Package: network-manager
Version: 1.23.90-1
Severity: important
Tags: ipv6
Symptom:
inet6 leases are acquired without (apparent) issue at connection start time,
but expire
without any attempt to renew (no log entry in journalctl containing "dhcp6").
inet4
addresses are properly acquired and renewed (with log entries including the
text dhcp4
in them).
Setup:
I am running behind a router with a dual ip4/ip6 setup. I get a long (90000sec)
lease
for ipv4, and a short (600sec) ipv6 lease.
This affects both ip6-privacy unset (which I believe means "on") and
ip6-privacy=0, as
set in the nm connection file:
# cat /proc/sys/net/ipv6/conf/default/use_tempaddr
2
This also happens 1.22.10-1.
Workaround:
Periodically call nmcli con up "${connection_name}"
Additional notes:
1. The inet6 address I get works: ping6 can get out.
2. This problem appeared after an upgrade to the router firmware. But, even if
there is an
issue with that firmware, why is no attempt made to renew the expiring inet6
address(es)?
Am I misunderstanding some detail about ipv6 address assignment?
3. Even before the router firmware upgrade, dhcp6 consistently connected with
some timeout issue:
Apr 07 06:36:46 hostname NetworkManager[2644]: <info> [1586263006.8448] dhcp6
(eth0): activation: beginning transaction (timeout in 45 seconds)
Apr 07 06:36:46 hostname NetworkManager[2644]: <info> [1586263006.8456]
policy: set 'eth0-dynamic' (eth0) as default for IPv6 routing and DNS
Apr 07 06:37:32 hostname NetworkManager[2644]: <warn> [1586263052.0247] dhcp6
(eth0): request timed out
Apr 07 06:37:32 hostname NetworkManager[2644]: <info> [1586263052.0247] dhcp6
(eth0): state changed unknown -> timeout
Apr 07 06:37:32 hostname NetworkManager[2644]: <info> [1586263052.0248] dhcp6
(eth0): canceled DHCP transaction
Apr 07 06:37:32 hostname NetworkManager[2644]: <info> [1586263052.0248] dhcp6
(eth0): state changed timeout -> done
This, however, did not cause trouble getting the inet6 address (I have other
records that indicate the address was
assigned).
Antonio
