dhcpcd-discuss

Hi Graham

On 27/09/2017 11:14, Graham Breed wrote:
> Hello!
>
> I'm running dhcpcd for plain DHCPv6 address assignment (along with
> DHCPv4 and SLAAC).  Everything works except the DHCPv6 addresses are
> being created with /128 prefixes where I want /64.

As Harald said in his reply, this is not possible.
DHCPv6 addresses are just an addresss, there is no way to associate it 
with a prefix route, even if one currently exists.
This means that the prefix route has to be correctly setup via RA.

>  The router
> advertisements look correct and if I enable "AdvAutonomous" on the same
> subnet on the same link I get SLAAC addresses (from dhcpcd) with the
> correct prefix.
>
> I don't know where the prefix should come from and why it's not working.
>   I found a section of code for "RFC 5942 Section 5" but I patched it and
> it doesn't help so I don't think that's the problem.
>
>
> Here are some logs as the address gets assigned:
>
> Sep 27 08:59:35 [5605]: eth1: broadcasting SOLICIT6 (xid 0x32576f), next
> in 0.9 seconds
> Sep 27 08:59:35 [5605]: eth1: ADV
> fd70:a6a5:a58b:0:ffff:ffff:ffff:fff8/128 from fe80::a00:27ff:fe5d:6607
> Sep 27 08:59:35 [5605]: eth1: broadcasting REQUEST6 (xid 0x8f2ce2), next
> in 1.0 seconds
> Sep 27 08:59:35 [5605]: eth1: REPLY6 received from fe80::a00:27ff:fe5d:6607
> Sep 27 08:59:35 [5605]: eth1: adding address
> fd70:a6a5:a58b:0:ffff:ffff:ffff:fff8/128
> Sep 27 08:59:35 [5605]: eth1: pltime 1125 seconds, vltime 1800 seconds
> Sep 27 08:59:35 [5605]: eth1: renew in 562, rebind in 900, expire in
> 1800 seconds
> Sep 27 08:59:35 [5605]: eth1: writing lease `/var/db/dhcpcd/eth1.lease6'
> Sep 27 08:59:35 [5605]: eth1: waiting for DHCPv6 DAD to complete
>
>
> Here is the configuration:
>
> debug
> script /usr/sbin/dhcpcd_hook.py
> logfile /var/log/dhcp_client.log
>
> # Request nameservers from the server
> option domain_name_servers
>
> # From the example config:
> # Use the hardware address of the interface for the Client ID.
> clientid
> # Rapid commit support.
> # Safe to enable by default because it requires the equivalent option set
> # on the server to actually work.
> option rapid_commit
> # Respect the network MTU. This is applied to DHCP routes.
> option interface_mtu
> # A ServerID is required by RFC2131.
> require dhcp_server_identifier
>
>
> I'm running Linux 4.1.35.  net.ipv6.conf.eth1.accept_ra is 0.

Nothing above looks wrong or out of place.

> The arch-rival dhclient works correctly on a different machine with the
> same DHCPv6 server.

You've not said what dhclient version is.
Early versions of DHCPv6 in dhclient did try to erroneously associate 
the DHCPv6 address with a prefix - later versions behave correctly like 
dhcpcd and do not. Is this what you are seeing?

If not, can post address and routing output from both please and 
describe what you think is correct and incorrect about it?

I'm glad you've tried dhcpcd-6 and dhcpcd-7.0.0-rc2, but any fixes if 
needed will go into dhcpcd-7

Roy