/*
* Given a source IPv6 address (and route, if available), determine the best
* interface to send the packet from. Checking for (and updating) the
* ROF_SRCIF_SELECTED flag in the pcb-supplied route placeholder is done
* without any locks, based on the assumption that in the event this is
* called from ip6_output(), the output operation is single-threaded per-pcb,
* i.e. for any given pcb there can only be one thread performing output at
* the IPv6 layer.
*
* This routine is analogous to in_selectsrcif() for IPv4.
*
* clone - meaningful only for bsdi and freebsd
*/
static int
selectroute(struct sockaddr_in6 *srcsock, struct sockaddr_in6 *dstsock,
struct ip6_pktopts *opts, struct ip6_moptions *mopts, struct route_in6 *ro,
struct ifnet **retifp, struct rtentry **retrt, int clone,
int norouteok, unsigned int ifscope, unsigned int nocell)
{
int error = 0;
struct ifnet *ifp = NULL;
struct route_in6 *route = NULL;
struct sockaddr_in6 *sin6_next;
struct in6_pktinfo *pi = NULL;
struct in6_addr *dst = &dstsock->sin6_addr;
struct ifaddr *ifa = NULL;
char s_src[MAX_IPv6_STR_LEN], s_dst[MAX_IPv6_STR_LEN];
boolean_t select_srcif;
#if 0
char ip6buf[INET6_ADDRSTRLEN];
if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
dstsock->sin6_addr.s6_addr32[1] == 0 &&
!IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
printf("in6_selectroute: strange destination %s\n",
ip6_sprintf(ip6buf, &dstsock->sin6_addr));
} else {
printf("in6_selectroute: destination = %s%%%d\n",
ip6_sprintf(ip6buf, &dstsock->sin6_addr),
dstsock->sin6_scope_id); /* for debug */
}
#endif
if (retifp != NULL)
*retifp = NULL;
if (retrt != NULL)
*retrt = NULL;
if (ip6_select_srcif_debug) {
struct in6_addr src;
src = (srcsock != NULL) ? srcsock->sin6_addr : in6addr_any;
(void) inet_ntop(AF_INET6, &src, s_src, sizeof (s_src));
(void) inet_ntop(AF_INET6, dst, s_dst, sizeof (s_dst));
}
/*
* If the destination address is UNSPECIFIED addr, bail out.
*/
if (IN6_IS_ADDR_UNSPECIFIED(dst)) {
error = EHOSTUNREACH;
goto done;
}
/*
* Perform source interface selection only if Scoped Routing
* is enabled and a source address that isn't unspecified.
*/
select_srcif = (ip6_doscopedroute && srcsock != NULL &&
!IN6_IS_ADDR_UNSPECIFIED(&srcsock->sin6_addr));
/*
* If Scoped Routing is disabled, ignore the given ifscope.
* Otherwise even if source selection won't be performed,
* we still obey IPV6_BOUND_IF.
*/
if (!ip6_doscopedroute && ifscope != IFSCOPE_NONE)
ifscope = IFSCOPE_NONE;
/* If the caller specified the outgoing interface explicitly, use it */
if (opts != NULL && (pi = opts->ip6po_pktinfo) != NULL &&
pi->ipi6_ifindex != 0) {
/*
* If IPV6_PKTINFO takes precedence over IPV6_BOUND_IF.
*/
ifscope = pi->ipi6_ifindex;
ifnet_head_lock_shared();
/* ifp may be NULL if detached or out of range */
ifp = (ifscope <= if_index) ? ifindex2ifnet[ifscope] : NULL;
ifnet_head_done();
if (norouteok || retrt == NULL || IN6_IS_ADDR_MULTICAST(dst)) {
/*
* We do not have to check or get the route for
* multicast. If the caller didn't ask/care for
* the route and we have no interface to use,
* it's an error.
*/
if (ifp == NULL)
error = EHOSTUNREACH;
goto done;
} else {
goto getsrcif;
}
}
/*
* If the destination address is a multicast address and the outgoing
* interface for the address is specified by the caller, use it.
*/
if (IN6_IS_ADDR_MULTICAST(dst) && mopts != NULL) {
IM6O_LOCK(mopts);
if ((ifp = mopts->im6o_multicast_ifp) != NULL) {
IM6O_UNLOCK(mopts);
goto done; /* we do not need a route for multicast. */
}
IM6O_UNLOCK(mopts);
}
getsrcif:
/*
* If the outgoing interface was not set via IPV6_BOUND_IF or
* IPV6_PKTINFO, use the scope ID in the destination address.
*/
if (ip6_doscopedroute && ifscope == IFSCOPE_NONE)
ifscope = dstsock->sin6_scope_id;
/*
* Perform source interface selection; the source IPv6 address
* must belong to one of the addresses of the interface used
* by the route. For performance reasons, do this only if
* there is no route, or if the routing table has changed,
* or if we haven't done source interface selection on this
* route (for this PCB instance) before.
*/
if (!select_srcif || (ro != NULL && ro->ro_rt != NULL &&
(ro->ro_rt->rt_flags & RTF_UP) &&
ro->ro_rt->generation_id == route_generation &&
(ro->ro_flags & ROF_SRCIF_SELECTED))) {
if (ro != NULL && ro->ro_rt != NULL) {
ifa = ro->ro_rt->rt_ifa;
IFA_ADDREF(ifa);
}
goto getroute;
}
/*
* Given the source IPv6 address, find a suitable source interface
* to use for transmission; if a scope ID has been specified,
* optimize the search by looking at the addresses only for that
* interface. This is still suboptimal, however, as we need to
* traverse the per-interface list.
*/
if (ifscope != IFSCOPE_NONE || (ro != NULL && ro->ro_rt != NULL)) {
unsigned int scope = ifscope;
struct ifnet *rt_ifp;
rt_ifp = (ro->ro_rt != NULL) ? ro->ro_rt->rt_ifp : NULL;
/*
* If no scope is specified and the route is stale (pointing
* to a defunct interface) use the current primary interface;
* this happens when switching between interfaces configured
* with the same IPv6 address. Otherwise pick up the scope
* information from the route; the ULP may have looked up a
* correct route and we just need to verify it here and mark
* it with the ROF_SRCIF_SELECTED flag below.
*/
if (scope == IFSCOPE_NONE) {
scope = rt_ifp->if_index;
if (scope != get_primary_ifscope(AF_INET6) &&
ro->ro_rt->generation_id != route_generation)
scope = get_primary_ifscope(AF_INET6);
}
ifa = (struct ifaddr *)
ifa_foraddr6_scoped(&srcsock->sin6_addr, scope);
if (ip6_select_srcif_debug && ifa != NULL) {
if (ro->ro_rt != NULL) {
printf("%s->%s ifscope %d->%d ifa_if %s "
"ro_if %s\n", s_src, s_dst, ifscope,
scope, if_name(ifa->ifa_ifp),
if_name(rt_ifp));
} else {
printf("%s->%s ifscope %d->%d ifa_if %s\n",
s_src, s_dst, ifscope, scope,
if_name(ifa->ifa_ifp));
}
}
}
/*
* Slow path; search for an interface having the corresponding source
* IPv6 address if the scope was not specified by the caller, and:
*
* 1) There currently isn't any route, or,
* 2) The interface used by the route does not own that source
* IPv6 address; in this case, the route will get blown away
* and we'll do a more specific scoped search using the newly
* found interface.
*/
if (ifa == NULL && ifscope == IFSCOPE_NONE) {
ifa = (struct ifaddr *)ifa_foraddr6(&srcsock->sin6_addr);
if (ip6_select_srcif_debug && ifa != NULL) {
printf("%s->%s ifscope %d ifa_if %s\n",
s_src, s_dst, ifscope, if_name(ifa->ifa_ifp));
}
}
getroute:
if (ifa != NULL)
ifscope = ifa->ifa_ifp->if_index;
/*
* If the next hop address for the packet is specified by the caller,
* use it as the gateway.
*/
if (opts != NULL && opts->ip6po_nexthop != NULL) {
struct route_in6 *ron;
sin6_next = satosin6(opts->ip6po_nexthop);
/* at this moment, we only support AF_INET6 next hops */
if (sin6_next->sin6_family != AF_INET6) {
error = EAFNOSUPPORT; /* or should we proceed? */
goto done;
}
/*
* If the next hop is an IPv6 address, then the node identified
* by that address must be a neighbor of the sending host.
*/
ron = &opts->ip6po_nextroute;
if (ron->ro_rt != NULL)
RT_LOCK(ron->ro_rt);
if ((ron->ro_rt != NULL &&
((ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) !=
(RTF_UP | RTF_LLINFO) ||
ron->ro_rt->generation_id != route_generation ||
(select_srcif && (ifa == NULL ||
ifa->ifa_ifp != ron->ro_rt->rt_ifp)))) ||
!IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr,
&sin6_next->sin6_addr)) {
if (ron->ro_rt != NULL) {
RT_UNLOCK(ron->ro_rt);
rtfree(ron->ro_rt);
ron->ro_rt = NULL;
}
*satosin6(&ron->ro_dst) = *sin6_next;
}
if (ron->ro_rt == NULL) {
rtalloc_scoped((struct route *)ron, ifscope);
if (ron->ro_rt != NULL)
RT_LOCK(ron->ro_rt);
if (ron->ro_rt == NULL ||
!(ron->ro_rt->rt_flags & RTF_LLINFO) ||
!IN6_ARE_ADDR_EQUAL(&satosin6(rt_key(ron->ro_rt))->
sin6_addr, &sin6_next->sin6_addr)) {
if (ron->ro_rt != NULL) {
RT_UNLOCK(ron->ro_rt);
rtfree(ron->ro_rt);
ron->ro_rt = NULL;
}
error = EHOSTUNREACH;
goto done;
}
}
route = ron;
ifp = ron->ro_rt->rt_ifp;
/*
* When cloning is required, try to allocate a route to the
* destination so that the caller can store path MTU
* information.
*/
if (!clone) {
if (select_srcif) {
/* Keep the route locked */
goto validateroute;
}
RT_UNLOCK(ron->ro_rt);
goto done;
}
RT_UNLOCK(ron->ro_rt);
}
/*
* Use a cached route if it exists and is valid, else try to allocate
* a new one. Note that we should check the address family of the
* cached destination, in case of sharing the cache with IPv4.
*/
if (ro == NULL)
goto done;
if (ro->ro_rt != NULL)
RT_LOCK(ro->ro_rt);
if (ro->ro_rt != NULL && (!(ro->ro_rt->rt_flags & RTF_UP) ||
satosin6(&ro->ro_dst)->sin6_family != AF_INET6 ||
ro->ro_rt->generation_id != route_generation ||
!IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, dst) ||
(select_srcif && (ifa == NULL ||
ifa->ifa_ifp != ro->ro_rt->rt_ifp)))) {
RT_UNLOCK(ro->ro_rt);
rtfree(ro->ro_rt);
ro->ro_rt = NULL;
}
if (ro->ro_rt == NULL) {
struct sockaddr_in6 *sa6;
if (ro->ro_rt != NULL)
RT_UNLOCK(ro->ro_rt);
/* No route yet, so try to acquire one */
bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
sa6->sin6_family = AF_INET6;
sa6->sin6_len = sizeof(struct sockaddr_in6);
sa6->sin6_addr = *dst;
if (IN6_IS_ADDR_MULTICAST(dst)) {
ro->ro_rt = rtalloc1_scoped(
&((struct route *)ro)->ro_dst, 0, 0, ifscope);
} else {
rtalloc_scoped((struct route *)ro, ifscope);
}
if (ro->ro_rt != NULL)
RT_LOCK(ro->ro_rt);
}
/*
* Do not care about the result if we have the nexthop
* explicitly specified (in case we're asked to clone.)
*/
if (opts != NULL && opts->ip6po_nexthop != NULL) {
if (ro->ro_rt != NULL)
RT_UNLOCK(ro->ro_rt);
goto done;
}
if (ro->ro_rt != NULL) {
RT_LOCK_ASSERT_HELD(ro->ro_rt);
ifp = ro->ro_rt->rt_ifp;
} else {
error = EHOSTUNREACH;
}
route = ro;
validateroute:
if (select_srcif) {
boolean_t has_route = (route != NULL && route->ro_rt != NULL);
if (has_route)
RT_LOCK_ASSERT_HELD(route->ro_rt);
/*
* If there is a non-loopback route with the wrong interface,
* or if there is no interface configured with such an address,
* blow it away. Except for local/loopback, we look for one
* with a matching interface scope/index.
*/
if (has_route && (ifa == NULL ||
(ifa->ifa_ifp != ifp && ifp != lo_ifp) ||
!(route->ro_rt->rt_flags & RTF_UP))) {
if (ip6_select_srcif_debug) {
if (ifa != NULL) {
printf("%s->%s ifscope %d ro_if %s "
"!= ifa_if %s (cached route "
"cleared)\n", s_src, s_dst,
ifscope, if_name(ifp),
if_name(ifa->ifa_ifp));
} else {
printf("%s->%s ifscope %d ro_if %s "
"(no ifa_if found)\n", s_src,
s_dst, ifscope, if_name(ifp));
}
}
RT_UNLOCK(route->ro_rt);
rtfree(route->ro_rt);
route->ro_rt = NULL;
route->ro_flags &= ~ROF_SRCIF_SELECTED;
error = EHOSTUNREACH;
/* Undo the settings done above */
route = NULL;
ifp = NULL;
} else if (has_route) {
route->ro_flags |= ROF_SRCIF_SELECTED;
route->ro_rt->generation_id = route_generation;
RT_UNLOCK(route->ro_rt);
}
} else {
if (ro->ro_rt != NULL)
RT_UNLOCK(ro->ro_rt);
if (ifp != NULL && opts != NULL &&
opts->ip6po_pktinfo != NULL &&
opts->ip6po_pktinfo->ipi6_ifindex != 0) {
/*
* Check if the outgoing interface conflicts with the
* interface specified by ipi6_ifindex (if specified).
* Note that loopback interface is always okay.
* (this may happen when we are sending a packet to
* one of our own addresses.)
*/
if (!(ifp->if_flags & IFF_LOOPBACK) && ifp->if_index !=
opts->ip6po_pktinfo->ipi6_ifindex) {
error = EHOSTUNREACH;
goto done;
}
}
}
done:
if (nocell && error == 0) {
if ((ifp != NULL && ifp->if_type == IFT_CELLULAR) ||
(route != NULL && route->ro_rt != NULL &&
route->ro_rt->rt_ifp->if_type == IFT_CELLULAR)) {
if (route != NULL && route->ro_rt != NULL) {
rtfree(route->ro_rt);
route->ro_rt = NULL;
route->ro_flags &= ~ROF_SRCIF_SELECTED;
route = NULL;
}
ifp = NULL;
error = EHOSTUNREACH;
}
}
if (ifp == NULL && (route == NULL || route->ro_rt == NULL)) {
/*
* This can happen if the caller did not pass a cached route
* nor any other hints. We treat this case an error.
*/
error = EHOSTUNREACH;
}
if (error == EHOSTUNREACH)
ip6stat.ip6s_noroute++;
if (error == 0) {
if (retifp != NULL) {
if (ifp != NULL)
ifnet_reference(ifp); /* for caller */
*retifp = ifp;
}
if (retrt != NULL && route != NULL)
*retrt = route->ro_rt; /* ro_rt may be NULL */
} else if (select_srcif && ip6_select_srcif_debug) {
printf("%s->%s ifscope %d ifa_if %s ro_if %s (error=%d)\n",
s_src, s_dst, ifscope,
(ifa != NULL) ? if_name(ifa->ifa_ifp) : "NONE",
(ifp != NULL) ? if_name(ifp) : "NONE", error);
}
if (ifa != NULL)
IFA_REMREF(ifa);
return (error);
}
/*
* 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;
}
