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 1.97 retrieving revision 1.141 diff -u -p -r1.97 -r1.141 --- src/sys/netinet/udp_usrreq.c 2003/01/20 00:05:46 1.97 +++ src/sys/netinet/udp_usrreq.c 2005/08/10 13:06:49 1.141 @@ -1,4 +1,4 @@ -/* $NetBSD: udp_usrreq.c,v 1.97 2003/01/20 00:05:46 simonb Exp $ */ +/* $NetBSD: udp_usrreq.c,v 1.141 2005/08/10 13:06:49 yamt Exp $ */ /* * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. @@ -41,11 +41,7 @@ * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. - * 3. All advertising materials mentioning features or use of this software - * must display the following acknowledgement: - * This product includes software developed by the University of - * California, Berkeley and its contributors. - * 4. Neither the name of the University nor the names of its contributors + * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * @@ -65,12 +61,13 @@ */ #include -__KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.97 2003/01/20 00:05:46 simonb Exp $"); +__KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.141 2005/08/10 13:06:49 yamt Exp $"); #include "opt_inet.h" #include "opt_ipsec.h" #include "opt_inet_csum.h" #include "opt_ipkdb.h" +#include "opt_mbuftrace.h" #include #include @@ -98,6 +95,11 @@ __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c #include #include +#ifdef IPSEC_NAT_T +#include +#include +#endif + #ifdef INET6 #include #include @@ -106,12 +108,10 @@ __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c #include #endif -#ifdef PULLDOWN_TEST #ifndef INET6 /* always need ip6.h for IP6_EXTHDR_GET */ #include #endif -#endif #include "faith.h" #if defined(NFAITH) && NFAITH > 0 @@ -120,6 +120,14 @@ __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c #include +#ifdef FAST_IPSEC +#include +#include /* XXX ipsecstat namespace */ +#ifdef INET6 +#include +#endif +#endif /* FAST_IPSEC*/ + #ifdef IPSEC #include #include @@ -138,24 +146,31 @@ int udpcksum = 1; #else int udpcksum = 0; /* XXX */ #endif +int udp_do_loopback_cksum = 0; struct inpcbtable udbtable; struct udpstat udpstat; #ifdef INET -static void udp4_sendup __P((struct mbuf *, int, struct sockaddr *, - struct socket *)); -static int udp4_realinput __P((struct sockaddr_in *, struct sockaddr_in *, - struct mbuf *, int)); +#ifdef IPSEC_NAT_T +static int udp4_espinudp (struct mbuf *, int, struct sockaddr *, + struct socket *); +#endif +static void udp4_sendup (struct mbuf *, int, struct sockaddr *, + struct socket *); +static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *, + struct mbuf *, int); +static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int); #endif #ifdef INET6 -static void udp6_sendup __P((struct mbuf *, int, struct sockaddr *, - struct socket *)); -static int udp6_realinput __P((int, struct sockaddr_in6 *, - struct sockaddr_in6 *, struct mbuf *, int)); +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 __P((struct inpcb *, int)); +static void udp_notify (struct inpcb *, int); #endif #ifndef UDBHASHSIZE @@ -163,9 +178,16 @@ static void udp_notify __P((struct inpcb #endif int udbhashsize = UDBHASHSIZE; +#ifdef MBUFTRACE +struct mowner udp_mowner = { "udp" }; +struct mowner udp_rx_mowner = { "udp", "rx" }; +struct mowner udp_tx_mowner = { "udp", "tx" }; +#endif + #ifdef UDP_CSUM_COUNTERS #include +#if defined(INET) struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, NULL, "udp", "hwcsum bad"); struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, @@ -175,6 +197,28 @@ struct evcnt udp_hwcsum_data = EVCNT_INI struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, NULL, "udp", "swcsum"); +EVCNT_ATTACH_STATIC(udp_hwcsum_bad); +EVCNT_ATTACH_STATIC(udp_hwcsum_ok); +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 @@ -184,30 +228,114 @@ struct evcnt udp_swcsum = EVCNT_INITIALI #endif /* UDP_CSUM_COUNTERS */ void -udp_init() +udp_init(void) { -#ifdef INET in_pcbinit(&udbtable, udbhashsize, udbhashsize); -#endif -#ifdef UDP_CSUM_COUNTERS - evcnt_attach_static(&udp_hwcsum_bad); - evcnt_attach_static(&udp_hwcsum_ok); - evcnt_attach_static(&udp_hwcsum_data); - evcnt_attach_static(&udp_swcsum); -#endif /* UDP_CSUM_COUNTERS */ + MOWNER_ATTACH(&udp_tx_mowner); + MOWNER_ATTACH(&udp_rx_mowner); + MOWNER_ATTACH(&udp_mowner); } +/* + * Checksum extended UDP header and data. + */ + +int +udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh, + int iphlen, int len) +{ + + switch (af) { #ifdef INET + case AF_INET: + return udp4_input_checksum(m, uh, iphlen, len); +#endif +#ifdef INET6 + case AF_INET6: + return udp6_input_checksum(m, uh, iphlen, len); +#endif + } +#ifdef DIAGNOSTIC + panic("udp_input_checksum: unknown af %d", af); +#endif + /* NOTREACHED */ + return -1; +} + +#ifdef INET + +/* + * Checksum extended UDP header and data. + */ + +static int +udp4_input_checksum(struct mbuf *m, const struct udphdr *uh, + int iphlen, int len) +{ + + /* + * XXX it's better to record and check if this mbuf is + * already checked. + */ + + if (uh->uh_sum == 0) + return 0; + + switch (m->m_pkthdr.csum_flags & + ((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); + goto badcsum; + + case M_CSUM_UDPv4|M_CSUM_DATA: { + u_int32_t hw_csum = m->m_pkthdr.csum_data; + + UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data); + if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) { + const struct ip *ip = + mtod(m, const struct ip *); + + hw_csum = in_cksum_phdr(ip->ip_src.s_addr, + ip->ip_dst.s_addr, + htons(hw_csum + len + IPPROTO_UDP)); + } + if ((hw_csum ^ 0xffff) != 0) + goto badcsum; + break; + } + + case M_CSUM_UDPv4: + /* Checksum was okay. */ + UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok); + break; + + default: + /* + * Need to compute it ourselves. Maybe skip checksum + * on loopback interfaces. + */ + if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & + IFF_LOOPBACK) || + udp_do_loopback_cksum)) { + UDP_CSUM_COUNTER_INCR(&udp_swcsum); + if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0) + goto badcsum; + } + break; + } + + return 0; + +badcsum: + udpstat.udps_badsum++; + return -1; +} + void -#if __STDC__ udp_input(struct mbuf *m, ...) -#else -udp_input(m, va_alist) - struct mbuf *m; - va_dcl -#endif { va_list ap; struct sockaddr_in src, dst; @@ -223,46 +351,18 @@ udp_input(m, va_alist) (void)va_arg(ap, int); /* ignore value, advance ap */ va_end(ap); + MCLAIM(m, &udp_rx_mowner); udpstat.udps_ipackets++; -#ifndef PULLDOWN_TEST - /* - * Strip IP options, if any; should skip this, - * make available to user, and use on returned packets, - * but we don't yet have a way to check the checksum - * with options still present. - */ - if (iphlen > sizeof (struct ip)) { - ip_stripoptions(m, (struct mbuf *)0); - iphlen = sizeof(struct ip); - } -#else - /* - * we may enable the above code if we save and pass IPv4 options - * to the userland. - */ -#endif - /* * Get IP and UDP header together in first mbuf. */ ip = mtod(m, struct ip *); -#ifndef PULLDOWN_TEST - if (m->m_len < iphlen + sizeof(struct udphdr)) { - if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) { - udpstat.udps_hdrops++; - return; - } - ip = mtod(m, struct ip *); - } - uh = (struct udphdr *)((caddr_t)ip + iphlen); -#else IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr)); if (uh == NULL) { udpstat.udps_hdrops++; return; } -#endif KASSERT(UDP_HDR_ALIGNED_P(uh)); /* destination port of 0 is illegal, based on RFC768. */ @@ -286,33 +386,8 @@ udp_input(m, va_alist) /* * Checksum extended UDP header and data. */ - if (uh->uh_sum) { - switch (m->m_pkthdr.csum_flags & - ((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); - goto badcsum; - - case M_CSUM_UDPv4|M_CSUM_DATA: - UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data); - if ((m->m_pkthdr.csum_data ^ 0xffff) != 0) - goto badcsum; - break; - - case M_CSUM_UDPv4: - /* Checksum was okay. */ - UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok); - break; - - default: - /* Need to compute it ourselves. */ - UDP_CSUM_COUNTER_INCR(&udp_swcsum); - if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0) - goto badcsum; - break; - } - } + if (udp4_input_checksum(m, uh, iphlen, len)) + goto badcsum; /* construct source and dst sockaddrs. */ bzero(&src, sizeof(src)); @@ -378,15 +453,64 @@ bad: badcsum: m_freem(m); - udpstat.udps_badsum++; } #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(mp, offp, proto) - struct mbuf **mp; - int *offp, proto; +udp6_input(struct mbuf **mp, int *offp, int proto) { struct mbuf *m = *mp; int off = *offp; @@ -395,9 +519,6 @@ udp6_input(mp, offp, proto) struct udphdr *uh; u_int32_t plen, ulen; -#ifndef PULLDOWN_TEST - IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE); -#endif ip6 = mtod(m, struct ip6_hdr *); #if defined(NFAITH) && 0 < NFAITH @@ -412,15 +533,11 @@ udp6_input(mp, offp, proto) /* check for jumbogram is done in ip6_input. we can trust pkthdr.len */ plen = m->m_pkthdr.len - off; -#ifndef PULLDOWN_TEST - uh = (struct udphdr *)((caddr_t)ip6 + off); -#else IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr)); if (uh == NULL) { ip6stat.ip6s_tooshort++; return IPPROTO_DONE; } -#endif KASSERT(UDP_HDR_ALIGNED_P(uh)); ulen = ntohs((u_short)uh->uh_ulen); /* @@ -447,14 +564,11 @@ udp6_input(mp, offp, proto) } /* - * Checksum extended UDP header and data. + * Checksum extended UDP header and data. Maybe skip checksum + * on loopback interfaces. */ - if (uh->uh_sum == 0) - udp6stat.udp6s_nosum++; - else if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) { - udp6stat.udp6s_badsum++; + if (udp6_input_checksum(m, uh, off, ulen)) goto bad; - } /* * Construct source and dst sockaddrs. @@ -492,11 +606,8 @@ bad: #ifdef INET static void -udp4_sendup(m, off, src, so) - struct mbuf *m; - int off; /* offset of data portion */ - struct sockaddr *src; - struct socket *so; +udp4_sendup(struct mbuf *m, int off /* offset of data portion */, + struct sockaddr *src, struct socket *so) { struct mbuf *opts = NULL; struct mbuf *n; @@ -516,10 +627,13 @@ udp4_sendup(m, off, src, so) return; } -#ifdef IPSEC +#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) + icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT, + 0, 0); return; } #endif /*IPSEC*/ @@ -537,6 +651,7 @@ udp4_sendup(m, off, src, so) m_freem(n); if (opts) m_freem(opts); + so->so_rcv.sb_overflowed++; udpstat.udps_fullsock++; } else sorwakeup(so); @@ -546,11 +661,8 @@ udp4_sendup(m, off, src, so) #ifdef INET6 static void -udp6_sendup(m, off, src, so) - struct mbuf *m; - int off; /* offset of data portion */ - struct sockaddr *src; - struct socket *so; +udp6_sendup(struct mbuf *m, int off /* offset of data portion */, + struct sockaddr *src, struct socket *so) { struct mbuf *opts = NULL; struct mbuf *n; @@ -562,10 +674,13 @@ udp6_sendup(m, off, src, so) return; in6p = sotoin6pcb(so); -#ifdef IPSEC +#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) + icmp6_error(n, ICMP6_DST_UNREACH, + ICMP6_DST_UNREACH_ADMIN, 0); return; } #endif /*IPSEC*/ @@ -582,6 +697,7 @@ udp6_sendup(m, off, src, so) m_freem(n); if (opts) m_freem(opts); + so->so_rcv.sb_overflowed++; udp6stat.udp6s_fullsock++; } else sorwakeup(so); @@ -591,15 +707,13 @@ udp6_sendup(m, off, src, so) #ifdef INET static int -udp4_realinput(src, dst, m, off) - struct sockaddr_in *src; - struct sockaddr_in *dst; - struct mbuf *m; - int off; /* offset of udphdr */ +udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst, + struct mbuf *m, int off /* offset of udphdr */) { u_int16_t *sport, *dport; int rcvcnt; struct in_addr *src4, *dst4; + struct inpcb_hdr *inph; struct inpcb *inp; rcvcnt = 0; @@ -638,7 +752,11 @@ udp4_realinput(src, dst, m, off) /* * Locate pcb(s) for datagram. */ - CIRCLEQ_FOREACH(inp, &udbtable.inpt_queue, inp_queue) { + CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) { + inp = (struct inpcb *)inph; + if (inp->inp_af != AF_INET) + continue; + if (inp->inp_lport != *dport) continue; if (!in_nullhost(inp->inp_laddr)) { @@ -679,6 +797,20 @@ udp4_realinput(src, dst, m, off) return rcvcnt; } +#ifdef IPSEC_NAT_T + /* Handle ESP over UDP */ + if (inp->inp_flags & INP_ESPINUDP_ALL) { + struct sockaddr *sa = (struct sockaddr *)src; + + if (udp4_espinudp(m, off, sa, inp->inp_socket) != 0) { + rcvcnt++; + goto bad; + } + + /* Normal UDP processing will take place */ + } +#endif + udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket); rcvcnt++; } @@ -690,17 +822,14 @@ bad: #ifdef INET6 static int -udp6_realinput(af, src, dst, m, off) - int af; /* af on packet */ - struct sockaddr_in6 *src; - struct sockaddr_in6 *dst; - struct mbuf *m; - int off; /* offset of udphdr */ +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; @@ -742,8 +871,11 @@ udp6_realinput(af, src, dst, m, off) /* * Locate pcb(s) for datagram. */ - for (in6p = udb6.in6p_next; in6p != &udb6; - in6p = in6p->in6p_next) { + 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)) { @@ -784,11 +916,11 @@ udp6_realinput(af, src, dst, m, off) /* * Locate pcb for datagram. */ - in6p = in6_pcblookup_connect(&udb6, &src6, sport, + in6p = in6_pcblookup_connect(&udbtable, &src6, sport, &dst6, dport, 0); if (in6p == 0) { ++udpstat.udps_pcbhashmiss; - in6p = in6_pcblookup_bind(&udb6, &dst6, dport, 0); + in6p = in6_pcblookup_bind(&udbtable, &dst6, dport, 0); if (in6p == 0) return rcvcnt; } @@ -808,25 +940,19 @@ bad: * just wake up so that he can collect error status. */ static void -udp_notify(inp, errno) - struct inpcb *inp; - int errno; +udp_notify(struct inpcb *inp, int errno) { - inp->inp_socket->so_error = errno; sorwakeup(inp->inp_socket); sowwakeup(inp->inp_socket); } void * -udp_ctlinput(cmd, sa, v) - int cmd; - struct sockaddr *sa; - void *v; +udp_ctlinput(int cmd, struct sockaddr *sa, void *v) { struct ip *ip = v; struct udphdr *uh; - void (*notify) __P((struct inpcb *, int)) = udp_notify; + void (*notify)(struct inpcb *, int) = udp_notify; int errno; if (sa->sa_family != AF_INET @@ -854,20 +980,110 @@ udp_ctlinput(cmd, sa, v) } int -#if __STDC__ -udp_output(struct mbuf *m, ...) -#else -udp_output(m, va_alist) +udp_ctloutput(op, so, level, optname, mp) + int op; + struct socket *so; + int level, optname; + struct mbuf **mp; +{ + int s; + int error = 0; struct mbuf *m; - va_dcl + struct inpcb *inp; + int family; + + family = so->so_proto->pr_domain->dom_family; + + s = splsoftnet(); + switch (family) { +#ifdef INET + case PF_INET: + if (level != IPPROTO_UDP) { + error = ip_ctloutput(op, so, level, optname, mp); + goto end; + } + break; +#endif +#ifdef INET6 + case PF_INET6: + if (level != IPPROTO_UDP) { + error = ip6_ctloutput(op, so, level, optname, mp); + goto end; + } + break; #endif + default: + error = EAFNOSUPPORT; + goto end; + break; + } + + + switch (op) { + case PRCO_SETOPT: + m = *mp; + inp = sotoinpcb(so); + + switch (optname) { + case UDP_ENCAP: + if (m == NULL || m->m_len < sizeof (int)) { + error = EINVAL; + goto end; + } + + switch(*mtod(m, int *)) { +#ifdef IPSEC_NAT_T + case 0: + inp->inp_flags &= ~INP_ESPINUDP_ALL; + break; + + case UDP_ENCAP_ESPINUDP: + inp->inp_flags &= ~INP_ESPINUDP_ALL; + inp->inp_flags |= INP_ESPINUDP; + break; + + case UDP_ENCAP_ESPINUDP_NON_IKE: + inp->inp_flags &= ~INP_ESPINUDP_ALL; + inp->inp_flags |= INP_ESPINUDP_NON_IKE; + break; +#endif + default: + error = EINVAL; + goto end; + break; + } + break; + + default: + error = ENOPROTOOPT; + goto end; + break; + } + break; + + default: + error = EINVAL; + goto end; + break; + } + +end: + splx(s); + return error; +} + + +int +udp_output(struct mbuf *m, ...) { struct inpcb *inp; struct udpiphdr *ui; + struct route *ro; int len = m->m_pkthdr.len; int error = 0; va_list ap; + MCLAIM(m, &udp_tx_mowner); va_start(ap, m); inp = va_arg(ap, struct inpcb *); va_end(ap); @@ -903,6 +1119,8 @@ udp_output(m, va_alist) ui->ui_dport = inp->inp_fport; ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr)); + ro = &inp->inp_route; + /* * Set up checksum and output datagram. */ @@ -923,16 +1141,9 @@ udp_output(m, va_alist) ((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos; /* XXX */ udpstat.udps_opackets++; -#ifdef IPSEC - if (ipsec_setsocket(m, inp->inp_socket) != 0) { - error = ENOBUFS; - goto release; - } -#endif /*IPSEC*/ - - return (ip_output(m, inp->inp_options, &inp->inp_route, + return (ip_output(m, inp->inp_options, ro, inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST), - inp->inp_moptions)); + inp->inp_moptions, inp->inp_socket)); release: m_freem(m); @@ -945,11 +1156,8 @@ int udp_recvspace = 40 * (1024 + sizeof( /*ARGSUSED*/ int -udp_usrreq(so, req, m, nam, control, p) - struct socket *so; - int req; - struct mbuf *m, *nam, *control; - struct proc *p; +udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam, + struct mbuf *control, struct proc *p) { struct inpcb *inp; int s; @@ -988,6 +1196,11 @@ udp_usrreq(so, req, m, nam, control, p) 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) @@ -1056,11 +1269,8 @@ udp_usrreq(so, req, m, nam, control, p) goto die; } error = in_pcbconnect(inp, nam); - if (error) { - die: - m_freem(m); - break; - } + if (error) + goto die; } else { if ((so->so_state & SS_ISCONNECTED) == 0) { error = ENOTCONN; @@ -1068,11 +1278,15 @@ udp_usrreq(so, req, m, nam, control, p) } } 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); } break; @@ -1113,31 +1327,204 @@ release: /* * Sysctl for udp variables. */ -int -udp_sysctl(name, namelen, oldp, oldlenp, newp, newlen) - int *name; - u_int namelen; - void *oldp; - size_t *oldlenp; - void *newp; - size_t newlen; +SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup") { - /* All sysctl names at this level are terminal. */ - if (namelen != 1) - return (ENOTDIR); - - switch (name[0]) { - case UDPCTL_CHECKSUM: - return (sysctl_int(oldp, oldlenp, newp, newlen, &udpcksum)); - case UDPCTL_SENDSPACE: - return (sysctl_int(oldp, oldlenp, newp, newlen, - &udp_sendspace)); - case UDPCTL_RECVSPACE: - return (sysctl_int(oldp, oldlenp, newp, newlen, - &udp_recvspace)); - default: - return (ENOPROTOOPT); + + 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); + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT, + CTLTYPE_NODE, "udp", + SYSCTL_DESCR("UDPv4 related settings"), + NULL, 0, NULL, 0, + CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL); + + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "checksum", + SYSCTL_DESCR("Compute UDP checksums"), + NULL, 0, &udpcksum, 0, + CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM, + CTL_EOL); + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "sendspace", + SYSCTL_DESCR("Default UDP send buffer size"), + NULL, 0, &udp_sendspace, 0, + CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE, + CTL_EOL); + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "recvspace", + SYSCTL_DESCR("Default UDP receive buffer size"), + NULL, 0, &udp_recvspace, 0, + CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE, + CTL_EOL); + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT|CTLFLAG_READWRITE, + CTLTYPE_INT, "do_loopback_cksum", + SYSCTL_DESCR("Perform UDP checksum on loopback"), + NULL, 0, &udp_do_loopback_cksum, 0, + CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM, + CTL_EOL); + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT, + CTLTYPE_STRUCT, "pcblist", + SYSCTL_DESCR("UDP protocol control block list"), + sysctl_inpcblist, 0, &udbtable, 0, + CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE, + CTL_EOL); + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT, + CTLTYPE_STRUCT, "stats", + SYSCTL_DESCR("UDP statistics"), + NULL, 0, &udpstat, sizeof(udpstat), + CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS, + CTL_EOL); +} +#endif + +#if (defined INET && defined IPSEC_NAT_T) +/* + * Returns: + * 1 if the packet was processed + * 0 if normal UDP processing should take place + */ +static int +udp4_espinudp(m, off, src, so) + struct mbuf *m; + int off; + struct sockaddr *src; + struct socket *so; +{ + size_t len; + caddr_t data; + struct inpcb *inp; + size_t skip = 0; + size_t minlen; + size_t iphdrlen; + struct ip *ip; + struct mbuf *n; + struct m_tag *tag; + struct udphdr *udphdr; + u_int16_t sport, dport; + + /* + * Collapse the mbuf chain if the first mbuf is too short + * The longest case is: UDP + non ESP marker + ESP + */ + minlen = off + sizeof(u_int64_t) + sizeof(struct esp); + if (minlen > m->m_pkthdr.len) + minlen = m->m_pkthdr.len; + + if (m->m_len < minlen) { + if ((m = m_pullup(m, minlen)) == NULL) { + printf("udp4_espinudp: m_pullup failed\n"); + return 0; + } } - /* NOTREACHED */ + + len = m->m_len - off; + data = mtod(m, caddr_t) + off; + inp = sotoinpcb(so); + + /* Ignore keepalive packets */ + if ((len == 1) && (data[0] == '\xff')) { + return 1; + } + + /* + * Check that the payload is long enough to hold + * an ESP header and compute the length of encapsulation + * header to remove + */ + if (inp->inp_flags & INP_ESPINUDP) { + u_int32_t *st = (u_int32_t *)data; + + if ((len <= sizeof(struct esp)) || (*st == 0)) + return 0; /* Normal UDP processing */ + + skip = sizeof(struct udphdr); + } + + if (inp->inp_flags & INP_ESPINUDP_NON_IKE) { + u_int64_t *st = (u_int64_t *)data; + + if ((len <= sizeof(u_int64_t) + sizeof(struct esp)) + || (*st != 0)) + return 0; /* Normal UDP processing */ + + skip = sizeof(struct udphdr) + sizeof(u_int64_t); + } + + /* + * Get the UDP ports. They are handled in network + * order everywhere in IPSEC_NAT_T code. + */ + udphdr = (struct udphdr *)(data - skip); + sport = udphdr->uh_sport; + dport = udphdr->uh_dport; + + /* + * Remove the UDP header (and possibly the non ESP marker) + * IP header lendth is iphdrlen + * Before: + * <--- off ---> + * +----+------+-----+ + * | IP | UDP | ESP | + * +----+------+-----+ + * <-skip-> + * After: + * +----+-----+ + * | IP | ESP | + * +----+-----+ + * <-skip-> + */ + iphdrlen = off - sizeof(struct udphdr); + memmove(mtod(m, caddr_t) + skip, mtod(m, caddr_t), iphdrlen); + m_adj(m, skip); + + ip = mtod(m, struct ip *); + ip->ip_len = htons(ntohs(ip->ip_len) - skip); + 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; + } + + /* + * 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 + * same NAT + */ + 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; + } + ((u_int16_t *)(tag + 1))[0] = sport; + ((u_int16_t *)(tag + 1))[1] = dport; + m_tag_prepend(n, tag); + + esp4_input(n, iphdrlen); + + /* We handled it, it shoudln't be handled by UDP */ + return 1; } #endif