Please note that diffs are not public domain; they are subject to the copyright notices on the relevant files. =================================================================== RCS file: /ftp/cvs/cvsroot/src/sys/netinet/ip_input.c,v rcsdiff: /ftp/cvs/cvsroot/src/sys/netinet/ip_input.c,v: warning: Unknown phrases like `commitid ...;' are present. retrieving revision 1.193 retrieving revision 1.218.2.3 diff -u -p -r1.193 -r1.218.2.3 --- src/sys/netinet/ip_input.c 2003/12/12 21:17:59 1.193 +++ src/sys/netinet/ip_input.c 2007/02/26 09:11:44 1.218.2.3 @@ -1,4 +1,4 @@ -/* $NetBSD: ip_input.c,v 1.193 2003/12/12 21:17:59 scw Exp $ */ +/* $NetBSD: ip_input.c,v 1.218.2.3 2007/02/26 09:11:44 yamt Exp $ */ /* * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. @@ -98,7 +98,7 @@ */ #include -__KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.193 2003/12/12 21:17:59 scw Exp $"); +__KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.218.2.3 2007/02/26 09:11:44 yamt Exp $"); #include "opt_inet.h" #include "opt_gateway.h" @@ -121,6 +121,7 @@ __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v #include #include #include +#include #include #include @@ -131,6 +132,7 @@ __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v #include #include #include +#include #include #include #include @@ -219,11 +221,10 @@ int ip_checkinterface = 0; struct rttimer_queue *ip_mtudisc_timeout_q = NULL; -extern struct domain inetdomain; int ipqmaxlen = IFQ_MAXLEN; u_long in_ifaddrhash; /* size of hash table - 1 */ int in_ifaddrentries; /* total number of addrs */ -struct in_ifaddrhead in_ifaddrhead; +struct in_ifaddrhead in_ifaddrhead; struct in_ifaddrhashhead *in_ifaddrhashtbl; u_long in_multihash; /* size of hash table - 1 */ int in_multientries; /* total number of addrs */ @@ -236,6 +237,19 @@ uint16_t ip_id; struct pfil_head inet_pfil_hook; #endif +/* + * Cached copy of nmbclusters. If nbclusters is different, + * recalculate IP parameters derived from nmbclusters. + */ +static int ip_nmbclusters; /* copy of nmbclusters */ +static void ip_nmbclusters_changed(void); /* recalc limits */ + +#define CHECK_NMBCLUSTER_PARAMS() \ +do { \ + if (__predict_false(ip_nmbclusters != nmbclusters)) \ + ip_nmbclusters_changed(); \ +} while (/*CONSTCOND*/0) + /* IP datagram reassembly queues (hashed) */ #define IPREASS_NHASH_LOG2 6 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) @@ -244,15 +258,33 @@ struct pfil_head inet_pfil_hook; (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) struct ipqhead ipq[IPREASS_NHASH]; int ipq_locked; -int ip_nfragpackets = 0; -int ip_maxfragpackets = 200; -int ip_nfrags = 0; /* total fragments in reass queues */ +static int ip_nfragpackets; /* packets in reass queue */ +static int ip_nfrags; /* total fragments in reass queues */ + +int ip_maxfragpackets = 200; /* limit on packets. XXX sysctl */ +int ip_maxfrags; /* limit on fragments. XXX sysctl */ + + +/* + * Additive-Increase/Multiplicative-Decrease (AIMD) strategy for + * IP reassembly queue buffer managment. + * + * We keep a count of total IP fragments (NB: not fragmented packets!) + * awaiting reassembly (ip_nfrags) and a limit (ip_maxfrags) on fragments. + * If ip_nfrags exceeds ip_maxfrags the limit, we drop half the + * total fragments in reassembly queues.This AIMD policy avoids + * repeatedly deleting single packets under heavy fragmentation load + * (e.g., from lossy NFS peers). + */ +static u_int ip_reass_ttl_decr(u_int ticks); +static void ip_reass_drophalf(void); + -static __inline int ipq_lock_try __P((void)); -static __inline void ipq_unlock __P((void)); +static inline int ipq_lock_try(void); +static inline void ipq_unlock(void); -static __inline int -ipq_lock_try() +static inline int +ipq_lock_try(void) { int s; @@ -270,8 +302,8 @@ ipq_lock_try() return (1); } -static __inline void -ipq_unlock() +static inline void +ipq_unlock(void) { int s; @@ -302,8 +334,8 @@ do { \ #define IPQ_UNLOCK() ipq_unlock() -struct pool inmulti_pool; -struct pool ipqent_pool; +POOL_INIT(inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl", NULL); +POOL_INIT(ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", NULL); #ifdef INET_CSUM_COUNTERS #include @@ -317,6 +349,10 @@ struct evcnt ip_swcsum = EVCNT_INITIALIZ #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ +EVCNT_ATTACH_STATIC(ip_hwcsum_bad); +EVCNT_ATTACH_STATIC(ip_hwcsum_ok); +EVCNT_ATTACH_STATIC(ip_swcsum); + #else #define INET_CSUM_COUNTER_INCR(ev) /* nothing */ @@ -338,28 +374,33 @@ static struct ip_srcrt { struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; } ip_srcrt; -static void save_rte __P((u_char *, struct in_addr)); +static void save_rte(u_char *, struct in_addr); #ifdef MBUFTRACE -struct mowner ip_rx_mowner = { "internet", "rx" }; -struct mowner ip_tx_mowner = { "internet", "tx" }; +struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx"); +struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx"); #endif /* + * Compute IP limits derived from the value of nmbclusters. + */ +static void +ip_nmbclusters_changed(void) +{ + ip_maxfrags = nmbclusters / 4; + ip_nmbclusters = nmbclusters; +} + +/* * IP initialization: fill in IP protocol switch table. * All protocols not implemented in kernel go to raw IP protocol handler. */ void -ip_init() +ip_init(void) { - struct protosw *pr; + const struct protosw *pr; int i; - pool_init(&inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl", - NULL); - pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", - NULL); - pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); if (pr == 0) panic("ip_init"); @@ -374,8 +415,11 @@ ip_init() for (i = 0; i < IPREASS_NHASH; i++) LIST_INIT(&ipq[i]); - ip_id = time.tv_sec & 0xfffff; + ip_id = time_second & 0xfffff; + ipintrq.ifq_maxlen = ipqmaxlen; + ip_nmbclusters_changed(); + TAILQ_INIT(&in_ifaddrhead); in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, M_WAITOK, &in_ifaddrhash); @@ -396,31 +440,28 @@ ip_init() "error %d\n", i); #endif /* PFIL_HOOKS */ -#ifdef INET_CSUM_COUNTERS - evcnt_attach_static(&ip_hwcsum_bad); - evcnt_attach_static(&ip_hwcsum_ok); - evcnt_attach_static(&ip_swcsum); -#endif /* INET_CSUM_COUNTERS */ - #ifdef MBUFTRACE MOWNER_ATTACH(&ip_tx_mowner); MOWNER_ATTACH(&ip_rx_mowner); #endif /* MBUFTRACE */ } -struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; +struct sockaddr_in ipaddr = { + .sin_len = sizeof(ipaddr), + .sin_family = AF_INET, +}; struct route ipforward_rt; /* * IP software interrupt routine */ void -ipintr() +ipintr(void) { int s; struct mbuf *m; - while (1) { + while (!IF_IS_EMPTY(&ipintrq)) { s = splnet(); IF_DEQUEUE(&ipintrq, m); splx(s); @@ -447,12 +488,13 @@ ip_input(struct mbuf *m) int downmatch; int checkif; int srcrt = 0; + int s; u_int hash; #ifdef FAST_IPSEC struct m_tag *mtag; struct tdb_ident *tdbi; struct secpolicy *sp; - int s, error; + int error; #endif /* FAST_IPSEC */ MCLAIM(m, &ip_rx_mowner); @@ -536,10 +578,16 @@ ip_input(struct mbuf *m) break; default: - /* Must compute it ourselves. */ - INET_CSUM_COUNTER_INCR(&ip_swcsum); - if (in_cksum(m, hlen) != 0) - goto bad; + /* + * Must compute it ourselves. Maybe skip checksum on + * loopback interfaces. + */ + if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & + IFF_LOOPBACK) || ip_do_loopback_cksum)) { + INET_CSUM_COUNTER_INCR(&ip_swcsum); + if (in_cksum(m, hlen) != 0) + goto badcsum; + } break; } @@ -613,6 +661,20 @@ ip_input(struct mbuf *m) return; ip = mtod(m, struct ip *); hlen = ip->ip_hl << 2; + /* + * XXX The setting of "srcrt" here is to prevent ip_forward() + * from generating ICMP redirects for packets that have + * been redirected by a hook back out on to the same LAN that + * they came from and is not an indication that the packet + * is being inffluenced by source routing options. This + * allows things like + * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp" + * where tlp0 is both on the 1.1.1.0/24 network and is the + * default route for hosts on 1.1.1.0/24. Of course this + * also requires a "map tlp0 ..." to complete the story. + * One might argue whether or not this kind of network config. + * should be supported in this manner... + */ srcrt = (odst.s_addr != ip->ip_dst.s_addr); } #endif /* PFIL_HOOKS */ @@ -675,8 +737,8 @@ ip_input(struct mbuf *m) } if (ia != NULL) goto ours; - if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { - TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) { + if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { + IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { if (ifa->ifa_addr->sa_family != AF_INET) continue; ia = ifatoia(ifa); @@ -702,14 +764,6 @@ ip_input(struct mbuf *m) #ifdef MROUTING extern struct socket *ip_mrouter; - if (M_READONLY(m)) { - if ((m = m_pullup(m, hlen)) == 0) { - ipstat.ips_toosmall++; - return; - } - ip = mtod(m, struct ip *); - } - if (ip_mrouter) { /* * If we are acting as a multicast router, all @@ -787,7 +841,7 @@ ip_input(struct mbuf *m) sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); } else { sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, - IP_FORWARDING, &error); + IP_FORWARDING, &error); } if (sp == NULL) { /* NB: can happen if error */ splx(s); @@ -841,13 +895,6 @@ ours: * but it's not worth the time; just let them time out.) */ if (ip->ip_off & ~htons(IP_DF|IP_RF)) { - if (M_READONLY(m)) { - if ((m = m_pullup(m, hlen)) == NULL) { - ipstat.ips_toosmall++; - goto bad; - } - ip = mtod(m, struct ip *); - } /* * Look for queue of fragments @@ -896,7 +943,9 @@ found: */ if (mff || ip->ip_off != htons(0)) { ipstat.ips_fragments++; + s = splvm(); ipqe = pool_get(&ipqent_pool, PR_NOWAIT); + splx(s); if (ipqe == NULL) { ipstat.ips_rcvmemdrop++; IPQ_UNLOCK(); @@ -932,7 +981,7 @@ found: goto bad; } #endif -#if FAST_IPSEC +#ifdef FAST_IPSEC /* * enforce IPsec policy checking if we are seeing last header. * note that we do not visit this with protocols with pcb layer @@ -944,7 +993,7 @@ found: * done. If so, then just pass it along. This tag gets * set during AH, ESP, etc. input handling, before the * packet is returned to the ip input queue for delivery. - */ + */ mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); s = splsoftnet(); if (mtag != NULL) { @@ -952,7 +1001,7 @@ found: sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); } else { sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, - IP_FORWARDING, &error); + IP_FORWARDING, &error); } if (sp != NULL) { /* @@ -1002,17 +1051,14 @@ badcsum: * is given as fp; otherwise have to make a chain. */ struct mbuf * -ip_reass(ipqe, fp, ipqhead) - struct ipqent *ipqe; - struct ipq *fp; - struct ipqhead *ipqhead; +ip_reass(struct ipqent *ipqe, struct ipq *fp, struct ipqhead *ipqhead) { struct mbuf *m = ipqe->ipqe_m; struct ipqent *nq, *p, *q; struct ip *ip; struct mbuf *t; int hlen = ipqe->ipqe_ip->ip_hl << 2; - int i, next; + int i, next, s; IPQ_LOCK_CHECK(); @@ -1023,11 +1069,20 @@ ip_reass(ipqe, fp, ipqhead) m->m_data += hlen; m->m_len -= hlen; +#ifdef notyet + /* make sure fragment limit is up-to-date */ + CHECK_NMBCLUSTER_PARAMS(); + + /* If we have too many fragments, drop the older half. */ + if (ip_nfrags >= ip_maxfrags) + ip_reass_drophalf(void); +#endif + /* * We are about to add a fragment; increment frag count. */ ip_nfrags++; - + /* * If first fragment to arrive, create a reassembly queue. */ @@ -1108,7 +1163,9 @@ ip_reass(ipqe, fp, ipqhead) nq = TAILQ_NEXT(q, ipqe_q); m_freem(q->ipqe_m); TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); + s = splvm(); pool_put(&ipqent_pool, q); + splx(s); fp->ipq_nfrags--; ip_nfrags--; } @@ -1149,11 +1206,15 @@ insert: m->m_next = 0; m_cat(m, t); nq = TAILQ_NEXT(q, ipqe_q); + s = splvm(); pool_put(&ipqent_pool, q); + splx(s); for (q = nq; q != NULL; q = nq) { t = q->ipqe_m; nq = TAILQ_NEXT(q, ipqe_q); + s = splvm(); pool_put(&ipqent_pool, q); + splx(s); m_cat(m, t); } ip_nfrags -= fp->ipq_nfrags; @@ -1178,6 +1239,7 @@ insert: for (t = m; t; t = t->m_next) plen += t->m_len; m->m_pkthdr.len = plen; + m->m_pkthdr.csum_flags = 0; } return (m); @@ -1187,7 +1249,9 @@ dropfrag: ip_nfrags--; ipstat.ips_fragdropped++; m_freem(m); + s = splvm(); pool_put(&ipqent_pool, ipqe); + splx(s); return (0); } @@ -1196,11 +1260,11 @@ dropfrag: * associated datagrams. */ void -ip_freef(fp) - struct ipq *fp; +ip_freef(struct ipq *fp) { struct ipqent *q, *p; u_int nfrags = 0; + int s; IPQ_LOCK_CHECK(); @@ -1209,7 +1273,9 @@ ip_freef(fp) m_freem(q->ipqe_m); nfrags++; TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); + s = splvm(); pool_put(&ipqent_pool, q); + splx(s); } if (nfrags != fp->ipq_nfrags) @@ -1221,30 +1287,100 @@ ip_freef(fp) } /* - * IP timer processing; - * if a timer expires on a reassembly - * queue, discard it. + * IP reassembly TTL machinery for multiplicative drop. */ -void -ip_slowtimo() +static u_int fragttl_histo[(IPFRAGTTL+1)]; + + +/* + * Decrement TTL of all reasembly queue entries by `ticks'. + * Count number of distinct fragments (as opposed to partial, fragmented + * datagrams) in the reassembly queue. While we traverse the entire + * reassembly queue, compute and return the median TTL over all fragments. + */ +static u_int +ip_reass_ttl_decr(u_int ticks) { - static u_int dropscanidx = 0; - u_int i; + u_int nfrags, median, dropfraction, keepfraction; struct ipq *fp, *nfp; - int s = splsoftnet(); + int i; + + nfrags = 0; + memset(fragttl_histo, 0, sizeof fragttl_histo); - IPQ_LOCK(); for (i = 0; i < IPREASS_NHASH; i++) { for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) { + fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ? + 0 : fp->ipq_ttl - ticks); nfp = LIST_NEXT(fp, ipq_q); - if (--fp->ipq_ttl == 0) { + if (fp->ipq_ttl == 0) { ipstat.ips_fragtimeout++; ip_freef(fp); + } else { + nfrags += fp->ipq_nfrags; + fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags; } } } + + KASSERT(ip_nfrags == nfrags); + + /* Find median (or other drop fraction) in histogram. */ + dropfraction = (ip_nfrags / 2); + keepfraction = ip_nfrags - dropfraction; + for (i = IPFRAGTTL, median = 0; i >= 0; i--) { + median += fragttl_histo[i]; + if (median >= keepfraction) + break; + } + + /* Return TTL of median (or other fraction). */ + return (u_int)i; +} + +void +ip_reass_drophalf(void) +{ + + u_int median_ticks; /* - * If we are over the maximum number of fragments + * Compute median TTL of all fragments, and count frags + * with that TTL or lower (roughly half of all fragments). + */ + median_ticks = ip_reass_ttl_decr(0); + + /* Drop half. */ + median_ticks = ip_reass_ttl_decr(median_ticks); + +} + +/* + * IP timer processing; + * if a timer expires on a reassembly + * queue, discard it. + */ +void +ip_slowtimo(void) +{ + static u_int dropscanidx = 0; + u_int i; + u_int median_ttl; + int s = splsoftnet(); + + IPQ_LOCK(); + + /* Age TTL of all fragments by 1 tick .*/ + median_ttl = ip_reass_ttl_decr(1); + + /* make sure fragment limit is up-to-date */ + CHECK_NMBCLUSTER_PARAMS(); + + /* If we have too many fragments, drop the older half. */ + if (ip_nfrags > ip_maxfrags) + ip_reass_ttl_decr(median_ttl); + + /* + * If we are over the maximum number of fragmented packets * (due to the limit being lowered), drain off * enough to get down to the new limit. Start draining * from the reassembly hashqueue most recently drained. @@ -1281,9 +1417,8 @@ ip_slowtimo() * Drain off all datagram fragments. */ void -ip_drain() +ip_drain(void) { - int i; /* * We may be called from a device's interrupt context. If @@ -1292,15 +1427,11 @@ ip_drain() if (ipq_lock_try() == 0) return; - for (i = 0; i < IPREASS_NHASH; i++) { - struct ipqhead *ipqh = &ipq[i]; - struct ipq *fp, *nfp; - for (fp = LIST_FIRST(ipqh); fp != NULL; fp = nfp) { - nfp = LIST_NEXT(fp, ipq_q); - ip_freef(fp); - ipstat.ips_fragdropped++; - } - } + /* + * Drop half the total fragments now. If more mbufs are needed, + * we will be called again soon. + */ + ip_reass_drophalf(); IPQ_UNLOCK(); } @@ -1313,8 +1444,7 @@ ip_drain() * 0 if the packet should be processed further. */ int -ip_dooptions(m) - struct mbuf *m; +ip_dooptions(struct mbuf *m) { struct ip *ip = mtod(m, struct ip *); u_char *cp, *cp0; @@ -1401,7 +1531,7 @@ ip_dooptions(m) bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, sizeof(ipaddr.sin_addr)); if (opt == IPOPT_SSRR) - ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))); + ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr))); else ia = ip_rtaddr(ipaddr.sin_addr); if (ia == 0) { @@ -1543,27 +1673,25 @@ bad: * return internet address info of interface to be used to get there. */ struct in_ifaddr * -ip_rtaddr(dst) - struct in_addr dst; +ip_rtaddr(struct in_addr dst) { - struct sockaddr_in *sin; + if (!in_hosteq(dst, satocsin(rtcache_getdst(&ipforward_rt))->sin_addr)) + rtcache_free(&ipforward_rt); + else + rtcache_check(&ipforward_rt); - sin = satosin(&ipforward_rt.ro_dst); + if (ipforward_rt.ro_rt == NULL) { + struct sockaddr_in *sin = satosin(&ipforward_rt.ro_dst); - if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) { - if (ipforward_rt.ro_rt) { - RTFREE(ipforward_rt.ro_rt); - ipforward_rt.ro_rt = 0; - } sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_addr = dst; - rtalloc(&ipforward_rt); + rtcache_init(&ipforward_rt); + if (ipforward_rt.ro_rt == NULL) + return NULL; } - if (ipforward_rt.ro_rt == 0) - return ((struct in_ifaddr *)0); - return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); + return ifatoia(ipforward_rt.ro_rt->rt_ifa); } /* @@ -1571,9 +1699,7 @@ ip_rtaddr(dst) * to be picked up later by ip_srcroute if the receiver is interested. */ void -save_rte(option, dst) - u_char *option; - struct in_addr dst; +save_rte(u_char *option, struct in_addr dst) { unsigned olen; @@ -1595,16 +1721,16 @@ save_rte(option, dst) * The first hop is placed before the options, will be removed later. */ struct mbuf * -ip_srcroute() +ip_srcroute(void) { struct in_addr *p, *q; struct mbuf *m; if (ip_nhops == 0) - return ((struct mbuf *)0); + return NULL; m = m_get(M_DONTWAIT, MT_SOOPTS); if (m == 0) - return ((struct mbuf *)0); + return NULL; MCLAIM(m, &inetdomain.dom_mowner); #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt)) @@ -1667,9 +1793,7 @@ ip_srcroute() * XXX should be deleted; last arg currently ignored. */ void -ip_stripoptions(m, mopt) - struct mbuf *m; - struct mbuf *mopt; +ip_stripoptions(struct mbuf *m, struct mbuf *mopt) { int i; struct ip *ip = mtod(m, struct ip *); @@ -1711,20 +1835,13 @@ const int inetctlerrmap[PRC_NCMDS] = { * via a source route. */ void -ip_forward(m, srcrt) - struct mbuf *m; - int srcrt; +ip_forward(struct mbuf *m, int srcrt) { struct ip *ip = mtod(m, struct ip *); - struct sockaddr_in *sin; struct rtentry *rt; - int error, type = 0, code = 0; + int error, type = 0, code = 0, destmtu = 0; struct mbuf *mcopy; n_long dest; - struct ifnet *destifp; -#if defined(IPSEC) || defined(FAST_IPSEC) - struct ifnet dummyifp; -#endif /* * We are now in the output path. @@ -1738,10 +1855,10 @@ ip_forward(m, srcrt) dest = 0; #ifdef DIAGNOSTIC - if (ipprintfs) - printf("forward: src %2.2x dst %2.2x ttl %x\n", - ntohl(ip->ip_src.s_addr), - ntohl(ip->ip_dst.s_addr), ip->ip_ttl); + if (ipprintfs) { + printf("forward: src %s ", inet_ntoa(ip->ip_src)); + printf("dst %s ttl %x\n", inet_ntoa(ip->ip_dst), ip->ip_ttl); + } #endif if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { ipstat.ips_cantforward++; @@ -1752,26 +1869,26 @@ ip_forward(m, srcrt) icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); return; } - ip->ip_ttl -= IPTTLDEC; - sin = satosin(&ipforward_rt.ro_dst); - if ((rt = ipforward_rt.ro_rt) == 0 || - !in_hosteq(ip->ip_dst, sin->sin_addr)) { - if (ipforward_rt.ro_rt) { - RTFREE(ipforward_rt.ro_rt); - ipforward_rt.ro_rt = 0; - } + if (!in_hosteq(ip->ip_dst, + satocsin(rtcache_getdst(&ipforward_rt))->sin_addr)) + rtcache_free(&ipforward_rt); + else + rtcache_check(&ipforward_rt); + if (ipforward_rt.ro_rt == NULL) { + struct sockaddr_in *sin = satosin(&ipforward_rt.ro_dst); + sin->sin_family = AF_INET; - sin->sin_len = sizeof(struct sockaddr_in); + sin->sin_len = sizeof(*sin); sin->sin_addr = ip->ip_dst; - rtalloc(&ipforward_rt); - if (ipforward_rt.ro_rt == 0) { - icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); + rtcache_init(&ipforward_rt); + if (ipforward_rt.ro_rt == NULL) { + icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); return; } - rt = ipforward_rt.ro_rt; } + rt = ipforward_rt.ro_rt; /* * Save at most 68 bytes of the packet in case @@ -1782,6 +1899,8 @@ ip_forward(m, srcrt) if (mcopy) mcopy = m_pullup(mcopy, ip->ip_hl << 2); + ip->ip_ttl -= IPTTLDEC; + /* * If forwarding packet using same interface that it came in on, * perhaps should send a redirect to sender to shortcut a hop. @@ -1815,7 +1934,7 @@ ip_forward(m, srcrt) } } - error = ip_output(m, (struct mbuf *)0, &ipforward_rt, + error = ip_output(m, NULL, &ipforward_rt, (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), (struct ip_moptions *)NULL, (struct socket *)NULL); @@ -1838,7 +1957,6 @@ ip_forward(m, srcrt) } if (mcopy == NULL) return; - destifp = NULL; switch (error) { @@ -1859,8 +1977,8 @@ ip_forward(m, srcrt) type = ICMP_UNREACH; code = ICMP_UNREACH_NEEDFRAG; #if !defined(IPSEC) && !defined(FAST_IPSEC) - if (ipforward_rt.ro_rt) - destifp = ipforward_rt.ro_rt->rt_ifp; + if (ipforward_rt.ro_rt != NULL) + destmtu = ipforward_rt.ro_rt->rt_ifp->if_mtu; #else /* * If the packet is routed over IPsec tunnel, tell the @@ -1868,7 +1986,7 @@ ip_forward(m, srcrt) * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz * XXX quickhack!!! */ - if (ipforward_rt.ro_rt) { + if (ipforward_rt.ro_rt != NULL) { struct secpolicy *sp; int ipsecerror; size_t ipsechdr; @@ -1879,7 +1997,7 @@ ip_forward(m, srcrt) &ipsecerror); if (sp == NULL) - destifp = ipforward_rt.ro_rt->rt_ifp; + destmtu = ipforward_rt.ro_rt->rt_ifp->if_mtu; else { /* count IPsec header size */ ipsechdr = ipsec4_hdrsiz(mcopy, @@ -1888,24 +2006,18 @@ ip_forward(m, srcrt) /* * find the correct route for outer IPv4 * header, compute tunnel MTU. - * - * XXX BUG ALERT - * The "dummyifp" code relies upon the fact - * that icmp_error() touches only ifp->if_mtu. */ - /*XXX*/ - destifp = NULL; + if (sp->req != NULL && sp->req->sav != NULL && sp->req->sav->sah != NULL) { ro = &sp->req->sav->sah->sa_route; if (ro->ro_rt && ro->ro_rt->rt_ifp) { - dummyifp.if_mtu = + destmtu = ro->ro_rt->rt_rmx.rmx_mtu ? ro->ro_rt->rt_rmx.rmx_mtu : ro->ro_rt->rt_ifp->if_mtu; - dummyifp.if_mtu -= ipsechdr; - destifp = &dummyifp; + destmtu -= ipsechdr; } } @@ -1937,15 +2049,12 @@ ip_forward(m, srcrt) break; #endif } - icmp_error(mcopy, type, code, dest, destifp); + icmp_error(mcopy, type, code, dest, destmtu); } void -ip_savecontrol(inp, mp, ip, m) - struct inpcb *inp; - struct mbuf **mp; - struct ip *ip; - struct mbuf *m; +ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, + struct mbuf *m) { if (inp->inp_socket->so_options & SO_TIMESTAMP) { @@ -2001,6 +2110,31 @@ ip_savecontrol(inp, mp, ip, m) } /* + * sysctl helper routine for net.inet.ip.forwsrcrt. + */ +static int +sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) +{ + int error, tmp; + struct sysctlnode node; + + node = *rnode; + tmp = ip_forwsrcrt; + node.sysctl_data = &tmp; + error = sysctl_lookup(SYSCTLFN_CALL(&node)); + if (error || newp == NULL) + return (error); + + if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT, + 0, NULL, NULL, NULL)) + return (EPERM); + + ip_forwsrcrt = tmp; + + return (0); +} + +/* * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the * range of the new value and tweaks timers if it changes. */ @@ -2038,7 +2172,7 @@ sysctl_net_inet_ip_maxflows(SYSCTLFN_ARG s = sysctl_lookup(SYSCTLFN_CALL(rnode)); if (s) return (s); - + s = splsoftnet(); ipflow_reap(0); splx(s); @@ -2052,132 +2186,199 @@ SYSCTL_SETUP(sysctl_net_inet_ip_setup, " { extern int subnetsarelocal, hostzeroisbroadcast; - sysctl_createv(SYSCTL_PERMANENT, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT, CTLTYPE_NODE, "net", NULL, NULL, 0, NULL, 0, CTL_NET, CTL_EOL); - sysctl_createv(SYSCTL_PERMANENT, - CTLTYPE_NODE, "inet", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT, + CTLTYPE_NODE, "inet", + SYSCTL_DESCR("PF_INET related settings"), NULL, 0, NULL, 0, CTL_NET, PF_INET, CTL_EOL); - sysctl_createv(SYSCTL_PERMANENT, - CTLTYPE_NODE, "ip", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT, + CTLTYPE_NODE, "ip", + SYSCTL_DESCR("IPv4 related settings"), NULL, 0, NULL, 0, CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); - - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "forwarding", NULL, + + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "forwarding", + SYSCTL_DESCR("Enable forwarding of INET datagrams"), NULL, 0, &ipforwarding, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_FORWARDING, CTL_EOL); - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "redirect", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "redirect", + SYSCTL_DESCR("Enable sending of ICMP redirect messages"), NULL, 0, &ipsendredirects, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_SENDREDIRECTS, CTL_EOL); - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "ttl", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "ttl", + SYSCTL_DESCR("Default TTL for an INET datagram"), NULL, 0, &ip_defttl, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_DEFTTL, CTL_EOL); #ifdef IPCTL_DEFMTU - sysctl_createv(SYSCTL_PERMANENT /* |SYSCTL_READWRITE? */, - CTLTYPE_INT, "mtu", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */, + CTLTYPE_INT, "mtu", + SYSCTL_DESCR("Default MTA for an INET route"), NULL, 0, &ip_mtu, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_DEFMTU, CTL_EOL); #endif /* IPCTL_DEFMTU */ - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READONLY1, - CTLTYPE_INT, "forwsrcrt", NULL, - NULL, 0, &ip_forwsrcrt, 0, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "forwsrcrt", + SYSCTL_DESCR("Enable forwarding of source-routed " + "datagrams"), + sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_FORWSRCRT, CTL_EOL); - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "directed-broadcast", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "directed-broadcast", + SYSCTL_DESCR("Enable forwarding of broadcast datagrams"), NULL, 0, &ip_directedbcast, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_DIRECTEDBCAST, CTL_EOL); - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "allowsrcrt", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "allowsrcrt", + SYSCTL_DESCR("Accept source-routed datagrams"), NULL, 0, &ip_allowsrcrt, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_ALLOWSRCRT, CTL_EOL); - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "subnetsarelocal", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "subnetsarelocal", + SYSCTL_DESCR("Whether logical subnets are considered " + "local"), NULL, 0, &subnetsarelocal, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_SUBNETSARELOCAL, CTL_EOL); - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "mtudisc", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "mtudisc", + SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"), NULL, 0, &ip_mtudisc, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_MTUDISC, CTL_EOL); - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "anonportmin", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "anonportmin", + SYSCTL_DESCR("Lowest ephemeral port number to assign"), sysctl_net_inet_ip_ports, 0, &anonportmin, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_ANONPORTMIN, CTL_EOL); - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "anonportmax", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "anonportmax", + SYSCTL_DESCR("Highest ephemeral port number to assign"), sysctl_net_inet_ip_ports, 0, &anonportmax, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_ANONPORTMAX, CTL_EOL); - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "mtudisctimeout", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "mtudisctimeout", + SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"), sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_MTUDISCTIMEOUT, CTL_EOL); #ifdef GATEWAY - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "maxflows", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "maxflows", + SYSCTL_DESCR("Number of flows for fast forwarding"), sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_MAXFLOWS, CTL_EOL); #endif /* GATEWAY */ - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "hostzerobroadcast", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "hostzerobroadcast", + SYSCTL_DESCR("All zeroes address is broadcast address"), NULL, 0, &hostzeroisbroadcast, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_HOSTZEROBROADCAST, CTL_EOL); #if NGIF > 0 - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "gifttl", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "gifttl", + SYSCTL_DESCR("Default TTL for a gif tunnel datagram"), NULL, 0, &ip_gif_ttl, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_GIF_TTL, CTL_EOL); #endif /* NGIF */ #ifndef IPNOPRIVPORTS - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "lowportmin", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "lowportmin", + SYSCTL_DESCR("Lowest privileged ephemeral port number " + "to assign"), sysctl_net_inet_ip_ports, 0, &lowportmin, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_LOWPORTMIN, CTL_EOL); - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "lowportmax", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "lowportmax", + SYSCTL_DESCR("Highest privileged ephemeral port number " + "to assign"), sysctl_net_inet_ip_ports, 0, &lowportmax, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_LOWPORTMAX, CTL_EOL); #endif /* IPNOPRIVPORTS */ - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "maxfragpackets", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "maxfragpackets", + SYSCTL_DESCR("Maximum number of fragments to retain for " + "possible reassembly"), NULL, 0, &ip_maxfragpackets, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_MAXFRAGPACKETS, CTL_EOL); #if NGRE > 0 - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "grettl", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "grettl", + SYSCTL_DESCR("Default TTL for a gre tunnel datagram"), NULL, 0, &ip_gre_ttl, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_GRE_TTL, CTL_EOL); #endif /* NGRE */ - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "checkinterface", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "checkinterface", + SYSCTL_DESCR("Enable receive side of Strong ES model " + "from RFC1122"), NULL, 0, &ip_checkinterface, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_CHECKINTERFACE, CTL_EOL); - sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, - CTLTYPE_INT, "random_id", NULL, + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "random_id", + SYSCTL_DESCR("Assign random ip_id values"), NULL, 0, &ip_do_randomid, 0, CTL_NET, PF_INET, IPPROTO_IP, IPCTL_RANDOMID, CTL_EOL); + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "do_loopback_cksum", + SYSCTL_DESCR("Perform IP checksum on loopback"), + NULL, 0, &ip_do_loopback_cksum, 0, + CTL_NET, PF_INET, IPPROTO_IP, + IPCTL_LOOPBACKCKSUM, CTL_EOL); + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT, + CTLTYPE_STRUCT, "stats", + SYSCTL_DESCR("IP statistics"), + NULL, 0, &ipstat, sizeof(ipstat), + CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS, + CTL_EOL); }