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 *net,
97 const struct in6_addr *prefix, int prefixlen,
98 const struct in6_addr *gwaddr, int ifindex,
100 static struct rt6_info *rt6_get_route_info(struct net *net,
101 const struct in6_addr *prefix, int prefixlen,
102 const struct in6_addr *gwaddr, int ifindex);
105 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
107 struct rt6_info *rt = (struct rt6_info *) dst;
108 struct inet_peer *peer;
111 if (!(rt->dst.flags & DST_HOST))
112 return dst_cow_metrics_generic(dst, old);
114 peer = rt6_get_peer_create(rt);
116 u32 *old_p = __DST_METRICS_PTR(old);
117 unsigned long prev, new;
120 if (inet_metrics_new(peer))
121 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
123 new = (unsigned long) p;
124 prev = cmpxchg(&dst->_metrics, old, new);
127 p = __DST_METRICS_PTR(prev);
128 if (prev & DST_METRICS_READ_ONLY)
135 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
139 struct in6_addr *p = &rt->rt6i_gateway;
141 if (!ipv6_addr_any(p))
142 return (const void *) p;
144 return &ipv6_hdr(skb)->daddr;
148 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
152 struct rt6_info *rt = (struct rt6_info *) dst;
155 daddr = choose_neigh_daddr(rt, skb, daddr);
156 n = __ipv6_neigh_lookup(dst->dev, daddr);
159 return neigh_create(&nd_tbl, daddr, dst->dev);
162 static struct dst_ops ip6_dst_ops_template = {
164 .protocol = cpu_to_be16(ETH_P_IPV6),
167 .check = ip6_dst_check,
168 .default_advmss = ip6_default_advmss,
170 .cow_metrics = ipv6_cow_metrics,
171 .destroy = ip6_dst_destroy,
172 .ifdown = ip6_dst_ifdown,
173 .negative_advice = ip6_negative_advice,
174 .link_failure = ip6_link_failure,
175 .update_pmtu = ip6_rt_update_pmtu,
176 .redirect = rt6_do_redirect,
177 .local_out = __ip6_local_out,
178 .neigh_lookup = ip6_neigh_lookup,
181 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
183 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
185 return mtu ? : dst->dev->mtu;
188 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
189 struct sk_buff *skb, u32 mtu)
193 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
198 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
204 static struct dst_ops ip6_dst_blackhole_ops = {
206 .protocol = cpu_to_be16(ETH_P_IPV6),
207 .destroy = ip6_dst_destroy,
208 .check = ip6_dst_check,
209 .mtu = ip6_blackhole_mtu,
210 .default_advmss = ip6_default_advmss,
211 .update_pmtu = ip6_rt_blackhole_update_pmtu,
212 .redirect = ip6_rt_blackhole_redirect,
213 .cow_metrics = ip6_rt_blackhole_cow_metrics,
214 .neigh_lookup = ip6_neigh_lookup,
217 static const u32 ip6_template_metrics[RTAX_MAX] = {
218 [RTAX_HOPLIMIT - 1] = 0,
221 static const struct rt6_info ip6_null_entry_template = {
223 .__refcnt = ATOMIC_INIT(1),
225 .obsolete = DST_OBSOLETE_FORCE_CHK,
226 .error = -ENETUNREACH,
227 .input = ip6_pkt_discard,
228 .output = ip6_pkt_discard_out,
230 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
231 .rt6i_protocol = RTPROT_KERNEL,
232 .rt6i_metric = ~(u32) 0,
233 .rt6i_ref = ATOMIC_INIT(1),
236 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
238 static const struct rt6_info ip6_prohibit_entry_template = {
240 .__refcnt = ATOMIC_INIT(1),
242 .obsolete = DST_OBSOLETE_FORCE_CHK,
244 .input = ip6_pkt_prohibit,
245 .output = ip6_pkt_prohibit_out,
247 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
248 .rt6i_protocol = RTPROT_KERNEL,
249 .rt6i_metric = ~(u32) 0,
250 .rt6i_ref = ATOMIC_INIT(1),
253 static const struct rt6_info ip6_blk_hole_entry_template = {
255 .__refcnt = ATOMIC_INIT(1),
257 .obsolete = DST_OBSOLETE_FORCE_CHK,
259 .input = dst_discard,
260 .output = dst_discard,
262 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
263 .rt6i_protocol = RTPROT_KERNEL,
264 .rt6i_metric = ~(u32) 0,
265 .rt6i_ref = ATOMIC_INIT(1),
270 /* allocate dst with ip6_dst_ops */
271 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
272 struct net_device *dev,
274 struct fib6_table *table)
276 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
277 0, DST_OBSOLETE_FORCE_CHK, flags);
280 struct dst_entry *dst = &rt->dst;
282 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
283 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
284 rt->rt6i_genid = rt_genid(net);
285 INIT_LIST_HEAD(&rt->rt6i_siblings);
286 rt->rt6i_nsiblings = 0;
291 static void ip6_dst_destroy(struct dst_entry *dst)
293 struct rt6_info *rt = (struct rt6_info *)dst;
294 struct inet6_dev *idev = rt->rt6i_idev;
295 struct dst_entry *from = dst->from;
297 if (!(rt->dst.flags & DST_HOST))
298 dst_destroy_metrics_generic(dst);
301 rt->rt6i_idev = NULL;
308 if (rt6_has_peer(rt)) {
309 struct inet_peer *peer = rt6_peer_ptr(rt);
314 void rt6_bind_peer(struct rt6_info *rt, int create)
316 struct inet_peer_base *base;
317 struct inet_peer *peer;
319 base = inetpeer_base_ptr(rt->_rt6i_peer);
323 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
325 if (!rt6_set_peer(rt, peer))
330 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
333 struct rt6_info *rt = (struct rt6_info *)dst;
334 struct inet6_dev *idev = rt->rt6i_idev;
335 struct net_device *loopback_dev =
336 dev_net(dev)->loopback_dev;
338 if (dev != loopback_dev) {
339 if (idev && idev->dev == dev) {
340 struct inet6_dev *loopback_idev =
341 in6_dev_get(loopback_dev);
343 rt->rt6i_idev = loopback_idev;
350 static bool rt6_check_expired(const struct rt6_info *rt)
352 if (rt->rt6i_flags & RTF_EXPIRES) {
353 if (time_after(jiffies, rt->dst.expires))
355 } else if (rt->dst.from) {
356 return rt6_check_expired((struct rt6_info *) rt->dst.from);
361 static bool rt6_need_strict(const struct in6_addr *daddr)
363 return ipv6_addr_type(daddr) &
364 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
367 /* Multipath route selection:
368 * Hash based function using packet header and flowlabel.
369 * Adapted from fib_info_hashfn()
371 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
372 const struct flowi6 *fl6)
374 unsigned int val = fl6->flowi6_proto;
376 val ^= ipv6_addr_hash(&fl6->daddr);
377 val ^= ipv6_addr_hash(&fl6->saddr);
379 /* Work only if this not encapsulated */
380 switch (fl6->flowi6_proto) {
384 val ^= (__force u16)fl6->fl6_sport;
385 val ^= (__force u16)fl6->fl6_dport;
389 val ^= (__force u16)fl6->fl6_icmp_type;
390 val ^= (__force u16)fl6->fl6_icmp_code;
393 /* RFC6438 recommands to use flowlabel */
394 val ^= (__force u32)fl6->flowlabel;
396 /* Perhaps, we need to tune, this function? */
397 val = val ^ (val >> 7) ^ (val >> 12);
398 return val % candidate_count;
401 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
404 struct rt6_info *sibling, *next_sibling;
407 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
408 /* Don't change the route, if route_choosen == 0
409 * (siblings does not include ourself)
412 list_for_each_entry_safe(sibling, next_sibling,
413 &match->rt6i_siblings, rt6i_siblings) {
415 if (route_choosen == 0) {
424 * Route lookup. Any table->tb6_lock is implied.
427 static inline struct rt6_info *rt6_device_match(struct net *net,
429 const struct in6_addr *saddr,
433 struct rt6_info *local = NULL;
434 struct rt6_info *sprt;
436 if (!oif && ipv6_addr_any(saddr))
439 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
440 struct net_device *dev = sprt->dst.dev;
443 if (dev->ifindex == oif)
445 if (dev->flags & IFF_LOOPBACK) {
446 if (!sprt->rt6i_idev ||
447 sprt->rt6i_idev->dev->ifindex != oif) {
448 if (flags & RT6_LOOKUP_F_IFACE && oif)
450 if (local && (!oif ||
451 local->rt6i_idev->dev->ifindex == oif))
457 if (ipv6_chk_addr(net, saddr, dev,
458 flags & RT6_LOOKUP_F_IFACE))
467 if (flags & RT6_LOOKUP_F_IFACE)
468 return net->ipv6.ip6_null_entry;
474 #ifdef CONFIG_IPV6_ROUTER_PREF
475 struct __rt6_probe_work {
476 struct work_struct work;
477 struct in6_addr target;
478 struct net_device *dev;
481 static void rt6_probe_deferred(struct work_struct *w)
483 struct in6_addr mcaddr;
484 struct __rt6_probe_work *work =
485 container_of(w, struct __rt6_probe_work, work);
487 addrconf_addr_solict_mult(&work->target, &mcaddr);
488 ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
493 static void rt6_probe(struct rt6_info *rt)
495 struct neighbour *neigh;
497 * Okay, this does not seem to be appropriate
498 * for now, however, we need to check if it
499 * is really so; aka Router Reachability Probing.
501 * Router Reachability Probe MUST be rate-limited
502 * to no more than one per minute.
504 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
507 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
509 write_lock(&neigh->lock);
510 if (neigh->nud_state & NUD_VALID)
515 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
516 struct __rt6_probe_work *work;
518 work = kmalloc(sizeof(*work), GFP_ATOMIC);
521 neigh->updated = jiffies;
524 write_unlock(&neigh->lock);
527 INIT_WORK(&work->work, rt6_probe_deferred);
528 work->target = rt->rt6i_gateway;
529 dev_hold(rt->dst.dev);
530 work->dev = rt->dst.dev;
531 schedule_work(&work->work);
535 write_unlock(&neigh->lock);
537 rcu_read_unlock_bh();
540 static inline void rt6_probe(struct rt6_info *rt)
546 * Default Router Selection (RFC 2461 6.3.6)
548 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
550 struct net_device *dev = rt->dst.dev;
551 if (!oif || dev->ifindex == oif)
553 if ((dev->flags & IFF_LOOPBACK) &&
554 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
559 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
561 struct neighbour *neigh;
562 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
564 if (rt->rt6i_flags & RTF_NONEXTHOP ||
565 !(rt->rt6i_flags & RTF_GATEWAY))
566 return RT6_NUD_SUCCEED;
569 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
571 read_lock(&neigh->lock);
572 if (neigh->nud_state & NUD_VALID)
573 ret = RT6_NUD_SUCCEED;
574 #ifdef CONFIG_IPV6_ROUTER_PREF
575 else if (!(neigh->nud_state & NUD_FAILED))
576 ret = RT6_NUD_SUCCEED;
578 read_unlock(&neigh->lock);
580 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
581 RT6_NUD_SUCCEED : RT6_NUD_FAIL_SOFT;
583 rcu_read_unlock_bh();
588 static int rt6_score_route(struct rt6_info *rt, int oif,
593 m = rt6_check_dev(rt, oif);
594 if (!m && (strict & RT6_LOOKUP_F_IFACE))
595 return RT6_NUD_FAIL_HARD;
596 #ifdef CONFIG_IPV6_ROUTER_PREF
597 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
599 if (strict & RT6_LOOKUP_F_REACHABLE) {
600 int n = rt6_check_neigh(rt);
607 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
608 int *mpri, struct rt6_info *match,
612 bool match_do_rr = false;
614 if (rt6_check_expired(rt))
617 m = rt6_score_route(rt, oif, strict);
618 if (m == RT6_NUD_FAIL_SOFT && !IS_ENABLED(CONFIG_IPV6_ROUTER_PREF)) {
620 m = 0; /* lowest valid score */
625 if (strict & RT6_LOOKUP_F_REACHABLE)
629 *do_rr = match_do_rr;
637 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
638 struct rt6_info *rr_head,
639 u32 metric, int oif, int strict,
642 struct rt6_info *rt, *match;
646 for (rt = rr_head; rt && rt->rt6i_metric == metric;
647 rt = rt->dst.rt6_next)
648 match = find_match(rt, oif, strict, &mpri, match, do_rr);
649 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
650 rt = rt->dst.rt6_next)
651 match = find_match(rt, oif, strict, &mpri, match, do_rr);
656 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
658 struct rt6_info *match, *rt0;
664 fn->rr_ptr = rt0 = fn->leaf;
666 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
670 struct rt6_info *next = rt0->dst.rt6_next;
672 /* no entries matched; do round-robin */
673 if (!next || next->rt6i_metric != rt0->rt6i_metric)
680 net = dev_net(rt0->dst.dev);
681 return match ? match : net->ipv6.ip6_null_entry;
684 #ifdef CONFIG_IPV6_ROUTE_INFO
685 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
686 const struct in6_addr *gwaddr)
688 struct net *net = dev_net(dev);
689 struct route_info *rinfo = (struct route_info *) opt;
690 struct in6_addr prefix_buf, *prefix;
692 unsigned long lifetime;
695 if (len < sizeof(struct route_info)) {
699 /* Sanity check for prefix_len and length */
700 if (rinfo->length > 3) {
702 } else if (rinfo->prefix_len > 128) {
704 } else if (rinfo->prefix_len > 64) {
705 if (rinfo->length < 2) {
708 } else if (rinfo->prefix_len > 0) {
709 if (rinfo->length < 1) {
714 pref = rinfo->route_pref;
715 if (pref == ICMPV6_ROUTER_PREF_INVALID)
718 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
720 if (rinfo->length == 3)
721 prefix = (struct in6_addr *)rinfo->prefix;
723 /* this function is safe */
724 ipv6_addr_prefix(&prefix_buf,
725 (struct in6_addr *)rinfo->prefix,
727 prefix = &prefix_buf;
730 if (rinfo->prefix_len == 0)
731 rt = rt6_get_dflt_router(gwaddr, dev);
733 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
734 gwaddr, dev->ifindex);
736 if (rt && !lifetime) {
742 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
745 rt->rt6i_flags = RTF_ROUTEINFO |
746 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
749 if (!addrconf_finite_timeout(lifetime))
750 rt6_clean_expires(rt);
752 rt6_set_expires(rt, jiffies + HZ * lifetime);
760 #define BACKTRACK(__net, saddr) \
762 if (rt == __net->ipv6.ip6_null_entry) { \
763 struct fib6_node *pn; \
765 if (fn->fn_flags & RTN_TL_ROOT) \
768 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
769 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
772 if (fn->fn_flags & RTN_RTINFO) \
778 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
779 struct fib6_table *table,
780 struct flowi6 *fl6, int flags)
782 struct fib6_node *fn;
785 read_lock_bh(&table->tb6_lock);
786 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
789 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
790 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
791 rt = rt6_multipath_select(rt, fl6);
792 BACKTRACK(net, &fl6->saddr);
794 dst_use(&rt->dst, jiffies);
795 read_unlock_bh(&table->tb6_lock);
800 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
803 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
805 EXPORT_SYMBOL_GPL(ip6_route_lookup);
807 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
808 const struct in6_addr *saddr, int oif, int strict)
810 struct flowi6 fl6 = {
814 struct dst_entry *dst;
815 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
818 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
819 flags |= RT6_LOOKUP_F_HAS_SADDR;
822 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
824 return (struct rt6_info *) dst;
831 EXPORT_SYMBOL(rt6_lookup);
833 /* ip6_ins_rt is called with FREE table->tb6_lock.
834 It takes new route entry, the addition fails by any reason the
835 route is freed. In any case, if caller does not hold it, it may
839 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
842 struct fib6_table *table;
844 table = rt->rt6i_table;
845 write_lock_bh(&table->tb6_lock);
846 err = fib6_add(&table->tb6_root, rt, info);
847 write_unlock_bh(&table->tb6_lock);
852 int ip6_ins_rt(struct rt6_info *rt)
854 struct nl_info info = {
855 .nl_net = dev_net(rt->dst.dev),
857 return __ip6_ins_rt(rt, &info);
860 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
861 const struct in6_addr *daddr,
862 const struct in6_addr *saddr)
870 rt = ip6_rt_copy(ort, daddr);
873 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
874 if (ort->rt6i_dst.plen != 128 &&
875 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
876 rt->rt6i_flags |= RTF_ANYCAST;
879 rt->rt6i_flags |= RTF_CACHE;
881 #ifdef CONFIG_IPV6_SUBTREES
882 if (rt->rt6i_src.plen && saddr) {
883 rt->rt6i_src.addr = *saddr;
884 rt->rt6i_src.plen = 128;
892 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
893 const struct in6_addr *daddr)
895 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
898 rt->rt6i_flags |= RTF_CACHE;
902 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
903 struct flowi6 *fl6, int flags)
905 struct fib6_node *fn;
906 struct rt6_info *rt, *nrt;
910 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
912 strict |= flags & RT6_LOOKUP_F_IFACE;
915 read_lock_bh(&table->tb6_lock);
918 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
921 rt = rt6_select(fn, oif, strict | reachable);
922 if (rt->rt6i_nsiblings && oif == 0)
923 rt = rt6_multipath_select(rt, fl6);
924 BACKTRACK(net, &fl6->saddr);
925 if (rt == net->ipv6.ip6_null_entry ||
926 rt->rt6i_flags & RTF_CACHE)
930 read_unlock_bh(&table->tb6_lock);
932 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
933 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
934 else if (!(rt->dst.flags & DST_HOST))
935 nrt = rt6_alloc_clone(rt, &fl6->daddr);
940 rt = nrt ? : net->ipv6.ip6_null_entry;
944 err = ip6_ins_rt(nrt);
953 * Race condition! In the gap, when table->tb6_lock was
954 * released someone could insert this route. Relookup.
965 read_unlock_bh(&table->tb6_lock);
967 rt->dst.lastuse = jiffies;
973 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
974 struct flowi6 *fl6, int flags)
976 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
979 static struct dst_entry *ip6_route_input_lookup(struct net *net,
980 struct net_device *dev,
981 struct flowi6 *fl6, int flags)
983 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
984 flags |= RT6_LOOKUP_F_IFACE;
986 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
989 void ip6_route_input(struct sk_buff *skb)
991 const struct ipv6hdr *iph = ipv6_hdr(skb);
992 struct net *net = dev_net(skb->dev);
993 int flags = RT6_LOOKUP_F_HAS_SADDR;
994 struct flowi6 fl6 = {
995 .flowi6_iif = skb->dev->ifindex,
998 .flowlabel = ip6_flowinfo(iph),
999 .flowi6_mark = skb->mark,
1000 .flowi6_proto = iph->nexthdr,
1003 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1006 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1007 struct flowi6 *fl6, int flags)
1009 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1012 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1017 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1019 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1020 flags |= RT6_LOOKUP_F_IFACE;
1022 if (!ipv6_addr_any(&fl6->saddr))
1023 flags |= RT6_LOOKUP_F_HAS_SADDR;
1025 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1027 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1030 EXPORT_SYMBOL(ip6_route_output);
1032 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1034 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1035 struct dst_entry *new = NULL;
1037 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1041 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1042 rt6_init_peer(rt, net->ipv6.peers);
1045 new->input = dst_discard;
1046 new->output = dst_discard;
1048 if (dst_metrics_read_only(&ort->dst))
1049 new->_metrics = ort->dst._metrics;
1051 dst_copy_metrics(new, &ort->dst);
1052 rt->rt6i_idev = ort->rt6i_idev;
1054 in6_dev_hold(rt->rt6i_idev);
1056 rt->rt6i_gateway = ort->rt6i_gateway;
1057 rt->rt6i_flags = ort->rt6i_flags;
1058 rt->rt6i_metric = 0;
1060 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1061 #ifdef CONFIG_IPV6_SUBTREES
1062 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1068 dst_release(dst_orig);
1069 return new ? new : ERR_PTR(-ENOMEM);
1073 * Destination cache support functions
1076 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1078 struct rt6_info *rt;
1080 rt = (struct rt6_info *) dst;
1082 /* All IPV6 dsts are created with ->obsolete set to the value
1083 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1084 * into this function always.
1086 if (rt->rt6i_genid != rt_genid(dev_net(rt->dst.dev)))
1089 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1092 if (rt6_check_expired(rt))
1098 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1100 struct rt6_info *rt = (struct rt6_info *) dst;
1103 if (rt->rt6i_flags & RTF_CACHE) {
1104 if (rt6_check_expired(rt)) {
1116 static void ip6_link_failure(struct sk_buff *skb)
1118 struct rt6_info *rt;
1120 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1122 rt = (struct rt6_info *) skb_dst(skb);
1124 if (rt->rt6i_flags & RTF_CACHE) {
1128 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1129 rt->rt6i_node->fn_sernum = -1;
1134 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1135 struct sk_buff *skb, u32 mtu)
1137 struct rt6_info *rt6 = (struct rt6_info*)dst;
1140 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1141 struct net *net = dev_net(dst->dev);
1143 rt6->rt6i_flags |= RTF_MODIFIED;
1144 if (mtu < IPV6_MIN_MTU)
1147 dst_metric_set(dst, RTAX_MTU, mtu);
1148 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1152 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1155 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1156 struct dst_entry *dst;
1159 memset(&fl6, 0, sizeof(fl6));
1160 fl6.flowi6_oif = oif;
1161 fl6.flowi6_mark = mark;
1162 fl6.flowi6_flags = 0;
1163 fl6.daddr = iph->daddr;
1164 fl6.saddr = iph->saddr;
1165 fl6.flowlabel = ip6_flowinfo(iph);
1167 dst = ip6_route_output(net, NULL, &fl6);
1169 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1172 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1174 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1176 ip6_update_pmtu(skb, sock_net(sk), mtu,
1177 sk->sk_bound_dev_if, sk->sk_mark);
1179 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1181 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1183 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1184 struct dst_entry *dst;
1187 memset(&fl6, 0, sizeof(fl6));
1188 fl6.flowi6_oif = oif;
1189 fl6.flowi6_mark = mark;
1190 fl6.flowi6_flags = 0;
1191 fl6.daddr = iph->daddr;
1192 fl6.saddr = iph->saddr;
1193 fl6.flowlabel = ip6_flowinfo(iph);
1195 dst = ip6_route_output(net, NULL, &fl6);
1197 rt6_do_redirect(dst, NULL, skb);
1200 EXPORT_SYMBOL_GPL(ip6_redirect);
1202 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1204 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1206 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1208 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1210 struct net_device *dev = dst->dev;
1211 unsigned int mtu = dst_mtu(dst);
1212 struct net *net = dev_net(dev);
1214 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1216 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1217 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1220 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1221 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1222 * IPV6_MAXPLEN is also valid and means: "any MSS,
1223 * rely only on pmtu discovery"
1225 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1230 static unsigned int ip6_mtu(const struct dst_entry *dst)
1232 struct inet6_dev *idev;
1233 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1241 idev = __in6_dev_get(dst->dev);
1243 mtu = idev->cnf.mtu6;
1247 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1250 static struct dst_entry *icmp6_dst_gc_list;
1251 static DEFINE_SPINLOCK(icmp6_dst_lock);
1253 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1256 struct dst_entry *dst;
1257 struct rt6_info *rt;
1258 struct inet6_dev *idev = in6_dev_get(dev);
1259 struct net *net = dev_net(dev);
1261 if (unlikely(!idev))
1262 return ERR_PTR(-ENODEV);
1264 rt = ip6_dst_alloc(net, dev, 0, NULL);
1265 if (unlikely(!rt)) {
1267 dst = ERR_PTR(-ENOMEM);
1271 rt->dst.flags |= DST_HOST;
1272 rt->dst.output = ip6_output;
1273 atomic_set(&rt->dst.__refcnt, 1);
1274 rt->rt6i_gateway = fl6->daddr;
1275 rt->rt6i_dst.addr = fl6->daddr;
1276 rt->rt6i_dst.plen = 128;
1277 rt->rt6i_idev = idev;
1278 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1280 spin_lock_bh(&icmp6_dst_lock);
1281 rt->dst.next = icmp6_dst_gc_list;
1282 icmp6_dst_gc_list = &rt->dst;
1283 spin_unlock_bh(&icmp6_dst_lock);
1285 fib6_force_start_gc(net);
1287 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1293 int icmp6_dst_gc(void)
1295 struct dst_entry *dst, **pprev;
1298 spin_lock_bh(&icmp6_dst_lock);
1299 pprev = &icmp6_dst_gc_list;
1301 while ((dst = *pprev) != NULL) {
1302 if (!atomic_read(&dst->__refcnt)) {
1311 spin_unlock_bh(&icmp6_dst_lock);
1316 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1319 struct dst_entry *dst, **pprev;
1321 spin_lock_bh(&icmp6_dst_lock);
1322 pprev = &icmp6_dst_gc_list;
1323 while ((dst = *pprev) != NULL) {
1324 struct rt6_info *rt = (struct rt6_info *) dst;
1325 if (func(rt, arg)) {
1332 spin_unlock_bh(&icmp6_dst_lock);
1335 static int ip6_dst_gc(struct dst_ops *ops)
1337 unsigned long now = jiffies;
1338 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1339 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1340 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1341 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1342 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1343 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1346 entries = dst_entries_get_fast(ops);
1347 if (time_after(rt_last_gc + rt_min_interval, now) &&
1348 entries <= rt_max_size)
1351 net->ipv6.ip6_rt_gc_expire++;
1352 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1353 net->ipv6.ip6_rt_last_gc = now;
1354 entries = dst_entries_get_slow(ops);
1355 if (entries < ops->gc_thresh)
1356 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1358 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1359 return entries > rt_max_size;
1362 int ip6_dst_hoplimit(struct dst_entry *dst)
1364 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1365 if (hoplimit == 0) {
1366 struct net_device *dev = dst->dev;
1367 struct inet6_dev *idev;
1370 idev = __in6_dev_get(dev);
1372 hoplimit = idev->cnf.hop_limit;
1374 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1379 EXPORT_SYMBOL(ip6_dst_hoplimit);
1385 int ip6_route_add(struct fib6_config *cfg)
1388 struct net *net = cfg->fc_nlinfo.nl_net;
1389 struct rt6_info *rt = NULL;
1390 struct net_device *dev = NULL;
1391 struct inet6_dev *idev = NULL;
1392 struct fib6_table *table;
1395 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1397 #ifndef CONFIG_IPV6_SUBTREES
1398 if (cfg->fc_src_len)
1401 if (cfg->fc_ifindex) {
1403 dev = dev_get_by_index(net, cfg->fc_ifindex);
1406 idev = in6_dev_get(dev);
1411 if (cfg->fc_metric == 0)
1412 cfg->fc_metric = IP6_RT_PRIO_USER;
1415 if (cfg->fc_nlinfo.nlh &&
1416 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1417 table = fib6_get_table(net, cfg->fc_table);
1419 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1420 table = fib6_new_table(net, cfg->fc_table);
1423 table = fib6_new_table(net, cfg->fc_table);
1429 rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1436 if (cfg->fc_flags & RTF_EXPIRES)
1437 rt6_set_expires(rt, jiffies +
1438 clock_t_to_jiffies(cfg->fc_expires));
1440 rt6_clean_expires(rt);
1442 if (cfg->fc_protocol == RTPROT_UNSPEC)
1443 cfg->fc_protocol = RTPROT_BOOT;
1444 rt->rt6i_protocol = cfg->fc_protocol;
1446 addr_type = ipv6_addr_type(&cfg->fc_dst);
1448 if (addr_type & IPV6_ADDR_MULTICAST)
1449 rt->dst.input = ip6_mc_input;
1450 else if (cfg->fc_flags & RTF_LOCAL)
1451 rt->dst.input = ip6_input;
1453 rt->dst.input = ip6_forward;
1455 rt->dst.output = ip6_output;
1457 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1458 rt->rt6i_dst.plen = cfg->fc_dst_len;
1459 if (rt->rt6i_dst.plen == 128)
1460 rt->dst.flags |= DST_HOST;
1462 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1463 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1468 dst_init_metrics(&rt->dst, metrics, 0);
1470 #ifdef CONFIG_IPV6_SUBTREES
1471 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1472 rt->rt6i_src.plen = cfg->fc_src_len;
1475 rt->rt6i_metric = cfg->fc_metric;
1477 /* We cannot add true routes via loopback here,
1478 they would result in kernel looping; promote them to reject routes
1480 if ((cfg->fc_flags & RTF_REJECT) ||
1481 (dev && (dev->flags & IFF_LOOPBACK) &&
1482 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1483 !(cfg->fc_flags & RTF_LOCAL))) {
1484 /* hold loopback dev/idev if we haven't done so. */
1485 if (dev != net->loopback_dev) {
1490 dev = net->loopback_dev;
1492 idev = in6_dev_get(dev);
1498 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1499 switch (cfg->fc_type) {
1501 rt->dst.error = -EINVAL;
1502 rt->dst.output = dst_discard;
1503 rt->dst.input = dst_discard;
1506 rt->dst.error = -EACCES;
1507 rt->dst.output = ip6_pkt_prohibit_out;
1508 rt->dst.input = ip6_pkt_prohibit;
1512 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1514 rt->dst.output = ip6_pkt_discard_out;
1515 rt->dst.input = ip6_pkt_discard;
1521 if (cfg->fc_flags & RTF_GATEWAY) {
1522 const struct in6_addr *gw_addr;
1525 gw_addr = &cfg->fc_gateway;
1526 rt->rt6i_gateway = *gw_addr;
1527 gwa_type = ipv6_addr_type(gw_addr);
1529 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1530 struct rt6_info *grt;
1532 /* IPv6 strictly inhibits using not link-local
1533 addresses as nexthop address.
1534 Otherwise, router will not able to send redirects.
1535 It is very good, but in some (rare!) circumstances
1536 (SIT, PtP, NBMA NOARP links) it is handy to allow
1537 some exceptions. --ANK
1540 if (!(gwa_type & IPV6_ADDR_UNICAST))
1543 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1545 err = -EHOSTUNREACH;
1549 if (dev != grt->dst.dev) {
1555 idev = grt->rt6i_idev;
1557 in6_dev_hold(grt->rt6i_idev);
1559 if (!(grt->rt6i_flags & RTF_GATEWAY))
1567 if (!dev || (dev->flags & IFF_LOOPBACK))
1575 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1576 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1580 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1581 rt->rt6i_prefsrc.plen = 128;
1583 rt->rt6i_prefsrc.plen = 0;
1585 rt->rt6i_flags = cfg->fc_flags;
1592 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1593 int type = nla_type(nla);
1596 if (type > RTAX_MAX) {
1601 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1607 rt->rt6i_idev = idev;
1608 rt->rt6i_table = table;
1610 cfg->fc_nlinfo.nl_net = dev_net(dev);
1612 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1624 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1627 struct fib6_table *table;
1628 struct net *net = dev_net(rt->dst.dev);
1630 if (rt == net->ipv6.ip6_null_entry) {
1635 table = rt->rt6i_table;
1636 write_lock_bh(&table->tb6_lock);
1637 err = fib6_del(rt, info);
1638 write_unlock_bh(&table->tb6_lock);
1645 int ip6_del_rt(struct rt6_info *rt)
1647 struct nl_info info = {
1648 .nl_net = dev_net(rt->dst.dev),
1650 return __ip6_del_rt(rt, &info);
1653 static int ip6_route_del(struct fib6_config *cfg)
1655 struct fib6_table *table;
1656 struct fib6_node *fn;
1657 struct rt6_info *rt;
1660 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1664 read_lock_bh(&table->tb6_lock);
1666 fn = fib6_locate(&table->tb6_root,
1667 &cfg->fc_dst, cfg->fc_dst_len,
1668 &cfg->fc_src, cfg->fc_src_len);
1671 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1672 if (cfg->fc_ifindex &&
1674 rt->dst.dev->ifindex != cfg->fc_ifindex))
1676 if (cfg->fc_flags & RTF_GATEWAY &&
1677 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1679 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1682 read_unlock_bh(&table->tb6_lock);
1684 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1687 read_unlock_bh(&table->tb6_lock);
1692 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1694 struct net *net = dev_net(skb->dev);
1695 struct netevent_redirect netevent;
1696 struct rt6_info *rt, *nrt = NULL;
1697 struct ndisc_options ndopts;
1698 struct inet6_dev *in6_dev;
1699 struct neighbour *neigh;
1701 int optlen, on_link;
1704 optlen = skb->tail - skb->transport_header;
1705 optlen -= sizeof(*msg);
1708 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1712 msg = (struct rd_msg *)icmp6_hdr(skb);
1714 if (ipv6_addr_is_multicast(&msg->dest)) {
1715 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1720 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1722 } else if (ipv6_addr_type(&msg->target) !=
1723 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1724 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1728 in6_dev = __in6_dev_get(skb->dev);
1731 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1735 * The IP source address of the Redirect MUST be the same as the current
1736 * first-hop router for the specified ICMP Destination Address.
1739 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1740 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1745 if (ndopts.nd_opts_tgt_lladdr) {
1746 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1749 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1754 rt = (struct rt6_info *) dst;
1755 if (rt == net->ipv6.ip6_null_entry) {
1756 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1760 /* Redirect received -> path was valid.
1761 * Look, redirects are sent only in response to data packets,
1762 * so that this nexthop apparently is reachable. --ANK
1764 dst_confirm(&rt->dst);
1766 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1771 * We have finally decided to accept it.
1774 neigh_update(neigh, lladdr, NUD_STALE,
1775 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1776 NEIGH_UPDATE_F_OVERRIDE|
1777 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1778 NEIGH_UPDATE_F_ISROUTER))
1781 nrt = ip6_rt_copy(rt, &msg->dest);
1785 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1787 nrt->rt6i_flags &= ~RTF_GATEWAY;
1789 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1791 if (ip6_ins_rt(nrt))
1794 netevent.old = &rt->dst;
1795 netevent.new = &nrt->dst;
1796 netevent.daddr = &msg->dest;
1797 netevent.neigh = neigh;
1798 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1800 if (rt->rt6i_flags & RTF_CACHE) {
1801 rt = (struct rt6_info *) dst_clone(&rt->dst);
1806 neigh_release(neigh);
1810 * Misc support functions
1813 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1814 const struct in6_addr *dest)
1816 struct net *net = dev_net(ort->dst.dev);
1817 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1821 rt->dst.input = ort->dst.input;
1822 rt->dst.output = ort->dst.output;
1823 rt->dst.flags |= DST_HOST;
1825 rt->rt6i_dst.addr = *dest;
1826 rt->rt6i_dst.plen = 128;
1827 dst_copy_metrics(&rt->dst, &ort->dst);
1828 rt->dst.error = ort->dst.error;
1829 rt->rt6i_idev = ort->rt6i_idev;
1831 in6_dev_hold(rt->rt6i_idev);
1832 rt->dst.lastuse = jiffies;
1834 if (ort->rt6i_flags & RTF_GATEWAY)
1835 rt->rt6i_gateway = ort->rt6i_gateway;
1837 rt->rt6i_gateway = *dest;
1838 rt->rt6i_flags = ort->rt6i_flags;
1839 rt6_set_from(rt, ort);
1840 rt->rt6i_metric = 0;
1842 #ifdef CONFIG_IPV6_SUBTREES
1843 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1845 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1846 rt->rt6i_table = ort->rt6i_table;
1851 #ifdef CONFIG_IPV6_ROUTE_INFO
1852 static struct rt6_info *rt6_get_route_info(struct net *net,
1853 const struct in6_addr *prefix, int prefixlen,
1854 const struct in6_addr *gwaddr, int ifindex)
1856 struct fib6_node *fn;
1857 struct rt6_info *rt = NULL;
1858 struct fib6_table *table;
1860 table = fib6_get_table(net, 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 != 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 *net,
1885 const struct in6_addr *prefix, int prefixlen,
1886 const struct in6_addr *gwaddr, int ifindex,
1889 struct fib6_config cfg = {
1890 .fc_table = RT6_TABLE_INFO,
1891 .fc_metric = IP6_RT_PRIO_USER,
1892 .fc_ifindex = ifindex,
1893 .fc_dst_len = prefixlen,
1894 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1895 RTF_UP | RTF_PREF(pref),
1896 .fc_nlinfo.portid = 0,
1897 .fc_nlinfo.nlh = NULL,
1898 .fc_nlinfo.nl_net = net,
1901 cfg.fc_dst = *prefix;
1902 cfg.fc_gateway = *gwaddr;
1904 /* We should treat it as a default route if prefix length is 0. */
1906 cfg.fc_flags |= RTF_DEFAULT;
1908 ip6_route_add(&cfg);
1910 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1914 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1916 struct rt6_info *rt;
1917 struct fib6_table *table;
1919 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
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 = 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);
1958 void rt6_purge_dflt_routers(struct net *net)
1960 struct rt6_info *rt;
1961 struct fib6_table *table;
1963 /* NOTE: Keep consistent with rt6_get_dflt_router */
1964 table = fib6_get_table(net, RT6_TABLE_DFLT);
1969 read_lock_bh(&table->tb6_lock);
1970 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1971 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
1972 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
1974 read_unlock_bh(&table->tb6_lock);
1979 read_unlock_bh(&table->tb6_lock);
1982 static void rtmsg_to_fib6_config(struct net *net,
1983 struct in6_rtmsg *rtmsg,
1984 struct fib6_config *cfg)
1986 memset(cfg, 0, sizeof(*cfg));
1988 cfg->fc_table = RT6_TABLE_MAIN;
1989 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1990 cfg->fc_metric = rtmsg->rtmsg_metric;
1991 cfg->fc_expires = rtmsg->rtmsg_info;
1992 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1993 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1994 cfg->fc_flags = rtmsg->rtmsg_flags;
1996 cfg->fc_nlinfo.nl_net = net;
1998 cfg->fc_dst = rtmsg->rtmsg_dst;
1999 cfg->fc_src = rtmsg->rtmsg_src;
2000 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2003 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2005 struct fib6_config cfg;
2006 struct in6_rtmsg rtmsg;
2010 case SIOCADDRT: /* Add a route */
2011 case SIOCDELRT: /* Delete a route */
2012 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2014 err = copy_from_user(&rtmsg, arg,
2015 sizeof(struct in6_rtmsg));
2019 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2024 err = ip6_route_add(&cfg);
2027 err = ip6_route_del(&cfg);
2041 * Drop the packet on the floor
2044 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2047 struct dst_entry *dst = skb_dst(skb);
2048 switch (ipstats_mib_noroutes) {
2049 case IPSTATS_MIB_INNOROUTES:
2050 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2051 if (type == IPV6_ADDR_ANY) {
2052 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2053 IPSTATS_MIB_INADDRERRORS);
2057 case IPSTATS_MIB_OUTNOROUTES:
2058 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2059 ipstats_mib_noroutes);
2062 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2067 static int ip6_pkt_discard(struct sk_buff *skb)
2069 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2072 static int ip6_pkt_discard_out(struct sk_buff *skb)
2074 skb->dev = skb_dst(skb)->dev;
2075 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2078 static int ip6_pkt_prohibit(struct sk_buff *skb)
2080 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2083 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2085 skb->dev = skb_dst(skb)->dev;
2086 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2090 * Allocate a dst for local (unicast / anycast) address.
2093 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2094 const struct in6_addr *addr,
2097 struct net *net = dev_net(idev->dev);
2098 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2101 return ERR_PTR(-ENOMEM);
2105 rt->dst.flags |= DST_HOST;
2106 rt->dst.input = ip6_input;
2107 rt->dst.output = ip6_output;
2108 rt->rt6i_idev = idev;
2110 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2112 rt->rt6i_flags |= RTF_ANYCAST;
2114 rt->rt6i_flags |= RTF_LOCAL;
2116 rt->rt6i_gateway = *addr;
2117 rt->rt6i_dst.addr = *addr;
2118 rt->rt6i_dst.plen = 128;
2119 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2121 atomic_set(&rt->dst.__refcnt, 1);
2126 int ip6_route_get_saddr(struct net *net,
2127 struct rt6_info *rt,
2128 const struct in6_addr *daddr,
2130 struct in6_addr *saddr)
2132 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2134 if (rt->rt6i_prefsrc.plen)
2135 *saddr = rt->rt6i_prefsrc.addr;
2137 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2138 daddr, prefs, saddr);
2142 /* remove deleted ip from prefsrc entries */
2143 struct arg_dev_net_ip {
2144 struct net_device *dev;
2146 struct in6_addr *addr;
2149 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2151 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2152 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2153 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2155 if (((void *)rt->dst.dev == dev || !dev) &&
2156 rt != net->ipv6.ip6_null_entry &&
2157 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2158 /* remove prefsrc entry */
2159 rt->rt6i_prefsrc.plen = 0;
2164 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2166 struct net *net = dev_net(ifp->idev->dev);
2167 struct arg_dev_net_ip adni = {
2168 .dev = ifp->idev->dev,
2172 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2175 struct arg_dev_net {
2176 struct net_device *dev;
2180 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2182 const struct arg_dev_net *adn = arg;
2183 const struct net_device *dev = adn->dev;
2185 if ((rt->dst.dev == dev || !dev) &&
2186 rt != adn->net->ipv6.ip6_null_entry)
2192 void rt6_ifdown(struct net *net, struct net_device *dev)
2194 struct arg_dev_net adn = {
2199 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2200 icmp6_clean_all(fib6_ifdown, &adn);
2203 struct rt6_mtu_change_arg {
2204 struct net_device *dev;
2208 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2210 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2211 struct inet6_dev *idev;
2213 /* In IPv6 pmtu discovery is not optional,
2214 so that RTAX_MTU lock cannot disable it.
2215 We still use this lock to block changes
2216 caused by addrconf/ndisc.
2219 idev = __in6_dev_get(arg->dev);
2223 /* For administrative MTU increase, there is no way to discover
2224 IPv6 PMTU increase, so PMTU increase should be updated here.
2225 Since RFC 1981 doesn't include administrative MTU increase
2226 update PMTU increase is a MUST. (i.e. jumbo frame)
2229 If new MTU is less than route PMTU, this new MTU will be the
2230 lowest MTU in the path, update the route PMTU to reflect PMTU
2231 decreases; if new MTU is greater than route PMTU, and the
2232 old MTU is the lowest MTU in the path, update the route PMTU
2233 to reflect the increase. In this case if the other nodes' MTU
2234 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2237 if (rt->dst.dev == arg->dev &&
2238 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2239 (dst_mtu(&rt->dst) >= arg->mtu ||
2240 (dst_mtu(&rt->dst) < arg->mtu &&
2241 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2242 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2247 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2249 struct rt6_mtu_change_arg arg = {
2254 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2257 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2258 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2259 [RTA_OIF] = { .type = NLA_U32 },
2260 [RTA_IIF] = { .type = NLA_U32 },
2261 [RTA_PRIORITY] = { .type = NLA_U32 },
2262 [RTA_METRICS] = { .type = NLA_NESTED },
2263 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2266 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2267 struct fib6_config *cfg)
2270 struct nlattr *tb[RTA_MAX+1];
2273 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2278 rtm = nlmsg_data(nlh);
2279 memset(cfg, 0, sizeof(*cfg));
2281 cfg->fc_table = rtm->rtm_table;
2282 cfg->fc_dst_len = rtm->rtm_dst_len;
2283 cfg->fc_src_len = rtm->rtm_src_len;
2284 cfg->fc_flags = RTF_UP;
2285 cfg->fc_protocol = rtm->rtm_protocol;
2286 cfg->fc_type = rtm->rtm_type;
2288 if (rtm->rtm_type == RTN_UNREACHABLE ||
2289 rtm->rtm_type == RTN_BLACKHOLE ||
2290 rtm->rtm_type == RTN_PROHIBIT ||
2291 rtm->rtm_type == RTN_THROW)
2292 cfg->fc_flags |= RTF_REJECT;
2294 if (rtm->rtm_type == RTN_LOCAL)
2295 cfg->fc_flags |= RTF_LOCAL;
2297 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2298 cfg->fc_nlinfo.nlh = nlh;
2299 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2301 if (tb[RTA_GATEWAY]) {
2302 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2303 cfg->fc_flags |= RTF_GATEWAY;
2307 int plen = (rtm->rtm_dst_len + 7) >> 3;
2309 if (nla_len(tb[RTA_DST]) < plen)
2312 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2316 int plen = (rtm->rtm_src_len + 7) >> 3;
2318 if (nla_len(tb[RTA_SRC]) < plen)
2321 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2324 if (tb[RTA_PREFSRC])
2325 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2328 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2330 if (tb[RTA_PRIORITY])
2331 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2333 if (tb[RTA_METRICS]) {
2334 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2335 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2339 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2341 if (tb[RTA_MULTIPATH]) {
2342 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2343 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2351 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2353 struct fib6_config r_cfg;
2354 struct rtnexthop *rtnh;
2357 int err = 0, last_err = 0;
2360 rtnh = (struct rtnexthop *)cfg->fc_mp;
2361 remaining = cfg->fc_mp_len;
2363 /* Parse a Multipath Entry */
2364 while (rtnh_ok(rtnh, remaining)) {
2365 memcpy(&r_cfg, cfg, sizeof(*cfg));
2366 if (rtnh->rtnh_ifindex)
2367 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2369 attrlen = rtnh_attrlen(rtnh);
2371 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2373 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2375 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2376 r_cfg.fc_flags |= RTF_GATEWAY;
2379 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2382 /* If we are trying to remove a route, do not stop the
2383 * loop when ip6_route_del() fails (because next hop is
2384 * already gone), we should try to remove all next hops.
2387 /* If add fails, we should try to delete all
2388 * next hops that have been already added.
2394 /* Because each route is added like a single route we remove
2395 * this flag after the first nexthop (if there is a collision,
2396 * we have already fail to add the first nexthop:
2397 * fib6_add_rt2node() has reject it).
2399 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2400 rtnh = rtnh_next(rtnh, &remaining);
2406 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2408 struct fib6_config cfg;
2411 err = rtm_to_fib6_config(skb, nlh, &cfg);
2416 return ip6_route_multipath(&cfg, 0);
2418 return ip6_route_del(&cfg);
2421 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2423 struct fib6_config cfg;
2426 err = rtm_to_fib6_config(skb, nlh, &cfg);
2431 return ip6_route_multipath(&cfg, 1);
2433 return ip6_route_add(&cfg);
2436 static inline size_t rt6_nlmsg_size(void)
2438 return NLMSG_ALIGN(sizeof(struct rtmsg))
2439 + nla_total_size(16) /* RTA_SRC */
2440 + nla_total_size(16) /* RTA_DST */
2441 + nla_total_size(16) /* RTA_GATEWAY */
2442 + nla_total_size(16) /* RTA_PREFSRC */
2443 + nla_total_size(4) /* RTA_TABLE */
2444 + nla_total_size(4) /* RTA_IIF */
2445 + nla_total_size(4) /* RTA_OIF */
2446 + nla_total_size(4) /* RTA_PRIORITY */
2447 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2448 + nla_total_size(sizeof(struct rta_cacheinfo));
2451 static int rt6_fill_node(struct net *net,
2452 struct sk_buff *skb, struct rt6_info *rt,
2453 struct in6_addr *dst, struct in6_addr *src,
2454 int iif, int type, u32 portid, u32 seq,
2455 int prefix, int nowait, unsigned int flags)
2458 struct nlmsghdr *nlh;
2462 if (prefix) { /* user wants prefix routes only */
2463 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2464 /* success since this is not a prefix route */
2469 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2473 rtm = nlmsg_data(nlh);
2474 rtm->rtm_family = AF_INET6;
2475 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2476 rtm->rtm_src_len = rt->rt6i_src.plen;
2479 table = rt->rt6i_table->tb6_id;
2481 table = RT6_TABLE_UNSPEC;
2482 rtm->rtm_table = table;
2483 if (nla_put_u32(skb, RTA_TABLE, table))
2484 goto nla_put_failure;
2485 if (rt->rt6i_flags & RTF_REJECT) {
2486 switch (rt->dst.error) {
2488 rtm->rtm_type = RTN_BLACKHOLE;
2491 rtm->rtm_type = RTN_PROHIBIT;
2494 rtm->rtm_type = RTN_THROW;
2497 rtm->rtm_type = RTN_UNREACHABLE;
2501 else if (rt->rt6i_flags & RTF_LOCAL)
2502 rtm->rtm_type = RTN_LOCAL;
2503 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2504 rtm->rtm_type = RTN_LOCAL;
2506 rtm->rtm_type = RTN_UNICAST;
2508 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2509 rtm->rtm_protocol = rt->rt6i_protocol;
2510 if (rt->rt6i_flags & RTF_DYNAMIC)
2511 rtm->rtm_protocol = RTPROT_REDIRECT;
2512 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2513 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2514 rtm->rtm_protocol = RTPROT_RA;
2516 rtm->rtm_protocol = RTPROT_KERNEL;
2519 if (rt->rt6i_flags & RTF_CACHE)
2520 rtm->rtm_flags |= RTM_F_CLONED;
2523 if (nla_put(skb, RTA_DST, 16, dst))
2524 goto nla_put_failure;
2525 rtm->rtm_dst_len = 128;
2526 } else if (rtm->rtm_dst_len)
2527 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2528 goto nla_put_failure;
2529 #ifdef CONFIG_IPV6_SUBTREES
2531 if (nla_put(skb, RTA_SRC, 16, src))
2532 goto nla_put_failure;
2533 rtm->rtm_src_len = 128;
2534 } else if (rtm->rtm_src_len &&
2535 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2536 goto nla_put_failure;
2539 #ifdef CONFIG_IPV6_MROUTE
2540 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2541 int err = ip6mr_get_route(net, skb, rtm, nowait);
2546 goto nla_put_failure;
2548 if (err == -EMSGSIZE)
2549 goto nla_put_failure;
2554 if (nla_put_u32(skb, RTA_IIF, iif))
2555 goto nla_put_failure;
2557 struct in6_addr saddr_buf;
2558 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2559 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2560 goto nla_put_failure;
2563 if (rt->rt6i_prefsrc.plen) {
2564 struct in6_addr saddr_buf;
2565 saddr_buf = rt->rt6i_prefsrc.addr;
2566 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2567 goto nla_put_failure;
2570 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2571 goto nla_put_failure;
2573 if (rt->rt6i_flags & RTF_GATEWAY) {
2574 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2575 goto nla_put_failure;
2579 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2580 goto nla_put_failure;
2581 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2582 goto nla_put_failure;
2584 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2586 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2587 goto nla_put_failure;
2589 return nlmsg_end(skb, nlh);
2592 nlmsg_cancel(skb, nlh);
2596 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2598 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2601 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2602 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2603 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2607 return rt6_fill_node(arg->net,
2608 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2609 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2610 prefix, 0, NLM_F_MULTI);
2613 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2615 struct net *net = sock_net(in_skb->sk);
2616 struct nlattr *tb[RTA_MAX+1];
2617 struct rt6_info *rt;
2618 struct sk_buff *skb;
2621 int err, iif = 0, oif = 0;
2623 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2628 memset(&fl6, 0, sizeof(fl6));
2631 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2634 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2638 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2641 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2645 iif = nla_get_u32(tb[RTA_IIF]);
2648 oif = nla_get_u32(tb[RTA_OIF]);
2651 struct net_device *dev;
2654 dev = __dev_get_by_index(net, iif);
2660 fl6.flowi6_iif = iif;
2662 if (!ipv6_addr_any(&fl6.saddr))
2663 flags |= RT6_LOOKUP_F_HAS_SADDR;
2665 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2668 fl6.flowi6_oif = oif;
2670 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2673 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2680 /* Reserve room for dummy headers, this skb can pass
2681 through good chunk of routing engine.
2683 skb_reset_mac_header(skb);
2684 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2686 skb_dst_set(skb, &rt->dst);
2688 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2689 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2690 nlh->nlmsg_seq, 0, 0, 0);
2696 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2701 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2703 struct sk_buff *skb;
2704 struct net *net = info->nl_net;
2709 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2711 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2715 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2716 event, info->portid, seq, 0, 0, 0);
2718 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2719 WARN_ON(err == -EMSGSIZE);
2723 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2724 info->nlh, gfp_any());
2728 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2731 static int ip6_route_dev_notify(struct notifier_block *this,
2732 unsigned long event, void *data)
2734 struct net_device *dev = (struct net_device *)data;
2735 struct net *net = dev_net(dev);
2737 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2738 net->ipv6.ip6_null_entry->dst.dev = dev;
2739 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2740 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2741 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2742 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2743 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2744 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2755 #ifdef CONFIG_PROC_FS
2766 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2768 struct seq_file *m = p_arg;
2770 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2772 #ifdef CONFIG_IPV6_SUBTREES
2773 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2775 seq_puts(m, "00000000000000000000000000000000 00 ");
2777 if (rt->rt6i_flags & RTF_GATEWAY) {
2778 seq_printf(m, "%pi6", &rt->rt6i_gateway);
2780 seq_puts(m, "00000000000000000000000000000000");
2782 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2783 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2784 rt->dst.__use, rt->rt6i_flags,
2785 rt->dst.dev ? rt->dst.dev->name : "");
2789 static int ipv6_route_show(struct seq_file *m, void *v)
2791 struct net *net = (struct net *)m->private;
2792 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2796 static int ipv6_route_open(struct inode *inode, struct file *file)
2798 return single_open_net(inode, file, ipv6_route_show);
2801 static const struct file_operations ipv6_route_proc_fops = {
2802 .owner = THIS_MODULE,
2803 .open = ipv6_route_open,
2805 .llseek = seq_lseek,
2806 .release = single_release_net,
2809 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2811 struct net *net = (struct net *)seq->private;
2812 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2813 net->ipv6.rt6_stats->fib_nodes,
2814 net->ipv6.rt6_stats->fib_route_nodes,
2815 net->ipv6.rt6_stats->fib_rt_alloc,
2816 net->ipv6.rt6_stats->fib_rt_entries,
2817 net->ipv6.rt6_stats->fib_rt_cache,
2818 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2819 net->ipv6.rt6_stats->fib_discarded_routes);
2824 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2826 return single_open_net(inode, file, rt6_stats_seq_show);
2829 static const struct file_operations rt6_stats_seq_fops = {
2830 .owner = THIS_MODULE,
2831 .open = rt6_stats_seq_open,
2833 .llseek = seq_lseek,
2834 .release = single_release_net,
2836 #endif /* CONFIG_PROC_FS */
2838 #ifdef CONFIG_SYSCTL
2841 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2842 void __user *buffer, size_t *lenp, loff_t *ppos)
2849 net = (struct net *)ctl->extra1;
2850 delay = net->ipv6.sysctl.flush_delay;
2851 proc_dointvec(ctl, write, buffer, lenp, ppos);
2852 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2856 ctl_table ipv6_route_table_template[] = {
2858 .procname = "flush",
2859 .data = &init_net.ipv6.sysctl.flush_delay,
2860 .maxlen = sizeof(int),
2862 .proc_handler = ipv6_sysctl_rtcache_flush
2865 .procname = "gc_thresh",
2866 .data = &ip6_dst_ops_template.gc_thresh,
2867 .maxlen = sizeof(int),
2869 .proc_handler = proc_dointvec,
2872 .procname = "max_size",
2873 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2874 .maxlen = sizeof(int),
2876 .proc_handler = proc_dointvec,
2879 .procname = "gc_min_interval",
2880 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2881 .maxlen = sizeof(int),
2883 .proc_handler = proc_dointvec_jiffies,
2886 .procname = "gc_timeout",
2887 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2888 .maxlen = sizeof(int),
2890 .proc_handler = proc_dointvec_jiffies,
2893 .procname = "gc_interval",
2894 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2895 .maxlen = sizeof(int),
2897 .proc_handler = proc_dointvec_jiffies,
2900 .procname = "gc_elasticity",
2901 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2902 .maxlen = sizeof(int),
2904 .proc_handler = proc_dointvec,
2907 .procname = "mtu_expires",
2908 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2909 .maxlen = sizeof(int),
2911 .proc_handler = proc_dointvec_jiffies,
2914 .procname = "min_adv_mss",
2915 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2916 .maxlen = sizeof(int),
2918 .proc_handler = proc_dointvec,
2921 .procname = "gc_min_interval_ms",
2922 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2923 .maxlen = sizeof(int),
2925 .proc_handler = proc_dointvec_ms_jiffies,
2930 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2932 struct ctl_table *table;
2934 table = kmemdup(ipv6_route_table_template,
2935 sizeof(ipv6_route_table_template),
2939 table[0].data = &net->ipv6.sysctl.flush_delay;
2940 table[0].extra1 = net;
2941 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2942 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2943 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2944 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2945 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2946 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2947 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2948 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2949 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2951 /* Don't export sysctls to unprivileged users */
2952 if (net->user_ns != &init_user_ns)
2953 table[0].procname = NULL;
2960 static int __net_init ip6_route_net_init(struct net *net)
2964 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2965 sizeof(net->ipv6.ip6_dst_ops));
2967 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2968 goto out_ip6_dst_ops;
2970 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2971 sizeof(*net->ipv6.ip6_null_entry),
2973 if (!net->ipv6.ip6_null_entry)
2974 goto out_ip6_dst_entries;
2975 net->ipv6.ip6_null_entry->dst.path =
2976 (struct dst_entry *)net->ipv6.ip6_null_entry;
2977 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2978 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2979 ip6_template_metrics, true);
2981 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2982 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2983 sizeof(*net->ipv6.ip6_prohibit_entry),
2985 if (!net->ipv6.ip6_prohibit_entry)
2986 goto out_ip6_null_entry;
2987 net->ipv6.ip6_prohibit_entry->dst.path =
2988 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2989 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2990 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2991 ip6_template_metrics, true);
2993 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2994 sizeof(*net->ipv6.ip6_blk_hole_entry),
2996 if (!net->ipv6.ip6_blk_hole_entry)
2997 goto out_ip6_prohibit_entry;
2998 net->ipv6.ip6_blk_hole_entry->dst.path =
2999 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3000 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3001 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3002 ip6_template_metrics, true);
3005 net->ipv6.sysctl.flush_delay = 0;
3006 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3007 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3008 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3009 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3010 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3011 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3012 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3014 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3020 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3021 out_ip6_prohibit_entry:
3022 kfree(net->ipv6.ip6_prohibit_entry);
3024 kfree(net->ipv6.ip6_null_entry);
3026 out_ip6_dst_entries:
3027 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3032 static void __net_exit ip6_route_net_exit(struct net *net)
3034 kfree(net->ipv6.ip6_null_entry);
3035 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3036 kfree(net->ipv6.ip6_prohibit_entry);
3037 kfree(net->ipv6.ip6_blk_hole_entry);
3039 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3042 static int __net_init ip6_route_net_init_late(struct net *net)
3044 #ifdef CONFIG_PROC_FS
3045 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3046 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3051 static void __net_exit ip6_route_net_exit_late(struct net *net)
3053 #ifdef CONFIG_PROC_FS
3054 remove_proc_entry("ipv6_route", net->proc_net);
3055 remove_proc_entry("rt6_stats", net->proc_net);
3059 static struct pernet_operations ip6_route_net_ops = {
3060 .init = ip6_route_net_init,
3061 .exit = ip6_route_net_exit,
3064 static int __net_init ipv6_inetpeer_init(struct net *net)
3066 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3070 inet_peer_base_init(bp);
3071 net->ipv6.peers = bp;
3075 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3077 struct inet_peer_base *bp = net->ipv6.peers;
3079 net->ipv6.peers = NULL;
3080 inetpeer_invalidate_tree(bp);
3084 static struct pernet_operations ipv6_inetpeer_ops = {
3085 .init = ipv6_inetpeer_init,
3086 .exit = ipv6_inetpeer_exit,
3089 static struct pernet_operations ip6_route_net_late_ops = {
3090 .init = ip6_route_net_init_late,
3091 .exit = ip6_route_net_exit_late,
3094 static struct notifier_block ip6_route_dev_notifier = {
3095 .notifier_call = ip6_route_dev_notify,
3099 int __init ip6_route_init(void)
3104 ip6_dst_ops_template.kmem_cachep =
3105 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3106 SLAB_HWCACHE_ALIGN, NULL);
3107 if (!ip6_dst_ops_template.kmem_cachep)
3110 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3112 goto out_kmem_cache;
3114 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3116 goto out_dst_entries;
3118 ret = register_pernet_subsys(&ip6_route_net_ops);
3120 goto out_register_inetpeer;
3122 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3124 /* Registering of the loopback is done before this portion of code,
3125 * the loopback reference in rt6_info will not be taken, do it
3126 * manually for init_net */
3127 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3128 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3129 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3130 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3131 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3132 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3133 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3137 goto out_register_subsys;
3143 ret = fib6_rules_init();
3147 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3149 goto fib6_rules_init;
3152 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3153 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3154 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3155 goto out_register_late_subsys;
3157 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3159 goto out_register_late_subsys;
3164 out_register_late_subsys:
3165 unregister_pernet_subsys(&ip6_route_net_late_ops);
3167 fib6_rules_cleanup();
3172 out_register_subsys:
3173 unregister_pernet_subsys(&ip6_route_net_ops);
3174 out_register_inetpeer:
3175 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3177 dst_entries_destroy(&ip6_dst_blackhole_ops);
3179 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3183 void ip6_route_cleanup(void)
3185 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3186 unregister_pernet_subsys(&ip6_route_net_late_ops);
3187 fib6_rules_cleanup();
3190 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3191 unregister_pernet_subsys(&ip6_route_net_ops);
3192 dst_entries_destroy(&ip6_dst_blackhole_ops);
3193 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);