RA not sent for PD

[timriker]

odhcp6c - 2023-05-12-bcd28363-20
odhcpd-ipv6only - 2023-10-24-d8118f6e-1

Log shows this:

Thu Sep 26 16:59:12 2024 daemon.info odhcpd[7832]: Address 2607:fa18:9ffe::1 (dprefix 0, valid 8617466) not suitable as RA route on lan

root@gw-xm-wr1200:~# ip a sh br-lan
15: br-lan: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP qlen 1000
    link/ether 00:01:9f:0a:1a:ec brd ff:ff:ff:ff:ff:ff
    inet 192.168.12.1/24 brd 192.168.12.255 scope global br-lan
       valid_lft forever preferred_lft forever
    inet 166.70.97.193/29 brd 166.70.97.199 scope global br-lan
       valid_lft forever preferred_lft forever
    inet6 2607:fa18:9ffe::1/56 scope global dynamic noprefixroute
       valid_lft 2578115sec preferred_lft 590915sec
    inet6 fe80::201:9fff:fe0a:1aec/64 scope link
       valid_lft forever preferred_lft forever

config dhcp 'lan'
        option interface 'lan'
        option start '100'
        option limit '150'
        option dhcpv4 'server'
        option leasetime '12h'
        option ra 'server'
        option dhcpv6 'server'
        option ra_default '2'

Why does it think that 2607:fa18:9ffe::1 is not suitable?

It looks like SLAAC and/or "auto" systems do get a route, but dhcpv6 systems do not. Perhaps the error log is not stopping RA packets?

[schuettecarsten]

I have the same problem (see #229). Have you been able to fix this in your config?

[timriker]

I have not. I'm now advertising the local ip instead of the ::1. I don't know why the ::1 won't work. It would be nice if there was an option to advertise fe80::1 instead.

[schuettecarsten]

I have not. I'm now advertising the local ip instead of the ::1. I don't know why the ::1 won't work. It would be nice if there was an option to advertise fe80::1 instead.

How do you do that?

[timriker]

My wan6 interface is set to: Request IPv6-prefix of length 48, as my ISP allows a /48 request.
lan has "IPv6 assignment length" as 56
lan IPv6 Settings has "RA-Service" set to server-mode and DHCPv6-Service set to server-mode
lan ipv6 RA Settings has "Default router" set to forced.

[schuettecarsten]

That's exactly my setup. It works for "guest" network with assignment hint 3, for "mgmt" with assignment hint a (10) but fails "lan" for assignment hint b2 (178). I can change that to 2, it fails. I compared everything from "lan" with "guest" and "mgmt" and cannot find any differences except the IP address ranges of course, and I cannot find the root cause. Weird....

I can see several of these linkes in logs, but I have no idea about it...

Tue Feb 18 22:06:37 2025 daemon.debug odhcpd[30643]: Using a RA lifetime of 1800 seconds on guest
Tue Feb 18 22:06:37 2025 daemon.info odhcpd[30643]: Address fd2b:43d7:e9a:3::1 (dprefix 0, valid-lifetime 4294967295) not suitable as RA route on guest
Tue Feb 18 22:06:37 2025 daemon.notice odhcpd[30643]: Sending a RA on guest
Tue Feb 18 22:06:37 2025 daemon.debug odhcpd[30643]: Sent 120 bytes to ff02::1%guest@br-guest
Tue Feb 18 22:06:37 2025 daemon.debug odhcpd[30643]: Using a RA lifetime of 1800 seconds on lan
Tue Feb 18 22:06:37 2025 daemon.info odhcpd[30643]: Address fd2b:43d7:e9a:b2::1 (dprefix 0, valid-lifetime 4294967295) not suitable as RA route on lan
Tue Feb 18 22:06:37 2025 daemon.notice odhcpd[30643]: Sending a RA on lan
Tue Feb 18 22:06:37 2025 daemon.debug odhcpd[30643]: Sent 120 bytes to ff02::1%lan@br-lan
Tue Feb 18 22:06:37 2025 daemon.debug odhcpd[30643]: Using a RA lifetime of 1800 seconds on mgmt
Tue Feb 18 22:06:37 2025 daemon.info odhcpd[30643]: Address fd2b:43d7:e9a:a::1 (dprefix 0, valid-lifetime 4294967295) not suitable as RA route on mgmt
Tue Feb 18 22:06:37 2025 daemon.notice odhcpd[30643]: Sending a RA on mgmt
Tue Feb 18 22:06:37 2025 daemon.debug odhcpd[30643]: Sent 120 bytes to ff02::1%mgmt@br-mgmt

[systemcrash]

I figure it's to do with noprefixroute. Was that delivered via upstream DHCPv6?

[schuettecarsten]

I figure it's to do with noprefixroute. Was that delivered via upstream DHCPv6?

I only work with local addresses and routes. odhcpd should announce them redardless of the upstream status. Also, it does on one interface (fd2b:43d7:e9a:be::1/64) but not on three others in a similar range (fd2b:43d7:e9a:3::1/64, fd2b:43d7:e9a:a::1/64, fd2b:43d7:e9a:b2::1/64). All interfaces have the exactly same dhcp configuration, only different VLANs on the interfaces and of course different ip subnets.

My IPv6 ULA prefix is fd2b:43d7:e9a::/56 and I get a /56 from my internet provider (Telekom). I use npt6 for address translation.

[schuettecarsten]

Maybe related to #199 and #219

[schuettecarsten]

I've tried to understand the source code. I have no idea if the following makes sense, maybe somebody (@systemcrash, @dedeckeh ?) can review this:

In line 820 of router.c, there is a check for addr->dprefix is 0, That causes my log messages:

odhcpd/src/router.c

Line 820 in 4308384

if (addr->dprefix >= 64 || addr->dprefix == 0 || addr->valid_lt <= (uint32_t)now) {

Somewhen earlier, when the routes are parsed, if code checks for default routes. But in case a default route was found, a flag is set (line 315), but dprefix is never set to a value, so it stays on the default value 0:

odhcpd/src/router.c

Line 315 in 4308384

found_default = true;

As line 327 is the only place where dprefix is set to something else than 0, I think that causes at least my problems with non-advertised RA routes:

odhcpd/src/router.c

Line 327 in 4308384

n[i].dprefix = p.prefix;

[systemcrash]

You might be onto something. I can say this, however: PD is intended to distribute /64 prefixes, especially those from upstream. But a /64 should be distributed based on your ULA. That is: the whole ULA, including the hint is /64, I think. In my config, I don't have as many interfaces as you have, but I have a /48 ULA and on interfaces:

IPv6 prefix filter = ULA
IPv6 assignment hint (whatever for that interface)
IPv6 assignment length = 64

Downstream devices get what they need.

noprefixroute is a flag used in network configurations that indicates that no route should be automatically created for a specific IP address in the routing table. This means that if the IP address is added, it won't have a corresponding route unless one is manually configured.

[schuettecarsten]

noprefixroute is a flag used in network configurations that indicates that no route should be automatically created for a specific IP address in the routing table. This means that if the IP address is added, it won't have a corresponding route unless one is manually configured.

That exactly what I do. I manage my local IPv6 addresses and use npt6.sh to translate them. So, the route that odhcpd should advertise to clients is to use the OpenWRT router as default gateway.