X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/1c79356b52d46aa6b508fb032f5ae709b1f2897b..b7266188b87f3620ec3f9f717e57194a7dd989fe:/bsd/netinet6/icmp6.c?ds=sidebyside diff --git a/bsd/netinet6/icmp6.c b/bsd/netinet6/icmp6.c index 1291a6ea6..02d19734f 100644 --- a/bsd/netinet6/icmp6.c +++ b/bsd/netinet6/icmp6.c @@ -1,4 +1,33 @@ -/* $KAME: icmp6.c,v 1.77 2000/04/13 11:31:39 itojun Exp $ */ +/* + * Copyright (c) 2008 Apple Inc. All rights reserved. + * + * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ + * + * This file contains Original Code and/or Modifications of Original Code + * as defined in and that are subject to the Apple Public Source License + * Version 2.0 (the 'License'). You may not use this file except in + * compliance with the License. The rights granted to you under the License + * may not be used to create, or enable the creation or redistribution of, + * unlawful or unlicensed copies of an Apple operating system, or to + * circumvent, violate, or enable the circumvention or violation of, any + * terms of an Apple operating system software license agreement. + * + * Please obtain a copy of the License at + * http://www.opensource.apple.com/apsl/ and read it before using this file. + * + * The Original Code and all software distributed under the License are + * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER + * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, + * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. + * Please see the License for the specific language governing rights and + * limitations under the License. + * + * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ + */ + +/* $FreeBSD: src/sys/netinet6/icmp6.c,v 1.6.2.6 2001/07/10 09:44:16 ume Exp $ */ +/* $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $ */ /* * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. @@ -64,15 +93,10 @@ * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 */ -#if (defined(__FreeBSD__) && __FreeBSD__ >= 3) || defined(__NetBSD__) -#include "opt_inet.h" -#ifdef __NetBSD__ /*XXX*/ -#include "opt_ipsec.h" -#endif -#endif #include #include +#include #include #include #include @@ -90,34 +114,22 @@ #include #include -#if defined(__OpenBSD__) || (defined(__bsdi__) && _BSDI_VERSION >= 199802) -#include -#include -#endif #include #include #include #include -#if !(defined(__FreeBSD__) && __FreeBSD__ >= 3) && !defined(__OpenBSD__) && !(defined(__bsdi__) && _BSDI_VERSION >= 199802) && !defined(__APPLE__) -#include -#else #include -#endif +#include #include #include #include -#ifdef __OpenBSD__ /*KAME IPSEC*/ -#undef IPSEC -#endif - #if IPSEC #include #include -#include -#endif -#include "faith.h" +extern int ipsec_bypass; +#endif #include @@ -125,54 +137,109 @@ extern struct domain inet6domain; extern struct ip6protosw inet6sw[]; extern struct ip6protosw *ip6_protox[]; +extern uint32_t rip_sendspace; +extern uint32_t rip_recvspace; + struct icmp6stat icmp6stat; -#if !(defined(__FreeBSD__) && __FreeBSD__ >= 3) && !defined(__APPLE__) -extern struct in6pcb rawin6pcb; -#else extern struct inpcbhead ripcb; -#endif -extern u_int icmp6errratelim; +extern int icmp6errppslim; +static int icmp6errpps_count = 0; +static struct timeval icmp6errppslim_last; extern int icmp6_nodeinfo; -#if defined(__NetBSD__) || defined(__OpenBSD__) -static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL; -extern int pmtu_expire; -#endif - -#ifndef HAVE_NRL_INPCB -static int icmp6_rip6_input __P((struct mbuf **, int)); -#endif -static void icmp6_mtudisc_update __P((struct in6_addr *, struct icmp6_hdr *, - struct mbuf *)); -static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int)); -static const char *icmp6_redirect_diag __P((struct in6_addr *, - struct in6_addr *, struct in6_addr *)); -static struct mbuf * ni6_input __P((struct mbuf *, int)); -static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *, - struct ifnet **)); -static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, - struct ifnet *, int)); -#if defined(__NetBSD__) || defined(__OpenBSD__) -static struct rtentry *icmp6_mtudisc_clone __P((struct sockaddr *)); -static void icmp6_mtudisc_timeout __P((struct rtentry *, struct rttimer *)); -#endif +extern struct inpcbinfo ripcbinfo; +extern lck_mtx_t *ip6_mutex; +extern lck_mtx_t *nd6_mutex; +extern lck_mtx_t *inet6_domain_mutex; + +static void icmp6_errcount(struct icmp6errstat *, int, int); +static int icmp6_rip6_input(struct mbuf **, int); +static int icmp6_ratelimit(const struct in6_addr *, const int, const int); +static const char *icmp6_redirect_diag(struct in6_addr *, + struct in6_addr *, struct in6_addr *); +#ifndef HAVE_PPSRATECHECK +static int ppsratecheck(struct timeval *, int *, int); +#endif +static struct mbuf *ni6_input(struct mbuf *, int); +static struct mbuf *ni6_nametodns(const char *, int, int); +static int ni6_dnsmatch(const char *, int, const char *, int); +static int ni6_addrs(struct icmp6_nodeinfo *, + struct ifnet **, char *); +static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, + struct ifnet *, int); +static int icmp6_notify_error(struct mbuf *, int, int, int); #ifdef COMPAT_RFC1885 +/* + * XXX: Compiled out for now, but if enabled we must use a lock for accesses, + * or easier, define it locally inside icmp6_reflect() and don't cache. + */ static struct route_in6 icmp6_reflect_rt; #endif -static struct timeval icmp6_nextsend = {0, 0}; -#if MIP6 -int (*mip6_icmp6_input_hook)(struct mbuf *m, int off) = NULL; -#endif /* MIP6 */ void icmp6_init() { mld6_init(); -#if defined(__NetBSD__) || defined(__OpenBSD__) - icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire); -#endif +} + +static void +icmp6_errcount(stat, type, code) + struct icmp6errstat *stat; + int type, code; +{ + switch (type) { + case ICMP6_DST_UNREACH: + switch (code) { + case ICMP6_DST_UNREACH_NOROUTE: + stat->icp6errs_dst_unreach_noroute++; + return; + case ICMP6_DST_UNREACH_ADMIN: + stat->icp6errs_dst_unreach_admin++; + return; + case ICMP6_DST_UNREACH_BEYONDSCOPE: + stat->icp6errs_dst_unreach_beyondscope++; + return; + case ICMP6_DST_UNREACH_ADDR: + stat->icp6errs_dst_unreach_addr++; + return; + case ICMP6_DST_UNREACH_NOPORT: + stat->icp6errs_dst_unreach_noport++; + return; + } + break; + case ICMP6_PACKET_TOO_BIG: + stat->icp6errs_packet_too_big++; + return; + case ICMP6_TIME_EXCEEDED: + switch (code) { + case ICMP6_TIME_EXCEED_TRANSIT: + stat->icp6errs_time_exceed_transit++; + return; + case ICMP6_TIME_EXCEED_REASSEMBLY: + stat->icp6errs_time_exceed_reassembly++; + return; + } + break; + case ICMP6_PARAM_PROB: + switch (code) { + case ICMP6_PARAMPROB_HEADER: + stat->icp6errs_paramprob_header++; + return; + case ICMP6_PARAMPROB_NEXTHEADER: + stat->icp6errs_paramprob_nextheader++; + return; + case ICMP6_PARAMPROB_OPTION: + stat->icp6errs_paramprob_option++; + return; + } + break; + case ND_REDIRECT: + stat->icp6errs_redirect++; + return; + } + stat->icp6errs_unknown++; } /* @@ -191,6 +258,10 @@ icmp6_error(m, type, code, param) icmp6stat.icp6s_error++; + lck_mtx_assert(ip6_mutex, LCK_MTX_ASSERT_NOTOWNED); + /* count per-type-code statistics */ + icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code); + #ifdef M_DECRYPTED /*not openbsd*/ if (m->m_flags & M_DECRYPTED) { icmp6stat.icp6s_canterror++; @@ -199,7 +270,7 @@ icmp6_error(m, type, code, param) #endif #ifndef PULLDOWN_TEST - IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), ); + IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), return); #else if (m->m_len < sizeof(struct ip6_hdr)) { m = m_pullup(m, sizeof(struct ip6_hdr)); @@ -236,7 +307,7 @@ icmp6_error(m, type, code, param) struct icmp6_hdr *icp; #ifndef PULLDOWN_TEST - IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), ); + IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), return); icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); #else IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, @@ -282,7 +353,7 @@ icmp6_error(m, type, code, param) if (m && m->m_len < preplen) m = m_pullup(m, preplen); if (m == NULL) { - printf("ENOBUFS in icmp6_error %d\n", __LINE__); + nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); return; } @@ -300,8 +371,17 @@ icmp6_error(m, type, code, param) icmp6->icmp6_code = code; icmp6->icmp6_pptr = htonl((u_int32_t)param); + /* + * icmp6_reflect() is designed to be in the input path. + * icmp6_error() can be called from both input and outut path, + * and if we are in output path rcvif could contain bogus value. + * clear m->m_pkthdr.rcvif for safety, we should have enough scope + * information in ip header (nip6). + */ + m->m_pkthdr.rcvif = NULL; + icmp6stat.icp6s_outhist[type]++; - icmp6_reflect(m, sizeof(struct ip6_hdr)); /*header order: IPv6 - ICMPv6*/ + icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ return; @@ -316,9 +396,9 @@ icmp6_error(m, type, code, param) * Process a received ICMP6 message. */ int -icmp6_input(mp, offp, proto) +icmp6_input(mp, offp) struct mbuf **mp; - int *offp, proto; + int *offp; { struct mbuf *m = *mp, *n; struct ip6_hdr *ip6, *nip6; @@ -326,11 +406,10 @@ icmp6_input(mp, offp, proto) int off = *offp; int icmp6len = m->m_pkthdr.len - *offp; int code, sum, noff; - struct sockaddr_in6 icmp6src; #ifndef PULLDOWN_TEST - IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE); - /* m might change if M_LOOP. So, call mtod after this */ + IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), return IPPROTO_DONE); + /* m might change if M_LOOP. So, call mtod after this */ #endif /* @@ -359,17 +438,15 @@ icmp6_input(mp, offp, proto) code = icmp6->icmp6_code; if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { - log(LOG_ERR, + nd6log((LOG_ERR, "ICMP6 checksum error(%d|%x) %s\n", - icmp6->icmp6_type, - sum, - ip6_sprintf(&ip6->ip6_src)); + icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src))); icmp6stat.icp6s_checksum++; goto freeit; } #if defined(NFAITH) && 0 < NFAITH - if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { + if (faithprefix(&ip6->ip6_dst)) { /* * Deliver very specific ICMP6 type only. * This is important to deilver TOOBIG. Otherwise PMTUD @@ -386,34 +463,13 @@ icmp6_input(mp, offp, proto) } #endif -#if IPSEC - /* drop it if it does not match the default policy */ - if (ipsec6_in_reject(m, NULL)) { - ipsecstat.in_polvio++; - goto freeit; - } -#endif - icmp6stat.icp6s_inhist[icmp6->icmp6_type]++; icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK) icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); -#if MIP6 - /* - * Mobile IPv6 - * - * Check for ICMP errors and modifications and extensions to Router - * Advertisement. - */ - if (mip6_icmp6_input_hook) { - if ((*mip6_icmp6_input_hook)(m, off) != 0) - goto freeit; - } -#endif /* MIP6 */ - + switch (icmp6->icmp6_type) { - case ICMP6_DST_UNREACH: icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); switch (code) { @@ -443,6 +499,7 @@ icmp6_input(mp, offp, proto) default: goto badcode; } + goto deliver; break; @@ -493,8 +550,15 @@ icmp6_input(mp, offp, proto) icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); if (code != 0) goto badcode; + + if (icmp6_ratelimit(&ip6->ip6_dst, icmp6->icmp6_type, code)) { + icmp6stat.icp6s_toofreq++; + goto freeit; + } + if ((n = m_copy(m, 0, M_COPYALL)) == NULL) { /* Give up remote */ + goto rate_limit_checked; break; } if ((n->m_flags & M_EXT) != 0 @@ -507,14 +571,12 @@ icmp6_input(mp, offp, proto) * always copy the length we specified. */ if (maxlen >= MCLBYTES) { -#if DIAGNOSTIC - printf("MCLBYTES too small\n"); -#endif /* Give up remote */ m_freem(n0); + goto rate_limit_checked; break; } - MGETHDR(n, M_DONTWAIT, n0->m_type); + MGETHDR(n, M_DONTWAIT, n0->m_type); /* MAC-OK */ if (n && maxlen >= MHLEN) { MCLGET(n, M_DONTWAIT); if ((n->m_flags & M_EXT) == 0) { @@ -525,6 +587,7 @@ icmp6_input(mp, offp, proto) if (n == NULL) { /* Give up remote */ m_freem(n0); + goto rate_limit_checked; break; } M_COPY_PKTHDR(n, n0); @@ -548,7 +611,8 @@ icmp6_input(mp, offp, proto) n0->m_flags &= ~M_PKTHDR; } else { nip6 = mtod(n, struct ip6_hdr *); - nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off); + IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off, + sizeof(*nicmp6)); noff = off; } nicmp6->icmp6_type = ICMP6_ECHO_REPLY; @@ -558,6 +622,7 @@ icmp6_input(mp, offp, proto) icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++; icmp6_reflect(n, noff); } + goto rate_limit_checked; break; case ICMP6_ECHO_REPLY: @@ -574,6 +639,12 @@ icmp6_input(mp, offp, proto) icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery); else icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport); + + if (icmp6_ratelimit(&ip6->ip6_dst, icmp6->icmp6_type, code)) { + icmp6stat.icp6s_toofreq++; + goto freeit; + } + if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { /* give up local */ mld6_input(m, off); @@ -582,6 +653,7 @@ icmp6_input(mp, offp, proto) } mld6_input(n, off); /* m stays. */ + goto rate_limit_checked; break; case MLD6_LISTENER_DONE: @@ -596,88 +668,37 @@ icmp6_input(mp, offp, proto) /* XXX: per-interface statistics? */ break; /* just pass it to applications */ - case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ - { - enum { WRU, FQDN } mode; - - if (code != 0) - goto badcode; + case ICMP6_NI_QUERY: if (!icmp6_nodeinfo) break; - if (icmp6len == sizeof(struct icmp6_hdr) + 4) - mode = WRU; - else if (icmp6len >= sizeof(struct icmp6_hdr) + 8) /* XXX */ - mode = FQDN; - else + /* By RFC 4620 refuse to answer queries from global scope addresses */ + if ((icmp6_nodeinfo & 8) != 8 && in6_addrscope(&ip6->ip6_src) == IPV6_ADDR_SCOPE_GLOBAL) + break; + + if (icmp6len < sizeof(struct icmp6_nodeinfo)) goto badlen; -#if defined( __FreeBSD__) || defined (__APPLE__) -#define hostnamelen strlen(hostname) -#endif - if (mode == FQDN) { #ifndef PULLDOWN_TEST - IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo), - IPPROTO_DONE); + IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo), + return IPPROTO_DONE); #endif - n = m_copy(m, 0, M_COPYALL); - if (n) - n = ni6_input(n, off); - if (n) - noff = sizeof(struct ip6_hdr); - } else { - u_char *p; - int maxlen, maxhlen; - - maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4; - if (maxlen >= MCLBYTES) { -#if DIAGNOSTIC - printf("MCLBYTES too small\n"); -#endif - /* Give up remote */ - break; - } - MGETHDR(n, M_DONTWAIT, m->m_type); - if (n && maxlen > MHLEN) { - MCLGET(n, M_DONTWAIT); - if ((n->m_flags & M_EXT) == 0) { - m_free(n); - n = NULL; - } - } - if (n == NULL) { - /* Give up remote */ - break; - } - n->m_len = 0; - maxhlen = M_TRAILINGSPACE(n) - maxlen; - if (maxhlen > hostnamelen) - maxhlen = hostnamelen; - /* - * Copy IPv6 and ICMPv6 only. - */ - nip6 = mtod(n, struct ip6_hdr *); - bcopy(ip6, nip6, sizeof(struct ip6_hdr)); - nicmp6 = (struct icmp6_hdr *)(nip6 + 1); - bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); - p = (u_char *)(nicmp6 + 1); - bzero(p, 4); - bcopy(hostname, p + 4, maxhlen); - noff = sizeof(struct ip6_hdr); - M_COPY_PKTHDR(n, m); /* just for recvif */ - n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + - sizeof(struct icmp6_hdr) + 4 + maxhlen; - nicmp6->icmp6_type = ICMP6_WRUREPLY; - nicmp6->icmp6_code = 0; + if (icmp6_ratelimit(&ip6->ip6_dst, icmp6->icmp6_type, code)) { + icmp6stat.icp6s_toofreq++; + goto freeit; } -#undef hostnamelen + + n = m_copy(m, 0, M_COPYALL); + if (n) + n = ni6_input(n, off); if (n) { + noff = sizeof(struct ip6_hdr); icmp6stat.icp6s_reflect++; icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++; icmp6_reflect(n, noff); } + goto rate_limit_checked; break; - } case ICMP6_WRUREPLY: if (code != 0) @@ -690,6 +711,12 @@ icmp6_input(mp, offp, proto) goto badcode; if (icmp6len < sizeof(struct nd_router_solicit)) goto badlen; + + if (icmp6_ratelimit(&ip6->ip6_dst, icmp6->icmp6_type, code)) { + icmp6stat.icp6s_toofreq++; + goto freeit; + } + if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { /* give up local */ nd6_rs_input(m, off, icmp6len); @@ -698,6 +725,7 @@ icmp6_input(mp, offp, proto) } nd6_rs_input(n, off, icmp6len); /* m stays. */ + goto rate_limit_checked; break; case ND_ROUTER_ADVERT: @@ -706,6 +734,12 @@ icmp6_input(mp, offp, proto) goto badcode; if (icmp6len < sizeof(struct nd_router_advert)) goto badlen; + + if (icmp6_ratelimit(&ip6->ip6_dst, icmp6->icmp6_type, code)) { + icmp6stat.icp6s_toofreq++; + goto freeit; + } + if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { /* give up local */ nd6_ra_input(m, off, icmp6len); @@ -714,6 +748,7 @@ icmp6_input(mp, offp, proto) } nd6_ra_input(n, off, icmp6len); /* m stays. */ + goto rate_limit_checked; break; case ND_NEIGHBOR_SOLICIT: @@ -722,6 +757,12 @@ icmp6_input(mp, offp, proto) goto badcode; if (icmp6len < sizeof(struct nd_neighbor_solicit)) goto badlen; + + if (icmp6_ratelimit(&ip6->ip6_dst, icmp6->icmp6_type, code)) { + icmp6stat.icp6s_toofreq++; + goto freeit; + } + if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { /* give up local */ nd6_ns_input(m, off, icmp6len); @@ -730,6 +771,7 @@ icmp6_input(mp, offp, proto) } nd6_ns_input(n, off, icmp6len); /* m stays. */ + goto rate_limit_checked; break; case ND_NEIGHBOR_ADVERT: @@ -738,6 +780,12 @@ icmp6_input(mp, offp, proto) goto badcode; if (icmp6len < sizeof(struct nd_neighbor_advert)) goto badlen; + + if (icmp6_ratelimit(&ip6->ip6_dst, icmp6->icmp6_type, code)) { + icmp6stat.icp6s_toofreq++; + goto freeit; + } + if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { /* give up local */ nd6_na_input(m, off, icmp6len); @@ -746,6 +794,7 @@ icmp6_input(mp, offp, proto) } nd6_na_input(n, off, icmp6len); /* m stays. */ + goto rate_limit_checked; break; case ND_REDIRECT: @@ -754,6 +803,12 @@ icmp6_input(mp, offp, proto) goto badcode; if (icmp6len < sizeof(struct nd_redirect)) goto badlen; + + if (icmp6_ratelimit(&ip6->ip6_dst, icmp6->icmp6_type, code)) { + icmp6stat.icp6s_toofreq++; + goto freeit; + } + if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { /* give up local */ icmp6_redirect_input(m, off); @@ -762,6 +817,7 @@ icmp6_input(mp, offp, proto) } icmp6_redirect_input(n, off); /* m stays. */ + goto rate_limit_checked; break; case ICMP6_ROUTER_RENUMBERING: @@ -773,45 +829,93 @@ icmp6_input(mp, offp, proto) break; default: - printf("icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", - icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), - ip6_sprintf(&ip6->ip6_dst), - m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0); + if (icmp6_ratelimit(&ip6->ip6_dst, icmp6->icmp6_type, code)) { + icmp6stat.icp6s_toofreq++; + goto freeit; + } + + nd6log((LOG_DEBUG, + "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", + icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), + ip6_sprintf(&ip6->ip6_dst), + m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { /* ICMPv6 error: MUST deliver it by spec... */ code = PRC_NCMDS; /* deliver */ } else { /* ICMPv6 informational: MUST not deliver */ + goto rate_limit_checked; break; } deliver: - if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { - icmp6stat.icp6s_tooshort++; + if (icmp6_ratelimit(&ip6->ip6_dst, icmp6->icmp6_type, code)) { + icmp6stat.icp6s_toofreq++; goto freeit; } + + if (icmp6_notify_error(m, off, icmp6len, code)) { + /* In this case, m should've been freed. */ + return(IPPROTO_DONE); + } + break; + + badcode: + icmp6stat.icp6s_badcode++; + break; + + badlen: + icmp6stat.icp6s_badlen++; + break; + } + + if (icmp6_ratelimit(&ip6->ip6_dst, icmp6->icmp6_type, code)) { + icmp6stat.icp6s_toofreq++; + goto freeit; + } +rate_limit_checked: + /* deliver the packet to appropriate sockets */ + icmp6_rip6_input(&m, *offp); + + return IPPROTO_DONE; + + freeit: + m_freem(m); + return IPPROTO_DONE; +} + +static int +icmp6_notify_error(m, off, icmp6len, code) + struct mbuf *m; + int off, icmp6len, code; +{ + struct icmp6_hdr *icmp6; + struct ip6_hdr *eip6; + u_int32_t notifymtu; + struct sockaddr_in6 icmp6src, icmp6dst; + + if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { + icmp6stat.icp6s_tooshort++; + goto freeit; + } #ifndef PULLDOWN_TEST - IP6_EXTHDR_CHECK(m, off, - sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), - IPPROTO_DONE); - icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); + IP6_EXTHDR_CHECK(m, off, + sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), + return -1); + icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); #else - IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, - sizeof(*icmp6) + sizeof(struct ip6_hdr)); - if (icmp6 == NULL) { - icmp6stat.icp6s_tooshort++; - return IPPROTO_DONE; - } + IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, + sizeof(*icmp6) + sizeof(struct ip6_hdr)); + if (icmp6 == NULL) { + icmp6stat.icp6s_tooshort++; + return(-1); + } #endif - bzero(&icmp6src, sizeof(icmp6src)); - icmp6src.sin6_len = sizeof(struct sockaddr_in6); - icmp6src.sin6_family = AF_INET6; - icmp6src.sin6_addr = ((struct ip6_hdr *)(icmp6 + 1))->ip6_dst; + eip6 = (struct ip6_hdr *)(icmp6 + 1); - /* Detect the upper level protocol */ - { - void (*ctlfunc) __P((int, struct sockaddr *, void *)); - struct ip6_hdr *eip6 = (struct ip6_hdr *)(icmp6 + 1); + /* Detect the upper level protocol */ + { + void (*ctlfunc)(int, struct sockaddr *, void *); u_int8_t nxt = eip6->ip6_nxt; int eoff = off + sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr); @@ -823,25 +927,25 @@ icmp6_input(mp, offp, proto) struct ip6_rthdr0 *rth0; int rthlen; - while (1) { /* XXX: should avoid inf. loop explicitly? */ + while (1) { /* XXX: should avoid infinite loop explicitly? */ struct ip6_ext *eh; - switch(nxt) { + switch (nxt) { case IPPROTO_HOPOPTS: case IPPROTO_DSTOPTS: case IPPROTO_AH: #ifndef PULLDOWN_TEST IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(struct ip6_ext), - IPPROTO_DONE); + return -1); eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff); #else IP6_EXTHDR_GET(eh, struct ip6_ext *, m, - eoff, sizeof(*eh)); + eoff, sizeof(*eh)); if (eh == NULL) { icmp6stat.icp6s_tooshort++; - return IPPROTO_DONE; + return(-1); } #endif @@ -862,15 +966,15 @@ icmp6_input(mp, offp, proto) */ #ifndef PULLDOWN_TEST IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), - IPPROTO_DONE); + return -1); rth = (struct ip6_rthdr *)(mtod(m, caddr_t) + eoff); #else IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, - eoff, sizeof(*rth)); + eoff, sizeof(*rth)); if (rth == NULL) { icmp6stat.icp6s_tooshort++; - return IPPROTO_DONE; + return(-1); } #endif rthlen = (rth->ip6r_len + 1) << 3; @@ -888,7 +992,7 @@ icmp6_input(mp, offp, proto) #ifndef PULLDOWN_TEST IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, - IPPROTO_DONE); + return -1); rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff); #else IP6_EXTHDR_GET(rth0, @@ -896,7 +1000,7 @@ icmp6_input(mp, offp, proto) eoff, rthlen); if (rth0 == NULL) { icmp6stat.icp6s_tooshort++; - return IPPROTO_DONE; + return(-1); } #endif /* just ignore a bogus header */ @@ -911,15 +1015,15 @@ icmp6_input(mp, offp, proto) #ifndef PULLDOWN_TEST IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(struct ip6_frag), - IPPROTO_DONE); + return -1); fh = (struct ip6_frag *)(mtod(m, caddr_t) + eoff); #else IP6_EXTHDR_GET(fh, struct ip6_frag *, m, - eoff, sizeof(*fh)); + eoff, sizeof(*fh)); if (fh == NULL) { icmp6stat.icp6s_tooshort++; - return IPPROTO_DONE; + return(-1); } #endif /* @@ -937,7 +1041,7 @@ icmp6_input(mp, offp, proto) default: /* * This case includes ESP and the No Next - * Header. In such cases going to the notify + * Header. In such cases going to the notify * label does not have any meaning * (i.e. ctlfunc will be NULL), but we go * anyway since we might have to update @@ -946,120 +1050,154 @@ icmp6_input(mp, offp, proto) goto notify; } } - notify: + notify: #ifndef PULLDOWN_TEST icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); #else IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, - sizeof(*icmp6) + sizeof(struct ip6_hdr)); + sizeof(*icmp6) + sizeof(struct ip6_hdr)); if (icmp6 == NULL) { icmp6stat.icp6s_tooshort++; - return IPPROTO_DONE; + return(-1); } #endif + + eip6 = (struct ip6_hdr *)(icmp6 + 1); + bzero(&icmp6dst, sizeof(icmp6dst)); + icmp6dst.sin6_len = sizeof(struct sockaddr_in6); + icmp6dst.sin6_family = AF_INET6; + if (finaldst == NULL) + icmp6dst.sin6_addr = eip6->ip6_dst; + else + icmp6dst.sin6_addr = *finaldst; + icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, + &icmp6dst.sin6_addr); +#ifndef SCOPEDROUTING + if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst, + NULL, NULL)) { + /* should be impossbile */ + nd6log((LOG_DEBUG, + "icmp6_notify_error: in6_embedscope failed\n")); + goto freeit; + } +#endif + + /* + * retrieve parameters from the inner IPv6 header, and convert + * them into sockaddr structures. + */ + bzero(&icmp6src, sizeof(icmp6src)); + icmp6src.sin6_len = sizeof(struct sockaddr_in6); + icmp6src.sin6_family = AF_INET6; + icmp6src.sin6_addr = eip6->ip6_src; + icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, + &icmp6src.sin6_addr); +#ifndef SCOPEDROUTING + if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src, + NULL, NULL)) { + /* should be impossbile */ + nd6log((LOG_DEBUG, + "icmp6_notify_error: in6_embedscope failed\n")); + goto freeit; + } +#endif + icmp6src.sin6_flowinfo = + (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); + + if (finaldst == NULL) + finaldst = &eip6->ip6_dst; + ip6cp.ip6c_m = m; + ip6cp.ip6c_icmp6 = icmp6; + ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); + ip6cp.ip6c_off = eoff; + ip6cp.ip6c_finaldst = finaldst; + ip6cp.ip6c_src = &icmp6src; + ip6cp.ip6c_nxt = nxt; + if (icmp6type == ICMP6_PACKET_TOO_BIG) { - if (finaldst == NULL) - finaldst = &((struct ip6_hdr *)(icmp6 + 1))->ip6_dst; - icmp6_mtudisc_update(finaldst, icmp6, m); + notifymtu = ntohl(icmp6->icmp6_mtu); + ip6cp.ip6c_cmdarg = (void *)¬ifymtu; + icmp6_mtudisc_update(&ip6cp, 1); /*XXX*/ } - ctlfunc = (void (*) __P((int, struct sockaddr *, void *))) + ctlfunc = (void (*)(int, struct sockaddr *, void *)) (ip6_protox[nxt]->pr_ctlinput); if (ctlfunc) { - ip6cp.ip6c_m = m; - ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); - ip6cp.ip6c_off = eoff; - (*ctlfunc)(code, (struct sockaddr *)&icmp6src, &ip6cp); + (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, + &ip6cp); } - } - break; - - badcode: - icmp6stat.icp6s_badcode++; - break; - - badlen: - icmp6stat.icp6s_badlen++; - break; } + return(0); -#if HAVE_NRL_INPCB - rip6_input(&m, offp, IPPROTO_ICMPV6); -#else - icmp6_rip6_input(&m, *offp); -#endif - return IPPROTO_DONE; - - freeit: +freeit: m_freem(m); - return IPPROTO_DONE; + return(-1); } -static void -icmp6_mtudisc_update(dst, icmp6, m) - struct in6_addr *dst; - struct icmp6_hdr *icmp6;/* we can assume the validity of the pointer */ - struct mbuf *m; /* currently unused but added for scoped addrs */ +void +icmp6_mtudisc_update(ip6cp, validated) + struct ip6ctlparam *ip6cp; + int validated; { + struct in6_addr *dst = ip6cp->ip6c_finaldst; + struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; + struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ u_int mtu = ntohl(icmp6->icmp6_mtu); struct rtentry *rt = NULL; struct sockaddr_in6 sin6; -#ifdef __bsdi__ - struct route_in6 ro6; -#endif + + if (!validated) + return; + /* + * In case the suggested mtu is less than IPV6_MMTU, we + * only need to remember that it was for above mentioned + * "alwaysfrag" case. + * Try to be as close to the spec as possible. + */ + if (mtu < IPV6_MMTU) + mtu = IPV6_MMTU - 8; bzero(&sin6, sizeof(sin6)); sin6.sin6_family = PF_INET6; sin6.sin6_len = sizeof(struct sockaddr_in6); sin6.sin6_addr = *dst; - /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */ -#if defined(__NetBSD__) || defined(__OpenBSD__) - rt = rtalloc1((struct sockaddr *)&sin6, 1); /*clone*/ - if (!rt || (rt->rt_flags & RTF_HOST) == 0) { - if (rt) - RTFREE(rt); - rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); + /* XXX normally, this won't happen */ + if (IN6_IS_ADDR_LINKLOCAL(dst)) { + sin6.sin6_addr.s6_addr16[1] = + htons(m->m_pkthdr.rcvif->if_index); } -#else -#if defined(__FreeBSD__) || defined (__APPLE__) - rt = rtalloc1((struct sockaddr *)&sin6, 0, - RTF_CLONING | RTF_PRCLONING); -#else -#ifdef __bsdi__ - bcopy(&sin6, &ro6.ro_dst, sizeof(struct sockaddr_in6)); - ro6.ro_rt = 0; - rtcalloc((struct route *)&ro6); - rt = ro6.ro_rt; -#else -#error no case for this particular operating system -#endif -#endif -#endif - - if (rt && (rt->rt_flags & RTF_HOST) - && !(rt->rt_rmx.rmx_locks & RTV_MTU)) { - if (mtu < IPV6_MMTU) { - /* xxx */ - rt->rt_rmx.rmx_locks |= RTV_MTU; - } else if (mtu < rt->rt_ifp->if_mtu && - rt->rt_rmx.rmx_mtu > mtu) { - rt->rt_rmx.rmx_mtu = mtu; + /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */ + rt = rtalloc1((struct sockaddr *)&sin6, 0, RTF_CLONING | RTF_PRCLONING); + if (rt != NULL) { + RT_LOCK(rt); + if ((rt->rt_flags & RTF_HOST) && + !(rt->rt_rmx.rmx_locks & RTV_MTU)) { + if (mtu < IPV6_MMTU) { + /* xxx */ + rt->rt_rmx.rmx_locks |= RTV_MTU; + } else if (mtu < rt->rt_ifp->if_mtu && + rt->rt_rmx.rmx_mtu > mtu) { + icmp6stat.icp6s_pmtuchg++; + rt->rt_rmx.rmx_mtu = mtu; + } } + RT_UNLOCK(rt); + rtfree(rt); } - if (rt) - RTFREE(rt); } /* - * Process a Node Information Query + * Process a Node Information Query packet, based on + * draft-ietf-ipngwg-icmp-name-lookups-07. + * + * Spec incompatibilities: + * - IPv6 Subject address handling + * - IPv4 Subject address handling support missing + * - Proxy reply (answer even if it's not for me) + * - joins NI group address at in6_ifattach() time only, does not cope + * with hostname changes by sethostname(3) */ -#if defined(__FreeBSD__) || defined (__APPLE__) #define hostnamelen strlen(hostname) -#endif -#ifndef offsetof /* XXX */ -#define offsetof(type, member) ((size_t)(&((type *)0)->member)) -#endif - static struct mbuf * ni6_input(m, off) struct mbuf *m; @@ -1068,11 +1206,19 @@ ni6_input(m, off) struct icmp6_nodeinfo *ni6, *nni6; struct mbuf *n = NULL; u_int16_t qtype; + int subjlen; int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); struct ni_reply_fqdn *fqdn; int addrs; /* for NI_QTYPE_NODEADDR */ struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ + struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */ + struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */ + struct ip6_hdr *ip6; + int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ + char *subj = NULL; + struct in6_ifaddr *ia6 = NULL; + ip6 = mtod(m, struct ip6_hdr *); #ifndef PULLDOWN_TEST ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off); #else @@ -1082,54 +1228,212 @@ ni6_input(m, off) return NULL; } #endif + + /* + * Validate IPv6 destination address. + * + * The Responder must discard the Query without further processing + * unless it is one of the Responder's unicast or anycast addresses, or + * a link-local scope multicast address which the Responder has joined. + * [icmp-name-lookups-07, Section 4.] + */ + bzero(&sin6, sizeof(sin6)); + sin6.sin6_family = AF_INET6; + sin6.sin6_len = sizeof(struct sockaddr_in6); + bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr)); + /* XXX scopeid */ + if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) { + /* unicast/anycast, fine */ + if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && + (icmp6_nodeinfo & 4) == 0) { + ifafree(&ia6->ia_ifa); + ia6 = NULL; + nd6log((LOG_DEBUG, "ni6_input: ignore node info to " + "a temporary address in %s:%d", + __FILE__, __LINE__)); + goto bad; + } + ifafree(&ia6->ia_ifa); + ia6 = NULL; + } else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) + ; /* link-local multicast, fine */ + else + goto bad; + + /* validate query Subject field. */ qtype = ntohs(ni6->ni_qtype); + subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); + switch (qtype) { + case NI_QTYPE_NOOP: + case NI_QTYPE_SUPTYPES: + /* 07 draft */ + if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) + break; + /* FALLTHROUGH */ + case NI_QTYPE_FQDN: + case NI_QTYPE_NODEADDR: + switch (ni6->ni_code) { + case ICMP6_NI_SUBJ_IPV6: +#if ICMP6_NI_SUBJ_IPV6 != 0 + case 0: +#endif + /* + * backward compatibility - try to accept 03 draft + * format, where no Subject is present. + */ + if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && + subjlen == 0) { + oldfqdn++; + break; + } +#if ICMP6_NI_SUBJ_IPV6 != 0 + if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) + goto bad; +#endif - switch(qtype) { - case NI_QTYPE_NOOP: - break; /* no reply data */ - case NI_QTYPE_SUPTYPES: - goto bad; /* xxx: to be implemented */ - break; - case NI_QTYPE_FQDN: - replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_name) + - hostnamelen; - break; - case NI_QTYPE_NODEADDR: - addrs = ni6_addrs(ni6, m, &ifp); - if ((replylen += addrs * sizeof(struct in6_addr)) > MCLBYTES) - replylen = MCLBYTES; /* XXX: we'll truncate later */ - - break; - default: - /* - * XXX: We must return a reply with the ICMP6 code - * `unknown Qtype' in this case. However we regard the case - * as an FQDN query for backward compatibility. - * Older versions set a random value to this field, - * so it rarely varies in the defined qtypes. - * But the mechanism is not reliable... - * maybe we should obsolete older versions. - */ - qtype = NI_QTYPE_FQDN; - replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_name) + - hostnamelen; - break; - } - - /* allocate a mbuf to reply. */ - MGETHDR(n, M_DONTWAIT, m->m_type); + if (subjlen != sizeof(sin6.sin6_addr)) + goto bad; + + /* + * Validate Subject address. + * + * Not sure what exactly "address belongs to the node" + * means in the spec, is it just unicast, or what? + * + * At this moment we consider Subject address as + * "belong to the node" if the Subject address equals + * to the IPv6 destination address; validation for + * IPv6 destination address should have done enough + * check for us. + * + * We do not do proxy at this moment. + */ + /* m_pulldown instead of copy? */ + m_copydata(m, off + sizeof(struct icmp6_nodeinfo), + subjlen, (caddr_t)&sin6.sin6_addr); + sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, + &sin6.sin6_addr); +#ifndef SCOPEDROUTING + in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL); +#endif + bzero(&sin6_d, sizeof(sin6_d)); + sin6_d.sin6_family = AF_INET6; /* not used, actually */ + sin6_d.sin6_len = sizeof(sin6_d); /* ditto */ + sin6_d.sin6_addr = ip6->ip6_dst; + sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, + &ip6->ip6_dst); +#ifndef SCOPEDROUTING + in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL); +#endif + subj = (char *)&sin6; + if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d)) + break; + + /* + * XXX if we are to allow other cases, we should really + * be careful about scope here. + * basically, we should disallow queries toward IPv6 + * destination X with subject Y, if scope(X) > scope(Y). + * if we allow scope(X) > scope(Y), it will result in + * information leakage across scope boundary. + */ + goto bad; + + case ICMP6_NI_SUBJ_FQDN: + /* + * Validate Subject name with gethostname(3). + * + * The behavior may need some debate, since: + * - we are not sure if the node has FQDN as + * hostname (returned by gethostname(3)). + * - the code does wildcard match for truncated names. + * however, we are not sure if we want to perform + * wildcard match, if gethostname(3) side has + * truncated hostname. + */ + n = ni6_nametodns(hostname, hostnamelen, 0); + if (!n || n->m_next || n->m_len == 0) + goto bad; + IP6_EXTHDR_GET(subj, char *, m, + off + sizeof(struct icmp6_nodeinfo), subjlen); + if (subj == NULL) + goto bad; + if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), + n->m_len)) { + goto bad; + } + m_freem(n); + n = NULL; + break; + + case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ + default: + goto bad; + } + break; + } + + /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ + switch (qtype) { + case NI_QTYPE_FQDN: + if ((icmp6_nodeinfo & 1) == 0) + goto bad; + break; + case NI_QTYPE_NODEADDR: + if ((icmp6_nodeinfo & 2) == 0) + goto bad; + break; + } + + /* guess reply length */ + switch (qtype) { + case NI_QTYPE_NOOP: + break; /* no reply data */ + case NI_QTYPE_SUPTYPES: + replylen += sizeof(u_int32_t); + break; + case NI_QTYPE_FQDN: + /* XXX will append an mbuf */ + replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); + break; + case NI_QTYPE_NODEADDR: + addrs = ni6_addrs(ni6, &ifp, subj); + if ((replylen += addrs * (sizeof(struct in6_addr) + + sizeof(u_int32_t))) > MCLBYTES) + replylen = MCLBYTES; /* XXX: will truncate pkt later */ + break; + default: + /* + * XXX: We must return a reply with the ICMP6 code + * `unknown Qtype' in this case. However we regard the case + * as an FQDN query for backward compatibility. + * Older versions set a random value to this field, + * so it rarely varies in the defined qtypes. + * But the mechanism is not reliable... + * maybe we should obsolete older versions. + */ + qtype = NI_QTYPE_FQDN; + /* XXX will append an mbuf */ + replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); + oldfqdn++; + break; + } + + /* allocate an mbuf to reply. */ + MGETHDR(n, M_DONTWAIT, m->m_type); /* MAC-OK */ if (n == NULL) { m_freem(m); return(NULL); } M_COPY_PKTHDR(n, m); /* just for recvif */ if (replylen > MHLEN) { - if (replylen > MCLBYTES) - /* - * XXX: should we try to allocate more? But MCLBYTES is - * probably much larger than IPV6_MMTU... - */ + if (replylen > MCLBYTES) { + /* + * XXX: should we try to allocate more? But MCLBYTES + * is probably much larger than IPV6_MMTU... + */ goto bad; + } MCLGET(n, M_DONTWAIT); if ((n->m_flags & M_EXT) == 0) { goto bad; @@ -1144,49 +1448,57 @@ ni6_input(m, off) /* qtype dependent procedure */ switch (qtype) { - case NI_QTYPE_NOOP: - nni6->ni_flags = 0; - break; - case NI_QTYPE_SUPTYPES: - goto bad; /* xxx: to be implemented */ - break; - case NI_QTYPE_FQDN: - if (hostnamelen > 255) { /* XXX: rare case, but may happen */ - printf("ni6_input: " - "hostname length(%d) is too large for reply\n", - hostnamelen); - goto bad; - } - fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + - sizeof(struct ip6_hdr) + - sizeof(struct icmp6_nodeinfo)); - nni6->ni_flags = 0; /* XXX: meaningless TTL */ - fqdn->ni_fqdn_ttl = 0; /* ditto. */ - fqdn->ni_fqdn_namelen = hostnamelen; - bcopy(hostname, &fqdn->ni_fqdn_name[0], hostnamelen); - break; - case NI_QTYPE_NODEADDR: - { - int lenlim, copied; - - if (n->m_flags & M_EXT) - lenlim = MCLBYTES - sizeof(struct ip6_hdr) - - sizeof(struct icmp6_nodeinfo); - else - lenlim = MHLEN - sizeof(struct ip6_hdr) - - sizeof(struct icmp6_nodeinfo); - copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); - /* XXX: reset mbuf length */ - n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + - sizeof(struct icmp6_nodeinfo) + copied; - break; - } - default: - break; /* XXX impossible! */ + case NI_QTYPE_NOOP: + nni6->ni_code = ICMP6_NI_SUCCESS; + nni6->ni_flags = 0; + break; + case NI_QTYPE_SUPTYPES: + { + u_int32_t v; + nni6->ni_code = ICMP6_NI_SUCCESS; + nni6->ni_flags = htons(0x0000); /* raw bitmap */ + /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ + v = (u_int32_t)htonl(0x0000000f); + bcopy(&v, nni6 + 1, sizeof(u_int32_t)); + break; + } + case NI_QTYPE_FQDN: + nni6->ni_code = ICMP6_NI_SUCCESS; + fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + + sizeof(struct ip6_hdr) + + sizeof(struct icmp6_nodeinfo)); + nni6->ni_flags = 0; /* XXX: meaningless TTL */ + fqdn->ni_fqdn_ttl = 0; /* ditto. */ + /* + * XXX do we really have FQDN in variable "hostname"? + */ + n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); + if (n->m_next == NULL) + goto bad; + /* XXX we assume that n->m_next is not a chain */ + if (n->m_next->m_next != NULL) + goto bad; + n->m_pkthdr.len += n->m_next->m_len; + break; + case NI_QTYPE_NODEADDR: + { + int lenlim, copied; + + nni6->ni_code = ICMP6_NI_SUCCESS; + n->m_pkthdr.len = n->m_len = + sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); + lenlim = M_TRAILINGSPACE(n); + copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); + /* XXX: reset mbuf length */ + n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + + sizeof(struct icmp6_nodeinfo) + copied; + break; + } + default: + break; /* XXX impossible! */ } nni6->ni_type = ICMP6_NI_REPLY; - nni6->ni_code = ICMP6_NI_SUCESS; m_freem(m); return(n); @@ -1198,41 +1510,217 @@ ni6_input(m, off) } #undef hostnamelen +/* + * make a mbuf with DNS-encoded string. no compression support. + * + * XXX names with less than 2 dots (like "foo" or "foo.section") will be + * treated as truncated name (two \0 at the end). this is a wild guess. + */ +static struct mbuf * +ni6_nametodns(name, namelen, old) + const char *name; + int namelen; + int old; /* return pascal string if non-zero */ +{ + struct mbuf *m; + char *cp, *ep; + const char *p, *q; + int i, len, nterm; + + if (old) + len = namelen + 1; + else + len = MCLBYTES; + + /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ + MGET(m, M_DONTWAIT, MT_DATA); + if (m && len > MLEN) { + MCLGET(m, M_DONTWAIT); + if ((m->m_flags & M_EXT) == 0) + goto fail; + } + if (!m) + goto fail; + m->m_next = NULL; + + if (old) { + m->m_len = len; + *mtod(m, char *) = namelen; + bcopy(name, mtod(m, char *) + 1, namelen); + return m; + } else { + m->m_len = 0; + cp = mtod(m, char *); + ep = mtod(m, char *) + M_TRAILINGSPACE(m); + + /* if not certain about my name, return empty buffer */ + if (namelen == 0) + return m; + + /* + * guess if it looks like shortened hostname, or FQDN. + * shortened hostname needs two trailing "\0". + */ + i = 0; + for (p = name; p < name + namelen; p++) { + if (*p && *p == '.') + i++; + } + if (i < 2) + nterm = 2; + else + nterm = 1; + + p = name; + while (cp < ep && p < name + namelen) { + i = 0; + for (q = p; q < name + namelen && *q && *q != '.'; q++) + i++; + /* result does not fit into mbuf */ + if (cp + i + 1 >= ep) + goto fail; + /* + * DNS label length restriction, RFC1035 page 8. + * "i == 0" case is included here to avoid returning + * 0-length label on "foo..bar". + */ + if (i <= 0 || i >= 64) + goto fail; + *cp++ = i; + bcopy(p, cp, i); + cp += i; + p = q; + if (p < name + namelen && *p == '.') + p++; + } + /* termination */ + if (cp + nterm >= ep) + goto fail; + while (nterm-- > 0) + *cp++ = '\0'; + m->m_len = cp - mtod(m, char *); + return m; + } + + panic("should not reach here"); + /* NOTREACHED */ + + fail: + if (m) + m_freem(m); + return NULL; +} + +/* + * check if two DNS-encoded string matches. takes care of truncated + * form (with \0\0 at the end). no compression support. + * XXX upper/lowercase match (see RFC2065) + */ +static int +ni6_dnsmatch(a, alen, b, blen) + const char *a; + int alen; + const char *b; + int blen; +{ + const char *a0, *b0; + int l; + + /* simplest case - need validation? */ + if (alen == blen && bcmp(a, b, alen) == 0) + return 1; + + a0 = a; + b0 = b; + + /* termination is mandatory */ + if (alen < 2 || blen < 2) + return 0; + if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') + return 0; + alen--; + blen--; + + while (a - a0 < alen && b - b0 < blen) { + if (a - a0 + 1 > alen || b - b0 + 1 > blen) + return 0; + + if ((signed char)a[0] < 0 || (signed char)b[0] < 0) + return 0; + /* we don't support compression yet */ + if (a[0] >= 64 || b[0] >= 64) + return 0; + + /* truncated case */ + if (a[0] == 0 && a - a0 == alen - 1) + return 1; + if (b[0] == 0 && b - b0 == blen - 1) + return 1; + if (a[0] == 0 || b[0] == 0) + return 0; + + if (a[0] != b[0]) + return 0; + l = a[0]; + if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) + return 0; + if (bcmp(a + 1, b + 1, l) != 0) + return 0; + + a += 1 + l; + b += 1 + l; + } + + if (a - a0 == alen && b - b0 == blen) + return 1; + else + return 0; +} + /* * calculate the number of addresses to be returned in the node info reply. */ static int -ni6_addrs(ni6, m, ifpp) +ni6_addrs(ni6, ifpp, subj) struct icmp6_nodeinfo *ni6; - struct mbuf *m; struct ifnet **ifpp; + char *subj; { - register struct ifnet *ifp; - register struct in6_ifaddr *ifa6; - register struct ifaddr *ifa; - struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); + struct ifnet *ifp; + struct in6_ifaddr *ifa6; + struct ifaddr *ifa; + struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ int addrs = 0, addrsofif, iffound = 0; + int niflags = ni6->ni_flags; + + if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { + switch (ni6->ni_code) { + case ICMP6_NI_SUBJ_IPV6: + if (subj == NULL) /* must be impossible... */ + return(0); + subj_ip6 = (struct sockaddr_in6 *)subj; + break; + default: + /* + * XXX: we only support IPv6 subject address for + * this Qtype. + */ + return(0); + } + } -#if defined(__bsdi__) || (defined(__FreeBSD__) && __FreeBSD__ < 3) - for (ifp = ifnet; ifp; ifp = ifp->if_next) -#else - for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) -#endif - { + ifnet_head_lock_shared(); + TAILQ_FOREACH(ifp, &ifnet_head, if_list) { addrsofif = 0; -#if defined(__bsdi__) || (defined(__FreeBSD__) && __FreeBSD__ < 3) - for (ifa = ifp->if_addrlist; ifa; ifa = ifa->ifa_next) -#else - for (ifa = ifp->if_addrlist.tqh_first; ifa; - ifa = ifa->ifa_list.tqe_next) -#endif + ifnet_lock_shared(ifp); + TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) { if (ifa->ifa_addr->sa_family != AF_INET6) continue; ifa6 = (struct in6_ifaddr *)ifa; - if (!(ni6->ni_flags & NI_NODEADDR_FLAG_ALL) && - IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, + if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && + IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, &ifa6->ia_addr.sin6_addr)) iffound = 1; @@ -1241,44 +1729,52 @@ ni6_addrs(ni6, m, ifpp) * Node Information proxy, since they represent * addresses of IPv4-only nodes, which perforce do * not implement this protocol. - * [icmp-name-lookups-05] + * [icmp-name-lookups-07, Section 5.4] * So we don't support NI_NODEADDR_FLAG_COMPAT in * this function at this moment. */ - if (ifa6->ia6_flags & IN6_IFF_ANYCAST) - continue; /* we need only unicast addresses */ - - if ((ni6->ni_flags & (NI_NODEADDR_FLAG_LINKLOCAL | - NI_NODEADDR_FLAG_SITELOCAL | - NI_NODEADDR_FLAG_GLOBAL)) == 0) - continue; - /* What do we have to do about ::1? */ - switch(in6_addrscope(&ifa6->ia_addr.sin6_addr)) { - case IPV6_ADDR_SCOPE_LINKLOCAL: - if (ni6->ni_flags & NI_NODEADDR_FLAG_LINKLOCAL) - addrsofif++; + switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { + case IPV6_ADDR_SCOPE_LINKLOCAL: + if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) + continue; + break; + case IPV6_ADDR_SCOPE_SITELOCAL: + if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) + continue; break; - case IPV6_ADDR_SCOPE_SITELOCAL: - if (ni6->ni_flags & NI_NODEADDR_FLAG_SITELOCAL) - addrsofif++; + case IPV6_ADDR_SCOPE_GLOBAL: + if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) + continue; break; - case IPV6_ADDR_SCOPE_GLOBAL: - if (ni6->ni_flags & NI_NODEADDR_FLAG_GLOBAL) - addrsofif++; - break; - default: - continue; + default: + continue; + } + + /* + * check if anycast is okay. + * XXX: just experimental. not in the spec. + */ + if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && + (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) + continue; /* we need only unicast addresses */ + if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && + (icmp6_nodeinfo & 4) == 0) { + continue; } + addrsofif++; /* count the address */ } + ifnet_lock_done(ifp); if (iffound) { *ifpp = ifp; + ifnet_head_done(); return(addrsofif); } addrs += addrsofif; } + ifnet_head_done(); return(addrs); } @@ -1289,98 +1785,152 @@ ni6_store_addrs(ni6, nni6, ifp0, resid) struct ifnet *ifp0; int resid; { -#if defined(__bsdi__) || (defined(__FreeBSD__) && __FreeBSD__ < 3) - register struct ifnet *ifp = ifp0 ? ifp0 : ifnet; -#else - register struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet); -#endif - register struct in6_ifaddr *ifa6; - register struct ifaddr *ifa; - int docopy, copied = 0; + struct ifnet *ifp = ifp0; + struct in6_ifaddr *ifa6; + struct ifaddr *ifa; + struct ifnet *ifp_dep = NULL; + int copied = 0, allow_deprecated = 0; u_char *cp = (u_char *)(nni6 + 1); + int niflags = ni6->ni_flags; + u_int32_t ltime; + struct timeval timenow; - if (ifp0 == NULL && !(ni6->ni_flags & NI_NODEADDR_FLAG_ALL)) + getmicrotime(&timenow); + + if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) return(0); /* needless to copy */ -#if defined(__bsdi__) || (defined(__FreeBSD__) && __FreeBSD__ < 3) - for (; ifp; ifp = ifp->if_next) -#else - for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) -#endif - { -#if defined(__bsdi__) || (defined(__FreeBSD__) && __FreeBSD__ < 3) - for (ifa = ifp->if_addrlist; ifa; ifa = ifa->ifa_next) -#else + again: + + ifnet_head_lock_shared(); + if (ifp == NULL) ifp = TAILQ_FIRST(&ifnet_head); + + for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) { + ifnet_lock_shared(ifp); for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa->ifa_list.tqe_next) -#endif { - docopy = 0; - if (ifa->ifa_addr->sa_family != AF_INET6) continue; ifa6 = (struct in6_ifaddr *)ifa; - if (ifa6->ia6_flags & IN6_IFF_ANYCAST) { - /* just experimental. not in the spec. */ - if (ni6->ni_flags & NI_NODEADDR_FLAG_ANYCAST) - docopy = 1; - else - continue; - } - else { /* unicast address */ - if (ni6->ni_flags & NI_NODEADDR_FLAG_ANYCAST) - continue; - else - docopy = 1; + if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && + allow_deprecated == 0) { + /* + * prefererred address should be put before + * deprecated addresses. + */ + + /* record the interface for later search */ + if (ifp_dep == NULL) + ifp_dep = ifp; + + continue; } + else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && + allow_deprecated != 0) + continue; /* we now collect deprecated addrs */ /* What do we have to do about ::1? */ - switch(in6_addrscope(&ifa6->ia_addr.sin6_addr)) { - case IPV6_ADDR_SCOPE_LINKLOCAL: - if (ni6->ni_flags & NI_NODEADDR_FLAG_LINKLOCAL) - docopy = 1; + switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { + case IPV6_ADDR_SCOPE_LINKLOCAL: + if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) + continue; + break; + case IPV6_ADDR_SCOPE_SITELOCAL: + if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) + continue; break; - case IPV6_ADDR_SCOPE_SITELOCAL: - if (ni6->ni_flags & NI_NODEADDR_FLAG_SITELOCAL) - docopy = 1; + case IPV6_ADDR_SCOPE_GLOBAL: + if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) + continue; break; - case IPV6_ADDR_SCOPE_GLOBAL: - if (ni6->ni_flags & NI_NODEADDR_FLAG_GLOBAL) - docopy = 1; - break; - default: - continue; + default: + continue; } - if (docopy) { - if (resid < sizeof(struct in6_addr)) { - /* - * We give up much more copy. - * Set the truncate flag and return. - */ - nni6->ni_flags |= - NI_NODEADDR_FLAG_TRUNCATE; - return(copied); - } - bcopy(&ifa6->ia_addr.sin6_addr, cp, - sizeof(struct in6_addr)); - /* XXX: KAME link-local hack; remove ifindex */ - if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr)) - ((struct in6_addr *)cp)->s6_addr16[1] = 0; - cp += sizeof(struct in6_addr); - resid -= sizeof(struct in6_addr); - copied += sizeof(struct in6_addr); + /* + * check if anycast is okay. + * XXX: just experimental. not in the spec. + */ + if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && + (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) + continue; + if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && + (icmp6_nodeinfo & 4) == 0) { + continue; } + + /* now we can copy the address */ + if (resid < sizeof(struct in6_addr) + + sizeof(u_int32_t)) { + /* + * We give up much more copy. + * Set the truncate flag and return. + */ + nni6->ni_flags |= + NI_NODEADDR_FLAG_TRUNCATE; + ifnet_lock_done(ifp); + ifnet_head_done(); + return(copied); + } + + /* + * Set the TTL of the address. + * The TTL value should be one of the following + * according to the specification: + * + * 1. The remaining lifetime of a DHCP lease on the + * address, or + * 2. The remaining Valid Lifetime of a prefix from + * which the address was derived through Stateless + * Autoconfiguration. + * + * Note that we currently do not support stateful + * address configuration by DHCPv6, so the former + * case can't happen. + */ + if (ifa6->ia6_lifetime.ia6t_expire == 0) + ltime = ND6_INFINITE_LIFETIME; + else { + if (ifa6->ia6_lifetime.ia6t_expire > + timenow.tv_sec) + ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - timenow.tv_sec); + else + ltime = 0; + } + + bcopy(<ime, cp, sizeof(u_int32_t)); + cp += sizeof(u_int32_t); + + /* copy the address itself */ + bcopy(&ifa6->ia_addr.sin6_addr, cp, + sizeof(struct in6_addr)); + /* XXX: KAME link-local hack; remove ifindex */ + if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr)) + ((struct in6_addr *)cp)->s6_addr16[1] = 0; + cp += sizeof(struct in6_addr); + + resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); + copied += (sizeof(struct in6_addr) + + sizeof(u_int32_t)); } + ifnet_lock_done(ifp); if (ifp0) /* we need search only on the specified IF */ break; } + ifnet_head_done(); + + if (allow_deprecated == 0 && ifp_dep != NULL) { + ifp = ifp_dep; + allow_deprecated = 1; + + goto again; + } return(copied); } -#ifndef HAVE_NRL_INPCB /* * XXX almost dup'ed code with rip6_input. */ @@ -1390,12 +1940,12 @@ icmp6_rip6_input(mp, off) int off; { struct mbuf *m = *mp; - register struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); - register struct in6pcb *in6p; + struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); + struct in6pcb *in6p; struct in6pcb *last = NULL; struct sockaddr_in6 rip6src; struct icmp6_hdr *icmp6; - struct ip6_recvpktopts opts; + struct mbuf *opts = NULL; #ifndef PULLDOWN_TEST /* this is assumed to be safe. */ @@ -1408,30 +1958,19 @@ icmp6_rip6_input(mp, off) } #endif - bzero(&opts, sizeof(opts)); bzero(&rip6src, sizeof(rip6src)); rip6src.sin6_len = sizeof(struct sockaddr_in6); rip6src.sin6_family = AF_INET6; - rip6src.sin6_addr = ip6->ip6_src; - if (IN6_IS_SCOPE_LINKLOCAL(&rip6src.sin6_addr)) - rip6src.sin6_addr.s6_addr16[1] = 0; - if (m->m_pkthdr.rcvif) { - if (IN6_IS_SCOPE_LINKLOCAL(&rip6src.sin6_addr)) - rip6src.sin6_scope_id = m->m_pkthdr.rcvif->if_index; - else - rip6src.sin6_scope_id = 0; - } else - rip6src.sin6_scope_id = 0; + /* KAME hack: recover scopeid */ + (void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif); -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) + lck_rw_lock_shared(ripcbinfo.mtx); LIST_FOREACH(in6p, &ripcb, inp_list) -#else - for (in6p = rawin6pcb.in6p_next; - in6p != &rawin6pcb; in6p = in6p->in6p_next) -#endif { -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - if ((in6p->inp_vflag & INP_IPV6) == NULL) + if ((in6p->inp_vflag & INP_IPV6) == 0) + continue; +#if HAVE_NRL_INPCB + if (!(in6p->in6p_flags & INP_IPV6)) continue; #endif if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6) @@ -1450,45 +1989,35 @@ icmp6_rip6_input(mp, off) struct mbuf *n; if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { if (last->in6p_flags & IN6P_CONTROLOPTS) - ip6_savecontrol(last, ip6, n, &opts, - NULL); + ip6_savecontrol(last, &opts, ip6, n); /* strip intermediate headers */ m_adj(n, off); if (sbappendaddr(&last->in6p_socket->so_rcv, (struct sockaddr *)&rip6src, - n, opts.head) == 0) { - /* should notify about lost packet */ - m_freem(n); - if (opts.head) { - m_freem(opts.head); - } - } else + n, opts, NULL) != 0) { sorwakeup(last->in6p_socket); - bzero(&opts, sizeof(opts)); + } + opts = NULL; } } last = in6p; } + lck_rw_done(ripcbinfo.mtx); if (last) { if (last->in6p_flags & IN6P_CONTROLOPTS) - ip6_savecontrol(last, ip6, m, &opts, NULL); + ip6_savecontrol(last, &opts, ip6, m); /* strip intermediate headers */ m_adj(m, off); if (sbappendaddr(&last->in6p_socket->so_rcv, - (struct sockaddr *)&rip6src, - m, opts.head) == 0) { - m_freem(m); - if (opts.head) - m_freem(opts.head); - } else + (struct sockaddr *)&rip6src, m, opts, NULL) != 0) { sorwakeup(last->in6p_socket); + } } else { m_freem(m); ip6stat.ip6s_delivered--; } return IPPROTO_DONE; } -#endif /*OpenBSD*/ /* * Reflect the ip6 packet back to the source. @@ -1502,20 +2031,23 @@ icmp6_reflect(m, off) struct ip6_hdr *ip6; struct icmp6_hdr *icmp6; struct in6_ifaddr *ia; - struct in6_addr t, *src = 0; + struct in6_addr t, src_storage, *src = 0; int plen; int type, code; struct ifnet *outif = NULL; + struct sockaddr_in6 sa6_src, sa6_dst; #ifdef COMPAT_RFC1885 int mtu = IPV6_MMTU; struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst; #endif + u_int32_t oflow; /* too short to reflect */ if (off < sizeof(struct ip6_hdr)) { - printf("sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", - (u_long)off, (u_long)sizeof(struct ip6_hdr), - __FILE__, __LINE__); + nd6log((LOG_DEBUG, + "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", + (u_int32_t)off, (u_int32_t)sizeof(struct ip6_hdr), + __FILE__, __LINE__)); goto bad; } @@ -1523,6 +2055,10 @@ icmp6_reflect(m, off) * If there are extra headers between IPv6 and ICMPv6, strip * off that header first. */ +#if DIAGNOSTIC + if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) + panic("assumption failed in icmp6_reflect"); +#endif if (off > sizeof(struct ip6_hdr)) { size_t l; struct ip6_hdr nip6; @@ -1558,12 +2094,24 @@ icmp6_reflect(m, off) */ ip6->ip6_dst = ip6->ip6_src; - /* XXX hack for link-local addresses */ - if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) - ip6->ip6_dst.s6_addr16[1] = - htons(m->m_pkthdr.rcvif->if_index); - if (IN6_IS_ADDR_LINKLOCAL(&t)) - t.s6_addr16[1] = htons(m->m_pkthdr.rcvif->if_index); + /* + * XXX: make sure to embed scope zone information, using + * already embedded IDs or the received interface (if any). + * Note that rcvif may be NULL. + * TODO: scoped routing case (XXX). + */ + bzero(&sa6_src, sizeof(sa6_src)); + sa6_src.sin6_family = AF_INET6; + sa6_src.sin6_len = sizeof(sa6_src); + sa6_src.sin6_addr = ip6->ip6_dst; + in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif); + in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL); + bzero(&sa6_dst, sizeof(sa6_dst)); + sa6_dst.sin6_family = AF_INET6; + sa6_dst.sin6_len = sizeof(sa6_dst); + sa6_dst.sin6_addr = t; + in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif); + in6_embedscope(&t, &sa6_dst, NULL, NULL); #ifdef COMPAT_RFC1885 /* @@ -1572,15 +2120,12 @@ icmp6_reflect(m, off) * does not fit in with (return) path MTU, but the description was * removed in the new spec. */ - if (icmp6_reflect_rt.ro_rt == 0 || + if (icmp6_reflect_rt.ro_rt == NULL || + !(icmp6_reflect_rt.ro_rt->rt_flags & RTF_UP) || + icmp6_reflect_rt.ro_rt->generation_id != route_generation || ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) { if (icmp6_reflect_rt.ro_rt) { -#if defined(__FreeBSD__) || defined(__APPLE__) - RTFREE(icmp6_reflect_rt.ro_rt); -#endif -#ifdef __bsdi__ rtfree(icmp6_reflect_rt.ro_rt); -#endif icmp6_reflect_rt.ro_rt = 0; } bzero(sin6, sizeof(*sin6)); @@ -1588,20 +2133,18 @@ icmp6_reflect(m, off) sin6->sin6_len = sizeof(struct sockaddr_in6); sin6->sin6_addr = ip6->ip6_dst; -#if defined(__FreeBSD__) || defined(__APPLE__) rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt, RTF_PRCLONING); -#else - rtalloc((struct route *)&icmp6_reflect_rt.ro_rt); -#endif } if (icmp6_reflect_rt.ro_rt == 0) goto bad; + RT_LOCK(icmp6_reflect_rt.ro_rt); if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST) && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu) mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu; + RT_UNLOCK(icmp6_reflect_rt.ro_rt); if (mtu < m->m_pkthdr.len) { plen -= (m->m_pkthdr.len - mtu); @@ -1611,67 +2154,86 @@ icmp6_reflect(m, off) /* * If the incoming packet was addressed directly to us(i.e. unicast), * use dst as the src for the reply. - * The IN6_IFF_NOTREADY case would be VERY rare, but is possible when + * The IN6_IFF_NOTREADY case would be VERY rare, but is possible * (for example) when we encounter an error while forwarding procedure * destined to a duplicated address of ours. */ - for (ia = in6_ifaddr; ia; ia = ia->ia_next) + lck_mtx_lock(nd6_mutex); + for (ia = in6_ifaddrs; ia; ia = ia->ia_next) if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) && (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) { src = &t; break; } + lck_mtx_unlock(nd6_mutex); if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) { /* * This is the case if the dst is our link-local address - * and the sender is also ourseleves. + * and the sender is also ourselves. */ src = &t; } - if (src == 0) + if (src == 0) { + int e; + struct route_in6 ro; + /* * This case matches to multicasts, our anycast, or unicasts - * that we do not own. Select a source address which has the - * same scope. - * XXX: for (non link-local) multicast addresses, this might - * not be a good choice. + * that we do not own. Select a source address based on the + * source address of the erroneous packet. */ - if ((ia = in6_ifawithscope(m->m_pkthdr.rcvif, &t)) != 0) - src = &IA6_SIN6(ia)->sin6_addr; - - if (src == 0) - goto bad; + bzero(&ro, sizeof(ro)); + src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &src_storage, &e); + if (ro.ro_rt) + rtfree(ro.ro_rt); /* XXX: we could use this */ + if (src == NULL) { + nd6log((LOG_DEBUG, + "icmp6_reflect: source can't be determined: " + "dst=%s, error=%d\n", + ip6_sprintf(&sa6_src.sin6_addr), e)); + goto bad; + } + } ip6->ip6_src = *src; + oflow = ip6->ip6_flow; /* Save for later */ ip6->ip6_flow = 0; ip6->ip6_vfc &= ~IPV6_VERSION_MASK; ip6->ip6_vfc |= IPV6_VERSION; + if (icmp6->icmp6_type == ICMP6_ECHO_REPLY && icmp6->icmp6_code == 0) { + ip6->ip6_flow |= (oflow & htonl(0x0ff00000)); + } ip6->ip6_nxt = IPPROTO_ICMPV6; - if (m->m_pkthdr.rcvif) { + if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_index < nd_ifinfo_indexlim) { /* XXX: This may not be the outgoing interface */ + lck_rw_lock_shared(nd_if_rwlock); ip6->ip6_hlim = nd_ifinfo[m->m_pkthdr.rcvif->if_index].chlim; + lck_rw_done(nd_if_rwlock); + } else { + ip6->ip6_hlim = ip6_defhlim; } - + /* Use the same traffic class as in the request to match IPv4 */ icmp6->icmp6_cksum = 0; icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), plen); /* - * xxx option handling + * XXX option handling */ m->m_flags &= ~(M_BCAST|M_MCAST); #if IPSEC /* Don't lookup socket */ - ipsec_setsocket(m, NULL); + if (ipsec_bypass == 0) + (void)ipsec_setsocket(m, NULL); #endif /*IPSEC*/ #ifdef COMPAT_RFC1885 - ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif); + ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, 0); #else - ip6_output(m, NULL, NULL, 0, NULL, &outif); + ip6_output(m, NULL, NULL, 0, NULL, &outif, 0); #endif if (outif) icmp6_ifoutstat_inc(outif, type, code); @@ -1686,6 +2248,7 @@ icmp6_reflect(m, off) void icmp6_fasttimo() { + mld6_fasttimeo(); } @@ -1696,19 +2259,14 @@ icmp6_redirect_diag(src6, dst6, tgt6) struct in6_addr *tgt6; { static char buf[1024]; -#if !defined(__OpenBSD__) && !defined(__bsdi__) snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); -#else - sprintf(buf, "(src=%s dst=%s tgt=%s)", - ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); -#endif return buf; } void icmp6_redirect_input(m, off) - register struct mbuf *m; + struct mbuf *m; int off; { struct ifnet *ifp = m->m_pkthdr.rcvif; @@ -1737,7 +2295,7 @@ icmp6_redirect_input(m, off) goto freeit; #ifndef PULLDOWN_TEST - IP6_EXTHDR_CHECK(m, off, icmp6len,); + IP6_EXTHDR_CHECK(m, off, icmp6len, return); nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off); #else IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); @@ -1756,17 +2314,17 @@ icmp6_redirect_input(m, off) /* validation */ if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { - log(LOG_ERR, + nd6log((LOG_ERR, "ICMP6 redirect sent from %s rejected; " - "must be from linklocal\n", ip6_sprintf(&src6)); - goto freeit; + "must be from linklocal\n", ip6_sprintf(&src6))); + goto bad; } if (ip6->ip6_hlim != 255) { - log(LOG_ERR, + nd6log((LOG_ERR, "ICMP6 redirect sent from %s rejected; " "hlim=%d (must be 255)\n", - ip6_sprintf(&src6), ip6->ip6_hlim); - goto freeit; + ip6_sprintf(&src6), ip6->ip6_hlim)); + goto bad; } { /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ @@ -1777,39 +2335,49 @@ icmp6_redirect_input(m, off) sin6.sin6_family = AF_INET6; sin6.sin6_len = sizeof(struct sockaddr_in6); bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6)); - rt = rtalloc1((struct sockaddr *)&sin6, 0 -#if defined(__FreeBSD__) || defined (__APPLE__) - , 0UL -#endif - ); + rt = rtalloc1((struct sockaddr *)&sin6, 0, 0); if (rt) { + RT_LOCK(rt); + if (rt->rt_gateway == NULL || + rt->rt_gateway->sa_family != AF_INET6) { + nd6log((LOG_ERR, + "ICMP6 redirect rejected; no route " + "with inet6 gateway found for redirect dst: %s\n", + icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); + RT_UNLOCK(rt); + rtfree(rt); + goto bad; + } + gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { - log(LOG_ERR, + nd6log((LOG_ERR, "ICMP6 redirect rejected; " "not equal to gw-for-src=%s (must be same): " "%s\n", ip6_sprintf(gw6), - icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); - RTFREE(rt); - goto freeit; + icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); + RT_UNLOCK(rt); + rtfree(rt); + goto bad; } } else { - log(LOG_ERR, + nd6log((LOG_ERR, "ICMP6 redirect rejected; " "no route found for redirect dst: %s\n", - icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); - goto freeit; + icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); + goto bad; } - RTFREE(rt); + RT_UNLOCK(rt); + rtfree(rt); rt = NULL; } if (IN6_IS_ADDR_MULTICAST(&reddst6)) { - log(LOG_ERR, + nd6log((LOG_ERR, "ICMP6 redirect rejected; " "redirect dst must be unicast: %s\n", - icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); - goto freeit; + icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); + goto bad; } is_router = is_onlink = 0; @@ -1818,20 +2386,21 @@ icmp6_redirect_input(m, off) if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) is_onlink = 1; /* on-link destination case */ if (!is_router && !is_onlink) { - log(LOG_ERR, + nd6log((LOG_ERR, "ICMP6 redirect rejected; " "neither router case nor onlink case: %s\n", - icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); - goto freeit; + icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); + goto bad; } /* validation passed */ icmp6len -= sizeof(*nd_rd); nd6_option_init(nd_rd + 1, icmp6len, &ndopts); if (nd6_options(&ndopts) < 0) { - log(LOG_INFO, "icmp6_redirect_input: " + nd6log((LOG_INFO, "icmp6_redirect_input: " "invalid ND option, rejected: %s\n", - icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); + icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); + /* nd6_options have incremented stats */ goto freeit; } @@ -1846,25 +2415,23 @@ icmp6_redirect_input(m, off) } if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { - log(LOG_INFO, + nd6log((LOG_INFO, "icmp6_redirect_input: lladdrlen mismatch for %s " "(if %d, icmp6 packet %d): %s\n", ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, - icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); + icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); + goto bad; } /* RFC 2461 8.3 */ nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); - if (!is_onlink) { /* better router case. perform rtredirect. */ + if (!is_onlink) { /* better router case. perform rtredirect. */ /* perform rtredirect */ struct sockaddr_in6 sdst; struct sockaddr_in6 sgw; struct sockaddr_in6 ssrc; -#ifdef __bsdi__ - extern int icmp_redirtimeout; /*XXX*/ -#endif bzero(&sdst, sizeof(sdst)); bzero(&sgw, sizeof(sgw)); @@ -1875,44 +2442,28 @@ icmp6_redirect_input(m, off) bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); - rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, - (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST, - (struct sockaddr *)&ssrc, -#ifdef __bsdi__ - icmp_redirtimeout -#else - (struct rtentry **)NULL -#endif /*__FreeBSD__, __NetBSD__, __bsdi__*/ - ); + rtredirect(ifp, (struct sockaddr *)&sdst, + (struct sockaddr *)&sgw, NULL, RTF_GATEWAY | RTF_HOST, + (struct sockaddr *)&ssrc, NULL); } /* finally update cached route in each socket via pfctlinput */ { struct sockaddr_in6 sdst; -#if 1 -#else - struct ip6protosw *pr; -#endif bzero(&sdst, sizeof(sdst)); sdst.sin6_family = AF_INET6; sdst.sin6_len = sizeof(struct sockaddr_in6); bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); -#if 1 - pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); -#else + /* - * do not use pfctlinput() here, we have different prototype for - * xx_ctlinput() in ip6proto. + * Radar 6843900 + * Release the IPv6 domain lock because we are going to take domain_proto_mtx + * and could otherwise cause a deadlock with other threads taking these locks + * in the reverse order -- e.g. frag6_slowtimo() from pfslowtimo() */ - for (pr = (struct ip6protosw *)inet6domain.dom_protosw; - pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; - pr++) { - if (pr->pr_ctlinput) { - (*pr->pr_ctlinput)(PRC_REDIRECT_HOST, - (struct sockaddr *)&sdst, NULL, NULL, 0); - } - } -#endif + lck_mtx_unlock(inet6_domain_mutex); + pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); + lck_mtx_lock(inet6_domain_mutex); #if IPSEC key_sa_routechange((struct sockaddr *)&sdst); #endif @@ -1920,6 +2471,11 @@ icmp6_redirect_input(m, off) freeit: m_freem(m); + return; + + bad: + icmp6stat.icp6s_badredirect++; + m_freem(m); } void @@ -1928,7 +2484,7 @@ icmp6_redirect_output(m0, rt) struct rtentry *rt; { struct ifnet *ifp; /* my outgoing interface */ - struct in6_addr *ifp_ll6; + struct in6_addr ifp_ll6; struct in6_addr *router_ll6; struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ struct mbuf *m = NULL; /* newly allocated one */ @@ -1937,15 +2493,21 @@ icmp6_redirect_output(m0, rt) size_t maxlen; u_char *p; struct ifnet *outif = NULL; + struct sockaddr_in6 src_sa; - /* if we are not router, we don't send icmp6 redirect */ - if (!ip6_forwarding || ip6_accept_rtadv) - goto fail; + icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0); + + if (rt != NULL) + RT_LOCK(rt); /* sanity check */ if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) goto fail; + /* if we are not router, we don't send icmp6 redirect */ + if (!ip6_forwarding || ip6_accept_rtadv || (ifp->if_eflags & IFEF_ACCEPT_RTADVD)) + goto fail; + /* * Address check: * the source address must identify a neighbor, and @@ -1953,8 +2515,19 @@ icmp6_redirect_output(m0, rt) * [RFC 2461, sec 8.2] */ sip6 = mtod(m0, struct ip6_hdr *); - if (nd6_is_addr_neighbor(&sip6->ip6_src, ifp) == 0) + bzero(&src_sa, sizeof(src_sa)); + src_sa.sin6_family = AF_INET6; + src_sa.sin6_len = sizeof(src_sa); + src_sa.sin6_addr = sip6->ip6_src; + /* we don't currently use sin6_scope_id, but eventually use it */ + src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src); + RT_UNLOCK(rt); + if (nd6_is_addr_neighbor(&src_sa, ifp, 0) == 0) { + /* already unlocked */ + rt = NULL; goto fail; + } + RT_LOCK(rt); if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) goto fail; /* what should we do here? */ @@ -1970,12 +2543,14 @@ icmp6_redirect_output(m0, rt) #if IPV6_MMTU >= MCLBYTES # error assumption failed about IPV6_MMTU and MCLBYTES #endif - MGETHDR(m, M_DONTWAIT, MT_HEADER); + MGETHDR(m, M_DONTWAIT, MT_HEADER); /* MAC-OK */ if (m && IPV6_MMTU >= MHLEN) MCLGET(m, M_DONTWAIT); if (!m) goto fail; - maxlen = (m->m_flags & M_EXT) ? MCLBYTES : MHLEN; + m->m_pkthdr.rcvif = NULL; + m->m_len = 0; + maxlen = M_TRAILINGSPACE(m); maxlen = min(IPV6_MMTU, maxlen); /* just for safety */ if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + @@ -1990,7 +2565,8 @@ icmp6_redirect_output(m0, rt) IN6_IFF_NOTREADY| IN6_IFF_ANYCAST)) == NULL) goto fail; - ifp_ll6 = &ia->ia_addr.sin6_addr; + ifp_ll6 = ia->ia_addr.sin6_addr; + ifafree(&ia->ia_ifa); } /* get ip6 linklocal address for the router. */ @@ -2012,7 +2588,7 @@ icmp6_redirect_output(m0, rt) ip6->ip6_nxt = IPPROTO_ICMPV6; ip6->ip6_hlim = 255; /* ip6->ip6_src must be linklocal addr for my outgoing if. */ - bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); + bcopy(&ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); /* ND Redirect */ @@ -2038,6 +2614,8 @@ icmp6_redirect_output(m0, rt) bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, sizeof(nd_rd->nd_rd_dst)); } + RT_UNLOCK(rt); + rt = NULL; p = (u_char *)(nd_rd + 1); @@ -2052,14 +2630,19 @@ icmp6_redirect_output(m0, rt) struct nd_opt_hdr *nd_opt; char *lladdr; - rt_router = nd6_lookup(router_ll6, 0, ifp); + /* Callee returns a locked route upon success */ + rt_router = nd6_lookup(router_ll6, 0, ifp, 0); if (!rt_router) goto nolladdropt; + RT_LOCK_ASSERT_HELD(rt_router); len = sizeof(*nd_opt) + ifp->if_addrlen; - len = (len + 7) & ~7; /*round by 8*/ + len = (len + 7) & ~7; /* round by 8 */ /* safety check */ - if (len + (p - (u_char *)ip6) > maxlen) + if (len + (p - (u_char *)ip6) > maxlen) { + RT_REMREF_LOCKED(rt_router); + RT_UNLOCK(rt_router); goto nolladdropt; + } if (!(rt_router->rt_flags & RTF_GATEWAY) && (rt_router->rt_flags & RTF_LLINFO) && (rt_router->rt_gateway->sa_family == AF_LINK) && @@ -2072,13 +2655,15 @@ icmp6_redirect_output(m0, rt) bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen); p += len; } + RT_REMREF_LOCKED(rt_router); + RT_UNLOCK(rt_router); } nolladdropt:; m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; /* just to be safe */ -#if M_DECRYPTED /*not openbsd*/ +#ifdef M_DECRYPTED /*not openbsd*/ if (m0->m_flags & M_DECRYPTED) goto noredhdropt; #endif @@ -2180,9 +2765,10 @@ noredhdropt:; /* send the packet to outside... */ #if IPSEC /* Don't lookup socket */ - ipsec_setsocket(m, NULL); + if (ipsec_bypass == 0) + (void)ipsec_setsocket(m, NULL); #endif /*IPSEC*/ - ip6_output(m, NULL, NULL, 0, NULL, &outif); + ip6_output(m, NULL, NULL, 0, NULL, &outif, 0); if (outif) { icmp6_ifstat_inc(outif, ifs6_out_msg); icmp6_ifstat_inc(outif, ifs6_out_redirect); @@ -2192,36 +2778,30 @@ noredhdropt:; return; fail: + if (rt != NULL) + RT_UNLOCK(rt); if (m) m_freem(m); if (m0) m_freem(m0); } -#ifndef HAVE_NRL_INPCB +#if HAVE_NRL_INPCB +#define sotoin6pcb sotoinpcb +#define in6pcb inpcb +#define in6p_icmp6filt inp_icmp6filt +#endif /* * ICMPv6 socket option processing. - * - * NOTE: for OSes that use NRL inpcb (bsdi4/openbsd), do not forget to modify - * sys/netinet6/raw_ipv6.c:rip6_ctloutput(). */ int -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) icmp6_ctloutput(so, sopt) struct socket *so; struct sockopt *sopt; -#else -icmp6_ctloutput(op, so, level, optname, mp) - int op; - struct socket *so; - int level, optname; - struct mbuf **mp; -#endif { int error = 0; int optlen; -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - register struct inpcb *inp = sotoinpcb(so); + struct inpcb *inp = sotoinpcb(so); int level, op, optname; if (sopt) { @@ -2231,22 +2811,12 @@ icmp6_ctloutput(op, so, level, optname, mp) optlen = sopt->sopt_valsize; } else level = op = optname = optlen = 0; -#else - register struct in6pcb *in6p = sotoin6pcb(so); - register struct mbuf *m = *mp; - - optlen = m ? m->m_len : 0; -#endif if (level != IPPROTO_ICMPV6) { -#if !(defined(__FreeBSD__) && __FreeBSD__ >= 3) && !defined(__APPLE__) - if (op == PRCO_SETOPT && m) - (void)m_free(m); -#endif return EINVAL; } - switch(op) { + switch (op) { case PRCO_SETOPT: switch (optname) { case ICMP6_FILTER: @@ -2257,23 +2827,12 @@ icmp6_ctloutput(op, so, level, optname, mp) error = EMSGSIZE; break; } -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) if (inp->in6p_icmp6filt == NULL) { error = EINVAL; break; } error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen, optlen); -#else - p = mtod(m, struct icmp6_filter *); - if (!p || !in6p->in6p_icmp6filt) { - error = EINVAL; - break; - } - bcopy(p, in6p->in6p_icmp6filt, - sizeof(struct icmp6_filter)); - error = 0; -#endif break; } @@ -2281,37 +2840,18 @@ icmp6_ctloutput(op, so, level, optname, mp) error = ENOPROTOOPT; break; } -#if !(defined(__FreeBSD__) && __FreeBSD__ >= 3) && !defined(__APPLE__) - if (m) - (void)m_freem(m); -#endif break; case PRCO_GETOPT: switch (optname) { case ICMP6_FILTER: { -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) if (inp->in6p_icmp6filt == NULL) { error = EINVAL; break; } error = sooptcopyout(sopt, inp->in6p_icmp6filt, sizeof(struct icmp6_filter)); -#else - struct icmp6_filter *p; - - if (!in6p->in6p_icmp6filt) { - error = EINVAL; - break; - } - *mp = m = m_get(M_WAIT, MT_SOOPTS); - m->m_len = sizeof(struct icmp6_filter); - p = mtod(m, struct icmp6_filter *); - bcopy(in6p->in6p_icmp6filt, p, - sizeof(struct icmp6_filter)); - error = 0; -#endif break; } @@ -2324,201 +2864,264 @@ icmp6_ctloutput(op, so, level, optname, mp) return(error); } -#endif /*NRL inpcb*/ +#if HAVE_NRL_INPCB +#undef sotoin6pcb +#undef in6pcb +#undef in6p_icmp6filt +#endif /* - * Perform rate limit check. - * Returns 0 if it is okay to send the icmp6 packet. - * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate - * limitation. - * - * XXX per-destination/type check necessary? + * ICMPv6 socket datagram option processing. */ -static int -icmp6_ratelimit(dst, type, code) - const struct in6_addr *dst; /* not used at this moment */ - const int type; /* not used at this moment */ - const int code; /* not used at this moment */ +int +icmp6_dgram_ctloutput(struct socket *so, struct sockopt *sopt) { - struct timeval tp; - long sec_diff, usec_diff; + if (so->so_uid == 0) + return icmp6_ctloutput(so, sopt); - /* If we are not doing rate limitation, it is always okay to send */ - if (!icmp6errratelim) - return 0; - -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - microtime(&tp); - tp.tv_sec = time_second; -#else - tp = time; -#endif - if (tp.tv_sec < icmp6_nextsend.tv_sec - || (tp.tv_sec == icmp6_nextsend.tv_sec - && tp.tv_usec < icmp6_nextsend.tv_usec)) { - /* The packet is subject to rate limit */ - return 1; + if (sopt->sopt_level == IPPROTO_ICMPV6) { + switch (sopt->sopt_name) { + case ICMP6_FILTER: + return icmp6_ctloutput(so, sopt); + default: + return EPERM; + } } - sec_diff = icmp6errratelim / 1000000; - usec_diff = icmp6errratelim % 1000000; - icmp6_nextsend.tv_sec = tp.tv_sec + sec_diff; - if ((tp.tv_usec = tp.tv_usec + usec_diff) >= 1000000) { - icmp6_nextsend.tv_sec++; - icmp6_nextsend.tv_usec -= 1000000; + + if (sopt->sopt_level != IPPROTO_IPV6) + return EINVAL; + + switch (sopt->sopt_name) { + case IPV6_PKTOPTIONS: + case IPV6_UNICAST_HOPS: + case IPV6_CHECKSUM: + case IPV6_FAITH: + case IPV6_V6ONLY: + case IPV6_PKTINFO: + case IPV6_HOPLIMIT: + case IPV6_HOPOPTS: + case IPV6_DSTOPTS: + case IPV6_RTHDR: + case IPV6_MULTICAST_IF: + case IPV6_MULTICAST_HOPS: + case IPV6_MULTICAST_LOOP: + case IPV6_JOIN_GROUP: + case IPV6_LEAVE_GROUP: + case IPV6_PORTRANGE: + case IPV6_IPSEC_POLICY: + case IPV6_RECVTCLASS: + case IPV6_TCLASS: + return ip6_ctloutput(so, sopt); + + default: + return EPERM; + + } - - /* it is okay to send this */ - return 0; } -#if defined(__NetBSD__) || defined(__OpenBSD__) -static struct rtentry * -icmp6_mtudisc_clone(dst) - struct sockaddr *dst; +__private_extern__ int +icmp6_dgram_send(struct socket *so, __unused int flags, struct mbuf *m, struct sockaddr *nam, + struct mbuf *control, __unused struct proc *p) { - struct rtentry *rt; - int error; + int error = 0; + struct inpcb *inp = sotoinpcb(so); + struct sockaddr_in6 tmp; + struct sockaddr_in6 *dst; + struct icmp6_hdr *icmp6; - rt = rtalloc1(dst, 1); - if (rt == 0) - return NULL; + if (so->so_uid == 0) + return rip6_output(m, so, (struct sockaddr_in6 *) nam, control); - /* If we didn't get a host route, allocate one */ - if ((rt->rt_flags & RTF_HOST) == 0) { - struct rtentry *nrt; + /* always copy sockaddr to avoid overwrites */ + if (so->so_state & SS_ISCONNECTED) { + if (nam) { + m_freem(m); + return EISCONN; + } + /* XXX */ + bzero(&tmp, sizeof(tmp)); + tmp.sin6_family = AF_INET6; + tmp.sin6_len = sizeof(struct sockaddr_in6); + bcopy(&inp->in6p_faddr, &tmp.sin6_addr, + sizeof(struct in6_addr)); + dst = &tmp; + } else { + if (nam == NULL) { + m_freem(m); + return ENOTCONN; + } + tmp = *(struct sockaddr_in6 *)nam; + dst = &tmp; + } - error = rtrequest((int) RTM_ADD, dst, - (struct sockaddr *) rt->rt_gateway, - (struct sockaddr *) 0, - RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); - if (error) { - rtfree(rt); - rtfree(nrt); - return NULL; + /* + * For an ICMPv6 packet, we should know its type and code + */ + if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) { + if (m->m_len < sizeof(struct icmp6_hdr) && + (m = m_pullup(m, sizeof(struct icmp6_hdr))) == NULL) { + error = ENOBUFS; + goto bad; + } + icmp6 = mtod(m, struct icmp6_hdr *); + + /* + * Allow only to send echo request and node information request + * See RFC 2463 for Echo Request Message format + */ + if ((icmp6->icmp6_type == ICMP6_ECHO_REQUEST && icmp6->icmp6_code == 0) || + (icmp6->icmp6_type == ICMP6_NI_QUERY && + (icmp6->icmp6_code == ICMP6_NI_SUBJ_IPV6 || + icmp6->icmp6_code == ICMP6_NI_SUBJ_FQDN))) { + /* Good */ + ; + } else { + error = EPERM; + goto bad; } - nrt->rt_rmx = rt->rt_rmx; - rtfree(rt); - rt = nrt; } - error = rt_timer_add(rt, icmp6_mtudisc_timeout, - icmp6_mtudisc_timeout_q); - if (error) { - rtfree(rt); - return NULL; + +#if ENABLE_DEFAULT_SCOPE + if (dst->sin6_scope_id == 0) { /* not change if specified */ + dst->sin6_scope_id = scope6_addr2default(&dst->sin6_addr); } +#endif - return rt; /* caller need to call rtfree() */ + return rip6_output(m, so, (struct sockaddr_in6 *) nam, control); +bad: + m_freem(m); + return error; } -static void -icmp6_mtudisc_timeout(rt, r) - struct rtentry *rt; - struct rttimer *r; +/* Like rip6_attach but without root privilege enforcement */ +__private_extern__ int +icmp6_dgram_attach(struct socket *so, int proto, struct proc *p) { - if (rt == NULL) - panic("icmp6_mtudisc_timeout: bad route to timeout"); - if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == - (RTF_DYNAMIC | RTF_HOST)) { - rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), - rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); - } else { - if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) { - rt->rt_rmx.rmx_mtu = 0; - } - } -} -#endif /*__NetBSD__ || __OpenBSD__*/ + struct inpcb *inp; + int error; -#ifdef __bsdi__ -void -icmp6_mtuexpire(rt, rtt) - struct rtentry *rt; - struct rttimer *rtt; -{ - rt->rt_flags |= RTF_PROBEMTU; - Free(rtt); + inp = sotoinpcb(so); + if (inp) + panic("icmp6_dgram_attach"); + + if (proto != IPPROTO_ICMPV6) + return EINVAL; + + error = soreserve(so, rip_sendspace, rip_recvspace); + if (error) + return error; + error = in_pcballoc(so, &ripcbinfo, p); + if (error) + return error; + inp = (struct inpcb *)so->so_pcb; + inp->inp_vflag |= INP_IPV6; + inp->in6p_ip6_nxt = IPPROTO_ICMPV6; + inp->in6p_hops = -1; /* use kernel default */ + inp->in6p_cksum = -1; + MALLOC(inp->in6p_icmp6filt, struct icmp6_filter *, + sizeof(struct icmp6_filter), M_PCB, M_WAITOK); + if (inp->in6p_icmp6filt == NULL) + return (ENOMEM); + ICMP6_FILTER_SETPASSALL(inp->in6p_icmp6filt); + return 0; } -int *icmp6_sysvars[] = ICMPV6CTL_VARS; -int -icmp6_sysctl(name, namelen, oldp, oldlenp, newp, newlen) - int *name; - u_int namelen; - void *oldp; - size_t *oldlenp; - void *newp; - size_t newlen; + +#ifndef HAVE_PPSRATECHECK +#ifndef timersub +#define timersub(tvp, uvp, vvp) \ + do { \ + (vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec; \ + (vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec; \ + if ((vvp)->tv_usec < 0) { \ + (vvp)->tv_sec--; \ + (vvp)->tv_usec += 1000000; \ + } \ + } while (0) +#endif + +/* + * ppsratecheck(): packets (or events) per second limitation. + */ +static int +ppsratecheck(lasttime, curpps, maxpps) + struct timeval *lasttime; + int *curpps; + int maxpps; /* maximum pps allowed */ { - if (name[0] >= ICMPV6CTL_MAXID) - return (EOPNOTSUPP); - switch (name[0]) { -#if 0 - ICMPV6CTL_ND6_PRUNE: - ICMPV6CTL_ND6_DELAY: - ICMPV6CTL_ND6_UMAXTRIES: - ICMPV6CTL_ND6_MMAXTRIES: - ICMPV6CTL_ND6_USELOOPBACK: - /* need to check the value. */ + struct timeval tv, delta; + int rv; + + microtime(&tv); + + timersub(&tv, lasttime, &delta); + + /* + * Check for 0,0 so that the message will be seen at least once. + * If more than one second has passed since the last update of + * lasttime, reset the counter. + * + * we do increment *curpps even in *curpps < maxpps case, as some may + * try to use *curpps for stat purposes as well. + */ + if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) || + delta.tv_sec >= 1) { + *lasttime = tv; + *curpps = 0; + rv = 1; + } else if (maxpps < 0) + rv = 1; + else if (*curpps < maxpps) + rv = 1; + else + rv = 0; + +#if 1 /* DIAGNOSTIC? */ + /* be careful about wrap-around */ + if (*curpps + 1 > 0) + *curpps = *curpps + 1; +#else + /* + * assume that there's not too many calls to this function. + * not sure if the assumption holds, as it depends on *caller's* + * behavior, not the behavior of this function. + * IMHO it is wrong to make assumption on the caller's behavior, + * so the above #if is #if 1, not #ifdef DIAGNOSTIC. + */ + *curpps = *curpps + 1; #endif - case ICMPV6CTL_STATS: - return sysctl_rdtrunc(oldp, oldlenp, newp, &icmp6stat, - sizeof(icmp6stat)); - default: - return (sysctl_int_arr(icmp6_sysvars, name, namelen, - oldp, oldlenp, newp, newlen)); - } + return (rv); } -#endif /*__bsdi__*/ +#endif -#if defined(__NetBSD__) || defined(__OpenBSD__) -#include -#include -int -icmp6_sysctl(name, namelen, oldp, oldlenp, newp, newlen) - int *name; - u_int namelen; - void *oldp; - size_t *oldlenp; - void *newp; - size_t newlen; +/* + * Perform rate limit check. + * Returns 0 if it is okay to send the icmp6 packet. + * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate + * limitation. + * + * XXX per-destination/type check necessary? + */ +static int +icmp6_ratelimit( + __unused const struct in6_addr *dst, /* not used at this moment */ + __unused const int type, /* not used at this moment */ + __unused const int code) /* not used at this moment */ { + int ret; - /* All sysctl names at this level are terminal. */ - if (namelen != 1) - return ENOTDIR; - - switch (name[0]) { - - case ICMPV6CTL_REDIRACCEPT: - return sysctl_int(oldp, oldlenp, newp, newlen, - &icmp6_rediraccept); - case ICMPV6CTL_REDIRTIMEOUT: - return sysctl_int(oldp, oldlenp, newp, newlen, - &icmp6_redirtimeout); - case ICMPV6CTL_STATS: - return sysctl_rdstruct(oldp, oldlenp, newp, - &icmp6stat, sizeof(icmp6stat)); - case ICMPV6CTL_ERRRATELIMIT: - return sysctl_int(oldp, oldlenp, newp, newlen, - &icmp6errratelim); - case ICMPV6CTL_ND6_PRUNE: - return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_prune); - case ICMPV6CTL_ND6_DELAY: - return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_delay); - case ICMPV6CTL_ND6_UMAXTRIES: - return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_umaxtries); - case ICMPV6CTL_ND6_MMAXTRIES: - return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_mmaxtries); - case ICMPV6CTL_ND6_USELOOPBACK: - return sysctl_int(oldp, oldlenp, newp, newlen, - &nd6_useloopback); - case ICMPV6CTL_NODEINFO: - return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6_nodeinfo); - default: - return ENOPROTOOPT; + ret = 0; /* okay to send */ + + /* PPS limit */ + if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, + icmp6errppslim)) { + /* The packet is subject to rate limit */ + ret++; } - /* NOTREACHED */ + + return ret; } -#endif /* __NetBSD__ */ +