2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
24 * Fixed routing subtrees.
27 #define pr_fmt(fmt) "IPv6: " fmt
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/export.h>
32 #include <linux/types.h>
33 #include <linux/times.h>
34 #include <linux/socket.h>
35 #include <linux/sockios.h>
36 #include <linux/net.h>
37 #include <linux/route.h>
38 #include <linux/netdevice.h>
39 #include <linux/in6.h>
40 #include <linux/mroute6.h>
41 #include <linux/init.h>
42 #include <linux/if_arp.h>
43 #include <linux/proc_fs.h>
44 #include <linux/seq_file.h>
45 #include <linux/nsproxy.h>
46 #include <linux/slab.h>
47 #include <net/net_namespace.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ndisc.h>
53 #include <net/addrconf.h>
55 #include <linux/rtnetlink.h>
58 #include <net/netevent.h>
59 #include <net/netlink.h>
60 #include <net/nexthop.h>
62 #include <asm/uaccess.h>
65 #include <linux/sysctl.h>
69 RT6_NUD_FAIL_HARD = -2,
70 RT6_NUD_FAIL_SOFT = -1,
74 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
75 const struct in6_addr *dest);
76 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
77 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
78 static unsigned int ip6_mtu(const struct dst_entry *dst);
79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
80 static void ip6_dst_destroy(struct dst_entry *);
81 static void ip6_dst_ifdown(struct dst_entry *,
82 struct net_device *dev, int how);
83 static int ip6_dst_gc(struct dst_ops *ops);
85 static int ip6_pkt_discard(struct sk_buff *skb);
86 static int ip6_pkt_discard_out(struct sk_buff *skb);
87 static int ip6_pkt_prohibit(struct sk_buff *skb);
88 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
89 static void ip6_link_failure(struct sk_buff *skb);
90 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
91 struct sk_buff *skb, u32 mtu);
92 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
95 #ifdef CONFIG_IPV6_ROUTE_INFO
96 static struct rt6_info *rt6_add_route_info(struct net_device *dev,
97 const struct in6_addr *prefix, int prefixlen,
98 const struct in6_addr *gwaddr, unsigned int pref);
99 static struct rt6_info *rt6_get_route_info(struct net_device *dev,
100 const struct in6_addr *prefix, int prefixlen,
101 const struct in6_addr *gwaddr);
104 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
106 struct rt6_info *rt = (struct rt6_info *) dst;
107 struct inet_peer *peer;
110 if (!(rt->dst.flags & DST_HOST))
113 peer = rt6_get_peer_create(rt);
115 u32 *old_p = __DST_METRICS_PTR(old);
116 unsigned long prev, new;
119 if (inet_metrics_new(peer))
120 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
122 new = (unsigned long) p;
123 prev = cmpxchg(&dst->_metrics, old, new);
126 p = __DST_METRICS_PTR(prev);
127 if (prev & DST_METRICS_READ_ONLY)
134 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
138 struct in6_addr *p = &rt->rt6i_gateway;
140 if (!ipv6_addr_any(p))
141 return (const void *) p;
143 return &ipv6_hdr(skb)->daddr;
147 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
151 struct rt6_info *rt = (struct rt6_info *) dst;
154 daddr = choose_neigh_daddr(rt, skb, daddr);
155 n = __ipv6_neigh_lookup(dst->dev, daddr);
158 return neigh_create(&nd_tbl, daddr, dst->dev);
161 static struct dst_ops ip6_dst_ops_template = {
163 .protocol = cpu_to_be16(ETH_P_IPV6),
166 .check = ip6_dst_check,
167 .default_advmss = ip6_default_advmss,
169 .cow_metrics = ipv6_cow_metrics,
170 .destroy = ip6_dst_destroy,
171 .ifdown = ip6_dst_ifdown,
172 .negative_advice = ip6_negative_advice,
173 .link_failure = ip6_link_failure,
174 .update_pmtu = ip6_rt_update_pmtu,
175 .redirect = rt6_do_redirect,
176 .local_out = __ip6_local_out,
177 .neigh_lookup = ip6_neigh_lookup,
180 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
182 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
184 return mtu ? : dst->dev->mtu;
187 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
188 struct sk_buff *skb, u32 mtu)
192 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
197 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
203 static struct dst_ops ip6_dst_blackhole_ops = {
205 .protocol = cpu_to_be16(ETH_P_IPV6),
206 .destroy = ip6_dst_destroy,
207 .check = ip6_dst_check,
208 .mtu = ip6_blackhole_mtu,
209 .default_advmss = ip6_default_advmss,
210 .update_pmtu = ip6_rt_blackhole_update_pmtu,
211 .redirect = ip6_rt_blackhole_redirect,
212 .cow_metrics = ip6_rt_blackhole_cow_metrics,
213 .neigh_lookup = ip6_neigh_lookup,
216 static const u32 ip6_template_metrics[RTAX_MAX] = {
217 [RTAX_HOPLIMIT - 1] = 0,
220 static const struct rt6_info ip6_null_entry_template = {
222 .__refcnt = ATOMIC_INIT(1),
224 .obsolete = DST_OBSOLETE_FORCE_CHK,
225 .error = -ENETUNREACH,
226 .input = ip6_pkt_discard,
227 .output = ip6_pkt_discard_out,
229 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
230 .rt6i_protocol = RTPROT_KERNEL,
231 .rt6i_metric = ~(u32) 0,
232 .rt6i_ref = ATOMIC_INIT(1),
235 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
237 static const struct rt6_info ip6_prohibit_entry_template = {
239 .__refcnt = ATOMIC_INIT(1),
241 .obsolete = DST_OBSOLETE_FORCE_CHK,
243 .input = ip6_pkt_prohibit,
244 .output = ip6_pkt_prohibit_out,
246 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
247 .rt6i_protocol = RTPROT_KERNEL,
248 .rt6i_metric = ~(u32) 0,
249 .rt6i_ref = ATOMIC_INIT(1),
252 static const struct rt6_info ip6_blk_hole_entry_template = {
254 .__refcnt = ATOMIC_INIT(1),
256 .obsolete = DST_OBSOLETE_FORCE_CHK,
258 .input = dst_discard,
259 .output = dst_discard,
261 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
262 .rt6i_protocol = RTPROT_KERNEL,
263 .rt6i_metric = ~(u32) 0,
264 .rt6i_ref = ATOMIC_INIT(1),
269 /* allocate dst with ip6_dst_ops */
270 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
271 struct net_device *dev,
273 struct fib6_table *table)
275 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
276 0, DST_OBSOLETE_FORCE_CHK, flags);
279 struct dst_entry *dst = &rt->dst;
281 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
282 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
283 rt->rt6i_genid = rt_genid(net);
284 INIT_LIST_HEAD(&rt->rt6i_siblings);
285 rt->rt6i_nsiblings = 0;
290 static void ip6_dst_destroy(struct dst_entry *dst)
292 struct rt6_info *rt = (struct rt6_info *)dst;
293 struct inet6_dev *idev = rt->rt6i_idev;
294 struct dst_entry *from = dst->from;
296 if (!(rt->dst.flags & DST_HOST))
297 dst_destroy_metrics_generic(dst);
300 rt->rt6i_idev = NULL;
307 if (rt6_has_peer(rt)) {
308 struct inet_peer *peer = rt6_peer_ptr(rt);
313 void rt6_bind_peer(struct rt6_info *rt, int create)
315 struct inet_peer_base *base;
316 struct inet_peer *peer;
318 base = inetpeer_base_ptr(rt->_rt6i_peer);
322 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
324 if (!rt6_set_peer(rt, peer))
329 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
332 struct rt6_info *rt = (struct rt6_info *)dst;
333 struct inet6_dev *idev = rt->rt6i_idev;
334 struct net_device *loopback_dev =
335 dev_net(dev)->loopback_dev;
337 if (dev != loopback_dev) {
338 if (idev && idev->dev == dev) {
339 struct inet6_dev *loopback_idev =
340 in6_dev_get(loopback_dev);
342 rt->rt6i_idev = loopback_idev;
349 static bool rt6_check_expired(const struct rt6_info *rt)
351 if (rt->rt6i_flags & RTF_EXPIRES) {
352 if (time_after(jiffies, rt->dst.expires))
354 } else if (rt->dst.from) {
355 return rt6_check_expired((struct rt6_info *) rt->dst.from);
360 static bool rt6_need_strict(const struct in6_addr *daddr)
362 return ipv6_addr_type(daddr) &
363 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
366 /* Multipath route selection:
367 * Hash based function using packet header and flowlabel.
368 * Adapted from fib_info_hashfn()
370 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
371 const struct flowi6 *fl6)
373 unsigned int val = fl6->flowi6_proto;
375 val ^= ipv6_addr_hash(&fl6->daddr);
376 val ^= ipv6_addr_hash(&fl6->saddr);
378 /* Work only if this not encapsulated */
379 switch (fl6->flowi6_proto) {
383 val ^= (__force u16)fl6->fl6_sport;
384 val ^= (__force u16)fl6->fl6_dport;
388 val ^= (__force u16)fl6->fl6_icmp_type;
389 val ^= (__force u16)fl6->fl6_icmp_code;
392 /* RFC6438 recommands to use flowlabel */
393 val ^= (__force u32)fl6->flowlabel;
395 /* Perhaps, we need to tune, this function? */
396 val = val ^ (val >> 7) ^ (val >> 12);
397 return val % candidate_count;
400 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
403 struct rt6_info *sibling, *next_sibling;
406 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
407 /* Don't change the route, if route_choosen == 0
408 * (siblings does not include ourself)
411 list_for_each_entry_safe(sibling, next_sibling,
412 &match->rt6i_siblings, rt6i_siblings) {
414 if (route_choosen == 0) {
423 * Route lookup. Any table->tb6_lock is implied.
426 static inline struct rt6_info *rt6_device_match(struct net *net,
428 const struct in6_addr *saddr,
432 struct rt6_info *local = NULL;
433 struct rt6_info *sprt;
435 if (!oif && ipv6_addr_any(saddr))
438 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
439 struct net_device *dev = sprt->dst.dev;
442 if (dev->ifindex == oif)
444 if (dev->flags & IFF_LOOPBACK) {
445 if (!sprt->rt6i_idev ||
446 sprt->rt6i_idev->dev->ifindex != oif) {
447 if (flags & RT6_LOOKUP_F_IFACE && oif)
449 if (local && (!oif ||
450 local->rt6i_idev->dev->ifindex == oif))
456 if (ipv6_chk_addr(net, saddr, dev,
457 flags & RT6_LOOKUP_F_IFACE))
466 if (flags & RT6_LOOKUP_F_IFACE)
467 return net->ipv6.ip6_null_entry;
473 #ifdef CONFIG_IPV6_ROUTER_PREF
474 struct __rt6_probe_work {
475 struct work_struct work;
476 struct in6_addr target;
477 struct net_device *dev;
480 static void rt6_probe_deferred(struct work_struct *w)
482 struct in6_addr mcaddr;
483 struct __rt6_probe_work *work =
484 container_of(w, struct __rt6_probe_work, work);
486 addrconf_addr_solict_mult(&work->target, &mcaddr);
487 ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
492 static void rt6_probe(struct rt6_info *rt)
494 struct neighbour *neigh;
496 * Okay, this does not seem to be appropriate
497 * for now, however, we need to check if it
498 * is really so; aka Router Reachability Probing.
500 * Router Reachability Probe MUST be rate-limited
501 * to no more than one per minute.
503 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
506 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
508 write_lock(&neigh->lock);
509 if (neigh->nud_state & NUD_VALID)
514 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
515 struct __rt6_probe_work *work;
517 work = kmalloc(sizeof(*work), GFP_ATOMIC);
520 neigh->updated = jiffies;
523 write_unlock(&neigh->lock);
526 INIT_WORK(&work->work, rt6_probe_deferred);
527 work->target = rt->rt6i_gateway;
528 dev_hold(rt->dst.dev);
529 work->dev = rt->dst.dev;
530 schedule_work(&work->work);
534 write_unlock(&neigh->lock);
536 rcu_read_unlock_bh();
539 static inline void rt6_probe(struct rt6_info *rt)
545 * Default Router Selection (RFC 2461 6.3.6)
547 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
549 struct net_device *dev = rt->dst.dev;
550 if (!oif || dev->ifindex == oif)
552 if ((dev->flags & IFF_LOOPBACK) &&
553 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
558 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
560 struct neighbour *neigh;
561 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
563 if (rt->rt6i_flags & RTF_NONEXTHOP ||
564 !(rt->rt6i_flags & RTF_GATEWAY))
565 return RT6_NUD_SUCCEED;
568 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
570 read_lock(&neigh->lock);
571 if (neigh->nud_state & NUD_VALID)
572 ret = RT6_NUD_SUCCEED;
573 #ifdef CONFIG_IPV6_ROUTER_PREF
574 else if (!(neigh->nud_state & NUD_FAILED))
575 ret = RT6_NUD_SUCCEED;
577 read_unlock(&neigh->lock);
579 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
580 RT6_NUD_SUCCEED : RT6_NUD_FAIL_SOFT;
582 rcu_read_unlock_bh();
587 static int rt6_score_route(struct rt6_info *rt, int oif,
592 m = rt6_check_dev(rt, oif);
593 if (!m && (strict & RT6_LOOKUP_F_IFACE))
594 return RT6_NUD_FAIL_HARD;
595 #ifdef CONFIG_IPV6_ROUTER_PREF
596 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
598 if (strict & RT6_LOOKUP_F_REACHABLE) {
599 int n = rt6_check_neigh(rt);
606 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
607 int *mpri, struct rt6_info *match,
611 bool match_do_rr = false;
613 if (rt6_check_expired(rt))
616 m = rt6_score_route(rt, oif, strict);
617 if (m == RT6_NUD_FAIL_SOFT && !IS_ENABLED(CONFIG_IPV6_ROUTER_PREF)) {
619 m = 0; /* lowest valid score */
624 if (strict & RT6_LOOKUP_F_REACHABLE)
628 *do_rr = match_do_rr;
636 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
637 struct rt6_info *rr_head,
638 u32 metric, int oif, int strict,
641 struct rt6_info *rt, *match;
645 for (rt = rr_head; rt && rt->rt6i_metric == metric;
646 rt = rt->dst.rt6_next)
647 match = find_match(rt, oif, strict, &mpri, match, do_rr);
648 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
649 rt = rt->dst.rt6_next)
650 match = find_match(rt, oif, strict, &mpri, match, do_rr);
655 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
657 struct rt6_info *match, *rt0;
663 fn->rr_ptr = rt0 = fn->leaf;
665 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
669 struct rt6_info *next = rt0->dst.rt6_next;
671 /* no entries matched; do round-robin */
672 if (!next || next->rt6i_metric != rt0->rt6i_metric)
679 net = dev_net(rt0->dst.dev);
680 return match ? match : net->ipv6.ip6_null_entry;
683 #ifdef CONFIG_IPV6_ROUTE_INFO
684 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
685 const struct in6_addr *gwaddr)
687 struct route_info *rinfo = (struct route_info *) opt;
688 struct in6_addr prefix_buf, *prefix;
690 unsigned long lifetime;
693 if (len < sizeof(struct route_info)) {
697 /* Sanity check for prefix_len and length */
698 if (rinfo->length > 3) {
700 } else if (rinfo->prefix_len > 128) {
702 } else if (rinfo->prefix_len > 64) {
703 if (rinfo->length < 2) {
706 } else if (rinfo->prefix_len > 0) {
707 if (rinfo->length < 1) {
712 pref = rinfo->route_pref;
713 if (pref == ICMPV6_ROUTER_PREF_INVALID)
716 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
718 if (rinfo->length == 3)
719 prefix = (struct in6_addr *)rinfo->prefix;
721 /* this function is safe */
722 ipv6_addr_prefix(&prefix_buf,
723 (struct in6_addr *)rinfo->prefix,
725 prefix = &prefix_buf;
728 if (rinfo->prefix_len == 0)
729 rt = rt6_get_dflt_router(gwaddr, dev);
731 rt = rt6_get_route_info(dev, prefix, rinfo->prefix_len, gwaddr);
733 if (rt && !lifetime) {
739 rt = rt6_add_route_info(dev, prefix, rinfo->prefix_len, gwaddr, pref);
741 rt->rt6i_flags = RTF_ROUTEINFO |
742 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
745 if (!addrconf_finite_timeout(lifetime))
746 rt6_clean_expires(rt);
748 rt6_set_expires(rt, jiffies + HZ * lifetime);
756 #define BACKTRACK(__net, saddr) \
758 if (rt == __net->ipv6.ip6_null_entry) { \
759 struct fib6_node *pn; \
761 if (fn->fn_flags & RTN_TL_ROOT) \
764 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
765 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
768 if (fn->fn_flags & RTN_RTINFO) \
774 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
775 struct fib6_table *table,
776 struct flowi6 *fl6, int flags)
778 struct fib6_node *fn;
781 read_lock_bh(&table->tb6_lock);
782 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
785 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
786 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
787 rt = rt6_multipath_select(rt, fl6);
788 BACKTRACK(net, &fl6->saddr);
790 dst_use(&rt->dst, jiffies);
791 read_unlock_bh(&table->tb6_lock);
796 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
799 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
801 EXPORT_SYMBOL_GPL(ip6_route_lookup);
803 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
804 const struct in6_addr *saddr, int oif, int strict)
806 struct flowi6 fl6 = {
810 struct dst_entry *dst;
811 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
814 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
815 flags |= RT6_LOOKUP_F_HAS_SADDR;
818 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
820 return (struct rt6_info *) dst;
827 EXPORT_SYMBOL(rt6_lookup);
829 /* ip6_ins_rt is called with FREE table->tb6_lock.
830 It takes new route entry, the addition fails by any reason the
831 route is freed. In any case, if caller does not hold it, it may
835 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
838 struct fib6_table *table;
840 table = rt->rt6i_table;
841 write_lock_bh(&table->tb6_lock);
842 err = fib6_add(&table->tb6_root, rt, info);
843 write_unlock_bh(&table->tb6_lock);
848 int ip6_ins_rt(struct rt6_info *rt)
850 struct nl_info info = {
851 .nl_net = dev_net(rt->dst.dev),
853 return __ip6_ins_rt(rt, &info);
856 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
857 const struct in6_addr *daddr,
858 const struct in6_addr *saddr)
866 rt = ip6_rt_copy(ort, daddr);
869 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
870 if (ort->rt6i_dst.plen != 128 &&
871 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
872 rt->rt6i_flags |= RTF_ANYCAST;
875 rt->rt6i_flags |= RTF_CACHE;
877 #ifdef CONFIG_IPV6_SUBTREES
878 if (rt->rt6i_src.plen && saddr) {
879 rt->rt6i_src.addr = *saddr;
880 rt->rt6i_src.plen = 128;
888 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
889 const struct in6_addr *daddr)
891 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
894 rt->rt6i_flags |= RTF_CACHE;
898 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
899 struct flowi6 *fl6, int flags)
901 struct fib6_node *fn;
902 struct rt6_info *rt, *nrt;
906 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
908 strict |= flags & RT6_LOOKUP_F_IFACE;
911 read_lock_bh(&table->tb6_lock);
914 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
917 rt = rt6_select(fn, oif, strict | reachable);
918 if (rt->rt6i_nsiblings && oif == 0)
919 rt = rt6_multipath_select(rt, fl6);
920 BACKTRACK(net, &fl6->saddr);
921 if (rt == net->ipv6.ip6_null_entry ||
922 rt->rt6i_flags & RTF_CACHE)
926 read_unlock_bh(&table->tb6_lock);
928 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
929 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
930 else if (!(rt->dst.flags & DST_HOST))
931 nrt = rt6_alloc_clone(rt, &fl6->daddr);
936 rt = nrt ? : net->ipv6.ip6_null_entry;
940 err = ip6_ins_rt(nrt);
949 * Race condition! In the gap, when table->tb6_lock was
950 * released someone could insert this route. Relookup.
961 read_unlock_bh(&table->tb6_lock);
963 rt->dst.lastuse = jiffies;
969 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
970 struct flowi6 *fl6, int flags)
972 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
975 static struct dst_entry *ip6_route_input_lookup(struct net *net,
976 struct net_device *dev,
977 struct flowi6 *fl6, int flags)
979 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
980 flags |= RT6_LOOKUP_F_IFACE;
982 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
985 void ip6_route_input(struct sk_buff *skb)
987 const struct ipv6hdr *iph = ipv6_hdr(skb);
988 struct net *net = dev_net(skb->dev);
989 int flags = RT6_LOOKUP_F_HAS_SADDR;
990 struct flowi6 fl6 = {
991 .flowi6_iif = skb->dev->ifindex,
994 .flowlabel = ip6_flowinfo(iph),
995 .flowi6_mark = skb->mark,
996 .flowi6_proto = iph->nexthdr,
999 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1002 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1003 struct flowi6 *fl6, int flags)
1005 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1008 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1013 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1015 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1016 flags |= RT6_LOOKUP_F_IFACE;
1018 if (!ipv6_addr_any(&fl6->saddr))
1019 flags |= RT6_LOOKUP_F_HAS_SADDR;
1021 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1023 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1026 EXPORT_SYMBOL(ip6_route_output);
1028 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1030 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1031 struct dst_entry *new = NULL;
1033 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1037 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1038 rt6_init_peer(rt, net->ipv6.peers);
1041 new->input = dst_discard;
1042 new->output = dst_discard;
1044 if (dst_metrics_read_only(&ort->dst))
1045 new->_metrics = ort->dst._metrics;
1047 dst_copy_metrics(new, &ort->dst);
1048 rt->rt6i_idev = ort->rt6i_idev;
1050 in6_dev_hold(rt->rt6i_idev);
1052 rt->rt6i_gateway = ort->rt6i_gateway;
1053 rt->rt6i_flags = ort->rt6i_flags;
1054 rt->rt6i_metric = 0;
1056 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1057 #ifdef CONFIG_IPV6_SUBTREES
1058 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1064 dst_release(dst_orig);
1065 return new ? new : ERR_PTR(-ENOMEM);
1069 * Destination cache support functions
1072 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1074 struct rt6_info *rt;
1076 rt = (struct rt6_info *) dst;
1078 /* All IPV6 dsts are created with ->obsolete set to the value
1079 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1080 * into this function always.
1082 if (rt->rt6i_genid != rt_genid(dev_net(rt->dst.dev)))
1085 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1088 if (rt6_check_expired(rt))
1094 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1096 struct rt6_info *rt = (struct rt6_info *) dst;
1099 if (rt->rt6i_flags & RTF_CACHE) {
1100 if (rt6_check_expired(rt)) {
1112 static void ip6_link_failure(struct sk_buff *skb)
1114 struct rt6_info *rt;
1116 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1118 rt = (struct rt6_info *) skb_dst(skb);
1120 if (rt->rt6i_flags & RTF_CACHE) {
1124 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1125 rt->rt6i_node->fn_sernum = -1;
1130 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1131 struct sk_buff *skb, u32 mtu)
1133 struct rt6_info *rt6 = (struct rt6_info*)dst;
1136 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1137 struct net *net = dev_net(dst->dev);
1139 rt6->rt6i_flags |= RTF_MODIFIED;
1140 if (mtu < IPV6_MIN_MTU) {
1141 u32 features = dst_metric(dst, RTAX_FEATURES);
1143 features |= RTAX_FEATURE_ALLFRAG;
1144 dst_metric_set(dst, RTAX_FEATURES, features);
1146 dst_metric_set(dst, RTAX_MTU, mtu);
1147 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1151 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1154 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1155 struct dst_entry *dst;
1158 memset(&fl6, 0, sizeof(fl6));
1159 fl6.flowi6_oif = oif;
1160 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1161 fl6.flowi6_flags = 0;
1162 fl6.daddr = iph->daddr;
1163 fl6.saddr = iph->saddr;
1164 fl6.flowlabel = ip6_flowinfo(iph);
1166 dst = ip6_route_output(net, NULL, &fl6);
1168 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1171 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1173 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1175 ip6_update_pmtu(skb, sock_net(sk), mtu,
1176 sk->sk_bound_dev_if, sk->sk_mark);
1178 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1180 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1182 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1183 struct dst_entry *dst;
1186 memset(&fl6, 0, sizeof(fl6));
1187 fl6.flowi6_oif = oif;
1188 fl6.flowi6_mark = mark;
1189 fl6.flowi6_flags = 0;
1190 fl6.daddr = iph->daddr;
1191 fl6.saddr = iph->saddr;
1192 fl6.flowlabel = ip6_flowinfo(iph);
1194 dst = ip6_route_output(net, NULL, &fl6);
1196 rt6_do_redirect(dst, NULL, skb);
1199 EXPORT_SYMBOL_GPL(ip6_redirect);
1201 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1203 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1205 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1207 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1209 struct net_device *dev = dst->dev;
1210 unsigned int mtu = dst_mtu(dst);
1211 struct net *net = dev_net(dev);
1213 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1215 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1216 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1219 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1220 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1221 * IPV6_MAXPLEN is also valid and means: "any MSS,
1222 * rely only on pmtu discovery"
1224 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1229 static unsigned int ip6_mtu(const struct dst_entry *dst)
1231 struct inet6_dev *idev;
1232 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1240 idev = __in6_dev_get(dst->dev);
1242 mtu = idev->cnf.mtu6;
1246 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1249 static struct dst_entry *icmp6_dst_gc_list;
1250 static DEFINE_SPINLOCK(icmp6_dst_lock);
1252 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1255 struct dst_entry *dst;
1256 struct rt6_info *rt;
1257 struct inet6_dev *idev = in6_dev_get(dev);
1258 struct net *net = dev_net(dev);
1260 if (unlikely(!idev))
1261 return ERR_PTR(-ENODEV);
1263 rt = ip6_dst_alloc(net, dev, 0, NULL);
1264 if (unlikely(!rt)) {
1266 dst = ERR_PTR(-ENOMEM);
1270 rt->dst.flags |= DST_HOST;
1271 rt->dst.output = ip6_output;
1272 atomic_set(&rt->dst.__refcnt, 1);
1273 rt->rt6i_gateway = fl6->daddr;
1274 rt->rt6i_dst.addr = fl6->daddr;
1275 rt->rt6i_dst.plen = 128;
1276 rt->rt6i_idev = idev;
1277 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1279 spin_lock_bh(&icmp6_dst_lock);
1280 rt->dst.next = icmp6_dst_gc_list;
1281 icmp6_dst_gc_list = &rt->dst;
1282 spin_unlock_bh(&icmp6_dst_lock);
1284 fib6_force_start_gc(net);
1286 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1292 int icmp6_dst_gc(void)
1294 struct dst_entry *dst, **pprev;
1297 spin_lock_bh(&icmp6_dst_lock);
1298 pprev = &icmp6_dst_gc_list;
1300 while ((dst = *pprev) != NULL) {
1301 if (!atomic_read(&dst->__refcnt)) {
1310 spin_unlock_bh(&icmp6_dst_lock);
1315 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1318 struct dst_entry *dst, **pprev;
1320 spin_lock_bh(&icmp6_dst_lock);
1321 pprev = &icmp6_dst_gc_list;
1322 while ((dst = *pprev) != NULL) {
1323 struct rt6_info *rt = (struct rt6_info *) dst;
1324 if (func(rt, arg)) {
1331 spin_unlock_bh(&icmp6_dst_lock);
1334 static int ip6_dst_gc(struct dst_ops *ops)
1336 unsigned long now = jiffies;
1337 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1338 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1339 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1340 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1341 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1342 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1345 entries = dst_entries_get_fast(ops);
1346 if (time_after(rt_last_gc + rt_min_interval, now) &&
1347 entries <= rt_max_size)
1350 net->ipv6.ip6_rt_gc_expire++;
1351 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1352 net->ipv6.ip6_rt_last_gc = now;
1353 entries = dst_entries_get_slow(ops);
1354 if (entries < ops->gc_thresh)
1355 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1357 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1358 return entries > rt_max_size;
1361 int ip6_dst_hoplimit(struct dst_entry *dst)
1363 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1364 if (hoplimit == 0) {
1365 struct net_device *dev = dst->dev;
1366 struct inet6_dev *idev;
1369 idev = __in6_dev_get(dev);
1371 hoplimit = idev->cnf.hop_limit;
1373 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1378 EXPORT_SYMBOL(ip6_dst_hoplimit);
1384 int ip6_route_add(struct fib6_config *cfg)
1387 struct net *net = cfg->fc_nlinfo.nl_net;
1388 struct rt6_info *rt = NULL;
1389 struct net_device *dev = NULL;
1390 struct inet6_dev *idev = NULL;
1391 struct fib6_table *table;
1394 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1396 #ifndef CONFIG_IPV6_SUBTREES
1397 if (cfg->fc_src_len)
1400 if (cfg->fc_ifindex) {
1402 dev = dev_get_by_index(net, cfg->fc_ifindex);
1405 idev = in6_dev_get(dev);
1410 if (cfg->fc_metric == 0)
1411 cfg->fc_metric = IP6_RT_PRIO_USER;
1414 if (cfg->fc_nlinfo.nlh &&
1415 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1416 table = fib6_get_table(net, cfg->fc_table);
1418 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1419 table = fib6_new_table(net, cfg->fc_table);
1422 table = fib6_new_table(net, cfg->fc_table);
1428 rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1435 if (cfg->fc_flags & RTF_EXPIRES)
1436 rt6_set_expires(rt, jiffies +
1437 clock_t_to_jiffies(cfg->fc_expires));
1439 rt6_clean_expires(rt);
1441 if (cfg->fc_protocol == RTPROT_UNSPEC)
1442 cfg->fc_protocol = RTPROT_BOOT;
1443 rt->rt6i_protocol = cfg->fc_protocol;
1445 addr_type = ipv6_addr_type(&cfg->fc_dst);
1447 if (addr_type & IPV6_ADDR_MULTICAST)
1448 rt->dst.input = ip6_mc_input;
1449 else if (cfg->fc_flags & RTF_LOCAL)
1450 rt->dst.input = ip6_input;
1452 rt->dst.input = ip6_forward;
1454 rt->dst.output = ip6_output;
1456 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1457 rt->rt6i_dst.plen = cfg->fc_dst_len;
1458 if (rt->rt6i_dst.plen == 128)
1459 rt->dst.flags |= DST_HOST;
1461 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1462 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1467 dst_init_metrics(&rt->dst, metrics, 0);
1469 #ifdef CONFIG_IPV6_SUBTREES
1470 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1471 rt->rt6i_src.plen = cfg->fc_src_len;
1474 rt->rt6i_metric = cfg->fc_metric;
1476 /* We cannot add true routes via loopback here,
1477 they would result in kernel looping; promote them to reject routes
1479 if ((cfg->fc_flags & RTF_REJECT) ||
1480 (dev && (dev->flags & IFF_LOOPBACK) &&
1481 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1482 !(cfg->fc_flags & RTF_LOCAL))) {
1483 /* hold loopback dev/idev if we haven't done so. */
1484 if (dev != net->loopback_dev) {
1489 dev = net->loopback_dev;
1491 idev = in6_dev_get(dev);
1497 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1498 switch (cfg->fc_type) {
1500 rt->dst.error = -EINVAL;
1501 rt->dst.output = dst_discard;
1502 rt->dst.input = dst_discard;
1505 rt->dst.error = -EACCES;
1506 rt->dst.output = ip6_pkt_prohibit_out;
1507 rt->dst.input = ip6_pkt_prohibit;
1511 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1513 rt->dst.output = ip6_pkt_discard_out;
1514 rt->dst.input = ip6_pkt_discard;
1520 if (cfg->fc_flags & RTF_GATEWAY) {
1521 const struct in6_addr *gw_addr;
1524 gw_addr = &cfg->fc_gateway;
1525 rt->rt6i_gateway = *gw_addr;
1526 gwa_type = ipv6_addr_type(gw_addr);
1528 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1529 struct rt6_info *grt;
1531 /* IPv6 strictly inhibits using not link-local
1532 addresses as nexthop address.
1533 Otherwise, router will not able to send redirects.
1534 It is very good, but in some (rare!) circumstances
1535 (SIT, PtP, NBMA NOARP links) it is handy to allow
1536 some exceptions. --ANK
1539 if (!(gwa_type & IPV6_ADDR_UNICAST))
1542 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1544 err = -EHOSTUNREACH;
1548 if (dev != grt->dst.dev) {
1554 idev = grt->rt6i_idev;
1556 in6_dev_hold(grt->rt6i_idev);
1558 if (!(grt->rt6i_flags & RTF_GATEWAY))
1566 if (!dev || (dev->flags & IFF_LOOPBACK))
1574 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1575 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1579 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1580 rt->rt6i_prefsrc.plen = 128;
1582 rt->rt6i_prefsrc.plen = 0;
1584 rt->rt6i_flags = cfg->fc_flags;
1591 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1592 int type = nla_type(nla);
1595 if (type > RTAX_MAX) {
1600 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1606 rt->rt6i_idev = idev;
1607 rt->rt6i_table = table;
1609 cfg->fc_nlinfo.nl_net = dev_net(dev);
1611 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1623 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1626 struct fib6_table *table;
1627 struct net *net = dev_net(rt->dst.dev);
1629 if (rt == net->ipv6.ip6_null_entry) {
1634 table = rt->rt6i_table;
1635 write_lock_bh(&table->tb6_lock);
1636 err = fib6_del(rt, info);
1637 write_unlock_bh(&table->tb6_lock);
1644 int ip6_del_rt(struct rt6_info *rt)
1646 struct nl_info info = {
1647 .nl_net = dev_net(rt->dst.dev),
1649 return __ip6_del_rt(rt, &info);
1652 static int ip6_route_del(struct fib6_config *cfg)
1654 struct fib6_table *table;
1655 struct fib6_node *fn;
1656 struct rt6_info *rt;
1659 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1663 read_lock_bh(&table->tb6_lock);
1665 fn = fib6_locate(&table->tb6_root,
1666 &cfg->fc_dst, cfg->fc_dst_len,
1667 &cfg->fc_src, cfg->fc_src_len);
1670 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1671 if (cfg->fc_ifindex &&
1673 rt->dst.dev->ifindex != cfg->fc_ifindex))
1675 if (cfg->fc_flags & RTF_GATEWAY &&
1676 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1678 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1681 read_unlock_bh(&table->tb6_lock);
1683 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1686 read_unlock_bh(&table->tb6_lock);
1691 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1693 struct net *net = dev_net(skb->dev);
1694 struct netevent_redirect netevent;
1695 struct rt6_info *rt, *nrt = NULL;
1696 struct ndisc_options ndopts;
1697 struct inet6_dev *in6_dev;
1698 struct neighbour *neigh;
1700 int optlen, on_link;
1703 optlen = skb->tail - skb->transport_header;
1704 optlen -= sizeof(*msg);
1707 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1711 msg = (struct rd_msg *)icmp6_hdr(skb);
1713 if (ipv6_addr_is_multicast(&msg->dest)) {
1714 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1719 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1721 } else if (ipv6_addr_type(&msg->target) !=
1722 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1723 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1727 in6_dev = __in6_dev_get(skb->dev);
1730 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1734 * The IP source address of the Redirect MUST be the same as the current
1735 * first-hop router for the specified ICMP Destination Address.
1738 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1739 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1744 if (ndopts.nd_opts_tgt_lladdr) {
1745 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1748 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1753 rt = (struct rt6_info *) dst;
1754 if (rt == net->ipv6.ip6_null_entry) {
1755 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1759 /* Redirect received -> path was valid.
1760 * Look, redirects are sent only in response to data packets,
1761 * so that this nexthop apparently is reachable. --ANK
1763 dst_confirm(&rt->dst);
1765 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1770 * We have finally decided to accept it.
1773 neigh_update(neigh, lladdr, NUD_STALE,
1774 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1775 NEIGH_UPDATE_F_OVERRIDE|
1776 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1777 NEIGH_UPDATE_F_ISROUTER))
1780 nrt = ip6_rt_copy(rt, &msg->dest);
1784 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1786 nrt->rt6i_flags &= ~RTF_GATEWAY;
1788 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1790 if (ip6_ins_rt(nrt))
1793 netevent.old = &rt->dst;
1794 netevent.new = &nrt->dst;
1795 netevent.daddr = &msg->dest;
1796 netevent.neigh = neigh;
1797 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1799 if (rt->rt6i_flags & RTF_CACHE) {
1800 rt = (struct rt6_info *) dst_clone(&rt->dst);
1805 neigh_release(neigh);
1809 * Misc support functions
1812 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1813 const struct in6_addr *dest)
1815 struct net *net = dev_net(ort->dst.dev);
1816 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1820 rt->dst.input = ort->dst.input;
1821 rt->dst.output = ort->dst.output;
1822 rt->dst.flags |= DST_HOST;
1824 rt->rt6i_dst.addr = *dest;
1825 rt->rt6i_dst.plen = 128;
1826 dst_copy_metrics(&rt->dst, &ort->dst);
1827 rt->dst.error = ort->dst.error;
1828 rt->rt6i_idev = ort->rt6i_idev;
1830 in6_dev_hold(rt->rt6i_idev);
1831 rt->dst.lastuse = jiffies;
1833 if (ort->rt6i_flags & RTF_GATEWAY)
1834 rt->rt6i_gateway = ort->rt6i_gateway;
1836 rt->rt6i_gateway = *dest;
1837 rt->rt6i_flags = ort->rt6i_flags;
1838 rt6_set_from(rt, ort);
1839 rt->rt6i_metric = 0;
1841 #ifdef CONFIG_IPV6_SUBTREES
1842 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1844 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1845 rt->rt6i_table = ort->rt6i_table;
1850 #ifdef CONFIG_IPV6_ROUTE_INFO
1851 static struct rt6_info *rt6_get_route_info(struct net_device *dev,
1852 const struct in6_addr *prefix, int prefixlen,
1853 const struct in6_addr *gwaddr)
1855 struct fib6_node *fn;
1856 struct rt6_info *rt = NULL;
1857 struct fib6_table *table;
1859 table = fib6_get_table(dev_net(dev),
1860 addrconf_rt_table(dev, RT6_TABLE_INFO));
1864 read_lock_bh(&table->tb6_lock);
1865 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1869 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1870 if (rt->dst.dev->ifindex != dev->ifindex)
1872 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1874 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1880 read_unlock_bh(&table->tb6_lock);
1884 static struct rt6_info *rt6_add_route_info(struct net_device *dev,
1885 const struct in6_addr *prefix, int prefixlen,
1886 const struct in6_addr *gwaddr, unsigned int pref)
1888 struct fib6_config cfg = {
1889 .fc_table = addrconf_rt_table(dev, RT6_TABLE_INFO),
1890 .fc_metric = IP6_RT_PRIO_USER,
1891 .fc_ifindex = dev->ifindex,
1892 .fc_dst_len = prefixlen,
1893 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1894 RTF_UP | RTF_PREF(pref),
1895 .fc_nlinfo.portid = 0,
1896 .fc_nlinfo.nlh = NULL,
1897 .fc_nlinfo.nl_net = dev_net(dev),
1900 cfg.fc_dst = *prefix;
1901 cfg.fc_gateway = *gwaddr;
1903 /* We should treat it as a default route if prefix length is 0. */
1905 cfg.fc_flags |= RTF_DEFAULT;
1907 ip6_route_add(&cfg);
1909 return rt6_get_route_info(dev, prefix, prefixlen, gwaddr);
1913 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1915 struct rt6_info *rt;
1916 struct fib6_table *table;
1918 table = fib6_get_table(dev_net(dev),
1919 addrconf_rt_table(dev, RT6_TABLE_MAIN));
1923 read_lock_bh(&table->tb6_lock);
1924 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1925 if (dev == rt->dst.dev &&
1926 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1927 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1932 read_unlock_bh(&table->tb6_lock);
1936 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1937 struct net_device *dev,
1940 struct fib6_config cfg = {
1941 .fc_table = addrconf_rt_table(dev, RT6_TABLE_DFLT),
1942 .fc_metric = IP6_RT_PRIO_USER,
1943 .fc_ifindex = dev->ifindex,
1944 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1945 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1946 .fc_nlinfo.portid = 0,
1947 .fc_nlinfo.nlh = NULL,
1948 .fc_nlinfo.nl_net = dev_net(dev),
1951 cfg.fc_gateway = *gwaddr;
1953 ip6_route_add(&cfg);
1955 return rt6_get_dflt_router(gwaddr, dev);
1959 int rt6_addrconf_purge(struct rt6_info *rt, void *arg) {
1960 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
1961 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2))
1966 void rt6_purge_dflt_routers(struct net *net)
1968 fib6_clean_all(net, rt6_addrconf_purge, 0, NULL);
1971 static void rtmsg_to_fib6_config(struct net *net,
1972 struct in6_rtmsg *rtmsg,
1973 struct fib6_config *cfg)
1975 memset(cfg, 0, sizeof(*cfg));
1977 cfg->fc_table = RT6_TABLE_MAIN;
1978 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1979 cfg->fc_metric = rtmsg->rtmsg_metric;
1980 cfg->fc_expires = rtmsg->rtmsg_info;
1981 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1982 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1983 cfg->fc_flags = rtmsg->rtmsg_flags;
1985 cfg->fc_nlinfo.nl_net = net;
1987 cfg->fc_dst = rtmsg->rtmsg_dst;
1988 cfg->fc_src = rtmsg->rtmsg_src;
1989 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1992 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1994 struct fib6_config cfg;
1995 struct in6_rtmsg rtmsg;
1999 case SIOCADDRT: /* Add a route */
2000 case SIOCDELRT: /* Delete a route */
2001 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2003 err = copy_from_user(&rtmsg, arg,
2004 sizeof(struct in6_rtmsg));
2008 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2013 err = ip6_route_add(&cfg);
2016 err = ip6_route_del(&cfg);
2030 * Drop the packet on the floor
2033 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2036 struct dst_entry *dst = skb_dst(skb);
2037 switch (ipstats_mib_noroutes) {
2038 case IPSTATS_MIB_INNOROUTES:
2039 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2040 if (type == IPV6_ADDR_ANY) {
2041 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2042 IPSTATS_MIB_INADDRERRORS);
2046 case IPSTATS_MIB_OUTNOROUTES:
2047 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2048 ipstats_mib_noroutes);
2051 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2056 static int ip6_pkt_discard(struct sk_buff *skb)
2058 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2061 static int ip6_pkt_discard_out(struct sk_buff *skb)
2063 skb->dev = skb_dst(skb)->dev;
2064 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2067 static int ip6_pkt_prohibit(struct sk_buff *skb)
2069 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2072 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2074 skb->dev = skb_dst(skb)->dev;
2075 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2079 * Allocate a dst for local (unicast / anycast) address.
2082 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2083 const struct in6_addr *addr,
2086 struct net *net = dev_net(idev->dev);
2087 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2090 return ERR_PTR(-ENOMEM);
2094 rt->dst.flags |= DST_HOST;
2095 rt->dst.input = ip6_input;
2096 rt->dst.output = ip6_output;
2097 rt->rt6i_idev = idev;
2099 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2101 rt->rt6i_flags |= RTF_ANYCAST;
2103 rt->rt6i_flags |= RTF_LOCAL;
2105 rt->rt6i_gateway = *addr;
2106 rt->rt6i_dst.addr = *addr;
2107 rt->rt6i_dst.plen = 128;
2108 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2110 atomic_set(&rt->dst.__refcnt, 1);
2115 int ip6_route_get_saddr(struct net *net,
2116 struct rt6_info *rt,
2117 const struct in6_addr *daddr,
2119 struct in6_addr *saddr)
2121 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2123 if (rt->rt6i_prefsrc.plen)
2124 *saddr = rt->rt6i_prefsrc.addr;
2126 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2127 daddr, prefs, saddr);
2131 /* remove deleted ip from prefsrc entries */
2132 struct arg_dev_net_ip {
2133 struct net_device *dev;
2135 struct in6_addr *addr;
2138 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2140 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2141 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2142 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2144 if (((void *)rt->dst.dev == dev || !dev) &&
2145 rt != net->ipv6.ip6_null_entry &&
2146 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2147 /* remove prefsrc entry */
2148 rt->rt6i_prefsrc.plen = 0;
2153 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2155 struct net *net = dev_net(ifp->idev->dev);
2156 struct arg_dev_net_ip adni = {
2157 .dev = ifp->idev->dev,
2161 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2164 struct arg_dev_net {
2165 struct net_device *dev;
2169 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2171 const struct arg_dev_net *adn = arg;
2172 const struct net_device *dev = adn->dev;
2174 if ((rt->dst.dev == dev || !dev) &&
2175 rt != adn->net->ipv6.ip6_null_entry)
2181 void rt6_ifdown(struct net *net, struct net_device *dev)
2183 struct arg_dev_net adn = {
2188 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2189 icmp6_clean_all(fib6_ifdown, &adn);
2192 struct rt6_mtu_change_arg {
2193 struct net_device *dev;
2197 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2199 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2200 struct inet6_dev *idev;
2202 /* In IPv6 pmtu discovery is not optional,
2203 so that RTAX_MTU lock cannot disable it.
2204 We still use this lock to block changes
2205 caused by addrconf/ndisc.
2208 idev = __in6_dev_get(arg->dev);
2212 /* For administrative MTU increase, there is no way to discover
2213 IPv6 PMTU increase, so PMTU increase should be updated here.
2214 Since RFC 1981 doesn't include administrative MTU increase
2215 update PMTU increase is a MUST. (i.e. jumbo frame)
2218 If new MTU is less than route PMTU, this new MTU will be the
2219 lowest MTU in the path, update the route PMTU to reflect PMTU
2220 decreases; if new MTU is greater than route PMTU, and the
2221 old MTU is the lowest MTU in the path, update the route PMTU
2222 to reflect the increase. In this case if the other nodes' MTU
2223 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2226 if (rt->dst.dev == arg->dev &&
2227 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2228 (dst_mtu(&rt->dst) >= arg->mtu ||
2229 (dst_mtu(&rt->dst) < arg->mtu &&
2230 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2231 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2236 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2238 struct rt6_mtu_change_arg arg = {
2243 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2246 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2247 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2248 [RTA_OIF] = { .type = NLA_U32 },
2249 [RTA_IIF] = { .type = NLA_U32 },
2250 [RTA_PRIORITY] = { .type = NLA_U32 },
2251 [RTA_METRICS] = { .type = NLA_NESTED },
2252 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2255 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2256 struct fib6_config *cfg)
2259 struct nlattr *tb[RTA_MAX+1];
2262 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2267 rtm = nlmsg_data(nlh);
2268 memset(cfg, 0, sizeof(*cfg));
2270 cfg->fc_table = rtm->rtm_table;
2271 cfg->fc_dst_len = rtm->rtm_dst_len;
2272 cfg->fc_src_len = rtm->rtm_src_len;
2273 cfg->fc_flags = RTF_UP;
2274 cfg->fc_protocol = rtm->rtm_protocol;
2275 cfg->fc_type = rtm->rtm_type;
2277 if (rtm->rtm_type == RTN_UNREACHABLE ||
2278 rtm->rtm_type == RTN_BLACKHOLE ||
2279 rtm->rtm_type == RTN_PROHIBIT ||
2280 rtm->rtm_type == RTN_THROW)
2281 cfg->fc_flags |= RTF_REJECT;
2283 if (rtm->rtm_type == RTN_LOCAL)
2284 cfg->fc_flags |= RTF_LOCAL;
2286 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2287 cfg->fc_nlinfo.nlh = nlh;
2288 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2290 if (tb[RTA_GATEWAY]) {
2291 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2292 cfg->fc_flags |= RTF_GATEWAY;
2296 int plen = (rtm->rtm_dst_len + 7) >> 3;
2298 if (nla_len(tb[RTA_DST]) < plen)
2301 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2305 int plen = (rtm->rtm_src_len + 7) >> 3;
2307 if (nla_len(tb[RTA_SRC]) < plen)
2310 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2313 if (tb[RTA_PREFSRC])
2314 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2317 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2319 if (tb[RTA_PRIORITY])
2320 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2322 if (tb[RTA_METRICS]) {
2323 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2324 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2328 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2330 if (tb[RTA_MULTIPATH]) {
2331 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2332 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2340 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2342 struct fib6_config r_cfg;
2343 struct rtnexthop *rtnh;
2346 int err = 0, last_err = 0;
2349 rtnh = (struct rtnexthop *)cfg->fc_mp;
2350 remaining = cfg->fc_mp_len;
2352 /* Parse a Multipath Entry */
2353 while (rtnh_ok(rtnh, remaining)) {
2354 memcpy(&r_cfg, cfg, sizeof(*cfg));
2355 if (rtnh->rtnh_ifindex)
2356 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2358 attrlen = rtnh_attrlen(rtnh);
2360 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2362 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2364 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2365 r_cfg.fc_flags |= RTF_GATEWAY;
2368 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2371 /* If we are trying to remove a route, do not stop the
2372 * loop when ip6_route_del() fails (because next hop is
2373 * already gone), we should try to remove all next hops.
2376 /* If add fails, we should try to delete all
2377 * next hops that have been already added.
2383 /* Because each route is added like a single route we remove
2384 * this flag after the first nexthop (if there is a collision,
2385 * we have already fail to add the first nexthop:
2386 * fib6_add_rt2node() has reject it).
2388 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2389 rtnh = rtnh_next(rtnh, &remaining);
2395 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2397 struct fib6_config cfg;
2400 err = rtm_to_fib6_config(skb, nlh, &cfg);
2405 return ip6_route_multipath(&cfg, 0);
2407 return ip6_route_del(&cfg);
2410 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2412 struct fib6_config cfg;
2415 err = rtm_to_fib6_config(skb, nlh, &cfg);
2420 return ip6_route_multipath(&cfg, 1);
2422 return ip6_route_add(&cfg);
2425 static inline size_t rt6_nlmsg_size(void)
2427 return NLMSG_ALIGN(sizeof(struct rtmsg))
2428 + nla_total_size(16) /* RTA_SRC */
2429 + nla_total_size(16) /* RTA_DST */
2430 + nla_total_size(16) /* RTA_GATEWAY */
2431 + nla_total_size(16) /* RTA_PREFSRC */
2432 + nla_total_size(4) /* RTA_TABLE */
2433 + nla_total_size(4) /* RTA_IIF */
2434 + nla_total_size(4) /* RTA_OIF */
2435 + nla_total_size(4) /* RTA_PRIORITY */
2436 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2437 + nla_total_size(sizeof(struct rta_cacheinfo));
2440 static int rt6_fill_node(struct net *net,
2441 struct sk_buff *skb, struct rt6_info *rt,
2442 struct in6_addr *dst, struct in6_addr *src,
2443 int iif, int type, u32 portid, u32 seq,
2444 int prefix, int nowait, unsigned int flags)
2447 struct nlmsghdr *nlh;
2451 if (prefix) { /* user wants prefix routes only */
2452 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2453 /* success since this is not a prefix route */
2458 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2462 rtm = nlmsg_data(nlh);
2463 rtm->rtm_family = AF_INET6;
2464 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2465 rtm->rtm_src_len = rt->rt6i_src.plen;
2468 table = rt->rt6i_table->tb6_id;
2470 table = RT6_TABLE_UNSPEC;
2471 rtm->rtm_table = table;
2472 if (nla_put_u32(skb, RTA_TABLE, table))
2473 goto nla_put_failure;
2474 if (rt->rt6i_flags & RTF_REJECT) {
2475 switch (rt->dst.error) {
2477 rtm->rtm_type = RTN_BLACKHOLE;
2480 rtm->rtm_type = RTN_PROHIBIT;
2483 rtm->rtm_type = RTN_THROW;
2486 rtm->rtm_type = RTN_UNREACHABLE;
2490 else if (rt->rt6i_flags & RTF_LOCAL)
2491 rtm->rtm_type = RTN_LOCAL;
2492 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2493 rtm->rtm_type = RTN_LOCAL;
2495 rtm->rtm_type = RTN_UNICAST;
2497 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2498 rtm->rtm_protocol = rt->rt6i_protocol;
2499 if (rt->rt6i_flags & RTF_DYNAMIC)
2500 rtm->rtm_protocol = RTPROT_REDIRECT;
2501 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2502 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2503 rtm->rtm_protocol = RTPROT_RA;
2505 rtm->rtm_protocol = RTPROT_KERNEL;
2508 if (rt->rt6i_flags & RTF_CACHE)
2509 rtm->rtm_flags |= RTM_F_CLONED;
2512 if (nla_put(skb, RTA_DST, 16, dst))
2513 goto nla_put_failure;
2514 rtm->rtm_dst_len = 128;
2515 } else if (rtm->rtm_dst_len)
2516 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2517 goto nla_put_failure;
2518 #ifdef CONFIG_IPV6_SUBTREES
2520 if (nla_put(skb, RTA_SRC, 16, src))
2521 goto nla_put_failure;
2522 rtm->rtm_src_len = 128;
2523 } else if (rtm->rtm_src_len &&
2524 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2525 goto nla_put_failure;
2528 #ifdef CONFIG_IPV6_MROUTE
2529 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2530 int err = ip6mr_get_route(net, skb, rtm, nowait);
2535 goto nla_put_failure;
2537 if (err == -EMSGSIZE)
2538 goto nla_put_failure;
2543 if (nla_put_u32(skb, RTA_IIF, iif))
2544 goto nla_put_failure;
2546 struct in6_addr saddr_buf;
2547 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2548 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2549 goto nla_put_failure;
2552 if (rt->rt6i_prefsrc.plen) {
2553 struct in6_addr saddr_buf;
2554 saddr_buf = rt->rt6i_prefsrc.addr;
2555 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2556 goto nla_put_failure;
2559 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2560 goto nla_put_failure;
2562 if (rt->rt6i_flags & RTF_GATEWAY) {
2563 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2564 goto nla_put_failure;
2568 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2569 goto nla_put_failure;
2570 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2571 goto nla_put_failure;
2573 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2575 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2576 goto nla_put_failure;
2578 return nlmsg_end(skb, nlh);
2581 nlmsg_cancel(skb, nlh);
2585 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2587 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2590 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2591 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2592 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2596 return rt6_fill_node(arg->net,
2597 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2598 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2599 prefix, 0, NLM_F_MULTI);
2602 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2604 struct net *net = sock_net(in_skb->sk);
2605 struct nlattr *tb[RTA_MAX+1];
2606 struct rt6_info *rt;
2607 struct sk_buff *skb;
2610 int err, iif = 0, oif = 0;
2612 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2617 memset(&fl6, 0, sizeof(fl6));
2620 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2623 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2627 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2630 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2634 iif = nla_get_u32(tb[RTA_IIF]);
2637 oif = nla_get_u32(tb[RTA_OIF]);
2640 struct net_device *dev;
2643 dev = __dev_get_by_index(net, iif);
2649 fl6.flowi6_iif = iif;
2651 if (!ipv6_addr_any(&fl6.saddr))
2652 flags |= RT6_LOOKUP_F_HAS_SADDR;
2654 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2657 fl6.flowi6_oif = oif;
2659 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2662 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2669 /* Reserve room for dummy headers, this skb can pass
2670 through good chunk of routing engine.
2672 skb_reset_mac_header(skb);
2673 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2675 skb_dst_set(skb, &rt->dst);
2677 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2678 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2679 nlh->nlmsg_seq, 0, 0, 0);
2685 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2690 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2692 struct sk_buff *skb;
2693 struct net *net = info->nl_net;
2698 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2700 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2704 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2705 event, info->portid, seq, 0, 0, 0);
2707 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2708 WARN_ON(err == -EMSGSIZE);
2712 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2713 info->nlh, gfp_any());
2717 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2720 static int ip6_route_dev_notify(struct notifier_block *this,
2721 unsigned long event, void *data)
2723 struct net_device *dev = (struct net_device *)data;
2724 struct net *net = dev_net(dev);
2726 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2727 net->ipv6.ip6_null_entry->dst.dev = dev;
2728 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2729 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2730 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2731 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2732 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2733 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2744 #ifdef CONFIG_PROC_FS
2755 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2757 struct seq_file *m = p_arg;
2759 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2761 #ifdef CONFIG_IPV6_SUBTREES
2762 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2764 seq_puts(m, "00000000000000000000000000000000 00 ");
2766 if (rt->rt6i_flags & RTF_GATEWAY) {
2767 seq_printf(m, "%pi6", &rt->rt6i_gateway);
2769 seq_puts(m, "00000000000000000000000000000000");
2771 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2772 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2773 rt->dst.__use, rt->rt6i_flags,
2774 rt->dst.dev ? rt->dst.dev->name : "");
2778 static int ipv6_route_show(struct seq_file *m, void *v)
2780 struct net *net = (struct net *)m->private;
2781 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2785 static int ipv6_route_open(struct inode *inode, struct file *file)
2787 return single_open_net(inode, file, ipv6_route_show);
2790 static const struct file_operations ipv6_route_proc_fops = {
2791 .owner = THIS_MODULE,
2792 .open = ipv6_route_open,
2794 .llseek = seq_lseek,
2795 .release = single_release_net,
2798 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2800 struct net *net = (struct net *)seq->private;
2801 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2802 net->ipv6.rt6_stats->fib_nodes,
2803 net->ipv6.rt6_stats->fib_route_nodes,
2804 net->ipv6.rt6_stats->fib_rt_alloc,
2805 net->ipv6.rt6_stats->fib_rt_entries,
2806 net->ipv6.rt6_stats->fib_rt_cache,
2807 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2808 net->ipv6.rt6_stats->fib_discarded_routes);
2813 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2815 return single_open_net(inode, file, rt6_stats_seq_show);
2818 static const struct file_operations rt6_stats_seq_fops = {
2819 .owner = THIS_MODULE,
2820 .open = rt6_stats_seq_open,
2822 .llseek = seq_lseek,
2823 .release = single_release_net,
2825 #endif /* CONFIG_PROC_FS */
2827 #ifdef CONFIG_SYSCTL
2830 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2831 void __user *buffer, size_t *lenp, loff_t *ppos)
2838 net = (struct net *)ctl->extra1;
2839 delay = net->ipv6.sysctl.flush_delay;
2840 proc_dointvec(ctl, write, buffer, lenp, ppos);
2841 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2845 ctl_table ipv6_route_table_template[] = {
2847 .procname = "flush",
2848 .data = &init_net.ipv6.sysctl.flush_delay,
2849 .maxlen = sizeof(int),
2851 .proc_handler = ipv6_sysctl_rtcache_flush
2854 .procname = "gc_thresh",
2855 .data = &ip6_dst_ops_template.gc_thresh,
2856 .maxlen = sizeof(int),
2858 .proc_handler = proc_dointvec,
2861 .procname = "max_size",
2862 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2863 .maxlen = sizeof(int),
2865 .proc_handler = proc_dointvec,
2868 .procname = "gc_min_interval",
2869 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2870 .maxlen = sizeof(int),
2872 .proc_handler = proc_dointvec_jiffies,
2875 .procname = "gc_timeout",
2876 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2877 .maxlen = sizeof(int),
2879 .proc_handler = proc_dointvec_jiffies,
2882 .procname = "gc_interval",
2883 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2884 .maxlen = sizeof(int),
2886 .proc_handler = proc_dointvec_jiffies,
2889 .procname = "gc_elasticity",
2890 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2891 .maxlen = sizeof(int),
2893 .proc_handler = proc_dointvec,
2896 .procname = "mtu_expires",
2897 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2898 .maxlen = sizeof(int),
2900 .proc_handler = proc_dointvec_jiffies,
2903 .procname = "min_adv_mss",
2904 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2905 .maxlen = sizeof(int),
2907 .proc_handler = proc_dointvec,
2910 .procname = "gc_min_interval_ms",
2911 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2912 .maxlen = sizeof(int),
2914 .proc_handler = proc_dointvec_ms_jiffies,
2919 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2921 struct ctl_table *table;
2923 table = kmemdup(ipv6_route_table_template,
2924 sizeof(ipv6_route_table_template),
2928 table[0].data = &net->ipv6.sysctl.flush_delay;
2929 table[0].extra1 = net;
2930 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2931 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2932 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2933 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2934 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2935 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2936 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2937 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2938 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2940 /* Don't export sysctls to unprivileged users */
2941 if (net->user_ns != &init_user_ns)
2942 table[0].procname = NULL;
2949 static int __net_init ip6_route_net_init(struct net *net)
2953 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2954 sizeof(net->ipv6.ip6_dst_ops));
2956 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2957 goto out_ip6_dst_ops;
2959 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2960 sizeof(*net->ipv6.ip6_null_entry),
2962 if (!net->ipv6.ip6_null_entry)
2963 goto out_ip6_dst_entries;
2964 net->ipv6.ip6_null_entry->dst.path =
2965 (struct dst_entry *)net->ipv6.ip6_null_entry;
2966 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2967 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2968 ip6_template_metrics, true);
2970 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2971 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2972 sizeof(*net->ipv6.ip6_prohibit_entry),
2974 if (!net->ipv6.ip6_prohibit_entry)
2975 goto out_ip6_null_entry;
2976 net->ipv6.ip6_prohibit_entry->dst.path =
2977 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2978 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2979 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2980 ip6_template_metrics, true);
2982 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2983 sizeof(*net->ipv6.ip6_blk_hole_entry),
2985 if (!net->ipv6.ip6_blk_hole_entry)
2986 goto out_ip6_prohibit_entry;
2987 net->ipv6.ip6_blk_hole_entry->dst.path =
2988 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2989 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2990 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2991 ip6_template_metrics, true);
2994 net->ipv6.sysctl.flush_delay = 0;
2995 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2996 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2997 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2998 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2999 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3000 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3001 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3003 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3009 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3010 out_ip6_prohibit_entry:
3011 kfree(net->ipv6.ip6_prohibit_entry);
3013 kfree(net->ipv6.ip6_null_entry);
3015 out_ip6_dst_entries:
3016 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3021 static void __net_exit ip6_route_net_exit(struct net *net)
3023 kfree(net->ipv6.ip6_null_entry);
3024 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3025 kfree(net->ipv6.ip6_prohibit_entry);
3026 kfree(net->ipv6.ip6_blk_hole_entry);
3028 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3031 static int __net_init ip6_route_net_init_late(struct net *net)
3033 #ifdef CONFIG_PROC_FS
3034 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3035 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3040 static void __net_exit ip6_route_net_exit_late(struct net *net)
3042 #ifdef CONFIG_PROC_FS
3043 remove_proc_entry("ipv6_route", net->proc_net);
3044 remove_proc_entry("rt6_stats", net->proc_net);
3048 static struct pernet_operations ip6_route_net_ops = {
3049 .init = ip6_route_net_init,
3050 .exit = ip6_route_net_exit,
3053 static int __net_init ipv6_inetpeer_init(struct net *net)
3055 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3059 inet_peer_base_init(bp);
3060 net->ipv6.peers = bp;
3064 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3066 struct inet_peer_base *bp = net->ipv6.peers;
3068 net->ipv6.peers = NULL;
3069 inetpeer_invalidate_tree(bp);
3073 static struct pernet_operations ipv6_inetpeer_ops = {
3074 .init = ipv6_inetpeer_init,
3075 .exit = ipv6_inetpeer_exit,
3078 static struct pernet_operations ip6_route_net_late_ops = {
3079 .init = ip6_route_net_init_late,
3080 .exit = ip6_route_net_exit_late,
3083 static struct notifier_block ip6_route_dev_notifier = {
3084 .notifier_call = ip6_route_dev_notify,
3088 int __init ip6_route_init(void)
3093 ip6_dst_ops_template.kmem_cachep =
3094 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3095 SLAB_HWCACHE_ALIGN, NULL);
3096 if (!ip6_dst_ops_template.kmem_cachep)
3099 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3101 goto out_kmem_cache;
3103 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3105 goto out_dst_entries;
3107 ret = register_pernet_subsys(&ip6_route_net_ops);
3109 goto out_register_inetpeer;
3111 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3113 /* Registering of the loopback is done before this portion of code,
3114 * the loopback reference in rt6_info will not be taken, do it
3115 * manually for init_net */
3116 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3117 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3118 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3119 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3120 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3121 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3122 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3126 goto out_register_subsys;
3132 ret = fib6_rules_init();
3136 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3138 goto fib6_rules_init;
3141 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3142 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3143 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3144 goto out_register_late_subsys;
3146 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3148 goto out_register_late_subsys;
3153 out_register_late_subsys:
3154 unregister_pernet_subsys(&ip6_route_net_late_ops);
3156 fib6_rules_cleanup();
3161 out_register_subsys:
3162 unregister_pernet_subsys(&ip6_route_net_ops);
3163 out_register_inetpeer:
3164 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3166 dst_entries_destroy(&ip6_dst_blackhole_ops);
3168 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3172 void ip6_route_cleanup(void)
3174 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3175 unregister_pernet_subsys(&ip6_route_net_late_ops);
3176 fib6_rules_cleanup();
3179 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3180 unregister_pernet_subsys(&ip6_route_net_ops);
3181 dst_entries_destroy(&ip6_dst_blackhole_ops);
3182 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);