]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/netinet6/esp_input.c
xnu-2782.30.5.tar.gz
[apple/xnu.git] / bsd / netinet6 / esp_input.c
index 7bd490e8f9573a055b9078d2f254e5f18e05aa97..c8a809490c863549cfd8ba5a469e9f0f45cbc516 100644 (file)
@@ -1,4 +1,33 @@
-/*     $KAME: esp_input.c,v 1.22 2000/03/21 05:14:49 itojun Exp $      */
+/*
+ * Copyright (c) 2008-2013 Apple Inc. All rights reserved.
+ *
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
+ * 
+ * This file contains Original Code and/or Modifications of Original Code
+ * as defined in and that are subject to the Apple Public Source License
+ * Version 2.0 (the 'License'). You may not use this file except in
+ * compliance with the License. The rights granted to you under the License
+ * may not be used to create, or enable the creation or redistribution of,
+ * unlawful or unlicensed copies of an Apple operating system, or to
+ * circumvent, violate, or enable the circumvention or violation of, any
+ * terms of an Apple operating system software license agreement.
+ * 
+ * Please obtain a copy of the License at
+ * http://www.opensource.apple.com/apsl/ and read it before using this file.
+ * 
+ * The Original Code and all software distributed under the License are
+ * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
+ * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
+ * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
+ * Please see the License for the specific language governing rights and
+ * limitations under the License.
+ * 
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
+ */
+
+/*     $FreeBSD: src/sys/netinet6/esp_input.c,v 1.1.2.3 2001/07/03 11:01:50 ume Exp $  */
+/*     $KAME: esp_input.c,v 1.55 2001/03/23 08:08:47 itojun Exp $      */
 
 /*
  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
  * RFC1827/2406 Encapsulated Security Payload.
  */
 
-#define _IP_VHL
-#if (defined(__FreeBSD__) && __FreeBSD__ >= 3) || defined(__NetBSD__)
-#include "opt_inet.h"
-#endif
-
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/malloc.h>
 #include <sys/mbuf.h>
+#include <sys/mcache.h>
 #include <sys/domain.h>
 #include <sys/protosw.h>
 #include <sys/socket.h>
 #include <sys/syslog.h>
 
 #include <net/if.h>
+#include <net/if_ipsec.h>
 #include <net/route.h>
-#include <net/netisr.h>
 #include <kern/cpu_number.h>
+#include <kern/locks.h>
 
 #include <netinet/in.h>
 #include <netinet/in_systm.h>
 #include <netinet/ip_var.h>
 #include <netinet/in_var.h>
 #include <netinet/ip_ecn.h>
+#include <netinet/in_pcb.h>
+#include <netinet/udp.h>
+#if INET6
+#include <netinet6/ip6_ecn.h>
+#endif
 
 #if INET6
 #include <netinet/ip6.h>
+#include <netinet6/in6_pcb.h>
 #include <netinet6/ip6_var.h>
 #include <netinet/icmp6.h>
+#include <netinet6/ip6protosw.h>
 #endif
 
 #include <netinet6/ipsec.h>
+#if INET6
+#include <netinet6/ipsec6.h>
+#endif
 #include <netinet6/ah.h>
+#if INET6
+#include <netinet6/ah6.h>
+#endif
 #include <netinet6/esp.h>
+#if INET6
+#include <netinet6/esp6.h>
+#endif
 #include <netkey/key.h>
 #include <netkey/keydb.h>
 #include <netkey/key_debug.h>
 
+#include <net/kpi_protocol.h>
+#include <netinet/kpi_ipfilter_var.h>
 
 #include <net/net_osdep.h>
+#include <mach/sdt.h>
 
+#include <sys/kdebug.h>
+#define DBG_LAYER_BEG          NETDBG_CODE(DBG_NETIPSEC, 1)
+#define DBG_LAYER_END          NETDBG_CODE(DBG_NETIPSEC, 3)
+#define DBG_FNC_ESPIN          NETDBG_CODE(DBG_NETIPSEC, (6 << 8))
+#define DBG_FNC_DECRYPT                NETDBG_CODE(DBG_NETIPSEC, (7 << 8))
 #define IPLEN_FLIPPED
 
-#if INET
-extern struct protosw inetsw[];
-#if defined(__bsdi__) || defined(__NetBSD__)
-extern u_char ip_protox[];
-#endif
+extern lck_mtx_t  *sadb_mutex;
 
+#if INET
 #define ESPMAXLEN \
        (sizeof(struct esp) < sizeof(struct newesp) \
                ? sizeof(struct newesp) : sizeof(struct esp))
 
+static struct ip *
+esp4_input_strip_UDP_encap (struct mbuf *m, int iphlen)
+{
+       // strip the udp header that's encapsulating ESP
+       struct ip *ip;
+       size_t     stripsiz = sizeof(struct udphdr);
+
+       ip = mtod(m, __typeof__(ip));
+       ovbcopy((caddr_t)ip, (caddr_t)(((u_char *)ip) + stripsiz), iphlen);
+       m->m_data += stripsiz;
+       m->m_len -= stripsiz;
+       m->m_pkthdr.len -= stripsiz;
+       ip = mtod(m, __typeof__(ip));
+       ip->ip_len = ip->ip_len - stripsiz;
+       ip->ip_p = IPPROTO_ESP;
+       return ip;
+}
+
 void
 esp4_input(m, off)
        struct mbuf *m;
        int off;
 {
        struct ip *ip;
+#if INET6
+       struct ip6_hdr *ip6;
+#endif /* INET6 */
        struct esp *esp;
        struct esptail esptail;
        u_int32_t spi;
+       u_int32_t seq;
        struct secasvar *sav = NULL;
        size_t taillen;
        u_int16_t nxt;
-       struct esp_algorithm *algo;
+       const struct esp_algorithm *algo;
        int ivlen;
        size_t hlen;
        size_t esplen;
-       int s;
+       sa_family_t     ifamily;
 
+       KERNEL_DEBUG(DBG_FNC_ESPIN | DBG_FUNC_START, 0,0,0,0,0);
        /* sanity check for alignment. */
        if (off % 4 != 0 || m->m_pkthdr.len % 4 != 0) {
                ipseclog((LOG_ERR, "IPv4 ESP input: packet alignment problem "
                        "(off=%d, pktlen=%d)\n", off, m->m_pkthdr.len));
-               ipsecstat.in_inval++;
+               IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
                goto bad;
        }
 
@@ -121,13 +191,24 @@ esp4_input(m, off)
                if (!m) {
                        ipseclog((LOG_DEBUG,
                            "IPv4 ESP input: can't pullup in esp4_input\n"));
-                       ipsecstat.in_inval++;
+                       IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
                        goto bad;
                }
        }
 
+       /* Expect 32-bit aligned data pointer on strict-align platforms */
+       MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
+
        ip = mtod(m, struct ip *);
-       esp = (struct esp *)(((u_int8_t *)ip) + off);
+       // expect udp-encap and esp packets only
+       if (ip->ip_p != IPPROTO_ESP &&
+           !(ip->ip_p == IPPROTO_UDP && off >= sizeof(struct udphdr))) {
+               ipseclog((LOG_DEBUG,
+                         "IPv4 ESP input: invalid protocol type\n"));
+               IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
+               goto bad;
+       }
+       esp = (struct esp *)(void *)(((u_int8_t *)ip) + off);
 #ifdef _IP_VHL
        hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 #else
@@ -143,52 +224,54 @@ esp4_input(m, off)
                ipseclog((LOG_WARNING,
                    "IPv4 ESP input: no key association found for spi %u\n",
                    (u_int32_t)ntohl(spi)));
-               ipsecstat.in_nosa++;
+               IPSEC_STAT_INCREMENT(ipsecstat.in_nosa);
                goto bad;
        }
        KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
-               printf("DP esp4_input called to allocate SA:%p\n", sav));
+           printf("DP esp4_input called to allocate SA:0x%llx\n",
+           (uint64_t)VM_KERNEL_ADDRPERM(sav)));
        if (sav->state != SADB_SASTATE_MATURE
         && sav->state != SADB_SASTATE_DYING) {
                ipseclog((LOG_DEBUG,
                    "IPv4 ESP input: non-mature/dying SA found for spi %u\n",
                    (u_int32_t)ntohl(spi)));
-               ipsecstat.in_badspi++;
+               IPSEC_STAT_INCREMENT(ipsecstat.in_badspi);
                goto bad;
        }
-       if (sav->alg_enc == SADB_EALG_NONE) {
+       algo = esp_algorithm_lookup(sav->alg_enc);
+       if (!algo) {
                ipseclog((LOG_DEBUG, "IPv4 ESP input: "
-                   "unspecified encryption algorithm for spi %u\n",
+                   "unsupported encryption algorithm for spi %u\n",
                    (u_int32_t)ntohl(spi)));
-               ipsecstat.in_badspi++;
+               IPSEC_STAT_INCREMENT(ipsecstat.in_badspi);
                goto bad;
        }
 
-       algo = &esp_algorithms[sav->alg_enc];   /*XXX*/
-
        /* check if we have proper ivlen information */
        ivlen = sav->ivlen;
        if (ivlen < 0) {
                ipseclog((LOG_ERR, "inproper ivlen in IPv4 ESP input: %s %s\n",
                    ipsec4_logpacketstr(ip, spi), ipsec_logsastr(sav)));
-               ipsecstat.in_inval++;
+               IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
                goto bad;
        }
 
+       seq = ntohl(((struct newesp *)esp)->esp_seq);
        if (!((sav->flags & SADB_X_EXT_OLD) == 0 && sav->replay
         && (sav->alg_auth && sav->key_auth)))
                goto noreplaycheck;
 
-       if (sav->alg_auth == SADB_AALG_NULL)
+       if (sav->alg_auth == SADB_X_AALG_NULL ||
+           sav->alg_auth == SADB_AALG_NONE)
                goto noreplaycheck;
 
        /*
         * check for sequence number.
         */
-       if (ipsec_chkreplay(ntohl(((struct newesp *)esp)->esp_seq), sav))
+       if (ipsec_chkreplay(seq, sav))
                ; /*okey*/
        else {
-               ipsecstat.in_espreplay++;
+               IPSEC_STAT_INCREMENT(ipsecstat.in_espreplay);
                ipseclog((LOG_WARNING,
                    "replay packet in IPv4 ESP input: %s %s\n",
                    ipsec4_logpacketstr(ip, spi), ipsec_logsastr(sav)));
@@ -197,34 +280,40 @@ esp4_input(m, off)
 
        /* check ICV */
     {
-       u_char sum0[AH_MAXSUMSIZE];
-       u_char sum[AH_MAXSUMSIZE];
-       struct ah_algorithm *sumalgo;
+       u_char sum0[AH_MAXSUMSIZE] __attribute__((aligned(4)));
+       u_char sum[AH_MAXSUMSIZE] __attribute__((aligned(4)));
+       const struct ah_algorithm *sumalgo;
        size_t siz;
 
-       sumalgo = &ah_algorithms[sav->alg_auth];
+       sumalgo = ah_algorithm_lookup(sav->alg_auth);
+       if (!sumalgo)
+               goto noreplaycheck;
        siz = (((*sumalgo->sumsiz)(sav) + 3) & ~(4 - 1));
+       if (m->m_pkthdr.len < off + ESPMAXLEN + siz) {
+               IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
+               goto bad;
+       }
        if (AH_MAXSUMSIZE < siz) {
                ipseclog((LOG_DEBUG,
                    "internal error: AH_MAXSUMSIZE must be larger than %lu\n",
-                   (u_long)siz));
-               ipsecstat.in_inval++;
+                   (u_int32_t)siz));
+               IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
                goto bad;
        }
 
-       m_copydata(m, m->m_pkthdr.len - siz, siz, &sum0[0]);
+       m_copydata(m, m->m_pkthdr.len - siz, siz, (caddr_t) &sum0[0]);
 
        if (esp_auth(m, off, m->m_pkthdr.len - off - siz, sav, sum)) {
                ipseclog((LOG_WARNING, "auth fail in IPv4 ESP input: %s %s\n",
                    ipsec4_logpacketstr(ip, spi), ipsec_logsastr(sav)));
-               ipsecstat.in_espauthfail++;
+               IPSEC_STAT_INCREMENT(ipsecstat.in_espauthfail);
                goto bad;
        }
 
        if (bcmp(sum0, sum, siz) != 0) {
                ipseclog((LOG_WARNING, "auth fail in IPv4 ESP input: %s %s\n",
                    ipsec4_logpacketstr(ip, spi), ipsec_logsastr(sav)));
-               ipsecstat.in_espauthfail++;
+               IPSEC_STAT_INCREMENT(ipsecstat.in_espauthfail);
                goto bad;
        }
 
@@ -237,15 +326,15 @@ esp4_input(m, off)
        ip->ip_len = htons(ntohs(ip->ip_len) - siz);
 #endif
        m->m_flags |= M_AUTHIPDGM;
-       ipsecstat.in_espauthsucc++;
+       IPSEC_STAT_INCREMENT(ipsecstat.in_espauthsucc);
     }
 
        /*
         * update sequence number.
         */
        if ((sav->flags & SADB_X_EXT_OLD) == 0 && sav->replay) {
-               if (ipsec_updatereplay(ntohl(((struct newesp *)esp)->esp_seq), sav)) {
-                       ipsecstat.in_espreplay++;
+               if (ipsec_updatereplay(seq, sav)) {
+                       IPSEC_STAT_INCREMENT(ipsecstat.in_espreplay);
                        goto bad;
                }
        }
@@ -267,7 +356,7 @@ noreplaycheck:
        if (m->m_pkthdr.len < off + esplen + ivlen + sizeof(esptail)) {
                ipseclog((LOG_WARNING,
                    "IPv4 ESP input: packet too short\n"));
-               ipsecstat.in_inval++;
+               IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
                goto bad;
        }
 
@@ -276,27 +365,38 @@ noreplaycheck:
                if (!m) {
                        ipseclog((LOG_DEBUG,
                            "IPv4 ESP input: can't pullup in esp4_input\n"));
-                       ipsecstat.in_inval++;
+                       IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
                        goto bad;
                }
        }
 
-    {
+       /*
+        * pre-compute and cache intermediate key
+        */
+       if (esp_schedule(algo, sav) != 0) {
+               IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
+               goto bad;
+       }
+
        /*
         * decrypt the packet.
         */
        if (!algo->decrypt)
                panic("internal error: no decrypt function");
+       KERNEL_DEBUG(DBG_FNC_DECRYPT | DBG_FUNC_START, 0,0,0,0,0);
        if ((*algo->decrypt)(m, off, sav, algo, ivlen)) {
-               ipseclog((LOG_ERR, "decrypt fail in IPv4 ESP input: %s %s\n",
-                   ipsec4_logpacketstr(ip, spi), ipsec_logsastr(sav)));
-               ipsecstat.in_inval++;
+               /* m is already freed */
+               m = NULL;
+               ipseclog((LOG_ERR, "decrypt fail in IPv4 ESP input: %s\n",
+                   ipsec_logsastr(sav)));
+               IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
+               KERNEL_DEBUG(DBG_FNC_DECRYPT | DBG_FUNC_END, 1,0,0,0,0);
                goto bad;
        }
-       ipsecstat.in_esphist[sav->alg_enc]++;
+       KERNEL_DEBUG(DBG_FNC_DECRYPT | DBG_FUNC_END, 2,0,0,0,0);
+       IPSEC_STAT_INCREMENT(ipsecstat.in_esphist[sav->alg_enc]);
 
        m->m_flags |= M_DECRYPTED;
-    }
 
        /*
         * find the trailer of the ESP.
@@ -311,21 +411,63 @@ noreplaycheck:
                ipseclog((LOG_WARNING,
                    "bad pad length in IPv4 ESP input: %s %s\n",
                    ipsec4_logpacketstr(ip, spi), ipsec_logsastr(sav)));
-               ipsecstat.in_inval++;
+               IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
                goto bad;
        }
 
        /* strip off the trailing pad area. */
        m_adj(m, -taillen);
-
+       ip = mtod(m, struct ip *);
 #ifdef IPLEN_FLIPPED
        ip->ip_len = ip->ip_len - taillen;
 #else
        ip->ip_len = htons(ntohs(ip->ip_len) - taillen);
 #endif
+       if (ip->ip_p == IPPROTO_UDP) {
+               // offset includes the outer ip and udp header lengths.
+               if (m->m_len < off) {
+                       m = m_pullup(m, off);
+                       if (!m) {
+                               ipseclog((LOG_DEBUG,
+                                         "IPv4 ESP input: invalid udp encapsulated ESP packet length \n"));
+                               IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
+                               goto bad;
+                       }
+               }
+
+               // check the UDP encap header to detect changes in the source port, and then strip the header
+               off -= sizeof(struct udphdr); // off no longer includes the udphdr's size
+               // if peer is behind nat and this is the latest esp packet
+               if ((sav->flags & SADB_X_EXT_NATT_DETECTED_PEER) != 0 &&
+                   (sav->flags & SADB_X_EXT_OLD) == 0 &&
+                   seq && sav->replay &&
+                   seq >= sav->replay->lastseq)  {
+                       struct udphdr *encap_uh = (__typeof__(encap_uh))(void *)((caddr_t)ip + off);
+                       if (encap_uh->uh_sport &&
+                           ntohs(encap_uh->uh_sport) != sav->remote_ike_port) {
+                               sav->remote_ike_port = ntohs(encap_uh->uh_sport);
+                       }
+               }
+               ip = esp4_input_strip_UDP_encap(m, off);
+               esp = (struct esp *)(void *)(((u_int8_t *)ip) + off);
+       }
+
+       if (sav->utun_is_keepalive_fn) {
+               if (sav->utun_is_keepalive_fn(sav->utun_pcb, &m, nxt, sav->flags, (off + esplen + ivlen))) {
+                       if (m) {
+                               // not really bad, we just wanna exit
+                               IPSEC_STAT_INCREMENT(ipsecstat.in_success);
+                               m = NULL;
+                       }
+                       goto bad;
+               }
+       }
 
        /* was it transmitted over the IPsec tunnel SA? */
-       if (ipsec4_tunnel_validate(ip, nxt, sav)) {
+       if (ipsec4_tunnel_validate(m, off + esplen + ivlen, nxt, sav, &ifamily)) {
+               ifaddr_t ifa;
+               struct sockaddr_storage addr;
+
                /*
                 * strip off all the headers that precedes ESP header.
                 *      IP4 xx ESP IP4' payload -> IP4' payload
@@ -337,45 +479,128 @@ noreplaycheck:
 
                tos = ip->ip_tos;
                m_adj(m, off + esplen + ivlen);
-               if (m->m_len < sizeof(*ip)) {
-                       m = m_pullup(m, sizeof(*ip));
-                       if (!m) {
-                               ipsecstat.in_inval++;
-                               goto bad;
+               if (ifamily == AF_INET) {
+                       struct sockaddr_in *ipaddr;
+
+                       if (m->m_len < sizeof(*ip)) {
+                               m = m_pullup(m, sizeof(*ip));
+                               if (!m) {
+                                       IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
+                                       goto bad;
+                               }
                        }
-               }
-               ip = mtod(m, struct ip *);
-               /* ECN consideration. */
-               ip_ecn_egress(ip4_ipsec_ecn, &tos, &ip->ip_tos);
-               if (!key_checktunnelsanity(sav, AF_INET,
+                       ip = mtod(m, struct ip *);
+                       /* ECN consideration. */
+                       ip_ecn_egress(ip4_ipsec_ecn, &tos, &ip->ip_tos);
+                       if (!key_checktunnelsanity(sav, AF_INET,
                            (caddr_t)&ip->ip_src, (caddr_t)&ip->ip_dst)) {
-                       ipseclog((LOG_ERR, "ipsec tunnel address mismatch "
-                           "in IPv4 ESP input: %s %s\n",
+                               ipseclog((LOG_ERR, "ipsec tunnel address mismatch "
+                           "in ESP input: %s %s\n",
                            ipsec4_logpacketstr(ip, spi), ipsec_logsastr(sav)));
-                       ipsecstat.in_inval++;
+                               IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
+                               goto bad;
+                       }
+
+                       if (ip_doscopedroute) {
+                               bzero(&addr, sizeof(addr));
+                               ipaddr = (__typeof__(ipaddr))&addr;
+                               ipaddr->sin_family = AF_INET;
+                               ipaddr->sin_len = sizeof(*ipaddr);
+                               ipaddr->sin_addr = ip->ip_dst;
+                       }
+#if INET6
+               } else if (ifamily == AF_INET6) {
+                       struct sockaddr_in6 *ip6addr;
+
+                       /*
+                        * m_pullup is prohibited in KAME IPv6 input processing
+                        * but there's no other way!
+                        */
+                       if (m->m_len < sizeof(*ip6)) {
+                               m = m_pullup(m, sizeof(*ip6));
+                               if (!m) {
+                                       IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
+                                       goto bad;
+                               }
+                       }
+
+                       /*
+                        * Expect 32-bit aligned data pointer on strict-align
+                        * platforms.
+                        */
+                       MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
+
+                       ip6 = mtod(m, struct ip6_hdr *);
+
+                       /* ECN consideration. */
+                       /* XXX To be fixed later if needed */
+                       //  ip_ecn_egress(ip4_ipsec_ecn, &tos, &ip->ip_tos);
+
+                       if (!key_checktunnelsanity(sav, AF_INET6,
+                           (caddr_t)&ip6->ip6_src, (caddr_t)&ip6->ip6_dst)) {
+                               ipseclog((LOG_ERR, "ipsec tunnel address mismatch "
+                           "in ESP input: %s %s\n",
+                           ipsec6_logpacketstr(ip6, spi), ipsec_logsastr(sav)));
+                               IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
+                               goto bad;
+                       }
+
+                       if (ip6_doscopedroute) {
+                               bzero(&addr, sizeof(addr));
+                               ip6addr = (__typeof__(ip6addr))&addr;
+                               ip6addr->sin6_family = AF_INET6;
+                               ip6addr->sin6_len = sizeof(*ip6addr);
+                               ip6addr->sin6_addr = ip6->ip6_dst;
+                       }
+#endif /* INET6 */
+               } else {
+                       ipseclog((LOG_ERR, "ipsec tunnel unsupported address family "
+                                 "in ESP input\n"));
                        goto bad;
                }
 
-#if 0 /* XXX should call ipfw rather than ipsec_in_reject, shouldn't it ? */
-               /* drop it if it does not match the default policy */
-               if (ipsec4_in_reject(m, NULL)) {
-                       ipsecstat.in_polvio++;
+               key_sa_recordxfer(sav, m);
+               if (ipsec_addhist(m, IPPROTO_ESP, spi) != 0 ||
+                   ipsec_addhist(m, IPPROTO_IPV4, 0) != 0) {
+                       IPSEC_STAT_INCREMENT(ipsecstat.in_nomem);
                        goto bad;
                }
-#endif
 
-               key_sa_recordxfer(sav, m);
+               if (ip_doscopedroute || ip6_doscopedroute) {
+                       // update the receiving interface address based on the inner address
+                       ifa = ifa_ifwithaddr((struct sockaddr *)&addr);
+                       if (ifa) {
+                               m->m_pkthdr.rcvif = ifa->ifa_ifp;
+                               IFA_REMREF(ifa);
+                       }
+               }
 
-               s = splimp();
-               if (IF_QFULL(&ipintrq)) {
-                       ipsecstat.in_inval++;
-                       goto bad;
+               /* Clear the csum flags, they can't be valid for the inner headers */
+               m->m_pkthdr.csum_flags = 0;
+
+               // Input via IPSec interface
+               if (sav->sah->ipsec_if != NULL) {
+                       if (ipsec_inject_inbound_packet(sav->sah->ipsec_if, m) == 0) {
+                               m = NULL;
+                               goto done;
+                       } else {
+                               goto bad;
+                       }
                }
-               IF_ENQUEUE(&ipintrq, m);
-               m = NULL;
-               schednetisr(NETISR_IP); /*can be skipped but to make sure*/
-               splx(s);
+               
+               if (sav->utun_in_fn) {
+                       if (!(sav->utun_in_fn(sav->utun_pcb, &m, ifamily == AF_INET ? PF_INET : PF_INET6))) {
+                               m = NULL;
+                               // we just wanna exit since packet has been completely processed
+                               goto bad;
+                       }
+               }
+
+               if (proto_input(ifamily == AF_INET ? PF_INET : PF_INET6, m) != 0)
+                       goto bad;
+
                nxt = IPPROTO_DONE;
+               KERNEL_DEBUG(DBG_FNC_ESPIN | DBG_FUNC_END, 2,0,0,0,0);
        } else {
                /*
                 * strip off ESP header and IV.
@@ -401,78 +626,166 @@ noreplaycheck:
                ip->ip_p = nxt;
 
                key_sa_recordxfer(sav, m);
+               if (ipsec_addhist(m, IPPROTO_ESP, spi) != 0) {
+                       IPSEC_STAT_INCREMENT(ipsecstat.in_nomem);
+                       goto bad;
+               }
+               
+               /*
+                * Set the csum valid flag, if we authenticated the
+                * packet, the payload shouldn't be corrupt unless
+                * it was corrupted before being signed on the other
+                * side.
+                */
+               if (nxt == IPPROTO_TCP || nxt == IPPROTO_UDP) {
+                       m->m_pkthdr.csum_flags = CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
+                       m->m_pkthdr.csum_data = 0xFFFF;
+               }
 
-               if (nxt != IPPROTO_DONE)
-                       (*ip_protox[nxt]->pr_input)(m, off);
-               else
+               if (nxt != IPPROTO_DONE) {
+                       if ((ip_protox[nxt]->pr_flags & PR_LASTHDR) != 0 &&
+                           ipsec4_in_reject(m, NULL)) {
+                               IPSEC_STAT_INCREMENT(ipsecstat.in_polvio);
+                               goto bad;
+                       }
+                       KERNEL_DEBUG(DBG_FNC_ESPIN | DBG_FUNC_END, 3,0,0,0,0);
+                       
+                       /* translate encapsulated UDP port ? */
+                       if ((sav->flags & SADB_X_EXT_NATT_MULTIPLEUSERS) != 0)  {
+                               struct udphdr   *udp;
+                               
+                               if (nxt != IPPROTO_UDP) {       /* not UPD packet - drop it */
+                                       IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
+                                       goto bad;
+                               }
+                                       
+                               if (m->m_len < off + sizeof(struct udphdr)) {
+                                       m = m_pullup(m, off + sizeof(struct udphdr));
+                                       if (!m) {
+                                               ipseclog((LOG_DEBUG,
+                                                       "IPv4 ESP input: can't pullup UDP header in esp4_input\n"));
+                                               IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
+                                               goto bad;
+                                       }
+                                       ip = mtod(m, struct ip *);
+                               }
+                               udp = (struct udphdr *)(void *)(((u_int8_t *)ip) + off);
+                       
+                               lck_mtx_lock(sadb_mutex);
+                               if (sav->natt_encapsulated_src_port == 0) {     
+                                       sav->natt_encapsulated_src_port = udp->uh_sport;
+                               } else if (sav->natt_encapsulated_src_port != udp->uh_sport) {  /* something wrong */
+                                       IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
+                                       lck_mtx_unlock(sadb_mutex);
+                                       goto bad;
+                               }
+                               lck_mtx_unlock(sadb_mutex);
+                               udp->uh_sport = htons(sav->remote_ike_port);
+                               udp->uh_sum = 0;
+                       }
+
+                       DTRACE_IP6(receive, struct mbuf *, m, struct inpcb *, NULL,
+                               struct ip *, ip, struct ifnet *, m->m_pkthdr.rcvif,
+                               struct ip *, ip, struct ip6_hdr *, NULL);
+
+                       // Input via IPSec interface
+                       if (sav->sah->ipsec_if != NULL) {
+                               ip->ip_len = htons(ip->ip_len + hlen);
+                               ip->ip_off = htons(ip->ip_off);
+                               ip->ip_sum = 0;
+                               ip->ip_sum = ip_cksum_hdr_in(m, hlen);
+                               if (ipsec_inject_inbound_packet(sav->sah->ipsec_if, m) == 0) {
+                                       m = NULL;
+                                       goto done;
+                               } else {
+                                       goto bad;
+                               }
+                       }
+                       
+                       if (sav->utun_in_fn) {
+                               if (!(sav->utun_in_fn(sav->utun_pcb, &m, PF_INET))) {
+                                       m = NULL;
+                                       // we just wanna exit since packet has been completely processed
+                                       goto bad;
+                               }
+                       }
+
+                       ip_proto_dispatch_in(m, off, nxt, 0);
+               } else
                        m_freem(m);
                m = NULL;
        }
 
+done:
        if (sav) {
                KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
-                       printf("DP esp4_input call free SA:%p\n", sav));
-               key_freesav(sav);
+                   printf("DP esp4_input call free SA:0x%llx\n",
+                   (uint64_t)VM_KERNEL_ADDRPERM(sav)));
+               key_freesav(sav, KEY_SADB_UNLOCKED);
        }
-       ipsecstat.in_success++;
+       IPSEC_STAT_INCREMENT(ipsecstat.in_success);
        return;
 
 bad:
        if (sav) {
                KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
-                       printf("DP esp4_input call free SA:%p\n", sav));
-               key_freesav(sav);
+                   printf("DP esp4_input call free SA:0x%llx\n",
+                   (uint64_t)VM_KERNEL_ADDRPERM(sav)));
+               key_freesav(sav, KEY_SADB_UNLOCKED);
        }
        if (m)
                m_freem(m);
+       KERNEL_DEBUG(DBG_FNC_ESPIN | DBG_FUNC_END, 4,0,0,0,0);
        return;
 }
 #endif /* INET */
 
 #if INET6
 int
-esp6_input(mp, offp, proto)
-       struct mbuf **mp;
-       int *offp, proto;
+esp6_input(struct mbuf **mp, int *offp, int proto)
 {
+#pragma unused(proto)
        struct mbuf *m = *mp;
        int off = *offp;
        struct ip6_hdr *ip6;
        struct esp *esp;
        struct esptail esptail;
        u_int32_t spi;
+       u_int32_t seq;
        struct secasvar *sav = NULL;
        size_t taillen;
        u_int16_t nxt;
-       struct esp_algorithm *algo;
+       const struct esp_algorithm *algo;
        int ivlen;
        size_t esplen;
-       int s;
 
        /* sanity check for alignment. */
        if (off % 4 != 0 || m->m_pkthdr.len % 4 != 0) {
                ipseclog((LOG_ERR, "IPv6 ESP input: packet alignment problem "
                        "(off=%d, pktlen=%d)\n", off, m->m_pkthdr.len));
-               ipsec6stat.in_inval++;
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_inval);
                goto bad;
        }
 
 #ifndef PULLDOWN_TEST
-       IP6_EXTHDR_CHECK(m, off, ESPMAXLEN, IPPROTO_DONE);
-       esp = (struct esp *)(mtod(m, caddr_t) + off);
+       IP6_EXTHDR_CHECK(m, off, ESPMAXLEN, {return IPPROTO_DONE;});
+       esp = (struct esp *)(void *)(mtod(m, caddr_t) + off);
 #else
        IP6_EXTHDR_GET(esp, struct esp *, m, off, ESPMAXLEN);
        if (esp == NULL) {
-               ipsec6stat.in_inval++;
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_inval);
                return IPPROTO_DONE;
        }
 #endif
+       /* Expect 32-bit data aligned pointer on strict-align platforms */
+       MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
+
        ip6 = mtod(m, struct ip6_hdr *);
 
        if (ntohs(ip6->ip6_plen) == 0) {
                ipseclog((LOG_ERR, "IPv6 ESP input: "
                    "ESP with IPv6 jumbogram is not supported.\n"));
-               ipsec6stat.in_inval++;
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_inval);
                goto bad;
        }
 
@@ -485,52 +798,55 @@ esp6_input(mp, offp, proto)
                ipseclog((LOG_WARNING,
                    "IPv6 ESP input: no key association found for spi %u\n",
                    (u_int32_t)ntohl(spi)));
-               ipsec6stat.in_nosa++;
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_nosa);
                goto bad;
        }
        KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
-               printf("DP esp6_input called to allocate SA:%p\n", sav));
+           printf("DP esp6_input called to allocate SA:0x%llx\n",
+           (uint64_t)VM_KERNEL_ADDRPERM(sav)));
        if (sav->state != SADB_SASTATE_MATURE
         && sav->state != SADB_SASTATE_DYING) {
                ipseclog((LOG_DEBUG,
                    "IPv6 ESP input: non-mature/dying SA found for spi %u\n",
                    (u_int32_t)ntohl(spi)));
-               ipsec6stat.in_badspi++;
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_badspi);
                goto bad;
        }
-       if (sav->alg_enc == SADB_EALG_NONE) {
+       algo = esp_algorithm_lookup(sav->alg_enc);
+       if (!algo) {
                ipseclog((LOG_DEBUG, "IPv6 ESP input: "
-                   "unspecified encryption algorithm for spi %u\n",
+                   "unsupported encryption algorithm for spi %u\n",
                    (u_int32_t)ntohl(spi)));
-               ipsec6stat.in_badspi++;
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_badspi);
                goto bad;
        }
 
-       algo = &esp_algorithms[sav->alg_enc];   /*XXX*/
-
        /* check if we have proper ivlen information */
        ivlen = sav->ivlen;
        if (ivlen < 0) {
                ipseclog((LOG_ERR, "inproper ivlen in IPv6 ESP input: %s %s\n",
                    ipsec6_logpacketstr(ip6, spi), ipsec_logsastr(sav)));
-               ipsec6stat.in_badspi++;
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_badspi);
                goto bad;
        }
 
+       seq = ntohl(((struct newesp *)esp)->esp_seq);
+
        if (!((sav->flags & SADB_X_EXT_OLD) == 0 && sav->replay
         && (sav->alg_auth && sav->key_auth)))
                goto noreplaycheck;
 
-       if (sav->alg_auth == SADB_AALG_NULL)
+       if (sav->alg_auth == SADB_X_AALG_NULL ||
+           sav->alg_auth == SADB_AALG_NONE)
                goto noreplaycheck;
 
        /*
         * check for sequence number.
         */
-       if (ipsec_chkreplay(ntohl(((struct newesp *)esp)->esp_seq), sav))
+       if (ipsec_chkreplay(seq, sav))
                ; /*okey*/
        else {
-               ipsec6stat.in_espreplay++;
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_espreplay);
                ipseclog((LOG_WARNING,
                    "replay packet in IPv6 ESP input: %s %s\n",
                    ipsec6_logpacketstr(ip6, spi), ipsec_logsastr(sav)));
@@ -539,34 +855,40 @@ esp6_input(mp, offp, proto)
 
        /* check ICV */
     {
-       u_char sum0[AH_MAXSUMSIZE];
-       u_char sum[AH_MAXSUMSIZE];
-       struct ah_algorithm *sumalgo;
+       u_char sum0[AH_MAXSUMSIZE] __attribute__((aligned(4)));
+       u_char sum[AH_MAXSUMSIZE] __attribute__((aligned(4)));
+       const struct ah_algorithm *sumalgo;
        size_t siz;
 
-       sumalgo = &ah_algorithms[sav->alg_auth];
+       sumalgo = ah_algorithm_lookup(sav->alg_auth);
+       if (!sumalgo)
+               goto noreplaycheck;
        siz = (((*sumalgo->sumsiz)(sav) + 3) & ~(4 - 1));
+       if (m->m_pkthdr.len < off + ESPMAXLEN + siz) {
+               IPSEC_STAT_INCREMENT(ipsecstat.in_inval);
+               goto bad;
+       }
        if (AH_MAXSUMSIZE < siz) {
                ipseclog((LOG_DEBUG,
                    "internal error: AH_MAXSUMSIZE must be larger than %lu\n",
-                   (u_long)siz));
-               ipsec6stat.in_inval++;
+                   (u_int32_t)siz));
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_inval);
                goto bad;
        }
 
-       m_copydata(m, m->m_pkthdr.len - siz, siz, &sum0[0]);
+       m_copydata(m, m->m_pkthdr.len - siz, siz, (caddr_t) &sum0[0]);
 
        if (esp_auth(m, off, m->m_pkthdr.len - off - siz, sav, sum)) {
                ipseclog((LOG_WARNING, "auth fail in IPv6 ESP input: %s %s\n",
                    ipsec6_logpacketstr(ip6, spi), ipsec_logsastr(sav)));
-               ipsec6stat.in_espauthfail++;
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_espauthfail);
                goto bad;
        }
 
        if (bcmp(sum0, sum, siz) != 0) {
                ipseclog((LOG_WARNING, "auth fail in IPv6 ESP input: %s %s\n",
                    ipsec6_logpacketstr(ip6, spi), ipsec_logsastr(sav)));
-               ipsec6stat.in_espauthfail++;
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_espauthfail);
                goto bad;
        }
 
@@ -576,15 +898,15 @@ esp6_input(mp, offp, proto)
        ip6->ip6_plen = htons(ntohs(ip6->ip6_plen) - siz);
 
        m->m_flags |= M_AUTHIPDGM;
-       ipsec6stat.in_espauthsucc++;
+       IPSEC_STAT_INCREMENT(ipsec6stat.in_espauthsucc);
     }
 
        /*
         * update sequence number.
         */
        if ((sav->flags & SADB_X_EXT_OLD) == 0 && sav->replay) {
-               if (ipsec_updatereplay(ntohl(((struct newesp *)esp)->esp_seq), sav)) {
-                       ipsec6stat.in_espreplay++;
+               if (ipsec_updatereplay(seq, sav)) {
+                       IPSEC_STAT_INCREMENT(ipsec6stat.in_espreplay);
                        goto bad;
                }
        }
@@ -606,34 +928,44 @@ noreplaycheck:
        if (m->m_pkthdr.len < off + esplen + ivlen + sizeof(esptail)) {
                ipseclog((LOG_WARNING,
                    "IPv6 ESP input: packet too short\n"));
-               ipsec6stat.in_inval++;
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_inval);
                goto bad;
        }
 
 #ifndef PULLDOWN_TEST
-       IP6_EXTHDR_CHECK(m, off, esplen + ivlen, IPPROTO_DONE); /*XXX*/
+       IP6_EXTHDR_CHECK(m, off, esplen + ivlen, return IPPROTO_DONE);  /*XXX*/
 #else
        IP6_EXTHDR_GET(esp, struct esp *, m, off, esplen + ivlen);
        if (esp == NULL) {
-               ipsec6stat.in_inval++;
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_inval);
                m = NULL;
                goto bad;
        }
 #endif
        ip6 = mtod(m, struct ip6_hdr *);        /*set it again just in case*/
 
+       /*
+        * pre-compute and cache intermediate key
+        */
+       if (esp_schedule(algo, sav) != 0) {
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_inval);
+               goto bad;
+       }
+
        /*
         * decrypt the packet.
         */
        if (!algo->decrypt)
                panic("internal error: no decrypt function");
        if ((*algo->decrypt)(m, off, sav, algo, ivlen)) {
-               ipseclog((LOG_ERR, "decrypt fail in IPv6 ESP input: %s %s\n",
-                   ipsec6_logpacketstr(ip6, spi), ipsec_logsastr(sav)));
-               ipsec6stat.in_inval++;
+               /* m is already freed */
+               m = NULL;
+               ipseclog((LOG_ERR, "decrypt fail in IPv6 ESP input: %s\n",
+                   ipsec_logsastr(sav)));
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_inval);
                goto bad;
        }
-       ipsec6stat.in_esphist[sav->alg_enc]++;
+       IPSEC_STAT_INCREMENT(ipsec6stat.in_esphist[sav->alg_enc]);
 
        m->m_flags |= M_DECRYPTED;
 
@@ -650,17 +982,31 @@ noreplaycheck:
                ipseclog((LOG_WARNING,
                    "bad pad length in IPv6 ESP input: %s %s\n",
                    ipsec6_logpacketstr(ip6, spi), ipsec_logsastr(sav)));
-               ipsec6stat.in_inval++;
+               IPSEC_STAT_INCREMENT(ipsec6stat.in_inval);
                goto bad;
        }
 
        /* strip off the trailing pad area. */
        m_adj(m, -taillen);
-
+       ip6 = mtod(m, struct ip6_hdr *);
        ip6->ip6_plen = htons(ntohs(ip6->ip6_plen) - taillen);
 
+       if (sav->utun_is_keepalive_fn) {
+               if (sav->utun_is_keepalive_fn(sav->utun_pcb, &m, nxt, sav->flags, (off + esplen + ivlen))) {
+                       if (m) {
+                               // not really bad, we just wanna exit
+                               IPSEC_STAT_INCREMENT(ipsec6stat.in_success);
+                               m = NULL;
+                       }
+                       goto bad;
+               }
+       }
+
        /* was it transmitted over the IPsec tunnel SA? */
-       if (ipsec6_tunnel_validate(ip6, nxt, sav)) {
+       if (ipsec6_tunnel_validate(m, off + esplen + ivlen, nxt, sav)) {
+               ifaddr_t ifa;
+               struct sockaddr_storage addr;
+
                /*
                 * strip off all the headers that precedes ESP header.
                 *      IP6 xx ESP IP6' payload -> IP6' payload
@@ -682,7 +1028,7 @@ noreplaycheck:
 #endif
                        m = m_pullup(m, sizeof(*ip6));
                        if (!m) {
-                               ipsec6stat.in_inval++;
+                               IPSEC_STAT_INCREMENT(ipsec6stat.in_inval);
                                goto bad;
                        }
                }
@@ -695,29 +1041,55 @@ noreplaycheck:
                            "in IPv6 ESP input: %s %s\n",
                            ipsec6_logpacketstr(ip6, spi),
                            ipsec_logsastr(sav)));
-                       ipsec6stat.in_inval++;
+                       IPSEC_STAT_INCREMENT(ipsec6stat.in_inval);
                        goto bad;
                }
 
-#if 0 /* XXX should call ipfw rather than ipsec_in_reject, shouldn't it ? */
-               /* drop it if it does not match the default policy */
-               if (ipsec6_in_reject(m, NULL)) {
-                       ipsec6stat.in_polvio++;
+               key_sa_recordxfer(sav, m);
+               if (ipsec_addhist(m, IPPROTO_ESP, spi) != 0 || 
+                   ipsec_addhist(m, IPPROTO_IPV6, 0) != 0) {
+                       IPSEC_STAT_INCREMENT(ipsec6stat.in_nomem);
                        goto bad;
                }
-#endif
 
-               key_sa_recordxfer(sav, m);
+               if (ip6_doscopedroute) {
+                       struct sockaddr_in6 *ip6addr;
 
-               s = splimp();
-               if (IF_QFULL(&ip6intrq)) {
-                       ipsec6stat.in_inval++;
-                       goto bad;
+                       bzero(&addr, sizeof(addr));
+                       ip6addr = (__typeof__(ip6addr))&addr;
+                       ip6addr->sin6_family = AF_INET6;
+                       ip6addr->sin6_len = sizeof(*ip6addr);
+                       ip6addr->sin6_addr = ip6->ip6_dst;
+
+                       // update the receiving interface address based on the inner address
+                       ifa = ifa_ifwithaddr((struct sockaddr *)&addr);
+                       if (ifa) {
+                               m->m_pkthdr.rcvif = ifa->ifa_ifp;
+                               IFA_REMREF(ifa);
+                       }
                }
-               IF_ENQUEUE(&ip6intrq, m);
-               m = NULL;
-               schednetisr(NETISR_IPV6); /*can be skipped but to make sure*/
-               splx(s);
+
+               // Input via IPSec interface
+               if (sav->sah->ipsec_if != NULL) {
+                       if (ipsec_inject_inbound_packet(sav->sah->ipsec_if, m) == 0) {
+                               m = NULL;
+                               nxt = IPPROTO_DONE;
+                               goto done;
+                       } else {
+                               goto bad;
+                       }
+               }
+               
+               if (sav->utun_in_fn) {
+                       if (!(sav->utun_in_fn(sav->utun_pcb, &m, PF_INET6))) {
+                               m = NULL;
+                               // we just wanna exit since packet has been completely processed
+                               goto bad;
+                       }
+               }
+
+               if (proto_input(PF_INET6, m) != 0)
+                       goto bad;
                nxt = IPPROTO_DONE;
        } else {
                /*
@@ -755,36 +1127,213 @@ noreplaycheck:
                                goto bad;
                        }
                        m_adj(n, stripsiz);
-                       m_cat(m, n);
                        /* m_cat does not update m_pkthdr.len */
                        m->m_pkthdr.len += n->m_pkthdr.len;
+                       m_cat(m, n);
+               }
+
+#ifndef PULLDOWN_TEST
+               /*
+                * KAME requires that the packet to be contiguous on the
+                * mbuf.  We need to make that sure.
+                * this kind of code should be avoided.
+                * XXX other conditions to avoid running this part?
+                */
+               if (m->m_len != m->m_pkthdr.len) {
+                       struct mbuf *n = NULL;
+                       int maxlen;
+
+                       MGETHDR(n, M_DONTWAIT, MT_HEADER);      /* MAC-OK */
+                       maxlen = MHLEN;
+                       if (n)
+                               M_COPY_PKTHDR(n, m);
+                       if (n && m->m_pkthdr.len > maxlen) {
+                               MCLGET(n, M_DONTWAIT);
+                               maxlen = MCLBYTES;
+                               if ((n->m_flags & M_EXT) == 0) {
+                                       m_free(n);
+                                       n = NULL;
+                               }
+                       }
+                       if (!n) {
+                               printf("esp6_input: mbuf allocation failed\n");
+                               goto bad;
+                       }
+
+                       if (m->m_pkthdr.len <= maxlen) {
+                               m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
+                               n->m_len = m->m_pkthdr.len;
+                               n->m_pkthdr.len = m->m_pkthdr.len;
+                               n->m_next = NULL;
+                               m_freem(m);
+                       } else {
+                               m_copydata(m, 0, maxlen, mtod(n, caddr_t));
+                               n->m_len = maxlen;
+                               n->m_pkthdr.len = m->m_pkthdr.len;
+                               n->m_next = m;
+                               m_adj(m, maxlen);
+                               m->m_flags &= ~M_PKTHDR;
+                       }
+                       m = n;
                }
+#endif
 
                ip6 = mtod(m, struct ip6_hdr *);
                ip6->ip6_plen = htons(ntohs(ip6->ip6_plen) - stripsiz);
 
                key_sa_recordxfer(sav, m);
+               if (ipsec_addhist(m, IPPROTO_ESP, spi) != 0) {
+                       IPSEC_STAT_INCREMENT(ipsec6stat.in_nomem);
+                       goto bad;
+               }
+
+               // Input via IPSec interface
+               if (sav->sah->ipsec_if != NULL) {
+                       if (ipsec_inject_inbound_packet(sav->sah->ipsec_if, m) == 0) {
+                               m = NULL;
+                               nxt = IPPROTO_DONE;
+                               goto done;
+                       } else {
+                               goto bad;
+                       }
+               }
+               
+               if (sav->utun_in_fn) {
+                       if (!(sav->utun_in_fn(sav->utun_pcb, &m, PF_INET6))) {
+                               m = NULL;
+                               // we just wanna exit since packet has been completely processed
+                               goto bad;
+                       }
+               }
        }
 
+done:
        *offp = off;
        *mp = m;
-
        if (sav) {
                KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
-                       printf("DP esp6_input call free SA:%p\n", sav));
-               key_freesav(sav);
+                   printf("DP esp6_input call free SA:0x%llx\n",
+                   (uint64_t)VM_KERNEL_ADDRPERM(sav)));
+               key_freesav(sav, KEY_SADB_UNLOCKED);
        }
-       ipsec6stat.in_success++;
+       IPSEC_STAT_INCREMENT(ipsec6stat.in_success);
        return nxt;
 
 bad:
        if (sav) {
                KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
-                       printf("DP esp6_input call free SA:%p\n", sav));
-               key_freesav(sav);
+                   printf("DP esp6_input call free SA:0x%llx\n",
+                   (uint64_t)VM_KERNEL_ADDRPERM(sav)));
+               key_freesav(sav, KEY_SADB_UNLOCKED);
        }
        if (m)
                m_freem(m);
        return IPPROTO_DONE;
 }
+
+void
+esp6_ctlinput(cmd, sa, d)
+       int cmd;
+       struct sockaddr *sa;
+       void *d;
+{
+       const struct newesp *espp;
+       struct newesp esp;
+       struct ip6ctlparam *ip6cp = NULL, ip6cp1;
+       struct secasvar *sav;
+       struct ip6_hdr *ip6;
+       struct mbuf *m;
+       int off;
+       struct sockaddr_in6 *sa6_src, *sa6_dst;
+
+       if (sa->sa_family != AF_INET6 ||
+           sa->sa_len != sizeof(struct sockaddr_in6))
+               return;
+       if ((unsigned)cmd >= PRC_NCMDS)
+               return;
+
+       /* if the parameter is from icmp6, decode it. */
+       if (d != NULL) {
+               ip6cp = (struct ip6ctlparam *)d;
+               m = ip6cp->ip6c_m;
+               ip6 = ip6cp->ip6c_ip6;
+               off = ip6cp->ip6c_off;
+       } else {
+               m = NULL;
+               ip6 = NULL;
+       }
+
+       if (ip6) {
+               /*
+                * Notify the error to all possible sockets via pfctlinput2.
+                * Since the upper layer information (such as protocol type,
+                * source and destination ports) is embedded in the encrypted
+                * data and might have been cut, we can't directly call
+                * an upper layer ctlinput function. However, the pcbnotify
+                * function will consider source and destination addresses
+                * as well as the flow info value, and may be able to find
+                * some PCB that should be notified.
+                * Although pfctlinput2 will call esp6_ctlinput(), there is
+                * no possibility of an infinite loop of function calls,
+                * because we don't pass the inner IPv6 header.
+                */
+               bzero(&ip6cp1, sizeof(ip6cp1));
+               ip6cp1.ip6c_src = ip6cp->ip6c_src;
+               pfctlinput2(cmd, sa, (void *)&ip6cp1);
+
+               /*
+                * Then go to special cases that need ESP header information.
+                * XXX: We assume that when ip6 is non NULL,
+                * M and OFF are valid.
+                */
+
+               /* check if we can safely examine src and dst ports */
+               if (m->m_pkthdr.len < off + sizeof(esp))
+                       return;
+
+               if (m->m_len < off + sizeof(esp)) {
+                       /*
+                        * this should be rare case,
+                        * so we compromise on this copy...
+                        */
+                       m_copydata(m, off, sizeof(esp), (caddr_t)&esp);
+                       espp = &esp;
+               } else
+                       espp = (struct newesp*)(void *)(mtod(m, caddr_t) + off);
+
+               if (cmd == PRC_MSGSIZE) {
+                       int valid = 0;
+
+                       /*
+                        * Check to see if we have a valid SA corresponding to
+                        * the address in the ICMP message payload.
+                        */
+                       sa6_src = ip6cp->ip6c_src;
+                       sa6_dst = (struct sockaddr_in6 *)(void *)sa;
+                       sav = key_allocsa(AF_INET6,
+                                         (caddr_t)&sa6_src->sin6_addr,
+                                         (caddr_t)&sa6_dst->sin6_addr,
+                                         IPPROTO_ESP, espp->esp_spi);
+                       if (sav) {
+                               if (sav->state == SADB_SASTATE_MATURE ||
+                                   sav->state == SADB_SASTATE_DYING)
+                                       valid++;
+                               key_freesav(sav, KEY_SADB_LOCKED);
+                       }
+
+                       /* XXX Further validation? */
+
+                       /*
+                        * Depending on the value of "valid" and routing table
+                        * size (mtudisc_{hi,lo}wat), we will:
+                        * - recalcurate the new MTU and create the
+                        *   corresponding routing entry, or
+                        * - ignore the MTU change notification.
+                        */
+                       icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
+               }
+       } else {
+               /* we normally notify any pcb here */
+       }
+}
 #endif /* INET6 */