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))
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) {
1145 u32 features = dst_metric(dst, RTAX_FEATURES);
1147 features |= RTAX_FEATURE_ALLFRAG;
1148 dst_metric_set(dst, RTAX_FEATURES, features);
1150 dst_metric_set(dst, RTAX_MTU, mtu);
1151 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1155 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1158 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1159 struct dst_entry *dst;
1162 memset(&fl6, 0, sizeof(fl6));
1163 fl6.flowi6_oif = oif;
1164 fl6.flowi6_mark = mark;
1165 fl6.flowi6_flags = 0;
1166 fl6.daddr = iph->daddr;
1167 fl6.saddr = iph->saddr;
1168 fl6.flowlabel = ip6_flowinfo(iph);
1170 dst = ip6_route_output(net, NULL, &fl6);
1172 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1175 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1177 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1179 ip6_update_pmtu(skb, sock_net(sk), mtu,
1180 sk->sk_bound_dev_if, sk->sk_mark);
1182 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1184 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1186 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1187 struct dst_entry *dst;
1190 memset(&fl6, 0, sizeof(fl6));
1191 fl6.flowi6_oif = oif;
1192 fl6.flowi6_mark = mark;
1193 fl6.flowi6_flags = 0;
1194 fl6.daddr = iph->daddr;
1195 fl6.saddr = iph->saddr;
1196 fl6.flowlabel = ip6_flowinfo(iph);
1198 dst = ip6_route_output(net, NULL, &fl6);
1200 rt6_do_redirect(dst, NULL, skb);
1203 EXPORT_SYMBOL_GPL(ip6_redirect);
1205 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1207 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1209 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1211 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1213 struct net_device *dev = dst->dev;
1214 unsigned int mtu = dst_mtu(dst);
1215 struct net *net = dev_net(dev);
1217 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1219 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1220 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1223 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1224 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1225 * IPV6_MAXPLEN is also valid and means: "any MSS,
1226 * rely only on pmtu discovery"
1228 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1233 static unsigned int ip6_mtu(const struct dst_entry *dst)
1235 struct inet6_dev *idev;
1236 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1244 idev = __in6_dev_get(dst->dev);
1246 mtu = idev->cnf.mtu6;
1250 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1253 static struct dst_entry *icmp6_dst_gc_list;
1254 static DEFINE_SPINLOCK(icmp6_dst_lock);
1256 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1259 struct dst_entry *dst;
1260 struct rt6_info *rt;
1261 struct inet6_dev *idev = in6_dev_get(dev);
1262 struct net *net = dev_net(dev);
1264 if (unlikely(!idev))
1265 return ERR_PTR(-ENODEV);
1267 rt = ip6_dst_alloc(net, dev, 0, NULL);
1268 if (unlikely(!rt)) {
1270 dst = ERR_PTR(-ENOMEM);
1274 rt->dst.flags |= DST_HOST;
1275 rt->dst.output = ip6_output;
1276 atomic_set(&rt->dst.__refcnt, 1);
1277 rt->rt6i_gateway = fl6->daddr;
1278 rt->rt6i_dst.addr = fl6->daddr;
1279 rt->rt6i_dst.plen = 128;
1280 rt->rt6i_idev = idev;
1281 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1283 spin_lock_bh(&icmp6_dst_lock);
1284 rt->dst.next = icmp6_dst_gc_list;
1285 icmp6_dst_gc_list = &rt->dst;
1286 spin_unlock_bh(&icmp6_dst_lock);
1288 fib6_force_start_gc(net);
1290 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1296 int icmp6_dst_gc(void)
1298 struct dst_entry *dst, **pprev;
1301 spin_lock_bh(&icmp6_dst_lock);
1302 pprev = &icmp6_dst_gc_list;
1304 while ((dst = *pprev) != NULL) {
1305 if (!atomic_read(&dst->__refcnt)) {
1314 spin_unlock_bh(&icmp6_dst_lock);
1319 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1322 struct dst_entry *dst, **pprev;
1324 spin_lock_bh(&icmp6_dst_lock);
1325 pprev = &icmp6_dst_gc_list;
1326 while ((dst = *pprev) != NULL) {
1327 struct rt6_info *rt = (struct rt6_info *) dst;
1328 if (func(rt, arg)) {
1335 spin_unlock_bh(&icmp6_dst_lock);
1338 static int ip6_dst_gc(struct dst_ops *ops)
1340 unsigned long now = jiffies;
1341 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1342 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1343 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1344 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1345 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1346 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1349 entries = dst_entries_get_fast(ops);
1350 if (time_after(rt_last_gc + rt_min_interval, now) &&
1351 entries <= rt_max_size)
1354 net->ipv6.ip6_rt_gc_expire++;
1355 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1356 net->ipv6.ip6_rt_last_gc = now;
1357 entries = dst_entries_get_slow(ops);
1358 if (entries < ops->gc_thresh)
1359 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1361 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1362 return entries > rt_max_size;
1365 int ip6_dst_hoplimit(struct dst_entry *dst)
1367 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1368 if (hoplimit == 0) {
1369 struct net_device *dev = dst->dev;
1370 struct inet6_dev *idev;
1373 idev = __in6_dev_get(dev);
1375 hoplimit = idev->cnf.hop_limit;
1377 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1382 EXPORT_SYMBOL(ip6_dst_hoplimit);
1388 int ip6_route_add(struct fib6_config *cfg)
1391 struct net *net = cfg->fc_nlinfo.nl_net;
1392 struct rt6_info *rt = NULL;
1393 struct net_device *dev = NULL;
1394 struct inet6_dev *idev = NULL;
1395 struct fib6_table *table;
1398 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1400 #ifndef CONFIG_IPV6_SUBTREES
1401 if (cfg->fc_src_len)
1404 if (cfg->fc_ifindex) {
1406 dev = dev_get_by_index(net, cfg->fc_ifindex);
1409 idev = in6_dev_get(dev);
1414 if (cfg->fc_metric == 0)
1415 cfg->fc_metric = IP6_RT_PRIO_USER;
1418 if (cfg->fc_nlinfo.nlh &&
1419 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1420 table = fib6_get_table(net, cfg->fc_table);
1422 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1423 table = fib6_new_table(net, cfg->fc_table);
1426 table = fib6_new_table(net, cfg->fc_table);
1432 rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1439 if (cfg->fc_flags & RTF_EXPIRES)
1440 rt6_set_expires(rt, jiffies +
1441 clock_t_to_jiffies(cfg->fc_expires));
1443 rt6_clean_expires(rt);
1445 if (cfg->fc_protocol == RTPROT_UNSPEC)
1446 cfg->fc_protocol = RTPROT_BOOT;
1447 rt->rt6i_protocol = cfg->fc_protocol;
1449 addr_type = ipv6_addr_type(&cfg->fc_dst);
1451 if (addr_type & IPV6_ADDR_MULTICAST)
1452 rt->dst.input = ip6_mc_input;
1453 else if (cfg->fc_flags & RTF_LOCAL)
1454 rt->dst.input = ip6_input;
1456 rt->dst.input = ip6_forward;
1458 rt->dst.output = ip6_output;
1460 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1461 rt->rt6i_dst.plen = cfg->fc_dst_len;
1462 if (rt->rt6i_dst.plen == 128)
1463 rt->dst.flags |= DST_HOST;
1465 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1466 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1471 dst_init_metrics(&rt->dst, metrics, 0);
1473 #ifdef CONFIG_IPV6_SUBTREES
1474 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1475 rt->rt6i_src.plen = cfg->fc_src_len;
1478 rt->rt6i_metric = cfg->fc_metric;
1480 /* We cannot add true routes via loopback here,
1481 they would result in kernel looping; promote them to reject routes
1483 if ((cfg->fc_flags & RTF_REJECT) ||
1484 (dev && (dev->flags & IFF_LOOPBACK) &&
1485 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1486 !(cfg->fc_flags & RTF_LOCAL))) {
1487 /* hold loopback dev/idev if we haven't done so. */
1488 if (dev != net->loopback_dev) {
1493 dev = net->loopback_dev;
1495 idev = in6_dev_get(dev);
1501 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1502 switch (cfg->fc_type) {
1504 rt->dst.error = -EINVAL;
1505 rt->dst.output = dst_discard;
1506 rt->dst.input = dst_discard;
1509 rt->dst.error = -EACCES;
1510 rt->dst.output = ip6_pkt_prohibit_out;
1511 rt->dst.input = ip6_pkt_prohibit;
1515 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1517 rt->dst.output = ip6_pkt_discard_out;
1518 rt->dst.input = ip6_pkt_discard;
1524 if (cfg->fc_flags & RTF_GATEWAY) {
1525 const struct in6_addr *gw_addr;
1528 gw_addr = &cfg->fc_gateway;
1529 rt->rt6i_gateway = *gw_addr;
1530 gwa_type = ipv6_addr_type(gw_addr);
1532 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1533 struct rt6_info *grt;
1535 /* IPv6 strictly inhibits using not link-local
1536 addresses as nexthop address.
1537 Otherwise, router will not able to send redirects.
1538 It is very good, but in some (rare!) circumstances
1539 (SIT, PtP, NBMA NOARP links) it is handy to allow
1540 some exceptions. --ANK
1543 if (!(gwa_type & IPV6_ADDR_UNICAST))
1546 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1548 err = -EHOSTUNREACH;
1552 if (dev != grt->dst.dev) {
1558 idev = grt->rt6i_idev;
1560 in6_dev_hold(grt->rt6i_idev);
1562 if (!(grt->rt6i_flags & RTF_GATEWAY))
1570 if (!dev || (dev->flags & IFF_LOOPBACK))
1578 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1579 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1583 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1584 rt->rt6i_prefsrc.plen = 128;
1586 rt->rt6i_prefsrc.plen = 0;
1588 rt->rt6i_flags = cfg->fc_flags;
1595 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1596 int type = nla_type(nla);
1599 if (type > RTAX_MAX) {
1604 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1610 rt->rt6i_idev = idev;
1611 rt->rt6i_table = table;
1613 cfg->fc_nlinfo.nl_net = dev_net(dev);
1615 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1627 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1630 struct fib6_table *table;
1631 struct net *net = dev_net(rt->dst.dev);
1633 if (rt == net->ipv6.ip6_null_entry) {
1638 table = rt->rt6i_table;
1639 write_lock_bh(&table->tb6_lock);
1640 err = fib6_del(rt, info);
1641 write_unlock_bh(&table->tb6_lock);
1648 int ip6_del_rt(struct rt6_info *rt)
1650 struct nl_info info = {
1651 .nl_net = dev_net(rt->dst.dev),
1653 return __ip6_del_rt(rt, &info);
1656 static int ip6_route_del(struct fib6_config *cfg)
1658 struct fib6_table *table;
1659 struct fib6_node *fn;
1660 struct rt6_info *rt;
1663 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1667 read_lock_bh(&table->tb6_lock);
1669 fn = fib6_locate(&table->tb6_root,
1670 &cfg->fc_dst, cfg->fc_dst_len,
1671 &cfg->fc_src, cfg->fc_src_len);
1674 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1675 if (cfg->fc_ifindex &&
1677 rt->dst.dev->ifindex != cfg->fc_ifindex))
1679 if (cfg->fc_flags & RTF_GATEWAY &&
1680 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1682 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1685 read_unlock_bh(&table->tb6_lock);
1687 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1690 read_unlock_bh(&table->tb6_lock);
1695 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1697 struct net *net = dev_net(skb->dev);
1698 struct netevent_redirect netevent;
1699 struct rt6_info *rt, *nrt = NULL;
1700 struct ndisc_options ndopts;
1701 struct inet6_dev *in6_dev;
1702 struct neighbour *neigh;
1704 int optlen, on_link;
1707 optlen = skb->tail - skb->transport_header;
1708 optlen -= sizeof(*msg);
1711 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1715 msg = (struct rd_msg *)icmp6_hdr(skb);
1717 if (ipv6_addr_is_multicast(&msg->dest)) {
1718 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1723 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1725 } else if (ipv6_addr_type(&msg->target) !=
1726 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1727 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1731 in6_dev = __in6_dev_get(skb->dev);
1734 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1738 * The IP source address of the Redirect MUST be the same as the current
1739 * first-hop router for the specified ICMP Destination Address.
1742 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1743 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1748 if (ndopts.nd_opts_tgt_lladdr) {
1749 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1752 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1757 rt = (struct rt6_info *) dst;
1758 if (rt == net->ipv6.ip6_null_entry) {
1759 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1763 /* Redirect received -> path was valid.
1764 * Look, redirects are sent only in response to data packets,
1765 * so that this nexthop apparently is reachable. --ANK
1767 dst_confirm(&rt->dst);
1769 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1774 * We have finally decided to accept it.
1777 neigh_update(neigh, lladdr, NUD_STALE,
1778 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1779 NEIGH_UPDATE_F_OVERRIDE|
1780 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1781 NEIGH_UPDATE_F_ISROUTER))
1784 nrt = ip6_rt_copy(rt, &msg->dest);
1788 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1790 nrt->rt6i_flags &= ~RTF_GATEWAY;
1792 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1794 if (ip6_ins_rt(nrt))
1797 netevent.old = &rt->dst;
1798 netevent.new = &nrt->dst;
1799 netevent.daddr = &msg->dest;
1800 netevent.neigh = neigh;
1801 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1803 if (rt->rt6i_flags & RTF_CACHE) {
1804 rt = (struct rt6_info *) dst_clone(&rt->dst);
1809 neigh_release(neigh);
1813 * Misc support functions
1816 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1817 const struct in6_addr *dest)
1819 struct net *net = dev_net(ort->dst.dev);
1820 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1824 rt->dst.input = ort->dst.input;
1825 rt->dst.output = ort->dst.output;
1826 rt->dst.flags |= DST_HOST;
1828 rt->rt6i_dst.addr = *dest;
1829 rt->rt6i_dst.plen = 128;
1830 dst_copy_metrics(&rt->dst, &ort->dst);
1831 rt->dst.error = ort->dst.error;
1832 rt->rt6i_idev = ort->rt6i_idev;
1834 in6_dev_hold(rt->rt6i_idev);
1835 rt->dst.lastuse = jiffies;
1837 if (ort->rt6i_flags & RTF_GATEWAY)
1838 rt->rt6i_gateway = ort->rt6i_gateway;
1840 rt->rt6i_gateway = *dest;
1841 rt->rt6i_flags = ort->rt6i_flags;
1842 rt6_set_from(rt, ort);
1843 rt->rt6i_metric = 0;
1845 #ifdef CONFIG_IPV6_SUBTREES
1846 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1848 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1849 rt->rt6i_table = ort->rt6i_table;
1854 #ifdef CONFIG_IPV6_ROUTE_INFO
1855 static struct rt6_info *rt6_get_route_info(struct net *net,
1856 const struct in6_addr *prefix, int prefixlen,
1857 const struct in6_addr *gwaddr, int ifindex)
1859 struct fib6_node *fn;
1860 struct rt6_info *rt = NULL;
1861 struct fib6_table *table;
1863 table = fib6_get_table(net, RT6_TABLE_INFO);
1867 read_lock_bh(&table->tb6_lock);
1868 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1872 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1873 if (rt->dst.dev->ifindex != ifindex)
1875 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1877 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1883 read_unlock_bh(&table->tb6_lock);
1887 static struct rt6_info *rt6_add_route_info(struct net *net,
1888 const struct in6_addr *prefix, int prefixlen,
1889 const struct in6_addr *gwaddr, int ifindex,
1892 struct fib6_config cfg = {
1893 .fc_table = RT6_TABLE_INFO,
1894 .fc_metric = IP6_RT_PRIO_USER,
1895 .fc_ifindex = ifindex,
1896 .fc_dst_len = prefixlen,
1897 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1898 RTF_UP | RTF_PREF(pref),
1899 .fc_nlinfo.portid = 0,
1900 .fc_nlinfo.nlh = NULL,
1901 .fc_nlinfo.nl_net = net,
1904 cfg.fc_dst = *prefix;
1905 cfg.fc_gateway = *gwaddr;
1907 /* We should treat it as a default route if prefix length is 0. */
1909 cfg.fc_flags |= RTF_DEFAULT;
1911 ip6_route_add(&cfg);
1913 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1917 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1919 struct rt6_info *rt;
1920 struct fib6_table *table;
1922 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1926 read_lock_bh(&table->tb6_lock);
1927 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1928 if (dev == rt->dst.dev &&
1929 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1930 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1935 read_unlock_bh(&table->tb6_lock);
1939 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1940 struct net_device *dev,
1943 struct fib6_config cfg = {
1944 .fc_table = RT6_TABLE_DFLT,
1945 .fc_metric = IP6_RT_PRIO_USER,
1946 .fc_ifindex = dev->ifindex,
1947 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1948 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1949 .fc_nlinfo.portid = 0,
1950 .fc_nlinfo.nlh = NULL,
1951 .fc_nlinfo.nl_net = dev_net(dev),
1954 cfg.fc_gateway = *gwaddr;
1956 ip6_route_add(&cfg);
1958 return rt6_get_dflt_router(gwaddr, dev);
1961 void rt6_purge_dflt_routers(struct net *net)
1963 struct rt6_info *rt;
1964 struct fib6_table *table;
1966 /* NOTE: Keep consistent with rt6_get_dflt_router */
1967 table = fib6_get_table(net, RT6_TABLE_DFLT);
1972 read_lock_bh(&table->tb6_lock);
1973 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1974 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
1975 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
1977 read_unlock_bh(&table->tb6_lock);
1982 read_unlock_bh(&table->tb6_lock);
1985 static void rtmsg_to_fib6_config(struct net *net,
1986 struct in6_rtmsg *rtmsg,
1987 struct fib6_config *cfg)
1989 memset(cfg, 0, sizeof(*cfg));
1991 cfg->fc_table = RT6_TABLE_MAIN;
1992 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1993 cfg->fc_metric = rtmsg->rtmsg_metric;
1994 cfg->fc_expires = rtmsg->rtmsg_info;
1995 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1996 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1997 cfg->fc_flags = rtmsg->rtmsg_flags;
1999 cfg->fc_nlinfo.nl_net = net;
2001 cfg->fc_dst = rtmsg->rtmsg_dst;
2002 cfg->fc_src = rtmsg->rtmsg_src;
2003 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2006 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2008 struct fib6_config cfg;
2009 struct in6_rtmsg rtmsg;
2013 case SIOCADDRT: /* Add a route */
2014 case SIOCDELRT: /* Delete a route */
2015 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2017 err = copy_from_user(&rtmsg, arg,
2018 sizeof(struct in6_rtmsg));
2022 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2027 err = ip6_route_add(&cfg);
2030 err = ip6_route_del(&cfg);
2044 * Drop the packet on the floor
2047 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2050 struct dst_entry *dst = skb_dst(skb);
2051 switch (ipstats_mib_noroutes) {
2052 case IPSTATS_MIB_INNOROUTES:
2053 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2054 if (type == IPV6_ADDR_ANY) {
2055 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2056 IPSTATS_MIB_INADDRERRORS);
2060 case IPSTATS_MIB_OUTNOROUTES:
2061 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2062 ipstats_mib_noroutes);
2065 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2070 static int ip6_pkt_discard(struct sk_buff *skb)
2072 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2075 static int ip6_pkt_discard_out(struct sk_buff *skb)
2077 skb->dev = skb_dst(skb)->dev;
2078 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2081 static int ip6_pkt_prohibit(struct sk_buff *skb)
2083 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2086 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2088 skb->dev = skb_dst(skb)->dev;
2089 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2093 * Allocate a dst for local (unicast / anycast) address.
2096 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2097 const struct in6_addr *addr,
2100 struct net *net = dev_net(idev->dev);
2101 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2104 return ERR_PTR(-ENOMEM);
2108 rt->dst.flags |= DST_HOST;
2109 rt->dst.input = ip6_input;
2110 rt->dst.output = ip6_output;
2111 rt->rt6i_idev = idev;
2113 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2115 rt->rt6i_flags |= RTF_ANYCAST;
2117 rt->rt6i_flags |= RTF_LOCAL;
2119 rt->rt6i_gateway = *addr;
2120 rt->rt6i_dst.addr = *addr;
2121 rt->rt6i_dst.plen = 128;
2122 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2124 atomic_set(&rt->dst.__refcnt, 1);
2129 int ip6_route_get_saddr(struct net *net,
2130 struct rt6_info *rt,
2131 const struct in6_addr *daddr,
2133 struct in6_addr *saddr)
2135 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2137 if (rt->rt6i_prefsrc.plen)
2138 *saddr = rt->rt6i_prefsrc.addr;
2140 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2141 daddr, prefs, saddr);
2145 /* remove deleted ip from prefsrc entries */
2146 struct arg_dev_net_ip {
2147 struct net_device *dev;
2149 struct in6_addr *addr;
2152 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2154 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2155 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2156 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2158 if (((void *)rt->dst.dev == dev || !dev) &&
2159 rt != net->ipv6.ip6_null_entry &&
2160 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2161 /* remove prefsrc entry */
2162 rt->rt6i_prefsrc.plen = 0;
2167 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2169 struct net *net = dev_net(ifp->idev->dev);
2170 struct arg_dev_net_ip adni = {
2171 .dev = ifp->idev->dev,
2175 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2178 struct arg_dev_net {
2179 struct net_device *dev;
2183 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2185 const struct arg_dev_net *adn = arg;
2186 const struct net_device *dev = adn->dev;
2188 if ((rt->dst.dev == dev || !dev) &&
2189 rt != adn->net->ipv6.ip6_null_entry)
2195 void rt6_ifdown(struct net *net, struct net_device *dev)
2197 struct arg_dev_net adn = {
2202 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2203 icmp6_clean_all(fib6_ifdown, &adn);
2206 struct rt6_mtu_change_arg {
2207 struct net_device *dev;
2211 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2213 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2214 struct inet6_dev *idev;
2216 /* In IPv6 pmtu discovery is not optional,
2217 so that RTAX_MTU lock cannot disable it.
2218 We still use this lock to block changes
2219 caused by addrconf/ndisc.
2222 idev = __in6_dev_get(arg->dev);
2226 /* For administrative MTU increase, there is no way to discover
2227 IPv6 PMTU increase, so PMTU increase should be updated here.
2228 Since RFC 1981 doesn't include administrative MTU increase
2229 update PMTU increase is a MUST. (i.e. jumbo frame)
2232 If new MTU is less than route PMTU, this new MTU will be the
2233 lowest MTU in the path, update the route PMTU to reflect PMTU
2234 decreases; if new MTU is greater than route PMTU, and the
2235 old MTU is the lowest MTU in the path, update the route PMTU
2236 to reflect the increase. In this case if the other nodes' MTU
2237 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2240 if (rt->dst.dev == arg->dev &&
2241 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2242 (dst_mtu(&rt->dst) >= arg->mtu ||
2243 (dst_mtu(&rt->dst) < arg->mtu &&
2244 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2245 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2250 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2252 struct rt6_mtu_change_arg arg = {
2257 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2260 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2261 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2262 [RTA_OIF] = { .type = NLA_U32 },
2263 [RTA_IIF] = { .type = NLA_U32 },
2264 [RTA_PRIORITY] = { .type = NLA_U32 },
2265 [RTA_METRICS] = { .type = NLA_NESTED },
2266 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2269 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2270 struct fib6_config *cfg)
2273 struct nlattr *tb[RTA_MAX+1];
2276 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2281 rtm = nlmsg_data(nlh);
2282 memset(cfg, 0, sizeof(*cfg));
2284 cfg->fc_table = rtm->rtm_table;
2285 cfg->fc_dst_len = rtm->rtm_dst_len;
2286 cfg->fc_src_len = rtm->rtm_src_len;
2287 cfg->fc_flags = RTF_UP;
2288 cfg->fc_protocol = rtm->rtm_protocol;
2289 cfg->fc_type = rtm->rtm_type;
2291 if (rtm->rtm_type == RTN_UNREACHABLE ||
2292 rtm->rtm_type == RTN_BLACKHOLE ||
2293 rtm->rtm_type == RTN_PROHIBIT ||
2294 rtm->rtm_type == RTN_THROW)
2295 cfg->fc_flags |= RTF_REJECT;
2297 if (rtm->rtm_type == RTN_LOCAL)
2298 cfg->fc_flags |= RTF_LOCAL;
2300 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2301 cfg->fc_nlinfo.nlh = nlh;
2302 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2304 if (tb[RTA_GATEWAY]) {
2305 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2306 cfg->fc_flags |= RTF_GATEWAY;
2310 int plen = (rtm->rtm_dst_len + 7) >> 3;
2312 if (nla_len(tb[RTA_DST]) < plen)
2315 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2319 int plen = (rtm->rtm_src_len + 7) >> 3;
2321 if (nla_len(tb[RTA_SRC]) < plen)
2324 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2327 if (tb[RTA_PREFSRC])
2328 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2331 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2333 if (tb[RTA_PRIORITY])
2334 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2336 if (tb[RTA_METRICS]) {
2337 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2338 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2342 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2344 if (tb[RTA_MULTIPATH]) {
2345 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2346 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2354 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2356 struct fib6_config r_cfg;
2357 struct rtnexthop *rtnh;
2360 int err = 0, last_err = 0;
2363 rtnh = (struct rtnexthop *)cfg->fc_mp;
2364 remaining = cfg->fc_mp_len;
2366 /* Parse a Multipath Entry */
2367 while (rtnh_ok(rtnh, remaining)) {
2368 memcpy(&r_cfg, cfg, sizeof(*cfg));
2369 if (rtnh->rtnh_ifindex)
2370 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2372 attrlen = rtnh_attrlen(rtnh);
2374 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2376 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2378 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2379 r_cfg.fc_flags |= RTF_GATEWAY;
2382 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2385 /* If we are trying to remove a route, do not stop the
2386 * loop when ip6_route_del() fails (because next hop is
2387 * already gone), we should try to remove all next hops.
2390 /* If add fails, we should try to delete all
2391 * next hops that have been already added.
2397 /* Because each route is added like a single route we remove
2398 * this flag after the first nexthop (if there is a collision,
2399 * we have already fail to add the first nexthop:
2400 * fib6_add_rt2node() has reject it).
2402 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2403 rtnh = rtnh_next(rtnh, &remaining);
2409 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2411 struct fib6_config cfg;
2414 err = rtm_to_fib6_config(skb, nlh, &cfg);
2419 return ip6_route_multipath(&cfg, 0);
2421 return ip6_route_del(&cfg);
2424 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2426 struct fib6_config cfg;
2429 err = rtm_to_fib6_config(skb, nlh, &cfg);
2434 return ip6_route_multipath(&cfg, 1);
2436 return ip6_route_add(&cfg);
2439 static inline size_t rt6_nlmsg_size(void)
2441 return NLMSG_ALIGN(sizeof(struct rtmsg))
2442 + nla_total_size(16) /* RTA_SRC */
2443 + nla_total_size(16) /* RTA_DST */
2444 + nla_total_size(16) /* RTA_GATEWAY */
2445 + nla_total_size(16) /* RTA_PREFSRC */
2446 + nla_total_size(4) /* RTA_TABLE */
2447 + nla_total_size(4) /* RTA_IIF */
2448 + nla_total_size(4) /* RTA_OIF */
2449 + nla_total_size(4) /* RTA_PRIORITY */
2450 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2451 + nla_total_size(sizeof(struct rta_cacheinfo));
2454 static int rt6_fill_node(struct net *net,
2455 struct sk_buff *skb, struct rt6_info *rt,
2456 struct in6_addr *dst, struct in6_addr *src,
2457 int iif, int type, u32 portid, u32 seq,
2458 int prefix, int nowait, unsigned int flags)
2461 struct nlmsghdr *nlh;
2465 if (prefix) { /* user wants prefix routes only */
2466 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2467 /* success since this is not a prefix route */
2472 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2476 rtm = nlmsg_data(nlh);
2477 rtm->rtm_family = AF_INET6;
2478 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2479 rtm->rtm_src_len = rt->rt6i_src.plen;
2482 table = rt->rt6i_table->tb6_id;
2484 table = RT6_TABLE_UNSPEC;
2485 rtm->rtm_table = table;
2486 if (nla_put_u32(skb, RTA_TABLE, table))
2487 goto nla_put_failure;
2488 if (rt->rt6i_flags & RTF_REJECT) {
2489 switch (rt->dst.error) {
2491 rtm->rtm_type = RTN_BLACKHOLE;
2494 rtm->rtm_type = RTN_PROHIBIT;
2497 rtm->rtm_type = RTN_THROW;
2500 rtm->rtm_type = RTN_UNREACHABLE;
2504 else if (rt->rt6i_flags & RTF_LOCAL)
2505 rtm->rtm_type = RTN_LOCAL;
2506 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2507 rtm->rtm_type = RTN_LOCAL;
2509 rtm->rtm_type = RTN_UNICAST;
2511 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2512 rtm->rtm_protocol = rt->rt6i_protocol;
2513 if (rt->rt6i_flags & RTF_DYNAMIC)
2514 rtm->rtm_protocol = RTPROT_REDIRECT;
2515 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2516 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2517 rtm->rtm_protocol = RTPROT_RA;
2519 rtm->rtm_protocol = RTPROT_KERNEL;
2522 if (rt->rt6i_flags & RTF_CACHE)
2523 rtm->rtm_flags |= RTM_F_CLONED;
2526 if (nla_put(skb, RTA_DST, 16, dst))
2527 goto nla_put_failure;
2528 rtm->rtm_dst_len = 128;
2529 } else if (rtm->rtm_dst_len)
2530 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2531 goto nla_put_failure;
2532 #ifdef CONFIG_IPV6_SUBTREES
2534 if (nla_put(skb, RTA_SRC, 16, src))
2535 goto nla_put_failure;
2536 rtm->rtm_src_len = 128;
2537 } else if (rtm->rtm_src_len &&
2538 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2539 goto nla_put_failure;
2542 #ifdef CONFIG_IPV6_MROUTE
2543 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2544 int err = ip6mr_get_route(net, skb, rtm, nowait);
2549 goto nla_put_failure;
2551 if (err == -EMSGSIZE)
2552 goto nla_put_failure;
2557 if (nla_put_u32(skb, RTA_IIF, iif))
2558 goto nla_put_failure;
2560 struct in6_addr saddr_buf;
2561 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2562 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2563 goto nla_put_failure;
2566 if (rt->rt6i_prefsrc.plen) {
2567 struct in6_addr saddr_buf;
2568 saddr_buf = rt->rt6i_prefsrc.addr;
2569 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2570 goto nla_put_failure;
2573 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2574 goto nla_put_failure;
2576 if (rt->rt6i_flags & RTF_GATEWAY) {
2577 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2578 goto nla_put_failure;
2582 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2583 goto nla_put_failure;
2584 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2585 goto nla_put_failure;
2587 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2589 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2590 goto nla_put_failure;
2592 return nlmsg_end(skb, nlh);
2595 nlmsg_cancel(skb, nlh);
2599 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2601 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2604 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2605 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2606 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2610 return rt6_fill_node(arg->net,
2611 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2612 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2613 prefix, 0, NLM_F_MULTI);
2616 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2618 struct net *net = sock_net(in_skb->sk);
2619 struct nlattr *tb[RTA_MAX+1];
2620 struct rt6_info *rt;
2621 struct sk_buff *skb;
2624 int err, iif = 0, oif = 0;
2626 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2631 memset(&fl6, 0, sizeof(fl6));
2634 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2637 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2641 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2644 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2648 iif = nla_get_u32(tb[RTA_IIF]);
2651 oif = nla_get_u32(tb[RTA_OIF]);
2654 struct net_device *dev;
2657 dev = __dev_get_by_index(net, iif);
2663 fl6.flowi6_iif = iif;
2665 if (!ipv6_addr_any(&fl6.saddr))
2666 flags |= RT6_LOOKUP_F_HAS_SADDR;
2668 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2671 fl6.flowi6_oif = oif;
2673 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2676 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2683 /* Reserve room for dummy headers, this skb can pass
2684 through good chunk of routing engine.
2686 skb_reset_mac_header(skb);
2687 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2689 skb_dst_set(skb, &rt->dst);
2691 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2692 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2693 nlh->nlmsg_seq, 0, 0, 0);
2699 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2704 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2706 struct sk_buff *skb;
2707 struct net *net = info->nl_net;
2712 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2714 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2718 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2719 event, info->portid, seq, 0, 0, 0);
2721 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2722 WARN_ON(err == -EMSGSIZE);
2726 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2727 info->nlh, gfp_any());
2731 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2734 static int ip6_route_dev_notify(struct notifier_block *this,
2735 unsigned long event, void *data)
2737 struct net_device *dev = (struct net_device *)data;
2738 struct net *net = dev_net(dev);
2740 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2741 net->ipv6.ip6_null_entry->dst.dev = dev;
2742 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2743 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2744 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2745 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2746 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2747 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2758 #ifdef CONFIG_PROC_FS
2769 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2771 struct seq_file *m = p_arg;
2773 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2775 #ifdef CONFIG_IPV6_SUBTREES
2776 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2778 seq_puts(m, "00000000000000000000000000000000 00 ");
2780 if (rt->rt6i_flags & RTF_GATEWAY) {
2781 seq_printf(m, "%pi6", &rt->rt6i_gateway);
2783 seq_puts(m, "00000000000000000000000000000000");
2785 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2786 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2787 rt->dst.__use, rt->rt6i_flags,
2788 rt->dst.dev ? rt->dst.dev->name : "");
2792 static int ipv6_route_show(struct seq_file *m, void *v)
2794 struct net *net = (struct net *)m->private;
2795 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2799 static int ipv6_route_open(struct inode *inode, struct file *file)
2801 return single_open_net(inode, file, ipv6_route_show);
2804 static const struct file_operations ipv6_route_proc_fops = {
2805 .owner = THIS_MODULE,
2806 .open = ipv6_route_open,
2808 .llseek = seq_lseek,
2809 .release = single_release_net,
2812 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2814 struct net *net = (struct net *)seq->private;
2815 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2816 net->ipv6.rt6_stats->fib_nodes,
2817 net->ipv6.rt6_stats->fib_route_nodes,
2818 net->ipv6.rt6_stats->fib_rt_alloc,
2819 net->ipv6.rt6_stats->fib_rt_entries,
2820 net->ipv6.rt6_stats->fib_rt_cache,
2821 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2822 net->ipv6.rt6_stats->fib_discarded_routes);
2827 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2829 return single_open_net(inode, file, rt6_stats_seq_show);
2832 static const struct file_operations rt6_stats_seq_fops = {
2833 .owner = THIS_MODULE,
2834 .open = rt6_stats_seq_open,
2836 .llseek = seq_lseek,
2837 .release = single_release_net,
2839 #endif /* CONFIG_PROC_FS */
2841 #ifdef CONFIG_SYSCTL
2844 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2845 void __user *buffer, size_t *lenp, loff_t *ppos)
2852 net = (struct net *)ctl->extra1;
2853 delay = net->ipv6.sysctl.flush_delay;
2854 proc_dointvec(ctl, write, buffer, lenp, ppos);
2855 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2859 ctl_table ipv6_route_table_template[] = {
2861 .procname = "flush",
2862 .data = &init_net.ipv6.sysctl.flush_delay,
2863 .maxlen = sizeof(int),
2865 .proc_handler = ipv6_sysctl_rtcache_flush
2868 .procname = "gc_thresh",
2869 .data = &ip6_dst_ops_template.gc_thresh,
2870 .maxlen = sizeof(int),
2872 .proc_handler = proc_dointvec,
2875 .procname = "max_size",
2876 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2877 .maxlen = sizeof(int),
2879 .proc_handler = proc_dointvec,
2882 .procname = "gc_min_interval",
2883 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2884 .maxlen = sizeof(int),
2886 .proc_handler = proc_dointvec_jiffies,
2889 .procname = "gc_timeout",
2890 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2891 .maxlen = sizeof(int),
2893 .proc_handler = proc_dointvec_jiffies,
2896 .procname = "gc_interval",
2897 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2898 .maxlen = sizeof(int),
2900 .proc_handler = proc_dointvec_jiffies,
2903 .procname = "gc_elasticity",
2904 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2905 .maxlen = sizeof(int),
2907 .proc_handler = proc_dointvec,
2910 .procname = "mtu_expires",
2911 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2912 .maxlen = sizeof(int),
2914 .proc_handler = proc_dointvec_jiffies,
2917 .procname = "min_adv_mss",
2918 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2919 .maxlen = sizeof(int),
2921 .proc_handler = proc_dointvec,
2924 .procname = "gc_min_interval_ms",
2925 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2926 .maxlen = sizeof(int),
2928 .proc_handler = proc_dointvec_ms_jiffies,
2933 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2935 struct ctl_table *table;
2937 table = kmemdup(ipv6_route_table_template,
2938 sizeof(ipv6_route_table_template),
2942 table[0].data = &net->ipv6.sysctl.flush_delay;
2943 table[0].extra1 = net;
2944 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2945 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2946 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2947 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2948 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2949 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2950 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2951 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2952 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2954 /* Don't export sysctls to unprivileged users */
2955 if (net->user_ns != &init_user_ns)
2956 table[0].procname = NULL;
2963 static int __net_init ip6_route_net_init(struct net *net)
2967 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2968 sizeof(net->ipv6.ip6_dst_ops));
2970 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2971 goto out_ip6_dst_ops;
2973 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2974 sizeof(*net->ipv6.ip6_null_entry),
2976 if (!net->ipv6.ip6_null_entry)
2977 goto out_ip6_dst_entries;
2978 net->ipv6.ip6_null_entry->dst.path =
2979 (struct dst_entry *)net->ipv6.ip6_null_entry;
2980 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2981 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2982 ip6_template_metrics, true);
2984 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2985 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2986 sizeof(*net->ipv6.ip6_prohibit_entry),
2988 if (!net->ipv6.ip6_prohibit_entry)
2989 goto out_ip6_null_entry;
2990 net->ipv6.ip6_prohibit_entry->dst.path =
2991 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2992 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2993 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2994 ip6_template_metrics, true);
2996 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2997 sizeof(*net->ipv6.ip6_blk_hole_entry),
2999 if (!net->ipv6.ip6_blk_hole_entry)
3000 goto out_ip6_prohibit_entry;
3001 net->ipv6.ip6_blk_hole_entry->dst.path =
3002 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3003 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3004 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3005 ip6_template_metrics, true);
3008 net->ipv6.sysctl.flush_delay = 0;
3009 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3010 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3011 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3012 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3013 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3014 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3015 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3017 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3023 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3024 out_ip6_prohibit_entry:
3025 kfree(net->ipv6.ip6_prohibit_entry);
3027 kfree(net->ipv6.ip6_null_entry);
3029 out_ip6_dst_entries:
3030 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3035 static void __net_exit ip6_route_net_exit(struct net *net)
3037 kfree(net->ipv6.ip6_null_entry);
3038 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3039 kfree(net->ipv6.ip6_prohibit_entry);
3040 kfree(net->ipv6.ip6_blk_hole_entry);
3042 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3045 static int __net_init ip6_route_net_init_late(struct net *net)
3047 #ifdef CONFIG_PROC_FS
3048 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3049 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3054 static void __net_exit ip6_route_net_exit_late(struct net *net)
3056 #ifdef CONFIG_PROC_FS
3057 remove_proc_entry("ipv6_route", net->proc_net);
3058 remove_proc_entry("rt6_stats", net->proc_net);
3062 static struct pernet_operations ip6_route_net_ops = {
3063 .init = ip6_route_net_init,
3064 .exit = ip6_route_net_exit,
3067 static int __net_init ipv6_inetpeer_init(struct net *net)
3069 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3073 inet_peer_base_init(bp);
3074 net->ipv6.peers = bp;
3078 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3080 struct inet_peer_base *bp = net->ipv6.peers;
3082 net->ipv6.peers = NULL;
3083 inetpeer_invalidate_tree(bp);
3087 static struct pernet_operations ipv6_inetpeer_ops = {
3088 .init = ipv6_inetpeer_init,
3089 .exit = ipv6_inetpeer_exit,
3092 static struct pernet_operations ip6_route_net_late_ops = {
3093 .init = ip6_route_net_init_late,
3094 .exit = ip6_route_net_exit_late,
3097 static struct notifier_block ip6_route_dev_notifier = {
3098 .notifier_call = ip6_route_dev_notify,
3102 int __init ip6_route_init(void)
3107 ip6_dst_ops_template.kmem_cachep =
3108 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3109 SLAB_HWCACHE_ALIGN, NULL);
3110 if (!ip6_dst_ops_template.kmem_cachep)
3113 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3115 goto out_kmem_cache;
3117 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3119 goto out_dst_entries;
3121 ret = register_pernet_subsys(&ip6_route_net_ops);
3123 goto out_register_inetpeer;
3125 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3127 /* Registering of the loopback is done before this portion of code,
3128 * the loopback reference in rt6_info will not be taken, do it
3129 * manually for init_net */
3130 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3131 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3132 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3133 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3134 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3135 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3136 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3140 goto out_register_subsys;
3146 ret = fib6_rules_init();
3150 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3152 goto fib6_rules_init;
3155 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3156 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3157 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3158 goto out_register_late_subsys;
3160 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3162 goto out_register_late_subsys;
3167 out_register_late_subsys:
3168 unregister_pernet_subsys(&ip6_route_net_late_ops);
3170 fib6_rules_cleanup();
3175 out_register_subsys:
3176 unregister_pernet_subsys(&ip6_route_net_ops);
3177 out_register_inetpeer:
3178 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3180 dst_entries_destroy(&ip6_dst_blackhole_ops);
3182 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3186 void ip6_route_cleanup(void)
3188 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3189 unregister_pernet_subsys(&ip6_route_net_late_ops);
3190 fib6_rules_cleanup();
3193 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3194 unregister_pernet_subsys(&ip6_route_net_ops);
3195 dst_entries_destroy(&ip6_dst_blackhole_ops);
3196 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);