]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/netinet/in.c
xnu-4903.241.1.tar.gz
[apple/xnu.git] / bsd / netinet / in.c
index 6686c77ee83af81528337119331bec843f9428a6..61de1526df04ec9f5e95c7dc0220f06bcc387112 100644 (file)
@@ -1,8 +1,8 @@
 /*
- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2016 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
  * unlawful or unlicensed copies of an Apple operating system, or to
  * circumvent, violate, or enable the circumvention or violation of, any
  * terms of an Apple operating system software license agreement.
- * 
+ *
  * Please obtain a copy of the License at
  * http://www.opensource.apple.com/apsl/ and read it before using this file.
- * 
+ *
  * The Original Code and all software distributed under the License are
  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
@@ -22,7 +22,7 @@
  * 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@
  */
 /*
@@ -58,7 +58,6 @@
  * SUCH DAMAGE.
  *
  *     @(#)in.c        8.4 (Berkeley) 1/9/95
- * $FreeBSD: src/sys/netinet/in.c,v 1.44.2.5 2001/08/13 16:26:17 ume Exp $
  */
 
 #include <sys/param.h>
 #include <sys/sysctl.h>
 #include <sys/kern_event.h>
 #include <sys/syslog.h>
+#include <sys/mcache.h>
+#include <sys/protosw.h>
+#include <sys/file.h>
+
+#include <kern/zalloc.h>
+#include <pexpert/pexpert.h>
 
 #include <net/if.h>
 #include <net/if_types.h>
 #include <net/route.h>
+#include <net/kpi_protocol.h>
+#include <net/dlil.h>
+#include <net/if_llatbl.h>
+#include <net/if_arp.h>
+#if PF
+#include <net/pfvar.h>
+#endif /* PF */
 
 #include <netinet/in.h>
 #include <netinet/in_var.h>
 #include <netinet/in_pcb.h>
-
 #include <netinet/igmp_var.h>
-#include <net/dlil.h>
-
 #include <netinet/ip_var.h>
-
 #include <netinet/tcp.h>
 #include <netinet/tcp_timer.h>
 #include <netinet/tcp_var.h>
+#include <netinet/if_ether.h>
+
+static int inctl_associd(struct socket *, u_long, caddr_t);
+static int inctl_connid(struct socket *, u_long, caddr_t);
+static int inctl_conninfo(struct socket *, u_long, caddr_t);
+static int inctl_autoaddr(struct ifnet *, struct ifreq *);
+static int inctl_arpipll(struct ifnet *, struct ifreq *);
+static int inctl_setrouter(struct ifnet *, struct ifreq *);
+static int inctl_ifaddr(struct ifnet *, struct in_ifaddr *, u_long,
+    struct ifreq *);
+static int inctl_ifdstaddr(struct ifnet *, struct in_ifaddr *, u_long,
+    struct ifreq *);
+static int inctl_ifbrdaddr(struct ifnet *, struct in_ifaddr *, u_long,
+    struct ifreq *);
+static int inctl_ifnetmask(struct ifnet *, struct in_ifaddr *, u_long,
+    struct ifreq *);
+
+static void in_socktrim(struct sockaddr_in *);
+static int in_ifinit(struct ifnet *, struct in_ifaddr *,
+    struct sockaddr_in *, int);
+
+#define        IA_HASH_INIT(ia) {                                      \
+       (ia)->ia_hash.tqe_next = (void *)(uintptr_t)-1;         \
+       (ia)->ia_hash.tqe_prev = (void *)(uintptr_t)-1;         \
+}
 
-#include <sys/file.h>
+#define        IA_IS_HASHED(ia)                                        \
+       (!((ia)->ia_hash.tqe_next == (void *)(uintptr_t)-1 ||   \
+       (ia)->ia_hash.tqe_prev == (void *)(uintptr_t)-1))
+
+static void in_iahash_remove(struct in_ifaddr *);
+static void in_iahash_insert(struct in_ifaddr *);
+static void in_iahash_insert_ptp(struct in_ifaddr *);
+static struct in_ifaddr *in_ifaddr_alloc(int);
+static void in_ifaddr_attached(struct ifaddr *);
+static void in_ifaddr_detached(struct ifaddr *);
+static void in_ifaddr_free(struct ifaddr *);
+static void in_ifaddr_trace(struct ifaddr *, int);
+
+static int in_getassocids(struct socket *, uint32_t *, user_addr_t);
+static int in_getconnids(struct socket *, sae_associd_t, uint32_t *, user_addr_t);
+
+/* IPv4 Layer 2 neighbor cache management routines */
+static void in_lltable_destroy_lle_unlocked(struct llentry *lle);
+static void in_lltable_destroy_lle(struct llentry *lle);
+static struct llentry *in_lltable_new(struct in_addr addr4, u_int flags);
+static int in_lltable_match_prefix(const struct sockaddr *saddr,
+    const struct sockaddr *smask, u_int flags, struct llentry *lle);
+static void in_lltable_free_entry(struct lltable *llt, struct llentry *lle);
+static int in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr);
+static inline uint32_t in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize);
+static uint32_t in_lltable_hash(const struct llentry *lle, uint32_t hsize);
+static void in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa);
+static inline struct llentry * in_lltable_find_dst(struct lltable *llt, struct in_addr dst);
+static void in_lltable_delete_entry(struct lltable *llt, struct llentry *lle);
+static struct llentry * in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr);
+static struct llentry * in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr);
+static int in_lltable_dump_entry(struct lltable *llt, struct llentry *lle, struct sysctl_req *wr);
+static struct lltable * in_lltattach(struct ifnet *ifp);
 
+static int subnetsarelocal = 0;
+SYSCTL_INT(_net_inet_ip, OID_AUTO, subnets_are_local,
+       CTLFLAG_RW | CTLFLAG_LOCKED, &subnetsarelocal, 0, "");
 
-static int in_mask2len(struct in_addr *);
-static void in_len2mask(struct in_addr *, int);
-static int in_lifaddr_ioctl(struct socket *, u_long, caddr_t,
-       struct ifnet *, struct proc *);
+/* Track whether or not the SIOCARPIPLL ioctl has been called */
+u_int32_t ipv4_ll_arp_aware = 0;
 
-static void    in_socktrim(struct sockaddr_in *);
-static int     in_ifinit(struct ifnet *,
-           struct in_ifaddr *, struct sockaddr_in *, int);
+#define        INIFA_TRACE_HIST_SIZE   32      /* size of trace history */
 
-static int subnetsarelocal = 0;
-SYSCTL_INT(_net_inet_ip, OID_AUTO, subnets_are_local, CTLFLAG_RW, 
-       &subnetsarelocal, 0, "");
+/* For gdb */
+__private_extern__ unsigned int inifa_trace_hist_size = INIFA_TRACE_HIST_SIZE;
 
-struct in_multihead in_multihead; /* XXX BSS initialization */
+struct in_ifaddr_dbg {
+       struct in_ifaddr        inifa;                  /* in_ifaddr */
+       struct in_ifaddr        inifa_old;              /* saved in_ifaddr */
+       u_int16_t               inifa_refhold_cnt;      /* # of IFA_ADDREF */
+       u_int16_t               inifa_refrele_cnt;      /* # of IFA_REMREF */
+       /*
+        * Alloc and free callers.
+        */
+       ctrace_t                inifa_alloc;
+       ctrace_t                inifa_free;
+       /*
+        * Circular lists of IFA_ADDREF and IFA_REMREF callers.
+        */
+       ctrace_t                inifa_refhold[INIFA_TRACE_HIST_SIZE];
+       ctrace_t                inifa_refrele[INIFA_TRACE_HIST_SIZE];
+       /*
+        * Trash list linkage
+        */
+       TAILQ_ENTRY(in_ifaddr_dbg) inifa_trash_link;
+};
 
-extern lck_mtx_t *rt_mtx;
+/* List of trash in_ifaddr entries protected by inifa_trash_lock */
+static TAILQ_HEAD(, in_ifaddr_dbg) inifa_trash_head;
+static decl_lck_mtx_data(, inifa_trash_lock);
 
-/* Track whether or not the SIOCARPIPLL ioctl has been called */
-__private_extern__     u_int32_t       ipv4_ll_arp_aware = 0;
+#if DEBUG
+static unsigned int inifa_debug = 1;           /* debugging (enabled) */
+#else
+static unsigned int inifa_debug;               /* debugging (disabled) */
+#endif /* !DEBUG */
+static unsigned int inifa_size;                        /* size of zone element */
+static struct zone *inifa_zone;                        /* zone for in_ifaddr */
+
+#define        INIFA_ZONE_MAX          64              /* maximum elements in zone */
+#define        INIFA_ZONE_NAME         "in_ifaddr"     /* zone name */
+
+static const unsigned int in_extra_size = sizeof (struct in_ifextra);
+static const unsigned int in_extra_bufsize = in_extra_size +
+    sizeof (void *) + sizeof (uint64_t);
+
+/*
+ * Return 1 if the address is
+ * - loopback
+ * - unicast or multicast link local
+ * - routed via a link level gateway
+ * - belongs to a directly connected (sub)net
+ */
+int
+inaddr_local(struct in_addr in)
+{
+       struct rtentry *rt;
+       struct sockaddr_in sin;
+       int local = 0;
+
+       if (ntohl(in.s_addr) == INADDR_LOOPBACK ||
+           IN_LINKLOCAL(ntohl(in.s_addr))) {
+               local = 1;
+       } else if (ntohl(in.s_addr) >= INADDR_UNSPEC_GROUP &&
+           ntohl(in.s_addr) <= INADDR_MAX_LOCAL_GROUP) {
+                       local = 1;
+       } else {
+               sin.sin_family = AF_INET;
+               sin.sin_len = sizeof (sin);
+               sin.sin_addr = in;
+               rt = rtalloc1((struct sockaddr *)&sin, 0, 0);
+
+               if (rt != NULL) {
+                       RT_LOCK_SPIN(rt);
+                       if (rt->rt_gateway->sa_family == AF_LINK ||
+                           (rt->rt_ifp->if_flags & IFF_LOOPBACK))
+                               local = 1;
+                       RT_UNLOCK(rt);
+                       rtfree(rt);
+               } else {
+                       local = in_localaddr(in);
+               }
+       }
+       return (local);
+}
 
 /*
  * Return 1 if an internet address is for a ``local'' host
  * (one to which we have a connection).  If subnetsarelocal
- * is true, this includes other subnets of the local net.
- * Otherwise, it includes only the directly-connected (sub)nets.
+ * is true, this includes other subnets of the local net,
+ * otherwise, it includes the directly-connected (sub)nets.
+ * The IPv4 link local prefix 169.254/16 is also included.
  */
 int
-in_localaddr(in)
-       struct in_addr in;
+in_localaddr(struct in_addr in)
 {
-       u_long i = ntohl(in.s_addr);
+       u_int32_t i = ntohl(in.s_addr);
        struct in_ifaddr *ia;
 
+       if (IN_LINKLOCAL(i))
+               return (1);
+
        if (subnetsarelocal) {
-               lck_mtx_lock(rt_mtx);
-               for (ia = in_ifaddrhead.tqh_first; ia; 
-                    ia = ia->ia_link.tqe_next)
+               lck_rw_lock_shared(in_ifaddr_rwlock);
+               for (ia = in_ifaddrhead.tqh_first; ia != NULL;
+                   ia = ia->ia_link.tqe_next) {
+                       IFA_LOCK(&ia->ia_ifa);
                        if ((i & ia->ia_netmask) == ia->ia_net) {
-                               lck_mtx_unlock(rt_mtx);
+                               IFA_UNLOCK(&ia->ia_ifa);
+                               lck_rw_done(in_ifaddr_rwlock);
                                return (1);
                        }
-               lck_mtx_unlock(rt_mtx);
+                       IFA_UNLOCK(&ia->ia_ifa);
+               }
+               lck_rw_done(in_ifaddr_rwlock);
        } else {
-               lck_mtx_lock(rt_mtx);
-               for (ia = in_ifaddrhead.tqh_first; ia;
-                    ia = ia->ia_link.tqe_next)
+               lck_rw_lock_shared(in_ifaddr_rwlock);
+               for (ia = in_ifaddrhead.tqh_first; ia != NULL;
+                   ia = ia->ia_link.tqe_next) {
+                       IFA_LOCK(&ia->ia_ifa);
                        if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
-                               lck_mtx_unlock(rt_mtx);
+                               IFA_UNLOCK(&ia->ia_ifa);
+                               lck_rw_done(in_ifaddr_rwlock);
                                return (1);
                        }
-               lck_mtx_unlock(rt_mtx);
+                       IFA_UNLOCK(&ia->ia_ifa);
+               }
+               lck_rw_done(in_ifaddr_rwlock);
        }
        return (0);
 }
@@ -153,884 +300,1339 @@ in_localaddr(in)
  * that may not be forwarded, or whether datagrams to that destination
  * may be forwarded.
  */
-int
-in_canforward(in)
-       struct in_addr in;
+boolean_t
+in_canforward(struct in_addr in)
 {
-       u_long i = ntohl(in.s_addr);
-       u_long net;
+       u_int32_t i = ntohl(in.s_addr);
+       u_int32_t net;
 
        if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i))
-               return (0);
+               return (FALSE);
        if (IN_CLASSA(i)) {
                net = i & IN_CLASSA_NET;
                if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
-                       return (0);
+                       return (FALSE);
        }
-       return (1);
+       return (TRUE);
 }
 
 /*
  * Trim a mask in a sockaddr
  */
 static void
-in_socktrim(ap)
-struct sockaddr_in *ap;
+in_socktrim(struct sockaddr_in *ap)
 {
-    char *cplim = (char *) &ap->sin_addr;
-    char *cp = (char *) (&ap->sin_addr + 1);
+       char *cplim = (char *)&ap->sin_addr;
+       char *cp = (char *)(&ap->sin_addr + 1);
 
-    ap->sin_len = 0;
-    while (--cp >= cplim)
-        if (*cp) {
-           (ap)->sin_len = cp - (char *) (ap) + 1;
-           break;
-       }
+       ap->sin_len = 0;
+       while (--cp >= cplim)
+               if (*cp) {
+                       (ap)->sin_len = cp - (char *)(ap) + 1;
+                       break;
+               }
 }
 
+static int in_interfaces;      /* number of external internet interfaces */
+
 static int
-in_mask2len(mask)
-       struct in_addr *mask;
+in_domifattach(struct ifnet *ifp)
 {
-       int x, y;
-       u_char *p;
+       int error;
 
-       p = (u_char *)mask;
-       for (x = 0; x < sizeof(*mask); x++) {
-               if (p[x] != 0xff)
-                       break;
-       }
-       y = 0;
-       if (x < sizeof(*mask)) {
-               for (y = 0; y < 8; y++) {
-                       if ((p[x] & (0x80 >> y)) == 0)
-                               break;
+       VERIFY(ifp != NULL);
+
+       if ((error = proto_plumb(PF_INET, ifp)) && error != EEXIST) {
+               log(LOG_ERR, "%s: proto_plumb returned %d if=%s\n",
+                   __func__, error, if_name(ifp));
+       } else if (error == 0 && ifp->if_inetdata == NULL) {
+               void **pbuf, *base;
+               struct in_ifextra *ext;
+               int errorx;
+
+               if ((ext = (struct in_ifextra *)_MALLOC(in_extra_bufsize,
+                   M_IFADDR, M_WAITOK|M_ZERO)) == NULL) {
+                       error = ENOMEM;
+                       errorx = proto_unplumb(PF_INET, ifp);
+                       if (errorx != 0) {
+                               log(LOG_ERR,
+                                   "%s: proto_unplumb returned %d if=%s%d\n",
+                                   __func__, errorx, ifp->if_name,
+                                   ifp->if_unit);
+                       }
+                       goto done;
                }
-       }
-       return x * 8 + y;
-}
 
-static void
-in_len2mask(mask, len)
-       struct in_addr *mask;
-       int len;
-{
-       int i;
-       u_char *p;
-
-       p = (u_char *)mask;
-       bzero(mask, sizeof(*mask));
-       for (i = 0; i < len / 8; i++)
-               p[i] = 0xff;
-       if (len % 8)
-               p[i] = (0xff00 >> (len % 8)) & 0xff;
+               /* Align on 64-bit boundary */
+               base = (void *)P2ROUNDUP((intptr_t)ext + sizeof (uint64_t),
+                   sizeof (uint64_t));
+               VERIFY(((intptr_t)base + in_extra_size) <=
+                   ((intptr_t)ext + in_extra_bufsize));
+               pbuf = (void **)((intptr_t)base - sizeof (void *));
+               *pbuf = ext;
+               ifp->if_inetdata = base;
+               IN_IFEXTRA(ifp)->ii_llt = in_lltattach(ifp);
+               VERIFY(IS_P2ALIGNED(ifp->if_inetdata, sizeof (uint64_t)));
+       }
+done:
+       if (error == 0 && ifp->if_inetdata != NULL) {
+               /*
+                * Since the structure is never freed, we need to
+                * zero out its contents to avoid reusing stale data.
+                * A little redundant with allocation above, but it
+                * keeps the code simpler for all cases.
+                */
+               bzero(ifp->if_inetdata, in_extra_size);
+       }
+       return (error);
 }
 
-static int in_interfaces;      /* number of external internet interfaces */
-
-/*
- * Generic internet control operations (ioctl's).
- * Ifp is 0 if not an interface-specific ioctl.
- */
-/* ARGSUSED */
-int
-in_control(
-       struct socket *so,
-       u_long cmd,
-       caddr_t data,
-       struct ifnet *ifp,
-       struct proc *p)
+static __attribute__((noinline)) int
+inctl_associd(struct socket *so, u_long cmd, caddr_t data)
 {
-       struct ifreq *ifr = (struct ifreq *)data;
-       struct in_ifaddr *ia = 0, *iap;
-       struct ifaddr *ifa;
-       struct in_ifaddr *oia;
-       struct in_aliasreq *ifra = (struct in_aliasreq *)data;
-       struct sockaddr_in oldaddr;
-       int error, hostIsNew, maskIsNew;
-       u_long i;
-       struct kev_msg        ev_msg;
-       struct kev_in_data    in_event_data;
-
+       int error = 0;
+       union {
+               struct so_aidreq32 a32;
+               struct so_aidreq64 a64;
+       } u;
 
-       switch (cmd) {
-       case SIOCALIFADDR:
-       case SIOCDLIFADDR:
-               if (p && (error = proc_suser(p)) != 0)
-                       return error;
-               /*fall through*/
-       case SIOCGLIFADDR:
-               if (!ifp)
-                       return EINVAL;
-               return in_lifaddr_ioctl(so, cmd, data, ifp, p);
-       }
+       VERIFY(so != NULL);
 
-       /*
-        * Find address for this interface, if it exists.
-        *
-        * If an alias address was specified, find that one instead of
-        * the first one on the interface.
-        */
-       if (ifp) {
-               lck_mtx_lock(rt_mtx);
-               for (iap = in_ifaddrhead.tqh_first; iap; 
-                    iap = iap->ia_link.tqe_next)
-                       if (iap->ia_ifp == ifp) {
-                               if (((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr.s_addr ==
-                                   iap->ia_addr.sin_addr.s_addr) {
-                                       ia = iap;
-                                       break;
-                               } else if (ia == NULL) {
-                                       ia = iap;
-                                       if (ifr->ifr_addr.sa_family != AF_INET)
-                                               break;
-                               }
-                       }
-               lck_mtx_unlock(rt_mtx);
-       }
        switch (cmd) {
-       case SIOCAUTOADDR:
-       case SIOCARPIPLL:
-               if (p && (error = proc_suser(p)) != 0)
-                       return error;
+       case SIOCGASSOCIDS32:           /* struct so_aidreq32 */
+               bcopy(data, &u.a32, sizeof (u.a32));
+               error = in_getassocids(so, &u.a32.sar_cnt, u.a32.sar_aidp);
+               if (error == 0)
+                       bcopy(&u.a32, data, sizeof (u.a32));
                break;
 
-       case SIOCAIFADDR:
-       case SIOCDIFADDR:
-               if (ifp == 0)
-                       return (EADDRNOTAVAIL);
-               if (ifra->ifra_addr.sin_family == AF_INET) {
-                       lck_mtx_lock(rt_mtx);
-                       for (oia = ia; ia; ia = ia->ia_link.tqe_next) {
-                               if (ia->ia_ifp == ifp  &&
-                                   ia->ia_addr.sin_addr.s_addr ==
-                                   ifra->ifra_addr.sin_addr.s_addr)
-                                       break;
-                       }
-                       lck_mtx_unlock(rt_mtx);
-                       if ((ifp->if_flags & IFF_POINTOPOINT)
-                           && (cmd == SIOCAIFADDR)
-                           && (ifra->ifra_dstaddr.sin_addr.s_addr
-                               == INADDR_ANY)) {
-                               return EDESTADDRREQ;
-                       }
-               }
-        else if (cmd == SIOCAIFADDR)
-            return (EINVAL);
-               if (cmd == SIOCDIFADDR && ia == 0)
-                       return (EADDRNOTAVAIL);
-               /* FALLTHROUGH */
-       case SIOCSIFADDR:
-       case SIOCSIFNETMASK:
-       case SIOCSIFDSTADDR:
-               if ((so->so_state & SS_PRIV) == 0)
-                       return (EPERM);
-
-               if (ifp == 0)
-                       return (EADDRNOTAVAIL);
-        if (ifra->ifra_addr.sin_family != AF_INET && cmd == SIOCSIFADDR)
-            return (EINVAL);
-               if (ia == (struct in_ifaddr *)0) {
-                       ia = (struct in_ifaddr *)
-                               _MALLOC(sizeof *ia, M_IFADDR, M_WAITOK);
-                       if (ia == (struct in_ifaddr *)NULL)
-                               return (ENOBUFS);
-                       bzero((caddr_t)ia, sizeof *ia);
-                       /*
-                        * Protect from ipintr() traversing address list
-                        * while we're modifying it.
-                        */
-                       
-                       ifa = &ia->ia_ifa;
+       case SIOCGASSOCIDS64:           /* struct so_aidreq64 */
+               bcopy(data, &u.a64, sizeof (u.a64));
+               error = in_getassocids(so, &u.a64.sar_cnt, u.a64.sar_aidp);
+               if (error == 0)
+                       bcopy(&u.a64, data, sizeof (u.a64));
+               break;
 
-                       ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
-                       ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
-                       ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
-                       ia->ia_sockmask.sin_len = 8;
-                       ifnet_lock_exclusive(ifp);
-                       if (ifp->if_flags & IFF_BROADCAST) {
-                               ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
-                               ia->ia_broadaddr.sin_family = AF_INET;
-                       }
-                       ia->ia_ifp = ifp;
-                       if (!(ifp->if_flags & IFF_LOOPBACK))
-                               in_interfaces++;
-                       if_attach_ifa(ifp, ifa);
-                       ifnet_lock_done(ifp);
+       default:
+               VERIFY(0);
+               /* NOTREACHED */
+       }
 
-                       lck_mtx_lock(rt_mtx);
-                       TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_link);
-                       lck_mtx_unlock(rt_mtx);
+       return (error);
+}
 
-                       /* Generic protocol plumbing */
+static __attribute__((noinline)) int
+inctl_connid(struct socket *so, u_long cmd, caddr_t data)
+{
+       int error = 0;
+       union {
+               struct so_cidreq32 c32;
+               struct so_cidreq64 c64;
+       } u;
 
-                       if (error = dlil_plumb_protocol(PF_INET, ifp)) {
-                               kprintf("in.c: warning can't plumb proto if=%s%n type %d error=%d\n",
-                                       ifp->if_name, ifp->if_unit, ifp->if_type, error);
-                               error = 0; /*discard error, can be cold with unsupported interfaces */
-                       }
+       VERIFY(so != NULL);
 
-               }
+       switch (cmd) {
+       case SIOCGCONNIDS32:            /* struct so_cidreq32 */
+               bcopy(data, &u.c32, sizeof (u.c32));
+               error = in_getconnids(so, u.c32.scr_aid, &u.c32.scr_cnt,
+                   u.c32.scr_cidp);
+               if (error == 0)
+                       bcopy(&u.c32, data, sizeof (u.c32));
                break;
 
-       case SIOCPROTOATTACH:
-       case SIOCPROTODETACH:
-               if (p && (error = proc_suser(p)) != 0)
-                       return error;
-               if (ifp == 0)
-                       return (EADDRNOTAVAIL);
+       case SIOCGCONNIDS64:            /* struct so_cidreq64 */
+               bcopy(data, &u.c64, sizeof (u.c64));
+               error = in_getconnids(so, u.c64.scr_aid, &u.c64.scr_cnt,
+                   u.c64.scr_cidp);
+               if (error == 0)
+                       bcopy(&u.c64, data, sizeof (u.c64));
                break;
-                
-       case SIOCSIFBRDADDR:
-#ifdef __APPLE__
-               if ((so->so_state & SS_PRIV) == 0)
-                       return (EPERM);
-#else
-               if (p && (error = suser(p)) != 0)
-                       return error;
-#endif
-               /* FALLTHROUGH */
 
-       case SIOCGIFADDR:
-       case SIOCGIFNETMASK:
-       case SIOCGIFDSTADDR:
-       case SIOCGIFBRDADDR:
-               if (ia == (struct in_ifaddr *)0)
-                       return (EADDRNOTAVAIL);
-               break;
+       default:
+               VERIFY(0);
+               /* NOTREACHED */
        }
+
+       return (error);
+}
+
+static __attribute__((noinline)) int
+inctl_conninfo(struct socket *so, u_long cmd, caddr_t data)
+{
+       int error = 0;
+       union {
+               struct so_cinforeq32 ci32;
+               struct so_cinforeq64 ci64;
+       } u;
+
+       VERIFY(so != NULL);
+
        switch (cmd) {
-       case SIOCAUTOADDR:
-               if (ifp == 0)
-                       return (EADDRNOTAVAIL);
-               ifnet_lock_exclusive(ifp);
-               if (ifr->ifr_intval)
-                       ifp->if_eflags |= IFEF_AUTOCONFIGURING;
-               else
-                       ifp->if_eflags &= ~IFEF_AUTOCONFIGURING;
-               ifnet_lock_done(ifp);
-               break;
-       
-       case SIOCARPIPLL:
-               if (ifp == 0)
-                       return (EADDRNOTAVAIL);
-               ipv4_ll_arp_aware = 1;
-               ifnet_lock_exclusive(ifp);
-               if (ifr->ifr_data)
-                       ifp->if_eflags |= IFEF_ARPLL;
-               else
-                       ifp->if_eflags &= ~IFEF_ARPLL;
-               ifnet_lock_done(ifp);
+       case SIOCGCONNINFO32:           /* struct so_cinforeq32 */
+               bcopy(data, &u.ci32, sizeof (u.ci32));
+               error = in_getconninfo(so, u.ci32.scir_cid, &u.ci32.scir_flags,
+                   &u.ci32.scir_ifindex, &u.ci32.scir_error, u.ci32.scir_src,
+                   &u.ci32.scir_src_len, u.ci32.scir_dst, &u.ci32.scir_dst_len,
+                   &u.ci32.scir_aux_type, u.ci32.scir_aux_data,
+                   &u.ci32.scir_aux_len);
+               if (error == 0)
+                       bcopy(&u.ci32, data, sizeof (u.ci32));
                break;
 
-       case SIOCGIFADDR:
-               *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
+       case SIOCGCONNINFO64:           /* struct so_cinforeq64 */
+               bcopy(data, &u.ci64, sizeof (u.ci64));
+               error = in_getconninfo(so, u.ci64.scir_cid, &u.ci64.scir_flags,
+                   &u.ci64.scir_ifindex, &u.ci64.scir_error, u.ci64.scir_src,
+                   &u.ci64.scir_src_len, u.ci64.scir_dst, &u.ci64.scir_dst_len,
+                   &u.ci64.scir_aux_type, u.ci64.scir_aux_data,
+                   &u.ci64.scir_aux_len);
+               if (error == 0)
+                       bcopy(&u.ci64, data, sizeof (u.ci64));
                break;
 
-       case SIOCGIFBRDADDR:
-               if ((ifp->if_flags & IFF_BROADCAST) == 0)
-                       return (EINVAL);
-               *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
-               break;
+       default:
+               VERIFY(0);
+               /* NOTREACHED */
+       }
 
-       case SIOCGIFDSTADDR:
-               if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
-                       return (EINVAL);
-               *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
-               break;
+       return (error);
+}
 
-       case SIOCGIFNETMASK:
-               *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
-               break;
+/*
+ * Caller passes in the ioctl data pointer directly via "ifr", with the
+ * expectation that this routine always uses bcopy() or other byte-aligned
+ * memory accesses.
+ */
+static __attribute__((noinline)) int
+inctl_autoaddr(struct ifnet *ifp, struct ifreq *ifr)
+{
+       int error = 0, intval;
 
-       case SIOCSIFDSTADDR:
-               if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
-                       return (EINVAL);
-               oldaddr = ia->ia_dstaddr;
-               ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
-               error = dlil_ioctl(PF_INET, ifp, SIOCSIFDSTADDR, (caddr_t)ia);
-               if (error == EOPNOTSUPP)
-                    error = 0;
+       VERIFY(ifp != NULL);
+
+       bcopy(&ifr->ifr_intval, &intval, sizeof (intval));
 
-               if (error) {
-                    ia->ia_dstaddr = oldaddr;
-                    return error;
+       ifnet_lock_exclusive(ifp);
+       if (intval) {
+               /*
+                * An interface in IPv4 router mode implies that it
+                * is configured with a static IP address and should
+                * not act as a DHCP client; prevent SIOCAUTOADDR from
+                * being set in that mode.
+                */
+               if (ifp->if_eflags & IFEF_IPV4_ROUTER) {
+                       intval = 0;     /* be safe; clear flag if set */
+                       error = EBUSY;
+               } else {
+                       ifp->if_eflags |= IFEF_AUTOCONFIGURING;
                }
+       }
+       if (!intval)
+               ifp->if_eflags &= ~IFEF_AUTOCONFIGURING;
+       ifnet_lock_done(ifp);
 
-               ev_msg.vendor_code    = KEV_VENDOR_APPLE;
-               ev_msg.kev_class      = KEV_NETWORK_CLASS;
-               ev_msg.kev_subclass   = KEV_INET_SUBCLASS;
-       
-               ev_msg.event_code = KEV_INET_SIFDSTADDR;
-
-               if (ia->ia_ifa.ifa_dstaddr)
-                    in_event_data.ia_dstaddr = 
-                         ((struct sockaddr_in *)ia->ia_ifa.ifa_dstaddr)->sin_addr;
-               else
-                    in_event_data.ia_dstaddr.s_addr  = 0;
-
-               in_event_data.ia_addr         = ia->ia_addr.sin_addr;
-               in_event_data.ia_net          = ia->ia_net;
-               in_event_data.ia_netmask      = ia->ia_netmask;
-               in_event_data.ia_subnet       = ia->ia_subnet;
-               in_event_data.ia_subnetmask   = ia->ia_subnetmask;
-               in_event_data.ia_netbroadcast = ia->ia_netbroadcast;
-               strncpy(&in_event_data.link_data.if_name[0], ifp->if_name, IFNAMSIZ);
-               in_event_data.link_data.if_family = ifp->if_family;
-               in_event_data.link_data.if_unit  = (unsigned long) ifp->if_unit;
+       return (error);
+}
 
-               ev_msg.dv[0].data_ptr    = &in_event_data;
-               ev_msg.dv[0].data_length      = sizeof(struct kev_in_data);
-               ev_msg.dv[1].data_length = 0;
+/*
+ * Caller passes in the ioctl data pointer directly via "ifr", with the
+ * expectation that this routine always uses bcopy() or other byte-aligned
+ * memory accesses.
+ */
+static __attribute__((noinline)) int
+inctl_arpipll(struct ifnet *ifp, struct ifreq *ifr)
+{
+       int error = 0, intval;
 
-               kev_post_msg(&ev_msg);
+       VERIFY(ifp != NULL);
 
+       bcopy(&ifr->ifr_intval, &intval, sizeof (intval));
+       ipv4_ll_arp_aware = 1;
 
-               if (ia->ia_flags & IFA_ROUTE) {
-                       ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
-                       rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
-                       ia->ia_ifa.ifa_dstaddr =
-                                       (struct sockaddr *)&ia->ia_dstaddr;
-                       rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
+       ifnet_lock_exclusive(ifp);
+       if (intval) {
+               /*
+                * An interface in IPv4 router mode implies that it
+                * is configured with a static IP address and should
+                * not have to deal with IPv4 Link-Local Address;
+                * prevent SIOCARPIPLL from being set in that mode.
+                */
+               if (ifp->if_eflags & IFEF_IPV4_ROUTER) {
+                       intval = 0;     /* be safe; clear flag if set */
+                       error = EBUSY;
+               } else {
+                       ifp->if_eflags |= IFEF_ARPLL;
                }
-               break;
+       }
+       if (!intval)
+               ifp->if_eflags &= ~IFEF_ARPLL;
+       ifnet_lock_done(ifp);
 
-       case SIOCSIFBRDADDR:
-               if ((ifp->if_flags & IFF_BROADCAST) == 0)
-                       return (EINVAL);
-               ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
+       return (error);
+}
 
-               ev_msg.vendor_code    = KEV_VENDOR_APPLE;
-               ev_msg.kev_class      = KEV_NETWORK_CLASS;
-               ev_msg.kev_subclass   = KEV_INET_SUBCLASS;
-       
-               ev_msg.event_code = KEV_INET_SIFBRDADDR;
+/*
+ * Handle SIOCSETROUTERMODE to set or clear the IPv4 router mode flag on
+ * the interface.  When in this mode, IPv4 Link-Local Address support is
+ * disabled in ARP, and DHCP client support is disabled in IP input; turning
+ * any of them on would cause an error to be returned.  Entering or exiting
+ * this mode will result in the removal of IPv4 addresses currently configured
+ * on the interface.
+ *
+ * Caller passes in the ioctl data pointer directly via "ifr", with the
+ * expectation that this routine always uses bcopy() or other byte-aligned
+ * memory accesses.
+ */
+static __attribute__((noinline)) int
+inctl_setrouter(struct ifnet *ifp, struct ifreq *ifr)
+{
+       int error = 0, intval;
 
-               if (ia->ia_ifa.ifa_dstaddr)
-                    in_event_data.ia_dstaddr = 
-                         ((struct sockaddr_in *)ia->ia_ifa.ifa_dstaddr)->sin_addr;
-               else
-                    in_event_data.ia_dstaddr.s_addr  = 0;
-
-               in_event_data.ia_addr         = ia->ia_addr.sin_addr;
-               in_event_data.ia_net          = ia->ia_net;
-               in_event_data.ia_netmask      = ia->ia_netmask;
-               in_event_data.ia_subnet       = ia->ia_subnet;
-               in_event_data.ia_subnetmask   = ia->ia_subnetmask;
-               in_event_data.ia_netbroadcast = ia->ia_netbroadcast;
-               strncpy(&in_event_data.link_data.if_name[0], ifp->if_name, IFNAMSIZ);
-               in_event_data.link_data.if_family = ifp->if_family;
-               in_event_data.link_data.if_unit  = (unsigned long) ifp->if_unit;
+       VERIFY(ifp != NULL);
 
-               ev_msg.dv[0].data_ptr    = &in_event_data;
-               ev_msg.dv[0].data_length      = sizeof(struct kev_in_data);
-               ev_msg.dv[1].data_length = 0;
+       /* Router mode isn't valid for loopback */
+       if (ifp->if_flags & IFF_LOOPBACK)
+               return (ENODEV);
 
-               kev_post_msg(&ev_msg);
+       bcopy(&ifr->ifr_intval, &intval, sizeof (intval));
 
-               break;
+       ifnet_lock_exclusive(ifp);
+       if (intval) {
+               ifp->if_eflags |= IFEF_IPV4_ROUTER;
+               ifp->if_eflags &= ~(IFEF_ARPLL | IFEF_AUTOCONFIGURING);
+       } else {
+               ifp->if_eflags &= ~IFEF_IPV4_ROUTER;
+       }
+       ifnet_lock_done(ifp);
 
-       case SIOCSIFADDR:
-               return (in_ifinit(ifp, ia,
-                   (struct sockaddr_in *) &ifr->ifr_addr, 1));
-
-       case SIOCPROTOATTACH:
-               error = dlil_plumb_protocol(PF_INET, ifp);
-               if (error)
-                       return(error);
-                break;
-                
-       case SIOCPROTODETACH:
-                // if an ip address is still present, refuse to detach
-               ifnet_lock_shared(ifp);
-               TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 
-                       if (ifa->ifa_addr->sa_family == AF_INET)
-                               break;
-               ifnet_lock_done(ifp);
-               if (ifa != 0)
-                       return EBUSY;
+       /* purge all IPv4 addresses configured on this interface */
+       in_purgeaddrs(ifp);
 
-               error = dlil_unplumb_protocol(PF_INET, ifp);
-               if (error)
-                       return(error);
-               break;
-               
+       return (error);
+}
 
-       case SIOCSIFNETMASK:
-               i = ifra->ifra_addr.sin_addr.s_addr;
-               ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr = i);
-               ev_msg.vendor_code    = KEV_VENDOR_APPLE;
-               ev_msg.kev_class      = KEV_NETWORK_CLASS;
-               ev_msg.kev_subclass   = KEV_INET_SUBCLASS;
-       
-               ev_msg.event_code = KEV_INET_SIFNETMASK;
+/*
+ * Caller passes in the ioctl data pointer directly via "ifr", with the
+ * expectation that this routine always uses bcopy() or other byte-aligned
+ * memory accesses.
+ */
+static __attribute__((noinline)) int
+inctl_ifaddr(struct ifnet *ifp, struct in_ifaddr *ia, u_long cmd,
+    struct ifreq *ifr)
+{
+       struct kev_in_data in_event_data;
+       struct kev_msg ev_msg;
+       struct sockaddr_in addr;
+       struct ifaddr *ifa;
+       int error = 0;
 
-               if (ia->ia_ifa.ifa_dstaddr)
-                    in_event_data.ia_dstaddr = 
-                         ((struct sockaddr_in *)ia->ia_ifa.ifa_dstaddr)->sin_addr;
-               else
-                    in_event_data.ia_dstaddr.s_addr  = 0;
-
-               in_event_data.ia_addr         = ia->ia_addr.sin_addr;
-               in_event_data.ia_net          = ia->ia_net;
-               in_event_data.ia_netmask      = ia->ia_netmask;
-               in_event_data.ia_subnet       = ia->ia_subnet;
-               in_event_data.ia_subnetmask   = ia->ia_subnetmask;
-               in_event_data.ia_netbroadcast = ia->ia_netbroadcast;
-               strncpy(&in_event_data.link_data.if_name[0], ifp->if_name, IFNAMSIZ);
-               in_event_data.link_data.if_family = ifp->if_family;
-               in_event_data.link_data.if_unit  = (unsigned long) ifp->if_unit;
+       VERIFY(ifp != NULL);
 
-               ev_msg.dv[0].data_ptr    = &in_event_data;
-               ev_msg.dv[0].data_length      = sizeof(struct kev_in_data);
-               ev_msg.dv[1].data_length = 0;
+       bzero(&in_event_data, sizeof (struct kev_in_data));
+       bzero(&ev_msg, sizeof (struct kev_msg));
 
-               kev_post_msg(&ev_msg);
+       switch (cmd) {
+       case SIOCGIFADDR:               /* struct ifreq */
+               if (ia == NULL) {
+                       error = EADDRNOTAVAIL;
+                       break;
+               }
+               IFA_LOCK(&ia->ia_ifa);
+               bcopy(&ia->ia_addr, &ifr->ifr_addr, sizeof (addr));
+               IFA_UNLOCK(&ia->ia_ifa);
+               break;
 
+       case SIOCSIFADDR:               /* struct ifreq */
+               VERIFY(ia != NULL);
+               bcopy(&ifr->ifr_addr, &addr, sizeof (addr));
+               /*
+                * If this is a new address, the reference count for the
+                * hash table has been taken at creation time above.
+                */
+               error = in_ifinit(ifp, ia, &addr, 1);
+               if (error == 0) {
+                       (void) ifnet_notify_address(ifp, AF_INET);
+               }
                break;
 
-       case SIOCAIFADDR:
+       case SIOCAIFADDR: {             /* struct {if,in_}aliasreq */
+               struct in_aliasreq *ifra = (struct in_aliasreq *)ifr;
+               struct sockaddr_in broadaddr, mask;
+               int hostIsNew, maskIsNew;
+
+               VERIFY(ia != NULL);
+               bcopy(&ifra->ifra_addr, &addr, sizeof (addr));
+               bcopy(&ifra->ifra_broadaddr, &broadaddr, sizeof (broadaddr));
+               bcopy(&ifra->ifra_mask, &mask, sizeof (mask));
+
                maskIsNew = 0;
                hostIsNew = 1;
                error = 0;
+
+               IFA_LOCK(&ia->ia_ifa);
                if (ia->ia_addr.sin_family == AF_INET) {
-                       if (ifra->ifra_addr.sin_len == 0) {
-                               ifra->ifra_addr = ia->ia_addr;
+                       if (addr.sin_len == 0) {
+                               addr = ia->ia_addr;
                                hostIsNew = 0;
-                       } else if (ifra->ifra_addr.sin_addr.s_addr ==
-                                              ia->ia_addr.sin_addr.s_addr)
+                       } else if (addr.sin_addr.s_addr ==
+                           ia->ia_addr.sin_addr.s_addr) {
                                hostIsNew = 0;
+                       }
                }
-               if (ifra->ifra_mask.sin_len) {
+               if (mask.sin_len) {
+                       IFA_UNLOCK(&ia->ia_ifa);
                        in_ifscrub(ifp, ia, 0);
-                       ia->ia_sockmask = ifra->ifra_mask;
+                       IFA_LOCK(&ia->ia_ifa);
+                       ia->ia_sockmask = mask;
                        ia->ia_subnetmask =
-                            ntohl(ia->ia_sockmask.sin_addr.s_addr);
+                           ntohl(ia->ia_sockmask.sin_addr.s_addr);
                        maskIsNew = 1;
                }
                if ((ifp->if_flags & IFF_POINTOPOINT) &&
-                   (ifra->ifra_dstaddr.sin_family == AF_INET)) {
+                   (broadaddr.sin_family == AF_INET)) {
+                       IFA_UNLOCK(&ia->ia_ifa);
                        in_ifscrub(ifp, ia, 0);
-                       ia->ia_dstaddr = ifra->ifra_dstaddr;
+                       IFA_LOCK(&ia->ia_ifa);
+                       ia->ia_dstaddr = broadaddr;
+                       ia->ia_dstaddr.sin_len = sizeof (struct sockaddr_in);
                        maskIsNew  = 1; /* We lie; but the effect's the same */
                }
-               if (ifra->ifra_addr.sin_family == AF_INET &&
-                   (hostIsNew || maskIsNew)) {
-                       error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
+               if (addr.sin_family == AF_INET && (hostIsNew || maskIsNew)) {
+                       IFA_UNLOCK(&ia->ia_ifa);
+                       error = in_ifinit(ifp, ia, &addr, 0);
+               } else {
+                       IFA_UNLOCK(&ia->ia_ifa);
+               }
+               if (error == 0) {
+                       (void) ifnet_notify_address(ifp, AF_INET);
                }
+               IFA_LOCK(&ia->ia_ifa);
                if ((ifp->if_flags & IFF_BROADCAST) &&
-                   (ifra->ifra_broadaddr.sin_family == AF_INET))
-                       ia->ia_broadaddr = ifra->ifra_broadaddr;
+                   (broadaddr.sin_family == AF_INET))
+                       ia->ia_broadaddr = broadaddr;
 
                /*
                 * Report event.
                 */
-
                if ((error == 0) || (error == EEXIST)) {
-                    ev_msg.vendor_code    = KEV_VENDOR_APPLE;
-                    ev_msg.kev_class      = KEV_NETWORK_CLASS;
-                    ev_msg.kev_subclass   = KEV_INET_SUBCLASS;
-
-                    if (hostIsNew)
-                         ev_msg.event_code = KEV_INET_NEW_ADDR;
-                    else
-                         ev_msg.event_code = KEV_INET_CHANGED_ADDR;
-
-                    if (ia->ia_ifa.ifa_dstaddr)
-                         in_event_data.ia_dstaddr = 
-                              ((struct sockaddr_in *)ia->ia_ifa.ifa_dstaddr)->sin_addr;
-                    else
-                         in_event_data.ia_dstaddr.s_addr  = 0;
-
-                    in_event_data.ia_addr         = ia->ia_addr.sin_addr;
-                    in_event_data.ia_net          = ia->ia_net;
-                    in_event_data.ia_netmask      = ia->ia_netmask;
-                    in_event_data.ia_subnet       = ia->ia_subnet;
-                    in_event_data.ia_subnetmask   = ia->ia_subnetmask;
-                    in_event_data.ia_netbroadcast = ia->ia_netbroadcast;
-                    strncpy(&in_event_data.link_data.if_name[0], ifp->if_name, IFNAMSIZ);
-                    in_event_data.link_data.if_family = ifp->if_family;
-                    in_event_data.link_data.if_unit  = (unsigned long) ifp->if_unit;
-
-                    ev_msg.dv[0].data_ptr    = &in_event_data;
-                    ev_msg.dv[0].data_length      = sizeof(struct kev_in_data);
-                    ev_msg.dv[1].data_length = 0;
-
-                    kev_post_msg(&ev_msg);
+                       ev_msg.vendor_code      = KEV_VENDOR_APPLE;
+                       ev_msg.kev_class        = KEV_NETWORK_CLASS;
+                       ev_msg.kev_subclass     = KEV_INET_SUBCLASS;
+
+                       if (hostIsNew)
+                               ev_msg.event_code = KEV_INET_NEW_ADDR;
+                       else
+                               ev_msg.event_code = KEV_INET_CHANGED_ADDR;
+
+                       if (ia->ia_ifa.ifa_dstaddr) {
+                               in_event_data.ia_dstaddr =
+                                   ((struct sockaddr_in *)(void *)ia->
+                                   ia_ifa.ifa_dstaddr)->sin_addr;
+                       } else {
+                               in_event_data.ia_dstaddr.s_addr = INADDR_ANY;
+                       }
+                       in_event_data.ia_addr           = ia->ia_addr.sin_addr;
+                       in_event_data.ia_net            = ia->ia_net;
+                       in_event_data.ia_netmask        = ia->ia_netmask;
+                       in_event_data.ia_subnet         = ia->ia_subnet;
+                       in_event_data.ia_subnetmask     = ia->ia_subnetmask;
+                       in_event_data.ia_netbroadcast   = ia->ia_netbroadcast;
+                       IFA_UNLOCK(&ia->ia_ifa);
+                       (void) strlcpy(&in_event_data.link_data.if_name[0],
+                           ifp->if_name, IFNAMSIZ);
+                       in_event_data.link_data.if_family = ifp->if_family;
+                       in_event_data.link_data.if_unit = ifp->if_unit;
+
+                       ev_msg.dv[0].data_ptr    = &in_event_data;
+                       ev_msg.dv[0].data_length = sizeof (struct kev_in_data);
+                       ev_msg.dv[1].data_length = 0;
+
+                       dlil_post_complete_msg(ifp, &ev_msg);
+               } else {
+                       IFA_UNLOCK(&ia->ia_ifa);
                }
+               break;
+       }
 
-               return (error);
-
-       case SIOCDIFADDR:
-               error = dlil_ioctl(PF_INET, ifp, SIOCDIFADDR, (caddr_t)ia);
+       case SIOCDIFADDR:               /* struct ifreq */
+               VERIFY(ia != NULL);
+               error = ifnet_ioctl(ifp, PF_INET, SIOCDIFADDR, ia);
                if (error == EOPNOTSUPP)
                        error = 0;
-               if (error)
-                   return error;
+               if (error != 0) {
+                       /* Reset the detaching flag */
+                       IFA_LOCK(&ia->ia_ifa);
+                       ia->ia_ifa.ifa_debug &= ~IFD_DETACHING;
+                       IFA_UNLOCK(&ia->ia_ifa);
+                       break;
+               }
 
                /* Fill out the kernel event information */
-               ev_msg.vendor_code    = KEV_VENDOR_APPLE;
-               ev_msg.kev_class      = KEV_NETWORK_CLASS;
-               ev_msg.kev_subclass   = KEV_INET_SUBCLASS;
-               
-               ev_msg.event_code = KEV_INET_ADDR_DELETED;
-
-               if (ia->ia_ifa.ifa_dstaddr)
-                    in_event_data.ia_dstaddr = 
-                         ((struct sockaddr_in *)ia->ia_ifa.ifa_dstaddr)->sin_addr;
-               else
-                    in_event_data.ia_dstaddr.s_addr  = 0;
-
-               in_event_data.ia_addr         = ia->ia_addr.sin_addr;
-               in_event_data.ia_net          = ia->ia_net;
-               in_event_data.ia_netmask      = ia->ia_netmask;
-               in_event_data.ia_subnet       = ia->ia_subnet;
-               in_event_data.ia_subnetmask   = ia->ia_subnetmask;
-               in_event_data.ia_netbroadcast = ia->ia_netbroadcast;
-               strncpy(&in_event_data.link_data.if_name[0], ifp->if_name, IFNAMSIZ);
+               ev_msg.vendor_code      = KEV_VENDOR_APPLE;
+               ev_msg.kev_class        = KEV_NETWORK_CLASS;
+               ev_msg.kev_subclass     = KEV_INET_SUBCLASS;
+
+               ev_msg.event_code       = KEV_INET_ADDR_DELETED;
+
+               IFA_LOCK(&ia->ia_ifa);
+               if (ia->ia_ifa.ifa_dstaddr) {
+                       in_event_data.ia_dstaddr = ((struct sockaddr_in *)
+                           (void *)ia->ia_ifa.ifa_dstaddr)->sin_addr;
+               } else {
+                       in_event_data.ia_dstaddr.s_addr = INADDR_ANY;
+               }
+               in_event_data.ia_addr           = ia->ia_addr.sin_addr;
+               in_event_data.ia_net            = ia->ia_net;
+               in_event_data.ia_netmask        = ia->ia_netmask;
+               in_event_data.ia_subnet         = ia->ia_subnet;
+               in_event_data.ia_subnetmask     = ia->ia_subnetmask;
+               in_event_data.ia_netbroadcast   = ia->ia_netbroadcast;
+               IFA_UNLOCK(&ia->ia_ifa);
+               (void) strlcpy(&in_event_data.link_data.if_name[0],
+                   ifp->if_name, IFNAMSIZ);
                in_event_data.link_data.if_family = ifp->if_family;
-               in_event_data.link_data.if_unit  = (unsigned long) ifp->if_unit;
+               in_event_data.link_data.if_unit  = (u_int32_t)ifp->if_unit;
 
                ev_msg.dv[0].data_ptr    = &in_event_data;
                ev_msg.dv[0].data_length = sizeof(struct kev_in_data);
                ev_msg.dv[1].data_length = 0;
 
-               lck_mtx_lock(rt_mtx);
+               ifa = &ia->ia_ifa;
+               lck_rw_lock_exclusive(in_ifaddr_rwlock);
+               /* Release ia_link reference */
+               IFA_REMREF(ifa);
                TAILQ_REMOVE(&in_ifaddrhead, ia, ia_link);
+               IFA_LOCK(ifa);
+               if (IA_IS_HASHED(ia))
+                       in_iahash_remove(ia);
+               IFA_UNLOCK(ifa);
+               lck_rw_done(in_ifaddr_rwlock);
+
                /*
                 * in_ifscrub kills the interface route.
                 */
-               in_ifscrub(ifp, ia, 1);
-               ifa = &ia->ia_ifa;
-               lck_mtx_unlock(rt_mtx);
+               in_ifscrub(ifp, ia, 0);
                ifnet_lock_exclusive(ifp);
+               IFA_LOCK(ifa);
+               /* if_detach_ifa() releases ifa_link reference */
                if_detach_ifa(ifp, ifa);
-               ifafree(&ia->ia_ifa);
-        
-#ifdef __APPLE__
-       /*
-        * If the interface supports multicast, and no address is left,
-        * remove the "all hosts" multicast group from that interface.
-        */
-               if (ifp->if_flags & IFF_MULTICAST) {
-                       struct in_addr addr;
-                       struct in_multi *inm = NULL;
-
-                       TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 
-                               if (ifa->ifa_addr->sa_family == AF_INET)
-                                       break;
+               /* Our reference to this address is dropped at the bottom */
+               IFA_UNLOCK(ifa);
+
+               /* invalidate route caches */
+               routegenid_inet_update();
 
-                       if (ifa == 0) {
-                               addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
-                               IN_LOOKUP_MULTI(addr, ifp, inm);
+               /*
+                * If the interface supports multicast, and no address is left,
+                * remove the "all hosts" multicast group from that interface.
+                */
+               if ((ifp->if_flags & IFF_MULTICAST) ||
+                   ifp->if_allhostsinm != NULL) {
+
+                       TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
+                               IFA_LOCK(ifa);
+                               if (ifa->ifa_addr->sa_family == AF_INET) {
+                                       IFA_UNLOCK(ifa);
+                                       break;
+                               }
+                               IFA_UNLOCK(ifa);
                        }
                        ifnet_lock_done(ifp);
-                       if (inm)
-                               in_delmulti(&inm);
-               } else 
+
+                       lck_mtx_lock(&ifp->if_addrconfig_lock);
+                       if (ifa == NULL && ifp->if_allhostsinm != NULL) {
+                               struct in_multi *inm = ifp->if_allhostsinm;
+                               ifp->if_allhostsinm = NULL;
+
+                               in_delmulti(inm);
+                               /* release the reference for allhostsinm */
+                               INM_REMREF(inm);
+                       }
+                       lck_mtx_unlock(&ifp->if_addrconfig_lock);
+               } else {
                        ifnet_lock_done(ifp);
-#endif
+               }
 
                /* Post the kernel event */
-               kev_post_msg(&ev_msg);
-               break;
+               dlil_post_complete_msg(ifp, &ev_msg);
 
-#ifdef __APPLE__
-    case SIOCSETOT: {
-        /*
-         * Inspiration from tcp_ctloutput() and ip_ctloutput()
-         * Special ioctl for OpenTransport sockets
-         */
-        struct inpcb   *inp, *cloned_inp;
-        int                    error2 = 0;
-        int                    cloned_fd = *(int *)data;
-
-        inp = sotoinpcb(so);
-        if (inp == NULL) {
-            break;
-        }
-
-        /* let's make sure it's either -1 or a valid file descriptor */
-        if (cloned_fd != -1) {
-            struct socket      *cloned_so;
-            error2 = file_socket(cloned_fd, &cloned_so);
-            if (error2){
-                break;
-            }
-            cloned_inp = sotoinpcb(cloned_so);
-                       file_drop(cloned_fd);
-        } else {
-            cloned_inp = NULL;
-        }
-
-        if (cloned_inp == NULL) {
-            /* OT always uses IP_PORTRANGE_HIGH */
-            inp->inp_flags &= ~(INP_LOWPORT);
-            inp->inp_flags |= INP_HIGHPORT;
-            /* For UDP, OT allows broadcast by default */
-            if (so->so_type == SOCK_DGRAM)
-                so->so_options |= SO_BROADCAST;
-            /* For TCP we want to see MSG_OOB when receive urgent data */
-            else if (so->so_type == SOCK_STREAM)
-                so->so_options |= SO_WANTOOBFLAG;
-        } else {
-            inp->inp_ip_tos = cloned_inp->inp_ip_tos;
-            inp->inp_ip_ttl = cloned_inp->inp_ip_ttl;
-            inp->inp_flags = cloned_inp->inp_flags;
-
-            /* Multicast options */
-            if (cloned_inp->inp_moptions != NULL) {
-                int                    i;
-                struct ip_moptions     *cloned_imo = cloned_inp->inp_moptions;
-                struct ip_moptions     *imo = inp->inp_moptions;
-
-                if (imo == NULL) {
-                    /*
-                     * No multicast option buffer attached to the pcb;
-                     * allocate one.
-                     */
-                    imo = (struct ip_moptions*)
-                        _MALLOC(sizeof(*imo), M_IPMOPTS, M_WAITOK);
-                    if (imo == NULL) {
-                        error2 = ENOBUFS;
-                        break;
-                    }
-                    inp->inp_moptions = imo;
-                }
-                imo->imo_multicast_ifp = cloned_imo->imo_multicast_ifp;
-                imo->imo_multicast_vif = cloned_imo->imo_multicast_vif;
-                imo->imo_multicast_ttl = cloned_imo->imo_multicast_ttl;
-                imo->imo_multicast_loop = cloned_imo->imo_multicast_loop;
-                imo->imo_num_memberships = cloned_imo->imo_num_memberships;
-                for (i = 0; i < cloned_imo->imo_num_memberships; i++) {
-                    imo->imo_membership[i] =
-                    in_addmulti(&cloned_imo->imo_membership[i]->inm_addr,
-                                cloned_imo->imo_membership[i]->inm_ifp);
-                                       if (imo->imo_membership[i] == NULL) {
-                                               error2 = ENOBUFS;
-                                               break;
-                                       }
-                }
-                if (i < cloned_imo->imo_num_memberships) {
-                       /* Failed, perform cleanup */
-                       for (i--; i >= 0; i--)
-                               in_delmulti(&imo->imo_membership[i]);
-                       imo->imo_num_memberships = 0;
-                       break;
-                }
-            }
-        }
-        break;
-    }
-#endif /* __APPLE__ */
+               /*
+                * See if there is any IPV4 address left and if so,
+                * reconfigure KDP to use current primary address.
+                */
+               ifa = ifa_ifpgetprimary(ifp, AF_INET);
+               if (ifa != NULL) {
+                       /*
+                        * NOTE: SIOCSIFADDR is defined with struct ifreq
+                        * as parameter, but here we are sending it down
+                        * to the interface with a pointer to struct ifaddr,
+                        * for legacy reasons.
+                        */
+                       error = ifnet_ioctl(ifp, PF_INET, SIOCSIFADDR, ifa);
+                       if (error == EOPNOTSUPP)
+                               error = 0;
+
+                       /* Release reference from ifa_ifpgetprimary() */
+                       IFA_REMREF(ifa);
+               }
+               (void) ifnet_notify_address(ifp, AF_INET);
+               break;
 
        default:
-               return EOPNOTSUPP;
+               VERIFY(0);
+               /* NOTREACHED */
        }
-       return (0);
+
+       return (error);
 }
 
 /*
- * SIOC[GAD]LIFADDR.
- *     SIOCGLIFADDR: get first address. (?!?)
- *     SIOCGLIFADDR with IFLR_PREFIX:
- *             get first address that matches the specified prefix.
- *     SIOCALIFADDR: add the specified address.
- *     SIOCALIFADDR with IFLR_PREFIX:
- *             EINVAL since we can't deduce hostid part of the address.
- *     SIOCDLIFADDR: delete the specified address.
- *     SIOCDLIFADDR with IFLR_PREFIX:
- *             delete the first address that matches the specified prefix.
- * return values:
- *     EINVAL on invalid parameters
- *     EADDRNOTAVAIL on prefix match failed/specified address not found
- *     other values may be returned from in_ioctl()
+ * Caller passes in the ioctl data pointer directly via "ifr", with the
+ * expectation that this routine always uses bcopy() or other byte-aligned
+ * memory accesses.
  */
-static int
-in_lifaddr_ioctl(
-       struct socket *so,
-       u_long cmd,
-       caddr_t data,
-       struct ifnet *ifp,
-       struct proc *p)
+static __attribute__((noinline)) int
+inctl_ifdstaddr(struct ifnet *ifp, struct in_ifaddr *ia, u_long cmd,
+    struct ifreq *ifr)
 {
-       struct if_laddrreq *iflr = (struct if_laddrreq *)data;
-       struct ifaddr *ifa;
+       struct kev_in_data in_event_data;
+       struct kev_msg ev_msg;
+       struct sockaddr_in dstaddr;
+       int error = 0;
 
-       /* sanity checks */
-       if (!data || !ifp) {
-               panic("invalid argument to in_lifaddr_ioctl");
-               /*NOTRECHED*/
-       }
+       VERIFY(ifp != NULL);
+
+       if (!(ifp->if_flags & IFF_POINTOPOINT))
+               return (EINVAL);
+
+       bzero(&in_event_data, sizeof (struct kev_in_data));
+       bzero(&ev_msg, sizeof (struct kev_msg));
 
        switch (cmd) {
-       case SIOCGLIFADDR:
-               /* address must be specified on GET with IFLR_PREFIX */
-               if ((iflr->flags & IFLR_PREFIX) == 0)
+       case SIOCGIFDSTADDR:            /* struct ifreq */
+               if (ia == NULL) {
+                       error = EADDRNOTAVAIL;
                        break;
-               /*FALLTHROUGH*/
-       case SIOCALIFADDR:
-       case SIOCDLIFADDR:
-               /* address must be specified on ADD and DELETE */
-               if (iflr->addr.ss_family != AF_INET)
-                       return EINVAL;
-               if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
-                       return EINVAL;
-               /* XXX need improvement */
-               if (iflr->dstaddr.ss_family
-                && iflr->dstaddr.ss_family != AF_INET)
-                       return EINVAL;
-               if (iflr->dstaddr.ss_family
-                && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
-                       return EINVAL;
+               }
+               IFA_LOCK(&ia->ia_ifa);
+               bcopy(&ia->ia_dstaddr, &ifr->ifr_dstaddr, sizeof (dstaddr));
+               IFA_UNLOCK(&ia->ia_ifa);
+               break;
+
+       case SIOCSIFDSTADDR:            /* struct ifreq */
+               VERIFY(ia != NULL);
+               IFA_LOCK(&ia->ia_ifa);
+               dstaddr = ia->ia_dstaddr;
+               bcopy(&ifr->ifr_dstaddr, &ia->ia_dstaddr, sizeof (dstaddr));
+               if (ia->ia_dstaddr.sin_family == AF_INET)
+                       ia->ia_dstaddr.sin_len = sizeof (struct sockaddr_in);
+               IFA_UNLOCK(&ia->ia_ifa);
+               /*
+                * NOTE: SIOCSIFDSTADDR is defined with struct ifreq
+                * as parameter, but here we are sending it down
+                * to the interface with a pointer to struct ifaddr,
+                * for legacy reasons.
+                */
+               error = ifnet_ioctl(ifp, PF_INET, SIOCSIFDSTADDR, ia);
+               IFA_LOCK(&ia->ia_ifa);
+               if (error == EOPNOTSUPP)
+                       error = 0;
+               if (error != 0) {
+                       ia->ia_dstaddr = dstaddr;
+                       IFA_UNLOCK(&ia->ia_ifa);
+                       break;
+               }
+               IFA_LOCK_ASSERT_HELD(&ia->ia_ifa);
+
+               ev_msg.vendor_code      = KEV_VENDOR_APPLE;
+               ev_msg.kev_class        = KEV_NETWORK_CLASS;
+               ev_msg.kev_subclass     = KEV_INET_SUBCLASS;
+
+               ev_msg.event_code       = KEV_INET_SIFDSTADDR;
+
+               if (ia->ia_ifa.ifa_dstaddr) {
+                       in_event_data.ia_dstaddr = ((struct sockaddr_in *)
+                           (void *)ia->ia_ifa.ifa_dstaddr)->sin_addr;
+               } else {
+                       in_event_data.ia_dstaddr.s_addr = INADDR_ANY;
+               }
+
+               in_event_data.ia_addr           = ia->ia_addr.sin_addr;
+               in_event_data.ia_net            = ia->ia_net;
+               in_event_data.ia_netmask        = ia->ia_netmask;
+               in_event_data.ia_subnet         = ia->ia_subnet;
+               in_event_data.ia_subnetmask     = ia->ia_subnetmask;
+               in_event_data.ia_netbroadcast   = ia->ia_netbroadcast;
+               IFA_UNLOCK(&ia->ia_ifa);
+               (void) strlcpy(&in_event_data.link_data.if_name[0],
+                   ifp->if_name, IFNAMSIZ);
+               in_event_data.link_data.if_family = ifp->if_family;
+               in_event_data.link_data.if_unit  = (u_int32_t)ifp->if_unit;
+
+               ev_msg.dv[0].data_ptr    = &in_event_data;
+               ev_msg.dv[0].data_length = sizeof (struct kev_in_data);
+               ev_msg.dv[1].data_length = 0;
+
+               dlil_post_complete_msg(ifp, &ev_msg);
+
+               lck_mtx_lock(rnh_lock);
+               IFA_LOCK(&ia->ia_ifa);
+               if (ia->ia_flags & IFA_ROUTE) {
+                       ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&dstaddr;
+                       IFA_UNLOCK(&ia->ia_ifa);
+                       rtinit_locked(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
+                       IFA_LOCK(&ia->ia_ifa);
+                       ia->ia_ifa.ifa_dstaddr =
+                           (struct sockaddr *)&ia->ia_dstaddr;
+                       IFA_UNLOCK(&ia->ia_ifa);
+                       rtinit_locked(&(ia->ia_ifa), (int)RTM_ADD,
+                           RTF_HOST|RTF_UP);
+               } else {
+                       IFA_UNLOCK(&ia->ia_ifa);
+               }
+               lck_mtx_unlock(rnh_lock);
                break;
-       default: /*shouldn't happen*/
-               return EOPNOTSUPP;
+
+
+
+       default:
+               VERIFY(0);
+               /* NOTREACHED */
        }
-       if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
-               return EINVAL;
+
+       return (error);
+}
+
+/*
+ * Caller passes in the ioctl data pointer directly via "ifr", with the
+ * expectation that this routine always uses bcopy() or other byte-aligned
+ * memory accesses.
+ */
+static __attribute__((noinline)) int
+inctl_ifbrdaddr(struct ifnet *ifp, struct in_ifaddr *ia, u_long cmd,
+    struct ifreq *ifr)
+{
+       struct kev_in_data in_event_data;
+       struct kev_msg ev_msg;
+       int error = 0;
+
+       VERIFY(ifp != NULL);
+
+       if (ia == NULL)
+               return (EADDRNOTAVAIL);
+
+       if (!(ifp->if_flags & IFF_BROADCAST))
+               return (EINVAL);
+
+       bzero(&in_event_data, sizeof (struct kev_in_data));
+       bzero(&ev_msg, sizeof (struct kev_msg));
 
        switch (cmd) {
-       case SIOCALIFADDR:
-           {
-               struct in_aliasreq ifra;
+       case SIOCGIFBRDADDR:            /* struct ifreq */
+               IFA_LOCK(&ia->ia_ifa);
+               bcopy(&ia->ia_broadaddr, &ifr->ifr_broadaddr,
+                   sizeof (struct sockaddr_in));
+               IFA_UNLOCK(&ia->ia_ifa);
+               break;
 
-               if (iflr->flags & IFLR_PREFIX)
-                       return EINVAL;
+       case SIOCSIFBRDADDR:            /* struct ifreq */
+               IFA_LOCK(&ia->ia_ifa);
+               bcopy(&ifr->ifr_broadaddr, &ia->ia_broadaddr,
+                   sizeof (struct sockaddr_in));
 
-               /* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
-               bzero(&ifra, sizeof(ifra));
-               bcopy(iflr->iflr_name, ifra.ifra_name,
-                       sizeof(ifra.ifra_name));
+               ev_msg.vendor_code      = KEV_VENDOR_APPLE;
+               ev_msg.kev_class        = KEV_NETWORK_CLASS;
+               ev_msg.kev_subclass     = KEV_INET_SUBCLASS;
 
-               bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
+               ev_msg.event_code = KEV_INET_SIFBRDADDR;
 
-               if (iflr->dstaddr.ss_family) {  /*XXX*/
-                       bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
-                               iflr->dstaddr.ss_len);
+               if (ia->ia_ifa.ifa_dstaddr) {
+                       in_event_data.ia_dstaddr = ((struct sockaddr_in *)
+                           (void *)ia->ia_ifa.ifa_dstaddr)->sin_addr;
+               } else {
+                       in_event_data.ia_dstaddr.s_addr = INADDR_ANY;
                }
+               in_event_data.ia_addr           = ia->ia_addr.sin_addr;
+               in_event_data.ia_net            = ia->ia_net;
+               in_event_data.ia_netmask        = ia->ia_netmask;
+               in_event_data.ia_subnet         = ia->ia_subnet;
+               in_event_data.ia_subnetmask     = ia->ia_subnetmask;
+               in_event_data.ia_netbroadcast   = ia->ia_netbroadcast;
+               IFA_UNLOCK(&ia->ia_ifa);
+               (void) strlcpy(&in_event_data.link_data.if_name[0],
+                   ifp->if_name, IFNAMSIZ);
+               in_event_data.link_data.if_family = ifp->if_family;
+               in_event_data.link_data.if_unit  = (u_int32_t)ifp->if_unit;
 
-               ifra.ifra_mask.sin_family = AF_INET;
-               ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
-               in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
-
-               return in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, p);
-           }
-       case SIOCGLIFADDR:
-       case SIOCDLIFADDR:
-           {
-               struct in_ifaddr *ia;
-               struct in_addr mask, candidate, match;
-               struct sockaddr_in *sin;
-               int cmp;
-
-               bzero(&mask, sizeof(mask));
-               if (iflr->flags & IFLR_PREFIX) {
-                       /* lookup a prefix rather than address. */
-                       in_len2mask(&mask, iflr->prefixlen);
-
-                       sin = (struct sockaddr_in *)&iflr->addr;
-                       match.s_addr = sin->sin_addr.s_addr;
-                       match.s_addr &= mask.s_addr;
-
-                       /* if you set extra bits, that's wrong */
-                       if (match.s_addr != sin->sin_addr.s_addr)
-                               return EINVAL;
-
-                       cmp = 1;
-               } else {
-                       if (cmd == SIOCGLIFADDR) {
-                               /* on getting an address, take the 1st match */
-                               cmp = 0;        /*XXX*/
-                       } else {
-                               /* on deleting an address, do exact match */
-                               in_len2mask(&mask, 32);
-                               sin = (struct sockaddr_in *)&iflr->addr;
-                               match.s_addr = sin->sin_addr.s_addr;
+               ev_msg.dv[0].data_ptr    = &in_event_data;
+               ev_msg.dv[0].data_length = sizeof (struct kev_in_data);
+               ev_msg.dv[1].data_length = 0;
 
-                               cmp = 1;
-                       }
+               dlil_post_complete_msg(ifp, &ev_msg);
+               break;
+
+       default:
+               VERIFY(0);
+               /* NOTREACHED */
+       }
+
+       return (error);
+}
+
+/*
+ * Caller passes in the ioctl data pointer directly via "ifr", with the
+ * expectation that this routine always uses bcopy() or other byte-aligned
+ * memory accesses.
+ */
+static __attribute__((noinline)) int
+inctl_ifnetmask(struct ifnet *ifp, struct in_ifaddr *ia, u_long cmd,
+    struct ifreq *ifr)
+{
+       struct kev_in_data in_event_data;
+       struct kev_msg ev_msg;
+       struct sockaddr_in mask;
+       int error = 0;
+
+       VERIFY(ifp != NULL);
+
+       bzero(&in_event_data, sizeof (struct kev_in_data));
+       bzero(&ev_msg, sizeof (struct kev_msg));
+
+       switch (cmd) {
+       case SIOCGIFNETMASK:            /* struct ifreq */
+               if (ia == NULL) {
+                       error = EADDRNOTAVAIL;
+                       break;
                }
+               IFA_LOCK(&ia->ia_ifa);
+               bcopy(&ia->ia_sockmask, &ifr->ifr_addr, sizeof (mask));
+               IFA_UNLOCK(&ia->ia_ifa);
+               break;
 
+       case SIOCSIFNETMASK: {          /* struct ifreq */
+               in_addr_t i;
+
+               bcopy(&ifr->ifr_addr, &mask, sizeof (mask));
+               i = mask.sin_addr.s_addr;
+
+               VERIFY(ia != NULL);
+               IFA_LOCK(&ia->ia_ifa);
+               ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr = i);
+               ev_msg.vendor_code      = KEV_VENDOR_APPLE;
+               ev_msg.kev_class        = KEV_NETWORK_CLASS;
+               ev_msg.kev_subclass     = KEV_INET_SUBCLASS;
+
+               ev_msg.event_code = KEV_INET_SIFNETMASK;
+
+               if (ia->ia_ifa.ifa_dstaddr) {
+                       in_event_data.ia_dstaddr = ((struct sockaddr_in *)
+                           (void *)ia->ia_ifa.ifa_dstaddr)->sin_addr;
+               } else {
+                       in_event_data.ia_dstaddr.s_addr = INADDR_ANY;
+               }
+               in_event_data.ia_addr           = ia->ia_addr.sin_addr;
+               in_event_data.ia_net            = ia->ia_net;
+               in_event_data.ia_netmask        = ia->ia_netmask;
+               in_event_data.ia_subnet         = ia->ia_subnet;
+               in_event_data.ia_subnetmask     = ia->ia_subnetmask;
+               in_event_data.ia_netbroadcast   = ia->ia_netbroadcast;
+               IFA_UNLOCK(&ia->ia_ifa);
+               (void) strlcpy(&in_event_data.link_data.if_name[0],
+                   ifp->if_name, IFNAMSIZ);
+               in_event_data.link_data.if_family = ifp->if_family;
+               in_event_data.link_data.if_unit  = (u_int32_t)ifp->if_unit;
+
+               ev_msg.dv[0].data_ptr    = &in_event_data;
+               ev_msg.dv[0].data_length = sizeof (struct kev_in_data);
+               ev_msg.dv[1].data_length = 0;
+
+               dlil_post_complete_msg(ifp, &ev_msg);
+               break;
+       }
+
+       default:
+               VERIFY(0);
+               /* NOTREACHED */
+       }
+
+       return (error);
+}
+
+/*
+ * Generic INET control operations (ioctl's).
+ *
+ * ifp is NULL if not an interface-specific ioctl.
+ *
+ * Most of the routines called to handle the ioctls would end up being
+ * tail-call optimized, which unfortunately causes this routine to
+ * consume too much stack space; this is the reason for the "noinline"
+ * attribute used on those routines.
+ *
+ * If called directly from within the networking stack (as opposed to via
+ * pru_control), the socket parameter may be NULL.
+ */
+int
+in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
+    struct proc *p)
+{
+       struct ifreq *ifr = (struct ifreq *)(void *)data;
+       struct sockaddr_in addr, dstaddr;
+       struct sockaddr_in sin, *sa = NULL;
+       boolean_t privileged = (proc_suser(p) == 0);
+       boolean_t so_unlocked = FALSE;
+       struct in_ifaddr *ia = NULL;
+       struct ifaddr *ifa;
+       int error = 0;
+
+       /* In case it's NULL, make sure it came from the kernel */
+       VERIFY(so != NULL || p == kernproc);
+
+       /*
+        * ioctls which don't require ifp, but require socket.
+        */
+       switch (cmd) {
+       case SIOCGASSOCIDS32:           /* struct so_aidreq32 */
+       case SIOCGASSOCIDS64:           /* struct so_aidreq64 */
+               return (inctl_associd(so, cmd, data));
+               /* NOTREACHED */
+
+       case SIOCGCONNIDS32:            /* struct so_cidreq32 */
+       case SIOCGCONNIDS64:            /* struct so_cidreq64 */
+               return (inctl_connid(so, cmd, data));
+               /* NOTREACHED */
+
+       case SIOCGCONNINFO32:           /* struct so_cinforeq32 */
+       case SIOCGCONNINFO64:           /* struct so_cinforeq64 */
+               return (inctl_conninfo(so, cmd, data));
+               /* NOTREACHED */
+       }
+
+       /*
+        * The rest of ioctls require ifp; reject if we don't have one;
+        * return ENXIO to be consistent with ifioctl().
+        */
+       if (ifp == NULL)
+               return (ENXIO);
+
+       /*
+        * ioctls which require ifp but not interface address.
+        */
+       switch (cmd) {
+       case SIOCAUTOADDR:              /* struct ifreq */
+               if (!privileged)
+                       return (EPERM);
+               return (inctl_autoaddr(ifp, ifr));
+               /* NOTREACHED */
+
+       case SIOCARPIPLL:               /* struct ifreq */
+               if (!privileged)
+                       return (EPERM);
+               return (inctl_arpipll(ifp, ifr));
+               /* NOTREACHED */
+
+       case SIOCSETROUTERMODE:         /* struct ifreq */
+               if (!privileged)
+                       return (EPERM);
+               return (inctl_setrouter(ifp, ifr));
+               /* NOTREACHED */
+
+       case SIOCPROTOATTACH:           /* struct ifreq */
+               if (!privileged)
+                       return (EPERM);
+               return (in_domifattach(ifp));
+               /* NOTREACHED */
+
+       case SIOCPROTODETACH:           /* struct ifreq */
+               if (!privileged)
+                       return (EPERM);
+
+               /*
+                * If an IPv4 address is still present, refuse to detach.
+                */
                ifnet_lock_shared(ifp);
-               TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
-                       if (ifa->ifa_addr->sa_family != AF_INET6)
-                               continue;
-                       if (!cmp)
-                               break;
-                       candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
-                       candidate.s_addr &= mask.s_addr;
-                       if (candidate.s_addr == match.s_addr)
+               TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
+                       IFA_LOCK(ifa);
+                       if (ifa->ifa_addr->sa_family == AF_INET) {
+                               IFA_UNLOCK(ifa);
                                break;
+                       }
+                       IFA_UNLOCK(ifa);
                }
                ifnet_lock_done(ifp);
-               if (!ifa)
-                       return EADDRNOTAVAIL;
-               ia = (struct in_ifaddr *)ifa;
+               return ((ifa == NULL) ? proto_unplumb(PF_INET, ifp) : EBUSY);
+               /* NOTREACHED */
+       }
 
-               if (cmd == SIOCGLIFADDR) {
-                       /* fill in the if_laddrreq structure */
-                       bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
+       /*
+        * ioctls which require interface address; obtain sockaddr_in.
+        */
+       switch (cmd) {
+       case SIOCAIFADDR:               /* struct {if,in_}aliasreq */
+               if (!privileged)
+                       return (EPERM);
+               bcopy(&((struct in_aliasreq *)(void *)data)->ifra_addr,
+                   &sin, sizeof (sin));
+               sa = &sin;
+               break;
 
-                       if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
-                               bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
-                                       ia->ia_dstaddr.sin_len);
-                       } else
-                               bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
+       case SIOCDIFADDR:               /* struct ifreq */
+       case SIOCSIFADDR:               /* struct ifreq */
+       case SIOCSIFDSTADDR:            /* struct ifreq */
+       case SIOCSIFNETMASK:            /* struct ifreq */
+       case SIOCSIFBRDADDR:            /* struct ifreq */
+               if (!privileged)
+                       return (EPERM);
+               /* FALLTHRU */
+       case SIOCGIFADDR:               /* struct ifreq */
+       case SIOCGIFDSTADDR:            /* struct ifreq */
+       case SIOCGIFNETMASK:            /* struct ifreq */
+       case SIOCGIFBRDADDR:            /* struct ifreq */
+               bcopy(&ifr->ifr_addr, &sin, sizeof (sin));
+               sa = &sin;
+               break;
+       }
 
-                       iflr->prefixlen =
-                               in_mask2len(&ia->ia_sockmask.sin_addr);
+       /*
+        * Find address for this interface, if it exists.
+        *
+        * If an alias address was specified, find that one instead of
+        * the first one on the interface, if possible.
+        */
+       VERIFY(ia == NULL);
+       if (sa != NULL) {
+               struct in_ifaddr *iap;
 
-                       iflr->flags = 0;        /*XXX*/
+               /*
+                * Any failures from this point on must take into account
+                * a non-NULL "ia" with an outstanding reference count, and
+                * therefore requires IFA_REMREF.  Jump to "done" label
+                * instead of calling return if "ia" is valid.
+                */
+               lck_rw_lock_shared(in_ifaddr_rwlock);
+               TAILQ_FOREACH(iap, INADDR_HASH(sa->sin_addr.s_addr), ia_hash) {
+                       IFA_LOCK(&iap->ia_ifa);
+                       if (iap->ia_ifp == ifp &&
+                           iap->ia_addr.sin_addr.s_addr ==
+                           sa->sin_addr.s_addr) {
+                               /*
+                                * Avoid the race condition seen when two
+                                * threads process SIOCDIFADDR command
+                                * at the same time (radar 28942007)
+                                */
+                               if (cmd == SIOCDIFADDR) {
+                                       if (iap->ia_ifa.ifa_debug &
+                                           IFD_DETACHING) {
+                                               IFA_UNLOCK(&iap->ia_ifa);
+                                               continue;
+                                       } else {
+                                               iap->ia_ifa.ifa_debug |=
+                                                   IFD_DETACHING;
+                                       }
+                               }
+                               ia = iap;
+                               IFA_ADDREF_LOCKED(&iap->ia_ifa);
+                               IFA_UNLOCK(&iap->ia_ifa);
+                               break;
+                       }
+                       IFA_UNLOCK(&iap->ia_ifa);
+               }
+               lck_rw_done(in_ifaddr_rwlock);
+
+               if (ia == NULL) {
+                       ifnet_lock_shared(ifp);
+                       TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
+                               iap = ifatoia(ifa);
+                               IFA_LOCK(&iap->ia_ifa);
+                               if (iap->ia_addr.sin_family == AF_INET) {
+                                       ia = iap;
+                                       IFA_UNLOCK(&iap->ia_ifa);
+                                       break;
+                               }
+                               IFA_UNLOCK(&iap->ia_ifa);
+                       }
+                       /* take a reference on ia before releasing lock */
+                       if (ia != NULL)
+                               IFA_ADDREF(&ia->ia_ifa);
+                       ifnet_lock_done(ifp);
+               }
+       }
+
+       /*
+        * Unlock the socket since ifnet_ioctl() may be invoked by
+        * one of the ioctl handlers below.  Socket will be re-locked
+        * prior to returning.
+        */
+       if (so != NULL) {
+               socket_unlock(so, 0);
+               so_unlocked = TRUE;
+       }
 
-                       return 0;
+       switch (cmd) {
+       case SIOCAIFADDR:               /* struct {if,in_}aliasreq */
+       case SIOCDIFADDR:               /* struct ifreq */
+               if (cmd == SIOCAIFADDR) {
+                       bcopy(&((struct in_aliasreq *)(void *)data)->
+                           ifra_addr, &addr, sizeof (addr));
+                       bcopy(&((struct in_aliasreq *)(void *)data)->
+                           ifra_dstaddr, &dstaddr, sizeof (dstaddr));
                } else {
-                       struct in_aliasreq ifra;
-
-                       /* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
-                       bzero(&ifra, sizeof(ifra));
-                       bcopy(iflr->iflr_name, ifra.ifra_name,
-                               sizeof(ifra.ifra_name));
-
-                       bcopy(&ia->ia_addr, &ifra.ifra_addr,
-                               ia->ia_addr.sin_len);
-                       if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
-                               bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
-                                       ia->ia_dstaddr.sin_len);
+                       VERIFY(cmd == SIOCDIFADDR);
+                       bcopy(&((struct ifreq *)(void *)data)->ifr_addr,
+                           &addr, sizeof (addr));
+                       bzero(&dstaddr, sizeof (dstaddr));
+               }
+
+               if (addr.sin_family == AF_INET) {
+                       struct in_ifaddr *oia;
+
+                       lck_rw_lock_shared(in_ifaddr_rwlock);
+                       for (oia = ia; ia; ia = ia->ia_link.tqe_next) {
+                               IFA_LOCK(&ia->ia_ifa);
+                               if (ia->ia_ifp == ifp &&
+                                   ia->ia_addr.sin_addr.s_addr ==
+                                   addr.sin_addr.s_addr) {
+                                       IFA_ADDREF_LOCKED(&ia->ia_ifa);
+                                       IFA_UNLOCK(&ia->ia_ifa);
+                                       break;
+                               }
+                               IFA_UNLOCK(&ia->ia_ifa);
                        }
-                       bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
-                               ia->ia_sockmask.sin_len);
+                       lck_rw_done(in_ifaddr_rwlock);
+                       if (oia != NULL)
+                               IFA_REMREF(&oia->ia_ifa);
+                       if ((ifp->if_flags & IFF_POINTOPOINT) &&
+                           (cmd == SIOCAIFADDR) &&
+                           (dstaddr.sin_addr.s_addr == INADDR_ANY)) {
+                               error = EDESTADDRREQ;
+                               goto done;
+                       }
+               } else if (cmd == SIOCAIFADDR) {
+                       error = EINVAL;
+                       goto done;
+               }
+               if (cmd == SIOCDIFADDR && ia == NULL) {
+                       error = EADDRNOTAVAIL;
+                       goto done;
+               }
+               /* FALLTHROUGH */
+       case SIOCSIFADDR:               /* struct ifreq */
+       case SIOCSIFDSTADDR:            /* struct ifreq */
+       case SIOCSIFNETMASK:            /* struct ifreq */
+               if (cmd == SIOCAIFADDR) {
+                       /* fell thru from above; just repeat it */
+                       bcopy(&((struct in_aliasreq *)(void *)data)->
+                           ifra_addr, &addr, sizeof (addr));
+               } else {
+                       VERIFY(cmd == SIOCDIFADDR || cmd == SIOCSIFADDR ||
+                           cmd == SIOCSIFNETMASK || cmd == SIOCSIFDSTADDR);
+                       bcopy(&((struct ifreq *)(void *)data)->ifr_addr,
+                           &addr, sizeof (addr));
+               }
 
-                       return in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
-                                         ifp, p);
+               if (addr.sin_family != AF_INET && cmd == SIOCSIFADDR) {
+                       error = EINVAL;
+                       goto done;
+               }
+               if (ia == NULL) {
+                       ia = in_ifaddr_alloc(M_WAITOK);
+                       if (ia == NULL) {
+                               error = ENOBUFS;
+                               goto done;
+                       }
+                       ifnet_lock_exclusive(ifp);
+                       ifa = &ia->ia_ifa;
+                       IFA_LOCK(ifa);
+                       /* Hold a reference for this routine */
+                       IFA_ADDREF_LOCKED(ifa);
+                       IA_HASH_INIT(ia);
+                       ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
+                       ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
+                       ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
+                       ia->ia_sockmask.sin_len = 8;
+                       if (ifp->if_flags & IFF_BROADCAST) {
+                               ia->ia_broadaddr.sin_len = sizeof (ia->ia_addr);
+                               ia->ia_broadaddr.sin_family = AF_INET;
+                       }
+                       ia->ia_ifp = ifp;
+                       if (!(ifp->if_flags & IFF_LOOPBACK))
+                               in_interfaces++;
+                       /* if_attach_ifa() holds a reference for ifa_link */
+                       if_attach_ifa(ifp, ifa);
+                       /*
+                        * If we have to go through in_ifinit(), make sure
+                        * to avoid installing route(s) based on this address
+                        * via PFC_IFUP event, before the link resolver (ARP)
+                        * initializes it.
+                        */
+                       if (cmd == SIOCAIFADDR || cmd == SIOCSIFADDR)
+                               ifa->ifa_debug |= IFD_NOTREADY;
+                       IFA_UNLOCK(ifa);
+                       ifnet_lock_done(ifp);
+                       lck_rw_lock_exclusive(in_ifaddr_rwlock);
+                       /* Hold a reference for ia_link */
+                       IFA_ADDREF(ifa);
+                       TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_link);
+                       lck_rw_done(in_ifaddr_rwlock);
+                       /* discard error */
+                       (void) in_domifattach(ifp);
+                       error = 0;
                }
-           }
+               break;
        }
 
-       return EOPNOTSUPP;      /*just for safety*/
+       switch (cmd) {
+       case SIOCGIFDSTADDR:            /* struct ifreq */
+       case SIOCSIFDSTADDR:            /* struct ifreq */
+               error = inctl_ifdstaddr(ifp, ia, cmd, ifr);
+               break;
+
+       case SIOCGIFBRDADDR:            /* struct ifreq */
+       case SIOCSIFBRDADDR:            /* struct ifreq */
+               error = inctl_ifbrdaddr(ifp, ia, cmd, ifr);
+               break;
+
+       case SIOCGIFNETMASK:            /* struct ifreq */
+       case SIOCSIFNETMASK:            /* struct ifreq */
+               error = inctl_ifnetmask(ifp, ia, cmd, ifr);
+               break;
+
+       case SIOCGIFADDR:               /* struct ifreq */
+       case SIOCSIFADDR:               /* struct ifreq */
+       case SIOCAIFADDR:               /* struct {if,in_}aliasreq */
+       case SIOCDIFADDR:               /* struct ifreq */
+               error = inctl_ifaddr(ifp, ia, cmd, ifr);
+               break;
+
+       default:
+               error = EOPNOTSUPP;
+               break;
+       }
+done:
+       if (ia != NULL)
+               IFA_REMREF(&ia->ia_ifa);
+       if (so_unlocked)
+               socket_lock(so, 0);
+
+       return (error);
 }
 
 /*
  * Delete any existing route for an interface.
  */
 void
-in_ifscrub(
-       struct ifnet *ifp,
-       struct in_ifaddr *ia,
-       int locked)
+in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia, int locked)
 {
-
-       if ((ia->ia_flags & IFA_ROUTE) == 0)
+       IFA_LOCK(&ia->ia_ifa);
+       if ((ia->ia_flags & IFA_ROUTE) == 0) {
+               IFA_UNLOCK(&ia->ia_ifa);
                return;
+       }
+       IFA_UNLOCK(&ia->ia_ifa);
        if (!locked)
-               lck_mtx_lock(rt_mtx);
+               lck_mtx_lock(rnh_lock);
        if (ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT))
                rtinit_locked(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
        else
                rtinit_locked(&(ia->ia_ifa), (int)RTM_DELETE, 0);
+       IFA_LOCK(&ia->ia_ifa);
        ia->ia_flags &= ~IFA_ROUTE;
+       IFA_UNLOCK(&ia->ia_ifa);
        if (!locked)
-               lck_mtx_unlock(rt_mtx);
+               lck_mtx_unlock(rnh_lock);
+}
+
+/*
+ * Caller must hold in_ifaddr_rwlock as writer.
+ */
+static void
+in_iahash_remove(struct in_ifaddr *ia)
+{
+       LCK_RW_ASSERT(in_ifaddr_rwlock, LCK_RW_ASSERT_EXCLUSIVE);
+       IFA_LOCK_ASSERT_HELD(&ia->ia_ifa);
+
+       if (!IA_IS_HASHED(ia)) {
+               panic("attempt to remove wrong ia %p from hash table\n", ia);
+               /* NOTREACHED */
+       }
+       TAILQ_REMOVE(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
+       IA_HASH_INIT(ia);
+       if (IFA_REMREF_LOCKED(&ia->ia_ifa) == NULL) {
+               panic("%s: unexpected (missing) refcnt ifa=%p", __func__,
+                   &ia->ia_ifa);
+               /* NOTREACHED */
+       }
+}
+
+/*
+ * Caller must hold in_ifaddr_rwlock as writer.
+ */
+static void
+in_iahash_insert(struct in_ifaddr *ia)
+{
+       LCK_RW_ASSERT(in_ifaddr_rwlock, LCK_RW_ASSERT_EXCLUSIVE);
+       IFA_LOCK_ASSERT_HELD(&ia->ia_ifa);
+
+       if (ia->ia_addr.sin_family != AF_INET) {
+               panic("attempt to insert wrong ia %p into hash table\n", ia);
+               /* NOTREACHED */
+       } else if (IA_IS_HASHED(ia)) {
+               panic("attempt to double-insert ia %p into hash table\n", ia);
+               /* NOTREACHED */
+       }
+       TAILQ_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
+           ia, ia_hash);
+       IFA_ADDREF_LOCKED(&ia->ia_ifa);
+}
+
+/*
+ * Some point to point interfaces that are tunnels borrow the address from
+ * an underlying interface (e.g. VPN server). In order for source address
+ * selection logic to find the underlying interface first, we add the address
+ * of borrowing point to point interfaces at the end of the list.
+ * (see rdar://6733789)
+ *
+ * Caller must hold in_ifaddr_rwlock as writer.
+ */
+static void
+in_iahash_insert_ptp(struct in_ifaddr *ia)
+{
+       struct in_ifaddr *tmp_ifa;
+       struct ifnet *tmp_ifp;
+
+       LCK_RW_ASSERT(in_ifaddr_rwlock, LCK_RW_ASSERT_EXCLUSIVE);
+       IFA_LOCK_ASSERT_HELD(&ia->ia_ifa);
+
+       if (ia->ia_addr.sin_family != AF_INET) {
+               panic("attempt to insert wrong ia %p into hash table\n", ia);
+               /* NOTREACHED */
+       } else if (IA_IS_HASHED(ia)) {
+               panic("attempt to double-insert ia %p into hash table\n", ia);
+               /* NOTREACHED */
+       }
+       IFA_UNLOCK(&ia->ia_ifa);
+       TAILQ_FOREACH(tmp_ifa, INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
+           ia_hash) {
+               IFA_LOCK(&tmp_ifa->ia_ifa);
+               /* ia->ia_addr won't change, so check without lock */
+               if (IA_SIN(tmp_ifa)->sin_addr.s_addr ==
+                   ia->ia_addr.sin_addr.s_addr) {
+                       IFA_UNLOCK(&tmp_ifa->ia_ifa);
+                       break;
+               }
+               IFA_UNLOCK(&tmp_ifa->ia_ifa);
+       }
+       tmp_ifp = (tmp_ifa == NULL) ? NULL : tmp_ifa->ia_ifp;
+
+       IFA_LOCK(&ia->ia_ifa);
+       if (tmp_ifp == NULL) {
+               TAILQ_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
+                   ia, ia_hash);
+       } else {
+               TAILQ_INSERT_TAIL(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
+                   ia, ia_hash);
+       }
+       IFA_ADDREF_LOCKED(&ia->ia_ifa);
 }
 
 /*
@@ -1038,36 +1640,108 @@ in_ifscrub(
  * and routing table entry.
  */
 static int
-in_ifinit(
-       struct ifnet *ifp,
-       struct in_ifaddr *ia,
-       struct sockaddr_in *sin,
-       int scrub)
+in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
+    int scrub)
 {
-       u_long i = ntohl(sin->sin_addr.s_addr);
+       u_int32_t i = ntohl(sin->sin_addr.s_addr);
        struct sockaddr_in oldaddr;
        int flags = RTF_UP, error;
+       struct ifaddr *ifa0;
+       unsigned int cmd;
+       int oldremoved = 0;
 
+       /* Take an extra reference for this routine */
+       IFA_ADDREF(&ia->ia_ifa);
+
+       lck_rw_lock_exclusive(in_ifaddr_rwlock);
+       IFA_LOCK(&ia->ia_ifa);
        oldaddr = ia->ia_addr;
+       if (IA_IS_HASHED(ia)) {
+               oldremoved = 1;
+               in_iahash_remove(ia);
+       }
        ia->ia_addr = *sin;
+       /*
+        * Interface addresses should not contain port or sin_zero information.
+        */
+       SIN(&ia->ia_addr)->sin_family = AF_INET;
+       SIN(&ia->ia_addr)->sin_len = sizeof (struct sockaddr_in);
+       SIN(&ia->ia_addr)->sin_port = 0;
+       bzero(&SIN(&ia->ia_addr)->sin_zero, sizeof (sin->sin_zero));
+       if ((ifp->if_flags & IFF_POINTOPOINT))
+               in_iahash_insert_ptp(ia);
+       else
+               in_iahash_insert(ia);
+       IFA_UNLOCK(&ia->ia_ifa);
+       lck_rw_done(in_ifaddr_rwlock);
 
        /*
-        * Give the interface a chance to initialize
-        * if this is its first address,
-        * and to validate the address if necessary.
+        * Give the interface a chance to initialize if this is its first
+        * address, and to validate the address if necessary.  Send down
+        * SIOCSIFADDR for first address, and SIOCAIFADDR for alias(es).
+        * We find the first IPV4 address assigned to it and check if this
+        * is the same as the one passed into this routine.
         */
-       error = dlil_ioctl(PF_INET, ifp, SIOCSIFADDR, (caddr_t)ia);
+       ifa0 = ifa_ifpgetprimary(ifp, AF_INET);
+       cmd = (&ia->ia_ifa == ifa0) ? SIOCSIFADDR : SIOCAIFADDR;
+       error = ifnet_ioctl(ifp, PF_INET, cmd, ia);
        if (error == EOPNOTSUPP)
-            error = 0;
+               error = 0;
+       /*
+        * If we've just sent down SIOCAIFADDR, send another ioctl down
+        * for SIOCSIFADDR for the first IPV4 address of the interface,
+        * because an address change on one of the addresses will result
+        * in the removal of the previous first IPV4 address.  KDP needs
+        * be reconfigured with the current primary IPV4 address.
+        */
+       if (error == 0 && cmd == SIOCAIFADDR) {
+               /*
+                * NOTE: SIOCSIFADDR is defined with struct ifreq
+                * as parameter, but here we are sending it down
+                * to the interface with a pointer to struct ifaddr,
+                * for legacy reasons.
+                */
+               error = ifnet_ioctl(ifp, PF_INET, SIOCSIFADDR, ifa0);
+               if (error == EOPNOTSUPP)
+                       error = 0;
+       }
+
+       /* Release reference from ifa_ifpgetprimary() */
+       IFA_REMREF(ifa0);
+
        if (error) {
+               lck_rw_lock_exclusive(in_ifaddr_rwlock);
+               IFA_LOCK(&ia->ia_ifa);
+               if (IA_IS_HASHED(ia))
+                       in_iahash_remove(ia);
                ia->ia_addr = oldaddr;
+               if (oldremoved) {
+                       if ((ifp->if_flags & IFF_POINTOPOINT))
+                               in_iahash_insert_ptp(ia);
+                       else
+                               in_iahash_insert(ia);
+               }
+               IFA_UNLOCK(&ia->ia_ifa);
+               lck_rw_done(in_ifaddr_rwlock);
+               /* Release extra reference taken above */
+               IFA_REMREF(&ia->ia_ifa);
                return (error);
        }
+       lck_mtx_lock(rnh_lock);
+       IFA_LOCK(&ia->ia_ifa);
+       /*
+        * Address has been initialized by the link resolver (ARP)
+        * via ifnet_ioctl() above; it may now generate route(s).
+        */
+       ia->ia_ifa.ifa_debug &= ~IFD_NOTREADY;
        if (scrub) {
                ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
-               in_ifscrub(ifp, ia, 0);
+               IFA_UNLOCK(&ia->ia_ifa);
+               in_ifscrub(ifp, ia, 1);
+               IFA_LOCK(&ia->ia_ifa);
                ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
        }
+       IFA_LOCK_ASSERT_HELD(&ia->ia_ifa);
        if (IN_CLASSA(i))
                ia->ia_netmask = IN_CLASSA_NET;
        else if (IN_CLASSB(i))
@@ -1093,19 +1767,32 @@ in_ifinit(
        ia->ia_ifa.ifa_metric = ifp->if_metric;
        if (ifp->if_flags & IFF_BROADCAST) {
                ia->ia_broadaddr.sin_addr.s_addr =
-                       htonl(ia->ia_subnet | ~ia->ia_subnetmask);
+                   htonl(ia->ia_subnet | ~ia->ia_subnetmask);
                ia->ia_netbroadcast.s_addr =
-                       htonl(ia->ia_net | ~ ia->ia_netmask);
+                   htonl(ia->ia_net | ~ ia->ia_netmask);
        } else if (ifp->if_flags & IFF_LOOPBACK) {
                ia->ia_ifa.ifa_dstaddr = ia->ia_ifa.ifa_addr;
                flags |= RTF_HOST;
        } else if (ifp->if_flags & IFF_POINTOPOINT) {
-               if (ia->ia_dstaddr.sin_family != AF_INET)
+               if (ia->ia_dstaddr.sin_family != AF_INET) {
+                       IFA_UNLOCK(&ia->ia_ifa);
+                       lck_mtx_unlock(rnh_lock);
+                       /* Release extra reference taken above */
+                       IFA_REMREF(&ia->ia_ifa);
                        return (0);
+               }
+               ia->ia_dstaddr.sin_len = sizeof (struct sockaddr_in);
                flags |= RTF_HOST;
        }
-       if ((error = rtinit(&(ia->ia_ifa), (int)RTM_ADD, flags)) == 0)
+       IFA_UNLOCK(&ia->ia_ifa);
+
+       if ((error = rtinit_locked(&(ia->ia_ifa), (int)RTM_ADD, flags)) == 0) {
+               IFA_LOCK(&ia->ia_ifa);
                ia->ia_flags |= IFA_ROUTE;
+               IFA_UNLOCK(&ia->ia_ifa);
+       }
+       lck_mtx_unlock(rnh_lock);
+
        /* XXX check if the subnet route points to the same interface */
        if (error == EEXIST)
                error = 0;
@@ -1115,48 +1802,64 @@ in_ifinit(
         * multicast group on that interface.
         */
        if (ifp->if_flags & IFF_MULTICAST) {
-               struct in_multi *inm;
                struct in_addr addr;
 
+               lck_mtx_lock(&ifp->if_addrconfig_lock);
                addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
-               ifnet_lock_shared(ifp);
-               IN_LOOKUP_MULTI(addr, ifp, inm);
-               ifnet_lock_done(ifp);
-               if (inm == 0)
-                       in_addmulti(&addr, ifp);
+               if (ifp->if_allhostsinm == NULL) {
+                       struct in_multi *inm;
+                       inm = in_addmulti(&addr, ifp);
+
+                       if (inm != NULL) {
+                               /*
+                                * Keep the reference on inm added by
+                                * in_addmulti above for storing the
+                                * pointer in allhostsinm.
+                                */
+                               ifp->if_allhostsinm = inm;
+                       } else {
+                               printf("%s: failed to add membership to "
+                                   "all-hosts multicast address on %s\n",
+                                   __func__, if_name(ifp));
+                       }
+               }
+               lck_mtx_unlock(&ifp->if_addrconfig_lock);
+       }
+
+       /* Release extra reference taken above */
+       IFA_REMREF(&ia->ia_ifa);
+
+       if (error == 0) {
+               /* invalidate route caches */
+               routegenid_inet_update();
        }
+
        return (error);
 }
 
-
 /*
- * Return 1 if the address might be a local broadcast address.
+ * Return TRUE if the address might be a local broadcast address.
  */
-int
-in_broadcast(
-       struct in_addr in,
-       struct ifnet *ifp)
+boolean_t
+in_broadcast(struct in_addr in, struct ifnet *ifp)
 {
        struct ifaddr *ifa;
-       u_long t;
+       u_int32_t t;
 
-       if (in.s_addr == INADDR_BROADCAST ||
-           in.s_addr == INADDR_ANY)
-               return 1;
-       if ((ifp->if_flags & IFF_BROADCAST) == 0)
-               return 0;
+       if (in.s_addr == INADDR_BROADCAST || in.s_addr == INADDR_ANY)
+               return (TRUE);
+       if (!(ifp->if_flags & IFF_BROADCAST))
+               return (FALSE);
        t = ntohl(in.s_addr);
+
        /*
         * Look through the list of addresses for a match
         * with a broadcast address.
         */
-#define ia ((struct in_ifaddr *)ifa)
+#define        ia ((struct in_ifaddr *)ifa)
        ifnet_lock_shared(ifp);
        TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
-               if (ifa->ifa_addr == NULL) {
-                       ifnet_lock_done(ifp);
-                       return (0);
-               }
+               IFA_LOCK(ifa);
                if (ifa->ifa_addr->sa_family == AF_INET &&
                    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
                     in.s_addr == ia->ia_netbroadcast.s_addr ||
@@ -1169,158 +1872,788 @@ in_broadcast(
                      * only exist when an interface gets a secondary
                      * address.
                      */
-                    ia->ia_subnetmask != (u_long)0xffffffff) {
+                    ia->ia_subnetmask != (u_int32_t)0xffffffff) {
+                       IFA_UNLOCK(ifa);
                        ifnet_lock_done(ifp);
-                       return 1;
+                       return (TRUE);
                }
+               IFA_UNLOCK(ifa);
        }
        ifnet_lock_done(ifp);
-       return (0);
+       return (FALSE);
 #undef ia
 }
 
-static void
-in_free_inm(
-       void*   ifma_protospec)
+void
+in_purgeaddrs(struct ifnet *ifp)
 {
-       struct in_multi *inm = ifma_protospec;
-       
+       struct ifaddr **ifap;
+       int err, i;
+
+       VERIFY(ifp != NULL);
+
        /*
-        * No remaining claims to this record; let IGMP know that
-        * we are leaving the multicast group.
+        * Be nice, and try the civilized way first.  If we can't get
+        * rid of them this way, then do it the rough way.  We must
+        * only get here during detach time, after the ifnet has been
+        * removed from the global list and arrays.
         */
-       igmp_leavegroup(inm);
-       lck_mtx_lock(rt_mtx);
-       LIST_REMOVE(inm, inm_link);
-       lck_mtx_unlock(rt_mtx);
-       FREE(inm, M_IPMADDR);
+       err = ifnet_get_address_list_family_internal(ifp, &ifap, AF_INET, 1,
+           M_WAITOK, 0);
+       if (err == 0 && ifap != NULL) {
+               struct ifreq ifr;
+
+               bzero(&ifr, sizeof (ifr));
+               (void) snprintf(ifr.ifr_name, sizeof (ifr.ifr_name),
+                   "%s", if_name(ifp));
+
+               for (i = 0; ifap[i] != NULL; i++) {
+                       struct ifaddr *ifa;
+
+                       ifa = ifap[i];
+                       IFA_LOCK(ifa);
+                       bcopy(ifa->ifa_addr, &ifr.ifr_addr,
+                           sizeof (struct sockaddr_in));
+                       IFA_UNLOCK(ifa);
+                       err = in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
+                           kernproc);
+                       /* if we lost the race, ignore it */
+                       if (err == EADDRNOTAVAIL)
+                               err = 0;
+                       if (err != 0) {
+                               char s_addr[MAX_IPv4_STR_LEN];
+                               char s_dstaddr[MAX_IPv4_STR_LEN];
+                               struct in_addr *s, *d;
+
+                               IFA_LOCK(ifa);
+                               s = &((struct sockaddr_in *)
+                                   (void *)ifa->ifa_addr)->sin_addr;
+                               d = &((struct sockaddr_in *)
+                                   (void *)ifa->ifa_dstaddr)->sin_addr;
+                               (void) inet_ntop(AF_INET, &s->s_addr, s_addr,
+                                   sizeof (s_addr));
+                               (void) inet_ntop(AF_INET, &d->s_addr, s_dstaddr,
+                                   sizeof (s_dstaddr));
+                               IFA_UNLOCK(ifa);
+
+                               printf("%s: SIOCDIFADDR ifp=%s ifa_addr=%s "
+                                   "ifa_dstaddr=%s (err=%d)\n", __func__,
+                                   ifp->if_xname, s_addr, s_dstaddr, err);
+                       }
+               }
+               ifnet_free_address_list(ifap);
+       } else if (err != 0 && err != ENXIO) {
+               printf("%s: error retrieving list of AF_INET addresses for "
+                   "ifp=%s (err=%d)\n", __func__, ifp->if_xname, err);
+       }
 }
 
 /*
- * Add an address to the list of IP multicast addresses for a given interface.
+ * Called as part of ip_init
  */
-struct in_multi *
-in_addmulti(
-       struct in_addr *ap,
-       struct ifnet *ifp)
+void
+in_ifaddr_init(void)
 {
-       struct in_multi *inm;
-       int error;
+       in_multi_init();
+
+       PE_parse_boot_argn("ifa_debug", &inifa_debug, sizeof (inifa_debug));
+
+       inifa_size = (inifa_debug == 0) ? sizeof (struct in_ifaddr) :
+           sizeof (struct in_ifaddr_dbg);
+
+       inifa_zone = zinit(inifa_size, INIFA_ZONE_MAX * inifa_size,
+           0, INIFA_ZONE_NAME);
+       if (inifa_zone == NULL) {
+               panic("%s: failed allocating %s", __func__, INIFA_ZONE_NAME);
+               /* NOTREACHED */
+       }
+       zone_change(inifa_zone, Z_EXPAND, TRUE);
+       zone_change(inifa_zone, Z_CALLERACCT, FALSE);
+
+       lck_mtx_init(&inifa_trash_lock, ifa_mtx_grp, ifa_mtx_attr);
+       TAILQ_INIT(&inifa_trash_head);
+}
+
+static struct in_ifaddr *
+in_ifaddr_alloc(int how)
+{
+       struct in_ifaddr *inifa;
+
+       inifa = (how == M_WAITOK) ? zalloc(inifa_zone) :
+           zalloc_noblock(inifa_zone);
+       if (inifa != NULL) {
+               bzero(inifa, inifa_size);
+               inifa->ia_ifa.ifa_free = in_ifaddr_free;
+               inifa->ia_ifa.ifa_debug |= IFD_ALLOC;
+               ifa_lock_init(&inifa->ia_ifa);
+               if (inifa_debug != 0) {
+                       struct in_ifaddr_dbg *inifa_dbg =
+                           (struct in_ifaddr_dbg *)inifa;
+                       inifa->ia_ifa.ifa_debug |= IFD_DEBUG;
+                       inifa->ia_ifa.ifa_trace = in_ifaddr_trace;
+                       inifa->ia_ifa.ifa_attached = in_ifaddr_attached;
+                       inifa->ia_ifa.ifa_detached = in_ifaddr_detached;
+                       ctrace_record(&inifa_dbg->inifa_alloc);
+               }
+       }
+       return (inifa);
+}
+
+static void
+in_ifaddr_free(struct ifaddr *ifa)
+{
+       IFA_LOCK_ASSERT_HELD(ifa);
+
+       if (ifa->ifa_refcnt != 0) {
+               panic("%s: ifa %p bad ref cnt", __func__, ifa);
+               /* NOTREACHED */
+       } if (!(ifa->ifa_debug & IFD_ALLOC)) {
+               panic("%s: ifa %p cannot be freed", __func__, ifa);
+               /* NOTREACHED */
+       }
+       if (ifa->ifa_debug & IFD_DEBUG) {
+               struct in_ifaddr_dbg *inifa_dbg = (struct in_ifaddr_dbg *)ifa;
+               ctrace_record(&inifa_dbg->inifa_free);
+               bcopy(&inifa_dbg->inifa, &inifa_dbg->inifa_old,
+                   sizeof (struct in_ifaddr));
+               if (ifa->ifa_debug & IFD_TRASHED) {
+                       /* Become a regular mutex, just in case */
+                       IFA_CONVERT_LOCK(ifa);
+                       lck_mtx_lock(&inifa_trash_lock);
+                       TAILQ_REMOVE(&inifa_trash_head, inifa_dbg,
+                           inifa_trash_link);
+                       lck_mtx_unlock(&inifa_trash_lock);
+                       ifa->ifa_debug &= ~IFD_TRASHED;
+               }
+       }
+       IFA_UNLOCK(ifa);
+       ifa_lock_destroy(ifa);
+       bzero(ifa, sizeof (struct in_ifaddr));
+       zfree(inifa_zone, ifa);
+}
+
+static void
+in_ifaddr_attached(struct ifaddr *ifa)
+{
+       struct in_ifaddr_dbg *inifa_dbg = (struct in_ifaddr_dbg *)ifa;
+
+       IFA_LOCK_ASSERT_HELD(ifa);
+
+       if (!(ifa->ifa_debug & IFD_DEBUG)) {
+               panic("%s: ifa %p has no debug structure", __func__, ifa);
+               /* NOTREACHED */
+       }
+       if (ifa->ifa_debug & IFD_TRASHED) {
+               /* Become a regular mutex, just in case */
+               IFA_CONVERT_LOCK(ifa);
+               lck_mtx_lock(&inifa_trash_lock);
+               TAILQ_REMOVE(&inifa_trash_head, inifa_dbg, inifa_trash_link);
+               lck_mtx_unlock(&inifa_trash_lock);
+               ifa->ifa_debug &= ~IFD_TRASHED;
+       }
+}
+
+static void
+in_ifaddr_detached(struct ifaddr *ifa)
+{
+       struct in_ifaddr_dbg *inifa_dbg = (struct in_ifaddr_dbg *)ifa;
+
+       IFA_LOCK_ASSERT_HELD(ifa);
+
+       if (!(ifa->ifa_debug & IFD_DEBUG)) {
+               panic("%s: ifa %p has no debug structure", __func__, ifa);
+               /* NOTREACHED */
+       } else if (ifa->ifa_debug & IFD_TRASHED) {
+               panic("%s: ifa %p is already in trash list", __func__, ifa);
+               /* NOTREACHED */
+       }
+       ifa->ifa_debug |= IFD_TRASHED;
+       /* Become a regular mutex, just in case */
+       IFA_CONVERT_LOCK(ifa);
+       lck_mtx_lock(&inifa_trash_lock);
+       TAILQ_INSERT_TAIL(&inifa_trash_head, inifa_dbg, inifa_trash_link);
+       lck_mtx_unlock(&inifa_trash_lock);
+}
+
+static void
+in_ifaddr_trace(struct ifaddr *ifa, int refhold)
+{
+       struct in_ifaddr_dbg *inifa_dbg = (struct in_ifaddr_dbg *)ifa;
+       ctrace_t *tr;
+       u_int32_t idx;
+       u_int16_t *cnt;
+
+       if (!(ifa->ifa_debug & IFD_DEBUG)) {
+               panic("%s: ifa %p has no debug structure", __func__, ifa);
+               /* NOTREACHED */
+       }
+       if (refhold) {
+               cnt = &inifa_dbg->inifa_refhold_cnt;
+               tr = inifa_dbg->inifa_refhold;
+       } else {
+               cnt = &inifa_dbg->inifa_refrele_cnt;
+               tr = inifa_dbg->inifa_refrele;
+       }
+
+       idx = atomic_add_16_ov(cnt, 1) % INIFA_TRACE_HIST_SIZE;
+       ctrace_record(&tr[idx]);
+}
+
+/*
+ * Handle SIOCGASSOCIDS ioctl for PF_INET domain.
+ */
+static int
+in_getassocids(struct socket *so, uint32_t *cnt, user_addr_t aidp)
+{
+       struct inpcb *inp = sotoinpcb(so);
+       sae_associd_t aid;
+
+       if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD)
+               return (EINVAL);
+
+       /* INPCB has no concept of association */
+       aid = SAE_ASSOCID_ANY;
+       *cnt = 0;
+
+       /* just asking how many there are? */
+       if (aidp == USER_ADDR_NULL)
+               return (0);
+
+       return (copyout(&aid, aidp, sizeof (aid)));
+}
+
+/*
+ * Handle SIOCGCONNIDS ioctl for PF_INET domain.
+ */
+static int
+in_getconnids(struct socket *so, sae_associd_t aid, uint32_t *cnt,
+    user_addr_t cidp)
+{
+       struct inpcb *inp = sotoinpcb(so);
+       sae_connid_t cid;
+
+       if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD)
+               return (EINVAL);
+
+       if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL)
+               return (EINVAL);
+
+       /* if connected, return 1 connection count */
+       *cnt = ((so->so_state & SS_ISCONNECTED) ? 1 : 0);
+
+       /* just asking how many there are? */
+       if (cidp == USER_ADDR_NULL)
+               return (0);
+
+       /* if INPCB is connected, assign it connid 1 */
+       cid = ((*cnt != 0) ? 1 : SAE_CONNID_ANY);
+
+       return (copyout(&cid, cidp, sizeof (cid)));
+}
+
+/*
+ * Handle SIOCGCONNINFO ioctl for PF_INET domain.
+ */
+int
+in_getconninfo(struct socket *so, sae_connid_t cid, uint32_t *flags,
+    uint32_t *ifindex, int32_t *soerror, user_addr_t src, socklen_t *src_len,
+    user_addr_t dst, socklen_t *dst_len, uint32_t *aux_type,
+    user_addr_t aux_data, uint32_t *aux_len)
+{
+       struct inpcb *inp = sotoinpcb(so);
        struct sockaddr_in sin;
-       struct ifmultiaddr *ifma;
+       struct ifnet *ifp = NULL;
+       int error = 0;
+       u_int32_t copy_len = 0;
 
        /*
-        * Call generic routine to add membership or increment
-        * refcount.  It wants addresses in the form of a sockaddr,
-        * so we build one here (being careful to zero the unused bytes).
+        * Don't test for INPCB_STATE_DEAD since this may be called
+        * after SOF_PCBCLEARING is set, e.g. after tcp_close().
         */
-       bzero(&sin, sizeof sin);
+       if (inp == NULL) {
+               error = EINVAL;
+               goto out;
+       }
+
+       if (cid != SAE_CONNID_ANY && cid != SAE_CONNID_ALL && cid != 1) {
+               error = EINVAL;
+               goto out;
+       }
+
+       ifp = inp->inp_last_outifp;
+       *ifindex = ((ifp != NULL) ? ifp->if_index : 0);
+       *soerror = so->so_error;
+       *flags = 0;
+       if (so->so_state & SS_ISCONNECTED)
+               *flags |= (CIF_CONNECTED | CIF_PREFERRED);
+       if (inp->inp_flags & INP_BOUND_IF)
+               *flags |= CIF_BOUND_IF;
+       if (!(inp->inp_flags & INP_INADDR_ANY))
+               *flags |= CIF_BOUND_IP;
+       if (!(inp->inp_flags & INP_ANONPORT))
+               *flags |= CIF_BOUND_PORT;
+
+       bzero(&sin, sizeof (sin));
+       sin.sin_len = sizeof (sin);
        sin.sin_family = AF_INET;
-       sin.sin_len = sizeof sin;
-       sin.sin_addr = *ap;
-       error = if_addmulti(ifp, (struct sockaddr *)&sin, &ifma);
-       if (error) {
-               return 0;
+
+       /* source address and port */
+       sin.sin_port = inp->inp_lport;
+       sin.sin_addr.s_addr = inp->inp_laddr.s_addr;
+       if (*src_len == 0) {
+               *src_len = sin.sin_len;
+       } else {
+               if (src != USER_ADDR_NULL) {
+                       copy_len = min(*src_len, sizeof (sin));
+                       error = copyout(&sin, src, copy_len);
+                       if (error != 0)
+                               goto out;
+                       *src_len = copy_len;
+               }
+       }
+
+       /* destination address and port */
+       sin.sin_port = inp->inp_fport;
+       sin.sin_addr.s_addr = inp->inp_faddr.s_addr;
+       if (*dst_len == 0) {
+               *dst_len = sin.sin_len;
+       } else {
+               if (dst != USER_ADDR_NULL) {
+                       copy_len = min(*dst_len, sizeof (sin));
+                       error = copyout(&sin, dst, copy_len);
+                       if (error != 0)
+                               goto out;
+                       *dst_len = copy_len;
+               }
+       }
+
+       if (SOCK_PROTO(so) == IPPROTO_TCP) {
+               struct conninfo_tcp tcp_ci;
+
+               *aux_type = CIAUX_TCP;
+               if (*aux_len == 0) {
+                       *aux_len = sizeof (tcp_ci);
+               } else {
+                       if (aux_data != USER_ADDR_NULL) {
+                               copy_len = min(*aux_len, sizeof (tcp_ci));
+                               bzero(&tcp_ci, sizeof (tcp_ci));
+                               tcp_getconninfo(so, &tcp_ci);
+                               error = copyout(&tcp_ci, aux_data, copy_len);
+                               if (error != 0)
+                                       goto out;
+                               *aux_len = copy_len;
+                       }
+               }
+       } else {
+               *aux_type = 0;
+               *aux_len = 0;
        }
 
+out:
+       return (error);
+}
+
+struct in_llentry {
+       struct llentry          base;
+};
+
+#define        IN_LLTBL_DEFAULT_HSIZE  32
+#define        IN_LLTBL_HASH(k, h) \
+    ((((((((k) >> 8) ^ (k)) >> 8) ^ (k)) >> 8) ^ (k)) & ((h) - 1))
+
+/*
+ * Do actual deallocation of @lle.
+ */
+static void
+in_lltable_destroy_lle_unlocked(struct llentry *lle)
+{
+       LLE_LOCK_DESTROY(lle);
+       LLE_REQ_DESTROY(lle);
+       FREE(lle, M_LLTABLE);
+}
+
+/*
+ * Called by LLE_FREE_LOCKED when number of references
+ * drops to zero.
+ */
+static void
+in_lltable_destroy_lle(struct llentry *lle)
+{
+       LLE_WUNLOCK(lle);
+       in_lltable_destroy_lle_unlocked(lle);
+}
+
+static struct llentry *
+in_lltable_new(struct in_addr addr4, u_int flags)
+{
+#pragma unused(flags)
+       struct in_llentry *lle;
+
+       MALLOC(lle, struct in_llentry *, sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
+       if (lle == NULL)                /* NB: caller generates msg */
+               return NULL;
+
        /*
-        * If ifma->ifma_protospec is null, then if_addmulti() created
-        * a new record.  Otherwise, we are done.
+        * For IPv4 this will trigger "arpresolve" to generate
+        * an ARP request.
         */
-       if (ifma->ifma_protospec != 0) {
-               return ifma->ifma_protospec;
+       lle->base.la_expire = net_uptime(); /* mark expired */
+       lle->base.r_l3addr.addr4 = addr4;
+       lle->base.lle_refcnt = 1;
+       lle->base.lle_free = in_lltable_destroy_lle;
+
+       LLE_LOCK_INIT(&lle->base);
+       LLE_REQ_INIT(&lle->base);
+       //callout_init(&lle->base.lle_timer, 1);
+
+       return (&lle->base);
+}
+
+#define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)      (               \
+    ((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 )
+
+static int
+in_lltable_match_prefix(const struct sockaddr *saddr,
+    const struct sockaddr *smask, u_int flags, struct llentry *lle)
+{
+       struct in_addr addr, mask, lle_addr;
+
+       addr = ((const struct sockaddr_in *)(const void *)saddr)->sin_addr;
+       mask = ((const struct sockaddr_in *)(const void *)smask)->sin_addr;
+       lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
+
+       if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
+               return (0);
+
+       if (lle->la_flags & LLE_IFADDR) {
+               /*
+                * Delete LLE_IFADDR records IFF address & flag matches.
+                * Note that addr is the interface address within prefix
+                * being matched.
+                * Note also we should handle 'ifdown' cases without removing
+                * ifaddr macs.
+                */
+               if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
+                       return (1);
+               return (0);
        }
 
-       inm = (struct in_multi *) _MALLOC(sizeof(*inm), M_IPMADDR, M_WAITOK);
-       if (inm == NULL) {
-               return (NULL);
+       /* flags & LLE_STATIC means deleting both dynamic and static entries */
+       if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
+               return (1);
+
+       return (0);
+}
+
+static void
+in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
+{
+       struct ifnet *ifp;
+       size_t pkts_dropped;
+
+       LLE_WLOCK_ASSERT(lle);
+       KASSERT(llt != NULL, ("lltable is NULL"));
+
+       /* Unlink entry from table if not already */
+       if ((lle->la_flags & LLE_LINKED) != 0) {
+               ifp = llt->llt_ifp;
+               IF_AFDATA_WLOCK_ASSERT(ifp, llt->llt_af);
+               lltable_unlink_entry(llt, lle);
        }
 
-       bzero(inm, sizeof *inm);
-       inm->inm_addr = *ap;
-       inm->inm_ifp = ifp;
-       inm->inm_ifma = ifma;
-       lck_mtx_lock(rt_mtx);
-       if (ifma->ifma_protospec == NULL) {
-               ifma->ifma_protospec = inm;
-               ifma->ifma_free = in_free_inm;
-               LIST_INSERT_HEAD(&in_multihead, inm, inm_link);
+#if 0
+       /* cancel timer */
+       if (callout_stop(&lle->lle_timer) > 0)
+               LLE_REMREF(lle);
+#endif
+       /* Drop hold queue */
+       pkts_dropped = llentry_free(lle);
+       arpstat.dropped += pkts_dropped;
+}
+
+
+static int
+in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
+{
+#pragma unused(flags)
+       struct rtentry *rt;
+
+       KASSERT(l3addr->sa_family == AF_INET,
+                       ("sin_family %d", l3addr->sa_family));
+
+       /* XXX rtalloc1 should take a const param */
+       rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
+       if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
+               log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
+                               inet_ntoa(((const struct sockaddr_in *)(const void *)l3addr)->sin_addr));
+               if (rt != NULL)
+                       rtfree_locked(rt);
+               return (EINVAL);
        }
-       lck_mtx_unlock(rt_mtx);
-       
-       if (ifma->ifma_protospec != inm) {
-               _FREE(inm, M_IPMADDR);
-               return ifma->ifma_protospec;
+       rtfree_locked(rt);
+       return 0;
+}
+
+static inline uint32_t
+in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
+{
+       return (IN_LLTBL_HASH(dst.s_addr, hsize));
+}
+
+static uint32_t
+in_lltable_hash(const struct llentry *lle, uint32_t hsize)
+{
+       return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
+}
+
+
+static void
+in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
+{
+       struct sockaddr_in *sin;
+
+       sin = (struct sockaddr_in *)(void *)sa;
+       bzero(sin, sizeof(*sin));
+       sin->sin_family = AF_INET;
+       sin->sin_len = sizeof(*sin);
+       sin->sin_addr = lle->r_l3addr.addr4;
+}
+
+static inline struct llentry *
+in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
+{
+       struct llentry *lle;
+       struct llentries *lleh;
+       u_int hashidx;
+
+       hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
+       lleh = &llt->lle_head[hashidx];
+       LIST_FOREACH(lle, lleh, lle_next) {
+               if (lle->la_flags & LLE_DELETED)
+                       continue;
+               if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
+                       break;
        }
 
+       return (lle);
+}
+
+static void
+in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
+{
+#pragma unused(llt)
+       lle->la_flags |= LLE_DELETED;
+       //EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
+#ifdef DIAGNOSTIC
+       log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
+#endif
+       llentry_free(lle);
+}
+
+static struct llentry *
+in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
+{
+       const struct sockaddr_in *sin = (const struct sockaddr_in *) (const void *)l3addr;
+       struct ifnet *ifp = llt->llt_ifp;
+       struct llentry *lle;
+
+       KASSERT(l3addr->sa_family == AF_INET,
+                       ("sin_family %d", l3addr->sa_family));
+
        /*
-        * Let IGMP know that we have joined a new IP multicast group.
+        * A route that covers the given address must have
+        * been installed 1st because we are doing a resolution,
+        * verify this.
         */
-       error = igmp_joingroup(inm);
-       if (error) {
-               char addrbuf[16];
-               
-               /*
-                * We can't free the inm because someone else may already be
-                * using it. Once we put it in to ifma->ifma_protospec, it
-                * must exist as long as the ifma does. Might be nice to flag
-                * the error so we can try igmp_joingroup the next time through.
-                */
-               log(LOG_ERR, "igmp_joingroup error %d joining multicast %s on %s%d\n",
-                       error, inet_ntop(AF_INET, &sin.sin_addr, addrbuf, sizeof(addrbuf)),
-                       ifp->if_name, ifp->if_unit);
+       if (!(flags & LLE_IFADDR) &&
+                       in_lltable_rtcheck(ifp, flags, l3addr) != 0)
+               return (NULL);
+
+       lle = in_lltable_new(sin->sin_addr, flags);
+       if (lle == NULL) {
+               log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
+               return (NULL);
+       }
+       lle->la_flags = flags & ~LLE_CREATE;
+       if (flags & LLE_STATIC)
+               lle->r_flags |= RLLE_VALID;
+       if ((flags & LLE_IFADDR) == LLE_IFADDR) {
+               lltable_set_entry_addr(ifp, lle, LLADDR(SDL(ifp->if_lladdr->ifa_addr)));
+               lle->la_flags |= LLE_STATIC;
+               lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
        }
-       
-       return (inm);
+       return (lle);
 }
 
 /*
- * Delete a multicast address record.
+ * Return NULL if not found or marked for deletion.
+ * If found return lle read locked.
  */
-void
-in_delmulti(
-       struct in_multi **inm)
+static struct llentry *
+in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
 {
-       struct in_multi *inm2;
-       
-       lck_mtx_lock(rt_mtx);
-       LIST_FOREACH(inm2, &in_multihead, inm_link) {
-               if (inm2 == *inm)
-                       break;
-       }
-       if (inm2 != *inm) {
-               lck_mtx_unlock(rt_mtx);
-               printf("in_delmulti - ignorning invalid inm (0x%x)\n", *inm);
-               return;
+       const struct sockaddr_in *sin = (const struct sockaddr_in *)(const void *)l3addr;
+       struct llentry *lle;
+
+       IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp, llt->llt_af);
+
+       KASSERT(l3addr->sa_family == AF_INET,
+                       ("sin_family %d", l3addr->sa_family));
+       lle = in_lltable_find_dst(llt, sin->sin_addr);
+
+       if (lle == NULL)
+               return (NULL);
+
+       KASSERT((flags & (LLE_UNLOCKED|LLE_EXCLUSIVE)) !=
+           (LLE_UNLOCKED|LLE_EXCLUSIVE),("wrong lle request flags: 0x%X",
+               flags));
+
+       if (flags & LLE_UNLOCKED)
+               return (lle);
+
+       if (flags & LLE_EXCLUSIVE)
+               LLE_WLOCK(lle);
+       else
+               LLE_RLOCK(lle);
+
+       return (lle);
+}
+
+static int
+in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
+    struct sysctl_req *wr)
+{
+       struct ifnet *ifp = llt->llt_ifp;
+       /* XXX stack use */
+       struct {
+               struct rt_msghdr        rtm;
+               struct sockaddr_in      sin;
+               struct sockaddr_dl      sdl;
+       } arpc;
+       struct sockaddr_dl *sdl;
+       int error;
+
+       bzero(&arpc, sizeof(arpc));
+       /* skip deleted entries */
+       if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
+               return (0);
+       /* Skip if jailed and not a valid IP of the prison. */
+       lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
+       /*
+        * produce a msg made of:
+        *  struct rt_msghdr;
+        *  struct sockaddr_in; (IPv4)
+        *  struct sockaddr_dl;
+        */
+       arpc.rtm.rtm_msglen = sizeof(arpc);
+       arpc.rtm.rtm_version = RTM_VERSION;
+       arpc.rtm.rtm_type = RTM_GET;
+       arpc.rtm.rtm_flags = RTF_UP;
+       arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
+
+       /* publish */
+       if (lle->la_flags & LLE_PUB)
+               arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
+
+       sdl = &arpc.sdl;
+       sdl->sdl_family = AF_LINK;
+       sdl->sdl_len = sizeof(*sdl);
+       sdl->sdl_index = ifp->if_index;
+       sdl->sdl_type = ifp->if_type;
+       if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
+               sdl->sdl_alen = ifp->if_addrlen;
+               bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
+       } else {
+               sdl->sdl_alen = 0;
+               bzero(LLADDR(sdl), ifp->if_addrlen);
        }
-       lck_mtx_unlock(rt_mtx);
-       
-       /* We intentionally do this a bit differently than BSD */
-       if ((*inm)->inm_ifma) {
-               if_delmultiaddr((*inm)->inm_ifma, 0);
-               ifma_release((*inm)->inm_ifma);
+
+       arpc.rtm.rtm_rmx.rmx_expire =
+               lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
+       arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
+       if (lle->la_flags & LLE_STATIC)
+               arpc.rtm.rtm_flags |= RTF_STATIC;
+       if (lle->la_flags & LLE_IFADDR)
+               arpc.rtm.rtm_flags |= RTF_PINNED;
+       arpc.rtm.rtm_flags |= RTF_PINNED;
+       arpc.rtm.rtm_index = ifp->if_index;
+       error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
+
+       return (error);
+}
+
+static struct lltable *
+in_lltattach(struct ifnet *ifp)
+{
+       struct lltable *llt;
+
+       llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
+       llt->llt_af = AF_INET;
+       llt->llt_ifp = ifp;
+
+       llt->llt_lookup = in_lltable_lookup;
+       llt->llt_alloc_entry = in_lltable_alloc;
+       llt->llt_delete_entry = in_lltable_delete_entry;
+       llt->llt_dump_entry = in_lltable_dump_entry;
+       llt->llt_hash = in_lltable_hash;
+       llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
+       llt->llt_free_entry = in_lltable_free_entry;
+       llt->llt_match_prefix = in_lltable_match_prefix;
+       lltable_link(llt);
+
+       return (llt);
+}
+
+struct in_ifaddr*
+inifa_ifpwithflag(struct ifnet * ifp, uint32_t flag)
+{
+       struct ifaddr *ifa;
+
+       ifnet_lock_shared(ifp);
+       TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_link)
+       {
+               IFA_LOCK_SPIN(ifa);
+               if (ifa->ifa_addr->sa_family != AF_INET) {
+                       IFA_UNLOCK(ifa);
+                       continue;
+               }
+               if ((((struct in_ifaddr *)ifa)->ia_flags & flag) == flag) {
+                       IFA_ADDREF_LOCKED(ifa);
+                       IFA_UNLOCK(ifa);
+                       break;
+               }
+               IFA_UNLOCK(ifa);
        }
-       *inm = NULL;
+       ifnet_lock_done(ifp);
+
+       return ((struct in_ifaddr *)ifa);
 }
 
-#if !NFSCLIENT
-int
-inet_aton(char * cp, struct in_addr * pin)
+struct in_ifaddr *
+inifa_ifpclatv4(struct ifnet * ifp)
 {
-    u_char * b = (char *)pin;
-    int           i;
-    char * p;
-
-    for (p = cp, i = 0; i < 4; i++) {
-       u_long l = strtoul(p, 0, 0);
-       if (l > 255)
-           return (FALSE);
-       b[i] = l;
-       p = strchr(p, '.');
-       if (i < 3 && p == NULL)
-           return (FALSE);
-       p++;
-    }
-    return (TRUE);
+       struct ifaddr *ifa;
+
+       ifnet_lock_shared(ifp);
+       TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_link)
+       {
+               uint32_t addr = 0;
+               IFA_LOCK_SPIN(ifa);
+               if (ifa->ifa_addr->sa_family != AF_INET) {
+                       IFA_UNLOCK(ifa);
+                       continue;
+               }
+
+               addr = ntohl(SIN(ifa->ifa_addr)->sin_addr.s_addr);
+               if (!IN_LINKLOCAL(addr) &&
+                   !IN_LOOPBACK(addr)) {
+                       IFA_ADDREF_LOCKED(ifa);
+                       IFA_UNLOCK(ifa);
+                       break;
+               }
+               IFA_UNLOCK(ifa);
+       }
+       ifnet_lock_done(ifp);
+
+       return ((struct in_ifaddr *)ifa);
 }
-#endif