/* * Copyright (c) 1983, 1988, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef lint static char sccsid[] = "@(#)tables.c 8.1 (Berkeley) 6/5/93"; #endif /* not lint */ /* * Routing Table Management Daemon */ #include "defs.h" #include #include #include #ifndef DEBUG #define DEBUG 0 #endif #ifdef RTM_ADD #define FIXLEN(s) {if ((s)->sa_len == 0) (s)->sa_len = sizeof *(s);} #else #define FIXLEN(s) { } #endif int install = !DEBUG; /* if 1 call kernel */ /* * Lookup dst in the tables for an exact match. */ struct rt_entry * rtlookup(dst) struct sockaddr *dst; { register struct rt_entry *rt; register struct rthash *rh; register u_int hash; struct afhash h; int doinghost = 1; if (dst->sa_family >= af_max) return (0); (*afswitch[dst->sa_family].af_hash)(dst, &h); hash = h.afh_hosthash; rh = &hosthash[hash & ROUTEHASHMASK]; again: for (rt = rh->rt_forw; rt != (struct rt_entry *)rh; rt = rt->rt_forw) { if (rt->rt_hash != hash) continue; if (equal(&rt->rt_dst, dst)) return (rt); } if (doinghost) { doinghost = 0; hash = h.afh_nethash; rh = &nethash[hash & ROUTEHASHMASK]; goto again; } return (0); } struct sockaddr wildcard; /* zero valued cookie for wildcard searches */ /* * Find a route to dst as the kernel would. */ struct rt_entry * rtfind(dst) struct sockaddr *dst; { register struct rt_entry *rt; register struct rthash *rh; register u_int hash; struct afhash h; int af = dst->sa_family; int doinghost = 1, (*match)(); if (af >= af_max) return (0); (*afswitch[af].af_hash)(dst, &h); hash = h.afh_hosthash; rh = &hosthash[hash & ROUTEHASHMASK]; again: for (rt = rh->rt_forw; rt != (struct rt_entry *)rh; rt = rt->rt_forw) { if (rt->rt_hash != hash) continue; if (doinghost) { if (equal(&rt->rt_dst, dst)) return (rt); } else { if (rt->rt_dst.sa_family == af && (*match)(&rt->rt_dst, dst)) return (rt); } } if (doinghost) { doinghost = 0; hash = h.afh_nethash; rh = &nethash[hash & ROUTEHASHMASK]; match = afswitch[af].af_netmatch; goto again; } #ifdef notyet /* * Check for wildcard gateway, by convention network 0. */ if (dst != &wildcard) { dst = &wildcard, hash = 0; goto again; } #endif return (0); } rtadd(dst, gate, metric, state) struct sockaddr *dst, *gate; int metric, state; { struct afhash h; register struct rt_entry *rt; struct rthash *rh; int af = dst->sa_family, flags; u_int hash; if (af >= af_max) return; (*afswitch[af].af_hash)(dst, &h); flags = (*afswitch[af].af_rtflags)(dst); /* * Subnet flag isn't visible to kernel, move to state. XXX */ FIXLEN(dst); FIXLEN(gate); if (flags & RTF_SUBNET) { state |= RTS_SUBNET; flags &= ~RTF_SUBNET; } if (flags & RTF_HOST) { hash = h.afh_hosthash; rh = &hosthash[hash & ROUTEHASHMASK]; } else { hash = h.afh_nethash; rh = &nethash[hash & ROUTEHASHMASK]; } rt = (struct rt_entry *)malloc(sizeof (*rt)); if (rt == 0) return; rt->rt_hash = hash; rt->rt_dst = *dst; rt->rt_router = *gate; rt->rt_timer = 0; rt->rt_flags = RTF_UP | flags; rt->rt_state = state | RTS_CHANGED; rt->rt_ifp = if_ifwithdstaddr(&rt->rt_dst); if (rt->rt_ifp == 0) rt->rt_ifp = if_ifwithnet(&rt->rt_router); if ((state & RTS_INTERFACE) == 0) rt->rt_flags |= RTF_GATEWAY; rt->rt_metric = metric; insque(rt, rh); TRACE_ACTION("ADD", rt); /* * If the ioctl fails because the gateway is unreachable * from this host, discard the entry. This should only * occur because of an incorrect entry in /etc/gateways. */ if ((rt->rt_state & (RTS_INTERNAL | RTS_EXTERNAL)) == 0 && rtioctl(ADD, &rt->rt_rt) < 0) { if (errno != EEXIST && gate->sa_family < af_max) syslog(LOG_ERR, "adding route to net/host %s through gateway %s: %m\n", (*afswitch[dst->sa_family].af_format)(dst), (*afswitch[gate->sa_family].af_format)(gate)); perror("ADD ROUTE"); if (errno == ENETUNREACH) { TRACE_ACTION("DELETE", rt); remque(rt); free((char *)rt); } } } rtchange(rt, gate, metric) struct rt_entry *rt; struct sockaddr *gate; short metric; { int add = 0, delete = 0, newgateway = 0; struct rtuentry oldroute; FIXLEN(gate); FIXLEN(&(rt->rt_router)); FIXLEN(&(rt->rt_dst)); if (!equal(&rt->rt_router, gate)) { newgateway++; TRACE_ACTION("CHANGE FROM ", rt); } else if (metric != rt->rt_metric) TRACE_NEWMETRIC(rt, metric); if ((rt->rt_state & RTS_INTERNAL) == 0) { /* * If changing to different router, we need to add * new route and delete old one if in the kernel. * If the router is the same, we need to delete * the route if has become unreachable, or re-add * it if it had been unreachable. */ if (newgateway) { add++; if (rt->rt_metric != HOPCNT_INFINITY) delete++; } else if (metric == HOPCNT_INFINITY) delete++; else if (rt->rt_metric == HOPCNT_INFINITY) add++; } if (delete) oldroute = rt->rt_rt; if ((rt->rt_state & RTS_INTERFACE) && delete) { rt->rt_state &= ~RTS_INTERFACE; rt->rt_flags |= RTF_GATEWAY; if (metric > rt->rt_metric && delete) syslog(LOG_ERR, "%s route to interface %s (timed out)", add? "changing" : "deleting", rt->rt_ifp ? rt->rt_ifp->int_name : "?"); } if (add) { rt->rt_router = *gate; rt->rt_ifp = if_ifwithdstaddr(&rt->rt_router); if (rt->rt_ifp == 0) rt->rt_ifp = if_ifwithnet(&rt->rt_router); } rt->rt_metric = metric; rt->rt_state |= RTS_CHANGED; if (newgateway) TRACE_ACTION("CHANGE TO ", rt); #ifndef RTM_ADD if (add && rtioctl(ADD, &rt->rt_rt) < 0) perror("ADD ROUTE"); if (delete && rtioctl(DELETE, &oldroute) < 0) perror("DELETE ROUTE"); #else if (delete && !add) { if (rtioctl(DELETE, &oldroute) < 0) perror("DELETE ROUTE"); } else if (!delete && add) { if (rtioctl(ADD, &rt->rt_rt) < 0) perror("ADD ROUTE"); } else if (delete && add) { if (rtioctl(CHANGE, &rt->rt_rt) < 0) perror("CHANGE ROUTE"); } #endif } rtdelete(rt) struct rt_entry *rt; { TRACE_ACTION("DELETE", rt); FIXLEN(&(rt->rt_router)); FIXLEN(&(rt->rt_dst)); if (rt->rt_metric < HOPCNT_INFINITY) { if ((rt->rt_state & (RTS_INTERFACE|RTS_INTERNAL)) == RTS_INTERFACE) syslog(LOG_ERR, "deleting route to interface %s? (timed out?)", rt->rt_ifp->int_name); if ((rt->rt_state & (RTS_INTERNAL | RTS_EXTERNAL)) == 0 && rtioctl(DELETE, &rt->rt_rt) < 0) perror("rtdelete"); } remque(rt); free((char *)rt); } rtdeleteall(sig) int sig; { register struct rthash *rh; register struct rt_entry *rt; struct rthash *base = hosthash; int doinghost = 1; again: for (rh = base; rh < &base[ROUTEHASHSIZ]; rh++) { rt = rh->rt_forw; for (; rt != (struct rt_entry *)rh; rt = rt->rt_forw) { if (rt->rt_state & RTS_INTERFACE || rt->rt_metric >= HOPCNT_INFINITY) continue; TRACE_ACTION("DELETE", rt); if ((rt->rt_state & (RTS_INTERNAL|RTS_EXTERNAL)) == 0 && rtioctl(DELETE, &rt->rt_rt) < 0) perror("rtdeleteall"); } } if (doinghost) { doinghost = 0; base = nethash; goto again; } exit(sig); } /* * If we have an interface to the wide, wide world, * add an entry for an Internet default route (wildcard) to the internal * tables and advertise it. This route is not added to the kernel routes, * but this entry prevents us from listening to other people's defaults * and installing them in the kernel here. */ rtdefault() { extern struct sockaddr inet_default; rtadd(&inet_default, &inet_default, 1, RTS_CHANGED | RTS_PASSIVE | RTS_INTERNAL); } rtinit() { register struct rthash *rh; for (rh = nethash; rh < &nethash[ROUTEHASHSIZ]; rh++) rh->rt_forw = rh->rt_back = (struct rt_entry *)rh; for (rh = hosthash; rh < &hosthash[ROUTEHASHSIZ]; rh++) rh->rt_forw = rh->rt_back = (struct rt_entry *)rh; } rtioctl(action, ort) int action; struct rtuentry *ort; { #ifndef RTM_ADD if (install == 0) return (errno = 0); ort->rtu_rtflags = ort->rtu_flags; switch (action) { case ADD: return (ioctl(s, SIOCADDRT, (char *)ort)); case DELETE: return (ioctl(s, SIOCDELRT, (char *)ort)); default: return (-1); } #else /* RTM_ADD */ struct { struct rt_msghdr w_rtm; struct sockaddr_in w_dst; struct sockaddr w_gate; struct sockaddr_in w_netmask; } w; #define rtm w.w_rtm bzero((char *)&w, sizeof(w)); rtm.rtm_msglen = sizeof(w); rtm.rtm_version = RTM_VERSION; rtm.rtm_type = (action == ADD ? RTM_ADD : (action == DELETE ? RTM_DELETE : RTM_CHANGE)); #undef rt_dst rtm.rtm_flags = ort->rtu_flags; rtm.rtm_seq = ++seqno; rtm.rtm_addrs = RTA_DST|RTA_GATEWAY; bcopy((char *)&ort->rtu_dst, (char *)&w.w_dst, sizeof(w.w_dst)); bcopy((char *)&ort->rtu_router, (char *)&w.w_gate, sizeof(w.w_gate)); w.w_dst.sin_family = AF_INET; w.w_dst.sin_len = sizeof(w.w_dst); w.w_gate.sa_family = AF_INET; w.w_gate.sa_len = sizeof(w.w_gate); if (rtm.rtm_flags & RTF_HOST) { rtm.rtm_msglen -= sizeof(w.w_netmask); } else { register char *cp; int len; rtm.rtm_addrs |= RTA_NETMASK; w.w_netmask.sin_addr.s_addr = inet_maskof(w.w_dst.sin_addr.s_addr); for (cp = (char *)(1 + &w.w_netmask.sin_addr); --cp > (char *) &w.w_netmask; ) if (*cp) break; len = cp - (char *)&w.w_netmask; if (len) { len++; w.w_netmask.sin_len = len; len = 1 + ((len - 1) | (sizeof(long) - 1)); } else len = sizeof(long); rtm.rtm_msglen -= (sizeof(w.w_netmask) - len); } errno = 0; return (install ? write(r, (char *)&w, rtm.rtm_msglen) : (errno = 0)); #endif /* RTM_ADD */ }