Re: [babel] About ICMP and unnumbered routers

[Donald Eastlake <[e-mail]>]

Hi Juliusz/David,

Thanks for the research. See below.

On Sun, Aug 2, 2020 at 3:08 PM Juliusz Chroboczek <[e-mail]> wrote:
> Dear all,
>
> This mail is about the consequences of draft-bastian-babel-v4ov6-00 on
> ICMPv4 generation.
>
> At last week's meeting, Margaret kindly pointed out that we didn't do our
> homework, and that we didn't consider how ICMPv4 is sent out when the
> sending router has no IPv4 addresses.  I've looked at a few RFCs, and here
> are my tentative conclusions.
>
> RFC 1812 Section 2.2.7 says:
>
>    In this scheme, a router that
>    has unnumbered point to point lines also has a special IP address,
>    called a router-id in this memo.  The router-id is one of the
>    router's IP addresses (a router is required to have at least one IP
>    address).  This router-id is used as if it is the IP address of all
>    unnumbered interfaces.
>
> So it looks like it was decided back in 1995 that all IPv4 routers need at
> least one IPv4 address.  Much as I tried, I couldn't find anything that
> would relax this requirement.  As expected, when an ICMPv4 packet is sent
> over an unnumbered interface, the distinguished "router-id" address is
> used as the source address:
>
>    the IP source address in an ICMP message originated by the router MUST
>    be one of the IP addresses associated with the physical interface over
>    which the ICMP message is transmitted.  If the interface has no IP
>    addresses associated with it, the router's router-id (see Section
>    [5.2.5]) is used instead.
>
> (This appears to be a typo -- there's nothing relevant in Section 5.2.5.)
>
> Interestingly, there appears to be nothing in RFC 1812 that specifies that
> the router-id is globally unique; thus, what we could do would be to put
> the same IPv4 address on all routers.  This would make traceroute a little
> confusing (RFC 5837 might help), but it would solve both the problem with
> ICMPv4 blackholes and compliance with RFC 1812.
>
> In the light of the above, and in the immortal words of Lenin, what shall we do?

> 1. Pick a private IPv4 (say 10.42.42.42) and put it on the loopback
>    interface of all unnumbered routers.

I would recommend against this due to possible collisions.

> 2. Request a dedicated IPv4 address (the "all-routers anycast address") to
>    be used as the loopback address of unnumbered routers.

I think this is the best way forward. Probably the "IPv4 dummy
address" David Schinazi found would do fine. (If there is any problem
with that, getting a new special use IPv4 address for this purpose
would be pretty easy. The assignment requirement is "IETF Review"
which means any document that goes through IETF Last call will do as
the basis for a new assignment.
https://www.iana.org/assignments/iana-ipv4-special-registry/iana-ipv4-special-registry.xhtml)

> 3. Update RFC 1812 to allow unnumbered routers and 0.0.0.0 as the source
>    address of ICMPv4 packets.

Updating RFC 1812 seems like a more difficult path and possibly there
are routers or middle boxes that discard packets with a 0.0.0.0 source
as bogus. I don't know if the text is still valid but RFC 1122 says of
0.0.0.0 source:

                 This host on this network.  MUST NOT be sent, except as
                 a source address as part of an initialization procedure
                 by which the host learns its own IP address.

Thanks,
Donald
===============================
 Donald E. Eastlake 3rd   [phone] (cell)
 2386 Panoramic Circle, Apopka, FL 32703 USA
 [e-mail]

> I'm interested in the comments of wise people.
>
> -- Juliusz