ip6_cantforward(const struct ip6_hdr *ip6, const struct ifnet *srcifp,
const struct ifnet *dstifp, const char *fmt, ...)
{
char sbuf[INET6_ADDRSTRLEN], dbuf[INET6_ADDRSTRLEN];
char reason[256];
va_list ap;
uint64_t *ip6s;
/* update statistics */
ip6s = IP6_STAT_GETREF();
ip6s[IP6_STAT_CANTFORWARD]++;
if (dstifp)
ip6s[IP6_STAT_BADSCOPE]++;
IP6_STAT_PUTREF();
if (dstifp)
in6_ifstat_inc(dstifp, ifs6_in_discard);
if (ip6_log_time + ip6_log_interval >= time_uptime)
return;
ip6_log_time = time_uptime;
va_start(ap, fmt);
vsnprintf(reason, sizeof(reason), fmt, ap);
va_end(ap);
log(LOG_DEBUG, "Cannot forward from %s@%s to %s@%s nxt %d (%s)\n",
IN6_PRINT(sbuf, &ip6->ip6_src), srcifp ? if_name(srcifp) : "?",
IN6_PRINT(dbuf, &ip6->ip6_dst), dstifp ? if_name(dstifp) : "?",
ip6->ip6_nxt, reason);
}
/*
* Forward a packet. If some error occurs return the sender
* an icmp packet. Note we can't always generate a meaningful
* icmp message because icmp doesn't have a large enough repertoire
* of codes and types.
*
* If not forwarding, just drop the packet. This could be confusing
* if ipforwarding was zero but some routing protocol was advancing
* us as a gateway to somewhere. However, we must let the routing
* protocol deal with that.
*
*/
void
ip6_forward(struct mbuf *m, int srcrt)
{
struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
struct sockaddr_in6 *dst = NULL;
struct rtentry *rt = NULL;
struct route_in6 rin6;
int error, type = 0, code = 0;
struct mbuf *mcopy = NULL;
struct ifnet *origifp; /* maybe unnecessary */
u_int32_t inzone, outzone;
struct in6_addr src_in6, dst_in6, odst;
struct m_tag *fwd_tag;
char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
/*
* Do not forward packets to multicast destination (should be handled
* by ip6_mforward().
* Do not forward packets with unspecified source. It was discussed
* in July 2000, on the ipngwg mailing list.
*/
if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
IP6STAT_INC(ip6s_cantforward);
/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
if (V_ip6_log_time + V_ip6_log_interval < time_uptime) {
V_ip6_log_time = time_uptime;
log(LOG_DEBUG,
"cannot forward "
"from %s to %s nxt %d received on %s\n",
ip6_sprintf(ip6bufs, &ip6->ip6_src),
ip6_sprintf(ip6bufd, &ip6->ip6_dst),
ip6->ip6_nxt,
if_name(m->m_pkthdr.rcvif));
}
m_freem(m);
return;
}
if (
#ifdef IPSTEALTH
V_ip6stealth == 0 &&
#endif
ip6->ip6_hlim <= IPV6_HLIMDEC) {
/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
icmp6_error(m, ICMP6_TIME_EXCEEDED,
ICMP6_TIME_EXCEED_TRANSIT, 0);
return;
}
/*
* Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
* size of IPv6 + ICMPv6 headers) bytes of the packet in case
* we need to generate an ICMP6 message to the src.
* Thanks to M_EXT, in most cases copy will not occur.
*
* It is important to save it before IPsec processing as IPsec
* processing may modify the mbuf.
*/
mcopy = m_copym(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN),
M_NOWAIT);
#ifdef IPSTEALTH
if (V_ip6stealth == 0)
#endif
ip6->ip6_hlim -= IPV6_HLIMDEC;
#if defined(IPSEC) || defined(IPSEC_SUPPORT)
if (IPSEC_ENABLED(ipv6)) {
if ((error = IPSEC_FORWARD(ipv6, m)) != 0) {
/* mbuf consumed by IPsec */
m_freem(mcopy);
if (error != EINPROGRESS)
IP6STAT_INC(ip6s_cantforward);
return;
}
/* No IPsec processing required */
}
#endif
again:
bzero(&rin6, sizeof(struct route_in6));
dst = (struct sockaddr_in6 *)&rin6.ro_dst;
dst->sin6_len = sizeof(struct sockaddr_in6);
dst->sin6_family = AF_INET6;
dst->sin6_addr = ip6->ip6_dst;
again2:
rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m));
rt = rin6.ro_rt;
if (rin6.ro_rt != NULL)
RT_UNLOCK(rin6.ro_rt);
else {
IP6STAT_INC(ip6s_noroute);
in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
if (mcopy) {
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_NOROUTE, 0);
}
goto bad;
}
/*
* 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). 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]
*/
src_in6 = ip6->ip6_src;
if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) {
/* XXX: this should not happen */
IP6STAT_INC(ip6s_cantforward);
IP6STAT_INC(ip6s_badscope);
goto bad;
}
if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) {
IP6STAT_INC(ip6s_cantforward);
IP6STAT_INC(ip6s_badscope);
goto bad;
}
if (inzone != outzone) {
IP6STAT_INC(ip6s_cantforward);
IP6STAT_INC(ip6s_badscope);
in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
if (V_ip6_log_time + V_ip6_log_interval < time_uptime) {
V_ip6_log_time = time_uptime;
log(LOG_DEBUG,
"cannot forward "
"src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
ip6_sprintf(ip6bufs, &ip6->ip6_src),
ip6_sprintf(ip6bufd, &ip6->ip6_dst),
ip6->ip6_nxt,
if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
}
if (mcopy)
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
goto bad;
}
/*
* 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_INC(ip6s_cantforward);
IP6STAT_INC(ip6s_badscope);
goto bad;
}
if (rt->rt_flags & RTF_GATEWAY)
dst = (struct sockaddr_in6 *)rt->rt_gateway;
/*
* If we are to forward the packet using the same interface
* as one we got the packet from, perhaps we should send a redirect
* to sender to shortcut a hop.
* Only send redirect if source is sending directly to us,
* and if packet was not source routed (or has any options).
* Also, don't send redirect if forwarding using a route
* modified by a redirect.
*/
if (V_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) {
/*
* If the incoming interface is equal to the outgoing
* one, and the link attached to the interface is
* point-to-point, then it will be highly probable
* that a routing loop occurs. Thus, we immediately
* drop the packet and send an ICMPv6 error message.
*
* type/code is based on suggestion by Rich Draves.
* not sure if it is the best pick.
*/
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_ADDR, 0);
goto bad;
}
type = ND_REDIRECT;
}
/*
* Fake scoped addresses. Note that even link-local source or
* destinaion can appear, if the originating node just sends the
* packet to us (without address resolution for the destination).
* Since both icmp6_error and icmp6_redirect_output fill the embedded
* link identifiers, we can do this stuff after making a copy for
* returning an error.
*/
if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
/*
* See corresponding comments in ip6_output.
* XXX: but is it possible that ip6_forward() sends a packet
* to a loopback interface? I don't think so, and thus
* I bark here. ([email protected])
* XXX: it is common to route invalid packets to loopback.
* also, the codepath will be visited on use of ::1 in
* rthdr. (itojun)
*/
#if 1
if (0)
#else
if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
#endif
{
printf("ip6_forward: outgoing interface is loopback. "
"src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
ip6_sprintf(ip6bufs, &ip6->ip6_src),
ip6_sprintf(ip6bufd, &ip6->ip6_dst),
ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
if_name(rt->rt_ifp));
}
/* we can just use rcvif in forwarding. */
origifp = m->m_pkthdr.rcvif;
}
else
origifp = rt->rt_ifp;
/*
* 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);
/* Jump over all PFIL processing if hooks are not active. */
if (!PFIL_HOOKED(&V_inet6_pfil_hook))
goto pass;
odst = ip6->ip6_dst;
/* Run through list of hooks for output packets. */
error = pfil_run_hooks(&V_inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL);
if (error != 0 || m == NULL)
goto freecopy; /* consumed by filter */
ip6 = mtod(m, struct ip6_hdr *);
/* See if destination IP address was changed by packet filter. */
if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
m->m_flags |= M_SKIP_FIREWALL;
/* If destination is now ourself drop to ip6_input(). */
if (in6_localip(&ip6->ip6_dst))
m->m_flags |= M_FASTFWD_OURS;
else {
RTFREE(rt);
goto again; /* Redo the routing table lookup. */
}
}
/* See if local, if yes, send it to netisr. */
if (m->m_flags & M_FASTFWD_OURS) {
if (m->m_pkthdr.rcvif == NULL)
m->m_pkthdr.rcvif = V_loif;
if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
m->m_pkthdr.csum_flags |=
CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
m->m_pkthdr.csum_data = 0xffff;
}
#ifdef SCTP
if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
#endif
error = netisr_queue(NETISR_IPV6, m);
goto out;
}
/* Or forward to some other address? */
if ((m->m_flags & M_IP6_NEXTHOP) &&
(fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
dst = (struct sockaddr_in6 *)&rin6.ro_dst;
bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in6));
m->m_flags |= M_SKIP_FIREWALL;
m->m_flags &= ~M_IP6_NEXTHOP;
m_tag_delete(m, fwd_tag);
RTFREE(rt);
goto again2;
}
pass:
/* See if the size was changed by the packet filter. */
if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
if (mcopy)
icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0,
IN6_LINKMTU(rt->rt_ifp));
goto bad;
}
error = nd6_output_ifp(rt->rt_ifp, origifp, m, dst, NULL);
if (error) {
in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
IP6STAT_INC(ip6s_cantforward);
} else {
IP6STAT_INC(ip6s_forward);
in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
if (type)
IP6STAT_INC(ip6s_redirectsent);
else {
if (mcopy)
goto freecopy;
}
}
if (mcopy == NULL)
goto out;
switch (error) {
case 0:
if (type == ND_REDIRECT) {
icmp6_redirect_output(mcopy, rt);
goto out;
}
goto freecopy;
case EMSGSIZE:
/* xxx MTU is constant in PPP? */
goto freecopy;
case ENOBUFS:
/* Tell source to slow down like source quench in IP? */
goto freecopy;
case ENETUNREACH: /* shouldn't happen, checked above */
case EHOSTUNREACH:
case ENETDOWN:
case EHOSTDOWN:
default:
type = ICMP6_DST_UNREACH;
code = ICMP6_DST_UNREACH_ADDR;
break;
}
icmp6_error(mcopy, type, code, 0);
goto out;
freecopy:
m_freem(mcopy);
goto out;
bad:
m_freem(m);
out:
if (rt != NULL)
RTFREE(rt);
}
/*
* Forward a packet. If some error occurs return the sender
* an icmp packet. Note we can't always generate a meaningful
* icmp message because icmp doesn't have a large enough repertoire
* of codes and types.
*
* If not forwarding, just drop the packet. This could be confusing
* if ipforwarding was zero but some routing protocol was advancing
* us as a gateway to somewhere. However, we must let the routing
* protocol deal with that.
*
*/
void
ip6_forward(struct mbuf *m, int srcrt)
{
struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
struct sockaddr_in6 *dst = NULL;
struct rtentry *rt = NULL;
struct route_in6 rin6;
int error, type = 0, code = 0;
struct mbuf *mcopy = NULL;
struct ifnet *origifp; /* maybe unnecessary */
u_int32_t inzone, outzone;
struct in6_addr src_in6, dst_in6, odst;
#ifdef IPSEC
struct secpolicy *sp = NULL;
#endif
#ifdef SCTP
int sw_csum;
#endif
struct m_tag *fwd_tag;
char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
#ifdef IPSEC
/*
* Check AH/ESP integrity.
*/
/*
* Don't increment ip6s_cantforward because this is the check
* before forwarding packet actually.
*/
if (ipsec6_in_reject(m, NULL)) {
IPSEC6STAT_INC(ips_in_polvio);
m_freem(m);
return;
}
#endif /* IPSEC */
/*
* Do not forward packets to multicast destination (should be handled
* by ip6_mforward().
* Do not forward packets with unspecified source. It was discussed
* in July 2000, on the ipngwg mailing list.
*/
if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
IP6STAT_INC(ip6s_cantforward);
/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
if (V_ip6_log_time + V_ip6_log_interval < time_uptime) {
V_ip6_log_time = time_uptime;
log(LOG_DEBUG,
"cannot forward "
"from %s to %s nxt %d received on %s\n",
ip6_sprintf(ip6bufs, &ip6->ip6_src),
ip6_sprintf(ip6bufd, &ip6->ip6_dst),
ip6->ip6_nxt,
if_name(m->m_pkthdr.rcvif));
}
m_freem(m);
return;
}
#ifdef IPSTEALTH
if (!V_ip6stealth) {
#endif
if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
icmp6_error(m, ICMP6_TIME_EXCEEDED,
ICMP6_TIME_EXCEED_TRANSIT, 0);
return;
}
ip6->ip6_hlim -= IPV6_HLIMDEC;
#ifdef IPSTEALTH
}
#endif
/*
* Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
* size of IPv6 + ICMPv6 headers) bytes of the packet in case
* we need to generate an ICMP6 message to the src.
* Thanks to M_EXT, in most cases copy will not occur.
*
* It is important to save it before IPsec processing as IPsec
* processing may modify the mbuf.
*/
mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
#ifdef IPSEC
/* get a security policy for this packet */
sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
IP_FORWARDING, &error);
if (sp == NULL) {
IPSEC6STAT_INC(ips_out_inval);
IP6STAT_INC(ip6s_cantforward);
if (mcopy) {
#if 0
/* XXX: what icmp ? */
#else
m_freem(mcopy);
#endif
}
m_freem(m);
return;
}
error = 0;
/* check policy */
switch (sp->policy) {
case IPSEC_POLICY_DISCARD:
/*
* This packet is just discarded.
*/
IPSEC6STAT_INC(ips_out_polvio);
IP6STAT_INC(ip6s_cantforward);
KEY_FREESP(&sp);
if (mcopy) {
#if 0
/* XXX: what icmp ? */
#else
m_freem(mcopy);
#endif
}
m_freem(m);
return;
case IPSEC_POLICY_BYPASS:
case IPSEC_POLICY_NONE:
/* no need to do IPsec. */
KEY_FREESP(&sp);
goto skip_ipsec;
case IPSEC_POLICY_IPSEC:
if (sp->req == NULL) {
/* XXX should be panic ? */
printf("ip6_forward: No IPsec request specified.\n");
IP6STAT_INC(ip6s_cantforward);
KEY_FREESP(&sp);
if (mcopy) {
#if 0
/* XXX: what icmp ? */
#else
m_freem(mcopy);
#endif
}
m_freem(m);
return;
}
/* do IPsec */
break;
case IPSEC_POLICY_ENTRUST:
default:
/* should be panic ?? */
printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
KEY_FREESP(&sp);
goto skip_ipsec;
}
{
struct ipsecrequest *isr = NULL;
/*
* when the kernel forwards a packet, it is not proper to apply
* IPsec transport mode to the packet is not proper. this check
* avoid from this.
* at present, if there is even a transport mode SA request in the
* security policy, the kernel does not apply IPsec to the packet.
* this check is not enough because the following case is valid.
* ipsec esp/tunnel/xxx-xxx/require esp/transport//require;
*/
for (isr = sp->req; isr; isr = isr->next) {
if (isr->saidx.mode == IPSEC_MODE_ANY)
goto doipsectunnel;
if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
goto doipsectunnel;
}
/*
* if there's no need for tunnel mode IPsec, skip.
*/
if (!isr)
goto skip_ipsec;
doipsectunnel:
/*
* All the extension headers will become inaccessible
* (since they can be encrypted).
* Don't panic, we need no more updates to extension headers
* on inner IPv6 packet (since they are now encapsulated).
*
* IPv6 [ESP|AH] IPv6 [extension headers] payload
*/
/*
* If we need to encapsulate the packet, do it here
* ipsec6_proces_packet will send the packet using ip6_output
*/
error = ipsec6_process_packet(m, sp->req);
KEY_FREESP(&sp);
if (error == EJUSTRETURN) {
/*
* We had a SP with a level of 'use' and no SA. We
* will just continue to process the packet without
* IPsec processing.
*/
error = 0;
goto skip_ipsec;
}
if (error) {
/* mbuf is already reclaimed in ipsec6_process_packet. */
switch (error) {
case EHOSTUNREACH:
case ENETUNREACH:
case EMSGSIZE:
case ENOBUFS:
case ENOMEM:
break;
default:
printf("ip6_output (ipsec): error code %d\n", error);
/* FALLTHROUGH */
case ENOENT:
/* don't show these error codes to the user */
break;
}
IP6STAT_INC(ip6s_cantforward);
if (mcopy) {
#if 0
/* XXX: what icmp ? */
#else
m_freem(mcopy);
#endif
}
return;
} else {
/*
* In the FAST IPSec case we have already
* re-injected the packet and it has been freed
* by the ipsec_done() function. So, just clean
* up after ourselves.
*/
m = NULL;
goto freecopy;
}
}
skip_ipsec:
#endif
again:
bzero(&rin6, sizeof(struct route_in6));
dst = (struct sockaddr_in6 *)&rin6.ro_dst;
dst->sin6_len = sizeof(struct sockaddr_in6);
dst->sin6_family = AF_INET6;
dst->sin6_addr = ip6->ip6_dst;
again2:
rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m));
if (rin6.ro_rt != NULL)
RT_UNLOCK(rin6.ro_rt);
else {
IP6STAT_INC(ip6s_noroute);
in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
if (mcopy) {
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_NOROUTE, 0);
}
goto bad;
}
rt = rin6.ro_rt;
/*
* 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). 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]
*/
src_in6 = ip6->ip6_src;
if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) {
/* XXX: this should not happen */
IP6STAT_INC(ip6s_cantforward);
IP6STAT_INC(ip6s_badscope);
goto bad;
}
if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) {
IP6STAT_INC(ip6s_cantforward);
IP6STAT_INC(ip6s_badscope);
goto bad;
}
if (inzone != outzone) {
IP6STAT_INC(ip6s_cantforward);
IP6STAT_INC(ip6s_badscope);
in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
if (V_ip6_log_time + V_ip6_log_interval < time_uptime) {
V_ip6_log_time = time_uptime;
log(LOG_DEBUG,
"cannot forward "
"src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
ip6_sprintf(ip6bufs, &ip6->ip6_src),
ip6_sprintf(ip6bufd, &ip6->ip6_dst),
ip6->ip6_nxt,
if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
}
if (mcopy)
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
goto bad;
}
/*
* 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_INC(ip6s_cantforward);
IP6STAT_INC(ip6s_badscope);
goto bad;
}
if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
if (mcopy) {
u_long mtu;
#ifdef IPSEC
struct secpolicy *sp;
int ipsecerror;
size_t ipsechdrsiz;
#endif /* IPSEC */
mtu = IN6_LINKMTU(rt->rt_ifp);
#ifdef IPSEC
/*
* When we do IPsec tunnel ingress, we need to play
* 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".
*/
sp = ipsec_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
IP_FORWARDING, &ipsecerror);
if (sp) {
ipsechdrsiz = ipsec_hdrsiz(mcopy,
IPSEC_DIR_OUTBOUND, NULL);
if (ipsechdrsiz < mtu)
mtu -= ipsechdrsiz;
}
/*
* if mtu becomes less than minimum MTU,
* tell minimum MTU (and I'll need to fragment it).
*/
if (mtu < IPV6_MMTU)
mtu = IPV6_MMTU;
#endif /* IPSEC */
icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
}
goto bad;
}
if (rt->rt_flags & RTF_GATEWAY)
dst = (struct sockaddr_in6 *)rt->rt_gateway;
/*
* If we are to forward the packet using the same interface
* as one we got the packet from, perhaps we should send a redirect
* to sender to shortcut a hop.
* Only send redirect if source is sending directly to us,
* and if packet was not source routed (or has any options).
* Also, don't send redirect if forwarding using a route
* modified by a redirect.
*/
if (V_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) {
/*
* If the incoming interface is equal to the outgoing
* one, and the link attached to the interface is
* point-to-point, then it will be highly probable
* that a routing loop occurs. Thus, we immediately
* drop the packet and send an ICMPv6 error message.
*
* type/code is based on suggestion by Rich Draves.
* not sure if it is the best pick.
*/
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_ADDR, 0);
goto bad;
}
type = ND_REDIRECT;
}
/*
* Fake scoped addresses. Note that even link-local source or
* destinaion can appear, if the originating node just sends the
* packet to us (without address resolution for the destination).
* Since both icmp6_error and icmp6_redirect_output fill the embedded
* link identifiers, we can do this stuff after making a copy for
* returning an error.
*/
if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
/*
* See corresponding comments in ip6_output.
* XXX: but is it possible that ip6_forward() sends a packet
* to a loopback interface? I don't think so, and thus
* I bark here. ([email protected])
* XXX: it is common to route invalid packets to loopback.
* also, the codepath will be visited on use of ::1 in
* rthdr. (itojun)
*/
#if 1
if (0)
#else
if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
#endif
{
printf("ip6_forward: outgoing interface is loopback. "
"src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
ip6_sprintf(ip6bufs, &ip6->ip6_src),
ip6_sprintf(ip6bufd, &ip6->ip6_dst),
ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
if_name(rt->rt_ifp));
}
/* we can just use rcvif in forwarding. */
origifp = m->m_pkthdr.rcvif;
}
else
origifp = rt->rt_ifp;
/*
* 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);
/* Jump over all PFIL processing if hooks are not active. */
if (!PFIL_HOOKED(&V_inet6_pfil_hook))
goto pass;
odst = ip6->ip6_dst;
/* Run through list of hooks for output packets. */
error = pfil_run_hooks(&V_inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL);
if (error != 0 || m == NULL)
goto freecopy; /* consumed by filter */
ip6 = mtod(m, struct ip6_hdr *);
/* See if destination IP address was changed by packet filter. */
if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
m->m_flags |= M_SKIP_FIREWALL;
/* If destination is now ourself drop to ip6_input(). */
if (in6_localip(&ip6->ip6_dst)) {
m->m_flags |= M_FASTFWD_OURS;
if (m->m_pkthdr.rcvif == NULL)
m->m_pkthdr.rcvif = V_loif;
if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
m->m_pkthdr.csum_flags |=
CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
m->m_pkthdr.csum_data = 0xffff;
}
#ifdef SCTP
if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
#endif
error = netisr_queue(NETISR_IPV6, m);
goto out;
} else
goto again; /* Redo the routing table lookup. */
}
/* See if local, if yes, send it to netisr. */
if (m->m_flags & M_FASTFWD_OURS) {
if (m->m_pkthdr.rcvif == NULL)
m->m_pkthdr.rcvif = V_loif;
if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
m->m_pkthdr.csum_flags |=
CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
m->m_pkthdr.csum_data = 0xffff;
}
#ifdef SCTP
if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
#endif
error = netisr_queue(NETISR_IPV6, m);
goto out;
}
/* Or forward to some other address? */
if ((m->m_flags & M_IP6_NEXTHOP) &&
(fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
dst = (struct sockaddr_in6 *)&rin6.ro_dst;
bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in6));
m->m_flags |= M_SKIP_FIREWALL;
m->m_flags &= ~M_IP6_NEXTHOP;
m_tag_delete(m, fwd_tag);
goto again2;
}
pass:
error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
if (error) {
in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
IP6STAT_INC(ip6s_cantforward);
} else {
IP6STAT_INC(ip6s_forward);
in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
if (type)
IP6STAT_INC(ip6s_redirectsent);
else {
if (mcopy)
goto freecopy;
}
}
if (mcopy == NULL)
goto out;
switch (error) {
case 0:
if (type == ND_REDIRECT) {
icmp6_redirect_output(mcopy, rt);
goto out;
}
goto freecopy;
case EMSGSIZE:
/* xxx MTU is constant in PPP? */
goto freecopy;
case ENOBUFS:
/* Tell source to slow down like source quench in IP? */
goto freecopy;
case ENETUNREACH: /* shouldn't happen, checked above */
case EHOSTUNREACH:
case ENETDOWN:
case EHOSTDOWN:
default:
type = ICMP6_DST_UNREACH;
code = ICMP6_DST_UNREACH_ADDR;
break;
}
icmp6_error(mcopy, type, code, 0);
goto out;
freecopy:
m_freem(mcopy);
goto out;
bad:
m_freem(m);
out:
if (rt != NULL)
RTFREE(rt);
}
void
ip6_forward(struct mbuf *m, int srcrt)
{
struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
struct sockaddr_in6 *dst;
struct rtentry *rt;
int error, type = 0, code = 0;
struct mbuf *mcopy = NULL;
struct ifnet *origifp; /* maybe unnecessary */
#ifdef IPSEC
struct secpolicy *sp = NULL;
#endif
#ifdef IPSEC
/*
* Check AH/ESP integrity.
*/
/*
* Don't increment ip6s_cantforward because this is the check
* before forwarding packet actually.
*/
if (ipsec6_in_reject(m, NULL)) {
#if !defined(FAST_IPSEC)
ipsec6stat.in_polvio++;
#endif
m_freem(m);
return;
}
#endif /* IPSEC */
/*
* Do not forward packets to multicast destination (should be handled
* by ip6_mforward().
* Do not forward packets with unspecified source. It was discussed
* in July 2000, on ipngwg mailing list.
*/
if ((m->m_flags & (M_BCAST | M_MCAST)) != 0 ||
IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
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 < time_second) {
ip6_log_time = time_second;
log(LOG_DEBUG,
"cannot forward "
"from %s to %s nxt %d received on %s\n",
ip6_sprintf(&ip6->ip6_src),
ip6_sprintf(&ip6->ip6_dst),
ip6->ip6_nxt,
if_name(m->m_pkthdr.rcvif));
}
m_freem(m);
return;
}
if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
icmp6_error(m, ICMP6_TIME_EXCEEDED,
ICMP6_TIME_EXCEED_TRANSIT, 0);
return;
}
ip6->ip6_hlim -= IPV6_HLIMDEC;
/*
* Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
* size of IPv6 + ICMPv6 headers) bytes of the packet in case
* we need to generate an ICMP6 message to the src.
* Thanks to M_EXT, in most cases copy will not occur.
*
* It is important to save it before IPsec processing as IPsec
* processing may modify the mbuf.
*/
mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
#ifdef IPSEC
/* get a security policy for this packet */
sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, IP_FORWARDING,
&error);
if (sp == NULL) {
ipsec6stat.out_inval++;
ip6stat.ip6s_cantforward++;
if (mcopy) {
#if 0
/* XXX: what icmp ? */
#else
m_freem(mcopy);
#endif
}
m_freem(m);
return;
}
error = 0;
/* check policy */
switch (sp->policy) {
case IPSEC_POLICY_DISCARD:
/*
* This packet is just discarded.
*/
ipsec6stat.out_polvio++;
ip6stat.ip6s_cantforward++;
key_freesp(sp);
if (mcopy) {
#if 0
/* XXX: what icmp ? */
#else
m_freem(mcopy);
#endif
}
m_freem(m);
return;
case IPSEC_POLICY_BYPASS:
case IPSEC_POLICY_NONE:
/* no need to do IPsec. */
key_freesp(sp);
goto skip_ipsec;
case IPSEC_POLICY_IPSEC:
if (sp->req == NULL) {
/* XXX should be panic ? */
kprintf("ip6_forward: No IPsec request specified.\n");
ip6stat.ip6s_cantforward++;
key_freesp(sp);
if (mcopy) {
#if 0
/* XXX: what icmp ? */
#else
m_freem(mcopy);
#endif
}
m_freem(m);
return;
}
/* do IPsec */
break;
case IPSEC_POLICY_ENTRUST:
default:
/* should be panic ?? */
kprintf("ip6_forward: Invalid policy found. %d\n", sp->policy);
key_freesp(sp);
goto skip_ipsec;
}
{
struct ipsec_output_state state;
/*
* All the extension headers will become inaccessible
* (since they can be encrypted).
* Don't panic, we need no more updates to extension headers
* on inner IPv6 packet (since they are now encapsulated).
*
* IPv6 [ESP|AH] IPv6 [extension headers] payload
*/
bzero(&state, sizeof(state));
state.m = m;
state.ro = NULL; /* update at ipsec6_output_tunnel() */
state.dst = NULL; /* update at ipsec6_output_tunnel() */
error = ipsec6_output_tunnel(&state, sp, 0);
m = state.m;
key_freesp(sp);
if (error) {
/* mbuf is already reclaimed in ipsec6_output_tunnel. */
switch (error) {
case EHOSTUNREACH:
case ENETUNREACH:
case EMSGSIZE:
case ENOBUFS:
case ENOMEM:
break;
default:
kprintf("ip6_output (ipsec): error code %d\n", error);
/* fall through */
case ENOENT:
/* don't show these error codes to the user */
break;
}
ip6stat.ip6s_cantforward++;
if (mcopy) {
#if 0
/* XXX: what icmp ? */
#else
m_freem(mcopy);
#endif
}
m_freem(m);
return;
}
}
skip_ipsec:
#endif /* IPSEC */
dst = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
if (!srcrt) {
/*
* ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst
*/
if (ip6_forward_rt.ro_rt == NULL ||
!(ip6_forward_rt.ro_rt->rt_flags & RTF_UP)) {
if (ip6_forward_rt.ro_rt) {
RTFREE(ip6_forward_rt.ro_rt);
ip6_forward_rt.ro_rt = 0;
}
/* this probably fails but give it a try again */
rtalloc_ign((struct route *)&ip6_forward_rt,
RTF_PRCLONING);
}
if (ip6_forward_rt.ro_rt == NULL) {
ip6stat.ip6s_noroute++;
in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
if (mcopy) {
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_NOROUTE, 0);
}
m_freem(m);
return;
}
} else if ((rt = ip6_forward_rt.ro_rt) == NULL ||
!IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) {
if (ip6_forward_rt.ro_rt) {
RTFREE(ip6_forward_rt.ro_rt);
ip6_forward_rt.ro_rt = 0;
}
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 *)&ip6_forward_rt, RTF_PRCLONING);
if (ip6_forward_rt.ro_rt == NULL) {
ip6stat.ip6s_noroute++;
in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
if (mcopy) {
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_NOROUTE, 0);
}
m_freem(m);
return;
}
}
rt = ip6_forward_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]
*/
if (in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_src) !=
in6_addr2scopeid(rt->rt_ifp, &ip6->ip6_src)) {
ip6stat.ip6s_cantforward++;
ip6stat.ip6s_badscope++;
in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
if (ip6_log_time + ip6_log_interval < time_second) {
ip6_log_time = time_second;
log(LOG_DEBUG,
"cannot forward "
"src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
ip6_sprintf(&ip6->ip6_src),
ip6_sprintf(&ip6->ip6_dst),
ip6->ip6_nxt,
if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
}
if (mcopy)
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
m_freem(m);
return;
}
if (m->m_pkthdr.len > rt->rt_ifp->if_mtu) {
in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
if (mcopy) {
u_long mtu;
#ifdef IPSEC
struct secpolicy *sp;
int ipsecerror;
size_t ipsechdrsiz;
#endif
mtu = rt->rt_ifp->if_mtu;
#ifdef IPSEC
/*
* When we do IPsec tunnel ingress, we need to play
* with if_mtu 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".
*/
sp = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
IP_FORWARDING, &ipsecerror);
if (sp) {
ipsechdrsiz = ipsec6_hdrsiz(mcopy,
IPSEC_DIR_OUTBOUND, NULL);
if (ipsechdrsiz < mtu)
mtu -= ipsechdrsiz;
}
/*
* if mtu becomes less than minimum MTU,
* tell minimum MTU (and I'll need to fragment it).
*/
if (mtu < IPV6_MMTU)
mtu = IPV6_MMTU;
#endif
icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
}
m_freem(m);
return;
}
if (rt->rt_flags & RTF_GATEWAY)
dst = (struct sockaddr_in6 *)rt->rt_gateway;
/*
* If we are to forward the packet using the same interface
* as one we got the packet from, perhaps we should send a redirect
* to sender to shortcut a hop.
* Only send redirect if source is sending directly to us,
* and if packet was not source routed (or has any options).
* Also, don't send redirect if forwarding using a route
* modified by a redirect.
*/
if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) == 0) {
if (rt->rt_ifp->if_flags & IFF_POINTOPOINT) {
/*
* If the incoming interface is equal to the outgoing
* one, and the link attached to the interface is
* point-to-point, then it will be highly probable
* that a routing loop occurs. Thus, we immediately
* drop the packet and send an ICMPv6 error message.
*
* type/code is based on suggestion by Rich Draves.
* not sure if it is the best pick.
*/
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_ADDR, 0);
m_freem(m);
return;
}
type = ND_REDIRECT;
}
/*
* Check with the firewall...
*/
if (ip6_fw_enable && ip6_fw_chk_ptr) {
u_short port = 0;
/* If ipfw says divert, we have to just drop packet */
if ((*ip6_fw_chk_ptr)(&ip6, rt->rt_ifp, &port, &m)) {
m_freem(m);
goto freecopy;
}
if (!m)
goto freecopy;
}
/*
* Fake scoped addresses. Note that even link-local source or
* destinaion can appear, if the originating node just sends the
* packet to us (without address resolution for the destination).
* Since both icmp6_error and icmp6_redirect_output fill the embedded
* link identifiers, we can do this stuff after making a copy for
* returning an error.
*/
if (rt->rt_ifp->if_flags & IFF_LOOPBACK) {
/*
* See corresponding comments in ip6_output.
* XXX: but is it possible that ip6_forward() sends a packet
* to a loopback interface? I don't think so, and thus
* I bark here. ([email protected])
* XXX: it is common to route invalid packets to loopback.
* also, the codepath will be visited on use of ::1 in
* rthdr. (itojun)
*/
#if 1
if (0)
#else
if ((rt->rt_flags & (RTF_BLACKHOLE | RTF_REJECT)) == 0)
#endif
{
kprintf("ip6_forward: outgoing interface is loopback. "
"src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
ip6_sprintf(&ip6->ip6_src),
ip6_sprintf(&ip6->ip6_dst),
ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
if_name(rt->rt_ifp));
}
/* we can just use rcvif in forwarding. */
origifp = m->m_pkthdr.rcvif;
}
else
origifp = rt->rt_ifp;
/*
* 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);
/*
* Run through list of hooks for output packets.
*/
if (pfil_has_hooks(&inet6_pfil_hook)) {
error = pfil_run_hooks(&inet6_pfil_hook, &m, rt->rt_ifp,
PFIL_OUT);
if (error != 0)
goto senderr;
if (m == NULL)
goto freecopy;
ip6 = mtod(m, struct ip6_hdr *);
}
/*
* Fragment input
*/
int
frag6_input(struct mbuf **mp, int *offp, int proto)
{
struct mbuf *m = *mp, *t;
struct ip6_hdr *ip6;
struct ip6_frag *ip6f;
struct ip6q *q6;
struct ip6asfrag *af6, *ip6af, *af6dwn;
struct in6_ifaddr *ia;
int offset = *offp, nxt, i, next;
int first_frag = 0;
int fragoff, frgpartlen; /* must be larger than u_int16_t */
struct ifnet *dstifp;
u_int8_t ecn, ecn0;
#if 0
char ip6buf[INET6_ADDRSTRLEN];
#endif
ip6 = mtod(m, struct ip6_hdr *);
#ifndef PULLDOWN_TEST
IP6_EXTHDR_CHECK(m, offset, sizeof(struct ip6_frag), IPPROTO_DONE);
ip6f = (struct ip6_frag *)((caddr_t)ip6 + offset);
#else
IP6_EXTHDR_GET(ip6f, struct ip6_frag *, m, offset, sizeof(*ip6f));
if (ip6f == NULL)
return (IPPROTO_DONE);
#endif
dstifp = NULL;
/* find the destination interface of the packet. */
ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
if (ia != NULL) {
dstifp = ia->ia_ifp;
ifa_free(&ia->ia_ifa);
}
/* jumbo payload can't contain a fragment header */
if (ip6->ip6_plen == 0) {
icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, offset);
in6_ifstat_inc(dstifp, ifs6_reass_fail);
return IPPROTO_DONE;
}
/*
* check whether fragment packet's fragment length is
* multiple of 8 octets.
* sizeof(struct ip6_frag) == 8
* sizeof(struct ip6_hdr) = 40
*/
if ((ip6f->ip6f_offlg & IP6F_MORE_FRAG) &&
(((ntohs(ip6->ip6_plen) - offset) & 0x7) != 0)) {
icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER,
offsetof(struct ip6_hdr, ip6_plen));
in6_ifstat_inc(dstifp, ifs6_reass_fail);
return IPPROTO_DONE;
}
IP6STAT_INC(ip6s_fragments);
in6_ifstat_inc(dstifp, ifs6_reass_reqd);
/* offset now points to data portion */
offset += sizeof(struct ip6_frag);
/*
* RFC 6946: Handle "atomic" fragments (offset and m bit set to 0)
* upfront, unrelated to any reassembly. Just skip the fragment header.
*/
if ((ip6f->ip6f_offlg & ~IP6F_RESERVED_MASK) == 0) {
/* XXX-BZ we want dedicated counters for this. */
IP6STAT_INC(ip6s_reassembled);
in6_ifstat_inc(dstifp, ifs6_reass_ok);
*offp = offset;
return (ip6f->ip6f_nxt);
}
IP6Q_LOCK();
/*
* Enforce upper bound on number of fragments.
* If maxfrag is 0, never accept fragments.
* If maxfrag is -1, accept all fragments without limitation.
*/
if (V_ip6_maxfrags < 0)
;
else if (V_frag6_nfrags >= (u_int)V_ip6_maxfrags)
goto dropfrag;
for (q6 = V_ip6q.ip6q_next; q6 != &V_ip6q; q6 = q6->ip6q_next)
if (ip6f->ip6f_ident == q6->ip6q_ident &&
IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &q6->ip6q_src) &&
IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &q6->ip6q_dst)
#ifdef MAC
&& mac_ip6q_match(m, q6)
#endif
)
break;
if (q6 == &V_ip6q) {
/*
* the first fragment to arrive, create a reassembly queue.
*/
first_frag = 1;
/*
* Enforce upper bound on number of fragmented packets
* for which we attempt reassembly;
* If maxfragpackets is 0, never accept fragments.
* If maxfragpackets is -1, accept all fragments without
* limitation.
*/
if (V_ip6_maxfragpackets < 0)
;
else if (V_frag6_nfragpackets >= (u_int)V_ip6_maxfragpackets)
goto dropfrag;
V_frag6_nfragpackets++;
q6 = (struct ip6q *)malloc(sizeof(struct ip6q), M_FTABLE,
M_NOWAIT);
if (q6 == NULL)
goto dropfrag;
bzero(q6, sizeof(*q6));
#ifdef MAC
if (mac_ip6q_init(q6, M_NOWAIT) != 0) {
free(q6, M_FTABLE);
goto dropfrag;
}
mac_ip6q_create(m, q6);
#endif
frag6_insque(q6, &V_ip6q);
/* ip6q_nxt will be filled afterwards, from 1st fragment */
q6->ip6q_down = q6->ip6q_up = (struct ip6asfrag *)q6;
#ifdef notyet
q6->ip6q_nxtp = (u_char *)nxtp;
#endif
q6->ip6q_ident = ip6f->ip6f_ident;
q6->ip6q_ttl = IPV6_FRAGTTL;
q6->ip6q_src = ip6->ip6_src;
q6->ip6q_dst = ip6->ip6_dst;
q6->ip6q_ecn =
(ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK;
q6->ip6q_unfrglen = -1; /* The 1st fragment has not arrived. */
q6->ip6q_nfrag = 0;
}
static
void
ip6_input(netmsg_t msg)
{
struct mbuf *m = msg->packet.nm_packet;
struct ip6_hdr *ip6;
int off = sizeof(struct ip6_hdr), nest;
u_int32_t plen;
u_int32_t rtalert = ~0;
int nxt, ours = 0, rh_present = 0;
struct ifnet *deliverifp = NULL;
struct in6_addr odst;
int srcrt = 0;
#ifdef IPSEC
/*
* should the inner packet be considered authentic?
* see comment in ah4_input().
*/
if (m) {
m->m_flags &= ~M_AUTHIPHDR;
m->m_flags &= ~M_AUTHIPDGM;
}
#endif
/*
* make sure we don't have onion peering information into m_aux.
*/
ip6_delaux(m);
/*
* mbuf statistics
*/
if (m->m_flags & M_EXT) {
if (m->m_next)
ip6stat.ip6s_mext2m++;
else
ip6stat.ip6s_mext1++;
} else {
#define M2MMAX NELEM(ip6stat.ip6s_m2m)
if (m->m_next) {
if (m->m_flags & M_LOOP) {
ip6stat.ip6s_m2m[loif[0].if_index]++; /* XXX */
} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
else
ip6stat.ip6s_m2m[0]++;
} else
ip6stat.ip6s_m1++;
#undef M2MMAX
}
in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
ip6stat.ip6s_total++;
#ifndef PULLDOWN_TEST
/*
* L2 bridge code and some other code can return mbuf chain
* that does not conform to KAME requirement. too bad.
* XXX: fails to join if interface MTU > MCLBYTES. jumbogram?
*/
if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
struct mbuf *n;
n = m_getb(m->m_pkthdr.len, MB_DONTWAIT, MT_HEADER, M_PKTHDR);
if (n == NULL)
goto bad;
M_MOVE_PKTHDR(n, m);
m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
n->m_len = n->m_pkthdr.len;
m_freem(m);
m = n;
}
IP6_EXTHDR_CHECK_VOIDRET(m, 0, sizeof(struct ip6_hdr));
#endif
if (m->m_len < sizeof(struct ip6_hdr)) {
struct ifnet *inifp;
inifp = m->m_pkthdr.rcvif;
if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
ip6stat.ip6s_toosmall++;
in6_ifstat_inc(inifp, ifs6_in_hdrerr);
goto bad2;
}
}
ip6 = mtod(m, struct ip6_hdr *);
if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
ip6stat.ip6s_badvers++;
in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
goto bad;
}
/*
* Run through list of hooks for input packets.
*
* NB: Beware of the destination address changing
* (e.g. by NAT rewriting). When this happens,
* tell ip6_forward to do the right thing.
*/
if (pfil_has_hooks(&inet6_pfil_hook)) {
odst = ip6->ip6_dst;
if (pfil_run_hooks(&inet6_pfil_hook, &m,
m->m_pkthdr.rcvif, PFIL_IN)) {
goto bad2;
}
if (m == NULL) /* consumed by filter */
goto bad2;
ip6 = mtod(m, struct ip6_hdr *);
srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
}
struct mbuf *
ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
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 */
u_int32_t inzone, outzone;
struct in6_addr src_in6, dst_in6;
#if IPSEC
struct secpolicy *sp = NULL;
#endif
struct timeval timenow;
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
/*
* Check AH/ESP integrity.
*/
/*
* Don't increment ip6s_cantforward because this is the check
* before forwarding packet actually.
*/
if (ipsec_bypass == 0) {
if (ipsec6_in_reject(m, NULL)) {
IPSEC_STAT_INCREMENT(ipsec6stat.in_polvio);
m_freem(m);
return (NULL);
}
}
#endif /*IPSEC*/
/*
* Do not forward packets to multicast destination (should be handled
* by ip6_mforward().
* Do not forward packets with unspecified source. It was discussed
* in July 2000, on ipngwg mailing list.
*/
if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
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;
log(LOG_DEBUG,
"cannot forward "
"from %s to %s nxt %d received on %s\n",
ip6_sprintf(&ip6->ip6_src),
ip6_sprintf(&ip6->ip6_dst),
ip6->ip6_nxt,
if_name(m->m_pkthdr.rcvif));
}
m_freem(m);
return (NULL);
}
if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
icmp6_error(m, ICMP6_TIME_EXCEEDED,
ICMP6_TIME_EXCEED_TRANSIT, 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;
/*
* Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
* size of IPv6 + ICMPv6 headers) bytes of the packet in case
* we need to generate an ICMP6 message to the src.
* Thanks to M_EXT, in most cases copy will not occur.
*
* It is important to save it before IPsec processing as IPsec
* processing may modify the mbuf.
*/
mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
#if IPSEC
if (ipsec_bypass != 0)
goto skip_ipsec;
/* get a security policy for this packet */
sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, IP_FORWARDING,
&error);
if (sp == NULL) {
IPSEC_STAT_INCREMENT(ipsec6stat.out_inval);
ip6stat.ip6s_cantforward++;
if (mcopy) {
#if 0
/* XXX: what icmp ? */
#else
m_freem(mcopy);
#endif
}
m_freem(m);
return (NULL);
}
error = 0;
/* check policy */
switch (sp->policy) {
case IPSEC_POLICY_DISCARD:
case IPSEC_POLICY_GENERATE:
/*
* This packet is just discarded.
*/
IPSEC_STAT_INCREMENT(ipsec6stat.out_polvio);
ip6stat.ip6s_cantforward++;
key_freesp(sp, KEY_SADB_UNLOCKED);
if (mcopy) {
#if 0
/* XXX: what icmp ? */
#else
m_freem(mcopy);
#endif
}
m_freem(m);
return (NULL);
case IPSEC_POLICY_BYPASS:
case IPSEC_POLICY_NONE:
/* no need to do IPsec. */
key_freesp(sp, KEY_SADB_UNLOCKED);
goto skip_ipsec;
case IPSEC_POLICY_IPSEC:
if (sp->req == NULL) {
/* XXX should be panic ? */
printf("ip6_forward: No IPsec request specified.\n");
ip6stat.ip6s_cantforward++;
key_freesp(sp, KEY_SADB_UNLOCKED);
if (mcopy) {
#if 0
/* XXX: what icmp ? */
#else
m_freem(mcopy);
#endif
}
m_freem(m);
return (NULL);
}
/* do IPsec */
break;
case IPSEC_POLICY_ENTRUST:
default:
/* should be panic ?? */
printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
key_freesp(sp, KEY_SADB_UNLOCKED);
goto skip_ipsec;
}
{
struct ipsec_output_state state;
/*
* All the extension headers will become inaccessible
* (since they can be encrypted).
* Don't panic, we need no more updates to extension headers
* on inner IPv6 packet (since they are now encapsulated).
*
* IPv6 [ESP|AH] IPv6 [extension headers] payload
*/
bzero(&state, sizeof(state));
state.m = m;
state.dst = NULL; /* update at ipsec6_output_tunnel() */
error = ipsec6_output_tunnel(&state, sp, 0);
key_freesp(sp, KEY_SADB_UNLOCKED);
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) {
case EHOSTUNREACH:
case ENETUNREACH:
case EMSGSIZE:
case ENOBUFS:
case ENOMEM:
break;
default:
printf("ip6_output (ipsec): error code %d\n", error);
/* fall through */
case ENOENT:
/* don't show these error codes to the user */
break;
}
ip6stat.ip6s_cantforward++;
if (mcopy) {
#if 0
/* XXX: what icmp ? */
#else
m_freem(mcopy);
#endif
}
m_freem(m);
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 (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_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 (rt == NULL) {
ip6stat.ip6s_noroute++;
in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
if (mcopy)
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_NOROUTE, 0);
m_freem(m);
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) {
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_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)
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_NOROUTE, 0);
m_freem(m);
return (NULL);
}
RT_LOCK(rt);
/* Take an extra ref for ourselves */
RT_ADDREF_LOCKED(rt);
}
/*
* 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]
*/
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);
if (ip6_log_time + ip6_log_interval < timenow.tv_sec) {
ip6_log_time = timenow.tv_sec;
log(LOG_DEBUG,
"cannot forward "
"src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
ip6_sprintf(&ip6->ip6_src),
ip6_sprintf(&ip6->ip6_dst),
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) {
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
}
m_freem(m);
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) {
uint32_t mtu;
#if IPSEC
struct secpolicy *sp2;
int ipsecerror;
size_t ipsechdrsiz;
#endif
mtu = rt->rt_ifp->if_mtu;
#if IPSEC
/*
* When we do IPsec tunnel ingress, we need to play
* 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".
*/
sp2 = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
IP_FORWARDING, &ipsecerror);
if (sp2) {
ipsechdrsiz = ipsec6_hdrsiz(mcopy,
IPSEC_DIR_OUTBOUND, NULL);
if (ipsechdrsiz < mtu)
mtu -= ipsechdrsiz;
key_freesp(sp2, KEY_SADB_UNLOCKED);
}
/*
* if mtu becomes less than minimum MTU,
* tell minimum MTU (and I'll need to fragment it).
*/
if (mtu < IPV6_MMTU)
mtu = IPV6_MMTU;
#endif
/* Release extra ref */
RT_REMREF_LOCKED(rt);
RT_UNLOCK(rt);
icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
} else {
/* Release extra ref */
RT_REMREF_LOCKED(rt);
RT_UNLOCK(rt);
}
m_freem(m);
return (NULL);
}
if (rt->rt_flags & RTF_GATEWAY)
dst = (struct sockaddr_in6 *)(void *)rt->rt_gateway;
/*
* If we are to forward the packet using the same interface
* as one we got the packet from, perhaps we should send a redirect
* to sender to shortcut a hop.
* Only send redirect if source is sending directly to us,
* and if packet was not source routed (or has any options).
* Also, don't send redirect if forwarding using a route
* modified by a redirect.
*/
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) {
/*
* If the incoming interface is equal to the outgoing
* one, and the link attached to the interface is
* point-to-point, then it will be highly probable
* that a routing loop occurs. Thus, we immediately
* drop the packet and send an ICMPv6 error message.
*
* type/code is based on suggestion by Rich Draves.
* not sure if it is the best pick.
*/
RT_REMREF_LOCKED(rt); /* Release extra ref */
RT_UNLOCK(rt);
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_ADDR, 0);
m_freem(m);
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, ifp, &port, &m)) {
m_freem(m);
goto freecopy;
}
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
* destinaion can appear, if the originating node just sends the
* packet to us (without address resolution for the destination).
* Since both icmp6_error and icmp6_redirect_output fill the embedded
* link identifiers, we can do this stuff after making a copy for
* returning an error.
*/
if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
/*
* See corresponding comments in ip6_output.
* XXX: but is it possible that ip6_forward() sends a packet
* to a loopback interface? I don't think so, and thus
* I bark here. ([email protected])
* XXX: it is common to route invalid packets to loopback.
* also, the codepath will be visited on use of ::1 in
* rthdr. (itojun)
*/
#if 1
if (0)
#else
if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
#endif
{
printf("ip6_forward: outgoing interface is loopback. "
"src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
ip6_sprintf(&ip6->ip6_src),
ip6_sprintf(&ip6->ip6_dst),
ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
if_name(rt->rt_ifp));
}
/* we can just use rcvif in forwarding. */
origifp = m->m_pkthdr.rcvif;
}
else
origifp = rt->rt_ifp;
/*
* 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);
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(ifp, ifs6_out_discard);
ip6stat.ip6s_cantforward++;
} else {
ip6stat.ip6s_forward++;
in6_ifstat_inc(ifp, ifs6_out_forward);
if (type)
ip6stat.ip6s_redirectsent++;
else {
if (mcopy) {
goto freecopy;
}
}
}
#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);
/* Release extra ref */
RT_REMREF(rt);
return (NULL);
}
#endif
goto freecopy;
case EMSGSIZE:
/* xxx MTU is constant in PPP? */
goto freecopy;
case ENOBUFS:
/* Tell source to slow down like source quench in IP? */
goto freecopy;
case ENETUNREACH: /* shouldn't happen, checked above */
case EHOSTUNREACH:
case ENETDOWN:
case EHOSTDOWN:
default:
type = ICMP6_DST_UNREACH;
code = ICMP6_DST_UNREACH_ADDR;
break;
}
icmp6_error(mcopy, type, code, 0);
/* Release extra ref */
RT_REMREF(rt);
return (NULL);
freecopy:
m_freem(mcopy);
/* Release extra ref */
RT_REMREF(rt);
return (NULL);
}
void
ip6_forward(struct mbuf *m, int srcrt)
{
struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
struct sockaddr_in6 *dst;
struct rtentry *rt;
int error = 0, type = 0, code = 0;
struct mbuf *mcopy = NULL;
struct ifnet *origifp; /* maybe unnecessary */
#ifdef IPSEC
u_int8_t sproto = 0;
struct m_tag *mtag;
union sockaddr_union sdst;
struct tdb_ident *tdbi;
u_int32_t sspi;
struct tdb *tdb;
int s;
#if NPF > 0
struct ifnet *encif;
#endif
#endif /* IPSEC */
u_int rtableid = 0;
/*
* Do not forward packets to multicast destination (should be handled
* by ip6_mforward().
* Do not forward packets with unspecified source. It was discussed
* in July 2000, on ipngwg mailing list.
*/
if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
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 < time_second) {
ip6_log_time = time_second;
log(LOG_DEBUG,
"cannot forward "
"from %s to %s nxt %d received on %s\n",
ip6_sprintf(&ip6->ip6_src),
ip6_sprintf(&ip6->ip6_dst),
ip6->ip6_nxt,
m->m_pkthdr.rcvif->if_xname);
}
m_freem(m);
return;
}
if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
icmp6_error(m, ICMP6_TIME_EXCEEDED,
ICMP6_TIME_EXCEED_TRANSIT, 0);
return;
}
ip6->ip6_hlim -= IPV6_HLIMDEC;
#if NPF > 0
reroute:
#endif
#ifdef IPSEC
if (!ipsec_in_use)
goto done_spd;
s = splnet();
/*
* Check if there was an outgoing SA bound to the flow
* from a transport protocol.
*/
/* Do we have any pending SAs to apply ? */
mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
if (mtag != NULL) {
#ifdef DIAGNOSTIC
if (mtag->m_tag_len != sizeof (struct tdb_ident))
panic("ip6_forward: tag of length %d (should be %d",
mtag->m_tag_len, sizeof (struct tdb_ident));
#endif
tdbi = (struct tdb_ident *)(mtag + 1);
tdb = gettdb(tdbi->rdomain, tdbi->spi, &tdbi->dst,
tdbi->proto);
if (tdb == NULL)
error = -EINVAL;
m_tag_delete(m, mtag);
} else
tdb = ipsp_spd_lookup(m, AF_INET6, sizeof(struct ip6_hdr),
&error, IPSP_DIRECTION_OUT, NULL, NULL);
if (tdb == NULL) {
splx(s);
if (error == 0) {
/*
* No IPsec processing required, we'll just send the
* packet out.
*/
sproto = 0;
/* Fall through to routing/multicast handling */
} else {
/*
* -EINVAL is used to indicate that the packet should
* be silently dropped, typically because we've asked
* key management for an SA.
*/
if (error == -EINVAL) /* Should silently drop packet */
error = 0;
goto freecopy;
}
} else {
/* Loop detection */
for (mtag = m_tag_first(m); mtag != NULL;
mtag = m_tag_next(m, mtag)) {
if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
mtag->m_tag_id !=
PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
continue;
tdbi = (struct tdb_ident *)(mtag + 1);
if (tdbi->spi == tdb->tdb_spi &&
tdbi->proto == tdb->tdb_sproto &&
tdbi->rdomain == tdb->tdb_rdomain &&
!bcmp(&tdbi->dst, &tdb->tdb_dst,
sizeof(union sockaddr_union))) {
splx(s);
sproto = 0; /* mark as no-IPsec-needed */
goto done_spd;
}
}
/* We need to do IPsec */
bcopy(&tdb->tdb_dst, &sdst, sizeof(sdst));
sspi = tdb->tdb_spi;
sproto = tdb->tdb_sproto;
splx(s);
}
/* Fall through to the routing/multicast handling code */
done_spd:
#endif /* IPSEC */
#if NPF > 0
rtableid = m->m_pkthdr.rdomain;
#endif
/*
* Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
* size of IPv6 + ICMPv6 headers) bytes of the packet in case
* we need to generate an ICMP6 message to the src.
* Thanks to M_EXT, in most cases copy will not occur.
*
* It is important to save it before IPsec processing as IPsec
* processing may modify the mbuf.
*/
mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
dst = &ip6_forward_rt.ro_dst;
if (!srcrt) {
/*
* ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst
*/
if (ip6_forward_rt.ro_rt == 0 ||
(ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0 ||
ip6_forward_rt.ro_tableid != rtableid) {
if (ip6_forward_rt.ro_rt) {
RTFREE(ip6_forward_rt.ro_rt);
ip6_forward_rt.ro_rt = 0;
}
/* this probably fails but give it a try again */
ip6_forward_rt.ro_tableid = rtableid;
rtalloc_mpath((struct route *)&ip6_forward_rt,
&ip6->ip6_src.s6_addr32[0]);
}
if (ip6_forward_rt.ro_rt == 0) {
ip6stat.ip6s_noroute++;
/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
if (mcopy) {
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_NOROUTE, 0);
}
m_freem(m);
return;
}
} else if (ip6_forward_rt.ro_rt == 0 ||
(ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0 ||
!IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr) ||
ip6_forward_rt.ro_tableid != rtableid) {
if (ip6_forward_rt.ro_rt) {
RTFREE(ip6_forward_rt.ro_rt);
ip6_forward_rt.ro_rt = 0;
}
bzero(dst, sizeof(*dst));
dst->sin6_len = sizeof(struct sockaddr_in6);
dst->sin6_family = AF_INET6;
dst->sin6_addr = ip6->ip6_dst;
ip6_forward_rt.ro_tableid = rtableid;
rtalloc_mpath((struct route *)&ip6_forward_rt,
&ip6->ip6_src.s6_addr32[0]);
if (ip6_forward_rt.ro_rt == 0) {
ip6stat.ip6s_noroute++;
/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
if (mcopy) {
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_NOROUTE, 0);
}
m_freem(m);
return;
}
}
rt = ip6_forward_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-v3-00.txt, Section 3.1]
*/
if (in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_src) !=
in6_addr2scopeid(rt->rt_ifp, &ip6->ip6_src)) {
ip6stat.ip6s_cantforward++;
ip6stat.ip6s_badscope++;
in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
if (ip6_log_time + ip6_log_interval < time_second) {
ip6_log_time = time_second;
log(LOG_DEBUG,
"cannot forward "
"src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
ip6_sprintf(&ip6->ip6_src),
ip6_sprintf(&ip6->ip6_dst),
ip6->ip6_nxt,
m->m_pkthdr.rcvif->if_xname, rt->rt_ifp->if_xname);
}
if (mcopy)
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
m_freem(m);
goto freert;
}
#ifdef IPSEC
/*
* Check if the packet needs encapsulation.
* ipsp_process_packet will never come back to here.
* XXX ipsp_process_packet() calls ip6_output(), and there'll be no
* PMTU notification. is it okay?
*/
if (sproto != 0) {
s = splnet();
tdb = gettdb(rtable_l2(m->m_pkthdr.rdomain),
sspi, &sdst, sproto);
if (tdb == NULL) {
splx(s);
error = EHOSTUNREACH;
m_freem(m);
goto senderr; /*XXX*/
}
#if NPF > 0
if ((encif = enc_getif(tdb->tdb_rdomain,
tdb->tdb_tap)) == NULL ||
pf_test6(PF_FWD, encif, &m, NULL) != PF_PASS) {
splx(s);
error = EHOSTUNREACH;
m_freem(m);
goto senderr;
}
if (m == NULL) {
splx(s);
goto senderr;
}
ip6 = mtod(m, struct ip6_hdr *);
/*
* PF_TAG_REROUTE handling or not...
* Packet is entering IPsec so the routing is
* already overruled by the IPsec policy.
* Until now the change was not reconsidered.
* What's the behaviour?
*/
#endif
m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
/* Callee frees mbuf */
error = ipsp_process_packet(m, tdb, AF_INET6, 0);
splx(s);
m_freem(mcopy);
goto freert;
}
void
ip6_forward(struct mbuf *m, int srcrt)
{
struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
const struct sockaddr_in6 *dst;
struct rtentry *rt;
int error = 0, type = 0, code = 0;
struct mbuf *mcopy = NULL;
struct ifnet *origifp; /* maybe unnecessary */
uint32_t inzone, outzone;
struct in6_addr src_in6, dst_in6;
#ifdef IPSEC
int needipsec = 0;
struct secpolicy *sp = NULL;
#endif
/*
* Clear any in-bound checksum flags for this packet.
*/
m->m_pkthdr.csum_flags = 0;
/*
* Do not forward packets to multicast destination (should be handled
* by ip6_mforward().
* Do not forward packets with unspecified source. It was discussed
* in July 2000, on ipngwg mailing list.
*/
if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
ip6_cantforward(ip6, m->m_pkthdr.rcvif, NULL,
((m->m_flags & (M_BCAST|M_MCAST)) != 0) ? "bcast/mcast" :
IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ? "mcast/dst" :
"unspec/src");
m_freem(m);
return;
}
if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
icmp6_error(m, ICMP6_TIME_EXCEEDED,
ICMP6_TIME_EXCEED_TRANSIT, 0);
return;
}
ip6->ip6_hlim -= IPV6_HLIMDEC;
/*
* Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
* size of IPv6 + ICMPv6 headers) bytes of the packet in case
* we need to generate an ICMP6 message to the src.
* Thanks to M_EXT, in most cases copy will not occur.
*
* It is important to save it before IPsec processing as IPsec
* processing may modify the mbuf.
*/
mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
#ifdef IPSEC
if (ipsec_used) {
/* Check the security policy (SP) for the packet */
sp = ipsec6_check_policy(m, NULL, 0, &needipsec, &error);
if (error != 0) {
/*
* Hack: -EINVAL is used to signal that a packet
* should be silently discarded. This is typically
* because we asked key management for an SA and
* it was delayed (e.g. kicked up to IKE).
*/
if (error == -EINVAL)
error = 0;
goto freecopy;
}
}
#endif /* IPSEC */
if (srcrt) {
union {
struct sockaddr dst;
struct sockaddr_in6 dst6;
} u;
sockaddr_in6_init(&u.dst6, &ip6->ip6_dst, 0, 0, 0);
if ((rt = rtcache_lookup(&ip6_forward_rt, &u.dst)) == NULL) {
IP6_STATINC(IP6_STAT_NOROUTE);
/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
if (mcopy) {
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_NOROUTE, 0);
}
m_freem(m);
return;
}
} else if ((rt = rtcache_validate(&ip6_forward_rt)) == NULL &&
(rt = rtcache_update(&ip6_forward_rt, 1)) == NULL) {
/*
* rtcache_getdst(ip6_forward_rt)->sin6_addr was equal to
* ip6->ip6_dst
*/
IP6_STATINC(IP6_STAT_NOROUTE);
/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
if (mcopy) {
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_NOROUTE, 0);
}
m_freem(m);
return;
}
dst = satocsin6(rtcache_getdst(&ip6_forward_rt));
/*
* 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). We use a local copy of ip6_src, since in6_setscope()
* will possibly modify its first argument.
* [draft-ietf-ipngwg-icmp-v3-07, Section 3.1]
*/
src_in6 = ip6->ip6_src;
inzone = outzone = ~0;
if (in6_setscope(&src_in6, rt->rt_ifp, &outzone) != 0 ||
in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
inzone != outzone) {
ip6_cantforward(ip6, m->m_pkthdr.rcvif, rt->rt_ifp,
"src[%s] inzone %d outzone %d",
in6_getscopename(&ip6->ip6_src), inzone, outzone);
if (mcopy)
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
m_freem(m);
return;
}
#ifdef IPSEC
/*
* If we need to encapsulate the packet, do it here
* ipsec6_proces_packet will send the packet using ip6_output
*/
if (needipsec) {
int s = splsoftnet();
error = ipsec6_process_packet(m, sp->req);
splx(s);
if (mcopy)
goto freecopy;
}
#endif
/*
* 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;
inzone = outzone = ~0;
if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
inzone != outzone) {
ip6_cantforward(ip6, m->m_pkthdr.rcvif, rt->rt_ifp,
"dst[%s] inzone %d outzone %d",
in6_getscopename(&ip6->ip6_dst), inzone, outzone);
if (mcopy)
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
m_freem(m);
return;
}
if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
if (mcopy) {
u_long mtu;
mtu = IN6_LINKMTU(rt->rt_ifp);
icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
}
m_freem(m);
return;
}
if (rt->rt_flags & RTF_GATEWAY)
dst = (struct sockaddr_in6 *)rt->rt_gateway;
/*
* If we are to forward the packet using the same interface
* as one we got the packet from, perhaps we should send a redirect
* to sender to shortcut a hop.
* Only send redirect if source is sending directly to us,
* and if packet was not source routed (or has any options).
* Also, don't send redirect if forwarding using a route
* modified by a redirect.
*/
if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && ip6_sendredirects &&
(rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) &&
nd6_is_addr_neighbor(
satocsin6(rtcache_getdst(&ip6_forward_rt)),
rt->rt_ifp)) {
/*
* If the incoming interface is equal to the outgoing
* one, the link attached to the interface is
* point-to-point, and the IPv6 destination is
* regarded as on-link on the link, then it will be
* highly probable that the destination address does
* not exist on the link and that the packet is going
* to loop. Thus, we immediately drop the packet and
* send an ICMPv6 error message.
* For other routing loops, we dare to let the packet
* go to the loop, so that a remote diagnosing host
* can detect the loop by traceroute.
* type/code is based on suggestion by Rich Draves.
* not sure if it is the best pick.
*/
icmp6_error(mcopy, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_ADDR, 0);
m_freem(m);
return;
}
type = ND_REDIRECT;
}
/*
* Fake scoped addresses. Note that even link-local source or
* destinaion can appear, if the originating node just sends the
* packet to us (without address resolution for the destination).
* Since both icmp6_error and icmp6_redirect_output fill the embedded
* link identifiers, we can do this stuff after making a copy for
* returning an error.
*/
if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
/*
* See corresponding comments in ip6_output.
* XXX: but is it possible that ip6_forward() sends a packet
* to a loopback interface? I don't think so, and thus
* I bark here. ([email protected])
* XXX: it is common to route invalid packets to loopback.
* also, the codepath will be visited on use of ::1 in
* rthdr. (itojun)
*/
#if 1
if (0)
#else
if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
#endif
{
printf("ip6_forward: outgoing interface is loopback. "
"src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
ip6_sprintf(&ip6->ip6_src),
ip6_sprintf(&ip6->ip6_dst),
ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
if_name(rt->rt_ifp));
}
/* we can just use rcvif in forwarding. */
origifp = m->m_pkthdr.rcvif;
}
else
origifp = rt->rt_ifp;
/*
* 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);
/*
* Run through list of hooks for output packets.
*/
if ((error = pfil_run_hooks(inet6_pfil_hook, &m, rt->rt_ifp,
PFIL_OUT)) != 0)
goto senderr;
if (m == NULL)
goto freecopy;
ip6 = mtod(m, struct ip6_hdr *);
error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
if (error) {
in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
IP6_STATINC(IP6_STAT_CANTFORWARD);
} else {
IP6_STATINC(IP6_STAT_FORWARD);
in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
if (type)
IP6_STATINC(IP6_STAT_REDIRECTSENT);
else {
#ifdef GATEWAY
if (m->m_flags & M_CANFASTFWD)
ip6flow_create(&ip6_forward_rt, m);
#endif
if (mcopy)
goto freecopy;
}
}
senderr:
if (mcopy == NULL)
return;
switch (error) {
case 0:
if (type == ND_REDIRECT) {
icmp6_redirect_output(mcopy, rt);
return;
}
goto freecopy;
case EMSGSIZE:
/* xxx MTU is constant in PPP? */
goto freecopy;
case ENOBUFS:
/* Tell source to slow down like source quench in IP? */
goto freecopy;
case ENETUNREACH: /* shouldn't happen, checked above */
case EHOSTUNREACH:
case ENETDOWN:
case EHOSTDOWN:
default:
type = ICMP6_DST_UNREACH;
code = ICMP6_DST_UNREACH_ADDR;
break;
}
icmp6_error(mcopy, type, code, 0);
return;
freecopy:
m_freem(mcopy);
return;
}
struct mbuf*
ip6_tryforward(struct mbuf *m)
{
struct sockaddr_in6 dst;
struct nhop6_basic nh;
struct m_tag *fwd_tag;
struct ip6_hdr *ip6;
struct ifnet *rcvif;
uint32_t plen;
int error;
/*
* Fallback conditions to ip6_input for slow path processing.
*/
ip6 = mtod(m, struct ip6_hdr *);
if (ip6->ip6_nxt == IPPROTO_HOPOPTS ||
IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst) ||
IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src) ||
IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) ||
in6_localip(&ip6->ip6_dst))
return (m);
/*
* Check that the amount of data in the buffers
* is as at least much as the IPv6 header would have us expect.
* Trim mbufs if longer than we expect.
* Drop packet if shorter than we expect.
*/
rcvif = m->m_pkthdr.rcvif;
plen = ntohs(ip6->ip6_plen);
if (plen == 0) {
/*
* Jumbograms must have hop-by-hop header and go via
* slow path.
*/
IP6STAT_INC(ip6s_badoptions);
goto dropin;
}
if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
IP6STAT_INC(ip6s_tooshort);
in6_ifstat_inc(rcvif, ifs6_in_truncated);
goto dropin;
}
if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
if (m->m_len == m->m_pkthdr.len) {
m->m_len = sizeof(struct ip6_hdr) + plen;
m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
} else
m_adj(m, sizeof(struct ip6_hdr) + plen -
m->m_pkthdr.len);
}
/*
* Hop limit.
*/
#ifdef IPSTEALTH
if (!V_ip6stealth)
#endif
if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
icmp6_error(m, ICMP6_TIME_EXCEEDED,
ICMP6_TIME_EXCEED_TRANSIT, 0);
m = NULL;
goto dropin;
}
bzero(&dst, sizeof(dst));
dst.sin6_family = AF_INET6;
dst.sin6_len = sizeof(dst);
dst.sin6_addr = ip6->ip6_dst;
/*
* Incoming packet firewall processing.
*/
if (!PFIL_HOOKED(&V_inet6_pfil_hook))
goto passin;
if (pfil_run_hooks(&V_inet6_pfil_hook, &m, rcvif, PFIL_IN,
NULL) != 0 || m == NULL)
goto dropin;
/*
* If packet filter sets the M_FASTFWD_OURS flag, this means
* that new destination or next hop is our local address.
* So, we can just go back to ip6_input.
* XXX: should we decrement ip6_hlim in such case?
*
* Also it can forward packet to another destination, e.g.
* M_IP6_NEXTHOP flag is set and fwd_tag is attached to mbuf.
*/
if (m->m_flags & M_FASTFWD_OURS)
return (m);
ip6 = mtod(m, struct ip6_hdr *);
if ((m->m_flags & M_IP6_NEXTHOP) &&
(fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
/*
* Now we will find route to forwarded by pfil destination.
*/
bcopy((fwd_tag + 1), &dst, sizeof(dst));
m->m_flags &= ~M_IP6_NEXTHOP;
m_tag_delete(m, fwd_tag);
} else {
/* Update dst since pfil could change it */
dst.sin6_addr = ip6->ip6_dst;
}
passin:
/*
* Find route to destination.
*/
if (ip6_findroute(&nh, &dst, m) != 0) {
m = NULL;
in6_ifstat_inc(rcvif, ifs6_in_noroute);
goto dropin;
}
/*
* We used slow path processing for packets with scoped addresses.
* So, scope checks aren't needed here.
*/
if (m->m_pkthdr.len > nh.nh_mtu) {
in6_ifstat_inc(nh.nh_ifp, ifs6_in_toobig);
icmp6_error(m, ICMP6_PACKET_TOO_BIG, 0, nh.nh_mtu);
m = NULL;
goto dropout;
}
/*
* Outgoing packet firewall processing.
*/
if (!PFIL_HOOKED(&V_inet6_pfil_hook))
goto passout;
if (pfil_run_hooks(&V_inet6_pfil_hook, &m, nh.nh_ifp, PFIL_OUT,
NULL) != 0 || m == NULL)
goto dropout;
/*
* If packet filter sets the M_FASTFWD_OURS flag, this means
* that new destination or next hop is our local address.
* So, we can just go back to ip6_input.
*
* Also it can forward packet to another destination, e.g.
* M_IP6_NEXTHOP flag is set and fwd_tag is attached to mbuf.
*/
if (m->m_flags & M_FASTFWD_OURS) {
/*
* XXX: we did one hop and should decrement hop limit. But
* now we are the destination and just don't pay attention.
*/
return (m);
}
/*
* Again. A packet filter could change the destination address.
*/
ip6 = mtod(m, struct ip6_hdr *);
if (m->m_flags & M_IP6_NEXTHOP)
fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
else
fwd_tag = NULL;
if (fwd_tag != NULL ||
!IN6_ARE_ADDR_EQUAL(&dst.sin6_addr, &ip6->ip6_dst)) {
if (fwd_tag != NULL) {
bcopy((fwd_tag + 1), &dst, sizeof(dst));
m->m_flags &= ~M_IP6_NEXTHOP;
m_tag_delete(m, fwd_tag);
} else
dst.sin6_addr = ip6->ip6_dst;
/*
* Redo route lookup with new destination address
*/
if (ip6_findroute(&nh, &dst, m) != 0) {
m = NULL;
goto dropout;
}
}
passout:
#ifdef IPSTEALTH
if (!V_ip6stealth)
#endif
{
ip6->ip6_hlim -= IPV6_HLIMDEC;
}
m_clrprotoflags(m); /* Avoid confusing lower layers. */
IP_PROBE(send, NULL, NULL, ip6, nh.nh_ifp, NULL, ip6);
/*
* XXX: we need to use destination address with embedded scope
* zone id, because LLTABLE uses such form of addresses for lookup.
*/
dst.sin6_addr = nh.nh_addr;
if (IN6_IS_SCOPE_LINKLOCAL(&dst.sin6_addr))
dst.sin6_addr.s6_addr16[1] = htons(nh.nh_ifp->if_index & 0xffff);
error = (*nh.nh_ifp->if_output)(nh.nh_ifp, m,
(struct sockaddr *)&dst, NULL);
if (error != 0) {
in6_ifstat_inc(nh.nh_ifp, ifs6_out_discard);
IP6STAT_INC(ip6s_cantforward);
} else {
in6_ifstat_inc(nh.nh_ifp, ifs6_out_forward);
IP6STAT_INC(ip6s_forward);
}
return (NULL);
dropin:
in6_ifstat_inc(rcvif, ifs6_in_discard);
goto drop;
dropout:
in6_ifstat_inc(nh.nh_ifp, ifs6_out_discard);
drop:
if (m != NULL)
m_freem(m);
return (NULL);
}
/*
* Fragment input
* NOTE: this function is called with the inet6_domain_mutex held from ip6_input.
* inet6_domain_mutex is protecting he frag6 queue manipulation.
*/
int
frag6_input(struct mbuf **mp, int *offp, int proto)
{
#pragma unused(proto)
struct mbuf *m = *mp, *t;
struct ip6_hdr *ip6;
struct ip6_frag *ip6f;
struct ip6q *q6;
struct ip6asfrag *af6, *ip6af, *af6dwn;
int offset = *offp, nxt, i, next;
int first_frag = 0;
int fragoff, frgpartlen; /* must be larger than u_int16_t */
struct ifnet *dstifp;
struct ifaddr *ifa = NULL;
u_int8_t ecn, ecn0;
#ifdef IN6_IFSTAT_STRICT
struct route_in6 ro;
struct sockaddr_in6 *dst;
#endif
/* Expect 32-bit aligned data pointer on strict-align platforms */
MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
ip6 = mtod(m, struct ip6_hdr *);
#ifndef PULLDOWN_TEST
IP6_EXTHDR_CHECK(m, offset, sizeof(struct ip6_frag), return IPPROTO_DONE);
ip6f = (struct ip6_frag *)((caddr_t)ip6 + offset);
#else
IP6_EXTHDR_GET(ip6f, struct ip6_frag *, m, offset, sizeof(*ip6f));
if (ip6f == NULL)
return IPPROTO_DONE;
#endif
dstifp = NULL;
#ifdef IN6_IFSTAT_STRICT
/* find the destination interface of the packet. */
bzero(&ro, sizeof (ro));
dst = (struct sockaddr_in6 *)&ro.ro_dst;
dst->sin6_family = AF_INET6;
dst->sin6_len = sizeof (struct sockaddr_in6);
dst->sin6_addr = ip6->ip6_dst;
rtalloc((struct route *)&ro);
if (ro.ro_rt != NULL) {
RT_LOCK(ro.ro_rt);
if ((ifa = ro.ro_rt->rt_ifa) != NULL) {
IFA_ADDREF(ifa);
dstifp = ((struct in6_ifaddr *)ro.ro_rt->rt_ifa)->ia_ifp;
}
RT_UNLOCK(ro.ro_rt);
rtfree(ro.ro_rt);
ro.ro_rt = NULL;
}
#else
/* we are violating the spec, this is not the destination interface */
if ((m->m_flags & M_PKTHDR) != 0)
dstifp = m->m_pkthdr.rcvif;
#endif
/* jumbo payload can't contain a fragment header */
if (ip6->ip6_plen == 0) {
icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, offset);
in6_ifstat_inc(dstifp, ifs6_reass_fail);
if (ifa != NULL)
IFA_REMREF(ifa);
return IPPROTO_DONE;
}
/*
* check whether fragment packet's fragment length is
* multiple of 8 octets.
* sizeof(struct ip6_frag) == 8
* sizeof(struct ip6_hdr) = 40
*/
if ((ip6f->ip6f_offlg & IP6F_MORE_FRAG) &&
(((ntohs(ip6->ip6_plen) - offset) & 0x7) != 0)) {
icmp6_error(m, ICMP6_PARAM_PROB,
ICMP6_PARAMPROB_HEADER,
offsetof(struct ip6_hdr, ip6_plen));
in6_ifstat_inc(dstifp, ifs6_reass_fail);
if (ifa != NULL)
IFA_REMREF(ifa);
return IPPROTO_DONE;
}
ip6stat.ip6s_fragments++;
in6_ifstat_inc(dstifp, ifs6_reass_reqd);
/* offset now points to data portion */
offset += sizeof(struct ip6_frag);
frag6_doing_reass = 1;
/*
* Enforce upper bound on number of fragments.
* If maxfrag is 0, never accept fragments.
* If maxfrag is -1, accept all fragments without limitation.
*/
if (ip6_maxfrags < 0)
;
else if (frag6_nfrags >= (u_int)ip6_maxfrags)
goto dropfrag;
for (q6 = ip6q.ip6q_next; q6 != &ip6q; q6 = q6->ip6q_next)
if (ip6f->ip6f_ident == q6->ip6q_ident &&
IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &q6->ip6q_src) &&
IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &q6->ip6q_dst))
break;
if (q6 == &ip6q) {
/*
* the first fragment to arrive, create a reassembly queue.
*/
first_frag = 1;
/*
* Enforce upper bound on number of fragmented packets
* for which we attempt reassembly;
* If maxfrag is 0, never accept fragments.
* If maxfrag is -1, accept all fragments without limitation.
*/
if (ip6_maxfragpackets < 0)
;
else if (frag6_nfragpackets >= (u_int)ip6_maxfragpackets)
goto dropfrag;
frag6_nfragpackets++;
q6 = (struct ip6q *)_MALLOC(sizeof(struct ip6q), M_FTABLE,
M_DONTWAIT);
if (q6 == NULL)
goto dropfrag;
bzero(q6, sizeof(*q6));
frag6_insque(q6, &ip6q);
/* ip6q_nxt will be filled afterwards, from 1st fragment */
q6->ip6q_down = q6->ip6q_up = (struct ip6asfrag *)q6;
#ifdef notyet
q6->ip6q_nxtp = (u_char *)nxtp;
#endif
q6->ip6q_ident = ip6f->ip6f_ident;
q6->ip6q_ttl = IPV6_FRAGTTL;
q6->ip6q_src = ip6->ip6_src;
q6->ip6q_dst = ip6->ip6_dst;
q6->ip6q_ecn =
(ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK;
q6->ip6q_unfrglen = -1; /* The 1st fragment has not arrived. */
q6->ip6q_nfrag = 0;
}
