2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
24 * Fixed routing subtrees.
27 #define pr_fmt(fmt) "IPv6: " fmt
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/export.h>
32 #include <linux/types.h>
33 #include <linux/times.h>
34 #include <linux/socket.h>
35 #include <linux/sockios.h>
36 #include <linux/net.h>
37 #include <linux/route.h>
38 #include <linux/netdevice.h>
39 #include <linux/in6.h>
40 #include <linux/mroute6.h>
41 #include <linux/init.h>
42 #include <linux/if_arp.h>
43 #include <linux/proc_fs.h>
44 #include <linux/seq_file.h>
45 #include <linux/nsproxy.h>
46 #include <linux/slab.h>
47 #include <net/net_namespace.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ndisc.h>
53 #include <net/addrconf.h>
55 #include <linux/rtnetlink.h>
58 #include <net/netevent.h>
59 #include <net/netlink.h>
60 #include <net/nexthop.h>
62 #include <asm/uaccess.h>
65 #include <linux/sysctl.h>
69 RT6_NUD_FAIL_HARD = -2,
70 RT6_NUD_FAIL_SOFT = -1,
74 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
75 const struct in6_addr *dest);
76 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
77 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
78 static unsigned int ip6_mtu(const struct dst_entry *dst);
79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
80 static void ip6_dst_destroy(struct dst_entry *);
81 static void ip6_dst_ifdown(struct dst_entry *,
82 struct net_device *dev, int how);
83 static int ip6_dst_gc(struct dst_ops *ops);
85 static int ip6_pkt_discard(struct sk_buff *skb);
86 static int ip6_pkt_discard_out(struct sk_buff *skb);
87 static int ip6_pkt_prohibit(struct sk_buff *skb);
88 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
89 static void ip6_link_failure(struct sk_buff *skb);
90 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
91 struct sk_buff *skb, u32 mtu);
92 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
95 #ifdef CONFIG_IPV6_ROUTE_INFO
96 static struct rt6_info *rt6_add_route_info(struct net_device *dev,
97 const struct in6_addr *prefix, int prefixlen,
98 const struct in6_addr *gwaddr, unsigned int pref);
99 static struct rt6_info *rt6_get_route_info(struct net_device *dev,
100 const struct in6_addr *prefix, int prefixlen,
101 const struct in6_addr *gwaddr);
104 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
106 struct rt6_info *rt = (struct rt6_info *) dst;
107 struct inet_peer *peer;
110 if (!(rt->dst.flags & DST_HOST))
111 return dst_cow_metrics_generic(dst, old);
113 peer = rt6_get_peer_create(rt);
115 u32 *old_p = __DST_METRICS_PTR(old);
116 unsigned long prev, new;
119 if (inet_metrics_new(peer))
120 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
122 new = (unsigned long) p;
123 prev = cmpxchg(&dst->_metrics, old, new);
126 p = __DST_METRICS_PTR(prev);
127 if (prev & DST_METRICS_READ_ONLY)
134 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
138 struct in6_addr *p = &rt->rt6i_gateway;
140 if (!ipv6_addr_any(p))
141 return (const void *) p;
143 return &ipv6_hdr(skb)->daddr;
147 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
151 struct rt6_info *rt = (struct rt6_info *) dst;
154 daddr = choose_neigh_daddr(rt, skb, daddr);
155 n = __ipv6_neigh_lookup(dst->dev, daddr);
158 return neigh_create(&nd_tbl, daddr, dst->dev);
161 static struct dst_ops ip6_dst_ops_template = {
163 .protocol = cpu_to_be16(ETH_P_IPV6),
166 .check = ip6_dst_check,
167 .default_advmss = ip6_default_advmss,
169 .cow_metrics = ipv6_cow_metrics,
170 .destroy = ip6_dst_destroy,
171 .ifdown = ip6_dst_ifdown,
172 .negative_advice = ip6_negative_advice,
173 .link_failure = ip6_link_failure,
174 .update_pmtu = ip6_rt_update_pmtu,
175 .redirect = rt6_do_redirect,
176 .local_out = __ip6_local_out,
177 .neigh_lookup = ip6_neigh_lookup,
180 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
182 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
184 return mtu ? : dst->dev->mtu;
187 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
188 struct sk_buff *skb, u32 mtu)
192 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
197 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
203 static struct dst_ops ip6_dst_blackhole_ops = {
205 .protocol = cpu_to_be16(ETH_P_IPV6),
206 .destroy = ip6_dst_destroy,
207 .check = ip6_dst_check,
208 .mtu = ip6_blackhole_mtu,
209 .default_advmss = ip6_default_advmss,
210 .update_pmtu = ip6_rt_blackhole_update_pmtu,
211 .redirect = ip6_rt_blackhole_redirect,
212 .cow_metrics = ip6_rt_blackhole_cow_metrics,
213 .neigh_lookup = ip6_neigh_lookup,
216 static const u32 ip6_template_metrics[RTAX_MAX] = {
217 [RTAX_HOPLIMIT - 1] = 0,
220 static const struct rt6_info ip6_null_entry_template = {
222 .__refcnt = ATOMIC_INIT(1),
224 .obsolete = DST_OBSOLETE_FORCE_CHK,
225 .error = -ENETUNREACH,
226 .input = ip6_pkt_discard,
227 .output = ip6_pkt_discard_out,
229 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
230 .rt6i_protocol = RTPROT_KERNEL,
231 .rt6i_metric = ~(u32) 0,
232 .rt6i_ref = ATOMIC_INIT(1),
235 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
237 static const struct rt6_info ip6_prohibit_entry_template = {
239 .__refcnt = ATOMIC_INIT(1),
241 .obsolete = DST_OBSOLETE_FORCE_CHK,
243 .input = ip6_pkt_prohibit,
244 .output = ip6_pkt_prohibit_out,
246 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
247 .rt6i_protocol = RTPROT_KERNEL,
248 .rt6i_metric = ~(u32) 0,
249 .rt6i_ref = ATOMIC_INIT(1),
252 static const struct rt6_info ip6_blk_hole_entry_template = {
254 .__refcnt = ATOMIC_INIT(1),
256 .obsolete = DST_OBSOLETE_FORCE_CHK,
258 .input = dst_discard,
259 .output = dst_discard,
261 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
262 .rt6i_protocol = RTPROT_KERNEL,
263 .rt6i_metric = ~(u32) 0,
264 .rt6i_ref = ATOMIC_INIT(1),
269 /* allocate dst with ip6_dst_ops */
270 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
271 struct net_device *dev,
273 struct fib6_table *table)
275 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
276 0, DST_OBSOLETE_FORCE_CHK, flags);
279 struct dst_entry *dst = &rt->dst;
281 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
282 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
283 rt->rt6i_genid = rt_genid(net);
284 INIT_LIST_HEAD(&rt->rt6i_siblings);
285 rt->rt6i_nsiblings = 0;
290 static void ip6_dst_destroy(struct dst_entry *dst)
292 struct rt6_info *rt = (struct rt6_info *)dst;
293 struct inet6_dev *idev = rt->rt6i_idev;
294 struct dst_entry *from = dst->from;
296 if (!(rt->dst.flags & DST_HOST))
297 dst_destroy_metrics_generic(dst);
300 rt->rt6i_idev = NULL;
307 if (rt6_has_peer(rt)) {
308 struct inet_peer *peer = rt6_peer_ptr(rt);
313 void rt6_bind_peer(struct rt6_info *rt, int create)
315 struct inet_peer_base *base;
316 struct inet_peer *peer;
318 base = inetpeer_base_ptr(rt->_rt6i_peer);
322 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
324 if (!rt6_set_peer(rt, peer))
329 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
332 struct rt6_info *rt = (struct rt6_info *)dst;
333 struct inet6_dev *idev = rt->rt6i_idev;
334 struct net_device *loopback_dev =
335 dev_net(dev)->loopback_dev;
337 if (dev != loopback_dev) {
338 if (idev && idev->dev == dev) {
339 struct inet6_dev *loopback_idev =
340 in6_dev_get(loopback_dev);
342 rt->rt6i_idev = loopback_idev;
349 static bool rt6_check_expired(const struct rt6_info *rt)
351 if (rt->rt6i_flags & RTF_EXPIRES) {
352 if (time_after(jiffies, rt->dst.expires))
354 } else if (rt->dst.from) {
355 return rt6_check_expired((struct rt6_info *) rt->dst.from);
360 static bool rt6_need_strict(const struct in6_addr *daddr)
362 return ipv6_addr_type(daddr) &
363 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
366 /* Multipath route selection:
367 * Hash based function using packet header and flowlabel.
368 * Adapted from fib_info_hashfn()
370 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
371 const struct flowi6 *fl6)
373 unsigned int val = fl6->flowi6_proto;
375 val ^= ipv6_addr_hash(&fl6->daddr);
376 val ^= ipv6_addr_hash(&fl6->saddr);
378 /* Work only if this not encapsulated */
379 switch (fl6->flowi6_proto) {
383 val ^= (__force u16)fl6->fl6_sport;
384 val ^= (__force u16)fl6->fl6_dport;
388 val ^= (__force u16)fl6->fl6_icmp_type;
389 val ^= (__force u16)fl6->fl6_icmp_code;
392 /* RFC6438 recommands to use flowlabel */
393 val ^= (__force u32)fl6->flowlabel;
395 /* Perhaps, we need to tune, this function? */
396 val = val ^ (val >> 7) ^ (val >> 12);
397 return val % candidate_count;
400 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
403 struct rt6_info *sibling, *next_sibling;
406 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
407 /* Don't change the route, if route_choosen == 0
408 * (siblings does not include ourself)
411 list_for_each_entry_safe(sibling, next_sibling,
412 &match->rt6i_siblings, rt6i_siblings) {
414 if (route_choosen == 0) {
423 * Route lookup. Any table->tb6_lock is implied.
426 static inline struct rt6_info *rt6_device_match(struct net *net,
428 const struct in6_addr *saddr,
432 struct rt6_info *local = NULL;
433 struct rt6_info *sprt;
435 if (!oif && ipv6_addr_any(saddr))
438 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
439 struct net_device *dev = sprt->dst.dev;
442 if (dev->ifindex == oif)
444 if (dev->flags & IFF_LOOPBACK) {
445 if (!sprt->rt6i_idev ||
446 sprt->rt6i_idev->dev->ifindex != oif) {
447 if (flags & RT6_LOOKUP_F_IFACE && oif)
449 if (local && (!oif ||
450 local->rt6i_idev->dev->ifindex == oif))
456 if (ipv6_chk_addr(net, saddr, dev,
457 flags & RT6_LOOKUP_F_IFACE))
466 if (flags & RT6_LOOKUP_F_IFACE)
467 return net->ipv6.ip6_null_entry;
473 #ifdef CONFIG_IPV6_ROUTER_PREF
474 struct __rt6_probe_work {
475 struct work_struct work;
476 struct in6_addr target;
477 struct net_device *dev;
480 static void rt6_probe_deferred(struct work_struct *w)
482 struct in6_addr mcaddr;
483 struct __rt6_probe_work *work =
484 container_of(w, struct __rt6_probe_work, work);
486 addrconf_addr_solict_mult(&work->target, &mcaddr);
487 ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
492 static void rt6_probe(struct rt6_info *rt)
494 struct neighbour *neigh;
496 * Okay, this does not seem to be appropriate
497 * for now, however, we need to check if it
498 * is really so; aka Router Reachability Probing.
500 * Router Reachability Probe MUST be rate-limited
501 * to no more than one per minute.
503 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
506 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
508 write_lock(&neigh->lock);
509 if (neigh->nud_state & NUD_VALID)
514 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
515 struct __rt6_probe_work *work;
517 work = kmalloc(sizeof(*work), GFP_ATOMIC);
520 neigh->updated = jiffies;
523 write_unlock(&neigh->lock);
526 INIT_WORK(&work->work, rt6_probe_deferred);
527 work->target = rt->rt6i_gateway;
528 dev_hold(rt->dst.dev);
529 work->dev = rt->dst.dev;
530 schedule_work(&work->work);
534 write_unlock(&neigh->lock);
536 rcu_read_unlock_bh();
539 static inline void rt6_probe(struct rt6_info *rt)
545 * Default Router Selection (RFC 2461 6.3.6)
547 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
549 struct net_device *dev = rt->dst.dev;
550 if (!oif || dev->ifindex == oif)
552 if ((dev->flags & IFF_LOOPBACK) &&
553 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
558 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
560 struct neighbour *neigh;
561 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
563 if (rt->rt6i_flags & RTF_NONEXTHOP ||
564 !(rt->rt6i_flags & RTF_GATEWAY))
565 return RT6_NUD_SUCCEED;
568 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
570 read_lock(&neigh->lock);
571 if (neigh->nud_state & NUD_VALID)
572 ret = RT6_NUD_SUCCEED;
573 #ifdef CONFIG_IPV6_ROUTER_PREF
574 else if (!(neigh->nud_state & NUD_FAILED))
575 ret = RT6_NUD_SUCCEED;
577 read_unlock(&neigh->lock);
579 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
580 RT6_NUD_SUCCEED : RT6_NUD_FAIL_SOFT;
582 rcu_read_unlock_bh();
587 static int rt6_score_route(struct rt6_info *rt, int oif,
592 m = rt6_check_dev(rt, oif);
593 if (!m && (strict & RT6_LOOKUP_F_IFACE))
594 return RT6_NUD_FAIL_HARD;
595 #ifdef CONFIG_IPV6_ROUTER_PREF
596 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
598 if (strict & RT6_LOOKUP_F_REACHABLE) {
599 int n = rt6_check_neigh(rt);
606 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
607 int *mpri, struct rt6_info *match,
611 bool match_do_rr = false;
613 if (rt6_check_expired(rt))
616 m = rt6_score_route(rt, oif, strict);
617 if (m == RT6_NUD_FAIL_SOFT && !IS_ENABLED(CONFIG_IPV6_ROUTER_PREF)) {
619 m = 0; /* lowest valid score */
624 if (strict & RT6_LOOKUP_F_REACHABLE)
628 *do_rr = match_do_rr;
636 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
637 struct rt6_info *rr_head,
638 u32 metric, int oif, int strict,
641 struct rt6_info *rt, *match;
645 for (rt = rr_head; rt && rt->rt6i_metric == metric;
646 rt = rt->dst.rt6_next)
647 match = find_match(rt, oif, strict, &mpri, match, do_rr);
648 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
649 rt = rt->dst.rt6_next)
650 match = find_match(rt, oif, strict, &mpri, match, do_rr);
655 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
657 struct rt6_info *match, *rt0;
663 fn->rr_ptr = rt0 = fn->leaf;
665 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
669 struct rt6_info *next = rt0->dst.rt6_next;
671 /* no entries matched; do round-robin */
672 if (!next || next->rt6i_metric != rt0->rt6i_metric)
679 net = dev_net(rt0->dst.dev);
680 return match ? match : net->ipv6.ip6_null_entry;
683 #ifdef CONFIG_IPV6_ROUTE_INFO
684 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
685 const struct in6_addr *gwaddr)
687 struct route_info *rinfo = (struct route_info *) opt;
688 struct in6_addr prefix_buf, *prefix;
690 unsigned long lifetime;
693 if (len < sizeof(struct route_info)) {
697 /* Sanity check for prefix_len and length */
698 if (rinfo->length > 3) {
700 } else if (rinfo->prefix_len > 128) {
702 } else if (rinfo->prefix_len > 64) {
703 if (rinfo->length < 2) {
706 } else if (rinfo->prefix_len > 0) {
707 if (rinfo->length < 1) {
712 pref = rinfo->route_pref;
713 if (pref == ICMPV6_ROUTER_PREF_INVALID)
716 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
718 if (rinfo->length == 3)
719 prefix = (struct in6_addr *)rinfo->prefix;
721 /* this function is safe */
722 ipv6_addr_prefix(&prefix_buf,
723 (struct in6_addr *)rinfo->prefix,
725 prefix = &prefix_buf;
728 if (rinfo->prefix_len == 0)
729 rt = rt6_get_dflt_router(gwaddr, dev);
731 rt = rt6_get_route_info(dev, prefix, rinfo->prefix_len, gwaddr);
733 if (rt && !lifetime) {
739 rt = rt6_add_route_info(dev, prefix, rinfo->prefix_len, gwaddr, pref);
741 rt->rt6i_flags = RTF_ROUTEINFO |
742 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
745 if (!addrconf_finite_timeout(lifetime))
746 rt6_clean_expires(rt);
748 rt6_set_expires(rt, jiffies + HZ * lifetime);
756 #define BACKTRACK(__net, saddr) \
758 if (rt == __net->ipv6.ip6_null_entry) { \
759 struct fib6_node *pn; \
761 if (fn->fn_flags & RTN_TL_ROOT) \
764 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
765 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
768 if (fn->fn_flags & RTN_RTINFO) \
774 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
775 struct fib6_table *table,
776 struct flowi6 *fl6, int flags)
778 struct fib6_node *fn;
781 read_lock_bh(&table->tb6_lock);
782 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
785 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
786 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
787 rt = rt6_multipath_select(rt, fl6);
788 BACKTRACK(net, &fl6->saddr);
790 dst_use(&rt->dst, jiffies);
791 read_unlock_bh(&table->tb6_lock);
796 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
799 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
801 EXPORT_SYMBOL_GPL(ip6_route_lookup);
803 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
804 const struct in6_addr *saddr, int oif, int strict)
806 struct flowi6 fl6 = {
810 struct dst_entry *dst;
811 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
814 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
815 flags |= RT6_LOOKUP_F_HAS_SADDR;
818 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
820 return (struct rt6_info *) dst;
827 EXPORT_SYMBOL(rt6_lookup);
829 /* ip6_ins_rt is called with FREE table->tb6_lock.
830 It takes new route entry, the addition fails by any reason the
831 route is freed. In any case, if caller does not hold it, it may
835 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
838 struct fib6_table *table;
840 table = rt->rt6i_table;
841 write_lock_bh(&table->tb6_lock);
842 err = fib6_add(&table->tb6_root, rt, info);
843 write_unlock_bh(&table->tb6_lock);
848 int ip6_ins_rt(struct rt6_info *rt)
850 struct nl_info info = {
851 .nl_net = dev_net(rt->dst.dev),
853 return __ip6_ins_rt(rt, &info);
856 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
857 const struct in6_addr *daddr,
858 const struct in6_addr *saddr)
866 rt = ip6_rt_copy(ort, daddr);
869 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
870 if (ort->rt6i_dst.plen != 128 &&
871 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
872 rt->rt6i_flags |= RTF_ANYCAST;
875 rt->rt6i_flags |= RTF_CACHE;
877 #ifdef CONFIG_IPV6_SUBTREES
878 if (rt->rt6i_src.plen && saddr) {
879 rt->rt6i_src.addr = *saddr;
880 rt->rt6i_src.plen = 128;
888 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
889 const struct in6_addr *daddr)
891 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
894 rt->rt6i_flags |= RTF_CACHE;
898 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
899 struct flowi6 *fl6, int flags)
901 struct fib6_node *fn;
902 struct rt6_info *rt, *nrt;
906 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
908 strict |= flags & RT6_LOOKUP_F_IFACE;
911 read_lock_bh(&table->tb6_lock);
914 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
917 rt = rt6_select(fn, oif, strict | reachable);
918 if (rt->rt6i_nsiblings && oif == 0)
919 rt = rt6_multipath_select(rt, fl6);
920 BACKTRACK(net, &fl6->saddr);
921 if (rt == net->ipv6.ip6_null_entry ||
922 rt->rt6i_flags & RTF_CACHE)
926 read_unlock_bh(&table->tb6_lock);
928 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
929 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
930 else if (!(rt->dst.flags & DST_HOST))
931 nrt = rt6_alloc_clone(rt, &fl6->daddr);
936 rt = nrt ? : net->ipv6.ip6_null_entry;
940 err = ip6_ins_rt(nrt);
949 * Race condition! In the gap, when table->tb6_lock was
950 * released someone could insert this route. Relookup.
961 read_unlock_bh(&table->tb6_lock);
963 rt->dst.lastuse = jiffies;
969 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
970 struct flowi6 *fl6, int flags)
972 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
975 static struct dst_entry *ip6_route_input_lookup(struct net *net,
976 struct net_device *dev,
977 struct flowi6 *fl6, int flags)
979 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
980 flags |= RT6_LOOKUP_F_IFACE;
982 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
985 void ip6_route_input(struct sk_buff *skb)
987 const struct ipv6hdr *iph = ipv6_hdr(skb);
988 struct net *net = dev_net(skb->dev);
989 int flags = RT6_LOOKUP_F_HAS_SADDR;
990 struct flowi6 fl6 = {
991 .flowi6_iif = skb->dev->ifindex,
994 .flowlabel = ip6_flowinfo(iph),
995 .flowi6_mark = skb->mark,
996 .flowi6_proto = iph->nexthdr,
999 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1002 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1003 struct flowi6 *fl6, int flags)
1005 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1008 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1013 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1015 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1016 flags |= RT6_LOOKUP_F_IFACE;
1018 if (!ipv6_addr_any(&fl6->saddr))
1019 flags |= RT6_LOOKUP_F_HAS_SADDR;
1021 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1023 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1026 EXPORT_SYMBOL(ip6_route_output);
1028 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1030 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1031 struct dst_entry *new = NULL;
1033 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1037 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1038 rt6_init_peer(rt, net->ipv6.peers);
1041 new->input = dst_discard;
1042 new->output = dst_discard;
1044 if (dst_metrics_read_only(&ort->dst))
1045 new->_metrics = ort->dst._metrics;
1047 dst_copy_metrics(new, &ort->dst);
1048 rt->rt6i_idev = ort->rt6i_idev;
1050 in6_dev_hold(rt->rt6i_idev);
1052 rt->rt6i_gateway = ort->rt6i_gateway;
1053 rt->rt6i_flags = ort->rt6i_flags;
1054 rt->rt6i_metric = 0;
1056 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1057 #ifdef CONFIG_IPV6_SUBTREES
1058 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1064 dst_release(dst_orig);
1065 return new ? new : ERR_PTR(-ENOMEM);
1069 * Destination cache support functions
1072 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1074 struct rt6_info *rt;
1076 rt = (struct rt6_info *) dst;
1078 /* All IPV6 dsts are created with ->obsolete set to the value
1079 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1080 * into this function always.
1082 if (rt->rt6i_genid != rt_genid(dev_net(rt->dst.dev)))
1085 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1088 if (rt6_check_expired(rt))
1094 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1096 struct rt6_info *rt = (struct rt6_info *) dst;
1099 if (rt->rt6i_flags & RTF_CACHE) {
1100 if (rt6_check_expired(rt)) {
1112 static void ip6_link_failure(struct sk_buff *skb)
1114 struct rt6_info *rt;
1116 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1118 rt = (struct rt6_info *) skb_dst(skb);
1120 if (rt->rt6i_flags & RTF_CACHE) {
1124 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1125 rt->rt6i_node->fn_sernum = -1;
1130 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1131 struct sk_buff *skb, u32 mtu)
1133 struct rt6_info *rt6 = (struct rt6_info*)dst;
1136 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1137 struct net *net = dev_net(dst->dev);
1139 rt6->rt6i_flags |= RTF_MODIFIED;
1140 if (mtu < IPV6_MIN_MTU)
1143 dst_metric_set(dst, RTAX_MTU, mtu);
1144 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1148 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1149 int oif, u32 mark, kuid_t uid)
1151 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1152 struct dst_entry *dst;
1155 memset(&fl6, 0, sizeof(fl6));
1156 fl6.flowi6_oif = oif;
1157 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1158 fl6.flowi6_flags = 0;
1159 fl6.daddr = iph->daddr;
1160 fl6.saddr = iph->saddr;
1161 fl6.flowlabel = ip6_flowinfo(iph);
1162 fl6.flowi6_uid = uid;
1164 dst = ip6_route_output(net, NULL, &fl6);
1166 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1169 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1171 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1173 ip6_update_pmtu(skb, sock_net(sk), mtu,
1174 sk->sk_bound_dev_if, sk->sk_mark, sock_i_uid(sk));
1176 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1178 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1180 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1181 struct dst_entry *dst;
1184 memset(&fl6, 0, sizeof(fl6));
1185 fl6.flowi6_oif = oif;
1186 fl6.flowi6_mark = mark;
1187 fl6.flowi6_flags = 0;
1188 fl6.daddr = iph->daddr;
1189 fl6.saddr = iph->saddr;
1190 fl6.flowlabel = ip6_flowinfo(iph);
1192 dst = ip6_route_output(net, NULL, &fl6);
1194 rt6_do_redirect(dst, NULL, skb);
1197 EXPORT_SYMBOL_GPL(ip6_redirect);
1199 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1201 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1203 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1205 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1207 struct net_device *dev = dst->dev;
1208 unsigned int mtu = dst_mtu(dst);
1209 struct net *net = dev_net(dev);
1211 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1213 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1214 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1217 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1218 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1219 * IPV6_MAXPLEN is also valid and means: "any MSS,
1220 * rely only on pmtu discovery"
1222 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1227 static unsigned int ip6_mtu(const struct dst_entry *dst)
1229 struct inet6_dev *idev;
1230 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1238 idev = __in6_dev_get(dst->dev);
1240 mtu = idev->cnf.mtu6;
1244 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1247 static struct dst_entry *icmp6_dst_gc_list;
1248 static DEFINE_SPINLOCK(icmp6_dst_lock);
1250 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1253 struct dst_entry *dst;
1254 struct rt6_info *rt;
1255 struct inet6_dev *idev = in6_dev_get(dev);
1256 struct net *net = dev_net(dev);
1258 if (unlikely(!idev))
1259 return ERR_PTR(-ENODEV);
1261 rt = ip6_dst_alloc(net, dev, 0, NULL);
1262 if (unlikely(!rt)) {
1264 dst = ERR_PTR(-ENOMEM);
1268 rt->dst.flags |= DST_HOST;
1269 rt->dst.output = ip6_output;
1270 atomic_set(&rt->dst.__refcnt, 1);
1271 rt->rt6i_gateway = fl6->daddr;
1272 rt->rt6i_dst.addr = fl6->daddr;
1273 rt->rt6i_dst.plen = 128;
1274 rt->rt6i_idev = idev;
1275 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1277 spin_lock_bh(&icmp6_dst_lock);
1278 rt->dst.next = icmp6_dst_gc_list;
1279 icmp6_dst_gc_list = &rt->dst;
1280 spin_unlock_bh(&icmp6_dst_lock);
1282 fib6_force_start_gc(net);
1284 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1290 int icmp6_dst_gc(void)
1292 struct dst_entry *dst, **pprev;
1295 spin_lock_bh(&icmp6_dst_lock);
1296 pprev = &icmp6_dst_gc_list;
1298 while ((dst = *pprev) != NULL) {
1299 if (!atomic_read(&dst->__refcnt)) {
1308 spin_unlock_bh(&icmp6_dst_lock);
1313 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1316 struct dst_entry *dst, **pprev;
1318 spin_lock_bh(&icmp6_dst_lock);
1319 pprev = &icmp6_dst_gc_list;
1320 while ((dst = *pprev) != NULL) {
1321 struct rt6_info *rt = (struct rt6_info *) dst;
1322 if (func(rt, arg)) {
1329 spin_unlock_bh(&icmp6_dst_lock);
1332 static int ip6_dst_gc(struct dst_ops *ops)
1334 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1335 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1336 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1337 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1338 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1339 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1342 entries = dst_entries_get_fast(ops);
1343 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1344 entries <= rt_max_size)
1347 net->ipv6.ip6_rt_gc_expire++;
1348 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1349 entries = dst_entries_get_slow(ops);
1350 if (entries < ops->gc_thresh)
1351 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1353 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1354 return entries > rt_max_size;
1357 int ip6_dst_hoplimit(struct dst_entry *dst)
1359 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1360 if (hoplimit == 0) {
1361 struct net_device *dev = dst->dev;
1362 struct inet6_dev *idev;
1365 idev = __in6_dev_get(dev);
1367 hoplimit = idev->cnf.hop_limit;
1369 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1374 EXPORT_SYMBOL(ip6_dst_hoplimit);
1380 int ip6_route_add(struct fib6_config *cfg)
1383 struct net *net = cfg->fc_nlinfo.nl_net;
1384 struct rt6_info *rt = NULL;
1385 struct net_device *dev = NULL;
1386 struct inet6_dev *idev = NULL;
1387 struct fib6_table *table;
1390 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1392 #ifndef CONFIG_IPV6_SUBTREES
1393 if (cfg->fc_src_len)
1396 if (cfg->fc_ifindex) {
1398 dev = dev_get_by_index(net, cfg->fc_ifindex);
1401 idev = in6_dev_get(dev);
1406 if (cfg->fc_metric == 0)
1407 cfg->fc_metric = IP6_RT_PRIO_USER;
1410 if (cfg->fc_nlinfo.nlh &&
1411 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1412 table = fib6_get_table(net, cfg->fc_table);
1414 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1415 table = fib6_new_table(net, cfg->fc_table);
1418 table = fib6_new_table(net, cfg->fc_table);
1424 rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1431 if (cfg->fc_flags & RTF_EXPIRES)
1432 rt6_set_expires(rt, jiffies +
1433 clock_t_to_jiffies(cfg->fc_expires));
1435 rt6_clean_expires(rt);
1437 if (cfg->fc_protocol == RTPROT_UNSPEC)
1438 cfg->fc_protocol = RTPROT_BOOT;
1439 rt->rt6i_protocol = cfg->fc_protocol;
1441 addr_type = ipv6_addr_type(&cfg->fc_dst);
1443 if (addr_type & IPV6_ADDR_MULTICAST)
1444 rt->dst.input = ip6_mc_input;
1445 else if (cfg->fc_flags & RTF_LOCAL)
1446 rt->dst.input = ip6_input;
1448 rt->dst.input = ip6_forward;
1450 rt->dst.output = ip6_output;
1452 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1453 rt->rt6i_dst.plen = cfg->fc_dst_len;
1454 if (rt->rt6i_dst.plen == 128)
1455 rt->dst.flags |= DST_HOST;
1457 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1458 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1463 dst_init_metrics(&rt->dst, metrics, 0);
1465 #ifdef CONFIG_IPV6_SUBTREES
1466 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1467 rt->rt6i_src.plen = cfg->fc_src_len;
1470 rt->rt6i_metric = cfg->fc_metric;
1472 /* We cannot add true routes via loopback here,
1473 they would result in kernel looping; promote them to reject routes
1475 if ((cfg->fc_flags & RTF_REJECT) ||
1476 (dev && (dev->flags & IFF_LOOPBACK) &&
1477 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1478 !(cfg->fc_flags & RTF_LOCAL))) {
1479 /* hold loopback dev/idev if we haven't done so. */
1480 if (dev != net->loopback_dev) {
1485 dev = net->loopback_dev;
1487 idev = in6_dev_get(dev);
1493 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1494 switch (cfg->fc_type) {
1496 rt->dst.error = -EINVAL;
1497 rt->dst.output = dst_discard;
1498 rt->dst.input = dst_discard;
1501 rt->dst.error = -EACCES;
1502 rt->dst.output = ip6_pkt_prohibit_out;
1503 rt->dst.input = ip6_pkt_prohibit;
1507 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1509 rt->dst.output = ip6_pkt_discard_out;
1510 rt->dst.input = ip6_pkt_discard;
1516 if (cfg->fc_flags & RTF_GATEWAY) {
1517 const struct in6_addr *gw_addr;
1520 gw_addr = &cfg->fc_gateway;
1521 rt->rt6i_gateway = *gw_addr;
1522 gwa_type = ipv6_addr_type(gw_addr);
1524 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1525 struct rt6_info *grt;
1527 /* IPv6 strictly inhibits using not link-local
1528 addresses as nexthop address.
1529 Otherwise, router will not able to send redirects.
1530 It is very good, but in some (rare!) circumstances
1531 (SIT, PtP, NBMA NOARP links) it is handy to allow
1532 some exceptions. --ANK
1535 if (!(gwa_type & IPV6_ADDR_UNICAST))
1538 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1540 err = -EHOSTUNREACH;
1544 if (dev != grt->dst.dev) {
1550 idev = grt->rt6i_idev;
1552 in6_dev_hold(grt->rt6i_idev);
1554 if (!(grt->rt6i_flags & RTF_GATEWAY))
1562 if (!dev || (dev->flags & IFF_LOOPBACK))
1570 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1571 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1575 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1576 rt->rt6i_prefsrc.plen = 128;
1578 rt->rt6i_prefsrc.plen = 0;
1580 rt->rt6i_flags = cfg->fc_flags;
1587 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1588 int type = nla_type(nla);
1591 if (type > RTAX_MAX) {
1596 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1602 rt->rt6i_idev = idev;
1603 rt->rt6i_table = table;
1605 cfg->fc_nlinfo.nl_net = dev_net(dev);
1607 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1619 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1622 struct fib6_table *table;
1623 struct net *net = dev_net(rt->dst.dev);
1625 if (rt == net->ipv6.ip6_null_entry) {
1630 table = rt->rt6i_table;
1631 write_lock_bh(&table->tb6_lock);
1632 err = fib6_del(rt, info);
1633 write_unlock_bh(&table->tb6_lock);
1640 int ip6_del_rt(struct rt6_info *rt)
1642 struct nl_info info = {
1643 .nl_net = dev_net(rt->dst.dev),
1645 return __ip6_del_rt(rt, &info);
1648 static int ip6_route_del(struct fib6_config *cfg)
1650 struct fib6_table *table;
1651 struct fib6_node *fn;
1652 struct rt6_info *rt;
1655 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1659 read_lock_bh(&table->tb6_lock);
1661 fn = fib6_locate(&table->tb6_root,
1662 &cfg->fc_dst, cfg->fc_dst_len,
1663 &cfg->fc_src, cfg->fc_src_len);
1666 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1667 if (cfg->fc_ifindex &&
1669 rt->dst.dev->ifindex != cfg->fc_ifindex))
1671 if (cfg->fc_flags & RTF_GATEWAY &&
1672 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1674 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1677 read_unlock_bh(&table->tb6_lock);
1679 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1682 read_unlock_bh(&table->tb6_lock);
1687 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1689 struct net *net = dev_net(skb->dev);
1690 struct netevent_redirect netevent;
1691 struct rt6_info *rt, *nrt = NULL;
1692 struct ndisc_options ndopts;
1693 struct inet6_dev *in6_dev;
1694 struct neighbour *neigh;
1696 int optlen, on_link;
1699 optlen = skb->tail - skb->transport_header;
1700 optlen -= sizeof(*msg);
1703 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1707 msg = (struct rd_msg *)icmp6_hdr(skb);
1709 if (ipv6_addr_is_multicast(&msg->dest)) {
1710 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1715 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1717 } else if (ipv6_addr_type(&msg->target) !=
1718 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1719 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1723 in6_dev = __in6_dev_get(skb->dev);
1726 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1730 * The IP source address of the Redirect MUST be the same as the current
1731 * first-hop router for the specified ICMP Destination Address.
1734 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1735 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1740 if (ndopts.nd_opts_tgt_lladdr) {
1741 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1744 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1749 rt = (struct rt6_info *) dst;
1750 if (rt == net->ipv6.ip6_null_entry) {
1751 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1755 /* Redirect received -> path was valid.
1756 * Look, redirects are sent only in response to data packets,
1757 * so that this nexthop apparently is reachable. --ANK
1759 dst_confirm(&rt->dst);
1761 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1766 * We have finally decided to accept it.
1769 neigh_update(neigh, lladdr, NUD_STALE,
1770 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1771 NEIGH_UPDATE_F_OVERRIDE|
1772 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1773 NEIGH_UPDATE_F_ISROUTER))
1776 nrt = ip6_rt_copy(rt, &msg->dest);
1780 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1782 nrt->rt6i_flags &= ~RTF_GATEWAY;
1784 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1786 if (ip6_ins_rt(nrt))
1789 netevent.old = &rt->dst;
1790 netevent.new = &nrt->dst;
1791 netevent.daddr = &msg->dest;
1792 netevent.neigh = neigh;
1793 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1795 if (rt->rt6i_flags & RTF_CACHE) {
1796 rt = (struct rt6_info *) dst_clone(&rt->dst);
1801 neigh_release(neigh);
1805 * Misc support functions
1808 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1809 const struct in6_addr *dest)
1811 struct net *net = dev_net(ort->dst.dev);
1812 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1816 rt->dst.input = ort->dst.input;
1817 rt->dst.output = ort->dst.output;
1818 rt->dst.flags |= DST_HOST;
1820 rt->rt6i_dst.addr = *dest;
1821 rt->rt6i_dst.plen = 128;
1822 dst_copy_metrics(&rt->dst, &ort->dst);
1823 rt->dst.error = ort->dst.error;
1824 rt->rt6i_idev = ort->rt6i_idev;
1826 in6_dev_hold(rt->rt6i_idev);
1827 rt->dst.lastuse = jiffies;
1829 if (ort->rt6i_flags & RTF_GATEWAY)
1830 rt->rt6i_gateway = ort->rt6i_gateway;
1832 rt->rt6i_gateway = *dest;
1833 rt->rt6i_flags = ort->rt6i_flags;
1834 rt6_set_from(rt, ort);
1835 rt->rt6i_metric = 0;
1837 #ifdef CONFIG_IPV6_SUBTREES
1838 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1840 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1841 rt->rt6i_table = ort->rt6i_table;
1846 #ifdef CONFIG_IPV6_ROUTE_INFO
1847 static struct rt6_info *rt6_get_route_info(struct net_device *dev,
1848 const struct in6_addr *prefix, int prefixlen,
1849 const struct in6_addr *gwaddr)
1851 struct fib6_node *fn;
1852 struct rt6_info *rt = NULL;
1853 struct fib6_table *table;
1855 table = fib6_get_table(dev_net(dev),
1856 addrconf_rt_table(dev, RT6_TABLE_INFO));
1860 read_lock_bh(&table->tb6_lock);
1861 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1865 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1866 if (rt->dst.dev->ifindex != dev->ifindex)
1868 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1870 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1876 read_unlock_bh(&table->tb6_lock);
1880 static struct rt6_info *rt6_add_route_info(struct net_device *dev,
1881 const struct in6_addr *prefix, int prefixlen,
1882 const struct in6_addr *gwaddr, unsigned int pref)
1884 struct fib6_config cfg = {
1885 .fc_table = addrconf_rt_table(dev, RT6_TABLE_INFO),
1886 .fc_metric = IP6_RT_PRIO_USER,
1887 .fc_ifindex = dev->ifindex,
1888 .fc_dst_len = prefixlen,
1889 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1890 RTF_UP | RTF_PREF(pref),
1891 .fc_nlinfo.portid = 0,
1892 .fc_nlinfo.nlh = NULL,
1893 .fc_nlinfo.nl_net = dev_net(dev),
1896 cfg.fc_dst = *prefix;
1897 cfg.fc_gateway = *gwaddr;
1899 /* We should treat it as a default route if prefix length is 0. */
1901 cfg.fc_flags |= RTF_DEFAULT;
1903 ip6_route_add(&cfg);
1905 return rt6_get_route_info(dev, prefix, prefixlen, gwaddr);
1909 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1911 struct rt6_info *rt;
1912 struct fib6_table *table;
1914 table = fib6_get_table(dev_net(dev),
1915 addrconf_rt_table(dev, RT6_TABLE_MAIN));
1919 read_lock_bh(&table->tb6_lock);
1920 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1921 if (dev == rt->dst.dev &&
1922 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1923 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1928 read_unlock_bh(&table->tb6_lock);
1932 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1933 struct net_device *dev,
1936 struct fib6_config cfg = {
1937 .fc_table = addrconf_rt_table(dev, RT6_TABLE_DFLT),
1938 .fc_metric = IP6_RT_PRIO_USER,
1939 .fc_ifindex = dev->ifindex,
1940 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1941 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1942 .fc_nlinfo.portid = 0,
1943 .fc_nlinfo.nlh = NULL,
1944 .fc_nlinfo.nl_net = dev_net(dev),
1947 cfg.fc_gateway = *gwaddr;
1949 ip6_route_add(&cfg);
1951 return rt6_get_dflt_router(gwaddr, dev);
1955 int rt6_addrconf_purge(struct rt6_info *rt, void *arg) {
1956 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
1957 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2))
1962 void rt6_purge_dflt_routers(struct net *net)
1964 fib6_clean_all(net, rt6_addrconf_purge, 0, NULL);
1967 static void rtmsg_to_fib6_config(struct net *net,
1968 struct in6_rtmsg *rtmsg,
1969 struct fib6_config *cfg)
1971 memset(cfg, 0, sizeof(*cfg));
1973 cfg->fc_table = RT6_TABLE_MAIN;
1974 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1975 cfg->fc_metric = rtmsg->rtmsg_metric;
1976 cfg->fc_expires = rtmsg->rtmsg_info;
1977 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1978 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1979 cfg->fc_flags = rtmsg->rtmsg_flags;
1981 cfg->fc_nlinfo.nl_net = net;
1983 cfg->fc_dst = rtmsg->rtmsg_dst;
1984 cfg->fc_src = rtmsg->rtmsg_src;
1985 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1988 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1990 struct fib6_config cfg;
1991 struct in6_rtmsg rtmsg;
1995 case SIOCADDRT: /* Add a route */
1996 case SIOCDELRT: /* Delete a route */
1997 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1999 err = copy_from_user(&rtmsg, arg,
2000 sizeof(struct in6_rtmsg));
2004 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2009 err = ip6_route_add(&cfg);
2012 err = ip6_route_del(&cfg);
2026 * Drop the packet on the floor
2029 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2032 struct dst_entry *dst = skb_dst(skb);
2033 switch (ipstats_mib_noroutes) {
2034 case IPSTATS_MIB_INNOROUTES:
2035 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2036 if (type == IPV6_ADDR_ANY) {
2037 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2038 IPSTATS_MIB_INADDRERRORS);
2042 case IPSTATS_MIB_OUTNOROUTES:
2043 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2044 ipstats_mib_noroutes);
2047 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2052 static int ip6_pkt_discard(struct sk_buff *skb)
2054 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2057 static int ip6_pkt_discard_out(struct sk_buff *skb)
2059 skb->dev = skb_dst(skb)->dev;
2060 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2063 static int ip6_pkt_prohibit(struct sk_buff *skb)
2065 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2068 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2070 skb->dev = skb_dst(skb)->dev;
2071 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2075 * Allocate a dst for local (unicast / anycast) address.
2078 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2079 const struct in6_addr *addr,
2082 struct net *net = dev_net(idev->dev);
2083 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2086 return ERR_PTR(-ENOMEM);
2090 rt->dst.flags |= DST_HOST;
2091 rt->dst.input = ip6_input;
2092 rt->dst.output = ip6_output;
2093 rt->rt6i_idev = idev;
2095 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2097 rt->rt6i_flags |= RTF_ANYCAST;
2099 rt->rt6i_flags |= RTF_LOCAL;
2101 rt->rt6i_gateway = *addr;
2102 rt->rt6i_dst.addr = *addr;
2103 rt->rt6i_dst.plen = 128;
2104 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2106 atomic_set(&rt->dst.__refcnt, 1);
2111 int ip6_route_get_saddr(struct net *net,
2112 struct rt6_info *rt,
2113 const struct in6_addr *daddr,
2115 struct in6_addr *saddr)
2117 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2119 if (rt->rt6i_prefsrc.plen)
2120 *saddr = rt->rt6i_prefsrc.addr;
2122 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2123 daddr, prefs, saddr);
2127 /* remove deleted ip from prefsrc entries */
2128 struct arg_dev_net_ip {
2129 struct net_device *dev;
2131 struct in6_addr *addr;
2134 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2136 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2137 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2138 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2140 if (((void *)rt->dst.dev == dev || !dev) &&
2141 rt != net->ipv6.ip6_null_entry &&
2142 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2143 /* remove prefsrc entry */
2144 rt->rt6i_prefsrc.plen = 0;
2149 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2151 struct net *net = dev_net(ifp->idev->dev);
2152 struct arg_dev_net_ip adni = {
2153 .dev = ifp->idev->dev,
2157 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2160 struct arg_dev_net {
2161 struct net_device *dev;
2165 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2167 const struct arg_dev_net *adn = arg;
2168 const struct net_device *dev = adn->dev;
2170 if ((rt->dst.dev == dev || !dev) &&
2171 rt != adn->net->ipv6.ip6_null_entry)
2177 void rt6_ifdown(struct net *net, struct net_device *dev)
2179 struct arg_dev_net adn = {
2184 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2185 icmp6_clean_all(fib6_ifdown, &adn);
2188 struct rt6_mtu_change_arg {
2189 struct net_device *dev;
2193 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2195 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2196 struct inet6_dev *idev;
2198 /* In IPv6 pmtu discovery is not optional,
2199 so that RTAX_MTU lock cannot disable it.
2200 We still use this lock to block changes
2201 caused by addrconf/ndisc.
2204 idev = __in6_dev_get(arg->dev);
2208 /* For administrative MTU increase, there is no way to discover
2209 IPv6 PMTU increase, so PMTU increase should be updated here.
2210 Since RFC 1981 doesn't include administrative MTU increase
2211 update PMTU increase is a MUST. (i.e. jumbo frame)
2214 If new MTU is less than route PMTU, this new MTU will be the
2215 lowest MTU in the path, update the route PMTU to reflect PMTU
2216 decreases; if new MTU is greater than route PMTU, and the
2217 old MTU is the lowest MTU in the path, update the route PMTU
2218 to reflect the increase. In this case if the other nodes' MTU
2219 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2222 if (rt->dst.dev == arg->dev &&
2223 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2224 (dst_mtu(&rt->dst) >= arg->mtu ||
2225 (dst_mtu(&rt->dst) < arg->mtu &&
2226 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2227 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2232 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2234 struct rt6_mtu_change_arg arg = {
2239 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2242 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2243 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2244 [RTA_OIF] = { .type = NLA_U32 },
2245 [RTA_IIF] = { .type = NLA_U32 },
2246 [RTA_PRIORITY] = { .type = NLA_U32 },
2247 [RTA_METRICS] = { .type = NLA_NESTED },
2248 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2249 [RTA_UID] = { .type = NLA_U32 },
2252 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2253 struct fib6_config *cfg)
2256 struct nlattr *tb[RTA_MAX+1];
2259 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2264 rtm = nlmsg_data(nlh);
2265 memset(cfg, 0, sizeof(*cfg));
2267 cfg->fc_table = rtm->rtm_table;
2268 cfg->fc_dst_len = rtm->rtm_dst_len;
2269 cfg->fc_src_len = rtm->rtm_src_len;
2270 cfg->fc_flags = RTF_UP;
2271 cfg->fc_protocol = rtm->rtm_protocol;
2272 cfg->fc_type = rtm->rtm_type;
2274 if (rtm->rtm_type == RTN_UNREACHABLE ||
2275 rtm->rtm_type == RTN_BLACKHOLE ||
2276 rtm->rtm_type == RTN_PROHIBIT ||
2277 rtm->rtm_type == RTN_THROW)
2278 cfg->fc_flags |= RTF_REJECT;
2280 if (rtm->rtm_type == RTN_LOCAL)
2281 cfg->fc_flags |= RTF_LOCAL;
2283 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2284 cfg->fc_nlinfo.nlh = nlh;
2285 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2287 if (tb[RTA_GATEWAY]) {
2288 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2289 cfg->fc_flags |= RTF_GATEWAY;
2293 int plen = (rtm->rtm_dst_len + 7) >> 3;
2295 if (nla_len(tb[RTA_DST]) < plen)
2298 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2302 int plen = (rtm->rtm_src_len + 7) >> 3;
2304 if (nla_len(tb[RTA_SRC]) < plen)
2307 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2310 if (tb[RTA_PREFSRC])
2311 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2314 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2316 if (tb[RTA_PRIORITY])
2317 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2319 if (tb[RTA_METRICS]) {
2320 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2321 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2325 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2327 if (tb[RTA_MULTIPATH]) {
2328 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2329 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2337 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2339 struct fib6_config r_cfg;
2340 struct rtnexthop *rtnh;
2343 int err = 0, last_err = 0;
2346 rtnh = (struct rtnexthop *)cfg->fc_mp;
2347 remaining = cfg->fc_mp_len;
2349 /* Parse a Multipath Entry */
2350 while (rtnh_ok(rtnh, remaining)) {
2351 memcpy(&r_cfg, cfg, sizeof(*cfg));
2352 if (rtnh->rtnh_ifindex)
2353 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2355 attrlen = rtnh_attrlen(rtnh);
2357 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2359 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2361 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2362 r_cfg.fc_flags |= RTF_GATEWAY;
2365 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2368 /* If we are trying to remove a route, do not stop the
2369 * loop when ip6_route_del() fails (because next hop is
2370 * already gone), we should try to remove all next hops.
2373 /* If add fails, we should try to delete all
2374 * next hops that have been already added.
2380 /* Because each route is added like a single route we remove
2381 * this flag after the first nexthop (if there is a collision,
2382 * we have already fail to add the first nexthop:
2383 * fib6_add_rt2node() has reject it).
2385 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2386 rtnh = rtnh_next(rtnh, &remaining);
2392 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2394 struct fib6_config cfg;
2397 err = rtm_to_fib6_config(skb, nlh, &cfg);
2402 return ip6_route_multipath(&cfg, 0);
2404 return ip6_route_del(&cfg);
2407 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2409 struct fib6_config cfg;
2412 err = rtm_to_fib6_config(skb, nlh, &cfg);
2417 return ip6_route_multipath(&cfg, 1);
2419 return ip6_route_add(&cfg);
2422 static inline size_t rt6_nlmsg_size(void)
2424 return NLMSG_ALIGN(sizeof(struct rtmsg))
2425 + nla_total_size(16) /* RTA_SRC */
2426 + nla_total_size(16) /* RTA_DST */
2427 + nla_total_size(16) /* RTA_GATEWAY */
2428 + nla_total_size(16) /* RTA_PREFSRC */
2429 + nla_total_size(4) /* RTA_TABLE */
2430 + nla_total_size(4) /* RTA_IIF */
2431 + nla_total_size(4) /* RTA_OIF */
2432 + nla_total_size(4) /* RTA_PRIORITY */
2433 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2434 + nla_total_size(sizeof(struct rta_cacheinfo));
2437 static int rt6_fill_node(struct net *net,
2438 struct sk_buff *skb, struct rt6_info *rt,
2439 struct in6_addr *dst, struct in6_addr *src,
2440 int iif, int type, u32 portid, u32 seq,
2441 int prefix, int nowait, unsigned int flags)
2444 struct nlmsghdr *nlh;
2448 if (prefix) { /* user wants prefix routes only */
2449 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2450 /* success since this is not a prefix route */
2455 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2459 rtm = nlmsg_data(nlh);
2460 rtm->rtm_family = AF_INET6;
2461 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2462 rtm->rtm_src_len = rt->rt6i_src.plen;
2465 table = rt->rt6i_table->tb6_id;
2467 table = RT6_TABLE_UNSPEC;
2468 rtm->rtm_table = table;
2469 if (nla_put_u32(skb, RTA_TABLE, table))
2470 goto nla_put_failure;
2471 if (rt->rt6i_flags & RTF_REJECT) {
2472 switch (rt->dst.error) {
2474 rtm->rtm_type = RTN_BLACKHOLE;
2477 rtm->rtm_type = RTN_PROHIBIT;
2480 rtm->rtm_type = RTN_THROW;
2483 rtm->rtm_type = RTN_UNREACHABLE;
2487 else if (rt->rt6i_flags & RTF_LOCAL)
2488 rtm->rtm_type = RTN_LOCAL;
2489 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2490 rtm->rtm_type = RTN_LOCAL;
2492 rtm->rtm_type = RTN_UNICAST;
2494 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2495 rtm->rtm_protocol = rt->rt6i_protocol;
2496 if (rt->rt6i_flags & RTF_DYNAMIC)
2497 rtm->rtm_protocol = RTPROT_REDIRECT;
2498 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2499 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2500 rtm->rtm_protocol = RTPROT_RA;
2502 rtm->rtm_protocol = RTPROT_KERNEL;
2505 if (rt->rt6i_flags & RTF_CACHE)
2506 rtm->rtm_flags |= RTM_F_CLONED;
2509 if (nla_put(skb, RTA_DST, 16, dst))
2510 goto nla_put_failure;
2511 rtm->rtm_dst_len = 128;
2512 } else if (rtm->rtm_dst_len)
2513 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2514 goto nla_put_failure;
2515 #ifdef CONFIG_IPV6_SUBTREES
2517 if (nla_put(skb, RTA_SRC, 16, src))
2518 goto nla_put_failure;
2519 rtm->rtm_src_len = 128;
2520 } else if (rtm->rtm_src_len &&
2521 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2522 goto nla_put_failure;
2525 #ifdef CONFIG_IPV6_MROUTE
2526 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2527 int err = ip6mr_get_route(net, skb, rtm, nowait);
2532 goto nla_put_failure;
2534 if (err == -EMSGSIZE)
2535 goto nla_put_failure;
2540 if (nla_put_u32(skb, RTA_IIF, iif))
2541 goto nla_put_failure;
2543 struct in6_addr saddr_buf;
2544 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2545 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2546 goto nla_put_failure;
2549 if (rt->rt6i_prefsrc.plen) {
2550 struct in6_addr saddr_buf;
2551 saddr_buf = rt->rt6i_prefsrc.addr;
2552 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2553 goto nla_put_failure;
2556 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2557 goto nla_put_failure;
2559 if (rt->rt6i_flags & RTF_GATEWAY) {
2560 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2561 goto nla_put_failure;
2565 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2566 goto nla_put_failure;
2567 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2568 goto nla_put_failure;
2570 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2572 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2573 goto nla_put_failure;
2575 return nlmsg_end(skb, nlh);
2578 nlmsg_cancel(skb, nlh);
2582 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2584 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2587 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2588 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2589 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2593 return rt6_fill_node(arg->net,
2594 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2595 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2596 prefix, 0, NLM_F_MULTI);
2599 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2601 struct net *net = sock_net(in_skb->sk);
2602 struct nlattr *tb[RTA_MAX+1];
2603 struct rt6_info *rt;
2604 struct sk_buff *skb;
2607 int err, iif = 0, oif = 0;
2609 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2614 memset(&fl6, 0, sizeof(fl6));
2617 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2620 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2624 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2627 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2631 iif = nla_get_u32(tb[RTA_IIF]);
2634 oif = nla_get_u32(tb[RTA_OIF]);
2637 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
2640 fl6.flowi6_uid = make_kuid(current_user_ns(),
2641 nla_get_u32(tb[RTA_UID]));
2643 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
2646 struct net_device *dev;
2649 dev = __dev_get_by_index(net, iif);
2655 fl6.flowi6_iif = iif;
2657 if (!ipv6_addr_any(&fl6.saddr))
2658 flags |= RT6_LOOKUP_F_HAS_SADDR;
2660 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2663 fl6.flowi6_oif = oif;
2665 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2668 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2675 /* Reserve room for dummy headers, this skb can pass
2676 through good chunk of routing engine.
2678 skb_reset_mac_header(skb);
2679 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2681 skb_dst_set(skb, &rt->dst);
2683 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2684 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2685 nlh->nlmsg_seq, 0, 0, 0);
2691 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2696 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2698 struct sk_buff *skb;
2699 struct net *net = info->nl_net;
2704 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2706 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2710 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2711 event, info->portid, seq, 0, 0, 0);
2713 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2714 WARN_ON(err == -EMSGSIZE);
2718 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2719 info->nlh, gfp_any());
2723 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2726 static int ip6_route_dev_notify(struct notifier_block *this,
2727 unsigned long event, void *data)
2729 struct net_device *dev = (struct net_device *)data;
2730 struct net *net = dev_net(dev);
2732 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2733 net->ipv6.ip6_null_entry->dst.dev = dev;
2734 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2735 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2736 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2737 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2738 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2739 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2750 #ifdef CONFIG_PROC_FS
2761 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2763 struct seq_file *m = p_arg;
2765 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2767 #ifdef CONFIG_IPV6_SUBTREES
2768 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2770 seq_puts(m, "00000000000000000000000000000000 00 ");
2772 if (rt->rt6i_flags & RTF_GATEWAY) {
2773 seq_printf(m, "%pi6", &rt->rt6i_gateway);
2775 seq_puts(m, "00000000000000000000000000000000");
2777 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2778 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2779 rt->dst.__use, rt->rt6i_flags,
2780 rt->dst.dev ? rt->dst.dev->name : "");
2784 static int ipv6_route_show(struct seq_file *m, void *v)
2786 struct net *net = (struct net *)m->private;
2787 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2791 static int ipv6_route_open(struct inode *inode, struct file *file)
2793 return single_open_net(inode, file, ipv6_route_show);
2796 static const struct file_operations ipv6_route_proc_fops = {
2797 .owner = THIS_MODULE,
2798 .open = ipv6_route_open,
2800 .llseek = seq_lseek,
2801 .release = single_release_net,
2804 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2806 struct net *net = (struct net *)seq->private;
2807 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2808 net->ipv6.rt6_stats->fib_nodes,
2809 net->ipv6.rt6_stats->fib_route_nodes,
2810 net->ipv6.rt6_stats->fib_rt_alloc,
2811 net->ipv6.rt6_stats->fib_rt_entries,
2812 net->ipv6.rt6_stats->fib_rt_cache,
2813 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2814 net->ipv6.rt6_stats->fib_discarded_routes);
2819 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2821 return single_open_net(inode, file, rt6_stats_seq_show);
2824 static const struct file_operations rt6_stats_seq_fops = {
2825 .owner = THIS_MODULE,
2826 .open = rt6_stats_seq_open,
2828 .llseek = seq_lseek,
2829 .release = single_release_net,
2831 #endif /* CONFIG_PROC_FS */
2833 #ifdef CONFIG_SYSCTL
2836 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2837 void __user *buffer, size_t *lenp, loff_t *ppos)
2844 net = (struct net *)ctl->extra1;
2845 delay = net->ipv6.sysctl.flush_delay;
2846 proc_dointvec(ctl, write, buffer, lenp, ppos);
2847 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2851 ctl_table ipv6_route_table_template[] = {
2853 .procname = "flush",
2854 .data = &init_net.ipv6.sysctl.flush_delay,
2855 .maxlen = sizeof(int),
2857 .proc_handler = ipv6_sysctl_rtcache_flush
2860 .procname = "gc_thresh",
2861 .data = &ip6_dst_ops_template.gc_thresh,
2862 .maxlen = sizeof(int),
2864 .proc_handler = proc_dointvec,
2867 .procname = "max_size",
2868 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2869 .maxlen = sizeof(int),
2871 .proc_handler = proc_dointvec,
2874 .procname = "gc_min_interval",
2875 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2876 .maxlen = sizeof(int),
2878 .proc_handler = proc_dointvec_jiffies,
2881 .procname = "gc_timeout",
2882 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2883 .maxlen = sizeof(int),
2885 .proc_handler = proc_dointvec_jiffies,
2888 .procname = "gc_interval",
2889 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2890 .maxlen = sizeof(int),
2892 .proc_handler = proc_dointvec_jiffies,
2895 .procname = "gc_elasticity",
2896 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2897 .maxlen = sizeof(int),
2899 .proc_handler = proc_dointvec,
2902 .procname = "mtu_expires",
2903 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2904 .maxlen = sizeof(int),
2906 .proc_handler = proc_dointvec_jiffies,
2909 .procname = "min_adv_mss",
2910 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2911 .maxlen = sizeof(int),
2913 .proc_handler = proc_dointvec,
2916 .procname = "gc_min_interval_ms",
2917 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2918 .maxlen = sizeof(int),
2920 .proc_handler = proc_dointvec_ms_jiffies,
2925 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2927 struct ctl_table *table;
2929 table = kmemdup(ipv6_route_table_template,
2930 sizeof(ipv6_route_table_template),
2934 table[0].data = &net->ipv6.sysctl.flush_delay;
2935 table[0].extra1 = net;
2936 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2937 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2938 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2939 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2940 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2941 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2942 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2943 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2944 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2946 /* Don't export sysctls to unprivileged users */
2947 if (net->user_ns != &init_user_ns)
2948 table[0].procname = NULL;
2955 static int __net_init ip6_route_net_init(struct net *net)
2959 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2960 sizeof(net->ipv6.ip6_dst_ops));
2962 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2963 goto out_ip6_dst_ops;
2965 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2966 sizeof(*net->ipv6.ip6_null_entry),
2968 if (!net->ipv6.ip6_null_entry)
2969 goto out_ip6_dst_entries;
2970 net->ipv6.ip6_null_entry->dst.path =
2971 (struct dst_entry *)net->ipv6.ip6_null_entry;
2972 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2973 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2974 ip6_template_metrics, true);
2976 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2977 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2978 sizeof(*net->ipv6.ip6_prohibit_entry),
2980 if (!net->ipv6.ip6_prohibit_entry)
2981 goto out_ip6_null_entry;
2982 net->ipv6.ip6_prohibit_entry->dst.path =
2983 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2984 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2985 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2986 ip6_template_metrics, true);
2988 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2989 sizeof(*net->ipv6.ip6_blk_hole_entry),
2991 if (!net->ipv6.ip6_blk_hole_entry)
2992 goto out_ip6_prohibit_entry;
2993 net->ipv6.ip6_blk_hole_entry->dst.path =
2994 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2995 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2996 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2997 ip6_template_metrics, true);
3000 net->ipv6.sysctl.flush_delay = 0;
3001 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3002 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3003 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3004 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3005 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3006 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3007 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3009 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3015 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3016 out_ip6_prohibit_entry:
3017 kfree(net->ipv6.ip6_prohibit_entry);
3019 kfree(net->ipv6.ip6_null_entry);
3021 out_ip6_dst_entries:
3022 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3027 static void __net_exit ip6_route_net_exit(struct net *net)
3029 kfree(net->ipv6.ip6_null_entry);
3030 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3031 kfree(net->ipv6.ip6_prohibit_entry);
3032 kfree(net->ipv6.ip6_blk_hole_entry);
3034 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3037 static int __net_init ip6_route_net_init_late(struct net *net)
3039 #ifdef CONFIG_PROC_FS
3040 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3041 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3046 static void __net_exit ip6_route_net_exit_late(struct net *net)
3048 #ifdef CONFIG_PROC_FS
3049 remove_proc_entry("ipv6_route", net->proc_net);
3050 remove_proc_entry("rt6_stats", net->proc_net);
3054 static struct pernet_operations ip6_route_net_ops = {
3055 .init = ip6_route_net_init,
3056 .exit = ip6_route_net_exit,
3059 static int __net_init ipv6_inetpeer_init(struct net *net)
3061 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3065 inet_peer_base_init(bp);
3066 net->ipv6.peers = bp;
3070 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3072 struct inet_peer_base *bp = net->ipv6.peers;
3074 net->ipv6.peers = NULL;
3075 inetpeer_invalidate_tree(bp);
3079 static struct pernet_operations ipv6_inetpeer_ops = {
3080 .init = ipv6_inetpeer_init,
3081 .exit = ipv6_inetpeer_exit,
3084 static struct pernet_operations ip6_route_net_late_ops = {
3085 .init = ip6_route_net_init_late,
3086 .exit = ip6_route_net_exit_late,
3089 static struct notifier_block ip6_route_dev_notifier = {
3090 .notifier_call = ip6_route_dev_notify,
3094 int __init ip6_route_init(void)
3099 ip6_dst_ops_template.kmem_cachep =
3100 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3101 SLAB_HWCACHE_ALIGN, NULL);
3102 if (!ip6_dst_ops_template.kmem_cachep)
3105 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3107 goto out_kmem_cache;
3109 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3111 goto out_dst_entries;
3113 ret = register_pernet_subsys(&ip6_route_net_ops);
3115 goto out_register_inetpeer;
3117 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3119 /* Registering of the loopback is done before this portion of code,
3120 * the loopback reference in rt6_info will not be taken, do it
3121 * manually for init_net */
3122 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3123 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3124 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3125 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3126 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3127 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3128 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3132 goto out_register_subsys;
3138 ret = fib6_rules_init();
3142 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3144 goto fib6_rules_init;
3147 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3148 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3149 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3150 goto out_register_late_subsys;
3152 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3154 goto out_register_late_subsys;
3159 out_register_late_subsys:
3160 unregister_pernet_subsys(&ip6_route_net_late_ops);
3162 fib6_rules_cleanup();
3167 out_register_subsys:
3168 unregister_pernet_subsys(&ip6_route_net_ops);
3169 out_register_inetpeer:
3170 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3172 dst_entries_destroy(&ip6_dst_blackhole_ops);
3174 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3178 void ip6_route_cleanup(void)
3180 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3181 unregister_pernet_subsys(&ip6_route_net_late_ops);
3182 fib6_rules_cleanup();
3185 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3186 unregister_pernet_subsys(&ip6_route_net_ops);
3187 dst_entries_destroy(&ip6_dst_blackhole_ops);
3188 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);