/*
* Check for alternatives when higher level complains
* about service problems. For now, invalidate cached
* routing information. If the route was created dynamically
* (by a redirect), time to try a default gateway again.
*/
void
in6_losing(struct in6pcb *in6p)
{
struct rtentry *rt;
struct rt_addrinfo info;
if (in6p->in6p_af != AF_INET6)
return;
if ((rt = rtcache_validate(&in6p->in6p_route)) == NULL)
return;
memset(&info, 0, sizeof(info));
info.rti_info[RTAX_DST] = rtcache_getdst(&in6p->in6p_route);
info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
info.rti_info[RTAX_NETMASK] = rt_mask(rt);
rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
if (rt->rt_flags & RTF_DYNAMIC) {
(void)rtrequest(RTM_DELETE, rt_getkey(rt),
rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
}
/*
* A new route can be allocated
* the next time output is attempted.
*/
rtcache_free(&in6p->in6p_route);
}
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;
}
/*
* Pass some notification to all connections of a protocol
* associated with address dst. The local address and/or port numbers
* may be specified to limit the search. The "usual action" will be
* taken, depending on the ctlinput cmd. The caller must filter any
* cmds that are uninteresting (e.g., no error in the map).
* Call the protocol specific routine (if any) to report
* any errors for each matching socket.
*
* Must be called at splsoftnet.
*
* Note: src (4th arg) carries the flowlabel value on the original IPv6
* header, in sin6_flowinfo member.
*/
int
in6_pcbnotify(struct inpcbtable *table, const struct sockaddr *dst,
u_int fport_arg, const struct sockaddr *src, u_int lport_arg, int cmd,
void *cmdarg, void (*notify)(struct in6pcb *, int))
{
struct rtentry *rt;
struct in6pcb *in6p, *nin6p;
struct sockaddr_in6 sa6_src;
const struct sockaddr_in6 *sa6_dst;
u_int16_t fport = fport_arg, lport = lport_arg;
int errno;
int nmatch = 0;
u_int32_t flowinfo;
if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
return 0;
sa6_dst = (const struct sockaddr_in6 *)dst;
if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
return 0;
/*
* note that src can be NULL when we get notify by local fragmentation.
*/
sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
flowinfo = sa6_src.sin6_flowinfo;
/*
* Redirects go to all references to the destination,
* and use in6_rtchange to invalidate the route cache.
* Dead host indications: also use in6_rtchange to invalidate
* the cache, and deliver the error to all the sockets.
* Otherwise, if we have knowledge of the local port and address,
* deliver only to that socket.
*/
if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
fport = 0;
lport = 0;
bzero((void *)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
if (cmd != PRC_HOSTDEAD)
notify = in6_rtchange;
}
errno = inet6ctlerrmap[cmd];
for (in6p = (struct in6pcb *)CIRCLEQ_FIRST(&table->inpt_queue);
in6p != (void *)&table->inpt_queue;
in6p = nin6p) {
nin6p = (struct in6pcb *)CIRCLEQ_NEXT(in6p, in6p_queue);
if (in6p->in6p_af != AF_INET6)
continue;
/*
* Under the following condition, notify of redirects
* to the pcb, without making address matches against inpcb.
* - redirect notification is arrived.
* - the inpcb is unconnected.
* - the inpcb is caching !RTF_HOST routing entry.
* - the ICMPv6 notification is from the gateway cached in the
* inpcb. i.e. ICMPv6 notification is from nexthop gateway
* the inpcb used very recently.
*
* This is to improve interaction between netbsd/openbsd
* redirect handling code, and inpcb route cache code.
* without the clause, !RTF_HOST routing entry (which carries
* gateway used by inpcb right before the ICMPv6 redirect)
* will be cached forever in unconnected inpcb.
*
* There still is a question regarding to what is TRT:
* - On bsdi/freebsd, RTF_HOST (cloned) routing entry will be
* generated on packet output. inpcb will always cache
* RTF_HOST routing entry so there's no need for the clause
* (ICMPv6 redirect will update RTF_HOST routing entry,
* and inpcb is caching it already).
* However, bsdi/freebsd are vulnerable to local DoS attacks
* due to the cloned routing entries.
* - Specwise, "destination cache" is mentioned in RFC2461.
* Jinmei says that it implies bsdi/freebsd behavior, itojun
* is not really convinced.
* - Having hiwat/lowat on # of cloned host route (redirect/
* pmtud) may be a good idea. netbsd/openbsd has it. see
* icmp6_mtudisc_update().
*/
if ((PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) &&
IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
(rt = rtcache_validate(&in6p->in6p_route)) != NULL &&
!(rt->rt_flags & RTF_HOST)) {
const struct sockaddr_in6 *dst6;
dst6 = (const struct sockaddr_in6 *)
rtcache_getdst(&in6p->in6p_route);
if (dst6 == NULL)
;
else if (IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr,
&sa6_dst->sin6_addr))
goto do_notify;
}
/*
* If the error designates a new path MTU for a destination
* and the application (associated with this socket) wanted to
* know the value, notify. Note that we notify for all
* disconnected sockets if the corresponding application
* wanted. This is because some UDP applications keep sending
* sockets disconnected.
* XXX: should we avoid to notify the value to TCP sockets?
*/
if (cmd == PRC_MSGSIZE && (in6p->in6p_flags & IN6P_MTU) != 0 &&
(IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) ||
IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &sa6_dst->sin6_addr))) {
ip6_notify_pmtu(in6p, (const struct sockaddr_in6 *)dst,
(u_int32_t *)cmdarg);
}
/*
* Detect if we should notify the error. If no source and
* destination ports are specified, but non-zero flowinfo and
* local address match, notify the error. This is the case
* when the error is delivered with an encrypted buffer
* by ESP. Otherwise, just compare addresses and ports
* as usual.
*/
if (lport == 0 && fport == 0 && flowinfo &&
in6p->in6p_socket != NULL &&
flowinfo == (in6p->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &sa6_src.sin6_addr))
goto do_notify;
else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
&sa6_dst->sin6_addr) ||
in6p->in6p_socket == 0 ||
(lport && in6p->in6p_lport != lport) ||
(!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
&sa6_src.sin6_addr)) ||
(fport && in6p->in6p_fport != fport))
continue;
do_notify:
if (notify)
(*notify)(in6p, errno);
nmatch++;
}
return nmatch;
}