]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/netinet6/ip6_forward.c
xnu-2050.48.11.tar.gz
[apple/xnu.git] / bsd / netinet6 / ip6_forward.c
index 9f6f70b5fe444a8bf4ac5098b4c63c26fa90fd6f..42156858ef82deff0a40b3cdfa33c546fccaac1e 100644 (file)
@@ -1,3 +1,31 @@
+/*
+ * Copyright (c) 2009-2011 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/ip6_forward.c,v 1.16 2002/10/16 02:25:05 sam Exp $   */
 /*     $KAME: ip6_forward.c,v 1.69 2001/05/17 03:48:30 itojun Exp $    */
 
@@ -56,6 +84,7 @@
 #include <netinet6/ip6_var.h>
 #include <netinet/icmp6.h>
 #include <netinet6/nd6.h>
+#include <netinet6/scope6_var.h>
 
 #include <netinet/in_pcb.h>
 
 #include <netkey/key.h>
 extern int ipsec_bypass;
 #endif /* IPSEC */
-extern lck_mtx_t *ip6_mutex;
 
 #include <netinet6/ip6_fw.h>
 
 #include <net/net_osdep.h>
 
+#if PF
+#include <net/pfvar.h>
+#endif /* PF */
+
 /*
  * Forward a packet.  If some error occurs return the sender
  * an icmp packet.  Note we can't always generate a meaningful
@@ -86,24 +118,50 @@ extern lck_mtx_t *ip6_mutex;
  *
  */
 
-void
+struct mbuf *
 ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
-    int srcrt, int locked)
+    int srcrt)
 {
        struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
        struct sockaddr_in6 *dst;
        struct rtentry *rt;
        int error, type = 0, code = 0;
+       boolean_t proxy = FALSE;
        struct mbuf *mcopy = NULL;
-       struct ifnet *origifp;  /* maybe unnecessary */
+       struct ifnet *ifp, *origifp;    /* maybe unnecessary */
+       u_int32_t inzone, outzone;
+       struct in6_addr src_in6, dst_in6;
 #if IPSEC
        struct secpolicy *sp = NULL;
 #endif
        struct timeval timenow;
-       int     tunneledv4 = 0;
+       unsigned int ifscope = IFSCOPE_NONE;
+#if PF
+       struct pf_mtag *pf_mtag;
+#endif /* PF */
 
        getmicrotime(&timenow);
+#if PF
+       pf_mtag = pf_find_mtag(m);
+       if (pf_mtag != NULL && pf_mtag->pftag_rtableid != IFSCOPE_NONE)
+               ifscope = pf_mtag->pftag_rtableid;
 
+       /*
+        * If the caller provides a route which is on a different interface
+        * than the one specified for scoped forwarding, discard the route
+        * and do a lookup below.
+        */
+       if (ifscope != IFSCOPE_NONE && (rt = ip6forward_rt->ro_rt) != NULL) {
+               RT_LOCK(rt);
+               if (rt->rt_ifp->if_index != ifscope) {
+                       RT_UNLOCK(rt);
+                       rtfree(rt);
+                       rt = ip6forward_rt->ro_rt = NULL;
+               } else {
+                       RT_UNLOCK(rt);
+               }
+       }
+#endif /* PF */
 
 #if IPSEC
        /*
@@ -117,7 +175,7 @@ ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
                if (ipsec6_in_reject(m, NULL)) {
                        IPSEC_STAT_INCREMENT(ipsec6stat.in_polvio);
                        m_freem(m);
-                       return;
+                       return (NULL);
                }
        }
 #endif /*IPSEC*/
@@ -144,19 +202,34 @@ ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
                            if_name(m->m_pkthdr.rcvif));
                }
                m_freem(m);
-               return;
+               return (NULL);
        }
 
        if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
                /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
-               if (locked)
-                       lck_mtx_unlock(ip6_mutex);
                icmp6_error(m, ICMP6_TIME_EXCEEDED,
                                ICMP6_TIME_EXCEED_TRANSIT, 0);
-               if (locked)
-                       lck_mtx_lock(ip6_mutex);
-               return;
+               return (NULL);
+       }
+
+       /*
+        * See if the destination is a proxied address, and if so pretend
+        * that it's for us.  This is mostly to handle NUD probes against
+        * the proxied addresses.  We filter for ICMPv6 here and will let
+        * icmp6_input handle the rest.
+        */
+       if (!srcrt && nd6_prproxy) {
+               VERIFY(!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst));
+               proxy = nd6_prproxy_isours(m, ip6, ip6forward_rt, ifscope);
+               /*
+                * Don't update hop limit while proxying; RFC 4389 4.1.
+                * Also skip IPsec forwarding path processing as this
+                * packet is not to be forwarded.
+                */
+               if (proxy)
+                       goto skip_ipsec;
        }
+
        ip6->ip6_hlim -= IPV6_HLIMDEC;
 
        /*
@@ -187,7 +260,7 @@ ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
 #endif
                }
                m_freem(m);
-               return;
+               return (NULL);
        }
 
        error = 0;
@@ -210,7 +283,7 @@ ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
 #endif
                }
                m_freem(m);
-               return;
+               return (NULL);
 
        case IPSEC_POLICY_BYPASS:
        case IPSEC_POLICY_NONE:
@@ -232,7 +305,7 @@ ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
 #endif
                        }
                        m_freem(m);
-                       return;
+                       return (NULL);
                }
                /* do IPsec */
                break;
@@ -258,19 +331,18 @@ ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
         */
        bzero(&state, sizeof(state));
        state.m = m;
-       state.ro = NULL;        /* update at ipsec6_output_tunnel() */
        state.dst = NULL;       /* update at ipsec6_output_tunnel() */
 
-       if (locked)
-               lck_mtx_unlock(ip6_mutex);
-       error = ipsec6_output_tunnel(&state, sp, 0, &tunneledv4);
-       if (locked)
-               lck_mtx_lock(ip6_mutex);
+       error = ipsec6_output_tunnel(&state, sp, 0);
        key_freesp(sp, KEY_SADB_UNLOCKED);
-       if (tunneledv4)
-               return;  /* packet is gone - sent over IPv4 */
+       if (state.tunneled == 4)
+               return (NULL);  /* packet is gone - sent over IPv4 */
                
        m = state.m;
+       if (state.ro.ro_rt) {
+               rtfree(state.ro.ro_rt);
+               state.ro.ro_rt = NULL;
+       }
        if (error) {
                /* mbuf is already reclaimed in ipsec6_output_tunnel. */
                switch (error) {
@@ -296,80 +368,109 @@ ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
 #endif
                }
                m_freem(m);
-               return;
+               return (NULL);
        }
     }
     skip_ipsec:
 #endif /* IPSEC */
 
        dst = (struct sockaddr_in6 *)&ip6forward_rt->ro_dst;
+       if ((rt = ip6forward_rt->ro_rt) != NULL) {
+               RT_LOCK(rt);
+               /* Take an extra ref for ourselves */
+               RT_ADDREF_LOCKED(rt);
+       }
+
        if (!srcrt) {
                /*
                 * ip6forward_rt->ro_dst.sin6_addr is equal to ip6->ip6_dst
                 */
-               if (ip6forward_rt->ro_rt == 0 ||
-                   (ip6forward_rt->ro_rt->rt_flags & RTF_UP) == 0) {
-                       if (ip6forward_rt->ro_rt) {
-                               rtfree(ip6forward_rt->ro_rt);
-                               ip6forward_rt->ro_rt = 0;
+               if (rt == NULL || !(rt->rt_flags & RTF_UP) ||
+                   rt->generation_id != route_generation) {
+                       if (rt != NULL) {
+                               /* Release extra ref */
+                               RT_REMREF_LOCKED(rt);
+                               RT_UNLOCK(rt);
+                               rtfree(rt);
+                               ip6forward_rt->ro_rt = NULL;
                        }
                        /* this probably fails but give it a try again */
-                       rtalloc_ign((struct route *)ip6forward_rt,
-                                   RTF_PRCLONING);
+                       rtalloc_scoped_ign((struct route *)ip6forward_rt,
+                           RTF_PRCLONING, ifscope);
+                       if ((rt = ip6forward_rt->ro_rt) != NULL) {
+                               RT_LOCK(rt);
+                               /* Take an extra ref for ourselves */
+                               RT_ADDREF_LOCKED(rt);
+                       }
                }
 
-               if (ip6forward_rt->ro_rt == 0) {
+               if (rt == NULL) {
                        ip6stat.ip6s_noroute++;
                        in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
-                       if (mcopy) {
-                               if (locked)
-                                       lck_mtx_unlock(ip6_mutex);
+                       if (mcopy)
                                icmp6_error(mcopy, ICMP6_DST_UNREACH,
                                            ICMP6_DST_UNREACH_NOROUTE, 0);
-                               if (locked)
-                                       lck_mtx_lock(ip6_mutex);
-                       }
                        m_freem(m);
-                       return;
+                       return (NULL);
                }
-       } else if ((rt = ip6forward_rt->ro_rt) == 0 ||
-                !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) {
-               if (ip6forward_rt->ro_rt) {
-                       rtfree(ip6forward_rt->ro_rt);
-                       ip6forward_rt->ro_rt = 0;
+               RT_LOCK_ASSERT_HELD(rt);
+       } else if (rt == NULL || !(rt->rt_flags & RTF_UP) ||
+           !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr) ||
+           rt->generation_id != route_generation) {
+               if (rt != NULL) {
+                       /* Release extra ref */
+                       RT_REMREF_LOCKED(rt);
+                       RT_UNLOCK(rt);
+                       rtfree(rt);
+                       ip6forward_rt->ro_rt = NULL;
                }
                bzero(dst, sizeof(*dst));
                dst->sin6_len = sizeof(struct sockaddr_in6);
                dst->sin6_family = AF_INET6;
                dst->sin6_addr = ip6->ip6_dst;
 
-               rtalloc_ign((struct route *)ip6forward_rt, RTF_PRCLONING);
-               if (ip6forward_rt->ro_rt == 0) {
+               rtalloc_scoped_ign((struct route *)ip6forward_rt,
+                   RTF_PRCLONING, ifscope);
+               if ((rt = ip6forward_rt->ro_rt) == NULL) {
                        ip6stat.ip6s_noroute++;
                        in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
-                       if (mcopy) {
-                               if (locked)
-                                       lck_mtx_unlock(ip6_mutex);
+                       if (mcopy)
                                icmp6_error(mcopy, ICMP6_DST_UNREACH,
-                                           ICMP6_DST_UNREACH_NOROUTE, 0);
-                               if (locked)
-                                       lck_mtx_lock(ip6_mutex);
-                       }
+                                   ICMP6_DST_UNREACH_NOROUTE, 0);
                        m_freem(m);
-                       return;
+                       return (NULL);
                }
+               RT_LOCK(rt);
+               /* Take an extra ref for ourselves */
+               RT_ADDREF_LOCKED(rt);
        }
-       rt = ip6forward_rt->ro_rt;
 
        /*
-        * Scope check: if a packet can't be delivered to its destination
-        * for the reason that the destination is beyond the scope of the
-        * source address, discard the packet and return an icmp6 destination
-        * unreachable error with Code 2 (beyond scope of source address).
-        * [draft-ietf-ipngwg-icmp-v3-02.txt, Section 3.1]
+        * Source scope check: if a packet can't be delivered to its
+        * destination for the reason that the destination is beyond the scope
+        * of the source address, discard the packet and return an icmp6
+        * destination unreachable error with Code 2 (beyond scope of source
+        * address) unless we are proxying (source address is link local
+        * for NUDs.)  We use a local copy of ip6_src, since in6_setscope()
+        * will possibly modify its first argument.
+        * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1]
         */
-       if (in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_src) !=
-           in6_addr2scopeid(rt->rt_ifp, &ip6->ip6_src)) {
+       src_in6 = ip6->ip6_src;
+       if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) {
+               /* XXX: this should not happen */
+               ip6stat.ip6s_cantforward++;
+               ip6stat.ip6s_badscope++;
+               m_freem(m);
+               return (NULL);
+       }
+       if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) {
+               ip6stat.ip6s_cantforward++;
+               ip6stat.ip6s_badscope++;
+               m_freem(m);
+               return (NULL);
+       }
+
+       if (inzone != outzone && !proxy) {
                ip6stat.ip6s_cantforward++;
                ip6stat.ip6s_badscope++;
                in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
@@ -384,22 +485,38 @@ ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
                            ip6->ip6_nxt,
                            if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
                }
+               /* Release extra ref */
+               RT_REMREF_LOCKED(rt);
+               RT_UNLOCK(rt);
                if (mcopy) {
-                       if (locked)
-                               lck_mtx_unlock(ip6_mutex);
                        icmp6_error(mcopy, ICMP6_DST_UNREACH,
                                    ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
-                       if (locked)
-                               lck_mtx_lock(ip6_mutex);
                }
                m_freem(m);
-               return;
+               return (NULL);
+       }
+
+       /*
+        * Destination scope check: if a packet is going to break the scope
+        * zone of packet's destination address, discard it.  This case should
+        * usually be prevented by appropriately-configured routing table, but
+        * we need an explicit check because we may mistakenly forward the
+        * packet to a different zone by (e.g.) a default route.
+        */
+       dst_in6 = ip6->ip6_dst;
+       if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
+           in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
+           inzone != outzone) {
+               ip6stat.ip6s_cantforward++;
+               ip6stat.ip6s_badscope++;
+               m_freem(m);
+               return (NULL);
        }
 
        if (m->m_pkthdr.len > rt->rt_ifp->if_mtu) {
                in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
                if (mcopy) {
-                       u_long mtu;
+                       uint32_t mtu;
 #if IPSEC
                        struct secpolicy *sp2;
                        int ipsecerror;
@@ -410,7 +527,7 @@ ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
 #if IPSEC
                        /*
                         * When we do IPsec tunnel ingress, we need to play
-                        * with if_mtu value (decrement IPsec header size
+                        * with the link value (decrement IPsec header size
                         * from mtu value).  The code is much simpler than v4
                         * case, as we have the outgoing interface for
                         * encapsulated packet as "rt->rt_ifp".
@@ -431,18 +548,21 @@ ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
                        if (mtu < IPV6_MMTU)
                                mtu = IPV6_MMTU;
 #endif
-                       if (locked)
-                               lck_mtx_unlock(ip6_mutex);
+                       /* Release extra ref */
+                       RT_REMREF_LOCKED(rt);
+                       RT_UNLOCK(rt);
                        icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
-                       if (locked)
-                               lck_mtx_lock(ip6_mutex);
+               } else {
+                       /* Release extra ref */
+                       RT_REMREF_LOCKED(rt);
+                       RT_UNLOCK(rt);
                }
                m_freem(m);
-               return;
+               return (NULL);
        }
 
        if (rt->rt_flags & RTF_GATEWAY)
-               dst = (struct sockaddr_in6 *)rt->rt_gateway;
+               dst = (struct sockaddr_in6 *)(void *)rt->rt_gateway;
 
        /*
         * If we are to forward the packet using the same interface
@@ -453,7 +573,8 @@ ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
         * Also, don't send redirect if forwarding using a route
         * modified by a redirect.
         */
-       if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
+       if (!proxy &&
+           ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
            (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
                if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) {
                        /*
@@ -466,31 +587,37 @@ ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
                         * type/code is based on suggestion by Rich Draves.
                         * not sure if it is the best pick.
                         */
-                       if (locked)
-                               lck_mtx_unlock(ip6_mutex);
+                       RT_REMREF_LOCKED(rt);   /* Release extra ref */
+                       RT_UNLOCK(rt);
                        icmp6_error(mcopy, ICMP6_DST_UNREACH,
                                    ICMP6_DST_UNREACH_ADDR, 0);
-                       if (locked)
-                               lck_mtx_lock(ip6_mutex);
                        m_freem(m);
-                       return;
+                       return (NULL);
                }
                type = ND_REDIRECT;
        }
 
+#if IPFW2
        /*
         * Check with the firewall...
         */
        if (ip6_fw_enable && ip6_fw_chk_ptr) {
                u_short port = 0;
+               ifp = rt->rt_ifp;
+               /* Drop the lock but retain the extra ref */
+               RT_UNLOCK(rt);
                /* If ipfw says divert, we have to just drop packet */
-               if (ip6_fw_chk_ptr(&ip6, rt->rt_ifp, &port, &m)) {
+               if (ip6_fw_chk_ptr(&ip6, ifp, &port, &m)) {
                        m_freem(m);
                        goto freecopy;
                }
-               if (!m)
+               if (!m) {
                        goto freecopy;
+               }
+               /* We still have the extra ref on rt */
+               RT_LOCK(rt);
        }
+#endif
 
        /*
         * Fake scoped addresses. Note that even link-local source or
@@ -529,38 +656,75 @@ ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
        }
        else
                origifp = rt->rt_ifp;
-#ifndef SCOPEDROUTING
        /*
         * clear embedded scope identifiers if necessary.
         * in6_clearscope will touch the addresses only when necessary.
         */
        in6_clearscope(&ip6->ip6_src);
        in6_clearscope(&ip6->ip6_dst);
-#endif
 
-       error = nd6_output(rt->rt_ifp, origifp, m, dst, rt, locked);
+       ifp = rt->rt_ifp;
+       /* Drop the lock but retain the extra ref */
+       RT_UNLOCK(rt);
+
+       /*
+        * If this is to be processed locally, let ip6_input have it.
+        */
+       if (proxy) {
+               VERIFY(m->m_pkthdr.aux_flags & MAUXF_PROXY_DST);
+               /* Release extra ref */
+               RT_REMREF(rt);
+               if (mcopy != NULL)
+                       m_freem(mcopy);
+               return (m);
+       }
+
+#if PF
+       /* Invoke outbound packet filter */
+       error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, NULL);
+
+       if (error != 0 || m == NULL) {
+               if (m != NULL) {
+                       panic("%s: unexpected packet %p\n", __func__, m);
+                       /* NOTREACHED */
+               }
+               /* Already freed by callee */
+               goto senderr;
+       }
+       ip6 = mtod(m, struct ip6_hdr *);
+#endif /* PF */
+
+       error = nd6_output(ifp, origifp, m, dst, rt, NULL);
        if (error) {
-               in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
+               in6_ifstat_inc(ifp, ifs6_out_discard);
                ip6stat.ip6s_cantforward++;
        } else {
                ip6stat.ip6s_forward++;
-               in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
+               in6_ifstat_inc(ifp, ifs6_out_forward);
                if (type)
                        ip6stat.ip6s_redirectsent++;
                else {
-                       if (mcopy)
+                       if (mcopy) {
                                goto freecopy;
+                       }
                }
        }
-       if (mcopy == NULL)
-               return;
-
+#if PF
+senderr:
+#endif /* PF */
+       if (mcopy == NULL) {
+               /* Release extra ref */
+               RT_REMREF(rt);
+               return (NULL);
+       }
        switch (error) {
        case 0:
 #if 1
                if (type == ND_REDIRECT) {
                        icmp6_redirect_output(mcopy, rt);
-                       return;
+                       /* Release extra ref */
+                       RT_REMREF(rt);
+                       return (NULL);
                }
 #endif
                goto freecopy;
@@ -582,14 +746,14 @@ ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
                code = ICMP6_DST_UNREACH_ADDR;
                break;
        }
-       if (locked)
-               lck_mtx_unlock(ip6_mutex);
        icmp6_error(mcopy, type, code, 0);
-       if (locked)
-               lck_mtx_lock(ip6_mutex);
-       return;
+       /* Release extra ref */
+       RT_REMREF(rt);
+       return (NULL);
 
  freecopy:
        m_freem(mcopy);
-       return;
+       /* Release extra ref */
+       RT_REMREF(rt);
+       return (NULL);
 }