Index: head/sys/netinet6/icmp6.c =================================================================== --- head/sys/netinet6/icmp6.c +++ head/sys/netinet6/icmp6.c @@ -2147,7 +2147,7 @@ * source address of the erroneous packet. */ in6_splitscope(&ip6->ip6_src, &dst6, &scopeid); - error = in6_selectsrc_addr(RT_DEFAULT_FIB, &dst6, + error = in6_selectsrc_addr(M_GETFIB(m), &dst6, scopeid, NULL, &src6, &hlim); if (error) { @@ -2289,7 +2289,7 @@ uint32_t scopeid; in6_splitscope(&reddst6, &kdst, &scopeid); - if (fib6_lookup_nh_basic(RT_DEFAULT_FIB, &kdst, scopeid, 0, 0,&nh6)==0){ + if (fib6_lookup_nh_basic(ifp->if_fib, &kdst, scopeid, 0, 0,&nh6)==0){ if ((nh6.nh_flags & NHF_GATEWAY) == 0) { nd6log((LOG_ERR, "ICMP6 redirect rejected; no route " Index: head/sys/netinet6/in6.c =================================================================== --- head/sys/netinet6/in6.c +++ head/sys/netinet6/in6.c @@ -159,6 +159,7 @@ struct sockaddr_dl gateway; struct sockaddr_in6 mask, addr; struct rtentry rt; + int fibnum; /* * initialize for rtmsg generation @@ -176,8 +177,9 @@ rt.rt_flags = RTF_HOST | RTF_STATIC; if (cmd == RTM_ADD) rt.rt_flags |= RTF_UP; - /* Announce arrival of local address to all FIBs. */ - rt_newaddrmsg(cmd, &ia->ia_ifa, 0, &rt); + fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ia62ifa(ia)->ifa_ifp->if_fib; + /* Announce arrival of local address to this FIB. */ + rt_newaddrmsg_fib(cmd, &ia->ia_ifa, 0, &rt, fibnum); } int @@ -2117,15 +2119,15 @@ uint32_t scopeid; int error; char ip6buf[INET6_ADDRSTRLEN]; + int fibnum; KASSERT(l3addr->sa_family == AF_INET6, ("sin_family %d", l3addr->sa_family)); - /* Our local addresses are always only installed on the default FIB. */ - sin6 = (const struct sockaddr_in6 *)l3addr; in6_splitscope(&sin6->sin6_addr, &dst, &scopeid); - error = fib6_lookup_nh_basic(RT_DEFAULT_FIB, &dst, scopeid, 0, 0, &nh6); + fibnum = V_rt_add_addr_allfibs ? RT_DEFAULT_FIB : ifp->if_fib; + error = fib6_lookup_nh_basic(fibnum, &dst, scopeid, 0, 0, &nh6); if (error != 0 || (nh6.nh_flags & NHF_GATEWAY) || nh6.nh_ifp != ifp) { struct ifaddr *ifa; /* Index: head/sys/netinet6/in6_src.c =================================================================== --- head/sys/netinet6/in6_src.c +++ head/sys/netinet6/in6_src.c @@ -297,7 +297,7 @@ */ /* get the outgoing interface */ if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp, - (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB)) != 0) + (inp != NULL) ? inp->inp_inc.inc_fibnum : fibnum)) != 0) return (error); #ifdef DIAGNOSTIC @@ -563,7 +563,7 @@ uint32_t fibnum; int error; - fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB; + fibnum = inp->inp_inc.inc_fibnum; retifp = NULL; error = in6_selectsrc(fibnum, dstsock, opts, inp, cred, &retifp, srcp); Index: head/sys/netinet6/nd6.h =================================================================== --- head/sys/netinet6/nd6.h +++ head/sys/netinet6/nd6.h @@ -469,6 +469,7 @@ void nd6_rs_input(struct mbuf *, int, int); void nd6_ra_input(struct mbuf *, int, int); void defrouter_reset(void); +void defrouter_select_fib(int fibnum); void defrouter_select(void); void defrouter_ref(struct nd_defrouter *); void defrouter_rele(struct nd_defrouter *); Index: head/sys/netinet6/nd6.c =================================================================== --- head/sys/netinet6/nd6.c +++ head/sys/netinet6/nd6.c @@ -157,6 +157,7 @@ struct sockaddr_dl gw; struct ifnet *ifp; int type; + int fibnum; LLE_WLOCK_ASSERT(lle); @@ -194,8 +195,9 @@ rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&dst; rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw; rtinfo.rti_addrs = RTA_DST | RTA_GATEWAY; + fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ifp->if_fib; rt_missmsg_fib(type, &rtinfo, RTF_HOST | RTF_LLDATA | ( - type == RTM_ADD ? RTF_UP: 0), 0, RT_DEFAULT_FIB); + type == RTM_ADD ? RTF_UP: 0), 0, fibnum); } /* @@ -1200,7 +1202,7 @@ if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { /* Refresh default router list. */ - defrouter_select(); + defrouter_select_fib(ifp->if_fib); } } @@ -1253,7 +1255,7 @@ nd6_is_new_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) { struct nd_prefix *pr; - struct ifaddr *dstaddr; + struct ifaddr *ifa; struct rt_addrinfo info; struct sockaddr_in6 rt_key; const struct sockaddr *dst6; @@ -1287,9 +1289,6 @@ bzero(&info, sizeof(info)); info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key; - /* Always use the default FIB here. XXME - why? */ - fibnum = RT_DEFAULT_FIB; - /* * If the address matches one of our addresses, * it should be a neighbor. @@ -1303,19 +1302,31 @@ continue; if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) { - /* Always use the default FIB here. */ dst6 = (const struct sockaddr *)&pr->ndpr_prefix; - genid = V_nd6_list_genid; - ND6_RUNLOCK(); - - /* Restore length field before retrying lookup */ - rt_key.sin6_len = sizeof(rt_key); - error = rib_lookup_info(fibnum, dst6, 0, 0, &info); + /* + * We only need to check all FIBs if add_addr_allfibs + * is unset. If set, checking any FIB will suffice. + */ + fibnum = V_rt_add_addr_allfibs ? rt_numfibs - 1 : 0; + for (; fibnum < rt_numfibs; fibnum++) { + genid = V_nd6_list_genid; + ND6_RUNLOCK(); - ND6_RLOCK(); - if (genid != V_nd6_list_genid) - goto restart; + /* + * Restore length field before + * retrying lookup + */ + rt_key.sin6_len = sizeof(rt_key); + error = rib_lookup_info(fibnum, dst6, 0, 0, + &info); + + ND6_RLOCK(); + if (genid != V_nd6_list_genid) + goto restart; + if (error == 0) + break; + } if (error != 0) continue; @@ -1346,13 +1357,18 @@ * If the address is assigned on the node of the other side of * a p2p interface, the address should be a neighbor. */ - dstaddr = ifa_ifwithdstaddr((const struct sockaddr *)addr, RT_ALL_FIBS); - if (dstaddr != NULL) { - if (dstaddr->ifa_ifp == ifp) { - ifa_free(dstaddr); - return (1); + if (ifp->if_flags & IFF_POINTOPOINT) { + IF_ADDR_RLOCK(ifp); + TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { + if (ifa->ifa_addr->sa_family != addr->sin6_family) + continue; + if (ifa->ifa_dstaddr != NULL && + sa_equal(addr, ifa->ifa_dstaddr)) { + IF_ADDR_RUNLOCK(ifp); + return 1; + } } - ifa_free(dstaddr); + IF_ADDR_RUNLOCK(ifp); } /* @@ -1485,7 +1501,7 @@ /* * We need to unlock to avoid a LOR with rt6_flush() with the * rnh and for the calls to pfxlist_onlink_check() and - * defrouter_select() in the block further down for calls + * defrouter_select_fib() in the block further down for calls * into nd6_lookup(). We still hold a ref. */ LLE_WUNLOCK(ln); @@ -1500,7 +1516,7 @@ if (dr) { /* - * Since defrouter_select() does not affect the + * Since defrouter_select_fib() does not affect the * on-link determination and MIP6 needs the check * before the default router selection, we perform * the check now. @@ -1510,7 +1526,7 @@ /* * Refresh default router list. */ - defrouter_select(); + defrouter_select_fib(dr->ifp->if_fib); } /* @@ -2104,11 +2120,11 @@ * Question: can we restrict the first condition to the "is_newentry" * case? * XXX: when we hear an RA from a new router with the link-layer - * address option, defrouter_select() is called twice, since + * address option, defrouter_select_fib() is called twice, since * defrtrlist_update called the function as well. However, I believe * we can compromise the overhead, since it only happens the first * time. - * XXX: although defrouter_select() should not have a bad effect + * XXX: although defrouter_select_fib() should not have a bad effect * for those are not autoconfigured hosts, we explicitly avoid such * cases for safety. */ @@ -2117,7 +2133,7 @@ /* * guaranteed recursion */ - defrouter_select(); + defrouter_select_fib(ifp->if_fib); } } Index: head/sys/netinet6/nd6_nbr.c =================================================================== --- head/sys/netinet6/nd6_nbr.c +++ head/sys/netinet6/nd6_nbr.c @@ -262,8 +262,7 @@ bzero(&info, sizeof(info)); info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway; - /* Always use the default FIB. */ - if (rib_lookup_info(RT_DEFAULT_FIB, (struct sockaddr *)&dst6, + if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6, 0, 0, &info) == 0) { if ((info.rti_flags & RTF_ANNOUNCE) != 0 && rt_gateway.sdl_family == AF_LINK) { @@ -485,7 +484,7 @@ uint32_t scopeid; in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid); - error = in6_selectsrc_addr(RT_DEFAULT_FIB, &dst6, + error = in6_selectsrc_addr(fibnum, &dst6, scopeid, ifp, &src6, NULL); if (error) { char ip6buf[INET6_ADDRSTRLEN]; @@ -982,7 +981,7 @@ * Select a source whose scope is the same as that of the dest. */ in6_splitscope(&daddr6, &dst6, &scopeid); - error = in6_selectsrc_addr(RT_DEFAULT_FIB, &dst6, + error = in6_selectsrc_addr(fibnum, &dst6, scopeid, ifp, &src6, NULL); if (error) { char ip6buf[INET6_ADDRSTRLEN]; Index: head/sys/netinet6/nd6_rtr.c =================================================================== --- head/sys/netinet6/nd6_rtr.c +++ head/sys/netinet6/nd6_rtr.c @@ -500,7 +500,7 @@ error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&def, (struct sockaddr *)&gate, (struct sockaddr *)&mask, - RTF_GATEWAY, &newrt, RT_DEFAULT_FIB); + RTF_GATEWAY, &newrt, new->ifp->if_fib); if (newrt) { nd6_rtmsg(RTM_ADD, newrt); /* tell user process */ RTFREE(newrt); @@ -551,8 +551,8 @@ /* * Remove the default route for a given router. - * This is just a subroutine function for defrouter_select(), and should - * not be called from anywhere else. + * This is just a subroutine function for defrouter_select_fib(), and + * should not be called from anywhere else. */ static void defrouter_delreq(struct nd_defrouter *dr) @@ -571,7 +571,7 @@ in6_rtrequest(RTM_DELETE, (struct sockaddr *)&def, (struct sockaddr *)&gate, - (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt, RT_DEFAULT_FIB); + (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt, dr->ifp->if_fib); if (oldrt) { nd6_rtmsg(RTM_DELETE, oldrt); RTFREE(oldrt); @@ -698,11 +698,11 @@ /* * If the router is the primary one, choose a new one. - * Note that defrouter_select() will remove the current gateway - * from the routing table. + * Note that defrouter_select_fib() will remove the current + * gateway from the routing table. */ if (deldr) - defrouter_select(); + defrouter_select_fib(deldr->ifp->if_fib); /* * Release the list reference. @@ -730,13 +730,23 @@ * even when the multipath routing is available, because we're not sure about * the benefits for stub hosts comparing to the risk of making the code * complicated and the possibility of introducing bugs. + * + * We maintain a single list of routers for multiple FIBs, only considering one + * at a time based on the receiving interface's FIB. If @fibnum is RT_ALL_FIBS, + * we do the whole thing multiple times. */ void -defrouter_select(void) +defrouter_select_fib(int fibnum) { struct nd_defrouter *dr, *selected_dr, *installed_dr; struct llentry *ln = NULL; + if (fibnum == RT_ALL_FIBS) { + for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { + defrouter_select_fib(fibnum); + } + } + ND6_RLOCK(); /* * Let's handle easy case (3) first: @@ -755,7 +765,7 @@ selected_dr = installed_dr = NULL; TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { IF_AFDATA_RLOCK(dr->ifp); - if (selected_dr == NULL && + if (selected_dr == NULL && dr->ifp->if_fib == fibnum && (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) && ND6_IS_LLINFO_PROBREACH(ln)) { selected_dr = dr; @@ -767,14 +777,17 @@ ln = NULL; } - if (dr->installed) { + if (dr->installed && dr->ifp->if_fib == fibnum) { if (installed_dr == NULL) { installed_dr = dr; defrouter_ref(installed_dr); } else { - /* this should not happen. warn for diagnosis. */ - log(LOG_ERR, - "defrouter_select: more than one router is installed\n"); + /* + * this should not happen. + * warn for diagnosis. + */ + log(LOG_ERR, "defrouter_select_fib: more than " + "one router is installed\n"); } } } @@ -789,14 +802,24 @@ if (selected_dr == NULL) { if (installed_dr == NULL || TAILQ_NEXT(installed_dr, dr_entry) == NULL) - selected_dr = TAILQ_FIRST(&V_nd_defrouter); + dr = TAILQ_FIRST(&V_nd_defrouter); else - selected_dr = TAILQ_NEXT(installed_dr, dr_entry); - defrouter_ref(selected_dr); + dr = TAILQ_NEXT(installed_dr, dr_entry); + + /* Ensure we select a router for this FIB. */ + TAILQ_FOREACH_FROM(dr, &V_nd_defrouter, dr_entry) { + if (dr->ifp->if_fib == fibnum) { + selected_dr = dr; + defrouter_ref(selected_dr); + break; + } + } } else if (installed_dr != NULL) { IF_AFDATA_RLOCK(installed_dr->ifp); - if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) && + if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, + installed_dr->ifp)) && ND6_IS_LLINFO_PROBREACH(ln) && + installed_dr->ifp->if_fib == fibnum && rtpref(selected_dr) <= rtpref(installed_dr)) { defrouter_rele(selected_dr); selected_dr = installed_dr; @@ -808,18 +831,30 @@ ND6_RUNLOCK(); /* - * If the selected router is different than the installed one, - * remove the installed router and install the selected one. - * Note that the selected router is never NULL here. + * If we selected a router for this FIB and it's different + * than the installed one, remove the installed router and + * install the selected one in its place. */ if (installed_dr != selected_dr) { if (installed_dr != NULL) { defrouter_delreq(installed_dr); defrouter_rele(installed_dr); } - defrouter_addreq(selected_dr); + if (selected_dr != NULL) + defrouter_addreq(selected_dr); } - defrouter_rele(selected_dr); + if (selected_dr != NULL) + defrouter_rele(selected_dr); +} + +/* + * Maintain old KPI for default router selection. + * If unspecified, we can re-select routers for all FIBs. + */ +void +defrouter_select(void) +{ + defrouter_select_fib(RT_ALL_FIBS); } /* @@ -942,7 +977,7 @@ V_nd6_list_genid++; ND6_WUNLOCK(); - defrouter_select(); + defrouter_select_fib(new->ifp->if_fib); return (n); } @@ -1731,7 +1766,7 @@ struct rtentry *rt; struct sockaddr_in6 mask6; u_long rtflags; - int error, a_failure, fibnum; + int error, a_failure, fibnum, maxfib; /* * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs. @@ -1742,8 +1777,15 @@ mask6.sin6_addr = pr->ndpr_mask; rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP; + if(V_rt_add_addr_allfibs) { + fibnum = 0; + maxfib = rt_numfibs; + } else { + fibnum = ifa->ifa_ifp->if_fib; + maxfib = fibnum + 1; + } a_failure = 0; - for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { + for (; fibnum < maxfib; fibnum++) { rt = NULL; error = in6_rtrequest(RTM_ADD, @@ -1831,6 +1873,10 @@ if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0) continue; + if (!V_rt_add_addr_allfibs && + opr->ndpr_ifp->if_fib != pr->ndpr_ifp->if_fib) + continue; + if (opr->ndpr_plen == pr->ndpr_plen && in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr, &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) { @@ -1891,7 +1937,7 @@ struct rtentry *rt; char ip6buf[INET6_ADDRSTRLEN]; uint64_t genid; - int fibnum, a_failure; + int fibnum, maxfib, a_failure; ND6_ONLINK_LOCK_ASSERT(); ND6_UNLOCK_ASSERT(); @@ -1909,8 +1955,16 @@ mask6.sin6_len = sizeof(sa6); bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr)); + if (V_rt_add_addr_allfibs) { + fibnum = 0; + maxfib = rt_numfibs; + } else { + fibnum = ifp->if_fib; + maxfib = fibnum + 1; + } + a_failure = 0; - for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { + for (; fibnum < maxfib; fibnum++) { rt = NULL; error = in6_rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL, (struct sockaddr *)&mask6, 0, &rt, fibnum); Index: head/tests/sys/netinet/fibs_test.sh =================================================================== --- head/tests/sys/netinet/fibs_test.sh +++ head/tests/sys/netinet/fibs_test.sh @@ -163,7 +163,6 @@ loopback_and_network_routes_on_nondefault_fib_inet6_body() { - atf_expect_fail "PR196361 IPv6 network routes don't respect net.add_addr_allfibs=0" # Configure the TAP interface to use a nonrouteable RFC3849 # address and a non-default fib ADDR="2001:db8::2" @@ -452,6 +451,7 @@ } slaac_on_nondefault_fib6_body() { + atf_skip "BUG217871 SLAAC on a newly created epair sometimes fails to add routes" # Configure the epair interfaces to use nonrouteable RFC3849 # addresses and non-default FIBs ADDR="2001:db8::2" @@ -459,8 +459,6 @@ SUBNET="2001:db8:" MASK="64" - atf_expect_fail "PR196361 IPv6 network routes don't respect net.add_addr_allfibs=0" - # Check system configuration if [ 0 != `sysctl -n net.add_addr_allfibs` ]; then atf_skip "This test requires net.add_addr_allfibs=0" @@ -478,8 +476,8 @@ # Configure epair interfaces get_epair setup_iface "$EPAIRA" "$FIB0" inet6 ${ADDR} ${MASK} - echo ifconfig "$EPAIRB" up inet6 fib $FIB1 -ifdisabled accept_rtadv - ifconfig "$EPAIRB" inet6 -ifdisabled accept_rtadv fib $FIB1 up + echo setfib $FIB1 ifconfig "$EPAIRB" inet6 -ifdisabled accept_rtadv fib $FIB1 up + setfib $FIB1 ifconfig "$EPAIRB" inet6 -ifdisabled accept_rtadv fib $FIB1 up rtadvd -p rtadvd.pid -C rtadvd.sock -c /dev/null "$EPAIRA" rtsol "$EPAIRB" @@ -514,7 +512,7 @@ cleanup_ifaces if [ -f "rtadvd.pid" ]; then pkill -F rtadvd.pid - rm rtadvd.pid + rm -f rtadvd.pid fi if [ -f "rfc6204w3.state" ] ; then sysctl "net.inet6.ip6.rfc6204w3"=`cat "rfc6204w3.state"` @@ -685,8 +683,6 @@ TARGET="2001:db8::100" SRCDIR=`atf_get_srcdir` - atf_expect_fail "PR196361 IPv6 network routes don't respect net.add_addr_allfibs=0" - # Check system configuration if [ 0 != `sysctl -n net.add_addr_allfibs` ]; then atf_skip "This test requires net.add_addr_allfibs=0" @@ -767,7 +763,7 @@ local EPAIRD if EPAIRD=`ifconfig epair create`; then - # Record the TAP device so we can clean it up later + # Record the epair device so we can clean it up later echo ${EPAIRD} >> "ifaces_to_cleanup" EPAIRA=${EPAIRD} EPAIRB=${EPAIRD%a}b