot is required to resolve this issue. Both systems are on the same Proxmox/kernel, only different hardware (NUC8 vs NUC13)

 

[Docop2]

this is only hardware. a usb4 cable do give full speed on a skull but nothing on nuc9 . just 1 time after like 5-6 reboot. tb is not to be pass.

 

[ualex]

@Docop2 i am not what you mean with your reply? Can you also assign an IP via USB4?

I got a few issues:
1) assign an IP address when plugging in the thunderbolt cable (using thunderbolt-net)
2) flaky behavior, with the "cannot obtain PCI resources, aborting" error message

 

[ualex]

Issue 1 I resolved using udev. I created "/etc/udev/rules.d/10-thunderbolt0.rules" with content:

ACTION=="add", SUBSYSTEM=="net", KERNEL=="thunderbolt0", RUN+="/usr/local/bin/pve-thunderbolt0.sh"

The script I created contains the following:

#!/bin/bash

/usr/sbin/ip link set thunderbolt0 up
/usr/sbin/ip address add 10.0.0.1/255.255.255.0 dev thunderbolt0

 

[scyto]

2) flaky behavior, with the "cannot obtain PCI resources, aborting" error message

Are you seeing this in syslog or somewhere else.
I haven't see this at all during the testing i have done today, but i can go looking to see if it's common and also on my system, if that would help?

 

[ualex]

@scyto I need to check where I seen it. I got a PiKVM connected to all my NUCs, so it could be the console too.

I also need to note I only seen it on my NUC8i5, not on the NUC13i3. The behavior when I see it, the thunderbolt0 device does *not* show up ever, a reboot is required. Thus if you always see the NUC, then you do not have the same problem as I noticed (which is good ;-)).

 

[scyto]

instead of using /usr/sbin/ip link set <IFACE> up if you switch to using /usr/sbin/ifup <IFACE> - this (seems) to have the advantage of applying whatever is /etc/network/interfaces and removes the need to set the IP in the script