X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/1c79356b52d46aa6b508fb032f5ae709b1f2897b..3e170ce000f1506b7b5d2c5c7faec85ceabb573d:/bsd/netinet6/ip6_output.c diff --git a/bsd/netinet6/ip6_output.c b/bsd/netinet6/ip6_output.c index a6e82dac0..7767822d7 100644 --- a/bsd/netinet6/ip6_output.c +++ b/bsd/netinet6/ip6_output.c @@ -1,4 +1,30 @@ -/* $KAME: ip6_output.c,v 1.94 2000/04/04 14:45:44 itojun Exp $ */ +/* + * Copyright (c) 2000-2015 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@ + */ /* * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. @@ -63,16 +89,12 @@ * * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 */ - -#if __FreeBSD__ -#include "opt_ip6fw.h" -#endif -#if (defined(__FreeBSD__) && __FreeBSD__ >= 3) || defined(__NetBSD__) -#include "opt_inet.h" -#if __NetBSD__ /*XXX*/ -#include "opt_ipsec.h" -#endif -#endif +/* + * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce + * support for mandatory and extensible security protections. This notice + * is included in support of clause 2.2 (b) of the Apple Public License, + * Version 2.0. + */ #include #include @@ -82,166 +104,450 @@ #include #include #include -#if (defined(__FreeBSD__) && __FreeBSD__ >= 3) || defined (__APPLE__) #include -#endif #include +#include +#include +#include +#include +#include + +#include +#include #include #include +#include +#include +#include #include #include -#if defined(__OpenBSD__) || (defined(__bsdi__) && _BSDI_VERSION >= 199802) -#include -#include -#endif +#include +#include #include +#include + +#include #include #include -#if (defined(__FreeBSD__) && __FreeBSD__ >= 3) || defined(__OpenBSD__) || (defined(__bsdi__) && _BSDI_VERSION >= 199802) || defined (__APPLE__) #include -#else -#include -#endif #include - +#include #if IPSEC #include +#include #include -#include +extern int ipsec_bypass; #endif /* IPSEC */ -#ifndef __bsdi__ -#include "loop.h" -#endif +#if NECP +#include +#endif /* NECP */ -#include +#if CONFIG_MACF_NET +#include +#endif /* CONFIG_MACF_NET */ -#if IPV6FIREWALL +#if DUMMYNET #include -#endif - -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 -static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options"); -#endif - -struct ip6_exthdrs { - struct mbuf *ip6e_ip6; - struct mbuf *ip6e_hbh; - struct mbuf *ip6e_dest1; - struct mbuf *ip6e_rthdr; - struct mbuf *ip6e_dest2; +#include +#include +#endif /* DUMMYNET */ + +#if PF +#include +#endif /* PF */ + +static int sysctl_reset_ip6_output_stats SYSCTL_HANDLER_ARGS; +static int sysctl_ip6_output_measure_bins SYSCTL_HANDLER_ARGS; +static int sysctl_ip6_output_getperf SYSCTL_HANDLER_ARGS; +static int ip6_copyexthdr(struct mbuf **, caddr_t, int); +static void ip6_out_cksum_stats(int, u_int32_t); +static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); +static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, + struct ip6_frag **); +static int ip6_getpmtu(struct route_in6 *, struct route_in6 *, + struct ifnet *, struct in6_addr *, u_int32_t *, boolean_t *); +static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, struct socket *, + struct sockopt *sopt); +static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, int); +static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *); +static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int); +static void im6o_trace(struct ip6_moptions *, int); +static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, int, + int, int); +static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); +static void ip6_output_checksum(struct ifnet *, uint32_t, struct mbuf *, + int, uint32_t, uint32_t); +extern int udp_ctloutput(struct socket *, struct sockopt *); +static int ip6_do_fragmentation(struct mbuf **morig, + uint32_t optlen, struct ifnet *ifp, uint32_t unfragpartlen, + struct ip6_hdr *ip6, struct ip6_exthdrs *exthdrsp, uint32_t mtu, + int nxt0); +static int ip6_fragment_packet(struct mbuf **m, + struct ip6_pktopts *opt, struct ip6_exthdrs *exthdrsp, struct ifnet *ifp, + uint32_t mtu, boolean_t alwaysfrag, uint32_t unfragpartlen, + struct route_in6 *ro_pmtu, int nxt0, uint32_t optlen); + +SYSCTL_DECL(_net_inet6_ip6); + +static int ip6_output_measure = 0; +SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, output_perf, + CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, + &ip6_output_measure, 0, sysctl_reset_ip6_output_stats, "I", "Do time measurement"); + +static uint64_t ip6_output_measure_bins = 0; +SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, output_perf_bins, + CTLTYPE_QUAD | CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_output_measure_bins, 0, + sysctl_ip6_output_measure_bins, "I", + "bins for chaining performance data histogram"); + +static net_perf_t net_perf; +SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, output_perf_data, + CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, + 0, 0, sysctl_ip6_output_getperf, "S,net_perf", + "IP6 output performance data (struct net_perf, net/net_perf.h)"); + +#define IM6O_TRACE_HIST_SIZE 32 /* size of trace history */ + +/* For gdb */ +__private_extern__ unsigned int im6o_trace_hist_size = IM6O_TRACE_HIST_SIZE; + +struct ip6_moptions_dbg { + struct ip6_moptions im6o; /* ip6_moptions */ + u_int16_t im6o_refhold_cnt; /* # of IM6O_ADDREF */ + u_int16_t im6o_refrele_cnt; /* # of IM6O_REMREF */ + /* + * Alloc and free callers. + */ + ctrace_t im6o_alloc; + ctrace_t im6o_free; + /* + * Circular lists of IM6O_ADDREF and IM6O_REMREF callers. + */ + ctrace_t im6o_refhold[IM6O_TRACE_HIST_SIZE]; + ctrace_t im6o_refrele[IM6O_TRACE_HIST_SIZE]; }; -static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **, int)); -static int ip6_getpcbopt __P((struct ip6_pktopts *, int, void **, int *)); -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined(__APPLE__) -static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, - struct socket *, struct sockopt *sopt)); +#if DEBUG +static unsigned int im6o_debug = 1; /* debugging (enabled) */ #else -static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *, - struct socket *)); -#endif -static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *)); -static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **)); -static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int)); -static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int, - struct ip6_frag **)); -static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t)); -static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *)); -#if defined(__bsdi__) || defined(__OpenBSD__) -extern struct ifnet loif; -#endif +static unsigned int im6o_debug; /* debugging (disabled) */ +#endif /* !DEBUG */ -#if __NetBSD__ -extern struct ifnet **ifindex2ifnet; -extern struct ifnet loif[NLOOP]; -#endif +static unsigned int im6o_size; /* size of zone element */ +static struct zone *im6o_zone; /* zone for ip6_moptions */ -#if MIP6 -int (*mip6_output_hook)(struct mbuf *m, struct ip6_pktopts **opt); -#endif /* MIP6 */ -static u_long lo_dl_tag = 0; +#define IM6O_ZONE_MAX 64 /* maximum elements in zone */ +#define IM6O_ZONE_NAME "ip6_moptions" /* zone name */ + +/* + * ip6_output() calls ip6_output_list() to do the work + */ +int +ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, + struct route_in6 *ro, int flags, struct ip6_moptions *im6o, + struct ifnet **ifpp, struct ip6_out_args *ip6oa) +{ + return ip6_output_list(m0, 0, opt, ro, flags, im6o, ifpp, ip6oa); +} /* - * IP6 output. The packet in mbuf chain m contains a skeletal IP6 + * IP6 output. Each packet in mbuf chain m contains a skeletal IP6 * header (with pri, len, nxt, hlim, src, dst). * This function may modify ver and hlim only. * The mbuf chain containing the packet will be freed. * The mbuf opt, if present, will not be freed. + * + * If ro is non-NULL and has valid ro->ro_rt, route lookup would be + * skipped and ro->ro_rt would be used. Otherwise the result of route + * lookup is stored in ro->ro_rt. + * + * type of "mtu": rt_rmx.rmx_mtu is u_int32_t, ifnet.ifr_mtu is int, and + * nd_ifinfo.linkmtu is u_int32_t. so we use u_int32_t to hold largest one, + * which is rt_rmx.rmx_mtu. */ int -ip6_output(m0, opt, ro, flags, im6o, ifpp) - struct mbuf *m0; - struct ip6_pktopts *opt; - struct route_in6 *ro; - int flags; - struct ip6_moptions *im6o; - struct ifnet **ifpp; /* XXX: just for statistics */ +ip6_output_list(struct mbuf *m0, int packetchain, struct ip6_pktopts *opt, + struct route_in6 *ro, int flags, struct ip6_moptions *im6o, + struct ifnet **ifpp, struct ip6_out_args *ip6oa) { - struct ip6_hdr *ip6, *mhip6; - struct ifnet *ifp; - struct mbuf *m = m0; - int hlen, tlen, len, off; - struct route_in6 ip6route; - struct sockaddr_in6 *dst; + struct ip6_hdr *ip6; + u_char *nexthdrp; + struct ifnet *ifp = NULL, *origifp = NULL; /* refcnt'd */ + struct ifnet **ifpp_save = ifpp; + struct mbuf *m, *mprev; + struct mbuf *sendchain = NULL, *sendchain_last = NULL; + struct mbuf *inputchain = NULL; + int nxt0; + struct route_in6 *ro_pmtu = NULL; + struct rtentry *rt = NULL; + struct sockaddr_in6 *dst, src_sa, dst_sa; int error = 0; - struct in6_ifaddr *ia; - u_long mtu; + struct in6_ifaddr *ia = NULL, *src_ia = NULL; + u_int32_t mtu; + boolean_t alwaysfrag = FALSE; u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; - struct ip6_exthdrs exthdrs; + struct ip6_rthdr *rh; struct in6_addr finaldst; - struct route_in6 *ro_pmtu = NULL; - int hdrsplit = 0; - int needipsec = 0; + ipfilter_t inject_filter_ref; + struct ipf_pktopts *ippo = NULL; + struct flowadv *adv = NULL; + uint32_t pktcnt = 0; + uint32_t packets_processed = 0; + struct timeval start_tv; +#if DUMMYNET + struct m_tag *tag; + struct ip6_out_args saved_ip6oa; + struct sockaddr_in6 dst_buf; +#endif /* DUMMYNET */ +#if IPSEC + struct socket *so = NULL; + struct secpolicy *sp = NULL; + struct route_in6 *ipsec_saved_route = NULL; + boolean_t needipsectun = FALSE; +#endif /* IPSEC */ +#if NECP + necp_kernel_policy_result necp_result = 0; + necp_kernel_policy_result_parameter necp_result_parameter; + necp_kernel_policy_id necp_matched_policy_id = 0; +#endif /* NECP */ + struct { + struct ipf_pktopts ipf_pktopts; + struct ip6_exthdrs exthdrs; + struct route_in6 ip6route; +#if IPSEC + struct ipsec_output_state ipsec_state; +#endif /* IPSEC */ +#if NECP + struct route_in6 necp_route; +#endif /* NECP */ +#if DUMMYNET + struct route_in6 saved_route; + struct route_in6 saved_ro_pmtu; + struct ip_fw_args args; +#endif /* DUMMYNET */ + } ip6obz; +#define ipf_pktopts ip6obz.ipf_pktopts +#define exthdrs ip6obz.exthdrs +#define ip6route ip6obz.ip6route +#define ipsec_state ip6obz.ipsec_state +#define necp_route ip6obz.necp_route +#define saved_route ip6obz.saved_route +#define saved_ro_pmtu ip6obz.saved_ro_pmtu +#define args ip6obz.args + union { + struct { + boolean_t select_srcif : 1; + boolean_t hdrsplit : 1; + boolean_t route_selected : 1; + boolean_t dontfrag : 1; +#if IPSEC + boolean_t needipsec : 1; + boolean_t noipsec : 1; +#endif /* IPSEC */ + }; + uint32_t raw; + } ip6obf = { .raw = 0 }; + + if (ip6_output_measure) + net_perf_start_time(&net_perf, &start_tv); + + VERIFY(m0->m_flags & M_PKTHDR); + + /* zero out {saved_route, saved_ro_pmtu, ip6route, exthdrs, args} */ + bzero(&ip6obz, sizeof (ip6obz)); + +#if DUMMYNET + if (SLIST_EMPTY(&m0->m_pkthdr.tags)) + goto tags_done; + + /* Grab info from mtags prepended to the chain */ + if ((tag = m_tag_locate(m0, KERNEL_MODULE_TAG_ID, + KERNEL_TAG_TYPE_DUMMYNET, NULL)) != NULL) { + struct dn_pkt_tag *dn_tag; + + /* + * ip6_output_list() cannot handle chains of packets reinjected + * by dummynet. The same restriction applies to + * ip_output_list(). + */ + VERIFY(0 == packetchain); + + dn_tag = (struct dn_pkt_tag *)(tag+1); + args.fwa_pf_rule = dn_tag->dn_pf_rule; + + bcopy(&dn_tag->dn_dst6, &dst_buf, sizeof (dst_buf)); + dst = &dst_buf; + ifp = dn_tag->dn_ifp; + if (ifp != NULL) + ifnet_reference(ifp); + flags = dn_tag->dn_flags; + if (dn_tag->dn_flags & IPV6_OUTARGS) { + saved_ip6oa = dn_tag->dn_ip6oa; + ip6oa = &saved_ip6oa; + } + + saved_route = dn_tag->dn_ro6; + ro = &saved_route; + saved_ro_pmtu = dn_tag->dn_ro6_pmtu; + ro_pmtu = &saved_ro_pmtu; + origifp = dn_tag->dn_origifp; + if (origifp != NULL) + ifnet_reference(origifp); + mtu = dn_tag->dn_mtu; + alwaysfrag = (dn_tag->dn_alwaysfrag != 0); + unfragpartlen = dn_tag->dn_unfragpartlen; + bcopy(&dn_tag->dn_exthdrs, &exthdrs, sizeof (exthdrs)); + + m_tag_delete(m0, tag); + } + +tags_done: +#endif /* DUMMYNET */ + + m = m0; #if IPSEC - int needipsectun = 0; - struct socket *so; - struct secpolicy *sp = NULL; + if (ipsec_bypass == 0) { + so = ipsec_getsocket(m); + if (so != NULL) { + (void) ipsec_setsocket(m, NULL); + } + /* If packet is bound to an interface, check bound policies */ + if ((flags & IPV6_OUTARGS) && + (ip6oa->ip6oa_flags & IPOAF_BOUND_IF) && + ip6oa->ip6oa_boundif != IFSCOPE_NONE) { + /* ip6obf.noipsec is a bitfield, use temp integer */ + int noipsec = 0; + + if (ipsec6_getpolicybyinterface(m, IPSEC_DIR_OUTBOUND, + flags, ip6oa, &noipsec, &sp) != 0) + goto bad; - /* for AH processing. stupid to have "socket" variable in IP layer... */ - so = ipsec_getsocket(m); - ipsec_setsocket(m, NULL); - ip6 = mtod(m, struct ip6_hdr *); + ip6obf.noipsec = (noipsec != 0); + } + } #endif /* IPSEC */ + + ippo = &ipf_pktopts; + + if (ip6_doscopedroute && (flags & IPV6_OUTARGS)) { + /* + * In the forwarding case, only the ifscope value is used, + * as source interface selection doesn't take place. + */ + if ((ip6obf.select_srcif = (!(flags & (IPV6_FORWARDING | + IPV6_UNSPECSRC | IPV6_FLAG_NOSRCIFSEL)) && + (ip6oa->ip6oa_flags & IP6OAF_SELECT_SRCIF)))) + ipf_pktopts.ippo_flags |= IPPOF_SELECT_SRCIF; + + if ((ip6oa->ip6oa_flags & IP6OAF_BOUND_IF) && + ip6oa->ip6oa_boundif != IFSCOPE_NONE) { + ipf_pktopts.ippo_flags |= (IPPOF_BOUND_IF | + (ip6oa->ip6oa_boundif << IPPOF_SHIFT_IFSCOPE)); + } + + if (ip6oa->ip6oa_flags & IP6OAF_BOUND_SRCADDR) + ipf_pktopts.ippo_flags |= IPPOF_BOUND_SRCADDR; + } else { + ip6obf.select_srcif = FALSE; + if (flags & IPV6_OUTARGS) { + ip6oa->ip6oa_boundif = IFSCOPE_NONE; + ip6oa->ip6oa_flags &= ~(IP6OAF_SELECT_SRCIF | + IP6OAF_BOUND_IF | IP6OAF_BOUND_SRCADDR); + } + } + + if (flags & IPV6_OUTARGS) { + if (ip6oa->ip6oa_flags & IP6OAF_NO_CELLULAR) + ipf_pktopts.ippo_flags |= IPPOF_NO_IFT_CELLULAR; + if (ip6oa->ip6oa_flags & IP6OAF_NO_EXPENSIVE) + ipf_pktopts.ippo_flags |= IPPOF_NO_IFF_EXPENSIVE; + adv = &ip6oa->ip6oa_flowadv; + adv->code = FADV_SUCCESS; + ip6oa->ip6oa_retflags = 0; + } + + /* + * Clear out ifpp to be filled in after determining route. ifpp_save is + * used to keep old value to release reference properly and dtrace + * ipsec tunnel traffic properly. + */ + if (ifpp != NULL && *ifpp != NULL) + *ifpp = NULL; + +#if DUMMYNET + if (args.fwa_pf_rule) { + ip6 = mtod(m, struct ip6_hdr *); + VERIFY(ro != NULL); /* ro == saved_route */ + goto check_with_pf; + } +#endif /* DUMMYNET */ + +#if NECP + /* + * Since all packets are assumed to come from same socket, necp lookup + * only needs to happen once per function entry. + */ + necp_matched_policy_id = necp_ip6_output_find_policy_match(m, flags, + (flags & IPV6_OUTARGS) ? ip6oa : NULL, &necp_result, + &necp_result_parameter); +#endif /* NECP */ + + /* + * If a chain was passed in, prepare for ther first iteration. For all + * other iterations, this work will be done at evaluateloop: label. + */ + if (packetchain) { + /* + * Remove m from the chain during processing to avoid + * accidental frees on entire list. + */ + inputchain = m->m_nextpkt; + m->m_nextpkt = NULL; + } + +loopit: + packets_processed++; + m->m_pkthdr.pkt_flags &= ~(PKTF_LOOP|PKTF_IFAINFO); + ip6 = mtod(m, struct ip6_hdr *); + nxt0 = ip6->ip6_nxt; + finaldst = ip6->ip6_dst; + ip6obf.hdrsplit = FALSE; + ro_pmtu = NULL; + + if (!SLIST_EMPTY(&m->m_pkthdr.tags)) + inject_filter_ref = ipf_get_inject_filter(m); + else + inject_filter_ref = NULL; -#define MAKE_EXTHDR(hp,mp) \ - { \ - if (hp) { \ +#define MAKE_EXTHDR(hp, mp) do { \ + if (hp != NULL) { \ struct ip6_ext *eh = (struct ip6_ext *)(hp); \ - error = ip6_copyexthdr((mp), (caddr_t)(hp), \ - ((eh)->ip6e_len + 1) << 3); \ + error = ip6_copyexthdr((mp), (caddr_t)(hp), \ + ((eh)->ip6e_len + 1) << 3); \ if (error) \ goto freehdrs; \ } \ - } - - bzero(&exthdrs, sizeof(exthdrs)); - -#if MIP6 - /* - * Mobile IPv6 - * - * Call Mobile IPv6 to check if there are any Destination Header - * options to add. - */ - if (mip6_output_hook) { - error = (*mip6_output_hook)(m, &opt); - if (error) - goto freehdrs; - } -#endif /* MIP6 */ +} while (0) - if (opt) { + if (opt != NULL) { /* Hop-by-Hop options header */ MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh); + /* Destination options header(1st part) */ if (opt->ip6po_rthdr) { /* * Destination options header(1st part) - * This only makes sence with a routing header. + * This only makes sense with a routing header. + * See Section 9.2 of RFC 3542. + * Disabling this part just for MIP6 convenience is + * a bad idea. We need to think carefully about a + * way to make the advanced API coexist with MIP6 + * options, which might automatically be inserted in + * the kernel. */ MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1); } @@ -251,16 +557,89 @@ ip6_output(m0, opt, ro, flags, im6o, ifpp) MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2); } +#undef MAKE_EXTHDR + +#if NECP + if (necp_matched_policy_id) { + necp_mark_packet_from_ip(m, necp_matched_policy_id); + + switch (necp_result) { + case NECP_KERNEL_POLICY_RESULT_PASS: + goto skip_ipsec; + case NECP_KERNEL_POLICY_RESULT_DROP: + case NECP_KERNEL_POLICY_RESULT_SOCKET_DIVERT: + /* + * Flow divert packets should be blocked at the IP + * layer. + */ + error = EHOSTUNREACH; + goto freehdrs; + case NECP_KERNEL_POLICY_RESULT_IP_TUNNEL: { + /* + * Verify that the packet is being routed to the tunnel + */ + struct ifnet *policy_ifp = + necp_get_ifnet_from_result_parameter( + &necp_result_parameter); + + if (policy_ifp == ifp) { + goto skip_ipsec; + } else { + if (necp_packet_can_rebind_to_ifnet(m, + policy_ifp, (struct route *)&necp_route, + AF_INET6)) { + /* + * Set scoped index to the tunnel + * interface, since it is compatible + * with the packet. This will only work + * for callers who pass IPV6_OUTARGS, + * but that covers all of the clients + * we care about today. + */ + if (flags & IPV6_OUTARGS) { + ip6oa->ip6oa_boundif = + policy_ifp->if_index; + ip6oa->ip6oa_flags |= + IP6OAF_BOUND_IF; + } + if (opt != NULL + && opt->ip6po_pktinfo != NULL) { + opt->ip6po_pktinfo-> + ipi6_ifindex = + policy_ifp->if_index; + } + ro = &necp_route; + goto skip_ipsec; + } else { + error = ENETUNREACH; + goto freehdrs; + } + } + break; + } + default: + break; + } + } +#endif /* NECP */ + #if IPSEC - /* get a security policy for this packet */ - if (so == NULL) - sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error); - else - sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); + if (ipsec_bypass != 0 || ip6obf.noipsec) + goto skip_ipsec; if (sp == NULL) { - ipsec6stat.out_inval++; - goto bad; + /* get a security policy for this packet */ + if (so != NULL) { + sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, + so, &error); + } else { + sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, + 0, &error); + } + if (sp == NULL) { + IPSEC_STAT_INCREMENT(ipsec6stat.out_inval); + goto freehdrs; + } } error = 0; @@ -268,31 +647,38 @@ ip6_output(m0, opt, ro, flags, im6o, ifpp) /* check policy */ switch (sp->policy) { case IPSEC_POLICY_DISCARD: + case IPSEC_POLICY_GENERATE: /* * This packet is just discarded. */ - ipsec6stat.out_polvio++; - goto bad; + IPSEC_STAT_INCREMENT(ipsec6stat.out_polvio); + goto freehdrs; case IPSEC_POLICY_BYPASS: case IPSEC_POLICY_NONE: /* no need to do IPsec. */ - needipsec = 0; + ip6obf.needipsec = FALSE; break; - + case IPSEC_POLICY_IPSEC: if (sp->req == NULL) { /* acquire a policy */ error = key_spdacquire(sp); - goto bad; + goto freehdrs; + } + if (sp->ipsec_if) { + goto skip_ipsec; + } else { + ip6obf.needipsec = TRUE; } - needipsec = 1; break; case IPSEC_POLICY_ENTRUST: default: - printf("ip6_output: Invalid policy found. %d\n", sp->policy); + printf("%s: Invalid policy found: %d\n", __func__, sp->policy); + break; } +skip_ipsec: #endif /* IPSEC */ /* @@ -300,24 +686,33 @@ ip6_output(m0, opt, ro, flags, im6o, ifpp) * Keep the length of the unfragmentable part for fragmentation. */ optlen = 0; - if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len; - if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len; - if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len; - unfragpartlen = optlen + sizeof(struct ip6_hdr); + if (exthdrs.ip6e_hbh != NULL) + optlen += exthdrs.ip6e_hbh->m_len; + if (exthdrs.ip6e_dest1 != NULL) + optlen += exthdrs.ip6e_dest1->m_len; + if (exthdrs.ip6e_rthdr != NULL) + optlen += exthdrs.ip6e_rthdr->m_len; + unfragpartlen = optlen + sizeof (struct ip6_hdr); + /* NOTE: we don't add AH/ESP length here. do that later. */ - if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len; + if (exthdrs.ip6e_dest2 != NULL) + optlen += exthdrs.ip6e_dest2->m_len; /* * If we need IPsec, or there is at least one extension header, * separate IP6 header from the payload. */ - if ((needipsec || optlen) && !hdrsplit) { + if (( +#if IPSEC + ip6obf.needipsec || +#endif /* IPSEC */ + optlen) && !ip6obf.hdrsplit) { if ((error = ip6_splithdr(m, &exthdrs)) != 0) { m = NULL; goto freehdrs; } m = exthdrs.ip6e_ip6; - hdrsplit++; + ip6obf.hdrsplit = TRUE; } /* adjust pointer */ @@ -325,26 +720,26 @@ ip6_output(m0, opt, ro, flags, im6o, ifpp) /* adjust mbuf packet header length */ m->m_pkthdr.len += optlen; - plen = m->m_pkthdr.len - sizeof(*ip6); + plen = m->m_pkthdr.len - sizeof (*ip6); /* If this is a jumbo payload, insert a jumbo payload option. */ if (plen > IPV6_MAXPACKET) { - if (!hdrsplit) { + if (!ip6obf.hdrsplit) { if ((error = ip6_splithdr(m, &exthdrs)) != 0) { m = NULL; goto freehdrs; } m = exthdrs.ip6e_ip6; - hdrsplit++; + ip6obf.hdrsplit = TRUE; } /* adjust pointer */ ip6 = mtod(m, struct ip6_hdr *); if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) goto freehdrs; ip6->ip6_plen = 0; - } else + } else { ip6->ip6_plen = htons(plen); - + } /* * Concatenate headers and fill in next header fields. * Here we have, on "m" @@ -355,54 +750,122 @@ ip6_output(m0, opt, ro, flags, im6o, ifpp) * during the header composing process, "m" points to IPv6 header. * "mprev" points to an extension header prior to esp. */ - { - u_char *nexthdrp = &ip6->ip6_nxt; - struct mbuf *mprev = m; + nexthdrp = &ip6->ip6_nxt; + mprev = m; - /* - * we treat dest2 specially. this makes IPsec processing - * much easier. - * - * result: IPv6 dest2 payload - * m and mprev will point to IPv6 header. - */ - if (exthdrs.ip6e_dest2) { - if (!hdrsplit) - panic("assumption failed: hdr not split"); - exthdrs.ip6e_dest2->m_next = m->m_next; - m->m_next = exthdrs.ip6e_dest2; - *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; - ip6->ip6_nxt = IPPROTO_DSTOPTS; - } - -#define MAKE_CHAIN(m,mp,p,i)\ - {\ - if (m) {\ - if (!hdrsplit) \ - panic("assumption failed: hdr not split"); \ - *mtod((m), u_char *) = *(p);\ - *(p) = (i);\ - p = mtod((m), u_char *);\ - (m)->m_next = (mp)->m_next;\ - (mp)->m_next = (m);\ - (mp) = (m);\ - }\ - } - /* - * result: IPv6 hbh dest1 rthdr dest2 payload - * m will point to IPv6 header. mprev will point to the - * extension header prior to dest2 (rthdr in the above case). - */ - MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, - nexthdrp, IPPROTO_HOPOPTS); - MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, - nexthdrp, IPPROTO_DSTOPTS); - MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, - nexthdrp, IPPROTO_ROUTING); + /* + * we treat dest2 specially. this makes IPsec processing + * much easier. the goal here is to make mprev point the + * mbuf prior to dest2. + * + * result: IPv6 dest2 payload + * m and mprev will point to IPv6 header. + */ + if (exthdrs.ip6e_dest2 != NULL) { + if (!ip6obf.hdrsplit) { + panic("assumption failed: hdr not split"); + /* NOTREACHED */ + } + exthdrs.ip6e_dest2->m_next = m->m_next; + m->m_next = exthdrs.ip6e_dest2; + *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt; + ip6->ip6_nxt = IPPROTO_DSTOPTS; + } + +#define MAKE_CHAIN(m, mp, p, i) do { \ + if (m != NULL) { \ + if (!ip6obf.hdrsplit) { \ + panic("assumption failed: hdr not split"); \ + /* NOTREACHED */ \ + } \ + *mtod((m), u_char *) = *(p); \ + *(p) = (i); \ + p = mtod((m), u_char *); \ + (m)->m_next = (mp)->m_next; \ + (mp)->m_next = (m); \ + (mp) = (m); \ + } \ +} while (0) + /* + * result: IPv6 hbh dest1 rthdr dest2 payload + * m will point to IPv6 header. mprev will point to the + * extension header prior to dest2 (rthdr in the above case). + */ + MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS); + MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, IPPROTO_DSTOPTS); + MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, IPPROTO_ROUTING); + + /* It is no longer safe to free the pointers in exthdrs. */ + exthdrs.merged = TRUE; + +#undef MAKE_CHAIN + +#if IPSEC + if (ip6obf.needipsec && (m->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA)) + in6_delayed_cksum_offset(m, 0, optlen, nxt0); +#endif /* IPSEC */ + + if (!TAILQ_EMPTY(&ipv6_filters)) { + struct ipfilter *filter; + int seen = (inject_filter_ref == NULL); + int fixscope = 0; + + if (im6o != NULL && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { + ippo->ippo_flags |= IPPOF_MCAST_OPTS; + IM6O_LOCK(im6o); + ippo->ippo_mcast_ifnet = im6o->im6o_multicast_ifp; + ippo->ippo_mcast_ttl = im6o->im6o_multicast_hlim; + ippo->ippo_mcast_loop = im6o->im6o_multicast_loop; + IM6O_UNLOCK(im6o); + } + + /* Hack: embed the scope_id in the destination */ + if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) && + (ip6->ip6_dst.s6_addr16[1] == 0) && (ro != NULL)) { + fixscope = 1; + ip6->ip6_dst.s6_addr16[1] = + htons(ro->ro_dst.sin6_scope_id); + } + + ipf_ref(); + TAILQ_FOREACH(filter, &ipv6_filters, ipf_link) { + /* + * Don't process packet twice if we've already seen it. + */ + if (seen == 0) { + if ((struct ipfilter *)inject_filter_ref == + filter) + seen = 1; + } else if (filter->ipf_filter.ipf_output != NULL) { + errno_t result; + + result = filter->ipf_filter.ipf_output( + filter->ipf_filter.cookie, + (mbuf_t *)&m, ippo); + if (result == EJUSTRETURN) { + ipf_unref(); + if (m != NULL) + m_freem(m); + m = NULL; + goto evaluateloop; + } + if (result != 0) { + ipf_unref(); + goto bad; + } + } + } + ipf_unref(); + + ip6 = mtod(m, struct ip6_hdr *); + /* Hack: cleanup embedded scope_id if we put it there */ + if (fixscope) + ip6->ip6_dst.s6_addr16[1] = 0; + } #if IPSEC - if (!needipsec) - goto skip_ipsec2; + if (ip6obf.needipsec) { + int segleft_org; /* * pointers after IPsec headers are not valid any more. @@ -411,22 +874,19 @@ ip6_output(m0, opt, ro, flags, im6o, ifpp) */ exthdrs.ip6e_dest2 = NULL; - { - struct ip6_rthdr *rh = NULL; - int segleft_org = 0; - struct ipsec_output_state state; - - if (exthdrs.ip6e_rthdr) { + if (exthdrs.ip6e_rthdr != NULL) { rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *); segleft_org = rh->ip6r_segleft; rh->ip6r_segleft = 0; + } else { + rh = NULL; + segleft_org = 0; } - bzero(&state, sizeof(state)); - state.m = m; - error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags, - &needipsectun); - m = state.m; + ipsec_state.m = m; + error = ipsec6_output_trans(&ipsec_state, nexthdrp, mprev, + sp, flags, &needipsectun); + m = ipsec_state.m; if (error) { /* mbuf is already reclaimed in ipsec6_output_trans. */ m = NULL; @@ -438,8 +898,9 @@ ip6_output(m0, opt, ro, flags, im6o, ifpp) case ENOMEM: break; default: - printf("ip6_output (ipsec): error code %d\n", error); - /*fall through*/ + printf("ip6_output (ipsec): error code %d\n", + error); + /* FALLTHRU */ case ENOENT: /* don't show these error codes to the user */ error = 0; @@ -447,47 +908,63 @@ ip6_output(m0, opt, ro, flags, im6o, ifpp) } goto bad; } - if (exthdrs.ip6e_rthdr) { + if (exthdrs.ip6e_rthdr != NULL) { /* ah6_output doesn't modify mbuf chain */ rh->ip6r_segleft = segleft_org; } - } -skip_ipsec2:; -#endif } +#endif /* IPSEC */ /* - * If there is a routing header, replace destination address field + * If there is a routing header, replace the destination address field * with the first hop of the routing header. */ - if (exthdrs.ip6e_rthdr) { - struct ip6_rthdr *rh = - (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr, - struct ip6_rthdr *)); + if (exthdrs.ip6e_rthdr != NULL) { struct ip6_rthdr0 *rh0; struct in6_addr *addr; + struct sockaddr_in6 sa; - finaldst = ip6->ip6_dst; - switch(rh->ip6r_type) { + rh = (struct ip6_rthdr *) + (mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *)); + switch (rh->ip6r_type) { case IPV6_RTHDR_TYPE_0: - rh0 = (struct ip6_rthdr0 *)rh; - addr = (struct in6_addr *)(rh0 + 1); - - ip6->ip6_dst = *addr; - bcopy((caddr_t)(addr + 1), (caddr_t)addr, - sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1) - ); - *(addr + rh0->ip6r0_segleft - 1) = finaldst; - break; + rh0 = (struct ip6_rthdr0 *)rh; + addr = (struct in6_addr *)(void *)(rh0 + 1); + + /* + * construct a sockaddr_in6 form of + * the first hop. + * + * XXX: we may not have enough + * information about its scope zone; + * there is no standard API to pass + * the information from the + * application. + */ + bzero(&sa, sizeof (sa)); + sa.sin6_family = AF_INET6; + sa.sin6_len = sizeof (sa); + sa.sin6_addr = addr[0]; + if ((error = sa6_embedscope(&sa, + ip6_use_defzone)) != 0) { + goto bad; + } + ip6->ip6_dst = sa.sin6_addr; + bcopy(&addr[1], &addr[0], sizeof (struct in6_addr) * + (rh0->ip6r0_segleft - 1)); + addr[rh0->ip6r0_segleft - 1] = finaldst; + /* XXX */ + in6_clearscope(addr + rh0->ip6r0_segleft - 1); + break; default: /* is it possible? */ - error = EINVAL; - goto bad; + error = EINVAL; + goto bad; } } /* Source address validation */ if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && - (flags & IPV6_DADOUTPUT) == 0) { + !(flags & IPV6_UNSPECSRC)) { error = EOPNOTSUPP; ip6stat.ip6s_badscope++; goto bad; @@ -503,34 +980,71 @@ skip_ipsec2:; /* * Route packet. */ - if (ro == 0) { + if (ro == NULL) { ro = &ip6route; - bzero((caddr_t)ro, sizeof(*ro)); + bzero((caddr_t)ro, sizeof (*ro)); } ro_pmtu = ro; - if (opt && opt->ip6po_rthdr) + if (opt != NULL && opt->ip6po_rthdr) ro = &opt->ip6po_route; - dst = (struct sockaddr_in6 *)&ro->ro_dst; + dst = SIN6(&ro->ro_dst); + + if (ro->ro_rt != NULL) + RT_LOCK_ASSERT_NOTHELD(ro->ro_rt); /* - * If there is a cached route, - * check that it is to the same destination - * and is still up. If not, free it and try again. + * if specified, try to fill in the traffic class field. + * do not override if a non-zero value is already set. + * we check the diffserv field and the ecn field separately. */ - if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || - !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) { - RTFREE(ro->ro_rt); - ro->ro_rt = (struct rtentry *)0; + if (opt != NULL && opt->ip6po_tclass >= 0) { + int mask = 0; + + if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) + mask |= 0xfc; + if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) + mask |= 0x03; + if (mask != 0) { + ip6->ip6_flow |= + htonl((opt->ip6po_tclass & mask) << 20); + } + } + + /* fill in or override the hop limit field, if necessary. */ + if (opt && opt->ip6po_hlim != -1) { + ip6->ip6_hlim = opt->ip6po_hlim & 0xff; + } else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { + if (im6o != NULL) { + IM6O_LOCK(im6o); + ip6->ip6_hlim = im6o->im6o_multicast_hlim; + IM6O_UNLOCK(im6o); + } else { + ip6->ip6_hlim = ip6_defmcasthlim; + } } - if (ro->ro_rt == 0) { - bzero(dst, sizeof(*dst)); + + /* + * If there is a cached route, check that it is to the same + * destination and is still up. If not, free it and try again. + * Test rt_flags without holding rt_lock for performance reasons; + * if the route is down it will hopefully be caught by the layer + * below (since it uses this route as a hint) or during the + * next transmit. + */ + if (ROUTE_UNUSABLE(ro) || dst->sin6_family != AF_INET6 || + !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst)) + ROUTE_RELEASE(ro); + + if (ro->ro_rt == NULL) { + bzero(dst, sizeof (*dst)); dst->sin6_family = AF_INET6; - dst->sin6_len = sizeof(struct sockaddr_in6); + dst->sin6_len = sizeof (struct sockaddr_in6); dst->sin6_addr = ip6->ip6_dst; } #if IPSEC - if (needipsec && needipsectun) { - struct ipsec_output_state state; - + if (ip6obf.needipsec && needipsectun) { +#if CONFIG_DTRACE + struct ifnet *trace_ifp = (ifpp_save != NULL) ? (*ifpp_save) : NULL; +#endif /* CONFIG_DTRACE */ /* * All the extension headers will become inaccessible * (since they can be encrypted). @@ -539,22 +1053,31 @@ skip_ipsec2:; * * IPv6 [ESP|AH] IPv6 [extension headers] payload */ - bzero(&exthdrs, sizeof(exthdrs)); + bzero(&exthdrs, sizeof (exthdrs)); exthdrs.ip6e_ip6 = m; - bzero(&state, sizeof(state)); - state.m = m; - state.ro = (struct route *)ro; - state.dst = (struct sockaddr *)dst; + ipsec_state.m = m; + route_copyout(&ipsec_state.ro, (struct route *)ro, + sizeof (ipsec_state.ro)); + ipsec_state.dst = SA(dst); - error = ipsec6_output_tunnel(&state, sp, flags); + /* So that we can see packets inside the tunnel */ + DTRACE_IP6(send, struct mbuf *, m, struct inpcb *, NULL, + struct ip6_hdr *, ip6, struct ifnet *, trace_ifp, + struct ip *, NULL, struct ip6_hdr *, ip6); - m = state.m; - ro = (struct route_in6 *)state.ro; - dst = (struct sockaddr_in6 *)state.dst; + error = ipsec6_output_tunnel(&ipsec_state, sp, flags); + /* tunneled in IPv4? packet is gone */ + if (ipsec_state.tunneled == 4) { + m = NULL; + goto evaluateloop; + } + m = ipsec_state.m; + ipsec_saved_route = ro; + ro = (struct route_in6 *)&ipsec_state.ro; + dst = SIN6(ipsec_state.dst); if (error) { /* mbuf is already reclaimed in ipsec6_output_tunnel. */ - m0 = m = NULL; m = NULL; switch (error) { case EHOSTUNREACH: @@ -564,8 +1087,9 @@ skip_ipsec2:; case ENOMEM: break; default: - printf("ip6_output (ipsec): error code %d\n", error); - /*fall through*/ + printf("ip6_output (ipsec): error code %d\n", + error); + /* FALLTHRU */ case ENOENT: /* don't show these error codes to the user */ error = 0; @@ -573,191 +1097,234 @@ skip_ipsec2:; } goto bad; } - - exthdrs.ip6e_ip6 = m; - } -#endif /*IPSEC*/ - - if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { - /* Unicast */ - -#define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa)) -#define sin6tosa(sin6) ((struct sockaddr *)(sin6)) - /* xxx - * interface selection comes here - * if an interface is specified from an upper layer, - * ifp must point it. + /* + * The packet has been encapsulated so the ifscope + * is no longer valid since it does not apply to the + * outer address: ignore the ifscope. */ - if (ro->ro_rt == 0) { -#ifndef __bsdi__ - /* - * non-bsdi always clone routes, if parent is - * PRF_CLONING. - */ - rtalloc((struct route *)ro); -#else - if (ro == &ip6route) /* xxx kazu */ - rtalloc((struct route *)ro); - else - rtcalloc((struct route *)ro); -#endif + if (flags & IPV6_OUTARGS) { + ip6oa->ip6oa_boundif = IFSCOPE_NONE; + ip6oa->ip6oa_flags &= ~IP6OAF_BOUND_IF; } - if (ro->ro_rt == 0) { - ip6stat.ip6s_noroute++; - error = EHOSTUNREACH; - /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */ - goto bad; + if (opt != NULL && opt->ip6po_pktinfo != NULL) { + if (opt->ip6po_pktinfo->ipi6_ifindex != IFSCOPE_NONE) + opt->ip6po_pktinfo->ipi6_ifindex = IFSCOPE_NONE; } - ia = ifatoia6(ro->ro_rt->rt_ifa); - ifp = ro->ro_rt->rt_ifp; - ro->ro_rt->rt_use++; - if (ro->ro_rt->rt_flags & RTF_GATEWAY) - dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway; - m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ + exthdrs.ip6e_ip6 = m; + } +#endif /* IPSEC */ - in6_ifstat_inc(ifp, ifs6_out_request); + /* + * ifp should only be filled in for dummy net packets which will jump + * to check_with_pf label. + */ + if (ifp != NULL) { + VERIFY(ip6obf.route_selected); + } - /* - * Check if the outgoing interface conflicts with - * the interface specified by ifi6_ifindex (if specified). - * Note that loopback interface is always okay. - * (this may happen when we are sending a packet to one of - * our own addresses.) - */ - if (opt && opt->ip6po_pktinfo - && opt->ip6po_pktinfo->ipi6_ifindex) { - if (!(ifp->if_flags & IFF_LOOPBACK) - && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) { - ip6stat.ip6s_noroute++; - in6_ifstat_inc(ifp, ifs6_out_discard); - error = EHOSTUNREACH; - goto bad; - } - } + /* adjust pointer */ + ip6 = mtod(m, struct ip6_hdr *); - if (opt && opt->ip6po_hlim != -1) - ip6->ip6_hlim = opt->ip6po_hlim & 0xff; - } else { - /* Multicast */ - struct in6_multi *in6m; + if (ip6obf.select_srcif) { + bzero(&src_sa, sizeof (src_sa)); + src_sa.sin6_family = AF_INET6; + src_sa.sin6_len = sizeof (src_sa); + src_sa.sin6_addr = ip6->ip6_src; + } + bzero(&dst_sa, sizeof (dst_sa)); + dst_sa.sin6_family = AF_INET6; + dst_sa.sin6_len = sizeof (dst_sa); + dst_sa.sin6_addr = ip6->ip6_dst; - m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; - - /* - * See if the caller provided any multicast options - */ - ifp = NULL; - if (im6o != NULL) { - ip6->ip6_hlim = im6o->im6o_multicast_hlim; - if (im6o->im6o_multicast_ifp != NULL) - ifp = im6o->im6o_multicast_ifp; - } else - ip6->ip6_hlim = ip6_defmcasthlim; + /* + * Only call in6_selectroute() on first iteration to avoid taking + * multiple references on ifp and rt. + * + * in6_selectroute() might return an ifp with its reference held + * even in the error case, so make sure to release its reference. + * ip6oa may be NULL if IPV6_OUTARGS isn't set. + */ + if (!ip6obf.route_selected) { + error = in6_selectroute( ip6obf.select_srcif ? &src_sa : NULL, + &dst_sa, opt, im6o, &src_ia, ro, &ifp, &rt, 0, ip6oa); + if (error != 0) { + switch (error) { + case EHOSTUNREACH: + ip6stat.ip6s_noroute++; + break; + case EADDRNOTAVAIL: + default: + break; /* XXX statistics? */ + } + if (ifp != NULL) + in6_ifstat_inc(ifp, ifs6_out_discard); + /* ifp (if non-NULL) will be released at the end */ + goto bad; + } + ip6obf.route_selected = TRUE; + } + if (rt == NULL) { /* - * See if the caller provided the outgoing interface - * as an ancillary data. - * Boundary check for ifindex is assumed to be already done. + * If in6_selectroute() does not return a route entry, + * dst may not have been updated. */ - if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex) - ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex]; + *dst = dst_sa; /* XXX */ + } - /* - * If the destination is a node-local scope multicast, - * the packet should be loop-backed only. - */ - if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) { - /* - * If the outgoing interface is already specified, - * it should be a loopback interface. - */ - if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) { - ip6stat.ip6s_badscope++; - error = ENETUNREACH; /* XXX: better error? */ - /* XXX correct ifp? */ - in6_ifstat_inc(ifp, ifs6_out_discard); - goto bad; - } else { -#ifdef __bsdi__ - ifp = loifp; -#else - ifp = &loif[0]; -#endif - } +#if NECP + /* Catch-all to check if the interface is allowed */ + if (!necp_packet_is_allowed_over_interface(m, ifp)) { + error = EHOSTUNREACH; + goto bad; + } +#endif /* NECP */ + + /* + * then rt (for unicast) and ifp must be non-NULL valid values. + */ + if (!(flags & IPV6_FORWARDING)) { + in6_ifstat_inc_na(ifp, ifs6_out_request); + } + if (rt != NULL) { + RT_LOCK(rt); + if (ia == NULL) { + ia = (struct in6_ifaddr *)(rt->rt_ifa); + if (ia != NULL) + IFA_ADDREF(&ia->ia_ifa); } + rt->rt_use++; + RT_UNLOCK(rt); + } + + /* + * The outgoing interface must be in the zone of source and + * destination addresses (except local/loopback). We should + * use ia_ifp to support the case of sending packets to an + * address of our own. + */ + if (ia != NULL && ia->ia_ifp) { + ifnet_reference(ia->ia_ifp); /* for origifp */ + if (origifp != NULL) + ifnet_release(origifp); + origifp = ia->ia_ifp; + } else { + if (ifp != NULL) + ifnet_reference(ifp); /* for origifp */ + if (origifp != NULL) + ifnet_release(origifp); + origifp = ifp; + } - if (opt && opt->ip6po_hlim != -1) - ip6->ip6_hlim = opt->ip6po_hlim & 0xff; + /* skip scope enforcements for local/loopback route */ + if (rt == NULL || !(rt->rt_ifp->if_flags & IFF_LOOPBACK)) { + struct in6_addr src0, dst0; + u_int32_t zone; + + src0 = ip6->ip6_src; + if (in6_setscope(&src0, origifp, &zone)) + goto badscope; + bzero(&src_sa, sizeof (src_sa)); + src_sa.sin6_family = AF_INET6; + src_sa.sin6_len = sizeof (src_sa); + src_sa.sin6_addr = ip6->ip6_src; + if ((sa6_recoverscope(&src_sa, TRUE) || + zone != src_sa.sin6_scope_id)) + goto badscope; + + dst0 = ip6->ip6_dst; + if ((in6_setscope(&dst0, origifp, &zone))) + goto badscope; + /* re-initialize to be sure */ + bzero(&dst_sa, sizeof (dst_sa)); + dst_sa.sin6_family = AF_INET6; + dst_sa.sin6_len = sizeof (dst_sa); + dst_sa.sin6_addr = ip6->ip6_dst; + if ((sa6_recoverscope(&dst_sa, TRUE) || + zone != dst_sa.sin6_scope_id)) + goto badscope; + + /* scope check is done. */ + goto routefound; + +badscope: + ip6stat.ip6s_badscope++; + in6_ifstat_inc(origifp, ifs6_out_discard); + if (error == 0) + error = EHOSTUNREACH; /* XXX */ + goto bad; + } +routefound: + if (rt != NULL && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { + if (opt != NULL && opt->ip6po_nextroute.ro_rt) { + /* + * The nexthop is explicitly specified by the + * application. We assume the next hop is an IPv6 + * address. + */ + dst = SIN6(opt->ip6po_nexthop); + } else if ((rt->rt_flags & RTF_GATEWAY)) { + dst = SIN6(rt->rt_gateway); + } /* - * If caller did not provide an interface lookup a - * default in the routing table. This is either a - * default for the speicfied group (i.e. a host - * route), or a multicast default (a route for the - * ``net'' ff00::/8). + * For packets destined to local/loopback, record the + * source the source interface (which owns the source + * address), as well as the output interface. This is + * needed to reconstruct the embedded zone for the + * link-local address case in ip6_input(). */ - if (ifp == NULL) { - if (ro->ro_rt == 0) { - ro->ro_rt = rtalloc1((struct sockaddr *) - &ro->ro_dst, 0 -#if __FreeBSD__ || defined (__APPLE__) - , 0UL -#endif - ); - } - if (ro->ro_rt == 0) { - ip6stat.ip6s_noroute++; - error = EHOSTUNREACH; - /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */ - goto bad; - } - ia = ifatoia6(ro->ro_rt->rt_ifa); - ifp = ro->ro_rt->rt_ifp; - ro->ro_rt->rt_use++; + if (ia != NULL && (ifp->if_flags & IFF_LOOPBACK)) { + uint32_t srcidx; + + if (src_ia != NULL) + srcidx = src_ia->ia_ifp->if_index; + else if (ro->ro_srcia != NULL) + srcidx = ro->ro_srcia->ifa_ifp->if_index; + else + srcidx = 0; + + ip6_setsrcifaddr_info(m, srcidx, NULL); + ip6_setdstifaddr_info(m, 0, ia); } + } + + if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { + m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ + } else { + struct in6_multi *in6m; - if ((flags & IPV6_FORWARDING) == 0) - in6_ifstat_inc(ifp, ifs6_out_request); - in6_ifstat_inc(ifp, ifs6_out_mcast); + m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; + in6_ifstat_inc_na(ifp, ifs6_out_mcast); /* * Confirm that the outgoing interface supports multicast. */ - if ((ifp->if_flags & IFF_MULTICAST) == 0) { + if (!(ifp->if_flags & IFF_MULTICAST)) { ip6stat.ip6s_noroute++; in6_ifstat_inc(ifp, ifs6_out_discard); error = ENETUNREACH; goto bad; } - IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m); + in6_multihead_lock_shared(); + IN6_LOOKUP_MULTI(&ip6->ip6_dst, ifp, in6m); + in6_multihead_lock_done(); + if (im6o != NULL) + IM6O_LOCK(im6o); if (in6m != NULL && - (im6o == NULL || im6o->im6o_multicast_loop)) { + (im6o == NULL || im6o->im6o_multicast_loop)) { + if (im6o != NULL) + IM6O_UNLOCK(im6o); /* * If we belong to the destination multicast group * on the outgoing interface, and the caller did not * forbid loopback, loop back a copy. */ - ip6_mloopback(ifp, m, dst); - } else { - /* - * If we are acting as a multicast router, perform - * multicast forwarding as if the packet had just - * arrived on the interface to which we are about - * to send. The multicast forwarding function - * recursively calls this function, using the - * IPV6_FORWARDING flag to prevent infinite recursion. - * - * Multicasts that are looped back by ip6_mloopback(), - * above, will be forwarded by the ip6_input() routine, - * if necessary. - */ - if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { - if (ip6_mforward(ip6, ifp, m) != NULL) { - m_freem(m); - goto done; - } - } - } + ip6_mloopback(NULL, ifp, m, dst, optlen, nxt0); + } else if (im6o != NULL) + IM6O_UNLOCK(im6o); + if (in6m != NULL) + IN6M_REMREF(in6m); /* * Multicasts with a hoplimit of zero may be looped back, * above, but must not be transmitted on a network. @@ -766,9 +1333,13 @@ skip_ipsec2:; * loop back a copy if this host actually belongs to the * destination group on the loopback interface. */ - if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) { - m_freem(m); - goto done; + if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || + IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { + /* remove m from the packetchain and continue looping */ + if (m != NULL) + m_freem(m); + m = NULL; + goto evaluateloop; } } @@ -776,187 +1347,417 @@ skip_ipsec2:; * Fill the outgoing inteface to tell the upper layer * to increment per-interface statistics. */ - if (ifpp) + if (ifpp != NULL && *ifpp == NULL) { + ifnet_reference(ifp); /* for caller */ *ifpp = ifp; - - /* - * Upper-layer reachability confirmation - */ - if (opt && (opt->ip6po_flags & IP6PO_REACHCONF)) - nd6_nud_hint(ro->ro_rt, NULL); - - /* - * Determine path MTU. - */ - if (ro_pmtu != ro) { - /* The first hop and the final destination may differ. */ - struct sockaddr_in6 *sin6_fin = - (struct sockaddr_in6 *)&ro_pmtu->ro_dst; - if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || - !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr, - &finaldst))) { - RTFREE(ro_pmtu->ro_rt); - ro_pmtu->ro_rt = (struct rtentry *)0; - } - if (ro_pmtu->ro_rt == 0) { - bzero(sin6_fin, sizeof(*sin6_fin)); - sin6_fin->sin6_family = AF_INET6; - sin6_fin->sin6_len = sizeof(struct sockaddr_in6); - sin6_fin->sin6_addr = finaldst; - -#ifdef __bsdi__ /* bsdi needs rtcalloc to clone a route. */ - rtcalloc((struct route *)ro_pmtu); -#else - rtalloc((struct route *)ro_pmtu); -#endif - } } - if (ro_pmtu->ro_rt != NULL) { - u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu; - mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; - if (mtu > ifmtu) { - /* - * The MTU on the route is larger than the MTU on - * the interface! This shouldn't happen, unless the - * MTU of the interface has been changed after the - * interface was brought up. Change the MTU in the - * route to match the interface MTU (as long as the - * field isn't locked). - */ - mtu = ifmtu; - if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0) - ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */ - } - } else { - mtu = nd_ifinfo[ifp->if_index].linkmtu; - } + /* Determine path MTU. */ + if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, + &alwaysfrag)) != 0) + goto bad; /* - * advanced API (IPV6_USE_MIN_MTU) overrides mtu setting + * The caller of this function may specify to use the minimum MTU + * in some cases. + * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU + * setting. The logic is a bit complicated; by default, unicast + * packets will follow path MTU while multicast packets will be sent at + * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets + * including unicast ones will be sent at the minimum MTU. Multicast + * packets will always be sent at the minimum MTU unless + * IP6PO_MINMTU_DISABLE is explicitly specified. + * See RFC 3542 for more details. */ if (mtu > IPV6_MMTU) { - if ((opt && (opt->ip6po_flags & IP6PO_MINMTU)) || - (flags & IPV6_MINMTU)) { + if ((flags & IPV6_MINMTU)) { + mtu = IPV6_MMTU; + } else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) { + mtu = IPV6_MMTU; + } else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && + (opt == NULL || + opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { mtu = IPV6_MMTU; } } /* - * Fake link-local scope-class addresses + * clear embedded scope identifiers if necessary. + * in6_clearscope will touch the addresses only when necessary. */ - if ((ifp->if_flags & IFF_LOOPBACK) == 0) { - if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) - ip6->ip6_src.s6_addr16[1] = 0; - if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) - ip6->ip6_dst.s6_addr16[1] = 0; - } + in6_clearscope(&ip6->ip6_src); + in6_clearscope(&ip6->ip6_dst); -#if IPV6FIREWALL +#if IPFW2 /* * Check with the firewall... */ - if (ip6_fw_chk_ptr) { + if (ip6_fw_enable && ip6_fw_chk_ptr) { u_short port = 0; + m->m_pkthdr.rcvif = NULL; /* XXX */ /* If ipfw says divert, we have to just drop packet */ - if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) { - m_freem(m); - goto done; - } - if (!m) { - error = EACCES; - goto done; + if (ip6_fw_chk_ptr(&ip6, ifp, &port, &m) || m == NULL) { + if (m != NULL) { + m_freem(m); + m = NULL; + goto evaluateloop; + } else { + error = EACCES; + goto bad; + } } } -#endif +#endif /* IPFW2 */ /* * If the outgoing packet contains a hop-by-hop options header, * it must be examined and processed even by the source node. * (RFC 2460, section 4.) */ - if (exthdrs.ip6e_hbh) { - struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, - struct ip6_hbh *); - u_int32_t dummy1; /* XXX unused */ - u_int32_t dummy2; /* XXX unused */ - + if (exthdrs.ip6e_hbh != NULL) { + struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *); + u_int32_t dummy; /* XXX unused */ + uint32_t oplen = 0; /* for ip6_process_hopopts() */ +#if DIAGNOSTIC + if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len) + panic("ip6e_hbh is not continuous"); +#endif /* - * XXX: if we have to send an ICMPv6 error to the sender, - * we need the M_LOOP flag since icmp6_error() expects - * the IPv6 and the hop-by-hop options header are - * continuous unless the flag is set. + * XXX: If we have to send an ICMPv6 error to the sender, + * we need the M_LOOP flag since icmp6_error() expects + * the IPv6 and the hop-by-hop options header are + * continuous unless the flag is set. */ m->m_flags |= M_LOOP; m->m_pkthdr.rcvif = ifp; - if (ip6_process_hopopts(m, - (u_int8_t *)(hbh + 1), - ((hbh->ip6h_len + 1) << 3) - - sizeof(struct ip6_hbh), - &dummy1, &dummy2) < 0) { - /* m was already freed at this point */ - error = EINVAL;/* better error? */ - goto done; + if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1), + ((hbh->ip6h_len + 1) << 3) - sizeof (struct ip6_hbh), + &dummy, &oplen) < 0) { + /* + * m was already freed at this point. Set to NULL so it + * is not re-freed at end of ip6_output_list. + */ + m = NULL; + error = EINVAL; /* better error? */ + goto bad; } m->m_flags &= ~M_LOOP; /* XXX */ m->m_pkthdr.rcvif = NULL; } +#if DUMMYNET +check_with_pf: +#endif /* DUMMYNET */ +#if PF + if (PF_IS_ENABLED) { +#if DUMMYNET + + /* + * TODO: Need to save opt->ip6po_flags for reinjection + * rdar://10434993 + */ + args.fwa_m = m; + args.fwa_oif = ifp; + args.fwa_oflags = flags; + if (flags & IPV6_OUTARGS) + args.fwa_ip6oa = ip6oa; + args.fwa_ro6 = ro; + args.fwa_dst6 = dst; + args.fwa_ro6_pmtu = ro_pmtu; + args.fwa_origifp = origifp; + args.fwa_mtu = mtu; + args.fwa_alwaysfrag = alwaysfrag; + args.fwa_unfragpartlen = unfragpartlen; + args.fwa_exthdrs = &exthdrs; + /* Invoke outbound packet filter */ + error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, &args); +#else /* !DUMMYNET */ + error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, NULL); +#endif /* !DUMMYNET */ + + if (error != 0 || m == NULL) { + if (m != NULL) { + panic("%s: unexpected packet %p\n", + __func__, m); + /* NOTREACHED */ + } + /* m was already freed by callee and is now NULL. */ + goto evaluateloop; + } + ip6 = mtod(m, struct ip6_hdr *); + } +#endif /* PF */ + +#ifdef IPSEC + /* clean ipsec history before fragmentation */ + ipsec_delaux(m); +#endif /* IPSEC */ + /* - * Send the packet to the outgoing interface. - * If necessary, do IPv6 fragmentation before sending. + * Determine whether fragmentation is necessary. If so, m is passed + * back as a chain of packets and original mbuf is freed. Otherwise, m + * is unchanged. */ - tlen = m->m_pkthdr.len; - if (tlen <= mtu -#if notyet - /* - * On any link that cannot convey a 1280-octet packet in one piece, - * link-specific fragmentation and reassembly must be provided at - * a layer below IPv6. [RFC 2460, sec.5] - * Thus if the interface has ability of link-level fragmentation, - * we can just send the packet even if the packet size is - * larger than the link's MTU. - * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet... - */ - - || ifp->if_flags & IFF_FRAGMENTABLE -#endif - ) - { -#if defined(__NetBSD__) && defined(IFA_STATS) - if (IFA_STATS) { - struct in6_ifaddr *ia6; - ip6 = mtod(m, struct ip6_hdr *); - ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); - if (ia6) { - ia->ia_ifa.ifa_data.ifad_outbytes += - m->m_pkthdr.len; - } + error = ip6_fragment_packet(&m, opt, + &exthdrs, ifp, mtu, alwaysfrag, unfragpartlen, ro_pmtu, nxt0, + optlen); + + if (error) + goto bad; + +/* + * The evaluateloop label is where we decide whether to continue looping over + * packets or call into nd code to send. + */ +evaluateloop: + + /* + * m may be NULL when we jump to the evaluateloop label from PF or + * other code that can drop packets. + */ + if (m != NULL) { + /* + * If we already have a chain to send, tack m onto the end. + * Otherwise make m the start and end of the to-be-sent chain. + */ + if (sendchain != NULL) { + sendchain_last->m_nextpkt = m; + } else { + sendchain = m; } -#endif -#if OLDIP6OUTPUT - error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, - ro->ro_rt); -#else - error = nd6_output(ifp, m, dst, ro->ro_rt); -#endif - goto done; - } else if (mtu < IPV6_MMTU) { + + /* Fragmentation may mean m is a chain. Find the last packet. */ + while (m->m_nextpkt) + m = m->m_nextpkt; + sendchain_last = m; + pktcnt++; + } + + /* Fill in next m from inputchain as appropriate. */ + m = inputchain; + if (m != NULL) { + /* Isolate m from rest of input chain. */ + inputchain = m->m_nextpkt; + m->m_nextpkt = NULL; + + /* + * Clear exthdrs and ipsec_state so stale contents are not + * reused. Note this also clears the exthdrs.merged flag. + */ + bzero(&exthdrs, sizeof(exthdrs)); + bzero(&ipsec_state, sizeof(ipsec_state)); + + /* Continue looping. */ + goto loopit; + } + + /* + * If we get here, there's no more mbufs in inputchain, so send the + * sendchain if there is one. + */ + if (pktcnt > 0) { + error = nd6_output_list(ifp, origifp, sendchain, dst, + ro->ro_rt, adv); + /* + * Fall through to done label even in error case because + * nd6_output_list frees packetchain in both success and + * failure cases. + */ + } + +done: + if (ifpp_save != NULL && *ifpp_save != NULL) { + ifnet_release(*ifpp_save); + *ifpp_save = NULL; + } + ROUTE_RELEASE(&ip6route); +#if IPSEC + ROUTE_RELEASE(&ipsec_state.ro); + if (sp != NULL) + key_freesp(sp, KEY_SADB_UNLOCKED); +#endif /* IPSEC */ +#if NECP + ROUTE_RELEASE(&necp_route); +#endif /* NECP */ +#if DUMMYNET + ROUTE_RELEASE(&saved_route); + ROUTE_RELEASE(&saved_ro_pmtu); +#endif /* DUMMYNET */ + + if (ia != NULL) + IFA_REMREF(&ia->ia_ifa); + if (src_ia != NULL) + IFA_REMREF(&src_ia->ia_ifa); + if (ifp != NULL) + ifnet_release(ifp); + if (origifp != NULL) + ifnet_release(origifp); + if (ip6_output_measure) { + net_perf_measure_time(&net_perf, &start_tv, packets_processed); + net_perf_histogram(&net_perf, packets_processed); + } + return (error); + +freehdrs: + if (exthdrs.ip6e_hbh != NULL) { + if (exthdrs.merged) + panic("Double free of ip6e_hbh"); + m_freem(exthdrs.ip6e_hbh); + } + if (exthdrs.ip6e_dest1 != NULL) { + if (exthdrs.merged) + panic("Double free of ip6e_dest1"); + m_freem(exthdrs.ip6e_dest1); + } + if (exthdrs.ip6e_rthdr != NULL) { + if (exthdrs.merged) + panic("Double free of ip6e_rthdr"); + m_freem(exthdrs.ip6e_rthdr); + } + if (exthdrs.ip6e_dest2 != NULL) { + if (exthdrs.merged) + panic("Double free of ip6e_dest2"); + m_freem(exthdrs.ip6e_dest2); + } + /* FALLTHRU */ +bad: + if (inputchain != NULL) + m_freem_list(inputchain); + if (sendchain != NULL) + m_freem_list(sendchain); + if (m != NULL) + m_freem(m); + + goto done; + +#undef ipf_pktopts +#undef exthdrs +#undef ip6route +#undef ipsec_state +#undef saved_route +#undef saved_ro_pmtu +#undef args +} + +/* ip6_fragment_packet + * + * The fragmentation logic is rather complex: + * 1: normal case (dontfrag == 0, alwaysfrag == 0) + * 1-a: send as is if tlen <= path mtu + * 1-b: fragment if tlen > path mtu + * + * 2: if user asks us not to fragment (dontfrag == 1) + * 2-a: send as is if tlen <= interface mtu + * 2-b: error if tlen > interface mtu + * + * 3: if we always need to attach fragment header (alwaysfrag == 1) + * always fragment + * + * 4: if dontfrag == 1 && alwaysfrag == 1 + * error, as we cannot handle this conflicting request + */ + +static int +ip6_fragment_packet(struct mbuf **mptr, struct ip6_pktopts *opt, + struct ip6_exthdrs *exthdrsp, struct ifnet *ifp, uint32_t mtu, + boolean_t alwaysfrag, uint32_t unfragpartlen, struct route_in6 *ro_pmtu, + int nxt0, uint32_t optlen) +{ + VERIFY(NULL != mptr); + struct mbuf *m = *mptr; + int error = 0; + size_t tlen = m->m_pkthdr.len; + boolean_t dontfrag = (opt != NULL && (opt->ip6po_flags & IP6PO_DONTFRAG)); + + if (dontfrag && alwaysfrag) { /* case 4 */ + /* conflicting request - can't transmit */ + return EMSGSIZE; + } + + /* Access without acquiring nd_ifinfo lock for performance */ + if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */ /* - * note that path MTU is never less than IPV6_MMTU - * (see icmp6_input). + * Even if the DONTFRAG option is specified, we cannot send the + * packet when the data length is larger than the MTU of the + * outgoing interface. + * Notify the error by sending IPV6_PATHMTU ancillary data as + * well as returning an error code (the latter is not described + * in the API spec.) */ - error = EMSGSIZE; + u_int32_t mtu32; + struct ip6ctlparam ip6cp; + + mtu32 = (u_int32_t)mtu; + bzero(&ip6cp, sizeof (ip6cp)); + ip6cp.ip6c_cmdarg = (void *)&mtu32; + pfctlinput2(PRC_MSGSIZE, SA(&ro_pmtu->ro_dst), (void *)&ip6cp); + return EMSGSIZE; + } + + /* + * transmit packet without fragmentation + */ + if (dontfrag || (!alwaysfrag && /* case 1-a and 2-a */ + (tlen <= mtu || TSO_IPV6_OK(ifp, m) || + (ifp->if_hwassist & CSUM_FRAGMENT_IPV6)))) { + /* + * mppn not updated in this case because no new chain is formed + * and inserted + */ + ip6_output_checksum(ifp, mtu, m, nxt0, tlen, optlen); + } else { + /* + * time to fragment - cases 1-b and 3 are handled inside + * ip6_do_fragmentation(). + * mppn is passed down to be updated to point at fragment chain. + */ + error = ip6_do_fragmentation(mptr, optlen, ifp, + unfragpartlen, mtod(m, struct ip6_hdr *), exthdrsp, mtu, nxt0); + } + + return error; +} + +/* + * ip6_do_fragmentation() is called by ip6_fragment_packet() after determining + * the packet needs to be fragmented. on success, morig is freed and a chain + * of fragments is linked into the packet chain where morig existed. Otherwise, + * an errno is returned. + */ +static int +ip6_do_fragmentation(struct mbuf **mptr, uint32_t optlen, struct ifnet *ifp, + uint32_t unfragpartlen, struct ip6_hdr *ip6, struct ip6_exthdrs *exthdrsp, + uint32_t mtu, int nxt0) +{ + VERIFY(NULL != mptr); + int error = 0; + + struct mbuf *morig = *mptr; + struct mbuf *first_mbufp = NULL; + struct mbuf *last_mbufp = NULL; + + size_t tlen = morig->m_pkthdr.len; + + /* + * try to fragment the packet. case 1-b and 3 + */ + if ((morig->m_pkthdr.csum_flags & CSUM_TSO_IPV6)) { + /* TSO and fragment aren't compatible */ in6_ifstat_inc(ifp, ifs6_out_fragfail); - goto bad; - } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */ - error = EMSGSIZE; + return EMSGSIZE; + } else if (mtu < IPV6_MMTU) { + /* path MTU cannot be less than IPV6_MMTU */ in6_ifstat_inc(ifp, ifs6_out_fragfail); - goto bad; + return EMSGSIZE; + } else if (ip6->ip6_plen == 0) { + /* jumbo payload cannot be fragmented */ + in6_ifstat_inc(ifp, ifs6_out_fragfail); + return EMSGSIZE; } else { - struct mbuf **mnext, *m_frgpart; + size_t hlen, len, off; + struct mbuf **mnext = NULL; struct ip6_frag *ip6f; - u_int32_t id = htonl(ip6_id++); + u_int32_t id = htonl(ip6_randomid()); u_char nextproto; /* @@ -967,73 +1768,101 @@ skip_ipsec2:; hlen = unfragpartlen; if (mtu > IPV6_MAXPACKET) mtu = IPV6_MAXPACKET; - len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7; + + len = (mtu - hlen - sizeof (struct ip6_frag)) & ~7; if (len < 8) { - error = EMSGSIZE; in6_ifstat_inc(ifp, ifs6_out_fragfail); - goto bad; + return EMSGSIZE; } - mnext = &m->m_nextpkt; - /* * Change the next header field of the last header in the * unfragmentable part. */ - if (exthdrs.ip6e_rthdr) { - nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); - *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; - } else if (exthdrs.ip6e_dest1) { - nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); - *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; - } else if (exthdrs.ip6e_hbh) { - nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); - *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; + if (exthdrsp->ip6e_rthdr != NULL) { + nextproto = *mtod(exthdrsp->ip6e_rthdr, u_char *); + *mtod(exthdrsp->ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; + } else if (exthdrsp->ip6e_dest1 != NULL) { + nextproto = *mtod(exthdrsp->ip6e_dest1, u_char *); + *mtod(exthdrsp->ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; + } else if (exthdrsp->ip6e_hbh != NULL) { + nextproto = *mtod(exthdrsp->ip6e_hbh, u_char *); + *mtod(exthdrsp->ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; } else { nextproto = ip6->ip6_nxt; ip6->ip6_nxt = IPPROTO_FRAGMENT; } + if (morig->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA) + in6_delayed_cksum_offset(morig, 0, optlen, nxt0); + /* * Loop through length of segment after first fragment, - * make new header and copy data of each part and link onto chain. + * make new header and copy data of each part and link onto + * chain. */ - m0 = m; for (off = hlen; off < tlen; off += len) { - MGETHDR(m, M_DONTWAIT, MT_HEADER); - if (!m) { + struct ip6_hdr *new_mhip6; + struct mbuf *new_m; + struct mbuf *m_frgpart; + + MGETHDR(new_m, M_DONTWAIT, MT_HEADER); /* MAC-OK */ + if (new_m == NULL) { error = ENOBUFS; ip6stat.ip6s_odropped++; - goto sendorfree; + break; + } + new_m->m_pkthdr.rcvif = NULL; + new_m->m_flags = morig->m_flags & M_COPYFLAGS; + + if (first_mbufp != NULL) { + /* Every pass through loop but first */ + *mnext = new_m; + last_mbufp = new_m; + } else { + /* This is the first element of the fragment chain */ + first_mbufp = new_m; + last_mbufp = new_m; } - m->m_flags = m0->m_flags & M_COPYFLAGS; - *mnext = m; - mnext = &m->m_nextpkt; - m->m_data += max_linkhdr; - mhip6 = mtod(m, struct ip6_hdr *); - *mhip6 = *ip6; - m->m_len = sizeof(*mhip6); - error = ip6_insertfraghdr(m0, m, hlen, &ip6f); - if (error) { + mnext = &new_m->m_nextpkt; + + new_m->m_data += max_linkhdr; + new_mhip6 = mtod(new_m, struct ip6_hdr *); + *new_mhip6 = *ip6; + new_m->m_len = sizeof (*new_mhip6); + + error = ip6_insertfraghdr(morig, new_m, hlen, &ip6f); + if (error) { ip6stat.ip6s_odropped++; - goto sendorfree; + break; } + ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7)); if (off + len >= tlen) len = tlen - off; else ip6f->ip6f_offlg |= IP6F_MORE_FRAG; - mhip6->ip6_plen = htons((u_short)(len + hlen + - sizeof(*ip6f) - - sizeof(struct ip6_hdr))); - if ((m_frgpart = m_copy(m0, off, len)) == 0) { + new_mhip6->ip6_plen = htons((u_short)(len + hlen + + sizeof (*ip6f) - sizeof (struct ip6_hdr))); + + if ((m_frgpart = m_copy(morig, off, len)) == NULL) { error = ENOBUFS; ip6stat.ip6s_odropped++; - goto sendorfree; + break; } - m_cat(m, m_frgpart); - m->m_pkthdr.len = len + hlen + sizeof(*ip6f); - m->m_pkthdr.rcvif = (struct ifnet *)0; + m_cat(new_m, m_frgpart); + new_m->m_pkthdr.len = len + hlen + sizeof (*ip6f); + new_m->m_pkthdr.rcvif = NULL; + + M_COPY_CLASSIFIER(new_m, morig); + M_COPY_PFTAG(new_m, morig); + +#ifdef notyet +#if CONFIG_MACF_NET + mac_create_fragment(morig, new_m); +#endif /* CONFIG_MACF_NET */ +#endif /* notyet */ + ip6f->ip6f_reserved = 0; ip6f->ip6f_ident = id; ip6f->ip6f_nxt = nextproto; @@ -1041,112 +1870,221 @@ skip_ipsec2:; in6_ifstat_inc(ifp, ifs6_out_fragcreat); } - in6_ifstat_inc(ifp, ifs6_out_fragok); - } - - /* - * Remove leading garbages. - */ -sendorfree: - m = m0->m_nextpkt; - m0->m_nextpkt = 0; - m_freem(m0); - for (m0 = m; m; m = m0) { - m0 = m->m_nextpkt; - m->m_nextpkt = 0; - if (error == 0) { -#if defined(__NetBSD__) && defined(IFA_STATS) - if (IFA_STATS) { - struct in6_ifaddr *ia6; - ip6 = mtod(m, struct ip6_hdr *); - ia6 = in6_ifawithifp(ifp, &ip6->ip6_src); - if (ia6) { - ia->ia_ifa.ifa_data.ifad_outbytes += - m->m_pkthdr.len; - } + if (error) { + /* free all the fragments created */ + if (first_mbufp != NULL) { + m_freem_list(first_mbufp); + first_mbufp = NULL; } -#endif -#if OLDIP6OUTPUT - error = (*ifp->if_output)(ifp, m, - (struct sockaddr *)dst, - ro->ro_rt); -#else - error = nd6_output(ifp, m, dst, ro->ro_rt); -#endif - } else - m_freem(m); + last_mbufp = NULL; + } else { + /* successful fragmenting */ + m_freem(morig); + *mptr = first_mbufp; + last_mbufp->m_nextpkt = NULL; + ip6stat.ip6s_fragmented++; + in6_ifstat_inc(ifp, ifs6_out_fragok); + } } + return error; +} - if (error == 0) - ip6stat.ip6s_fragmented++; +static int +ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen) +{ + struct mbuf *m; -done: - if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */ - RTFREE(ro->ro_rt); - } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) { - RTFREE(ro_pmtu->ro_rt); - } - -#if IPSEC - if (sp != NULL) - key_freesp(sp); -#endif /* IPSEC */ - - return(error); - -freehdrs: - m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */ - m_freem(exthdrs.ip6e_dest1); - m_freem(exthdrs.ip6e_rthdr); - m_freem(exthdrs.ip6e_dest2); - /* fall through */ -bad: - m_freem(m); - goto done; -} - -static int -ip6_copyexthdr(mp, hdr, hlen) - struct mbuf **mp; - caddr_t hdr; - int hlen; -{ - struct mbuf *m; - - if (hlen > MCLBYTES) - return(ENOBUFS); /* XXX */ + if (hlen > MCLBYTES) + return (ENOBUFS); /* XXX */ MGET(m, M_DONTWAIT, MT_DATA); - if (!m) - return(ENOBUFS); + if (m == NULL) + return (ENOBUFS); if (hlen > MLEN) { MCLGET(m, M_DONTWAIT); - if ((m->m_flags & M_EXT) == 0) { + if (!(m->m_flags & M_EXT)) { m_free(m); - return(ENOBUFS); + return (ENOBUFS); } } m->m_len = hlen; - if (hdr) + if (hdr != NULL) bcopy(hdr, mtod(m, caddr_t), hlen); *mp = m; - return(0); + return (0); +} + +static void +ip6_out_cksum_stats(int proto, u_int32_t len) +{ + switch (proto) { + case IPPROTO_TCP: + tcp_out6_cksum_stats(len); + break; + case IPPROTO_UDP: + udp_out6_cksum_stats(len); + break; + default: + /* keep only TCP or UDP stats for now */ + break; + } +} + +/* + * Process a delayed payload checksum calculation (outbound path.) + * + * hoff is the number of bytes beyond the mbuf data pointer which + * points to the IPv6 header. optlen is the number of bytes, if any, + * between the end of IPv6 header and the beginning of the ULP payload + * header, which represents the extension headers. If optlen is less + * than zero, this routine will bail when it detects extension headers. + * + * Returns a bitmask representing all the work done in software. + */ +uint32_t +in6_finalize_cksum(struct mbuf *m, uint32_t hoff, int32_t optlen, + int32_t nxt0, uint32_t csum_flags) +{ + unsigned char buf[sizeof (struct ip6_hdr)] __attribute__((aligned(8))); + struct ip6_hdr *ip6; + uint32_t offset, mlen, hlen, olen, sw_csum; + uint16_t csum, ulpoff, plen; + uint8_t nxt; + + _CASSERT(sizeof (csum) == sizeof (uint16_t)); + VERIFY(m->m_flags & M_PKTHDR); + + sw_csum = (csum_flags & m->m_pkthdr.csum_flags); + + if ((sw_csum &= CSUM_DELAY_IPV6_DATA) == 0) + goto done; + + mlen = m->m_pkthdr.len; /* total mbuf len */ + hlen = sizeof (*ip6); /* IPv6 header len */ + + /* sanity check (need at least IPv6 header) */ + if (mlen < (hoff + hlen)) { + panic("%s: mbuf %p pkt len (%u) < hoff+ip6_hdr " + "(%u+%u)\n", __func__, m, mlen, hoff, hlen); + /* NOTREACHED */ + } + + /* + * In case the IPv6 header is not contiguous, or not 32-bit + * aligned, copy it to a local buffer. + */ + if ((hoff + hlen) > m->m_len || + !IP6_HDR_ALIGNED_P(mtod(m, caddr_t) + hoff)) { + m_copydata(m, hoff, hlen, (caddr_t)buf); + ip6 = (struct ip6_hdr *)(void *)buf; + } else { + ip6 = (struct ip6_hdr *)(void *)(m->m_data + hoff); + } + + nxt = ip6->ip6_nxt; + plen = ntohs(ip6->ip6_plen); + if (plen != (mlen - (hoff + hlen))) { + plen = OSSwapInt16(plen); + if (plen != (mlen - (hoff + hlen))) { + /* Don't complain for jumbograms */ + if (plen != 0 || nxt != IPPROTO_HOPOPTS) { + printf("%s: mbuf 0x%llx proto %d IPv6 " + "plen %d (%x) [swapped %d (%x)] doesn't " + "match actual packet length; %d is used " + "instead\n", __func__, + (uint64_t)VM_KERNEL_ADDRPERM(m), nxt, + ip6->ip6_plen, ip6->ip6_plen, plen, plen, + (mlen - (hoff + hlen))); + } + plen = mlen - (hoff + hlen); + } + } + + if (optlen < 0) { + /* next header isn't TCP/UDP and we don't know optlen, bail */ + if (nxt != IPPROTO_TCP && nxt != IPPROTO_UDP) { + sw_csum = 0; + goto done; + } + olen = 0; + } else { + /* caller supplied the original transport number; use it */ + if (nxt0 >= 0) + nxt = nxt0; + olen = optlen; + } + + offset = hoff + hlen + olen; /* ULP header */ + + /* sanity check */ + if (mlen < offset) { + panic("%s: mbuf %p pkt len (%u) < hoff+ip6_hdr+ext_hdr " + "(%u+%u+%u)\n", __func__, m, mlen, hoff, hlen, olen); + /* NOTREACHED */ + } + + /* + * offset is added to the lower 16-bit value of csum_data, + * which is expected to contain the ULP offset; therefore + * CSUM_PARTIAL offset adjustment must be undone. + */ + if ((m->m_pkthdr.csum_flags & (CSUM_PARTIAL|CSUM_DATA_VALID)) == + (CSUM_PARTIAL|CSUM_DATA_VALID)) { + /* + * Get back the original ULP offset (this will + * undo the CSUM_PARTIAL logic in ip6_output.) + */ + m->m_pkthdr.csum_data = (m->m_pkthdr.csum_tx_stuff - + m->m_pkthdr.csum_tx_start); + } + + ulpoff = (m->m_pkthdr.csum_data & 0xffff); /* ULP csum offset */ + + if (mlen < (ulpoff + sizeof (csum))) { + panic("%s: mbuf %p pkt len (%u) proto %d invalid ULP " + "cksum offset (%u) cksum flags 0x%x\n", __func__, + m, mlen, nxt, ulpoff, m->m_pkthdr.csum_flags); + /* NOTREACHED */ + } + + csum = inet6_cksum(m, 0, offset, plen - olen); + + /* Update stats */ + ip6_out_cksum_stats(nxt, plen - olen); + + /* RFC1122 4.1.3.4 */ + if (csum == 0 && (m->m_pkthdr.csum_flags & CSUM_UDPIPV6)) + csum = 0xffff; + + /* Insert the checksum in the ULP csum field */ + offset += ulpoff; + if ((offset + sizeof (csum)) > m->m_len) { + m_copyback(m, offset, sizeof (csum), &csum); + } else if (IP6_HDR_ALIGNED_P(mtod(m, char *) + hoff)) { + *(uint16_t *)(void *)(mtod(m, char *) + offset) = csum; + } else { + bcopy(&csum, (mtod(m, char *) + offset), sizeof (csum)); + } + m->m_pkthdr.csum_flags &= + ~(CSUM_DELAY_IPV6_DATA | CSUM_DATA_VALID | CSUM_PARTIAL); + +done: + return (sw_csum); } /* * Insert jumbo payload option. */ static int -ip6_insert_jumboopt(exthdrs, plen) - struct ip6_exthdrs *exthdrs; - u_int32_t plen; +ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) { struct mbuf *mopt; u_char *optbuf; + u_int32_t v; -#define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ +#define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */ /* * If there is no hop-by-hop options header, allocate new one. @@ -1154,10 +2092,10 @@ ip6_insert_jumboopt(exthdrs, plen) * jumbo payload option, allocate a cluster to store the whole options. * Otherwise, use it to store the options. */ - if (exthdrs->ip6e_hbh == 0) { + if (exthdrs->ip6e_hbh == NULL) { MGET(mopt, M_DONTWAIT, MT_DATA); - if (mopt == 0) - return(ENOBUFS); + if (mopt == NULL) + return (ENOBUFS); mopt->m_len = JUMBOOPTLEN; optbuf = mtod(mopt, u_char *); optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ @@ -1167,18 +2105,42 @@ ip6_insert_jumboopt(exthdrs, plen) mopt = exthdrs->ip6e_hbh; if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) { - caddr_t oldoptp = mtod(mopt, caddr_t); - int oldoptlen = mopt->m_len; - - if (mopt->m_flags & M_EXT) - return(ENOBUFS); /* XXX */ - MCLGET(mopt, M_DONTWAIT); - if ((mopt->m_flags & M_EXT) == 0) - return(ENOBUFS); - - bcopy(oldoptp, mtod(mopt, caddr_t), oldoptlen); - optbuf = mtod(mopt, caddr_t) + oldoptlen; - mopt->m_len = oldoptlen + JUMBOOPTLEN; + /* + * XXX assumption: + * - exthdrs->ip6e_hbh is not referenced from places + * other than exthdrs. + * - exthdrs->ip6e_hbh is not an mbuf chain. + */ + u_int32_t oldoptlen = mopt->m_len; + struct mbuf *n; + + /* + * XXX: give up if the whole (new) hbh header does + * not fit even in an mbuf cluster. + */ + if (oldoptlen + JUMBOOPTLEN > MCLBYTES) + return (ENOBUFS); + + /* + * As a consequence, we must always prepare a cluster + * at this point. + */ + MGET(n, M_DONTWAIT, MT_DATA); + if (n != NULL) { + MCLGET(n, M_DONTWAIT); + if (!(n->m_flags & M_EXT)) { + m_freem(n); + n = NULL; + } + } + if (n == NULL) + return (ENOBUFS); + n->m_len = oldoptlen + JUMBOOPTLEN; + bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), + oldoptlen); + optbuf = mtod(n, u_char *) + oldoptlen; + m_freem(mopt); + mopt = exthdrs->ip6e_hbh = n; } else { optbuf = mtod(mopt, u_char *) + mopt->m_len; mopt->m_len += JUMBOOPTLEN; @@ -1197,12 +2159,13 @@ ip6_insert_jumboopt(exthdrs, plen) /* fill in the option. */ optbuf[2] = IP6OPT_JUMBO; optbuf[3] = 4; - *(u_int32_t *)&optbuf[4] = htonl(plen + JUMBOOPTLEN); + v = (u_int32_t)htonl(plen + JUMBOOPTLEN); + bcopy(&v, &optbuf[4], sizeof (u_int32_t)); /* finally, adjust the packet header length */ exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; - return(0); + return (0); #undef JUMBOOPTLEN } @@ -1210,18 +2173,16 @@ ip6_insert_jumboopt(exthdrs, plen) * Insert fragment header and copy unfragmentable header portions. */ static int -ip6_insertfraghdr(m0, m, hlen, frghdrp) - struct mbuf *m0, *m; - int hlen; - struct ip6_frag **frghdrp; +ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, + struct ip6_frag **frghdrp) { struct mbuf *n, *mlast; - if (hlen > sizeof(struct ip6_hdr)) { - n = m_copym(m0, sizeof(struct ip6_hdr), - hlen - sizeof(struct ip6_hdr), M_DONTWAIT); - if (n == 0) - return(ENOBUFS); + if (hlen > sizeof (struct ip6_hdr)) { + n = m_copym(m0, sizeof (struct ip6_hdr), + hlen - sizeof (struct ip6_hdr), M_DONTWAIT); + if (n == NULL) + return (ENOBUFS); m->m_next = n; } else n = m; @@ -1230,131 +2191,156 @@ ip6_insertfraghdr(m0, m, hlen, frghdrp) for (mlast = n; mlast->m_next; mlast = mlast->m_next) ; - if ((mlast->m_flags & M_EXT) == 0 && - M_TRAILINGSPACE(mlast) < sizeof(struct ip6_frag)) { - /* use the trailing space of the last mbuf for the fragment hdr */ - *frghdrp = - (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len); - mlast->m_len += sizeof(struct ip6_frag); - m->m_pkthdr.len += sizeof(struct ip6_frag); + if (!(mlast->m_flags & M_EXT) && + M_TRAILINGSPACE(mlast) >= sizeof (struct ip6_frag)) { + /* use the trailing space of the last mbuf for the frag hdr */ + *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) + + mlast->m_len); + mlast->m_len += sizeof (struct ip6_frag); + m->m_pkthdr.len += sizeof (struct ip6_frag); } else { /* allocate a new mbuf for the fragment header */ struct mbuf *mfrg; MGET(mfrg, M_DONTWAIT, MT_DATA); - if (mfrg == 0) - return(ENOBUFS); - mfrg->m_len = sizeof(struct ip6_frag); + if (mfrg == NULL) + return (ENOBUFS); + mfrg->m_len = sizeof (struct ip6_frag); *frghdrp = mtod(mfrg, struct ip6_frag *); mlast->m_next = mfrg; } - return(0); + return (0); +} + +static int +ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, + struct ifnet *ifp, struct in6_addr *dst, u_int32_t *mtup, + boolean_t *alwaysfragp) +{ + u_int32_t mtu = 0; + boolean_t alwaysfrag = FALSE; + int error = 0; + + if (ro_pmtu != ro) { + /* The first hop and the final destination may differ. */ + struct sockaddr_in6 *sa6_dst = SIN6(&ro_pmtu->ro_dst); + if (ROUTE_UNUSABLE(ro_pmtu) || + !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst)) + ROUTE_RELEASE(ro_pmtu); + + if (ro_pmtu->ro_rt == NULL) { + bzero(sa6_dst, sizeof (*sa6_dst)); + sa6_dst->sin6_family = AF_INET6; + sa6_dst->sin6_len = sizeof (struct sockaddr_in6); + sa6_dst->sin6_addr = *dst; + + rtalloc_scoped((struct route *)ro_pmtu, + ifp != NULL ? ifp->if_index : IFSCOPE_NONE); + } + } + + if (ro_pmtu->ro_rt != NULL) { + u_int32_t ifmtu; + + if (ifp == NULL) + ifp = ro_pmtu->ro_rt->rt_ifp; + /* Access without acquiring nd_ifinfo lock for performance */ + ifmtu = IN6_LINKMTU(ifp); + + /* + * Access rmx_mtu without holding the route entry lock, + * for performance; this isn't something that changes + * often, so optimize. + */ + mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu; + if (mtu > ifmtu || mtu == 0) { + /* + * The MTU on the route is larger than the MTU on + * the interface! This shouldn't happen, unless the + * MTU of the interface has been changed after the + * interface was brought up. Change the MTU in the + * route to match the interface MTU (as long as the + * field isn't locked). + * + * if MTU on the route is 0, we need to fix the MTU. + * this case happens with path MTU discovery timeouts. + */ + mtu = ifmtu; + if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) + ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */ + } else if (mtu < IPV6_MMTU) { + /* + * RFC2460 section 5, last paragraph: + * if we record ICMPv6 too big message with + * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU + * or smaller, with framgent header attached. + * (fragment header is needed regardless from the + * packet size, for translators to identify packets) + */ + alwaysfrag = TRUE; + mtu = IPV6_MMTU; + } + } else { + if (ifp) { + /* Don't hold nd_ifinfo lock for performance */ + mtu = IN6_LINKMTU(ifp); + } else { + error = EHOSTUNREACH; /* XXX */ + } + } + + *mtup = mtu; + if (alwaysfragp != NULL) + *alwaysfragp = alwaysfrag; + return (error); } /* * IP6 socket option processing. */ -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) -int -ip6_ctloutput(so, sopt) - struct socket *so; - struct sockopt *sopt; -#else int -ip6_ctloutput(op, so, level, optname, mp) - int op; - struct socket *so; - int level, optname; - struct mbuf **mp; -#endif +ip6_ctloutput(struct socket *so, struct sockopt *sopt) { - int privileged, optdatalen; + int optdatalen, uproto; void *optdata; - struct ip6_recvpktopts *rcvopts; -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - register struct inpcb *in6p = sotoinpcb(so); - int error, optval; - int level, op, optname; - int optlen; + int privileged; + struct inpcb *in6p = sotoinpcb(so); + int error = 0, optval = 0; + int level, op = -1, optname = 0; + int optlen = 0; struct proc *p; - if (sopt) { - level = sopt->sopt_level; - op = sopt->sopt_dir; - optname = sopt->sopt_name; - optlen = sopt->sopt_valsize; - p = sopt->sopt_p; - } else { - panic("ip6_ctloutput: arg soopt is NULL"); - } -#else -#if HAVE_NRL_INPCB - register struct inpcb *inp = sotoinpcb(so); -#else - register struct in6pcb *in6p = sotoin6pcb(so); -#endif - register struct mbuf *m = *mp; - int error, optval; - int optlen; -#if !defined(__bsdi__) && !(defined(__FreeBSD__) && __FreeBSD__ < 3) && !defined (__APPLE__) - struct proc *p = curproc; /* XXX */ -#endif - - optlen = m ? m->m_len : 0; -#endif - error = optval = 0; + VERIFY(sopt != NULL); -#if !defined(__bsdi__) && !(defined(__FreeBSD__) && __FreeBSD__ < 3) && !defined (__APPLE__) - privileged = (p == 0 || suser(p->p_ucred, &p->p_acflag)) ? 0 : 1; -#else -#if HAVE_NRL_INPCB - privileged = (inp->inp_socket->so_state & SS_PRIV); -#else - privileged = (in6p->in6p_socket->so_state & SS_PRIV); -#endif -#endif + level = sopt->sopt_level; + op = sopt->sopt_dir; + optname = sopt->sopt_name; + optlen = sopt->sopt_valsize; + p = sopt->sopt_p; + uproto = (int)SOCK_PROTO(so); -#if defined(HAVE_NRL_INPCB) - rcvopts = inp->inp_inputopts6; -#else - rcvopts = in6p->in6p_inputopts; -#endif + privileged = (proc_suser(p) == 0); if (level == IPPROTO_IPV6) { switch (op) { - -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) case SOPT_SET: -#else - case PRCO_SETOPT: -#endif switch (optname) { - case IPV6_PKTOPTIONS: - { -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) + case IPV6_2292PKTOPTIONS: { struct mbuf *m; - error = sooptgetm(sopt, &m); /* XXX */ - if (error != NULL) + error = soopt_getm(sopt, &m); + if (error != 0) break; - error = sooptmcopyin(sopt, m); /* XXX */ - if (error != NULL) + error = soopt_mcopyin(sopt, m); + if (error != 0) break; error = ip6_pcbopts(&in6p->in6p_outputopts, - m, so, sopt); - m_freem(m); /* XXX */ -#else -#if HAVE_NRL_INPCB - error = ip6_pcbopts(&inp->inp_outputopts6, - m, so); -#else - error = ip6_pcbopts(&in6p->in6p_outputopts, - m, so); -#endif /* HAVE_NRL_INPCB */ -#endif /* FreeBSD >= 3 */ + m, so, sopt); + m_freem(m); break; } + /* * Use of some Hop-by-Hop options or some * Destination options, might require special @@ -1371,487 +2357,420 @@ ip6_ctloutput(op, so, level, optname, mp) case IPV6_RECVHOPOPTS: case IPV6_RECVDSTOPTS: case IPV6_RECVRTHDRDSTOPTS: - if (!privileged) { - error = EPERM; + if (!privileged) break; - } - /* fall through */ + /* FALLTHROUGH */ case IPV6_UNICAST_HOPS: case IPV6_HOPLIMIT: - case IPV6_CHECKSUM: - case IPV6_FAITH: - case IPV6_RECVPKTINFO: case IPV6_RECVHOPLIMIT: case IPV6_RECVRTHDR: - case IPV6_USE_MIN_MTU: -#ifdef notyet /* To be implemented */ case IPV6_RECVPATHMTU: -#endif -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - case IPV6_BINDV6ONLY: -#endif - if (optlen != sizeof(int)) + case IPV6_RECVTCLASS: + case IPV6_V6ONLY: + case IPV6_AUTOFLOWLABEL: + if (optlen != sizeof (int)) { error = EINVAL; - else { -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - error = sooptcopyin(sopt, &optval, - sizeof optval, sizeof optval); - if (error) - break; -#else - optval = *mtod(m, int *); -#endif - switch (optname) { - - case IPV6_UNICAST_HOPS: - if (optval < -1 || optval >= 256) - error = EINVAL; - else { - /* -1 = kernel default */ -#if HAVE_NRL_INPCB - inp->inp_hops = optval; -#else - in6p->in6p_hops = optval; + break; + } + error = sooptcopyin(sopt, &optval, + sizeof (optval), sizeof (optval)); + if (error) + break; -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - if ((in6p->in6p_vflag & - INP_IPV4) != 0) - in6p->inp_ip_ttl = optval; -#endif -#endif + switch (optname) { + case IPV6_UNICAST_HOPS: + if (optval < -1 || optval >= 256) { + error = EINVAL; + } else { + /* -1 = kernel default */ + in6p->in6p_hops = optval; + if (in6p->inp_vflag & + INP_IPV4) { + in6p->inp_ip_ttl = + optval; } - break; -#if HAVE_NRL_INPCB -#define OPTSET(bit) \ - if (optval) \ - inp->inp_flags |= (bit); \ - else \ - inp->inp_flags &= ~(bit); -#else -#define OPTSET(bit) \ - if (optval) \ - in6p->in6p_flags |= (bit); \ - else \ - in6p->in6p_flags &= ~(bit); -#endif -#if HAVE_NRL_INPCB -#define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0) -#else -#define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0) -#endif - - case IPV6_RECVPKTINFO: - OPTSET(IN6P_PKTINFO); - if (OPTBIT(IN6P_PKTINFO) == 0) - ip6_reset_rcvopt(rcvopts, IPV6_RECVPKTINFO); - break; - - case IPV6_HOPLIMIT: - { -#if COMPAT_RFC2292 - OPTSET(IN6P_HOPLIMIT); - if (OPTBIT(IN6P_HOPLIMIT) == 0) - ip6_reset_rcvopt(rcvopts, IPV6_RECVHOPLIMIT); - break; -#else /* new advanced API (2292bis) */ - struct ip6_pktopts **optp; -#if HAVE_NRL_INPCB - optp = &inp->inp_outputopts6; -#else - optp = &in6p->in6p_outputopts; -#endif + } + break; +#define OPTSET(bit) do { \ + if (optval) \ + in6p->inp_flags |= (bit); \ + else \ + in6p->inp_flags &= ~(bit); \ +} while (0) + +#define OPTSET2292(bit) do { \ + in6p->inp_flags |= IN6P_RFC2292; \ + if (optval) \ + in6p->inp_flags |= (bit); \ + else \ + in6p->inp_flags &= ~(bit); \ +} while (0) + +#define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0) - error = ip6_pcbopt(IPV6_HOPLIMIT, - (u_char *)&optval, - sizeof(optval), - optp, - privileged); + case IPV6_RECVPKTINFO: + /* cannot mix with RFC2292 */ + if (OPTBIT(IN6P_RFC2292)) { + error = EINVAL; break; -#endif } + OPTSET(IN6P_PKTINFO); + break; - case IPV6_RECVHOPLIMIT: - OPTSET(IN6P_HOPLIMIT); - if (OPTBIT(IN6P_HOPLIMIT) == 0) - ip6_reset_rcvopt(rcvopts, IPV6_RECVHOPLIMIT); - break; + case IPV6_HOPLIMIT: { + struct ip6_pktopts **optp; - case IPV6_RECVHOPOPTS: - OPTSET(IN6P_HOPOPTS); - if (OPTBIT(IN6P_HOPOPTS) == 0) - ip6_reset_rcvopt(rcvopts, IPV6_RECVHOPOPTS); + /* cannot mix with RFC2292 */ + if (OPTBIT(IN6P_RFC2292)) { + error = EINVAL; break; + } + optp = &in6p->in6p_outputopts; + error = ip6_pcbopt(IPV6_HOPLIMIT, + (u_char *)&optval, sizeof (optval), + optp, uproto); + break; + } - case IPV6_RECVDSTOPTS: - OPTSET(IN6P_DSTOPTS); - if (OPTBIT(IN6P_DSTOPTS) == 0) - ip6_reset_rcvopt(rcvopts, IPV6_RECVDSTOPTS); + case IPV6_RECVHOPLIMIT: + /* cannot mix with RFC2292 */ + if (OPTBIT(IN6P_RFC2292)) { + error = EINVAL; break; + } + OPTSET(IN6P_HOPLIMIT); + break; - case IPV6_RECVRTHDRDSTOPTS: - OPTSET(IN6P_RTHDRDSTOPTS); - if (OPTBIT(IN6P_RTHDRDSTOPTS) == 0) - ip6_reset_rcvopt(rcvopts, IPV6_RECVRTHDRDSTOPTS); + case IPV6_RECVHOPOPTS: + /* cannot mix with RFC2292 */ + if (OPTBIT(IN6P_RFC2292)) { + error = EINVAL; break; + } + OPTSET(IN6P_HOPOPTS); + break; - case IPV6_RECVRTHDR: - OPTSET(IN6P_RTHDR); - if (OPTBIT(IN6P_RTHDR) == 0) - ip6_reset_rcvopt(rcvopts, IPV6_RECVRTHDR); + case IPV6_RECVDSTOPTS: + /* cannot mix with RFC2292 */ + if (OPTBIT(IN6P_RFC2292)) { + error = EINVAL; break; + } + OPTSET(IN6P_DSTOPTS); + break; - case IPV6_CHECKSUM: -#if HAVE_NRL_INPCB - inp->inp_csumoffset = optval; -#else - in6p->in6p_cksum = optval; -#endif + case IPV6_RECVRTHDRDSTOPTS: + /* cannot mix with RFC2292 */ + if (OPTBIT(IN6P_RFC2292)) { + error = EINVAL; break; + } + OPTSET(IN6P_RTHDRDSTOPTS); + break; - case IPV6_FAITH: - OPTSET(IN6P_FAITH); + case IPV6_RECVRTHDR: + /* cannot mix with RFC2292 */ + if (OPTBIT(IN6P_RFC2292)) { + error = EINVAL; break; + } + OPTSET(IN6P_RTHDR); + break; - case IPV6_USE_MIN_MTU: - OPTSET(IN6P_MINMTU); - break; + case IPV6_RECVPATHMTU: + /* + * We ignore this option for TCP + * sockets. + * (RFC3542 leaves this case + * unspecified.) + */ + if (uproto != IPPROTO_TCP) + OPTSET(IN6P_MTU); + break; -#if (defined(__FreeBSD__) && __FreeBSD__ >= 3) || (defined(__NetBSD__) && !defined(INET6_BINDV6ONLY)) || defined (__APPLE__) - case IPV6_BINDV6ONLY: - OPTSET(IN6P_BINDV6ONLY); - break; -#endif - } - } - break; - case IPV6_PKTINFO: - case IPV6_HOPOPTS: - case IPV6_RTHDR: - case IPV6_DSTOPTS: - case IPV6_RTHDRDSTOPTS: - if (optlen == sizeof(int)) { - /* RFC 2292 */ -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - error = sooptcopyin(sopt, &optval, - sizeof optval, sizeof optval); - if (error == 0) - break; -#else - optval = *mtod(m, int *); -#endif - switch(optname) { - case IPV6_PKTINFO: - OPTSET(IN6P_PKTINFO); - if (OPTBIT(IN6P_PKTINFO) == 0) - ip6_reset_rcvopt(rcvopts, IPV6_RECVPKTINFO); - break; - case IPV6_HOPOPTS: - /* - * Check super-user privilege. - * See comments for - * IPV6_RECVHOPOPTS. - */ - if (!privileged) - return(EPERM); - OPTSET(IN6P_HOPOPTS); - if (OPTBIT(IN6P_HOPOPTS) == 0) - ip6_reset_rcvopt(rcvopts, IPV6_RECVHOPOPTS); - break; - case IPV6_DSTOPTS: - if (!privileged) - return(EPERM); - OPTSET(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ - if (OPTBIT(IN6P_DSTOPTS) == 0) { - ip6_reset_rcvopt(rcvopts, IPV6_RECVDSTOPTS); - ip6_reset_rcvopt(rcvopts, IPV6_RECVRTHDRDSTOPTS); - } - break; - case IPV6_RTHDR: - OPTSET(IN6P_RTHDR); - if (OPTBIT(IN6P_RTHDR) == 0) - ip6_reset_rcvopt(rcvopts, IPV6_RECVRTHDR); + case IPV6_V6ONLY: + /* + * make setsockopt(IPV6_V6ONLY) + * available only prior to bind(2). + * see ipng mailing list, Jun 22 2001. + */ + if (in6p->inp_lport || + !IN6_IS_ADDR_UNSPECIFIED( + &in6p->in6p_laddr)) { + error = EINVAL; break; } + OPTSET(IN6P_IPV6_V6ONLY); + if (optval) + in6p->inp_vflag &= ~INP_IPV4; + else + in6p->inp_vflag |= INP_IPV4; break; - } else { - /* new advanced API (2292bis) */ - u_char *optbuf; - int optlen; - struct ip6_pktopts **optp; -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - optbuf = sopt->sopt_val; - optlen = sopt->sopt_valsize; -#else /* !fbsd3 */ - if (m && m->m_next) { - error = EINVAL; /* XXX */ - break; - } - if (m) { - optbuf = mtod(m, u_char *); - optlen = m->m_len; - } else { - optbuf = NULL; - optlen = 0; - } -#endif + case IPV6_RECVTCLASS: + /* we can mix with RFC2292 */ + OPTSET(IN6P_TCLASS); + break; -#if HAVE_NRL_INPCB - optp = &inp->inp_outputopts6; -#else - optp = &in6p->in6p_outputopts; -#endif + case IPV6_AUTOFLOWLABEL: + OPTSET(IN6P_AUTOFLOWLABEL); + break; - error = ip6_pcbopt(optname, - optbuf, optlen, - optp, privileged); } break; -#undef OPTSET - case IPV6_MULTICAST_IF: - case IPV6_MULTICAST_HOPS: - case IPV6_MULTICAST_LOOP: - case IPV6_JOIN_GROUP: - case IPV6_LEAVE_GROUP: -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - { - struct mbuf *m; - if (sopt->sopt_valsize > MLEN) { - error = EMSGSIZE; + case IPV6_TCLASS: + case IPV6_DONTFRAG: + case IPV6_USE_MIN_MTU: + case IPV6_PREFER_TEMPADDR: { + struct ip6_pktopts **optp; + + if (optlen != sizeof (optval)) { + error = EINVAL; break; } - /* XXX */ - MGET(m, sopt->sopt_p ? M_WAIT : M_DONTWAIT, MT_HEADER); - if (m == 0) { - error = ENOBUFS; + error = sooptcopyin(sopt, &optval, + sizeof (optval), sizeof (optval)); + if (error) break; - } - m->m_len = sopt->sopt_valsize; - error = sooptcopyin(sopt, mtod(m, char *), - m->m_len, m->m_len); - error = ip6_setmoptions(sopt->sopt_name, - &in6p->in6p_moptions, - m); - (void)m_free(m); - } -#else -#if HAVE_NRL_INPCB - error = ip6_setmoptions(optname, - &inp->inp_moptions6, m); - /* - * XXX: setting the flag would be redundant - * except at the first time. Also, we - * actually don't have to reset the flag, - * since ip6_freemoptions() would simply - * return when the inp_moptions6 is NULL. - */ - if (inp->inp_moptions6) - inp->inp_flags |= INP_IPV6_MCAST; - else - inp->inp_flags &= ~INP_IPV6_MCAST; -#else - error = ip6_setmoptions(optname, - &in6p->in6p_moptions, m); -#endif -#endif - break; -#ifndef __bsdi__ - case IPV6_PORTRANGE: -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - error = sooptcopyin(sopt, &optval, sizeof optval, - sizeof optval); - if (error) + optp = &in6p->in6p_outputopts; + error = ip6_pcbopt(optname, (u_char *)&optval, + sizeof (optval), optp, uproto); break; -#else - optval = *mtod(m, int *); -#endif + } -#if HAVE_NRL_INPCB -# define in6p inp -# define in6p_flags inp_flags -#endif - switch (optval) { - case IPV6_PORTRANGE_DEFAULT: - in6p->in6p_flags &= ~(IN6P_LOWPORT); - in6p->in6p_flags &= ~(IN6P_HIGHPORT); - break; - - case IPV6_PORTRANGE_HIGH: - in6p->in6p_flags &= ~(IN6P_LOWPORT); - in6p->in6p_flags |= IN6P_HIGHPORT; + case IPV6_2292PKTINFO: + case IPV6_2292HOPLIMIT: + case IPV6_2292HOPOPTS: + case IPV6_2292DSTOPTS: + case IPV6_2292RTHDR: + /* RFC 2292 */ + if (optlen != sizeof (int)) { + error = EINVAL; + break; + } + error = sooptcopyin(sopt, &optval, + sizeof (optval), sizeof (optval)); + if (error) + break; + switch (optname) { + case IPV6_2292PKTINFO: + OPTSET2292(IN6P_PKTINFO); + break; + case IPV6_2292HOPLIMIT: + OPTSET2292(IN6P_HOPLIMIT); + break; + case IPV6_2292HOPOPTS: + /* + * Check super-user privilege. + * See comments for IPV6_RECVHOPOPTS. + */ + if (!privileged) + return (EPERM); + OPTSET2292(IN6P_HOPOPTS); + break; + case IPV6_2292DSTOPTS: + if (!privileged) + return (EPERM); + OPTSET2292(IN6P_DSTOPTS| + IN6P_RTHDRDSTOPTS); /* XXX */ + break; + case IPV6_2292RTHDR: + OPTSET2292(IN6P_RTHDR); + break; + } break; - case IPV6_PORTRANGE_LOW: - in6p->in6p_flags &= ~(IN6P_HIGHPORT); - in6p->in6p_flags |= IN6P_LOWPORT; - break; + case IPV6_3542PKTINFO: + case IPV6_3542HOPOPTS: + case IPV6_3542RTHDR: + case IPV6_3542DSTOPTS: + case IPV6_RTHDRDSTOPTS: + case IPV6_3542NEXTHOP: { + struct ip6_pktopts **optp; + /* new advanced API (RFC3542) */ + struct mbuf *m; - default: - error = EINVAL; + /* cannot mix with RFC2292 */ + if (OPTBIT(IN6P_RFC2292)) { + error = EINVAL; + break; + } + error = soopt_getm(sopt, &m); + if (error != 0) + break; + error = soopt_mcopyin(sopt, m); + if (error != 0) + break; + + optp = &in6p->in6p_outputopts; + error = ip6_pcbopt(optname, mtod(m, u_char *), + m->m_len, optp, uproto); + m_freem(m); break; } -#if HAVE_NRL_INPCB -# undef in6p -# undef in6p_flags -#endif - break; -#endif +#undef OPTSET + case IPV6_MULTICAST_IF: + case IPV6_MULTICAST_HOPS: + case IPV6_MULTICAST_LOOP: + case IPV6_JOIN_GROUP: + case IPV6_LEAVE_GROUP: + case IPV6_MSFILTER: + case MCAST_BLOCK_SOURCE: + case MCAST_UNBLOCK_SOURCE: + case MCAST_JOIN_GROUP: + case MCAST_LEAVE_GROUP: + case MCAST_JOIN_SOURCE_GROUP: + case MCAST_LEAVE_SOURCE_GROUP: + error = ip6_setmoptions(in6p, sopt); + break; + + case IPV6_PORTRANGE: + error = sooptcopyin(sopt, &optval, + sizeof (optval), sizeof (optval)); + if (error) + break; + + switch (optval) { + case IPV6_PORTRANGE_DEFAULT: + in6p->inp_flags &= ~(INP_LOWPORT); + in6p->inp_flags &= ~(INP_HIGHPORT); + break; + + case IPV6_PORTRANGE_HIGH: + in6p->inp_flags &= ~(INP_LOWPORT); + in6p->inp_flags |= INP_HIGHPORT; + break; + + case IPV6_PORTRANGE_LOW: + in6p->inp_flags &= ~(INP_HIGHPORT); + in6p->inp_flags |= INP_LOWPORT; + break; + default: + error = EINVAL; + break; + } + break; #if IPSEC - case IPV6_IPSEC_POLICY: - { + case IPV6_IPSEC_POLICY: { caddr_t req = NULL; size_t len = 0; -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) struct mbuf *m; -#endif - -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - if (error = sooptgetm(sopt, &m)) /* XXX */ + + if ((error = soopt_getm(sopt, &m)) != 0) break; - if (error = sooptmcopyin(sopt, m)) /* XXX */ + if ((error = soopt_mcopyin(sopt, m)) != 0) break; -#endif - if (m) { - req = mtod(m, caddr_t); - len = m->m_len; - } -#if HAVE_NRL_INPCB - error = ipsec6_set_policy(inp, optname, req, - len, privileged); -#else + + req = mtod(m, caddr_t); + len = m->m_len; error = ipsec6_set_policy(in6p, optname, req, - len, privileged); -#endif -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) + len, privileged); m_freem(m); -#endif - } break; + } #endif /* IPSEC */ - -#if IPV6FIREWALL +#if IPFIREWALL case IPV6_FW_ADD: case IPV6_FW_DEL: case IPV6_FW_FLUSH: - case IPV6_FW_ZERO: - { -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - struct mbuf *m; - struct mbuf **mp = &m; -#endif - -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) + case IPV6_FW_ZERO: { if (ip6_fw_ctl_ptr == NULL) - return EINVAL; - if (error = sooptgetm(sopt, &m)) /* XXX */ + load_ip6fw(); + if (ip6_fw_ctl_ptr != NULL) + error = (*ip6_fw_ctl_ptr)(sopt); + else + error = ENOPROTOOPT; + break; + } +#endif /* IPFIREWALL */ + /* + * IPv6 variant of IP_BOUND_IF; for details see + * comments on IP_BOUND_IF in ip_ctloutput(). + */ + case IPV6_BOUND_IF: + /* This option is settable only on IPv6 */ + if (!(in6p->inp_vflag & INP_IPV6)) { + error = EINVAL; break; - if (error = sooptmcopyin(sopt, m)) /* XXX */ + } + + error = sooptcopyin(sopt, &optval, + sizeof (optval), sizeof (optval)); + + if (error) + break; + + error = inp_bindif(in6p, optval, NULL); + break; + + case IPV6_NO_IFT_CELLULAR: + /* This option is settable only for IPv6 */ + if (!(in6p->inp_vflag & INP_IPV6)) { + error = EINVAL; break; -#else - if (ip6_fw_ctl_ptr == NULL) { - if (m) (void)m_free(m); - return EINVAL; } -#endif - error = (*ip6_fw_ctl_ptr)(optname, mp); - m = *mp; - } + + error = sooptcopyin(sopt, &optval, + sizeof (optval), sizeof (optval)); + + if (error) + break; + + /* once set, it cannot be unset */ + if (!optval && INP_NO_CELLULAR(in6p)) { + error = EINVAL; + break; + } + + error = so_set_restrictions(so, + SO_RESTRICT_DENY_CELLULAR); + break; + + case IPV6_OUT_IF: + /* This option is not settable */ + error = EINVAL; break; -#endif default: error = ENOPROTOOPT; break; } -#if !(defined(__FreeBSD__) && __FreeBSD__ >= 3) && !defined(__APPLE__) - if (m) - (void)m_free(m); -#endif break; -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) case SOPT_GET: -#else - case PRCO_GETOPT: -#endif switch (optname) { - case IPV6_PKTOPTIONS: -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - if (in6p->in6p_inputopts && - in6p->in6p_inputopts->head) { - error = sooptmcopyout(sopt, - in6p->in6p_inputopts->head); - } else - sopt->sopt_valsize = 0; -#elif defined(HAVE_NRL_INPCB) - if (inp->inp_options) { - *mp = m_copym(inp->inp_options, 0, - M_COPYALL, M_WAIT); - } else { - *mp = m_get(M_WAIT, MT_SOOPTS); - (*mp)->m_len = 0; - } -#else - if (in6p->in6p_inputopts && - in6p->in6p_inputopts->head) { - *mp = m_copym(in6p->in6p_inputopts->head, - 0, M_COPYALL, M_WAIT); - } else { - *mp = m_get(M_WAIT, MT_SOOPTS); - (*mp)->m_len = 0; - } -#endif + case IPV6_2292PKTOPTIONS: + /* + * RFC3542 (effectively) deprecated the + * semantics of the 2292-style pktoptions. + * Since it was not reliable in nature (i.e., + * applications had to expect the lack of some + * information after all), it would make sense + * to simplify this part by always returning + * empty data. + */ + sopt->sopt_valsize = 0; break; case IPV6_RECVHOPOPTS: case IPV6_RECVDSTOPTS: case IPV6_RECVRTHDRDSTOPTS: - if (!privileged) { - error = EPERM; - break; - } - /* fall through */ case IPV6_UNICAST_HOPS: - case IPV6_CHECKSUM: - case IPV6_RECVPKTINFO: case IPV6_RECVHOPLIMIT: case IPV6_RECVRTHDR: - case IPV6_USE_MIN_MTU: -#ifdef notyet /* To be implemented */ case IPV6_RECVPATHMTU: -#endif - - case IPV6_FAITH: -#if (defined(__FreeBSD__) && __FreeBSD__ >= 3) || (defined(__NetBSD__) && !defined(INET6_BINDV6ONLY)) || defined(__APPLE__) - case IPV6_BINDV6ONLY: -#endif -#ifndef __bsdi__ + case IPV6_V6ONLY: case IPV6_PORTRANGE: -#endif + case IPV6_RECVTCLASS: + case IPV6_AUTOFLOWLABEL: switch (optname) { - case IPV6_UNICAST_HOPS: -#if HAVE_NRL_INPCB - optval = inp->inp_hops; -#else - optval = in6p->in6p_hops; -#endif - break; - - case IPV6_RECVPKTINFO: - optval = OPTBIT(IN6P_PKTINFO); - break; - - case IPV6_RECVHOPLIMIT: - optval = OPTBIT(IN6P_HOPLIMIT); - break; - case IPV6_RECVHOPOPTS: optval = OPTBIT(IN6P_HOPOPTS); break; @@ -1864,207 +2783,169 @@ ip6_ctloutput(op, so, level, optname, mp) optval = OPTBIT(IN6P_RTHDRDSTOPTS); break; - case IPV6_CHECKSUM: -#if HAVE_NRL_INPCB - optval = inp->inp_csumoffset; -#else - optval = in6p->in6p_cksum; -#endif + case IPV6_UNICAST_HOPS: + optval = in6p->in6p_hops; + break; + + case IPV6_RECVPKTINFO: + optval = OPTBIT(IN6P_PKTINFO); break; - case IPV6_USE_MIN_MTU: - optval = OPTBIT(IN6P_MINMTU); + case IPV6_RECVHOPLIMIT: + optval = OPTBIT(IN6P_HOPLIMIT); break; - case IPV6_FAITH: - optval = OPTBIT(IN6P_FAITH); + case IPV6_RECVRTHDR: + optval = OPTBIT(IN6P_RTHDR); break; -#if (defined(__FreeBSD__) && __FreeBSD__ >= 3) || (defined(__NetBSD__) && !defined(INET6_BINDV6ONLY)) || defined (__APPLE__) - case IPV6_BINDV6ONLY: - optval = OPTBIT(IN6P_BINDV6ONLY); + case IPV6_RECVPATHMTU: + optval = OPTBIT(IN6P_MTU); + break; + + case IPV6_V6ONLY: + optval = OPTBIT(IN6P_IPV6_V6ONLY); break; -#endif -#ifndef __bsdi__ - case IPV6_PORTRANGE: - { + case IPV6_PORTRANGE: { int flags; -#if HAVE_NRL_INPCB - flags = inp->inp_flags; -#else - flags = in6p->in6p_flags; -#endif - if (flags & IN6P_HIGHPORT) + flags = in6p->inp_flags; + if (flags & INP_HIGHPORT) optval = IPV6_PORTRANGE_HIGH; - else if (flags & IN6P_LOWPORT) + else if (flags & INP_LOWPORT) optval = IPV6_PORTRANGE_LOW; else optval = 0; break; - } -#endif } -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) + case IPV6_RECVTCLASS: + optval = OPTBIT(IN6P_TCLASS); + break; + + case IPV6_AUTOFLOWLABEL: + optval = OPTBIT(IN6P_AUTOFLOWLABEL); + break; + } + if (error) + break; error = sooptcopyout(sopt, &optval, - sizeof optval); -#else - *mp = m = m_get(M_WAIT, MT_SOOPTS); - m->m_len = sizeof(int); - *mtod(m, int *) = optval; -#endif + sizeof (optval)); break; - case IPV6_PKTINFO: - case IPV6_HOPOPTS: - case IPV6_RTHDR: - case IPV6_DSTOPTS: - case IPV6_RTHDRDSTOPTS: -#if COMPAT_RFC2292 - if (optname == IPV6_HOPOPTS || - optname == IPV6_DSTOPTS || - !privileged) - return(EPERM); - switch(optname) { - case IPV6_PKTINFO: - optbit = OPTBIT(IN6P_PKTINFO); + case IPV6_PATHMTU: { + u_int32_t pmtu = 0; + struct ip6_mtuinfo mtuinfo; + struct route_in6 sro; + + bzero(&sro, sizeof (sro)); + + if (!(so->so_state & SS_ISCONNECTED)) + return (ENOTCONN); + /* + * XXX: we dot not consider the case of source + * routing, or optional information to specify + * the outgoing interface. + */ + error = ip6_getpmtu(&sro, NULL, NULL, + &in6p->in6p_faddr, &pmtu, NULL); + ROUTE_RELEASE(&sro); + if (error) break; - case IPV6_HOPLIMIT: + if (pmtu > IPV6_MAXPACKET) + pmtu = IPV6_MAXPACKET; + + bzero(&mtuinfo, sizeof (mtuinfo)); + mtuinfo.ip6m_mtu = (u_int32_t)pmtu; + optdata = (void *)&mtuinfo; + optdatalen = sizeof (mtuinfo); + error = sooptcopyout(sopt, optdata, + optdatalen); + break; + } + + case IPV6_2292PKTINFO: + case IPV6_2292HOPLIMIT: + case IPV6_2292HOPOPTS: + case IPV6_2292RTHDR: + case IPV6_2292DSTOPTS: + switch (optname) { + case IPV6_2292PKTINFO: + optval = OPTBIT(IN6P_PKTINFO); + break; + case IPV6_2292HOPLIMIT: optval = OPTBIT(IN6P_HOPLIMIT); break; - case IPV6_HOPOPTS: - optbit = OPTBIT(IN6P_HOPOPTS); + case IPV6_2292HOPOPTS: + optval = OPTBIT(IN6P_HOPOPTS); break; - case IPV6_RTHDR: - optbit = OPTBIT(IN6P_RTHDR); + case IPV6_2292RTHDR: + optval = OPTBIT(IN6P_RTHDR); break; - case IPV6_DSTOPTS: - optbit = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); + case IPV6_2292DSTOPTS: + optval = OPTBIT(IN6P_DSTOPTS| + IN6P_RTHDRDSTOPTS); break; - case IPV6_RTHDRDSTOPTS: /* in 2292bis only */ - return(EOPNOTSUPP); } -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) error = sooptcopyout(sopt, &optval, - sizeof optval); -#else - *mp = m = m_get(M_WAIT, MT_SOOPTS); - m->m_len = sizeof(int); - *mtod(m, int *) = optval; -#endif /* FreeBSD3 */ -#else /* new advanced API */ -#if HAVE_NRL_INPCB -#define in6p inp -#define in6p_outputopts inp_outputopts6 -#endif + sizeof (optval)); + break; + + case IPV6_PKTINFO: + case IPV6_HOPOPTS: + case IPV6_RTHDR: + case IPV6_DSTOPTS: + case IPV6_RTHDRDSTOPTS: + case IPV6_NEXTHOP: + case IPV6_TCLASS: + case IPV6_DONTFRAG: + case IPV6_USE_MIN_MTU: + case IPV6_PREFER_TEMPADDR: error = ip6_getpcbopt(in6p->in6p_outputopts, - optname, &optdata, - &optdatalen); - if (error == 0) { -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - /* note that optdatalen maybe 0 */ - error = sooptcopyout(sopt, optdata, - optdatalen); -#else /* !FreeBSD3 */ - if (optdatalen > MCLBYTES) - return(EMSGSIZE); /* XXX */ - *mp = m = m_get(M_WAIT, MT_SOOPTS); - if (optdatalen > MLEN) - MCLGET(m, M_WAIT); - m->m_len = optdatalen; - bcopy(optdata, mtod(m, void *), - optdatalen); -#endif /* FreeBSD3 */ - } -#if HAVE_NRL_INPCB -#undef in6p -#undef in6p_outputopts -#endif -#endif /* COMPAT_RFC2292 */ + optname, sopt); break; case IPV6_MULTICAST_IF: case IPV6_MULTICAST_HOPS: case IPV6_MULTICAST_LOOP: - case IPV6_JOIN_GROUP: - case IPV6_LEAVE_GROUP: -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - { - struct mbuf *m; - error = ip6_getmoptions(sopt->sopt_name, - in6p->in6p_moptions, &m); - if (error == 0) - error = sooptcopyout(sopt, - mtod(m, char *), m->m_len); - m_freem(m); - } -#elif defined(HAVE_NRL_INPCB) - error = ip6_getmoptions(optname, inp->inp_moptions6, mp); -#else - error = ip6_getmoptions(optname, in6p->in6p_moptions, mp); -#endif + case IPV6_MSFILTER: + error = ip6_getmoptions(in6p, sopt); break; - #if IPSEC - case IPV6_IPSEC_POLICY: - { - caddr_t req = NULL; - size_t len = 0; -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - struct mbuf *m = NULL; - struct mbuf **mp = &m; - - error = sooptgetm(sopt, &m); /* XXX */ - if (error != NULL) - break; - error = sooptmcopyin(sopt, m); /* XXX */ - if (error != NULL) - break; -#endif - if (m) { - req = mtod(m, caddr_t); - len = m->m_len; - } -#if HAVE_NRL_INPCB - error = ipsec6_get_policy(inp, req, len, mp); -#else - error = ipsec6_get_policy(in6p, req, len, mp); -#endif -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - if (error == 0) - error = sooptmcopyout(sopt, m); /*XXX*/ - m_freem(m); -#endif + case IPV6_IPSEC_POLICY: { + error = 0; /* This option is no longer supported */ break; - } + } #endif /* IPSEC */ +#if IPFIREWALL + case IPV6_FW_GET: { + if (ip6_fw_ctl_ptr == NULL) + load_ip6fw(); + if (ip6_fw_ctl_ptr != NULL) + error = (*ip6_fw_ctl_ptr)(sopt); + else + error = ENOPROTOOPT; + break; + } +#endif /* IPFIREWALL */ + case IPV6_BOUND_IF: + if (in6p->inp_flags & INP_BOUND_IF) + optval = in6p->inp_boundifp->if_index; + error = sooptcopyout(sopt, &optval, + sizeof (optval)); + break; -#if IPV6FIREWALL - case IPV6_FW_GET: - { -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - struct mbuf *m; - struct mbuf **mp = &m; -#endif + case IPV6_NO_IFT_CELLULAR: + optval = INP_NO_CELLULAR(in6p) ? 1 : 0; + error = sooptcopyout(sopt, &optval, + sizeof (optval)); + break; - if (ip6_fw_ctl_ptr == NULL) - { -#if !(defined(__FreeBSD__) && __FreeBSD__ >= 3) || defined (__APPLE__) - if (m) - (void)m_free(m); -#endif - return EINVAL; - } - error = (*ip6_fw_ctl_ptr)(optname, mp); -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - if (error == 0) - error = sooptmcopyout(sopt, m); /* XXX */ - if (m) - m_freem(m); -#endif - } + case IPV6_OUT_IF: + optval = (in6p->in6p_last_outifp != NULL) ? + in6p->in6p_last_outifp->if_index : 0; + error = sooptcopyout(sopt, &optval, + sizeof (optval)); break; -#endif default: error = ENOPROTOOPT; @@ -2072,14 +2953,82 @@ ip6_ctloutput(op, so, level, optname, mp) } break; } + } else if (level == IPPROTO_UDP) { + error = udp_ctloutput(so, sopt); } else { error = EINVAL; -#if !(defined(__FreeBSD__) && __FreeBSD__ >= 3) && !defined(__APPLE__) - if (op == PRCO_SETOPT && *mp) - (void)m_free(*mp); -#endif } - return(error); + return (error); +} + +int +ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) +{ + int error = 0, optval, optlen; + const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); + struct inpcb *in6p = sotoinpcb(so); + int level, op, optname; + + level = sopt->sopt_level; + op = sopt->sopt_dir; + optname = sopt->sopt_name; + optlen = sopt->sopt_valsize; + + if (level != IPPROTO_IPV6) + return (EINVAL); + + switch (optname) { + case IPV6_CHECKSUM: + /* + * For ICMPv6 sockets, no modification allowed for checksum + * offset, permit "no change" values to help existing apps. + * + * RFC3542 says: "An attempt to set IPV6_CHECKSUM + * for an ICMPv6 socket will fail." + * The current behavior does not meet RFC3542. + */ + switch (op) { + case SOPT_SET: + if (optlen != sizeof (int)) { + error = EINVAL; + break; + } + error = sooptcopyin(sopt, &optval, sizeof (optval), + sizeof (optval)); + if (error) + break; + if ((optval % 2) != 0) { + /* the API assumes even offset values */ + error = EINVAL; + } else if (SOCK_PROTO(so) == IPPROTO_ICMPV6) { + if (optval != icmp6off) + error = EINVAL; + } else { + in6p->in6p_cksum = optval; + } + break; + + case SOPT_GET: + if (SOCK_PROTO(so) == IPPROTO_ICMPV6) + optval = icmp6off; + else + optval = in6p->in6p_cksum; + + error = sooptcopyout(sopt, &optval, sizeof (optval)); + break; + + default: + error = EINVAL; + break; + } + break; + + default: + error = ENOPROTOOPT; + break; + } + + return (error); } /* @@ -2087,1059 +3036,897 @@ ip6_ctloutput(op, so, level, optname, mp) * specifying behavior of outgoing packets. */ static int -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) -ip6_pcbopts(pktopt, m, so, sopt) -#else -ip6_pcbopts(pktopt, m, so) -#endif - struct ip6_pktopts **pktopt; - register struct mbuf *m; - struct socket *so; -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - struct sockopt *sopt; -#endif +ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so, + struct sockopt *sopt) { - register struct ip6_pktopts *opt = *pktopt; +#pragma unused(sopt) + struct ip6_pktopts *opt = *pktopt; int error = 0; -#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined (__APPLE__) - struct proc *p = sopt->sopt_p; -#else - struct proc *p = curproc; /* XXX */ -#endif - int priv = 0; /* turn off any old options. */ - if (opt) { + if (opt != NULL) { #if DIAGNOSTIC - if (opt->ip6po_pktinfo || opt->ip6po_nexthop || - opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || - opt->ip6po_rhinfo.ip6po_rhi_rthdr) - printf("ip6_pcbopts: all specified options are cleared.\n"); + if (opt->ip6po_pktinfo || opt->ip6po_nexthop || + opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || + opt->ip6po_rhinfo.ip6po_rhi_rthdr) + printf("%s: all specified options are cleared.\n", + __func__); #endif - ip6_clearpktopts(opt, 1, -1); + ip6_clearpktopts(opt, -1); + } else { + opt = _MALLOC(sizeof (*opt), M_IP6OPT, M_WAITOK); + if (opt == NULL) + return (ENOBUFS); } - else - opt = _MALLOC(sizeof(*opt), M_IP6OPT, M_WAITOK); *pktopt = NULL; - if (!m || m->m_len == 0) { + if (m == NULL || m->m_len == 0) { /* - * Only turning off any previous options. + * Only turning off any previous options, regardless of + * whether the opt is just created or given. */ - if (opt) - _FREE(opt, M_IP6OPT); - return(0); + if (opt != NULL) + FREE(opt, M_IP6OPT); + return (0); } /* set options specified by user. */ -#if 0 - if (p && !suser(p->p_ucred, &p->p_acflag)) - priv = 1; -#endif - if ((error = ip6_setpktoptions(m, opt, priv, 1)) != 0) { - ip6_clearpktopts(opt, 1, -1); /* XXX: discard all options */ - return(error); + if ((error = ip6_setpktopts(m, opt, NULL, SOCK_PROTO(so))) != 0) { + ip6_clearpktopts(opt, -1); /* XXX: discard all options */ + FREE(opt, M_IP6OPT); + return (error); } *pktopt = opt; - return(0); + return (0); } /* - * Set up an IP6 option in pcb for insertion in output packets or - * specifying behavior of outgoing packets. - * XXX: The logic of this function is very similar to ip6_setpktoptions(). + * initialize ip6_pktopts. beware that there are non-zero default values in + * the struct. */ -static int -ip6_pcbopt(optname, buf, len, pktopt, priv) - int optname, len, priv; - u_char *buf; - struct ip6_pktopts **pktopt; +void +ip6_initpktopts(struct ip6_pktopts *opt) { - struct ip6_pktopts *opt; - struct in6_pktinfo *pktinfo; - - if (*pktopt == NULL) { - *pktopt = _MALLOC(sizeof(struct ip6_pktopts), M_IP6OPT, - M_WAITOK); - bzero(*pktopt, sizeof(struct ip6_pktopts)); - (*pktopt)->ip6po_hlim = -1; - } - opt = *pktopt; - - switch(optname) { - case IPV6_PKTINFO: - if (len == 0) { /* just remove the option */ - ip6_clearpktopts(opt, 1, IPV6_PKTINFO); - break; - } - - if (len != sizeof(struct in6_pktinfo)) - return EINVAL; - pktinfo = (struct in6_pktinfo *)buf; - - /* - * An application can clear any sticky IPV6_PKTINFO option by - * doing a "regular" setsockopt with ipi6_addr being - * in6addr_any and ipi6_ifindex being zero. - * [rfc2292bis-01, Section 6] - * XXX: Is this a good feature?? (jinmei@kame.net) - */ - if (pktinfo->ipi6_ifindex == 0 && - IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { - ip6_clearpktopts(opt, 1, IPV6_PKTINFO); - break; - } - - /* XXX: this overrides the original data space */ - if (pktinfo->ipi6_ifindex && - IN6_IS_ADDR_LINKLOCAL(&pktinfo->ipi6_addr)) - pktinfo->ipi6_addr.s6_addr16[1] = - htons(pktinfo->ipi6_ifindex); - - if (pktinfo->ipi6_ifindex > if_index || - pktinfo->ipi6_ifindex < 0) - return(ENXIO); - - /* - * Check if the requested source address is indeed a unicast - * address assigned to the node. - */ - if (!IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { - struct ifaddr *ia; - struct sockaddr_in6 sin6; - - bzero(&sin6, sizeof(sin6)); - sin6.sin6_len = sizeof(sin6); - sin6.sin6_family = AF_INET6; - sin6.sin6_addr = pktinfo->ipi6_addr; - ia = ifa_ifwithaddr(sin6tosa(&sin6)); - if (ia == NULL) - return(EADDRNOTAVAIL); - } - - if (opt->ip6po_pktinfo == NULL) - opt->ip6po_pktinfo = _MALLOC(sizeof(struct in6_pktinfo), - M_IP6OPT, M_WAITOK); - bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo)); - - break; - case IPV6_HOPLIMIT: - { - int *hlimp; - - if (len != sizeof(int)) - return(EINVAL); - hlimp = (int *)buf; - if (*hlimp < -1 || *hlimp > 255) - return(EINVAL); - - opt->ip6po_hlim = *hlimp; - break; - } - case IPV6_NEXTHOP: - if (!priv) - return(EPERM); - - if (len == 0) { /* just remove the option */ - ip6_clearpktopts(opt, 1, IPV6_NEXTHOP); - break; - } - - /* check if cmsg_len is large enough for sa_len */ - if (len < sizeof(u_char) || - len < *buf) - return(EINVAL); - - /* turn off the previous option */ - ip6_clearpktopts(opt, 1, IPV6_NEXTHOP); - - opt->ip6po_nexthop = _MALLOC(*buf, M_IP6OPT, M_WAITOK); - bcopy(buf, opt->ip6po_nexthop, *buf); - break; - case IPV6_HOPOPTS: - { - struct ip6_hbh *hbh; - int hbhlen; - - /* - * XXX: We don't allow a non-privileged user to set ANY HbH - * options, since per-option restriction has too much - * overhead. - */ - if (!priv) - return(EPERM); - - if (len == 0) { - ip6_clearpktopts(opt, 1, IPV6_HOPOPTS); - break; /* just remove the option */ - } - - if (len < sizeof(struct ip6_hbh)) - return(EINVAL); - hbh = (struct ip6_hbh *)buf; - hbhlen = (hbh->ip6h_len + 1) << 3; - if (len != hbhlen) - return(EINVAL); - - /* turn off the previous option */ - ip6_clearpktopts(opt, 1, IPV6_HOPOPTS); - - opt->ip6po_hbh = _MALLOC(hbhlen, M_IP6OPT, M_WAITOK); - bcopy(buf, opt->ip6po_hbh, hbhlen); - - break; - } - case IPV6_DSTOPTS: - case IPV6_RTHDRDSTOPTS: - { - struct ip6_dest *dest, *newdest; - int destlen; - - if (!priv) /* XXX: see the comment for IPV6_HOPOPTS */ - return(EPERM); - - if (len == 0) { - ip6_clearpktopts(opt, 1, optname); - break; /* just remove the option */ - } - - if (len < sizeof(struct ip6_dest)) - return(EINVAL); - dest = (struct ip6_dest *)buf; - destlen = (dest->ip6d_len + 1) << 3; - if (len != destlen) - return(EINVAL); - - /* turn off the previous option */ - ip6_clearpktopts(opt, 1, optname); - - newdest = _MALLOC(destlen, M_IP6OPT, M_WAITOK); - bcopy(buf, newdest, destlen); - - if (optname == IPV6_DSTOPTS) - opt->ip6po_dest2 = newdest; - else - opt->ip6po_dest1 = newdest; - break; - } - case IPV6_RTHDR: - { - struct ip6_rthdr *rth; - int rthlen; - - if (len == 0) { - ip6_clearpktopts(opt, 1, IPV6_RTHDR); - break; /* just remove the option */ - } - - if (len < sizeof(struct ip6_rthdr)) - return(EINVAL); - rth = (struct ip6_rthdr *)buf; - rthlen = (rth->ip6r_len + 1) << 3; - if (len != rthlen) - return(EINVAL); - - switch(rth->ip6r_type) { - case IPV6_RTHDR_TYPE_0: - if (rth->ip6r_len == 0) /* must contain one addr */ - return(EINVAL); - if (rth->ip6r_len % 2) /* length must be even */ - return(EINVAL); - if (rth->ip6r_len / 2 != rth->ip6r_segleft) - return(EINVAL); - break; - default: - return(EINVAL); /* not supported */ - } + bzero(opt, sizeof (*opt)); + opt->ip6po_hlim = -1; /* -1 means default hop limit */ + opt->ip6po_tclass = -1; /* -1 means default traffic class */ + opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; + opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; +} - /* turn off the previous option */ - ip6_clearpktopts(opt, 1, IPV6_RTHDR); +static int +ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, + int uproto) +{ + struct ip6_pktopts *opt; - opt->ip6po_rthdr = _MALLOC(rthlen, M_IP6OPT, M_WAITOK); - bcopy(buf, opt->ip6po_rthdr, rthlen); - - break; + opt = *pktopt; + if (opt == NULL) { + opt = _MALLOC(sizeof (*opt), M_IP6OPT, M_WAITOK); + if (opt == NULL) + return (ENOBUFS); + ip6_initpktopts(opt); + *pktopt = opt; } - default: - return(ENOPROTOOPT); - } /* end of switch */ - return(0); + return (ip6_setpktopt(optname, buf, len, opt, 1, 0, uproto)); } static int -ip6_getpcbopt(pktopt, optname, datap, datalenp) - struct ip6_pktopts *pktopt; - int optname, *datalenp; - void **datap; +ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) { void *optdata = NULL; - struct ip6_ext *ip6e; int optdatalen = 0; + struct ip6_ext *ip6e; + struct in6_pktinfo null_pktinfo; + int deftclass = 0, on; + int defminmtu = IP6PO_MINMTU_MCASTONLY; + int defpreftemp = IP6PO_TEMPADDR_SYSTEM; - if (pktopt == NULL) - goto end; - switch(optname) { + switch (optname) { case IPV6_PKTINFO: - if (pktopt->ip6po_pktinfo) { + if (pktopt && pktopt->ip6po_pktinfo) optdata = (void *)pktopt->ip6po_pktinfo; - optdatalen = sizeof(struct in6_pktinfo); + else { + /* XXX: we don't have to do this every time... */ + bzero(&null_pktinfo, sizeof (null_pktinfo)); + optdata = (void *)&null_pktinfo; } + optdatalen = sizeof (struct in6_pktinfo); break; - case IPV6_HOPLIMIT: - optdata = (void *)&pktopt->ip6po_hlim; - optdatalen = sizeof(int); + + case IPV6_TCLASS: + if (pktopt && pktopt->ip6po_tclass >= 0) + optdata = (void *)&pktopt->ip6po_tclass; + else + optdata = (void *)&deftclass; + optdatalen = sizeof (int); break; + case IPV6_HOPOPTS: - if (pktopt->ip6po_hbh) { + if (pktopt && pktopt->ip6po_hbh) { optdata = (void *)pktopt->ip6po_hbh; ip6e = (struct ip6_ext *)pktopt->ip6po_hbh; optdatalen = (ip6e->ip6e_len + 1) << 3; } break; + case IPV6_RTHDR: - if (pktopt->ip6po_rthdr) { + if (pktopt && pktopt->ip6po_rthdr) { optdata = (void *)pktopt->ip6po_rthdr; ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr; optdatalen = (ip6e->ip6e_len + 1) << 3; } break; + case IPV6_RTHDRDSTOPTS: - if (pktopt->ip6po_dest1) { + if (pktopt && pktopt->ip6po_dest1) { optdata = (void *)pktopt->ip6po_dest1; ip6e = (struct ip6_ext *)pktopt->ip6po_dest1; optdatalen = (ip6e->ip6e_len + 1) << 3; } break; + case IPV6_DSTOPTS: - if (pktopt->ip6po_dest2) { + if (pktopt && pktopt->ip6po_dest2) { optdata = (void *)pktopt->ip6po_dest2; ip6e = (struct ip6_ext *)pktopt->ip6po_dest2; optdatalen = (ip6e->ip6e_len + 1) << 3; } break; - } - end: - *datap = optdata; - *datalenp = optdatalen; + case IPV6_NEXTHOP: + if (pktopt && pktopt->ip6po_nexthop) { + optdata = (void *)pktopt->ip6po_nexthop; + optdatalen = pktopt->ip6po_nexthop->sa_len; + } + break; + + case IPV6_USE_MIN_MTU: + if (pktopt) + optdata = (void *)&pktopt->ip6po_minmtu; + else + optdata = (void *)&defminmtu; + optdatalen = sizeof (int); + break; + + case IPV6_DONTFRAG: + if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) + on = 1; + else + on = 0; + optdata = (void *)&on; + optdatalen = sizeof (on); + break; + + case IPV6_PREFER_TEMPADDR: + if (pktopt) + optdata = (void *)&pktopt->ip6po_prefer_tempaddr; + else + optdata = (void *)&defpreftemp; + optdatalen = sizeof (int); + break; + + default: /* should not happen */ +#ifdef DIAGNOSTIC + panic("ip6_getpcbopt: unexpected option\n"); +#endif + return (ENOPROTOOPT); + } - return(0); + return (sooptcopyout(sopt, optdata, optdatalen)); } void -ip6_clearpktopts(pktopt, needfree, optname) - struct ip6_pktopts *pktopt; - int needfree, optname; +ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) { if (pktopt == NULL) return; if (optname == -1 || optname == IPV6_PKTINFO) { - if (needfree && pktopt->ip6po_pktinfo) - _FREE(pktopt->ip6po_pktinfo, M_IP6OPT); + if (pktopt->ip6po_pktinfo) + FREE(pktopt->ip6po_pktinfo, M_IP6OPT); pktopt->ip6po_pktinfo = NULL; } if (optname == -1 || optname == IPV6_HOPLIMIT) pktopt->ip6po_hlim = -1; + if (optname == -1 || optname == IPV6_TCLASS) + pktopt->ip6po_tclass = -1; if (optname == -1 || optname == IPV6_NEXTHOP) { - if (needfree && pktopt->ip6po_nexthop) - _FREE(pktopt->ip6po_nexthop, M_IP6OPT); + ROUTE_RELEASE(&pktopt->ip6po_nextroute); + if (pktopt->ip6po_nexthop) + FREE(pktopt->ip6po_nexthop, M_IP6OPT); pktopt->ip6po_nexthop = NULL; } if (optname == -1 || optname == IPV6_HOPOPTS) { - if (needfree && pktopt->ip6po_hbh) - _FREE(pktopt->ip6po_hbh, M_IP6OPT); + if (pktopt->ip6po_hbh) + FREE(pktopt->ip6po_hbh, M_IP6OPT); pktopt->ip6po_hbh = NULL; } if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { - if (needfree && pktopt->ip6po_dest1) - _FREE(pktopt->ip6po_dest1, M_IP6OPT); + if (pktopt->ip6po_dest1) + FREE(pktopt->ip6po_dest1, M_IP6OPT); pktopt->ip6po_dest1 = NULL; } if (optname == -1 || optname == IPV6_RTHDR) { - if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) - _FREE(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); + if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) + FREE(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; - if (pktopt->ip6po_route.ro_rt) { - RTFREE(pktopt->ip6po_route.ro_rt); - pktopt->ip6po_route.ro_rt = NULL; - } + ROUTE_RELEASE(&pktopt->ip6po_route); } if (optname == -1 || optname == IPV6_DSTOPTS) { - if (needfree && pktopt->ip6po_dest2) - _FREE(pktopt->ip6po_dest2, M_IP6OPT); + if (pktopt->ip6po_dest2) + FREE(pktopt->ip6po_dest2, M_IP6OPT); pktopt->ip6po_dest2 = NULL; } } -#define PKTOPT_EXTHDRCPY(type) if (src->type) {\ - int hlen =\ - (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\ - dst->type = _MALLOC(hlen, M_IP6OPT, canwait);\ - if (dst->type == NULL && canwait == M_NOWAIT)\ - goto bad;\ - bcopy(src->type, dst->type, hlen);\ - } +#define PKTOPT_EXTHDRCPY(type) do { \ + if (src->type) { \ + int hlen = \ + (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3; \ + dst->type = _MALLOC(hlen, M_IP6OPT, canwait); \ + if (dst->type == NULL && canwait == M_NOWAIT) \ + goto bad; \ + bcopy(src->type, dst->type, hlen); \ + } \ +} while (0) -struct ip6_pktopts * -ip6_copypktopts(src, canwait) - struct ip6_pktopts *src; - int canwait; +static int +copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) { - struct ip6_pktopts *dst; - - if (src == NULL) { - printf("ip6_clearpktopts: invalid argument\n"); - return(NULL); + if (dst == NULL || src == NULL) { + printf("copypktopts: invalid argument\n"); + return (EINVAL); } - dst = _MALLOC(sizeof(*dst), M_IP6OPT, canwait); - if (dst == NULL && canwait == M_NOWAIT) - goto bad; - bzero(dst, sizeof(*dst)); - dst->ip6po_hlim = src->ip6po_hlim; + dst->ip6po_tclass = src->ip6po_tclass; dst->ip6po_flags = src->ip6po_flags; if (src->ip6po_pktinfo) { - dst->ip6po_pktinfo = _MALLOC(sizeof(*dst->ip6po_pktinfo), - M_IP6OPT, canwait); + dst->ip6po_pktinfo = _MALLOC(sizeof (*dst->ip6po_pktinfo), + M_IP6OPT, canwait); if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT) goto bad; *dst->ip6po_pktinfo = *src->ip6po_pktinfo; } if (src->ip6po_nexthop) { dst->ip6po_nexthop = _MALLOC(src->ip6po_nexthop->sa_len, - M_IP6OPT, canwait); + M_IP6OPT, canwait); if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) goto bad; bcopy(src->ip6po_nexthop, dst->ip6po_nexthop, - src->ip6po_nexthop->sa_len); + src->ip6po_nexthop->sa_len); } PKTOPT_EXTHDRCPY(ip6po_hbh); PKTOPT_EXTHDRCPY(ip6po_dest1); PKTOPT_EXTHDRCPY(ip6po_dest2); PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ - return(dst); - - bad: - printf("ip6_copypktopts: copy failed"); - if (dst->ip6po_pktinfo) _FREE(dst->ip6po_pktinfo, M_IP6OPT); - if (dst->ip6po_nexthop) _FREE(dst->ip6po_nexthop, M_IP6OPT); - if (dst->ip6po_hbh) _FREE(dst->ip6po_hbh, M_IP6OPT); - if (dst->ip6po_dest1) _FREE(dst->ip6po_dest1, M_IP6OPT); - if (dst->ip6po_dest2) _FREE(dst->ip6po_dest2, M_IP6OPT); - if (dst->ip6po_rthdr) _FREE(dst->ip6po_rthdr, M_IP6OPT); - return(NULL); + return (0); + +bad: + ip6_clearpktopts(dst, -1); + return (ENOBUFS); } #undef PKTOPT_EXTHDRCPY +struct ip6_pktopts * +ip6_copypktopts(struct ip6_pktopts *src, int canwait) +{ + int error; + struct ip6_pktopts *dst; + + dst = _MALLOC(sizeof (*dst), M_IP6OPT, canwait); + if (dst == NULL) + return (NULL); + ip6_initpktopts(dst); + + if ((error = copypktopts(dst, src, canwait)) != 0) { + FREE(dst, M_IP6OPT); + return (NULL); + } + + return (dst); +} + void -ip6_freepcbopts(pktopt) - struct ip6_pktopts *pktopt; +ip6_freepcbopts(struct ip6_pktopts *pktopt) { if (pktopt == NULL) return; - ip6_clearpktopts(pktopt, 1, -1); + ip6_clearpktopts(pktopt, -1); + + FREE(pktopt, M_IP6OPT); +} + +void +ip6_moptions_init(void) +{ + PE_parse_boot_argn("ifa_debug", &im6o_debug, sizeof (im6o_debug)); + + im6o_size = (im6o_debug == 0) ? sizeof (struct ip6_moptions) : + sizeof (struct ip6_moptions_dbg); + + im6o_zone = zinit(im6o_size, IM6O_ZONE_MAX * im6o_size, 0, + IM6O_ZONE_NAME); + if (im6o_zone == NULL) { + panic("%s: failed allocating %s", __func__, IM6O_ZONE_NAME); + /* NOTREACHED */ + } + zone_change(im6o_zone, Z_EXPAND, TRUE); +} + +void +im6o_addref(struct ip6_moptions *im6o, int locked) +{ + if (!locked) + IM6O_LOCK(im6o); + else + IM6O_LOCK_ASSERT_HELD(im6o); + + if (++im6o->im6o_refcnt == 0) { + panic("%s: im6o %p wraparound refcnt\n", __func__, im6o); + /* NOTREACHED */ + } else if (im6o->im6o_trace != NULL) { + (*im6o->im6o_trace)(im6o, TRUE); + } + + if (!locked) + IM6O_UNLOCK(im6o); +} + +void +im6o_remref(struct ip6_moptions *im6o) +{ + int i; + + IM6O_LOCK(im6o); + if (im6o->im6o_refcnt == 0) { + panic("%s: im6o %p negative refcnt", __func__, im6o); + /* NOTREACHED */ + } else if (im6o->im6o_trace != NULL) { + (*im6o->im6o_trace)(im6o, FALSE); + } + + --im6o->im6o_refcnt; + if (im6o->im6o_refcnt > 0) { + IM6O_UNLOCK(im6o); + return; + } + + for (i = 0; i < im6o->im6o_num_memberships; ++i) { + struct in6_mfilter *imf; + + imf = im6o->im6o_mfilters ? &im6o->im6o_mfilters[i] : NULL; + if (imf != NULL) + im6f_leave(imf); + + (void) in6_mc_leave(im6o->im6o_membership[i], imf); + + if (imf != NULL) + im6f_purge(imf); + + IN6M_REMREF(im6o->im6o_membership[i]); + im6o->im6o_membership[i] = NULL; + } + im6o->im6o_num_memberships = 0; + if (im6o->im6o_mfilters != NULL) { + FREE(im6o->im6o_mfilters, M_IN6MFILTER); + im6o->im6o_mfilters = NULL; + } + if (im6o->im6o_membership != NULL) { + FREE(im6o->im6o_membership, M_IP6MOPTS); + im6o->im6o_membership = NULL; + } + IM6O_UNLOCK(im6o); + + lck_mtx_destroy(&im6o->im6o_lock, ifa_mtx_grp); + + if (!(im6o->im6o_debug & IFD_ALLOC)) { + panic("%s: im6o %p cannot be freed", __func__, im6o); + /* NOTREACHED */ + } + zfree(im6o_zone, im6o); +} + +static void +im6o_trace(struct ip6_moptions *im6o, int refhold) +{ + struct ip6_moptions_dbg *im6o_dbg = (struct ip6_moptions_dbg *)im6o; + ctrace_t *tr; + u_int32_t idx; + u_int16_t *cnt; + + if (!(im6o->im6o_debug & IFD_DEBUG)) { + panic("%s: im6o %p has no debug structure", __func__, im6o); + /* NOTREACHED */ + } + if (refhold) { + cnt = &im6o_dbg->im6o_refhold_cnt; + tr = im6o_dbg->im6o_refhold; + } else { + cnt = &im6o_dbg->im6o_refrele_cnt; + tr = im6o_dbg->im6o_refrele; + } + + idx = atomic_add_16_ov(cnt, 1) % IM6O_TRACE_HIST_SIZE; + ctrace_record(&tr[idx]); +} + +struct ip6_moptions * +ip6_allocmoptions(int how) +{ + struct ip6_moptions *im6o; + + im6o = (how == M_WAITOK) ? + zalloc(im6o_zone) : zalloc_noblock(im6o_zone); + if (im6o != NULL) { + bzero(im6o, im6o_size); + lck_mtx_init(&im6o->im6o_lock, ifa_mtx_grp, ifa_mtx_attr); + im6o->im6o_debug |= IFD_ALLOC; + if (im6o_debug != 0) { + im6o->im6o_debug |= IFD_DEBUG; + im6o->im6o_trace = im6o_trace; + } + IM6O_ADDREF(im6o); + } - _FREE(pktopt, M_IP6OPT); + return (im6o); } /* - * Set the IP6 multicast options in response to user setsockopt(). + * Set IPv6 outgoing packet options based on advanced API. */ -static int -ip6_setmoptions(optname, im6op, m) - int optname; - struct ip6_moptions **im6op; - struct mbuf *m; +int +ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, + struct ip6_pktopts *stickyopt, int uproto) { - int error = 0; - u_int loop, ifindex; - struct ipv6_mreq *mreq; - struct ifnet *ifp; - struct ip6_moptions *im6o = *im6op; - struct route_in6 ro; - struct sockaddr_in6 *dst; - struct in6_multi_mship *imm; + struct cmsghdr *cm = NULL; + + if (control == NULL || opt == NULL) + return (EINVAL); - struct proc *p = current_proc(); /* ### */ + ip6_initpktopts(opt); + if (stickyopt) { + int error; - if (im6o == NULL) { /* - * No multicast option buffer attached to the pcb; - * allocate one and initialize to default values. + * If stickyopt is provided, make a local copy of the options + * for this particular packet, then override them by ancillary + * objects. + * XXX: copypktopts() does not copy the cached route to a next + * hop (if any). This is not very good in terms of efficiency, + * but we can allow this since this option should be rarely + * used. */ - im6o = (struct ip6_moptions *) - _MALLOC(sizeof(*im6o), M_IPMOPTS, M_WAITOK); + if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) + return (error); + } + + /* + * XXX: Currently, we assume all the optional information is stored + * in a single mbuf. + */ + if (control->m_next) + return (EINVAL); + + if (control->m_len < CMSG_LEN(0)) + return (EINVAL); + + for (cm = M_FIRST_CMSGHDR(control); cm != NULL; + cm = M_NXT_CMSGHDR(control, cm)) { + int error; + + if (cm->cmsg_len < sizeof (struct cmsghdr) || + cm->cmsg_len > control->m_len) + return (EINVAL); + if (cm->cmsg_level != IPPROTO_IPV6) + continue; + + error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), + cm->cmsg_len - CMSG_LEN(0), opt, 0, 1, uproto); + if (error) + return (error); + } + + return (0); +} +/* + * Set a particular packet option, as a sticky option or an ancillary data + * item. "len" can be 0 only when it's a sticky option. + * We have 4 cases of combination of "sticky" and "cmsg": + * "sticky=0, cmsg=0": impossible + * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data + * "sticky=1, cmsg=0": RFC3542 socket option + * "sticky=1, cmsg=1": RFC2292 socket option + */ +static int +ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, + int sticky, int cmsg, int uproto) +{ + int minmtupolicy, preftemp; + int error; + + if (!sticky && !cmsg) { +#ifdef DIAGNOSTIC + printf("ip6_setpktopt: impossible case\n"); +#endif + return (EINVAL); + } + + /* + * Caller must have ensured that the buffer is at least + * aligned on 32-bit boundary. + */ + VERIFY(IS_P2ALIGNED(buf, sizeof (u_int32_t))); - if (im6o == NULL) - return(ENOBUFS); - *im6op = im6o; - im6o->im6o_multicast_ifp = NULL; - im6o->im6o_multicast_hlim = ip6_defmcasthlim; - im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP; - LIST_INIT(&im6o->im6o_memberships); + /* + * IPV6_2292xxx is for backward compatibility to RFC2292, and should + * not be specified in the context of RFC3542. Conversely, + * RFC3542 types should not be specified in the context of RFC2292. + */ + if (!cmsg) { + switch (optname) { + case IPV6_2292PKTINFO: + case IPV6_2292HOPLIMIT: + case IPV6_2292NEXTHOP: + case IPV6_2292HOPOPTS: + case IPV6_2292DSTOPTS: + case IPV6_2292RTHDR: + case IPV6_2292PKTOPTIONS: + return (ENOPROTOOPT); + } + } + if (sticky && cmsg) { + switch (optname) { + case IPV6_PKTINFO: + case IPV6_HOPLIMIT: + case IPV6_NEXTHOP: + case IPV6_HOPOPTS: + case IPV6_DSTOPTS: + case IPV6_RTHDRDSTOPTS: + case IPV6_RTHDR: + case IPV6_USE_MIN_MTU: + case IPV6_DONTFRAG: + case IPV6_TCLASS: + case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */ + return (ENOPROTOOPT); + } } switch (optname) { + case IPV6_2292PKTINFO: + case IPV6_PKTINFO: { + struct ifnet *ifp = NULL; + struct in6_pktinfo *pktinfo; + + if (len != sizeof (struct in6_pktinfo)) + return (EINVAL); + + pktinfo = (struct in6_pktinfo *)(void *)buf; - case IPV6_MULTICAST_IF: /* - * Select the interface for outgoing multicast packets. + * An application can clear any sticky IPV6_PKTINFO option by + * doing a "regular" setsockopt with ipi6_addr being + * in6addr_any and ipi6_ifindex being zero. + * [RFC 3542, Section 6] */ - if (m == NULL || m->m_len != sizeof(u_int)) { - error = EINVAL; + if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo && + pktinfo->ipi6_ifindex == 0 && + IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { + ip6_clearpktopts(opt, optname); break; } - ifindex = *(mtod(m, u_int *)); - if (ifindex < 0 || if_index < ifindex) { - error = ENXIO; /* XXX EINVAL? */ - break; - } - ifp = ifindex2ifnet[ifindex]; - if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { - error = EADDRNOTAVAIL; - break; - } - im6o->im6o_multicast_ifp = ifp; - break; - case IPV6_MULTICAST_HOPS: - { - /* - * Set the IP6 hoplimit for outgoing multicast packets. - */ - int optval; - if (m == NULL || m->m_len != sizeof(int)) { - error = EINVAL; - break; + if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && + sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { + return (EINVAL); } - optval = *(mtod(m, u_int *)); - if (optval < -1 || optval >= 256) - error = EINVAL; - else if (optval == -1) - im6o->im6o_multicast_hlim = ip6_defmcasthlim; - else - im6o->im6o_multicast_hlim = optval; - break; - } - case IPV6_MULTICAST_LOOP: - /* - * Set the loopback flag for outgoing multicast packets. - * Must be zero or one. - */ - if (m == NULL || m->m_len != sizeof(u_int) || - (loop = *(mtod(m, u_int *))) > 1) { - error = EINVAL; - break; - } - im6o->im6o_multicast_loop = loop; - break; + /* validate the interface index if specified. */ + ifnet_head_lock_shared(); - case IPV6_JOIN_GROUP: - /* - * Add a multicast group membership. - * Group must be a valid IP6 multicast address. - */ - if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { - error = EINVAL; - break; + if (pktinfo->ipi6_ifindex > if_index) { + ifnet_head_done(); + return (ENXIO); } - mreq = mtod(m, struct ipv6_mreq *); - if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { - /* - * We use the unspecified address to specify to accept - * all multicast addresses. Only super user is allowed - * to do this. - */ -#if ISFB31 - if (suser(p->p_ucred, &p->p_acflag)) { - error = EACCES; - break; + + if (pktinfo->ipi6_ifindex) { + ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; + if (ifp == NULL) { + ifnet_head_done(); + return (ENXIO); } -#endif - } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { - error = EINVAL; - break; } - /* - * If the interface is specified, validate it. - */ - if (mreq->ipv6mr_interface < 0 - || if_index < mreq->ipv6mr_interface) { - error = ENXIO; /* XXX EINVAL? */ - break; - } - /* - * If no interface was explicitly specified, choose an - * appropriate one according to the given multicast address. - */ - if (mreq->ipv6mr_interface == 0) { - /* - * If the multicast address is in node-local scope, - * the interface should be a loopback interface. - * Otherwise, look up the routing table for the - * address, and choose the outgoing interface. - * XXX: is it a good approach? - */ - if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) { -#ifdef __bsdi__ - ifp = loifp; -#else - ifp = &loif[0]; -#endif - } else { - ro.ro_rt = NULL; - dst = (struct sockaddr_in6 *)&ro.ro_dst; - bzero(dst, sizeof(*dst)); - dst->sin6_len = sizeof(struct sockaddr_in6); - dst->sin6_family = AF_INET6; - dst->sin6_addr = mreq->ipv6mr_multiaddr; - rtalloc((struct route *)&ro); - if (ro.ro_rt == NULL) { - error = EADDRNOTAVAIL; - break; - } - ifp = ro.ro_rt->rt_ifp; - rtfree(ro.ro_rt); - } - } else - ifp = ifindex2ifnet[mreq->ipv6mr_interface]; + ifnet_head_done(); /* - * See if we found an interface, and confirm that it - * supports multicast + * We store the address anyway, and let in6_selectsrc() + * validate the specified address. This is because ipi6_addr + * may not have enough information about its scope zone, and + * we may need additional information (such as outgoing + * interface or the scope zone of a destination address) to + * disambiguate the scope. + * XXX: the delay of the validation may confuse the + * application when it is used as a sticky option. */ - if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { - error = EADDRNOTAVAIL; - break; - } - /* - * Put interface index into the multicast address, - * if the address has link-local scope. - */ - if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { - mreq->ipv6mr_multiaddr.s6_addr16[1] - = htons(mreq->ipv6mr_interface); - } - /* - * See if the membership already exists. - */ - for (imm = im6o->im6o_memberships.lh_first; - imm != NULL; imm = imm->i6mm_chain.le_next) - if (imm->i6mm_maddr->in6m_ifp == ifp && - IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, - &mreq->ipv6mr_multiaddr)) - break; - if (imm != NULL) { - error = EADDRINUSE; - break; - } - /* - * Everything looks good; add a new record to the multicast - * address list for the given interface. - */ - imm = _MALLOC(sizeof(*imm), M_IPMADDR, M_WAITOK); - if (imm == NULL) { - error = ENOBUFS; - break; - } - if ((imm->i6mm_maddr = - in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) { - _FREE(imm, M_IPMADDR); - break; + if (opt->ip6po_pktinfo == NULL) { + opt->ip6po_pktinfo = _MALLOC(sizeof (*pktinfo), + M_IP6OPT, M_NOWAIT); + if (opt->ip6po_pktinfo == NULL) + return (ENOBUFS); } - LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); + bcopy(pktinfo, opt->ip6po_pktinfo, sizeof (*pktinfo)); break; + } + + case IPV6_2292HOPLIMIT: + case IPV6_HOPLIMIT: { + int *hlimp; - case IPV6_LEAVE_GROUP: - /* - * Drop a multicast group membership. - * Group must be a valid IP6 multicast address. - */ - if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) { - error = EINVAL; - break; - } - mreq = mtod(m, struct ipv6_mreq *); - if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) { - if (suser(p->p_ucred, &p->p_acflag)) { - error = EACCES; - break; - } - } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) { - error = EINVAL; - break; - } - /* - * If an interface address was specified, get a pointer - * to its ifnet structure. - */ - if (mreq->ipv6mr_interface < 0 - || if_index < mreq->ipv6mr_interface) { - error = ENXIO; /* XXX EINVAL? */ - break; - } - ifp = ifindex2ifnet[mreq->ipv6mr_interface]; - /* - * Put interface index into the multicast address, - * if the address has link-local scope. - */ - if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) { - mreq->ipv6mr_multiaddr.s6_addr16[1] - = htons(mreq->ipv6mr_interface); - } - /* - * Find the membership in the membership list. - */ - for (imm = im6o->im6o_memberships.lh_first; - imm != NULL; imm = imm->i6mm_chain.le_next) { - if ((ifp == NULL || - imm->i6mm_maddr->in6m_ifp == ifp) && - IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr, - &mreq->ipv6mr_multiaddr)) - break; - } - if (imm == NULL) { - /* Unable to resolve interface */ - error = EADDRNOTAVAIL; - break; - } /* - * Give up the multicast address record to which the - * membership points. + * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT + * to simplify the ordering among hoplimit options. */ - LIST_REMOVE(imm, i6mm_chain); - in6_delmulti(imm->i6mm_maddr); - _FREE(imm, M_IPMADDR); - break; + if (optname == IPV6_HOPLIMIT && sticky) + return (ENOPROTOOPT); - default: - error = EOPNOTSUPP; + if (len != sizeof (int)) + return (EINVAL); + hlimp = (int *)(void *)buf; + if (*hlimp < -1 || *hlimp > 255) + return (EINVAL); + + opt->ip6po_hlim = *hlimp; break; } - /* - * If all options have default values, no need to keep the mbuf. - */ - if (im6o->im6o_multicast_ifp == NULL && - im6o->im6o_multicast_hlim == ip6_defmcasthlim && - im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP && - im6o->im6o_memberships.lh_first == NULL) { - _FREE(*im6op, M_IPMOPTS); - *im6op = NULL; + case IPV6_TCLASS: { + int tclass; + + if (len != sizeof (int)) + return (EINVAL); + tclass = *(int *)(void *)buf; + if (tclass < -1 || tclass > 255) + return (EINVAL); + + opt->ip6po_tclass = tclass; + break; } - return(error); -} + case IPV6_2292NEXTHOP: + case IPV6_NEXTHOP: + error = suser(kauth_cred_get(), 0); + if (error) + return (EACCES); -/* - * Return the IP6 multicast options in response to user getsockopt(). - */ -static int -ip6_getmoptions(optname, im6o, mp) - int optname; - register struct ip6_moptions *im6o; - register struct mbuf **mp; -{ - u_int *hlim, *loop, *ifindex; + if (len == 0) { /* just remove the option */ + ip6_clearpktopts(opt, IPV6_NEXTHOP); + break; + } -#if __FreeBSD__ || defined (__APPLE__) - *mp = m_get(M_WAIT, MT_HEADER); /*XXX*/ -#else - *mp = m_get(M_WAIT, MT_SOOPTS); -#endif + /* check if cmsg_len is large enough for sa_len */ + if (len < sizeof (struct sockaddr) || len < *buf) + return (EINVAL); - switch (optname) { + switch (SA(buf)->sa_family) { + case AF_INET6: { + struct sockaddr_in6 *sa6 = SIN6(buf); - case IPV6_MULTICAST_IF: - ifindex = mtod(*mp, u_int *); - (*mp)->m_len = sizeof(u_int); - if (im6o == NULL || im6o->im6o_multicast_ifp == NULL) - *ifindex = 0; - else - *ifindex = im6o->im6o_multicast_ifp->if_index; - return(0); - - case IPV6_MULTICAST_HOPS: - hlim = mtod(*mp, u_int *); - (*mp)->m_len = sizeof(u_int); - if (im6o == NULL) - *hlim = ip6_defmcasthlim; - else - *hlim = im6o->im6o_multicast_hlim; - return(0); - - case IPV6_MULTICAST_LOOP: - loop = mtod(*mp, u_int *); - (*mp)->m_len = sizeof(u_int); - if (im6o == NULL) - *loop = ip6_defmcasthlim; - else - *loop = im6o->im6o_multicast_loop; - return(0); + if (sa6->sin6_len != sizeof (struct sockaddr_in6)) + return (EINVAL); - default: - return(EOPNOTSUPP); - } -} + if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || + IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { + return (EINVAL); + } + if ((error = sa6_embedscope(sa6, ip6_use_defzone)) + != 0) { + return (error); + } + break; + } + case AF_LINK: /* should eventually be supported */ + default: + return (EAFNOSUPPORT); + } -/* - * Discard the IP6 multicast options. - */ -void -ip6_freemoptions(im6o) - register struct ip6_moptions *im6o; -{ - struct in6_multi_mship *imm; + /* turn off the previous option, then set the new option. */ + ip6_clearpktopts(opt, IPV6_NEXTHOP); + opt->ip6po_nexthop = _MALLOC(*buf, M_IP6OPT, M_NOWAIT); + if (opt->ip6po_nexthop == NULL) + return (ENOBUFS); + bcopy(buf, opt->ip6po_nexthop, *buf); + break; - if (im6o == NULL) - return; + case IPV6_2292HOPOPTS: + case IPV6_HOPOPTS: { + struct ip6_hbh *hbh; + int hbhlen; - while ((imm = im6o->im6o_memberships.lh_first) != NULL) { - LIST_REMOVE(imm, i6mm_chain); - if (imm->i6mm_maddr) - in6_delmulti(imm->i6mm_maddr); - _FREE(imm, M_IPMADDR); - } - _FREE(im6o, M_IPMOPTS); -} + /* + * XXX: We don't allow a non-privileged user to set ANY HbH + * options, since per-option restriction has too much + * overhead. + */ + error = suser(kauth_cred_get(), 0); + if (error) + return (EACCES); -/* - * Set IPv6 outgoing packet options based on advanced API. - */ -int -ip6_setpktoptions(control, opt, priv, needcopy) - struct mbuf *control; - struct ip6_pktopts *opt; - int priv, needcopy; -{ - register struct cmsghdr *cm = 0; + if (len == 0) { + ip6_clearpktopts(opt, IPV6_HOPOPTS); + break; /* just remove the option */ + } - if (control == 0 || opt == 0) - return(EINVAL); + /* message length validation */ + if (len < sizeof (struct ip6_hbh)) + return (EINVAL); + hbh = (struct ip6_hbh *)(void *)buf; + hbhlen = (hbh->ip6h_len + 1) << 3; + if (len != hbhlen) + return (EINVAL); - bzero(opt, sizeof(*opt)); - opt->ip6po_hlim = -1; /* -1 means to use default hop limit */ + /* turn off the previous option, then set the new option. */ + ip6_clearpktopts(opt, IPV6_HOPOPTS); + opt->ip6po_hbh = _MALLOC(hbhlen, M_IP6OPT, M_NOWAIT); + if (opt->ip6po_hbh == NULL) + return (ENOBUFS); + bcopy(hbh, opt->ip6po_hbh, hbhlen); - /* - * XXX: Currently, we assume all the optional information is stored - * in a single mbuf. - */ - if (control->m_next) - return(EINVAL); + break; + } - for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len), - control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { - cm = mtod(control, struct cmsghdr *); - if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) - return(EINVAL); - if (cm->cmsg_level != IPPROTO_IPV6) - continue; + case IPV6_2292DSTOPTS: + case IPV6_DSTOPTS: + case IPV6_RTHDRDSTOPTS: { + struct ip6_dest *dest, **newdest = NULL; + int destlen; - switch(cm->cmsg_type) { - case IPV6_PKTINFO: - if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo))) - return(EINVAL); - if (needcopy) { - /* XXX: Is it really WAITOK? */ - opt->ip6po_pktinfo = - _MALLOC(sizeof(struct in6_pktinfo), - M_IP6OPT, M_WAITOK); - *opt->ip6po_pktinfo = - *(struct in6_pktinfo *)CMSG_DATA(cm); - } else - opt->ip6po_pktinfo = - (struct in6_pktinfo *)CMSG_DATA(cm); - if (opt->ip6po_pktinfo->ipi6_ifindex && - IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr)) - opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] = - htons(opt->ip6po_pktinfo->ipi6_ifindex); - - if (opt->ip6po_pktinfo->ipi6_ifindex > if_index - || opt->ip6po_pktinfo->ipi6_ifindex < 0) { - return(ENXIO); - } + error = suser(kauth_cred_get(), 0); + if (error) + return (EACCES); - /* - * Check if the requested source address is indeed a - * unicast address assigned to the node. - */ - if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) { - struct ifaddr *ia; - struct sockaddr_in6 sin6; - - bzero(&sin6, sizeof(sin6)); - sin6.sin6_len = sizeof(sin6); - sin6.sin6_family = AF_INET6; - sin6.sin6_addr = - opt->ip6po_pktinfo->ipi6_addr; - ia = ifa_ifwithaddr(sin6tosa(&sin6)); - if (ia == NULL) - return(EADDRNOTAVAIL); - } - break; + if (len == 0) { + ip6_clearpktopts(opt, optname); + break; /* just remove the option */ + } - case IPV6_HOPLIMIT: - if (cm->cmsg_len != CMSG_LEN(sizeof(int))) - return(EINVAL); + /* message length validation */ + if (len < sizeof (struct ip6_dest)) + return (EINVAL); + dest = (struct ip6_dest *)(void *)buf; + destlen = (dest->ip6d_len + 1) << 3; + if (len != destlen) + return (EINVAL); - opt->ip6po_hlim = *(int *)CMSG_DATA(cm); - if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255) - return(EINVAL); + /* + * Determine the position that the destination options header + * should be inserted; before or after the routing header. + */ + switch (optname) { + case IPV6_2292DSTOPTS: + /* + * The old advacned API is ambiguous on this point. + * Our approach is to determine the position based + * according to the existence of a routing header. + * Note, however, that this depends on the order of the + * extension headers in the ancillary data; the 1st + * part of the destination options header must appear + * before the routing header in the ancillary data, + * too. + * RFC3542 solved the ambiguity by introducing + * separate ancillary data or option types. + */ + if (opt->ip6po_rthdr == NULL) + newdest = &opt->ip6po_dest1; + else + newdest = &opt->ip6po_dest2; break; - - case IPV6_NEXTHOP: - if (!priv) - return(EPERM); - - if (cm->cmsg_len < sizeof(u_char) || - /* check if cmsg_len is large enough for sa_len */ - cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm))) - return(EINVAL); - - if (needcopy) { - opt->ip6po_nexthop = - _MALLOC(*CMSG_DATA(cm), - M_IP6OPT, M_WAITOK); - bcopy(CMSG_DATA(cm), - opt->ip6po_nexthop, - *CMSG_DATA(cm)); - } else - opt->ip6po_nexthop = - (struct sockaddr *)CMSG_DATA(cm); + case IPV6_RTHDRDSTOPTS: + newdest = &opt->ip6po_dest1; break; - - case IPV6_HOPOPTS: - { - struct ip6_hbh *hbh; - int hbhlen; - - if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh))) - return(EINVAL); - hbh = (struct ip6_hbh *)CMSG_DATA(cm); - hbhlen = (hbh->ip6h_len + 1) << 3; - if (cm->cmsg_len != CMSG_LEN(hbhlen)) - return(EINVAL); - - if (needcopy) { - opt->ip6po_hbh = - _MALLOC(hbhlen, M_IP6OPT, M_WAITOK); - bcopy(hbh, opt->ip6po_hbh, hbhlen); - } else - opt->ip6po_hbh = hbh; + case IPV6_DSTOPTS: + newdest = &opt->ip6po_dest2; break; } - case IPV6_DSTOPTS: - { - struct ip6_dest *dest; - int destlen; + /* turn off the previous option, then set the new option. */ + ip6_clearpktopts(opt, optname); + *newdest = _MALLOC(destlen, M_IP6OPT, M_NOWAIT); + if (*newdest == NULL) + return (ENOBUFS); + bcopy(dest, *newdest, destlen); + break; + } - if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest))) - return(EINVAL); - dest = (struct ip6_dest *)CMSG_DATA(cm); - destlen = (dest->ip6d_len + 1) << 3; - if (cm->cmsg_len != CMSG_LEN(destlen)) - return(EINVAL); + case IPV6_2292RTHDR: + case IPV6_RTHDR: { + struct ip6_rthdr *rth; + int rthlen; - /* - * If there is no routing header yet, the destination - * options header should be put on the 1st part. - * Otherwise, the header should be on the 2nd part. - * (See RFC 2460, section 4.1) - */ - if (opt->ip6po_rthdr == NULL) { - if (needcopy) { - opt->ip6po_dest1 = - _MALLOC(destlen, M_IP6OPT, - M_WAITOK); - bcopy(dest, opt->ip6po_dest1, destlen); - } else - opt->ip6po_dest1 = dest; - } else { - if (needcopy) { - opt->ip6po_dest2 = - _MALLOC(destlen, M_IP6OPT, - M_WAITOK); - bcopy(dest, opt->ip6po_dest2, destlen); - } else - opt->ip6po_dest2 = dest; - } - break; + if (len == 0) { + ip6_clearpktopts(opt, IPV6_RTHDR); + break; /* just remove the option */ } - case IPV6_RTHDR: - { - struct ip6_rthdr *rth; - int rthlen; - - if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr))) - return(EINVAL); - rth = (struct ip6_rthdr *)CMSG_DATA(cm); - rthlen = (rth->ip6r_len + 1) << 3; - if (cm->cmsg_len != CMSG_LEN(rthlen)) - return(EINVAL); - - switch(rth->ip6r_type) { - case IPV6_RTHDR_TYPE_0: - /* must contain one addr */ - if (rth->ip6r_len == 0) - return(EINVAL); - /* length must be even */ - if (rth->ip6r_len % 2) - return(EINVAL); - if (rth->ip6r_len / 2 != rth->ip6r_segleft) - return(EINVAL); - break; - default: - return(EINVAL); /* not supported */ - } - - if (needcopy) { - opt->ip6po_rthdr = _MALLOC(rthlen, M_IP6OPT, - M_WAITOK); - bcopy(rth, opt->ip6po_rthdr, rthlen); - } else - opt->ip6po_rthdr = rth; + /* message length validation */ + if (len < sizeof (struct ip6_rthdr)) + return (EINVAL); + rth = (struct ip6_rthdr *)(void *)buf; + rthlen = (rth->ip6r_len + 1) << 3; + if (len != rthlen) + return (EINVAL); + switch (rth->ip6r_type) { + case IPV6_RTHDR_TYPE_0: + if (rth->ip6r_len == 0) /* must contain one addr */ + return (EINVAL); + if (rth->ip6r_len % 2) /* length must be even */ + return (EINVAL); + if (rth->ip6r_len / 2 != rth->ip6r_segleft) + return (EINVAL); break; + default: + return (EINVAL); /* not supported */ + } + + /* turn off the previous option */ + ip6_clearpktopts(opt, IPV6_RTHDR); + opt->ip6po_rthdr = _MALLOC(rthlen, M_IP6OPT, M_NOWAIT); + if (opt->ip6po_rthdr == NULL) + return (ENOBUFS); + bcopy(rth, opt->ip6po_rthdr, rthlen); + break; + } + + case IPV6_USE_MIN_MTU: + if (len != sizeof (int)) + return (EINVAL); + minmtupolicy = *(int *)(void *)buf; + if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && + minmtupolicy != IP6PO_MINMTU_DISABLE && + minmtupolicy != IP6PO_MINMTU_ALL) { + return (EINVAL); } + opt->ip6po_minmtu = minmtupolicy; + break; + + case IPV6_DONTFRAG: + if (len != sizeof (int)) + return (EINVAL); - case IPV6_REACHCONF: -#if 1 + if (uproto == IPPROTO_TCP || *(int *)(void *)buf == 0) { /* - * it looks dangerous to allow IPV6_REACHCONF to - * normal user. it affects the ND state (system state) - * and can affect communication by others - jinmei + * we ignore this option for TCP sockets. + * (RFC3542 leaves this case unspecified.) */ - if (!priv) - return(EPERM); -#endif - - if (cm->cmsg_len != CMSG_LEN(0)) - return(EINVAL); - opt->ip6po_flags |= IP6PO_REACHCONF; - break; - - case IPV6_USE_MIN_MTU: - if (cm->cmsg_len != CMSG_LEN(0)) - return(EINVAL); - opt->ip6po_flags |= IP6PO_MINMTU; - break; + opt->ip6po_flags &= ~IP6PO_DONTFRAG; + } else { + opt->ip6po_flags |= IP6PO_DONTFRAG; + } + break; - default: - return(ENOPROTOOPT); + case IPV6_PREFER_TEMPADDR: + if (len != sizeof (int)) + return (EINVAL); + preftemp = *(int *)(void *)buf; + if (preftemp != IP6PO_TEMPADDR_SYSTEM && + preftemp != IP6PO_TEMPADDR_NOTPREFER && + preftemp != IP6PO_TEMPADDR_PREFER) { + return (EINVAL); } - } + opt->ip6po_prefer_tempaddr = preftemp; + break; + + default: + return (ENOPROTOOPT); + } /* end of switch */ - return(0); + return (0); } /* @@ -3149,100 +3936,249 @@ ip6_setpktoptions(control, opt, priv, needcopy) * pointer that might NOT be &loif -- easier than replicating that code here. */ void -ip6_mloopback(ifp, m, dst) - struct ifnet *ifp; - register struct mbuf *m; - register struct sockaddr_in6 *dst; +ip6_mloopback(struct ifnet *srcifp, struct ifnet *origifp, struct mbuf *m, + struct sockaddr_in6 *dst, uint32_t optlen, int32_t nxt0) { - struct mbuf *copym; - - copym = m_copy(m, 0, M_COPYALL); - if (copym != NULL) { -#ifdef __APPLE__ - /* - * TedW: - * We need to send all loopback traffic down to dlil in case - * a filter has tapped-in. - */ - - if (lo_dl_tag == 0) - dlil_find_dltag(APPLE_IF_FAM_LOOPBACK, 0, PF_INET6, &lo_dl_tag); - - if (lo_dl_tag) - dlil_output(lo_dl_tag, copym, 0, (struct sockaddr *) dst, 0); - else { - printf("Warning: ip6_mloopback call to dlil_find_dltag failed!\n"); - m_freem(copym); - } -#else - (void)if_simloop(ifp, copym, (struct sockaddr *)dst, NULL); - (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL); -#endif + struct mbuf *copym; + struct ip6_hdr *ip6; + struct in6_addr src; + + if (lo_ifp == NULL) + return; + + /* + * Copy the packet header as it's needed for the checksum. + * Make sure to deep-copy IPv6 header portion in case the data + * is in an mbuf cluster, so that we can safely override the IPv6 + * header portion later. + */ + copym = m_copym_mode(m, 0, M_COPYALL, M_DONTWAIT, M_COPYM_COPY_HDR); + if (copym != NULL && ((copym->m_flags & M_EXT) || + copym->m_len < sizeof (struct ip6_hdr))) + copym = m_pullup(copym, sizeof (struct ip6_hdr)); + + if (copym == NULL) + return; + + ip6 = mtod(copym, struct ip6_hdr *); + src = ip6->ip6_src; + /* + * clear embedded scope identifiers if necessary. + * in6_clearscope will touch the addresses only when necessary. + */ + in6_clearscope(&ip6->ip6_src); + in6_clearscope(&ip6->ip6_dst); + + if (copym->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA) + in6_delayed_cksum_offset(copym, 0, optlen, nxt0); + + /* + * Stuff the 'real' ifp into the pkthdr, to be used in matching + * in ip6_input(); we need the loopback ifp/dl_tag passed as args + * to make the loopback driver compliant with the data link + * requirements. + */ + copym->m_pkthdr.rcvif = origifp; + + /* + * Also record the source interface (which owns the source address). + * This is basically a stripped down version of ifa_foraddr6(). + */ + if (srcifp == NULL) { + struct in6_ifaddr *ia; + + lck_rw_lock_shared(&in6_ifaddr_rwlock); + for (ia = in6_ifaddrs; ia != NULL; ia = ia->ia_next) { + IFA_LOCK_SPIN(&ia->ia_ifa); + /* compare against src addr with embedded scope */ + if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &src)) { + srcifp = ia->ia_ifp; + IFA_UNLOCK(&ia->ia_ifa); + break; + } + IFA_UNLOCK(&ia->ia_ifa); + } + lck_rw_done(&in6_ifaddr_rwlock); } + if (srcifp != NULL) + ip6_setsrcifaddr_info(copym, srcifp->if_index, NULL); + ip6_setdstifaddr_info(copym, origifp->if_index, NULL); + + dlil_output(lo_ifp, PF_INET6, copym, NULL, SA(dst), 0, NULL); } /* * Chop IPv6 header off from the payload. */ static int -ip6_splithdr(m, exthdrs) - struct mbuf *m; - struct ip6_exthdrs *exthdrs; +ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) { struct mbuf *mh; struct ip6_hdr *ip6; ip6 = mtod(m, struct ip6_hdr *); - if (m->m_len > sizeof(*ip6)) { - MGETHDR(mh, M_DONTWAIT, MT_HEADER); - if (mh == 0) { + if (m->m_len > sizeof (*ip6)) { + MGETHDR(mh, M_DONTWAIT, MT_HEADER); /* MAC-OK */ + if (mh == NULL) { m_freem(m); - return ENOBUFS; + return (ENOBUFS); } M_COPY_PKTHDR(mh, m); - MH_ALIGN(mh, sizeof(*ip6)); + MH_ALIGN(mh, sizeof (*ip6)); m->m_flags &= ~M_PKTHDR; - m->m_len -= sizeof(*ip6); - m->m_data += sizeof(*ip6); + m->m_len -= sizeof (*ip6); + m->m_data += sizeof (*ip6); mh->m_next = m; m = mh; - m->m_len = sizeof(*ip6); - bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); + m->m_len = sizeof (*ip6); + bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof (*ip6)); } exthdrs->ip6e_ip6 = m; - return 0; + return (0); +} + +static void +ip6_output_checksum(struct ifnet *ifp, uint32_t mtu, struct mbuf *m, + int nxt0, uint32_t tlen, uint32_t optlen) +{ + uint32_t sw_csum, hwcap = ifp->if_hwassist; + int tso = TSO_IPV6_OK(ifp, m); + + if (!hwcksum_tx) { + /* do all in software; checksum offload is disabled */ + sw_csum = CSUM_DELAY_IPV6_DATA & m->m_pkthdr.csum_flags; + } else { + /* do in software what the hardware cannot */ + sw_csum = m->m_pkthdr.csum_flags & + ~IF_HWASSIST_CSUM_FLAGS(hwcap); + } + + if (optlen != 0) { + sw_csum |= (CSUM_DELAY_IPV6_DATA & + m->m_pkthdr.csum_flags); + } else if (!(sw_csum & CSUM_DELAY_IPV6_DATA) && + (hwcap & CSUM_PARTIAL)) { + /* + * Partial checksum offload, ere), if no extension + * headers, and TCP only (no UDP support, as the + * hardware may not be able to convert +0 to + * -0 (0xffff) per RFC1122 4.1.3.4.) + */ + if (hwcksum_tx && !tso && + (m->m_pkthdr.csum_flags & CSUM_TCPIPV6) && + tlen <= mtu) { + uint16_t start = sizeof (struct ip6_hdr); + uint16_t ulpoff = + m->m_pkthdr.csum_data & 0xffff; + m->m_pkthdr.csum_flags |= + (CSUM_DATA_VALID | CSUM_PARTIAL); + m->m_pkthdr.csum_tx_stuff = (ulpoff + start); + m->m_pkthdr.csum_tx_start = start; + sw_csum = 0; + } else { + sw_csum |= (CSUM_DELAY_IPV6_DATA & + m->m_pkthdr.csum_flags); + } + } + + if (sw_csum & CSUM_DELAY_IPV6_DATA) { + in6_delayed_cksum_offset(m, 0, optlen, nxt0); + sw_csum &= ~CSUM_DELAY_IPV6_DATA; + } + + if (hwcksum_tx) { + /* + * Drop off bits that aren't supported by hardware; + * also make sure to preserve non-checksum related bits. + */ + m->m_pkthdr.csum_flags = + ((m->m_pkthdr.csum_flags & + (IF_HWASSIST_CSUM_FLAGS(hwcap) | CSUM_DATA_VALID)) | + (m->m_pkthdr.csum_flags & ~IF_HWASSIST_CSUM_MASK)); + } else { + /* drop all bits; checksum offload is disabled */ + m->m_pkthdr.csum_flags = 0; + } } /* * Compute IPv6 extension header length. */ -#if HAVE_NRL_INPCB -# define in6pcb inpcb -# define in6p_outputopts inp_outputopts6 -#endif int -ip6_optlen(in6p) - struct in6pcb *in6p; +ip6_optlen(struct in6pcb *in6p) { int len; if (!in6p->in6p_outputopts) - return 0; + return (0); len = 0; -#define elen(x) \ - (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) +#define elen(x) \ + (((struct ip6_ext *)(x)) ? \ + (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) len += elen(in6p->in6p_outputopts->ip6po_hbh); - if (in6p->in6p_outputopts->ip6po_rthdr) + if (in6p->in6p_outputopts->ip6po_rthdr) { /* dest1 is valid with rthdr only */ len += elen(in6p->in6p_outputopts->ip6po_dest1); + } len += elen(in6p->in6p_outputopts->ip6po_rthdr); len += elen(in6p->in6p_outputopts->ip6po_dest2); - return len; + return (len); #undef elen } -#if HAVE_NRL_INPCB -# undef in6pcb -# undef in6p_outputopts -#endif + +static int +sysctl_reset_ip6_output_stats SYSCTL_HANDLER_ARGS +{ +#pragma unused(arg1, arg2) + int error, i; + + i = ip6_output_measure; + error = sysctl_handle_int(oidp, &i, 0, req); + if (error || req->newptr == USER_ADDR_NULL) + goto done; + /* impose bounds */ + if (i < 0 || i > 1) { + error = EINVAL; + goto done; + } + if (ip6_output_measure != i && i == 1) { + net_perf_initialize(&net_perf, ip6_output_measure_bins); + } + ip6_output_measure = i; +done: + return (error); +} + +static int +sysctl_ip6_output_measure_bins SYSCTL_HANDLER_ARGS +{ +#pragma unused(arg1, arg2) + int error; + uint64_t i; + + i = ip6_output_measure_bins; + error = sysctl_handle_quad(oidp, &i, 0, req); + if (error || req->newptr == USER_ADDR_NULL) + goto done; + /* validate data */ + if (!net_perf_validate_bins(i)) { + error = EINVAL; + goto done; + } + ip6_output_measure_bins = i; +done: + return (error); +} + +static int +sysctl_ip6_output_getperf SYSCTL_HANDLER_ARGS +{ +#pragma unused(oidp, arg1, arg2) + if (req->oldptr == USER_ADDR_NULL) + req->oldlen = (size_t)sizeof (struct ipstat); + + return (SYSCTL_OUT(req, &net_perf, MIN(sizeof (net_perf), req->oldlen))); +} +