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 1.93.2.6 retrieving revision 1.151 diff -u -p -r1.93.2.6 -r1.151 --- src/sys/netinet/ip_input.c 2001/03/12 13:31:50 1.93.2.6 +++ src/sys/netinet/ip_input.c 2002/06/07 13:43:47 1.151 @@ -1,4 +1,4 @@ -/* $NetBSD: ip_input.c,v 1.93.2.6 2001/03/12 13:31:50 bouyer Exp $ */ +/* $NetBSD: ip_input.c,v 1.151 2002/06/07 13:43:47 itojun Exp $ */ /* * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. @@ -101,10 +101,14 @@ * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 */ +#include +__KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.151 2002/06/07 13:43:47 itojun Exp $"); + #include "opt_gateway.h" #include "opt_pfil_hooks.h" #include "opt_ipsec.h" #include "opt_mrouting.h" +#include "opt_inet_csum.h" #include #include @@ -117,11 +121,7 @@ #include #include #include -#include #include - -#include - #include #include @@ -139,6 +139,8 @@ /* just for gif_ttl */ #include #include "gif.h" +#include +#include "gre.h" #ifdef MROUTING #include @@ -199,6 +201,8 @@ struct rttimer_queue *ip_mtudisc_timeout 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_ifaddrhashhead *in_ifaddrhashtbl; struct ifqueue ipintrq; @@ -211,6 +215,8 @@ struct pfil_head inet_pfil_hook; struct ipqhead ipq; int ipq_locked; +int ip_nfragpackets = 0; +int ip_maxfragpackets = 200; static __inline int ipq_lock_try __P((void)); static __inline void ipq_unlock __P((void)); @@ -220,7 +226,11 @@ ipq_lock_try() { int s; - s = splimp(); + /* + * Use splvm() -- we're blocking things that would cause + * mbuf allocation. + */ + s = splvm(); if (ipq_locked) { splx(s); return (0); @@ -235,7 +245,7 @@ ipq_unlock() { int s; - s = splimp(); + s = splvm(); ipq_locked = 0; splx(s); } @@ -264,6 +274,24 @@ do { \ struct pool ipqent_pool; +#ifdef INET_CSUM_COUNTERS +#include + +struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, + NULL, "inet", "hwcsum bad"); +struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, + NULL, "inet", "hwcsum ok"); +struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, + NULL, "inet", "swcsum"); + +#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ + +#else + +#define INET_CSUM_COUNTER_INCR(ev) /* nothing */ + +#endif /* INET_CSUM_COUNTERS */ + /* * We need to save the IP options in case a protocol wants to respond * to an incoming packet over the same route if the packet got here @@ -292,7 +320,7 @@ ip_init() int i; pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", - 0, NULL, NULL, M_IPQ); + NULL); pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); if (pr == 0) @@ -326,6 +354,12 @@ ip_init() printf("ip_init: WARNING: unable to register pfil hook, " "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 */ } struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET }; @@ -341,7 +375,7 @@ ipintr() struct mbuf *m; while (1) { - s = splimp(); + s = splnet(); IF_DEQUEUE(&ipintrq, m); splx(s); if (m == 0) @@ -383,7 +417,7 @@ 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 (in_ifaddr.tqh_first == 0) + if (TAILQ_FIRST(&in_ifaddr) == 0) goto bad; ipstat.ips_total++; if (m->m_len < sizeof (struct ip) && @@ -427,9 +461,24 @@ ip_input(struct mbuf *m) } } - if (in_cksum(m, hlen) != 0) { - ipstat.ips_badsum++; - goto bad; + switch (m->m_pkthdr.csum_flags & + ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) | + M_CSUM_IPv4_BAD)) { + case M_CSUM_IPv4|M_CSUM_IPv4_BAD: + INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); + goto badcsum; + + case M_CSUM_IPv4: + /* Checksum was okay. */ + INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); + break; + + default: + /* Must compute it ourselves. */ + INET_CSUM_COUNTER_INCR(&ip_swcsum); + if (in_cksum(m, hlen) != 0) + goto bad; + break; } /* Retrieve the packet length. */ @@ -462,7 +511,7 @@ ip_input(struct mbuf *m) } #ifdef IPSEC - /* ipflow (IP fast fowarding) is not compatible with IPsec. */ + /* ipflow (IP fast forwarding) is not compatible with IPsec. */ m->m_flags &= ~M_CANFASTFWD; #else /* @@ -485,7 +534,7 @@ ip_input(struct mbuf *m) * not the decapsulated packet. */ #ifdef IPSEC - if (!ipsec_gethist(m, NULL)) + if (!ipsec_getnhist(m)) #else if (1) #endif @@ -496,6 +545,7 @@ ip_input(struct mbuf *m) if (m == NULL) return; ip = mtod(m, struct ip *); + hlen = ip->ip_hl << 2; } #endif /* PFIL_HOOKS */ @@ -531,9 +581,7 @@ ip_input(struct mbuf *m) * as not mine. */ downmatch = 0; - for (ia = IN_IFADDR_HASH(ip->ip_dst.s_addr).lh_first; - ia != NULL; - ia = ia->ia_hash.le_next) { + LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) { if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) { if ((ia->ia_ifp->if_flags & IFF_UP) != 0) break; @@ -544,9 +592,9 @@ ip_input(struct mbuf *m) if (ia != NULL) goto ours; if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { - for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first; - ifa != NULL; ifa = ifa->ifa_list.tqe_next) { - if (ifa->ifa_addr->sa_family != AF_INET) continue; + TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) { + if (ifa->ifa_addr->sa_family != AF_INET) + continue; ia = ifatoia(ifa); if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) || in_hosteq(ip->ip_dst, ia->ia_netbroadcast) || @@ -570,7 +618,7 @@ ip_input(struct mbuf *m) #ifdef MROUTING extern struct socket *ip_mrouter; - if (m->m_flags & M_EXT) { + if (M_READONLY(m)) { if ((m = m_pullup(m, hlen)) == 0) { ipstat.ips_toosmall++; return; @@ -641,6 +689,13 @@ ip_input(struct mbuf *m) ipstat.ips_cantforward++; return; } +#ifdef IPSEC + if (ipsec4_in_reject(m, NULL)) { + ipsecstat.in_polvio++; + goto bad; + } +#endif + ip_forward(m, 0); } return; @@ -659,7 +714,7 @@ ours: * of this datagram. */ IPQ_LOCK(); - for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next) + LIST_FOREACH(fp, &ipq, 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) && @@ -748,6 +803,11 @@ found: } bad: m_freem(m); + return; + +badcsum: + ipstat.ips_badsum++; + m_freem(m); } /* @@ -781,6 +841,17 @@ ip_reass(ipqe, fp) * If first fragment to arrive, create a reassembly queue. */ if (fp == 0) { + /* + * Enforce upper bound on number of fragmented packets + * for which we attempt reassembly; + * If maxfrag is 0, never accept fragments. + * If maxfrag is -1, accept all fragments without limitation. + */ + if (ip_maxfragpackets < 0) + ; + else if (ip_nfragpackets >= ip_maxfragpackets) + goto dropfrag; + ip_nfragpackets++; MALLOC(fp, struct ipq *, sizeof (struct ipq), M_FTABLE, M_NOWAIT); if (fp == NULL) @@ -789,7 +860,7 @@ ip_reass(ipqe, fp) fp->ipq_ttl = IPFRAGTTL; fp->ipq_p = ipqe->ipqe_ip->ip_p; fp->ipq_id = ipqe->ipqe_ip->ip_id; - LIST_INIT(&fp->ipq_fragq); + TAILQ_INIT(&fp->ipq_fragq); fp->ipq_src = ipqe->ipqe_ip->ip_src; fp->ipq_dst = ipqe->ipqe_ip->ip_dst; p = NULL; @@ -799,8 +870,8 @@ ip_reass(ipqe, fp) /* * Find a segment which begins after this one does. */ - for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; - p = q, q = q->ipqe_q.le_next) + for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; + p = q, q = TAILQ_NEXT(q, ipqe_q)) if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off) break; @@ -835,9 +906,9 @@ ip_reass(ipqe, fp) m_adj(q->ipqe_m, i); break; } - nq = q->ipqe_q.le_next; + nq = TAILQ_NEXT(q, ipqe_q); m_freem(q->ipqe_m); - LIST_REMOVE(q, ipqe_q); + TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); pool_put(&ipqent_pool, q); } @@ -847,13 +918,13 @@ insert: * check for complete reassembly. */ if (p == NULL) { - LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); + TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); } else { - LIST_INSERT_AFTER(p, ipqe, ipqe_q); + TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q); } next = 0; - for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL; - p = q, q = q->ipqe_q.le_next) { + for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; + p = q, q = TAILQ_NEXT(q, ipqe_q)) { if (q->ipqe_ip->ip_off != next) return (0); next += q->ipqe_ip->ip_len; @@ -865,7 +936,7 @@ insert: * Reassembly is complete. Check for a bogus message size and * concatenate fragments. */ - q = fp->ipq_fragq.lh_first; + q = TAILQ_FIRST(&fp->ipq_fragq); ip = q->ipqe_ip; if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { ipstat.ips_toolong++; @@ -876,11 +947,11 @@ insert: t = m->m_next; m->m_next = 0; m_cat(m, t); - nq = q->ipqe_q.le_next; + nq = TAILQ_NEXT(q, ipqe_q); pool_put(&ipqent_pool, q); for (q = nq; q != NULL; q = nq) { t = q->ipqe_m; - nq = q->ipqe_q.le_next; + nq = TAILQ_NEXT(q, ipqe_q); pool_put(&ipqent_pool, q); m_cat(m, t); } @@ -896,6 +967,7 @@ insert: ip->ip_dst = fp->ipq_dst; LIST_REMOVE(fp, ipq_q); FREE(fp, M_FTABLE); + ip_nfragpackets--; m->m_len += (ip->ip_hl << 2); m->m_data -= (ip->ip_hl << 2); /* some debugging cruft by sklower, below, will go away soon */ @@ -926,14 +998,15 @@ ip_freef(fp) IPQ_LOCK_CHECK(); - for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) { - p = q->ipqe_q.le_next; + for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) { + p = TAILQ_NEXT(q, ipqe_q); m_freem(q->ipqe_m); - LIST_REMOVE(q, ipqe_q); + TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q); pool_put(&ipqent_pool, q); } LIST_REMOVE(fp, ipq_q); FREE(fp, M_FTABLE); + ip_nfragpackets--; } /* @@ -948,13 +1021,24 @@ ip_slowtimo() int s = splsoftnet(); IPQ_LOCK(); - for (fp = ipq.lh_first; fp != NULL; fp = nfp) { - nfp = fp->ipq_q.le_next; + 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); } } + /* + * If we are over the maximum number of fragments + * (due to the limit being lowered), drain off + * enough to get down to the new limit. + */ + if (ip_maxfragpackets < 0) + ; + else { + while (ip_nfragpackets > ip_maxfragpackets && LIST_FIRST(&ipq)) + ip_freef(LIST_FIRST(&ipq)); + } IPQ_UNLOCK(); #ifdef GATEWAY ipflow_slowtimo(); @@ -976,9 +1060,9 @@ ip_drain() if (ipq_lock_try() == 0) return; - while (ipq.lh_first != NULL) { + while (LIST_FIRST(&ipq) != NULL) { ipstat.ips_fragdropped++; - ip_freef(ipq.lh_first); + ip_freef(LIST_FIRST(&ipq)); } IPQ_UNLOCK(); @@ -1365,7 +1449,7 @@ ip_stripoptions(m, mopt) ip->ip_hl = sizeof (struct ip) >> 2; } -int inetctlerrmap[PRC_NCMDS] = { +const int inetctlerrmap[PRC_NCMDS] = { 0, 0, 0, 0, 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, @@ -1404,6 +1488,11 @@ ip_forward(m, srcrt) struct ifnet dummyifp; #endif + /* + * Clear any in-bound checksum flags for this packet. + */ + m->m_pkthdr.csum_flags = 0; + dest = 0; #ifdef DIAGNOSTIC if (ipprintfs) @@ -1484,7 +1573,7 @@ ip_forward(m, srcrt) } #ifdef IPSEC - /* Don't lookup socket in forwading case */ + /* Don't lookup socket in forwarding case */ (void)ipsec_setsocket(m, NULL); #endif error = ip_output(m, (struct mbuf *)0, &ipforward_rt, @@ -1573,6 +1662,8 @@ ip_forward(m, srcrt) ro = &sp->req->sav->sah->sa_route; if (ro->ro_rt && ro->ro_rt->rt_ifp) { dummyifp.if_mtu = + 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; @@ -1587,9 +1678,21 @@ ip_forward(m, srcrt) break; case ENOBUFS: +#if 1 + /* + * a router should not generate ICMP_SOURCEQUENCH as + * required in RFC1812 Requirements for IP Version 4 Routers. + * source quench could be a big problem under DoS attacks, + * or if the underlying interface is rate-limited. + */ + if (mcopy) + m_freem(mcopy); + return; +#else type = ICMP_SOURCEQUENCH; code = 0; break; +#endif } icmp_error(mcopy, type, code, dest, destifp); } @@ -1766,6 +1869,12 @@ ip_sysctl(name, namelen, oldp, oldlenp, &ip_gif_ttl)); #endif +#if NGRE > 0 + case IPCTL_GRE_TTL: + return(sysctl_int(oldp, oldlenp, newp, newlen, + &ip_gre_ttl)); +#endif + #ifndef IPNOPRIVPORTS case IPCTL_LOWPORTMIN: old = lowportmin; @@ -1791,6 +1900,10 @@ ip_sysctl(name, namelen, oldp, oldlenp, return (error); #endif + case IPCTL_MAXFRAGPACKETS: + return (sysctl_int(oldp, oldlenp, newp, newlen, + &ip_maxfragpackets)); + default: return (EOPNOTSUPP); }