Bug#892794: systemd-networkd fails to configure IPv6 without MTU from RA

[Cyril Brulebois]

Hi Russ,

Russ Allbery <rra at debian.org> (2018-03-12):
> Package: systemd
> Version: 232-25+deb9u2
> Severity: important
> Tags: ipv6
> 
> systemd 232-25+deb9u2 breaks IPv6 configuration with systemd-networkd
> if the RA doesn't include an MTU.  This is a regression from 232-25+deb9u1
> and breaks all IPv6 networking on the system in this network configuration.

I'm very sorry for that…

> This is presumbly due to the fix for #878162.
> 
> After upgrading to 232-25+deb9u2, I got error messages in the logs every
> minute:
> 
> Mar 12 18:02:08 haven systemd-networkd[25396]: eth0: Failed to get default router MTU from RA: No data available

Yeah, that seems very likely… Even if the word regression or the commit
ID doesn't appear in the commit message, the follow-up commit seems to
fix this specific issue:

    commit 29b5ad083a6925efec8e188013d1298742e0baaa
    Author: Jörg Thalheim <joerg at higgsboson.tk>
    Date:   Mon Dec 19 15:34:07 2016 +0100
    
        networkd-ndisc: handle missing mtu gracefully (#4913)
        
        At least bird's implementation of router advertisement does not
        set MTU option by default (instead it supplies an option to the user).
        In this case just leave MTU as it is.

 → https://github.com/systemd/systemd/commit/29b5ad083a6925efec8e188013d1298742e0baaa

Regarding the other pair of functions getting the return code checked
(sd_netlink_message_open_container & sd_netlink_message_close_container)
there doesn't seem to be any follow-up commits touching that file in
these areas… I'm not sure netlink operations are likely to fail anyway
(except for ENOMEM situations), so that's probably fine?

> It's possible that this means the local IPv6 routing environment is
> not correctly configured, but since this change was made in a stable
> update and IPv6 was working perfectly prior to this change, it still
> feels like a relatively serious regression to me.

Definitely, sorry for having proposed that commit without spotting it.

I can build a patched systemd with this extra cherry-pick for you to
double check; I'll post a link once it's available.


Cheers,
-- 
Cyril Brulebois (kibi at debian.org)            <https://debamax.com/>
D-I release manager -- Release team member -- Freelance Consultant
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://alioth-lists.debian.net/pipermail/pkg-systemd-maintainers/attachments/20180314/9b66b29a/attachment-0002.sig>