ot-br-posix
ot-br-posix copied to clipboard
openthread
otbr + systemd-networkd results in failure of infrastructure network
Describe the bug
There seems to be a serious conflict between otbr-agent & systemd-networkd (from Ubuntu 21.04 on pi 4b) that causes failure of the infrastructure network on the BR.
To Reproduce
raspberry pi 4b (Ubuntu 21.04), ot-br-posix @ b66e46416a34dc25ec17693affab006629db1385 nrf54840-dongle, ot-nrf528xx @ 063233bf4a0833bda55295bc89890a96abb5aeb8
connect to ethernet & configure via netplan
ubuntu@ubuntu:/tmp$ cat /etc/netplan/50-cloud-init.yaml
# This file is generated from information provided by the datasource. Changes
# to it will not persist across an instance reboot. To disable cloud-init's
# network configuration capabilities, write a file
# /etc/cloud/cloud.cfg.d/99-disable-network-config.cfg with the following:
# network: {config: disabled}
network:
ethernets:
eth0:
dhcp4: true
optional: true
version: 2
Prior to starting BR, network is in a healthy state:
ubuntu@ubuntu:~$ sudo networkctl list eth0
IDX LINK TYPE OPERATIONAL SETUP
2 eth0 ether routable configured
1 links listed.
Start otbr-agent
sudo systemctl start otbr-agent
The backbone network fails:
ubuntu@ubuntu:~$ sudo networkctl list eth0
IDX LINK TYPE OPERATIONAL SETUP
2 eth0 ether routable failed
1 links listed.
ubuntu@ubuntu:~$ sudo networkctl status eth0
● 2: eth0
Link File: /usr/lib/systemd/network/99-default.link
Network File: /run/systemd/network/10-netplan-eth0.network
Type: ether
State: routable (failed)
Path: platform-fd580000.ethernet
Driver: bcmgenet
HW Address: dc:a6:32:8c:93:bf (Raspberry Pi Trading Ltd)
MTU: 1500 (min: 68, max: 1500)
QDisc: mq
IPv6 Address Generation Mode: eui64
Queue Length (Tx/Rx): 5/5
Auto negotiation: yes
Speed: 1Gbps
Duplex: full
Port: tp
Address: 10.1.0.50 (DHCP4 via 10.1.0.2)
fe80::5873:39b0:f14e:c1b5
fe80::dea6:32ff:fe8c:93bf
Gateway: 10.1.0.2 (Intel Corporate)
DNS: 10.1.0.2
Activation Policy: up
Required For Online: no
DHCP4 Client ID: IAID:0x396e6af3/DUID
DHCP6 Client DUID: DUID-EN/Vendor:0000ab11d8a291b15382e5900000
Aug 17 22:26:47 ubuntu systemd-networkd[1643]: eth0: Link UP
Aug 17 22:26:51 ubuntu systemd-networkd[1643]: eth0: Gained carrier
Aug 17 22:26:51 ubuntu systemd-networkd[1643]: eth0: DHCPv4 address 10.1.0.50/24 via 10.1.0.2
Aug 17 22:26:52 ubuntu systemd-networkd[1643]: eth0: Gained IPv6LL
Aug 17 22:30:21 ubuntu systemd-networkd[1643]: eth0: Could not set NDisc route: Gateway can not be a local address. Invalid argument
Aug 17 22:30:21 ubuntu systemd-networkd[1643]: eth0: Failed
A failed interface does not renew DHCP leases so this can cause a loss of network connectivity.
It appears to be related to RA processing in systemd (and may be a bug in systemd rather than otbr).
Expected behavior A clear and concise description of what you expected to happen.
The network should not fail.
Console/log output If applicable, add console/log output to help explain your problem.
Additional context Add any other context about the problem here.
This should be an RA processing bug at the systemd-networkd side since otbr-agent works well with dhcpcd.
There have been similar bug reports (e.g. https://github.com/coreos/bugs/issues/1478#issuecomment-234772765)
Seems systemd-networkd failed to handle RIO in the RA advertised by otbr-agent, because kernel refused to add the route with a local gateway address.
systemd-networkd should ignore RIO sent by the device itself (otbr-agent) because such route does not make sense to kernel. We actually have a dhcpcd fix for handling RIO, in which we explicitly ignored RIO sent by the device itself.
