return success from mc_multicst(9e)

[rcgoodfellow]

This came up in IPv6 e2e testing with omicron probes.

Wiring up mc_multicst(9e) is required for omicron probes setting up addresses directly on opte devices. The NeighborAdvertisement HairpinAction expects multicast macs. But without mc_multicst(9e) plumbed, multicast macs will fall back to broadcast macs, the HairpinAction predicates will fail and ndp will go nowhere in the probe zone.

My read of mc_multicst(9e) is that it is both an opportunity for a driver to set up multicast filtering on a device, and to signal to the kernel that multicast addressing is supported for data links on said device. Since we're not doing any hardware-based multicast filtering at the moment, I think it's sufficient to simply signal multicast L2 address support to the kernel.

When we return ENOTSUP for the mc_multicst entry point, we see the following in the kernel logs when multicast addresses are assigned to OPTE data links.

ip: joining multicasts failed (4) on opte1 - will use link layer broadcasts for multicast

Which results in NDP traffic using broadcast macs instead of multicast macs.

I've taken this patch for a spin in a4x2. Everything comes up. NDP traffic flows over OPTE devices use multicast addressing and address setup works as expected.