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>
57 #include <net/dst_metadata.h>
59 #include <net/netevent.h>
60 #include <net/netlink.h>
61 #include <net/nexthop.h>
62 #include <net/lwtunnel.h>
63 #include <net/ip_tunnels.h>
64 #include <net/l3mdev.h>
65 #include <trace/events/fib6.h>
67 #include <asm/uaccess.h>
70 #include <linux/sysctl.h>
74 RT6_NUD_FAIL_HARD = -3,
75 RT6_NUD_FAIL_PROBE = -2,
76 RT6_NUD_FAIL_DO_RR = -1,
80 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
81 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
82 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
83 static unsigned int ip6_mtu(const struct dst_entry *dst);
84 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
85 static void ip6_dst_destroy(struct dst_entry *);
86 static void ip6_dst_ifdown(struct dst_entry *,
87 struct net_device *dev, int how);
88 static int ip6_dst_gc(struct dst_ops *ops);
90 static int ip6_pkt_discard(struct sk_buff *skb);
91 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
92 static int ip6_pkt_prohibit(struct sk_buff *skb);
93 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
94 static void ip6_link_failure(struct sk_buff *skb);
95 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
96 struct sk_buff *skb, u32 mtu);
97 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
99 static void rt6_dst_from_metrics_check(struct rt6_info *rt);
100 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
102 #ifdef CONFIG_IPV6_ROUTE_INFO
103 static struct rt6_info *rt6_add_route_info(struct net *net,
104 const struct in6_addr *prefix, int prefixlen,
105 const struct in6_addr *gwaddr, int ifindex,
107 static struct rt6_info *rt6_get_route_info(struct net *net,
108 const struct in6_addr *prefix, int prefixlen,
109 const struct in6_addr *gwaddr, int ifindex);
112 struct uncached_list {
114 struct list_head head;
117 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
119 static void rt6_uncached_list_add(struct rt6_info *rt)
121 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
123 rt->dst.flags |= DST_NOCACHE;
124 rt->rt6i_uncached_list = ul;
126 spin_lock_bh(&ul->lock);
127 list_add_tail(&rt->rt6i_uncached, &ul->head);
128 spin_unlock_bh(&ul->lock);
131 static void rt6_uncached_list_del(struct rt6_info *rt)
133 if (!list_empty(&rt->rt6i_uncached)) {
134 struct uncached_list *ul = rt->rt6i_uncached_list;
136 spin_lock_bh(&ul->lock);
137 list_del(&rt->rt6i_uncached);
138 spin_unlock_bh(&ul->lock);
142 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
144 struct net_device *loopback_dev = net->loopback_dev;
147 if (dev == loopback_dev)
150 for_each_possible_cpu(cpu) {
151 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
154 spin_lock_bh(&ul->lock);
155 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
156 struct inet6_dev *rt_idev = rt->rt6i_idev;
157 struct net_device *rt_dev = rt->dst.dev;
159 if (rt_idev->dev == dev) {
160 rt->rt6i_idev = in6_dev_get(loopback_dev);
161 in6_dev_put(rt_idev);
165 rt->dst.dev = loopback_dev;
166 dev_hold(rt->dst.dev);
170 spin_unlock_bh(&ul->lock);
174 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
176 return dst_metrics_write_ptr(rt->dst.from);
179 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
181 struct rt6_info *rt = (struct rt6_info *)dst;
183 if (rt->rt6i_flags & RTF_PCPU)
184 return rt6_pcpu_cow_metrics(rt);
185 else if (rt->rt6i_flags & RTF_CACHE)
188 return dst_cow_metrics_generic(dst, old);
191 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
195 struct in6_addr *p = &rt->rt6i_gateway;
197 if (!ipv6_addr_any(p))
198 return (const void *) p;
200 return &ipv6_hdr(skb)->daddr;
204 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
208 struct rt6_info *rt = (struct rt6_info *) dst;
211 daddr = choose_neigh_daddr(rt, skb, daddr);
212 n = __ipv6_neigh_lookup(dst->dev, daddr);
215 return neigh_create(&nd_tbl, daddr, dst->dev);
218 static struct dst_ops ip6_dst_ops_template = {
222 .check = ip6_dst_check,
223 .default_advmss = ip6_default_advmss,
225 .cow_metrics = ipv6_cow_metrics,
226 .destroy = ip6_dst_destroy,
227 .ifdown = ip6_dst_ifdown,
228 .negative_advice = ip6_negative_advice,
229 .link_failure = ip6_link_failure,
230 .update_pmtu = ip6_rt_update_pmtu,
231 .redirect = rt6_do_redirect,
232 .local_out = __ip6_local_out,
233 .neigh_lookup = ip6_neigh_lookup,
236 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
238 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
240 return mtu ? : dst->dev->mtu;
243 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
244 struct sk_buff *skb, u32 mtu)
248 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
253 static struct dst_ops ip6_dst_blackhole_ops = {
255 .destroy = ip6_dst_destroy,
256 .check = ip6_dst_check,
257 .mtu = ip6_blackhole_mtu,
258 .default_advmss = ip6_default_advmss,
259 .update_pmtu = ip6_rt_blackhole_update_pmtu,
260 .redirect = ip6_rt_blackhole_redirect,
261 .cow_metrics = dst_cow_metrics_generic,
262 .neigh_lookup = ip6_neigh_lookup,
265 static const u32 ip6_template_metrics[RTAX_MAX] = {
266 [RTAX_HOPLIMIT - 1] = 0,
269 static const struct rt6_info ip6_null_entry_template = {
271 .__refcnt = ATOMIC_INIT(1),
273 .obsolete = DST_OBSOLETE_FORCE_CHK,
274 .error = -ENETUNREACH,
275 .input = ip6_pkt_discard,
276 .output = ip6_pkt_discard_out,
278 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
279 .rt6i_protocol = RTPROT_KERNEL,
280 .rt6i_metric = ~(u32) 0,
281 .rt6i_ref = ATOMIC_INIT(1),
284 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
286 static const struct rt6_info ip6_prohibit_entry_template = {
288 .__refcnt = ATOMIC_INIT(1),
290 .obsolete = DST_OBSOLETE_FORCE_CHK,
292 .input = ip6_pkt_prohibit,
293 .output = ip6_pkt_prohibit_out,
295 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
296 .rt6i_protocol = RTPROT_KERNEL,
297 .rt6i_metric = ~(u32) 0,
298 .rt6i_ref = ATOMIC_INIT(1),
301 static const struct rt6_info ip6_blk_hole_entry_template = {
303 .__refcnt = ATOMIC_INIT(1),
305 .obsolete = DST_OBSOLETE_FORCE_CHK,
307 .input = dst_discard,
308 .output = dst_discard_out,
310 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
311 .rt6i_protocol = RTPROT_KERNEL,
312 .rt6i_metric = ~(u32) 0,
313 .rt6i_ref = ATOMIC_INIT(1),
318 static void rt6_info_init(struct rt6_info *rt)
320 struct dst_entry *dst = &rt->dst;
322 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
323 INIT_LIST_HEAD(&rt->rt6i_siblings);
324 INIT_LIST_HEAD(&rt->rt6i_uncached);
327 /* allocate dst with ip6_dst_ops */
328 static struct rt6_info *__ip6_dst_alloc(struct net *net,
329 struct net_device *dev,
332 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
333 0, DST_OBSOLETE_FORCE_CHK, flags);
341 struct rt6_info *ip6_dst_alloc(struct net *net,
342 struct net_device *dev,
345 struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
348 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
352 for_each_possible_cpu(cpu) {
355 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
356 /* no one shares rt */
360 dst_destroy((struct dst_entry *)rt);
367 EXPORT_SYMBOL(ip6_dst_alloc);
369 static void ip6_dst_destroy(struct dst_entry *dst)
371 struct rt6_info *rt = (struct rt6_info *)dst;
372 struct dst_entry *from = dst->from;
373 struct inet6_dev *idev;
375 dst_destroy_metrics_generic(dst);
376 free_percpu(rt->rt6i_pcpu);
377 rt6_uncached_list_del(rt);
379 idev = rt->rt6i_idev;
381 rt->rt6i_idev = NULL;
389 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
392 struct rt6_info *rt = (struct rt6_info *)dst;
393 struct inet6_dev *idev = rt->rt6i_idev;
394 struct net_device *loopback_dev =
395 dev_net(dev)->loopback_dev;
397 if (dev != loopback_dev) {
398 if (idev && idev->dev == dev) {
399 struct inet6_dev *loopback_idev =
400 in6_dev_get(loopback_dev);
402 rt->rt6i_idev = loopback_idev;
409 static bool __rt6_check_expired(const struct rt6_info *rt)
411 if (rt->rt6i_flags & RTF_EXPIRES)
412 return time_after(jiffies, rt->dst.expires);
417 static bool rt6_check_expired(const struct rt6_info *rt)
419 if (rt->rt6i_flags & RTF_EXPIRES) {
420 if (time_after(jiffies, rt->dst.expires))
422 } else if (rt->dst.from) {
423 return rt6_check_expired((struct rt6_info *) rt->dst.from);
428 /* Multipath route selection:
429 * Hash based function using packet header and flowlabel.
430 * Adapted from fib_info_hashfn()
432 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
433 const struct flowi6 *fl6)
435 return get_hash_from_flowi6(fl6) % candidate_count;
438 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
439 struct flowi6 *fl6, int oif,
442 struct rt6_info *sibling, *next_sibling;
445 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
446 /* Don't change the route, if route_choosen == 0
447 * (siblings does not include ourself)
450 list_for_each_entry_safe(sibling, next_sibling,
451 &match->rt6i_siblings, rt6i_siblings) {
453 if (route_choosen == 0) {
454 if (rt6_score_route(sibling, oif, strict) < 0)
464 * Route lookup. Any table->tb6_lock is implied.
467 static inline struct rt6_info *rt6_device_match(struct net *net,
469 const struct in6_addr *saddr,
473 struct rt6_info *local = NULL;
474 struct rt6_info *sprt;
476 if (!oif && ipv6_addr_any(saddr))
479 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
480 struct net_device *dev = sprt->dst.dev;
483 if (dev->ifindex == oif)
485 if (dev->flags & IFF_LOOPBACK) {
486 if (!sprt->rt6i_idev ||
487 sprt->rt6i_idev->dev->ifindex != oif) {
488 if (flags & RT6_LOOKUP_F_IFACE)
491 local->rt6i_idev->dev->ifindex == oif)
497 if (ipv6_chk_addr(net, saddr, dev,
498 flags & RT6_LOOKUP_F_IFACE))
507 if (flags & RT6_LOOKUP_F_IFACE)
508 return net->ipv6.ip6_null_entry;
514 #ifdef CONFIG_IPV6_ROUTER_PREF
515 struct __rt6_probe_work {
516 struct work_struct work;
517 struct in6_addr target;
518 struct net_device *dev;
521 static void rt6_probe_deferred(struct work_struct *w)
523 struct in6_addr mcaddr;
524 struct __rt6_probe_work *work =
525 container_of(w, struct __rt6_probe_work, work);
527 addrconf_addr_solict_mult(&work->target, &mcaddr);
528 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL);
533 static void rt6_probe(struct rt6_info *rt)
535 struct __rt6_probe_work *work;
536 struct neighbour *neigh;
538 * Okay, this does not seem to be appropriate
539 * for now, however, we need to check if it
540 * is really so; aka Router Reachability Probing.
542 * Router Reachability Probe MUST be rate-limited
543 * to no more than one per minute.
545 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
548 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
550 if (neigh->nud_state & NUD_VALID)
554 write_lock(&neigh->lock);
555 if (!(neigh->nud_state & NUD_VALID) &&
558 rt->rt6i_idev->cnf.rtr_probe_interval)) {
559 work = kmalloc(sizeof(*work), GFP_ATOMIC);
561 __neigh_set_probe_once(neigh);
563 write_unlock(&neigh->lock);
565 work = kmalloc(sizeof(*work), GFP_ATOMIC);
569 INIT_WORK(&work->work, rt6_probe_deferred);
570 work->target = rt->rt6i_gateway;
571 dev_hold(rt->dst.dev);
572 work->dev = rt->dst.dev;
573 schedule_work(&work->work);
577 rcu_read_unlock_bh();
580 static inline void rt6_probe(struct rt6_info *rt)
586 * Default Router Selection (RFC 2461 6.3.6)
588 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
590 struct net_device *dev = rt->dst.dev;
591 if (!oif || dev->ifindex == oif)
593 if ((dev->flags & IFF_LOOPBACK) &&
594 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
599 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
601 struct neighbour *neigh;
602 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
604 if (rt->rt6i_flags & RTF_NONEXTHOP ||
605 !(rt->rt6i_flags & RTF_GATEWAY))
606 return RT6_NUD_SUCCEED;
609 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
611 read_lock(&neigh->lock);
612 if (neigh->nud_state & NUD_VALID)
613 ret = RT6_NUD_SUCCEED;
614 #ifdef CONFIG_IPV6_ROUTER_PREF
615 else if (!(neigh->nud_state & NUD_FAILED))
616 ret = RT6_NUD_SUCCEED;
618 ret = RT6_NUD_FAIL_PROBE;
620 read_unlock(&neigh->lock);
622 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
623 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
625 rcu_read_unlock_bh();
630 static int rt6_score_route(struct rt6_info *rt, int oif,
635 m = rt6_check_dev(rt, oif);
636 if (!m && (strict & RT6_LOOKUP_F_IFACE))
637 return RT6_NUD_FAIL_HARD;
638 #ifdef CONFIG_IPV6_ROUTER_PREF
639 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
641 if (strict & RT6_LOOKUP_F_REACHABLE) {
642 int n = rt6_check_neigh(rt);
649 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
650 int *mpri, struct rt6_info *match,
654 bool match_do_rr = false;
655 struct inet6_dev *idev = rt->rt6i_idev;
656 struct net_device *dev = rt->dst.dev;
658 if (dev && !netif_carrier_ok(dev) &&
659 idev->cnf.ignore_routes_with_linkdown)
662 if (rt6_check_expired(rt))
665 m = rt6_score_route(rt, oif, strict);
666 if (m == RT6_NUD_FAIL_DO_RR) {
668 m = 0; /* lowest valid score */
669 } else if (m == RT6_NUD_FAIL_HARD) {
673 if (strict & RT6_LOOKUP_F_REACHABLE)
676 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
678 *do_rr = match_do_rr;
686 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
687 struct rt6_info *rr_head,
688 u32 metric, int oif, int strict,
691 struct rt6_info *rt, *match, *cont;
696 for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
697 if (rt->rt6i_metric != metric) {
702 match = find_match(rt, oif, strict, &mpri, match, do_rr);
705 for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
706 if (rt->rt6i_metric != metric) {
711 match = find_match(rt, oif, strict, &mpri, match, do_rr);
717 for (rt = cont; rt; rt = rt->dst.rt6_next)
718 match = find_match(rt, oif, strict, &mpri, match, do_rr);
723 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
725 struct rt6_info *match, *rt0;
731 fn->rr_ptr = rt0 = fn->leaf;
733 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
737 struct rt6_info *next = rt0->dst.rt6_next;
739 /* no entries matched; do round-robin */
740 if (!next || next->rt6i_metric != rt0->rt6i_metric)
747 net = dev_net(rt0->dst.dev);
748 return match ? match : net->ipv6.ip6_null_entry;
751 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
753 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
756 #ifdef CONFIG_IPV6_ROUTE_INFO
757 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
758 const struct in6_addr *gwaddr)
760 struct net *net = dev_net(dev);
761 struct route_info *rinfo = (struct route_info *) opt;
762 struct in6_addr prefix_buf, *prefix;
764 unsigned long lifetime;
767 if (len < sizeof(struct route_info)) {
771 /* Sanity check for prefix_len and length */
772 if (rinfo->length > 3) {
774 } else if (rinfo->prefix_len > 128) {
776 } else if (rinfo->prefix_len > 64) {
777 if (rinfo->length < 2) {
780 } else if (rinfo->prefix_len > 0) {
781 if (rinfo->length < 1) {
786 pref = rinfo->route_pref;
787 if (pref == ICMPV6_ROUTER_PREF_INVALID)
790 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
792 if (rinfo->length == 3)
793 prefix = (struct in6_addr *)rinfo->prefix;
795 /* this function is safe */
796 ipv6_addr_prefix(&prefix_buf,
797 (struct in6_addr *)rinfo->prefix,
799 prefix = &prefix_buf;
802 if (rinfo->prefix_len == 0)
803 rt = rt6_get_dflt_router(gwaddr, dev);
805 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
806 gwaddr, dev->ifindex);
808 if (rt && !lifetime) {
814 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
817 rt->rt6i_flags = RTF_ROUTEINFO |
818 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
821 if (!addrconf_finite_timeout(lifetime))
822 rt6_clean_expires(rt);
824 rt6_set_expires(rt, jiffies + HZ * lifetime);
832 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
833 struct in6_addr *saddr)
835 struct fib6_node *pn;
837 if (fn->fn_flags & RTN_TL_ROOT)
840 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
841 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
844 if (fn->fn_flags & RTN_RTINFO)
849 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
850 struct fib6_table *table,
851 struct flowi6 *fl6, int flags)
853 struct fib6_node *fn;
856 read_lock_bh(&table->tb6_lock);
857 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
860 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
861 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
862 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
863 if (rt == net->ipv6.ip6_null_entry) {
864 fn = fib6_backtrack(fn, &fl6->saddr);
868 dst_use(&rt->dst, jiffies);
869 read_unlock_bh(&table->tb6_lock);
871 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
877 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
880 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
882 EXPORT_SYMBOL_GPL(ip6_route_lookup);
884 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
885 const struct in6_addr *saddr, int oif, int strict)
887 struct flowi6 fl6 = {
891 struct dst_entry *dst;
892 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
895 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
896 flags |= RT6_LOOKUP_F_HAS_SADDR;
899 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
901 return (struct rt6_info *) dst;
907 EXPORT_SYMBOL(rt6_lookup);
909 /* ip6_ins_rt is called with FREE table->tb6_lock.
910 It takes new route entry, the addition fails by any reason the
911 route is freed. In any case, if caller does not hold it, it may
915 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
916 struct mx6_config *mxc)
919 struct fib6_table *table;
921 table = rt->rt6i_table;
922 write_lock_bh(&table->tb6_lock);
923 err = fib6_add(&table->tb6_root, rt, info, mxc);
924 write_unlock_bh(&table->tb6_lock);
929 int ip6_ins_rt(struct rt6_info *rt)
931 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
932 struct mx6_config mxc = { .mx = NULL, };
934 return __ip6_ins_rt(rt, &info, &mxc);
937 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
938 const struct in6_addr *daddr,
939 const struct in6_addr *saddr)
947 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
948 ort = (struct rt6_info *)ort->dst.from;
950 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
955 ip6_rt_copy_init(rt, ort);
956 rt->rt6i_flags |= RTF_CACHE;
958 rt->dst.flags |= DST_HOST;
959 rt->rt6i_dst.addr = *daddr;
960 rt->rt6i_dst.plen = 128;
962 if (!rt6_is_gw_or_nonexthop(ort)) {
963 if (ort->rt6i_dst.plen != 128 &&
964 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
965 rt->rt6i_flags |= RTF_ANYCAST;
966 #ifdef CONFIG_IPV6_SUBTREES
967 if (rt->rt6i_src.plen && saddr) {
968 rt->rt6i_src.addr = *saddr;
969 rt->rt6i_src.plen = 128;
977 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
979 struct rt6_info *pcpu_rt;
981 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
982 rt->dst.dev, rt->dst.flags);
986 ip6_rt_copy_init(pcpu_rt, rt);
987 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
988 pcpu_rt->rt6i_flags |= RTF_PCPU;
992 /* It should be called with read_lock_bh(&tb6_lock) acquired */
993 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
995 struct rt6_info *pcpu_rt, **p;
997 p = this_cpu_ptr(rt->rt6i_pcpu);
1001 dst_hold(&pcpu_rt->dst);
1002 rt6_dst_from_metrics_check(pcpu_rt);
1007 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1009 struct fib6_table *table = rt->rt6i_table;
1010 struct rt6_info *pcpu_rt, *prev, **p;
1012 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1014 struct net *net = dev_net(rt->dst.dev);
1016 dst_hold(&net->ipv6.ip6_null_entry->dst);
1017 return net->ipv6.ip6_null_entry;
1020 read_lock_bh(&table->tb6_lock);
1021 if (rt->rt6i_pcpu) {
1022 p = this_cpu_ptr(rt->rt6i_pcpu);
1023 prev = cmpxchg(p, NULL, pcpu_rt);
1025 /* If someone did it before us, return prev instead */
1026 dst_destroy(&pcpu_rt->dst);
1030 /* rt has been removed from the fib6 tree
1031 * before we have a chance to acquire the read_lock.
1032 * In this case, don't brother to create a pcpu rt
1033 * since rt is going away anyway. The next
1034 * dst_check() will trigger a re-lookup.
1036 dst_destroy(&pcpu_rt->dst);
1039 dst_hold(&pcpu_rt->dst);
1040 rt6_dst_from_metrics_check(pcpu_rt);
1041 read_unlock_bh(&table->tb6_lock);
1045 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1046 struct flowi6 *fl6, int flags)
1048 struct fib6_node *fn, *saved_fn;
1049 struct rt6_info *rt;
1052 strict |= flags & RT6_LOOKUP_F_IFACE;
1053 if (net->ipv6.devconf_all->forwarding == 0)
1054 strict |= RT6_LOOKUP_F_REACHABLE;
1056 read_lock_bh(&table->tb6_lock);
1058 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1061 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1065 rt = rt6_select(fn, oif, strict);
1066 if (rt->rt6i_nsiblings)
1067 rt = rt6_multipath_select(rt, fl6, oif, strict);
1068 if (rt == net->ipv6.ip6_null_entry) {
1069 fn = fib6_backtrack(fn, &fl6->saddr);
1071 goto redo_rt6_select;
1072 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1073 /* also consider unreachable route */
1074 strict &= ~RT6_LOOKUP_F_REACHABLE;
1076 goto redo_rt6_select;
1081 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1082 dst_use(&rt->dst, jiffies);
1083 read_unlock_bh(&table->tb6_lock);
1085 rt6_dst_from_metrics_check(rt);
1087 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1089 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1090 !(rt->rt6i_flags & RTF_GATEWAY))) {
1091 /* Create a RTF_CACHE clone which will not be
1092 * owned by the fib6 tree. It is for the special case where
1093 * the daddr in the skb during the neighbor look-up is different
1094 * from the fl6->daddr used to look-up route here.
1097 struct rt6_info *uncached_rt;
1099 dst_use(&rt->dst, jiffies);
1100 read_unlock_bh(&table->tb6_lock);
1102 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1103 dst_release(&rt->dst);
1106 rt6_uncached_list_add(uncached_rt);
1108 uncached_rt = net->ipv6.ip6_null_entry;
1110 dst_hold(&uncached_rt->dst);
1112 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1116 /* Get a percpu copy */
1118 struct rt6_info *pcpu_rt;
1120 rt->dst.lastuse = jiffies;
1122 pcpu_rt = rt6_get_pcpu_route(rt);
1125 read_unlock_bh(&table->tb6_lock);
1127 /* We have to do the read_unlock first
1128 * because rt6_make_pcpu_route() may trigger
1129 * ip6_dst_gc() which will take the write_lock.
1132 read_unlock_bh(&table->tb6_lock);
1133 pcpu_rt = rt6_make_pcpu_route(rt);
1134 dst_release(&rt->dst);
1137 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1143 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1144 struct flowi6 *fl6, int flags)
1146 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1149 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1150 struct net_device *dev,
1151 struct flowi6 *fl6, int flags)
1153 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1154 flags |= RT6_LOOKUP_F_IFACE;
1156 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1159 void ip6_route_input(struct sk_buff *skb)
1161 const struct ipv6hdr *iph = ipv6_hdr(skb);
1162 struct net *net = dev_net(skb->dev);
1163 int flags = RT6_LOOKUP_F_HAS_SADDR;
1164 struct ip_tunnel_info *tun_info;
1165 struct flowi6 fl6 = {
1166 .flowi6_iif = l3mdev_fib_oif(skb->dev),
1167 .daddr = iph->daddr,
1168 .saddr = iph->saddr,
1169 .flowlabel = ip6_flowinfo(iph),
1170 .flowi6_mark = skb->mark,
1171 .flowi6_proto = iph->nexthdr,
1174 tun_info = skb_tunnel_info(skb);
1175 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1176 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1178 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1181 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1182 struct flowi6 *fl6, int flags)
1184 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1187 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1188 struct flowi6 *fl6, int flags)
1190 struct dst_entry *dst;
1193 dst = l3mdev_rt6_dst_by_oif(net, fl6);
1197 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1199 any_src = ipv6_addr_any(&fl6->saddr);
1200 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1201 (fl6->flowi6_oif && any_src))
1202 flags |= RT6_LOOKUP_F_IFACE;
1205 flags |= RT6_LOOKUP_F_HAS_SADDR;
1207 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1209 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1211 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1213 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1215 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1216 struct dst_entry *new = NULL;
1218 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1224 new->input = dst_discard;
1225 new->output = dst_discard_out;
1227 dst_copy_metrics(new, &ort->dst);
1228 rt->rt6i_idev = ort->rt6i_idev;
1230 in6_dev_hold(rt->rt6i_idev);
1232 rt->rt6i_gateway = ort->rt6i_gateway;
1233 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1234 rt->rt6i_metric = 0;
1236 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1237 #ifdef CONFIG_IPV6_SUBTREES
1238 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1244 dst_release(dst_orig);
1245 return new ? new : ERR_PTR(-ENOMEM);
1249 * Destination cache support functions
1252 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1255 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1256 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1259 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1261 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1264 if (rt6_check_expired(rt))
1270 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1272 if (!__rt6_check_expired(rt) &&
1273 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1274 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1280 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1282 struct rt6_info *rt;
1284 rt = (struct rt6_info *) dst;
1286 /* All IPV6 dsts are created with ->obsolete set to the value
1287 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1288 * into this function always.
1291 rt6_dst_from_metrics_check(rt);
1293 if (rt->rt6i_flags & RTF_PCPU ||
1294 (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1295 return rt6_dst_from_check(rt, cookie);
1297 return rt6_check(rt, cookie);
1300 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1302 struct rt6_info *rt = (struct rt6_info *) dst;
1305 if (rt->rt6i_flags & RTF_CACHE) {
1306 if (rt6_check_expired(rt)) {
1318 static void ip6_link_failure(struct sk_buff *skb)
1320 struct rt6_info *rt;
1322 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1324 rt = (struct rt6_info *) skb_dst(skb);
1326 if (rt->rt6i_flags & RTF_CACHE) {
1329 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1330 rt->rt6i_node->fn_sernum = -1;
1335 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1337 struct net *net = dev_net(rt->dst.dev);
1339 rt->rt6i_flags |= RTF_MODIFIED;
1340 rt->rt6i_pmtu = mtu;
1341 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1344 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1346 return !(rt->rt6i_flags & RTF_CACHE) &&
1347 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1350 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1351 const struct ipv6hdr *iph, u32 mtu)
1353 struct rt6_info *rt6 = (struct rt6_info *)dst;
1355 if (rt6->rt6i_flags & RTF_LOCAL)
1359 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1360 if (mtu >= dst_mtu(dst))
1363 if (!rt6_cache_allowed_for_pmtu(rt6)) {
1364 rt6_do_update_pmtu(rt6, mtu);
1366 const struct in6_addr *daddr, *saddr;
1367 struct rt6_info *nrt6;
1370 daddr = &iph->daddr;
1371 saddr = &iph->saddr;
1373 daddr = &sk->sk_v6_daddr;
1374 saddr = &inet6_sk(sk)->saddr;
1378 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1380 rt6_do_update_pmtu(nrt6, mtu);
1382 /* ip6_ins_rt(nrt6) will bump the
1383 * rt6->rt6i_node->fn_sernum
1384 * which will fail the next rt6_check() and
1385 * invalidate the sk->sk_dst_cache.
1392 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1393 struct sk_buff *skb, u32 mtu)
1395 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1398 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1401 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1402 struct dst_entry *dst;
1405 memset(&fl6, 0, sizeof(fl6));
1406 fl6.flowi6_oif = oif;
1407 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1408 fl6.daddr = iph->daddr;
1409 fl6.saddr = iph->saddr;
1410 fl6.flowlabel = ip6_flowinfo(iph);
1412 dst = ip6_route_output(net, NULL, &fl6);
1414 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1417 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1419 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1421 struct dst_entry *dst;
1423 ip6_update_pmtu(skb, sock_net(sk), mtu,
1424 sk->sk_bound_dev_if, sk->sk_mark);
1426 dst = __sk_dst_get(sk);
1427 if (!dst || !dst->obsolete ||
1428 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1432 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1433 ip6_datagram_dst_update(sk, false);
1436 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1438 /* Handle redirects */
1439 struct ip6rd_flowi {
1441 struct in6_addr gateway;
1444 static struct rt6_info *__ip6_route_redirect(struct net *net,
1445 struct fib6_table *table,
1449 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1450 struct rt6_info *rt;
1451 struct fib6_node *fn;
1453 /* Get the "current" route for this destination and
1454 * check if the redirect has come from approriate router.
1456 * RFC 4861 specifies that redirects should only be
1457 * accepted if they come from the nexthop to the target.
1458 * Due to the way the routes are chosen, this notion
1459 * is a bit fuzzy and one might need to check all possible
1463 read_lock_bh(&table->tb6_lock);
1464 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1466 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1467 if (rt6_check_expired(rt))
1471 if (!(rt->rt6i_flags & RTF_GATEWAY))
1473 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1475 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1481 rt = net->ipv6.ip6_null_entry;
1482 else if (rt->dst.error) {
1483 rt = net->ipv6.ip6_null_entry;
1487 if (rt == net->ipv6.ip6_null_entry) {
1488 fn = fib6_backtrack(fn, &fl6->saddr);
1496 read_unlock_bh(&table->tb6_lock);
1498 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1502 static struct dst_entry *ip6_route_redirect(struct net *net,
1503 const struct flowi6 *fl6,
1504 const struct in6_addr *gateway)
1506 int flags = RT6_LOOKUP_F_HAS_SADDR;
1507 struct ip6rd_flowi rdfl;
1510 rdfl.gateway = *gateway;
1512 return fib6_rule_lookup(net, &rdfl.fl6,
1513 flags, __ip6_route_redirect);
1516 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1518 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1519 struct dst_entry *dst;
1522 memset(&fl6, 0, sizeof(fl6));
1523 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1524 fl6.flowi6_oif = oif;
1525 fl6.flowi6_mark = mark;
1526 fl6.daddr = iph->daddr;
1527 fl6.saddr = iph->saddr;
1528 fl6.flowlabel = ip6_flowinfo(iph);
1530 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1531 rt6_do_redirect(dst, NULL, skb);
1534 EXPORT_SYMBOL_GPL(ip6_redirect);
1536 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1539 const struct ipv6hdr *iph = ipv6_hdr(skb);
1540 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1541 struct dst_entry *dst;
1544 memset(&fl6, 0, sizeof(fl6));
1545 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1546 fl6.flowi6_oif = oif;
1547 fl6.flowi6_mark = mark;
1548 fl6.daddr = msg->dest;
1549 fl6.saddr = iph->daddr;
1551 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1552 rt6_do_redirect(dst, NULL, skb);
1556 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1558 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1560 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1562 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1564 struct net_device *dev = dst->dev;
1565 unsigned int mtu = dst_mtu(dst);
1566 struct net *net = dev_net(dev);
1568 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1570 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1571 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1574 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1575 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1576 * IPV6_MAXPLEN is also valid and means: "any MSS,
1577 * rely only on pmtu discovery"
1579 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1584 static unsigned int ip6_mtu(const struct dst_entry *dst)
1586 const struct rt6_info *rt = (const struct rt6_info *)dst;
1587 unsigned int mtu = rt->rt6i_pmtu;
1588 struct inet6_dev *idev;
1593 mtu = dst_metric_raw(dst, RTAX_MTU);
1600 idev = __in6_dev_get(dst->dev);
1602 mtu = idev->cnf.mtu6;
1606 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1609 static struct dst_entry *icmp6_dst_gc_list;
1610 static DEFINE_SPINLOCK(icmp6_dst_lock);
1612 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1615 struct dst_entry *dst;
1616 struct rt6_info *rt;
1617 struct inet6_dev *idev = in6_dev_get(dev);
1618 struct net *net = dev_net(dev);
1620 if (unlikely(!idev))
1621 return ERR_PTR(-ENODEV);
1623 rt = ip6_dst_alloc(net, dev, 0);
1624 if (unlikely(!rt)) {
1626 dst = ERR_PTR(-ENOMEM);
1630 rt->dst.flags |= DST_HOST;
1631 rt->dst.output = ip6_output;
1632 atomic_set(&rt->dst.__refcnt, 1);
1633 rt->rt6i_gateway = fl6->daddr;
1634 rt->rt6i_dst.addr = fl6->daddr;
1635 rt->rt6i_dst.plen = 128;
1636 rt->rt6i_idev = idev;
1637 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1639 spin_lock_bh(&icmp6_dst_lock);
1640 rt->dst.next = icmp6_dst_gc_list;
1641 icmp6_dst_gc_list = &rt->dst;
1642 spin_unlock_bh(&icmp6_dst_lock);
1644 fib6_force_start_gc(net);
1646 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1652 int icmp6_dst_gc(void)
1654 struct dst_entry *dst, **pprev;
1657 spin_lock_bh(&icmp6_dst_lock);
1658 pprev = &icmp6_dst_gc_list;
1660 while ((dst = *pprev) != NULL) {
1661 if (!atomic_read(&dst->__refcnt)) {
1670 spin_unlock_bh(&icmp6_dst_lock);
1675 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1678 struct dst_entry *dst, **pprev;
1680 spin_lock_bh(&icmp6_dst_lock);
1681 pprev = &icmp6_dst_gc_list;
1682 while ((dst = *pprev) != NULL) {
1683 struct rt6_info *rt = (struct rt6_info *) dst;
1684 if (func(rt, arg)) {
1691 spin_unlock_bh(&icmp6_dst_lock);
1694 static int ip6_dst_gc(struct dst_ops *ops)
1696 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1697 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1698 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1699 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1700 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1701 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1704 entries = dst_entries_get_fast(ops);
1705 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1706 entries <= rt_max_size)
1709 net->ipv6.ip6_rt_gc_expire++;
1710 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1711 entries = dst_entries_get_slow(ops);
1712 if (entries < ops->gc_thresh)
1713 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1715 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1716 return entries > rt_max_size;
1719 static int ip6_convert_metrics(struct mx6_config *mxc,
1720 const struct fib6_config *cfg)
1722 bool ecn_ca = false;
1730 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1734 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1735 int type = nla_type(nla);
1740 if (unlikely(type > RTAX_MAX))
1743 if (type == RTAX_CC_ALGO) {
1744 char tmp[TCP_CA_NAME_MAX];
1746 nla_strlcpy(tmp, nla, sizeof(tmp));
1747 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1748 if (val == TCP_CA_UNSPEC)
1751 val = nla_get_u32(nla);
1753 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1757 __set_bit(type - 1, mxc->mx_valid);
1761 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1762 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1772 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
1773 struct fib6_config *cfg,
1774 const struct in6_addr *gw_addr)
1776 struct flowi6 fl6 = {
1777 .flowi6_oif = cfg->fc_ifindex,
1779 .saddr = cfg->fc_prefsrc,
1781 struct fib6_table *table;
1782 struct rt6_info *rt;
1785 table = fib6_get_table(net, cfg->fc_table);
1789 if (!ipv6_addr_any(&cfg->fc_prefsrc))
1790 flags |= RT6_LOOKUP_F_HAS_SADDR;
1792 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
1794 /* if table lookup failed, fall back to full lookup */
1795 if (rt == net->ipv6.ip6_null_entry) {
1803 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1805 struct net *net = cfg->fc_nlinfo.nl_net;
1806 struct rt6_info *rt = NULL;
1807 struct net_device *dev = NULL;
1808 struct inet6_dev *idev = NULL;
1809 struct fib6_table *table;
1813 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1815 #ifndef CONFIG_IPV6_SUBTREES
1816 if (cfg->fc_src_len)
1819 if (cfg->fc_ifindex) {
1821 dev = dev_get_by_index(net, cfg->fc_ifindex);
1824 idev = in6_dev_get(dev);
1829 if (cfg->fc_metric == 0)
1830 cfg->fc_metric = IP6_RT_PRIO_USER;
1833 if (cfg->fc_nlinfo.nlh &&
1834 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1835 table = fib6_get_table(net, cfg->fc_table);
1837 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1838 table = fib6_new_table(net, cfg->fc_table);
1841 table = fib6_new_table(net, cfg->fc_table);
1847 rt = ip6_dst_alloc(net, NULL,
1848 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1855 if (cfg->fc_flags & RTF_EXPIRES)
1856 rt6_set_expires(rt, jiffies +
1857 clock_t_to_jiffies(cfg->fc_expires));
1859 rt6_clean_expires(rt);
1861 if (cfg->fc_protocol == RTPROT_UNSPEC)
1862 cfg->fc_protocol = RTPROT_BOOT;
1863 rt->rt6i_protocol = cfg->fc_protocol;
1865 addr_type = ipv6_addr_type(&cfg->fc_dst);
1867 if (addr_type & IPV6_ADDR_MULTICAST)
1868 rt->dst.input = ip6_mc_input;
1869 else if (cfg->fc_flags & RTF_LOCAL)
1870 rt->dst.input = ip6_input;
1872 rt->dst.input = ip6_forward;
1874 rt->dst.output = ip6_output;
1876 if (cfg->fc_encap) {
1877 struct lwtunnel_state *lwtstate;
1879 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1880 cfg->fc_encap, AF_INET6, cfg,
1884 rt->dst.lwtstate = lwtstate_get(lwtstate);
1885 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1886 rt->dst.lwtstate->orig_output = rt->dst.output;
1887 rt->dst.output = lwtunnel_output;
1889 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1890 rt->dst.lwtstate->orig_input = rt->dst.input;
1891 rt->dst.input = lwtunnel_input;
1895 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1896 rt->rt6i_dst.plen = cfg->fc_dst_len;
1897 if (rt->rt6i_dst.plen == 128)
1898 rt->dst.flags |= DST_HOST;
1900 #ifdef CONFIG_IPV6_SUBTREES
1901 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1902 rt->rt6i_src.plen = cfg->fc_src_len;
1905 rt->rt6i_metric = cfg->fc_metric;
1907 /* We cannot add true routes via loopback here,
1908 they would result in kernel looping; promote them to reject routes
1910 if ((cfg->fc_flags & RTF_REJECT) ||
1911 (dev && (dev->flags & IFF_LOOPBACK) &&
1912 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1913 !(cfg->fc_flags & RTF_LOCAL))) {
1914 /* hold loopback dev/idev if we haven't done so. */
1915 if (dev != net->loopback_dev) {
1920 dev = net->loopback_dev;
1922 idev = in6_dev_get(dev);
1928 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1929 switch (cfg->fc_type) {
1931 rt->dst.error = -EINVAL;
1932 rt->dst.output = dst_discard_out;
1933 rt->dst.input = dst_discard;
1936 rt->dst.error = -EACCES;
1937 rt->dst.output = ip6_pkt_prohibit_out;
1938 rt->dst.input = ip6_pkt_prohibit;
1941 case RTN_UNREACHABLE:
1943 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1944 : (cfg->fc_type == RTN_UNREACHABLE)
1945 ? -EHOSTUNREACH : -ENETUNREACH;
1946 rt->dst.output = ip6_pkt_discard_out;
1947 rt->dst.input = ip6_pkt_discard;
1953 if (cfg->fc_flags & RTF_GATEWAY) {
1954 const struct in6_addr *gw_addr;
1957 gw_addr = &cfg->fc_gateway;
1958 gwa_type = ipv6_addr_type(gw_addr);
1960 /* if gw_addr is local we will fail to detect this in case
1961 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1962 * will return already-added prefix route via interface that
1963 * prefix route was assigned to, which might be non-loopback.
1966 if (ipv6_chk_addr_and_flags(net, gw_addr,
1967 gwa_type & IPV6_ADDR_LINKLOCAL ?
1971 rt->rt6i_gateway = *gw_addr;
1973 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1974 struct rt6_info *grt = NULL;
1976 /* IPv6 strictly inhibits using not link-local
1977 addresses as nexthop address.
1978 Otherwise, router will not able to send redirects.
1979 It is very good, but in some (rare!) circumstances
1980 (SIT, PtP, NBMA NOARP links) it is handy to allow
1981 some exceptions. --ANK
1983 if (!(gwa_type & IPV6_ADDR_UNICAST))
1987 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
1990 grt = rt6_lookup(net, gw_addr, NULL,
1991 cfg->fc_ifindex, 1);
1993 err = -EHOSTUNREACH;
1997 if (dev != grt->dst.dev) {
2003 idev = grt->rt6i_idev;
2005 in6_dev_hold(grt->rt6i_idev);
2007 if (!(grt->rt6i_flags & RTF_GATEWAY))
2015 if (!dev || (dev->flags & IFF_LOOPBACK))
2023 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2024 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2028 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2029 rt->rt6i_prefsrc.plen = 128;
2031 rt->rt6i_prefsrc.plen = 0;
2033 rt->rt6i_flags = cfg->fc_flags;
2037 rt->rt6i_idev = idev;
2038 rt->rt6i_table = table;
2040 cfg->fc_nlinfo.nl_net = dev_net(dev);
2051 return ERR_PTR(err);
2054 int ip6_route_add(struct fib6_config *cfg)
2056 struct mx6_config mxc = { .mx = NULL, };
2057 struct rt6_info *rt;
2060 rt = ip6_route_info_create(cfg);
2067 err = ip6_convert_metrics(&mxc, cfg);
2071 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2083 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2086 struct fib6_table *table;
2087 struct net *net = dev_net(rt->dst.dev);
2089 if (rt == net->ipv6.ip6_null_entry ||
2090 rt->dst.flags & DST_NOCACHE) {
2095 table = rt->rt6i_table;
2096 write_lock_bh(&table->tb6_lock);
2097 err = fib6_del(rt, info);
2098 write_unlock_bh(&table->tb6_lock);
2105 int ip6_del_rt(struct rt6_info *rt)
2107 struct nl_info info = {
2108 .nl_net = dev_net(rt->dst.dev),
2110 return __ip6_del_rt(rt, &info);
2113 static int ip6_route_del(struct fib6_config *cfg)
2115 struct fib6_table *table;
2116 struct fib6_node *fn;
2117 struct rt6_info *rt;
2120 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2124 read_lock_bh(&table->tb6_lock);
2126 fn = fib6_locate(&table->tb6_root,
2127 &cfg->fc_dst, cfg->fc_dst_len,
2128 &cfg->fc_src, cfg->fc_src_len);
2131 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2132 if ((rt->rt6i_flags & RTF_CACHE) &&
2133 !(cfg->fc_flags & RTF_CACHE))
2135 if (cfg->fc_ifindex &&
2137 rt->dst.dev->ifindex != cfg->fc_ifindex))
2139 if (cfg->fc_flags & RTF_GATEWAY &&
2140 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2142 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2145 read_unlock_bh(&table->tb6_lock);
2147 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2150 read_unlock_bh(&table->tb6_lock);
2155 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2157 struct netevent_redirect netevent;
2158 struct rt6_info *rt, *nrt = NULL;
2159 struct ndisc_options ndopts;
2160 struct inet6_dev *in6_dev;
2161 struct neighbour *neigh;
2163 int optlen, on_link;
2166 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2167 optlen -= sizeof(*msg);
2170 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2174 msg = (struct rd_msg *)icmp6_hdr(skb);
2176 if (ipv6_addr_is_multicast(&msg->dest)) {
2177 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2182 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2184 } else if (ipv6_addr_type(&msg->target) !=
2185 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2186 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2190 in6_dev = __in6_dev_get(skb->dev);
2193 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2197 * The IP source address of the Redirect MUST be the same as the current
2198 * first-hop router for the specified ICMP Destination Address.
2201 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2202 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2207 if (ndopts.nd_opts_tgt_lladdr) {
2208 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2211 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2216 rt = (struct rt6_info *) dst;
2217 if (rt->rt6i_flags & RTF_REJECT) {
2218 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2222 /* Redirect received -> path was valid.
2223 * Look, redirects are sent only in response to data packets,
2224 * so that this nexthop apparently is reachable. --ANK
2226 dst_confirm(&rt->dst);
2228 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2233 * We have finally decided to accept it.
2236 neigh_update(neigh, lladdr, NUD_STALE,
2237 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2238 NEIGH_UPDATE_F_OVERRIDE|
2239 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2240 NEIGH_UPDATE_F_ISROUTER))
2243 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2247 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2249 nrt->rt6i_flags &= ~RTF_GATEWAY;
2251 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2253 if (ip6_ins_rt(nrt))
2256 netevent.old = &rt->dst;
2257 netevent.new = &nrt->dst;
2258 netevent.daddr = &msg->dest;
2259 netevent.neigh = neigh;
2260 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2262 if (rt->rt6i_flags & RTF_CACHE) {
2263 rt = (struct rt6_info *) dst_clone(&rt->dst);
2268 neigh_release(neigh);
2272 * Misc support functions
2275 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2277 BUG_ON(from->dst.from);
2279 rt->rt6i_flags &= ~RTF_EXPIRES;
2280 dst_hold(&from->dst);
2281 rt->dst.from = &from->dst;
2282 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2285 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2287 rt->dst.input = ort->dst.input;
2288 rt->dst.output = ort->dst.output;
2289 rt->rt6i_dst = ort->rt6i_dst;
2290 rt->dst.error = ort->dst.error;
2291 rt->rt6i_idev = ort->rt6i_idev;
2293 in6_dev_hold(rt->rt6i_idev);
2294 rt->dst.lastuse = jiffies;
2295 rt->rt6i_gateway = ort->rt6i_gateway;
2296 rt->rt6i_flags = ort->rt6i_flags;
2297 rt6_set_from(rt, ort);
2298 rt->rt6i_metric = ort->rt6i_metric;
2299 #ifdef CONFIG_IPV6_SUBTREES
2300 rt->rt6i_src = ort->rt6i_src;
2302 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2303 rt->rt6i_table = ort->rt6i_table;
2304 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2307 #ifdef CONFIG_IPV6_ROUTE_INFO
2308 static struct rt6_info *rt6_get_route_info(struct net *net,
2309 const struct in6_addr *prefix, int prefixlen,
2310 const struct in6_addr *gwaddr, int ifindex)
2312 struct fib6_node *fn;
2313 struct rt6_info *rt = NULL;
2314 struct fib6_table *table;
2316 table = fib6_get_table(net, RT6_TABLE_INFO);
2320 read_lock_bh(&table->tb6_lock);
2321 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2325 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2326 if (rt->dst.dev->ifindex != ifindex)
2328 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2330 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2336 read_unlock_bh(&table->tb6_lock);
2340 static struct rt6_info *rt6_add_route_info(struct net *net,
2341 const struct in6_addr *prefix, int prefixlen,
2342 const struct in6_addr *gwaddr, int ifindex,
2345 struct fib6_config cfg = {
2346 .fc_metric = IP6_RT_PRIO_USER,
2347 .fc_ifindex = ifindex,
2348 .fc_dst_len = prefixlen,
2349 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2350 RTF_UP | RTF_PREF(pref),
2351 .fc_nlinfo.portid = 0,
2352 .fc_nlinfo.nlh = NULL,
2353 .fc_nlinfo.nl_net = net,
2356 cfg.fc_table = l3mdev_fib_table_by_index(net, ifindex) ? : RT6_TABLE_INFO;
2357 cfg.fc_dst = *prefix;
2358 cfg.fc_gateway = *gwaddr;
2360 /* We should treat it as a default route if prefix length is 0. */
2362 cfg.fc_flags |= RTF_DEFAULT;
2364 ip6_route_add(&cfg);
2366 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2370 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2372 struct rt6_info *rt;
2373 struct fib6_table *table;
2375 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2379 read_lock_bh(&table->tb6_lock);
2380 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2381 if (dev == rt->dst.dev &&
2382 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2383 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2388 read_unlock_bh(&table->tb6_lock);
2392 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2393 struct net_device *dev,
2396 struct fib6_config cfg = {
2397 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2398 .fc_metric = IP6_RT_PRIO_USER,
2399 .fc_ifindex = dev->ifindex,
2400 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2401 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2402 .fc_nlinfo.portid = 0,
2403 .fc_nlinfo.nlh = NULL,
2404 .fc_nlinfo.nl_net = dev_net(dev),
2407 cfg.fc_gateway = *gwaddr;
2409 ip6_route_add(&cfg);
2411 return rt6_get_dflt_router(gwaddr, dev);
2414 void rt6_purge_dflt_routers(struct net *net)
2416 struct rt6_info *rt;
2417 struct fib6_table *table;
2419 /* NOTE: Keep consistent with rt6_get_dflt_router */
2420 table = fib6_get_table(net, RT6_TABLE_DFLT);
2425 read_lock_bh(&table->tb6_lock);
2426 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2427 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2428 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2430 read_unlock_bh(&table->tb6_lock);
2435 read_unlock_bh(&table->tb6_lock);
2438 static void rtmsg_to_fib6_config(struct net *net,
2439 struct in6_rtmsg *rtmsg,
2440 struct fib6_config *cfg)
2442 memset(cfg, 0, sizeof(*cfg));
2444 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2446 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2447 cfg->fc_metric = rtmsg->rtmsg_metric;
2448 cfg->fc_expires = rtmsg->rtmsg_info;
2449 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2450 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2451 cfg->fc_flags = rtmsg->rtmsg_flags;
2453 cfg->fc_nlinfo.nl_net = net;
2455 cfg->fc_dst = rtmsg->rtmsg_dst;
2456 cfg->fc_src = rtmsg->rtmsg_src;
2457 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2460 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2462 struct fib6_config cfg;
2463 struct in6_rtmsg rtmsg;
2467 case SIOCADDRT: /* Add a route */
2468 case SIOCDELRT: /* Delete a route */
2469 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2471 err = copy_from_user(&rtmsg, arg,
2472 sizeof(struct in6_rtmsg));
2476 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2481 err = ip6_route_add(&cfg);
2484 err = ip6_route_del(&cfg);
2498 * Drop the packet on the floor
2501 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2504 struct dst_entry *dst = skb_dst(skb);
2505 switch (ipstats_mib_noroutes) {
2506 case IPSTATS_MIB_INNOROUTES:
2507 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2508 if (type == IPV6_ADDR_ANY) {
2509 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2510 IPSTATS_MIB_INADDRERRORS);
2514 case IPSTATS_MIB_OUTNOROUTES:
2515 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2516 ipstats_mib_noroutes);
2519 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2524 static int ip6_pkt_discard(struct sk_buff *skb)
2526 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2529 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2531 skb->dev = skb_dst(skb)->dev;
2532 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2535 static int ip6_pkt_prohibit(struct sk_buff *skb)
2537 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2540 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2542 skb->dev = skb_dst(skb)->dev;
2543 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2547 * Allocate a dst for local (unicast / anycast) address.
2550 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2551 const struct in6_addr *addr,
2555 struct net *net = dev_net(idev->dev);
2556 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2559 return ERR_PTR(-ENOMEM);
2563 rt->dst.flags |= DST_HOST;
2564 rt->dst.input = ip6_input;
2565 rt->dst.output = ip6_output;
2566 rt->rt6i_idev = idev;
2568 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2570 rt->rt6i_flags |= RTF_ANYCAST;
2572 rt->rt6i_flags |= RTF_LOCAL;
2574 rt->rt6i_gateway = *addr;
2575 rt->rt6i_dst.addr = *addr;
2576 rt->rt6i_dst.plen = 128;
2577 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2578 rt->rt6i_table = fib6_get_table(net, tb_id);
2579 rt->dst.flags |= DST_NOCACHE;
2581 atomic_set(&rt->dst.__refcnt, 1);
2586 int ip6_route_get_saddr(struct net *net,
2587 struct rt6_info *rt,
2588 const struct in6_addr *daddr,
2590 struct in6_addr *saddr)
2592 struct inet6_dev *idev =
2593 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2595 if (rt && rt->rt6i_prefsrc.plen)
2596 *saddr = rt->rt6i_prefsrc.addr;
2598 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2599 daddr, prefs, saddr);
2603 /* remove deleted ip from prefsrc entries */
2604 struct arg_dev_net_ip {
2605 struct net_device *dev;
2607 struct in6_addr *addr;
2610 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2612 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2613 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2614 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2616 if (((void *)rt->dst.dev == dev || !dev) &&
2617 rt != net->ipv6.ip6_null_entry &&
2618 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2619 /* remove prefsrc entry */
2620 rt->rt6i_prefsrc.plen = 0;
2625 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2627 struct net *net = dev_net(ifp->idev->dev);
2628 struct arg_dev_net_ip adni = {
2629 .dev = ifp->idev->dev,
2633 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2636 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2637 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2639 /* Remove routers and update dst entries when gateway turn into host. */
2640 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2642 struct in6_addr *gateway = (struct in6_addr *)arg;
2644 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2645 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2646 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2652 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2654 fib6_clean_all(net, fib6_clean_tohost, gateway);
2657 struct arg_dev_net {
2658 struct net_device *dev;
2662 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2664 const struct arg_dev_net *adn = arg;
2665 const struct net_device *dev = adn->dev;
2667 if ((rt->dst.dev == dev || !dev) &&
2668 rt != adn->net->ipv6.ip6_null_entry)
2674 void rt6_ifdown(struct net *net, struct net_device *dev)
2676 struct arg_dev_net adn = {
2681 fib6_clean_all(net, fib6_ifdown, &adn);
2682 icmp6_clean_all(fib6_ifdown, &adn);
2684 rt6_uncached_list_flush_dev(net, dev);
2687 struct rt6_mtu_change_arg {
2688 struct net_device *dev;
2692 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2694 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2695 struct inet6_dev *idev;
2697 /* In IPv6 pmtu discovery is not optional,
2698 so that RTAX_MTU lock cannot disable it.
2699 We still use this lock to block changes
2700 caused by addrconf/ndisc.
2703 idev = __in6_dev_get(arg->dev);
2707 /* For administrative MTU increase, there is no way to discover
2708 IPv6 PMTU increase, so PMTU increase should be updated here.
2709 Since RFC 1981 doesn't include administrative MTU increase
2710 update PMTU increase is a MUST. (i.e. jumbo frame)
2713 If new MTU is less than route PMTU, this new MTU will be the
2714 lowest MTU in the path, update the route PMTU to reflect PMTU
2715 decreases; if new MTU is greater than route PMTU, and the
2716 old MTU is the lowest MTU in the path, update the route PMTU
2717 to reflect the increase. In this case if the other nodes' MTU
2718 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2721 if (rt->dst.dev == arg->dev &&
2722 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2723 if (rt->rt6i_flags & RTF_CACHE) {
2724 /* For RTF_CACHE with rt6i_pmtu == 0
2725 * (i.e. a redirected route),
2726 * the metrics of its rt->dst.from has already
2729 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2730 rt->rt6i_pmtu = arg->mtu;
2731 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2732 (dst_mtu(&rt->dst) < arg->mtu &&
2733 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2734 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2740 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2742 struct rt6_mtu_change_arg arg = {
2747 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2750 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2751 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2752 [RTA_OIF] = { .type = NLA_U32 },
2753 [RTA_IIF] = { .type = NLA_U32 },
2754 [RTA_PRIORITY] = { .type = NLA_U32 },
2755 [RTA_METRICS] = { .type = NLA_NESTED },
2756 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2757 [RTA_PREF] = { .type = NLA_U8 },
2758 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2759 [RTA_ENCAP] = { .type = NLA_NESTED },
2760 [RTA_EXPIRES] = { .type = NLA_U32 },
2763 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2764 struct fib6_config *cfg)
2767 struct nlattr *tb[RTA_MAX+1];
2771 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2776 rtm = nlmsg_data(nlh);
2777 memset(cfg, 0, sizeof(*cfg));
2779 cfg->fc_table = rtm->rtm_table;
2780 cfg->fc_dst_len = rtm->rtm_dst_len;
2781 cfg->fc_src_len = rtm->rtm_src_len;
2782 cfg->fc_flags = RTF_UP;
2783 cfg->fc_protocol = rtm->rtm_protocol;
2784 cfg->fc_type = rtm->rtm_type;
2786 if (rtm->rtm_type == RTN_UNREACHABLE ||
2787 rtm->rtm_type == RTN_BLACKHOLE ||
2788 rtm->rtm_type == RTN_PROHIBIT ||
2789 rtm->rtm_type == RTN_THROW)
2790 cfg->fc_flags |= RTF_REJECT;
2792 if (rtm->rtm_type == RTN_LOCAL)
2793 cfg->fc_flags |= RTF_LOCAL;
2795 if (rtm->rtm_flags & RTM_F_CLONED)
2796 cfg->fc_flags |= RTF_CACHE;
2798 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2799 cfg->fc_nlinfo.nlh = nlh;
2800 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2802 if (tb[RTA_GATEWAY]) {
2803 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2804 cfg->fc_flags |= RTF_GATEWAY;
2808 int plen = (rtm->rtm_dst_len + 7) >> 3;
2810 if (nla_len(tb[RTA_DST]) < plen)
2813 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2817 int plen = (rtm->rtm_src_len + 7) >> 3;
2819 if (nla_len(tb[RTA_SRC]) < plen)
2822 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2825 if (tb[RTA_PREFSRC])
2826 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2829 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2831 if (tb[RTA_PRIORITY])
2832 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2834 if (tb[RTA_METRICS]) {
2835 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2836 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2840 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2842 if (tb[RTA_MULTIPATH]) {
2843 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2844 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2848 pref = nla_get_u8(tb[RTA_PREF]);
2849 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2850 pref != ICMPV6_ROUTER_PREF_HIGH)
2851 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2852 cfg->fc_flags |= RTF_PREF(pref);
2856 cfg->fc_encap = tb[RTA_ENCAP];
2858 if (tb[RTA_ENCAP_TYPE])
2859 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2861 if (tb[RTA_EXPIRES]) {
2862 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
2864 if (addrconf_finite_timeout(timeout)) {
2865 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
2866 cfg->fc_flags |= RTF_EXPIRES;
2876 struct rt6_info *rt6_info;
2877 struct fib6_config r_cfg;
2878 struct mx6_config mxc;
2879 struct list_head next;
2882 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2886 list_for_each_entry(nh, rt6_nh_list, next) {
2887 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2888 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2889 nh->r_cfg.fc_ifindex);
2893 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2894 struct rt6_info *rt, struct fib6_config *r_cfg)
2897 struct rt6_info *rtnh;
2900 list_for_each_entry(nh, rt6_nh_list, next) {
2901 /* check if rt6_info already exists */
2902 rtnh = nh->rt6_info;
2904 if (rtnh->dst.dev == rt->dst.dev &&
2905 rtnh->rt6i_idev == rt->rt6i_idev &&
2906 ipv6_addr_equal(&rtnh->rt6i_gateway,
2911 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2915 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2920 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2921 list_add_tail(&nh->next, rt6_nh_list);
2926 static int ip6_route_multipath_add(struct fib6_config *cfg)
2928 struct fib6_config r_cfg;
2929 struct rtnexthop *rtnh;
2930 struct rt6_info *rt;
2931 struct rt6_nh *err_nh;
2932 struct rt6_nh *nh, *nh_safe;
2937 int replace = (cfg->fc_nlinfo.nlh &&
2938 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2939 LIST_HEAD(rt6_nh_list);
2941 remaining = cfg->fc_mp_len;
2942 rtnh = (struct rtnexthop *)cfg->fc_mp;
2944 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2945 * rt6_info structs per nexthop
2947 while (rtnh_ok(rtnh, remaining)) {
2948 memcpy(&r_cfg, cfg, sizeof(*cfg));
2949 if (rtnh->rtnh_ifindex)
2950 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2952 attrlen = rtnh_attrlen(rtnh);
2954 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2956 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2958 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2959 r_cfg.fc_flags |= RTF_GATEWAY;
2961 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2962 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2964 r_cfg.fc_encap_type = nla_get_u16(nla);
2967 rt = ip6_route_info_create(&r_cfg);
2974 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2980 rtnh = rtnh_next(rtnh, &remaining);
2984 list_for_each_entry(nh, &rt6_nh_list, next) {
2985 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2986 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2987 nh->rt6_info = NULL;
2990 ip6_print_replace_route_err(&rt6_nh_list);
2995 /* Because each route is added like a single route we remove
2996 * these flags after the first nexthop: if there is a collision,
2997 * we have already failed to add the first nexthop:
2998 * fib6_add_rt2node() has rejected it; when replacing, old
2999 * nexthops have been replaced by first new, the rest should
3002 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
3010 /* Delete routes that were already added */
3011 list_for_each_entry(nh, &rt6_nh_list, next) {
3014 ip6_route_del(&nh->r_cfg);
3018 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3020 dst_free(&nh->rt6_info->dst);
3022 list_del(&nh->next);
3029 static int ip6_route_multipath_del(struct fib6_config *cfg)
3031 struct fib6_config r_cfg;
3032 struct rtnexthop *rtnh;
3035 int err = 1, last_err = 0;
3037 remaining = cfg->fc_mp_len;
3038 rtnh = (struct rtnexthop *)cfg->fc_mp;
3040 /* Parse a Multipath Entry */
3041 while (rtnh_ok(rtnh, remaining)) {
3042 memcpy(&r_cfg, cfg, sizeof(*cfg));
3043 if (rtnh->rtnh_ifindex)
3044 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3046 attrlen = rtnh_attrlen(rtnh);
3048 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3050 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3052 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3053 r_cfg.fc_flags |= RTF_GATEWAY;
3056 err = ip6_route_del(&r_cfg);
3060 rtnh = rtnh_next(rtnh, &remaining);
3066 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3068 struct fib6_config cfg;
3071 err = rtm_to_fib6_config(skb, nlh, &cfg);
3076 return ip6_route_multipath_del(&cfg);
3078 return ip6_route_del(&cfg);
3081 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3083 struct fib6_config cfg;
3086 err = rtm_to_fib6_config(skb, nlh, &cfg);
3091 return ip6_route_multipath_add(&cfg);
3093 return ip6_route_add(&cfg);
3096 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3098 return NLMSG_ALIGN(sizeof(struct rtmsg))
3099 + nla_total_size(16) /* RTA_SRC */
3100 + nla_total_size(16) /* RTA_DST */
3101 + nla_total_size(16) /* RTA_GATEWAY */
3102 + nla_total_size(16) /* RTA_PREFSRC */
3103 + nla_total_size(4) /* RTA_TABLE */
3104 + nla_total_size(4) /* RTA_IIF */
3105 + nla_total_size(4) /* RTA_OIF */
3106 + nla_total_size(4) /* RTA_PRIORITY */
3107 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3108 + nla_total_size(sizeof(struct rta_cacheinfo))
3109 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3110 + nla_total_size(1) /* RTA_PREF */
3111 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3114 static int rt6_fill_node(struct net *net,
3115 struct sk_buff *skb, struct rt6_info *rt,
3116 struct in6_addr *dst, struct in6_addr *src,
3117 int iif, int type, u32 portid, u32 seq,
3118 int prefix, int nowait, unsigned int flags)
3120 u32 metrics[RTAX_MAX];
3122 struct nlmsghdr *nlh;
3126 if (prefix) { /* user wants prefix routes only */
3127 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3128 /* success since this is not a prefix route */
3133 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3137 rtm = nlmsg_data(nlh);
3138 rtm->rtm_family = AF_INET6;
3139 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3140 rtm->rtm_src_len = rt->rt6i_src.plen;
3143 table = rt->rt6i_table->tb6_id;
3145 table = RT6_TABLE_UNSPEC;
3146 rtm->rtm_table = table;
3147 if (nla_put_u32(skb, RTA_TABLE, table))
3148 goto nla_put_failure;
3149 if (rt->rt6i_flags & RTF_REJECT) {
3150 switch (rt->dst.error) {
3152 rtm->rtm_type = RTN_BLACKHOLE;
3155 rtm->rtm_type = RTN_PROHIBIT;
3158 rtm->rtm_type = RTN_THROW;
3161 rtm->rtm_type = RTN_UNREACHABLE;
3165 else if (rt->rt6i_flags & RTF_LOCAL)
3166 rtm->rtm_type = RTN_LOCAL;
3167 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3168 rtm->rtm_type = RTN_LOCAL;
3170 rtm->rtm_type = RTN_UNICAST;
3172 if (!netif_carrier_ok(rt->dst.dev)) {
3173 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3174 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3175 rtm->rtm_flags |= RTNH_F_DEAD;
3177 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3178 rtm->rtm_protocol = rt->rt6i_protocol;
3179 if (rt->rt6i_flags & RTF_DYNAMIC)
3180 rtm->rtm_protocol = RTPROT_REDIRECT;
3181 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3182 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3183 rtm->rtm_protocol = RTPROT_RA;
3185 rtm->rtm_protocol = RTPROT_KERNEL;
3188 if (rt->rt6i_flags & RTF_CACHE)
3189 rtm->rtm_flags |= RTM_F_CLONED;
3192 if (nla_put_in6_addr(skb, RTA_DST, dst))
3193 goto nla_put_failure;
3194 rtm->rtm_dst_len = 128;
3195 } else if (rtm->rtm_dst_len)
3196 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3197 goto nla_put_failure;
3198 #ifdef CONFIG_IPV6_SUBTREES
3200 if (nla_put_in6_addr(skb, RTA_SRC, src))
3201 goto nla_put_failure;
3202 rtm->rtm_src_len = 128;
3203 } else if (rtm->rtm_src_len &&
3204 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3205 goto nla_put_failure;
3208 #ifdef CONFIG_IPV6_MROUTE
3209 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3210 int err = ip6mr_get_route(net, skb, rtm, nowait);
3215 goto nla_put_failure;
3217 if (err == -EMSGSIZE)
3218 goto nla_put_failure;
3223 if (nla_put_u32(skb, RTA_IIF, iif))
3224 goto nla_put_failure;
3226 struct in6_addr saddr_buf;
3227 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3228 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3229 goto nla_put_failure;
3232 if (rt->rt6i_prefsrc.plen) {
3233 struct in6_addr saddr_buf;
3234 saddr_buf = rt->rt6i_prefsrc.addr;
3235 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3236 goto nla_put_failure;
3239 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3241 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3242 if (rtnetlink_put_metrics(skb, metrics) < 0)
3243 goto nla_put_failure;
3245 if (rt->rt6i_flags & RTF_GATEWAY) {
3246 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3247 goto nla_put_failure;
3251 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3252 goto nla_put_failure;
3253 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3254 goto nla_put_failure;
3256 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3258 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3259 goto nla_put_failure;
3261 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3262 goto nla_put_failure;
3264 lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3266 nlmsg_end(skb, nlh);
3270 nlmsg_cancel(skb, nlh);
3274 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3276 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3279 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3280 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3281 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3285 return rt6_fill_node(arg->net,
3286 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3287 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3288 prefix, 0, NLM_F_MULTI);
3291 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3293 struct net *net = sock_net(in_skb->sk);
3294 struct nlattr *tb[RTA_MAX+1];
3295 struct rt6_info *rt;
3296 struct sk_buff *skb;
3299 int err, iif = 0, oif = 0;
3301 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3306 memset(&fl6, 0, sizeof(fl6));
3309 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3312 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3316 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3319 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3323 iif = nla_get_u32(tb[RTA_IIF]);
3326 oif = nla_get_u32(tb[RTA_OIF]);
3329 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3332 struct net_device *dev;
3335 dev = __dev_get_by_index(net, iif);
3341 fl6.flowi6_iif = iif;
3343 if (!ipv6_addr_any(&fl6.saddr))
3344 flags |= RT6_LOOKUP_F_HAS_SADDR;
3346 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3349 fl6.flowi6_oif = oif;
3351 if (netif_index_is_l3_master(net, oif)) {
3352 fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3353 FLOWI_FLAG_SKIP_NH_OIF;
3356 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3359 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3366 /* Reserve room for dummy headers, this skb can pass
3367 through good chunk of routing engine.
3369 skb_reset_mac_header(skb);
3370 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3372 skb_dst_set(skb, &rt->dst);
3374 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3375 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3376 nlh->nlmsg_seq, 0, 0, 0);
3382 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3387 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3388 unsigned int nlm_flags)
3390 struct sk_buff *skb;
3391 struct net *net = info->nl_net;
3396 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3398 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3402 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3403 event, info->portid, seq, 0, 0, nlm_flags);
3405 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3406 WARN_ON(err == -EMSGSIZE);
3410 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3411 info->nlh, gfp_any());
3415 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3418 static int ip6_route_dev_notify(struct notifier_block *this,
3419 unsigned long event, void *ptr)
3421 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3422 struct net *net = dev_net(dev);
3424 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3425 net->ipv6.ip6_null_entry->dst.dev = dev;
3426 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3427 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3428 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3429 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3430 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3431 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3442 #ifdef CONFIG_PROC_FS
3444 static const struct file_operations ipv6_route_proc_fops = {
3445 .owner = THIS_MODULE,
3446 .open = ipv6_route_open,
3448 .llseek = seq_lseek,
3449 .release = seq_release_net,
3452 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3454 struct net *net = (struct net *)seq->private;
3455 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3456 net->ipv6.rt6_stats->fib_nodes,
3457 net->ipv6.rt6_stats->fib_route_nodes,
3458 net->ipv6.rt6_stats->fib_rt_alloc,
3459 net->ipv6.rt6_stats->fib_rt_entries,
3460 net->ipv6.rt6_stats->fib_rt_cache,
3461 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3462 net->ipv6.rt6_stats->fib_discarded_routes);
3467 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3469 return single_open_net(inode, file, rt6_stats_seq_show);
3472 static const struct file_operations rt6_stats_seq_fops = {
3473 .owner = THIS_MODULE,
3474 .open = rt6_stats_seq_open,
3476 .llseek = seq_lseek,
3477 .release = single_release_net,
3479 #endif /* CONFIG_PROC_FS */
3481 #ifdef CONFIG_SYSCTL
3484 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3485 void __user *buffer, size_t *lenp, loff_t *ppos)
3492 net = (struct net *)ctl->extra1;
3493 delay = net->ipv6.sysctl.flush_delay;
3494 proc_dointvec(ctl, write, buffer, lenp, ppos);
3495 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3499 struct ctl_table ipv6_route_table_template[] = {
3501 .procname = "flush",
3502 .data = &init_net.ipv6.sysctl.flush_delay,
3503 .maxlen = sizeof(int),
3505 .proc_handler = ipv6_sysctl_rtcache_flush
3508 .procname = "gc_thresh",
3509 .data = &ip6_dst_ops_template.gc_thresh,
3510 .maxlen = sizeof(int),
3512 .proc_handler = proc_dointvec,
3515 .procname = "max_size",
3516 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3517 .maxlen = sizeof(int),
3519 .proc_handler = proc_dointvec,
3522 .procname = "gc_min_interval",
3523 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3524 .maxlen = sizeof(int),
3526 .proc_handler = proc_dointvec_jiffies,
3529 .procname = "gc_timeout",
3530 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3531 .maxlen = sizeof(int),
3533 .proc_handler = proc_dointvec_jiffies,
3536 .procname = "gc_interval",
3537 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3538 .maxlen = sizeof(int),
3540 .proc_handler = proc_dointvec_jiffies,
3543 .procname = "gc_elasticity",
3544 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3545 .maxlen = sizeof(int),
3547 .proc_handler = proc_dointvec,
3550 .procname = "mtu_expires",
3551 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3552 .maxlen = sizeof(int),
3554 .proc_handler = proc_dointvec_jiffies,
3557 .procname = "min_adv_mss",
3558 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3559 .maxlen = sizeof(int),
3561 .proc_handler = proc_dointvec,
3564 .procname = "gc_min_interval_ms",
3565 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3566 .maxlen = sizeof(int),
3568 .proc_handler = proc_dointvec_ms_jiffies,
3573 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3575 struct ctl_table *table;
3577 table = kmemdup(ipv6_route_table_template,
3578 sizeof(ipv6_route_table_template),
3582 table[0].data = &net->ipv6.sysctl.flush_delay;
3583 table[0].extra1 = net;
3584 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3585 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3586 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3587 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3588 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3589 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3590 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3591 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3592 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3594 /* Don't export sysctls to unprivileged users */
3595 if (net->user_ns != &init_user_ns)
3596 table[0].procname = NULL;
3603 static int __net_init ip6_route_net_init(struct net *net)
3607 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3608 sizeof(net->ipv6.ip6_dst_ops));
3610 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3611 goto out_ip6_dst_ops;
3613 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3614 sizeof(*net->ipv6.ip6_null_entry),
3616 if (!net->ipv6.ip6_null_entry)
3617 goto out_ip6_dst_entries;
3618 net->ipv6.ip6_null_entry->dst.path =
3619 (struct dst_entry *)net->ipv6.ip6_null_entry;
3620 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3621 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3622 ip6_template_metrics, true);
3624 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3625 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3626 sizeof(*net->ipv6.ip6_prohibit_entry),
3628 if (!net->ipv6.ip6_prohibit_entry)
3629 goto out_ip6_null_entry;
3630 net->ipv6.ip6_prohibit_entry->dst.path =
3631 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3632 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3633 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3634 ip6_template_metrics, true);
3636 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3637 sizeof(*net->ipv6.ip6_blk_hole_entry),
3639 if (!net->ipv6.ip6_blk_hole_entry)
3640 goto out_ip6_prohibit_entry;
3641 net->ipv6.ip6_blk_hole_entry->dst.path =
3642 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3643 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3644 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3645 ip6_template_metrics, true);
3648 net->ipv6.sysctl.flush_delay = 0;
3649 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3650 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3651 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3652 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3653 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3654 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3655 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3657 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3663 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3664 out_ip6_prohibit_entry:
3665 kfree(net->ipv6.ip6_prohibit_entry);
3667 kfree(net->ipv6.ip6_null_entry);
3669 out_ip6_dst_entries:
3670 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3675 static void __net_exit ip6_route_net_exit(struct net *net)
3677 kfree(net->ipv6.ip6_null_entry);
3678 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3679 kfree(net->ipv6.ip6_prohibit_entry);
3680 kfree(net->ipv6.ip6_blk_hole_entry);
3682 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3685 static int __net_init ip6_route_net_init_late(struct net *net)
3687 #ifdef CONFIG_PROC_FS
3688 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3689 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3694 static void __net_exit ip6_route_net_exit_late(struct net *net)
3696 #ifdef CONFIG_PROC_FS
3697 remove_proc_entry("ipv6_route", net->proc_net);
3698 remove_proc_entry("rt6_stats", net->proc_net);
3702 static struct pernet_operations ip6_route_net_ops = {
3703 .init = ip6_route_net_init,
3704 .exit = ip6_route_net_exit,
3707 static int __net_init ipv6_inetpeer_init(struct net *net)
3709 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3713 inet_peer_base_init(bp);
3714 net->ipv6.peers = bp;
3718 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3720 struct inet_peer_base *bp = net->ipv6.peers;
3722 net->ipv6.peers = NULL;
3723 inetpeer_invalidate_tree(bp);
3727 static struct pernet_operations ipv6_inetpeer_ops = {
3728 .init = ipv6_inetpeer_init,
3729 .exit = ipv6_inetpeer_exit,
3732 static struct pernet_operations ip6_route_net_late_ops = {
3733 .init = ip6_route_net_init_late,
3734 .exit = ip6_route_net_exit_late,
3737 static struct notifier_block ip6_route_dev_notifier = {
3738 .notifier_call = ip6_route_dev_notify,
3742 int __init ip6_route_init(void)
3748 ip6_dst_ops_template.kmem_cachep =
3749 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3750 SLAB_HWCACHE_ALIGN, NULL);
3751 if (!ip6_dst_ops_template.kmem_cachep)
3754 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3756 goto out_kmem_cache;
3758 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3760 goto out_dst_entries;
3762 ret = register_pernet_subsys(&ip6_route_net_ops);
3764 goto out_register_inetpeer;
3766 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3768 /* Registering of the loopback is done before this portion of code,
3769 * the loopback reference in rt6_info will not be taken, do it
3770 * manually for init_net */
3771 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3772 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3773 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3774 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3775 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3776 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3777 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3781 goto out_register_subsys;
3787 ret = fib6_rules_init();
3791 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3793 goto fib6_rules_init;
3796 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3797 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3798 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3799 goto out_register_late_subsys;
3801 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3803 goto out_register_late_subsys;
3805 for_each_possible_cpu(cpu) {
3806 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3808 INIT_LIST_HEAD(&ul->head);
3809 spin_lock_init(&ul->lock);
3815 out_register_late_subsys:
3816 unregister_pernet_subsys(&ip6_route_net_late_ops);
3818 fib6_rules_cleanup();
3823 out_register_subsys:
3824 unregister_pernet_subsys(&ip6_route_net_ops);
3825 out_register_inetpeer:
3826 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3828 dst_entries_destroy(&ip6_dst_blackhole_ops);
3830 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3834 void ip6_route_cleanup(void)
3836 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3837 unregister_pernet_subsys(&ip6_route_net_late_ops);
3838 fib6_rules_cleanup();
3841 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3842 unregister_pernet_subsys(&ip6_route_net_ops);
3843 dst_entries_destroy(&ip6_dst_blackhole_ops);
3844 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);