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 1.197.2.1 retrieving revision 1.212.2.3 diff -u -p -r1.197.2.1 -r1.212.2.3 --- src/sys/netinet/ip_input.c 2004/05/28 07:25:05 1.197.2.1 +++ src/sys/netinet/ip_input.c 2007/09/17 20:14:21 1.212.2.3 @@ -1,4 +1,4 @@ -/* $NetBSD: ip_input.c,v 1.197.2.1 2004/05/28 07:25:05 tron Exp $ */ +/* $NetBSD: ip_input.c,v 1.212.2.3 2007/09/17 20:14:21 bouyer Exp $ */ /* * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. @@ -98,7 +98,7 @@ */ #include -__KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.197.2.1 2004/05/28 07:25:05 tron Exp $"); +__KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.212.2.3 2007/09/17 20:14:21 bouyer Exp $"); #include "opt_inet.h" #include "opt_gateway.h" @@ -219,11 +219,10 @@ 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 */ @@ -241,7 +240,7 @@ struct pfil_head inet_pfil_hook; * recalculate IP parameters derived from nmbclusters. */ static int ip_nmbclusters; /* copy of nmbclusters */ -static void ip_nmbclusters_changed __P((void)); /* recalc limits */ +static void ip_nmbclusters_changed(void); /* recalc limits */ #define CHECK_NMBCLUSTER_PARAMS() \ do { \ @@ -257,7 +256,7 @@ do { \ (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) struct ipqhead ipq[IPREASS_NHASH]; int ipq_locked; -static int ip_nfragpackets; /* packets in reass queue */ +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 */ @@ -267,7 +266,7 @@ int ip_maxfrags; /* limit on fr /* * 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 @@ -275,15 +274,15 @@ int ip_maxfrags; /* limit on fr * repeatedly deleting single packets under heavy fragmentation load * (e.g., from lossy NFS peers). */ -static u_int ip_reass_ttl_decr __P((u_int ticks)); -static void ip_reass_drophalf __P((void)); +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() +ipq_lock_try(void) { int s; @@ -302,7 +301,7 @@ ipq_lock_try() } static __inline void -ipq_unlock() +ipq_unlock(void) { int s; @@ -333,8 +332,8 @@ 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 @@ -348,6 +347,10 @@ 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 */ @@ -369,7 +372,7 @@ 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" }; @@ -391,16 +394,11 @@ ip_nmbclusters_changed(void) * 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"); @@ -440,12 +438,6 @@ 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); @@ -459,7 +451,7 @@ struct route ipforward_rt; * IP software interrupt routine */ void -ipintr() +ipintr(void) { int s; struct mbuf *m; @@ -580,10 +572,16 @@ 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; } @@ -657,6 +655,20 @@ 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 */ @@ -720,7 +732,7 @@ 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) { + IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { if (ifa->ifa_addr->sa_family != AF_INET) continue; ia = ifatoia(ifa); @@ -746,14 +758,6 @@ 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 @@ -831,7 +835,7 @@ 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); @@ -885,13 +889,6 @@ 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 @@ -988,7 +985,7 @@ 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) { @@ -996,7 +993,7 @@ 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) { /* @@ -1008,7 +1005,6 @@ found: /* XXX error stat??? */ error = EINVAL; DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ - goto bad; } splx(s); if (error) @@ -1046,10 +1042,7 @@ 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; @@ -1080,7 +1073,7 @@ ip_reass(ipqe, fp, ipqhead) * We are about to add a fragment; increment frag count. */ ip_nfrags++; - + /* * If first fragment to arrive, create a reassembly queue. */ @@ -1231,6 +1224,7 @@ 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); @@ -1249,8 +1243,7 @@ dropfrag: * associated datagrams. */ void -ip_freef(fp) - struct ipq *fp; +ip_freef(struct ipq *fp) { struct ipqent *q, *p; u_int nfrags = 0; @@ -1288,13 +1281,13 @@ static u_int fragttl_histo[(IPFRAGTTL+1) static u_int ip_reass_ttl_decr(u_int ticks) { - u_int i, nfrags, median; + u_int nfrags, median, dropfraction, keepfraction; struct ipq *fp, *nfp; - u_int dropfraction, keepfraction; - + 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) ? @@ -1347,7 +1340,7 @@ ip_reass_drophalf(void) * queue, discard it. */ void -ip_slowtimo() +ip_slowtimo(void) { static u_int dropscanidx = 0; u_int i; @@ -1404,7 +1397,7 @@ ip_slowtimo() * Drain off all datagram fragments. */ void -ip_drain() +ip_drain(void) { /* @@ -1431,8 +1424,7 @@ 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; @@ -1661,8 +1653,7 @@ 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; @@ -1689,9 +1680,7 @@ 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; @@ -1713,7 +1702,7 @@ 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; @@ -1785,9 +1774,7 @@ 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 *); @@ -1829,9 +1816,7 @@ 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; @@ -2059,11 +2044,8 @@ ip_forward(m, srcrt) } 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) { @@ -2156,7 +2138,7 @@ sysctl_net_inet_ip_maxflows(SYSCTLFN_ARG s = sysctl_lookup(SYSCTLFN_CALL(rnode)); if (s) return (s); - + s = splsoftnet(); ipflow_reap(0); splx(s); @@ -2187,7 +2169,7 @@ SYSCTL_SETUP(sysctl_net_inet_ip_setup, " 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", @@ -2351,4 +2333,11 @@ SYSCTL_SETUP(sysctl_net_inet_ip_setup, " 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); }