Re: [B.A.T.M.A.N.] [RFC v3 14/19] batman-adv: Add multicast_mode mesh genl configuration

["Linus Lüssing" <linus.luessing@c0d3.blue>]

On Fri, Jan 04, 2019 at 08:51:56AM +0000, Jiri Pirko wrote:
> Fri, Jan 04, 2019 at 08:44:38AM CET, sven@narfation.org wrote:
> >On Friday, 4 January 2019 02.57.26 CET Linus Lüssing wrote:
> >> On Fri, Dec 07, 2018 at 02:58:41PM +0100, Sven Eckelmann wrote:
> >> > diff --git a/net/batman-adv/netlink.c b/net/batman-adv/netlink.c
> >> > index 8e3a1b6e..32762003 100644
> >> > --- a/net/batman-adv/netlink.c
> >> > +++ b/net/batman-adv/netlink.c
> >> > @@ -155,6 +155,7 @@ static const struct nla_policy batadv_netlink_policy[NUM_BATADV_ATTR] = {
> >> >  	[BATADV_ATTR_GW_SEL_CLASS]		= { .type = NLA_U32 },
> >> >  	[BATADV_ATTR_HOP_PENALTY]		= { .type = NLA_U8 },
> >> >  	[BATADV_ATTR_LOG_LEVEL]			= { .type = NLA_U32 },
> >> > +	[BATADV_ATTR_MULTICAST_MODE]		= { .type = NLA_U8 },
> >> >  };
> >> 
> >> Since the name is more generic than just a flag, would it make
> >> sense to make this a U32, just to stay flexible for the future?
> 
> If the attribute is supposed to be just a bool, the name would be
> probably better to change to "BATADV_ATTR_MULTICAST_ENABLED" or
> something.

Hm, BATADV_ATTR_MULTICAST_ENABLED sounds confusing to me. In a
layer 2 mesh network, multicast always needs to be enabled for
things like IP to work.

So setting this to 0 does not disable multicast.

What this option does is using "classic flooding" (0) for
multicast packets or dropping/unicasting if possible, if IGMP/MLD
snooping detected such potential (1).

And since there were many discussions regarding which additional
techniques and algorithms could be implemented I would expect more
options to follow.

And for complementary techniques, a bitfield of 32 bits size
could be useful, I guess?