]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/netinet/ip_icmp.c
xnu-1228.tar.gz
[apple/xnu.git] / bsd / netinet / ip_icmp.c
index 8ecf5d4a2f60fa631b572858c8dd79e8d21e00c8..3ed8a2d459e42e5c2993095ec8694bc4f242e3ef 100644 (file)
@@ -1,23 +1,29 @@
 /*
- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2007 Apple Inc. All rights reserved.
  *
- * @APPLE_LICENSE_HEADER_START@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
- * The contents of this file constitute Original Code as defined in and
- * are subject to the Apple Public Source License Version 1.1 (the
- * "License").  You may not use this file except in compliance with the
- * License.  Please obtain a copy of the License at
- * http://www.apple.com/publicsource and read it before using this file.
+ * This file contains Original Code and/or Modifications of Original Code
+ * as defined in and that are subject to the Apple Public Source License
+ * Version 2.0 (the 'License'). You may not use this file except in
+ * compliance with the License. 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.
  * 
- * This Original Code and all software distributed under the License are
- * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
+ * 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 OR NON-INFRINGEMENT.  Please see the
- * License for the specific language governing rights and limitations
- * under the License.
+ * 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_LICENSE_HEADER_END@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /*
  * Copyright (c) 1982, 1986, 1988, 1993
  *
  *     @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
  */
+/*
+ * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
+ * support for mandatory and extensible security protections.  This notice
+ * is included in support of clause 2.2 (b) of the Apple Public License,
+ * Version 2.0.
+ */
 
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <netinet/ip_icmp.h>
 #include <netinet/ip_var.h>
 #include <netinet/icmp_var.h>
+#include <netinet/tcp.h>
+#include <netinet/tcp_fsm.h>
+#include <netinet/tcp_seq.h>
+#include <netinet/tcp_timer.h>
+#include <netinet/tcp_var.h>
+#include <netinet/tcpip.h>
 
 #if IPSEC
 #include <netinet6/ipsec.h>
 #include <net/if_types.h>
 #endif
 
+ /* XXX This one should go in sys/mbuf.h. It is used to avoid that
+ * a firewall-generated packet loops forever through the firewall.
+ */
+#ifndef M_SKIP_FIREWALL
+#define M_SKIP_FIREWALL         0x4000
+#endif 
+
+#if CONFIG_MACF_NET
+#include <security/mac_framework.h>
+#endif /* MAC_NET */
+
 /*
  * ICMP routines: error generation, receive packet processing, and
  * routines to turnaround packets back to the originator, and
@@ -99,6 +128,18 @@ static int  icmpmaskrepl = 0;
 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
        &icmpmaskrepl, 0, "");
 
+static int     icmptimestamp = 0;
+SYSCTL_INT(_net_inet_icmp, ICMPCTL_TIMESTAMP, timestamp, CTLFLAG_RW,
+       &icmptimestamp, 0, "");
+
+static int     drop_redirect = 0;
+SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW, 
+       &drop_redirect, 0, "");
+
+static int     log_redirect = 0;
+SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW, 
+       &log_redirect, 0, "");
+
 #if ICMP_BANDLIM 
  
 /*    
@@ -106,7 +147,7 @@ SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
  *      variable content is -1 and read-only.
  */     
     
-static int      icmplim = 100;
+static int      icmplim = 250;
 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
        &icmplim, 0, "");
 #else
@@ -121,18 +162,18 @@ SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD,
  * ICMP broadcast echo sysctl
  */
 
-static int     icmpbmcastecho = 0;
-SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW, &icmpbmcastecho,
-          0, "");
+static int     icmpbmcastecho = 1;
+SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
+       &icmpbmcastecho, 0, "");
 
 
 #if ICMPPRINTFS
 int    icmpprintfs = 0;
 #endif
 
-static void    icmp_reflect __P((struct mbuf *));
-static void    icmp_send __P((struct mbuf *, struct mbuf *));
-int    ip_next_mtu __P((int, int));
+static void    icmp_reflect(struct mbuf *);
+static void    icmp_send(struct mbuf *, struct mbuf *);
+static int     ip_next_mtu(int, int);
 
 extern struct protosw inetsw[];
 
@@ -141,16 +182,17 @@ extern    struct protosw inetsw[];
  * in response to bad packet ip.
  */
 void
-icmp_error(n, type, code, dest, destifp)
-       struct mbuf *n;
-       int type, code;
-       n_long dest;
-       struct ifnet *destifp;
+icmp_error(
+       struct mbuf *n,
+       int type,
+       int code,
+       n_long dest,
+       struct ifnet *destifp)
 {
-       register struct ip *oip = mtod(n, struct ip *), *nip;
-       register unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
-       register struct icmp *icp;
-       register struct mbuf *m;
+       struct ip *oip = mtod(n, struct ip *), *nip;
+       unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
+       struct icmp *icp;
+       struct mbuf *m;
        unsigned icmplen;
 
 #if ICMPPRINTFS
@@ -160,13 +202,10 @@ icmp_error(n, type, code, dest, destifp)
        if (type != ICMP_REDIRECT)
                icmpstat.icps_error++;
        /*
-        * Don't send error if the original packet was encrypted.
         * Don't send error if not the first fragment of message.
         * Don't error if the old packet protocol was ICMP
         * error message, only known informational types.
         */
-       if (n->m_flags & M_DECRYPTED)
-               goto freeit;
        if (oip->ip_off &~ (IP_MF|IP_DF))
                goto freeit;
        if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
@@ -181,10 +220,24 @@ icmp_error(n, type, code, dest, destifp)
        /*
         * First, formulate icmp message
         */
-       m = m_gethdr(M_DONTWAIT, MT_HEADER);
+       m = m_gethdr(M_DONTWAIT, MT_HEADER);    /* MAC-OK */
        if (m == NULL)
                goto freeit;
-       icmplen = oiplen + min(8, oip->ip_len);
+
+        if (n->m_flags & M_SKIP_FIREWALL) {
+               /* set M_SKIP_FIREWALL to skip firewall check, since we're called from firewall */
+               m->m_flags |= M_SKIP_FIREWALL;
+       }
+
+#if CONFIG_MACF_NET
+       mac_mbuf_label_associate_netlayer(n, m);
+#endif
+       icmplen = min(oiplen + 8, oip->ip_len);
+       if (icmplen < sizeof(struct ip)) {
+               printf("icmp_error: bad length\n");
+               m_free(m);
+               goto freeit;
+       }
        m->m_len = icmplen + ICMP_MINLEN;
        MH_ALIGN(m, m->m_len);
        icp = mtod(m, struct icmp *);
@@ -210,9 +263,14 @@ icmp_error(n, type, code, dest, destifp)
        }
 
        icp->icmp_code = code;
-       bcopy((caddr_t)oip, (caddr_t)&icp->icmp_ip, icmplen);
+       m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
        nip = &icp->icmp_ip;
-       nip->ip_len = htons((u_short)(nip->ip_len + oiplen));
+
+       /*
+        * Convert fields to network representation.
+        */
+       HTONS(nip->ip_len);
+       HTONS(nip->ip_off);
 
        /*
         * Now, copy old ip header (without options)
@@ -224,7 +282,6 @@ icmp_error(n, type, code, dest, destifp)
        m->m_len += sizeof(struct ip);
        m->m_pkthdr.len = m->m_len;
        m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
-       m->m_pkthdr.aux = NULL; /* for IPsec */
        nip = mtod(m, struct ip *);
        bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
        nip->ip_len = m->m_len;
@@ -237,24 +294,25 @@ freeit:
        m_freem(n);
 }
 
-static struct sockaddr_in icmpsrc = { sizeof (struct sockaddr_in), AF_INET };
-static struct sockaddr_in icmpdst = { sizeof (struct sockaddr_in), AF_INET };
-static struct sockaddr_in icmpgw = { sizeof (struct sockaddr_in), AF_INET };
+static struct sockaddr_in icmpsrc = { sizeof (struct sockaddr_in), AF_INET, 
+                                                                               0 , { 0 }, { 0,0,0,0,0,0,0,0 } };
+static struct sockaddr_in icmpdst = { sizeof (struct sockaddr_in), AF_INET, 
+                                                                               0 , { 0 }, { 0,0,0,0,0,0,0,0 } };
+static struct sockaddr_in icmpgw = { sizeof (struct sockaddr_in), AF_INET, 
+                                                                               0 , { 0 }, { 0,0,0,0,0,0,0,0 } };
 
 /*
  * Process a received ICMP message.
  */
 void
-icmp_input(m, hlen)
-       register struct mbuf *m;
-       int hlen;
+icmp_input(struct mbuf *m, int hlen)
 {
-       register struct icmp *icp;
-       register struct ip *ip = mtod(m, struct ip *);
+       struct icmp *icp;
+       struct ip *ip = mtod(m, struct ip *);
        int icmplen = ip->ip_len;
-       register int i;
+       int i;
        struct in_ifaddr *ia;
-       void (*ctlfunc) __P((int, struct sockaddr *, void *));
+       void (*ctlfunc)(int, struct sockaddr *, void *);
        int code;
 
        /*
@@ -263,10 +321,13 @@ icmp_input(m, hlen)
         */
 #if ICMPPRINTFS
        if (icmpprintfs) {
-               char buf[4 * sizeof "123"];
-               strcpy(buf, inet_ntoa(ip->ip_src));
+               char buf[MAX_IPv4_STR_LEN];
+               char ipv4str[MAX_IPv4_STR_LEN];
+
                printf("icmp_input from %s to %s, len %d\n",
-                      buf, inet_ntoa(ip->ip_dst), icmplen);
+                      inet_ntop(AF_INET, &ip->ip_src, buf, sizeof(buf)),
+                      inet_ntop(AF_INET, &ip->ip_dst, ipv4str, sizeof(ipv4str)),
+                      icmplen);
        }
 #endif
        if (icmplen < ICMP_MINLEN) {
@@ -310,15 +371,6 @@ icmp_input(m, hlen)
                    icp->icmp_code);
 #endif
 
-#if IPSEC
-       /* drop it if it does not match the policy */
-       /* XXX Is there meaning of check in here ? */
-       if (ipsec4_in_reject(m, NULL)) {
-               ipsecstat.in_polvio++;
-               goto freeit;
-       }
-#endif
-
        /*
         * Message type specific processing.
         */
@@ -332,33 +384,34 @@ icmp_input(m, hlen)
                switch (code) {
                        case ICMP_UNREACH_NET:
                        case ICMP_UNREACH_HOST:
-                       case ICMP_UNREACH_PROTOCOL:
-                       case ICMP_UNREACH_PORT:
                        case ICMP_UNREACH_SRCFAIL:
-                               code += PRC_UNREACH_NET;
+                       case ICMP_UNREACH_NET_UNKNOWN:
+                       case ICMP_UNREACH_HOST_UNKNOWN:
+                       case ICMP_UNREACH_ISOLATED:
+                       case ICMP_UNREACH_TOSNET:
+                       case ICMP_UNREACH_TOSHOST:
+                       case ICMP_UNREACH_HOST_PRECEDENCE:
+                       case ICMP_UNREACH_PRECEDENCE_CUTOFF:
+                               code = PRC_UNREACH_NET;
                                break;
 
                        case ICMP_UNREACH_NEEDFRAG:
                                code = PRC_MSGSIZE;
                                break;
 
-                       case ICMP_UNREACH_NET_UNKNOWN:
-                       case ICMP_UNREACH_NET_PROHIB:
-                       case ICMP_UNREACH_TOSNET:
-                               code = PRC_UNREACH_NET;
+                       /*
+                        * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
+                        * Treat subcodes 2,3 as immediate RST
+                        */
+                       case ICMP_UNREACH_PROTOCOL:
+                       case ICMP_UNREACH_PORT:
+                               code = PRC_UNREACH_PORT;
                                break;
 
-                       case ICMP_UNREACH_HOST_UNKNOWN:
-                       case ICMP_UNREACH_ISOLATED:
+                       case ICMP_UNREACH_NET_PROHIB:
                        case ICMP_UNREACH_HOST_PROHIB:
-                       case ICMP_UNREACH_TOSHOST:
-                               code = PRC_UNREACH_HOST;
-                               break;
-
                        case ICMP_UNREACH_FILTER_PROHIB:
-                       case ICMP_UNREACH_HOST_PRECEDENCE:
-                       case ICMP_UNREACH_PRECEDENCE_CUTOFF:
-                               code = PRC_UNREACH_PORT;
+                               code = PRC_UNREACH_ADMIN_PROHIB;
                                break;
 
                        default:
@@ -410,7 +463,7 @@ icmp_input(m, hlen)
                 * notice that the MTU has changed and adapt accordingly.
                 * If no new MTU was suggested, then we guess a new one
                 * less than the current value.  If the new MTU is 
-                * unreasonably small (arbitrarily set at 296), then
+                * unreasonably small (defined by sysctl tcp_minmss), then
                 * we reset the MTU to the interface value and enable the
                 * lock bit, indicating that we are no longer doing MTU
                 * discovery.
@@ -429,9 +482,11 @@ icmp_input(m, hlen)
                                                          1);
 #if DEBUG_MTUDISC
                                printf("MTU for %s reduced to %d\n",
-                                       inet_ntoa(icmpsrc.sin_addr), mtu);
+                                          inet_ntop(AF_INET, &icmpsrc.sin_addr, ipv4str,
+                                                                sizeof(ipv4str)),
+                                          mtu);
 #endif
-                               if (mtu < 296) {
+                               if (mtu < max(296, (tcp_minmss + sizeof(struct tcpiphdr)))) {
                                        /* rt->rt_rmx.rmx_mtu =
                                                rt->rt_ifp->if_mtu; */
                                        rt->rt_rmx.rmx_locks |= RTV_MTU;
@@ -440,7 +495,7 @@ icmp_input(m, hlen)
                                }
                        }
                        if (rt)
-                               RTFREE(rt);
+                               rtfree(rt);
                }
 
 #endif
@@ -465,9 +520,18 @@ icmp_input(m, hlen)
                        break;
                }
                icp->icmp_type = ICMP_ECHOREPLY;
-               goto reflect;
+#if ICMP_BANDLIM
+               if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
+                       goto freeit;
+               else
+#endif
+                       goto reflect;
 
        case ICMP_TSTAMP:
+
+               if (icmptimestamp == 0)
+                       break;
+
                if (!icmpbmcastecho
                    && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
                        icmpstat.icps_bmcasttstamp++;
@@ -480,7 +544,12 @@ icmp_input(m, hlen)
                icp->icmp_type = ICMP_TSTAMPREPLY;
                icp->icmp_rtime = iptime();
                icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
-               goto reflect;
+#if ICMP_BANDLIM
+               if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
+                       goto freeit;
+               else
+#endif
+                       goto reflect;
 
        case ICMP_MASKREQ:
 #define        satosin(sa)     ((struct sockaddr_in *)(sa))
@@ -506,8 +575,11 @@ icmp_input(m, hlen)
                            (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
                if (ia == 0)
                        break;
-               if (ia->ia_ifp == 0)
+               if (ia->ia_ifp == 0) {
+                       ifafree(&ia->ia_ifa);
+                       ia = NULL;
                        break;
+               }
                icp->icmp_type = ICMP_MASKREPLY;
                icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
                if (ip->ip_src.s_addr == 0) {
@@ -516,6 +588,7 @@ icmp_input(m, hlen)
                        else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
                            ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
                }
+               ifafree(&ia->ia_ifa);
 reflect:
                ip->ip_len += hlen;     /* since ip_input deducts this */
                icmpstat.icps_reflect++;
@@ -524,6 +597,23 @@ reflect:
                return;
 
        case ICMP_REDIRECT:
+               if (log_redirect) {
+                       u_long src, dst, gw;
+
+                       src = ntohl(ip->ip_src.s_addr);
+                       dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
+                       gw = ntohl(icp->icmp_gwaddr.s_addr);
+                       printf("icmp redirect from %d.%d.%d.%d: "
+                              "%d.%d.%d.%d => %d.%d.%d.%d\n",
+                              (int)(src >> 24), (int)((src >> 16) & 0xff),
+                              (int)((src >> 8) & 0xff), (int)(src & 0xff),
+                              (int)(dst >> 24), (int)((dst >> 16) & 0xff),
+                              (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
+                              (int)(gw >> 24), (int)((gw >> 16) & 0xff),
+                              (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
+               }
+               if (drop_redirect)
+                       break;
                if (code > 3)
                        goto badcode;
                if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
@@ -542,11 +632,12 @@ reflect:
                icmpdst.sin_addr = icp->icmp_gwaddr;
 #if    ICMPPRINTFS
                if (icmpprintfs) {
-                       char buf[4 * sizeof "123"];
-                       strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
+                       char buf[MAX_IPv4_STR_LEN];
 
                        printf("redirect dst %s to %s\n",
-                              buf, inet_ntoa(icp->icmp_gwaddr));
+                              inet_ntop(AF_INET, &icp->icmp_ip.ip_dst, buf, sizeof(buf)),
+                              inet_ntop(AF_INET, &icp->icmp_gwaddr, ipv4str,
+                                                sizeof(ipv4str)));
                }
 #endif
                icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
@@ -586,13 +677,12 @@ freeit:
  * Reflect the ip packet back to the source
  */
 static void
-icmp_reflect(m)
-       struct mbuf *m;
+icmp_reflect(struct mbuf *m)
 {
-       register struct ip *ip = mtod(m, struct ip *);
-       register struct in_ifaddr *ia;
+       struct ip *ip = mtod(m, struct ip *);
+       struct in_ifaddr *ia;
        struct in_addr t;
-       struct mbuf *opts = 0;
+       struct mbuf *opts = NULL;
        int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
 
        if (!in_canforward(ip->ip_src) &&
@@ -609,6 +699,7 @@ icmp_reflect(m)
         * or anonymous), use the address which corresponds
         * to the incoming interface.
         */
+       lck_mtx_lock(rt_mtx);
        for (ia = in_ifaddrhead.tqh_first; ia; ia = ia->ia_link.tqe_next) {
                if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
                        break;
@@ -616,6 +707,8 @@ icmp_reflect(m)
                    t.s_addr == satosin(&ia->ia_broadaddr)->sin_addr.s_addr)
                        break;
        }
+       if (ia)
+               ifaref(&ia->ia_ifa);
        icmpdst.sin_addr = t;
        if ((ia == (struct in_ifaddr *)0) && m->m_pkthdr.rcvif)
                ia = (struct in_ifaddr *)ifaof_ifpforaddr(
@@ -624,14 +717,27 @@ icmp_reflect(m)
         * The following happens if the packet was not addressed to us,
         * and was received on an interface with no IP address.
         */
-       if (ia == (struct in_ifaddr *)0)
+       if (ia == (struct in_ifaddr *)0) {
                ia = in_ifaddrhead.tqh_first;
+               if (ia == (struct in_ifaddr *)0) {/* no address yet, bail out */
+                       m_freem(m);
+                       lck_mtx_unlock(rt_mtx);
+                       goto done;
+               }
+               ifaref(&ia->ia_ifa);
+       }
+       lck_mtx_unlock(rt_mtx);
+#if CONFIG_MACF_NET
+       mac_netinet_icmp_reply(m);
+#endif
        t = IA_SIN(ia)->sin_addr;
        ip->ip_src = t;
-       ip->ip_ttl = MAXTTL;
+       ip->ip_ttl = ip_defttl;
+       ifafree(&ia->ia_ifa);
+       ia = NULL;
 
        if (optlen > 0) {
-               register u_char *cp;
+               u_char *cp;
                int opt, cnt;
                u_int len;
 
@@ -641,7 +747,7 @@ icmp_reflect(m)
                 */
                cp = (u_char *) (ip + 1);
                if ((opts = ip_srcroute()) == 0 &&
-                   (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
+                   (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) { /* MAC-OK */
                        opts->m_len = sizeof(struct in_addr);
                        mtod(opts, struct in_addr *)->s_addr = 0;
                }
@@ -662,7 +768,7 @@ icmp_reflect(m)
                                            break;
                                    len = cp[IPOPT_OLEN];
                                    if (len < IPOPT_OLEN + sizeof(*cp) ||
-                                       len > cnt)
+                                       len > cnt)
                                            break;
                            }
                            /*
@@ -714,13 +820,11 @@ done:
  * after supplying a checksum.
  */
 static void
-icmp_send(m, opts)
-       register struct mbuf *m;
-       struct mbuf *opts;
+icmp_send(struct mbuf *m, struct mbuf *opts)
 {
-       register struct ip *ip = mtod(m, struct ip *);
-       register int hlen;
-       register struct icmp *icp;
+       struct ip *ip = mtod(m, struct ip *);
+       int hlen;
+       struct icmp *icp;
        struct route ro;
 
        hlen = IP_VHL_HL(ip->ip_vhl) << 2;
@@ -731,30 +835,27 @@ icmp_send(m, opts)
        icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
        m->m_data -= hlen;
        m->m_len += hlen;
-       m->m_pkthdr.rcvif = (struct ifnet *)0;
-       m->m_pkthdr.aux = NULL;
+       m->m_pkthdr.rcvif = NULL;
        m->m_pkthdr.csum_data = 0;
        m->m_pkthdr.csum_flags = 0;
 #if ICMPPRINTFS
        if (icmpprintfs) {
-               char buf[4 * sizeof "123"];
-               strcpy(buf, inet_ntoa(ip->ip_dst));
+               char buf[MAX_IPv4_STR_LEN];
+               char ipv4str[MAX_IPv4_STR_LEN];
+
                printf("icmp_send dst %s src %s\n",
-                      buf, inet_ntoa(ip->ip_src));
+                      inet_ntop(AF_INET, &ip->ip_dst, buf, sizeof(buf)),
+                      inet_ntop(AF_INET, &ip->ip_src, ipv4str, sizeof(ipv4str)));
        }
 #endif
        bzero(&ro, sizeof ro);
-
-#ifdef IPSEC
-       ipsec_setsocket(m, NULL);
-#endif /*IPSEC*/
-       (void) ip_output(m, opts, &ro, 0, NULL);
+       (void) ip_output(m, opts, &ro, 0, NULL, NULL);
        if (ro.ro_rt)
-               RTFREE(ro.ro_rt);
+               rtfree(ro.ro_rt);
 }
 
 n_time
-iptime()
+iptime(void)
 {
        struct timeval atv;
        u_long t;
@@ -770,10 +871,8 @@ iptime()
  * given current value MTU.  If DIR is less than zero, a larger plateau
  * is returned; otherwise, a smaller value is returned.
  */
-/* static */ int
-ip_next_mtu(mtu, dir)
-       int mtu;
-       int dir;
+static int
+ip_next_mtu(int mtu, int dir)
 {
        static int mtutab[] = {
                65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
@@ -828,31 +927,45 @@ ip_next_mtu(mtu, dir)
 int
 badport_bandlim(int which)
 {
-       static int lticks[2];
-       static int lpackets[2];
-       int dticks;
+       static struct timeval lticks[BANDLIM_MAX + 1];
+       static int lpackets[BANDLIM_MAX + 1];
+       struct timeval time;
+       int secs;
+
+       const char *bandlimittype[] = {
+               "Limiting icmp unreach response",
+               "Limiting icmp ping response",
+               "Limiting icmp tstamp response",
+               "Limiting closed port RST response",
+               "Limiting open port RST response"
+               };
 
        /*
         * Return ok status if feature disabled or argument out of
         * ranage.
         */
 
-       if (icmplim <= 0 || which >= 2 || which < 0)
+       if (icmplim <= 0 || which > BANDLIM_MAX || which < 0)
                return(0);
-       dticks = ticks - lticks[which];
 
+       getmicrouptime(&time);
+
+       secs = time.tv_sec - lticks[which].tv_sec ;
+                       
        /*
-        * reset stats when cumulative dt exceeds one second.
+        * reset stats when cumulative delta exceeds one second.
         */
 
-       if ((unsigned int)dticks > hz) {
+       if ((secs > 1) || (secs == 1 && (lticks[which].tv_usec > time.tv_usec))) { 
                if (lpackets[which] > icmplim) {
-                       printf("icmp-response bandwidth limit %d/%d pps\n",
+                       printf("%s from %d to %d packets per second\n",
+                               bandlimittype[which],
                                lpackets[which],
                                icmplim
                        );
                }
-               lticks[which] = ticks;
+               lticks[which].tv_sec = time.tv_sec;
+               lticks[which].tv_usec = time.tv_usec;
                lpackets[which] = 0;
        }
 
@@ -868,4 +981,199 @@ badport_bandlim(int which)
 
 #endif
 
+#if __APPLE__
+
+/*
+ * Non-privileged ICMP socket operations
+ * - send ICMP echo request
+ * - all ICMP
+ * - limited socket options
+ */
+
+#include <netinet/ip_icmp.h>
+#include <netinet/in_pcb.h>
+
+extern struct domain inetdomain;
+extern u_long rip_sendspace;
+extern u_long rip_recvspace;
+extern struct inpcbinfo ripcbinfo;
+
+int rip_abort(struct socket *);
+int rip_bind(struct socket *, struct sockaddr *, struct proc *);
+int rip_connect(struct socket *, struct sockaddr *, struct proc *);
+int rip_detach(struct socket *);
+int rip_disconnect(struct socket *);
+int rip_shutdown(struct socket *);
+
+__private_extern__ int icmp_dgram_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct proc *p);
+__private_extern__ int icmp_dgram_attach(struct socket *so, int proto, struct proc *p);
+__private_extern__ int icmp_dgram_ctloutput(struct socket *so, struct sockopt *sopt);
+
+__private_extern__ struct pr_usrreqs icmp_dgram_usrreqs = {
+        rip_abort, pru_accept_notsupp, icmp_dgram_attach, rip_bind, rip_connect,
+        pru_connect2_notsupp, in_control, rip_detach, rip_disconnect,
+        pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp,
+        pru_rcvoob_notsupp, icmp_dgram_send, pru_sense_null, rip_shutdown,
+        in_setsockaddr, sosend, soreceive, pru_sopoll_notsupp
+};
+
+/* Like rip_attach but without root privilege enforcement */
+__private_extern__ int
+icmp_dgram_attach(struct socket *so, __unused int proto, struct proc *p)
+{
+        struct inpcb *inp;
+        int error;
+
+        inp = sotoinpcb(so);
+        if (inp)
+                panic("icmp_dgram_attach");
+
+        error = soreserve(so, rip_sendspace, rip_recvspace);
+        if (error)
+                return error;
+        error = in_pcballoc(so, &ripcbinfo, p);
+        if (error)
+                return error;
+        inp = (struct inpcb *)so->so_pcb;       
+        inp->inp_vflag |= INP_IPV4;
+        inp->inp_ip_p = IPPROTO_ICMP;
+        inp->inp_ip_ttl = ip_defttl;
+        return 0;
+}
+
+/*
+ * Raw IP socket option processing.
+ */
+__private_extern__ int
+icmp_dgram_ctloutput(struct socket *so, struct sockopt *sopt)
+{
+       int     error;
+
+       if (sopt->sopt_level != IPPROTO_IP)
+               return (EINVAL);
+
+       switch (sopt->sopt_name) {
+               case IP_OPTIONS:
+               case IP_HDRINCL:
+               case IP_TOS:
+               case IP_TTL:
+               case IP_RECVOPTS:
+               case IP_RECVRETOPTS:
+               case IP_RECVDSTADDR:
+               case IP_RETOPTS:
+               case IP_MULTICAST_IF:
+               case IP_MULTICAST_TTL:
+               case IP_MULTICAST_LOOP:
+               case IP_ADD_MEMBERSHIP:
+               case IP_DROP_MEMBERSHIP:
+               case IP_MULTICAST_VIF:
+               case IP_PORTRANGE:
+               case IP_RECVIF:
+               case IP_IPSEC_POLICY:
+#if defined(NFAITH) && NFAITH > 0
+               case IP_FAITH:
+#endif
+               case IP_STRIPHDR:
+               case IP_RECVTTL:
+                       error = rip_ctloutput(so, sopt);
+                       break;
+               
+               default:
+                       error = EINVAL;
+                       break;
+       }
+
+       return (error);
+}
+
+__private_extern__ int
+icmp_dgram_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
+         struct mbuf *control, struct proc *p)
+{
+       struct ip *ip;
+       struct inpcb *inp = sotoinpcb(so);
+       int hlen;
+       struct icmp *icp;
+        struct in_ifaddr *ia = NULL;
+       int icmplen;
+
+       if ((inp->inp_flags & INP_HDRINCL) != 0) {
+               /*
+                * This is not raw IP, we liberal only for fields TOS, id and TTL 
+                */
+               ip = mtod(m, struct ip *);
+
+               hlen = IP_VHL_HL(ip->ip_vhl) << 2;
+               /* Some sanity checks */
+               if (m->m_pkthdr.len < hlen + ICMP_MINLEN) {
+                       goto bad;
+               }
+               /* Only IPv4 */
+               if (IP_VHL_V(ip->ip_vhl) != 4)
+                       goto bad;
+               if (hlen < 20 || hlen > 40 || ip->ip_len != m->m_pkthdr.len)
+                       goto bad;
+               /* Bogus fragments can tie up peer resources */ 
+               if (ip->ip_off != 0)
+                       goto bad;
+               /* Allow only ICMP even for user provided IP header */
+               if (ip->ip_p != IPPROTO_ICMP)
+                       goto bad;
+               /* To prevent spoofing, specified source address must be one of ours */
+               if (ip->ip_src.s_addr != INADDR_ANY) {
+                       socket_unlock(so, 0);
+                       lck_mtx_lock(rt_mtx);
+                       if (TAILQ_EMPTY(&in_ifaddrhead)) {
+                               lck_mtx_unlock(rt_mtx);
+                               socket_lock(so, 0);
+                               goto bad;
+                       }
+                       TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
+                               if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_src.s_addr) {
+                                       lck_mtx_unlock(rt_mtx);
+                                       socket_lock(so, 0);
+                                       goto ours;
+                               }
+                       }
+                       lck_mtx_unlock(rt_mtx);
+                       socket_lock(so, 0);
+                       goto bad;
+               }
+ours:
+               /* Do not trust we got a valid checksum */
+               ip->ip_sum = 0;
+               
+               icp = (struct icmp *)(((char *)m->m_data) + hlen);
+               icmplen = m->m_pkthdr.len - hlen;
+       } else {
+               if ((icmplen = m->m_pkthdr.len) < ICMP_MINLEN) {
+                       goto bad;
+               }
+               icp = mtod(m, struct icmp *);
+       }
+       /*
+        * Allow only to send request types with code 0
+        */
+       if (icp->icmp_code != 0)
+               goto bad;
+       switch (icp->icmp_type) {
+               case ICMP_ECHO:
+                       break;
+               case ICMP_TSTAMP:
+                       if (icmplen != 20)
+                               goto bad;
+                       break;
+               case ICMP_MASKREQ:
+                       if (icmplen != 12)
+                               goto bad;
+                       break;
+               default:
+                       goto bad;
+       }
+       return rip_send(so, flags, m, nam, control, p);
+bad:
+       m_freem(m);
+       return EINVAL;
+}
 
+#endif /* __APPLE__ */