Re: [PATCH] babel: Drop check for IF_MULTICAST interface flag

From: Ondrej Zajicek <santiago@crfreenet.org>
To: "Toke Høiland-Jørgensen" <toke@toke.dk>
Cc: bird-users@network.cz, wireguard@lists.zx2c4.com,
	Stefan Haller <stefan.haller@stha.de>
Subject: Re: [PATCH] babel: Drop check for IF_MULTICAST interface flag
Date: Mon, 19 Apr 2021 18:32:59 +0200	[thread overview]
Message-ID: <YH2wux4jIJuD4RS2@feanor.crfreenet.org> (raw)
In-Reply-To: <87y2degz09.fsf@toke.dk>

On Mon, Apr 19, 2021 at 03:55:18PM +0200, Toke Høiland-Jørgensen wrote:
> Ondrej Zajicek <santiago@crfreenet.org> writes:
> 
> > Is there a reason why to disregard the IF_MULTICAST flag? This seems to me
> > more like a bug in FreeBSD Wireguard implementation that should be fixed
> > there. Is this flag properly checked on Linux, or is there some reason why
> > the flag is missing?
> 
> We did fix Wireguard - see:
> https://git.zx2c4.com/wireguard-freebsd/patch/?id=a7a84a17faf784857f076e37aa4818f6b6c12a95
> 
> However, that didn't help, Babel still refused to use the interface.
> Looking at krt-sock.c, the IF_MULTICAST flag is only set on
> IFF_POINTOPOINT or IFF_BROADCAST on bsd. The Linux code (in netlink.c)
> has a further:
> 
>       if (fl & IFF_MULTICAST)
> 	f.flags |= IF_MULTICAST;
> 
> beneath the other flag checks, so maybe that's really what's missing on
> the BSD side?

Yes, it is likely that it is an issue in sysdep/bsd code.

> > Routing protocols in BIRD generally follow this flag (and perhaps use
> > it to switch to unicast-only mode), so i do not see why Babel should
> > behave differently.
> 
> Yeah, I do believe I originally copied that check from one of the other
> protocols. I can see how it makes sense to check the flag and change
> operation mode based on it, but given that Babel doesn't do that it just
> seems kinda redundant? If the interface *actually* is unable to send
> multicast packets, the subsequent socket operation is going to fail, and
> at least that produces an error message instead of just silently
> ignoring the interface like that flag check does :)

Well, i am OK with generating a warning in cases of non-matching interface
type, instead of ignoring it silently. (In contrast to iface down or missing
lladdr, which should be silent, as it may correct later.)

-- 
Elen sila lumenn' omentielvo

Ondrej 'Santiago' Zajicek (email: santiago@crfreenet.org)
OpenPGP encrypted e-mails preferred (KeyID 0x11DEADC3, wwwkeys.pgp.net)
"To err is human -- to blame it on a computer is even more so."