X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/fe8ab488e9161c46dd9885d58fc52996dc0249ff..HEAD:/bsd/netinet6/ip6_output.c diff --git a/bsd/netinet6/ip6_output.c b/bsd/netinet6/ip6_output.c index 812bf2b3a..92ee660ed 100644 --- a/bsd/netinet6/ip6_output.c +++ b/bsd/netinet6/ip6_output.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2000-2014 Apple Inc. All rights reserved. + * Copyright (c) 2000-2020 Apple Inc. All rights reserved. * * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ * @@ -118,14 +118,18 @@ #include #include #include +#include #include +#include +#include #include #include #include #include #include #include +#include #include #include @@ -144,13 +148,7 @@ extern int ipsec_bypass; #include #endif /* NECP */ -#if CONFIG_MACF_NET -#include -#endif /* CONFIG_MACF_NET */ - #if DUMMYNET -#include -#include #include #endif /* DUMMYNET */ @@ -158,13 +156,16 @@ extern int ipsec_bypass; #include #endif /* PF */ +static int sysctl_reset_ip6_output_stats SYSCTL_HANDLER_ARGS; +static int sysctl_ip6_output_measure_bins SYSCTL_HANDLER_ARGS; +static int sysctl_ip6_output_getperf SYSCTL_HANDLER_ARGS; static int ip6_copyexthdr(struct mbuf **, caddr_t, int); static void ip6_out_cksum_stats(int, u_int32_t); static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t); static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int, struct ip6_frag **); static int ip6_getpmtu(struct route_in6 *, struct route_in6 *, - struct ifnet *, struct in6_addr *, u_int32_t *, boolean_t *); + struct ifnet *, struct in6_addr *, u_int32_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); @@ -176,81 +177,75 @@ static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, int, static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *); static void ip6_output_checksum(struct ifnet *, uint32_t, struct mbuf *, int, uint32_t, uint32_t); +extern int udp_ctloutput(struct socket *, struct sockopt *); +static int ip6_fragment_packet(struct mbuf **m, + struct ip6_pktopts *opt, struct ip6_out_args * ip6oa, + struct ip6_exthdrs *exthdrsp, struct ifnet *ifp, + uint32_t mtu, 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 */ +#define IM6O_TRACE_HIST_SIZE 32 /* size of trace history */ /* For gdb */ __private_extern__ unsigned int im6o_trace_hist_size = IM6O_TRACE_HIST_SIZE; struct ip6_moptions_dbg { - struct ip6_moptions im6o; /* ip6_moptions */ - u_int16_t im6o_refhold_cnt; /* # of IM6O_ADDREF */ - u_int16_t im6o_refrele_cnt; /* # of IM6O_REMREF */ + struct ip6_moptions im6o; /* ip6_moptions */ + u_int16_t im6o_refhold_cnt; /* # of IM6O_ADDREF */ + u_int16_t im6o_refrele_cnt; /* # of IM6O_REMREF */ /* * Alloc and free callers. */ - ctrace_t im6o_alloc; - ctrace_t im6o_free; + ctrace_t im6o_alloc; + ctrace_t im6o_free; /* * Circular lists of IM6O_ADDREF and IM6O_REMREF callers. */ - ctrace_t im6o_refhold[IM6O_TRACE_HIST_SIZE]; - ctrace_t im6o_refrele[IM6O_TRACE_HIST_SIZE]; + ctrace_t im6o_refhold[IM6O_TRACE_HIST_SIZE]; + ctrace_t im6o_refrele[IM6O_TRACE_HIST_SIZE]; }; #if DEBUG -static unsigned int im6o_debug = 1; /* debugging (enabled) */ +static unsigned int im6o_debug = 1; /* debugging (enabled) */ #else -static unsigned int im6o_debug; /* debugging (disabled) */ +static unsigned int im6o_debug; /* debugging (disabled) */ #endif /* !DEBUG */ -static unsigned int im6o_size; /* size of zone element */ -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"); +static struct zone *im6o_zone; /* zone for ip6_moptions */ +#define IM6O_ZONE_NAME "ip6_moptions" /* zone name */ /* - * 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, +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 != NULL) { - /* - * 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, ip6oa); - if (error != 0) { - if (nextpkt != NULL) - 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. @@ -265,28 +260,38 @@ ip6_output_list(struct mbuf *m0, int packetlist, struct ip6_pktopts *opt, * 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, - struct ip6_out_args *ip6oa) +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; u_char *nexthdrp; - struct ifnet *ifp = NULL, *origifp = NULL; /* refcnt'd */ + struct ifnet *ifp = NULL, *origifp = NULL; /* refcnt'd */ + struct ifnet **ifpp_save = ifpp; struct mbuf *m, *mprev; - int hlen, tlen, len, off, nxt0; + 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, *src_ia = NULL; - u_int32_t mtu; - boolean_t alwaysfrag = FALSE; + u_int32_t mtu = 0; u_int32_t optlen = 0, plen = 0, unfragpartlen = 0; struct ip6_rthdr *rh; struct in6_addr finaldst; ipfilter_t inject_filter_ref; struct ipf_pktopts *ippo = NULL; struct flowadv *adv = NULL; + uint32_t pktcnt = 0; + uint32_t packets_processed = 0; + struct timeval start_tv; +#if PF + boolean_t skip_pf = (ip6oa != NULL) && + (ip6oa->ip6oa_flags & IP6OAF_SKIP_PF); +#endif + #if DUMMYNET struct m_tag *tag; struct ip6_out_args saved_ip6oa; @@ -319,18 +324,19 @@ ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro, 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 +#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; @@ -340,28 +346,41 @@ ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro, uint32_t raw; } ip6obf = { .raw = 0 }; + if (ip6_output_measure) { + net_perf_start_time(&net_perf, &start_tv); + } + VERIFY(m0->m_flags & M_PKTHDR); /* zero out {saved_route, saved_ro_pmtu, ip6route, exthdrs, args} */ - bzero(&ip6obz, sizeof (ip6obz)); + bzero(&ip6obz, sizeof(ip6obz)); #if DUMMYNET - if (SLIST_EMPTY(&m0->m_pkthdr.tags)) + if (SLIST_EMPTY(&m0->m_pkthdr.tags)) { goto tags_done; + } /* Grab info from mtags prepended to the chain */ if ((tag = m_tag_locate(m0, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_DUMMYNET, NULL)) != NULL) { - struct dn_pkt_tag *dn_tag; + struct dn_pkt_tag *dn_tag; - dn_tag = (struct dn_pkt_tag *)(tag+1); + /* + * 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 != NULL) + if (ifp != NULL) { ifnet_reference(ifp); + } flags = dn_tag->dn_flags; if (dn_tag->dn_flags & IPV6_OUTARGS) { saved_ip6oa = dn_tag->dn_ip6oa; @@ -373,13 +392,13 @@ ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro, saved_ro_pmtu = dn_tag->dn_ro6_pmtu; ro_pmtu = &saved_ro_pmtu; origifp = dn_tag->dn_origifp; - if (origifp != NULL) + if (origifp != NULL) { ifnet_reference(origifp); + } mtu = dn_tag->dn_mtu; - alwaysfrag = (dn_tag->dn_alwaysfrag != 0); unfragpartlen = dn_tag->dn_unfragpartlen; - bcopy(&dn_tag->dn_exthdrs, &exthdrs, sizeof (exthdrs)); + bcopy(&dn_tag->dn_exthdrs, &exthdrs, sizeof(exthdrs)); m_tag_delete(m0, tag); } @@ -388,7 +407,6 @@ tags_done: #endif /* DUMMYNET */ m = m0; - m->m_pkthdr.pkt_flags &= ~(PKTF_LOOP|PKTF_IFAINFO); #if IPSEC if (ipsec_bypass == 0) { @@ -398,35 +416,33 @@ tags_done: } /* If packet is bound to an interface, check bound policies */ if ((flags & IPV6_OUTARGS) && - (ip6oa->ip6oa_flags & IPOAF_BOUND_IF) && - ip6oa->ip6oa_boundif != IFSCOPE_NONE) { + (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) + flags, ip6oa, &noipsec, &sp) != 0) { goto bad; + } ip6obf.noipsec = (noipsec != 0); } } #endif /* IPSEC */ - - ip6 = mtod(m, struct ip6_hdr *); - nxt0 = ip6->ip6_nxt; - finaldst = ip6->ip6_dst; - inject_filter_ref = ipf_get_inject_filter(m); + ippo = &ipf_pktopts; - if (ip6_doscopedroute && (flags & IPV6_OUTARGS)) { + 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) { @@ -434,8 +450,9 @@ tags_done: (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 { ip6obf.select_srcif = FALSE; if (flags & IPV6_OUTARGS) { @@ -446,31 +463,83 @@ tags_done: } if (flags & IPV6_OUTARGS) { - if (ip6oa->ip6oa_flags & IP6OAF_NO_CELLULAR) + if (ip6oa->ip6oa_flags & IP6OAF_NO_CELLULAR) { ipf_pktopts.ippo_flags |= IPPOF_NO_IFT_CELLULAR; - if (ip6oa->ip6oa_flags & IP6OAF_NO_EXPENSIVE) + } + if (ip6oa->ip6oa_flags & IP6OAF_NO_EXPENSIVE) { ipf_pktopts.ippo_flags |= IPPOF_NO_IFF_EXPENSIVE; + } + if (ip6oa->ip6oa_flags & IP6OAF_NO_CONSTRAINED) { + ipf_pktopts.ippo_flags |= IPPOF_NO_IFF_CONSTRAINED; + } 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 */ + VERIFY(ro != NULL); /* ro == saved_route */ goto check_with_pf; } #endif /* DUMMYNET */ -#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); \ - if (error) \ - goto freehdrs; \ - } \ +#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, ro ? ro->ro_rt : 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); \ + if (error) \ + goto freehdrs; \ + } \ } while (0) if (opt != NULL) { @@ -499,62 +568,106 @@ tags_done: #undef MAKE_EXTHDR #if NECP - necp_matched_policy_id = necp_ip6_output_find_policy_match (m, flags, (flags & IPV6_OUTARGS) ? ip6oa : NULL, - &necp_result, &necp_result_parameter); if (necp_matched_policy_id) { necp_mark_packet_from_ip(m, necp_matched_policy_id); + switch (necp_result) { - case NECP_KERNEL_POLICY_RESULT_PASS: + case NECP_KERNEL_POLICY_RESULT_PASS: + if (necp_result_parameter.pass_flags & NECP_KERNEL_POLICY_PASS_NO_SKIP_IPSEC) { + break; + } + goto skip_ipsec; + case NECP_KERNEL_POLICY_RESULT_DROP: + error = EHOSTUNREACH; + ip6stat.ip6s_necp_policy_drop++; + goto freehdrs; + 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); + + /* + * Update the QOS marking policy if + * 1. upper layer asks it to do so + * 2. net_qos_policy_restricted is not set + * 3. qos_marking_gencount doesn't match necp_kernel_socket_policies_gencount (checked in necp_lookup_current_qos_marking) + */ + if (ip6oa != NULL && (ip6oa->ip6oa_flags & IP6OAF_REDO_QOSMARKING_POLICY) && + net_qos_policy_restricted != 0) { + bool qos_marking = (ip6oa->ip6oa_flags & IP6OAF_QOSMARKING_ALLOWED) != 0; + qos_marking = necp_lookup_current_qos_marking(&ip6oa->qos_marking_gencount, NULL, policy_ifp, necp_result_parameter.route_rule_id, qos_marking); + if (qos_marking) { + ip6oa->ip6oa_flags |= IP6OAF_QOSMARKING_ALLOWED; + } else { + ip6oa->ip6oa_flags &= ~IP6OAF_QOSMARKING_ALLOWED; + } + } + + if (policy_ifp == ifp) { goto skip_ipsec; - case NECP_KERNEL_POLICY_RESULT_DROP: - case NECP_KERNEL_POLICY_RESULT_SOCKET_DIVERT: - /* Flow divert packets should be blocked at the IP layer */ - error = EHOSTUNREACH; - goto bad; - 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) { + } 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 { - if (necp_packet_can_rebind_to_ifnet(m, policy_ifp, (struct route *)&necp_route, AF_INET6)) { - /* Set scoped index to the tunnel interface, since it is compatible with the packet */ - /* This will only work for callers who pass IPV6_OUTARGS, but that covers all of the - clients we care about today */ - if (flags & IPV6_OUTARGS) { - ip6oa->ip6oa_boundif = policy_ifp->if_index; - ip6oa->ip6oa_flags |= IP6OAF_BOUND_IF; - } - if (opt != NULL && opt->ip6po_pktinfo != NULL) { - opt->ip6po_pktinfo->ipi6_ifindex = policy_ifp->if_index; - } - ro = &necp_route; - goto skip_ipsec; - } else { - error = ENETUNREACH; - goto bad; - } + error = ENETUNREACH; + ip6stat.ip6s_necp_policy_drop++; + goto freehdrs; } - break; } - default: - break; + } + default: + break; } } #endif /* NECP */ - + #if IPSEC - if (ipsec_bypass != 0 || ip6obf.noipsec) + if (ipsec_bypass != 0 || ip6obf.noipsec) { goto skip_ipsec; + } if (sp == NULL) { /* get a security policy for this packet */ if (so != NULL) { sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, - so, &error); + so, &error); } else { sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, - 0, &error); + 0, &error); } if (sp == NULL) { IPSEC_STAT_INCREMENT(ipsec6stat.out_inval); @@ -606,17 +719,21 @@ skip_ipsec: * Keep the length of the unfragmentable part for fragmentation. */ optlen = 0; - if (exthdrs.ip6e_hbh != NULL) + if (exthdrs.ip6e_hbh != NULL) { optlen += exthdrs.ip6e_hbh->m_len; - if (exthdrs.ip6e_dest1 != NULL) + } + if (exthdrs.ip6e_dest1 != NULL) { optlen += exthdrs.ip6e_dest1->m_len; - if (exthdrs.ip6e_rthdr != NULL) + } + 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 != NULL) + if (exthdrs.ip6e_dest2 != NULL) { optlen += exthdrs.ip6e_dest2->m_len; + } /* * If we need IPsec, or there is at least one extension header, @@ -624,9 +741,9 @@ skip_ipsec: */ if (( #if IPSEC - ip6obf.needipsec || + ip6obf.needipsec || #endif /* IPSEC */ - optlen) && !ip6obf.hdrsplit) { + optlen) && !ip6obf.hdrsplit) { if ((error = ip6_splithdr(m, &exthdrs)) != 0) { m = NULL; goto freehdrs; @@ -640,7 +757,7 @@ skip_ipsec: /* 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) { @@ -654,11 +771,12 @@ skip_ipsec: } /* adjust pointer */ ip6 = mtod(m, struct ip6_hdr *); - if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) + if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0) { goto freehdrs; + } ip6->ip6_plen = 0; } else { - ip6->ip6_plen = htons(plen); + ip6->ip6_plen = htons((uint16_t)plen); } /* * Concatenate headers and fill in next header fields. @@ -692,19 +810,19 @@ skip_ipsec: ip6->ip6_nxt = IPPROTO_DSTOPTS; } -#define MAKE_CHAIN(m, mp, p, i) do { \ - if (m != NULL) { \ - if (!ip6obf.hdrsplit) { \ - panic("assumption failed: hdr not split"); \ - /* NOTREACHED */ \ - } \ - *mtod((m), u_char *) = *(p); \ - *(p) = (i); \ - p = mtod((m), u_char *); \ - (m)->m_next = (mp)->m_next; \ - (mp)->m_next = (m); \ - (mp) = (m); \ - } \ +#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 @@ -715,15 +833,25 @@ skip_ipsec: 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)) + if (ip6obf.needipsec && (m->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA)) { in6_delayed_cksum_offset(m, 0, optlen, nxt0); + } #endif /* IPSEC */ - if (!TAILQ_EMPTY(&ipv6_filters)) { - struct ipfilter *filter; + if (!TAILQ_EMPTY(&ipv6_filters) && + !((flags & IPV6_OUTARGS) && + (ip6oa->ip6oa_flags & IP6OAF_INTCOPROC_ALLOWED) +#if NECP + && !necp_packet_should_skip_filters(m) +#endif // NECP + )) { + struct ipfilter *filter; int seen = (inject_filter_ref == NULL); int fixscope = 0; @@ -741,7 +869,7 @@ skip_ipsec: (ip6->ip6_dst.s6_addr16[1] == 0) && (ro != NULL)) { fixscope = 1; ip6->ip6_dst.s6_addr16[1] = - htons(ro->ro_dst.sin6_scope_id); + htons((uint16_t)ro->ro_dst.sin6_scope_id); } ipf_ref(); @@ -751,17 +879,19 @@ skip_ipsec: */ if (seen == 0) { if ((struct ipfilter *)inject_filter_ref == - filter) + 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); + filter->ipf_filter.cookie, + (mbuf_t *)&m, ippo); if (result == EJUSTRETURN) { ipf_unref(); - goto done; + m = NULL; + goto evaluateloop; } if (result != 0) { ipf_unref(); @@ -773,13 +903,14 @@ skip_ipsec: ip6 = mtod(m, struct ip6_hdr *); /* Hack: cleanup embedded scope_id if we put it there */ - if (fixscope) + if (fixscope) { ip6->ip6_dst.s6_addr16[1] = 0; + } } #if IPSEC if (ip6obf.needipsec) { - int segleft_org; + uint8_t segleft_org; /* * pointers after IPsec headers are not valid any more. @@ -814,7 +945,7 @@ skip_ipsec: default: printf("ip6_output (ipsec): error code %d\n", error); - /* FALLTHRU */ + OS_FALLTHROUGH; case ENOENT: /* don't show these error codes to the user */ error = 0; @@ -829,51 +960,10 @@ skip_ipsec: } #endif /* IPSEC */ - /* - * If there is a routing header, replace the destination address field - * with the first hop of the routing header. - */ + /* If there is a routing header, discard the packet. */ 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; - default: /* is it possible? */ - error = EINVAL; - goto bad; - } + error = EINVAL; + goto bad; } /* Source address validation */ @@ -896,16 +986,17 @@ skip_ipsec: */ if (ro == NULL) { ro = &ip6route; - bzero((caddr_t)ro, sizeof (*ro)); + bzero((caddr_t)ro, sizeof(*ro)); } - VERIFY(ro_pmtu == NULL); /* must not get here if dummynet */ ro_pmtu = ro; - if (opt != NULL && opt->ip6po_rthdr) + if (opt != NULL && opt->ip6po_rthdr) { ro = &opt->ip6po_route; + } dst = SIN6(&ro->ro_dst); - if (ro->ro_rt != NULL) + 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. @@ -914,10 +1005,12 @@ skip_ipsec: if (opt != NULL && opt->ip6po_tclass >= 0) { int mask = 0; - if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) + if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0) { mask |= 0xfc; - if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) + } + if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0) { mask |= 0x03; + } if (mask != 0) { ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20); @@ -933,7 +1026,7 @@ skip_ipsec: ip6->ip6_hlim = im6o->im6o_multicast_hlim; IM6O_UNLOCK(im6o); } else { - ip6->ip6_hlim = ip6_defmcasthlim; + ip6->ip6_hlim = (uint8_t)ip6_defmcasthlim; } } @@ -946,19 +1039,20 @@ skip_ipsec: * next transmit. */ if (ROUTE_UNUSABLE(ro) || dst->sin6_family != AF_INET6 || - !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst)) + !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 (ip6obf.needipsec && needipsectun) { #if CONFIG_DTRACE - struct ifnet *trace_ifp = (ifpp != NULL) ? (*ifpp) : NULL; + struct ifnet *trace_ifp = (ifpp_save != NULL) ? (*ifpp_save) : NULL; #endif /* CONFIG_DTRACE */ /* * All the extension headers will become inaccessible @@ -968,12 +1062,12 @@ skip_ipsec: * * 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)); + route_copyout((struct route *)&ipsec_state.ro, (struct route *)ro, + sizeof(struct route_in6)); ipsec_state.dst = SA(dst); /* So that we can see packets inside the tunnel */ @@ -983,15 +1077,16 @@ skip_ipsec: error = ipsec6_output_tunnel(&ipsec_state, sp, flags); /* tunneled in IPv4? packet is gone */ - if (ipsec_state.tunneled == 4) - goto done; + if (ipsec_state.tunneled == 4) { + m = NULL; + goto evaluateloop; + } m = ipsec_state.m; ipsec_saved_route = ro; ro = (struct route_in6 *)&ipsec_state.ro; dst = SIN6(ipsec_state.dst); if (error) { /* mbuf is already reclaimed in ipsec6_output_tunnel. */ - m0 = m = NULL; m = NULL; switch (error) { case EHOSTUNREACH: @@ -1003,7 +1098,7 @@ skip_ipsec: default: printf("ip6_output (ipsec): error code %d\n", error); - /* FALLTHRU */ + OS_FALLTHROUGH; case ENOENT: /* don't show these error codes to the user */ error = 0; @@ -1021,61 +1116,82 @@ skip_ipsec: ip6oa->ip6oa_flags &= ~IP6OAF_BOUND_IF; } if (opt != NULL && opt->ip6po_pktinfo != NULL) { - if (opt->ip6po_pktinfo->ipi6_ifindex != IFSCOPE_NONE) + if (opt->ip6po_pktinfo->ipi6_ifindex != IFSCOPE_NONE) { opt->ip6po_pktinfo->ipi6_ifindex = IFSCOPE_NONE; + } } exthdrs.ip6e_ip6 = m; } #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 (ip6obf.select_srcif) { - bzero(&src_sa, sizeof (src_sa)); + 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(ip6obf.select_srcif ? &src_sa : NULL, - &dst_sa, opt, im6o, &src_ia, 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) { /* * If in6_selectroute() does not return a route entry, * dst may not have been updated. */ - *dst = dst_sa; /* XXX */ + *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. */ @@ -1084,9 +1200,12 @@ skip_ipsec: } 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); } @@ -1098,15 +1217,18 @@ skip_ipsec: * address of our own. */ if (ia != NULL && ia->ia_ifp) { - ifnet_reference(ia->ia_ifp); /* for origifp */ - if (origifp != NULL) + ifnet_reference(ia->ia_ifp); /* for origifp */ + if (origifp != NULL) { ifnet_release(origifp); + } origifp = ia->ia_ifp; } else { - if (ifp != NULL) - ifnet_reference(ifp); /* for origifp */ - if (origifp != NULL) + if (ifp != NULL) { + ifnet_reference(ifp); /* for origifp */ + } + if (origifp != NULL) { ifnet_release(origifp); + } origifp = ifp; } @@ -1116,27 +1238,31 @@ skip_ipsec: u_int32_t zone; src0 = ip6->ip6_src; - if (in6_setscope(&src0, origifp, &zone)) + if (in6_setscope(&src0, origifp, &zone)) { goto badscope; - bzero(&src_sa, sizeof (src_sa)); + } + 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; if ((sa6_recoverscope(&src_sa, TRUE) || - zone != src_sa.sin6_scope_id)) + zone != src_sa.sin6_scope_id)) { goto badscope; + } dst0 = ip6->ip6_dst; - if ((in6_setscope(&dst0, origifp, &zone))) + if ((in6_setscope(&dst0, origifp, &zone))) { goto badscope; + } /* re-initialize to be sure */ - 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; if ((sa6_recoverscope(&dst_sa, TRUE) || - zone != dst_sa.sin6_scope_id)) + zone != dst_sa.sin6_scope_id)) { goto badscope; + } /* scope check is done. */ goto routefound; @@ -1144,8 +1270,9 @@ skip_ipsec: badscope: ip6stat.ip6s_badscope++; in6_ifstat_inc(origifp, ifs6_out_discard); - if (error == 0) + if (error == 0) { error = EHOSTUNREACH; /* XXX */ + } goto bad; } @@ -1171,12 +1298,13 @@ routefound: if (ia != NULL && (ifp->if_flags & IFF_LOOPBACK)) { uint32_t srcidx; - if (src_ia != NULL) + if (src_ia != NULL) { srcidx = src_ia->ia_ifp->if_index; - else if (ro->ro_srcia != NULL) + } else if (ro->ro_srcia != NULL) { srcidx = ro->ro_srcia->ifa_ifp->if_index; - else + } else { srcidx = 0; + } ip6_setsrcifaddr_info(m, srcidx, NULL); ip6_setdstifaddr_info(m, 0, ia); @@ -1186,7 +1314,7 @@ routefound: if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */ } else { - struct in6_multi *in6m; + struct in6_multi *in6m; m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST; in6_ifstat_inc_na(ifp, ifs6_out_mcast); @@ -1203,22 +1331,26 @@ routefound: in6_multihead_lock_shared(); IN6_LOOKUP_MULTI(&ip6->ip6_dst, ifp, in6m); in6_multihead_lock_done(); - if (im6o != NULL) + if (im6o != NULL) { IM6O_LOCK(im6o); + } if (in6m != NULL && (im6o == NULL || im6o->im6o_multicast_loop)) { - if (im6o != NULL) + if (im6o != NULL) { IM6O_UNLOCK(im6o); + } /* * If we belong to the destination multicast group * on the outgoing interface, and the caller did not * forbid loopback, loop back a copy. */ ip6_mloopback(NULL, ifp, m, dst, optlen, nxt0); - } else if (im6o != NULL) + } else if (im6o != NULL) { IM6O_UNLOCK(im6o); - if (in6m != NULL) + } + if (in6m != NULL) { IN6M_REMREF(in6m); + } /* * Multicasts with a hoplimit of zero may be looped back, * above, but must not be transmitted on a network. @@ -1229,8 +1361,12 @@ routefound: */ 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; } } @@ -1238,17 +1374,15 @@ routefound: * Fill the outgoing inteface to tell the upper layer * to increment per-interface statistics. */ - if (ifpp != NULL) { - ifnet_reference(ifp); /* for caller */ - if (*ifpp != NULL) - ifnet_release(*ifpp); + if (ifpp != NULL && *ifpp == NULL) { + ifnet_reference(ifp); /* for caller */ *ifpp = ifp; } /* Determine path MTU. */ - if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu, - &alwaysfrag)) != 0) + if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu)) != 0) { goto bad; + } /* * The caller of this function may specify to use the minimum MTU @@ -1280,26 +1414,6 @@ routefound: */ 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 == NULL) { - error = EACCES; - goto done; - } - } -#endif /* IPFW2 */ - /* * If the outgoing packet contains a hop-by-hop options header, * it must be examined and processed even by the source node. @@ -1310,8 +1424,9 @@ routefound: 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) + 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, @@ -1322,11 +1437,15 @@ routefound: 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), + ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh), &dummy, &oplen) < 0) { - /* m was already freed at this point */ - error = EINVAL; /* better error? */ - goto done; + /* + * 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; @@ -1336,8 +1455,9 @@ routefound: check_with_pf: #endif /* DUMMYNET */ #if PF - if (PF_IS_ENABLED) { + if (PF_IS_ENABLED && !skip_pf) { #if DUMMYNET + /* * TODO: Need to save opt->ip6po_flags for reinjection * rdar://10434993 @@ -1345,14 +1465,14 @@ check_with_pf: args.fwa_m = m; args.fwa_oif = ifp; args.fwa_oflags = flags; - if (flags & IPV6_OUTARGS) + if (flags & IPV6_OUTARGS) { args.fwa_ip6oa = ip6oa; + } args.fwa_ro6 = ro; args.fwa_dst6 = dst; args.fwa_ro6_pmtu = ro_pmtu; args.fwa_origifp = origifp; args.fwa_mtu = mtu; - args.fwa_alwaysfrag = alwaysfrag; args.fwa_unfragpartlen = unfragpartlen; args.fwa_exthdrs = &exthdrs; /* Invoke outbound packet filter */ @@ -1362,58 +1482,240 @@ check_with_pf: #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); /* 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; - - if (opt != NULL && (opt->ip6po_flags & IP6PO_DONTFRAG)) - ip6obf.dontfrag = TRUE; - else - ip6obf.dontfrag = FALSE; - if (ip6obf.dontfrag && alwaysfrag) { /* case 4 */ - /* conflicting request - can't transmit */ - error = EMSGSIZE; + error = ip6_fragment_packet(&m, opt, ip6oa, + &exthdrs, ifp, mtu, unfragpartlen, ro_pmtu, nxt0, + optlen); + + if (error) { goto bad; } - lck_rw_lock_shared(nd_if_rwlock); +/* + * 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) + * 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 + */ + +static int +ip6_fragment_packet(struct mbuf **mptr, struct ip6_pktopts *opt, + struct ip6_out_args *ip6oa, struct ip6_exthdrs *exthdrsp, + struct ifnet *ifp, uint32_t mtu, uint32_t unfragpartlen, + struct route_in6 *ro_pmtu, int nxt0, uint32_t optlen) +{ + VERIFY(NULL != mptr); + struct mbuf *m = *mptr; + int error = 0; + uint32_t tlen = m->m_pkthdr.len; + boolean_t dontfrag = (opt != NULL && (opt->ip6po_flags & IP6PO_DONTFRAG)) || + (ip6oa != NULL && (ip6oa->ip6oa_flags & IP6OAF_DONT_FRAG)); + + 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; + } + } + /* Access without acquiring nd_ifinfo lock for performance */ - if (ip6obf.dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */ - lck_rw_done(nd_if_rwlock); + 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 @@ -1426,56 +1728,93 @@ 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, SA(&ro_pmtu->ro_dst), (void *)&ip6cp); - error = EMSGSIZE; - goto bad; - } else { - lck_rw_done(nd_if_rwlock); + return EMSGSIZE; } /* * transmit packet without fragmentation */ - if (ip6obf.dontfrag || (!alwaysfrag && /* case 1-a and 2-a */ + if (dontfrag || (tlen <= mtu || TSO_IPV6_OK(ifp, m) || - (ifp->if_hwassist & CSUM_FRAGMENT_IPV6)))) { -#ifdef IPSEC - /* clean ipsec history once it goes out of the node */ - ipsec_delaux(m); -#endif /* IPSEC */ - + (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); - - if (ro->ro_rt) - RT_LOCK_ASSERT_NOTHELD(ro->ro_rt); - error = nd6_output(ifp, origifp, m, dst, ro->ro_rt, adv); - goto done; + } else { + /* + * time to fragment - cases 1-b is handled inside + * ip6_do_fragmentation(). + * mppn is passed down to be updated to point at fragment chain. + */ + u_int8_t *lexthdrsp; + + if (exthdrsp->ip6e_rthdr != NULL) { + lexthdrsp = mtod(exthdrsp->ip6e_rthdr, uint8_t *); + } else if (exthdrsp->ip6e_dest1 != NULL) { + lexthdrsp = mtod(exthdrsp->ip6e_dest1, uint8_t *); + } else if (exthdrsp->ip6e_hbh != NULL) { + lexthdrsp = mtod(exthdrsp->ip6e_hbh, uint8_t *); + } else { + lexthdrsp = NULL; + } + error = ip6_do_fragmentation(mptr, optlen, ifp, + unfragpartlen, mtod(m, struct ip6_hdr *), lexthdrsp, mtu, + nxt0, htonl(ip6_randomid())); } - /* - * try to fragment the packet. case 1-b and 3 - */ - if ((m->m_pkthdr.csum_flags & CSUM_TSO_IPV6)) { + return error; +} + +/* + * ip6_do_fragmentation() is called by ip6_fragment_packet() after determining + * the packet needs to be fragmented. on success, morig is freed and a chain + * of fragments is linked into the packet chain where morig existed. Otherwise, + * an errno is returned. + * optlen: total length of all extension headers (excludes the IPv6 header). + * unfragpartlen: length of the per-fragment headers which consist of the IPv6 + * header plus any extension headers that must be processed by nodes + * en route to the destination. + * lexthdrsp: pointer to the last extension header in the unfragmentable part + * or NULL. + * nxt0: upper-layer protocol number. + * id: Identification value to be used in the fragment header. + */ +int +ip6_do_fragmentation(struct mbuf **mptr, uint32_t optlen, struct ifnet *ifp, + uint32_t unfragpartlen, struct ip6_hdr *ip6, uint8_t *lexthdrsp, + uint32_t mtu, int nxt0, uint32_t id) +{ + VERIFY(NULL != mptr); + int error = 0; + + struct mbuf *morig = *mptr; + struct mbuf *first_mbufp = NULL; + struct mbuf *last_mbufp = NULL; + + uint32_t tlen = morig->m_pkthdr.len; + + /* try to fragment the packet. case 1-b */ + if ((morig->m_pkthdr.csum_flags & CSUM_TSO_IPV6)) { /* TSO and fragment aren't compatible */ - error = EMSGSIZE; in6_ifstat_inc(ifp, ifs6_out_fragfail); - goto bad; + 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; + uint32_t hlen, off, len; + struct mbuf **mnext = NULL; struct ip6_frag *ip6f; - u_int32_t id = htonl(ip6_randomid()); u_char nextproto; /* @@ -1484,91 +1823,94 @@ check_with_pf: * Must be able to put at least 8 bytes per fragment. */ hlen = unfragpartlen; - if (mtu > IPV6_MAXPACKET) + 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 != NULL) { - nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *); - *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT; - } else if (exthdrs.ip6e_dest1 != NULL) { - nextproto = *mtod(exthdrs.ip6e_dest1, u_char *); - *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT; - } else if (exthdrs.ip6e_hbh != NULL) { - nextproto = *mtod(exthdrs.ip6e_hbh, u_char *); - *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT; + if (lexthdrsp != NULL) { + nextproto = *lexthdrsp; + *lexthdrsp = IPPROTO_FRAGMENT; } else { nextproto = ip6->ip6_nxt; ip6->ip6_nxt = IPPROTO_FRAGMENT; } - if (m->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA) - in6_delayed_cksum_offset(m, 0, optlen, nxt0); + 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) { - struct ip6_hdr *mhip6; + struct ip6_hdr *new_mhip6; + struct mbuf *new_m; + struct mbuf *m_frgpart; - MGETHDR(m, M_DONTWAIT, MT_HEADER); /* MAC-OK */ - if (m == NULL) { + MGETHDR(new_m, M_DONTWAIT, MT_HEADER); /* MAC-OK */ + if (new_m == NULL) { error = ENOBUFS; ip6stat.ip6s_odropped++; - goto sendorfree; + break; + } + new_m->m_pkthdr.rcvif = NULL; + new_m->m_flags = morig->m_flags & M_COPYFLAGS; + + if (first_mbufp != NULL) { + /* Every pass through loop but first */ + *mnext = new_m; + last_mbufp = new_m; + } else { + /* This is the first element of the fragment chain */ + first_mbufp = new_m; + last_mbufp = new_m; } - m->m_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); + 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) + if (off + len >= tlen) { len = tlen - off; - else + } 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)) == NULL) { + } + 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 = NULL; - - M_COPY_CLASSIFIER(m, m0); - M_COPY_PFTAG(m, m0); + m_cat(new_m, m_frgpart); + new_m->m_pkthdr.len = len + hlen + sizeof(*ip6f); + new_m->m_pkthdr.rcvif = NULL; -#ifdef notyet -#if CONFIG_MACF_NET - mac_create_fragment(m0, m); -#endif /* CONFIG_MACF_NET */ -#endif /* notyet */ + M_COPY_CLASSIFIER(new_m, morig); + M_COPY_PFTAG(new_m, morig); + M_COPY_NECPTAG(new_m, morig); ip6f->ip6f_reserved = 0; ip6f->ip6f_ident = id; @@ -1577,81 +1919,23 @@ check_with_pf: 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 = NULL; - m_freem(m0); - for (m0 = m; m != NULL; m = m0) { - m0 = m->m_nextpkt; - m->m_nextpkt = NULL; - if (error == 0) { -#if IPSEC - /* clean ipsec history once it goes out of the node */ - ipsec_delaux(m); -#endif /* IPSEC */ - error = nd6_output(ifp, origifp, m, dst, ro->ro_rt, - adv); + if (error) { + /* free all the fragments created */ + if (first_mbufp != NULL) { + m_freem_list(first_mbufp); + first_mbufp = NULL; + } + last_mbufp = NULL; } else { - m_freem(m); + /* successful fragmenting */ + m_freem(morig); + *mptr = first_mbufp; + last_mbufp->m_nextpkt = NULL; + ip6stat.ip6s_fragmented++; + in6_ifstat_inc(ifp, ifs6_out_fragok); } } - - if (error == 0) - ip6stat.ip6s_fragmented++; - -done: - 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); - return (error); - -freehdrs: - if (exthdrs.ip6e_hbh != NULL) - m_freem(exthdrs.ip6e_hbh); - if (exthdrs.ip6e_dest1 != NULL) - m_freem(exthdrs.ip6e_dest1); - if (exthdrs.ip6e_rthdr != NULL) - m_freem(exthdrs.ip6e_rthdr); - if (exthdrs.ip6e_dest2 != NULL) - m_freem(exthdrs.ip6e_dest2); - /* FALLTHRU */ -bad: - 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 + return error; } static int @@ -1659,26 +1943,28 @@ ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen) { struct mbuf *m; - if (hlen > MCLBYTES) - return (ENOBUFS); /* XXX */ - + if (hlen > MCLBYTES) { + return ENOBUFS; /* XXX */ + } MGET(m, M_DONTWAIT, MT_DATA); - if (m == NULL) - return (ENOBUFS); + if (m == NULL) { + return ENOBUFS; + } if (hlen > MLEN) { MCLGET(m, M_DONTWAIT); if (!(m->m_flags & M_EXT)) { m_free(m); - return (ENOBUFS); + return ENOBUFS; } } m->m_len = hlen; - if (hdr != NULL) + if (hdr != NULL) { bcopy(hdr, mtod(m, caddr_t), hlen); + } *mp = m; - return (0); + return 0; } static void @@ -1712,22 +1998,23 @@ uint32_t in6_finalize_cksum(struct mbuf *m, uint32_t hoff, int32_t optlen, int32_t nxt0, uint32_t csum_flags) { - unsigned char buf[sizeof (struct ip6_hdr)] __attribute__((aligned(8))); + unsigned char buf[sizeof(struct ip6_hdr)] __attribute__((aligned(8))); struct ip6_hdr *ip6; uint32_t offset, mlen, hlen, olen, sw_csum; uint16_t csum, ulpoff, plen; uint8_t nxt; - _CASSERT(sizeof (csum) == sizeof (uint16_t)); + _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) + if ((sw_csum &= CSUM_DELAY_IPV6_DATA) == 0) { goto done; + } - mlen = m->m_pkthdr.len; /* total mbuf len */ - hlen = sizeof (*ip6); /* IPv6 header len */ + 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)) { @@ -1763,7 +2050,7 @@ in6_finalize_cksum(struct mbuf *m, uint32_t hoff, int32_t optlen, ip6->ip6_plen, ip6->ip6_plen, plen, plen, (mlen - (hoff + hlen))); } - plen = mlen - (hoff + hlen); + plen = (uint16_t)(mlen - (hoff + hlen)); } } @@ -1776,12 +2063,13 @@ in6_finalize_cksum(struct mbuf *m, uint32_t hoff, int32_t optlen, olen = 0; } else { /* caller supplied the original transport number; use it */ - if (nxt0 >= 0) - nxt = nxt0; + if (nxt0 >= 0) { + nxt = (uint8_t)nxt0; + } olen = optlen; } - offset = hoff + hlen + olen; /* ULP header */ + offset = hoff + hlen + olen; /* ULP header */ /* sanity check */ if (mlen < offset) { @@ -1795,8 +2083,8 @@ in6_finalize_cksum(struct mbuf *m, uint32_t hoff, int32_t optlen, * 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)) { + 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.) @@ -1805,9 +2093,9 @@ in6_finalize_cksum(struct mbuf *m, uint32_t hoff, int32_t optlen, m->m_pkthdr.csum_tx_start); } - ulpoff = (m->m_pkthdr.csum_data & 0xffff); /* ULP csum offset */ + ulpoff = (m->m_pkthdr.csum_data & 0xffff); /* ULP csum offset */ - if (mlen < (ulpoff + sizeof (csum))) { + 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); @@ -1820,23 +2108,25 @@ in6_finalize_cksum(struct mbuf *m, uint32_t hoff, int32_t optlen, ip6_out_cksum_stats(nxt, plen - olen); /* RFC1122 4.1.3.4 */ - if (csum == 0 && (m->m_pkthdr.csum_flags & CSUM_UDPIPV6)) + 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); + 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)); + bcopy(&csum, (mtod(m, char *) + offset), sizeof(csum)); } - m->m_pkthdr.csum_flags &= - ~(CSUM_DELAY_IPV6_DATA | CSUM_DATA_VALID | CSUM_PARTIAL); + m->m_pkthdr.csum_flags &= ~(CSUM_DELAY_IPV6_DATA | CSUM_DATA_VALID | + CSUM_PARTIAL | CSUM_ZERO_INVERT); done: - return (sw_csum); + return sw_csum; } /* @@ -1849,7 +2139,7 @@ ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) 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. @@ -1859,11 +2149,12 @@ ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) */ if (exthdrs->ip6e_hbh == NULL) { MGET(mopt, M_DONTWAIT, MT_DATA); - if (mopt == NULL) - return (ENOBUFS); + if (mopt == NULL) { + return ENOBUFS; + } mopt->m_len = JUMBOOPTLEN; optbuf = mtod(mopt, u_char *); - optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ + optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */ exthdrs->ip6e_hbh = mopt; } else { struct ip6_hbh *hbh; @@ -1883,8 +2174,9 @@ ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) * XXX: give up if the whole (new) hbh header does * not fit even in an mbuf cluster. */ - if (oldoptlen + JUMBOOPTLEN > MCLBYTES) - return (ENOBUFS); + if (oldoptlen + JUMBOOPTLEN > MCLBYTES) { + return ENOBUFS; + } /* * As a consequence, we must always prepare a cluster @@ -1898,8 +2190,9 @@ ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen) n = NULL; } } - if (n == NULL) - return (ENOBUFS); + if (n == NULL) { + return ENOBUFS; + } n->m_len = oldoptlen + JUMBOOPTLEN; bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), oldoptlen); @@ -1925,12 +2218,12 @@ ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t 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; - return (0); + return 0; #undef JUMBOOPTLEN } @@ -1943,61 +2236,65 @@ ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen, { 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 == NULL) - return (ENOBUFS); + if (hlen > sizeof(struct ip6_hdr)) { + n = m_copym(m0, sizeof(struct ip6_hdr), + hlen - sizeof(struct ip6_hdr), M_DONTWAIT); + if (n == NULL) { + return ENOBUFS; + } m->m_next = n; - } else + } else { n = m; + } /* Search for the last mbuf of unfragmentable part. */ - for (mlast = n; mlast->m_next; mlast = mlast->m_next) + for (mlast = n; mlast->m_next; mlast = mlast->m_next) { ; + } if (!(mlast->m_flags & M_EXT) && - M_TRAILINGSPACE(mlast) >= sizeof (struct ip6_frag)) { + 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 == NULL) - return (ENOBUFS); - mfrg->m_len = sizeof (struct ip6_frag); + if (mfrg == NULL) { + return ENOBUFS; + } + mfrg->m_len = sizeof(struct ip6_frag); *frghdrp = mtod(mfrg, struct ip6_frag *); mlast->m_next = mfrg; } - return (0); + return 0; } static int ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, - struct ifnet *ifp, struct in6_addr *dst, u_int32_t *mtup, - boolean_t *alwaysfragp) + struct ifnet *ifp, struct in6_addr *dst, u_int32_t *mtup) { u_int32_t mtu = 0; - boolean_t alwaysfrag = FALSE; int error = 0; + if (ro_pmtu != ro) { /* The first hop and the final destination may differ. */ struct sockaddr_in6 *sa6_dst = SIN6(&ro_pmtu->ro_dst); if (ROUTE_UNUSABLE(ro_pmtu) || - !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst)) + !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, @@ -2008,12 +2305,11 @@ ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, if (ro_pmtu->ro_rt != NULL) { u_int32_t ifmtu; - if (ifp == NULL) + if (ifp == NULL) { ifp = ro_pmtu->ro_rt->rt_ifp; - 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); /* * Access rmx_mtu without holding the route entry lock, @@ -2034,35 +2330,21 @@ ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, * this case happens with path MTU discovery timeouts. */ mtu = ifmtu; - if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) + if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU)) { ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */ - } else if (mtu < IPV6_MMTU) { - /* - * RFC2460 section 5, last paragraph: - * if we record ICMPv6 too big message with - * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU - * or smaller, with framgent header attached. - * (fragment header is needed regardless from the - * packet size, for translators to identify packets) - */ - alwaysfrag = TRUE; - mtu = IPV6_MMTU; + } } } else { if (ifp) { - 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 { error = EHOSTUNREACH; /* XXX */ } } *mtup = mtu; - if (alwaysfragp != NULL) - *alwaysfragp = alwaysfrag; - return (error); + return error; } /* @@ -2077,8 +2359,9 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) struct inpcb *in6p = sotoinpcb(so); int error = 0, optval = 0; int level, op = -1, optname = 0; - int optlen = 0; + size_t optlen = 0; struct proc *p; + lck_mtx_t *mutex_held = NULL; VERIFY(sopt != NULL); @@ -2092,18 +2375,37 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) privileged = (proc_suser(p) == 0); if (level == IPPROTO_IPV6) { + boolean_t capture_exthdrstat_in = FALSE; switch (op) { case SOPT_SET: + mutex_held = socket_getlock(so, PR_F_WILLUNLOCK); + /* + * Wait if we are in the middle of ip6_output + * as we unlocked the socket there and don't + * want to overwrite the IP options + */ + if (in6p->inp_sndinprog_cnt > 0) { + in6p->inp_sndingprog_waiters++; + + while (in6p->inp_sndinprog_cnt > 0) { + msleep(&in6p->inp_sndinprog_cnt, mutex_held, + PSOCK | PCATCH, "inp_sndinprog_cnt", + NULL); + } + in6p->inp_sndingprog_waiters--; + } switch (optname) { case IPV6_2292PKTOPTIONS: { struct mbuf *m; error = soopt_getm(sopt, &m); - if (error != 0) + if (error != 0) { break; + } error = soopt_mcopyin(sopt, m); - if (error != 0) + if (error != 0) { break; + } error = ip6_pcbopts(&in6p->in6p_outputopts, m, so, sopt); m_freem(m); @@ -2126,9 +2428,10 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) case IPV6_RECVHOPOPTS: case IPV6_RECVDSTOPTS: case IPV6_RECVRTHDRDSTOPTS: - if (!privileged) + if (!privileged) { break; - /* FALLTHROUGH */ + } + OS_FALLTHROUGH; case IPV6_UNICAST_HOPS: case IPV6_HOPLIMIT: case IPV6_RECVPKTINFO: @@ -2138,14 +2441,15 @@ ip6_ctloutput(struct socket *so, struct sockopt *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)); - if (error) + sizeof(optval), sizeof(optval)); + if (error) { break; + } switch (optname) { case IPV6_UNICAST_HOPS: @@ -2153,30 +2457,30 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) error = EINVAL; } else { /* -1 = kernel default */ - in6p->in6p_hops = optval; + in6p->in6p_hops = (short)optval; if (in6p->inp_vflag & INP_IPV4) { in6p->inp_ip_ttl = - optval; + (uint8_t)optval; } } break; -#define OPTSET(bit) do { \ - if (optval) \ - in6p->inp_flags |= (bit); \ - else \ - in6p->inp_flags &= ~(bit); \ +#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); \ +#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) +#define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0) case IPV6_RECVPKTINFO: /* cannot mix with RFC2292 */ @@ -2197,7 +2501,7 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) } optp = &in6p->in6p_outputopts; error = ip6_pcbopt(IPV6_HOPLIMIT, - (u_char *)&optval, sizeof (optval), + (u_char *)&optval, sizeof(optval), optp, uproto); break; } @@ -2218,6 +2522,7 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) break; } OPTSET(IN6P_HOPOPTS); + capture_exthdrstat_in = TRUE; break; case IPV6_RECVDSTOPTS: @@ -2227,6 +2532,7 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) break; } OPTSET(IN6P_DSTOPTS); + capture_exthdrstat_in = TRUE; break; case IPV6_RECVRTHDRDSTOPTS: @@ -2236,6 +2542,7 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) break; } OPTSET(IN6P_RTHDRDSTOPTS); + capture_exthdrstat_in = TRUE; break; case IPV6_RECVRTHDR: @@ -2245,6 +2552,7 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) break; } OPTSET(IN6P_RTHDR); + capture_exthdrstat_in = TRUE; break; case IPV6_RECVPATHMTU: @@ -2254,8 +2562,9 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) * (RFC3542 leaves this case * unspecified.) */ - if (uproto != IPPROTO_TCP) + if (uproto != IPPROTO_TCP) { OPTSET(IN6P_MTU); + } break; case IPV6_V6ONLY: @@ -2266,15 +2575,16 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) */ if (in6p->inp_lport || !IN6_IS_ADDR_UNSPECIFIED( - &in6p->in6p_laddr)) { + &in6p->in6p_laddr)) { error = EINVAL; break; } OPTSET(IN6P_IPV6_V6ONLY); - if (optval) + if (optval) { in6p->inp_vflag &= ~INP_IPV4; - else + } else { in6p->inp_vflag |= INP_IPV4; + } break; case IPV6_RECVTCLASS: @@ -2285,7 +2595,6 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) case IPV6_AUTOFLOWLABEL: OPTSET(IN6P_AUTOFLOWLABEL); break; - } break; @@ -2295,18 +2604,26 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) case IPV6_PREFER_TEMPADDR: { struct ip6_pktopts **optp; - if (optlen != sizeof (optval)) { + if (optlen != sizeof(optval)) { error = EINVAL; break; } error = sooptcopyin(sopt, &optval, - sizeof (optval), sizeof (optval)); - if (error) + sizeof(optval), sizeof(optval)); + if (error) { break; + } optp = &in6p->in6p_outputopts; error = ip6_pcbopt(optname, (u_char *)&optval, - sizeof (optval), optp, uproto); + 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; } @@ -2316,14 +2633,15 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) 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)); - if (error) + sizeof(optval), sizeof(optval)); + if (error) { break; + } switch (optname) { case IPV6_2292PKTINFO: OPTSET2292(IN6P_PKTINFO); @@ -2336,18 +2654,23 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) * Check super-user privilege. * See comments for IPV6_RECVHOPOPTS. */ - if (!privileged) - return (EPERM); + if (!privileged) { + return EPERM; + } OPTSET2292(IN6P_HOPOPTS); + capture_exthdrstat_in = TRUE; break; case IPV6_2292DSTOPTS: - if (!privileged) - return (EPERM); - OPTSET2292(IN6P_DSTOPTS| + if (!privileged) { + 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; @@ -2368,11 +2691,13 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) break; } error = soopt_getm(sopt, &m); - if (error != 0) + if (error != 0) { break; + } error = soopt_mcopyin(sopt, m); - if (error != 0) + if (error != 0) { break; + } optp = &in6p->in6p_outputopts; error = ip6_pcbopt(optname, mtod(m, u_char *), @@ -2398,9 +2723,10 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) case IPV6_PORTRANGE: error = sooptcopyin(sopt, &optval, - sizeof (optval), sizeof (optval)); - if (error) + sizeof(optval), sizeof(optval)); + if (error) { break; + } switch (optval) { case IPV6_PORTRANGE_DEFAULT: @@ -2428,34 +2754,22 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) caddr_t req = NULL; size_t len = 0; struct mbuf *m; - - if ((error = soopt_getm(sopt, &m)) != 0) + + if ((error = soopt_getm(sopt, &m)) != 0) { break; - if ((error = soopt_mcopyin(sopt, m)) != 0) + } + if ((error = soopt_mcopyin(sopt, m)) != 0) { break; - + } + 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 /* 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 - error = ENOPROTOOPT; - break; - } -#endif /* IPFIREWALL */ /* * IPv6 variant of IP_BOUND_IF; for details see * comments on IP_BOUND_IF in ip_ctloutput(). @@ -2468,10 +2782,11 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) } error = sooptcopyin(sopt, &optval, - sizeof (optval), sizeof (optval)); + sizeof(optval), sizeof(optval)); - if (error) + if (error) { break; + } error = inp_bindif(in6p, optval, NULL); break; @@ -2484,10 +2799,11 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) } error = sooptcopyin(sopt, &optval, - sizeof (optval), sizeof (optval)); + sizeof(optval), sizeof(optval)); - if (error) + if (error) { break; + } /* once set, it cannot be unset */ if (!optval && INP_NO_CELLULAR(in6p)) { @@ -2508,11 +2824,17 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) 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: switch (optname) { - case IPV6_2292PKTOPTIONS: /* * RFC3542 (effectively) deprecated the @@ -2539,7 +2861,6 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) case IPV6_RECVTCLASS: case IPV6_AUTOFLOWLABEL: switch (optname) { - case IPV6_RECVHOPOPTS: optval = OPTBIT(IN6P_HOPOPTS); break; @@ -2579,12 +2900,13 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) case IPV6_PORTRANGE: { int flags; flags = in6p->inp_flags; - if (flags & INP_HIGHPORT) + if (flags & INP_HIGHPORT) { optval = IPV6_PORTRANGE_HIGH; - else if (flags & INP_LOWPORT) + } else if (flags & INP_LOWPORT) { optval = IPV6_PORTRANGE_LOW; - else + } else { optval = 0; + } break; } case IPV6_RECVTCLASS: @@ -2595,10 +2917,11 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) optval = OPTBIT(IN6P_AUTOFLOWLABEL); break; } - if (error) + if (error) { break; + } error = sooptcopyout(sopt, &optval, - sizeof (optval)); + sizeof(optval)); break; case IPV6_PATHMTU: { @@ -2606,27 +2929,30 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) struct ip6_mtuinfo mtuinfo; struct route_in6 sro; - bzero(&sro, sizeof (sro)); + bzero(&sro, sizeof(sro)); - if (!(so->so_state & SS_ISCONNECTED)) - return (ENOTCONN); + if (!(so->so_state & SS_ISCONNECTED)) { + return ENOTCONN; + } /* * XXX: we dot not consider the case of source * routing, or optional information to specify * the outgoing interface. */ error = ip6_getpmtu(&sro, NULL, NULL, - &in6p->in6p_faddr, &pmtu, NULL); + &in6p->in6p_faddr, &pmtu); ROUTE_RELEASE(&sro); - if (error) + if (error) { break; - if (pmtu > IPV6_MAXPACKET) + } + 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; @@ -2651,12 +2977,12 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) optval = OPTBIT(IN6P_RTHDR); break; case IPV6_2292DSTOPTS: - optval = OPTBIT(IN6P_DSTOPTS| + optval = OPTBIT(IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS); break; } error = sooptcopyout(sopt, &optval, - sizeof (optval)); + sizeof(optval)); break; case IPV6_PKTINFO: @@ -2685,35 +3011,25 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) break; } #endif /* IPSEC */ -#if IPFIREWALL - case IPV6_FW_GET: { - if (ip6_fw_ctl_ptr == NULL) - load_ip6fw(); - if (ip6_fw_ctl_ptr != NULL) - error = (*ip6_fw_ctl_ptr)(sopt); - else - error = ENOPROTOOPT; - break; - } -#endif /* IPFIREWALL */ case IPV6_BOUND_IF: - if (in6p->inp_flags & INP_BOUND_IF) + if (in6p->inp_flags & INP_BOUND_IF) { optval = in6p->inp_boundifp->if_index; + } error = sooptcopyout(sopt, &optval, - sizeof (optval)); + sizeof(optval)); break; case IPV6_NO_IFT_CELLULAR: optval = INP_NO_CELLULAR(in6p) ? 1 : 0; error = sooptcopyout(sopt, &optval, - sizeof (optval)); + sizeof(optval)); break; case IPV6_OUT_IF: optval = (in6p->in6p_last_outifp != NULL) ? in6p->in6p_last_outifp->if_index : 0; error = sooptcopyout(sopt, &optval, - sizeof (optval)); + sizeof(optval)); break; default: @@ -2722,16 +3038,19 @@ ip6_ctloutput(struct socket *so, struct sockopt *sopt) } break; } + } else if (level == IPPROTO_UDP) { + error = udp_ctloutput(so, sopt); } else { error = EINVAL; } - return (error); + return error; } int ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) { - int error = 0, optval, optlen; + int error = 0, optval; + size_t optlen; const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum); struct inpcb *in6p = sotoinpcb(so); int level, op, optname; @@ -2741,8 +3060,9 @@ ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) optname = sopt->sopt_name; optlen = sopt->sopt_valsize; - if (level != IPPROTO_IPV6) - return (EINVAL); + if (level != IPPROTO_IPV6) { + return EINVAL; + } switch (optname) { case IPV6_CHECKSUM: @@ -2756,32 +3076,35 @@ 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)); - if (error) + error = sooptcopyin(sopt, &optval, sizeof(optval), + sizeof(optval)); + if (error) { break; + } if ((optval % 2) != 0) { /* the API assumes even offset values */ error = EINVAL; } else if (SOCK_PROTO(so) == IPPROTO_ICMPV6) { - if (optval != icmp6off) + if (optval != icmp6off) { error = EINVAL; + } } else { in6p->in6p_cksum = optval; } break; case SOPT_GET: - if (SOCK_PROTO(so) == IPPROTO_ICMPV6) + if (SOCK_PROTO(so) == IPPROTO_ICMPV6) { optval = icmp6off; - else + } else { optval = in6p->in6p_cksum; + } - error = sooptcopyout(sopt, &optval, sizeof (optval)); + error = sooptcopyout(sopt, &optval, sizeof(optval)); break; default: @@ -2795,7 +3118,7 @@ ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt) break; } - return (error); + return error; } /* @@ -2815,15 +3138,17 @@ ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so, #if DIAGNOSTIC if (opt->ip6po_pktinfo || opt->ip6po_nexthop || opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 || - opt->ip6po_rhinfo.ip6po_rhi_rthdr) + opt->ip6po_rhinfo.ip6po_rhi_rthdr) { printf("%s: all specified options are cleared.\n", __func__); + } #endif ip6_clearpktopts(opt, -1); } else { - opt = _MALLOC(sizeof (*opt), M_IP6OPT, M_WAITOK); - if (opt == NULL) - return (ENOBUFS); + opt = _MALLOC(sizeof(*opt), M_IP6OPT, M_WAITOK); + if (opt == NULL) { + return ENOBUFS; + } } *pktopt = NULL; @@ -2832,19 +3157,20 @@ ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so, * Only turning off any previous options, regardless of * whether the opt is just created or given. */ - if (opt != NULL) + if (opt != NULL) { FREE(opt, M_IP6OPT); - return (0); + } + return 0; } /* set options specified by user. */ 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; } /* @@ -2854,10 +3180,9 @@ ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so, void ip6_initpktopts(struct ip6_pktopts *opt) { - - bzero(opt, sizeof (*opt)); - opt->ip6po_hlim = -1; /* -1 means default hop limit */ - opt->ip6po_tclass = -1; /* -1 means default traffic class */ + bzero(opt, sizeof(*opt)); + opt->ip6po_hlim = -1; /* -1 means default hop limit */ + opt->ip6po_tclass = -1; /* -1 means default traffic class */ opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY; opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM; } @@ -2870,14 +3195,15 @@ 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); - if (opt == NULL) - return (ENOBUFS); + opt = _MALLOC(sizeof(*opt), M_IP6OPT, M_WAITOK); + if (opt == NULL) { + return ENOBUFS; + } ip6_initpktopts(opt); *pktopt = opt; } - return (ip6_setpktopt(optname, buf, len, opt, 1, 0, uproto)); + return ip6_setpktopt(optname, buf, len, opt, 1, 0, uproto); } static int @@ -2894,22 +3220,23 @@ ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) switch (optname) { case IPV6_PKTINFO: - if (pktopt && pktopt->ip6po_pktinfo) + if (pktopt && pktopt->ip6po_pktinfo) { optdata = (void *)pktopt->ip6po_pktinfo; - else { + } 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) + if (pktopt && pktopt->ip6po_tclass >= 0) { optdata = (void *)&pktopt->ip6po_tclass; - else + } else { optdata = (void *)&deftclass; - optdatalen = sizeof (int); + } + optdatalen = sizeof(int); break; case IPV6_HOPOPTS: @@ -2952,118 +3279,132 @@ ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt) break; case IPV6_USE_MIN_MTU: - if (pktopt) + if (pktopt) { optdata = (void *)&pktopt->ip6po_minmtu; - else + } else { optdata = (void *)&defminmtu; - optdatalen = sizeof (int); + } + optdatalen = sizeof(int); break; case IPV6_DONTFRAG: - if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) + if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG)) { on = 1; - else + } else { on = 0; + } optdata = (void *)&on; - optdatalen = sizeof (on); + optdatalen = sizeof(on); break; case IPV6_PREFER_TEMPADDR: - if (pktopt) + if (pktopt) { optdata = (void *)&pktopt->ip6po_prefer_tempaddr; - else + } else { optdata = (void *)&defpreftemp; - optdatalen = sizeof (int); + } + optdatalen = sizeof(int); break; - default: /* should not happen */ + default: /* should not happen */ #ifdef DIAGNOSTIC panic("ip6_getpcbopt: unexpected option\n"); #endif - return (ENOPROTOOPT); + return ENOPROTOOPT; } - return (sooptcopyout(sopt, optdata, optdatalen)); + return sooptcopyout(sopt, optdata, optdatalen); } void ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname) { - if (pktopt == NULL) + if (pktopt == NULL) { return; + } if (optname == -1 || optname == IPV6_PKTINFO) { - if (pktopt->ip6po_pktinfo) + if (pktopt->ip6po_pktinfo) { FREE(pktopt->ip6po_pktinfo, M_IP6OPT); + } pktopt->ip6po_pktinfo = NULL; } - if (optname == -1 || optname == IPV6_HOPLIMIT) + if (optname == -1 || optname == IPV6_HOPLIMIT) { pktopt->ip6po_hlim = -1; - if (optname == -1 || optname == IPV6_TCLASS) + } + if (optname == -1 || optname == IPV6_TCLASS) { pktopt->ip6po_tclass = -1; + } if (optname == -1 || optname == IPV6_NEXTHOP) { ROUTE_RELEASE(&pktopt->ip6po_nextroute); - if (pktopt->ip6po_nexthop) + if (pktopt->ip6po_nexthop) { FREE(pktopt->ip6po_nexthop, M_IP6OPT); + } pktopt->ip6po_nexthop = NULL; } if (optname == -1 || optname == IPV6_HOPOPTS) { - if (pktopt->ip6po_hbh) + if (pktopt->ip6po_hbh) { FREE(pktopt->ip6po_hbh, M_IP6OPT); + } pktopt->ip6po_hbh = NULL; } if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) { - if (pktopt->ip6po_dest1) + if (pktopt->ip6po_dest1) { FREE(pktopt->ip6po_dest1, M_IP6OPT); + } pktopt->ip6po_dest1 = NULL; } if (optname == -1 || optname == IPV6_RTHDR) { - if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) + if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr) { FREE(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT); + } pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL; ROUTE_RELEASE(&pktopt->ip6po_route); } if (optname == -1 || optname == IPV6_DSTOPTS) { - if (pktopt->ip6po_dest2) + if (pktopt->ip6po_dest2) { FREE(pktopt->ip6po_dest2, M_IP6OPT); + } pktopt->ip6po_dest2 = NULL; } } -#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 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) { - if (dst == NULL || src == NULL) { + if (dst == NULL || src == NULL) { printf("copypktopts: invalid argument\n"); - return (EINVAL); + return EINVAL; } dst->ip6po_hlim = src->ip6po_hlim; dst->ip6po_tclass = src->ip6po_tclass; dst->ip6po_flags = src->ip6po_flags; if (src->ip6po_pktinfo) { - dst->ip6po_pktinfo = _MALLOC(sizeof (*dst->ip6po_pktinfo), + dst->ip6po_pktinfo = _MALLOC(sizeof(*dst->ip6po_pktinfo), M_IP6OPT, canwait); - if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT) + 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); - if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) + if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT) { goto bad; + } bcopy(src->ip6po_nexthop, dst->ip6po_nexthop, src->ip6po_nexthop->sa_len); } @@ -3071,11 +3412,11 @@ copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait) PKTOPT_EXTHDRCPY(ip6po_dest1); PKTOPT_EXTHDRCPY(ip6po_dest2); PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */ - return (0); + return 0; bad: ip6_clearpktopts(dst, -1); - return (ENOBUFS); + return ENOBUFS; } #undef PKTOPT_EXTHDRCPY @@ -3085,24 +3426,26 @@ ip6_copypktopts(struct ip6_pktopts *src, int canwait) int error; struct ip6_pktopts *dst; - dst = _MALLOC(sizeof (*dst), M_IP6OPT, canwait); - if (dst == NULL) - return (NULL); + dst = _MALLOC(sizeof(*dst), M_IP6OPT, canwait); + if (dst == NULL) { + return NULL; + } ip6_initpktopts(dst); if ((error = copypktopts(dst, src, canwait)) != 0) { FREE(dst, M_IP6OPT); - return (NULL); + return NULL; } - return (dst); + return dst; } void ip6_freepcbopts(struct ip6_pktopts *pktopt) { - if (pktopt == NULL) + if (pktopt == NULL) { return; + } ip6_clearpktopts(pktopt, -1); @@ -3112,27 +3455,22 @@ ip6_freepcbopts(struct ip6_pktopts *pktopt) void ip6_moptions_init(void) { - PE_parse_boot_argn("ifa_debug", &im6o_debug, sizeof (im6o_debug)); + PE_parse_boot_argn("ifa_debug", &im6o_debug, sizeof(im6o_debug)); - im6o_size = (im6o_debug == 0) ? sizeof (struct ip6_moptions) : - sizeof (struct ip6_moptions_dbg); + vm_size_t im6o_size = (im6o_debug == 0) ? sizeof(struct ip6_moptions) : + sizeof(struct ip6_moptions_dbg); - im6o_zone = zinit(im6o_size, IM6O_ZONE_MAX * im6o_size, 0, - IM6O_ZONE_NAME); - if (im6o_zone == NULL) { - panic("%s: failed allocating %s", __func__, IM6O_ZONE_NAME); - /* NOTREACHED */ - } - zone_change(im6o_zone, Z_EXPAND, TRUE); + im6o_zone = zone_create(IM6O_ZONE_NAME, im6o_size, ZC_ZFREE_CLEARMEM); } void im6o_addref(struct ip6_moptions *im6o, int locked) { - if (!locked) + if (!locked) { IM6O_LOCK(im6o); - else + } else { IM6O_LOCK_ASSERT_HELD(im6o); + } if (++im6o->im6o_refcnt == 0) { panic("%s: im6o %p wraparound refcnt\n", __func__, im6o); @@ -3141,8 +3479,9 @@ im6o_addref(struct ip6_moptions *im6o, int locked) (*im6o->im6o_trace)(im6o, TRUE); } - if (!locked) + if (!locked) { IM6O_UNLOCK(im6o); + } } void @@ -3168,13 +3507,15 @@ im6o_remref(struct ip6_moptions *im6o) struct in6_mfilter *imf; imf = im6o->im6o_mfilters ? &im6o->im6o_mfilters[i] : NULL; - if (imf != NULL) + if (imf != NULL) { im6f_leave(imf); + } (void) in6_mc_leave(im6o->im6o_membership[i], imf); - if (imf != NULL) + if (imf != NULL) { im6f_purge(imf); + } IN6M_REMREF(im6o->im6o_membership[i]); im6o->im6o_membership[i] = NULL; @@ -3224,14 +3565,12 @@ im6o_trace(struct ip6_moptions *im6o, int refhold) } struct ip6_moptions * -ip6_allocmoptions(int how) +ip6_allocmoptions(zalloc_flags_t how) { struct ip6_moptions *im6o; - im6o = (how == M_WAITOK) ? - zalloc(im6o_zone) : zalloc_noblock(im6o_zone); + im6o = zalloc_flags(im6o_zone, how | Z_ZERO); if (im6o != NULL) { - bzero(im6o, im6o_size); lck_mtx_init(&im6o->im6o_lock, ifa_mtx_grp, ifa_mtx_attr); im6o->im6o_debug |= IFD_ALLOC; if (im6o_debug != 0) { @@ -3241,7 +3580,7 @@ ip6_allocmoptions(int how) IM6O_ADDREF(im6o); } - return (im6o); + return im6o; } /* @@ -3253,8 +3592,9 @@ ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, { struct cmsghdr *cm = NULL; - if (control == NULL || opt == NULL) - return (EINVAL); + if (control == NULL || opt == NULL) { + return EINVAL; + } ip6_initpktopts(opt); if (stickyopt) { @@ -3269,37 +3609,40 @@ ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt, * but we can allow this since this option should be rarely * used. */ - if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) - return (error); + if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0) { + return error; + } } /* * XXX: Currently, we assume all the optional information is stored * in a single mbuf. */ - if (control->m_next) - return (EINVAL); + if (control->m_next) { + return EINVAL; + } - if (control->m_len < CMSG_LEN(0)) - return (EINVAL); + if (control->m_len < CMSG_LEN(0)) { + return EINVAL; + } - for (cm = M_FIRST_CMSGHDR(control); cm != NULL; + for (cm = M_FIRST_CMSGHDR(control); + is_cmsg_valid(control, cm); cm = M_NXT_CMSGHDR(control, cm)) { int error; - if (cm->cmsg_len < sizeof (struct cmsghdr) || - cm->cmsg_len > control->m_len) - return (EINVAL); - if (cm->cmsg_level != IPPROTO_IPV6) + if (cm->cmsg_level != IPPROTO_IPV6) { continue; + } error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm), cm->cmsg_len - CMSG_LEN(0), opt, 0, 1, uproto); - if (error) - return (error); + if (error) { + return error; + } } - return (0); + return 0; } /* * Set a particular packet option, as a sticky option or an ancillary data @@ -3316,19 +3659,20 @@ 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 printf("ip6_setpktopt: impossible case\n"); #endif - return (EINVAL); + return EINVAL; } /* * Caller must have ensured that the buffer is at least * aligned on 32-bit boundary. */ - VERIFY(IS_P2ALIGNED(buf, sizeof (u_int32_t))); + VERIFY(IS_P2ALIGNED(buf, sizeof(u_int32_t))); /* * IPV6_2292xxx is for backward compatibility to RFC2292, and should @@ -3344,7 +3688,7 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, case IPV6_2292DSTOPTS: case IPV6_2292RTHDR: case IPV6_2292PKTOPTIONS: - return (ENOPROTOOPT); + return ENOPROTOOPT; } } if (sticky && cmsg) { @@ -3360,7 +3704,7 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, case IPV6_DONTFRAG: case IPV6_TCLASS: case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */ - return (ENOPROTOOPT); + return ENOPROTOOPT; } } @@ -3370,8 +3714,9 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, struct ifnet *ifp = NULL; struct in6_pktinfo *pktinfo; - if (len != sizeof (struct in6_pktinfo)) - return (EINVAL); + if (len != sizeof(struct in6_pktinfo)) { + return EINVAL; + } pktinfo = (struct in6_pktinfo *)(void *)buf; @@ -3390,7 +3735,7 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO && sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) { - return (EINVAL); + return EINVAL; } /* validate the interface index if specified. */ @@ -3398,14 +3743,14 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, if (pktinfo->ipi6_ifindex > if_index) { ifnet_head_done(); - return (ENXIO); + return ENXIO; } if (pktinfo->ipi6_ifindex) { ifp = ifindex2ifnet[pktinfo->ipi6_ifindex]; if (ifp == NULL) { ifnet_head_done(); - return (ENXIO); + return ENXIO; } } @@ -3422,12 +3767,13 @@ 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); + if (opt->ip6po_pktinfo == NULL) { + return ENOBUFS; + } } - bcopy(pktinfo, opt->ip6po_pktinfo, sizeof (*pktinfo)); + bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo)); break; } @@ -3439,14 +3785,17 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT * to simplify the ordering among hoplimit options. */ - if (optname == IPV6_HOPLIMIT && sticky) - return (ENOPROTOOPT); + if (optname == IPV6_HOPLIMIT && sticky) { + return ENOPROTOOPT; + } - if (len != sizeof (int)) - return (EINVAL); + if (len != sizeof(int)) { + return EINVAL; + } hlimp = (int *)(void *)buf; - if (*hlimp < -1 || *hlimp > 255) - return (EINVAL); + if (*hlimp < -1 || *hlimp > IPV6_MAXHLIM) { + return EINVAL; + } opt->ip6po_hlim = *hlimp; break; @@ -3455,11 +3804,13 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, case IPV6_TCLASS: { int tclass; - if (len != sizeof (int)) - return (EINVAL); + if (len != sizeof(int)) { + return EINVAL; + } tclass = *(int *)(void *)buf; - if (tclass < -1 || tclass > 255) - return (EINVAL); + if (tclass < -1 || tclass > 255) { + return EINVAL; + } opt->ip6po_tclass = tclass; break; @@ -3468,45 +3819,49 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, case IPV6_2292NEXTHOP: case IPV6_NEXTHOP: error = suser(kauth_cred_get(), 0); - if (error) - return (EACCES); + if (error) { + return EACCES; + } - if (len == 0) { /* just remove the option */ + if (len == 0) { /* just remove the option */ ip6_clearpktopts(opt, IPV6_NEXTHOP); break; } /* check if cmsg_len is large enough for sa_len */ - if (len < sizeof (struct sockaddr) || len < *buf) - return (EINVAL); + if (len < sizeof(struct sockaddr) || len < *buf) { + return EINVAL; + } switch (SA(buf)->sa_family) { case AF_INET6: { struct sockaddr_in6 *sa6 = SIN6(buf); - if (sa6->sin6_len != sizeof (struct sockaddr_in6)) - return (EINVAL); + if (sa6->sin6_len != sizeof(struct sockaddr_in6)) { + return EINVAL; + } if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) || IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) { - return (EINVAL); + return EINVAL; } if ((error = sa6_embedscope(sa6, ip6_use_defzone)) != 0) { - return (error); + return error; } break; } - case AF_LINK: /* should eventually be supported */ + case AF_LINK: /* should eventually be supported */ default: - return (EAFNOSUPPORT); + return EAFNOSUPPORT; } /* turn off the previous option, then set the new option. */ ip6_clearpktopts(opt, IPV6_NEXTHOP); opt->ip6po_nexthop = _MALLOC(*buf, M_IP6OPT, M_NOWAIT); - if (opt->ip6po_nexthop == NULL) - return (ENOBUFS); + if (opt->ip6po_nexthop == NULL) { + return ENOBUFS; + } bcopy(buf, opt->ip6po_nexthop, *buf); break; @@ -3521,29 +3876,33 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, * overhead. */ error = suser(kauth_cred_get(), 0); - if (error) - return (EACCES); + if (error) { + return EACCES; + } if (len == 0) { ip6_clearpktopts(opt, IPV6_HOPOPTS); - break; /* just remove the option */ + break; /* just remove the option */ } /* message length validation */ - if (len < sizeof (struct ip6_hbh)) - return (EINVAL); + if (len < sizeof(struct ip6_hbh)) { + return EINVAL; + } hbh = (struct ip6_hbh *)(void *)buf; hbhlen = (hbh->ip6h_len + 1) << 3; - if (len != hbhlen) - return (EINVAL); + if (len != hbhlen) { + return EINVAL; + } /* turn off the previous option, then set the new option. */ ip6_clearpktopts(opt, IPV6_HOPOPTS); opt->ip6po_hbh = _MALLOC(hbhlen, M_IP6OPT, M_NOWAIT); - if (opt->ip6po_hbh == NULL) - return (ENOBUFS); + if (opt->ip6po_hbh == NULL) { + return ENOBUFS; + } bcopy(hbh, opt->ip6po_hbh, hbhlen); - + capture_exthdrstat_out = TRUE; break; } @@ -3554,21 +3913,24 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, int destlen; error = suser(kauth_cred_get(), 0); - if (error) - return (EACCES); + if (error) { + return EACCES; + } if (len == 0) { ip6_clearpktopts(opt, optname); - break; /* just remove the option */ + break; /* just remove the option */ } /* message length validation */ - if (len < sizeof (struct ip6_dest)) - return (EINVAL); + if (len < sizeof(struct ip6_dest)) { + return EINVAL; + } dest = (struct ip6_dest *)(void *)buf; destlen = (dest->ip6d_len + 1) << 3; - if (len != destlen) - return (EINVAL); + if (len != destlen) { + return EINVAL; + } /* * Determine the position that the destination options header @@ -3588,10 +3950,11 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, * RFC3542 solved the ambiguity by introducing * separate ancillary data or option types. */ - if (opt->ip6po_rthdr == NULL) + if (opt->ip6po_rthdr == NULL) { newdest = &opt->ip6po_dest1; - else + } else { newdest = &opt->ip6po_dest2; + } break; case IPV6_RTHDRDSTOPTS: newdest = &opt->ip6po_dest1; @@ -3604,9 +3967,11 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, /* turn off the previous option, then set the new option. */ ip6_clearpktopts(opt, optname); *newdest = _MALLOC(destlen, M_IP6OPT, M_NOWAIT); - if (*newdest == NULL) - return (ENOBUFS); + if (*newdest == NULL) { + return ENOBUFS; + } bcopy(dest, *newdest, destlen); + capture_exthdrstat_out = TRUE; break; } @@ -3617,54 +3982,63 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, if (len == 0) { ip6_clearpktopts(opt, IPV6_RTHDR); - break; /* just remove the option */ + break; /* just remove the option */ } /* message length validation */ - if (len < sizeof (struct ip6_rthdr)) - return (EINVAL); + if (len < sizeof(struct ip6_rthdr)) { + return EINVAL; + } rth = (struct ip6_rthdr *)(void *)buf; rthlen = (rth->ip6r_len + 1) << 3; - if (len != rthlen) - return (EINVAL); + if (len != rthlen) { + return EINVAL; + } switch (rth->ip6r_type) { case IPV6_RTHDR_TYPE_0: - if (rth->ip6r_len == 0) /* must contain one addr */ - return (EINVAL); - if (rth->ip6r_len % 2) /* length must be even */ - return (EINVAL); - if (rth->ip6r_len / 2 != rth->ip6r_segleft) - return (EINVAL); + if (rth->ip6r_len == 0) { /* must contain one addr */ + return EINVAL; + } + if (rth->ip6r_len % 2) { /* length must be even */ + return EINVAL; + } + if (rth->ip6r_len / 2 != rth->ip6r_segleft) { + return EINVAL; + } break; default: - return (EINVAL); /* not supported */ + return EINVAL; /* not supported */ } /* turn off the previous option */ ip6_clearpktopts(opt, IPV6_RTHDR); opt->ip6po_rthdr = _MALLOC(rthlen, M_IP6OPT, M_NOWAIT); - if (opt->ip6po_rthdr == NULL) - return (ENOBUFS); + 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)) - return (EINVAL); + if (len != sizeof(int)) { + return EINVAL; + } minmtupolicy = *(int *)(void *)buf; if (minmtupolicy != IP6PO_MINMTU_MCASTONLY && minmtupolicy != IP6PO_MINMTU_DISABLE && minmtupolicy != IP6PO_MINMTU_ALL) { - return (EINVAL); + return EINVAL; } opt->ip6po_minmtu = minmtupolicy; break; case IPV6_DONTFRAG: - if (len != sizeof (int)) - return (EINVAL); + if (len != sizeof(int)) { + return EINVAL; + } if (uproto == IPPROTO_TCP || *(int *)(void *)buf == 0) { /* @@ -3678,22 +4052,31 @@ ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt, break; case IPV6_PREFER_TEMPADDR: - if (len != sizeof (int)) - return (EINVAL); + if (len != sizeof(int)) { + return EINVAL; + } preftemp = *(int *)(void *)buf; if (preftemp != IP6PO_TEMPADDR_SYSTEM && preftemp != IP6PO_TEMPADDR_NOTPREFER && preftemp != IP6PO_TEMPADDR_PREFER) { - return (EINVAL); + return EINVAL; } opt->ip6po_prefer_tempaddr = preftemp; break; default: - return (ENOPROTOOPT); + return ENOPROTOOPT; } /* end of switch */ - return (0); + 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; } /* @@ -3710,8 +4093,9 @@ ip6_mloopback(struct ifnet *srcifp, struct ifnet *origifp, struct mbuf *m, struct ip6_hdr *ip6; struct in6_addr src; - if (lo_ifp == NULL) + if (lo_ifp == NULL) { return; + } /* * Copy the packet header as it's needed for the checksum. @@ -3721,11 +4105,13 @@ ip6_mloopback(struct ifnet *srcifp, struct ifnet *origifp, struct mbuf *m, */ 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)); + copym->m_len < sizeof(struct ip6_hdr))) { + copym = m_pullup(copym, sizeof(struct ip6_hdr)); + } - if (copym == NULL) + if (copym == NULL) { return; + } ip6 = mtod(copym, struct ip6_hdr *); src = ip6->ip6_src; @@ -3736,8 +4122,9 @@ ip6_mloopback(struct ifnet *srcifp, struct ifnet *origifp, struct mbuf *m, in6_clearscope(&ip6->ip6_src); in6_clearscope(&ip6->ip6_dst); - if (copym->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA) + if (copym->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA) { in6_delayed_cksum_offset(copym, 0, optlen, nxt0); + } /* * Stuff the 'real' ifp into the pkthdr, to be used in matching @@ -3755,7 +4142,7 @@ ip6_mloopback(struct ifnet *srcifp, struct ifnet *origifp, struct mbuf *m, struct in6_ifaddr *ia; lck_rw_lock_shared(&in6_ifaddr_rwlock); - for (ia = in6_ifaddrs; ia != NULL; ia = ia->ia_next) { + TAILQ_FOREACH(ia, IN6ADDR_HASH(&src), ia6_hash) { IFA_LOCK_SPIN(&ia->ia_ifa); /* compare against src addr with embedded scope */ if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &src)) { @@ -3767,8 +4154,9 @@ ip6_mloopback(struct ifnet *srcifp, struct ifnet *origifp, struct mbuf *m, } lck_rw_done(&in6_ifaddr_rwlock); } - if (srcifp != NULL) + if (srcifp != NULL) { ip6_setsrcifaddr_info(copym, srcifp->if_index, NULL); + } ip6_setdstifaddr_info(copym, origifp->if_index, NULL); dlil_output(lo_ifp, PF_INET6, copym, NULL, SA(dst), 0, NULL); @@ -3784,24 +4172,24 @@ ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs) struct ip6_hdr *ip6; ip6 = mtod(m, struct ip6_hdr *); - if (m->m_len > sizeof (*ip6)) { - MGETHDR(mh, M_DONTWAIT, MT_HEADER); /* MAC-OK */ + if (m->m_len > sizeof(*ip6)) { + MGETHDR(mh, M_DONTWAIT, MT_HEADER); /* MAC-OK */ if (mh == NULL) { m_freem(m); - return (ENOBUFS); + return ENOBUFS; } M_COPY_PKTHDR(mh, m); - MH_ALIGN(mh, sizeof (*ip6)); + MH_ALIGN(mh, sizeof(*ip6)); m->m_flags &= ~M_PKTHDR; - m->m_len -= sizeof (*ip6); - m->m_data += sizeof (*ip6); + m->m_len -= sizeof(*ip6); + m->m_data += sizeof(*ip6); mh->m_next = m; m = mh; - m->m_len = sizeof (*ip6); - bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof (*ip6)); + m->m_len = sizeof(*ip6); + bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6)); } exthdrs->ip6e_ip6 = m; - return (0); + return 0; } static void @@ -3826,15 +4214,17 @@ ip6_output_checksum(struct ifnet *ifp, uint32_t mtu, struct mbuf *m, } 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.) + * 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) && + ((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 start = sizeof(struct ip6_hdr); uint16_t ulpoff = m->m_pkthdr.csum_data & 0xffff; m->m_pkthdr.csum_flags |= @@ -3876,12 +4266,13 @@ ip6_optlen(struct in6pcb *in6p) { int len; - if (!in6p->in6p_outputopts) - return (0); + if (!in6p->in6p_outputopts) { + return 0; + } len = 0; -#define elen(x) \ - (((struct ip6_ext *)(x)) ? \ +#define elen(x) \ + (((struct ip6_ext *)(x)) ? \ (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0) len += elen(in6p->in6p_outputopts->ip6po_hbh); @@ -3891,6 +4282,63 @@ ip6_optlen(struct in6pcb *in6p) } 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)); +}