]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/netinet6/ip6_output.c
xnu-4903.221.2.tar.gz
[apple/xnu.git] / bsd / netinet6 / ip6_output.c
index c8b578e6cb2909510c5e76affc99c2f1cb4182b3..0720d6809f59992495cfab7ec091fa4aaba186b1 100644 (file)
@@ -1,5 +1,30 @@
-/*     $FreeBSD: src/sys/netinet6/ip6_output.c,v 1.43 2002/10/31 19:45:48 ume Exp $    */
-/*     $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $    */
+/*
+ * Copyright (c) 2000-2017 Apple Inc. All rights reserved.
+ *
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
+ *
+ * This file contains Original Code and/or Modifications of Original Code
+ * as defined in and that are subject to the Apple Public Source License
+ * Version 2.0 (the 'License'). You may not use this file except in
+ * compliance with the License. The rights granted to you under the License
+ * may not be used to create, or enable the creation or redistribution of,
+ * unlawful or unlicensed copies of an Apple operating system, or to
+ * circumvent, violate, or enable the circumvention or violation of, any
+ * terms of an Apple operating system software license agreement.
+ *
+ * Please obtain a copy of the License at
+ * http://www.opensource.apple.com/apsl/ and read it before using this file.
+ *
+ * The Original Code and all software distributed under the License are
+ * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
+ * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
+ * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
+ * Please see the License for the specific language governing rights and
+ * limitations under the License.
+ *
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
+ */
 
 /*
  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
 
 /*
  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
  *
  *     @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  */
  *
  *     @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  */
-
+/*
+ * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
+ * support for mandatory and extensible security protections.  This notice
+ * is included in support of clause 2.2 (b) of the Apple Public License,
+ * Version 2.0.
+ */
 
 #include <sys/param.h>
 #include <sys/malloc.h>
 
 #include <sys/param.h>
 #include <sys/malloc.h>
 #include <sys/kernel.h>
 #include <sys/proc.h>
 #include <sys/kauth.h>
 #include <sys/kernel.h>
 #include <sys/proc.h>
 #include <sys/kauth.h>
+#include <sys/mcache.h>
+#include <sys/sysctl.h>
+#include <kern/zalloc.h>
+#include <libkern/OSByteOrder.h>
+
+#include <pexpert/pexpert.h>
+#include <mach/sdt.h>
 
 #include <net/if.h>
 #include <net/route.h>
 
 #include <net/if.h>
 #include <net/route.h>
+#include <net/dlil.h>
+#include <net/net_api_stats.h>
+#include <net/net_osdep.h>
+#include <net/net_perf.h>
 
 
+#include <netinet/ip.h>
 #include <netinet/in.h>
 #include <netinet/in_var.h>
 #include <netinet/ip_var.h>
 #include <netinet6/in6_var.h>
 #include <netinet/ip6.h>
 #include <netinet/in.h>
 #include <netinet/in_var.h>
 #include <netinet/ip_var.h>
 #include <netinet6/in6_var.h>
 #include <netinet/ip6.h>
+#include <netinet/kpi_ipfilter_var.h>
+#include <netinet/in_tclass.h>
+
+#include <netinet6/ip6protosw.h>
 #include <netinet/icmp6.h>
 #include <netinet6/ip6_var.h>
 #include <netinet/in_pcb.h>
 #include <netinet6/nd6.h>
 #include <netinet/icmp6.h>
 #include <netinet6/ip6_var.h>
 #include <netinet/in_pcb.h>
 #include <netinet6/nd6.h>
-
+#include <netinet6/scope6_var.h>
 #if IPSEC
 #include <netinet6/ipsec.h>
 #if IPSEC
 #include <netinet6/ipsec.h>
-#if INET6
 #include <netinet6/ipsec6.h>
 #include <netinet6/ipsec6.h>
-#endif
 #include <netkey/key.h>
 extern int ipsec_bypass;
 #include <netkey/key.h>
 extern int ipsec_bypass;
-extern lck_mtx_t *sadb_mutex;
-extern lck_mtx_t *nd6_mutex;
 #endif /* IPSEC */
 
 #endif /* IPSEC */
 
-#include <netinet6/ip6_fw.h>
-
-#include <net/net_osdep.h>
-
-#include <netinet/kpi_ipfilter_var.h>
-
-#ifndef __APPLE__
-static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
-#endif
+#if NECP
+#include <net/necp.h>
+#endif /* NECP */
 
 
+#if CONFIG_MACF_NET
+#include <security/mac.h>
+#endif /* CONFIG_MACF_NET */
 
 
-extern u_long  route_generation;
+#if DUMMYNET
+#include <netinet/ip_fw.h>
+#include <netinet/ip_dummynet.h>
+#endif /* DUMMYNET */
 
 
-struct ip6_exthdrs {
-       struct mbuf *ip6e_ip6;
-       struct mbuf *ip6e_hbh;
-       struct mbuf *ip6e_dest1;
-       struct mbuf *ip6e_rthdr;
-       struct mbuf *ip6e_dest2;
-};
+#if PF
+#include <net/pfvar.h>
+#endif /* PF */
 
 
-static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
-                           struct socket *, struct sockopt *sopt);
-static int ip6_setmoptions(int, struct inpcb *, struct mbuf *);
-static int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf **);
+static int sysctl_reset_ip6_output_stats SYSCTL_HANDLER_ARGS;
+static int sysctl_ip6_output_measure_bins SYSCTL_HANDLER_ARGS;
+static int sysctl_ip6_output_getperf SYSCTL_HANDLER_ARGS;
 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
-static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
-                                 struct ip6_frag **);
+static void ip6_out_cksum_stats(int, u_int32_t);
 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
+static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
+    struct ip6_frag **);
+static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
+    struct ifnet *, struct in6_addr *, u_int32_t *, boolean_t *);
+static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, struct socket *,
+    struct sockopt *sopt);
+static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, int);
+static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
+static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
+static void im6o_trace(struct ip6_moptions *, int);
+static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, int,
+    int, int);
 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
 static int ip6_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_exthdrs *exthdrsp, struct ifnet *ifp,
+    uint32_t mtu, boolean_t alwaysfrag, uint32_t unfragpartlen,
+    struct route_in6 *ro_pmtu, int nxt0, uint32_t optlen);
+
+SYSCTL_DECL(_net_inet6_ip6);
+
+static int ip6_output_measure = 0;
+SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, output_perf,
+       CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
+       &ip6_output_measure, 0, sysctl_reset_ip6_output_stats, "I", "Do time measurement");
+
+static uint64_t ip6_output_measure_bins = 0;
+SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, output_perf_bins,
+       CTLTYPE_QUAD | CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_output_measure_bins, 0,
+       sysctl_ip6_output_measure_bins, "I",
+       "bins for chaining performance data histogram");
+
+static net_perf_t net_perf;
+SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, output_perf_data,
+       CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED,
+       0, 0, sysctl_ip6_output_getperf, "S,net_perf",
+       "IP6 output performance data (struct net_perf, net/net_perf.h)");
+
+#define        IM6O_TRACE_HIST_SIZE    32      /* size of trace history */
+
+/* For gdb */
+__private_extern__ unsigned int im6o_trace_hist_size = IM6O_TRACE_HIST_SIZE;
+
+struct ip6_moptions_dbg {
+       struct ip6_moptions     im6o;                   /* ip6_moptions */
+       u_int16_t               im6o_refhold_cnt;       /* # of IM6O_ADDREF */
+       u_int16_t               im6o_refrele_cnt;       /* # of IM6O_REMREF */
+       /*
+        * Alloc and free callers.
+        */
+       ctrace_t                im6o_alloc;
+       ctrace_t                im6o_free;
+       /*
+        * Circular lists of IM6O_ADDREF and IM6O_REMREF callers.
+        */
+       ctrace_t                im6o_refhold[IM6O_TRACE_HIST_SIZE];
+       ctrace_t                im6o_refrele[IM6O_TRACE_HIST_SIZE];
+};
+
+#if DEBUG
+static unsigned int im6o_debug = 1;    /* debugging (enabled) */
+#else
+static unsigned int im6o_debug;                /* debugging (disabled) */
+#endif /* !DEBUG */
 
 
-extern int ip_createmoptions(struct ip_moptions **imop);
-extern int ip_addmembership(struct ip_moptions *imo, struct ip_mreq *mreq);
-extern int ip_dropmembership(struct ip_moptions *imo, struct ip_mreq *mreq);
-extern lck_mtx_t *ip6_mutex;
+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 */
+
+/*
+ * ip6_output() calls ip6_output_list() to do the work
+ */
+int
+ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
+    struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
+    struct ifnet **ifpp, struct ip6_out_args *ip6oa)
+{
+       return ip6_output_list(m0, 0, opt, ro, flags, im6o, ifpp, ip6oa);
+}
 
 /*
 
 /*
- * IP6 output. The packet in mbuf chain m contains a skeletal IP6
+ * IP6 output. Each packet in mbuf chain m contains a skeletal IP6
  * header (with pri, len, nxt, hlim, src, dst).
  * This function may modify ver and hlim only.
  * The mbuf chain containing the packet will be freed.
  * The mbuf opt, if present, will not be freed.
  *
  * header (with pri, len, nxt, hlim, src, dst).
  * This function may modify ver and hlim only.
  * The mbuf chain containing the packet will be freed.
  * The mbuf opt, if present, will not be freed.
  *
- * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
- * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
+ * If ro is non-NULL and has valid ro->ro_rt, route lookup would be
+ * skipped and ro->ro_rt would be used.  Otherwise the result of route
+ * lookup is stored in ro->ro_rt.
+ *
+ * type of "mtu": rt_rmx.rmx_mtu is u_int32_t, ifnet.ifr_mtu is int, and
+ * nd_ifinfo.linkmtu is u_int32_t.  so we use u_int32_t to hold largest one,
  * which is rt_rmx.rmx_mtu.
  */
 int
  * which is rt_rmx.rmx_mtu.
  */
 int
-ip6_output(
-       struct mbuf *m0,
-       struct ip6_pktopts *opt,
-       struct route_in6 *ro,
-       int flags,
-       struct ip6_moptions *im6o,
-       struct ifnet **ifpp,    /* XXX: just for statistics */
-       int    locked)          
+ip6_output_list(struct mbuf *m0, int packetchain, struct ip6_pktopts *opt,
+    struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
+    struct ifnet **ifpp, struct ip6_out_args *ip6oa)
 {
 {
-       struct ip6_hdr *ip6, *mhip6;
-       struct ifnet *ifp, *origifp;
-       struct mbuf *m = m0;
-       int hlen, tlen, len, off;
-       struct route_in6 ip6route;
-       struct sockaddr_in6 *dst;
+       struct ip6_hdr *ip6;
+       u_char *nexthdrp;
+       struct ifnet *ifp = NULL, *origifp = NULL;      /* refcnt'd */
+       struct ifnet **ifpp_save = ifpp;
+       struct mbuf *m, *mprev;
+       struct mbuf *sendchain = NULL, *sendchain_last = NULL;
+       struct mbuf *inputchain = NULL;
+       int nxt0 = 0;
+       struct route_in6 *ro_pmtu = NULL;
+       struct rtentry *rt = NULL;
+       struct sockaddr_in6 *dst = NULL, src_sa, dst_sa;
        int error = 0;
        int error = 0;
-       struct in6_ifaddr *ia = NULL;
-       u_long mtu;
+       struct in6_ifaddr *ia = NULL, *src_ia = NULL;
+       u_int32_t mtu = 0;
+       boolean_t alwaysfrag = FALSE;
        u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
        u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
-       struct ip6_exthdrs exthdrs;
+       struct ip6_rthdr *rh;
        struct in6_addr finaldst;
        struct in6_addr finaldst;
-       struct route_in6 *ro_pmtu = NULL;
-       int hdrsplit = 0;
-       int needipsec = 0;
        ipfilter_t inject_filter_ref;
        ipfilter_t inject_filter_ref;
-       
+       struct ipf_pktopts *ippo = NULL;
+       struct flowadv *adv = NULL;
+       uint32_t pktcnt = 0;
+       uint32_t packets_processed = 0;
+       struct timeval start_tv;
+#if DUMMYNET
+       struct m_tag *tag;
+       struct ip6_out_args saved_ip6oa;
+       struct sockaddr_in6 dst_buf;
+#endif /* DUMMYNET */
 #if IPSEC
 #if IPSEC
-       int needipsectun = 0;
        struct socket *so = NULL;
        struct secpolicy *sp = NULL;
        struct socket *so = NULL;
        struct secpolicy *sp = NULL;
+       struct route_in6 *ipsec_saved_route = NULL;
+       boolean_t needipsectun = FALSE;
+#endif /* IPSEC */
+#if NECP
+       necp_kernel_policy_result necp_result = 0;
+       necp_kernel_policy_result_parameter necp_result_parameter;
+       necp_kernel_policy_id necp_matched_policy_id = 0;
+#endif /* NECP */
+       struct {
+               struct ipf_pktopts ipf_pktopts;
+               struct ip6_exthdrs exthdrs;
+               struct route_in6 ip6route;
+#if IPSEC
+               struct ipsec_output_state ipsec_state;
+#endif /* IPSEC */
+#if NECP
+               struct route_in6 necp_route;
+#endif /* NECP */
+#if DUMMYNET
+               struct route_in6 saved_route;
+               struct route_in6 saved_ro_pmtu;
+               struct ip_fw_args args;
+#endif /* DUMMYNET */
+       } ip6obz;
+#define        ipf_pktopts     ip6obz.ipf_pktopts
+#define        exthdrs         ip6obz.exthdrs
+#define        ip6route        ip6obz.ip6route
+#define        ipsec_state     ip6obz.ipsec_state
+#define        necp_route      ip6obz.necp_route
+#define        saved_route     ip6obz.saved_route
+#define        saved_ro_pmtu   ip6obz.saved_ro_pmtu
+#define        args            ip6obz.args
+       union {
+               struct {
+                       boolean_t select_srcif : 1;
+                       boolean_t hdrsplit : 1;
+                       boolean_t route_selected : 1;
+                       boolean_t dontfrag : 1;
+#if IPSEC
+                       boolean_t needipsec : 1;
+                       boolean_t noipsec : 1;
+#endif /* IPSEC */
+               };
+               uint32_t raw;
+       } ip6obf = { .raw = 0 };
 
 
-       if (!locked)
-               lck_mtx_lock(ip6_mutex);
-       /* for AH processing. stupid to have "socket" variable in IP layer... */
-       if (ipsec_bypass == 0)
-       {
+       if (ip6_output_measure)
+               net_perf_start_time(&net_perf, &start_tv);
+
+       VERIFY(m0->m_flags & M_PKTHDR);
+
+       /* zero out {saved_route, saved_ro_pmtu, ip6route, exthdrs, args} */
+       bzero(&ip6obz, sizeof (ip6obz));
+
+#if DUMMYNET
+       if (SLIST_EMPTY(&m0->m_pkthdr.tags))
+               goto tags_done;
+
+       /* Grab info from mtags prepended to the chain */
+       if ((tag = m_tag_locate(m0, KERNEL_MODULE_TAG_ID,
+           KERNEL_TAG_TYPE_DUMMYNET, NULL)) != NULL) {
+               struct dn_pkt_tag       *dn_tag;
+
+               /*
+                * ip6_output_list() cannot handle chains of packets reinjected
+                * by dummynet. The same restriction applies to
+                * ip_output_list().
+                */
+               VERIFY(0 == packetchain);
+
+               dn_tag = (struct dn_pkt_tag *)(tag+1);
+               args.fwa_pf_rule = dn_tag->dn_pf_rule;
+
+               bcopy(&dn_tag->dn_dst6, &dst_buf, sizeof (dst_buf));
+               dst = &dst_buf;
+               ifp = dn_tag->dn_ifp;
+               if (ifp != NULL)
+                       ifnet_reference(ifp);
+               flags = dn_tag->dn_flags;
+               if (dn_tag->dn_flags & IPV6_OUTARGS) {
+                       saved_ip6oa = dn_tag->dn_ip6oa;
+                       ip6oa = &saved_ip6oa;
+               }
+
+               saved_route = dn_tag->dn_ro6;
+               ro = &saved_route;
+               saved_ro_pmtu = dn_tag->dn_ro6_pmtu;
+               ro_pmtu = &saved_ro_pmtu;
+               origifp = dn_tag->dn_origifp;
+               if (origifp != NULL)
+                       ifnet_reference(origifp);
+               mtu = dn_tag->dn_mtu;
+               alwaysfrag = (dn_tag->dn_alwaysfrag != 0);
+               unfragpartlen = dn_tag->dn_unfragpartlen;
+
+               bcopy(&dn_tag->dn_exthdrs, &exthdrs, sizeof (exthdrs));
+
+               m_tag_delete(m0, tag);
+       }
+
+tags_done:
+#endif /* DUMMYNET */
+
+       m = m0;
+
+#if IPSEC
+       if (ipsec_bypass == 0) {
                so = ipsec_getsocket(m);
                so = ipsec_getsocket(m);
-               (void)ipsec_setsocket(m, NULL);
+               if (so != NULL) {
+                       (void) ipsec_setsocket(m, NULL);
+               }
+               /* If packet is bound to an interface, check bound policies */
+               if ((flags & IPV6_OUTARGS) &&
+                       (ip6oa->ip6oa_flags & IP6OAF_BOUND_IF) &&
+                       ip6oa->ip6oa_boundif != IFSCOPE_NONE) {
+                       /* ip6obf.noipsec is a bitfield, use temp integer */
+                       int noipsec = 0;
+
+                       if (ipsec6_getpolicybyinterface(m, IPSEC_DIR_OUTBOUND,
+                               flags, ip6oa, &noipsec, &sp) != 0)
+                               goto bad;
+
+                       ip6obf.noipsec = (noipsec != 0);
+               }
        }
 #endif /* IPSEC */
 
        }
 #endif /* IPSEC */
 
+       ippo = &ipf_pktopts;
+
+       if (flags & IPV6_OUTARGS) {
+               /*
+                * In the forwarding case, only the ifscope value is used,
+                * as source interface selection doesn't take place.
+                */
+               if ((ip6obf.select_srcif = (!(flags & (IPV6_FORWARDING |
+                   IPV6_UNSPECSRC | IPV6_FLAG_NOSRCIFSEL)) &&
+                   (ip6oa->ip6oa_flags & IP6OAF_SELECT_SRCIF))))
+                       ipf_pktopts.ippo_flags |= IPPOF_SELECT_SRCIF;
+
+               if ((ip6oa->ip6oa_flags & IP6OAF_BOUND_IF) &&
+                   ip6oa->ip6oa_boundif != IFSCOPE_NONE) {
+                       ipf_pktopts.ippo_flags |= (IPPOF_BOUND_IF |
+                           (ip6oa->ip6oa_boundif << IPPOF_SHIFT_IFSCOPE));
+               }
+
+               if (ip6oa->ip6oa_flags & IP6OAF_BOUND_SRCADDR)
+                       ipf_pktopts.ippo_flags |= IPPOF_BOUND_SRCADDR;
+       } else {
+               ip6obf.select_srcif = FALSE;
+               if (flags & IPV6_OUTARGS) {
+                       ip6oa->ip6oa_boundif = IFSCOPE_NONE;
+                       ip6oa->ip6oa_flags &= ~(IP6OAF_SELECT_SRCIF |
+                           IP6OAF_BOUND_IF | IP6OAF_BOUND_SRCADDR);
+               }
+       }
+
+       if (flags & IPV6_OUTARGS) {
+               if (ip6oa->ip6oa_flags & IP6OAF_NO_CELLULAR)
+                       ipf_pktopts.ippo_flags |= IPPOF_NO_IFT_CELLULAR;
+               if (ip6oa->ip6oa_flags & IP6OAF_NO_EXPENSIVE)
+                       ipf_pktopts.ippo_flags |= IPPOF_NO_IFF_EXPENSIVE;
+               adv = &ip6oa->ip6oa_flowadv;
+               adv->code = FADV_SUCCESS;
+               ip6oa->ip6oa_retflags = 0;
+       }
+
+       /*
+        * Clear out ifpp to be filled in after determining route. ifpp_save is
+        * used to keep old value to release reference properly and dtrace
+        * ipsec tunnel traffic properly.
+        */
+       if (ifpp != NULL && *ifpp != NULL)
+               *ifpp = NULL;
+
+#if DUMMYNET
+       if (args.fwa_pf_rule) {
+               ip6 = mtod(m, struct ip6_hdr *);
+               VERIFY(ro != NULL);     /* ro == saved_route */
+               goto check_with_pf;
+       }
+#endif /* DUMMYNET */
+
+#if NECP
+       /*
+        * Since all packets are assumed to come from same socket, necp lookup
+        * only needs to happen once per function entry.
+        */
+       necp_matched_policy_id = necp_ip6_output_find_policy_match(m, flags,
+           (flags & IPV6_OUTARGS) ? ip6oa : NULL, &necp_result,
+           &necp_result_parameter);
+#endif /* NECP */
+
+       /*
+        * If a chain was passed in, prepare for ther first iteration. For all
+        * other iterations, this work will be done at evaluateloop: label.
+        */
+       if (packetchain) {
+               /*
+                * Remove m from the chain during processing to avoid
+                * accidental frees on entire list.
+                */
+               inputchain = m->m_nextpkt;
+               m->m_nextpkt = NULL;
+       }
+
+loopit:
+       packets_processed++;
+       m->m_pkthdr.pkt_flags &= ~(PKTF_LOOP|PKTF_IFAINFO);
        ip6 = mtod(m, struct ip6_hdr *);
        ip6 = mtod(m, struct ip6_hdr *);
-       inject_filter_ref = ipf_get_inject_filter(m);
+       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) {                                                       \
+#define        MAKE_EXTHDR(hp, mp) do {                                        \
+       if (hp != NULL) {                                               \
                struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
                struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
-               error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
-                                      ((eh)->ip6e_len + 1) << 3);      \
+               error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
+                   ((eh)->ip6e_len + 1) << 3);                         \
                if (error)                                              \
                        goto freehdrs;                                  \
        }                                                               \
                if (error)                                              \
                        goto freehdrs;                                  \
        }                                                               \
-    } while (0)
-       
-       bzero(&exthdrs, sizeof(exthdrs));
-       
-       if (opt) {
+} while (0)
+
+       if (opt != NULL) {
                /* Hop-by-Hop options header */
                MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
                /* Destination options header(1st part) */
                /* Hop-by-Hop options header */
                MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
                /* Destination options header(1st part) */
-               MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
+               if (opt->ip6po_rthdr) {
+                       /*
+                        * Destination options header(1st part)
+                        * This only makes sense with a routing header.
+                        * See Section 9.2 of RFC 3542.
+                        * Disabling this part just for MIP6 convenience is
+                        * a bad idea.  We need to think carefully about a
+                        * way to make the advanced API coexist with MIP6
+                        * options, which might automatically be inserted in
+                        * the kernel.
+                        */
+                       MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
+               }
                /* Routing header */
                MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
                /* Destination options header(2nd part) */
                MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
        }
 
                /* Routing header */
                MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
                /* Destination options header(2nd part) */
                MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
        }
 
+#undef MAKE_EXTHDR
+
+#if NECP
+       if (necp_matched_policy_id) {
+               necp_mark_packet_from_ip(m, necp_matched_policy_id);
+
+               switch (necp_result) {
+               case NECP_KERNEL_POLICY_RESULT_PASS:
+                       goto skip_ipsec;
+               case NECP_KERNEL_POLICY_RESULT_DROP:
+               case NECP_KERNEL_POLICY_RESULT_SOCKET_DIVERT:
+                       /*
+                        * Flow divert packets should be blocked at the IP
+                        * layer.
+                        */
+                       error = EHOSTUNREACH;
+                       ip6stat.ip6s_necp_policy_drop++;
+                       goto freehdrs;
+               case NECP_KERNEL_POLICY_RESULT_IP_TUNNEL: {
+                       /*
+                        * Verify that the packet is being routed to the tunnel
+                        */
+                       struct ifnet *policy_ifp =
+                           necp_get_ifnet_from_result_parameter(
+                               &necp_result_parameter);
+
+                       if (policy_ifp == ifp) {
+                               goto skip_ipsec;
+                       } else {
+                               if (necp_packet_can_rebind_to_ifnet(m,
+                                   policy_ifp, (struct route *)&necp_route,
+                                   AF_INET6)) {
+                                       /*
+                                        * Set scoped index to the tunnel
+                                        * interface, since it is compatible
+                                        * with the packet. This will only work
+                                        * for callers who pass IPV6_OUTARGS,
+                                        * but that covers all of the clients
+                                        * we care about today.
+                                        */
+                                       if (flags & IPV6_OUTARGS) {
+                                               ip6oa->ip6oa_boundif =
+                                                   policy_ifp->if_index;
+                                               ip6oa->ip6oa_flags |=
+                                                   IP6OAF_BOUND_IF;
+                                       }
+                                       if (opt != NULL
+                                           && opt->ip6po_pktinfo != NULL) {
+                                               opt->ip6po_pktinfo->
+                                                   ipi6_ifindex =
+                                                       policy_ifp->if_index;
+                                       }
+                                       ro = &necp_route;
+                                       goto skip_ipsec;
+                               } else {
+                                       error = ENETUNREACH;
+                                       ip6stat.ip6s_necp_policy_drop++;
+                                       goto freehdrs;
+                               }
+                       }
+               }
+               default:
+                       break;
+               }
+       }
+#endif /* NECP */
+
 #if IPSEC
 #if IPSEC
-       if (ipsec_bypass != 0)
+       if (ipsec_bypass != 0 || ip6obf.noipsec)
                goto skip_ipsec;
                goto skip_ipsec;
-       
-       lck_mtx_lock(sadb_mutex);
-       /* get a security policy for this packet */
-       if (so == NULL)
-               sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
-       else
-               sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
 
        if (sp == NULL) {
 
        if (sp == NULL) {
-               ipsec6stat.out_inval++;
-               lck_mtx_unlock(sadb_mutex);
-               goto freehdrs;
+               /* get a security policy for this packet */
+               if (so != NULL) {
+                       sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND,
+                               so, &error);
+               } else {
+                       sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
+                               0, &error);
+               }
+               if (sp == NULL) {
+                       IPSEC_STAT_INCREMENT(ipsec6stat.out_inval);
+                       goto freehdrs;
+               }
        }
 
        error = 0;
        }
 
        error = 0;
@@ -240,35 +646,38 @@ ip6_output(
        /* check policy */
        switch (sp->policy) {
        case IPSEC_POLICY_DISCARD:
        /* check policy */
        switch (sp->policy) {
        case IPSEC_POLICY_DISCARD:
+       case IPSEC_POLICY_GENERATE:
                /*
                 * This packet is just discarded.
                 */
                /*
                 * This packet is just discarded.
                 */
-               ipsec6stat.out_polvio++;
-               lck_mtx_unlock(sadb_mutex);
+               IPSEC_STAT_INCREMENT(ipsec6stat.out_polvio);
                goto freehdrs;
 
        case IPSEC_POLICY_BYPASS:
        case IPSEC_POLICY_NONE:
                /* no need to do IPsec. */
                goto freehdrs;
 
        case IPSEC_POLICY_BYPASS:
        case IPSEC_POLICY_NONE:
                /* no need to do IPsec. */
-               needipsec = 0;
+               ip6obf.needipsec = FALSE;
                break;
                break;
-       
+
        case IPSEC_POLICY_IPSEC:
                if (sp->req == NULL) {
                        /* acquire a policy */
                        error = key_spdacquire(sp);
        case IPSEC_POLICY_IPSEC:
                if (sp->req == NULL) {
                        /* acquire a policy */
                        error = key_spdacquire(sp);
-                       lck_mtx_unlock(sadb_mutex);
                        goto freehdrs;
                }
                        goto freehdrs;
                }
-               needipsec = 1;
+               if (sp->ipsec_if) {
+                       goto skip_ipsec;
+               } else {
+                       ip6obf.needipsec = TRUE;
+               }
                break;
 
        case IPSEC_POLICY_ENTRUST:
        default:
                break;
 
        case IPSEC_POLICY_ENTRUST:
        default:
-               printf("ip6_output: Invalid policy found. %d\n", sp->policy);
+               printf("%s: Invalid policy found: %d\n", __func__, sp->policy);
+               break;
        }
        }
-       lck_mtx_unlock(sadb_mutex);
-       skip_ipsec:
+skip_ipsec:
 #endif /* IPSEC */
 
        /*
 #endif /* IPSEC */
 
        /*
@@ -276,24 +685,33 @@ ip6_output(
         * Keep the length of the unfragmentable part for fragmentation.
         */
        optlen = 0;
         * Keep the length of the unfragmentable part for fragmentation.
         */
        optlen = 0;
-       if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
-       if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
-       if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
-       unfragpartlen = optlen + sizeof(struct ip6_hdr);
+       if (exthdrs.ip6e_hbh != NULL)
+               optlen += exthdrs.ip6e_hbh->m_len;
+       if (exthdrs.ip6e_dest1 != NULL)
+               optlen += exthdrs.ip6e_dest1->m_len;
+       if (exthdrs.ip6e_rthdr != NULL)
+               optlen += exthdrs.ip6e_rthdr->m_len;
+       unfragpartlen = optlen + sizeof (struct ip6_hdr);
+
        /* NOTE: we don't add AH/ESP length here. do that later. */
        /* NOTE: we don't add AH/ESP length here. do that later. */
-       if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
+       if (exthdrs.ip6e_dest2 != NULL)
+               optlen += exthdrs.ip6e_dest2->m_len;
 
        /*
         * If we need IPsec, or there is at least one extension header,
         * separate IP6 header from the payload.
         */
 
        /*
         * If we need IPsec, or there is at least one extension header,
         * separate IP6 header from the payload.
         */
-       if ((needipsec || optlen) && !hdrsplit) {
+       if ((
+#if IPSEC
+           ip6obf.needipsec ||
+#endif /* IPSEC */
+           optlen) && !ip6obf.hdrsplit) {
                if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
                        m = NULL;
                        goto freehdrs;
                }
                m = exthdrs.ip6e_ip6;
                if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
                        m = NULL;
                        goto freehdrs;
                }
                m = exthdrs.ip6e_ip6;
-               hdrsplit++;
+               ip6obf.hdrsplit = TRUE;
        }
 
        /* adjust pointer */
        }
 
        /* adjust pointer */
@@ -301,26 +719,26 @@ ip6_output(
 
        /* adjust mbuf packet header length */
        m->m_pkthdr.len += optlen;
 
        /* adjust mbuf packet header length */
        m->m_pkthdr.len += optlen;
-       plen = m->m_pkthdr.len - sizeof(*ip6);
+       plen = m->m_pkthdr.len - sizeof (*ip6);
 
        /* If this is a jumbo payload, insert a jumbo payload option. */
        if (plen > IPV6_MAXPACKET) {
 
        /* If this is a jumbo payload, insert a jumbo payload option. */
        if (plen > IPV6_MAXPACKET) {
-               if (!hdrsplit) {
+               if (!ip6obf.hdrsplit) {
                        if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
                                m = NULL;
                                goto freehdrs;
                        }
                        m = exthdrs.ip6e_ip6;
                        if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
                                m = NULL;
                                goto freehdrs;
                        }
                        m = exthdrs.ip6e_ip6;
-                       hdrsplit++;
+                       ip6obf.hdrsplit = TRUE;
                }
                /* adjust pointer */
                ip6 = mtod(m, struct ip6_hdr *);
                if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
                        goto freehdrs;
                ip6->ip6_plen = 0;
                }
                /* adjust pointer */
                ip6 = mtod(m, struct ip6_hdr *);
                if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
                        goto freehdrs;
                ip6->ip6_plen = 0;
-       } else
+       } else {
                ip6->ip6_plen = htons(plen);
                ip6->ip6_plen = htons(plen);
-
+       }
        /*
         * Concatenate headers and fill in next header fields.
         * Here we have, on "m"
        /*
         * Concatenate headers and fill in next header fields.
         * Here we have, on "m"
@@ -331,110 +749,122 @@ ip6_output(
         * during the header composing process, "m" points to IPv6 header.
         * "mprev" points to an extension header prior to esp.
         */
         * during the header composing process, "m" points to IPv6 header.
         * "mprev" points to an extension header prior to esp.
         */
-       {
-               u_char *nexthdrp = &ip6->ip6_nxt;
-               struct mbuf *mprev = m;
+       nexthdrp = &ip6->ip6_nxt;
+       mprev = m;
 
 
-               /*
-                * we treat dest2 specially.  this makes IPsec processing
-                * much easier.  the goal here is to make mprev point the
-                * mbuf prior to dest2.
-                *
-                * result: IPv6 dest2 payload
-                * m and mprev will point to IPv6 header.
-                */
-               if (exthdrs.ip6e_dest2) {
-                       if (!hdrsplit)
-                               panic("assumption failed: hdr not split");
-                       exthdrs.ip6e_dest2->m_next = m->m_next;
-                       m->m_next = exthdrs.ip6e_dest2;
-                       *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
-                       ip6->ip6_nxt = IPPROTO_DSTOPTS;
-               }
-
-#define MAKE_CHAIN(m, mp, p, i)\
-    do {\
-       if (m) {\
-               if (!hdrsplit) \
-                       panic("assumption failed: hdr not split"); \
-               *mtod((m), u_char *) = *(p);\
-               *(p) = (i);\
-               p = mtod((m), u_char *);\
-               (m)->m_next = (mp)->m_next;\
-               (mp)->m_next = (m);\
-               (mp) = (m);\
-       }\
-    } while (0)
-               /*
-                * result: IPv6 hbh dest1 rthdr dest2 payload
-                * m will point to IPv6 header.  mprev will point to the
-                * extension header prior to dest2 (rthdr in the above case).
-                */
-               MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
-                          nexthdrp, IPPROTO_HOPOPTS);
-               MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
-                          nexthdrp, IPPROTO_DSTOPTS);
-               MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
-                          nexthdrp, IPPROTO_ROUTING);
-
-               if (!TAILQ_EMPTY(&ipv6_filters)) {
-                       struct ipfilter *filter;
-                       int seen = (inject_filter_ref == 0);
-                       int     fixscope = 0;
-                       struct ipf_pktopts *ippo = 0, ipf_pktopts;
-                                               
-                       if (im6o != NULL && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
-                               ippo = &ipf_pktopts;
-                               ippo->ippo_flags = IPPOF_MCAST_OPTS;
-                               ippo->ippo_mcast_ifnet = im6o->im6o_multicast_ifp;
-                               ippo->ippo_mcast_ttl = im6o->im6o_multicast_hlim;
-                               ippo->ippo_mcast_loop = im6o->im6o_multicast_loop;
-                       }
+       /*
+        * we treat dest2 specially.  this makes IPsec processing
+        * much easier.  the goal here is to make mprev point the
+        * mbuf prior to dest2.
+        *
+        * result: IPv6 dest2 payload
+        * m and mprev will point to IPv6 header.
+        */
+       if (exthdrs.ip6e_dest2 != NULL) {
+               if (!ip6obf.hdrsplit) {
+                       panic("assumption failed: hdr not split");
+                       /* NOTREACHED */
+               }
+               exthdrs.ip6e_dest2->m_next = m->m_next;
+               m->m_next = exthdrs.ip6e_dest2;
+               *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
+               ip6->ip6_nxt = IPPROTO_DSTOPTS;
+       }
 
 
-                       /* Hack: embed the scope_id in the destination */
-                       if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
-                               (ip6->ip6_dst.s6_addr16[1] == 0) && (ro != NULL)) {
-                               fixscope = 1;
-                               ip6->ip6_dst.s6_addr16[1] = htons(ro->ro_dst.sin6_scope_id);
-                       }
-                       {
-                               lck_mtx_unlock(ip6_mutex);
-                               ipf_ref();
-                               TAILQ_FOREACH(filter, &ipv6_filters, ipf_link) {
-                                       /*
-                                        * No need to proccess packet twice if we've 
-                                        * already seen it
-                                        */
-                                       if (seen == 0) {
-                                               if ((struct ipfilter *)inject_filter_ref == filter)
-                                                       seen = 1;
-                                       } else if (filter->ipf_filter.ipf_output) {
-                                               errno_t result;
-                                               
-                                               result = filter->ipf_filter.ipf_output(filter->ipf_filter.cookie, (mbuf_t*)&m, ippo);
-                                               if (result == EJUSTRETURN) {
-                                                       ipf_unref();
-                                                       locked = 1; /* Don't want to take lock to unlock it right away */
-                                                       goto done;
-                                               }
-                                               if (result != 0) {
-                                                       ipf_unref();
-                                                       locked = 1; /* Don't want to take lock to unlock it right away */
-                                                       goto bad;
-                                               }
-                                       }
+#define        MAKE_CHAIN(m, mp, p, i) do {                                    \
+       if (m != NULL) {                                                \
+               if (!ip6obf.hdrsplit) {                                 \
+                       panic("assumption failed: hdr not split");      \
+                       /* NOTREACHED */                                \
+               }                                                       \
+               *mtod((m), u_char *) = *(p);                            \
+               *(p) = (i);                                             \
+               p = mtod((m), u_char *);                                \
+               (m)->m_next = (mp)->m_next;                             \
+               (mp)->m_next = (m);                                     \
+               (mp) = (m);                                             \
+       }                                                               \
+} while (0)
+       /*
+        * result: IPv6 hbh dest1 rthdr dest2 payload
+        * m will point to IPv6 header.  mprev will point to the
+        * extension header prior to dest2 (rthdr in the above case).
+        */
+       MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
+       MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, IPPROTO_DSTOPTS);
+       MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, IPPROTO_ROUTING);
+
+       /* It is no longer safe to free the pointers in exthdrs. */
+       exthdrs.merged = TRUE;
+
+#undef MAKE_CHAIN
+
+#if IPSEC
+       if (ip6obf.needipsec && (m->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA))
+               in6_delayed_cksum_offset(m, 0, optlen, nxt0);
+#endif /* IPSEC */
+
+       if (!TAILQ_EMPTY(&ipv6_filters) &&
+           !((flags & IPV6_OUTARGS) &&
+           (ip6oa->ip6oa_flags & IP6OAF_INTCOPROC_ALLOWED))) {
+               struct ipfilter *filter;
+               int seen = (inject_filter_ref == NULL);
+               int fixscope = 0;
+
+               if (im6o != NULL && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
+                       ippo->ippo_flags |= IPPOF_MCAST_OPTS;
+                       IM6O_LOCK(im6o);
+                       ippo->ippo_mcast_ifnet = im6o->im6o_multicast_ifp;
+                       ippo->ippo_mcast_ttl = im6o->im6o_multicast_hlim;
+                       ippo->ippo_mcast_loop = im6o->im6o_multicast_loop;
+                       IM6O_UNLOCK(im6o);
+               }
+
+               /* Hack: embed the scope_id in the destination */
+               if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
+                   (ip6->ip6_dst.s6_addr16[1] == 0) && (ro != NULL)) {
+                       fixscope = 1;
+                       ip6->ip6_dst.s6_addr16[1] =
+                           htons(ro->ro_dst.sin6_scope_id);
+               }
+
+               ipf_ref();
+               TAILQ_FOREACH(filter, &ipv6_filters, ipf_link) {
+                       /*
+                        * Don't process packet twice if we've already seen it.
+                        */
+                       if (seen == 0) {
+                               if ((struct ipfilter *)inject_filter_ref ==
+                                   filter)
+                                       seen = 1;
+                       } else if (filter->ipf_filter.ipf_output != NULL) {
+                               errno_t result;
+
+                               result = filter->ipf_filter.ipf_output(
+                                   filter->ipf_filter.cookie,
+                                   (mbuf_t *)&m, ippo);
+                               if (result == EJUSTRETURN) {
+                                       ipf_unref();
+                                       m = NULL;
+                                       goto evaluateloop;
+                               }
+                               if (result != 0) {
+                                       ipf_unref();
+                                       goto bad;
                                }
                                }
-                               ipf_unref();
-                               lck_mtx_lock(ip6_mutex);
                        }
                        }
-                       /* Hack: cleanup embedded scope_id if we put it there */
-                       if (fixscope)
-                               ip6->ip6_dst.s6_addr16[1] = 0;
                }
                }
+               ipf_unref();
+
+               ip6 = mtod(m, struct ip6_hdr *);
+               /* Hack: cleanup embedded scope_id if we put it there */
+               if (fixscope)
+                       ip6->ip6_dst.s6_addr16[1] = 0;
+       }
 
 #if IPSEC
 
 #if IPSEC
-               if (!needipsec)
-                       goto skip_ipsec2;
+       if (ip6obf.needipsec) {
+               int segleft_org;
 
                /*
                 * pointers after IPsec headers are not valid any more.
 
                /*
                 * pointers after IPsec headers are not valid any more.
@@ -443,26 +873,19 @@ ip6_output(
                 */
                exthdrs.ip6e_dest2 = NULL;
 
                 */
                exthdrs.ip6e_dest2 = NULL;
 
-           {
-               struct ip6_rthdr *rh = NULL;
-               int segleft_org = 0;
-               struct ipsec_output_state state;
-
-               if (exthdrs.ip6e_rthdr) {
+               if (exthdrs.ip6e_rthdr != NULL) {
                        rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
                        segleft_org = rh->ip6r_segleft;
                        rh->ip6r_segleft = 0;
                        rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
                        segleft_org = rh->ip6r_segleft;
                        rh->ip6r_segleft = 0;
+               } else {
+                       rh = NULL;
+                       segleft_org = 0;
                }
 
                }
 
-               bzero(&state, sizeof(state));
-               state.m = m;
-               lck_mtx_unlock(ip6_mutex);
-               lck_mtx_lock(sadb_mutex);
-               error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
-                       &needipsectun);
-               lck_mtx_unlock(sadb_mutex);
-               lck_mtx_lock(ip6_mutex);
-               m = state.m;
+               ipsec_state.m = m;
+               error = ipsec6_output_trans(&ipsec_state, nexthdrp, mprev,
+                   sp, flags, &needipsectun);
+               m = ipsec_state.m;
                if (error) {
                        /* mbuf is already reclaimed in ipsec6_output_trans. */
                        m = NULL;
                if (error) {
                        /* mbuf is already reclaimed in ipsec6_output_trans. */
                        m = NULL;
@@ -474,8 +897,9 @@ ip6_output(
                        case ENOMEM:
                                break;
                        default:
                        case ENOMEM:
                                break;
                        default:
-                               printf("ip6_output (ipsec): error code %d\n", error);
-                               /* fall through */
+                               printf("ip6_output (ipsec): error code %d\n",
+                                   error);
+                               /* FALLTHRU */
                        case ENOENT:
                                /* don't show these error codes to the user */
                                error = 0;
                        case ENOENT:
                                /* don't show these error codes to the user */
                                error = 0;
@@ -483,45 +907,63 @@ ip6_output(
                        }
                        goto bad;
                }
                        }
                        goto bad;
                }
-               if (exthdrs.ip6e_rthdr) {
+               if (exthdrs.ip6e_rthdr != NULL) {
                        /* ah6_output doesn't modify mbuf chain */
                        rh->ip6r_segleft = segleft_org;
                }
                        /* ah6_output doesn't modify mbuf chain */
                        rh->ip6r_segleft = segleft_org;
                }
-           }
-skip_ipsec2:;
-#endif
        }
        }
+#endif /* IPSEC */
 
        /*
 
        /*
-        * If there is a routing header, replace destination address field
+        * If there is a routing header, replace the destination address field
         * with the first hop of the routing header.
         */
         * with the first hop of the routing header.
         */
-       if (exthdrs.ip6e_rthdr) {
-               struct ip6_rthdr *rh =
-                       (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
-                                                 struct ip6_rthdr *));
+       if (exthdrs.ip6e_rthdr != NULL) {
                struct ip6_rthdr0 *rh0;
                struct ip6_rthdr0 *rh0;
+               struct in6_addr *addr;
+               struct sockaddr_in6 sa;
 
 
-               finaldst = ip6->ip6_dst;
+               rh = (struct ip6_rthdr *)
+                   (mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *));
                switch (rh->ip6r_type) {
                case IPV6_RTHDR_TYPE_0:
                switch (rh->ip6r_type) {
                case IPV6_RTHDR_TYPE_0:
-                        rh0 = (struct ip6_rthdr0 *)rh;
-                        ip6->ip6_dst = rh0->ip6r0_addr[0];
-                        bcopy((caddr_t)&rh0->ip6r0_addr[1],
-                              (caddr_t)&rh0->ip6r0_addr[0],
-                              sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
-                                );
-                        rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
-                        break;
+                       rh0 = (struct ip6_rthdr0 *)rh;
+                       addr = (struct in6_addr *)(void *)(rh0 + 1);
+
+                       /*
+                        * construct a sockaddr_in6 form of
+                        * the first hop.
+                        *
+                        * XXX: we may not have enough
+                        * information about its scope zone;
+                        * there is no standard API to pass
+                        * the information from the
+                        * application.
+                        */
+                       bzero(&sa, sizeof (sa));
+                       sa.sin6_family = AF_INET6;
+                       sa.sin6_len = sizeof (sa);
+                       sa.sin6_addr = addr[0];
+                       if ((error = sa6_embedscope(&sa,
+                           ip6_use_defzone)) != 0) {
+                               goto bad;
+                       }
+                       ip6->ip6_dst = sa.sin6_addr;
+                       bcopy(&addr[1], &addr[0], sizeof (struct in6_addr) *
+                           (rh0->ip6r0_segleft - 1));
+                       addr[rh0->ip6r0_segleft - 1] = finaldst;
+                       /* XXX */
+                       in6_clearscope(addr + rh0->ip6r0_segleft - 1);
+                       break;
                default:        /* is it possible? */
                default:        /* is it possible? */
-                        error = EINVAL;
-                        goto bad;
+                       error = EINVAL;
+                       goto bad;
                }
        }
 
        /* Source address validation */
        if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
                }
        }
 
        /* Source address validation */
        if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
-           (flags & IPV6_DADOUTPUT) == 0) {
+           !(flags & IPV6_UNSPECSRC)) {
                error = EOPNOTSUPP;
                ip6stat.ip6s_badscope++;
                goto bad;
                error = EOPNOTSUPP;
                ip6stat.ip6s_badscope++;
                goto bad;
@@ -537,41 +979,71 @@ skip_ipsec2:;
        /*
         * Route packet.
         */
        /*
         * Route packet.
         */
-       if (ro == 0) {
+       if (ro == NULL) {
                ro = &ip6route;
                ro = &ip6route;
-               bzero((caddr_t)ro, sizeof(*ro));
+               bzero((caddr_t)ro, sizeof (*ro));
        }
        ro_pmtu = ro;
        }
        ro_pmtu = ro;
-       if (opt && opt->ip6po_rthdr)
+       if (opt != NULL && opt->ip6po_rthdr)
                ro = &opt->ip6po_route;
                ro = &opt->ip6po_route;
-       dst = (struct sockaddr_in6 *)&ro->ro_dst;
+       dst = SIN6(&ro->ro_dst);
+
+       if (ro->ro_rt != NULL)
+               RT_LOCK_ASSERT_NOTHELD(ro->ro_rt);
        /*
        /*
-        * If there is a cached route,
-        * check that it is to the same destination
-        * and is still up. If not, free it and try again.
+        * if specified, try to fill in the traffic class field.
+        * do not override if a non-zero value is already set.
+        * we check the diffserv field and the ecn field separately.
         */
         */
-       if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
-                        dst->sin6_family != AF_INET6 ||
-                        !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst) ||
-                        ro->ro_rt->generation_id != route_generation)) {
-               rtfree(ro->ro_rt);
-               ro->ro_rt = (struct rtentry *)0;
-       }
-       if (ro->ro_rt == 0) {
-               bzero(dst, sizeof(*dst));
+       if (opt != NULL && opt->ip6po_tclass >= 0) {
+               int mask = 0;
+
+               if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
+                       mask |= 0xfc;
+               if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
+                       mask |= 0x03;
+               if (mask != 0) {
+                       ip6->ip6_flow |=
+                           htonl((opt->ip6po_tclass & mask) << 20);
+               }
+       }
+
+       /* fill in or override the hop limit field, if necessary. */
+       if (opt && opt->ip6po_hlim != -1) {
+               ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
+       } else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
+               if (im6o != NULL) {
+                       IM6O_LOCK(im6o);
+                       ip6->ip6_hlim = im6o->im6o_multicast_hlim;
+                       IM6O_UNLOCK(im6o);
+               } else {
+                       ip6->ip6_hlim = ip6_defmcasthlim;
+               }
+       }
+
+       /*
+        * If there is a cached route, check that it is to the same
+        * destination and is still up. If not, free it and try again.
+        * Test rt_flags without holding rt_lock for performance reasons;
+        * if the route is down it will hopefully be caught by the layer
+        * below (since it uses this route as a hint) or during the
+        * next transmit.
+        */
+       if (ROUTE_UNUSABLE(ro) || dst->sin6_family != AF_INET6 ||
+           !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))
+               ROUTE_RELEASE(ro);
+
+       if (ro->ro_rt == NULL) {
+               bzero(dst, sizeof (*dst));
                dst->sin6_family = AF_INET6;
                dst->sin6_family = AF_INET6;
-               dst->sin6_len = sizeof(struct sockaddr_in6);
+               dst->sin6_len = sizeof (struct sockaddr_in6);
                dst->sin6_addr = ip6->ip6_dst;
                dst->sin6_addr = ip6->ip6_dst;
-#if SCOPEDROUTING
-               /* XXX: sin6_scope_id should already be fixed at this point */
-               if (IN6_IS_SCOPE_LINKLOCAL(&dst->sin6_addr))
-                       dst->sin6_scope_id = ntohs(dst->sin6_addr.s6_addr16[1]);
-#endif
        }
 #if IPSEC
        }
 #if IPSEC
-       if (needipsec && needipsectun) {
-               struct ipsec_output_state state;
-
+       if (ip6obf.needipsec && needipsectun) {
+#if CONFIG_DTRACE
+               struct ifnet *trace_ifp = (ifpp_save != NULL) ? (*ifpp_save) : NULL;
+#endif /* CONFIG_DTRACE */
                /*
                 * All the extension headers will become inaccessible
                 * (since they can be encrypted).
                /*
                 * All the extension headers will become inaccessible
                 * (since they can be encrypted).
@@ -580,24 +1052,31 @@ skip_ipsec2:;
                 *
                 * IPv6 [ESP|AH] IPv6 [extension headers] payload
                 */
                 *
                 * IPv6 [ESP|AH] IPv6 [extension headers] payload
                 */
-               bzero(&exthdrs, sizeof(exthdrs));
+               bzero(&exthdrs, sizeof (exthdrs));
                exthdrs.ip6e_ip6 = m;
 
                exthdrs.ip6e_ip6 = m;
 
-               bzero(&state, sizeof(state));
-               state.m = m;
-               state.ro = (struct route *)ro;
-               state.dst = (struct sockaddr *)dst;
-               lck_mtx_unlock(ip6_mutex);
-               lck_mtx_lock(sadb_mutex);
-               error = ipsec6_output_tunnel(&state, sp, flags);
-               lck_mtx_unlock(sadb_mutex);
-               lck_mtx_lock(ip6_mutex);
-               m = state.m;
-               ro = (struct route_in6 *)state.ro;
-               dst = (struct sockaddr_in6 *)state.dst;
+               ipsec_state.m = m;
+               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 */
+               DTRACE_IP6(send, struct mbuf *, m, struct inpcb *, NULL,
+                   struct ip6_hdr *, ip6, struct ifnet *, trace_ifp,
+                   struct ip *, NULL, struct ip6_hdr *, ip6);
+
+               error = ipsec6_output_tunnel(&ipsec_state, sp, flags);
+               /* tunneled in IPv4? packet is gone */
+               if (ipsec_state.tunneled == 4) {
+                       m = NULL;
+                       goto evaluateloop;
+               }
+               m = ipsec_state.m;
+               ipsec_saved_route = ro;
+               ro = (struct route_in6 *)&ipsec_state.ro;
+               dst = SIN6(ipsec_state.dst);
                if (error) {
                        /* mbuf is already reclaimed in ipsec6_output_tunnel. */
                if (error) {
                        /* mbuf is already reclaimed in ipsec6_output_tunnel. */
-                       m0 = m = NULL;
                        m = NULL;
                        switch (error) {
                        case EHOSTUNREACH:
                        m = NULL;
                        switch (error) {
                        case EHOSTUNREACH:
@@ -607,194 +1086,245 @@ skip_ipsec2:;
                        case ENOMEM:
                                break;
                        default:
                        case ENOMEM:
                                break;
                        default:
-                               printf("ip6_output (ipsec): error code %d\n", error);
-                               /* fall through */
+                               printf("ip6_output (ipsec): error code %d\n",
+                                   error);
+                               /* FALLTHRU */
                        case ENOENT:
                                /* don't show these error codes to the user */
                                error = 0;
                                break;
                        }
                        case ENOENT:
                                /* don't show these error codes to the user */
                                error = 0;
                                break;
                        }
-                       lck_mtx_unlock(sadb_mutex);
                        goto bad;
                }
                        goto bad;
                }
-
+               /*
+                * The packet has been encapsulated so the ifscope
+                * is no longer valid since it does not apply to the
+                * outer address: ignore the ifscope.
+                */
+               if (flags & IPV6_OUTARGS) {
+                       ip6oa->ip6oa_boundif = IFSCOPE_NONE;
+                       ip6oa->ip6oa_flags &= ~IP6OAF_BOUND_IF;
+               }
+               if (opt != NULL && opt->ip6po_pktinfo != NULL) {
+                       if (opt->ip6po_pktinfo->ipi6_ifindex != IFSCOPE_NONE)
+                               opt->ip6po_pktinfo->ipi6_ifindex = IFSCOPE_NONE;
+               }
                exthdrs.ip6e_ip6 = m;
        }
 #endif /* IPSEC */
 
                exthdrs.ip6e_ip6 = m;
        }
 #endif /* IPSEC */
 
-       if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
-               /* Unicast */
-
-#define ifatoia6(ifa)  ((struct in6_ifaddr *)(ifa))
-#define sin6tosa(sin6) ((struct sockaddr *)(sin6))
-               /* xxx
-                * interface selection comes here
-                * if an interface is specified from an upper layer,
-                * ifp must point it.
-                */
-               lck_mtx_lock(rt_mtx);
-               if (ro->ro_rt == 0) {
-                       /*
-                        * non-bsdi always clone routes, if parent is
-                        * PRF_CLONING.
-                        */
-                       rtalloc_ign_locked((struct route *)ro, 0UL);
-               }
-               if (ro->ro_rt == 0) {
-                       ip6stat.ip6s_noroute++;
-                       error = EHOSTUNREACH;
-                       lck_mtx_unlock(rt_mtx);
-                       /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
-                       goto bad;
-               }
-               ia = ifatoia6(ro->ro_rt->rt_ifa);
-               ifp = ro->ro_rt->rt_ifp;
-               ro->ro_rt->rt_use++;
-               if (ro->ro_rt->rt_flags & RTF_GATEWAY)
-                       dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
-               lck_mtx_unlock(rt_mtx);
-               m->m_flags &= ~(M_BCAST | M_MCAST);     /* just in case */
+       /*
+        * ifp should only be filled in for dummy net packets which will jump
+        * to check_with_pf label.
+        */
+       if (ifp != NULL) {
+               VERIFY(ip6obf.route_selected);
+       }
 
 
-               in6_ifstat_inc(ifp, ifs6_out_request);
+       /* adjust pointer */
+       ip6 = mtod(m, struct ip6_hdr *);
 
 
-               /*
-                * Check if the outgoing interface conflicts with
-                * the interface specified by ifi6_ifindex (if specified).
-                * Note that loopback interface is always okay.
-                * (this may happen when we are sending a packet to one of
-                *  our own addresses.)
-                */
-               if (opt && opt->ip6po_pktinfo
-                && opt->ip6po_pktinfo->ipi6_ifindex) {
-                       if (!(ifp->if_flags & IFF_LOOPBACK)
-                        && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
+       if (ip6obf.select_srcif) {
+               bzero(&src_sa, sizeof (src_sa));
+               src_sa.sin6_family = AF_INET6;
+               src_sa.sin6_len = sizeof (src_sa);
+               src_sa.sin6_addr = ip6->ip6_src;
+       }
+       bzero(&dst_sa, sizeof (dst_sa));
+       dst_sa.sin6_family = AF_INET6;
+       dst_sa.sin6_len = sizeof (dst_sa);
+       dst_sa.sin6_addr = ip6->ip6_dst;
+
+       /*
+        * Only call in6_selectroute() on first iteration to avoid taking
+        * multiple references on ifp and rt.
+        *
+        * in6_selectroute() might return an ifp with its reference held
+        * even in the error case, so make sure to release its reference.
+        * ip6oa may be NULL if IPV6_OUTARGS isn't set.
+        */
+       if (!ip6obf.route_selected) {
+               error = in6_selectroute( ip6obf.select_srcif ? &src_sa : NULL,
+                   &dst_sa, opt, im6o, &src_ia, ro, &ifp, &rt, 0, ip6oa);
+
+               if (error != 0) {
+                       switch (error) {
+                       case EHOSTUNREACH:
                                ip6stat.ip6s_noroute++;
                                ip6stat.ip6s_noroute++;
-                               in6_ifstat_inc(ifp, ifs6_out_discard);
-                               error = EHOSTUNREACH;
-                               goto bad;
+                               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 (opt && opt->ip6po_hlim != -1)
-                       ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
-       } else {
-               /* Multicast */
-               struct  in6_multi *in6m;
-
-               m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
-
+               ip6obf.route_selected = TRUE;
+       }
+       if (rt == NULL) {
                /*
                /*
-                * See if the caller provided any multicast options
+                * If in6_selectroute() does not return a route entry,
+                * dst may not have been updated.
                 */
                 */
-               ifp = NULL;
-               if (im6o != NULL) {
-                       ip6->ip6_hlim = im6o->im6o_multicast_hlim;
-                       if (im6o->im6o_multicast_ifp != NULL)
-                               ifp = im6o->im6o_multicast_ifp;
-               } else
-                       ip6->ip6_hlim = ip6_defmcasthlim;
+               *dst = dst_sa;  /* XXX */
+       }
 
 
-               /*
-                * See if the caller provided the outgoing interface
-                * as an ancillary data.
-                * Boundary check for ifindex is assumed to be already done.
-                */
-               if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
-                       ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
+#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 */
 
 
-               /*
-                * If the destination is a node-local scope multicast,
-                * the packet should be loop-backed only.
-                */
-               if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
-                       /*
-                        * If the outgoing interface is already specified,
-                        * it should be a loopback interface.
-                        */
-                       if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
-                               ip6stat.ip6s_badscope++;
-                               error = ENETUNREACH; /* XXX: better error? */
-                               /* XXX correct ifp? */
-                               in6_ifstat_inc(ifp, ifs6_out_discard);
-                               goto bad;
-                       } else {
-                               ifp = &loif[0];
-                       }
+       /*
+        * then rt (for unicast) and ifp must be non-NULL valid values.
+        */
+       if (!(flags & IPV6_FORWARDING)) {
+               in6_ifstat_inc_na(ifp, ifs6_out_request);
+       }
+       if (rt != NULL) {
+               RT_LOCK(rt);
+               if (ia == NULL) {
+                       ia = (struct in6_ifaddr *)(rt->rt_ifa);
+                       if (ia != NULL)
+                               IFA_ADDREF(&ia->ia_ifa);
                }
                }
+               rt->rt_use++;
+               RT_UNLOCK(rt);
+       }
+
+       /*
+        * The outgoing interface must be in the zone of source and
+        * destination addresses (except local/loopback).  We should
+        * use ia_ifp to support the case of sending packets to an
+        * address of our own.
+        */
+       if (ia != NULL && ia->ia_ifp) {
+               ifnet_reference(ia->ia_ifp);    /* for origifp */
+               if (origifp != NULL)
+                       ifnet_release(origifp);
+               origifp = ia->ia_ifp;
+       } else {
+               if (ifp != NULL)
+                       ifnet_reference(ifp);   /* for origifp */
+               if (origifp != NULL)
+                       ifnet_release(origifp);
+               origifp = ifp;
+       }
 
 
-               if (opt && opt->ip6po_hlim != -1)
-                       ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
+       /* skip scope enforcements for local/loopback route */
+       if (rt == NULL || !(rt->rt_ifp->if_flags & IFF_LOOPBACK)) {
+               struct in6_addr src0, dst0;
+               u_int32_t zone;
+
+               src0 = ip6->ip6_src;
+               if (in6_setscope(&src0, origifp, &zone))
+                       goto badscope;
+               bzero(&src_sa, sizeof (src_sa));
+               src_sa.sin6_family = AF_INET6;
+               src_sa.sin6_len = sizeof (src_sa);
+               src_sa.sin6_addr = ip6->ip6_src;
+               if ((sa6_recoverscope(&src_sa, TRUE) ||
+                   zone != src_sa.sin6_scope_id))
+                       goto badscope;
+
+               dst0 = ip6->ip6_dst;
+               if ((in6_setscope(&dst0, origifp, &zone)))
+                       goto badscope;
+               /* re-initialize to be sure */
+               bzero(&dst_sa, sizeof (dst_sa));
+               dst_sa.sin6_family = AF_INET6;
+               dst_sa.sin6_len = sizeof (dst_sa);
+               dst_sa.sin6_addr = ip6->ip6_dst;
+               if ((sa6_recoverscope(&dst_sa, TRUE) ||
+                   zone != dst_sa.sin6_scope_id))
+                       goto badscope;
+
+               /* scope check is done. */
+               goto routefound;
+
+badscope:
+               ip6stat.ip6s_badscope++;
+               in6_ifstat_inc(origifp, ifs6_out_discard);
+               if (error == 0)
+                       error = EHOSTUNREACH; /* XXX */
+               goto bad;
+       }
 
 
+routefound:
+       if (rt != NULL && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
+               if (opt != NULL && opt->ip6po_nextroute.ro_rt) {
+                       /*
+                        * The nexthop is explicitly specified by the
+                        * application.  We assume the next hop is an IPv6
+                        * address.
+                        */
+                       dst = SIN6(opt->ip6po_nexthop);
+               } else if ((rt->rt_flags & RTF_GATEWAY)) {
+                       dst = SIN6(rt->rt_gateway);
+               }
                /*
                /*
-                * If caller did not provide an interface lookup a
-                * default in the routing table.  This is either a
-                * default for the speicfied group (i.e. a host
-                * route), or a multicast default (a route for the
-                * ``net'' ff00::/8).
+                * For packets destined to local/loopback, record the
+                * source the source interface (which owns the source
+                * address), as well as the output interface.  This is
+                * needed to reconstruct the embedded zone for the
+                * link-local address case in ip6_input().
                 */
                 */
-               if (ifp == NULL) {
-                       lck_mtx_lock(rt_mtx);
-                       if (ro->ro_rt == 0) {
-                               ro->ro_rt = rtalloc1_locked((struct sockaddr *)
-                                               &ro->ro_dst, 0, 0UL);
-                       }
-                       if (ro->ro_rt == 0) {
-                               ip6stat.ip6s_noroute++;
-                               lck_mtx_unlock(rt_mtx);
-                               error = EHOSTUNREACH;
-                               /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
-                               goto bad;
-                       }
-                       ia = ifatoia6(ro->ro_rt->rt_ifa);
-                       ifp = ro->ro_rt->rt_ifp;
-                       ro->ro_rt->rt_use++;
-                       lck_mtx_unlock(rt_mtx);
+               if (ia != NULL && (ifp->if_flags & IFF_LOOPBACK)) {
+                       uint32_t srcidx;
+
+                       if (src_ia != NULL)
+                               srcidx = src_ia->ia_ifp->if_index;
+                       else if (ro->ro_srcia != NULL)
+                               srcidx = ro->ro_srcia->ifa_ifp->if_index;
+                       else
+                               srcidx = 0;
+
+                       ip6_setsrcifaddr_info(m, srcidx, NULL);
+                       ip6_setdstifaddr_info(m, 0, ia);
                }
                }
+       }
+
+       if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
+               m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
+       } else {
+               struct  in6_multi *in6m;
 
 
-               if ((flags & IPV6_FORWARDING) == 0)
-                       in6_ifstat_inc(ifp, ifs6_out_request);
-               in6_ifstat_inc(ifp, ifs6_out_mcast);
+               m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
+               in6_ifstat_inc_na(ifp, ifs6_out_mcast);
 
                /*
                 * Confirm that the outgoing interface supports multicast.
                 */
 
                /*
                 * Confirm that the outgoing interface supports multicast.
                 */
-               if ((ifp->if_flags & IFF_MULTICAST) == 0) {
+               if (!(ifp->if_flags & IFF_MULTICAST)) {
                        ip6stat.ip6s_noroute++;
                        in6_ifstat_inc(ifp, ifs6_out_discard);
                        error = ENETUNREACH;
                        goto bad;
                }
                        ip6stat.ip6s_noroute++;
                        in6_ifstat_inc(ifp, ifs6_out_discard);
                        error = ENETUNREACH;
                        goto bad;
                }
-               ifnet_lock_shared(ifp);
-               IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
-               ifnet_lock_done(ifp);
+               in6_multihead_lock_shared();
+               IN6_LOOKUP_MULTI(&ip6->ip6_dst, ifp, in6m);
+               in6_multihead_lock_done();
+               if (im6o != NULL)
+                       IM6O_LOCK(im6o);
                if (in6m != NULL &&
                if (in6m != NULL &&
-                  (im6o == NULL || im6o->im6o_multicast_loop)) {
+                   (im6o == NULL || im6o->im6o_multicast_loop)) {
+                       if (im6o != NULL)
+                               IM6O_UNLOCK(im6o);
                        /*
                         * If we belong to the destination multicast group
                         * on the outgoing interface, and the caller did not
                         * forbid loopback, loop back a copy.
                         */
                        /*
                         * If we belong to the destination multicast group
                         * on the outgoing interface, and the caller did not
                         * forbid loopback, loop back a copy.
                         */
-                       ip6_mloopback(ifp, m, dst);
-               } else {
-                       /*
-                        * If we are acting as a multicast router, perform
-                        * multicast forwarding as if the packet had just
-                        * arrived on the interface to which we are about
-                        * to send.  The multicast forwarding function
-                        * recursively calls this function, using the
-                        * IPV6_FORWARDING flag to prevent infinite recursion.
-                        *
-                        * Multicasts that are looped back by ip6_mloopback(),
-                        * above, will be forwarded by the ip6_input() routine,
-                        * if necessary.
-                        */
-                       if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
-                               if (ip6_mforward(ip6, ifp, m) != NULL) {
-                                       m_freem(m);
-                                       goto done;
-                               }
-                       }
-               }
+                       ip6_mloopback(NULL, ifp, m, dst, optlen, nxt0);
+               } else if (im6o != NULL)
+                       IM6O_UNLOCK(im6o);
+               if (in6m != NULL)
+                       IN6M_REMREF(in6m);
                /*
                 * Multicasts with a hoplimit of zero may be looped back,
                 * above, but must not be transmitted on a network.
                /*
                 * Multicasts with a hoplimit of zero may be looped back,
                 * above, but must not be transmitted on a network.
@@ -803,9 +1333,13 @@ skip_ipsec2:;
                 * loop back a copy if this host actually belongs to the
                 * destination group on the loopback interface.
                 */
                 * loop back a copy if this host actually belongs to the
                 * destination group on the loopback interface.
                 */
-               if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
-                       m_freem(m);
-                       goto done;
+               if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
+                   IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
+                       /* remove m from the packetchain and continue looping */
+                       if (m != NULL)
+                               m_freem(m);
+                       m = NULL;
+                       goto evaluateloop;
                }
        }
 
                }
        }
 
@@ -813,211 +1347,425 @@ skip_ipsec2:;
         * Fill the outgoing inteface to tell the upper layer
         * to increment per-interface statistics.
         */
         * Fill the outgoing inteface to tell the upper layer
         * to increment per-interface statistics.
         */
-       if (ifpp)
+       if (ifpp != NULL && *ifpp == NULL) {
+               ifnet_reference(ifp);   /* for caller */
                *ifpp = ifp;
                *ifpp = ifp;
-
-       /*
-        * Determine path MTU.
-        */
-       if (ro_pmtu != ro) {
-               /* The first hop and the final destination may differ. */
-               struct sockaddr_in6 *sin6_fin =
-                       (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
-               if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
-                                      !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
-                                                          &finaldst))) {
-                       rtfree(ro_pmtu->ro_rt);
-                       ro_pmtu->ro_rt = (struct rtentry *)0;
-               }
-               if (ro_pmtu->ro_rt == 0) {
-                       bzero(sin6_fin, sizeof(*sin6_fin));
-                       sin6_fin->sin6_family = AF_INET6;
-                       sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
-                       sin6_fin->sin6_addr = finaldst;
-
-                       rtalloc((struct route *)ro_pmtu);
-               }
        }
        }
-       if (ro_pmtu->ro_rt != NULL) {
-               u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
 
 
-               mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
-               if (mtu > ifmtu || mtu == 0) {
-                       /*
-                        * The MTU on the route is larger than the MTU on
-                        * the interface!  This shouldn't happen, unless the
-                        * MTU of the interface has been changed after the
-                        * interface was brought up.  Change the MTU in the
-                        * route to match the interface MTU (as long as the
-                        * field isn't locked).
-                        *
-                        * if MTU on the route is 0, we need to fix the MTU.
-                        * this case happens with path MTU discovery timeouts.
-                        */
-                        mtu = ifmtu;
-                        if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
-                                ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
-               }
-       } else {
-               mtu = nd_ifinfo[ifp->if_index].linkmtu;
-       }
+       /* Determine path MTU. */
+       if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
+           &alwaysfrag)) != 0)
+               goto bad;
 
        /*
 
        /*
-        * advanced API (IPV6_USE_MIN_MTU) overrides mtu setting
+        * The caller of this function may specify to use the minimum MTU
+        * in some cases.
+        * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
+        * setting.  The logic is a bit complicated; by default, unicast
+        * packets will follow path MTU while multicast packets will be sent at
+        * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
+        * including unicast ones will be sent at the minimum MTU.  Multicast
+        * packets will always be sent at the minimum MTU unless
+        * IP6PO_MINMTU_DISABLE is explicitly specified.
+        * See RFC 3542 for more details.
         */
         */
-       if ((flags & IPV6_MINMTU) != 0 && mtu > IPV6_MMTU)
-               mtu = IPV6_MMTU;
-
-       /* Fake scoped addresses */
-       if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
-               /*
-                * If source or destination address is a scoped address, and
-                * the packet is going to be sent to a loopback interface,
-                * we should keep the original interface.
-                */
-
-               /*
-                * XXX: this is a very experimental and temporary solution.
-                * We eventually have sockaddr_in6 and use the sin6_scope_id
-                * field of the structure here.
-                * We rely on the consistency between two scope zone ids
-                * of source and destination, which should already be assured.
-                * Larger scopes than link will be supported in the future. 
-                */
-               origifp = NULL;
-               if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
-                       origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
-               else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
-                       origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
-               /*
-                * XXX: origifp can be NULL even in those two cases above.
-                * For example, if we remove the (only) link-local address
-                * from the loopback interface, and try to send a link-local
-                * address without link-id information.  Then the source
-                * address is ::1, and the destination address is the
-                * link-local address with its s6_addr16[1] being zero.
-                * What is worse, if the packet goes to the loopback interface
-                * by a default rejected route, the null pointer would be
-                * passed to looutput, and the kernel would hang.
-                * The following last resort would prevent such disaster.
-                */
-               if (origifp == NULL)
-                       origifp = ifp;
+       if (mtu > IPV6_MMTU) {
+               if ((flags & IPV6_MINMTU)) {
+                       mtu = IPV6_MMTU;
+               } else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL) {
+                       mtu = IPV6_MMTU;
+               } else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
+                   (opt == NULL ||
+                   opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
+                       mtu = IPV6_MMTU;
+               }
        }
        }
-       else
-               origifp = ifp;
-#ifndef SCOPEDROUTING
+
        /*
         * clear embedded scope identifiers if necessary.
         * in6_clearscope will touch the addresses only when necessary.
         */
        in6_clearscope(&ip6->ip6_src);
        in6_clearscope(&ip6->ip6_dst);
        /*
         * clear embedded scope identifiers if necessary.
         * in6_clearscope will touch the addresses only when necessary.
         */
        in6_clearscope(&ip6->ip6_src);
        in6_clearscope(&ip6->ip6_dst);
-#endif
-
-       /*
-        * Check with the firewall...
-        */
-        if (ip6_fw_enable && ip6_fw_chk_ptr) {
-               u_short port = 0;
-               m->m_pkthdr.rcvif = NULL;       /* XXX */
-               /* If ipfw says divert, we have to just drop packet */
-               if (ip6_fw_chk_ptr(&ip6, ifp, &port, &m)) {
-                       m_freem(m);
-                       goto done;
-               }
-               if (!m) {
-                       error = EACCES;
-                       goto done;
-               }
-       }
-
        /*
         * If the outgoing packet contains a hop-by-hop options header,
         * it must be examined and processed even by the source node.
         * (RFC 2460, section 4.)
         */
        /*
         * If the outgoing packet contains a hop-by-hop options header,
         * it must be examined and processed even by the source node.
         * (RFC 2460, section 4.)
         */
-       if (exthdrs.ip6e_hbh) {
+       if (exthdrs.ip6e_hbh != NULL) {
                struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
                struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
-               u_int32_t dummy1; /* XXX unused */
-               u_int32_t dummy2; /* XXX unused */
-
+               u_int32_t dummy; /* XXX unused */
+               uint32_t oplen = 0; /* for ip6_process_hopopts() */
 #if DIAGNOSTIC
                if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
                        panic("ip6e_hbh is not continuous");
 #endif
                /*
 #if DIAGNOSTIC
                if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
                        panic("ip6e_hbh is not continuous");
 #endif
                /*
-                *  XXX: if we have to send an ICMPv6 error to the sender,
-                *       we need the M_LOOP flag since icmp6_error() expects
-                *       the IPv6 and the hop-by-hop options header are
-                *       continuous unless the flag is set.
+                * XXX: If we have to send an ICMPv6 error to the sender,
+                * we need the M_LOOP flag since icmp6_error() expects
+                * the IPv6 and the hop-by-hop options header are
+                * continuous unless the flag is set.
                 */
                m->m_flags |= M_LOOP;
                m->m_pkthdr.rcvif = ifp;
                 */
                m->m_flags |= M_LOOP;
                m->m_pkthdr.rcvif = ifp;
-               if (ip6_process_hopopts(m,
-                                       (u_int8_t *)(hbh + 1),
-                                       ((hbh->ip6h_len + 1) << 3) -
-                                       sizeof(struct ip6_hbh),
-                                       &dummy1, &dummy2) < 0) {
-                       /* m was already freed at this point */
-                       error = EINVAL;/* better error? */
-                       goto done;
+               if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
+                   ((hbh->ip6h_len + 1) << 3) - sizeof (struct ip6_hbh),
+                   &dummy, &oplen) < 0) {
+                       /*
+                        * m was already freed at this point. Set to NULL so it
+                        * is not re-freed at end of ip6_output_list.
+                        */
+                       m = NULL;
+                       error = EINVAL; /* better error? */
+                       goto bad;
                }
                m->m_flags &= ~M_LOOP; /* XXX */
                m->m_pkthdr.rcvif = NULL;
        }
 
                }
                m->m_flags &= ~M_LOOP; /* XXX */
                m->m_pkthdr.rcvif = NULL;
        }
 
+#if DUMMYNET
+check_with_pf:
+#endif /* DUMMYNET */
+#if PF
+       if (PF_IS_ENABLED) {
+#if DUMMYNET
+
+               /*
+                * TODO: Need to save opt->ip6po_flags for reinjection
+                * rdar://10434993
+                */
+               args.fwa_m = m;
+               args.fwa_oif = ifp;
+               args.fwa_oflags = flags;
+               if (flags & IPV6_OUTARGS)
+                       args.fwa_ip6oa = ip6oa;
+               args.fwa_ro6 = ro;
+               args.fwa_dst6 = dst;
+               args.fwa_ro6_pmtu = ro_pmtu;
+               args.fwa_origifp = origifp;
+               args.fwa_mtu = mtu;
+               args.fwa_alwaysfrag = alwaysfrag;
+               args.fwa_unfragpartlen = unfragpartlen;
+               args.fwa_exthdrs = &exthdrs;
+               /* Invoke outbound packet filter */
+               error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, &args);
+#else /* !DUMMYNET */
+               error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, NULL);
+#endif /* !DUMMYNET */
+
+               if (error != 0 || m == NULL) {
+                       if (m != NULL) {
+                               panic("%s: unexpected packet %p\n",
+                                   __func__, m);
+                               /* NOTREACHED */
+                       }
+                       /* m was already freed by callee and is now NULL.  */
+                       goto evaluateloop;
+               }
+               ip6 = mtod(m, struct ip6_hdr *);
+       }
+#endif /* PF */
+
+#ifdef IPSEC
+       /* clean ipsec history before fragmentation */
+       ipsec_delaux(m);
+#endif /* IPSEC */
+
+       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.
+        * Determine whether fragmentation is necessary. If so, m is passed
+        * back as a chain of packets and original mbuf is freed. Otherwise, m
+        * is unchanged.
         */
         */
-       tlen = m->m_pkthdr.len;
-       if (tlen <= mtu
-#if notyet
-           /*
-            * On any link that cannot convey a 1280-octet packet in one piece,
-            * link-specific fragmentation and reassembly must be provided at
-            * a layer below IPv6. [RFC 2460, sec.5]
-            * Thus if the interface has ability of link-level fragmentation,
-            * we can just send the packet even if the packet size is
-            * larger than the link's MTU.
-            * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
-            */
-       
-           || ifp->if_flags & IFF_FRAGMENTABLE
-#endif
-           )
-       {
-               /* Record statistics for this interface address. */
-               if (ia && !(flags & IPV6_FORWARDING)) {
-#ifndef __APPLE__
-                       ia->ia_ifa.if_opackets++;
-                       ia->ia_ifa.if_obytes += m->m_pkthdr.len;
-#endif
-               }
-#ifdef IPSEC
-               /* clean ipsec history once it goes out of the node */
-               ipsec_delaux(m);
-#endif
+       error = ip6_fragment_packet(&m, opt,
+           &exthdrs, ifp, mtu, alwaysfrag, unfragpartlen, ro_pmtu, nxt0,
+           optlen);
 
 
-               error = nd6_output(ifp, origifp, m, dst, ro->ro_rt, 1);
-               goto done;
-       } else if (mtu < IPV6_MMTU) {
+       if (error)
+               goto bad;
+
+/*
+ * The evaluateloop label is where we decide whether to continue looping over
+ * packets or call into nd code to send.
+ */
+evaluateloop:
+
+       /*
+        * m may be NULL when we jump to the evaluateloop label from PF or
+        * other code that can drop packets.
+        */
+       if (m != NULL) {
+               /*
+                * If we already have a chain to send, tack m onto the end.
+                * Otherwise make m the start and end of the to-be-sent chain.
+                */
+               if (sendchain != NULL) {
+                       sendchain_last->m_nextpkt = m;
+               } else {
+                       sendchain = m;
+               }
+
+               /* Fragmentation may mean m is a chain. Find the last packet. */
+               while (m->m_nextpkt)
+                       m = m->m_nextpkt;
+               sendchain_last = m;
+               pktcnt++;
+       }
+
+       /* Fill in next m from inputchain as appropriate. */
+       m = inputchain;
+       if (m != NULL) {
+               /* Isolate m from rest of input chain. */
+               inputchain = m->m_nextpkt;
+               m->m_nextpkt = NULL;
+
+               /*
+                * Clear exthdrs and ipsec_state so stale contents are not
+                * reused. Note this also clears the exthdrs.merged flag.
+                */
+               bzero(&exthdrs, sizeof(exthdrs));
+               bzero(&ipsec_state, sizeof(ipsec_state));
+
+               /* Continue looping. */
+               goto loopit;
+       }
+
+       /*
+        * If we get here, there's no more mbufs in inputchain, so send the
+        * sendchain if there is one.
+        */
+       if (pktcnt > 0) {
+               error = nd6_output_list(ifp, origifp, sendchain, dst,
+                   ro->ro_rt, adv);
+               /*
+                * Fall through to done label even in error case because
+                * nd6_output_list frees packetchain in both success and
+                * failure cases.
+                */
+       }
+
+done:
+       if (ifpp_save != NULL && *ifpp_save != NULL) {
+               ifnet_release(*ifpp_save);
+               *ifpp_save = NULL;
+       }
+       ROUTE_RELEASE(&ip6route);
+#if IPSEC
+       ROUTE_RELEASE(&ipsec_state.ro);
+       if (sp != NULL)
+               key_freesp(sp, KEY_SADB_UNLOCKED);
+#endif /* IPSEC */
+#if NECP
+       ROUTE_RELEASE(&necp_route);
+#endif /* NECP */
+#if DUMMYNET
+       ROUTE_RELEASE(&saved_route);
+       ROUTE_RELEASE(&saved_ro_pmtu);
+#endif /* DUMMYNET */
+
+       if (ia != NULL)
+               IFA_REMREF(&ia->ia_ifa);
+       if (src_ia != NULL)
+               IFA_REMREF(&src_ia->ia_ifa);
+       if (ifp != NULL)
+               ifnet_release(ifp);
+       if (origifp != NULL)
+               ifnet_release(origifp);
+       if (ip6_output_measure) {
+               net_perf_measure_time(&net_perf, &start_tv, packets_processed);
+               net_perf_histogram(&net_perf, packets_processed);
+       }
+       return (error);
+
+freehdrs:
+       if (exthdrs.ip6e_hbh != NULL) {
+               if (exthdrs.merged)
+                       panic("Double free of ip6e_hbh");
+               m_freem(exthdrs.ip6e_hbh);
+       }
+       if (exthdrs.ip6e_dest1 != NULL) {
+               if (exthdrs.merged)
+                       panic("Double free of ip6e_dest1");
+               m_freem(exthdrs.ip6e_dest1);
+       }
+       if (exthdrs.ip6e_rthdr != NULL) {
+               if (exthdrs.merged)
+                       panic("Double free of ip6e_rthdr");
+               m_freem(exthdrs.ip6e_rthdr);
+       }
+       if (exthdrs.ip6e_dest2 != NULL) {
+               if (exthdrs.merged)
+                       panic("Double free of ip6e_dest2");
+               m_freem(exthdrs.ip6e_dest2);
+       }
+       /* FALLTHRU */
+bad:
+       if (inputchain != NULL)
+               m_freem_list(inputchain);
+       if (sendchain != NULL)
+               m_freem_list(sendchain);
+       if (m != NULL)
+               m_freem(m);
+
+       goto done;
+
+#undef ipf_pktopts
+#undef exthdrs
+#undef ip6route
+#undef ipsec_state
+#undef saved_route
+#undef saved_ro_pmtu
+#undef args
+}
+
+/* ip6_fragment_packet
+ *
+ * The fragmentation logic is rather complex:
+ * 1: normal case (dontfrag == 0, alwaysfrag == 0)
+ * 1-a:        send as is if tlen <= path mtu
+ * 1-b:        fragment if tlen > path mtu
+ *
+ * 2: if user asks us not to fragment (dontfrag == 1)
+ * 2-a:        send as is if tlen <= interface mtu
+ * 2-b:        error if tlen > interface mtu
+ *
+ * 3: if we always need to attach fragment header (alwaysfrag == 1)
+ *     always fragment
+ *
+ * 4: if dontfrag == 1 && alwaysfrag == 1
+ *     error, as we cannot handle this conflicting request
+ */
+
+static int
+ip6_fragment_packet(struct mbuf **mptr, struct ip6_pktopts *opt,
+     struct ip6_exthdrs *exthdrsp, struct ifnet *ifp, uint32_t mtu,
+     boolean_t alwaysfrag, uint32_t unfragpartlen, struct route_in6 *ro_pmtu,
+     int nxt0, uint32_t optlen)
+{
+       VERIFY(NULL != mptr);
+       struct mbuf *m = *mptr;
+       int error = 0;
+       size_t tlen = m->m_pkthdr.len;
+       boolean_t dontfrag = (opt != NULL && (opt->ip6po_flags & IP6PO_DONTFRAG));
+
+       if (m->m_pkthdr.pkt_flags & PKTF_FORWARDED) {
+               dontfrag = TRUE;
+               /*
+                * Discard partial sum information if this packet originated
+                * from another interface; the packet would already have the
+                * final checksum and we shouldn't recompute it.
+                */
+               if ((m->m_pkthdr.csum_flags & (CSUM_DATA_VALID|CSUM_PARTIAL)) ==
+                   (CSUM_DATA_VALID|CSUM_PARTIAL)) {
+                       m->m_pkthdr.csum_flags &= ~CSUM_TX_FLAGS;
+                       m->m_pkthdr.csum_data = 0;
+               }
+       }
+
+       if (dontfrag && alwaysfrag) {   /* case 4 */
+               /* conflicting request - can't transmit */
+               return EMSGSIZE;
+       }
+
+       /* Access without acquiring nd_ifinfo lock for performance */
+       if (dontfrag && tlen > IN6_LINKMTU(ifp)) {      /* case 2-b */
+               /*
+                * Even if the DONTFRAG option is specified, we cannot send the
+                * packet when the data length is larger than the MTU of the
+                * outgoing interface.
+                * Notify the error by sending IPV6_PATHMTU ancillary data as
+                * well as returning an error code (the latter is not described
+                * in the API spec.)
+                */
+               u_int32_t mtu32;
+               struct ip6ctlparam ip6cp;
+
+               mtu32 = (u_int32_t)mtu;
+               bzero(&ip6cp, sizeof (ip6cp));
+               ip6cp.ip6c_cmdarg = (void *)&mtu32;
+               pfctlinput2(PRC_MSGSIZE, SA(&ro_pmtu->ro_dst), (void *)&ip6cp);
+               return EMSGSIZE;
+       }
+
+       /*
+        * transmit packet without fragmentation
+        */
+       if (dontfrag || (!alwaysfrag &&         /* case 1-a and 2-a */
+           (tlen <= mtu || TSO_IPV6_OK(ifp, m) ||
+           (ifp->if_hwassist & CSUM_FRAGMENT_IPV6)))) {
+               /*
+                * mppn not updated in this case because no new chain is formed
+                * and inserted
+                */
+               ip6_output_checksum(ifp, mtu, m, nxt0, tlen, optlen);
+       } else {
                /*
                /*
-                * note that path MTU is never less than IPV6_MMTU
-                * (see icmp6_input).
+                * time to fragment - cases 1-b and 3 are handled inside
+                * ip6_do_fragmentation().
+                * mppn is passed down to be updated to point at fragment chain.
                 */
                 */
-               error = EMSGSIZE;
+               error = ip6_do_fragmentation(mptr, optlen, ifp,
+                   unfragpartlen, mtod(m, struct ip6_hdr *), exthdrsp, mtu, nxt0);
+       }
+
+       return error;
+}
+
+/*
+ * ip6_do_fragmentation() is called by ip6_fragment_packet() after determining
+ * the packet needs to be fragmented. on success, morig is freed and a chain
+ * of fragments is linked into the packet chain where morig existed. Otherwise,
+ * an errno is returned.
+ */
+int
+ip6_do_fragmentation(struct mbuf **mptr, uint32_t optlen, struct ifnet *ifp,
+    uint32_t unfragpartlen, struct ip6_hdr *ip6, struct ip6_exthdrs *exthdrsp,
+    uint32_t mtu, int nxt0)
+{
+       VERIFY(NULL != mptr);
+       int error = 0;
+
+       struct mbuf *morig = *mptr;
+       struct mbuf *first_mbufp = NULL;
+       struct mbuf *last_mbufp = NULL;
+
+       size_t tlen = morig->m_pkthdr.len;
+
+       /*
+        * try to fragment the packet.  case 1-b and 3
+        */
+       if ((morig->m_pkthdr.csum_flags & CSUM_TSO_IPV6)) {
+               /* TSO and fragment aren't compatible */
                in6_ifstat_inc(ifp, ifs6_out_fragfail);
                in6_ifstat_inc(ifp, ifs6_out_fragfail);
-               goto bad;
-       } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
-               error = EMSGSIZE;
+               return EMSGSIZE;
+       } else if (mtu < IPV6_MMTU) {
+               /* path MTU cannot be less than IPV6_MMTU */
                in6_ifstat_inc(ifp, ifs6_out_fragfail);
                in6_ifstat_inc(ifp, ifs6_out_fragfail);
-               goto bad;
+               return EMSGSIZE;
+       } else if (ip6->ip6_plen == 0) {
+               /* jumbo payload cannot be fragmented */
+               in6_ifstat_inc(ifp, ifs6_out_fragfail);
+               return EMSGSIZE;
        } else {
        } else {
-               struct mbuf **mnext, *m_frgpart;
+               size_t hlen, len, off;
+               struct mbuf **mnext = NULL;
                struct ip6_frag *ip6f;
                struct ip6_frag *ip6f;
-               u_int32_t id = htonl(ip6_id++);
+               u_int32_t id = htonl(ip6_randomid());
                u_char nextproto;
 
                /*
                u_char nextproto;
 
                /*
@@ -1029,76 +1777,100 @@ skip_ipsec2:;
                if (mtu > IPV6_MAXPACKET)
                        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) {
                if (len < 8) {
-                       error = EMSGSIZE;
                        in6_ifstat_inc(ifp, ifs6_out_fragfail);
                        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.
                 */
                /*
                 * Change the next header field of the last header in the
                 * unfragmentable part.
                 */
-               if (exthdrs.ip6e_rthdr) {
-                       nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
-                       *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
-               } else if (exthdrs.ip6e_dest1) {
-                       nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
-                       *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
-               } else if (exthdrs.ip6e_hbh) {
-                       nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
-                       *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
+               if (exthdrsp->ip6e_rthdr != NULL) {
+                       nextproto = *mtod(exthdrsp->ip6e_rthdr, u_char *);
+                       *mtod(exthdrsp->ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
+               } else if (exthdrsp->ip6e_dest1 != NULL) {
+                       nextproto = *mtod(exthdrsp->ip6e_dest1, u_char *);
+                       *mtod(exthdrsp->ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
+               } else if (exthdrsp->ip6e_hbh != NULL) {
+                       nextproto = *mtod(exthdrsp->ip6e_hbh, u_char *);
+                       *mtod(exthdrsp->ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
                } else {
                        nextproto = ip6->ip6_nxt;
                        ip6->ip6_nxt = IPPROTO_FRAGMENT;
                }
 
                } else {
                        nextproto = ip6->ip6_nxt;
                        ip6->ip6_nxt = IPPROTO_FRAGMENT;
                }
 
+               if (morig->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA)
+                       in6_delayed_cksum_offset(morig, 0, optlen, nxt0);
+
                /*
                 * Loop through length of segment after first fragment,
                 * make new header and copy data of each part and link onto
                 * chain.
                 */
                /*
                 * 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) {
                for (off = hlen; off < tlen; off += len) {
-                       MGETHDR(m, M_DONTWAIT, MT_HEADER);
-                       if (!m) {
+                       struct ip6_hdr *new_mhip6;
+                       struct mbuf *new_m;
+                       struct mbuf *m_frgpart;
+
+                       MGETHDR(new_m, M_DONTWAIT, MT_HEADER);  /* MAC-OK */
+                       if (new_m == NULL) {
                                error = ENOBUFS;
                                ip6stat.ip6s_odropped++;
                                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);
-                       if (error) {
+                       mnext = &new_m->m_nextpkt;
+
+                       new_m->m_data += max_linkhdr;
+                       new_mhip6 = mtod(new_m, struct ip6_hdr *);
+                       *new_mhip6 = *ip6;
+                       new_m->m_len = sizeof (*new_mhip6);
+
+                       error = ip6_insertfraghdr(morig, new_m, hlen, &ip6f);
+                       if (error) {
                                ip6stat.ip6s_odropped++;
                                ip6stat.ip6s_odropped++;
-                               goto sendorfree;
+                               break;
                        }
                        }
+
                        ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
                        if (off + len >= tlen)
                                len = tlen - off;
                        else
                                ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
                        ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
                        if (off + len >= tlen)
                                len = tlen - off;
                        else
                                ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
-                       mhip6->ip6_plen = htons((u_short)(len + hlen +
-                                                         sizeof(*ip6f) -
-                                                         sizeof(struct ip6_hdr)));
-                       if ((m_frgpart = m_copy(m0, off, len)) == 0) {
+                       new_mhip6->ip6_plen = htons((u_short)(len + hlen +
+                           sizeof (*ip6f) - sizeof (struct ip6_hdr)));
+
+                       if ((m_frgpart = m_copy(morig, off, len)) == NULL) {
                                error = ENOBUFS;
                                ip6stat.ip6s_odropped++;
                                error = ENOBUFS;
                                ip6stat.ip6s_odropped++;
-                               goto sendorfree;
+                               break;
                        }
                        }
-                       m_cat(m, m_frgpart);
-                       m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
-                       m->m_pkthdr.rcvif = 0;
-                       m->m_pkthdr.socket_id = m0->m_pkthdr.socket_id;
+                       m_cat(new_m, m_frgpart);
+                       new_m->m_pkthdr.len = len + hlen + sizeof (*ip6f);
+                       new_m->m_pkthdr.rcvif = NULL;
+
+                       M_COPY_CLASSIFIER(new_m, morig);
+                       M_COPY_PFTAG(new_m, morig);
+
+#ifdef notyet
+#if CONFIG_MACF_NET
+                       mac_create_fragment(morig, new_m);
+#endif /* CONFIG_MACF_NET */
+#endif /* notyet */
+
                        ip6f->ip6f_reserved = 0;
                        ip6f->ip6f_ident = id;
                        ip6f->ip6f_nxt = nextproto;
                        ip6f->ip6f_reserved = 0;
                        ip6f->ip6f_ident = id;
                        ip6f->ip6f_nxt = nextproto;
@@ -1106,113 +1878,222 @@ skip_ipsec2:;
                        in6_ifstat_inc(ifp, ifs6_out_fragcreat);
                }
 
                        in6_ifstat_inc(ifp, ifs6_out_fragcreat);
                }
 
-               in6_ifstat_inc(ifp, ifs6_out_fragok);
+               if (error) {
+                       /* free all the fragments created */
+                       if (first_mbufp != NULL) {
+                               m_freem_list(first_mbufp);
+                               first_mbufp = NULL;
+                       }
+                       last_mbufp = NULL;
+               } else {
+                       /* successful fragmenting */
+                       m_freem(morig);
+                       *mptr = first_mbufp;
+                       last_mbufp->m_nextpkt = NULL;
+                       ip6stat.ip6s_fragmented++;
+                       in6_ifstat_inc(ifp, ifs6_out_fragok);
+               }
        }
        }
+       return error;
+}
 
 
-       /*
-        * Remove leading garbages.
-        */
-sendorfree:
-       m = m0->m_nextpkt;
-       m0->m_nextpkt = 0;
-       m_freem(m0);
-       for (m0 = m; m; m = m0) {
-               m0 = m->m_nextpkt;
-               m->m_nextpkt = 0;
-               if (error == 0) {
-                       /* Record statistics for this interface address. */
-                       if (ia) {
-#ifndef __APPLE__
-                               ia->ia_ifa.if_opackets++;
-                               ia->ia_ifa.if_obytes += m->m_pkthdr.len;
-#endif
-                       }
-#if IPSEC
-                       /* clean ipsec history once it goes out of the node */
-                       ipsec_delaux(m);
-#endif
-                       error = nd6_output(ifp, origifp, m, dst, ro->ro_rt, 1);
+static int
+ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
+{
+       struct mbuf *m;
 
 
-               } else
-                       m_freem(m);
-       }
+       if (hlen > MCLBYTES)
+               return (ENOBUFS); /* XXX */
 
 
-       if (error == 0)
-               ip6stat.ip6s_fragmented++;
+       MGET(m, M_DONTWAIT, MT_DATA);
+       if (m == NULL)
+               return (ENOBUFS);
 
 
-done:
-       if (!locked)
-               lck_mtx_unlock(ip6_mutex);
-       if (ro == &ip6route && ro->ro_rt) { /* brace necessary for rtfree */
-               rtfree(ro->ro_rt);
-       } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
-               rtfree(ro_pmtu->ro_rt);
-       }
-
-#if IPSEC
-       if (sp != NULL) {
-               lck_mtx_lock(sadb_mutex);
-               key_freesp(sp);
-               lck_mtx_unlock(sadb_mutex);
+       if (hlen > MLEN) {
+               MCLGET(m, M_DONTWAIT);
+               if (!(m->m_flags & M_EXT)) {
+                       m_free(m);
+                       return (ENOBUFS);
+               }
        }
        }
-#endif /* IPSEC */
+       m->m_len = hlen;
+       if (hdr != NULL)
+               bcopy(hdr, mtod(m, caddr_t), hlen);
 
 
-       return(error);
+       *mp = m;
+       return (0);
+}
 
 
-freehdrs:
-       m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
-       m_freem(exthdrs.ip6e_dest1);
-       m_freem(exthdrs.ip6e_rthdr);
-       m_freem(exthdrs.ip6e_dest2);
-       /* fall through */
-bad:
-       m_freem(m);
-       goto done;
+static void
+ip6_out_cksum_stats(int proto, u_int32_t len)
+{
+       switch (proto) {
+       case IPPROTO_TCP:
+               tcp_out6_cksum_stats(len);
+               break;
+       case IPPROTO_UDP:
+               udp_out6_cksum_stats(len);
+               break;
+       default:
+               /* keep only TCP or UDP stats for now */
+               break;
+       }
 }
 
 }
 
-static int
-ip6_copyexthdr(mp, hdr, hlen)
-       struct mbuf **mp;
-       caddr_t hdr;
-       int hlen;
+/*
+ * Process a delayed payload checksum calculation (outbound path.)
+ *
+ * hoff is the number of bytes beyond the mbuf data pointer which
+ * points to the IPv6 header.  optlen is the number of bytes, if any,
+ * between the end of IPv6 header and the beginning of the ULP payload
+ * header, which represents the extension headers.  If optlen is less
+ * than zero, this routine will bail when it detects extension headers.
+ *
+ * Returns a bitmask representing all the work done in software.
+ */
+uint32_t
+in6_finalize_cksum(struct mbuf *m, uint32_t hoff, int32_t optlen,
+    int32_t nxt0, uint32_t csum_flags)
 {
 {
-       struct mbuf *m;
+       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;
 
 
-       if (hlen > MCLBYTES)
-               return(ENOBUFS); /* XXX */
+       _CASSERT(sizeof (csum) == sizeof (uint16_t));
+       VERIFY(m->m_flags & M_PKTHDR);
 
 
-       MGET(m, M_DONTWAIT, MT_DATA);
-       if (!m)
-               return(ENOBUFS);
+       sw_csum = (csum_flags & m->m_pkthdr.csum_flags);
 
 
-       if (hlen > MLEN) {
-               MCLGET(m, M_DONTWAIT);
-               if ((m->m_flags & M_EXT) == 0) {
-                       m_free(m);
-                       return(ENOBUFS);
+       if ((sw_csum &= CSUM_DELAY_IPV6_DATA) == 0)
+               goto done;
+
+       mlen = m->m_pkthdr.len;                         /* total mbuf len */
+       hlen = sizeof (*ip6);                           /* IPv6 header len */
+
+       /* sanity check (need at least IPv6 header) */
+       if (mlen < (hoff + hlen)) {
+               panic("%s: mbuf %p pkt len (%u) < hoff+ip6_hdr "
+                   "(%u+%u)\n", __func__, m, mlen, hoff, hlen);
+               /* NOTREACHED */
+       }
+
+       /*
+        * In case the IPv6 header is not contiguous, or not 32-bit
+        * aligned, copy it to a local buffer.
+        */
+       if ((hoff + hlen) > m->m_len ||
+           !IP6_HDR_ALIGNED_P(mtod(m, caddr_t) + hoff)) {
+               m_copydata(m, hoff, hlen, (caddr_t)buf);
+               ip6 = (struct ip6_hdr *)(void *)buf;
+       } else {
+               ip6 = (struct ip6_hdr *)(void *)(m->m_data + hoff);
+       }
+
+       nxt = ip6->ip6_nxt;
+       plen = ntohs(ip6->ip6_plen);
+       if (plen != (mlen - (hoff + hlen))) {
+               plen = OSSwapInt16(plen);
+               if (plen != (mlen - (hoff + hlen))) {
+                       /* Don't complain for jumbograms */
+                       if (plen != 0 || nxt != IPPROTO_HOPOPTS) {
+                               printf("%s: mbuf 0x%llx proto %d IPv6 "
+                                   "plen %d (%x) [swapped %d (%x)] doesn't "
+                                   "match actual packet length; %d is used "
+                                   "instead\n", __func__,
+                                   (uint64_t)VM_KERNEL_ADDRPERM(m), nxt,
+                                   ip6->ip6_plen, ip6->ip6_plen, plen, plen,
+                                   (mlen - (hoff + hlen)));
+                       }
+                       plen = mlen - (hoff + hlen);
                }
        }
                }
        }
-       m->m_len = hlen;
-       if (hdr)
-               bcopy(hdr, mtod(m, caddr_t), hlen);
 
 
-       *mp = m;
-       return(0);
+       if (optlen < 0) {
+               /* next header isn't TCP/UDP and we don't know optlen, bail */
+               if (nxt != IPPROTO_TCP && nxt != IPPROTO_UDP) {
+                       sw_csum = 0;
+                       goto done;
+               }
+               olen = 0;
+       } else {
+               /* caller supplied the original transport number; use it */
+               if (nxt0 >= 0)
+                       nxt = nxt0;
+               olen = optlen;
+       }
+
+       offset = hoff + hlen + olen;                    /* ULP header */
+
+       /* sanity check */
+       if (mlen < offset) {
+               panic("%s: mbuf %p pkt len (%u) < hoff+ip6_hdr+ext_hdr "
+                   "(%u+%u+%u)\n", __func__, m, mlen, hoff, hlen, olen);
+               /* NOTREACHED */
+       }
+
+       /*
+        * offset is added to the lower 16-bit value of csum_data,
+        * which is expected to contain the ULP offset; therefore
+        * CSUM_PARTIAL offset adjustment must be undone.
+        */
+       if ((m->m_pkthdr.csum_flags & (CSUM_PARTIAL|CSUM_DATA_VALID)) ==
+           (CSUM_PARTIAL|CSUM_DATA_VALID)) {
+               /*
+                * Get back the original ULP offset (this will
+                * undo the CSUM_PARTIAL logic in ip6_output.)
+                */
+               m->m_pkthdr.csum_data = (m->m_pkthdr.csum_tx_stuff -
+                   m->m_pkthdr.csum_tx_start);
+       }
+
+       ulpoff = (m->m_pkthdr.csum_data & 0xffff);      /* ULP csum offset */
+
+       if (mlen < (ulpoff + sizeof (csum))) {
+               panic("%s: mbuf %p pkt len (%u) proto %d invalid ULP "
+                   "cksum offset (%u) cksum flags 0x%x\n", __func__,
+                   m, mlen, nxt, ulpoff, m->m_pkthdr.csum_flags);
+               /* NOTREACHED */
+       }
+
+       csum = inet6_cksum(m, 0, offset, plen - olen);
+
+       /* Update stats */
+       ip6_out_cksum_stats(nxt, plen - olen);
+
+       /* RFC1122 4.1.3.4 */
+       if (csum == 0 &&
+           (m->m_pkthdr.csum_flags & (CSUM_UDPIPV6|CSUM_ZERO_INVERT)))
+               csum = 0xffff;
+
+       /* Insert the checksum in the ULP csum field */
+       offset += ulpoff;
+       if ((offset + sizeof (csum)) > m->m_len) {
+               m_copyback(m, offset, sizeof (csum), &csum);
+       } else if (IP6_HDR_ALIGNED_P(mtod(m, char *) + hoff)) {
+               *(uint16_t *)(void *)(mtod(m, char *) + offset) = csum;
+       } else {
+               bcopy(&csum, (mtod(m, char *) + offset), sizeof (csum));
+       }
+       m->m_pkthdr.csum_flags &= ~(CSUM_DELAY_IPV6_DATA | CSUM_DATA_VALID |
+           CSUM_PARTIAL | CSUM_ZERO_INVERT);
+
+done:
+       return (sw_csum);
 }
 
 /*
  * Insert jumbo payload option.
  */
 static int
 }
 
 /*
  * Insert jumbo payload option.
  */
 static int
-ip6_insert_jumboopt(exthdrs, plen)
-       struct ip6_exthdrs *exthdrs;
-       u_int32_t plen;
+ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
 {
        struct mbuf *mopt;
        u_char *optbuf;
        u_int32_t v;
 
 {
        struct mbuf *mopt;
        u_char *optbuf;
        u_int32_t v;
 
-#define JUMBOOPTLEN    8       /* length of jumbo payload option and padding */
+#define        JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
 
        /*
         * If there is no hop-by-hop options header, allocate new one.
 
        /*
         * If there is no hop-by-hop options header, allocate new one.
@@ -1220,10 +2101,10 @@ ip6_insert_jumboopt(exthdrs, plen)
         * jumbo payload option, allocate a cluster to store the whole options.
         * Otherwise, use it to store the options.
         */
         * jumbo payload option, allocate a cluster to store the whole options.
         * Otherwise, use it to store the options.
         */
-       if (exthdrs->ip6e_hbh == 0) {
+       if (exthdrs->ip6e_hbh == NULL) {
                MGET(mopt, M_DONTWAIT, MT_DATA);
                MGET(mopt, M_DONTWAIT, MT_DATA);
-               if (mopt == 0)
-                       return(ENOBUFS);
+               if (mopt == NULL)
+                       return (ENOBUFS);
                mopt->m_len = JUMBOOPTLEN;
                optbuf = mtod(mopt, u_char *);
                optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
                mopt->m_len = JUMBOOPTLEN;
                optbuf = mtod(mopt, u_char *);
                optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
@@ -1239,7 +2120,7 @@ ip6_insert_jumboopt(exthdrs, plen)
                         *   other than exthdrs.
                         * - exthdrs->ip6e_hbh is not an mbuf chain.
                         */
                         *   other than exthdrs.
                         * - exthdrs->ip6e_hbh is not an mbuf chain.
                         */
-                       int oldoptlen = mopt->m_len;
+                       u_int32_t oldoptlen = mopt->m_len;
                        struct mbuf *n;
 
                        /*
                        struct mbuf *n;
 
                        /*
@@ -1247,26 +2128,26 @@ ip6_insert_jumboopt(exthdrs, plen)
                         * not fit even in an mbuf cluster.
                         */
                        if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
                         * not fit even in an mbuf cluster.
                         */
                        if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
-                               return(ENOBUFS);
+                               return (ENOBUFS);
 
                        /*
                         * As a consequence, we must always prepare a cluster
                         * at this point.
                         */
                        MGET(n, M_DONTWAIT, MT_DATA);
 
                        /*
                         * As a consequence, we must always prepare a cluster
                         * at this point.
                         */
                        MGET(n, M_DONTWAIT, MT_DATA);
-                       if (n) {
+                       if (n != NULL) {
                                MCLGET(n, M_DONTWAIT);
                                MCLGET(n, M_DONTWAIT);
-                               if ((n->m_flags & M_EXT) == 0) {
+                               if (!(n->m_flags & M_EXT)) {
                                        m_freem(n);
                                        n = NULL;
                                }
                        }
                                        m_freem(n);
                                        n = NULL;
                                }
                        }
-                       if (!n)
-                               return(ENOBUFS);
+                       if (n == NULL)
+                               return (ENOBUFS);
                        n->m_len = oldoptlen + JUMBOOPTLEN;
                        bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
                        n->m_len = oldoptlen + JUMBOOPTLEN;
                        bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
-                             oldoptlen);
-                       optbuf = mtod(n, caddr_t) + oldoptlen;
+                           oldoptlen);
+                       optbuf = mtod(n, u_char *) + oldoptlen;
                        m_freem(mopt);
                        mopt = exthdrs->ip6e_hbh = n;
                } else {
                        m_freem(mopt);
                        mopt = exthdrs->ip6e_hbh = n;
                } else {
@@ -1288,12 +2169,12 @@ ip6_insert_jumboopt(exthdrs, plen)
        optbuf[2] = IP6OPT_JUMBO;
        optbuf[3] = 4;
        v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
        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;
 
 
        /* finally, adjust the packet header length */
        exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
 
-       return(0);
+       return (0);
 #undef JUMBOOPTLEN
 }
 
 #undef JUMBOOPTLEN
 }
 
@@ -1301,18 +2182,16 @@ ip6_insert_jumboopt(exthdrs, plen)
  * Insert fragment header and copy unfragmentable header portions.
  */
 static int
  * Insert fragment header and copy unfragmentable header portions.
  */
 static int
-ip6_insertfraghdr(m0, m, hlen, frghdrp)
-       struct mbuf *m0, *m;
-       int hlen;
-       struct ip6_frag **frghdrp;
+ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
+    struct ip6_frag **frghdrp)
 {
        struct mbuf *n, *mlast;
 
 {
        struct mbuf *n, *mlast;
 
-       if (hlen > sizeof(struct ip6_hdr)) {
-               n = m_copym(m0, sizeof(struct ip6_hdr),
-                           hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
-               if (n == 0)
-                       return(ENOBUFS);
+       if (hlen > sizeof (struct ip6_hdr)) {
+               n = m_copym(m0, sizeof (struct ip6_hdr),
+                   hlen - sizeof (struct ip6_hdr), M_DONTWAIT);
+               if (n == NULL)
+                       return (ENOBUFS);
                m->m_next = n;
        } else
                n = m;
                m->m_next = n;
        } else
                n = m;
@@ -1321,80 +2200,158 @@ ip6_insertfraghdr(m0, m, hlen, frghdrp)
        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) == 0 &&
-           M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
-               /* use the trailing space of the last mbuf for the fragment hdr */
-               *frghdrp =
-                       (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
-               mlast->m_len += sizeof(struct ip6_frag);
-               m->m_pkthdr.len += sizeof(struct ip6_frag);
+       if (!(mlast->m_flags & M_EXT) &&
+           M_TRAILINGSPACE(mlast) >= sizeof (struct ip6_frag)) {
+               /* use the trailing space of the last mbuf for the frag hdr */
+               *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
+                   mlast->m_len);
+               mlast->m_len += sizeof (struct ip6_frag);
+               m->m_pkthdr.len += sizeof (struct ip6_frag);
        } else {
                /* allocate a new mbuf for the fragment header */
                struct mbuf *mfrg;
 
                MGET(mfrg, M_DONTWAIT, MT_DATA);
        } else {
                /* allocate a new mbuf for the fragment header */
                struct mbuf *mfrg;
 
                MGET(mfrg, M_DONTWAIT, MT_DATA);
-               if (mfrg == 0)
-                       return(ENOBUFS);
-               mfrg->m_len = sizeof(struct ip6_frag);
+               if (mfrg == NULL)
+                       return (ENOBUFS);
+               mfrg->m_len = sizeof (struct ip6_frag);
                *frghdrp = mtod(mfrg, struct ip6_frag *);
                mlast->m_next = mfrg;
        }
 
                *frghdrp = mtod(mfrg, struct ip6_frag *);
                mlast->m_next = mfrg;
        }
 
-       return(0);
+       return (0);
 }
 
 }
 
-extern int load_ipfw();
+static int
+ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
+    struct ifnet *ifp, struct in6_addr *dst, u_int32_t *mtup,
+    boolean_t *alwaysfragp)
+{
+       u_int32_t mtu = 0;
+       boolean_t alwaysfrag = FALSE;
+       int error = 0;
+       boolean_t is_local = FALSE;
+
+       if (IN6_IS_SCOPE_LINKLOCAL(dst))
+               is_local = TRUE;
+
+       if (ro_pmtu != ro) {
+               /* The first hop and the final destination may differ. */
+               struct sockaddr_in6 *sa6_dst = SIN6(&ro_pmtu->ro_dst);
+               if (ROUTE_UNUSABLE(ro_pmtu) ||
+                   !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
+                       ROUTE_RELEASE(ro_pmtu);
+
+               if (ro_pmtu->ro_rt == NULL) {
+                       bzero(sa6_dst, sizeof (*sa6_dst));
+                       sa6_dst->sin6_family = AF_INET6;
+                       sa6_dst->sin6_len = sizeof (struct sockaddr_in6);
+                       sa6_dst->sin6_addr = *dst;
+
+                       rtalloc_scoped((struct route *)ro_pmtu,
+                           ifp != NULL ? ifp->if_index : IFSCOPE_NONE);
+               }
+       }
+
+       if (ro_pmtu->ro_rt != NULL) {
+               u_int32_t ifmtu;
+
+               if (ifp == NULL)
+                       ifp = ro_pmtu->ro_rt->rt_ifp;
+               /* Access without acquiring nd_ifinfo lock for performance */
+               ifmtu = IN6_LINKMTU(ifp);
+
+               /*
+                * Access rmx_mtu without holding the route entry lock,
+                * for performance; this isn't something that changes
+                * often, so optimize.
+                */
+               mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
+               if (mtu > ifmtu || mtu == 0) {
+                       /*
+                        * The MTU on the route is larger than the MTU on
+                        * the interface!  This shouldn't happen, unless the
+                        * MTU of the interface has been changed after the
+                        * interface was brought up.  Change the MTU in the
+                        * route to match the interface MTU (as long as the
+                        * field isn't locked).
+                        *
+                        * if MTU on the route is 0, we need to fix the MTU.
+                        * this case happens with path MTU discovery timeouts.
+                        */
+                       mtu = ifmtu;
+                       if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU))
+                               ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
+               } else if (mtu < IPV6_MMTU) {
+                       /*
+                        * RFC2460 section 5, last paragraph:
+                        * if we record ICMPv6 too big message with
+                        * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
+                        * or smaller, with framgent header attached.
+                        * (fragment header is needed regardless from the
+                        * packet size, for translators to identify packets)
+                        */
+                       alwaysfrag = TRUE;
+                       mtu = IPV6_MMTU;
+               }
+       } else {
+               if (ifp) {
+                       /* Don't hold nd_ifinfo lock for performance */
+                       mtu = IN6_LINKMTU(ifp);
+               } else {
+                       error = EHOSTUNREACH; /* XXX */
+               }
+       }
+
+       *mtup = mtu;
+       if ((alwaysfragp != NULL) && !is_local)
+               *alwaysfragp = alwaysfrag;
+       return (error);
+}
 
 /*
  * IP6 socket option processing.
  */
 int
 
 /*
  * IP6 socket option processing.
  */
 int
-ip6_ctloutput(so, sopt)
-       struct socket *so;
-       struct sockopt *sopt;
+ip6_ctloutput(struct socket *so, struct sockopt *sopt)
 {
 {
+       int optdatalen, uproto;
+       void *optdata;
        int privileged;
        struct inpcb *in6p = sotoinpcb(so);
        int privileged;
        struct inpcb *in6p = sotoinpcb(so);
-       int error, optval;
-       int level, op, optname;
-       int optlen;
+       int error = 0, optval = 0;
+       int level, op = -1, optname = 0;
+       int optlen = 0;
        struct proc *p;
 
        struct proc *p;
 
-       level = error = optval = 0;
-       if (sopt == NULL) 
-               panic("ip6_ctloutput: arg soopt is NULL");
-       else {
-               level = sopt->sopt_level;
-               op = sopt->sopt_dir;
-               optname = sopt->sopt_name;
-               optlen = sopt->sopt_valsize;
-               p = sopt->sopt_p;
-       }
+       VERIFY(sopt != NULL);
+
+       level = sopt->sopt_level;
+       op = sopt->sopt_dir;
+       optname = sopt->sopt_name;
+       optlen = sopt->sopt_valsize;
+       p = sopt->sopt_p;
+       uproto = (int)SOCK_PROTO(so);
 
 
-       privileged = (p == 0 || proc_suser(p)) ? 0 : 1;
+       privileged = (proc_suser(p) == 0);
 
        if (level == IPPROTO_IPV6) {
 
        if (level == IPPROTO_IPV6) {
+               boolean_t capture_exthdrstat_in = FALSE;
                switch (op) {
                switch (op) {
-
                case SOPT_SET:
                        switch (optname) {
                case SOPT_SET:
                        switch (optname) {
-                       case IPV6_PKTOPTIONS:
-                       {
+                       case IPV6_2292PKTOPTIONS: {
                                struct mbuf *m;
 
                                struct mbuf *m;
 
-                               if (sopt->sopt_valsize > MCLBYTES) {
-                                       error = EMSGSIZE;
-                                       break;
-                               }
-                               error = soopt_getm(sopt, &m); /* XXX */
-                               if (error != NULL)
+                               error = soopt_getm(sopt, &m);
+                               if (error != 0)
                                        break;
                                        break;
-                               error = soopt_mcopyin(sopt, m); /* XXX */
-                               if (error != NULL)
+                               error = soopt_mcopyin(sopt, m);
+                               if (error != 0)
                                        break;
                                error = ip6_pcbopts(&in6p->in6p_outputopts,
                                        break;
                                error = ip6_pcbopts(&in6p->in6p_outputopts,
-                                                   m, so, sopt);
-                               m_freem(m); /* XXX */
+                                   m, so, sopt);
+                               m_freem(m);
                                break;
                        }
 
                                break;
                        }
 
@@ -1411,48 +2368,143 @@ ip6_ctloutput(so, sopt)
                         *  receiving ANY hbh/dst options in order to avoid
                         *  overhead of parsing options in the kernel.
                         */
                         *  receiving ANY hbh/dst options in order to avoid
                         *  overhead of parsing options in the kernel.
                         */
+                       case IPV6_RECVHOPOPTS:
+                       case IPV6_RECVDSTOPTS:
+                       case IPV6_RECVRTHDRDSTOPTS:
+                               if (!privileged)
+                                       break;
+                               /* FALLTHROUGH */
                        case IPV6_UNICAST_HOPS:
                        case IPV6_UNICAST_HOPS:
-                       case IPV6_CHECKSUM:
-                       case IPV6_FAITH:
-
+                       case IPV6_HOPLIMIT:
+                       case IPV6_RECVPKTINFO:
+                       case IPV6_RECVHOPLIMIT:
+                       case IPV6_RECVRTHDR:
+                       case IPV6_RECVPATHMTU:
+                       case IPV6_RECVTCLASS:
                        case IPV6_V6ONLY:
                        case IPV6_V6ONLY:
-                               if (optlen != sizeof(int)) {
+                       case IPV6_AUTOFLOWLABEL:
+                               if (optlen != sizeof (int)) {
                                        error = EINVAL;
                                        break;
                                }
                                error = sooptcopyin(sopt, &optval,
                                        error = EINVAL;
                                        break;
                                }
                                error = sooptcopyin(sopt, &optval,
-                                       sizeof optval, sizeof optval);
+                                   sizeof (optval), sizeof (optval));
                                if (error)
                                        break;
                                if (error)
                                        break;
-                               switch (optname) {
 
 
+                               switch (optname) {
                                case IPV6_UNICAST_HOPS:
                                case IPV6_UNICAST_HOPS:
-                                       if (optval < -1 || optval >= 256)
+                                       if (optval < -1 || optval >= 256) {
                                                error = EINVAL;
                                                error = EINVAL;
-                                       else {
+                                       else {
                                                /* -1 = kernel default */
                                                in6p->in6p_hops = optval;
                                                /* -1 = kernel default */
                                                in6p->in6p_hops = optval;
-
-                                               if ((in6p->in6p_vflag &
-                                                    INP_IPV4) != 0)
-                                                       in6p->inp_ip_ttl = optval;
+                                               if (in6p->inp_vflag &
+                                                   INP_IPV4) {
+                                                       in6p->inp_ip_ttl =
+                                                           optval;
+                                               }
                                        }
                                        break;
                                        }
                                        break;
-#define OPTSET(bit) \
-do { \
-       if (optval) \
-               in6p->in6p_flags |= (bit); \
-       else \
-               in6p->in6p_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);                              \
 } while (0)
 } while (0)
-#define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
 
 
-                               case IPV6_CHECKSUM:
-                                       in6p->in6p_cksum = optval;
+#define        OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
+
+                               case IPV6_RECVPKTINFO:
+                                       /* cannot mix with RFC2292 */
+                                       if (OPTBIT(IN6P_RFC2292)) {
+                                               error = EINVAL;
+                                               break;
+                                       }
+                                       OPTSET(IN6P_PKTINFO);
+                                       break;
+
+                               case IPV6_HOPLIMIT: {
+                                       struct ip6_pktopts **optp;
+
+                                       /* cannot mix with RFC2292 */
+                                       if (OPTBIT(IN6P_RFC2292)) {
+                                               error = EINVAL;
+                                               break;
+                                       }
+                                       optp = &in6p->in6p_outputopts;
+                                       error = ip6_pcbopt(IPV6_HOPLIMIT,
+                                           (u_char *)&optval, sizeof (optval),
+                                           optp, uproto);
+                                       break;
+                               }
+
+                               case IPV6_RECVHOPLIMIT:
+                                       /* cannot mix with RFC2292 */
+                                       if (OPTBIT(IN6P_RFC2292)) {
+                                               error = EINVAL;
+                                               break;
+                                       }
+                                       OPTSET(IN6P_HOPLIMIT);
+                                       break;
+
+                               case IPV6_RECVHOPOPTS:
+                                       /* cannot mix with RFC2292 */
+                                       if (OPTBIT(IN6P_RFC2292)) {
+                                               error = EINVAL;
+                                               break;
+                                       }
+                                       OPTSET(IN6P_HOPOPTS);
+                                       capture_exthdrstat_in = TRUE;
+                                       break;
+
+                               case IPV6_RECVDSTOPTS:
+                                       /* cannot mix with RFC2292 */
+                                       if (OPTBIT(IN6P_RFC2292)) {
+                                               error = EINVAL;
+                                               break;
+                                       }
+                                       OPTSET(IN6P_DSTOPTS);
+                                       capture_exthdrstat_in = TRUE;
+                                       break;
+
+                               case IPV6_RECVRTHDRDSTOPTS:
+                                       /* cannot mix with RFC2292 */
+                                       if (OPTBIT(IN6P_RFC2292)) {
+                                               error = EINVAL;
+                                               break;
+                                       }
+                                       OPTSET(IN6P_RTHDRDSTOPTS);
+                                       capture_exthdrstat_in = TRUE;
+                                       break;
+
+                               case IPV6_RECVRTHDR:
+                                       /* cannot mix with RFC2292 */
+                                       if (OPTBIT(IN6P_RFC2292)) {
+                                               error = EINVAL;
+                                               break;
+                                       }
+                                       OPTSET(IN6P_RTHDR);
+                                       capture_exthdrstat_in = TRUE;
                                        break;
 
                                        break;
 
-                               case IPV6_FAITH:
-                                       OPTSET(IN6P_FAITH);
+                               case IPV6_RECVPATHMTU:
+                                       /*
+                                        * We ignore this option for TCP
+                                        * sockets.
+                                        * (RFC3542 leaves this case
+                                        * unspecified.)
+                                        */
+                                       if (uproto != IPPROTO_TCP)
+                                               OPTSET(IN6P_MTU);
                                        break;
 
                                case IPV6_V6ONLY:
                                        break;
 
                                case IPV6_V6ONLY:
@@ -1461,108 +2513,168 @@ do { \
                                         * available only prior to bind(2).
                                         * see ipng mailing list, Jun 22 2001.
                                         */
                                         * available only prior to bind(2).
                                         * see ipng mailing list, Jun 22 2001.
                                         */
-                                       if (in6p->in6p_lport ||
-                                           !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
-                                       {
+                                       if (in6p->inp_lport ||
+                                           !IN6_IS_ADDR_UNSPECIFIED(
+                                           &in6p->in6p_laddr)) {
                                                error = EINVAL;
                                                break;
                                        }
                                        OPTSET(IN6P_IPV6_V6ONLY);
                                        if (optval)
                                                error = EINVAL;
                                                break;
                                        }
                                        OPTSET(IN6P_IPV6_V6ONLY);
                                        if (optval)
-                                               in6p->in6p_vflag &= ~INP_IPV4;
+                                               in6p->inp_vflag &= ~INP_IPV4;
                                        else
                                        else
-                                               in6p->in6p_vflag |= INP_IPV4;
+                                               in6p->inp_vflag |= INP_IPV4;
+                                       break;
+
+                               case IPV6_RECVTCLASS:
+                                       /* we can mix with RFC2292 */
+                                       OPTSET(IN6P_TCLASS);
+                                       break;
+
+                               case IPV6_AUTOFLOWLABEL:
+                                       OPTSET(IN6P_AUTOFLOWLABEL);
                                        break;
                                        break;
+
                                }
                                break;
 
                                }
                                break;
 
-                       case IPV6_PKTINFO:
-                       case IPV6_HOPLIMIT:
-                       case IPV6_HOPOPTS:
-                       case IPV6_DSTOPTS:
-                       case IPV6_RTHDR:
+                       case IPV6_TCLASS:
+                       case IPV6_DONTFRAG:
+                       case IPV6_USE_MIN_MTU:
+                       case IPV6_PREFER_TEMPADDR: {
+                               struct ip6_pktopts **optp;
+
+                               if (optlen != sizeof (optval)) {
+                                       error = EINVAL;
+                                       break;
+                               }
+                               error = sooptcopyin(sopt, &optval,
+                                   sizeof (optval), sizeof (optval));
+                               if (error)
+                                       break;
+
+                               optp = &in6p->in6p_outputopts;
+                               error = ip6_pcbopt(optname, (u_char *)&optval,
+                                   sizeof (optval), optp, uproto);
+
+                               if (optname == IPV6_TCLASS) {
+                                       // Add in the ECN flags
+                                       u_int8_t tos = (in6p->inp_ip_tos & ~IPTOS_ECN_MASK);
+                                       u_int8_t ecn = optval & IPTOS_ECN_MASK;
+                                       in6p->inp_ip_tos = tos | ecn;
+                               }
+                               break;
+                       }
+
+                       case IPV6_2292PKTINFO:
+                       case IPV6_2292HOPLIMIT:
+                       case IPV6_2292HOPOPTS:
+                       case IPV6_2292DSTOPTS:
+                       case IPV6_2292RTHDR:
                                /* RFC 2292 */
                                /* RFC 2292 */
-                               if (optlen != sizeof(int)) {
+                               if (optlen != sizeof (int)) {
                                        error = EINVAL;
                                        break;
                                }
                                error = sooptcopyin(sopt, &optval,
                                        error = EINVAL;
                                        break;
                                }
                                error = sooptcopyin(sopt, &optval,
-                                       sizeof optval, sizeof optval);
+                                   sizeof (optval), sizeof (optval));
                                if (error)
                                        break;
                                switch (optname) {
                                if (error)
                                        break;
                                switch (optname) {
-                               case IPV6_PKTINFO:
-                                       OPTSET(IN6P_PKTINFO);
+                               case IPV6_2292PKTINFO:
+                                       OPTSET2292(IN6P_PKTINFO);
                                        break;
                                        break;
-                               case IPV6_HOPLIMIT:
-                                       OPTSET(IN6P_HOPLIMIT);
+                               case IPV6_2292HOPLIMIT:
+                                       OPTSET2292(IN6P_HOPLIMIT);
                                        break;
                                        break;
-                               case IPV6_HOPOPTS:
+                               case IPV6_2292HOPOPTS:
                                        /*
                                         * Check super-user privilege.
                                         * See comments for IPV6_RECVHOPOPTS.
                                         */
                                        if (!privileged)
                                        /*
                                         * Check super-user privilege.
                                         * See comments for IPV6_RECVHOPOPTS.
                                         */
                                        if (!privileged)
-                                               return(EPERM);
-                                       OPTSET(IN6P_HOPOPTS);
+                                               return (EPERM);
+                                       OPTSET2292(IN6P_HOPOPTS);
+                                       capture_exthdrstat_in = TRUE;
                                        break;
                                        break;
-                               case IPV6_DSTOPTS:
+                               case IPV6_2292DSTOPTS:
                                        if (!privileged)
                                        if (!privileged)
-                                               return(EPERM);
-                                       OPTSET(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
+                                               return (EPERM);
+                                       OPTSET2292(IN6P_DSTOPTS|
+                                           IN6P_RTHDRDSTOPTS); /* XXX */
+                                       capture_exthdrstat_in = TRUE;
                                        break;
                                        break;
-                               case IPV6_RTHDR:
-                                       OPTSET(IN6P_RTHDR);
+                               case IPV6_2292RTHDR:
+                                       OPTSET2292(IN6P_RTHDR);
+                                       capture_exthdrstat_in = TRUE;
                                        break;
                                }
                                break;
                                        break;
                                }
                                break;
-#undef OPTSET
 
 
+                       case IPV6_3542PKTINFO:
+                       case IPV6_3542HOPOPTS:
+                       case IPV6_3542RTHDR:
+                       case IPV6_3542DSTOPTS:
+                       case IPV6_RTHDRDSTOPTS:
+                       case IPV6_3542NEXTHOP: {
+                               struct ip6_pktopts **optp;
+                               /* new advanced API (RFC3542) */
+                               struct mbuf *m;
+
+                               /* cannot mix with RFC2292 */
+                               if (OPTBIT(IN6P_RFC2292)) {
+                                       error = EINVAL;
+                                       break;
+                               }
+                               error = soopt_getm(sopt, &m);
+                               if (error != 0)
+                                       break;
+                               error = soopt_mcopyin(sopt, m);
+                               if (error != 0)
+                                       break;
+
+                               optp = &in6p->in6p_outputopts;
+                               error = ip6_pcbopt(optname, mtod(m, u_char *),
+                                   m->m_len, optp, uproto);
+                               m_freem(m);
+                               break;
+                       }
+#undef OPTSET
                        case IPV6_MULTICAST_IF:
                        case IPV6_MULTICAST_HOPS:
                        case IPV6_MULTICAST_LOOP:
                        case IPV6_JOIN_GROUP:
                        case IPV6_LEAVE_GROUP:
                        case IPV6_MULTICAST_IF:
                        case IPV6_MULTICAST_HOPS:
                        case IPV6_MULTICAST_LOOP:
                        case IPV6_JOIN_GROUP:
                        case IPV6_LEAVE_GROUP:
-                           {
-                               struct mbuf *m;
-                               if (sopt->sopt_valsize > MLEN) {
-                                       error = EMSGSIZE;
-                                       break;
-                               }
-                               /* XXX */
-                               MGET(m, sopt->sopt_p ? M_WAIT : M_DONTWAIT, MT_HEADER);
-                               if (m == 0) {
-                                       error = ENOBUFS;
-                                       break;
-                               }
-                               m->m_len = sopt->sopt_valsize;
-                               error = sooptcopyin(sopt, mtod(m, char *),
-                                                   m->m_len, m->m_len);
-                               error = ip6_setmoptions(sopt->sopt_name, in6p, m);
-                               (void)m_free(m);
-                           }
+                       case IPV6_MSFILTER:
+                       case MCAST_BLOCK_SOURCE:
+                       case MCAST_UNBLOCK_SOURCE:
+                       case MCAST_JOIN_GROUP:
+                       case MCAST_LEAVE_GROUP:
+                       case MCAST_JOIN_SOURCE_GROUP:
+                       case MCAST_LEAVE_SOURCE_GROUP:
+                               error = ip6_setmoptions(in6p, sopt);
                                break;
 
                        case IPV6_PORTRANGE:
                                error = sooptcopyin(sopt, &optval,
                                break;
 
                        case IPV6_PORTRANGE:
                                error = sooptcopyin(sopt, &optval,
-                                   sizeof optval, sizeof optval);
+                                   sizeof (optval), sizeof (optval));
                                if (error)
                                        break;
 
                                switch (optval) {
                                case IPV6_PORTRANGE_DEFAULT:
                                if (error)
                                        break;
 
                                switch (optval) {
                                case IPV6_PORTRANGE_DEFAULT:
-                                       in6p->in6p_flags &= ~(IN6P_LOWPORT);
-                                       in6p->in6p_flags &= ~(IN6P_HIGHPORT);
+                                       in6p->inp_flags &= ~(INP_LOWPORT);
+                                       in6p->inp_flags &= ~(INP_HIGHPORT);
                                        break;
 
                                case IPV6_PORTRANGE_HIGH:
                                        break;
 
                                case IPV6_PORTRANGE_HIGH:
-                                       in6p->in6p_flags &= ~(IN6P_LOWPORT);
-                                       in6p->in6p_flags |= IN6P_HIGHPORT;
+                                       in6p->inp_flags &= ~(INP_LOWPORT);
+                                       in6p->inp_flags |= INP_HIGHPORT;
                                        break;
 
                                case IPV6_PORTRANGE_LOW:
                                        break;
 
                                case IPV6_PORTRANGE_LOW:
-                                       in6p->in6p_flags &= ~(IN6P_HIGHPORT);
-                                       in6p->in6p_flags |= IN6P_LOWPORT;
+                                       in6p->inp_flags &= ~(INP_HIGHPORT);
+                                       in6p->inp_flags |= INP_LOWPORT;
                                        break;
 
                                default:
                                        break;
 
                                default:
@@ -1570,199 +2682,279 @@ do { \
                                        break;
                                }
                                break;
                                        break;
                                }
                                break;
-
 #if IPSEC
 #if IPSEC
-                       case IPV6_IPSEC_POLICY:
-                           {
+                       case IPV6_IPSEC_POLICY: {
                                caddr_t req = NULL;
                                size_t len = 0;
                                struct mbuf *m;
 
                                caddr_t req = NULL;
                                size_t len = 0;
                                struct mbuf *m;
 
-                                if (sopt->sopt_valsize > MCLBYTES) {
-                                        error = EMSGSIZE;
-                                        break;
-                                }
-                               if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
+                               if ((error = soopt_getm(sopt, &m)) != 0)
                                        break;
                                        break;
-                               if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
+                               if ((error = soopt_mcopyin(sopt, m)) != 0)
                                        break;
                                        break;
-                               if (m) {
-                                       req = mtod(m, caddr_t);
-                                       len = m->m_len;
-                               }
-                               lck_mtx_lock(sadb_mutex);
+
+                               req = mtod(m, caddr_t);
+                               len = m->m_len;
                                error = ipsec6_set_policy(in6p, optname, req,
                                error = ipsec6_set_policy(in6p, optname, req,
-                                                         len, privileged);
-                               lck_mtx_unlock(sadb_mutex);
+                                       len, privileged);
                                m_freem(m);
                                m_freem(m);
-                           }
                                break;
                                break;
-#endif /* KAME IPSEC */
+                       }
+#endif /* IPSEC */
+                       /*
+                        * IPv6 variant of IP_BOUND_IF; for details see
+                        * comments on IP_BOUND_IF in ip_ctloutput().
+                        */
+                       case IPV6_BOUND_IF:
+                               /* This option is settable only on IPv6 */
+                               if (!(in6p->inp_vflag & INP_IPV6)) {
+                                       error = EINVAL;
+                                       break;
+                               }
+
+                               error = sooptcopyin(sopt, &optval,
+                                   sizeof (optval), sizeof (optval));
+
+                               if (error)
+                                       break;
+
+                               error = inp_bindif(in6p, optval, NULL);
+                               break;
+
+                       case IPV6_NO_IFT_CELLULAR:
+                               /* This option is settable only for IPv6 */
+                               if (!(in6p->inp_vflag & INP_IPV6)) {
+                                       error = EINVAL;
+                                       break;
+                               }
+
+                               error = sooptcopyin(sopt, &optval,
+                                   sizeof (optval), sizeof (optval));
 
 
-                       case IPV6_FW_ADD:
-                       case IPV6_FW_DEL:
-                       case IPV6_FW_FLUSH:
-                       case IPV6_FW_ZERO:
-                               {
-                               if (ip6_fw_ctl_ptr == NULL && load_ipfw() != 0)
-                                       return EINVAL;
+                               if (error)
+                                       break;
 
 
-                               error = (*ip6_fw_ctl_ptr)(sopt);
+                               /* once set, it cannot be unset */
+                               if (!optval && INP_NO_CELLULAR(in6p)) {
+                                       error = EINVAL;
+                                       break;
                                }
                                }
+
+                               error = so_set_restrictions(so,
+                                   SO_RESTRICT_DENY_CELLULAR);
+                               break;
+
+                       case IPV6_OUT_IF:
+                               /* This option is not settable */
+                               error = EINVAL;
                                break;
 
                        default:
                                error = ENOPROTOOPT;
                                break;
                        }
                                break;
 
                        default:
                                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) {
 
                        break;
 
                case SOPT_GET:
                        switch (optname) {
 
-                       case IPV6_PKTOPTIONS:
-                               if (in6p->in6p_options) {
-                                       struct mbuf *m;
-                                       m = m_copym(in6p->in6p_options,
-                                           0, M_COPYALL, M_WAIT);
-                                       error = soopt_mcopyout(sopt, m);
-                                       if (error == 0)
-                                               m_freem(m);
-                               } else
-                                       sopt->sopt_valsize = 0;
+                       case IPV6_2292PKTOPTIONS:
+                               /*
+                                * RFC3542 (effectively) deprecated the
+                                * semantics of the 2292-style pktoptions.
+                                * Since it was not reliable in nature (i.e.,
+                                * applications had to expect the lack of some
+                                * information after all), it would make sense
+                                * to simplify this part by always returning
+                                * empty data.
+                                */
+                               sopt->sopt_valsize = 0;
                                break;
 
                                break;
 
+                       case IPV6_RECVHOPOPTS:
+                       case IPV6_RECVDSTOPTS:
+                       case IPV6_RECVRTHDRDSTOPTS:
                        case IPV6_UNICAST_HOPS:
                        case IPV6_UNICAST_HOPS:
-                       case IPV6_CHECKSUM:
-
-                       case IPV6_FAITH:
+                       case IPV6_RECVPKTINFO:
+                       case IPV6_RECVHOPLIMIT:
+                       case IPV6_RECVRTHDR:
+                       case IPV6_RECVPATHMTU:
                        case IPV6_V6ONLY:
                        case IPV6_PORTRANGE:
                        case IPV6_V6ONLY:
                        case IPV6_PORTRANGE:
+                       case IPV6_RECVTCLASS:
+                       case IPV6_AUTOFLOWLABEL:
                                switch (optname) {
 
                                switch (optname) {
 
+                               case IPV6_RECVHOPOPTS:
+                                       optval = OPTBIT(IN6P_HOPOPTS);
+                                       break;
+
+                               case IPV6_RECVDSTOPTS:
+                                       optval = OPTBIT(IN6P_DSTOPTS);
+                                       break;
+
+                               case IPV6_RECVRTHDRDSTOPTS:
+                                       optval = OPTBIT(IN6P_RTHDRDSTOPTS);
+                                       break;
+
                                case IPV6_UNICAST_HOPS:
                                        optval = in6p->in6p_hops;
                                        break;
 
                                case IPV6_UNICAST_HOPS:
                                        optval = in6p->in6p_hops;
                                        break;
 
-                               case IPV6_CHECKSUM:
-                                       optval = in6p->in6p_cksum;
+                               case IPV6_RECVPKTINFO:
+                                       optval = OPTBIT(IN6P_PKTINFO);
                                        break;
 
                                        break;
 
-                               case IPV6_FAITH:
-                                       optval = OPTBIT(IN6P_FAITH);
+                               case IPV6_RECVHOPLIMIT:
+                                       optval = OPTBIT(IN6P_HOPLIMIT);
                                        break;
 
                                        break;
 
-                               case IPV6_V6ONLY:
-                                       optval = OPTBIT(IN6P_IPV6_V6ONLY);
+                               case IPV6_RECVRTHDR:
+                                       optval = OPTBIT(IN6P_RTHDR);
                                        break;
 
                                        break;
 
-                               case IPV6_PORTRANGE:
-                                   {
-                                       int flags;
-                                       flags = in6p->in6p_flags;
-                                       if (flags & IN6P_HIGHPORT)
+                               case IPV6_RECVPATHMTU:
+                                       optval = OPTBIT(IN6P_MTU);
+                                       break;
+
+                               case IPV6_V6ONLY:
+                                       optval = OPTBIT(IN6P_IPV6_V6ONLY);
+                                       break;
+
+                               case IPV6_PORTRANGE: {
+                                       int flags;
+                                       flags = in6p->inp_flags;
+                                       if (flags & INP_HIGHPORT)
                                                optval = IPV6_PORTRANGE_HIGH;
                                                optval = IPV6_PORTRANGE_HIGH;
-                                       else if (flags & IN6P_LOWPORT)
+                                       else if (flags & INP_LOWPORT)
                                                optval = IPV6_PORTRANGE_LOW;
                                        else
                                                optval = 0;
                                        break;
                                                optval = IPV6_PORTRANGE_LOW;
                                        else
                                                optval = 0;
                                        break;
-                                   }
                                }
                                }
+                               case IPV6_RECVTCLASS:
+                                       optval = OPTBIT(IN6P_TCLASS);
+                                       break;
+
+                               case IPV6_AUTOFLOWLABEL:
+                                       optval = OPTBIT(IN6P_AUTOFLOWLABEL);
+                                       break;
+                               }
+                               if (error)
+                                       break;
                                error = sooptcopyout(sopt, &optval,
                                error = sooptcopyout(sopt, &optval,
-                                       sizeof optval);
+                                   sizeof (optval));
                                break;
 
                                break;
 
-                       case IPV6_PKTINFO:
-                       case IPV6_HOPLIMIT:
-                       case IPV6_HOPOPTS:
-                       case IPV6_RTHDR:
-                       case IPV6_DSTOPTS:
-                               if ((optname == IPV6_HOPOPTS ||
-                                   optname == IPV6_DSTOPTS) &&
-                                   !privileged)
-                                       return(EPERM);
+                       case IPV6_PATHMTU: {
+                               u_int32_t pmtu = 0;
+                               struct ip6_mtuinfo mtuinfo;
+                               struct route_in6 sro;
+
+                               bzero(&sro, sizeof (sro));
+
+                               if (!(so->so_state & SS_ISCONNECTED))
+                                       return (ENOTCONN);
+                               /*
+                                * XXX: we dot not consider the case of source
+                                * routing, or optional information to specify
+                                * the outgoing interface.
+                                */
+                               error = ip6_getpmtu(&sro, NULL, NULL,
+                                   &in6p->in6p_faddr, &pmtu, NULL);
+                               ROUTE_RELEASE(&sro);
+                               if (error)
+                                       break;
+                               if (pmtu > IPV6_MAXPACKET)
+                                       pmtu = IPV6_MAXPACKET;
+
+                               bzero(&mtuinfo, sizeof (mtuinfo));
+                               mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
+                               optdata = (void *)&mtuinfo;
+                               optdatalen = sizeof (mtuinfo);
+                               error = sooptcopyout(sopt, optdata,
+                                   optdatalen);
+                               break;
+                       }
+
+                       case IPV6_2292PKTINFO:
+                       case IPV6_2292HOPLIMIT:
+                       case IPV6_2292HOPOPTS:
+                       case IPV6_2292RTHDR:
+                       case IPV6_2292DSTOPTS:
                                switch (optname) {
                                switch (optname) {
-                               case IPV6_PKTINFO:
+                               case IPV6_2292PKTINFO:
                                        optval = OPTBIT(IN6P_PKTINFO);
                                        break;
                                        optval = OPTBIT(IN6P_PKTINFO);
                                        break;
-                               case IPV6_HOPLIMIT:
+                               case IPV6_2292HOPLIMIT:
                                        optval = OPTBIT(IN6P_HOPLIMIT);
                                        break;
                                        optval = OPTBIT(IN6P_HOPLIMIT);
                                        break;
-                               case IPV6_HOPOPTS:
-                                       if (!privileged)
-                                               return(EPERM);
+                               case IPV6_2292HOPOPTS:
                                        optval = OPTBIT(IN6P_HOPOPTS);
                                        break;
                                        optval = OPTBIT(IN6P_HOPOPTS);
                                        break;
-                               case IPV6_RTHDR:
+                               case IPV6_2292RTHDR:
                                        optval = OPTBIT(IN6P_RTHDR);
                                        break;
                                        optval = OPTBIT(IN6P_RTHDR);
                                        break;
-                               case IPV6_DSTOPTS:
-                                       if (!privileged)
-                                               return(EPERM);
-                                       optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
+                               case IPV6_2292DSTOPTS:
+                                       optval = OPTBIT(IN6P_DSTOPTS|
+                                           IN6P_RTHDRDSTOPTS);
                                        break;
                                }
                                error = sooptcopyout(sopt, &optval,
                                        break;
                                }
                                error = sooptcopyout(sopt, &optval,
-                                       sizeof optval);
+                                   sizeof (optval));
+                               break;
+
+                       case IPV6_PKTINFO:
+                       case IPV6_HOPOPTS:
+                       case IPV6_RTHDR:
+                       case IPV6_DSTOPTS:
+                       case IPV6_RTHDRDSTOPTS:
+                       case IPV6_NEXTHOP:
+                       case IPV6_TCLASS:
+                       case IPV6_DONTFRAG:
+                       case IPV6_USE_MIN_MTU:
+                       case IPV6_PREFER_TEMPADDR:
+                               error = ip6_getpcbopt(in6p->in6p_outputopts,
+                                   optname, sopt);
                                break;
 
                        case IPV6_MULTICAST_IF:
                        case IPV6_MULTICAST_HOPS:
                        case IPV6_MULTICAST_LOOP:
                                break;
 
                        case IPV6_MULTICAST_IF:
                        case IPV6_MULTICAST_HOPS:
                        case IPV6_MULTICAST_LOOP:
-                       case IPV6_JOIN_GROUP:
-                       case IPV6_LEAVE_GROUP:
-                           {
-                               struct mbuf *m;
-                               error = ip6_getmoptions(sopt->sopt_name,
-                                               in6p->in6p_moptions, &m);
-                               if (error == 0)
-                                       error = sooptcopyout(sopt,
-                                               mtod(m, char *), m->m_len);
-                               m_freem(m);
-                           }
+                       case IPV6_MSFILTER:
+                               error = ip6_getmoptions(in6p, sopt);
                                break;
                                break;
-
 #if IPSEC
 #if IPSEC
-                       case IPV6_IPSEC_POLICY:
-                         {
-                               caddr_t req = NULL;
-                               size_t len = 0;
-                               struct mbuf *m = NULL;
-                               struct mbuf **mp = &m;
-
-                                if (sopt->sopt_valsize > MCLBYTES) {
-                                        error = EMSGSIZE;
-                                        break;
-                                }
-                               error = soopt_getm(sopt, &m); /* XXX */
-                               if (error != NULL)
-                                       break;
-                               error = soopt_mcopyin(sopt, m); /* XXX */
-                               if (error != NULL)
-                                       break;
-                               if (m) {
-                                       req = mtod(m, caddr_t);
-                                       len = m->m_len;
-                               }
-                               lck_mtx_lock(sadb_mutex);
-                               error = ipsec6_get_policy(in6p, req, len, mp);
-                               lck_mtx_unlock(sadb_mutex);
-                               if (error == 0)
-                                       error = soopt_mcopyout(sopt, m); /*XXX*/
-                               if (error == 0 && m)
-                                       m_freem(m);
+                       case IPV6_IPSEC_POLICY: {
+                               error = 0; /* This option is no longer supported */
+                               break;
+                       }
+#endif /* IPSEC */
+                       case IPV6_BOUND_IF:
+                               if (in6p->inp_flags & INP_BOUND_IF)
+                                       optval = in6p->inp_boundifp->if_index;
+                               error = sooptcopyout(sopt, &optval,
+                                   sizeof (optval));
                                break;
                                break;
-                         }
-#endif /* KAME IPSEC */
 
 
-                       case IPV6_FW_GET:
-                               {
-                               if (ip6_fw_ctl_ptr == NULL && load_ipfw() != 0)
-                                       return EINVAL;
+                       case IPV6_NO_IFT_CELLULAR:
+                               optval = INP_NO_CELLULAR(in6p) ? 1 : 0;
+                               error = sooptcopyout(sopt, &optval,
+                                   sizeof (optval));
+                               break;
 
 
-                               error = (*ip6_fw_ctl_ptr)(sopt);
-                               }
+                       case IPV6_OUT_IF:
+                               optval = (in6p->in6p_last_outifp != NULL) ?
+                                   in6p->in6p_last_outifp->if_index : 0;
+                               error = sooptcopyout(sopt, &optval,
+                                   sizeof (optval));
                                break;
 
                        default:
                                break;
 
                        default:
@@ -1771,10 +2963,82 @@ do { \
                        }
                        break;
                }
                        }
                        break;
                }
+       } else if (level == IPPROTO_UDP) {
+               error = udp_ctloutput(so, sopt);
        } else {
                error = EINVAL;
        }
        } else {
                error = EINVAL;
        }
-       return(error);
+       return (error);
+}
+
+int
+ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
+{
+       int error = 0, optval, optlen;
+       const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
+       struct inpcb *in6p = sotoinpcb(so);
+       int level, op, optname;
+
+       level = sopt->sopt_level;
+       op = sopt->sopt_dir;
+       optname = sopt->sopt_name;
+       optlen = sopt->sopt_valsize;
+
+       if (level != IPPROTO_IPV6)
+               return (EINVAL);
+
+       switch (optname) {
+       case IPV6_CHECKSUM:
+               /*
+                * For ICMPv6 sockets, no modification allowed for checksum
+                * offset, permit "no change" values to help existing apps.
+                *
+                * RFC3542 says: "An attempt to set IPV6_CHECKSUM
+                * for an ICMPv6 socket will fail."
+                * The current behavior does not meet RFC3542.
+                */
+               switch (op) {
+               case SOPT_SET:
+                       if (optlen != sizeof (int)) {
+                               error = EINVAL;
+                               break;
+                       }
+                       error = sooptcopyin(sopt, &optval, sizeof (optval),
+                           sizeof (optval));
+                       if (error)
+                               break;
+                       if ((optval % 2) != 0) {
+                               /* the API assumes even offset values */
+                               error = EINVAL;
+                       } else if (SOCK_PROTO(so) == IPPROTO_ICMPV6) {
+                               if (optval != icmp6off)
+                                       error = EINVAL;
+                       } else {
+                               in6p->in6p_cksum = optval;
+                       }
+                       break;
+
+               case SOPT_GET:
+                       if (SOCK_PROTO(so) == IPPROTO_ICMPV6)
+                               optval = icmp6off;
+                       else
+                               optval = in6p->in6p_cksum;
+
+                       error = sooptcopyout(sopt, &optval, sizeof (optval));
+                       break;
+
+               default:
+                       error = EINVAL;
+                       break;
+               }
+               break;
+
+       default:
+               error = ENOPROTOOPT;
+               break;
+       }
+
+       return (error);
 }
 
 /*
 }
 
 /*
@@ -1782,50 +3046,48 @@ do { \
  * specifying behavior of outgoing packets.
  */
 static int
  * specifying behavior of outgoing packets.
  */
 static int
-ip6_pcbopts(pktopt, m, so, sopt)
-       struct ip6_pktopts **pktopt;
-       struct mbuf *m;
-       struct socket *so;
-       struct sockopt *sopt;
+ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so,
+    struct sockopt *sopt)
 {
 {
+#pragma unused(sopt)
        struct ip6_pktopts *opt = *pktopt;
        int error = 0;
        struct ip6_pktopts *opt = *pktopt;
        int error = 0;
-       struct proc *p = sopt->sopt_p;
-       int priv = 0;
 
        /* turn off any old options. */
 
        /* turn off any old options. */
-       if (opt) {
+       if (opt != NULL) {
 #if DIAGNOSTIC
                if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
                    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
                    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
 #if DIAGNOSTIC
                if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
                    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
                    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
-                       printf("ip6_pcbopts: all specified options are cleared.\n");
+                       printf("%s: all specified options are cleared.\n",
+                           __func__);
 #endif
 #endif
-               ip6_clearpktopts(opt, 1, -1);
-       } else
-               opt = _MALLOC(sizeof(*opt), M_IP6OPT, M_WAITOK);
+               ip6_clearpktopts(opt, -1);
+       } else {
+               opt = _MALLOC(sizeof (*opt), M_IP6OPT, M_WAITOK);
+               if (opt == NULL)
+                       return (ENOBUFS);
+       }
        *pktopt = NULL;
 
        *pktopt = NULL;
 
-       if (!m || m->m_len == 0) {
+       if (m == NULL || m->m_len == 0) {
                /*
                 * Only turning off any previous options, regardless of
                 * whether the opt is just created or given.
                 */
                /*
                 * Only turning off any previous options, regardless of
                 * whether the opt is just created or given.
                 */
-               if (opt)
+               if (opt != NULL)
                        FREE(opt, M_IP6OPT);
                        FREE(opt, M_IP6OPT);
-               return(0);
+               return (0);
        }
 
        /*  set options specified by user. */
        }
 
        /*  set options specified by user. */
-       if (p && !proc_suser(p))
-               priv = 1;
-       if ((error = ip6_setpktoptions(m, opt, priv, 1)) != 0) {
-               ip6_clearpktopts(opt, 1, -1); /* XXX: discard all options */
+       if ((error = ip6_setpktopts(m, opt, NULL, SOCK_PROTO(so))) != 0) {
+               ip6_clearpktopts(opt, -1); /* XXX: discard all options */
                FREE(opt, M_IP6OPT);
                FREE(opt, M_IP6OPT);
-               return(error);
+               return (error);
        }
        *pktopt = opt;
        }
        *pktopt = opt;
-       return(0);
+       return (0);
 }
 
 /*
 }
 
 /*
@@ -1833,800 +3095,859 @@ ip6_pcbopts(pktopt, m, so, sopt)
  * the struct.
  */
 void
  * the struct.
  */
 void
-init_ip6pktopts(opt)
-       struct ip6_pktopts *opt;
+ip6_initpktopts(struct ip6_pktopts *opt)
 {
 
 {
 
-       bzero(opt, sizeof(*opt));
+       bzero(opt, sizeof (*opt));
        opt->ip6po_hlim = -1;   /* -1 means default hop limit */
        opt->ip6po_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;
+}
+
+static int
+ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
+    int uproto)
+{
+       struct ip6_pktopts *opt;
+
+       opt = *pktopt;
+       if (opt == NULL) {
+               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));
+}
+
+static int
+ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
+{
+       void *optdata = NULL;
+       int optdatalen = 0;
+       struct ip6_ext *ip6e;
+       struct in6_pktinfo null_pktinfo;
+       int deftclass = 0, on;
+       int defminmtu = IP6PO_MINMTU_MCASTONLY;
+       int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
+
+
+       switch (optname) {
+       case IPV6_PKTINFO:
+               if (pktopt && pktopt->ip6po_pktinfo)
+                       optdata = (void *)pktopt->ip6po_pktinfo;
+               else {
+                       /* XXX: we don't have to do this every time... */
+                       bzero(&null_pktinfo, sizeof (null_pktinfo));
+                       optdata = (void *)&null_pktinfo;
+               }
+               optdatalen = sizeof (struct in6_pktinfo);
+               break;
+
+       case IPV6_TCLASS:
+               if (pktopt && pktopt->ip6po_tclass >= 0)
+                       optdata = (void *)&pktopt->ip6po_tclass;
+               else
+                       optdata = (void *)&deftclass;
+               optdatalen = sizeof (int);
+               break;
+
+       case IPV6_HOPOPTS:
+               if (pktopt && pktopt->ip6po_hbh) {
+                       optdata = (void *)pktopt->ip6po_hbh;
+                       ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
+                       optdatalen = (ip6e->ip6e_len + 1) << 3;
+               }
+               break;
+
+       case IPV6_RTHDR:
+               if (pktopt && pktopt->ip6po_rthdr) {
+                       optdata = (void *)pktopt->ip6po_rthdr;
+                       ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
+                       optdatalen = (ip6e->ip6e_len + 1) << 3;
+               }
+               break;
+
+       case IPV6_RTHDRDSTOPTS:
+               if (pktopt && pktopt->ip6po_dest1) {
+                       optdata = (void *)pktopt->ip6po_dest1;
+                       ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
+                       optdatalen = (ip6e->ip6e_len + 1) << 3;
+               }
+               break;
+
+       case IPV6_DSTOPTS:
+               if (pktopt && pktopt->ip6po_dest2) {
+                       optdata = (void *)pktopt->ip6po_dest2;
+                       ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
+                       optdatalen = (ip6e->ip6e_len + 1) << 3;
+               }
+               break;
+
+       case IPV6_NEXTHOP:
+               if (pktopt && pktopt->ip6po_nexthop) {
+                       optdata = (void *)pktopt->ip6po_nexthop;
+                       optdatalen = pktopt->ip6po_nexthop->sa_len;
+               }
+               break;
+
+       case IPV6_USE_MIN_MTU:
+               if (pktopt)
+                       optdata = (void *)&pktopt->ip6po_minmtu;
+               else
+                       optdata = (void *)&defminmtu;
+               optdatalen = sizeof (int);
+               break;
+
+       case IPV6_DONTFRAG:
+               if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
+                       on = 1;
+               else
+                       on = 0;
+               optdata = (void *)&on;
+               optdatalen = sizeof (on);
+               break;
+
+       case IPV6_PREFER_TEMPADDR:
+               if (pktopt)
+                       optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
+               else
+                       optdata = (void *)&defpreftemp;
+               optdatalen = sizeof (int);
+               break;
+
+       default:                /* should not happen */
+#ifdef DIAGNOSTIC
+               panic("ip6_getpcbopt: unexpected option\n");
+#endif
+               return (ENOPROTOOPT);
+       }
+
+       return (sooptcopyout(sopt, optdata, optdatalen));
 }
 
 void
 }
 
 void
-ip6_clearpktopts(pktopt, needfree, optname)
-       struct ip6_pktopts *pktopt;
-       int needfree, optname;
+ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
 {
        if (pktopt == NULL)
                return;
 
 {
        if (pktopt == NULL)
                return;
 
-       if (optname == -1) {
-               if (needfree && pktopt->ip6po_pktinfo)
+       if (optname == -1 || optname == IPV6_PKTINFO) {
+               if (pktopt->ip6po_pktinfo)
                        FREE(pktopt->ip6po_pktinfo, M_IP6OPT);
                pktopt->ip6po_pktinfo = NULL;
        }
                        FREE(pktopt->ip6po_pktinfo, M_IP6OPT);
                pktopt->ip6po_pktinfo = NULL;
        }
-       if (optname == -1)
+       if (optname == -1 || optname == IPV6_HOPLIMIT)
                pktopt->ip6po_hlim = -1;
                pktopt->ip6po_hlim = -1;
-       if (optname == -1) {
-               if (needfree && pktopt->ip6po_nexthop)
+       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)
                        FREE(pktopt->ip6po_nexthop, M_IP6OPT);
                pktopt->ip6po_nexthop = NULL;
        }
                        FREE(pktopt->ip6po_nexthop, M_IP6OPT);
                pktopt->ip6po_nexthop = NULL;
        }
-       if (optname == -1) {
-               if (needfree && pktopt->ip6po_hbh)
+       if (optname == -1 || optname == IPV6_HOPOPTS) {
+               if (pktopt->ip6po_hbh)
                        FREE(pktopt->ip6po_hbh, M_IP6OPT);
                pktopt->ip6po_hbh = NULL;
        }
                        FREE(pktopt->ip6po_hbh, M_IP6OPT);
                pktopt->ip6po_hbh = NULL;
        }
-       if (optname == -1) {
-               if (needfree && pktopt->ip6po_dest1)
+       if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
+               if (pktopt->ip6po_dest1)
                        FREE(pktopt->ip6po_dest1, M_IP6OPT);
                pktopt->ip6po_dest1 = NULL;
        }
                        FREE(pktopt->ip6po_dest1, M_IP6OPT);
                pktopt->ip6po_dest1 = NULL;
        }
-       if (optname == -1) {
-               if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
+       if (optname == -1 || optname == IPV6_RTHDR) {
+               if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
                        FREE(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
                pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
                        FREE(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
                pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
-               if (pktopt->ip6po_route.ro_rt) {
-                       rtfree(pktopt->ip6po_route.ro_rt);
-                       pktopt->ip6po_route.ro_rt = NULL;
-               }
+               ROUTE_RELEASE(&pktopt->ip6po_route);
        }
        }
-       if (optname == -1) {
-               if (needfree && pktopt->ip6po_dest2)
+       if (optname == -1 || optname == IPV6_DSTOPTS) {
+               if (pktopt->ip6po_dest2)
                        FREE(pktopt->ip6po_dest2, M_IP6OPT);
                pktopt->ip6po_dest2 = NULL;
        }
 }
 
                        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)
 
 } while (0)
 
-struct ip6_pktopts *
-ip6_copypktopts(src, canwait)
-       struct ip6_pktopts *src;
-       int canwait;
+static int
+copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
 {
 {
-       struct ip6_pktopts *dst;
-
-       if (src == NULL) {
-               printf("ip6_clearpktopts: invalid argument\n");
-               return(NULL);
+       if (dst == NULL || src == NULL)  {
+               printf("copypktopts: invalid argument\n");
+               return (EINVAL);
        }
 
        }
 
-       dst = _MALLOC(sizeof(*dst), M_IP6OPT, canwait);
-       if (dst == NULL && canwait == M_NOWAIT)
-               return (NULL);
-       bzero(dst, sizeof(*dst));
-
        dst->ip6po_hlim = src->ip6po_hlim;
        dst->ip6po_hlim = src->ip6po_hlim;
+       dst->ip6po_tclass = src->ip6po_tclass;
+       dst->ip6po_flags = src->ip6po_flags;
        if (src->ip6po_pktinfo) {
        if (src->ip6po_pktinfo) {
-               dst->ip6po_pktinfo = _MALLOC(sizeof(*dst->ip6po_pktinfo),
-                                           M_IP6OPT, canwait);
+               dst->ip6po_pktinfo = _MALLOC(sizeof (*dst->ip6po_pktinfo),
+                   M_IP6OPT, canwait);
                if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
                        goto bad;
                *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
        }
        if (src->ip6po_nexthop) {
                dst->ip6po_nexthop = _MALLOC(src->ip6po_nexthop->sa_len,
                if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
                        goto bad;
                *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
        }
        if (src->ip6po_nexthop) {
                dst->ip6po_nexthop = _MALLOC(src->ip6po_nexthop->sa_len,
-                                           M_IP6OPT, canwait);
+                   M_IP6OPT, canwait);
                if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
                        goto bad;
                bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
                if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
                        goto bad;
                bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
-                     src->ip6po_nexthop->sa_len);
+                   src->ip6po_nexthop->sa_len);
        }
        PKTOPT_EXTHDRCPY(ip6po_hbh);
        PKTOPT_EXTHDRCPY(ip6po_dest1);
        PKTOPT_EXTHDRCPY(ip6po_dest2);
        PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
        }
        PKTOPT_EXTHDRCPY(ip6po_hbh);
        PKTOPT_EXTHDRCPY(ip6po_dest1);
        PKTOPT_EXTHDRCPY(ip6po_dest2);
        PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
-       return(dst);
-
-  bad:
-       if (dst->ip6po_pktinfo) FREE(dst->ip6po_pktinfo, M_IP6OPT);
-       if (dst->ip6po_nexthop) FREE(dst->ip6po_nexthop, M_IP6OPT);
-       if (dst->ip6po_hbh) FREE(dst->ip6po_hbh, M_IP6OPT);
-       if (dst->ip6po_dest1) FREE(dst->ip6po_dest1, M_IP6OPT);
-       if (dst->ip6po_dest2) FREE(dst->ip6po_dest2, M_IP6OPT);
-       if (dst->ip6po_rthdr) FREE(dst->ip6po_rthdr, M_IP6OPT);
-       FREE(dst, M_IP6OPT);
-       return(NULL);
+       return (0);
+
+bad:
+       ip6_clearpktopts(dst, -1);
+       return (ENOBUFS);
 }
 #undef PKTOPT_EXTHDRCPY
 
 }
 #undef PKTOPT_EXTHDRCPY
 
-void
-ip6_freepcbopts(pktopt)
-       struct ip6_pktopts *pktopt;
+struct ip6_pktopts *
+ip6_copypktopts(struct ip6_pktopts *src, int canwait)
 {
 {
-       if (pktopt == NULL)
-               return;
-
-       ip6_clearpktopts(pktopt, 1, -1);
+       int error;
+       struct ip6_pktopts *dst;
 
 
-       FREE(pktopt, M_IP6OPT);
-}
+       dst = _MALLOC(sizeof (*dst), M_IP6OPT, canwait);
+       if (dst == NULL)
+               return (NULL);
+       ip6_initpktopts(dst);
 
 
-/*
- * Set the IP6 multicast options in response to user setsockopt().
- */
-static int
-ip6_setmoptions(
-       int optname,
-       struct inpcb* in6p,
-       struct mbuf *m)
-{
-       int error = 0;
-       u_int loop, ifindex;
-       struct ipv6_mreq *mreq;
-       struct ifnet *ifp;
-       struct ip6_moptions **im6op = &in6p->in6p_moptions;
-       struct ip6_moptions *im6o = *im6op;
-       struct ip_moptions *imo;
-       struct route_in6 ro;
-       struct sockaddr_in6 *dst;
-       struct in6_multi_mship *imm;
-       struct proc *p = current_proc();        /* XXX */
-
-       if (im6o == NULL) {
-               /*
-                * No multicast option buffer attached to the pcb;
-                * allocate one and initialize to default values.
-                */
-               im6o = (struct ip6_moptions *)
-                       _MALLOC(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
-
-               if (im6o == NULL)
-                       return(ENOBUFS);
-               *im6op = im6o;
-               im6o->im6o_multicast_ifp = NULL;
-               im6o->im6o_multicast_hlim = ip6_defmcasthlim;
-               im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
-               LIST_INIT(&im6o->im6o_memberships);
-       }
-       
-       if (in6p->inp_moptions == NULL) {
-               /*
-                * No IPv4 multicast option buffer attached to the pcb;
-                * call ip_createmoptions to allocate one and initialize
-                * to default values.
-                */
-               error = ip_createmoptions(&in6p->inp_moptions);
-               if (error != 0)
-                       return error;
+       if ((error = copypktopts(dst, src, canwait)) != 0) {
+               FREE(dst, M_IP6OPT);
+               return (NULL);
        }
        }
-       imo = in6p->inp_moptions;
-
-       switch (optname) {
-
-       case IPV6_MULTICAST_IF:
-               /*
-                * Select the interface for outgoing multicast packets.
-                */
-               if (m == NULL || m->m_len != sizeof(u_int)) {
-                       error = EINVAL;
-                       break;
-               }
-               bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
-               if (ifindex < 0 || if_index < ifindex) {
-                       error = ENXIO;  /* XXX EINVAL? */
-                       break;
-               }
-               ifp = ifindex2ifnet[ifindex];
-               if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
-                       error = EADDRNOTAVAIL;
-                       break;
-               }
-               im6o->im6o_multicast_ifp = ifp;
-               imo->imo_multicast_ifp = ifp;
-               break;
-
-       case IPV6_MULTICAST_HOPS:
-           {
-               /*
-                * Set the IP6 hoplimit for outgoing multicast packets.
-                */
-               int optval;
-               if (m == NULL || m->m_len != sizeof(int)) {
-                       error = EINVAL;
-                       break;
-               }
-               bcopy(mtod(m, u_int *), &optval, sizeof(optval));
-               if (optval < -1 || optval >= 256)
-                       error = EINVAL;
-               else if (optval == -1) {
-                       im6o->im6o_multicast_hlim = ip6_defmcasthlim;
-                       imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
-               } else {
-                       im6o->im6o_multicast_hlim = optval;
-                       imo->imo_multicast_ttl = optval;
-               }
-               break;
-           }
 
 
-       case IPV6_MULTICAST_LOOP:
-               /*
-                * Set the loopback flag for outgoing multicast packets.
-                * Must be zero or one.
-                */
-               if (m == NULL || m->m_len != sizeof(u_int)) {
-                       error = EINVAL;
-                       break;
-               }
-               bcopy(mtod(m, u_int *), &loop, sizeof(loop));
-               if (loop > 1) {
-                       error = EINVAL;
-                       break;
-               }
-               im6o->im6o_multicast_loop = loop;
-               imo->imo_multicast_loop = loop;
-               break;
-
-       case IPV6_JOIN_GROUP:
-               /*
-                * Add a multicast group membership.
-                * Group must be a valid IP6 multicast address.
-                */
-               if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
-                       error = EINVAL;
-                       break;
-               }
-               mreq = mtod(m, struct ipv6_mreq *);
-               /*
-                * If the interface is specified, validate it.
-                */
-               if (mreq->ipv6mr_interface < 0
-                || if_index < mreq->ipv6mr_interface) {
-                       error = ENXIO;  /* XXX EINVAL? */
-                       break;
-               }
-               
-               if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
-                       /*
-                        * We use the unspecified address to specify to accept
-                        * all multicast addresses. Only super user is allowed
-                        * to do this.
-                        */
-                       if (suser(kauth_cred_get(), 0))
-                       {
-                               error = EACCES;
-                               break;
-                       }
-               } else if (IN6_IS_ADDR_V4MAPPED(&mreq->ipv6mr_multiaddr)) {
-                       struct ip_mreq v4req;
-                       
-                       v4req.imr_multiaddr.s_addr = mreq->ipv6mr_multiaddr.s6_addr32[3];
-                       v4req.imr_interface.s_addr = INADDR_ANY;
-                       
-                       /* Find an IPv4 address on the specified interface. */
-                       if (mreq->ipv6mr_interface != 0) {
-                               struct in_ifaddr *ifa;
-
-                               ifp = ifindex2ifnet[mreq->ipv6mr_interface];
-
-                               lck_mtx_lock(rt_mtx);
-                               TAILQ_FOREACH(ifa, &in_ifaddrhead, ia_link) {
-                                       if (ifa->ia_ifp == ifp) {
-                                               v4req.imr_interface = IA_SIN(ifa)->sin_addr;
-                                               break;
-                                       }
-                               }
-                               lck_mtx_unlock(rt_mtx);
-                               
-                               if (v4req.imr_multiaddr.s_addr == 0) {
-                                       /* Interface has no IPv4 address. */
-                                       error = EINVAL;
-                                       break;
-                               }
-                       }
-                       
-                       error = ip_addmembership(imo, &v4req);
-                       break;
-               } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
-                       error = EINVAL;
-                       break;
-               }
-               /*
-                * If no interface was explicitly specified, choose an
-                * appropriate one according to the given multicast address.
-                */
-               if (mreq->ipv6mr_interface == 0) {
-                       /*
-                        * If the multicast address is in node-local scope,
-                        * the interface should be a loopback interface.
-                        * Otherwise, look up the routing table for the
-                        * address, and choose the outgoing interface.
-                        *   XXX: is it a good approach?
-                        */
-                       if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
-                               ifp = &loif[0];
-                       } else {
-                               ro.ro_rt = NULL;
-                               dst = (struct sockaddr_in6 *)&ro.ro_dst;
-                               bzero(dst, sizeof(*dst));
-                               dst->sin6_len = sizeof(struct sockaddr_in6);
-                               dst->sin6_family = AF_INET6;
-                               dst->sin6_addr = mreq->ipv6mr_multiaddr;
-                               rtalloc((struct route *)&ro);
-                               if (ro.ro_rt == NULL) {
-                                       error = EADDRNOTAVAIL;
-                                       break;
-                               }
-                               ifp = ro.ro_rt->rt_ifp;
-                               rtfree(ro.ro_rt);
-                       }
-               } else
-                       ifp = ifindex2ifnet[mreq->ipv6mr_interface];
-
-               /*
-                * See if we found an interface, and confirm that it
-                * supports multicast
-                */
-               if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
-                       error = EADDRNOTAVAIL;
-                       break;
-               }
-               /*
-                * Put interface index into the multicast address,
-                * if the address has link-local scope.
-                */
-               if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
-                       mreq->ipv6mr_multiaddr.s6_addr16[1]
-                               = htons(mreq->ipv6mr_interface);
-               }
-               /*
-                * See if the membership already exists.
-                */
-               lck_mtx_lock(nd6_mutex);
-               for (imm = im6o->im6o_memberships.lh_first;
-                    imm != NULL; imm = imm->i6mm_chain.le_next)
-                       if (imm->i6mm_maddr->in6m_ifp == ifp &&
-                           IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
-                                              &mreq->ipv6mr_multiaddr))
-                               break;
-               if (imm != NULL) {
-                       error = EADDRINUSE;
-                       lck_mtx_unlock(nd6_mutex);
-                       break;
-               }
-               /*
-                * Everything looks good; add a new record to the multicast
-                * address list for the given interface.
-                */
-               imm = _MALLOC(sizeof(*imm), M_IPMADDR, M_WAITOK);
-               if (imm == NULL) {
-                       error = ENOBUFS;
-                       lck_mtx_unlock(nd6_mutex);
-                       break;
-               }
-               if ((imm->i6mm_maddr =
-                    in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error, 1)) == NULL) {
-                       FREE(imm, M_IPMADDR);
-                       lck_mtx_unlock(nd6_mutex);
-                       break;
-               }
-               LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
-               lck_mtx_unlock(nd6_mutex);
-               break;
-
-       case IPV6_LEAVE_GROUP:
-               /*
-                * Drop a multicast group membership.
-                * Group must be a valid IP6 multicast address.
-                */
-               if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
-                       error = EINVAL;
-                       break;
-               }
-               mreq = mtod(m, struct ipv6_mreq *);
-               /*
-                * If an interface address was specified, get a pointer
-                * to its ifnet structure.
-                */
-               if (mreq->ipv6mr_interface < 0
-                || if_index < mreq->ipv6mr_interface) {
-                       error = ENXIO;  /* XXX EINVAL? */
-                       break;
-               }
-               ifp = ifindex2ifnet[mreq->ipv6mr_interface];
-               
-               if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
-                       if (suser(kauth_cred_get(), 0)) {
-                               error = EACCES;
-                               break;
-                       }
-               } else if (IN6_IS_ADDR_V4MAPPED(&mreq->ipv6mr_multiaddr)) {
-                       struct ip_mreq v4req;
-                       
-                       v4req.imr_multiaddr.s_addr = mreq->ipv6mr_multiaddr.s6_addr32[3];
-                       v4req.imr_interface.s_addr = INADDR_ANY;
-                       
-                       if (ifp != NULL) {
-                               struct in_ifaddr *ifa;
-                               
-                               lck_mtx_lock(rt_mtx);
-                               TAILQ_FOREACH(ifa, &in_ifaddrhead, ia_link) {
-                                       if (ifa->ia_ifp == ifp) {
-                                               v4req.imr_interface = IA_SIN(ifa)->sin_addr;
-                                               break;
-                                       }
-                               }
-                               lck_mtx_unlock(rt_mtx);
-                       }
-                       
-                       error = ip_dropmembership(imo, &v4req);
-                       break;
-               } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
-                       error = EINVAL;
-                       break;
-               }
-               /*
-                * Put interface index into the multicast address,
-                * if the address has link-local scope.
-                */
-               if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
-                       mreq->ipv6mr_multiaddr.s6_addr16[1]
-                               = htons(mreq->ipv6mr_interface);
-               }
-               /*
-                * Find the membership in the membership list.
-                */
-               lck_mtx_lock(nd6_mutex);
-               for (imm = im6o->im6o_memberships.lh_first;
-                    imm != NULL; imm = imm->i6mm_chain.le_next) {
-                       if ((ifp == NULL ||
-                            imm->i6mm_maddr->in6m_ifp == ifp) &&
-                           IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
-                                              &mreq->ipv6mr_multiaddr))
-                               break;
-               }
-               if (imm == NULL) {
-                       /* Unable to resolve interface */
-                       error = EADDRNOTAVAIL;
-                       lck_mtx_unlock(nd6_mutex);
-                       break;
-               }
-               /*
-                * Give up the multicast address record to which the
-                * membership points.
-                */
-               LIST_REMOVE(imm, i6mm_chain);
-               in6_delmulti(imm->i6mm_maddr, 1);
-               lck_mtx_unlock(nd6_mutex);
-               FREE(imm, M_IPMADDR);
-               break;
+       return (dst);
+}
 
 
-       default:
-               error = EOPNOTSUPP;
-               break;
-       }
+void
+ip6_freepcbopts(struct ip6_pktopts *pktopt)
+{
+       if (pktopt == NULL)
+               return;
 
 
-       /*
-        * If all options have default values, no need to keep the mbuf.
-        */
-       lck_mtx_lock(nd6_mutex);
-       if (im6o->im6o_multicast_ifp == NULL &&
-           im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
-           im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
-           im6o->im6o_memberships.lh_first == NULL) {
-               FREE(*im6op, M_IPMOPTS);
-               *im6op = NULL;
-       }
-       if (imo->imo_multicast_ifp == NULL &&
-           imo->imo_multicast_vif == -1 &&
-           imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
-           imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
-           imo->imo_num_memberships == 0) {
-               ip_freemoptions(imo);
-               in6p->inp_moptions = 0;
-       }
-       lck_mtx_unlock(nd6_mutex);
-
-       return(error);
+       ip6_clearpktopts(pktopt, -1);
+
+       FREE(pktopt, M_IP6OPT);
 }
 
 }
 
-/*
- * Return the IP6 multicast options in response to user getsockopt().
- */
-static int
-ip6_getmoptions(optname, im6o, mp)
-       int optname;
-       struct ip6_moptions *im6o;
-       struct mbuf **mp;
+void
+ip6_moptions_init(void)
 {
 {
-       u_int *hlim, *loop, *ifindex;
+       PE_parse_boot_argn("ifa_debug", &im6o_debug, sizeof (im6o_debug));
 
 
-       *mp = m_get(M_WAIT, MT_HEADER);         /*XXX*/
+       im6o_size = (im6o_debug == 0) ? sizeof (struct ip6_moptions) :
+           sizeof (struct ip6_moptions_dbg);
 
 
-       switch (optname) {
+       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);
+}
 
 
-       case IPV6_MULTICAST_IF:
-               ifindex = mtod(*mp, u_int *);
-               (*mp)->m_len = sizeof(u_int);
-               if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
-                       *ifindex = 0;
-               else
-                       *ifindex = im6o->im6o_multicast_ifp->if_index;
-               return(0);
-
-       case IPV6_MULTICAST_HOPS:
-               hlim = mtod(*mp, u_int *);
-               (*mp)->m_len = sizeof(u_int);
-               if (im6o == NULL)
-                       *hlim = ip6_defmcasthlim;
-               else
-                       *hlim = im6o->im6o_multicast_hlim;
-               return(0);
-
-       case IPV6_MULTICAST_LOOP:
-               loop = mtod(*mp, u_int *);
-               (*mp)->m_len = sizeof(u_int);
-               if (im6o == NULL)
-                       *loop = ip6_defmcasthlim;
-               else
-                       *loop = im6o->im6o_multicast_loop;
-               return(0);
+void
+im6o_addref(struct ip6_moptions *im6o, int locked)
+{
+       if (!locked)
+               IM6O_LOCK(im6o);
+       else
+               IM6O_LOCK_ASSERT_HELD(im6o);
 
 
-       default:
-               return(EOPNOTSUPP);
+       if (++im6o->im6o_refcnt == 0) {
+               panic("%s: im6o %p wraparound refcnt\n", __func__, im6o);
+               /* NOTREACHED */
+       } else if (im6o->im6o_trace != NULL) {
+               (*im6o->im6o_trace)(im6o, TRUE);
        }
        }
+
+       if (!locked)
+               IM6O_UNLOCK(im6o);
 }
 
 }
 
-/*
- * Discard the IP6 multicast options.
- */
 void
 void
-ip6_freemoptions(im6o)
-       struct ip6_moptions *im6o;
+im6o_remref(struct ip6_moptions *im6o)
 {
 {
-       struct in6_multi_mship *imm;
+       int i;
+
+       IM6O_LOCK(im6o);
+       if (im6o->im6o_refcnt == 0) {
+               panic("%s: im6o %p negative refcnt", __func__, im6o);
+               /* NOTREACHED */
+       } else if (im6o->im6o_trace != NULL) {
+               (*im6o->im6o_trace)(im6o, FALSE);
+       }
 
 
-       if (im6o == NULL)
+       --im6o->im6o_refcnt;
+       if (im6o->im6o_refcnt > 0) {
+               IM6O_UNLOCK(im6o);
                return;
                return;
-       
-       lck_mtx_lock(nd6_mutex);
-       while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
-               LIST_REMOVE(imm, i6mm_chain);
-               if (imm->i6mm_maddr)
-                       in6_delmulti(imm->i6mm_maddr, 1);
-               FREE(imm, M_IPMADDR);
-       }
-       lck_mtx_unlock(nd6_mutex);
-       FREE(im6o, M_IPMOPTS);
+       }
+
+       for (i = 0; i < im6o->im6o_num_memberships; ++i) {
+               struct in6_mfilter *imf;
+
+               imf = im6o->im6o_mfilters ? &im6o->im6o_mfilters[i] : NULL;
+               if (imf != NULL)
+                       im6f_leave(imf);
+
+               (void) in6_mc_leave(im6o->im6o_membership[i], imf);
+
+               if (imf != NULL)
+                       im6f_purge(imf);
+
+               IN6M_REMREF(im6o->im6o_membership[i]);
+               im6o->im6o_membership[i] = NULL;
+       }
+       im6o->im6o_num_memberships = 0;
+       if (im6o->im6o_mfilters != NULL) {
+               FREE(im6o->im6o_mfilters, M_IN6MFILTER);
+               im6o->im6o_mfilters = NULL;
+       }
+       if (im6o->im6o_membership != NULL) {
+               FREE(im6o->im6o_membership, M_IP6MOPTS);
+               im6o->im6o_membership = NULL;
+       }
+       IM6O_UNLOCK(im6o);
+
+       lck_mtx_destroy(&im6o->im6o_lock, ifa_mtx_grp);
+
+       if (!(im6o->im6o_debug & IFD_ALLOC)) {
+               panic("%s: im6o %p cannot be freed", __func__, im6o);
+               /* NOTREACHED */
+       }
+       zfree(im6o_zone, im6o);
+}
+
+static void
+im6o_trace(struct ip6_moptions *im6o, int refhold)
+{
+       struct ip6_moptions_dbg *im6o_dbg = (struct ip6_moptions_dbg *)im6o;
+       ctrace_t *tr;
+       u_int32_t idx;
+       u_int16_t *cnt;
+
+       if (!(im6o->im6o_debug & IFD_DEBUG)) {
+               panic("%s: im6o %p has no debug structure", __func__, im6o);
+               /* NOTREACHED */
+       }
+       if (refhold) {
+               cnt = &im6o_dbg->im6o_refhold_cnt;
+               tr = im6o_dbg->im6o_refhold;
+       } else {
+               cnt = &im6o_dbg->im6o_refrele_cnt;
+               tr = im6o_dbg->im6o_refrele;
+       }
+
+       idx = atomic_add_16_ov(cnt, 1) % IM6O_TRACE_HIST_SIZE;
+       ctrace_record(&tr[idx]);
+}
+
+struct ip6_moptions *
+ip6_allocmoptions(int how)
+{
+       struct ip6_moptions *im6o;
+
+       im6o = (how == M_WAITOK) ?
+           zalloc(im6o_zone) : zalloc_noblock(im6o_zone);
+       if (im6o != NULL) {
+               bzero(im6o, im6o_size);
+               lck_mtx_init(&im6o->im6o_lock, ifa_mtx_grp, ifa_mtx_attr);
+               im6o->im6o_debug |= IFD_ALLOC;
+               if (im6o_debug != 0) {
+                       im6o->im6o_debug |= IFD_DEBUG;
+                       im6o->im6o_trace = im6o_trace;
+               }
+               IM6O_ADDREF(im6o);
+       }
+
+       return (im6o);
 }
 
 /*
  * Set IPv6 outgoing packet options based on advanced API.
  */
 int
 }
 
 /*
  * Set IPv6 outgoing packet options based on advanced API.
  */
 int
-ip6_setpktoptions(control, opt, priv, needcopy)
-       struct mbuf *control;
-       struct ip6_pktopts *opt;
-       int priv, needcopy;
+ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
+    struct ip6_pktopts *stickyopt, int uproto)
 {
 {
-       struct cmsghdr *cm = 0;
+       struct cmsghdr *cm = NULL;
 
 
-       if (control == 0 || opt == 0)
-               return(EINVAL);
+       if (control == NULL || opt == NULL)
+               return (EINVAL);
 
 
-       init_ip6pktopts(opt);
+       ip6_initpktopts(opt);
+       if (stickyopt) {
+               int error;
+
+               /*
+                * If stickyopt is provided, make a local copy of the options
+                * for this particular packet, then override them by ancillary
+                * objects.
+                * XXX: copypktopts() does not copy the cached route to a next
+                * hop (if any).  This is not very good in terms of efficiency,
+                * but we can allow this since this option should be rarely
+                * used.
+                */
+               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)
 
        /*
         * XXX: Currently, we assume all the optional information is stored
         * in a single mbuf.
         */
        if (control->m_next)
-               return(EINVAL);
+               return (EINVAL);
 
 
-       for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
-                    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
-               cm = mtod(control, struct cmsghdr *);
-               if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
-                       return(EINVAL);
+       if (control->m_len < CMSG_LEN(0))
+               return (EINVAL);
+
+       for (cm = M_FIRST_CMSGHDR(control); cm != NULL;
+           cm = M_NXT_CMSGHDR(control, cm)) {
+               int error;
+
+               if (cm->cmsg_len < sizeof (struct cmsghdr) ||
+                   cm->cmsg_len > control->m_len)
+                       return (EINVAL);
                if (cm->cmsg_level != IPPROTO_IPV6)
                        continue;
 
                if (cm->cmsg_level != IPPROTO_IPV6)
                        continue;
 
+               error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
+                   cm->cmsg_len - CMSG_LEN(0), opt, 0, 1, uproto);
+               if (error)
+                       return (error);
+       }
+
+       return (0);
+}
+/*
+ * Set a particular packet option, as a sticky option or an ancillary data
+ * item.  "len" can be 0 only when it's a sticky option.
+ * We have 4 cases of combination of "sticky" and "cmsg":
+ * "sticky=0, cmsg=0": impossible
+ * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
+ * "sticky=1, cmsg=0": RFC3542 socket option
+ * "sticky=1, cmsg=1": RFC2292 socket option
+ */
+static int
+ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
+    int sticky, int cmsg, int uproto)
+{
+       int minmtupolicy, preftemp;
+       int error;
+       boolean_t capture_exthdrstat_out = FALSE;
+
+       if (!sticky && !cmsg) {
+#ifdef DIAGNOSTIC
+               printf("ip6_setpktopt: impossible case\n");
+#endif
+               return (EINVAL);
+       }
+
+       /*
+        * Caller must have ensured that the buffer is at least
+        * aligned on 32-bit boundary.
+        */
+       VERIFY(IS_P2ALIGNED(buf, sizeof (u_int32_t)));
+
+       /*
+        * IPV6_2292xxx is for backward compatibility to RFC2292, and should
+        * not be specified in the context of RFC3542.  Conversely,
+        * RFC3542 types should not be specified in the context of RFC2292.
+        */
+       if (!cmsg) {
+               switch (optname) {
+               case IPV6_2292PKTINFO:
+               case IPV6_2292HOPLIMIT:
+               case IPV6_2292NEXTHOP:
+               case IPV6_2292HOPOPTS:
+               case IPV6_2292DSTOPTS:
+               case IPV6_2292RTHDR:
+               case IPV6_2292PKTOPTIONS:
+                       return (ENOPROTOOPT);
+               }
+       }
+       if (sticky && cmsg) {
+               switch (optname) {
+               case IPV6_PKTINFO:
+               case IPV6_HOPLIMIT:
+               case IPV6_NEXTHOP:
+               case IPV6_HOPOPTS:
+               case IPV6_DSTOPTS:
+               case IPV6_RTHDRDSTOPTS:
+               case IPV6_RTHDR:
+               case IPV6_USE_MIN_MTU:
+               case IPV6_DONTFRAG:
+               case IPV6_TCLASS:
+               case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
+                       return (ENOPROTOOPT);
+               }
+       }
+
+       switch (optname) {
+       case IPV6_2292PKTINFO:
+       case IPV6_PKTINFO: {
+               struct ifnet *ifp = NULL;
+               struct in6_pktinfo *pktinfo;
+
+               if (len != sizeof (struct in6_pktinfo))
+                       return (EINVAL);
+
+               pktinfo = (struct in6_pktinfo *)(void *)buf;
+
                /*
                /*
-                * XXX should check if RFC2292 API is mixed with 2292bis API
+                * An application can clear any sticky IPV6_PKTINFO option by
+                * doing a "regular" setsockopt with ipi6_addr being
+                * in6addr_any and ipi6_ifindex being zero.
+                * [RFC 3542, Section 6]
                 */
                 */
-               switch (cm->cmsg_type) {
-               case IPV6_PKTINFO:
-                       if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
-                               return(EINVAL);
-                       if (needcopy) {
-                               /* XXX: Is it really WAITOK? */
-                               opt->ip6po_pktinfo =
-                                       _MALLOC(sizeof(struct in6_pktinfo),
-                                              M_IP6OPT, M_WAITOK);
-                               bcopy(CMSG_DATA(cm), opt->ip6po_pktinfo,
-                                   sizeof(struct in6_pktinfo));
-                       } else
-                               opt->ip6po_pktinfo =
-                                       (struct in6_pktinfo *)CMSG_DATA(cm);
-                       if (opt->ip6po_pktinfo->ipi6_ifindex &&
-                           IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
-                               opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
-                                       htons(opt->ip6po_pktinfo->ipi6_ifindex);
-
-                       if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
-                        || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
-                               return(ENXIO);
-                       }
+               if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
+                   pktinfo->ipi6_ifindex == 0 &&
+                   IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
+                       ip6_clearpktopts(opt, optname);
+                       break;
+               }
 
 
-                       /*
-                        * Check if the requested source address is indeed a
-                        * unicast address assigned to the node, and can be
-                        * used as the packet's source address.
-                        */
-                       if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
-                               struct in6_ifaddr *ia6;
-                               struct sockaddr_in6 sin6;
-
-                               bzero(&sin6, sizeof(sin6));
-                               sin6.sin6_len = sizeof(sin6);
-                               sin6.sin6_family = AF_INET6;
-                               sin6.sin6_addr =
-                                       opt->ip6po_pktinfo->ipi6_addr;
-                               ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
-                               if (ia6 == NULL ||
-                                   (ia6->ia6_flags & (IN6_IFF_ANYCAST |
-                                                      IN6_IFF_NOTREADY)) != 0) {
-                                       if (ia6) ifafree(&ia6->ia_ifa);
-                                       return(EADDRNOTAVAIL);
-                               }
-                               ifafree(&ia6->ia_ifa);
-                               ia6 = NULL;
+               if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
+                   sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
+                       return (EINVAL);
+               }
+
+               /* validate the interface index if specified. */
+               ifnet_head_lock_shared();
+
+               if (pktinfo->ipi6_ifindex > if_index) {
+                       ifnet_head_done();
+                       return (ENXIO);
+               }
+
+               if (pktinfo->ipi6_ifindex) {
+                       ifp = ifindex2ifnet[pktinfo->ipi6_ifindex];
+                       if (ifp == NULL) {
+                               ifnet_head_done();
+                               return (ENXIO);
                        }
                        }
-                       break;
+               }
 
 
-               case IPV6_HOPLIMIT:
-                       if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
-                               return(EINVAL);
+               ifnet_head_done();
 
 
-                       opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
-                       if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
-                               return(EINVAL);
-                       break;
+               /*
+                * We store the address anyway, and let in6_selectsrc()
+                * validate the specified address.  This is because ipi6_addr
+                * may not have enough information about its scope zone, and
+                * we may need additional information (such as outgoing
+                * interface or the scope zone of a destination address) to
+                * disambiguate the scope.
+                * XXX: the delay of the validation may confuse the
+                * application when it is used as a sticky option.
+                */
+               if (opt->ip6po_pktinfo == NULL) {
+                       opt->ip6po_pktinfo = _MALLOC(sizeof (*pktinfo),
+                           M_IP6OPT, M_NOWAIT);
+                       if (opt->ip6po_pktinfo == NULL)
+                               return (ENOBUFS);
+               }
+               bcopy(pktinfo, opt->ip6po_pktinfo, sizeof (*pktinfo));
+               break;
+       }
 
 
-               case IPV6_NEXTHOP:
-                       if (!priv)
-                               return(EPERM);
-
-                       if (cm->cmsg_len < sizeof(u_char) ||
-                           /* check if cmsg_len is large enough for sa_len */
-                           cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
-                               return(EINVAL);
-
-                       if (needcopy) {
-                               opt->ip6po_nexthop =
-                                       _MALLOC(*CMSG_DATA(cm),
-                                              M_IP6OPT, M_WAITOK);
-                               bcopy(CMSG_DATA(cm),
-                                     opt->ip6po_nexthop,
-                                     *CMSG_DATA(cm));
-                       } else
-                               opt->ip6po_nexthop =
-                                       (struct sockaddr *)CMSG_DATA(cm);
+       case IPV6_2292HOPLIMIT:
+       case IPV6_HOPLIMIT: {
+               int *hlimp;
+
+               /*
+                * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
+                * to simplify the ordering among hoplimit options.
+                */
+               if (optname == IPV6_HOPLIMIT && sticky)
+                       return (ENOPROTOOPT);
+
+               if (len != sizeof (int))
+                       return (EINVAL);
+               hlimp = (int *)(void *)buf;
+               if (*hlimp < -1 || *hlimp > 255)
+                       return (EINVAL);
+
+               opt->ip6po_hlim = *hlimp;
+               break;
+       }
+
+       case IPV6_TCLASS: {
+               int tclass;
+
+               if (len != sizeof (int))
+                       return (EINVAL);
+               tclass = *(int *)(void *)buf;
+               if (tclass < -1 || tclass > 255)
+                       return (EINVAL);
+
+               opt->ip6po_tclass = tclass;
+               break;
+       }
+
+       case IPV6_2292NEXTHOP:
+       case IPV6_NEXTHOP:
+               error = suser(kauth_cred_get(), 0);
+               if (error)
+                       return (EACCES);
+
+               if (len == 0) { /* just remove the option */
+                       ip6_clearpktopts(opt, IPV6_NEXTHOP);
                        break;
                        break;
+               }
 
 
-               case IPV6_HOPOPTS:
-               {
-                       struct ip6_hbh *hbh;
-                       int hbhlen;
-
-                       if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
-                               return(EINVAL);
-                       hbh = (struct ip6_hbh *)CMSG_DATA(cm);
-                       hbhlen = (hbh->ip6h_len + 1) << 3;
-                       if (cm->cmsg_len != CMSG_LEN(hbhlen))
-                               return(EINVAL);
-
-                       if (needcopy) {
-                               opt->ip6po_hbh =
-                                       _MALLOC(hbhlen, M_IP6OPT, M_WAITOK);
-                               bcopy(hbh, opt->ip6po_hbh, hbhlen);
-                       } else
-                               opt->ip6po_hbh = hbh;
+               /* check if cmsg_len is large enough for sa_len */
+               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 (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
+                           IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
+                               return (EINVAL);
+                       }
+                       if ((error = sa6_embedscope(sa6, ip6_use_defzone))
+                           != 0) {
+                               return (error);
+                       }
                        break;
                }
                        break;
                }
+               case AF_LINK:   /* should eventually be supported */
+               default:
+                       return (EAFNOSUPPORT);
+               }
 
 
-               case IPV6_DSTOPTS:
-               {
-                       struct ip6_dest *dest, **newdest;
-                       int destlen;
-
-                       if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
-                               return(EINVAL);
-                       dest = (struct ip6_dest *)CMSG_DATA(cm);
-                       destlen = (dest->ip6d_len + 1) << 3;
-                       if (cm->cmsg_len != CMSG_LEN(destlen))
-                               return(EINVAL);
-
-                       /* 
-                        * The old advacned API is ambiguous on this
-                        * point. Our approach is to determine the
-                        * position based according to the existence
-                        * of a routing header. Note, however, that
-                        * this depends on the order of the extension
-                        * headers in the ancillary data; the 1st part
-                        * of the destination options header must
-                        * appear before the routing header in the
-                        * ancillary data, too.
-                        * RFC2292bis solved the ambiguity by
-                        * introducing separate cmsg types.
+               /* turn off the previous option, then set the new option. */
+               ip6_clearpktopts(opt, IPV6_NEXTHOP);
+               opt->ip6po_nexthop = _MALLOC(*buf, M_IP6OPT, M_NOWAIT);
+               if (opt->ip6po_nexthop == NULL)
+                       return (ENOBUFS);
+               bcopy(buf, opt->ip6po_nexthop, *buf);
+               break;
+
+       case IPV6_2292HOPOPTS:
+       case IPV6_HOPOPTS: {
+               struct ip6_hbh *hbh;
+               int hbhlen;
+
+               /*
+                * XXX: We don't allow a non-privileged user to set ANY HbH
+                * options, since per-option restriction has too much
+                * overhead.
+                */
+               error = suser(kauth_cred_get(), 0);
+               if (error)
+                       return (EACCES);
+
+               if (len == 0) {
+                       ip6_clearpktopts(opt, IPV6_HOPOPTS);
+                       break;  /* just remove the option */
+               }
+
+               /* message length validation */
+               if (len < sizeof (struct ip6_hbh))
+                       return (EINVAL);
+               hbh = (struct ip6_hbh *)(void *)buf;
+               hbhlen = (hbh->ip6h_len + 1) << 3;
+               if (len != hbhlen)
+                       return (EINVAL);
+
+               /* turn off the previous option, then set the new option. */
+               ip6_clearpktopts(opt, IPV6_HOPOPTS);
+               opt->ip6po_hbh = _MALLOC(hbhlen, M_IP6OPT, M_NOWAIT);
+               if (opt->ip6po_hbh == NULL)
+                       return (ENOBUFS);
+               bcopy(hbh, opt->ip6po_hbh, hbhlen);
+               capture_exthdrstat_out = TRUE;
+               break;
+       }
+
+       case IPV6_2292DSTOPTS:
+       case IPV6_DSTOPTS:
+       case IPV6_RTHDRDSTOPTS: {
+               struct ip6_dest *dest, **newdest = NULL;
+               int destlen;
+
+               error = suser(kauth_cred_get(), 0);
+               if (error)
+                       return (EACCES);
+
+               if (len == 0) {
+                       ip6_clearpktopts(opt, optname);
+                       break;  /* just remove the option */
+               }
+
+               /* message length validation */
+               if (len < sizeof (struct ip6_dest))
+                       return (EINVAL);
+               dest = (struct ip6_dest *)(void *)buf;
+               destlen = (dest->ip6d_len + 1) << 3;
+               if (len != destlen)
+                       return (EINVAL);
+
+               /*
+                * Determine the position that the destination options header
+                * should be inserted; before or after the routing header.
+                */
+               switch (optname) {
+               case IPV6_2292DSTOPTS:
+                       /*
+                        * The old advacned API is ambiguous on this point.
+                        * Our approach is to determine the position based
+                        * according to the existence of a routing header.
+                        * Note, however, that this depends on the order of the
+                        * extension headers in the ancillary data; the 1st
+                        * part of the destination options header must appear
+                        * before the routing header in the ancillary data,
+                        * too.
+                        * RFC3542 solved the ambiguity by introducing
+                        * separate ancillary data or option types.
                         */
                        if (opt->ip6po_rthdr == NULL)
                                newdest = &opt->ip6po_dest1;
                        else
                                newdest = &opt->ip6po_dest2;
                         */
                        if (opt->ip6po_rthdr == NULL)
                                newdest = &opt->ip6po_dest1;
                        else
                                newdest = &opt->ip6po_dest2;
+                       break;
+               case IPV6_RTHDRDSTOPTS:
+                       newdest = &opt->ip6po_dest1;
+                       break;
+               case IPV6_DSTOPTS:
+                       newdest = &opt->ip6po_dest2;
+                       break;
+               }
+
+               /* turn off the previous option, then set the new option. */
+               ip6_clearpktopts(opt, optname);
+               *newdest = _MALLOC(destlen, M_IP6OPT, M_NOWAIT);
+               if (*newdest == NULL)
+                       return (ENOBUFS);
+               bcopy(dest, *newdest, destlen);
+               capture_exthdrstat_out = TRUE;
+               break;
+       }
+
+       case IPV6_2292RTHDR:
+       case IPV6_RTHDR: {
+               struct ip6_rthdr *rth;
+               int rthlen;
 
 
-                       if (needcopy) {
-                               *newdest = _MALLOC(destlen, M_IP6OPT, M_WAITOK);
-                               bcopy(dest, *newdest, destlen);
-                       } else
-                               *newdest = dest;
+               if (len == 0) {
+                       ip6_clearpktopts(opt, IPV6_RTHDR);
+                       break;  /* just remove the option */
+               }
+
+               /* message length validation */
+               if (len < sizeof (struct ip6_rthdr))
+                       return (EINVAL);
+               rth = (struct ip6_rthdr *)(void *)buf;
+               rthlen = (rth->ip6r_len + 1) << 3;
+               if (len != rthlen)
+                       return (EINVAL);
 
 
+               switch (rth->ip6r_type) {
+               case IPV6_RTHDR_TYPE_0:
+                       if (rth->ip6r_len == 0) /* must contain one addr */
+                               return (EINVAL);
+                       if (rth->ip6r_len % 2) /* length must be even */
+                               return (EINVAL);
+                       if (rth->ip6r_len / 2 != rth->ip6r_segleft)
+                               return (EINVAL);
                        break;
                        break;
+               default:
+                       return (EINVAL);        /* not supported */
                }
 
                }
 
-               case IPV6_RTHDR:
-               {
-                       struct ip6_rthdr *rth;
-                       int rthlen;
-
-                       if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
-                               return(EINVAL);
-                       rth = (struct ip6_rthdr *)CMSG_DATA(cm);
-                       rthlen = (rth->ip6r_len + 1) << 3;
-                       if (cm->cmsg_len != CMSG_LEN(rthlen))
-                               return(EINVAL);
-
-                       switch (rth->ip6r_type) {
-                       case IPV6_RTHDR_TYPE_0:
-                               /* must contain one addr */
-                               if (rth->ip6r_len == 0)
-                                       return(EINVAL);
-                               /* length must be even */
-                               if (rth->ip6r_len % 2)
-                                       return(EINVAL);
-                               if (rth->ip6r_len / 2 != rth->ip6r_segleft)
-                                       return(EINVAL);
-                               break;
-                       default:
-                               return(EINVAL); /* not supported */
-                       }
+               /* turn off the previous option */
+               ip6_clearpktopts(opt, IPV6_RTHDR);
+               opt->ip6po_rthdr = _MALLOC(rthlen, M_IP6OPT, M_NOWAIT);
+               if (opt->ip6po_rthdr == NULL)
+                       return (ENOBUFS);
+               bcopy(rth, opt->ip6po_rthdr, rthlen);
+               capture_exthdrstat_out = TRUE;
+               break;
+       }
+
+       case IPV6_USE_MIN_MTU:
+               if (len != sizeof (int))
+                       return (EINVAL);
+               minmtupolicy = *(int *)(void *)buf;
+               if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
+                   minmtupolicy != IP6PO_MINMTU_DISABLE &&
+                   minmtupolicy != IP6PO_MINMTU_ALL) {
+                       return (EINVAL);
+               }
+               opt->ip6po_minmtu = minmtupolicy;
+               break;
 
 
-                       if (needcopy) {
-                               opt->ip6po_rthdr = _MALLOC(rthlen, M_IP6OPT,
-                                                         M_WAITOK);
-                               bcopy(rth, opt->ip6po_rthdr, rthlen);
-                       } else
-                               opt->ip6po_rthdr = rth;
+       case IPV6_DONTFRAG:
+               if (len != sizeof (int))
+                       return (EINVAL);
 
 
-                       break;
+               if (uproto == IPPROTO_TCP || *(int *)(void *)buf == 0) {
+                       /*
+                        * we ignore this option for TCP sockets.
+                        * (RFC3542 leaves this case unspecified.)
+                        */
+                       opt->ip6po_flags &= ~IP6PO_DONTFRAG;
+               } else {
+                       opt->ip6po_flags |= IP6PO_DONTFRAG;
                }
                }
+               break;
 
 
-               default:
-                       return(ENOPROTOOPT);
+       case IPV6_PREFER_TEMPADDR:
+               if (len != sizeof (int))
+                       return (EINVAL);
+               preftemp = *(int *)(void *)buf;
+               if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
+                   preftemp != IP6PO_TEMPADDR_NOTPREFER &&
+                   preftemp != IP6PO_TEMPADDR_PREFER) {
+                       return (EINVAL);
+               }
+               opt->ip6po_prefer_tempaddr = preftemp;
+               break;
+
+       default:
+               return (ENOPROTOOPT);
+       } /* end of switch */
+
+       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);
+       return (0);
 }
 
 /*
 }
 
 /*
@@ -2636,121 +3957,250 @@ ip6_setpktoptions(control, opt, priv, needcopy)
  * pointer that might NOT be &loif -- easier than replicating that code here.
  */
 void
  * pointer that might NOT be &loif -- easier than replicating that code here.
  */
 void
-ip6_mloopback(
-       struct ifnet *ifp,
-       struct mbuf *m,
-       struct sockaddr_in6 *dst)
+ip6_mloopback(struct ifnet *srcifp, struct ifnet *origifp, struct mbuf *m,
+    struct sockaddr_in6 *dst, uint32_t optlen, int32_t nxt0)
 {
        struct mbuf *copym;
        struct ip6_hdr *ip6;
 {
        struct mbuf *copym;
        struct ip6_hdr *ip6;
+       struct in6_addr src;
 
 
-       copym = m_copy(m, 0, M_COPYALL);
-       if (copym == NULL)
+       if (lo_ifp == NULL)
                return;
 
        /*
                return;
 
        /*
+        * Copy the packet header as it's needed for the checksum.
         * Make sure to deep-copy IPv6 header portion in case the data
         * is in an mbuf cluster, so that we can safely override the IPv6
         * header portion later.
         */
         * Make sure to deep-copy IPv6 header portion in case the data
         * is in an mbuf cluster, so that we can safely override the IPv6
         * header portion later.
         */
-       if ((copym->m_flags & M_EXT) != 0 ||
-           copym->m_len < sizeof(struct ip6_hdr)) {
-               copym = m_pullup(copym, sizeof(struct ip6_hdr));
-               if (copym == NULL)
-                       return;
-       }
+       copym = m_copym_mode(m, 0, M_COPYALL, M_DONTWAIT, M_COPYM_COPY_HDR);
+       if (copym != NULL && ((copym->m_flags & M_EXT) ||
+           copym->m_len < sizeof (struct ip6_hdr)))
+               copym = m_pullup(copym, sizeof (struct ip6_hdr));
 
 
-#if DIAGNOSTIC
-       if (copym->m_len < sizeof(*ip6)) {
-               m_freem(copym);
+       if (copym == NULL)
                return;
                return;
-       }
-#endif
 
        ip6 = mtod(copym, struct ip6_hdr *);
 
        ip6 = mtod(copym, struct ip6_hdr *);
-#ifndef SCOPEDROUTING
+       src = ip6->ip6_src;
        /*
         * clear embedded scope identifiers if necessary.
         * in6_clearscope will touch the addresses only when necessary.
         */
        in6_clearscope(&ip6->ip6_src);
        in6_clearscope(&ip6->ip6_dst);
        /*
         * clear embedded scope identifiers if necessary.
         * in6_clearscope will touch the addresses only when necessary.
         */
        in6_clearscope(&ip6->ip6_src);
        in6_clearscope(&ip6->ip6_dst);
-#endif
 
 
-#ifdef __APPLE__
+       if (copym->m_pkthdr.csum_flags & CSUM_DELAY_IPV6_DATA)
+               in6_delayed_cksum_offset(copym, 0, optlen, nxt0);
 
 
-       /* Makes sure the HW checksum flags are cleaned before sending the packet */
+       /*
+        * Stuff the 'real' ifp into the pkthdr, to be used in matching
+        * in ip6_input(); we need the loopback ifp/dl_tag passed as args
+        * to make the loopback driver compliant with the data link
+        * requirements.
+        */
+       copym->m_pkthdr.rcvif = origifp;
 
 
-       copym->m_pkthdr.rcvif = 0;
-       copym->m_pkthdr.csum_data = 0;
-       copym->m_pkthdr.csum_flags = 0;
+       /*
+        * Also record the source interface (which owns the source address).
+        * This is basically a stripped down version of ifa_foraddr6().
+        */
+       if (srcifp == NULL) {
+               struct in6_ifaddr *ia;
+
+               lck_rw_lock_shared(&in6_ifaddr_rwlock);
+               for (ia = in6_ifaddrs; ia != NULL; ia = ia->ia_next) {
+                       IFA_LOCK_SPIN(&ia->ia_ifa);
+                       /* compare against src addr with embedded scope */
+                       if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &src)) {
+                               srcifp = ia->ia_ifp;
+                               IFA_UNLOCK(&ia->ia_ifa);
+                               break;
+                       }
+                       IFA_UNLOCK(&ia->ia_ifa);
+               }
+               lck_rw_done(&in6_ifaddr_rwlock);
+       }
+       if (srcifp != NULL)
+               ip6_setsrcifaddr_info(copym, srcifp->if_index, NULL);
+       ip6_setdstifaddr_info(copym, origifp->if_index, NULL);
 
 
-       if (lo_ifp) {
-               copym->m_pkthdr.rcvif = ifp;
-               lck_mtx_unlock(ip6_mutex);
-               dlil_output(lo_ifp, PF_INET6, copym, 0, (struct sockaddr *)dst, 0);
-               lck_mtx_lock(ip6_mutex);
-       } else
-               m_free(copym);
-#else
-       (void)if_simloop(ifp, copym, dst->sin6_family, NULL);
-#endif
+       dlil_output(lo_ifp, PF_INET6, copym, NULL, SA(dst), 0, NULL);
 }
 
 /*
  * Chop IPv6 header off from the payload.
  */
 static int
 }
 
 /*
  * Chop IPv6 header off from the payload.
  */
 static int
-ip6_splithdr(m, exthdrs)
-       struct mbuf *m;
-       struct ip6_exthdrs *exthdrs;
+ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
 {
        struct mbuf *mh;
        struct ip6_hdr *ip6;
 
        ip6 = mtod(m, struct ip6_hdr *);
 {
        struct mbuf *mh;
        struct ip6_hdr *ip6;
 
        ip6 = mtod(m, struct ip6_hdr *);
-       if (m->m_len > sizeof(*ip6)) {
-               MGETHDR(mh, M_DONTWAIT, MT_HEADER);
-               if (mh == 0) {
+       if (m->m_len > sizeof (*ip6)) {
+               MGETHDR(mh, M_DONTWAIT, MT_HEADER);     /* MAC-OK */
+               if (mh == NULL) {
                        m_freem(m);
                        m_freem(m);
-                       return ENOBUFS;
+                       return (ENOBUFS);
                }
                M_COPY_PKTHDR(mh, m);
                }
                M_COPY_PKTHDR(mh, m);
-               MH_ALIGN(mh, sizeof(*ip6));
+               MH_ALIGN(mh, sizeof (*ip6));
                m->m_flags &= ~M_PKTHDR;
                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;
                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;
        }
        exthdrs->ip6e_ip6 = m;
-       return 0;
+       return (0);
+}
+
+static void
+ip6_output_checksum(struct ifnet *ifp, uint32_t mtu, struct mbuf *m,
+    int nxt0, uint32_t tlen, uint32_t optlen)
+{
+       uint32_t sw_csum, hwcap = ifp->if_hwassist;
+       int tso = TSO_IPV6_OK(ifp, m);
+
+       if (!hwcksum_tx) {
+               /* do all in software; checksum offload is disabled */
+               sw_csum = CSUM_DELAY_IPV6_DATA & m->m_pkthdr.csum_flags;
+       } else {
+               /* do in software what the hardware cannot */
+               sw_csum = m->m_pkthdr.csum_flags &
+                   ~IF_HWASSIST_CSUM_FLAGS(hwcap);
+       }
+
+       if (optlen != 0) {
+               sw_csum |= (CSUM_DELAY_IPV6_DATA &
+                   m->m_pkthdr.csum_flags);
+       } else if (!(sw_csum & CSUM_DELAY_IPV6_DATA) &&
+           (hwcap & CSUM_PARTIAL)) {
+               /*
+                * Partial checksum offload, ere), if no extension headers,
+                * and TCP only (no UDP support, as the hardware may not be
+                * able to convert +0 to -0 (0xffff) per RFC1122 4.1.3.4.
+                * unless the interface supports "invert zero" capability.)
+                */
+               if (hwcksum_tx && !tso &&
+                   ((m->m_pkthdr.csum_flags & CSUM_TCPIPV6) ||
+                   ((hwcap & CSUM_ZERO_INVERT) &&
+                   (m->m_pkthdr.csum_flags & CSUM_ZERO_INVERT))) &&
+                   tlen <= mtu) {
+                       uint16_t start = sizeof (struct ip6_hdr);
+                       uint16_t ulpoff =
+                           m->m_pkthdr.csum_data & 0xffff;
+                       m->m_pkthdr.csum_flags |=
+                           (CSUM_DATA_VALID | CSUM_PARTIAL);
+                       m->m_pkthdr.csum_tx_stuff = (ulpoff + start);
+                       m->m_pkthdr.csum_tx_start = start;
+                       sw_csum = 0;
+               } else {
+                       sw_csum |= (CSUM_DELAY_IPV6_DATA &
+                           m->m_pkthdr.csum_flags);
+               }
+       }
+
+       if (sw_csum & CSUM_DELAY_IPV6_DATA) {
+               in6_delayed_cksum_offset(m, 0, optlen, nxt0);
+               sw_csum &= ~CSUM_DELAY_IPV6_DATA;
+       }
+
+       if (hwcksum_tx) {
+               /*
+                * Drop off bits that aren't supported by hardware;
+                * also make sure to preserve non-checksum related bits.
+                */
+               m->m_pkthdr.csum_flags =
+                   ((m->m_pkthdr.csum_flags &
+                   (IF_HWASSIST_CSUM_FLAGS(hwcap) | CSUM_DATA_VALID)) |
+                   (m->m_pkthdr.csum_flags & ~IF_HWASSIST_CSUM_MASK));
+       } else {
+               /* drop all bits; checksum offload is disabled */
+               m->m_pkthdr.csum_flags = 0;
+       }
 }
 
 /*
  * Compute IPv6 extension header length.
  */
 int
 }
 
 /*
  * Compute IPv6 extension header length.
  */
 int
-ip6_optlen(in6p)
-       struct in6pcb *in6p;
+ip6_optlen(struct in6pcb *in6p)
 {
        int len;
 
        if (!in6p->in6p_outputopts)
 {
        int len;
 
        if (!in6p->in6p_outputopts)
-               return 0;
+               return (0);
 
        len = 0;
 
        len = 0;
-#define elen(x) \
-    (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
+#define        elen(x)                                                         \
+       (((struct ip6_ext *)(x)) ?                                      \
+       (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
 
        len += elen(in6p->in6p_outputopts->ip6po_hbh);
 
        len += elen(in6p->in6p_outputopts->ip6po_hbh);
-       if (in6p->in6p_outputopts->ip6po_rthdr)
+       if (in6p->in6p_outputopts->ip6po_rthdr) {
                /* dest1 is valid with rthdr only */
                len += elen(in6p->in6p_outputopts->ip6po_dest1);
                /* dest1 is valid with rthdr only */
                len += elen(in6p->in6p_outputopts->ip6po_dest1);
+       }
        len += elen(in6p->in6p_outputopts->ip6po_rthdr);
        len += elen(in6p->in6p_outputopts->ip6po_dest2);
        len += elen(in6p->in6p_outputopts->ip6po_rthdr);
        len += elen(in6p->in6p_outputopts->ip6po_dest2);
-       return len;
+       return (len);
 #undef elen
 }
 #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)));
+}