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 1.66 retrieving revision 1.84 diff -u -p -r1.66 -r1.84 --- src/sys/netinet/udp_usrreq.c 2000/03/30 13:25:11 1.66 +++ src/sys/netinet/udp_usrreq.c 2001/09/17 17:27:01 1.84 @@ -1,4 +1,4 @@ -/* $NetBSD: udp_usrreq.c,v 1.66 2000/03/30 13:25:11 augustss Exp $ */ +/* $NetBSD: udp_usrreq.c,v 1.84 2001/09/17 17:27:01 thorpej Exp $ */ /* * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. @@ -64,7 +64,9 @@ * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95 */ +#include "opt_inet.h" #include "opt_ipsec.h" +#include "opt_inet_csum.h" #include "opt_ipkdb.h" #include @@ -79,7 +81,7 @@ #include #include -#include +#include #include #include @@ -110,12 +112,16 @@ #endif #endif +#include "faith.h" +#if defined(NFAITH) && NFAITH > 0 +#include +#endif + #include #ifdef IPSEC #include #include -#include #endif /*IPSEC*/ #ifdef IPKDB @@ -132,10 +138,12 @@ int udpcksum = 1; int udpcksum = 0; /* XXX */ #endif +#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)); +#endif #ifdef INET6 static void udp6_sendup __P((struct mbuf *, int, struct sockaddr *, struct socket *)); @@ -144,21 +152,53 @@ static int in6_mcmatch __P((struct in6pc static int udp6_realinput __P((int, struct sockaddr_in6 *, struct sockaddr_in6 *, struct mbuf *, int)); #endif +#ifdef INET static void udp_notify __P((struct inpcb *, int)); +#endif #ifndef UDBHASHSIZE #define UDBHASHSIZE 128 #endif int udbhashsize = UDBHASHSIZE; +#ifdef UDP_CSUM_COUNTERS +#include + +struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, + NULL, "udp", "hwcsum bad"); +struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, + NULL, "udp", "hwcsum ok"); +struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, + NULL, "udp", "hwcsum data"); +struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, + NULL, "udp", "swcsum"); + +#define UDP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ + +#else + +#define UDP_CSUM_COUNTER_INCR(ev) /* nothing */ + +#endif /* UDP_CSUM_COUNTERS */ + void udp_init() { +#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 */ } #ifndef UDP6 +#ifdef INET void #if __STDC__ udp_input(struct mbuf *m, ...) @@ -232,7 +272,7 @@ udp_input(m, va_alist) */ len = ntohs((u_int16_t)uh->uh_ulen); if (ip->ip_len != iphlen + len) { - if (ip->ip_len < iphlen + len) { + if (ip->ip_len < iphlen + len || len < sizeof(struct udphdr)) { udpstat.udps_badlen++; goto bad; } @@ -243,10 +283,30 @@ udp_input(m, va_alist) * Checksum extended UDP header and data. */ if (uh->uh_sum) { - if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0) { - udpstat.udps_badsum++; - m_freem(m); - return; + 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; } } @@ -310,7 +370,13 @@ udp_input(m, va_alist) bad: if (m) m_freem(m); + return; + +badcsum: + m_freem(m); + udpstat.udps_badsum++; } +#endif #ifdef INET6 int @@ -325,23 +391,21 @@ 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 - if (m->m_pkthdr.rcvif) { - if (m->m_pkthdr.rcvif->if_type == IFT_FAITH) { - /* send icmp6 host unreach? */ - m_freem(m); - return IPPROTO_DONE; - } + if (faithprefix(&ip6->ip6_dst)) { + /* send icmp6 host unreach? */ + m_freem(m); + return IPPROTO_DONE; } #endif udp6stat.udp6s_ipackets++; -#ifndef PULLDOWN_TEST - IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE); -#endif - - ip6 = mtod(m, struct ip6_hdr *); /* check for jumbogram is done in ip6_input. we can trust pkthdr.len */ plen = m->m_pkthdr.len - off; #ifndef PULLDOWN_TEST @@ -354,6 +418,10 @@ udp6_input(mp, offp, proto) } #endif 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; @@ -390,28 +458,14 @@ udp6_input(mp, offp, proto) bzero(&src, sizeof(src)); src.sin6_family = AF_INET6; src.sin6_len = sizeof(struct sockaddr_in6); - bcopy(&ip6->ip6_src, &src.sin6_addr, sizeof(src.sin6_addr)); - if (IN6_IS_SCOPE_LINKLOCAL(&src.sin6_addr)) - src.sin6_addr.s6_addr16[1] = 0; - if (m->m_pkthdr.rcvif) { - if (IN6_IS_SCOPE_LINKLOCAL(&src.sin6_addr)) - src.sin6_scope_id = m->m_pkthdr.rcvif->if_index; - else - src.sin6_scope_id = 0; - } + /* KAME hack: recover scopeid */ + (void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif); src.sin6_port = uh->uh_sport; bzero(&dst, sizeof(dst)); dst.sin6_family = AF_INET6; dst.sin6_len = sizeof(struct sockaddr_in6); - bcopy(&ip6->ip6_dst, &dst.sin6_addr, sizeof(dst.sin6_addr)); - if (IN6_IS_SCOPE_LINKLOCAL(&dst.sin6_addr)) - dst.sin6_addr.s6_addr16[1] = 0; - if (m->m_pkthdr.rcvif) { - if (IN6_IS_SCOPE_LINKLOCAL(&dst.sin6_addr)) - dst.sin6_scope_id = m->m_pkthdr.rcvif->if_index; - else - dst.sin6_scope_id = 0; - } + /* KAME hack: recover scopeid */ + (void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif); dst.sin6_port = uh->uh_dport; if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) { @@ -431,6 +485,7 @@ bad: } #endif +#ifdef INET static void udp4_sendup(m, off, src, so) struct mbuf *m; @@ -481,10 +536,12 @@ udp4_sendup(m, off, src, so) m_freem(n); if (opts) m_freem(opts); + udpstat.udps_fullsock++; } else sorwakeup(so); } } +#endif #ifdef INET6 static void @@ -531,6 +588,7 @@ udp6_sendup(m, off, src, so) } #endif +#ifdef INET static int udp4_realinput(src, dst, m, off) struct sockaddr_in *src; @@ -554,7 +612,7 @@ udp4_realinput(src, dst, m, off) dst4 = &dst->sin_addr; dport = &dst->sin_port; - if (IN_MULTICAST(src4->s_addr) || + if (IN_MULTICAST(dst4->s_addr) || in_broadcast(*dst4, m->m_pkthdr.rcvif)) { struct inpcb *last; /* @@ -575,7 +633,7 @@ udp4_realinput(src, dst, m, off) /* * KAME note: usually we drop udpiphdr from mbuf here. - * we need udpiphdr for iPsec processing so we do that later. + * we need udpiphdr for IPsec processing so we do that later. */ /* * Locate pcb(s) for datagram. @@ -612,18 +670,6 @@ udp4_realinput(src, dst, m, off) (SO_REUSEPORT|SO_REUSEADDR)) == 0) break; } - -#if 0 - if (last == NULL) { - /* - * No matching pcb found; discard datagram. - * (No need to send an ICMP Port Unreachable - * for a broadcast or multicast datgram.) - */ - udpstat.udps_noportbcast++; - goto bad; - } -#endif } else { /* * Locate pcb for datagram. @@ -632,32 +678,8 @@ udp4_realinput(src, dst, m, off) if (inp == 0) { ++udpstat.udps_pcbhashmiss; inp = in_pcblookup_bind(&udbtable, *dst4, *dport); - if (inp == 0) { -#if 0 - struct mbuf *n; - - if (m->m_flags & (M_BCAST | M_MCAST)) { - udpstat.udps_noportbcast++; - goto bad; - } - udpstat.udps_noport++; -#ifdef IPKDB - if (checkipkdb(src4, *sport, *dport, m, off, - m->m_pkthdr.len - off)) { - /* - * It was a debugger connect packet, - * just drop it now - */ - goto bad; - } -#endif - if ((n = m_copy(m, 0, M_COPYALL)) != NULL) { - icmp_error(n, ICMP_UNREACH, - ICMP_UNREACH_PORT, 0, 0); - } -#endif + if (inp == 0) return rcvcnt; - } } udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket); @@ -667,6 +689,7 @@ udp4_realinput(src, dst, m, off) bad: return rcvcnt; } +#endif #ifdef INET6 static int @@ -700,10 +723,10 @@ udp6_realinput(af, src, dst, m, off) struct mbuf *m; int off; /* offset of udphdr */ { - u_int16_t *sport, *dport; + u_int16_t sport, dport; int rcvcnt; - struct in6_addr *src6, *dst6; - struct in_addr *src4; + struct in6_addr src6, dst6; + const struct in_addr *dst4; struct in6pcb *in6p; rcvcnt = 0; @@ -714,14 +737,14 @@ udp6_realinput(af, src, dst, m, off) if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6) goto bad; - src6 = &src->sin6_addr; - sport = &src->sin6_port; - dst6 = &dst->sin6_addr; - dport = &dst->sin6_port; - src4 = (struct in_addr *)&src->sin6_addr.s6_addr32[12]; + in6_embedscope(&src6, src, NULL, NULL); + sport = src->sin6_port; + in6_embedscope(&dst6, dst, NULL, NULL); + dport = dst->sin6_port; + dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr32[12]; - if (IN6_IS_ADDR_MULTICAST(dst6) - || (af == AF_INET && IN_MULTICAST(src4->s_addr))) { + if (IN6_IS_ADDR_MULTICAST(&dst6) || + (af == AF_INET && IN_MULTICAST(dst4->s_addr))) { struct in6pcb *last; /* * Deliver a multicast or broadcast datagram to *all* sockets @@ -741,36 +764,36 @@ udp6_realinput(af, src, dst, m, off) /* * KAME note: usually we drop udpiphdr from mbuf here. - * we need udpiphdr for iPsec processing so we do that later. + * we need udpiphdr for IPsec processing so we do that later. */ /* * Locate pcb(s) for datagram. */ for (in6p = udb6.in6p_next; in6p != &udb6; in6p = in6p->in6p_next) { - if (in6p->in6p_lport != *dport) + if (in6p->in6p_lport != dport) continue; if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { - if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, dst6) - && !in6_mcmatch(in6p, dst6, m->m_pkthdr.rcvif)) + if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6) && + !in6_mcmatch(in6p, &dst6, m->m_pkthdr.rcvif)) continue; } #ifndef INET6_BINDV6ONLY else { - if (IN6_IS_ADDR_V4MAPPED(dst6) - && (in6p->in6p_flags & IN6P_BINDV6ONLY)) + if (IN6_IS_ADDR_V4MAPPED(&dst6) && + (in6p->in6p_flags & IN6P_BINDV6ONLY)) continue; } #endif if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) { - if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, src6) - || in6p->in6p_fport != *sport) + if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, + &src6) || in6p->in6p_fport != sport) continue; } #ifndef INET6_BINDV6ONLY else { - if (IN6_IS_ADDR_V4MAPPED(src6) - && (in6p->in6p_flags & IN6P_BINDV6ONLY)) + if (IN6_IS_ADDR_V4MAPPED(&src6) && + (in6p->in6p_flags & IN6P_BINDV6ONLY)) continue; } #endif @@ -792,64 +815,17 @@ udp6_realinput(af, src, dst, m, off) (SO_REUSEPORT|SO_REUSEADDR)) == 0) break; } - -#if 0 - if (last == NULL) { - /* - * No matching pcb found; discard datagram. - * (No need to send an ICMP Port Unreachable - * for a broadcast or multicast datgram.) - */ - switch (af) { - case AF_INET: - udpstat.udps_noportbcast++; - break; - case AF_INET6: - udp6stat.udp6s_noportmcast++; - break; - } - goto bad; - } -#endif } else { /* * Locate pcb for datagram. */ - in6p = in6_pcblookup_connect(&udb6, src6, *sport, - dst6, *dport, 0); + in6p = in6_pcblookup_connect(&udb6, &src6, sport, + &dst6, dport, 0); if (in6p == 0) { ++udpstat.udps_pcbhashmiss; - in6p = in6_pcblookup_bind(&udb6, dst6, *dport, 0); - if (in6p == 0) { -#if 0 - struct mbuf *n; - n = m_copy(m, 0, M_COPYALL); - switch (af) { - case AF_INET: - if (m->m_flags & (M_BCAST | M_MCAST)) { - udpstat.udps_noportbcast++; - goto bad; - } - udpstat.udps_noport++; - if (n != NULL) - icmp_error(n, ICMP_UNREACH, - ICMP_UNREACH_PORT, 0, 0); - break; - case AF_INET6: - if (m->m_flags & M_MCAST) { - udp6stat.udp6s_noportmcast++; - goto bad; - } - udp6stat.udp6s_noport++; - if (n != NULL) - icmp6_error(n, ICMP6_DST_UNREACH, - ICMP6_DST_UNREACH_NOPORT, 0); - break; - } -#endif - + in6p = in6_pcblookup_bind(&udb6, &dst6, dport, 0); + if (in6p == 0) return rcvcnt; - } } udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket); @@ -925,7 +901,7 @@ udp_input(m, va_alist) */ len = ntohs((u_int16_t)uh->uh_ulen); if (ip->ip_len != iphlen + len) { - if (ip->ip_len < iphlen + len) { + if (ip->ip_len < iphlen + len || len < sizeof(struct udphdr)) { udpstat.udps_badlen++; goto bad; } @@ -941,13 +917,33 @@ udp_input(m, va_alist) * Checksum extended UDP header and data. */ if (uh->uh_sum) { - bzero(((struct ipovly *)ip)->ih_x1, - sizeof ((struct ipovly *)ip)->ih_x1); - ((struct ipovly *)ip)->ih_len = uh->uh_ulen; - if (in_cksum(m, len + sizeof (struct ip)) != 0) { - udpstat.udps_badsum++; - m_freem(m); - return; + switch (m->m_pkthdr.csum_flags & + ((m->m_pkthdr.rcvif->if_csum_flags & 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); + bzero(((struct ipovly *)ip)->ih_x1, + sizeof ((struct ipovly *)ip)->ih_x1); + ((struct ipovly *)ip)->ih_len = uh->uh_ulen; + if (in_cksum(m, len + sizeof (struct ip)) != 0) + goto badcsum; + break; } } @@ -982,7 +978,7 @@ udp_input(m, va_alist) iphlen += sizeof(struct udphdr); /* * KAME note: usually we drop udpiphdr from mbuf here. - * we need udpiphdr for iPsec processing so we do that later. + * we need udpiphdr for IPsec processing so we do that later. */ /* * Locate pcb(s) for datagram. @@ -1029,6 +1025,7 @@ udp_input(m, va_alist) m_freem(n); if (opts) m_freem(opts); + udpstat.udps_fullsock++; } else sorwakeup(last->inp_socket); opts = 0; @@ -1135,9 +1132,15 @@ bad: m_freem(m); if (opts) m_freem(opts); + return; + +badcsum: + udpstat.udps_badsum++; + m_freem(m); } #endif /*UDP6*/ +#ifdef INET /* * Notify a udp user of an asynchronous error; * just wake up so that he can collect error status. @@ -1231,30 +1234,38 @@ udp_output(m, va_alist) * and addresses and length put into network format. */ ui = mtod(m, struct udpiphdr *); - bzero(ui->ui_x1, sizeof ui->ui_x1); ui->ui_pr = IPPROTO_UDP; - ui->ui_len = htons((u_int16_t)len + sizeof (struct udphdr)); ui->ui_src = inp->inp_laddr; ui->ui_dst = inp->inp_faddr; ui->ui_sport = inp->inp_lport; ui->ui_dport = inp->inp_fport; - ui->ui_ulen = ui->ui_len; + ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr)); /* - * Stuff checksum and output datagram. + * Set up checksum and output datagram. */ - ui->ui_sum = 0; if (udpcksum) { - if ((ui->ui_sum = in_cksum(m, sizeof (struct udpiphdr) + len)) == 0) - ui->ui_sum = 0xffff; - } + /* + * XXX Cache pseudo-header checksum part for + * XXX "connected" UDP sockets. + */ + ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr, + ui->ui_dst.s_addr, htons((u_int16_t)len + + sizeof(struct udphdr) + IPPROTO_UDP)); + m->m_pkthdr.csum_flags = M_CSUM_UDPv4; + m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); + } else + ui->ui_sum = 0; ((struct ip *)ui)->ip_len = 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++; #ifdef IPSEC - ipsec_setsocket(m, inp->inp_socket); + if (ipsec_setsocket(m, inp->inp_socket) != 0) { + error = ENOBUFS; + goto release; + } #endif /*IPSEC*/ return (ip_output(m, inp->inp_options, &inp->inp_route, @@ -1287,6 +1298,7 @@ udp_usrreq(so, req, m, nam, control, p) (struct ifnet *)control, p)); if (req == PRU_PURGEIF) { + in_pcbpurgeif0(&udbtable, (struct ifnet *)control); in_purgeif((struct ifnet *)control); in_pcbpurgeif(&udbtable, (struct ifnet *)control); return (0); @@ -1324,13 +1336,6 @@ udp_usrreq(so, req, m, nam, control, p) break; inp = sotoinpcb(so); inp->inp_ip.ip_ttl = ip_defttl; -#ifdef IPSEC - error = ipsec_init_policy(so, &inp->inp_sp); - if (error != 0) { - in_pcbdetach(inp); - break; - } -#endif /*IPSEC*/ break; case PRU_DETACH: @@ -1473,3 +1478,4 @@ udp_sysctl(name, namelen, oldp, oldlenp, } /* NOTREACHED */ } +#endif