X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/316670eb35587141e969394ae8537d66b9211e80..d9a64523371fa019c4575bb400cbbc3a50ac9903:/bsd/netinet6/ip6_output.c?ds=sidebyside diff --git a/bsd/netinet6/ip6_output.c b/bsd/netinet6/ip6_output.c index d96db58f7..0720d6809 100644 --- a/bsd/netinet6/ip6_output.c +++ b/bsd/netinet6/ip6_output.c @@ -1,8 +1,8 @@ /* - * Copyright (c) 2000-2012 Apple Inc. All rights reserved. + * Copyright (c) 2000-2017 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 @@ -11,10 +11,10 @@ * 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, @@ -22,13 +22,10 @@ * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. * Please see the License for the specific language governing rights and * limitations under the License. - * + * * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ */ -/* $FreeBSD: src/sys/netinet6/ip6_output.c,v 1.43 2002/10/31 19:45:48 ume Exp $ */ -/* $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $ */ - /* * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * All rights reserved. @@ -113,72 +110,102 @@ #include #include #include +#include #include +#include #include #include #include +#include +#include +#include +#include #include #include #include #include #include +#include +#include + #include #include #include #include #include #include -#include - #if IPSEC #include -#if INET6 #include -#endif #include extern int ipsec_bypass; #endif /* IPSEC */ +#if NECP +#include +#endif /* NECP */ + #if CONFIG_MACF_NET #include -#endif /* MAC_NET */ - -#include +#endif /* CONFIG_MACF_NET */ #if DUMMYNET #include #include #endif /* DUMMYNET */ -#include - -#include - #if PF #include #endif /* PF */ -#ifndef __APPLE__ -static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options"); -#endif - -int ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt); -static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, - struct socket *, struct sockopt *sopt); -static int ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, int uproto); -static int ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt); -static int ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, int sticky, int cmsg, int uproto); -static void im6o_trace(struct ip6_moptions *, int); +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 int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, - struct ip6_frag **); +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 int ip6_getpmtu (struct route_in6 *, struct route_in6 *, - struct ifnet *, struct in6_addr *, u_int32_t *, int *); +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_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 */ @@ -213,218 +240,299 @@ static struct zone *im6o_zone; /* zone for ip6_moptions */ #define IM6O_ZONE_MAX 64 /* maximum elements in zone */ #define IM6O_ZONE_NAME "ip6_moptions" /* zone name */ -SYSCTL_DECL(_net_inet6_ip6); - -static int ip6_maxchainsent = 0; -SYSCTL_INT(_net_inet6_ip6, OID_AUTO, maxchainsent, CTLFLAG_RW | CTLFLAG_LOCKED, - &ip6_maxchainsent, 0, "use dlil_output_list"); - /* - * XXX we don't handle mbuf chains yet in nd6_output() so ip6_output_list() only - * walks through the packet chain and sends each mbuf separately. + * ip6_output() calls ip6_output_list() to do the work */ int -ip6_output_list( - struct mbuf *m0, - int packetlist, - struct ip6_pktopts *opt, - struct route_in6 *ro, - int flags, - struct ip6_moptions *im6o, - struct ifnet **ifpp, /* XXX: just for statistics */ - struct ip6_out_args *ip6oap) +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) { -#pragma unused(packetlist) - struct mbuf *m = m0, *nextpkt; - int error = 0; - - while (m) { - /* - * Break the chain before calling ip6_output() and free the - * mbufs if there was an error. - */ - nextpkt = m->m_nextpkt; - m->m_nextpkt = NULL; - error = ip6_output(m, opt, ro, flags, im6o, ifpp, ip6oap); - if (error) { - if (nextpkt) - m_freem_list(nextpkt); - return (error); - } - m = nextpkt; - } - - return (error); + 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( - struct mbuf *m0, - struct ip6_pktopts *opt, - struct route_in6 *ro, - int flags, - struct ip6_moptions *im6o, - struct ifnet **ifpp, /* XXX: just for statistics */ - struct ip6_out_args *ip6oap) +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 = NULL, *origifp = NULL; - struct mbuf *m = m0; - int hlen, tlen, len, off; - struct route_in6 ip6route; + 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 = 0; + struct route_in6 *ro_pmtu = NULL; struct rtentry *rt = NULL; - struct sockaddr_in6 *dst, src_sa, dst_sa; + struct sockaddr_in6 *dst = NULL, src_sa, dst_sa; int error = 0; - struct in6_ifaddr *ia = NULL; - u_int32_t mtu; - int alwaysfrag = 0, dontfrag = 0; + struct in6_ifaddr *ia = NULL, *src_ia = NULL; + u_int32_t mtu = 0; + boolean_t alwaysfrag = FALSE; u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; - struct ip6_exthdrs exthdrs; - struct in6_addr finaldst, src0, dst0; - u_int32_t zone; - struct route_in6 *ro_pmtu = NULL; - int hdrsplit = 0; - int needipsec = 0; + struct ip6_rthdr *rh; + struct in6_addr finaldst; ipfilter_t inject_filter_ref; - int tso; - boolean_t select_srcif; - struct ipf_pktopts *ippo = NULL, ipf_pktopts; - struct ip6_out_args ip6oa = { IFSCOPE_NONE, { 0 }, 0 }; + struct ipf_pktopts *ippo = NULL; struct flowadv *adv = NULL; - u_int32_t ifmtu; + uint32_t pktcnt = 0; + uint32_t packets_processed = 0; + struct timeval start_tv; #if DUMMYNET struct m_tag *tag; - struct route_in6 saved_route; - struct route_in6 saved_ro_pmtu; - struct ip_fw_args args; + struct ip6_out_args saved_ip6oa; struct sockaddr_in6 dst_buf; - - bzero(&args, sizeof(struct ip_fw_args)); #endif /* DUMMYNET */ - - if ((flags & IPV6_OUTARGS) && ip6oap != NULL) { - ip6oa = *ip6oap; - adv = &ip6oap->ip6oa_flowadv; - adv->code = FADV_SUCCESS; - } - #if IPSEC - int needipsectun = 0; struct socket *so = NULL; struct secpolicy *sp = NULL; struct route_in6 *ipsec_saved_route = NULL; - struct ipsec_output_state ipsec_state; + 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 }; - bzero(&ipsec_state, sizeof(ipsec_state)); + if (ip6_output_measure) + net_perf_start_time(&net_perf, &start_tv); - /* for AH processing. stupid to have "socket" variable in IP layer... */ - if (ipsec_bypass == 0) - { - so = ipsec_getsocket(m); - (void)ipsec_setsocket(m, NULL); - } -#endif /* IPSEC */ + VERIFY(m0->m_flags & M_PKTHDR); - bzero(&ipf_pktopts, sizeof(struct ipf_pktopts)); - ippo = &ipf_pktopts; + /* zero out {saved_route, saved_ro_pmtu, ip6route, exthdrs, args} */ + bzero(&ip6obz, sizeof (ip6obz)); - ip6 = mtod(m, struct ip6_hdr *); - inject_filter_ref = ipf_get_inject_filter(m); +#if DUMMYNET + if (SLIST_EMPTY(&m0->m_pkthdr.tags)) + goto tags_done; /* Grab info from mtags prepended to the chain */ -#if DUMMYNET 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)); + bcopy(&dn_tag->dn_dst6, &dst_buf, sizeof (dst_buf)); dst = &dst_buf; ifp = dn_tag->dn_ifp; - if (ifp) + if (ifp != NULL) ifnet_reference(ifp); flags = dn_tag->dn_flags; - if (dn_tag->dn_flags & IPV6_OUTARGS) - ip6oa = dn_tag->dn_ip6oa; + 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) + if (origifp != NULL) ifnet_reference(origifp); mtu = dn_tag->dn_mtu; - alwaysfrag = dn_tag->dn_alwaysfrag; + alwaysfrag = (dn_tag->dn_alwaysfrag != 0); unfragpartlen = dn_tag->dn_unfragpartlen; - bcopy(&dn_tag->dn_exthdrs, &exthdrs, sizeof(exthdrs)); + bcopy(&dn_tag->dn_exthdrs, &exthdrs, sizeof (exthdrs)); m_tag_delete(m0, tag); } + +tags_done: #endif /* DUMMYNET */ - finaldst = ip6->ip6_dst; + m = m0; + +#if IPSEC + 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 & IP6OAF_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; + + ip6obf.noipsec = (noipsec != 0); + } + } +#endif /* IPSEC */ - if (ip6_doscopedroute && (flags & IPV6_OUTARGS)) { - if ((select_srcif = (!(flags & (IPV6_FORWARDING | + ippo = &ipf_pktopts; + + if (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)))) + (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) { + 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)); + (ip6oa->ip6oa_boundif << IPPOF_SHIFT_IFSCOPE)); } - if (ip6oa.ip6oa_flags & IP6OAF_BOUND_SRCADDR) + if (ip6oa->ip6oa_flags & IP6OAF_BOUND_SRCADDR) ipf_pktopts.ippo_flags |= IPPOF_BOUND_SRCADDR; } else { - select_srcif = FALSE; - ip6oa.ip6oa_boundif = IFSCOPE_NONE; - ip6oa.ip6oa_flags &= ~(IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_IF | - IP6OAF_BOUND_SRCADDR); + 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) && (ip6oa.ip6oa_flags & IP6OAF_NO_CELLULAR)) - ipf_pktopts.ippo_flags |= IPPOF_NO_IFT_CELLULAR; + 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 */ -#define MAKE_EXTHDR(hp, mp) \ - do { \ - if (hp) { \ +#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) 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; \ } \ - } while (0) - - bzero(&exthdrs, sizeof(exthdrs)); +} 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) */ @@ -447,19 +555,90 @@ ip6_output( 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; + ip6stat.ip6s_necp_policy_drop++; + 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; + ip6stat.ip6s_necp_policy_drop++; + goto freehdrs; + } + } + } + default: + break; + } + } +#endif /* NECP */ + #if IPSEC - if (ipsec_bypass != 0) + if (ipsec_bypass != 0 || ip6obf.noipsec) goto skip_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 (sp == NULL) { - IPSEC_STAT_INCREMENT(ipsec6stat.out_inval); - goto freehdrs; + /* 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; @@ -477,7 +656,7 @@ ip6_output( case IPSEC_POLICY_BYPASS: case IPSEC_POLICY_NONE: /* no need to do IPsec. */ - needipsec = 0; + ip6obf.needipsec = FALSE; break; case IPSEC_POLICY_IPSEC: @@ -486,14 +665,19 @@ ip6_output( error = key_spdacquire(sp); goto freehdrs; } - needipsec = 1; + if (sp->ipsec_if) { + goto skip_ipsec; + } else { + ip6obf.needipsec = TRUE; + } 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: +skip_ipsec: #endif /* IPSEC */ /* @@ -501,36 +685,33 @@ ip6_output( * Keep the length of the unfragmentable part for fragmentation. */ optlen = 0; - if (exthdrs.ip6e_hbh) + if (exthdrs.ip6e_hbh != NULL) optlen += exthdrs.ip6e_hbh->m_len; - if (exthdrs.ip6e_dest1) + if (exthdrs.ip6e_dest1 != NULL) optlen += exthdrs.ip6e_dest1->m_len; - if (exthdrs.ip6e_rthdr) + if (exthdrs.ip6e_rthdr != NULL) optlen += exthdrs.ip6e_rthdr->m_len; - unfragpartlen = optlen + sizeof(struct ip6_hdr); + unfragpartlen = optlen + sizeof (struct ip6_hdr); /* NOTE: we don't add AH/ESP length here. do that later. */ - if (exthdrs.ip6e_dest2) + if (exthdrs.ip6e_dest2 != NULL) optlen += exthdrs.ip6e_dest2->m_len; - - if (needipsec && - (m->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA) != 0) { - in6_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen); - m->m_pkthdr.csum_flags &= ~CSUM_DELAY_IPV6_DATA; - } - /* * 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 */ @@ -538,26 +719,26 @@ ip6_output( /* 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" @@ -568,107 +749,122 @@ ip6_output( * 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. 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) { - 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)\ - do {\ - 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);\ - }\ - } 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); - - if (!TAILQ_EMPTY(&ipv6_filters)) { - struct ipfilter *filter; - int seen = (inject_filter_ref == 0); - 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); - } + /* + * 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; + } - /* 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) { - /* - * No need to proccess 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) { - errno_t result; - - result = filter->ipf_filter.ipf_output(filter->ipf_filter.cookie, (mbuf_t*)&m, ippo); - if (result == EJUSTRETURN) { - ipf_unref(); - goto done; - } - if (result != 0) { - ipf_unref(); - goto bad; - } - } +#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) && + !((flags & IPV6_OUTARGS) && + (ip6oa->ip6oa_flags & IP6OAF_INTCOPROC_ALLOWED))) { + 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(); + 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; } + 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. @@ -677,19 +873,18 @@ ip6_output( */ exthdrs.ip6e_dest2 = NULL; - { - struct ip6_rthdr *rh = NULL; - int segleft_org = 0; - - 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; } ipsec_state.m = m; - error = ipsec6_output_trans(&ipsec_state, nexthdrp, mprev, sp, flags, - &needipsectun); + 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. */ @@ -702,8 +897,9 @@ ip6_output( 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; @@ -711,66 +907,63 @@ ip6_output( } 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 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; + 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 *)(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; + 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_UNSPECSRC) == 0) { + !(flags & IPV6_UNSPECSRC)) { error = EOPNOTSUPP; ip6stat.ip6s_badscope++; goto bad; @@ -786,37 +979,39 @@ 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->ro_rt) + if (ro->ro_rt != NULL) RT_LOCK_ASSERT_NOTHELD(ro->ro_rt); /* * 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 (opt && opt->ip6po_tclass >= 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); + 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) + if (opt && opt->ip6po_hlim != -1) { ip6->ip6_hlim = opt->ip6po_hlim & 0xff; - else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { + } else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { if (im6o != NULL) { IM6O_LOCK(im6o); ip6->ip6_hlim = im6o->im6o_multicast_hlim; @@ -834,24 +1029,21 @@ skip_ipsec2: * below (since it uses this route as a hint) or during the * next transmit. */ - if (ro->ro_rt != NULL && (!(ro->ro_rt->rt_flags & RTF_UP) || - dst->sin6_family != AF_INET6 || - !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst) || - ro->ro_rt->generation_id != route_generation)) { - rtfree(ro->ro_rt); - ro->ro_rt = NULL; - } + 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)); + 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) { + if (ip6obf.needipsec && needipsectun) { #if CONFIG_DTRACE - struct ifnet *trace_ifp = (ifpp != NULL) ? (*ifpp) : NULL; -#endif + 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). @@ -860,28 +1052,31 @@ skip_ipsec2: * * IPv6 [ESP|AH] IPv6 [extension headers] payload */ - bzero(&exthdrs, sizeof(exthdrs)); + bzero(&exthdrs, sizeof (exthdrs)); exthdrs.ip6e_ip6 = m; ipsec_state.m = m; - route_copyout(&ipsec_state.ro, (struct route *)ro, sizeof(ipsec_state.ro)); - ipsec_state.dst = (struct sockaddr *)dst; + route_copyout((struct route *)&ipsec_state.ro, (struct route *)ro, + sizeof (struct route_in6)); + ipsec_state.dst = SA(dst); - /* Added a trace here so that we can see packets inside a tunnel */ + /* 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); + struct ip6_hdr *, ip6, struct ifnet *, trace_ifp, + struct ip *, NULL, struct ip6_hdr *, ip6); error = ipsec6_output_tunnel(&ipsec_state, sp, flags); - if (ipsec_state.tunneled == 4) /* tunneled in IPv4 - packet is gone */ - goto done; + /* 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 = (struct sockaddr_in6 *)(void *)ipsec_state.dst; + dst = SIN6(ipsec_state.dst); if (error) { /* mbuf is already reclaimed in ipsec6_output_tunnel. */ - m0 = m = NULL; m = NULL; switch (error) { case EHOSTUNREACH: @@ -891,8 +1086,9 @@ 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; @@ -901,11 +1097,14 @@ skip_ipsec2: goto bad; } /* - * The packet has been encapsulated so the ifscope is no longer valid - * since it does not apply to the outer address: ignore the ifscope. + * The packet has been encapsulated so the ifscope + * is no longer valid since it does not apply to the + * outer address: ignore the ifscope. */ - ip6oa.ip6oa_boundif = IFSCOPE_NONE; - ip6oa.ip6oa_flags &= ~IP6OAF_BOUND_IF; + if (flags & IPV6_OUTARGS) { + ip6oa->ip6oa_boundif = IFSCOPE_NONE; + ip6oa->ip6oa_flags &= ~IP6OAF_BOUND_IF; + } if (opt != NULL && opt->ip6po_pktinfo != NULL) { if (opt->ip6po_pktinfo->ipi6_ifindex != IFSCOPE_NONE) opt->ip6po_pktinfo->ipi6_ifindex = IFSCOPE_NONE; @@ -914,44 +1113,55 @@ skip_ipsec2: } #endif /* IPSEC */ - /* for safety */ + /* + * ifp should only be filled in for dummy net packets which will jump + * to check_with_pf label. + */ if (ifp != NULL) { - ifnet_release(ifp); - ifp = NULL; + VERIFY(ip6obf.route_selected); } /* adjust pointer */ ip6 = mtod(m, struct ip6_hdr *); - if (select_srcif) { - bzero(&src_sa, sizeof(src_sa)); + 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_len = sizeof (src_sa); src_sa.sin6_addr = ip6->ip6_src; } - bzero(&dst_sa, sizeof(dst_sa)); + bzero(&dst_sa, sizeof (dst_sa)); dst_sa.sin6_family = AF_INET6; - dst_sa.sin6_len = sizeof(dst_sa); + dst_sa.sin6_len = sizeof (dst_sa); dst_sa.sin6_addr = ip6->ip6_dst; /* + * 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 ((error = in6_selectroute(select_srcif ? &src_sa : NULL, - &dst_sa, opt, im6o, ro, &ifp, &rt, 0, &ip6oa)) != 0) { - switch (error) { - case EHOSTUNREACH: - ip6stat.ip6s_noroute++; - break; - case EADDRNOTAVAIL: - default: - break; /* XXX statistics? */ + 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; } - 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) { /* @@ -961,26 +1171,37 @@ skip_ipsec2: *dst = dst_sa; /* XXX */ } +#if NECP + /* Catch-all to check if the interface is allowed */ + if (!necp_packet_is_allowed_over_interface(m, ifp)) { + error = EHOSTUNREACH; + ip6stat.ip6s_necp_policy_drop++; + goto bad; + } +#endif /* NECP */ + /* * then rt (for unicast) and ifp must be non-NULL valid values. */ - if ((flags & IPV6_FORWARDING) == 0) { - /* XXX: the FORWARDING flag can be set for mrouting. */ - in6_ifstat_inc(ifp, ifs6_out_request); + if (!(flags & IPV6_FORWARDING)) { + in6_ifstat_inc_na(ifp, ifs6_out_request); } if (rt != NULL) { RT_LOCK(rt); - ia = (struct in6_ifaddr *)(rt->rt_ifa); - if (ia != NULL) - IFA_ADDREF(&ia->ia_ifa); + 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. We should use ia_ifp to support the - * case of sending packets to an address of our own. + * 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 */ @@ -994,50 +1215,78 @@ skip_ipsec2: ifnet_release(origifp); origifp = ifp; } - 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; + /* 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; - badscope: - ip6stat.ip6s_badscope++; - in6_ifstat_inc(origifp, ifs6_out_discard); - if (error == 0) - error = EHOSTUNREACH; /* XXX */ - goto bad; + 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 && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { - if (opt && opt->ip6po_nextroute.ro_rt) { +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 = (struct sockaddr_in6 *)(void *)opt->ip6po_nexthop; + dst = SIN6(opt->ip6po_nexthop); + } else if ((rt->rt_flags & RTF_GATEWAY)) { + dst = SIN6(rt->rt_gateway); + } + /* + * 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 (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); } - else if ((rt->rt_flags & RTF_GATEWAY)) - dst = (struct sockaddr_in6 *)(void *)rt->rt_gateway; } if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { @@ -1046,13 +1295,12 @@ skip_ipsec2: struct in6_multi *in6m; m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; - - in6_ifstat_inc(ifp, ifs6_out_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; @@ -1064,7 +1312,7 @@ skip_ipsec2: 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); /* @@ -1072,41 +1320,9 @@ skip_ipsec2: * on the outgoing interface, and the caller did not * forbid loopback, loop back a copy. */ - ip6_mloopback(ifp, m, dst); - } else { - if (im6o != NULL) - IM6O_UNLOCK(im6o); - /* - * 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 MROUTING - if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) { - /* - * XXX: ip6_mforward expects that rcvif is NULL - * when it is called from the originating path. - * However, it is not always the case, since - * some versions of MGETHDR() does not - * initialize the field. - */ - m->m_pkthdr.rcvif = NULL; - if (ip6_mforward(ip6, ifp, m) != 0) { - m_freem(m); - if (in6m != NULL) - IN6M_REMREF(in6m); - goto done; - } - } -#endif - } + ip6_mloopback(NULL, ifp, m, dst, optlen, nxt0); + } else if (im6o != NULL) + IM6O_UNLOCK(im6o); if (in6m != NULL) IN6M_REMREF(in6m); /* @@ -1119,8 +1335,11 @@ skip_ipsec2: */ if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) || IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) { - m_freem(m); - goto done; + /* remove m from the packetchain and continue looping */ + if (m != NULL) + m_freem(m); + m = NULL; + goto evaluateloop; } } @@ -1128,10 +1347,8 @@ skip_ipsec2: * Fill the outgoing inteface to tell the upper layer * to increment per-interface statistics. */ - if (ifpp != NULL) { + if (ifpp != NULL && *ifpp == NULL) { ifnet_reference(ifp); /* for caller */ - if (*ifpp != NULL) - ifnet_release(*ifpp); *ifpp = ifp; } @@ -1153,13 +1370,13 @@ skip_ipsec2: * See RFC 3542 for more details. */ if (mtu > IPV6_MMTU) { - if ((flags & IPV6_MINMTU)) + if ((flags & IPV6_MINMTU)) { mtu = IPV6_MMTU; - else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) + } 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)) { + } else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && + (opt == NULL || + opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) { mtu = IPV6_MMTU; } } @@ -1170,53 +1387,37 @@ skip_ipsec2: */ in6_clearscope(&ip6->ip6_src); in6_clearscope(&ip6->ip6_dst); - -#if IPFW2 - /* - * Check with the firewall... - */ - 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; - } - } -#endif - /* * 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) { + 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), - &dummy, &plen) < 0) { - /* m was already freed at this point */ - error = EINVAL;/* better error? */ - goto done; + ((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; @@ -1224,18 +1425,20 @@ skip_ipsec2: #if DUMMYNET check_with_pf: -#endif +#endif /* DUMMYNET */ #if PF if (PF_IS_ENABLED) { #if DUMMYNET + /* - * TBD: Need to save opt->ip6po_flags for reinjection rdar://10434993 + * 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; + if (flags & IPV6_OUTARGS) + args.fwa_ip6oa = ip6oa; args.fwa_ro6 = ro; args.fwa_dst6 = dst; args.fwa_ro6_pmtu = ro_pmtu; @@ -1246,64 +1449,241 @@ check_with_pf: args.fwa_exthdrs = &exthdrs; /* Invoke outbound packet filter */ error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, &args); -#else +#else /* !DUMMYNET */ error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, NULL); -#endif /* DUMMYNET */ +#endif /* !DUMMYNET */ if (error != 0 || m == NULL) { - /* - * Note that if we ever handle packet chain, we will - * have to restore the linkage from the previous - * packet to the next like in ip_outout_list() - */ if (m != NULL) { - panic("%s: unexpected packet %p\n", __func__, m); + panic("%s: unexpected packet %p\n", + __func__, m); /* NOTREACHED */ } - /* Already freed by callee */ - goto done; + /* 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 */ + + if (ip6oa != NULL) { + u_int8_t dscp; + + dscp = (ntohl(ip6->ip6_flow) & IP6FLOW_DSCP_MASK) >> IP6FLOW_DSCP_SHIFT; + + error = set_packet_qos(m, ifp, + ip6oa->ip6oa_flags & IP6OAF_QOSMARKING_ALLOWED ? TRUE : FALSE, + ip6oa->ip6oa_sotc, ip6oa->ip6oa_netsvctype, &dscp); + if (error == 0) { + ip6->ip6_flow &= ~htonl(IP6FLOW_DSCP_MASK); + ip6->ip6_flow |= htonl((u_int32_t)dscp << IP6FLOW_DSCP_SHIFT); + } else { + printf("%s if_dscp_for_mbuf() error %d\n", __func__, error); + error = 0; + } + } /* - * Send the packet to the outgoing interface. - * If necessary, do IPv6 fragmentation before sending. - * - * the logic here 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 + * 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; + 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; + } + + /* 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 (m->m_pkthdr.pkt_flags & PKTF_FORWARDED) { + dontfrag = TRUE; + /* + * Discard partial sum information if this packet originated + * from another interface; the packet would already have the + * final checksum and we shouldn't recompute it. + */ + if ((m->m_pkthdr.csum_flags & (CSUM_DATA_VALID|CSUM_PARTIAL)) == + (CSUM_DATA_VALID|CSUM_PARTIAL)) { + m->m_pkthdr.csum_flags &= ~CSUM_TX_FLAGS; + m->m_pkthdr.csum_data = 0; + } + } - if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) - dontfrag = 1; - else - dontfrag = 0; if (dontfrag && alwaysfrag) { /* case 4 */ /* conflicting request - can't transmit */ - error = EMSGSIZE; - goto bad; + return EMSGSIZE; } - lck_rw_lock_shared(nd_if_rwlock); /* Access without acquiring nd_ifinfo lock for performance */ - ifmtu = IN6_LINKMTU(ifp); - lck_rw_done(nd_if_rwlock); - - if (dontfrag && tlen > ifmtu) { /* case 2-b */ + if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */ /* * Even if the DONTFRAG option is specified, we cannot send the * packet when the data length is larger than the MTU of the @@ -1316,61 +1696,74 @@ check_with_pf: struct ip6ctlparam ip6cp; mtu32 = (u_int32_t)mtu; - bzero(&ip6cp, sizeof(ip6cp)); + bzero(&ip6cp, sizeof (ip6cp)); ip6cp.ip6c_cmdarg = (void *)&mtu32; - pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst, - (void *)&ip6cp); - - error = EMSGSIZE; - goto bad; + pfctlinput2(PRC_MSGSIZE, SA(&ro_pmtu->ro_dst), (void *)&ip6cp); + return EMSGSIZE; } /* * transmit packet without fragmentation */ - tso = (ifp->if_hwassist & IFNET_TSO_IPV6) && - (m->m_pkthdr.csum_flags & CSUM_TSO_IPV6); if (dontfrag || (!alwaysfrag && /* case 1-a and 2-a */ - (tlen <= mtu || tso || (ifp->if_hwassist & CSUM_FRAGMENT_IPV6)))) { - int sw_csum; + (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); + } - ip6 = mtod(m, struct ip6_hdr *); -#ifdef IPSEC - /* clean ipsec history once it goes out of the node */ - ipsec_delaux(m); -#endif + return error; +} - if (apple_hwcksum_tx == 0) /* Do not let HW handle cksum */ - sw_csum = m->m_pkthdr.csum_flags; - else - sw_csum = m->m_pkthdr.csum_flags & - ~IF_HWASSIST_CSUM_FLAGS(ifp->if_hwassist); +/* + * 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. + */ +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; - if ((sw_csum & CSUM_DELAY_IPV6_DATA) != 0) { - in6_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen); - m->m_pkthdr.csum_flags &= ~CSUM_DELAY_IPV6_DATA; - } - if (ro->ro_rt) - RT_LOCK_ASSERT_NOTHELD(ro->ro_rt); - error = nd6_output(ifp, origifp, m, dst, ro->ro_rt, adv); - goto done; - } + 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 (mtu < IPV6_MMTU) { + if ((morig->m_pkthdr.csum_flags & CSUM_TSO_IPV6)) { + /* TSO and fragment aren't compatible */ + in6_ifstat_inc(ifp, ifs6_out_fragfail); + return EMSGSIZE; + } else if (mtu < IPV6_MMTU) { /* path MTU cannot be less than IPV6_MMTU */ - error = EMSGSIZE; in6_ifstat_inc(ifp, ifs6_out_fragfail); - goto bad; + return EMSGSIZE; } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */ - error = EMSGSIZE; in6_ifstat_inc(ifp, ifs6_out_fragfail); - goto bad; + 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_randomid()); u_char nextproto; @@ -1384,236 +1777,323 @@ check_with_pf: 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 ((m->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA) != 0) { - in6_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen); - m->m_pkthdr.csum_flags &= ~CSUM_DELAY_IPV6_DATA; - } + 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. */ - m0 = m; for (off = hlen; off < tlen; off += len) { - MGETHDR(m, M_DONTWAIT, MT_HEADER); /* MAC-OK */ - 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; } - m->m_pkthdr.rcvif = NULL; - 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) { + 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; + } + 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 = 0; - m->m_pkthdr.socket_id = m0->m_pkthdr.socket_id; + m_cat(new_m, m_frgpart); + new_m->m_pkthdr.len = len + hlen + sizeof (*ip6f); + new_m->m_pkthdr.rcvif = NULL; - M_COPY_PFTAG(m, m0); - m_set_service_class(m, m0->m_pkthdr.svc); + M_COPY_CLASSIFIER(new_m, morig); + M_COPY_PFTAG(new_m, morig); -#ifdef __darwin8_notyet +#ifdef notyet #if CONFIG_MACF_NET - mac_create_fragment(m0, m); -#endif -#endif + mac_create_fragment(morig, new_m); +#endif /* CONFIG_MACF_NET */ +#endif /* notyet */ + ip6f->ip6f_reserved = 0; ip6f->ip6f_ident = id; ip6f->ip6f_nxt = nextproto; ip6stat.ip6s_ofragments++; - 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) { - /* Record statistics for this interface address. */ - if (ia) { -#ifndef __APPLE__ - ia->ia_ifa.if_opackets++; - ia->ia_ifa.if_obytes += m->m_pkthdr.len; -#endif - } -#if IPSEC - /* clean ipsec history once it goes out of the node */ - ipsec_delaux(m); -#endif - error = nd6_output(ifp, origifp, m, dst, ro->ro_rt, - adv); - - } else - m_freem(m); - } - - if (error == 0) - ip6stat.ip6s_fragmented++; - -done: -#if IPSEC - if (ipsec_saved_route) { - ro = ipsec_saved_route; - if (ipsec_state.ro.ro_rt) { - rtfree(ipsec_state.ro.ro_rt); - } - } -#endif /* IPSEC */ - 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, KEY_SADB_UNLOCKED); -#endif /* IPSEC */ - - if (ia != NULL) - IFA_REMREF(&ia->ia_ifa); - if (ifp != NULL) - ifnet_release(ifp); - if (origifp != NULL) - ifnet_release(origifp); - return (error); + in6_ifstat_inc(ifp, ifs6_out_fragcreat); + } -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; + if (error) { + /* free all the fragments created */ + if (first_mbufp != NULL) { + m_freem_list(first_mbufp); + first_mbufp = NULL; + } + 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; } static int -ip6_copyexthdr(mp, hdr, hlen) - struct mbuf **mp; - caddr_t hdr; - int hlen; +ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen) { struct mbuf *m; if (hlen > MCLBYTES) - return(ENOBUFS); /* XXX */ + 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); } } m->m_len = hlen; - if (hdr) + if (hdr != NULL) bcopy(hdr, mtod(m, caddr_t), hlen); *mp = m; 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. + * 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. */ -void -in6_delayed_cksum(struct mbuf *m, uint16_t offset) +uint32_t +in6_finalize_cksum(struct mbuf *m, uint32_t hoff, int32_t optlen, + int32_t nxt0, uint32_t csum_flags) { - uint16_t csum; + 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; - csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset); - if (csum == 0 && (m->m_pkthdr.csum_flags & CSUM_UDPIPV6) != 0) { - csum = 0xffff; + _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 += (m->m_pkthdr.csum_data & 0xffff); - if ((offset + sizeof(csum)) > m->m_len) { - m_copyback(m, offset, sizeof(csum), &csum); - } else if (IP6_HDR_ALIGNED_P(mtod(m, char *))) { + /* + * 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_ZERO_INVERT))) + 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 | CSUM_ZERO_INVERT); + +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. @@ -1621,9 +2101,9 @@ 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) + if (mopt == NULL) return (ENOBUFS); mopt->m_len = JUMBOOPTLEN; optbuf = mtod(mopt, u_char *); @@ -1655,14 +2135,14 @@ ip6_insert_jumboopt(exthdrs, plen) * at this point. */ MGET(n, M_DONTWAIT, MT_DATA); - if (n) { + if (n != NULL) { MCLGET(n, M_DONTWAIT); - if ((n->m_flags & M_EXT) == 0) { + if (!(n->m_flags & M_EXT)) { m_freem(n); n = NULL; } } - if (!n) + if (n == NULL) return (ENOBUFS); n->m_len = oldoptlen + JUMBOOPTLEN; bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), @@ -1689,7 +2169,7 @@ ip6_insert_jumboopt(exthdrs, plen) optbuf[2] = IP6OPT_JUMBO; optbuf[3] = 4; v = (u_int32_t)htonl(plen + JUMBOOPTLEN); - bcopy(&v, &optbuf[4], sizeof(u_int32_t)); + bcopy(&v, &optbuf[4], sizeof (u_int32_t)); /* finally, adjust the packet header length */ exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN; @@ -1702,17 +2182,15 @@ 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) + 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 @@ -1722,21 +2200,21 @@ 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 */ + 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); + 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) + if (mfrg == NULL) return (ENOBUFS); - mfrg->m_len = sizeof(struct ip6_frag); + mfrg->m_len = sizeof (struct ip6_frag); *frghdrp = mtod(mfrg, struct ip6_frag *); mlast->m_next = mfrg; } @@ -1744,31 +2222,30 @@ ip6_insertfraghdr(m0, m, hlen, frghdrp) return (0); } -extern int load_ipfw(void); static int ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, struct ifnet *ifp, struct in6_addr *dst, u_int32_t *mtup, - int *alwaysfragp) + boolean_t *alwaysfragp) { u_int32_t mtu = 0; - int alwaysfrag = 0; + boolean_t alwaysfrag = FALSE; int error = 0; + boolean_t is_local = FALSE; + + if (IN6_IS_SCOPE_LINKLOCAL(dst)) + is_local = TRUE; if (ro_pmtu != ro) { /* The first hop and the final destination may differ. */ - struct sockaddr_in6 *sa6_dst = - (struct sockaddr_in6 *)&ro_pmtu->ro_dst; - if (ro_pmtu->ro_rt && - ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 || - ro_pmtu->ro_rt->generation_id != route_generation || - !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) { - rtfree(ro_pmtu->ro_rt); - ro_pmtu->ro_rt = (struct rtentry *)NULL; - } + 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)); + bzero(sa6_dst, sizeof (*sa6_dst)); sa6_dst->sin6_family = AF_INET6; - sa6_dst->sin6_len = sizeof(struct sockaddr_in6); + sa6_dst->sin6_len = sizeof (struct sockaddr_in6); sa6_dst->sin6_addr = *dst; rtalloc_scoped((struct route *)ro_pmtu, @@ -1776,16 +2253,19 @@ ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, } } - if (ro_pmtu->ro_rt != NULL) { u_int32_t ifmtu; - lck_rw_lock_shared(nd_if_rwlock); + if (ifp == NULL) + ifp = ro_pmtu->ro_rt->rt_ifp; /* Access without acquiring nd_ifinfo lock for performance */ ifmtu = IN6_LINKMTU(ifp); - lck_rw_done(nd_if_rwlock); - RT_LOCK_SPIN(ro_pmtu->ro_rt); + /* + * 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) { /* @@ -1799,11 +2279,10 @@ ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, * 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) == 0) - ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */ - } - else if (mtu < IPV6_MMTU) { + 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 @@ -1812,22 +2291,20 @@ ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, * (fragment header is needed regardless from the * packet size, for translators to identify packets) */ - alwaysfrag = 1; + alwaysfrag = TRUE; mtu = IPV6_MMTU; - } - RT_UNLOCK(ro_pmtu->ro_rt); + } } else { if (ifp) { - lck_rw_lock_shared(nd_if_rwlock); /* Don't hold nd_ifinfo lock for performance */ mtu = IN6_LINKMTU(ifp); - lck_rw_done(nd_if_rwlock); - } else + } else { error = EHOSTUNREACH; /* XXX */ + } } *mtup = mtu; - if (alwaysfragp) + if ((alwaysfragp != NULL) && !is_local) *alwaysfragp = alwaysfrag; return (error); } @@ -1836,9 +2313,7 @@ ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, * IP6 socket option processing. */ int -ip6_ctloutput(so, sopt) - struct socket *so; - struct sockopt *sopt; +ip6_ctloutput(struct socket *so, struct sockopt *sopt) { int optdatalen, uproto; void *optdata; @@ -1849,37 +2324,34 @@ ip6_ctloutput(so, sopt) int optlen = 0; struct proc *p; - if (sopt == NULL) { - panic("ip6_ctloutput: arg soopt is NULL"); - /* NOTREACHED */ - } + VERIFY(sopt != NULL); + level = sopt->sopt_level; op = sopt->sopt_dir; optname = sopt->sopt_name; optlen = sopt->sopt_valsize; p = sopt->sopt_p; - uproto = (int)so->so_proto->pr_protocol; + uproto = (int)SOCK_PROTO(so); privileged = (proc_suser(p) == 0); if (level == IPPROTO_IPV6) { + boolean_t capture_exthdrstat_in = FALSE; switch (op) { - case SOPT_SET: switch (optname) { - case IPV6_2292PKTOPTIONS: - { + case IPV6_2292PKTOPTIONS: { struct mbuf *m; - error = soopt_getm(sopt, &m); /* XXX */ + error = soopt_getm(sopt, &m); if (error != 0) break; - error = soopt_mcopyin(sopt, m); /* XXX */ + error = soopt_mcopyin(sopt, m); if (error != 0) break; error = ip6_pcbopts(&in6p->in6p_outputopts, - m, so, sopt); - m_freem(m); /* XXX */ + m, so, sopt); + m_freem(m); break; } @@ -1899,12 +2371,11 @@ ip6_ctloutput(so, sopt) case IPV6_RECVHOPOPTS: case IPV6_RECVDSTOPTS: case IPV6_RECVRTHDRDSTOPTS: - if (!privileged) - break; + if (!privileged) + break; /* FALLTHROUGH */ case IPV6_UNICAST_HOPS: case IPV6_HOPLIMIT: - case IPV6_RECVPKTINFO: case IPV6_RECVHOPLIMIT: case IPV6_RECVRTHDR: @@ -1912,43 +2383,45 @@ ip6_ctloutput(so, sopt) case IPV6_RECVTCLASS: case IPV6_V6ONLY: case IPV6_AUTOFLOWLABEL: - if (optlen != sizeof(int)) { + if (optlen != sizeof (int)) { error = EINVAL; break; } error = sooptcopyin(sopt, &optval, - sizeof optval, sizeof optval); + sizeof (optval), sizeof (optval)); if (error) break; - switch (optname) { + switch (optname) { case IPV6_UNICAST_HOPS: - if (optval < -1 || optval >= 256) + if (optval < -1 || optval >= 256) { error = EINVAL; - else { + } else { /* -1 = kernel default */ in6p->in6p_hops = optval; - if ((in6p->inp_vflag & - INP_IPV4) != 0) - in6p->inp_ip_ttl = optval; + if (in6p->inp_vflag & + INP_IPV4) { + in6p->inp_ip_ttl = + optval; + } } break; -#define OPTSET(bit) \ -do { \ - if (optval) \ - in6p->inp_flags |= (bit); \ - else \ - in6p->inp_flags &= ~(bit); \ -} while (/*CONSTCOND*/ 0) -#define OPTSET2292(bit) \ -do { \ - in6p->inp_flags |= IN6P_RFC2292; \ - if (optval) \ - in6p->inp_flags |= (bit); \ - else \ - in6p->inp_flags &= ~(bit); \ -} while (/*CONSTCOND*/ 0) -#define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0) +#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) case IPV6_RECVPKTINFO: /* cannot mix with RFC2292 */ @@ -1959,8 +2432,7 @@ do { \ OPTSET(IN6P_PKTINFO); break; - case IPV6_HOPLIMIT: - { + case IPV6_HOPLIMIT: { struct ip6_pktopts **optp; /* cannot mix with RFC2292 */ @@ -1970,7 +2442,7 @@ do { \ } optp = &in6p->in6p_outputopts; error = ip6_pcbopt(IPV6_HOPLIMIT, - (u_char *)&optval, sizeof(optval), + (u_char *)&optval, sizeof (optval), optp, uproto); break; } @@ -1991,6 +2463,7 @@ do { \ break; } OPTSET(IN6P_HOPOPTS); + capture_exthdrstat_in = TRUE; break; case IPV6_RECVDSTOPTS: @@ -2000,6 +2473,7 @@ do { \ break; } OPTSET(IN6P_DSTOPTS); + capture_exthdrstat_in = TRUE; break; case IPV6_RECVRTHDRDSTOPTS: @@ -2009,6 +2483,7 @@ do { \ break; } OPTSET(IN6P_RTHDRDSTOPTS); + capture_exthdrstat_in = TRUE; break; case IPV6_RECVRTHDR: @@ -2018,6 +2493,7 @@ do { \ break; } OPTSET(IN6P_RTHDR); + capture_exthdrstat_in = TRUE; break; case IPV6_RECVPATHMTU: @@ -2038,7 +2514,8 @@ do { \ * see ipng mailing list, Jun 22 2001. */ if (in6p->inp_lport || - !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) { + !IN6_IS_ADDR_UNSPECIFIED( + &in6p->in6p_laddr)) { error = EINVAL; break; } @@ -2048,10 +2525,12 @@ do { \ else in6p->inp_vflag |= INP_IPV4; break; + case IPV6_RECVTCLASS: /* we can mix with RFC2292 */ OPTSET(IN6P_TCLASS); break; + case IPV6_AUTOFLOWLABEL: OPTSET(IN6P_AUTOFLOWLABEL); break; @@ -2062,23 +2541,30 @@ do { \ case IPV6_TCLASS: case IPV6_DONTFRAG: case IPV6_USE_MIN_MTU: - case IPV6_PREFER_TEMPADDR: - if (optlen != sizeof(optval)) { + case IPV6_PREFER_TEMPADDR: { + struct ip6_pktopts **optp; + + if (optlen != sizeof (optval)) { error = EINVAL; break; } error = sooptcopyin(sopt, &optval, - sizeof optval, sizeof optval); + sizeof (optval), sizeof (optval)); if (error) break; - { - struct ip6_pktopts **optp; - optp = &in6p->in6p_outputopts; - error = ip6_pcbopt(optname, - (u_char *)&optval, sizeof(optval), - optp, uproto); - break; + + optp = &in6p->in6p_outputopts; + error = ip6_pcbopt(optname, (u_char *)&optval, + sizeof (optval), optp, uproto); + + if (optname == IPV6_TCLASS) { + // Add in the ECN flags + u_int8_t tos = (in6p->inp_ip_tos & ~IPTOS_ECN_MASK); + u_int8_t ecn = optval & IPTOS_ECN_MASK; + in6p->inp_ip_tos = tos | ecn; } + break; + } case IPV6_2292PKTINFO: case IPV6_2292HOPLIMIT: @@ -2086,12 +2572,12 @@ do { \ case IPV6_2292DSTOPTS: case IPV6_2292RTHDR: /* RFC 2292 */ - if (optlen != sizeof(int)) { + if (optlen != sizeof (int)) { error = EINVAL; break; } error = sooptcopyin(sopt, &optval, - sizeof optval, sizeof optval); + sizeof (optval), sizeof (optval)); if (error) break; switch (optname) { @@ -2107,26 +2593,30 @@ do { \ * See comments for IPV6_RECVHOPOPTS. */ if (!privileged) - return(EPERM); + return (EPERM); OPTSET2292(IN6P_HOPOPTS); + capture_exthdrstat_in = TRUE; break; case IPV6_2292DSTOPTS: if (!privileged) - return(EPERM); - OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */ + return (EPERM); + OPTSET2292(IN6P_DSTOPTS| + IN6P_RTHDRDSTOPTS); /* XXX */ + capture_exthdrstat_in = TRUE; break; case IPV6_2292RTHDR: OPTSET2292(IN6P_RTHDR); + capture_exthdrstat_in = TRUE; break; } break; + case IPV6_3542PKTINFO: case IPV6_3542HOPOPTS: case IPV6_3542RTHDR: case IPV6_3542DSTOPTS: case IPV6_RTHDRDSTOPTS: - case IPV6_3542NEXTHOP: - { + case IPV6_3542NEXTHOP: { struct ip6_pktopts **optp; /* new advanced API (RFC3542) */ struct mbuf *m; @@ -2140,18 +2630,16 @@ do { \ if (error != 0) break; error = soopt_mcopyin(sopt, m); - if (error) { - m_freem(m); + if (error != 0) break; - } + optp = &in6p->in6p_outputopts; - error = ip6_pcbopt(optname, mtod(m, u_char *), - m->m_len, optp, uproto); + error = ip6_pcbopt(optname, mtod(m, u_char *), + m->m_len, optp, uproto); m_freem(m); break; } #undef OPTSET - case IPV6_MULTICAST_IF: case IPV6_MULTICAST_HOPS: case IPV6_MULTICAST_LOOP: @@ -2169,7 +2657,7 @@ do { \ case IPV6_PORTRANGE: error = sooptcopyin(sopt, &optval, - sizeof optval, sizeof optval); + sizeof (optval), sizeof (optval)); if (error) break; @@ -2194,45 +2682,25 @@ do { \ break; } break; - #if IPSEC - case IPV6_IPSEC_POLICY: - { + case IPV6_IPSEC_POLICY: { caddr_t req = NULL; size_t len = 0; struct mbuf *m; - if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ + if ((error = soopt_getm(sopt, &m)) != 0) break; - if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ + if ((error = soopt_mcopyin(sopt, m)) != 0) break; - if (m) { - req = mtod(m, caddr_t); - len = m->m_len; - } + + req = mtod(m, caddr_t); + len = m->m_len; error = ipsec6_set_policy(in6p, optname, req, - len, privileged); + len, privileged); m_freem(m); - } - break; -#endif /* KAME IPSEC */ - -#if IPFIREWALL - case IPV6_FW_ADD: - case IPV6_FW_DEL: - case IPV6_FW_FLUSH: - case IPV6_FW_ZERO: - { - if (ip6_fw_ctl_ptr == NULL) - load_ip6fw(); - if (ip6_fw_ctl_ptr != NULL) - error = (*ip6_fw_ctl_ptr)(sopt); - else - return ENOPROTOOPT; - } break; -#endif /* IPFIREWALL */ - + } +#endif /* IPSEC */ /* * IPv6 variant of IP_BOUND_IF; for details see * comments on IP_BOUND_IF in ip_ctloutput(). @@ -2250,7 +2718,7 @@ do { \ if (error) break; - error = inp_bindif(in6p, optval); + error = inp_bindif(in6p, optval, NULL); break; case IPV6_NO_IFT_CELLULAR: @@ -2266,7 +2734,14 @@ do { \ if (error) break; - error = inp_nocellular(in6p, optval); + /* 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: @@ -2278,6 +2753,13 @@ do { \ error = ENOPROTOOPT; break; } + if (capture_exthdrstat_in) { + if (uproto == IPPROTO_TCP) { + INC_ATOMIC_INT64_LIM(net_api_stats.nas_sock_inet6_stream_exthdr_in); + } else if (uproto == IPPROTO_UDP) { + INC_ATOMIC_INT64_LIM(net_api_stats.nas_sock_inet6_dgram_exthdr_in); + } + } break; case SOPT_GET: @@ -2304,7 +2786,6 @@ do { \ case IPV6_RECVHOPLIMIT: case IPV6_RECVRTHDR: case IPV6_RECVPATHMTU: - case IPV6_V6ONLY: case IPV6_PORTRANGE: case IPV6_RECVTCLASS: @@ -2347,8 +2828,7 @@ do { \ optval = OPTBIT(IN6P_IPV6_V6ONLY); break; - case IPV6_PORTRANGE: - { + case IPV6_PORTRANGE: { int flags; flags = in6p->inp_flags; if (flags & INP_HIGHPORT) @@ -2358,7 +2838,7 @@ do { \ else optval = 0; break; - } + } case IPV6_RECVTCLASS: optval = OPTBIT(IN6P_TCLASS); break; @@ -2370,16 +2850,15 @@ do { \ if (error) break; error = sooptcopyout(sopt, &optval, - sizeof optval); + sizeof (optval)); break; - case IPV6_PATHMTU: - { + case IPV6_PATHMTU: { u_int32_t pmtu = 0; struct ip6_mtuinfo mtuinfo; struct route_in6 sro; - bzero(&sro, sizeof(sro)); + bzero(&sro, sizeof (sro)); if (!(so->so_state & SS_ISCONNECTED)) return (ENOTCONN); @@ -2390,17 +2869,16 @@ do { \ */ error = ip6_getpmtu(&sro, NULL, NULL, &in6p->in6p_faddr, &pmtu, NULL); - if (sro.ro_rt) - rtfree(sro.ro_rt); + ROUTE_RELEASE(&sro); if (error) break; if (pmtu > IPV6_MAXPACKET) pmtu = IPV6_MAXPACKET; - bzero(&mtuinfo, sizeof(mtuinfo)); + bzero(&mtuinfo, sizeof (mtuinfo)); mtuinfo.ip6m_mtu = (u_int32_t)pmtu; optdata = (void *)&mtuinfo; - optdatalen = sizeof(mtuinfo); + optdatalen = sizeof (mtuinfo); error = sooptcopyout(sopt, optdata, optdatalen); break; @@ -2425,12 +2903,14 @@ do { \ optval = OPTBIT(IN6P_RTHDR); break; case IPV6_2292DSTOPTS: - optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); + optval = OPTBIT(IN6P_DSTOPTS| + IN6P_RTHDRDSTOPTS); break; } error = sooptcopyout(sopt, &optval, - sizeof optval); + sizeof (optval)); break; + case IPV6_PKTINFO: case IPV6_HOPOPTS: case IPV6_RTHDR: @@ -2451,47 +2931,12 @@ do { \ case IPV6_MSFILTER: error = ip6_getmoptions(in6p, sopt); break; - #if IPSEC - case IPV6_IPSEC_POLICY: - { - caddr_t req = NULL; - size_t len = 0; - struct mbuf *m = NULL; - struct mbuf **mp = &m; - - error = soopt_getm(sopt, &m); /* XXX */ - if (error != 0) - break; - error = soopt_mcopyin(sopt, m); /* XXX */ - if (error != 0) - break; - if (m) { - req = mtod(m, caddr_t); - len = m->m_len; - } - error = ipsec6_get_policy(in6p, req, len, mp); - if (error == 0) - error = soopt_mcopyout(sopt, m); /*XXX*/ - if (error == 0 && m) - m_freem(m); - break; - } -#endif /* KAME 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 - return ENOPROTOOPT; - } + case IPV6_IPSEC_POLICY: { + error = 0; /* This option is no longer supported */ break; -#endif /* IPFIREWALL */ - + } +#endif /* IPSEC */ case IPV6_BOUND_IF: if (in6p->inp_flags & INP_BOUND_IF) optval = in6p->inp_boundifp->if_index; @@ -2500,8 +2945,7 @@ do { \ break; case IPV6_NO_IFT_CELLULAR: - optval = (in6p->inp_flags & INP_NO_IFT_CELLULAR) - ? 1 : 0; + optval = INP_NO_CELLULAR(in6p) ? 1 : 0; error = sooptcopyout(sopt, &optval, sizeof (optval)); break; @@ -2519,10 +2963,12 @@ do { \ } break; } + } else if (level == IPPROTO_UDP) { + error = udp_ctloutput(so, sopt); } else { error = EINVAL; } - return(error); + return (error); } int @@ -2538,9 +2984,8 @@ ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) optname = sopt->sopt_name; optlen = sopt->sopt_valsize; - if (level != IPPROTO_IPV6) { + if (level != IPPROTO_IPV6) return (EINVAL); - } switch (optname) { case IPV6_CHECKSUM: @@ -2554,32 +2999,32 @@ ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) */ switch (op) { case SOPT_SET: - if (optlen != sizeof(int)) { + if (optlen != sizeof (int)) { error = EINVAL; break; } - error = sooptcopyin(sopt, &optval, sizeof(optval), - sizeof(optval)); + 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 (so->so_proto->pr_protocol == - IPPROTO_ICMPV6) { + } else if (SOCK_PROTO(so) == IPPROTO_ICMPV6) { if (optval != icmp6off) error = EINVAL; - } else + } else { in6p->in6p_cksum = optval; + } break; case SOPT_GET: - if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) + if (SOCK_PROTO(so) == IPPROTO_ICMPV6) optval = icmp6off; else optval = in6p->in6p_cksum; - error = sooptcopyout(sopt, &optval, sizeof(optval)); + error = sooptcopyout(sopt, &optval, sizeof (optval)); break; default: @@ -2601,49 +3046,48 @@ ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) * specifying behavior of outgoing packets. */ static int -ip6_pcbopts( - struct ip6_pktopts **pktopt, - struct mbuf *m, - __unused struct socket *so, - __unused struct sockopt *sopt) +ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so, + struct sockopt *sopt) { +#pragma unused(sopt) struct ip6_pktopts *opt = *pktopt; int error = 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"); + printf("%s: all specified options are cleared.\n", + __func__); #endif ip6_clearpktopts(opt, -1); } else { - opt = _MALLOC(sizeof(*opt), M_IP6OPT, M_WAITOK); + opt = _MALLOC(sizeof (*opt), M_IP6OPT, M_WAITOK); if (opt == NULL) - return ENOBUFS; + return (ENOBUFS); } *pktopt = NULL; - if (!m || m->m_len == 0) { + if (m == NULL || m->m_len == 0) { /* * Only turning off any previous options, regardless of * whether the opt is just created or given. */ - if (opt) + if (opt != NULL) FREE(opt, M_IP6OPT); - return(0); + return (0); } /* set options specified by user. */ - if ((error = ip6_setpktopts(m, opt, NULL, so->so_proto->pr_protocol)) != 0) { + 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); + return (error); } *pktopt = opt; - return(0); + return (0); } /* @@ -2654,7 +3098,7 @@ void ip6_initpktopts(struct ip6_pktopts *opt) { - bzero(opt, sizeof(*opt)); + 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; @@ -2669,9 +3113,9 @@ ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, opt = *pktopt; if (opt == NULL) { - opt = _MALLOC(sizeof(*opt), M_IP6OPT, M_WAITOK); + opt = _MALLOC(sizeof (*opt), M_IP6OPT, M_WAITOK); if (opt == NULL) - return(ENOBUFS); + return (ENOBUFS); ip6_initpktopts(opt); *pktopt = opt; } @@ -2685,7 +3129,6 @@ ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) void *optdata = NULL; int optdatalen = 0; struct ip6_ext *ip6e; - int error = 0; struct in6_pktinfo null_pktinfo; int deftclass = 0, on; int defminmtu = IP6PO_MINMTU_MCASTONLY; @@ -2698,18 +3141,20 @@ ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) optdata = (void *)pktopt->ip6po_pktinfo; else { /* XXX: we don't have to do this every time... */ - bzero(&null_pktinfo, sizeof(null_pktinfo)); + bzero(&null_pktinfo, sizeof (null_pktinfo)); optdata = (void *)&null_pktinfo; } - optdatalen = sizeof(struct in6_pktinfo); + optdatalen = sizeof (struct in6_pktinfo); break; + case IPV6_TCLASS: if (pktopt && pktopt->ip6po_tclass >= 0) optdata = (void *)&pktopt->ip6po_tclass; else optdata = (void *)&deftclass; - optdatalen = sizeof(int); + optdatalen = sizeof (int); break; + case IPV6_HOPOPTS: if (pktopt && pktopt->ip6po_hbh) { optdata = (void *)pktopt->ip6po_hbh; @@ -2717,6 +3162,7 @@ ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) optdatalen = (ip6e->ip6e_len + 1) << 3; } break; + case IPV6_RTHDR: if (pktopt && pktopt->ip6po_rthdr) { optdata = (void *)pktopt->ip6po_rthdr; @@ -2724,6 +3170,7 @@ ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) optdatalen = (ip6e->ip6e_len + 1) << 3; } break; + case IPV6_RTHDRDSTOPTS: if (pktopt && pktopt->ip6po_dest1) { optdata = (void *)pktopt->ip6po_dest1; @@ -2731,6 +3178,7 @@ ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) optdatalen = (ip6e->ip6e_len + 1) << 3; } break; + case IPV6_DSTOPTS: if (pktopt && pktopt->ip6po_dest2) { optdata = (void *)pktopt->ip6po_dest2; @@ -2738,34 +3186,39 @@ ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) optdatalen = (ip6e->ip6e_len + 1) << 3; } break; + 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); + 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); + optdatalen = sizeof (on); break; + case IPV6_PREFER_TEMPADDR: if (pktopt) optdata = (void *)&pktopt->ip6po_prefer_tempaddr; else optdata = (void *)&defpreftemp; - optdatalen = sizeof(int); + optdatalen = sizeof (int); break; + default: /* should not happen */ #ifdef DIAGNOSTIC panic("ip6_getpcbopt: unexpected option\n"); @@ -2773,9 +3226,7 @@ ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) return (ENOPROTOOPT); } - error = sooptcopyout(sopt, optdata, optdatalen); - - return (error); + return (sooptcopyout(sopt, optdata, optdatalen)); } void @@ -2794,10 +3245,7 @@ ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) if (optname == -1 || optname == IPV6_TCLASS) pktopt->ip6po_tclass = -1; if (optname == -1 || optname == IPV6_NEXTHOP) { - if (pktopt->ip6po_nextroute.ro_rt) { - rtfree(pktopt->ip6po_nextroute.ro_rt); - pktopt->ip6po_nextroute.ro_rt = NULL; - } + ROUTE_RELEASE(&pktopt->ip6po_nextroute); if (pktopt->ip6po_nexthop) FREE(pktopt->ip6po_nexthop, M_IP6OPT); pktopt->ip6po_nexthop = NULL; @@ -2816,10 +3264,7 @@ ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) 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 (pktopt->ip6po_dest2) @@ -2828,16 +3273,15 @@ ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) } } -#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);\ - }\ +#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) static int @@ -2852,19 +3296,19 @@ copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) 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); @@ -2872,7 +3316,7 @@ copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ return (0); - bad: +bad: ip6_clearpktopts(dst, -1); return (ENOBUFS); } @@ -2884,7 +3328,7 @@ ip6_copypktopts(struct ip6_pktopts *src, int canwait) int error; struct ip6_pktopts *dst; - dst = _MALLOC(sizeof(*dst), M_IP6OPT, canwait); + dst = _MALLOC(sizeof (*dst), M_IP6OPT, canwait); if (dst == NULL) return (NULL); ip6_initpktopts(dst); @@ -3050,7 +3494,7 @@ int ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, struct ip6_pktopts *stickyopt, int uproto) { - struct cmsghdr *cm = 0; + struct cmsghdr *cm = NULL; if (control == NULL || opt == NULL) return (EINVAL); @@ -3082,10 +3526,12 @@ ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, if (control->m_len < CMSG_LEN(0)) return (EINVAL); - for (cm = M_FIRST_CMSGHDR(control); cm; cm = M_NXT_CMSGHDR(control, cm)) { + 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) + if (cm->cmsg_len < sizeof (struct cmsghdr) || + cm->cmsg_len > control->m_len) return (EINVAL); if (cm->cmsg_level != IPPROTO_IPV6) continue; @@ -3113,6 +3559,7 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, { int minmtupolicy, preftemp; int error; + boolean_t capture_exthdrstat_out = FALSE; if (!sticky && !cmsg) { #ifdef DIAGNOSTIC @@ -3163,12 +3610,11 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, switch (optname) { case IPV6_2292PKTINFO: - case IPV6_PKTINFO: - { + case IPV6_PKTINFO: { struct ifnet *ifp = NULL; struct in6_pktinfo *pktinfo; - if (len != sizeof(struct in6_pktinfo)) + if (len != sizeof (struct in6_pktinfo)) return (EINVAL); pktinfo = (struct in6_pktinfo *)(void *)buf; @@ -3198,7 +3644,7 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, ifnet_head_done(); return (ENXIO); } - + if (pktinfo->ipi6_ifindex) { ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; if (ifp == NULL) { @@ -3206,7 +3652,7 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, return (ENXIO); } } - + ifnet_head_done(); /* @@ -3220,18 +3666,17 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, * application when it is used as a sticky option. */ if (opt->ip6po_pktinfo == NULL) { - opt->ip6po_pktinfo = _MALLOC(sizeof(*pktinfo), + opt->ip6po_pktinfo = _MALLOC(sizeof (*pktinfo), M_IP6OPT, M_NOWAIT); if (opt->ip6po_pktinfo == NULL) return (ENOBUFS); } - bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo)); + bcopy(pktinfo, opt->ip6po_pktinfo, sizeof (*pktinfo)); break; } case IPV6_2292HOPLIMIT: - case IPV6_HOPLIMIT: - { + case IPV6_HOPLIMIT: { int *hlimp; /* @@ -3241,7 +3686,7 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, if (optname == IPV6_HOPLIMIT && sticky) return (ENOPROTOOPT); - if (len != sizeof(int)) + if (len != sizeof (int)) return (EINVAL); hlimp = (int *)(void *)buf; if (*hlimp < -1 || *hlimp > 255) @@ -3251,11 +3696,10 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, break; } - case IPV6_TCLASS: - { + case IPV6_TCLASS: { int tclass; - if (len != sizeof(int)) + if (len != sizeof (int)) return (EINVAL); tclass = *(int *)(void *)buf; if (tclass < -1 || tclass > 255) @@ -3277,16 +3721,14 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, } /* check if cmsg_len is large enough for sa_len */ - if (len < sizeof(struct sockaddr) || len < *buf) + if (len < sizeof (struct sockaddr) || len < *buf) return (EINVAL); - switch (((struct sockaddr *)buf)->sa_family) { - case AF_INET6: - { - struct sockaddr_in6 *sa6 = - (struct sockaddr_in6 *)(void *)buf; + switch (SA(buf)->sa_family) { + case AF_INET6: { + struct sockaddr_in6 *sa6 = SIN6(buf); - if (sa6->sin6_len != sizeof(struct sockaddr_in6)) + if (sa6->sin6_len != sizeof (struct sockaddr_in6)) return (EINVAL); if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || @@ -3313,8 +3755,7 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, break; case IPV6_2292HOPOPTS: - case IPV6_HOPOPTS: - { + case IPV6_HOPOPTS: { struct ip6_hbh *hbh; int hbhlen; @@ -3333,7 +3774,7 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, } /* message length validation */ - if (len < sizeof(struct ip6_hbh)) + if (len < sizeof (struct ip6_hbh)) return (EINVAL); hbh = (struct ip6_hbh *)(void *)buf; hbhlen = (hbh->ip6h_len + 1) << 3; @@ -3346,14 +3787,13 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, if (opt->ip6po_hbh == NULL) return (ENOBUFS); bcopy(hbh, opt->ip6po_hbh, hbhlen); - + capture_exthdrstat_out = TRUE; break; } case IPV6_2292DSTOPTS: case IPV6_DSTOPTS: - case IPV6_RTHDRDSTOPTS: - { + case IPV6_RTHDRDSTOPTS: { struct ip6_dest *dest, **newdest = NULL; int destlen; @@ -3367,7 +3807,7 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, } /* message length validation */ - if (len < sizeof(struct ip6_dest)) + if (len < sizeof (struct ip6_dest)) return (EINVAL); dest = (struct ip6_dest *)(void *)buf; destlen = (dest->ip6d_len + 1) << 3; @@ -3411,13 +3851,12 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, if (*newdest == NULL) return (ENOBUFS); bcopy(dest, *newdest, destlen); - + capture_exthdrstat_out = TRUE; break; } case IPV6_2292RTHDR: - case IPV6_RTHDR: - { + case IPV6_RTHDR: { struct ip6_rthdr *rth; int rthlen; @@ -3427,7 +3866,7 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, } /* message length validation */ - if (len < sizeof(struct ip6_rthdr)) + if (len < sizeof (struct ip6_rthdr)) return (EINVAL); rth = (struct ip6_rthdr *)(void *)buf; rthlen = (rth->ip6r_len + 1) << 3; @@ -3453,12 +3892,12 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, if (opt->ip6po_rthdr == NULL) return (ENOBUFS); bcopy(rth, opt->ip6po_rthdr, rthlen); - + capture_exthdrstat_out = TRUE; break; } case IPV6_USE_MIN_MTU: - if (len != sizeof(int)) + if (len != sizeof (int)) return (EINVAL); minmtupolicy = *(int *)(void *)buf; if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && @@ -3470,7 +3909,7 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, break; case IPV6_DONTFRAG: - if (len != sizeof(int)) + if (len != sizeof (int)) return (EINVAL); if (uproto == IPPROTO_TCP || *(int *)(void *)buf == 0) { @@ -3479,12 +3918,13 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, * (RFC3542 leaves this case unspecified.) */ opt->ip6po_flags &= ~IP6PO_DONTFRAG; - } else + } else { opt->ip6po_flags |= IP6PO_DONTFRAG; + } break; case IPV6_PREFER_TEMPADDR: - if (len != sizeof(int)) + if (len != sizeof (int)) return (EINVAL); preftemp = *(int *)(void *)buf; if (preftemp != IP6PO_TEMPADDR_SYSTEM && @@ -3499,6 +3939,14 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, return (ENOPROTOOPT); } /* end of switch */ + if (capture_exthdrstat_out) { + if (uproto == IPPROTO_TCP) { + INC_ATOMIC_INT64_LIM(net_api_stats.nas_sock_inet6_stream_exthdr_out); + } else if (uproto == IPPROTO_UDP) { + INC_ATOMIC_INT64_LIM(net_api_stats.nas_sock_inet6_dgram_exthdr_out); + } + } + return (0); } @@ -3509,38 +3957,32 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, * pointer that might NOT be &loif -- easier than replicating that code here. */ void -ip6_mloopback( - struct ifnet *ifp, - struct mbuf *m, - 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; struct ip6_hdr *ip6; + struct in6_addr src; - copym = m_copy(m, 0, M_COPYALL); - if (copym == NULL) + 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. */ - if ((copym->m_flags & M_EXT) != 0 || - copym->m_len < sizeof(struct ip6_hdr)) { - copym = m_pullup(copym, sizeof(struct ip6_hdr)); - if (copym == NULL) - return; - } + 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 DIAGNOSTIC - if (copym->m_len < sizeof(*ip6)) { - m_freem(copym); + if (copym == NULL) return; - } -#endif 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. @@ -3548,83 +3990,217 @@ ip6_mloopback( in6_clearscope(&ip6->ip6_src); in6_clearscope(&ip6->ip6_dst); -#ifdef __APPLE__ + if (copym->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA) + in6_delayed_cksum_offset(copym, 0, optlen, nxt0); - /* Makes sure the HW checksum flags are cleaned before sending the packet */ + /* + * 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; - if ((copym->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA) != 0) { - in6_delayed_cksum(copym, sizeof(struct ip6_hdr)); - copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_IPV6_DATA; + /* + * 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); } - copym->m_pkthdr.rcvif = 0; - copym->m_pkthdr.csum_data = 0; - copym->m_pkthdr.csum_flags = 0; + if (srcifp != NULL) + ip6_setsrcifaddr_info(copym, srcifp->if_index, NULL); + ip6_setdstifaddr_info(copym, origifp->if_index, NULL); - if (lo_ifp) { - copym->m_pkthdr.rcvif = ifp; - dlil_output(lo_ifp, PF_INET6, copym, 0, - (struct sockaddr *)dst, 0, NULL); - } else - m_free(copym); -#else - (void)if_simloop(ifp, copym, dst->sin6_family, NULL); -#endif + 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)) { + if (m->m_len > sizeof (*ip6)) { MGETHDR(mh, M_DONTWAIT, MT_HEADER); /* MAC-OK */ - if (mh == 0) { + 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. + * unless the interface supports "invert zero" capability.) + */ + if (hwcksum_tx && !tso && + ((m->m_pkthdr.csum_flags & CSUM_TCPIPV6) || + ((hwcap & CSUM_ZERO_INVERT) && + (m->m_pkthdr.csum_flags & CSUM_ZERO_INVERT))) && + 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. */ 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 } + +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))); +}