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 1.144 retrieving revision 1.169 diff -u -p -r1.144 -r1.169 --- src/sys/netinet/udp_usrreq.c 2005/12/09 15:36:34 1.144 +++ src/sys/netinet/udp_usrreq.c 2008/04/23 06:09:05 1.169 @@ -1,4 +1,4 @@ -/* $NetBSD: udp_usrreq.c,v 1.144 2005/12/09 15:36:34 manu Exp $ */ +/* $NetBSD: udp_usrreq.c,v 1.169 2008/04/23 06:09:05 thorpej Exp $ */ /* * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. @@ -61,7 +61,7 @@ */ #include -__KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.144 2005/12/09 15:36:34 manu Exp $"); +__KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.169 2008/04/23 06:09:05 thorpej Exp $"); #include "opt_inet.h" #include "opt_ipsec.h" @@ -94,18 +94,17 @@ __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c #include #include #include - -#ifdef IPSEC_NAT_T -#include -#include -#endif +#include #ifdef INET6 #include #include #include +#include #include #include +#include +#include #endif #ifndef INET6 @@ -122,16 +121,20 @@ __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c #ifdef FAST_IPSEC #include -#include /* XXX ipsecstat namespace */ +#include +#include +#include #ifdef INET6 #include #endif -#endif /* FAST_IPSEC*/ +#endif /* FAST_IPSEC */ #ifdef IPSEC #include +#include +#include #include -#endif /*IPSEC*/ +#endif /* IPSEC */ #ifdef IPKDB #include @@ -141,15 +144,12 @@ __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c * UDP protocol implementation. * Per RFC 768, August, 1980. */ -#ifndef COMPAT_42 int udpcksum = 1; -#else -int udpcksum = 0; /* XXX */ -#endif int udp_do_loopback_cksum = 0; struct inpcbtable udbtable; -struct udpstat udpstat; + +percpu_t *udpstat_percpu; #ifdef INET #ifdef IPSEC_NAT_T @@ -179,9 +179,9 @@ static void udp_notify (struct inpcb *, int udbhashsize = UDBHASHSIZE; #ifdef MBUFTRACE -struct mowner udp_mowner = { "udp" }; -struct mowner udp_rx_mowner = { "udp", "rx" }; -struct mowner udp_tx_mowner = { "udp", "tx" }; +struct mowner udp_mowner = MOWNER_INIT("udp", ""); +struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx"); +struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx"); #endif #ifdef UDP_CSUM_COUNTERS @@ -236,6 +236,8 @@ udp_init(void) MOWNER_ATTACH(&udp_tx_mowner); MOWNER_ATTACH(&udp_rx_mowner); MOWNER_ATTACH(&udp_mowner); + + udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS); } /* @@ -330,7 +332,7 @@ udp4_input_checksum(struct mbuf *m, cons return 0; badcsum: - udpstat.udps_badsum++; + UDP_STATINC(UDP_STAT_BADSUM); return -1; } @@ -352,7 +354,7 @@ 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. @@ -360,7 +362,7 @@ 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)); @@ -377,7 +379,7 @@ 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); @@ -390,19 +392,11 @@ udp_input(struct mbuf *m, ...) goto badcsum; /* construct source and dst sockaddrs. */ - bzero(&src, sizeof(src)); - src.sin_family = AF_INET; - src.sin_len = sizeof(struct sockaddr_in); - bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr)); - src.sin_port = uh->uh_sport; - bzero(&dst, sizeof(dst)); - dst.sin_family = AF_INET; - dst.sin_len = sizeof(struct sockaddr_in); - bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr)); - dst.sin_port = uh->uh_dport; + sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport); + 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; } #ifdef INET6 @@ -430,10 +424,10 @@ 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), @@ -473,7 +467,7 @@ udp6_input_checksum(struct mbuf *m, cons goto good; } if (uh->uh_sum == 0) { - udp6stat.udp6s_nosum++; + UDP6_STATINC(UDP6_STAT_NOSUM); goto bad; } @@ -482,7 +476,7 @@ 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 */ @@ -501,7 +495,7 @@ 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; } } @@ -532,13 +526,13 @@ 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)); @@ -551,7 +545,7 @@ udp6_input(struct mbuf **mp, int *offp, ulen = plen; if (plen != ulen) { - udp6stat.udp6s_badlen++; + UDP6_STATINC(UDP6_STAT_BADLEN); goto bad; } @@ -575,27 +569,24 @@ udp6_input(struct mbuf **mp, int *offp, /* * Construct source and dst sockaddrs. - * Note that ifindex (s6_addr16[1]) is already filled. */ bzero(&src, sizeof(src)); src.sin6_family = AF_INET6; src.sin6_len = sizeof(struct sockaddr_in6); - /* KAME hack: recover scopeid */ - (void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif); + src.sin6_addr = ip6->ip6_src; src.sin6_port = uh->uh_sport; bzero(&dst, sizeof(dst)); dst.sin6_family = AF_INET6; dst.sin6_len = sizeof(struct sockaddr_in6); - /* KAME hack: recover scopeid */ - (void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif); + dst.sin6_addr = ip6->ip6_dst; dst.sin6_port = uh->uh_dport; 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; } @@ -633,15 +624,15 @@ udp4_sendup(struct mbuf *m, int off /* o #if defined(IPSEC) || defined(FAST_IPSEC) /* check AH/ESP integrity. */ if (so != NULL && ipsec4_in_reject_so(m, so)) { - ipsecstat.in_polvio++; - if ((n = m_copy(m, 0, M_COPYALL)) != NULL) + 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); return; } #endif /*IPSEC*/ - if ((n = m_copy(m, 0, M_COPYALL)) != NULL) { + if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) { if (inp && (inp->inp_flags & INP_CONTROLOPTS || so->so_options & SO_TIMESTAMP)) { struct ip *ip = mtod(n, struct ip *); @@ -655,7 +646,7 @@ 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); } @@ -680,15 +671,15 @@ udp6_sendup(struct mbuf *m, int off /* o #if defined(IPSEC) || defined(FAST_IPSEC) /* check AH/ESP integrity. */ if (so != NULL && ipsec6_in_reject_so(m, so)) { - ipsec6stat.in_polvio++; - if ((n = m_copy(m, 0, M_COPYALL)) != NULL) + 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); return; } #endif /*IPSEC*/ - if ((n = m_copy(m, 0, M_COPYALL)) != NULL) { + if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) { if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS || in6p->in6p_socket->so_options & SO_TIMESTAMP)) { struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *); @@ -701,7 +692,7 @@ 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); } @@ -795,7 +786,7 @@ udp4_realinput(struct sockaddr_in *src, */ inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport); if (inp == 0) { - ++udpstat.udps_pcbhashmiss; + UDP_STATINC(UDP_STAT_PCBHASHMISS); inp = in_pcblookup_bind(&udbtable, *dst4, *dport); if (inp == 0) return rcvcnt; @@ -845,7 +836,7 @@ udp6_realinput(int af, struct sockaddr_i { u_int16_t sport, dport; int rcvcnt; - struct in6_addr src6, dst6; + struct in6_addr src6, *dst6; const struct in_addr *dst4; struct inpcb_hdr *inph; struct in6pcb *in6p; @@ -858,13 +849,18 @@ udp6_realinput(int af, struct sockaddr_i if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6) goto bad; - in6_embedscope(&src6, src, NULL, NULL); + src6 = src->sin6_addr; + if (sa6_recoverscope(src) != 0) { + /* XXX: should be impossible. */ + goto bad; + } sport = src->sin6_port; - in6_embedscope(&dst6, dst, NULL, NULL); + dport = dst->sin6_port; dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12]; + dst6 = &dst->sin6_addr; - if (IN6_IS_ADDR_MULTICAST(&dst6) || + if (IN6_IS_ADDR_MULTICAST(dst6) || (af == AF_INET && IN_MULTICAST(dst4->s_addr))) { /* * Deliver a multicast or broadcast datagram to *all* sockets @@ -897,10 +893,11 @@ udp6_realinput(int af, struct sockaddr_i if (in6p->in6p_lport != dport) continue; if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { - if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6)) + if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, + dst6)) continue; } else { - if (IN6_IS_ADDR_V4MAPPED(&dst6) && + if (IN6_IS_ADDR_V4MAPPED(dst6) && (in6p->in6p_flags & IN6P_IPV6_V6ONLY)) continue; } @@ -934,11 +931,11 @@ udp6_realinput(int af, struct sockaddr_i /* * Locate pcb for datagram. */ - in6p = in6_pcblookup_connect(&udbtable, &src6, sport, - &dst6, dport, 0); + in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6, + dport, 0); if (in6p == 0) { - ++udpstat.udps_pcbhashmiss; - in6p = in6_pcblookup_bind(&udbtable, &dst6, dport, 0); + UDP_STATINC(UDP_STAT_PCBHASHMISS); + in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0); if (in6p == 0) return rcvcnt; } @@ -966,7 +963,7 @@ udp_notify(struct inpcb *inp, int errno) } void * -udp_ctlinput(int cmd, struct sockaddr *sa, void *v) +udp_ctlinput(int cmd, const struct sockaddr *sa, void *v) { struct ip *ip = v; struct udphdr *uh; @@ -986,23 +983,20 @@ udp_ctlinput(int cmd, struct sockaddr *s else if (errno == 0) return NULL; if (ip) { - uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2)); - in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport, + uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2)); + in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport, ip->ip_src, uh->uh_sport, errno, notify); /* XXX mapped address case */ } else - in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno, + in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno, notify); return NULL; } int -udp_ctloutput(op, so, level, optname, mp) - int op; - struct socket *so; - int level, optname; - struct mbuf **mp; +udp_ctloutput(int op, struct socket *so, int level, int optname, + struct mbuf **mp) { int s; int error = 0; @@ -1033,7 +1027,6 @@ udp_ctloutput(op, so, level, optname, mp default: error = EAFNOSUPPORT; goto end; - break; } @@ -1044,9 +1037,9 @@ udp_ctloutput(op, so, level, optname, mp switch (optname) { case UDP_ENCAP: - if (m == NULL || m->m_len < sizeof (int)) { + if (m == NULL || m->m_len != sizeof(int)) { error = EINVAL; - goto end; + break; } switch(*mtod(m, int *)) { @@ -1067,21 +1060,21 @@ udp_ctloutput(op, so, level, optname, mp #endif default: error = EINVAL; - goto end; break; } break; default: error = ENOPROTOOPT; - goto end; break; } + if (m != NULL) { + m_free(m); + } break; default: error = EINVAL; - goto end; break; } @@ -1157,7 +1150,7 @@ 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), @@ -1175,24 +1168,26 @@ int udp_recvspace = 40 * (1024 + sizeof( /*ARGSUSED*/ int udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam, - struct mbuf *control, struct proc *p) + struct mbuf *control, struct lwp *l) { struct inpcb *inp; int s; int error = 0; if (req == PRU_CONTROL) - return (in_control(so, (long)m, (caddr_t)nam, - (struct ifnet *)control, p)); + return (in_control(so, (long)m, (void *)nam, + (struct ifnet *)control, l)); + + s = splsoftnet(); if (req == PRU_PURGEIF) { in_pcbpurgeif0(&udbtable, (struct ifnet *)control); in_purgeif((struct ifnet *)control); in_pcbpurgeif(&udbtable, (struct ifnet *)control); + splx(s); return (0); } - s = splsoftnet(); inp = sotoinpcb(so); #ifdef DIAGNOSTIC if (req != PRU_SEND && req != PRU_SENDOOB && control) @@ -1236,7 +1231,7 @@ udp_usrreq(struct socket *so, int req, s break; case PRU_BIND: - error = in_pcbbind(inp, nam, p); + error = in_pcbbind(inp, nam, l); break; case PRU_LISTEN: @@ -1244,7 +1239,7 @@ udp_usrreq(struct socket *so, int req, s break; case PRU_CONNECT: - error = in_pcbconnect(inp, nam, p); + error = in_pcbconnect(inp, nam, l); if (error) break; soisconnected(so); @@ -1286,7 +1281,7 @@ udp_usrreq(struct socket *so, int req, s error = EISCONN; goto die; } - error = in_pcbconnect(inp, nam, p); + error = in_pcbconnect(inp, nam, l); if (error) goto die; } else { @@ -1342,6 +1337,18 @@ release: return (error); } +static int +sysctl_net_inet_udp_stats(SYSCTLFN_ARGS) +{ + netstat_sysctl_context ctx; + uint64_t udps[UDP_NSTATS]; + + ctx.ctx_stat = udpstat_percpu; + ctx.ctx_counters = udps; + ctx.ctx_ncounters = UDP_NSTATS; + return (NETSTAT_SYSCTL(&ctx)); +} + /* * Sysctl for udp variables. */ @@ -1404,12 +1411,20 @@ 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 +void +udp_statinc(u_int stat) +{ + + KASSERT(stat < UDP_NSTATS); + UDP_STATINC(stat); +} + #if (defined INET && defined IPSEC_NAT_T) /* * Returns: @@ -1418,14 +1433,11 @@ SYSCTL_SETUP(sysctl_net_inet_udp_setup, * -1 if an error occurent and m was freed */ static int -udp4_espinudp(mp, off, src, so) - struct mbuf **mp; - int off; - struct sockaddr *src; - struct socket *so; +udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src, + struct socket *so) { size_t len; - caddr_t data; + void *data; struct inpcb *inp; size_t skip = 0; size_t minlen; @@ -1454,11 +1466,11 @@ udp4_espinudp(mp, off, src, so) } len = m->m_len - off; - data = mtod(m, caddr_t) + off; + data = mtod(m, char *) + off; inp = sotoinpcb(so); /* Ignore keepalive packets */ - if ((len == 1) && (data[0] == '\xff')) { + if ((len == 1) && (*(unsigned char *)data == 0xff)) { return 1; } @@ -1490,7 +1502,7 @@ udp4_espinudp(mp, off, src, so) * Get the UDP ports. They are handled in network * order everywhere in IPSEC_NAT_T code. */ - udphdr = (struct udphdr *)(data - skip); + udphdr = (struct udphdr *)((char *)data - skip); sport = udphdr->uh_sport; dport = udphdr->uh_dport; @@ -1510,7 +1522,7 @@ udp4_espinudp(mp, off, src, so) * <-skip-> */ iphdrlen = off - sizeof(struct udphdr); - memmove(mtod(m, caddr_t) + skip, mtod(m, caddr_t), iphdrlen); + memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen); m_adj(m, skip); ip = mtod(m, struct ip *); @@ -1543,7 +1555,11 @@ udp4_espinudp(mp, off, src, so) ((u_int16_t *)(tag + 1))[1] = dport; m_tag_prepend(n, tag); +#ifdef FAST_IPSEC + ipsec4_common_input(n, iphdrlen, IPPROTO_ESP); +#else esp4_input(n, iphdrlen); +#endif /* We handled it, it shoudln't be handled by UDP */ return 1;