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/udp_usrreq.c,v rcsdiff: /ftp/cvs/cvsroot/src/sys/netinet/udp_usrreq.c,v: warning: Unknown phrases like `commitid ...;' are present. retrieving revision 1.160.8.1 retrieving revision 1.189 diff -u -p -r1.160.8.1 -r1.189 --- src/sys/netinet/udp_usrreq.c 2007/11/06 23:33:54 1.160.8.1 +++ src/sys/netinet/udp_usrreq.c 2013/06/05 00:48:32 1.189 @@ -1,4 +1,4 @@ -/* $NetBSD: udp_usrreq.c,v 1.160.8.1 2007/11/06 23:33:54 matt Exp $ */ +/* $NetBSD: udp_usrreq.c,v 1.189 2013/06/05 00:48:32 christos Exp $ */ /* * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. @@ -61,9 +61,10 @@ */ #include -__KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.160.8.1 2007/11/06 23:33:54 matt Exp $"); +__KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.189 2013/06/05 00:48:32 christos Exp $"); #include "opt_inet.h" +#include "opt_compat_netbsd.h" #include "opt_ipsec.h" #include "opt_inet_csum.h" #include "opt_ipkdb.h" @@ -94,13 +95,16 @@ __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c #include #include #include +#include #ifdef INET6 #include #include #include +#include #include #include +#include #include #endif @@ -114,22 +118,19 @@ __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c #include #endif -#include - #ifdef FAST_IPSEC #include -#include /* XXX ipsecstat namespace */ +#include +#include #include #ifdef INET6 #include #endif -#endif /* FAST_IPSEC*/ +#endif /* FAST_IPSEC */ -#ifdef IPSEC -#include -#include -#include -#endif /*IPSEC*/ +#ifdef COMPAT_50 +#include +#endif #ifdef IPKDB #include @@ -143,10 +144,11 @@ int udpcksum = 1; int udp_do_loopback_cksum = 0; struct inpcbtable udbtable; -struct udpstat udpstat; + +percpu_t *udpstat_percpu; #ifdef INET -#ifdef IPSEC_NAT_T +#ifdef FAST_IPSEC static int udp4_espinudp (struct mbuf **, int, struct sockaddr *, struct socket *); #endif @@ -221,15 +223,26 @@ EVCNT_ATTACH_STATIC(udp6_swcsum); #endif /* UDP_CSUM_COUNTERS */ +static void sysctl_net_inet_udp_setup(struct sysctllog **); + void udp_init(void) { + sysctl_net_inet_udp_setup(NULL); + in_pcbinit(&udbtable, udbhashsize, udbhashsize); MOWNER_ATTACH(&udp_tx_mowner); MOWNER_ATTACH(&udp_rx_mowner); MOWNER_ATTACH(&udp_mowner); + +#ifdef INET + udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS); +#endif +#ifdef INET6 + udp6stat_percpu = percpu_alloc(sizeof(uint64_t) * UDP6_NSTATS); +#endif } /* @@ -324,7 +337,7 @@ udp4_input_checksum(struct mbuf *m, cons return 0; badcsum: - udpstat.udps_badsum++; + UDP_STATINC(UDP_STAT_BADSUM); return -1; } @@ -346,7 +359,7 @@ udp_input(struct mbuf *m, ...) va_end(ap); MCLAIM(m, &udp_rx_mowner); - udpstat.udps_ipackets++; + UDP_STATINC(UDP_STAT_IPACKETS); /* * Get IP and UDP header together in first mbuf. @@ -354,7 +367,7 @@ udp_input(struct mbuf *m, ...) ip = mtod(m, struct ip *); IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr)); if (uh == NULL) { - udpstat.udps_hdrops++; + UDP_STATINC(UDP_STAT_HDROPS); return; } KASSERT(UDP_HDR_ALIGNED_P(uh)); @@ -371,7 +384,7 @@ udp_input(struct mbuf *m, ...) len = ntohs((u_int16_t)uh->uh_ulen); if (ip_len != iphlen + len) { if (ip_len < iphlen + len || len < sizeof(struct udphdr)) { - udpstat.udps_badlen++; + UDP_STATINC(UDP_STAT_BADLEN); goto bad; } m_adj(m, iphlen + len - ip_len); @@ -388,25 +401,33 @@ udp_input(struct mbuf *m, ...) sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport); if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) { - udpstat.udps_hdrops++; + UDP_STATINC(UDP_STAT_HDROPS); + return; + } + if (m == NULL) { + /* + * packet has been processed by ESP stuff - + * e.g. dropped NAT-T-keep-alive-packet ... + */ return; } + ip = mtod(m, struct ip *); #ifdef INET6 if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) { struct sockaddr_in6 src6, dst6; - bzero(&src6, sizeof(src6)); + memset(&src6, 0, sizeof(src6)); src6.sin6_family = AF_INET6; src6.sin6_len = sizeof(struct sockaddr_in6); src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff; - bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12], + memcpy(&src6.sin6_addr.s6_addr[12], &ip->ip_src, sizeof(ip->ip_src)); src6.sin6_port = uh->uh_sport; - bzero(&dst6, sizeof(dst6)); + memset(&dst6, 0, sizeof(dst6)); dst6.sin6_family = AF_INET6; dst6.sin6_len = sizeof(struct sockaddr_in6); dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff; - bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12], + memcpy(&dst6.sin6_addr.s6_addr[12], &ip->ip_dst, sizeof(ip->ip_dst)); dst6.sin6_port = uh->uh_dport; @@ -416,10 +437,10 @@ udp_input(struct mbuf *m, ...) if (n == 0) { if (m->m_flags & (M_BCAST | M_MCAST)) { - udpstat.udps_noportbcast++; + UDP_STATINC(UDP_STAT_NOPORTBCAST); goto bad; } - udpstat.udps_noport++; + UDP_STATINC(UDP_STAT_NOPORT); #ifdef IPKDB if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport, m, iphlen + sizeof(struct udphdr), @@ -459,7 +480,7 @@ udp6_input_checksum(struct mbuf *m, cons goto good; } if (uh->uh_sum == 0) { - udp6stat.udp6s_nosum++; + UDP6_STATINC(UDP6_STAT_NOSUM); goto bad; } @@ -468,7 +489,7 @@ udp6_input_checksum(struct mbuf *m, cons M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD: UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad); - udp6stat.udp6s_badsum++; + UDP6_STATINC(UDP6_STAT_BADSUM); goto bad; #if 0 /* notyet */ @@ -487,7 +508,7 @@ udp6_input_checksum(struct mbuf *m, cons */ UDP_CSUM_COUNTER_INCR(&udp6_swcsum); if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) { - udp6stat.udp6s_badsum++; + UDP6_STATINC(UDP6_STAT_BADSUM); goto bad; } } @@ -518,13 +539,13 @@ udp6_input(struct mbuf **mp, int *offp, } #endif - udp6stat.udp6s_ipackets++; + UDP6_STATINC(UDP6_STAT_IPACKETS); /* check for jumbogram is done in ip6_input. we can trust pkthdr.len */ plen = m->m_pkthdr.len - off; IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr)); if (uh == NULL) { - ip6stat.ip6s_tooshort++; + IP6_STATINC(IP6_STAT_TOOSHORT); return IPPROTO_DONE; } KASSERT(UDP_HDR_ALIGNED_P(uh)); @@ -537,7 +558,7 @@ udp6_input(struct mbuf **mp, int *offp, ulen = plen; if (plen != ulen) { - udp6stat.udp6s_badlen++; + UDP6_STATINC(UDP6_STAT_BADLEN); goto bad; } @@ -562,12 +583,12 @@ udp6_input(struct mbuf **mp, int *offp, /* * Construct source and dst sockaddrs. */ - bzero(&src, sizeof(src)); + memset(&src, 0, sizeof(src)); src.sin6_family = AF_INET6; src.sin6_len = sizeof(struct sockaddr_in6); src.sin6_addr = ip6->ip6_src; src.sin6_port = uh->uh_sport; - bzero(&dst, sizeof(dst)); + memset(&dst, 0, sizeof(dst)); dst.sin6_family = AF_INET6; dst.sin6_len = sizeof(struct sockaddr_in6); dst.sin6_addr = ip6->ip6_dst; @@ -575,10 +596,10 @@ udp6_input(struct mbuf **mp, int *offp, if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) { if (m->m_flags & M_MCAST) { - udp6stat.udp6s_noportmcast++; + UDP6_STATINC(UDP6_STAT_NOPORTMCAST); goto bad; } - udp6stat.udp6s_noport++; + UDP6_STATINC(UDP6_STAT_NOPORT); icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0); m = NULL; } @@ -613,10 +634,10 @@ udp4_sendup(struct mbuf *m, int off /* o return; } -#if defined(IPSEC) || defined(FAST_IPSEC) +#if defined(FAST_IPSEC) /* check AH/ESP integrity. */ if (so != NULL && ipsec4_in_reject_so(m, so)) { - ipsecstat.in_polvio++; + IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT, 0, 0); @@ -626,6 +647,9 @@ udp4_sendup(struct mbuf *m, int off /* o if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) { if (inp && (inp->inp_flags & INP_CONTROLOPTS +#ifdef SO_OTIMESTAMP + || so->so_options & SO_OTIMESTAMP +#endif || so->so_options & SO_TIMESTAMP)) { struct ip *ip = mtod(n, struct ip *); ip_savecontrol(inp, &opts, ip, n); @@ -638,7 +662,7 @@ udp4_sendup(struct mbuf *m, int off /* o if (opts) m_freem(opts); so->so_rcv.sb_overflowed++; - udpstat.udps_fullsock++; + UDP_STATINC(UDP_STAT_FULLSOCK); } else sorwakeup(so); } @@ -660,10 +684,10 @@ udp6_sendup(struct mbuf *m, int off /* o return; in6p = sotoin6pcb(so); -#if defined(IPSEC) || defined(FAST_IPSEC) +#if defined(FAST_IPSEC) /* check AH/ESP integrity. */ if (so != NULL && ipsec6_in_reject_so(m, so)) { - ipsec6stat.in_polvio++; + IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO); if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) icmp6_error(n, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADMIN, 0); @@ -673,7 +697,10 @@ udp6_sendup(struct mbuf *m, int off /* o if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) { if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS - || in6p->in6p_socket->so_options & SO_TIMESTAMP)) { +#ifdef SO_OTIMESTAMP + || in6p->in6p_socket->so_options & SO_OTIMESTAMP +#endif + || in6p->in6p_socket->so_options & SO_TIMESTAMP)) { struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *); ip6_savecontrol(in6p, &opts, ip6, n); } @@ -684,7 +711,7 @@ udp6_sendup(struct mbuf *m, int off /* o if (opts) m_freem(opts); so->so_rcv.sb_overflowed++; - udp6stat.udp6s_fullsock++; + UDP6_STATINC(UDP6_STAT_FULLSOCK); } else sorwakeup(so); } @@ -776,15 +803,16 @@ udp4_realinput(struct sockaddr_in *src, /* * Locate pcb for datagram. */ - inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport); + inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, + *dport, 0); if (inp == 0) { - ++udpstat.udps_pcbhashmiss; + UDP_STATINC(UDP_STAT_PCBHASHMISS); inp = in_pcblookup_bind(&udbtable, *dst4, *dport); if (inp == 0) return rcvcnt; } -#ifdef IPSEC_NAT_T +#ifdef FAST_IPSEC /* Handle ESP over UDP */ if (inp->inp_flags & INP_ESPINUDP_ALL) { struct sockaddr *sa = (struct sockaddr *)src; @@ -812,6 +840,12 @@ udp4_realinput(struct sockaddr_in *src, } #endif + /* + * Check the minimum TTL for socket. + */ + if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl) + goto bad; + udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket); rcvcnt++; } @@ -924,9 +958,9 @@ udp6_realinput(int af, struct sockaddr_i * Locate pcb for datagram. */ in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6, - dport, 0); + dport, 0, 0); if (in6p == 0) { - ++udpstat.udps_pcbhashmiss; + UDP_STATINC(UDP_STAT_PCBHASHMISS); in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0); if (in6p == 0) return rcvcnt; @@ -987,14 +1021,13 @@ udp_ctlinput(int cmd, const struct socka } int -udp_ctloutput(int op, struct socket *so, int level, int optname, - struct mbuf **mp) +udp_ctloutput(int op, struct socket *so, struct sockopt *sopt) { int s; int error = 0; - struct mbuf *m; struct inpcb *inp; int family; + int optval; family = so->so_proto->pr_domain->dom_family; @@ -1002,16 +1035,16 @@ udp_ctloutput(int op, struct socket *so, switch (family) { #ifdef INET case PF_INET: - if (level != IPPROTO_UDP) { - error = ip_ctloutput(op, so, level, optname, mp); + if (sopt->sopt_level != IPPROTO_UDP) { + error = ip_ctloutput(op, so, sopt); goto end; } break; #endif #ifdef INET6 case PF_INET6: - if (level != IPPROTO_UDP) { - error = ip6_ctloutput(op, so, level, optname, mp); + if (sopt->sopt_level != IPPROTO_UDP) { + error = ip6_ctloutput(op, so, sopt); goto end; } break; @@ -1024,18 +1057,15 @@ udp_ctloutput(int op, struct socket *so, switch (op) { case PRCO_SETOPT: - m = *mp; inp = sotoinpcb(so); - switch (optname) { + switch (sopt->sopt_name) { case UDP_ENCAP: - if (m == NULL || m->m_len < sizeof (int)) { - error = EINVAL; + error = sockopt_getint(sopt, &optval); + if (error) break; - } - switch(*mtod(m, int *)) { -#ifdef IPSEC_NAT_T + switch(optval) { case 0: inp->inp_flags &= ~INP_ESPINUDP_ALL; break; @@ -1049,20 +1079,16 @@ udp_ctloutput(int op, struct socket *so, inp->inp_flags &= ~INP_ESPINUDP_ALL; inp->inp_flags |= INP_ESPINUDP_NON_IKE; break; -#endif default: error = EINVAL; break; } break; - + default: error = ENOPROTOOPT; break; } - if (m != NULL) { - m_free(m); - } break; default: @@ -1142,7 +1168,7 @@ udp_output(struct mbuf *m, ...) ((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len); ((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl; /* XXX */ ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */ - udpstat.udps_opackets++; + UDP_STATINC(UDP_STAT_OPACKETS); return (ip_output(m, inp->inp_options, ro, inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST), @@ -1173,9 +1199,11 @@ udp_usrreq(struct socket *so, int req, s s = splsoftnet(); if (req == PRU_PURGEIF) { + mutex_enter(softnet_lock); in_pcbpurgeif0(&udbtable, (struct ifnet *)control); in_purgeif((struct ifnet *)control); in_pcbpurgeif(&udbtable, (struct ifnet *)control); + mutex_exit(softnet_lock); splx(s); return (0); } @@ -1185,7 +1213,9 @@ udp_usrreq(struct socket *so, int req, s if (req != PRU_SEND && req != PRU_SENDOOB && control) panic("udp_usrreq: unexpected control mbuf"); #endif - if (inp == 0 && req != PRU_ATTACH) { + if (req == PRU_ATTACH) { + sosetlock(so); + } else if (inp == 0) { error = EINVAL; goto release; } @@ -1267,6 +1297,7 @@ udp_usrreq(struct socket *so, int req, s { struct in_addr laddr; /* XXX */ + memset(&laddr, 0, sizeof laddr); if (nam) { laddr = inp->inp_laddr; /* XXX */ if ((so->so_state & SS_ISCONNECTED) != 0) { @@ -1329,12 +1360,19 @@ release: return (error); } +static int +sysctl_net_inet_udp_stats(SYSCTLFN_ARGS) +{ + + return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS)); +} + /* * Sysctl for udp variables. */ -SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup") +static void +sysctl_net_inet_udp_setup(struct sysctllog **clog) { - sysctl_createv(clog, 0, NULL, NULL, CTLFLAG_PERMANENT, CTLTYPE_NODE, "net", NULL, @@ -1391,13 +1429,21 @@ SYSCTL_SETUP(sysctl_net_inet_udp_setup, CTLFLAG_PERMANENT, CTLTYPE_STRUCT, "stats", SYSCTL_DESCR("UDP statistics"), - NULL, 0, &udpstat, sizeof(udpstat), + sysctl_net_inet_udp_stats, 0, NULL, 0, CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS, CTL_EOL); } #endif -#if (defined INET && defined IPSEC_NAT_T) +void +udp_statinc(u_int stat) +{ + + KASSERT(stat < UDP_NSTATS); + UDP_STATINC(stat); +} + +#if defined(INET) && defined(FAST_IPSEC) /* * Returns: * 1 if the packet was processed @@ -1415,7 +1461,6 @@ udp4_espinudp(struct mbuf **mp, int off, size_t minlen; size_t iphdrlen; struct ip *ip; - struct mbuf *n; struct m_tag *tag; struct udphdr *udphdr; u_int16_t sport, dport; @@ -1443,6 +1488,8 @@ udp4_espinudp(struct mbuf **mp, int off, /* Ignore keepalive packets */ if ((len == 1) && (*(unsigned char *)data == 0xff)) { + m_free(m); + *mp = NULL; /* avoid any further processiong by caller ... */ return 1; } @@ -1502,16 +1549,9 @@ udp4_espinudp(struct mbuf **mp, int off, ip->ip_p = IPPROTO_ESP; /* - * Copy the mbuf to avoid multiple free, as both - * esp4_input (which we call) and udp_input (which - * called us) free the mbuf. - */ - if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { - printf("udp4_espinudp: m_dup failed\n"); - return 0; - } - - /* + * We have modified the packet - it is now ESP, so we should not + * return to UDP processing ... + * * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember * the source UDP port. This is required if we want * to select the right SPD for multiple hosts behind @@ -1520,20 +1560,21 @@ udp4_espinudp(struct mbuf **mp, int off, if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS, sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) { printf("udp4_espinudp: m_tag_get failed\n"); - m_freem(n); - return 0; + m_freem(m); + return -1; } ((u_int16_t *)(tag + 1))[0] = sport; ((u_int16_t *)(tag + 1))[1] = dport; - m_tag_prepend(n, tag); + m_tag_prepend(m, tag); #ifdef FAST_IPSEC - ipsec4_common_input(n, iphdrlen, IPPROTO_ESP); + ipsec4_common_input(m, iphdrlen, IPPROTO_ESP); #else - esp4_input(n, iphdrlen); + esp4_input(m, iphdrlen); #endif - /* We handled it, it shoudln't be handled by UDP */ + /* We handled it, it shouldn't be handled by UDP */ + *mp = NULL; /* avoid free by caller ... */ return 1; } #endif