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 1.138.2.5 retrieving revision 1.213 diff -u -p -r1.138.2.5 -r1.213 --- src/sys/netinet/udp_usrreq.c 2007/12/07 17:34:33 1.138.2.5 +++ src/sys/netinet/udp_usrreq.c 2014/08/02 03:55:26 1.213 @@ -1,4 +1,4 @@ -/* $NetBSD: udp_usrreq.c,v 1.138.2.5 2007/12/07 17:34:33 yamt Exp $ */ +/* $NetBSD: udp_usrreq.c,v 1.213 2014/08/02 03:55:26 rtr Exp $ */ /* * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. @@ -60,23 +60,27 @@ * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95 */ +/* + * UDP protocol implementation. + * Per RFC 768, August, 1980. + */ + #include -__KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.138.2.5 2007/12/07 17:34:33 yamt Exp $"); +__KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.213 2014/08/02 03:55:26 rtr Exp $"); #include "opt_inet.h" +#include "opt_compat_netbsd.h" #include "opt_ipsec.h" #include "opt_inet_csum.h" #include "opt_ipkdb.h" #include "opt_mbuftrace.h" #include -#include #include +#include #include #include #include -#include -#include #include #include #include @@ -94,14 +98,16 @@ __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c #include #include #include +#include #ifdef INET6 #include #include #include +#include #include #include -#include +#include #endif #ifndef INET6 @@ -109,44 +115,33 @@ __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c #include #endif -#include "faith.h" -#if defined(NFAITH) && NFAITH > 0 -#include -#endif - -#include - -#ifdef FAST_IPSEC +#ifdef IPSEC #include -#include /* XXX ipsecstat namespace */ +#include +#include #include #ifdef INET6 #include #endif -#endif /* FAST_IPSEC*/ +#endif /* IPSEC */ -#ifdef IPSEC -#include -#include -#include -#endif /*IPSEC*/ +#ifdef COMPAT_50 +#include +#endif #ifdef IPKDB #include #endif -/* - * UDP protocol implementation. - * Per RFC 768, August, 1980. - */ 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 IPSEC static int udp4_espinudp (struct mbuf **, int, struct sockaddr *, struct socket *); #endif @@ -156,13 +151,6 @@ static int udp4_realinput (struct sockad struct mbuf **, int); static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int); #endif -#ifdef INET6 -static void udp6_sendup (struct mbuf *, int, struct sockaddr *, - struct socket *); -static int udp6_realinput (int, struct sockaddr_in6 *, - struct sockaddr_in6 *, struct mbuf *, int); -static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int); -#endif #ifdef INET static void udp_notify (struct inpcb *, int); #endif @@ -172,6 +160,12 @@ static void udp_notify (struct inpcb *, #endif int udbhashsize = UDBHASHSIZE; +/* + * For send - really max datagram size; for receive - 40 1K datagrams. + */ +static int udp_sendspace = 9216; +static int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in)); + #ifdef MBUFTRACE struct mowner udp_mowner = MOWNER_INIT("udp", ""); struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx"); @@ -197,39 +191,42 @@ EVCNT_ATTACH_STATIC(udp_hwcsum_data); EVCNT_ATTACH_STATIC(udp_swcsum); #endif /* defined(INET) */ -#if defined(INET6) -struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, - NULL, "udp6", "hwcsum bad"); -struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, - NULL, "udp6", "hwcsum ok"); -struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, - NULL, "udp6", "hwcsum data"); -struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, - NULL, "udp6", "swcsum"); - -EVCNT_ATTACH_STATIC(udp6_hwcsum_bad); -EVCNT_ATTACH_STATIC(udp6_hwcsum_ok); -EVCNT_ATTACH_STATIC(udp6_hwcsum_data); -EVCNT_ATTACH_STATIC(udp6_swcsum); -#endif /* defined(INET6) */ - #define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ - #else - #define UDP_CSUM_COUNTER_INCR(ev) /* nothing */ - #endif /* UDP_CSUM_COUNTERS */ -void -udp_init(void) +static void sysctl_net_inet_udp_setup(struct sysctllog **); + +static int +do_udpinit(void) { in_pcbinit(&udbtable, udbhashsize, udbhashsize); + udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS); MOWNER_ATTACH(&udp_tx_mowner); MOWNER_ATTACH(&udp_rx_mowner); MOWNER_ATTACH(&udp_mowner); + + return 0; +} + +void +udp_init_common(void) +{ + static ONCE_DECL(doudpinit); + + RUN_ONCE(&doudpinit, do_udpinit); +} + +void +udp_init(void) +{ + + sysctl_net_inet_udp_setup(NULL); + + udp_init_common(); } /* @@ -324,7 +321,7 @@ udp4_input_checksum(struct mbuf *m, cons return 0; badcsum: - udpstat.udps_badsum++; + UDP_STATINC(UDP_STAT_BADSUM); return -1; } @@ -346,7 +343,7 @@ 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 +351,7 @@ 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 +368,7 @@ 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 +385,33 @@ 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 +421,10 @@ 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), @@ -445,151 +450,6 @@ badcsum: } #endif -#ifdef INET6 -static int -udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len) -{ - - /* - * XXX it's better to record and check if this mbuf is - * already checked. - */ - - if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) { - goto good; - } - if (uh->uh_sum == 0) { - udp6stat.udp6s_nosum++; - goto bad; - } - - switch (m->m_pkthdr.csum_flags & - ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) | - 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++; - goto bad; - -#if 0 /* notyet */ - case M_CSUM_UDPv6|M_CSUM_DATA: -#endif - - case M_CSUM_UDPv6: - /* Checksum was okay. */ - UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok); - break; - - default: - /* - * Need to compute it ourselves. Maybe skip checksum - * on loopback interfaces. - */ - UDP_CSUM_COUNTER_INCR(&udp6_swcsum); - if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) { - udp6stat.udp6s_badsum++; - goto bad; - } - } - -good: - return 0; -bad: - return -1; -} - -int -udp6_input(struct mbuf **mp, int *offp, int proto) -{ - struct mbuf *m = *mp; - int off = *offp; - struct sockaddr_in6 src, dst; - struct ip6_hdr *ip6; - struct udphdr *uh; - u_int32_t plen, ulen; - - ip6 = mtod(m, struct ip6_hdr *); - -#if defined(NFAITH) && 0 < NFAITH - if (faithprefix(&ip6->ip6_dst)) { - /* send icmp6 host unreach? */ - m_freem(m); - return IPPROTO_DONE; - } -#endif - - udp6stat.udp6s_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++; - return IPPROTO_DONE; - } - KASSERT(UDP_HDR_ALIGNED_P(uh)); - ulen = ntohs((u_short)uh->uh_ulen); - /* - * RFC2675 section 4: jumbograms will have 0 in the UDP header field, - * iff payload length > 0xffff. - */ - if (ulen == 0 && plen > 0xffff) - ulen = plen; - - if (plen != ulen) { - udp6stat.udp6s_badlen++; - goto bad; - } - - /* destination port of 0 is illegal, based on RFC768. */ - if (uh->uh_dport == 0) - goto bad; - - /* Be proactive about malicious use of IPv4 mapped address */ - if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || - IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { - /* XXX stat */ - goto bad; - } - - /* - * Checksum extended UDP header and data. Maybe skip checksum - * on loopback interfaces. - */ - if (udp6_input_checksum(m, uh, off, ulen)) - goto bad; - - /* - * Construct source and dst sockaddrs. - */ - bzero(&src, 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)); - dst.sin6_family = AF_INET6; - dst.sin6_len = sizeof(struct sockaddr_in6); - 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++; - goto bad; - } - udp6stat.udp6s_noport++; - icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0); - m = NULL; - } - -bad: - if (m) - m_freem(m); - return IPPROTO_DONE; -} -#endif - #ifdef INET static void udp4_sendup(struct mbuf *m, int off /* offset of data portion */, @@ -613,10 +473,10 @@ udp4_sendup(struct mbuf *m, int off /* o return; } -#if defined(IPSEC) || defined(FAST_IPSEC) +#if defined(IPSEC) /* check AH/ESP integrity. */ - if (so != NULL && ipsec4_in_reject_so(m, so)) { - ipsecstat.in_polvio++; + if (ipsec_used && so != NULL && ipsec4_in_reject_so(m, so)) { + 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 +486,9 @@ 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); @@ -638,53 +501,7 @@ udp4_sendup(struct mbuf *m, int off /* o if (opts) m_freem(opts); so->so_rcv.sb_overflowed++; - udpstat.udps_fullsock++; - } else - sorwakeup(so); - } -} -#endif - -#ifdef INET6 -static void -udp6_sendup(struct mbuf *m, int off /* offset of data portion */, - struct sockaddr *src, struct socket *so) -{ - struct mbuf *opts = NULL; - struct mbuf *n; - struct in6pcb *in6p = NULL; - - if (!so) - return; - if (so->so_proto->pr_domain->dom_family != AF_INET6) - return; - in6p = sotoin6pcb(so); - -#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_copypacket(m, M_DONTWAIT)) != NULL) - icmp6_error(n, ICMP6_DST_UNREACH, - ICMP6_DST_UNREACH_ADMIN, 0); - return; - } -#endif /*IPSEC*/ - - 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 *); - ip6_savecontrol(in6p, &opts, ip6, n); - } - - m_adj(n, off); - if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) { - m_freem(n); - if (opts) - m_freem(opts); - so->so_rcv.sb_overflowed++; - udp6stat.udp6s_fullsock++; + UDP_STATINC(UDP_STAT_FULLSOCK); } else sorwakeup(so); } @@ -739,7 +556,7 @@ udp4_realinput(struct sockaddr_in *src, /* * Locate pcb(s) for datagram. */ - CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { + TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { inp = (struct inpcb *)inph; if (inp->inp_af != AF_INET) continue; @@ -776,15 +593,16 @@ 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 IPSEC /* Handle ESP over UDP */ if (inp->inp_flags & INP_ESPINUDP_ALL) { struct sockaddr *sa = (struct sockaddr *)src; @@ -812,127 +630,13 @@ udp4_realinput(struct sockaddr_in *src, } #endif - udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket); - rcvcnt++; - } - -bad: - return rcvcnt; -} -#endif - -#ifdef INET6 -static int -udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst, - struct mbuf *m, int off) -{ - u_int16_t sport, dport; - int rcvcnt; - struct in6_addr src6, *dst6; - const struct in_addr *dst4; - struct inpcb_hdr *inph; - struct in6pcb *in6p; - - rcvcnt = 0; - off += sizeof(struct udphdr); /* now, offset of payload */ - - if (af != AF_INET && af != AF_INET6) - goto bad; - if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6) - goto bad; - - src6 = src->sin6_addr; - if (sa6_recoverscope(src) != 0) { - /* XXX: should be impossible. */ - goto bad; - } - sport = src->sin6_port; - - dport = dst->sin6_port; - dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12]; - dst6 = &dst->sin6_addr; - - if (IN6_IS_ADDR_MULTICAST(dst6) || - (af == AF_INET && IN_MULTICAST(dst4->s_addr))) { - /* - * Deliver a multicast or broadcast datagram to *all* sockets - * for which the local and remote addresses and ports match - * those of the incoming datagram. This allows more than - * one process to receive multi/broadcasts on the same port. - * (This really ought to be done for unicast datagrams as - * well, but that would cause problems with existing - * applications that open both address-specific sockets and - * a wildcard socket listening to the same port -- they would - * end up receiving duplicates of every unicast datagram. - * Those applications open the multiple sockets to overcome an - * inadequacy of the UDP socket interface, but for backwards - * compatibility we avoid the problem here rather than - * fixing the interface. Maybe 4.5BSD will remedy this?) - */ - - /* - * KAME note: traditionally we dropped udpiphdr from mbuf here. - * we need udpiphdr for IPsec processing so we do that later. - */ - /* - * Locate pcb(s) for datagram. - */ - CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { - in6p = (struct in6pcb *)inph; - if (in6p->in6p_af != AF_INET6) - continue; - - if (in6p->in6p_lport != dport) - continue; - if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { - if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, - dst6)) - continue; - } else { - if (IN6_IS_ADDR_V4MAPPED(dst6) && - (in6p->in6p_flags & IN6P_IPV6_V6ONLY)) - continue; - } - if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { - if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, - &src6) || in6p->in6p_fport != sport) - continue; - } else { - if (IN6_IS_ADDR_V4MAPPED(&src6) && - (in6p->in6p_flags & IN6P_IPV6_V6ONLY)) - continue; - } - - udp6_sendup(m, off, (struct sockaddr *)src, - in6p->in6p_socket); - rcvcnt++; - - /* - * Don't look for additional matches if this one does - * not have either the SO_REUSEPORT or SO_REUSEADDR - * socket options set. This heuristic avoids searching - * through all pcbs in the common case of a non-shared - * port. It assumes that an application will never - * clear these options after setting them. - */ - if ((in6p->in6p_socket->so_options & - (SO_REUSEPORT|SO_REUSEADDR)) == 0) - break; - } - } else { /* - * Locate pcb for datagram. + * Check the minimum TTL for socket. */ - in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6, - dport, 0); - if (in6p == 0) { - ++udpstat.udps_pcbhashmiss; - in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0); - if (in6p == 0) - return rcvcnt; - } + if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl) + goto bad; - udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket); + udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket); rcvcnt++; } @@ -987,14 +691,13 @@ 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 +705,16 @@ 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 +727,15 @@ 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 +749,16 @@ 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 +838,7 @@ 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), @@ -1153,41 +849,257 @@ release: return (error); } -int udp_sendspace = 9216; /* really max datagram size */ -int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in)); - /* 40 1K datagrams */ +static int +udp_attach(struct socket *so, int proto) +{ + struct inpcb *inp; + int error; + + KASSERT(sotoinpcb(so) == NULL); -/*ARGSUSED*/ -int -udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam, - struct mbuf *control, struct lwp *l) + /* Assign the lock (must happen even if we will error out). */ + sosetlock(so); + +#ifdef MBUFTRACE + so->so_mowner = &udp_mowner; + so->so_rcv.sb_mowner = &udp_rx_mowner; + so->so_snd.sb_mowner = &udp_tx_mowner; +#endif + if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { + error = soreserve(so, udp_sendspace, udp_recvspace); + if (error) { + return error; + } + } + + error = in_pcballoc(so, &udbtable); + if (error) { + return error; + } + inp = sotoinpcb(so); + inp->inp_ip.ip_ttl = ip_defttl; + KASSERT(solocked(so)); + + return error; +} + +static void +udp_detach(struct socket *so) { struct inpcb *inp; + + KASSERT(solocked(so)); + inp = sotoinpcb(so); + KASSERT(inp != NULL); + in_pcbdetach(inp); +} + +static int +udp_accept(struct socket *so, struct mbuf *nam) +{ + KASSERT(solocked(so)); + + panic("udp_accept"); + + return EOPNOTSUPP; +} + +static int +udp_bind(struct socket *so, struct mbuf *nam) +{ + struct inpcb *inp = sotoinpcb(so); + int error = 0; int s; + + KASSERT(solocked(so)); + KASSERT(inp != NULL); + KASSERT(nam != NULL); + + s = splsoftnet(); + error = in_pcbbind(inp, nam); + splx(s); + + return error; +} + +static int +udp_listen(struct socket *so) +{ + KASSERT(solocked(so)); + + return EOPNOTSUPP; +} + +static int +udp_connect(struct socket *so, struct mbuf *nam) +{ + struct inpcb *inp = sotoinpcb(so); int error = 0; + int s; + + KASSERT(solocked(so)); + KASSERT(inp != NULL); + KASSERT(nam != NULL); + + s = splsoftnet(); + error = in_pcbconnect(inp, nam, curlwp); + if (! error) + soisconnected(so); + splx(s); + return error; +} + +static int +udp_disconnect(struct socket *so) +{ + struct inpcb *inp = sotoinpcb(so); + int s; + + KASSERT(solocked(so)); + KASSERT(inp != NULL); + + s = splsoftnet(); + /*soisdisconnected(so);*/ + so->so_state &= ~SS_ISCONNECTED; /* XXX */ + in_pcbdisconnect(inp); + inp->inp_laddr = zeroin_addr; /* XXX */ + in_pcbstate(inp, INP_BOUND); /* XXX */ + splx(s); + + return 0; +} + +static int +udp_shutdown(struct socket *so) +{ + int s; + + KASSERT(solocked(so)); + + s = splsoftnet(); + socantsendmore(so); + splx(s); + + return 0; +} + +static int +udp_abort(struct socket *so) +{ + KASSERT(solocked(so)); + + panic("udp_abort"); + + return EOPNOTSUPP; +} + +static int +udp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp) +{ + return in_control(so, cmd, nam, ifp); +} + +static int +udp_stat(struct socket *so, struct stat *ub) +{ + KASSERT(solocked(so)); + + /* stat: don't bother with a blocksize. */ + return 0; +} - if (req == PRU_CONTROL) - return (in_control(so, (long)m, (void *)nam, - (struct ifnet *)control, l)); +static int +udp_peeraddr(struct socket *so, struct mbuf *nam) +{ + int s; + + KASSERT(solocked(so)); + KASSERT(sotoinpcb(so) != NULL); + KASSERT(nam != NULL); + + s = splsoftnet(); + in_setpeeraddr(sotoinpcb(so), nam); + splx(s); + + return 0; +} + +static int +udp_sockaddr(struct socket *so, struct mbuf *nam) +{ + int s; + + KASSERT(solocked(so)); + KASSERT(sotoinpcb(so) != NULL); + KASSERT(nam != NULL); s = splsoftnet(); + in_setsockaddr(sotoinpcb(so), nam); + splx(s); + + return 0; +} + +static int +udp_recvoob(struct socket *so, struct mbuf *m, int flags) +{ + KASSERT(solocked(so)); + + return EOPNOTSUPP; +} + +static int +udp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control) +{ + KASSERT(solocked(so)); + + m_freem(m); + m_freem(control); + return EOPNOTSUPP; +} + +static int +udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam, + struct mbuf *control, struct lwp *l) +{ + struct inpcb *inp; + int s, error = 0; + + KASSERT(req != PRU_ATTACH); + KASSERT(req != PRU_DETACH); + KASSERT(req != PRU_ACCEPT); + KASSERT(req != PRU_BIND); + KASSERT(req != PRU_LISTEN); + KASSERT(req != PRU_CONNECT); + KASSERT(req != PRU_DISCONNECT); + KASSERT(req != PRU_SHUTDOWN); + KASSERT(req != PRU_ABORT); + KASSERT(req != PRU_CONTROL); + KASSERT(req != PRU_SENSE); + KASSERT(req != PRU_PEERADDR); + KASSERT(req != PRU_SOCKADDR); + KASSERT(req != PRU_RCVOOB); + KASSERT(req != PRU_SENDOOB); + + 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); + return 0; } + KASSERT(solocked(so)); inp = sotoinpcb(so); -#ifdef DIAGNOSTIC - if (req != PRU_SEND && req != PRU_SENDOOB && control) - panic("udp_usrreq: unexpected control mbuf"); -#endif - if (inp == 0 && req != PRU_ATTACH) { - error = EINVAL; - goto release; + + KASSERT(!control || req == PRU_SEND); + if (inp == NULL) { + splx(s); + return EINVAL; } /* @@ -1195,64 +1107,10 @@ udp_usrreq(struct socket *so, int req, s * the udp pcb queue and/or pcb addresses. */ switch (req) { - - case PRU_ATTACH: - if (inp != 0) { - error = EISCONN; - break; - } -#ifdef MBUFTRACE - so->so_mowner = &udp_mowner; - so->so_rcv.sb_mowner = &udp_rx_mowner; - so->so_snd.sb_mowner = &udp_tx_mowner; -#endif - if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { - error = soreserve(so, udp_sendspace, udp_recvspace); - if (error) - break; - } - error = in_pcballoc(so, &udbtable); - if (error) - break; - inp = sotoinpcb(so); - inp->inp_ip.ip_ttl = ip_defttl; - break; - - case PRU_DETACH: - in_pcbdetach(inp); - break; - - case PRU_BIND: - error = in_pcbbind(inp, nam, l); - break; - - case PRU_LISTEN: - error = EOPNOTSUPP; - break; - - case PRU_CONNECT: - error = in_pcbconnect(inp, nam, l); - if (error) - break; - soisconnected(so); - break; - case PRU_CONNECT2: error = EOPNOTSUPP; break; - case PRU_DISCONNECT: - /*soisdisconnected(so);*/ - so->so_state &= ~SS_ISCONNECTED; /* XXX */ - in_pcbdisconnect(inp); - inp->inp_laddr = zeroin_addr; /* XXX */ - in_pcbstate(inp, INP_BOUND); /* XXX */ - break; - - case PRU_SHUTDOWN: - socantsendmore(so); - break; - case PRU_RCVD: error = EOPNOTSUPP; break; @@ -1267,6 +1125,7 @@ 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) { @@ -1295,53 +1154,30 @@ udp_usrreq(struct socket *so, int req, s } break; - case PRU_SENSE: - /* - * stat: don't bother with a blocksize. - */ - splx(s); - return (0); - - case PRU_RCVOOB: - error = EOPNOTSUPP; - break; - - case PRU_SENDOOB: - m_freem(control); - m_freem(m); - error = EOPNOTSUPP; - break; - - case PRU_SOCKADDR: - in_setsockaddr(inp, nam); - break; - - case PRU_PEERADDR: - in_setpeeraddr(inp, nam); - break; - default: panic("udp_usrreq"); } - -release: splx(s); - return (error); + + 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, - NULL, 0, NULL, 0, - CTL_NET, CTL_EOL); - sysctl_createv(clog, 0, NULL, NULL, - CTLFLAG_PERMANENT, CTLTYPE_NODE, "inet", NULL, NULL, 0, NULL, 0, CTL_NET, PF_INET, CTL_EOL); @@ -1391,13 +1227,21 @@ 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(IPSEC) /* * Returns: * 1 if the packet was processed @@ -1415,7 +1259,6 @@ 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 +1286,8 @@ 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 +1347,9 @@ 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 +1358,60 @@ 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); +#ifdef IPSEC + if (ipsec_used) + ipsec4_common_input(m, iphdrlen, IPPROTO_ESP); + /* XXX: else */ #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 + +PR_WRAP_USRREQS(udp) +#define udp_attach udp_attach_wrapper +#define udp_detach udp_detach_wrapper +#define udp_accept udp_accept_wrapper +#define udp_bind udp_bind_wrapper +#define udp_listen udp_listen_wrapper +#define udp_connect udp_connect_wrapper +#define udp_disconnect udp_disconnect_wrapper +#define udp_shutdown udp_shutdown_wrapper +#define udp_abort udp_abort_wrapper +#define udp_ioctl udp_ioctl_wrapper +#define udp_stat udp_stat_wrapper +#define udp_peeraddr udp_peeraddr_wrapper +#define udp_sockaddr udp_sockaddr_wrapper +#define udp_recvoob udp_recvoob_wrapper +#define udp_sendoob udp_sendoob_wrapper +#define udp_usrreq udp_usrreq_wrapper + +const struct pr_usrreqs udp_usrreqs = { + .pr_attach = udp_attach, + .pr_detach = udp_detach, + .pr_accept = udp_accept, + .pr_bind = udp_bind, + .pr_listen = udp_listen, + .pr_connect = udp_connect, + .pr_disconnect = udp_disconnect, + .pr_shutdown = udp_shutdown, + .pr_abort = udp_abort, + .pr_ioctl = udp_ioctl, + .pr_stat = udp_stat, + .pr_peeraddr = udp_peeraddr, + .pr_sockaddr = udp_sockaddr, + .pr_recvoob = udp_recvoob, + .pr_sendoob = udp_sendoob, + .pr_generic = udp_usrreq, +};