[mantic] usb ethernet not working

Bug #2036841 reported by Paul Larson
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
linux-raspi (Ubuntu)
New
Undecided
Unassigned

Bug Description

Tested with the 20230919.1 arm64 image on rpi-zero2
This device does not have built-in ethernet, so it relies on USB ethernet to connect. I noticed on the initial boot that I was unable to login as ubuntu/ubuntu, and I suspect this might be loosely related to this because I also noticed that there were no lights on my ethernet dongle. If I unplugged it and replugged it, it would light up for a minute then stop. After reboot, I was able to login but still no ethernet. The dongle is detected, but it gets renamed from eth0 to something else, which I think is what's causing it to fail to be setup properly.

ubuntu@ubuntu:~$ uname -a
Linux ubuntu 6.5.0-1002-raspi #2-Ubuntu SMP PREEMPT_DYNAMIC Fri Sep 8 15:39:40 UTC 2023 aarch64 aarch64 aarch64 GNU/Linux

[ 6.990858] ax88179_178a 1-1.3.4.3:1.0 eth0: register 'ax88179_178a' at usb-3f980000.usb-1.3.4.3, ASIX AX88179 USB 3.0 Gigabit Ethernet, 00:0e:c6:c4:1f:57
...
[ 7.026290] ax88179_178a 1-1.3.4.3:1.0 enx000ec6c41f57: renamed from eth0

ubuntu@ubuntu:~$ ip a
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: enx000ec6c41f57: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN group default qlen 1000
    link/ether 00:0e:c6:c4:1f:57 brd ff:ff:ff:ff:ff:ff
3: wlan0: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN group default qlen 1000
    link/ether e4:5f:01:10:82:e1 brd ff:ff:ff:ff:ff:ff

I am able to manually configure ethernet though, so the device itself is working with this kernel.

Revision history for this message
Ubuntu QA Website (ubuntuqa) wrote :

This bug has been reported on the Ubuntu ISO testing tracker.

A list of all reports related to this bug can be found here:
https://iso.qa.ubuntu.com/qatracker/reports/bugs/2036841

tags: added: iso-testing
tags: added: rls-mm-incoming rpi
Manuel Diewald (diewald)
tags: added: kern-8058
Revision history for this message
Dave Jones (waveform) wrote :

What netplan configuration were you using? I managed to get an ethernet adapter working on mine but had to use something to match the enMACMACMAC name of the card:

network:
    ethernets:
        eth0:
            dhcp4: true
            optional: true
            set-name: eth0
            match:
              name: en*

The "set-name" is entirely optional, but if it's not there the card remains named "enMACMACMAC". The default netplan configuration won't work as the interface isn't named "eth0".

Revision history for this message
Paul Larson (pwlars) wrote :

I haven't altered the netplan config, but I thought this used to work on past releases without manually modifying the netplan didn't it?

Revision history for this message
Dave Jones (waveform) wrote :

It may have worked years ago when we just mapped absolutely anything with an "en*" name to eth0, but that caused issues with people that wanted multiple eth interfaces on their Pi (e.g. the built-in one and another plugged into a USB port). Sometime around jammy (I think?) we changed to only map interfaces matching "en*" and *also* matching one of the drivers used by the Pi's built-in ethernet (currently bcmgenet, smsc95xx, and lan78xx).

However, this does mean any port that's plugged into USB that doesn't match this spec, needs an updated netplan config.

Revision history for this message
Dave Jones (waveform) wrote :

I'm seeing something bizarrely similar to this on the 3B+ with its built-in ethernet during ISO testing the mantic final server images, that I could swear wasn't occurring during the beta. The ethernet adapter is coming up with an enMACMACMAC name on boot, so the netplan configuration doesn't apply and cloud-init falls apart if it tries to do anything on the network..

The bizarre bit is that the netplan (and systemd network) configuration appear correct because "sudo netplan apply" after boot causes the enMACMACMAC interface to rename to eth0 (as it should be) and everything applies happily. But that's no good if the Pi is headless and someone's relying on ethernet to reach it!

Revision history for this message
Juerg Haefliger (juergh) wrote :

These are not good:

Oct 10 15:07:15 ubuntu (udev-worker)[367]: eth0: /usr/lib/udev/rules.d/75-net-description.rules:6 Importing properties from results of builtin command 'net_id'
Oct 10 15:07:15 ubuntu (udev-worker)[367]: eth0: /usr/lib/udev/rules.d/75-net-description.rules:6 Failed to run builtin 'net_id': No such file or directory
Oct 10 15:07:15 ubuntu (udev-worker)[367]: eth0: /usr/lib/udev/rules.d/75-net-description.rules:8 Importing properties from results of builtin command 'usb_id'
Oct 10 15:07:15 ubuntu (udev-worker)[367]: eth0: Failed to access usb_interface: No such file or directory
Oct 10 15:07:15 ubuntu (udev-worker)[367]: eth0: /usr/lib/udev/rules.d/75-net-description.rules:8 Failed to run builtin 'usb_id': No such file or directory
Oct 10 15:07:15 ubuntu (udev-worker)[367]: eth0: /usr/lib/udev/rules.d/80-net-setup-link.rules:5 Importing properties from results of builtin command 'path_id'
Oct 10 15:07:15 ubuntu (udev-worker)[367]: eth0: /usr/lib/udev/rules.d/80-net-setup-link.rules:9 Importing properties from results of builtin command 'net_setup_link'
Oct 10 15:07:15 ubuntu (udev-worker)[367]: eth0: Failed to query name_assign_type: No such file or directory
Oct 10 15:07:15 ubuntu (udev-worker)[367]: eth0: Failed to get "name_assign_type" attribute, ignoring: No such file or directory
Oct 10 15:07:15 ubuntu (udev-worker)[367]: eth0: Failed to query addr_assign_type: No such file or directory
Oct 10 15:07:15 ubuntu (udev-worker)[367]: eth0: Failed to get "addr_assign_type" attribute, ignoring: No such file or directory
Oct 10 15:07:15 ubuntu (udev-worker)[367]: eth0: Link vanished while getting information, ignoring.

Revision history for this message
Dave Jones (waveform) wrote :

I've opened a separate issue (LP: #2038964) for my particular circumstance, as I don't think it's what Paul was seeing originally. It appears to be a udev race during boot causing multiple renames of the interface (with netplan losing the race and trying to apply the configuration to the "wrong" adapter).

Revision history for this message
Dave Jones (waveform) wrote :

Sorry, jiggling priorities on the wrong bug!

Changed in linux-raspi (Ubuntu):
importance: Undecided → High
status: New → Confirmed
status: Confirmed → New
importance: High → Undecided
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.