Fix source address selection on unbound sockets with multiple fibs.

[asomers]

Summary

Fix source address selection on unbound sockets in the presence of multiple
fibs. Use the mbuf's or the socket's fib instead of RT_ALL_FIBS. Fixes PR

187553. Also fixes netperf's UDP_STREAM test on a nondefault fib.

sys/netinet/ip_output.c
In ip_output, lookup the source address using the mbuf's fib instead
of RT_ALL_FIBS.

sys/netinet/in_pcb.c
in in_pcbladdr, lookup the source address using the socket's fib,
because we don't seem to have the mbuf fib. They should be the
same, though.

tests/sys/net/fibs_test.sh
Clear the expected failure on udp_dontroute.

Test Plan

sys/net/fibs_test:udp_dontroute

[hrs]

This patch looks good to me.

[melifaro]

Looks good

[bz]

I think I'd go ahead; I left the comment for the one ip_output.c case just in case we'll find other problems in the future.

sys/netinet/ip_output.c

256

I am not actually sure if these two instances in this case are correct. I wonder if in this case we should get the FIB from the interface? IP_ROUTETOIF is meant to bypass the routing table(s).

[asomers]

sys/netinet/ip_output.c

256

I think the mbuf fib is correct. We certainly can't use the interface fib, because at this point we don't yet know which interface to use. Furthermore, despite its description IP_ROUTETOIF can't mean to completely bypass the routing tables; we need the routing tables to do ifa_ifwithdstaddr and ifa_ifwithnet. IP_ROUTETOIF really just means to send the packet to a locally connected network instead of forwarding it through a router. Given that, our only real choice is whether to use the mbuf fib or RT_ALL_FIBS. If the system has multiple interfaces on the same network but with different fibs, then the mbuf fib (or the process fib) is the only way that the application has to request which interface to use. So I think that the mbuf fib is the best choice.

[asomers]

Committed as r271675.