/*
- * Copyright (c) 2008 Apple Inc. All rights reserved.
+ * Copyright (c) 2009-2016 Apple Inc. All rights reserved.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_START@
- *
+ *
* This file contains Original Code and/or Modifications of Original Code
* as defined in and that are subject to the Apple Public Source License
* Version 2.0 (the 'License'). You may not use this file except in
* unlawful or unlicensed copies of an Apple operating system, or to
* circumvent, violate, or enable the circumvention or violation of, any
* terms of an Apple operating system software license agreement.
- *
+ *
* Please obtain a copy of the License at
* http://www.opensource.apple.com/apsl/ and read it before using this file.
- *
+ *
* The Original Code and all software distributed under the License are
* distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
* EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
* 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@
*/
#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 DUMMYNET
+#include <netinet/ip_fw.h>
+#include <netinet/ip_dummynet.h>
+#endif /* DUMMYNET */
+
#if PF
#include <net/pfvar.h>
#endif /* PF */
*
*/
-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 *ifp, *origifp; /* maybe unnecessary */
+ struct ifnet *ifp, *rcvifp, *origifp; /* maybe unnecessary */
+ u_int32_t inzone, outzone, len;
+ struct in6_addr src_in6, dst_in6;
+ uint64_t curtime = net_uptime();
#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 */
+
+ /*
+ * In the prefix proxying case, the route to the proxied node normally
+ * gets created by nd6_prproxy_ns_output(), as part of forwarding a
+ * NS (NUD/AR) packet to the proxied node. In the event that such
+ * packet did not arrive in time before the correct route gets created,
+ * ip6_input() would have performed a rtalloc() which most likely will
+ * create the wrong cloned route; this route points back to the same
+ * interface as the inbound interface, since the parent non-scoped
+ * prefix route points there. Therefore we check if that is the case
+ * and perform the necessary fixup to get the correct route installed.
+ */
+ if (!srcrt && nd6_prproxy &&
+ (rt = ip6forward_rt->ro_rt) != NULL && (rt->rt_flags & RTF_PROXY)) {
+ nd6_proxy_find_fwdroute(m->m_pkthdr.rcvif, ip6forward_rt);
+ if ((rt = ip6forward_rt->ro_rt) != NULL)
+ ifscope = rt->rt_ifp->if_index;
+ }
- 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);
+ ROUTE_RELEASE(ip6forward_rt);
+ rt = NULL;
+ } else {
+ RT_UNLOCK(rt);
+ }
+ }
+#endif /* PF */
#if IPSEC
/*
if (ipsec6_in_reject(m, NULL)) {
IPSEC_STAT_INCREMENT(ipsec6stat.in_polvio);
m_freem(m);
- return;
+ return (NULL);
}
}
#endif /*IPSEC*/
/*
- * Do not forward packets to multicast destination (should be handled
- * by ip6_mforward().
+ * Do not forward packets to multicast destination.
* Do not forward packets with unspecified source. It was discussed
* in July 2000, on ipngwg mailing list.
*/
IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
ip6stat.ip6s_cantforward++;
/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
- if (ip6_log_time + ip6_log_interval < timenow.tv_sec) {
- ip6_log_time = timenow.tv_sec;
+ if (ip6_log_time + ip6_log_interval < curtime) {
+ ip6_log_time = curtime;
log(LOG_DEBUG,
"cannot forward "
"from %s to %s nxt %d received on %s\n",
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;
+ icmp6_error_flag(m, ICMP6_TIME_EXCEEDED,
+ ICMP6_TIME_EXCEED_TRANSIT, 0, 0);
+ 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;
/*
#endif
}
m_freem(m);
- return;
+ return (NULL);
}
error = 0;
#endif
}
m_freem(m);
- return;
+ return (NULL);
case IPSEC_POLICY_BYPASS:
case IPSEC_POLICY_NONE:
#endif
}
m_freem(m);
- return;
+ return (NULL);
}
/* do IPsec */
break;
*/
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) {
+ ROUTE_RELEASE(&state.ro);
+ return (NULL); /* packet is gone - sent over IPv4 */
+ }
+
m = state.m;
+ ROUTE_RELEASE(&state.ro);
+
if (error) {
/* mbuf is already reclaimed in ipsec6_output_tunnel. */
switch (error) {
#endif
}
m_freem(m);
- return;
+ return (NULL);
}
}
- skip_ipsec:
#endif /* IPSEC */
+ skip_ipsec:
- /*
- * If "locked", ip6forward_rt points to the globally defined
- * struct route cache which requires ip6_mutex, e.g. when this
- * is called from ip6_input(). Else the caller is responsible
- * for the struct route and its serialization (if needed), e.g.
- * when this is called from ip6_rthdr0().
- */
- if (locked)
- lck_mtx_assert(ip6_mutex, LCK_MTX_ASSERT_OWNED);
dst = (struct sockaddr_in6 *)&ip6forward_rt->ro_dst;
if ((rt = ip6forward_rt->ro_rt) != NULL) {
RT_LOCK(rt);
RT_ADDREF_LOCKED(rt);
}
+ VERIFY(rt == NULL || rt == ip6forward_rt->ro_rt);
if (!srcrt) {
/*
* ip6forward_rt->ro_dst.sin6_addr is equal to ip6->ip6_dst
*/
- if (rt == NULL || !(rt->rt_flags & RTF_UP) ||
- rt->generation_id != route_generation) {
+ if (ROUTE_UNUSABLE(ip6forward_rt)) {
if (rt != NULL) {
/* Release extra ref */
RT_REMREF_LOCKED(rt);
RT_UNLOCK(rt);
- rtfree(rt);
- ip6forward_rt->ro_rt = NULL;
}
+ ROUTE_RELEASE(ip6forward_rt);
+
/* 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 */
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);
}
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) {
+ } else if (ROUTE_UNUSABLE(ip6forward_rt) ||
+ !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) {
if (rt != NULL) {
/* Release extra ref */
RT_REMREF_LOCKED(rt);
RT_UNLOCK(rt);
- rtfree(rt);
- ip6forward_rt->ro_rt = NULL;
}
+ ROUTE_RELEASE(ip6forward_rt);
+
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);
+ 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);
- }
m_freem(m);
- return;
+ return (NULL);
}
RT_LOCK(rt);
/* Take an extra ref for ourselves */
}
/*
- * 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)) {
+ RT_REMREF_LOCKED(rt);
+ RT_UNLOCK(rt);
+ /* 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)) {
+ RT_REMREF_LOCKED(rt);
+ RT_UNLOCK(rt);
+ 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);
- if (ip6_log_time + ip6_log_interval < timenow.tv_sec) {
- ip6_log_time = timenow.tv_sec;
+ if (ip6_log_time + ip6_log_interval < curtime) {
+ ip6_log_time = curtime;
log(LOG_DEBUG,
"cannot forward "
"src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
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) {
+ RT_REMREF_LOCKED(rt);
+ RT_UNLOCK(rt);
+ ip6stat.ip6s_cantforward++;
+ ip6stat.ip6s_badscope++;
+ m_freem(m);
+ return (NULL);
}
if (m->m_pkthdr.len > rt->rt_ifp->if_mtu) {
#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".
/* Release extra ref */
RT_REMREF_LOCKED(rt);
RT_UNLOCK(rt);
- if (locked)
- lck_mtx_unlock(ip6_mutex);
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
* 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) {
/*
*/
RT_REMREF_LOCKED(rt); /* Release extra ref */
RT_UNLOCK(rt);
- if (locked)
- lck_mtx_unlock(ip6_mutex);
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;
}
-
- /*
- * 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, ifp, &port, &m)) {
- m_freem(m);
- goto freecopy;
- }
- if (!m) {
- goto freecopy;
- }
- /* We still have the extra ref on rt */
- RT_LOCK(rt);
- }
-
/*
* Fake scoped addresses. Note that even link-local source or
* destinaion can appear, if the originating node just sends the
* rthdr. (itojun)
*/
#if 1
- if (0)
+ if ((0))
#else
if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
#endif
}
/* we can just use rcvif in forwarding. */
- origifp = m->m_pkthdr.rcvif;
- }
- else
+ origifp = rcvifp = m->m_pkthdr.rcvif;
+ } else if (nd6_prproxy) {
+ /*
+ * In the prefix proxying case, we need to inform nd6_output()
+ * about the inbound interface, so that any subsequent NS
+ * packets generated by nd6_prproxy_ns_output() will not be
+ * sent back to that same interface.
+ */
+ origifp = rcvifp = m->m_pkthdr.rcvif;
+ } else {
+ rcvifp = m->m_pkthdr.rcvif;
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
ifp = rt->rt_ifp;
/* Drop the lock but retain the extra ref */
RT_UNLOCK(rt);
-#if PF
- if (locked)
- lck_mtx_unlock(ip6_mutex);
-
- /* Invoke outbound packet filter */
- error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE);
+ /*
+ * If this is to be processed locally, let ip6_input have it.
+ */
+ if (proxy) {
+ VERIFY(m->m_pkthdr.pkt_flags & PKTF_PROXY_DST);
+ /* Release extra ref */
+ RT_REMREF(rt);
+ if (mcopy != NULL)
+ m_freem(mcopy);
+ return (m);
+ }
- if (locked)
- lck_mtx_lock(ip6_mutex);
+ /* Mark this packet as being forwarded from another interface */
+ m->m_pkthdr.pkt_flags |= PKTF_FORWARDED;
- if (error) {
- if (m != NULL) {
- panic("%s: unexpected packet %p\n", __func__, m);
- /* NOTREACHED */
+#if PF
+ if (PF_IS_ENABLED) {
+#if DUMMYNET
+ struct ip_fw_args args;
+ bzero(&args, sizeof(args));
+
+ args.fwa_m = m;
+ args.fwa_oif = ifp;
+ args.fwa_oflags = 0;
+ args.fwa_ro6 = ip6forward_rt;
+ args.fwa_ro6_pmtu = ip6forward_rt;
+ args.fwa_mtu = rt->rt_ifp->if_mtu;
+ args.fwa_dst6 = dst;
+ args.fwa_origifp = origifp;
+ /* Invoke outbound packet filter */
+ error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, &args);
+#else /* !DUMMYNET */
+ error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, NULL);
+#endif /* !DUMMYNET */
+ if (error != 0 || m == NULL) {
+ if (m != NULL) {
+ panic("%s: unexpected packet %p\n", __func__, m);
+ /* NOTREACHED */
+ }
+ /* Already freed by callee */
+ goto senderr;
}
- /* Already freed by callee */
- goto senderr;
+ /*
+ * We do not use ip6 header again in the code below,
+ * however still adding the bit here so that any new
+ * code in future doesn't end up working with the
+ * wrong pointer
+ */
+ ip6 = mtod(m, struct ip6_hdr *);
}
- ip6 = mtod(m, struct ip6_hdr *);
#endif /* PF */
- error = nd6_output(ifp, origifp, m, dst, rt, locked);
+ len = m_pktlen(m);
+ error = nd6_output(ifp, origifp, m, dst, rt, NULL);
if (error) {
in6_ifstat_inc(ifp, ifs6_out_discard);
ip6stat.ip6s_cantforward++;
} else {
+ /*
+ * Increment stats on the source interface; the ones
+ * for destination interface has been taken care of
+ * during output above by virtue of PKTF_FORWARDED.
+ */
+ rcvifp->if_fpackets++;
+ rcvifp->if_fbytes += len;
+
ip6stat.ip6s_forward++;
in6_ifstat_inc(ifp, ifs6_out_forward);
if (type)
if (mcopy == NULL) {
/* Release extra ref */
RT_REMREF(rt);
- return;
+ return (NULL);
}
switch (error) {
case 0:
icmp6_redirect_output(mcopy, rt);
/* Release extra ref */
RT_REMREF(rt);
- return;
+ return (NULL);
}
#endif
goto freecopy;
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);
/* Release extra ref */
RT_REMREF(rt);
- return;
+ return (NULL);
freecopy:
m_freem(mcopy);
/* Release extra ref */
RT_REMREF(rt);
- return;
+ return (NULL);
}
projects / apple / xnu.git / blobdiff