]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/netinet/ip_output.c
xnu-792.6.70.tar.gz
[apple/xnu.git] / bsd / netinet / ip_output.c
index e365ce7f187e1914362a842ec7847599611890d3..44f89c3cfb9af8f3c60bdaf3333ac1cf389e75e2 100644 (file)
@@ -1,24 +1,21 @@
 /*
- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
  *
  * @APPLE_LICENSE_HEADER_START@
  * 
- * Copyright (c) 1999-2003 Apple Computer, Inc.  All Rights Reserved.
+ * The contents of this file constitute Original Code as defined in and
+ * are subject to the Apple Public Source License Version 1.1 (the
+ * "License").  You may not use this file except in compliance with the
+ * License.  Please obtain a copy of the License at
+ * http://www.apple.com/publicsource and read it before using this file.
  * 
- * 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. 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
+ * This 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.
+ * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
+ * License for the specific language governing rights and limitations
+ * under the License.
  * 
  * @APPLE_LICENSE_HEADER_END@
  */
@@ -68,6 +65,8 @@
 #include <sys/protosw.h>
 #include <sys/socket.h>
 #include <sys/socketvar.h>
+#include <kern/locks.h>
+#include <sys/sysctl.h>
 
 #include <net/if.h>
 #include <net/route.h>
@@ -79,6 +78,8 @@
 #include <netinet/in_var.h>
 #include <netinet/ip_var.h>
 
+#include <netinet/kpi_ipfilter_var.h>
+
 #include "faith.h"
 
 #include <net/dlil.h>
 #define DBG_LAYER_BEG          NETDBG_CODE(DBG_NETIP, 1)
 #define DBG_LAYER_END          NETDBG_CODE(DBG_NETIP, 3)
 #define DBG_FNC_IP_OUTPUT      NETDBG_CODE(DBG_NETIP, (1 << 8) | 1)
+#define DBG_FNC_IPSEC4_OUTPUT  NETDBG_CODE(DBG_NETIP, (2 << 8) | 1)
 
 
-#if vax
-#include <machine/mtpr.h>
-#endif
-
-#if __FreeBSD__
-#include <machine/in_cksum.h>
-
-static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
-#endif
-
 #if IPSEC
 #include <netinet6/ipsec.h>
 #include <netkey/key.h>
@@ -110,6 +102,7 @@ static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
 #endif /*IPSEC*/
 
 #include <netinet/ip_fw.h>
+#include <netinet/ip_divert.h>
 
 #if DUMMYNET
 #include <netinet/ip_dummynet.h>
@@ -122,38 +115,45 @@ static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
                                                  (ntohl(a.s_addr))&0xFF);
 #endif
 
+#if IPSEC
+extern lck_mtx_t    *sadb_mutex;
+#endif
+
 u_short ip_id;
 
-static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
-static struct ifnet *ip_multicast_if __P((struct in_addr *, int *));
-static void    ip_mloopback
-       __P((struct ifnet *, struct mbuf *, struct sockaddr_in *, int));
-static int     ip_getmoptions
-       __P((struct sockopt *, struct ip_moptions *));
-static int     ip_pcbopts __P((int, struct mbuf **, struct mbuf *));
-static int     ip_setmoptions
-       __P((struct sockopt *, struct ip_moptions **));
-
-int    ip_optcopy __P((struct ip *, struct ip *));
-extern int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **));
+static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
+static struct ifnet *ip_multicast_if(struct in_addr *, int *);
+static void    ip_mloopback(struct ifnet *, struct mbuf *,
+       struct sockaddr_in *, int);
+static int     ip_getmoptions(struct sockopt *, struct ip_moptions *);
+static int     ip_pcbopts(int, struct mbuf **, struct mbuf *);
+static int     ip_setmoptions(struct sockopt *, struct ip_moptions **);
+
+int ip_createmoptions(struct ip_moptions **imop);
+int ip_addmembership(struct ip_moptions *imo, struct ip_mreq *mreq);
+int ip_dropmembership(struct ip_moptions *imo, struct ip_mreq *mreq);
+int    ip_optcopy(struct ip *, struct ip *);
+extern int (*fr_checkp)(struct ip *, int, struct ifnet *, int, struct mbuf **);
 #ifdef __APPLE__
 extern struct mbuf*    m_dup(register struct mbuf *m, int how);
 #endif
 
-static u_long  lo_dl_tag = 0;
-
-void in_delayed_cksum(struct mbuf *m);
 extern int apple_hwcksum_tx;
+extern u_long  route_generation;
 
 extern struct protosw inetsw[];
 
 extern struct ip_linklocal_stat ip_linklocal_stat;
+extern lck_mtx_t *ip_mutex;
 
 /* temporary: for testing */
 #if IPSEC
 extern int ipsec_bypass;
 #endif
 
+static int     ip_maxchainsent = 0;
+SYSCTL_INT(_net_inet_ip, OID_AUTO, maxchainsent, CTLFLAG_RW,
+    &ip_maxchainsent, 0, "use dlil_output_list");
 /*
  * IP output.  The packet in mbuf chain m contains a skeletal IP
  * header (with len, off, ttl, proto, tos, src, dst).
@@ -161,96 +161,145 @@ extern int ipsec_bypass;
  * The mbuf opt, if present, will not be freed.
  */
 int
-ip_output(m0, opt, ro, flags, imo)
-       struct mbuf *m0;
-       struct mbuf *opt;
-       struct route *ro;
-       int flags;
-       struct ip_moptions *imo;
+ip_output(
+       struct mbuf *m0,
+       struct mbuf *opt,
+       struct route *ro,
+       int flags,
+       struct ip_moptions *imo)
+{
+       int error;
+       error = ip_output_list(m0, 0, opt, ro, flags, imo);
+       return error;
+}
+
+int
+ip_output_list(
+       struct mbuf *m0,
+       int packetchain,
+       struct mbuf *opt,
+       struct route *ro,
+       int flags,
+       struct ip_moptions *imo)
 {
        struct ip *ip, *mhip;
-       struct ifnet *ifp;
-       u_long       dl_tag;
+       struct ifnet *ifp = NULL;
        struct mbuf *m = m0;
        int hlen = sizeof (struct ip);
        int len, off, error = 0;
-       struct sockaddr_in *dst;
+       struct sockaddr_in *dst = NULL;
        struct in_ifaddr *ia = NULL;
        int isbroadcast, sw_csum;
+       struct in_addr pkt_dst;
 #if IPSEC
        struct route iproute;
        struct socket *so = NULL;
        struct secpolicy *sp = NULL;
 #endif
-       u_int16_t divert_cookie;                /* firewall cookie */
 #if IPFIREWALL_FORWARD
        int fwd_rewrite_src = 0;
 #endif
-       struct ip_fw_chain *rule = NULL;
-  
-#if IPDIVERT
-       /* Get and reset firewall cookie */
-       divert_cookie = ip_divert_cookie;
-       ip_divert_cookie = 0;
-#else
-       divert_cookie = 0;
-#endif
+       struct ip_fw_args args;
+       int didfilter = 0;
+       ipfilter_t inject_filter_ref = 0;
+       struct m_tag    *tag;
+       struct route    dn_route;
+       struct mbuf * packetlist;
+       int pktcnt = 0;
+       
+       lck_mtx_lock(ip_mutex); 
 
        KERNEL_DEBUG(DBG_FNC_IP_OUTPUT | DBG_FUNC_START, 0,0,0,0,0);
 
-#if IPFIREWALL && DUMMYNET
-        /*  
-         * dummynet packet are prepended a vestigial mbuf with
-         * m_type = MT_DUMMYNET and m_data pointing to the matching
-         * rule.
-         */ 
-        if (m->m_type == MT_DUMMYNET) {
-            /*
-             * the packet was already tagged, so part of the
-             * processing was already done, and we need to go down.
-             * Get parameters from the header.
-             */
-                       rule = (struct ip_fw_chain *)(m->m_data) ;
-                       opt = NULL ;
-                       ro = & ( ((struct dn_pkt *)m)->ro ) ;
-                       imo = NULL ;
-                       dst = ((struct dn_pkt *)m)->dn_dst ;
-                       ifp = ((struct dn_pkt *)m)->ifp ;
-                       flags = ((struct dn_pkt *)m)->flags ;
-                       m0 = m = m->m_next ;
-#if IPSEC
-           if (ipsec_bypass == 0) {
-               so = ipsec_getsocket(m);
-               (void)ipsec_setsocket(m, NULL);
-           }
+       packetlist = m0;
+       args.eh = NULL;
+       args.rule = NULL;
+       args.next_hop = NULL;
+       args.divert_rule = 0;                   /* divert cookie */
+       
+       /* Grab info from mtags prepended to the chain */
+#if DUMMYNET
+       if ((tag = m_tag_locate(m0, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_DUMMYNET, NULL)) != NULL) {
+               struct dn_pkt_tag       *dn_tag;
+               
+               dn_tag = (struct dn_pkt_tag *)(tag+1);
+               args.rule = dn_tag->rule;
+               opt = NULL;
+               dn_route = dn_tag->ro;
+               ro = &dn_route;
+                       
+               imo = NULL;
+               dst = dn_tag->dn_dst;
+               ifp = dn_tag->ifp;
+               flags = dn_tag->flags;
+               
+               m_tag_delete(m0, tag);
+       }
+#endif /* DUMMYNET */
+
+       if ((tag = m_tag_locate(m0, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_DIVERT, NULL)) != NULL) {
+               struct divert_tag       *div_tag;
+               
+               div_tag = (struct divert_tag *)(tag+1);
+               args.divert_rule = div_tag->cookie;
+
+               m_tag_delete(m0, tag);
+       }
+       if ((tag = m_tag_locate(m0, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_IPFORWARD, NULL)) != NULL) {
+               struct ip_fwd_tag       *ipfwd_tag;
+               
+               ipfwd_tag = (struct ip_fwd_tag *)(tag+1);
+               args.next_hop = ipfwd_tag->next_hop;
+               
+               m_tag_delete(m0, tag);
+       }
+
+       m = m0;
+       
+#if    DIAGNOSTIC
+       if ( !m || (m->m_flags & M_PKTHDR) != 0)
+               panic("ip_output no HDR");
+       if (!ro)
+               panic("ip_output no route, proto = %d",
+                     mtod(m, struct ip *)->ip_p);
 #endif
+
+       if (args.rule != NULL) {        /* dummynet already saw us */
             ip = mtod(m, struct ip *);
             hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
             if (ro->ro_rt != NULL)
                 ia = (struct in_ifaddr *)ro->ro_rt->rt_ifa;
-            goto sendit;
-        } else
-            rule = NULL ;
+                       if (ia)
+                               ifaref(&ia->ia_ifa);
+#if IPSEC
+                       if (ipsec_bypass == 0 && (flags & IP_NOIPSEC) == 0) {
+                               so = ipsec_getsocket(m);
+                               (void)ipsec_setsocket(m, NULL);
+                       }
 #endif
+            goto sendit;
+       }
+
 #if IPSEC
-       if (ipsec_bypass == 0) {
+       if (ipsec_bypass == 0 && (flags & IP_NOIPSEC) == 0) {
                so = ipsec_getsocket(m);
                (void)ipsec_setsocket(m, NULL);
        }
 #endif
+loopit:        
+       /*
+        * No need to proccess packet twice if we've 
+        * already seen it
+        */
+       inject_filter_ref = ipf_get_inject_filter(m);
 
-#if    DIAGNOSTIC
-       if ((m->m_flags & M_PKTHDR) == 0)
-               panic("ip_output no HDR");
-       if (!ro)
-               panic("ip_output no route, proto = %d",
-                     mtod(m, struct ip *)->ip_p);
-#endif
        if (opt) {
                m = ip_insertoptions(m, opt, &len);
                hlen = len;
        }
        ip = mtod(m, struct ip *);
+       pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
+
        /*
         * Fill in IP header.
         */
@@ -271,20 +320,34 @@ ip_output(m0, opt, ro, flags, imo)
                     ip->ip_src.s_addr, ip->ip_p, ip->ip_off, ip->ip_len);
 
        dst = (struct sockaddr_in *)&ro->ro_dst;
+
        /*
         * If there is a cached route,
         * check that it is to the same destination
         * and is still up.  If not, free it and try again.
+        * The address family should also be checked in case of sharing the
+        * cache with IPv6.
         */
+
+       {
+               if (ro->ro_rt && (ro->ro_rt->generation_id != route_generation) &&
+                       ((flags & (IP_ROUTETOIF | IP_FORWARDING)) == 0) && (ip->ip_src.s_addr != INADDR_ANY) &&
+                       (ifa_foraddr(ip->ip_src.s_addr) == 0)) {
+                       error = EADDRNOTAVAIL;
+                       goto bad;
+               }
+       }
        if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
-          dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
+          dst->sin_family != AF_INET ||
+          dst->sin_addr.s_addr != pkt_dst.s_addr)) {
                rtfree(ro->ro_rt);
                ro->ro_rt = (struct rtentry *)0;
        }
        if (ro->ro_rt == 0) {
+               bzero(dst, sizeof(*dst));
                dst->sin_family = AF_INET;
                dst->sin_len = sizeof(*dst);
-               dst->sin_addr = ip->ip_dst;
+               dst->sin_addr = pkt_dst;
        }
        /*
         * If routing to interface only,
@@ -293,14 +356,16 @@ ip_output(m0, opt, ro, flags, imo)
 #define ifatoia(ifa)   ((struct in_ifaddr *)(ifa))
 #define sintosa(sin)   ((struct sockaddr *)(sin))
        if (flags & IP_ROUTETOIF) {
-               if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 &&
-                   (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
-                       ipstat.ips_noroute++;
-                       error = ENETUNREACH;
-                       goto bad;
+               if (ia)
+                       ifafree(&ia->ia_ifa);
+               if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0) {
+                       if ((ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
+                               ipstat.ips_noroute++;
+                               error = ENETUNREACH;
+                               goto bad;
+                       }
                }
                ifp = ia->ia_ifp;
-               dl_tag = ia->ia_ifa.ifa_dlt;
                ip->ip_ttl = 1;
                isbroadcast = in_broadcast(dst->sin_addr, ifp);
        } else {
@@ -320,9 +385,12 @@ ip_output(m0, opt, ro, flags, imo)
                        error = EHOSTUNREACH;
                        goto bad;
                }
+               if (ia)
+                       ifafree(&ia->ia_ifa);
                ia = ifatoia(ro->ro_rt->rt_ifa);
+               if (ia)
+                       ifaref(&ia->ia_ifa);
                ifp = ro->ro_rt->rt_ifp;
-               dl_tag = ro->ro_rt->rt_dlt;
                ro->ro_rt->rt_use++;
                if (ro->ro_rt->rt_flags & RTF_GATEWAY)
                        dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
@@ -331,7 +399,7 @@ ip_output(m0, opt, ro, flags, imo)
                else
                        isbroadcast = in_broadcast(dst->sin_addr, ifp);
        }
-       if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
+       if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
                struct in_multi *inm;
 
                m->m_flags |= M_MCAST;
@@ -345,16 +413,16 @@ ip_output(m0, opt, ro, flags, imo)
                 * See if the caller provided any multicast options
                 */
                if (imo != NULL) {
-                       ip->ip_ttl = imo->imo_multicast_ttl;
+                       if ((flags & IP_RAWOUTPUT) == 0) ip->ip_ttl = imo->imo_multicast_ttl;
                        if (imo->imo_multicast_ifp != NULL) {
                                ifp = imo->imo_multicast_ifp;
-                               dl_tag = ifp->if_data.default_proto;
                        }
-                       if (imo->imo_multicast_vif != -1)
+                       if (imo->imo_multicast_vif != -1 && 
+                               ((flags & IP_RAWOUTPUT) == 0 || ip->ip_src.s_addr == INADDR_ANY))
                                ip->ip_src.s_addr =
-                                   ip_mcast_src(imo->imo_multicast_vif);
+                                       ip_mcast_src(imo->imo_multicast_vif);
                } else
-                       ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
+                       if ((flags & IP_RAWOUTPUT) == 0) ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
                /*
                 * Confirm that the outgoing interface supports multicast.
                 */
@@ -375,11 +443,18 @@ ip_output(m0, opt, ro, flags, imo)
                        TAILQ_FOREACH(ia1, &in_ifaddrhead, ia_link)
                                if (ia1->ia_ifp == ifp) {
                                        ip->ip_src = IA_SIN(ia1)->sin_addr;
+                                       
                                        break;
                                }
+                       if (ip->ip_src.s_addr == INADDR_ANY) {
+                               error = ENETUNREACH;
+                               goto bad;
+                       }
                }
 
-               IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
+               ifnet_lock_shared(ifp);
+               IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
+               ifnet_lock_done(ifp);
                if (inm != NULL &&
                   (imo == NULL || imo->imo_multicast_loop)) {
                        /*
@@ -387,6 +462,42 @@ ip_output(m0, opt, ro, flags, imo)
                         * on the outgoing interface, and the caller did not
                         * forbid loopback, loop back a copy.
                         */
+                       if (!TAILQ_EMPTY(&ipv4_filters)) {
+                               struct ipfilter *filter;
+                               int seen = (inject_filter_ref == 0);
+                               struct ipf_pktopts *ippo = 0, ipf_pktopts;
+
+                               if (imo) {                                              
+                                       ippo = &ipf_pktopts;
+                                       ipf_pktopts.ippo_mcast_ifnet = imo->imo_multicast_ifp;
+                                       ipf_pktopts.ippo_mcast_ttl = imo->imo_multicast_ttl;
+                                       ipf_pktopts.ippo_mcast_loop = imo->imo_multicast_loop;
+                               }
+                               
+                               lck_mtx_unlock(ip_mutex);
+                               ipf_ref();
+                               TAILQ_FOREACH(filter, &ipv4_filters, ipf_link) {
+                                       if (seen == 0) {
+                                               if ((struct ipfilter *)inject_filter_ref == filter)
+                                                       seen = 1;
+                                       } else if (filter->ipf_filter.ipf_output) {
+                                               errno_t result;
+                                               result = filter->ipf_filter.ipf_output(filter->ipf_filter.cookie, (mbuf_t*)&m, ippo);
+                                               if (result == EJUSTRETURN) {
+                                                       ipf_unref();
+                                                       goto done;
+                                               }
+                                               if (result != 0) {
+                                                       ipf_unref();
+                                                       lck_mtx_lock(ip_mutex);
+                                                       goto bad;
+                                               }
+                                       }
+                               }
+                               lck_mtx_lock(ip_mutex);
+                               ipf_unref();
+                               didfilter = 1;
+                       }
                        ip_mloopback(ifp, m, dst, hlen);
                }
                else {
@@ -413,6 +524,7 @@ ip_output(m0, opt, ro, flags, imo)
                                  imo = NULL;
                                if (ip_mforward(ip, ifp, m, imo) != 0) {
                                        m_freem(m);
+                                       lck_mtx_unlock(ip_mutex);
                                        goto done;
                                }
                        }
@@ -428,6 +540,7 @@ ip_output(m0, opt, ro, flags, imo)
                 */
                if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
                        m_freem(m);
+                       lck_mtx_unlock(ip_mutex);
                        goto done;
                }
 
@@ -450,15 +563,6 @@ ip_output(m0, opt, ro, flags, imo)
 #endif /* IPFIREWALL_FORWARD */
        }
 #endif /* notdef */
-       /*
-        * Verify that we have any chance at all of being able to queue
-        *      the packet or packet fragments
-        */
-       if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
-               ifp->if_snd.ifq_maxlen) {
-                       error = ENOBUFS;
-                       goto bad;
-       }
 
        /*
         * Look for broadcast address and
@@ -496,12 +600,45 @@ sendit:
                }
         }
 
+injectit:
+       if (!didfilter && !TAILQ_EMPTY(&ipv4_filters)) {
+               struct ipfilter *filter;
+               int seen = (inject_filter_ref == 0);
+               
+               lck_mtx_unlock(ip_mutex);
+               ipf_ref();
+               TAILQ_FOREACH(filter, &ipv4_filters, ipf_link) {
+                       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, 0);
+                               if (result == EJUSTRETURN) {
+                                       ipf_unref();
+                                       goto done;
+                               }
+                               if (result != 0) {
+                                       ipf_unref();
+                                       lck_mtx_lock(ip_mutex);
+                                       goto bad;
+                               }
+                       }
+               }
+               ipf_unref();
+               lck_mtx_lock(ip_mutex);
+       }
+
 #if IPSEC
        /* temporary for testing only: bypass ipsec alltogether */
 
-       if (ipsec_bypass != 0)
+       if (ipsec_bypass != 0 || (flags & IP_NOIPSEC) != 0)
                goto skip_ipsec;
 
+       KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_START, 0,0,0,0,0);
+
+       lck_mtx_lock(sadb_mutex);
+
        /* get SP for this packet */
        if (so == NULL)
                sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
@@ -510,6 +647,8 @@ sendit:
 
        if (sp == NULL) {
                ipsecstat.out_inval++;
+               KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 0,0,0,0,0);
+               lck_mtx_unlock(sadb_mutex);
                goto bad;
        }
 
@@ -522,17 +661,23 @@ sendit:
                 * This packet is just discarded.
                 */
                ipsecstat.out_polvio++;
+               KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 1,0,0,0,0);
+               lck_mtx_unlock(sadb_mutex);
                goto bad;
 
        case IPSEC_POLICY_BYPASS:
        case IPSEC_POLICY_NONE:
                /* no need to do IPsec. */
+               KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 2,0,0,0,0);
+               lck_mtx_unlock(sadb_mutex);
                goto skip_ipsec;
        
        case IPSEC_POLICY_IPSEC:
                if (sp->req == NULL) {
                        /* acquire a policy */
                        error = key_spdacquire(sp);
+                       KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 3,0,0,0,0);
+                       lck_mtx_unlock(sadb_mutex);
                        goto bad;
                }
                break;
@@ -566,9 +711,13 @@ sendit:
        HTONS(ip->ip_len);
        HTONS(ip->ip_off);
 
+       lck_mtx_unlock(ip_mutex);
        error = ipsec4_output(&state, sp, flags);
-
-       m = state.m;
+       lck_mtx_unlock(sadb_mutex);
+       lck_mtx_lock(ip_mutex);
+    
+       m0 = m = state.m;
+       
        if (flags & IP_ROUTETOIF) {
                /*
                 * if we have tunnel mode SA, we may need to ignore
@@ -580,6 +729,7 @@ sendit:
                }
        } else
                ro = state.ro;
+
        dst = (struct sockaddr_in *)state.dst;
        if (error) {
                /* mbuf is already reclaimed in ipsec4_output. */
@@ -599,33 +749,77 @@ sendit:
                        error = 0;
                        break;
                }
+               KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 4,0,0,0,0);
                goto bad;
        }
     }
 
        /* be sure to update variables that are affected by ipsec4_output() */
        ip = mtod(m, struct ip *);
+       
 #ifdef _IP_VHL
        hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 #else
        hlen = ip->ip_hl << 2;
 #endif
+       /* Check that there wasn't a route change and src is still valid */
+
+       if (ro->ro_rt->generation_id != route_generation) {
+               if (ifa_foraddr(ip->ip_src.s_addr) == 0 && ((flags & (IP_ROUTETOIF | IP_FORWARDING)) == 0)) {
+                       error = EADDRNOTAVAIL;
+                       KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 5,0,0,0,0);
+                       goto bad;
+               }
+               rtfree(ro->ro_rt);
+               ro->ro_rt = NULL;
+       }
+
        if (ro->ro_rt == NULL) {
                if ((flags & IP_ROUTETOIF) == 0) {
                        printf("ip_output: "
                                "can't update route after IPsec processing\n");
-                       error = EHOSTUNREACH;   /*XXX*/
+                       error = EHOSTUNREACH;   /*XXX*/ 
+                       KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 6,0,0,0,0);
                        goto bad;
                }
        } else {
+               if (ia)
+                       ifafree(&ia->ia_ifa);
                ia = ifatoia(ro->ro_rt->rt_ifa);
+               if (ia)
+                       ifaref(&ia->ia_ifa);
                ifp = ro->ro_rt->rt_ifp;
-               dl_tag = ia->ia_ifa.ifa_dlt;
        }
 
        /* make it flipped, again. */
        NTOHS(ip->ip_len);
        NTOHS(ip->ip_off);
+       KERNEL_DEBUG(DBG_FNC_IPSEC4_OUTPUT | DBG_FUNC_END, 7,0xff,0xff,0xff,0xff);
+       
+       /* Pass to filters again */
+       if (!TAILQ_EMPTY(&ipv4_filters)) {
+               struct ipfilter *filter;
+               
+               lck_mtx_unlock(ip_mutex);
+               ipf_ref();
+               TAILQ_FOREACH(filter, &ipv4_filters, ipf_link) {
+                       if (filter->ipf_filter.ipf_output) {
+                               errno_t result;
+                               result = filter->ipf_filter.ipf_output(filter->ipf_filter.cookie, (mbuf_t*)&m, 0);
+                               if (result == EJUSTRETURN) {
+                                       ipf_unref();
+                                       goto done;
+                               }
+                               if (result != 0) {
+                                       ipf_unref();
+                                       lck_mtx_lock(ip_mutex);
+                                       goto bad;
+                               }
+                       }
+               }
+               ipf_unref();
+               lck_mtx_lock(ip_mutex);
+       }
 skip_ipsec:
 #endif /*IPSEC*/
 
@@ -639,19 +833,28 @@ skip_ipsec:
        if (fr_checkp) {
                struct  mbuf    *m1 = m;
 
-               if ((error = (*fr_checkp)(ip, hlen, ifp, 1, &m1)) || !m1)
+               if ((error = (*fr_checkp)(ip, hlen, ifp, 1, &m1)) || !m1) {
+                       lck_mtx_unlock(ip_mutex);
                        goto done;
-               ip = mtod(m = m1, struct ip *);
+               }
+               ip = mtod(m0 = m = m1, struct ip *);
        }
 
        /*
         * Check with the firewall...
+        * but not if we are already being fwd'd from a firewall.
         */
-       if (fw_enable && ip_fw_chk_ptr) {
+       if (fw_enable && IPFW_LOADED && !args.next_hop) {
                struct sockaddr_in *old = dst;
 
-               off = (*ip_fw_chk_ptr)(&ip,
-                   hlen, ifp, &divert_cookie, &m, &rule, &dst);
+               args.m = m;
+               args.next_hop = dst;
+               args.oif = ifp;
+               lck_mtx_unlock(ip_mutex);
+               off = ip_fw_chk_ptr(&args);
+               m = args.m;
+               dst = args.next_hop;
+
                 /*
                  * On return we must do the following:
                  * IP_FW_PORT_DENY_FLAG                -> drop the pkt (XXX new)
@@ -666,6 +869,7 @@ skip_ipsec:
                  * unsupported rules), but better play safe and drop
                  * packets in case of doubt.
                  */
+               m0 = m;
                if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) {
                        if (m)
                                m_freem(m);
@@ -673,10 +877,12 @@ skip_ipsec:
                        goto done ;
                }
                ip = mtod(m, struct ip *);
-               if (off == 0 && dst == old) /* common case */
+               if (off == 0 && dst == old) {/* common case */
+                       lck_mtx_lock(ip_mutex);
                        goto pass ;
+               }
 #if DUMMYNET
-                if ((off & IP_FW_PORT_DYNT_FLAG) != 0) {
+                if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) {
                     /*
                      * pass the pkt to dummynet. Need to include
                      * pipe number, m, ifp, ro, dst because these are
@@ -686,11 +892,16 @@ skip_ipsec:
                      * XXX note: if the ifp or ro entry are deleted
                      * while a pkt is in dummynet, we are in trouble!
                      */ 
-                   error = dummynet_io(off & 0xffff, DN_TO_IP_OUT, m,
-                               ifp,ro,dst,rule, flags);
+                   args.ro = ro;
+                   args.dst = dst;
+                   args.flags = flags;
+
+                   error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT,
+                               &args);
                    goto done;
                }
-#endif   
+#endif /* DUMMYNET */
+               lck_mtx_lock(ip_mutex);
 #if IPDIVERT
                if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) {
                        struct mbuf *clone = NULL;
@@ -713,15 +924,15 @@ skip_ipsec:
                        HTONS(ip->ip_off);
 
                        /* Deliver packet to divert input routine */
-                       ip_divert_cookie = divert_cookie;
-                       divert_packet(m, 0, off & 0xffff);
+                       divert_packet(m, 0, off & 0xffff, args.divert_rule);
 
                        /* If 'tee', continue with original packet */
                        if (clone != NULL) {
-                               m = clone;
+                               m0 = m = clone;
                                ip = mtod(m, struct ip *);
                                goto pass;
                        }
+                       lck_mtx_unlock(ip_mutex);
                        goto done;
                }
 #endif
@@ -737,7 +948,7 @@ skip_ipsec:
                 * And I'm babbling.
                 */
                if (off == 0 && old != dst) {
-                       struct in_ifaddr *ia;
+                       struct in_ifaddr *ia_fw;
 
                        /* It's changed... */
                        /* There must be a better way to do this next line... */
@@ -760,31 +971,65 @@ skip_ipsec:
                         * as the packet runs through ip_input() as
                         * it is done through a ISR.
                         */
-                       TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
+                       TAILQ_FOREACH(ia_fw, &in_ifaddrhead, ia_link) {
                                /*
                                 * If the addr to forward to is one
                                 * of ours, we pretend to
                                 * be the destination for this packet.
                                 */
-                               if (IA_SIN(ia)->sin_addr.s_addr ==
+                               if (IA_SIN(ia_fw)->sin_addr.s_addr ==
                                                 dst->sin_addr.s_addr)
                                        break;
                        }
                        if (ia) {
                                /* tell ip_input "dont filter" */
-                               ip_fw_fwd_addr = dst;
+                               struct m_tag            *fwd_tag;
+                               struct ip_fwd_tag       *ipfwd_tag;
+                               
+                               fwd_tag = m_tag_alloc(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_IPFORWARD,
+                                               sizeof(struct sockaddr_in), M_NOWAIT);
+                               if (fwd_tag == NULL) {
+                                       error = ENOBUFS;
+                                       goto bad;
+                               }
+                               
+                               ipfwd_tag = (struct ip_fwd_tag *)(fwd_tag+1);
+                               ipfwd_tag->next_hop = args.next_hop;
+
+                               m_tag_prepend(m, fwd_tag);
+
                                if (m->m_pkthdr.rcvif == NULL)
                                        m->m_pkthdr.rcvif = ifunit("lo0");
-                               if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
+                               if ((~IF_HWASSIST_CSUM_FLAGS(m->m_pkthdr.rcvif->if_hwassist) & 
+                                               m->m_pkthdr.csum_flags) == 0) {
+                                       if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
+                                               m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
+                                               m->m_pkthdr.csum_flags |=
+                                                       CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
+                                               m->m_pkthdr.csum_data = 0xffff;
+                                       }
                                        m->m_pkthdr.csum_flags |=
-                                           CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
-                                       m0->m_pkthdr.csum_data = 0xffff;
+                                               CSUM_IP_CHECKED | CSUM_IP_VALID;
+                               }
+                               else if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
+                                       in_delayed_cksum(m);
+                                       m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
+                                       ip->ip_sum = in_cksum(m, hlen);
                                }
-                               m->m_pkthdr.csum_flags |=
-                                   CSUM_IP_CHECKED | CSUM_IP_VALID;
                                HTONS(ip->ip_len);
                                HTONS(ip->ip_off);
-                               ip_input(m);
+                               
+                               lck_mtx_unlock(ip_mutex);
+                               
+                               /*  we need to call dlil_output to run filters
+                                *      and resync to avoid recursion loops.
+                                */
+                               if (lo_ifp) {
+                                       dlil_output(lo_ifp, PF_INET, m, 0, (struct sockaddr *)dst, 0);
+                               }
+                               else {
+                                       printf("ip_output: no loopback ifp for forwarding!!!\n");
+                               }
                                goto done;
                        }
                        /* Some of the logic for this was
@@ -804,9 +1049,8 @@ skip_ipsec:
                                goto bad;
                        }
 
-                       ia = ifatoia(ro_fwd->ro_rt->rt_ifa);
+                       ia_fw = ifatoia(ro_fwd->ro_rt->rt_ifa);
                        ifp = ro_fwd->ro_rt->rt_ifp;
-                       dl_tag = ro_fwd->ro_rt->rt_dlt;
                        ro_fwd->ro_rt->rt_use++;
                        if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
                                dst = (struct sockaddr_in *)ro_fwd->ro_rt->rt_gateway;
@@ -825,7 +1069,7 @@ skip_ipsec:
                         * interface, do it again, from the new one.
                         */
                        if (fwd_rewrite_src)
-                               ip->ip_src = IA_SIN(ia)->sin_addr;
+                               ip->ip_src = IA_SIN(ia_fw)->sin_addr;
                        goto pass ;
                }
 #endif /* IPFIREWALL_FORWARD */
@@ -834,13 +1078,34 @@ skip_ipsec:
                  * we have to drop the pkt
                  */
                m_freem(m);
-                error = EACCES; /* not sure this is the right error msg */
-                goto done;
+               error = EACCES; /* not sure this is the right error msg */
+               lck_mtx_unlock(ip_mutex);
+               goto done;
        }
 
 pass:
+#if __APPLE__
+       /* Do not allow loopback address to wind up on a wire */
+       if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
+                ((ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
+                 (ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)) {
+               ipstat.ips_badaddr++;
+               m_freem(m);
+               /* 
+                * Do not simply drop the packet just like a firewall -- we want the 
+                * the application to feel the pain.
+                * Return ENETUNREACH like ip6_output does in some similar cases. 
+                * This can startle the otherwise clueless process that specifies
+                * loopback as the source address.
+                */
+               error = ENETUNREACH;
+               lck_mtx_unlock(ip_mutex);
+               goto done;
+       }
+#endif
        m->m_pkthdr.csum_flags |= CSUM_IP;
-       sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
+       sw_csum = m->m_pkthdr.csum_flags 
+               & ~IF_HWASSIST_CSUM_FLAGS(ifp->if_hwassist);
 
        if ((ifp->if_hwassist & CSUM_TCP_SUM16) != 0) {
                /*
@@ -870,7 +1135,7 @@ pass:
                m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
        }
        
-       m->m_pkthdr.csum_flags &= ifp->if_hwassist;
+       m->m_pkthdr.csum_flags &= IF_HWASSIST_CSUM_FLAGS(ifp->if_hwassist);
 
        /*
         * If small enough for interface, or the interface will take
@@ -895,17 +1160,32 @@ pass:
 
 #if IPSEC
                /* clean ipsec history once it goes out of the node */
-               if (ipsec_bypass == 0)
+               if (ipsec_bypass == 0 && (flags & IP_NOIPSEC) == 0)
                        ipsec_delaux(m);
 #endif
-#if __APPLE__
-               error = dlil_output(dl_tag, m, (void *) ro->ro_rt,
+               if (packetchain == 0) {
+                       lck_mtx_unlock(ip_mutex);
+                       error = dlil_output(ifp, PF_INET, m, (void *) ro->ro_rt,
                                    (struct sockaddr *)dst, 0);
-#else
-               error = (*ifp->if_output)(ifp, m,
-                               (struct sockaddr *)dst, ro->ro_rt);
-#endif
-               goto done;
+                       goto done;      
+               }
+               else { /* packet chaining allows us to reuse the route for all packets */
+                       m = m->m_nextpkt;
+                       if (m == NULL) {
+                               if (pktcnt > ip_maxchainsent)
+                                       ip_maxchainsent = pktcnt;
+                               //send
+                               lck_mtx_unlock(ip_mutex);
+                               error = dlil_output_list(ifp, PF_INET, packetlist, (void *) ro->ro_rt,
+                                       (struct sockaddr *)dst, 0);
+                               pktcnt = 0;
+                               goto done;
+       
+                       }
+                       m0 = m;
+                       pktcnt++;
+                       goto loopit;
+               }
        }
        /*
         * Too large for interface; fragment if possible.
@@ -941,8 +1221,10 @@ pass:
        if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
            (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) {
                in_delayed_cksum(m);
-               if (m == NULL)
+               if (m == NULL) {
+                       lck_mtx_unlock(ip_mutex);
                        return(ENOMEM);
+               }
                m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
        }
 
@@ -990,8 +1272,9 @@ pass:
                        goto sendorfree;
                }
                m->m_pkthdr.len = mhlen + len;
-               m->m_pkthdr.rcvif = (struct ifnet *)0;
+               m->m_pkthdr.rcvif = 0;
                m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
+               m->m_pkthdr.socket_id = m0->m_pkthdr.socket_id;
                HTONS(mhip->ip_off);
                mhip->ip_sum = 0;
                if (sw_csum & CSUM_DELAY_IP) {
@@ -1027,12 +1310,13 @@ sendorfree:
        KERNEL_DEBUG(DBG_LAYER_END, ip->ip_dst.s_addr, 
                     ip->ip_src.s_addr, ip->ip_p, ip->ip_off, ip->ip_len);
 
+       lck_mtx_unlock(ip_mutex);
        for (m = m0; m; m = m0) {
                m0 = m->m_nextpkt;
                m->m_nextpkt = 0;
 #if IPSEC
                /* clean ipsec history once it goes out of the node */
-               if (ipsec_bypass == 0)
+               if (ipsec_bypass == 0 && (flags & IP_NOIPSEC) == 0)
                        ipsec_delaux(m);
 #endif
                if (error == 0) {
@@ -1043,14 +1327,10 @@ sendorfree:
                                ia->ia_ifa.if_obytes += m->m_pkthdr.len;
                        }
 #endif
-                       
-#if __APPLE__
-                   error = dlil_output(dl_tag, m, (void *) ro->ro_rt,
+                   if ((packetchain != 0)  && (pktcnt > 0))    
+                               panic("ip_output: mix of packet in packetlist is wrong=%x", packetlist);
+                   error = dlil_output(ifp, PF_INET, m, (void *) ro->ro_rt,
                                        (struct sockaddr *)dst, 0);
-#else
-                       error = (*ifp->if_output)(ifp, m,
-                           (struct sockaddr *)dst, ro->ro_rt);
-#endif
                } else
                        m_freem(m);
        }
@@ -1059,8 +1339,12 @@ sendorfree:
                ipstat.ips_fragmented++;
     }
 done:
+       if (ia) {
+               ifafree(&ia->ia_ifa);
+               ia = NULL;
+       }
 #if IPSEC
-       if (ipsec_bypass == 0) {
+       if (ipsec_bypass == 0 && (flags & IP_NOIPSEC) == 0) {
        if (ro == &iproute && ro->ro_rt) {
                rtfree(ro->ro_rt);
                ro->ro_rt = NULL;
@@ -1068,7 +1352,9 @@ done:
        if (sp != NULL) {
                KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
                        printf("DP ip_output call free SP:%x\n", sp));
+               lck_mtx_lock(sadb_mutex);
                key_freesp(sp);
+               lck_mtx_unlock(sadb_mutex);
        }
        }
 #endif /* IPSEC */
@@ -1077,36 +1363,128 @@ done:
        return (error);
 bad:
        m_freem(m0);
+       lck_mtx_unlock(ip_mutex);
        goto done;
 }
 
 void
-in_delayed_cksum(struct mbuf *m)
+in_delayed_cksum_offset(struct mbuf *m, int ip_offset)
 {
        struct ip *ip;
        u_short csum, offset;
-       ip = mtod(m, struct ip *);
+       
+       while (ip_offset > m->m_len) {
+               ip_offset -= m->m_len;
+               m = m->m_next;
+               if (m) {
+                       printf("in_delayed_cksum_withoffset failed - ip_offset wasn't in the packet\n");
+                       return;
+               }
+       }
+       
+       if (ip_offset + sizeof(struct ip) > m->m_len) {
+               printf("delayed m_pullup, m->len: %d  off: %d\n",
+                       m->m_len, ip_offset);
+               /*
+                * XXX
+                * this shouldn't happen
+                */
+               m = m_pullup(m, ip_offset + sizeof(struct ip));
+       }
+       
+       /* Gross */
+       if (ip_offset) {
+               m->m_len -= ip_offset;
+               m->m_data += ip_offset;
+       }
+       
+       ip = mtod(m, struct ip*);
        offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
        csum = in_cksum_skip(m, ip->ip_len, offset);
        if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
                csum = 0xffff;
        offset += m->m_pkthdr.csum_data & 0xFFFF;        /* checksum offset */
+       
+       /* Gross */
+       if (ip_offset) {
+               if (M_LEADINGSPACE(m) < ip_offset)
+                       panic("in_delayed_cksum_withoffset - chain modified!\n");
+               m->m_len += ip_offset;
+               m->m_data -= ip_offset;
+       }
 
        if (offset > ip->ip_len) /* bogus offset */
                return;
 
-       if (offset + sizeof(u_short) > m->m_len) {
+       if (offset + ip_offset + sizeof(u_short) > m->m_len) {
                printf("delayed m_pullup, m->len: %d  off: %d  p: %d\n",
-                   m->m_len, offset, ip->ip_p);
+                   m->m_len, offset + ip_offset, ip->ip_p);
                /*
                 * XXX
                 * this shouldn't happen, but if it does, the
                 * correct behavior may be to insert the checksum
                 * in the existing chain instead of rearranging it.
                 */
-               m = m_pullup(m, offset + sizeof(u_short));
+               m = m_pullup(m, offset + ip_offset + sizeof(u_short));
+       }
+       *(u_short *)(m->m_data + offset + ip_offset) = csum;
+}
+
+void
+in_delayed_cksum(struct mbuf *m)
+{
+       in_delayed_cksum_offset(m, 0);
+}
+
+void
+in_cksum_offset(struct mbuf* m, size_t ip_offset)
+{
+       struct ip* ip = NULL;
+       int hlen = 0;
+       
+       while (ip_offset > m->m_len) {
+               ip_offset -= m->m_len;
+               m = m->m_next;
+               if (m) {
+                       printf("in_cksum_offset failed - ip_offset wasn't in the packet\n");
+                       return;
+               }
+       }
+       
+       if (ip_offset + sizeof(struct ip) > m->m_len) {
+               printf("in_cksum_offset - delayed m_pullup, m->len: %d  off: %d\n",
+                       m->m_len, ip_offset);
+               /*
+                * XXX
+                * this shouldn't happen
+                */
+               m = m_pullup(m, ip_offset + sizeof(struct ip));
+       }
+       
+       /* Gross */
+       if (ip_offset) {
+               m->m_len -= ip_offset;
+               m->m_data += ip_offset;
+       }
+
+       ip = mtod(m, struct ip*);
+
+#ifdef _IP_VHL
+       hlen = IP_VHL_HL(ip->ip_vhl) << 2;
+#else
+       hlen = ip->ip_hl << 2;
+#endif
+       
+       ip->ip_sum = 0;
+       ip->ip_sum = in_cksum(m, hlen);
+       
+       /* Gross */
+       if (ip_offset) {
+               if (M_LEADINGSPACE(m) < ip_offset)
+                       panic("in_cksum_offset - chain modified!\n");
+               m->m_len += ip_offset;
+               m->m_data -= ip_offset;
        }
-       *(u_short *)(m->m_data + offset) = csum;
 }
 
 /*
@@ -1136,7 +1514,7 @@ ip_insertoptions(m, opt, phlen)
                MGETHDR(n, M_DONTWAIT, MT_HEADER);
                if (n == 0)
                        return (m);
-               n->m_pkthdr.rcvif = (struct ifnet *)0;
+               n->m_pkthdr.rcvif = 0;
                n->m_pkthdr.len = m->m_pkthdr.len + optlen;
                m->m_len -= sizeof(struct ip);
                m->m_data += sizeof(struct ip);
@@ -1255,6 +1633,7 @@ ip_ctloutput(so, sopt)
                case IP_RECVRETOPTS:
                case IP_RECVDSTADDR:
                case IP_RECVIF:
+               case IP_RECVTTL:
 #if defined(NFAITH) && NFAITH > 0
                case IP_FAITH:
 #endif
@@ -1293,6 +1672,10 @@ ip_ctloutput(so, sopt)
                                OPTSET(INP_RECVIF);
                                break;
 
+                       case IP_RECVTTL:
+                               OPTSET(INP_RECVTTL);
+                               break;
+
 #if defined(NFAITH) && NFAITH > 0
                        case IP_FAITH:
                                OPTSET(INP_FAITH);
@@ -1348,19 +1731,24 @@ ip_ctloutput(so, sopt)
                        struct mbuf *m;
                        int optname;
 
+                        if (sopt->sopt_valsize > MCLBYTES) {
+                                error = EMSGSIZE;
+                                break;
+                        }
                        if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
                                break;
                        if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
                                break;
                        priv = (sopt->sopt_p != NULL &&
-                               suser(sopt->sopt_p->p_ucred,
-                                     &sopt->sopt_p->p_acflag) != 0) ? 0 : 1;
+                               proc_suser(sopt->sopt_p) != 0) ? 0 : 1;
                        if (m) {
                                req = mtod(m, caddr_t);
                                len = m->m_len;
                        }
                        optname = sopt->sopt_name;
+                       lck_mtx_lock(sadb_mutex);
                        error = ipsec4_set_policy(inp, optname, req, len, priv);
+                       lck_mtx_unlock(sadb_mutex);
                        m_freem(m);
                        break;
                }
@@ -1391,6 +1779,7 @@ ip_ctloutput(so, sopt)
                case IP_RECVRETOPTS:
                case IP_RECVDSTADDR:
                case IP_RECVIF:
+               case IP_RECVTTL:
                case IP_PORTRANGE:
 #if defined(NFAITH) && NFAITH > 0
                case IP_FAITH:
@@ -1423,6 +1812,10 @@ ip_ctloutput(so, sopt)
                                optval = OPTBIT(INP_RECVIF);
                                break;
 
+                       case IP_RECVTTL:
+                               optval = OPTBIT(INP_RECVTTL);
+                               break;
+
                        case IP_PORTRANGE:
                                if (inp->inp_flags & INP_HIGHPORT)
                                        optval = IP_PORTRANGE_HIGH;
@@ -1461,7 +1854,9 @@ ip_ctloutput(so, sopt)
                                req = mtod(m, caddr_t);
                                len = m->m_len;
                        }
+                       lck_mtx_lock(sadb_mutex);
                        error = ipsec4_get_policy(sotoinpcb(so), req, len, &m);
+                       lck_mtx_unlock(sadb_mutex);
                        if (error == 0)
                                error = soopt_mcopyout(sopt, m); /* XXX */
                        if (error == 0)
@@ -1607,9 +2002,13 @@ ip_multicast_if(a, ifindexp)
                *ifindexp = 0;
        if (ntohl(a->s_addr) >> 24 == 0) {
                ifindex = ntohl(a->s_addr) & 0xffffff;
-               if (ifindex < 0 || if_index < ifindex)
+               ifnet_head_lock_shared();
+               if (ifindex < 0 || if_index < ifindex) {
+                       ifnet_head_done();
                        return NULL;
+               }
                ifp = ifindex2ifnet[ifindex];
+               ifnet_head_done();
                if (ifindexp)
                        *ifindexp = ifindex;
        } else {
@@ -1632,28 +2031,17 @@ ip_setmoptions(sopt, imop)
        struct ip_mreq mreq;
        struct ifnet *ifp = NULL;
        struct ip_moptions *imo = *imop;
-       struct route ro;
-       struct sockaddr_in *dst;
        int ifindex;
-       int s;
 
        if (imo == NULL) {
                /*
                 * No multicast option buffer attached to the pcb;
                 * allocate one and initialize to default values.
                 */
-               imo = (struct ip_moptions*) _MALLOC(sizeof(*imo), M_IPMOPTS,
-                   M_WAITOK);
-
-               if (imo == NULL)
-                       return (ENOBUFS);
-               *imop = imo;
-               imo->imo_multicast_ifp = NULL;
-               imo->imo_multicast_addr.s_addr = INADDR_ANY;
-               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;
+               error = ip_createmoptions(imop);
+               if (error != 0)
+                       return error;
+               imo = *imop;
        }
 
        switch (sopt->sopt_name) {
@@ -1694,10 +2082,8 @@ ip_setmoptions(sopt, imop)
                 * IP address.  Find the interface and confirm that
                 * it supports multicasting.
                 */
-               s = splimp();
                ifp = ip_multicast_if(&addr, &ifindex);
                if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
-                       splx(s);
                        error = EADDRNOTAVAIL;
                        break;
                }
@@ -1706,7 +2092,6 @@ ip_setmoptions(sopt, imop)
                        imo->imo_multicast_addr = addr;
                else
                        imo->imo_multicast_addr.s_addr = INADDR_ANY;
-               splx(s);
                break;
 
        case IP_MULTICAST_TTL:
@@ -1766,78 +2151,8 @@ ip_setmoptions(sopt, imop)
                error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
                if (error)
                        break;
-
-               if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
-                       error = EINVAL;
-                       break;
-               }
-               s = splimp();
-               /*
-                * If no interface address was provided, use the interface of
-                * the route to the given multicast address.
-                */
-               if (mreq.imr_interface.s_addr == INADDR_ANY) {
-                       bzero((caddr_t)&ro, sizeof(ro));
-                       dst = (struct sockaddr_in *)&ro.ro_dst;
-                       dst->sin_len = sizeof(*dst);
-                       dst->sin_family = AF_INET;
-                       dst->sin_addr = mreq.imr_multiaddr;
-                       rtalloc(&ro);
-                       if (ro.ro_rt != NULL) {
-                               ifp = ro.ro_rt->rt_ifp;
-                               rtfree(ro.ro_rt);
-                       }
-                       else {
-                               /* If there's no default route, try using loopback */
-                               mreq.imr_interface.s_addr = INADDR_LOOPBACK;
-                       }
-               }
                
-               if (ifp == NULL) {
-                       ifp = ip_multicast_if(&mreq.imr_interface, NULL);
-               }
-
-               /*
-                * See if we found an interface, and confirm that it
-                * supports multicast.
-                */
-               if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
-                       error = EADDRNOTAVAIL;
-                       splx(s);
-                       break;
-               }
-               /*
-                * See if the membership already exists or if all the
-                * membership slots are full.
-                */
-               for (i = 0; i < imo->imo_num_memberships; ++i) {
-                       if (imo->imo_membership[i]->inm_ifp == ifp &&
-                           imo->imo_membership[i]->inm_addr.s_addr
-                                               == mreq.imr_multiaddr.s_addr)
-                               break;
-               }
-               if (i < imo->imo_num_memberships) {
-                       error = EADDRINUSE;
-                       splx(s);
-                       break;
-               }
-               if (i == IP_MAX_MEMBERSHIPS) {
-                       error = ETOOMANYREFS;
-                       splx(s);
-                       break;
-               }
-               /*
-                * Everything looks good; add a new record to the multicast
-                * address list for the given interface.
-                */
-               if ((imo->imo_membership[i] =
-                   in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
-                       error = ENOBUFS;
-                       splx(s);
-                       break;
-               }
-               ++imo->imo_num_memberships;
-               splx(s);
+               error = ip_addmembership(imo, &mreq);
                break;
 
        case IP_DROP_MEMBERSHIP:
@@ -1848,54 +2163,8 @@ ip_setmoptions(sopt, imop)
                error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
                if (error)
                        break;
-
-               if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
-                       error = EINVAL;
-                       break;
-               }
-
-               s = splimp();
-               /*
-                * If an interface address was specified, get a pointer
-                * to its ifnet structure.
-                */
-               if (mreq.imr_interface.s_addr == INADDR_ANY)
-                       ifp = NULL;
-               else {
-                       ifp = ip_multicast_if(&mreq.imr_interface, NULL);
-                       if (ifp == NULL) {
-                               error = EADDRNOTAVAIL;
-                               splx(s);
-                               break;
-                       }
-               }
-               /*
-                * Find the membership in the membership array.
-                */
-               for (i = 0; i < imo->imo_num_memberships; ++i) {
-                       if ((ifp == NULL ||
-                            imo->imo_membership[i]->inm_ifp == ifp) &&
-                            imo->imo_membership[i]->inm_addr.s_addr ==
-                            mreq.imr_multiaddr.s_addr)
-                               break;
-               }
-               if (i == imo->imo_num_memberships) {
-                       error = EADDRNOTAVAIL;
-                       splx(s);
-                       break;
-               }
-               /*
-                * Give up the multicast address record to which the
-                * membership points.
-                */
-               in_delmulti(imo->imo_membership[i]);
-               /*
-                * Remove the gap in the membership array.
-                */
-               for (++i; i < imo->imo_num_memberships; ++i)
-                       imo->imo_membership[i-1] = imo->imo_membership[i];
-               --imo->imo_num_memberships;
-               splx(s);
+               
+               error = ip_dropmembership(imo, &mreq);
                break;
 
        default:
@@ -1918,6 +2187,172 @@ ip_setmoptions(sopt, imop)
        return (error);
 }
 
+/*
+ * Set the IP multicast options in response to user setsockopt().
+ */
+__private_extern__ int
+ip_createmoptions(
+       struct ip_moptions **imop)
+{
+       struct ip_moptions *imo;
+       imo = (struct ip_moptions*) _MALLOC(sizeof(*imo), M_IPMOPTS,
+               M_WAITOK);
+
+       if (imo == NULL)
+               return (ENOBUFS);
+       *imop = imo;
+       imo->imo_multicast_ifp = NULL;
+       imo->imo_multicast_addr.s_addr = INADDR_ANY;
+       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;
+       
+       return 0;
+}
+
+/*
+ * Add membership to an IPv4 multicast.
+ */
+__private_extern__ int
+ip_addmembership(
+       struct ip_moptions *imo,
+       struct ip_mreq *mreq)
+{
+       struct route ro;
+       struct sockaddr_in *dst;
+       struct ifnet *ifp = NULL;
+       int error = 0;
+       int i;
+       
+       if (!IN_MULTICAST(ntohl(mreq->imr_multiaddr.s_addr))) {
+               error = EINVAL;
+               return error;
+       }
+       /*
+        * If no interface address was provided, use the interface of
+        * the route to the given multicast address.
+        */
+       if (mreq->imr_interface.s_addr == INADDR_ANY) {
+               bzero((caddr_t)&ro, sizeof(ro));
+               dst = (struct sockaddr_in *)&ro.ro_dst;
+               dst->sin_len = sizeof(*dst);
+               dst->sin_family = AF_INET;
+               dst->sin_addr = mreq->imr_multiaddr;
+               rtalloc(&ro);
+               if (ro.ro_rt != NULL) {
+                       ifp = ro.ro_rt->rt_ifp;
+                       rtfree(ro.ro_rt);
+               }
+               else {
+                       /* If there's no default route, try using loopback */
+                       mreq->imr_interface.s_addr = INADDR_LOOPBACK;
+               }
+       }
+       
+       if (ifp == NULL) {
+               ifp = ip_multicast_if(&mreq->imr_interface, NULL);
+       }
+
+       /*
+        * See if we found an interface, and confirm that it
+        * supports multicast.
+        */
+       if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
+               error = EADDRNOTAVAIL;
+               return error;
+       }
+       /*
+        * See if the membership already exists or if all the
+        * membership slots are full.
+        */
+       for (i = 0; i < imo->imo_num_memberships; ++i) {
+               if (imo->imo_membership[i]->inm_ifp == ifp &&
+                       imo->imo_membership[i]->inm_addr.s_addr
+                                       == mreq->imr_multiaddr.s_addr)
+                       break;
+       }
+       if (i < imo->imo_num_memberships) {
+               error = EADDRINUSE;
+               return error;
+       }
+       if (i == IP_MAX_MEMBERSHIPS) {
+               error = ETOOMANYREFS;
+               return error;
+       }
+       /*
+        * Everything looks good; add a new record to the multicast
+        * address list for the given interface.
+        */
+       if ((imo->imo_membership[i] =
+               in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
+               error = ENOBUFS;
+               return error;
+       }
+       ++imo->imo_num_memberships;
+       
+       return error;
+}
+
+/*
+ * Drop membership of an IPv4 multicast.
+ */
+__private_extern__ int
+ip_dropmembership(
+       struct ip_moptions *imo,
+       struct ip_mreq *mreq)
+{
+       int error = 0;
+       struct ifnet* ifp = NULL;
+       int i;
+       
+       if (!IN_MULTICAST(ntohl(mreq->imr_multiaddr.s_addr))) {
+               error = EINVAL;
+               return error;
+       }
+
+       /*
+        * If an interface address was specified, get a pointer
+        * to its ifnet structure.
+        */
+       if (mreq->imr_interface.s_addr == INADDR_ANY)
+               ifp = NULL;
+       else {
+               ifp = ip_multicast_if(&mreq->imr_interface, NULL);
+               if (ifp == NULL) {
+                       error = EADDRNOTAVAIL;
+                       return error;
+               }
+       }
+       /*
+        * Find the membership in the membership array.
+        */
+       for (i = 0; i < imo->imo_num_memberships; ++i) {
+               if ((ifp == NULL ||
+                        imo->imo_membership[i]->inm_ifp == ifp) &&
+                        imo->imo_membership[i]->inm_addr.s_addr ==
+                        mreq->imr_multiaddr.s_addr)
+                       break;
+       }
+       if (i == imo->imo_num_memberships) {
+               error = EADDRNOTAVAIL;
+               return error;
+       }
+       /*
+        * Give up the multicast address record to which the
+        * membership points.
+        */
+       in_delmulti(&imo->imo_membership[i]);
+       /*
+        * Remove the gap in the membership array.
+        */
+       for (++i; i < imo->imo_num_memberships; ++i)
+               imo->imo_membership[i-1] = imo->imo_membership[i];
+       --imo->imo_num_memberships;
+       
+       return error;
+}
+
 /*
  * Return the IP multicast options in response to user getsockopt().
  */
@@ -1995,8 +2430,7 @@ ip_freemoptions(imo)
 
        if (imo != NULL) {
                for (i = 0; i < imo->imo_num_memberships; ++i)
-                       if (imo->imo_membership[i] != NULL)
-                               in_delmulti(imo->imo_membership[i]);
+                       in_delmulti(&imo->imo_membership[i]);
                FREE(imo, M_IPMOPTS);
        }
 }
@@ -2061,7 +2495,7 @@ ip_mloopback(ifp, m, dst, hlen)
         * The UDP checksum has not been calculated yet.
         */
         if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
-            if (ifp->if_hwassist) {
+            if (IF_HWASSIST_CSUM_FLAGS(ifp->if_hwassist)) {
                 copym->m_pkthdr.csum_flags |=
                     CSUM_DATA_VALID | CSUM_PSEUDO_HDR |
                     CSUM_IP_CHECKED | CSUM_IP_VALID;
@@ -2080,18 +2514,15 @@ ip_mloopback(ifp, m, dst, hlen)
                 * a filter has tapped-in.
                 */
 
-               if (lo_dl_tag == 0)
-                   dlil_find_dltag(APPLE_IF_FAM_LOOPBACK, 0, PF_INET, &lo_dl_tag);
-
                /*
                 * Stuff the 'real' ifp into the pkthdr, to be used in matching
                 *  in ip_input(); we need the loopback ifp/dl_tag passed as args
                 *  to make the loopback driver compliant with the data link
                 *  requirements.
                 */
-               if (lo_dl_tag) {
+               if (lo_ifp) {
                        copym->m_pkthdr.rcvif = ifp;
-                   dlil_output(lo_dl_tag, copym, 0, (struct sockaddr *) dst, 0);
+                   dlil_output(lo_ifp, PF_INET, copym, 0, (struct sockaddr *) dst, 0);
                } else {
                    printf("Warning: ip_output call to dlil_find_dltag failed!\n");
                    m_freem(copym);