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 1.187.2.3 retrieving revision 1.189 diff -u -p -r1.187.2.3 -r1.189 --- src/sys/netinet/udp_usrreq.c 2017/12/03 11:39:04 1.187.2.3 +++ src/sys/netinet/udp_usrreq.c 2013/06/05 00:48:32 1.189 @@ -1,4 +1,4 @@ -/* $NetBSD: udp_usrreq.c,v 1.187.2.3 2017/12/03 11:39:04 jdolecek 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. @@ -60,35 +60,31 @@ * @(#)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 1.187.2.3 2017/12/03 11:39:04 jdolecek Exp $"); +__KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.189 2013/06/05 00:48:32 christos Exp $"); -#ifdef _KERNEL_OPT #include "opt_inet.h" +#include "opt_compat_netbsd.h" #include "opt_ipsec.h" #include "opt_inet_csum.h" #include "opt_ipkdb.h" #include "opt_mbuftrace.h" -#include "opt_net_mpsafe.h" -#endif #include +#include #include -#include #include #include #include +#include +#include #include #include #include #include #include +#include #include #include @@ -109,6 +105,7 @@ __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c #include #include #include +#include #endif #ifndef INET6 @@ -116,7 +113,12 @@ __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c #include #endif -#ifdef IPSEC +#include "faith.h" +#if defined(NFAITH) && NFAITH > 0 +#include +#endif + +#ifdef FAST_IPSEC #include #include #include @@ -124,12 +126,20 @@ __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c #ifdef INET6 #include #endif -#endif /* IPSEC */ +#endif /* FAST_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; @@ -138,7 +148,7 @@ struct inpcbtable udbtable; percpu_t *udpstat_percpu; #ifdef INET -#ifdef IPSEC +#ifdef FAST_IPSEC static int udp4_espinudp (struct mbuf **, int, struct sockaddr *, struct socket *); #endif @@ -148,6 +158,13 @@ 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 @@ -157,12 +174,6 @@ 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"); @@ -188,42 +199,50 @@ 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 */ static void sysctl_net_inet_udp_setup(struct sysctllog **); -static int -do_udpinit(void) +void +udp_init(void) { + sysctl_net_inet_udp_setup(NULL); + 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(); +#ifdef INET + udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS); +#endif +#ifdef INET6 + udp6stat_percpu = percpu_alloc(sizeof(uint64_t) * UDP6_NSTATS); +#endif } /* @@ -272,7 +291,7 @@ udp4_input_checksum(struct mbuf *m, cons return 0; switch (m->m_pkthdr.csum_flags & - ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_UDPv4) | + ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) | M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD: UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad); @@ -305,7 +324,7 @@ udp4_input_checksum(struct mbuf *m, cons * Need to compute it ourselves. Maybe skip checksum * on loopback interfaces. */ - if (__predict_true(!(m_get_rcvif_NOMPSAFE(m)->if_flags & + if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) || udp_do_loopback_cksum)) { UDP_CSUM_COUNTER_INCR(&udp_swcsum); @@ -351,19 +370,6 @@ udp_input(struct mbuf *m, ...) UDP_STATINC(UDP_STAT_HDROPS); return; } - /* - * Enforce alignment requirements that are violated in - * some cases, see kern/50766 for details. - */ - if (UDP_HDR_ALIGNED_P(uh) == 0) { - m = m_copyup(m, iphlen + sizeof(struct udphdr), 0); - if (m == NULL) { - UDP_STATINC(UDP_STAT_HDROPS); - return; - } - ip = mtod(m, struct ip *); - uh = (struct udphdr *)(mtod(m, char *) + iphlen); - } KASSERT(UDP_HDR_ALIGNED_P(uh)); /* destination port of 0 is illegal, based on RFC768. */ @@ -413,12 +419,16 @@ udp_input(struct mbuf *m, ...) memset(&src6, 0, sizeof(src6)); src6.sin6_family = AF_INET6; src6.sin6_len = sizeof(struct sockaddr_in6); - in6_in_2_v4mapin6(&ip->ip_src, &src6.sin6_addr); + src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff; + memcpy(&src6.sin6_addr.s6_addr[12], &ip->ip_src, + sizeof(ip->ip_src)); src6.sin6_port = uh->uh_sport; memset(&dst6, 0, sizeof(dst6)); dst6.sin6_family = AF_INET6; dst6.sin6_len = sizeof(struct sockaddr_in6); - in6_in_2_v4mapin6(&ip->ip_dst, &dst6.sin6_addr); + dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff; + memcpy(&dst6.sin6_addr.s6_addr[12], &ip->ip_dst, + sizeof(ip->ip_dst)); dst6.sin6_port = uh->uh_dport; n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen); @@ -456,6 +466,151 @@ 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) { + UDP6_STATINC(UDP6_STAT_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); + UDP6_STATINC(UDP6_STAT_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) { + UDP6_STATINC(UDP6_STAT_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 + + 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) { + IP6_STATINC(IP6_STAT_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) { + UDP6_STATINC(UDP6_STAT_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. + */ + 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; + memset(&dst, 0, 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) { + UDP6_STATINC(UDP6_STAT_NOPORTMCAST); + goto bad; + } + UDP6_STATINC(UDP6_STAT_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 */, @@ -463,16 +618,25 @@ udp4_sendup(struct mbuf *m, int off /* o { struct mbuf *opts = NULL; struct mbuf *n; - struct inpcb *inp; + struct inpcb *inp = NULL; - KASSERT(so != NULL); - KASSERT(so->so_proto->pr_domain->dom_family == AF_INET); - inp = sotoinpcb(so); - KASSERT(inp != NULL); + if (!so) + return; + switch (so->so_proto->pr_domain->dom_family) { + case AF_INET: + inp = sotoinpcb(so); + break; +#ifdef INET6 + case AF_INET6: + break; +#endif + default: + return; + } -#if defined(IPSEC) +#if defined(FAST_IPSEC) /* check AH/ESP integrity. */ - if (ipsec_used && ipsec4_in_reject(m, inp)) { + if (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, @@ -482,8 +646,11 @@ udp4_sendup(struct mbuf *m, int off /* o #endif /*IPSEC*/ if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) { - if (inp->inp_flags & INP_CONTROLOPTS - || SOOPT_TIMESTAMP(so->so_options)) { + 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); } @@ -502,6 +669,55 @@ udp4_sendup(struct mbuf *m, int off /* o } #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(FAST_IPSEC) + /* check AH/ESP integrity. */ + if (so != NULL && ipsec6_in_reject_so(m, so)) { + 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_copypacket(m, M_DONTWAIT)) != NULL) { + if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS +#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); + } + + 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++; + UDP6_STATINC(UDP6_STAT_FULLSOCK); + } else + sorwakeup(so); + } +} +#endif + #ifdef INET static int udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst, @@ -526,7 +742,7 @@ udp4_realinput(struct sockaddr_in *src, dport = &dst->sin_port; if (IN_MULTICAST(dst4->s_addr) || - in_broadcast(*dst4, m_get_rcvif_NOMPSAFE(m))) { + in_broadcast(*dst4, m->m_pkthdr.rcvif)) { /* * Deliver a multicast or broadcast datagram to *all* sockets * for which the local and remote addresses and ports match @@ -550,7 +766,7 @@ udp4_realinput(struct sockaddr_in *src, /* * Locate pcb(s) for datagram. */ - TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { + CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { inp = (struct inpcb *)inph; if (inp->inp_af != AF_INET) continue; @@ -596,7 +812,7 @@ udp4_realinput(struct sockaddr_in *src, return rcvcnt; } -#ifdef IPSEC +#ifdef FAST_IPSEC /* Handle ESP over UDP */ if (inp->inp_flags & INP_ESPINUDP_ALL) { struct sockaddr *sa = (struct sockaddr *)src; @@ -639,6 +855,126 @@ bad: } #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. + */ + in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6, + dport, 0, 0); + if (in6p == 0) { + UDP_STATINC(UDP_STAT_PCBHASHMISS); + in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0); + if (in6p == 0) + return rcvcnt; + } + + udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket); + rcvcnt++; + } + +bad: + return rcvcnt; +} +#endif + #ifdef INET /* * Notify a udp user of an asynchronous error; @@ -765,18 +1101,21 @@ end: return error; } + int -udp_output(struct mbuf *m, struct inpcb *inp, struct mbuf *control, - struct lwp *l) +udp_output(struct mbuf *m, ...) { + struct inpcb *inp; struct udpiphdr *ui; struct route *ro; - struct ip_pktopts pktopts; - kauth_cred_t cred; int len = m->m_pkthdr.len; - int error, flags = 0; + int error = 0; + va_list ap; MCLAIM(m, &udp_tx_mowner); + va_start(ap, m); + inp = va_arg(ap, struct inpcb *); + va_end(ap); /* * Calculate data length and get a mbuf @@ -797,30 +1136,13 @@ udp_output(struct mbuf *m, struct inpcb goto release; } - if (l == NULL) - cred = NULL; - else - cred = l->l_cred; - - /* Setup IP outgoing packet options */ - memset(&pktopts, 0, sizeof(pktopts)); - error = ip_setpktopts(control, &pktopts, &flags, inp, cred, - IPPROTO_UDP); - if (error != 0) - goto release; - - if (control != NULL) { - m_freem(control); - control = NULL; - } - /* * Fill in mbuf with extended UDP header * and addresses and length put into network format. */ ui = mtod(m, struct udpiphdr *); ui->ui_pr = IPPROTO_UDP; - ui->ui_src = pktopts.ippo_laddr.sin_addr; + ui->ui_src = inp->inp_laddr; ui->ui_dst = inp->inp_faddr; ui->ui_sport = inp->inp_lport; ui->ui_dport = inp->inp_fport; @@ -848,312 +1170,194 @@ udp_output(struct mbuf *m, struct inpcb ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */ UDP_STATINC(UDP_STAT_OPACKETS); - flags |= inp->inp_socket->so_options & (SO_DONTROUTE|SO_BROADCAST); - return ip_output(m, inp->inp_options, ro, flags, pktopts.ippo_imo, inp); + return (ip_output(m, inp->inp_options, ro, + inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST), + inp->inp_moptions, inp->inp_socket)); - release: - if (control != NULL) - m_freem(control); +release: m_freem(m); - return error; + return (error); } -static int -udp_attach(struct socket *so, int proto) -{ - struct inpcb *inp; - int error; - - KASSERT(sotoinpcb(so) == NULL); +int udp_sendspace = 9216; /* really max datagram size */ +int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in)); + /* 40 1K datagrams */ - /* 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) +/*ARGSUSED*/ +int +udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam, + struct mbuf *control, struct lwp *l) { struct inpcb *inp; - - KASSERT(solocked(so)); - inp = sotoinpcb(so); - KASSERT(inp != NULL); - in_pcbdetach(inp); -} - -static int -udp_accept(struct socket *so, struct sockaddr *nam) -{ - KASSERT(solocked(so)); - - panic("udp_accept"); - - return EOPNOTSUPP; -} - -static int -udp_bind(struct socket *so, struct sockaddr *nam, struct lwp *l) -{ - struct inpcb *inp = sotoinpcb(so); - struct sockaddr_in *sin = (struct sockaddr_in *)nam; - int error = 0; int s; - - KASSERT(solocked(so)); - KASSERT(inp != NULL); - KASSERT(nam != NULL); - - s = splsoftnet(); - error = in_pcbbind(inp, sin, l); - splx(s); - - return error; -} - -static int -udp_listen(struct socket *so, struct lwp *l) -{ - KASSERT(solocked(so)); - - return EOPNOTSUPP; -} - -static int -udp_connect(struct socket *so, struct sockaddr *nam, struct lwp *l) -{ - 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, (struct sockaddr_in *)nam, l); - if (! error) - soisconnected(so); - splx(s); - return error; -} -static int -udp_connect2(struct socket *so, struct socket *so2) -{ - KASSERT(solocked(so)); - - return EOPNOTSUPP; -} - -static int -udp_disconnect(struct socket *so) -{ - struct inpcb *inp = sotoinpcb(so); - int s; - - KASSERT(solocked(so)); - KASSERT(inp != NULL); + if (req == PRU_CONTROL) + return (in_control(so, (long)m, (void *)nam, + (struct ifnet *)control, l)); 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; -} - -static int -udp_peeraddr(struct socket *so, struct sockaddr *nam) -{ - int s; - KASSERT(solocked(so)); - KASSERT(sotoinpcb(so) != NULL); - KASSERT(nam != NULL); - - s = splsoftnet(); - in_setpeeraddr(sotoinpcb(so), (struct sockaddr_in *)nam); - splx(s); + 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; -} + inp = sotoinpcb(so); +#ifdef DIAGNOSTIC + if (req != PRU_SEND && req != PRU_SENDOOB && control) + panic("udp_usrreq: unexpected control mbuf"); +#endif + if (req == PRU_ATTACH) { + sosetlock(so); + } else if (inp == 0) { + error = EINVAL; + goto release; + } -static int -udp_sockaddr(struct socket *so, struct sockaddr *nam) -{ - int s; + /* + * Note: need to block udp_input while changing + * the udp pcb queue and/or pcb addresses. + */ + switch (req) { - KASSERT(solocked(so)); - KASSERT(sotoinpcb(so) != NULL); - KASSERT(nam != NULL); + 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; - s = splsoftnet(); - in_setsockaddr(sotoinpcb(so), (struct sockaddr_in *)nam); - splx(s); + case PRU_DETACH: + in_pcbdetach(inp); + break; - return 0; -} + case PRU_BIND: + error = in_pcbbind(inp, nam, l); + break; -static int -udp_rcvd(struct socket *so, int flags, struct lwp *l) -{ - KASSERT(solocked(so)); + case PRU_LISTEN: + error = EOPNOTSUPP; + break; - return EOPNOTSUPP; -} + case PRU_CONNECT: + error = in_pcbconnect(inp, nam, l); + if (error) + break; + soisconnected(so); + break; -static int -udp_recvoob(struct socket *so, struct mbuf *m, int flags) -{ - KASSERT(solocked(so)); + case PRU_CONNECT2: + error = EOPNOTSUPP; + break; - return EOPNOTSUPP; -} + 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; -static int -udp_send(struct socket *so, struct mbuf *m, struct sockaddr *nam, - struct mbuf *control, struct lwp *l) -{ - struct inpcb *inp = sotoinpcb(so); - int error = 0; - struct in_addr laddr; /* XXX */ - int s; + case PRU_SHUTDOWN: + socantsendmore(so); + break; - KASSERT(solocked(so)); - KASSERT(inp != NULL); - KASSERT(m != NULL); + case PRU_RCVD: + error = EOPNOTSUPP; + break; - memset(&laddr, 0, sizeof laddr); + case PRU_SEND: + if (control && control->m_len) { + m_freem(control); + m_freem(m); + error = EINVAL; + break; + } + { + struct in_addr laddr; /* XXX */ - s = splsoftnet(); - if (nam) { - laddr = inp->inp_laddr; /* XXX */ - if ((so->so_state & SS_ISCONNECTED) != 0) { - error = EISCONN; - goto die; + memset(&laddr, 0, sizeof laddr); + if (nam) { + laddr = inp->inp_laddr; /* XXX */ + if ((so->so_state & SS_ISCONNECTED) != 0) { + error = EISCONN; + goto die; + } + error = in_pcbconnect(inp, nam, l); + if (error) + goto die; + } else { + if ((so->so_state & SS_ISCONNECTED) == 0) { + error = ENOTCONN; + goto die; + } } - error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l); - if (error) - goto die; - } else { - if ((so->so_state & SS_ISCONNECTED) == 0) { - error = ENOTCONN; - goto die; + error = udp_output(m, inp); + m = NULL; + if (nam) { + in_pcbdisconnect(inp); + inp->inp_laddr = laddr; /* XXX */ + in_pcbstate(inp, INP_BOUND); /* XXX */ } + die: + if (m) + m_freem(m); } - error = udp_output(m, inp, control, l); - m = NULL; - control = NULL; - if (nam) { - in_pcbdisconnect(inp); - inp->inp_laddr = laddr; /* XXX */ - in_pcbstate(inp, INP_BOUND); /* XXX */ - } - die: - if (m != NULL) - m_freem(m); - if (control != NULL) - m_freem(control); + break; - splx(s); - return error; -} + case PRU_SENSE: + /* + * stat: don't bother with a blocksize. + */ + splx(s); + return (0); -static int -udp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control) -{ - KASSERT(solocked(so)); + case PRU_RCVOOB: + error = EOPNOTSUPP; + break; - m_freem(m); - m_freem(control); + case PRU_SENDOOB: + m_freem(control); + m_freem(m); + error = EOPNOTSUPP; + break; - return EOPNOTSUPP; -} + case PRU_SOCKADDR: + in_setsockaddr(inp, nam); + break; -static int -udp_purgeif(struct socket *so, struct ifnet *ifp) -{ - int s; + case PRU_PEERADDR: + in_setpeeraddr(inp, nam); + break; - s = splsoftnet(); - mutex_enter(softnet_lock); - in_pcbpurgeif0(&udbtable, ifp); -#ifdef NET_MPSAFE - mutex_exit(softnet_lock); -#endif - in_purgeif(ifp); -#ifdef NET_MPSAFE - mutex_enter(softnet_lock); -#endif - in_pcbpurgeif(&udbtable, ifp); - mutex_exit(softnet_lock); - splx(s); + default: + panic("udp_usrreq"); + } - return 0; +release: + splx(s); + return (error); } static int @@ -1169,7 +1373,11 @@ sysctl_net_inet_udp_stats(SYSCTLFN_ARGS) 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, @@ -1235,7 +1443,7 @@ udp_statinc(u_int stat) UDP_STATINC(stat); } -#if defined(INET) && defined(IPSEC) +#if defined(INET) && defined(FAST_IPSEC) /* * Returns: * 1 if the packet was processed @@ -1359,55 +1567,14 @@ udp4_espinudp(struct mbuf **mp, int off, ((u_int16_t *)(tag + 1))[1] = dport; m_tag_prepend(m, tag); - if (ipsec_used) - ipsec4_common_input(m, iphdrlen, IPPROTO_ESP); - /* XXX: else */ +#ifdef FAST_IPSEC + ipsec4_common_input(m, iphdrlen, IPPROTO_ESP); +#else + esp4_input(m, iphdrlen); +#endif /* 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_connect2 udp_connect2_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_rcvd udp_rcvd_wrapper -#define udp_recvoob udp_recvoob_wrapper -#define udp_send udp_send_wrapper -#define udp_sendoob udp_sendoob_wrapper -#define udp_purgeif udp_purgeif_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_connect2 = udp_connect2, - .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_rcvd = udp_rcvd, - .pr_recvoob = udp_recvoob, - .pr_send = udp_send, - .pr_sendoob = udp_sendoob, - .pr_purgeif = udp_purgeif, -};