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 1.180 retrieving revision 1.250.4.2 diff -u -p -r1.180 -r1.250.4.2 --- src/sys/netinet/ip_input.c 2003/11/10 20:03:29 1.180 +++ src/sys/netinet/ip_input.c 2007/09/03 16:49:04 1.250.4.2 @@ -1,4 +1,4 @@ -/* $NetBSD: ip_input.c,v 1.180 2003/11/10 20:03:29 jonathan Exp $ */ +/* $NetBSD: ip_input.c,v 1.250.4.2 2007/09/03 16:49:04 jmcneill Exp $ */ /* * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. @@ -98,8 +98,9 @@ */ #include -__KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.180 2003/11/10 20:03:29 jonathan Exp $"); +__KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.250.4.2 2007/09/03 16:49:04 jmcneill Exp $"); +#include "opt_inet.h" #include "opt_gateway.h" #include "opt_pfil_hooks.h" #include "opt_ipsec.h" @@ -120,6 +121,7 @@ __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v #include #include #include +#include #include #include @@ -130,6 +132,7 @@ __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v #include #include #include +#include #include #include #include @@ -197,6 +200,9 @@ int ip_mtudisc_timeout = IPMTUDISCTIMEOU #ifdef DIAGNOSTIC int ipprintfs = 0; #endif + +int ip_do_randomid = 0; + /* * XXX - Setting ip_checkinterface mostly implements the receive side of * the Strong ES model described in RFC 1122, but since the routing table @@ -215,32 +221,70 @@ 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_ifaddr; +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 */ struct in_multihashhead *in_multihashtbl; struct ifqueue ipintrq; struct ipstat ipstat; +uint16_t ip_id; #ifdef PFIL_HOOKS struct pfil_head inet_pfil_hook; #endif -struct ipqhead ipq; +/* + * 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) +#define IPREASS_HMASK (IPREASS_NHASH - 1) +#define IPREASS_HASH(x,y) \ + (((((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; +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() +static inline int ipq_lock_try(void); +static inline void ipq_unlock(void); + +static inline int +ipq_lock_try(void) { int s; @@ -258,8 +302,8 @@ ipq_lock_try() return (1); } -static __inline void -ipq_unlock() +static inline void +ipq_unlock(void) { int s; @@ -290,8 +334,10 @@ 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, + IPL_SOFTNET); +POOL_INIT(ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", NULL, + IPL_VM); #ifdef INET_CSUM_COUNTERS #include @@ -305,6 +351,10 @@ 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 */ @@ -326,28 +376,33 @@ 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"); @@ -358,16 +413,23 @@ ip_init() if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) ip_protox[pr->pr_protocol] = pr - inetsw; - LIST_INIT(&ipq); + + for (i = 0; i < IPREASS_NHASH; i++) + LIST_INIT(&ipq[i]); + + ip_id = time_second & 0xfffff; + ipintrq.ifq_maxlen = ipqmaxlen; - TAILQ_INIT(&in_ifaddr); + ip_nmbclusters_changed(); + + TAILQ_INIT(&in_ifaddrhead); in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR, M_WAITOK, &in_ifaddrhash); in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IPMADDR, M_WAITOK, &in_multihash); ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout); #ifdef GATEWAY - ipflow_init(); + ipflow_init(ip_hashsize); #endif #ifdef PFIL_HOOKS @@ -380,31 +442,28 @@ 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); @@ -431,11 +490,13 @@ 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); @@ -448,7 +509,7 @@ ip_input(struct mbuf *m) * If no IP addresses have been set yet but the interfaces * are receiving, can't do anything with incoming packets yet. */ - if (TAILQ_FIRST(&in_ifaddr) == 0) + if (TAILQ_FIRST(&in_ifaddrhead) == 0) goto bad; ipstat.ips_total++; /* @@ -457,7 +518,7 @@ ip_input(struct mbuf *m) * it. Otherwise, if it is aligned, make sure the entire * base IP header is in the first mbuf of the chain. */ - if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) { + if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) { if ((m = m_copyup(m, sizeof(struct ip), (max_linkhdr + 3) & ~3)) == NULL) { /* XXXJRT new stat, please */ @@ -519,10 +580,16 @@ 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; } @@ -555,7 +622,7 @@ ip_input(struct mbuf *m) m_adj(m, len - m->m_pkthdr.len); } -#ifdef IPSEC +#if defined(IPSEC) /* ipflow (IP fast forwarding) is not compatible with IPsec. */ m->m_flags &= ~M_CANFASTFWD; #else @@ -580,6 +647,8 @@ ip_input(struct mbuf *m) */ #ifdef IPSEC if (!ipsec_getnhist(m)) +#elif defined(FAST_IPSEC) + if (!ipsec_indone(m)) #else if (1) #endif @@ -594,6 +663,20 @@ 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 */ @@ -656,8 +739,8 @@ 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); @@ -683,14 +766,6 @@ 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 @@ -768,7 +843,7 @@ 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); @@ -787,6 +862,26 @@ ip_input(struct mbuf *m) ipstat.ips_cantforward++; goto bad; } + + /* + * Peek at the outbound SP for this packet to determine if + * it's a Fast Forward candidate. + */ + mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); + if (mtag != NULL) + m->m_flags &= ~M_CANFASTFWD; + else { + s = splsoftnet(); + sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, + (IP_FORWARDING | + (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), + &error, NULL); + if (sp != NULL) { + m->m_flags &= ~M_CANFASTFWD; + KEY_FREESP(&sp); + } + splx(s); + } #endif /* FAST_IPSEC */ ip_forward(m, srcrt); @@ -802,25 +897,21 @@ 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 * of this datagram. */ IPQ_LOCK(); - LIST_FOREACH(fp, &ipq, ipq_q) + hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); + LIST_FOREACH(fp, &ipq[hash], ipq_q) { if (ip->ip_id == fp->ipq_id && in_hosteq(ip->ip_src, fp->ipq_src) && in_hosteq(ip->ip_dst, fp->ipq_dst) && ip->ip_p == fp->ipq_p) goto found; + + } fp = 0; found: @@ -852,7 +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(); @@ -861,7 +954,7 @@ found: ipqe->ipqe_mff = mff; ipqe->ipqe_m = m; ipqe->ipqe_ip = ip; - m = ip_reass(ipqe, fp); + m = ip_reass(ipqe, fp, &ipq[hash]); if (m == 0) { IPQ_UNLOCK(); return; @@ -888,7 +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 @@ -900,7 +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) { @@ -908,7 +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) { /* @@ -958,16 +1051,14 @@ badcsum: * is given as fp; otherwise have to make a chain. */ struct mbuf * -ip_reass(ipqe, fp) - struct ipqent *ipqe; - struct ipq *fp; +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(); @@ -978,6 +1069,20 @@ ip_reass(ipqe, fp) 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. */ @@ -997,7 +1102,8 @@ ip_reass(ipqe, fp) M_FTABLE, M_NOWAIT); if (fp == NULL) goto dropfrag; - LIST_INSERT_HEAD(&ipq, fp, ipq_q); + LIST_INSERT_HEAD(ipqhead, fp, ipq_q); + fp->ipq_nfrags = 1; fp->ipq_ttl = IPFRAGTTL; fp->ipq_p = ipqe->ipqe_ip->ip_p; fp->ipq_id = ipqe->ipqe_ip->ip_id; @@ -1006,6 +1112,8 @@ ip_reass(ipqe, fp) fp->ipq_dst = ipqe->ipqe_ip->ip_dst; p = NULL; goto insert; + } else { + fp->ipq_nfrags++; } /* @@ -1055,7 +1163,11 @@ ip_reass(ipqe, fp) 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--; } insert: @@ -1094,13 +1206,18 @@ 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; /* * Create header for new ip packet by @@ -1122,13 +1239,19 @@ 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); dropfrag: + if (fp != 0) + fp->ipq_nfrags--; + ip_nfrags--; ipstat.ips_fragdropped++; m_freem(m); + s = splvm(); pool_put(&ipqent_pool, ipqe); + splx(s); return (0); } @@ -1137,58 +1260,153 @@ 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(); for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { p = TAILQ_NEXT(q, ipqe_q); 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) + printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags); + ip_nfrags -= nfrags; LIST_REMOVE(fp, ipq_q); FREE(fp, M_FTABLE); ip_nfragpackets--; } /* + * IP reassembly TTL machinery for multiplicative drop. + */ +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) +{ + u_int nfrags, median, dropfraction, keepfraction; + struct ipq *fp, *nfp; + int i; + + nfrags = 0; + memset(fragttl_histo, 0, sizeof fragttl_histo); + + 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) { + 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; + /* + * 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() +ip_slowtimo(void) { - struct ipq *fp, *nfp; + static u_int dropscanidx = 0; + u_int i; + u_int median_ttl; int s = splsoftnet(); IPQ_LOCK(); - for (fp = LIST_FIRST(&ipq); fp != NULL; fp = nfp) { - nfp = LIST_NEXT(fp, ipq_q); - if (--fp->ipq_ttl == 0) { - ipstat.ips_fragtimeout++; - ip_freef(fp); - } - } + + /* 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 fragments + * 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. + * enough to get down to the new limit. Start draining + * from the reassembly hashqueue most recently drained. */ if (ip_maxfragpackets < 0) ; else { - while (ip_nfragpackets > ip_maxfragpackets && LIST_FIRST(&ipq)) - ip_freef(LIST_FIRST(&ipq)); + int wrapped = 0; + + i = dropscanidx; + while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) { + while (LIST_FIRST(&ipq[i]) != NULL) + ip_freef(LIST_FIRST(&ipq[i])); + if (++i >= IPREASS_NHASH) { + i = 0; + } + /* + * Dont scan forever even if fragment counters are + * wrong: stop after scanning entire reassembly queue. + */ + if (i == dropscanidx) + wrapped = 1; + } + dropscanidx = i; } IPQ_UNLOCK(); -#ifdef GATEWAY - ipflow_slowtimo(); -#endif splx(s); } @@ -1196,7 +1414,7 @@ ip_slowtimo() * Drain off all datagram fragments. */ void -ip_drain() +ip_drain(void) { /* @@ -1206,10 +1424,11 @@ ip_drain() if (ipq_lock_try() == 0) return; - while (LIST_FIRST(&ipq) != NULL) { - ipstat.ips_fragdropped++; - ip_freef(LIST_FIRST(&ipq)); - } + /* + * Drop half the total fragments now. If more mbufs are needed, + * we will be called again soon. + */ + ip_reass_drophalf(); IPQ_UNLOCK(); } @@ -1222,8 +1441,7 @@ 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; @@ -1307,10 +1525,10 @@ ip_dooptions(m) /* * locate outgoing interface */ - bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr, + bcopy((void *)(cp + off), (void *)&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) { @@ -1319,8 +1537,8 @@ ip_dooptions(m) goto bad; } ip->ip_dst = ipaddr.sin_addr; - bcopy((caddr_t)&ia->ia_addr.sin_addr, - (caddr_t)(cp + off), sizeof(struct in_addr)); + bcopy((void *)&ia->ia_addr.sin_addr, + (void *)(cp + off), sizeof(struct in_addr)); cp[IPOPT_OFFSET] += sizeof(struct in_addr); /* * Let ip_intr's mcast routing check handle mcast pkts @@ -1343,7 +1561,7 @@ ip_dooptions(m) off--; /* 0 origin */ if ((off + sizeof(struct in_addr)) > optlen) break; - bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr, + bcopy((void *)(&ip->ip_dst), (void *)&ipaddr.sin_addr, sizeof(ipaddr.sin_addr)); /* * locate outgoing interface; if we're the destination, @@ -1356,8 +1574,8 @@ ip_dooptions(m) code = ICMP_UNREACH_HOST; goto bad; } - bcopy((caddr_t)&ia->ia_addr.sin_addr, - (caddr_t)(cp + off), sizeof(struct in_addr)); + bcopy((void *)&ia->ia_addr.sin_addr, + (void *)(cp + off), sizeof(struct in_addr)); cp[IPOPT_OFFSET] += sizeof(struct in_addr); break; @@ -1426,7 +1644,7 @@ ip_dooptions(m) } ntime = iptime(); cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */ - bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1, + memmove((char *)cp + ipt->ipt_ptr - 1, cp0, sizeof(n_time)); ipt->ipt_ptr += sizeof(n_time); } @@ -1452,27 +1670,20 @@ 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; + struct rtentry *rt; + union { + struct sockaddr dst; + struct sockaddr_in dst4; + } u; - sin = satosin(&ipforward_rt.ro_dst); + sockaddr_in_init(&u.dst4, &dst, 0); - 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; + if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) + return NULL; - rtalloc(&ipforward_rt); - } - if (ipforward_rt.ro_rt == 0) - return ((struct in_ifaddr *)0); - return (ifatoia(ipforward_rt.ro_rt->rt_ifa)); + return ifatoia(rt->rt_ifa); } /* @@ -1480,9 +1691,7 @@ 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; @@ -1493,7 +1702,7 @@ save_rte(option, dst) #endif /* 0 */ if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst))) return; - bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen); + bcopy((void *)option, (void *)ip_srcrt.srcopt, olen); ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); ip_srcrt.dst = dst; } @@ -1504,16 +1713,16 @@ 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)) @@ -1541,9 +1750,9 @@ ip_srcroute() */ ip_srcrt.nop = IPOPT_NOP; ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; - bcopy((caddr_t)&ip_srcrt.nop, - mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ); - q = (struct in_addr *)(mtod(m, caddr_t) + + memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop, + OPTSIZ); + q = (struct in_addr *)(mtod(m, char *) + sizeof(struct in_addr) + OPTSIZ); #undef OPTSIZ /* @@ -1576,19 +1785,17 @@ 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 *); - caddr_t opts; + void *opts; int olen; olen = (ip->ip_hl << 2) - sizeof (struct ip); - opts = (caddr_t)(ip + 1); + opts = (void *)(ip + 1); i = m->m_len - (sizeof (struct ip) + olen); - bcopy(opts + olen, opts, (unsigned)i); + memmove(opts, (char *)opts + olen, (unsigned)i); m->m_len -= olen; if (m->m_flags & M_PKTHDR) m->m_pkthdr.len -= olen; @@ -1620,20 +1827,17 @@ 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 + union { + struct sockaddr dst; + struct sockaddr_in dst4; + } u; /* * We are now in the output path. @@ -1647,10 +1851,10 @@ 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++; @@ -1661,25 +1865,11 @@ 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; - } - sin->sin_family = AF_INET; - sin->sin_len = sizeof(struct sockaddr_in); - 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); - return; - } - rt = ipforward_rt.ro_rt; + sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); + if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) { + icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0); + return; } /* @@ -1691,6 +1881,8 @@ 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. @@ -1701,7 +1893,7 @@ ip_forward(m, srcrt) */ if (rt->rt_ifp == m->m_pkthdr.rcvif && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && - !in_nullhost(satosin(rt_key(rt))->sin_addr) && + !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) && ipsendredirects && !srcrt) { if (rt->rt_ifa && (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) == @@ -1724,7 +1916,7 @@ 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); @@ -1747,7 +1939,6 @@ ip_forward(m, srcrt) } if (mcopy == NULL) return; - destifp = NULL; switch (error) { @@ -1768,8 +1959,8 @@ 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 @@ -1777,7 +1968,7 @@ 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; @@ -1788,7 +1979,7 @@ 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, @@ -1797,24 +1988,18 @@ 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; } } @@ -1846,28 +2031,25 @@ 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) { struct timeval tv; microtime(&tv); - *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), + *mp = sbcreatecontrol((void *) &tv, sizeof(tv), SCM_TIMESTAMP, SOL_SOCKET); if (*mp) mp = &(*mp)->m_next; } if (inp->inp_flags & INP_RECVDSTADDR) { - *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, + *mp = sbcreatecontrol((void *) &ip->ip_dst, sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; @@ -1881,14 +2063,14 @@ ip_savecontrol(inp, mp, ip, m) */ /* options were tossed already */ if (inp->inp_flags & INP_RECVOPTS) { - *mp = sbcreatecontrol((caddr_t) opts_deleted_above, + *mp = sbcreatecontrol((void *) opts_deleted_above, sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } /* ip_srcroute doesn't do what we want here, need to fix */ if (inp->inp_flags & INP_RECVRETOPTS) { - *mp = sbcreatecontrol((caddr_t) ip_srcroute(), + *mp = sbcreatecontrol((void *) ip_srcroute(), sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; @@ -1897,175 +2079,322 @@ ip_savecontrol(inp, mp, ip, m) if (inp->inp_flags & INP_RECVIF) { struct sockaddr_dl sdl; - sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); - sdl.sdl_family = AF_LINK; - sdl.sdl_index = m->m_pkthdr.rcvif ? - m->m_pkthdr.rcvif->if_index : 0; - sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; - *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, - IP_RECVIF, IPPROTO_IP); + sockaddr_dl_init(&sdl, sizeof(sdl), + (m->m_pkthdr.rcvif != NULL) + ? m->m_pkthdr.rcvif->if_index + : 0, + 0, NULL, 0, NULL, 0); + *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP); if (*mp) mp = &(*mp)->m_next; } } -int -ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen) - int *name; - u_int namelen; - void *oldp; - size_t *oldlenp; - void *newp; - size_t newlen; +/* + * sysctl helper routine for net.inet.ip.forwsrcrt. + */ +static int +sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS) { - extern int subnetsarelocal, hostzeroisbroadcast; - - int error, old; + int error, tmp; + struct sysctlnode node; - /* All sysctl names (except ifq.*) at this level are terminal. */ - if ((namelen != 1) && !(namelen == 2 && name[0] == IPCTL_IFQ)) - return (ENOTDIR); - - switch (name[0]) { - case IPCTL_FORWARDING: - return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding)); - case IPCTL_SENDREDIRECTS: - return (sysctl_int(oldp, oldlenp, newp, newlen, - &ipsendredirects)); - case IPCTL_DEFTTL: - return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl)); -#ifdef notyet - case IPCTL_DEFMTU: - return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); -#endif - case IPCTL_FORWSRCRT: - /* Don't allow this to change in a secure environment. */ - if (securelevel > 0) - return (sysctl_rdint(oldp, oldlenp, newp, - ip_forwsrcrt)); - else - return (sysctl_int(oldp, oldlenp, newp, newlen, - &ip_forwsrcrt)); - case IPCTL_DIRECTEDBCAST: - return (sysctl_int(oldp, oldlenp, newp, newlen, - &ip_directedbcast)); - case IPCTL_ALLOWSRCRT: - return (sysctl_int(oldp, oldlenp, newp, newlen, - &ip_allowsrcrt)); - case IPCTL_SUBNETSARELOCAL: - return (sysctl_int(oldp, oldlenp, newp, newlen, - &subnetsarelocal)); - case IPCTL_MTUDISC: - error = sysctl_int(oldp, oldlenp, newp, newlen, - &ip_mtudisc); - if (error == 0 && ip_mtudisc == 0) - rt_timer_queue_remove_all(ip_mtudisc_timeout_q, TRUE); - return error; - case IPCTL_ANONPORTMIN: - old = anonportmin; - error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin); - if (anonportmin >= anonportmax || anonportmin < 0 - || anonportmin > 65535 -#ifndef IPNOPRIVPORTS - || anonportmin < IPPORT_RESERVED -#endif - ) { - anonportmin = old; - return (EINVAL); - } + node = *rnode; + tmp = ip_forwsrcrt; + node.sysctl_data = &tmp; + error = sysctl_lookup(SYSCTLFN_CALL(&node)); + if (error || newp == NULL) return (error); - case IPCTL_ANONPORTMAX: - old = anonportmax; - error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax); - if (anonportmin >= anonportmax || anonportmax < 0 - || anonportmax > 65535 -#ifndef IPNOPRIVPORTS - || anonportmax < IPPORT_RESERVED -#endif - ) { - anonportmax = old; - return (EINVAL); - } - return (error); - case IPCTL_MTUDISCTIMEOUT: - old = ip_mtudisc_timeout; - error = sysctl_int(oldp, oldlenp, newp, newlen, - &ip_mtudisc_timeout); - if (ip_mtudisc_timeout < 0) { - ip_mtudisc_timeout = old; - return (EINVAL); - } - if (error == 0) - rt_timer_queue_change(ip_mtudisc_timeout_q, - ip_mtudisc_timeout); + + 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. + */ +static int +sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS) +{ + int error, tmp; + struct sysctlnode node; + + node = *rnode; + tmp = ip_mtudisc_timeout; + node.sysctl_data = &tmp; + error = sysctl_lookup(SYSCTLFN_CALL(&node)); + if (error || newp == NULL) return (error); + if (tmp < 0) + return (EINVAL); + + ip_mtudisc_timeout = tmp; + rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout); + + return (0); +} + #ifdef GATEWAY - case IPCTL_MAXFLOWS: - { - int s; +/* + * sysctl helper routine for net.inet.ip.maxflows. + */ +static int +sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS) +{ + int s; - error = sysctl_int(oldp, oldlenp, newp, newlen, - &ip_maxflows); - s = splsoftnet(); - ipflow_reap(0); - splx(s); - return (error); - } -#endif - case IPCTL_HOSTZEROBROADCAST: - return (sysctl_int(oldp, oldlenp, newp, newlen, - &hostzeroisbroadcast)); -#if NGIF > 0 - case IPCTL_GIF_TTL: - return (sysctl_int(oldp, oldlenp, newp, newlen, - &ip_gif_ttl)); -#endif + s = sysctl_lookup(SYSCTLFN_CALL(rnode)); + if (s || newp == NULL) + return (s); -#if NGRE > 0 - case IPCTL_GRE_TTL: - return (sysctl_int(oldp, oldlenp, newp, newlen, - &ip_gre_ttl)); -#endif + s = splsoftnet(); + ipflow_reap(0); + splx(s); -#ifndef IPNOPRIVPORTS - case IPCTL_LOWPORTMIN: - old = lowportmin; - error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin); - if (lowportmin >= lowportmax - || lowportmin > IPPORT_RESERVEDMAX - || lowportmin < IPPORT_RESERVEDMIN - ) { - lowportmin = old; - return (EINVAL); - } - return (error); - case IPCTL_LOWPORTMAX: - old = lowportmax; - error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax); - if (lowportmin >= lowportmax - || lowportmax > IPPORT_RESERVEDMAX - || lowportmax < IPPORT_RESERVEDMIN - ) { - lowportmax = old; - return (EINVAL); - } + return (0); +} + +static int +sysctl_net_inet_ip_hashsize(SYSCTLFN_ARGS) +{ + int error, tmp; + struct sysctlnode node; + + node = *rnode; + tmp = ip_hashsize; + node.sysctl_data = &tmp; + error = sysctl_lookup(SYSCTLFN_CALL(&node)); + if (error || newp == NULL) return (error); -#endif - case IPCTL_MAXFRAGPACKETS: - return (sysctl_int(oldp, oldlenp, newp, newlen, - &ip_maxfragpackets)); - - case IPCTL_CHECKINTERFACE: - return (sysctl_int(oldp, oldlenp, newp, newlen, - &ip_checkinterface)); - case IPCTL_IFQ: - return (sysctl_ifq(name+1, namelen-1, - oldp, oldlenp, newp, newlen, - &ipintrq)); + if ((tmp & (tmp - 1)) == 0 && tmp != 0) { + /* + * Can only fail due to malloc() + */ + if (ipflow_invalidate_all(tmp)) + return ENOMEM; + } else { + /* + * EINVAL if not a power of 2 + */ + return EINVAL; + } - default: - return (EOPNOTSUPP); - } - /* NOTREACHED */ + return (0); +} +#endif /* GATEWAY */ + + +SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup") +{ + extern int subnetsarelocal, hostzeroisbroadcast; + + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT, + CTLTYPE_NODE, "net", NULL, + NULL, 0, NULL, 0, + CTL_NET, CTL_EOL); + 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(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(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(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(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(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(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(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(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(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(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(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(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(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(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); + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "hashsize", + SYSCTL_DESCR("Size of hash table for fast forwarding (IPv4)"), + sysctl_net_inet_ip_hashsize, 0, &ip_hashsize, 0, + CTL_NET, PF_INET, IPPROTO_IP, + CTL_CREATE, CTL_EOL); +#endif /* GATEWAY */ + 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(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(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(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(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(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(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(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); }