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 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1046 int oif, 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);
1142 EXPORT_SYMBOL_GPL(ip6_pol_route);
1144 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1145 struct flowi6 *fl6, int flags)
1147 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1150 struct dst_entry *ip6_route_input_lookup(struct net *net,
1151 struct net_device *dev,
1152 struct flowi6 *fl6, int flags)
1154 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1155 flags |= RT6_LOOKUP_F_IFACE;
1157 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1159 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
1161 void ip6_route_input(struct sk_buff *skb)
1163 const struct ipv6hdr *iph = ipv6_hdr(skb);
1164 struct net *net = dev_net(skb->dev);
1165 int flags = RT6_LOOKUP_F_HAS_SADDR;
1166 struct ip_tunnel_info *tun_info;
1167 struct flowi6 fl6 = {
1168 .flowi6_iif = skb->dev->ifindex,
1169 .daddr = iph->daddr,
1170 .saddr = iph->saddr,
1171 .flowlabel = ip6_flowinfo(iph),
1172 .flowi6_mark = skb->mark,
1173 .flowi6_proto = iph->nexthdr,
1176 tun_info = skb_tunnel_info(skb);
1177 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1178 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1180 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1183 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1184 struct flowi6 *fl6, int flags)
1186 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1189 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1190 struct flowi6 *fl6, int flags)
1194 if (rt6_need_strict(&fl6->daddr)) {
1195 struct dst_entry *dst;
1197 dst = l3mdev_link_scope_lookup(net, fl6);
1202 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1204 any_src = ipv6_addr_any(&fl6->saddr);
1205 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1206 (fl6->flowi6_oif && any_src))
1207 flags |= RT6_LOOKUP_F_IFACE;
1210 flags |= RT6_LOOKUP_F_HAS_SADDR;
1212 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1214 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1216 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1218 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1220 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1221 struct dst_entry *new = NULL;
1223 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1229 new->input = dst_discard;
1230 new->output = dst_discard_out;
1232 dst_copy_metrics(new, &ort->dst);
1233 rt->rt6i_idev = ort->rt6i_idev;
1235 in6_dev_hold(rt->rt6i_idev);
1237 rt->rt6i_gateway = ort->rt6i_gateway;
1238 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1239 rt->rt6i_metric = 0;
1241 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1242 #ifdef CONFIG_IPV6_SUBTREES
1243 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1249 dst_release(dst_orig);
1250 return new ? new : ERR_PTR(-ENOMEM);
1254 * Destination cache support functions
1257 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1260 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1261 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1264 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1266 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1269 if (rt6_check_expired(rt))
1275 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1277 if (!__rt6_check_expired(rt) &&
1278 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1279 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1285 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1287 struct rt6_info *rt;
1289 rt = (struct rt6_info *) dst;
1291 /* All IPV6 dsts are created with ->obsolete set to the value
1292 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1293 * into this function always.
1296 rt6_dst_from_metrics_check(rt);
1298 if (rt->rt6i_flags & RTF_PCPU ||
1299 (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1300 return rt6_dst_from_check(rt, cookie);
1302 return rt6_check(rt, cookie);
1305 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1307 struct rt6_info *rt = (struct rt6_info *) dst;
1310 if (rt->rt6i_flags & RTF_CACHE) {
1311 if (rt6_check_expired(rt)) {
1323 static void ip6_link_failure(struct sk_buff *skb)
1325 struct rt6_info *rt;
1327 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1329 rt = (struct rt6_info *) skb_dst(skb);
1331 if (rt->rt6i_flags & RTF_CACHE) {
1334 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1335 rt->rt6i_node->fn_sernum = -1;
1340 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1342 struct net *net = dev_net(rt->dst.dev);
1344 rt->rt6i_flags |= RTF_MODIFIED;
1345 rt->rt6i_pmtu = mtu;
1346 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1349 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1351 return !(rt->rt6i_flags & RTF_CACHE) &&
1352 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1355 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1356 const struct ipv6hdr *iph, u32 mtu)
1358 struct rt6_info *rt6 = (struct rt6_info *)dst;
1360 if (rt6->rt6i_flags & RTF_LOCAL)
1364 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1365 if (mtu >= dst_mtu(dst))
1368 if (!rt6_cache_allowed_for_pmtu(rt6)) {
1369 rt6_do_update_pmtu(rt6, mtu);
1371 const struct in6_addr *daddr, *saddr;
1372 struct rt6_info *nrt6;
1375 daddr = &iph->daddr;
1376 saddr = &iph->saddr;
1378 daddr = &sk->sk_v6_daddr;
1379 saddr = &inet6_sk(sk)->saddr;
1383 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1385 rt6_do_update_pmtu(nrt6, mtu);
1387 /* ip6_ins_rt(nrt6) will bump the
1388 * rt6->rt6i_node->fn_sernum
1389 * which will fail the next rt6_check() and
1390 * invalidate the sk->sk_dst_cache.
1397 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1398 struct sk_buff *skb, u32 mtu)
1400 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1403 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1406 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1407 struct dst_entry *dst;
1410 memset(&fl6, 0, sizeof(fl6));
1411 fl6.flowi6_oif = oif;
1412 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1413 fl6.daddr = iph->daddr;
1414 fl6.saddr = iph->saddr;
1415 fl6.flowlabel = ip6_flowinfo(iph);
1417 dst = ip6_route_output(net, NULL, &fl6);
1419 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1422 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1424 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1426 struct dst_entry *dst;
1428 ip6_update_pmtu(skb, sock_net(sk), mtu,
1429 sk->sk_bound_dev_if, sk->sk_mark);
1431 dst = __sk_dst_get(sk);
1432 if (!dst || !dst->obsolete ||
1433 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1437 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1438 ip6_datagram_dst_update(sk, false);
1441 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1443 /* Handle redirects */
1444 struct ip6rd_flowi {
1446 struct in6_addr gateway;
1449 static struct rt6_info *__ip6_route_redirect(struct net *net,
1450 struct fib6_table *table,
1454 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1455 struct rt6_info *rt;
1456 struct fib6_node *fn;
1458 /* Get the "current" route for this destination and
1459 * check if the redirect has come from approriate router.
1461 * RFC 4861 specifies that redirects should only be
1462 * accepted if they come from the nexthop to the target.
1463 * Due to the way the routes are chosen, this notion
1464 * is a bit fuzzy and one might need to check all possible
1468 read_lock_bh(&table->tb6_lock);
1469 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1471 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1472 if (rt6_check_expired(rt))
1476 if (!(rt->rt6i_flags & RTF_GATEWAY))
1478 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1480 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1486 rt = net->ipv6.ip6_null_entry;
1487 else if (rt->dst.error) {
1488 rt = net->ipv6.ip6_null_entry;
1492 if (rt == net->ipv6.ip6_null_entry) {
1493 fn = fib6_backtrack(fn, &fl6->saddr);
1501 read_unlock_bh(&table->tb6_lock);
1503 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1507 static struct dst_entry *ip6_route_redirect(struct net *net,
1508 const struct flowi6 *fl6,
1509 const struct in6_addr *gateway)
1511 int flags = RT6_LOOKUP_F_HAS_SADDR;
1512 struct ip6rd_flowi rdfl;
1515 rdfl.gateway = *gateway;
1517 return fib6_rule_lookup(net, &rdfl.fl6,
1518 flags, __ip6_route_redirect);
1521 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1523 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1524 struct dst_entry *dst;
1527 memset(&fl6, 0, sizeof(fl6));
1528 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1529 fl6.flowi6_oif = oif;
1530 fl6.flowi6_mark = mark;
1531 fl6.daddr = iph->daddr;
1532 fl6.saddr = iph->saddr;
1533 fl6.flowlabel = ip6_flowinfo(iph);
1535 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1536 rt6_do_redirect(dst, NULL, skb);
1539 EXPORT_SYMBOL_GPL(ip6_redirect);
1541 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1544 const struct ipv6hdr *iph = ipv6_hdr(skb);
1545 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1546 struct dst_entry *dst;
1549 memset(&fl6, 0, sizeof(fl6));
1550 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1551 fl6.flowi6_oif = oif;
1552 fl6.flowi6_mark = mark;
1553 fl6.daddr = msg->dest;
1554 fl6.saddr = iph->daddr;
1556 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1557 rt6_do_redirect(dst, NULL, skb);
1561 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1563 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1565 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1567 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1569 struct net_device *dev = dst->dev;
1570 unsigned int mtu = dst_mtu(dst);
1571 struct net *net = dev_net(dev);
1573 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1575 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1576 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1579 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1580 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1581 * IPV6_MAXPLEN is also valid and means: "any MSS,
1582 * rely only on pmtu discovery"
1584 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1589 static unsigned int ip6_mtu(const struct dst_entry *dst)
1591 const struct rt6_info *rt = (const struct rt6_info *)dst;
1592 unsigned int mtu = rt->rt6i_pmtu;
1593 struct inet6_dev *idev;
1598 mtu = dst_metric_raw(dst, RTAX_MTU);
1605 idev = __in6_dev_get(dst->dev);
1607 mtu = idev->cnf.mtu6;
1611 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1613 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
1616 static struct dst_entry *icmp6_dst_gc_list;
1617 static DEFINE_SPINLOCK(icmp6_dst_lock);
1619 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1622 struct dst_entry *dst;
1623 struct rt6_info *rt;
1624 struct inet6_dev *idev = in6_dev_get(dev);
1625 struct net *net = dev_net(dev);
1627 if (unlikely(!idev))
1628 return ERR_PTR(-ENODEV);
1630 rt = ip6_dst_alloc(net, dev, 0);
1631 if (unlikely(!rt)) {
1633 dst = ERR_PTR(-ENOMEM);
1637 rt->dst.flags |= DST_HOST;
1638 rt->dst.output = ip6_output;
1639 atomic_set(&rt->dst.__refcnt, 1);
1640 rt->rt6i_gateway = fl6->daddr;
1641 rt->rt6i_dst.addr = fl6->daddr;
1642 rt->rt6i_dst.plen = 128;
1643 rt->rt6i_idev = idev;
1644 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1646 spin_lock_bh(&icmp6_dst_lock);
1647 rt->dst.next = icmp6_dst_gc_list;
1648 icmp6_dst_gc_list = &rt->dst;
1649 spin_unlock_bh(&icmp6_dst_lock);
1651 fib6_force_start_gc(net);
1653 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1659 int icmp6_dst_gc(void)
1661 struct dst_entry *dst, **pprev;
1664 spin_lock_bh(&icmp6_dst_lock);
1665 pprev = &icmp6_dst_gc_list;
1667 while ((dst = *pprev) != NULL) {
1668 if (!atomic_read(&dst->__refcnt)) {
1677 spin_unlock_bh(&icmp6_dst_lock);
1682 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1685 struct dst_entry *dst, **pprev;
1687 spin_lock_bh(&icmp6_dst_lock);
1688 pprev = &icmp6_dst_gc_list;
1689 while ((dst = *pprev) != NULL) {
1690 struct rt6_info *rt = (struct rt6_info *) dst;
1691 if (func(rt, arg)) {
1698 spin_unlock_bh(&icmp6_dst_lock);
1701 static int ip6_dst_gc(struct dst_ops *ops)
1703 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1704 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1705 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1706 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1707 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1708 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1711 entries = dst_entries_get_fast(ops);
1712 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1713 entries <= rt_max_size)
1716 net->ipv6.ip6_rt_gc_expire++;
1717 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1718 entries = dst_entries_get_slow(ops);
1719 if (entries < ops->gc_thresh)
1720 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1722 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1723 return entries > rt_max_size;
1726 static int ip6_convert_metrics(struct mx6_config *mxc,
1727 const struct fib6_config *cfg)
1729 bool ecn_ca = false;
1737 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1741 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1742 int type = nla_type(nla);
1747 if (unlikely(type > RTAX_MAX))
1750 if (type == RTAX_CC_ALGO) {
1751 char tmp[TCP_CA_NAME_MAX];
1753 nla_strlcpy(tmp, nla, sizeof(tmp));
1754 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1755 if (val == TCP_CA_UNSPEC)
1758 val = nla_get_u32(nla);
1760 if (type == RTAX_HOPLIMIT && val > 255)
1762 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1766 __set_bit(type - 1, mxc->mx_valid);
1770 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1771 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1781 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
1782 struct fib6_config *cfg,
1783 const struct in6_addr *gw_addr)
1785 struct flowi6 fl6 = {
1786 .flowi6_oif = cfg->fc_ifindex,
1788 .saddr = cfg->fc_prefsrc,
1790 struct fib6_table *table;
1791 struct rt6_info *rt;
1792 int flags = RT6_LOOKUP_F_IFACE;
1794 table = fib6_get_table(net, cfg->fc_table);
1798 if (!ipv6_addr_any(&cfg->fc_prefsrc))
1799 flags |= RT6_LOOKUP_F_HAS_SADDR;
1801 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
1803 /* if table lookup failed, fall back to full lookup */
1804 if (rt == net->ipv6.ip6_null_entry) {
1812 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1814 struct net *net = cfg->fc_nlinfo.nl_net;
1815 struct rt6_info *rt = NULL;
1816 struct net_device *dev = NULL;
1817 struct inet6_dev *idev = NULL;
1818 struct fib6_table *table;
1822 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1824 #ifndef CONFIG_IPV6_SUBTREES
1825 if (cfg->fc_src_len)
1828 if (cfg->fc_ifindex) {
1830 dev = dev_get_by_index(net, cfg->fc_ifindex);
1833 idev = in6_dev_get(dev);
1838 if (cfg->fc_metric == 0)
1839 cfg->fc_metric = IP6_RT_PRIO_USER;
1842 if (cfg->fc_nlinfo.nlh &&
1843 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1844 table = fib6_get_table(net, cfg->fc_table);
1846 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1847 table = fib6_new_table(net, cfg->fc_table);
1850 table = fib6_new_table(net, cfg->fc_table);
1856 rt = ip6_dst_alloc(net, NULL,
1857 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1864 if (cfg->fc_flags & RTF_EXPIRES)
1865 rt6_set_expires(rt, jiffies +
1866 clock_t_to_jiffies(cfg->fc_expires));
1868 rt6_clean_expires(rt);
1870 if (cfg->fc_protocol == RTPROT_UNSPEC)
1871 cfg->fc_protocol = RTPROT_BOOT;
1872 rt->rt6i_protocol = cfg->fc_protocol;
1874 addr_type = ipv6_addr_type(&cfg->fc_dst);
1876 if (addr_type & IPV6_ADDR_MULTICAST)
1877 rt->dst.input = ip6_mc_input;
1878 else if (cfg->fc_flags & RTF_LOCAL)
1879 rt->dst.input = ip6_input;
1881 rt->dst.input = ip6_forward;
1883 rt->dst.output = ip6_output;
1885 if (cfg->fc_encap) {
1886 struct lwtunnel_state *lwtstate;
1888 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1889 cfg->fc_encap, AF_INET6, cfg,
1893 rt->dst.lwtstate = lwtstate_get(lwtstate);
1894 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1895 rt->dst.lwtstate->orig_output = rt->dst.output;
1896 rt->dst.output = lwtunnel_output;
1898 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1899 rt->dst.lwtstate->orig_input = rt->dst.input;
1900 rt->dst.input = lwtunnel_input;
1904 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1905 rt->rt6i_dst.plen = cfg->fc_dst_len;
1906 if (rt->rt6i_dst.plen == 128)
1907 rt->dst.flags |= DST_HOST;
1909 #ifdef CONFIG_IPV6_SUBTREES
1910 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1911 rt->rt6i_src.plen = cfg->fc_src_len;
1914 rt->rt6i_metric = cfg->fc_metric;
1916 /* We cannot add true routes via loopback here,
1917 they would result in kernel looping; promote them to reject routes
1919 if ((cfg->fc_flags & RTF_REJECT) ||
1920 (dev && (dev->flags & IFF_LOOPBACK) &&
1921 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1922 !(cfg->fc_flags & RTF_LOCAL))) {
1923 /* hold loopback dev/idev if we haven't done so. */
1924 if (dev != net->loopback_dev) {
1929 dev = net->loopback_dev;
1931 idev = in6_dev_get(dev);
1937 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1938 switch (cfg->fc_type) {
1940 rt->dst.error = -EINVAL;
1941 rt->dst.output = dst_discard_out;
1942 rt->dst.input = dst_discard;
1945 rt->dst.error = -EACCES;
1946 rt->dst.output = ip6_pkt_prohibit_out;
1947 rt->dst.input = ip6_pkt_prohibit;
1950 case RTN_UNREACHABLE:
1952 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1953 : (cfg->fc_type == RTN_UNREACHABLE)
1954 ? -EHOSTUNREACH : -ENETUNREACH;
1955 rt->dst.output = ip6_pkt_discard_out;
1956 rt->dst.input = ip6_pkt_discard;
1962 if (cfg->fc_flags & RTF_GATEWAY) {
1963 const struct in6_addr *gw_addr;
1966 gw_addr = &cfg->fc_gateway;
1967 gwa_type = ipv6_addr_type(gw_addr);
1969 /* if gw_addr is local we will fail to detect this in case
1970 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1971 * will return already-added prefix route via interface that
1972 * prefix route was assigned to, which might be non-loopback.
1975 if (ipv6_chk_addr_and_flags(net, gw_addr,
1976 gwa_type & IPV6_ADDR_LINKLOCAL ?
1980 rt->rt6i_gateway = *gw_addr;
1982 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1983 struct rt6_info *grt = NULL;
1985 /* IPv6 strictly inhibits using not link-local
1986 addresses as nexthop address.
1987 Otherwise, router will not able to send redirects.
1988 It is very good, but in some (rare!) circumstances
1989 (SIT, PtP, NBMA NOARP links) it is handy to allow
1990 some exceptions. --ANK
1992 if (!(gwa_type & IPV6_ADDR_UNICAST))
1995 if (cfg->fc_table) {
1996 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
1999 if (grt->rt6i_flags & RTF_GATEWAY ||
2000 (dev && dev != grt->dst.dev)) {
2008 grt = rt6_lookup(net, gw_addr, NULL,
2009 cfg->fc_ifindex, 1);
2011 err = -EHOSTUNREACH;
2015 if (dev != grt->dst.dev) {
2021 idev = grt->rt6i_idev;
2023 in6_dev_hold(grt->rt6i_idev);
2025 if (!(grt->rt6i_flags & RTF_GATEWAY))
2033 if (!dev || (dev->flags & IFF_LOOPBACK))
2041 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2042 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2046 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2047 rt->rt6i_prefsrc.plen = 128;
2049 rt->rt6i_prefsrc.plen = 0;
2051 rt->rt6i_flags = cfg->fc_flags;
2055 rt->rt6i_idev = idev;
2056 rt->rt6i_table = table;
2058 cfg->fc_nlinfo.nl_net = dev_net(dev);
2069 return ERR_PTR(err);
2072 int ip6_route_add(struct fib6_config *cfg)
2074 struct mx6_config mxc = { .mx = NULL, };
2075 struct rt6_info *rt;
2078 rt = ip6_route_info_create(cfg);
2085 err = ip6_convert_metrics(&mxc, cfg);
2089 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2101 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2104 struct fib6_table *table;
2105 struct net *net = dev_net(rt->dst.dev);
2107 if (rt == net->ipv6.ip6_null_entry ||
2108 rt->dst.flags & DST_NOCACHE) {
2113 table = rt->rt6i_table;
2114 write_lock_bh(&table->tb6_lock);
2115 err = fib6_del(rt, info);
2116 write_unlock_bh(&table->tb6_lock);
2123 int ip6_del_rt(struct rt6_info *rt)
2125 struct nl_info info = {
2126 .nl_net = dev_net(rt->dst.dev),
2128 return __ip6_del_rt(rt, &info);
2131 static int ip6_route_del(struct fib6_config *cfg)
2133 struct fib6_table *table;
2134 struct fib6_node *fn;
2135 struct rt6_info *rt;
2138 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2142 read_lock_bh(&table->tb6_lock);
2144 fn = fib6_locate(&table->tb6_root,
2145 &cfg->fc_dst, cfg->fc_dst_len,
2146 &cfg->fc_src, cfg->fc_src_len);
2149 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2150 if ((rt->rt6i_flags & RTF_CACHE) &&
2151 !(cfg->fc_flags & RTF_CACHE))
2153 if (cfg->fc_ifindex &&
2155 rt->dst.dev->ifindex != cfg->fc_ifindex))
2157 if (cfg->fc_flags & RTF_GATEWAY &&
2158 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2160 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2163 read_unlock_bh(&table->tb6_lock);
2165 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2168 read_unlock_bh(&table->tb6_lock);
2173 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2175 struct netevent_redirect netevent;
2176 struct rt6_info *rt, *nrt = NULL;
2177 struct ndisc_options ndopts;
2178 struct inet6_dev *in6_dev;
2179 struct neighbour *neigh;
2181 int optlen, on_link;
2184 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2185 optlen -= sizeof(*msg);
2188 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2192 msg = (struct rd_msg *)icmp6_hdr(skb);
2194 if (ipv6_addr_is_multicast(&msg->dest)) {
2195 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2200 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2202 } else if (ipv6_addr_type(&msg->target) !=
2203 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2204 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2208 in6_dev = __in6_dev_get(skb->dev);
2211 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2215 * The IP source address of the Redirect MUST be the same as the current
2216 * first-hop router for the specified ICMP Destination Address.
2219 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
2220 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2225 if (ndopts.nd_opts_tgt_lladdr) {
2226 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2229 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2234 rt = (struct rt6_info *) dst;
2235 if (rt->rt6i_flags & RTF_REJECT) {
2236 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2240 /* Redirect received -> path was valid.
2241 * Look, redirects are sent only in response to data packets,
2242 * so that this nexthop apparently is reachable. --ANK
2244 dst_confirm(&rt->dst);
2246 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2251 * We have finally decided to accept it.
2254 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
2255 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2256 NEIGH_UPDATE_F_OVERRIDE|
2257 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2258 NEIGH_UPDATE_F_ISROUTER)),
2259 NDISC_REDIRECT, &ndopts);
2261 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2265 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2267 nrt->rt6i_flags &= ~RTF_GATEWAY;
2269 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2271 if (ip6_ins_rt(nrt))
2274 netevent.old = &rt->dst;
2275 netevent.new = &nrt->dst;
2276 netevent.daddr = &msg->dest;
2277 netevent.neigh = neigh;
2278 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2280 if (rt->rt6i_flags & RTF_CACHE) {
2281 rt = (struct rt6_info *) dst_clone(&rt->dst);
2286 neigh_release(neigh);
2290 * Misc support functions
2293 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2295 BUG_ON(from->dst.from);
2297 rt->rt6i_flags &= ~RTF_EXPIRES;
2298 dst_hold(&from->dst);
2299 rt->dst.from = &from->dst;
2300 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2303 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2305 rt->dst.input = ort->dst.input;
2306 rt->dst.output = ort->dst.output;
2307 rt->rt6i_dst = ort->rt6i_dst;
2308 rt->dst.error = ort->dst.error;
2309 rt->rt6i_idev = ort->rt6i_idev;
2311 in6_dev_hold(rt->rt6i_idev);
2312 rt->dst.lastuse = jiffies;
2313 rt->rt6i_gateway = ort->rt6i_gateway;
2314 rt->rt6i_flags = ort->rt6i_flags;
2315 rt6_set_from(rt, ort);
2316 rt->rt6i_metric = ort->rt6i_metric;
2317 #ifdef CONFIG_IPV6_SUBTREES
2318 rt->rt6i_src = ort->rt6i_src;
2320 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2321 rt->rt6i_table = ort->rt6i_table;
2322 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2325 #ifdef CONFIG_IPV6_ROUTE_INFO
2326 static struct rt6_info *rt6_get_route_info(struct net *net,
2327 const struct in6_addr *prefix, int prefixlen,
2328 const struct in6_addr *gwaddr, int ifindex)
2330 struct fib6_node *fn;
2331 struct rt6_info *rt = NULL;
2332 struct fib6_table *table;
2334 table = fib6_get_table(net, RT6_TABLE_INFO);
2338 read_lock_bh(&table->tb6_lock);
2339 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2343 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2344 if (rt->dst.dev->ifindex != ifindex)
2346 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2348 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2354 read_unlock_bh(&table->tb6_lock);
2358 static struct rt6_info *rt6_add_route_info(struct net *net,
2359 const struct in6_addr *prefix, int prefixlen,
2360 const struct in6_addr *gwaddr, int ifindex,
2363 struct fib6_config cfg = {
2364 .fc_metric = IP6_RT_PRIO_USER,
2365 .fc_ifindex = ifindex,
2366 .fc_dst_len = prefixlen,
2367 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2368 RTF_UP | RTF_PREF(pref),
2369 .fc_nlinfo.portid = 0,
2370 .fc_nlinfo.nlh = NULL,
2371 .fc_nlinfo.nl_net = net,
2374 cfg.fc_table = l3mdev_fib_table_by_index(net, ifindex) ? : RT6_TABLE_INFO;
2375 cfg.fc_dst = *prefix;
2376 cfg.fc_gateway = *gwaddr;
2378 /* We should treat it as a default route if prefix length is 0. */
2380 cfg.fc_flags |= RTF_DEFAULT;
2382 ip6_route_add(&cfg);
2384 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2388 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2390 struct rt6_info *rt;
2391 struct fib6_table *table;
2393 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2397 read_lock_bh(&table->tb6_lock);
2398 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2399 if (dev == rt->dst.dev &&
2400 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2401 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2406 read_unlock_bh(&table->tb6_lock);
2410 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2411 struct net_device *dev,
2414 struct fib6_config cfg = {
2415 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2416 .fc_metric = IP6_RT_PRIO_USER,
2417 .fc_ifindex = dev->ifindex,
2418 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2419 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2420 .fc_nlinfo.portid = 0,
2421 .fc_nlinfo.nlh = NULL,
2422 .fc_nlinfo.nl_net = dev_net(dev),
2425 cfg.fc_gateway = *gwaddr;
2427 ip6_route_add(&cfg);
2429 return rt6_get_dflt_router(gwaddr, dev);
2432 void rt6_purge_dflt_routers(struct net *net)
2434 struct rt6_info *rt;
2435 struct fib6_table *table;
2437 /* NOTE: Keep consistent with rt6_get_dflt_router */
2438 table = fib6_get_table(net, RT6_TABLE_DFLT);
2443 read_lock_bh(&table->tb6_lock);
2444 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2445 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2446 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2448 read_unlock_bh(&table->tb6_lock);
2453 read_unlock_bh(&table->tb6_lock);
2456 static void rtmsg_to_fib6_config(struct net *net,
2457 struct in6_rtmsg *rtmsg,
2458 struct fib6_config *cfg)
2460 memset(cfg, 0, sizeof(*cfg));
2462 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2464 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2465 cfg->fc_metric = rtmsg->rtmsg_metric;
2466 cfg->fc_expires = rtmsg->rtmsg_info;
2467 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2468 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2469 cfg->fc_flags = rtmsg->rtmsg_flags;
2471 cfg->fc_nlinfo.nl_net = net;
2473 cfg->fc_dst = rtmsg->rtmsg_dst;
2474 cfg->fc_src = rtmsg->rtmsg_src;
2475 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2478 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2480 struct fib6_config cfg;
2481 struct in6_rtmsg rtmsg;
2485 case SIOCADDRT: /* Add a route */
2486 case SIOCDELRT: /* Delete a route */
2487 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2489 err = copy_from_user(&rtmsg, arg,
2490 sizeof(struct in6_rtmsg));
2494 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2499 err = ip6_route_add(&cfg);
2502 err = ip6_route_del(&cfg);
2516 * Drop the packet on the floor
2519 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2522 struct dst_entry *dst = skb_dst(skb);
2523 switch (ipstats_mib_noroutes) {
2524 case IPSTATS_MIB_INNOROUTES:
2525 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2526 if (type == IPV6_ADDR_ANY) {
2527 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2528 IPSTATS_MIB_INADDRERRORS);
2532 case IPSTATS_MIB_OUTNOROUTES:
2533 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2534 ipstats_mib_noroutes);
2537 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2542 static int ip6_pkt_discard(struct sk_buff *skb)
2544 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2547 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2549 skb->dev = skb_dst(skb)->dev;
2550 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2553 static int ip6_pkt_prohibit(struct sk_buff *skb)
2555 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2558 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2560 skb->dev = skb_dst(skb)->dev;
2561 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2565 * Allocate a dst for local (unicast / anycast) address.
2568 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2569 const struct in6_addr *addr,
2573 struct net *net = dev_net(idev->dev);
2574 struct net_device *dev = net->loopback_dev;
2575 struct rt6_info *rt;
2577 /* use L3 Master device as loopback for host routes if device
2578 * is enslaved and address is not link local or multicast
2580 if (!rt6_need_strict(addr))
2581 dev = l3mdev_master_dev_rcu(idev->dev) ? : dev;
2583 rt = ip6_dst_alloc(net, dev, DST_NOCOUNT);
2585 return ERR_PTR(-ENOMEM);
2589 rt->dst.flags |= DST_HOST;
2590 rt->dst.input = ip6_input;
2591 rt->dst.output = ip6_output;
2592 rt->rt6i_idev = idev;
2594 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2596 rt->rt6i_flags |= RTF_ANYCAST;
2598 rt->rt6i_flags |= RTF_LOCAL;
2600 rt->rt6i_gateway = *addr;
2601 rt->rt6i_dst.addr = *addr;
2602 rt->rt6i_dst.plen = 128;
2603 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2604 rt->rt6i_table = fib6_get_table(net, tb_id);
2605 rt->dst.flags |= DST_NOCACHE;
2607 atomic_set(&rt->dst.__refcnt, 1);
2612 /* remove deleted ip from prefsrc entries */
2613 struct arg_dev_net_ip {
2614 struct net_device *dev;
2616 struct in6_addr *addr;
2619 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2621 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2622 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2623 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2625 if (((void *)rt->dst.dev == dev || !dev) &&
2626 rt != net->ipv6.ip6_null_entry &&
2627 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2628 /* remove prefsrc entry */
2629 rt->rt6i_prefsrc.plen = 0;
2634 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2636 struct net *net = dev_net(ifp->idev->dev);
2637 struct arg_dev_net_ip adni = {
2638 .dev = ifp->idev->dev,
2642 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2645 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2646 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2648 /* Remove routers and update dst entries when gateway turn into host. */
2649 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2651 struct in6_addr *gateway = (struct in6_addr *)arg;
2653 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2654 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2655 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2661 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2663 fib6_clean_all(net, fib6_clean_tohost, gateway);
2666 struct arg_dev_net {
2667 struct net_device *dev;
2671 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2673 const struct arg_dev_net *adn = arg;
2674 const struct net_device *dev = adn->dev;
2676 if ((rt->dst.dev == dev || !dev) &&
2677 rt != adn->net->ipv6.ip6_null_entry)
2683 void rt6_ifdown(struct net *net, struct net_device *dev)
2685 struct arg_dev_net adn = {
2690 fib6_clean_all(net, fib6_ifdown, &adn);
2691 icmp6_clean_all(fib6_ifdown, &adn);
2693 rt6_uncached_list_flush_dev(net, dev);
2696 struct rt6_mtu_change_arg {
2697 struct net_device *dev;
2701 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2703 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2704 struct inet6_dev *idev;
2706 /* In IPv6 pmtu discovery is not optional,
2707 so that RTAX_MTU lock cannot disable it.
2708 We still use this lock to block changes
2709 caused by addrconf/ndisc.
2712 idev = __in6_dev_get(arg->dev);
2716 /* For administrative MTU increase, there is no way to discover
2717 IPv6 PMTU increase, so PMTU increase should be updated here.
2718 Since RFC 1981 doesn't include administrative MTU increase
2719 update PMTU increase is a MUST. (i.e. jumbo frame)
2722 If new MTU is less than route PMTU, this new MTU will be the
2723 lowest MTU in the path, update the route PMTU to reflect PMTU
2724 decreases; if new MTU is greater than route PMTU, and the
2725 old MTU is the lowest MTU in the path, update the route PMTU
2726 to reflect the increase. In this case if the other nodes' MTU
2727 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2730 if (rt->dst.dev == arg->dev &&
2731 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2732 if (rt->rt6i_flags & RTF_CACHE) {
2733 /* For RTF_CACHE with rt6i_pmtu == 0
2734 * (i.e. a redirected route),
2735 * the metrics of its rt->dst.from has already
2738 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2739 rt->rt6i_pmtu = arg->mtu;
2740 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2741 (dst_mtu(&rt->dst) < arg->mtu &&
2742 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2743 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2749 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2751 struct rt6_mtu_change_arg arg = {
2756 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2759 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2760 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2761 [RTA_OIF] = { .type = NLA_U32 },
2762 [RTA_IIF] = { .type = NLA_U32 },
2763 [RTA_PRIORITY] = { .type = NLA_U32 },
2764 [RTA_METRICS] = { .type = NLA_NESTED },
2765 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2766 [RTA_PREF] = { .type = NLA_U8 },
2767 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2768 [RTA_ENCAP] = { .type = NLA_NESTED },
2769 [RTA_EXPIRES] = { .type = NLA_U32 },
2772 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2773 struct fib6_config *cfg)
2776 struct nlattr *tb[RTA_MAX+1];
2780 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2785 rtm = nlmsg_data(nlh);
2786 memset(cfg, 0, sizeof(*cfg));
2788 cfg->fc_table = rtm->rtm_table;
2789 cfg->fc_dst_len = rtm->rtm_dst_len;
2790 cfg->fc_src_len = rtm->rtm_src_len;
2791 cfg->fc_flags = RTF_UP;
2792 cfg->fc_protocol = rtm->rtm_protocol;
2793 cfg->fc_type = rtm->rtm_type;
2795 if (rtm->rtm_type == RTN_UNREACHABLE ||
2796 rtm->rtm_type == RTN_BLACKHOLE ||
2797 rtm->rtm_type == RTN_PROHIBIT ||
2798 rtm->rtm_type == RTN_THROW)
2799 cfg->fc_flags |= RTF_REJECT;
2801 if (rtm->rtm_type == RTN_LOCAL)
2802 cfg->fc_flags |= RTF_LOCAL;
2804 if (rtm->rtm_flags & RTM_F_CLONED)
2805 cfg->fc_flags |= RTF_CACHE;
2807 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2808 cfg->fc_nlinfo.nlh = nlh;
2809 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2811 if (tb[RTA_GATEWAY]) {
2812 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2813 cfg->fc_flags |= RTF_GATEWAY;
2817 int plen = (rtm->rtm_dst_len + 7) >> 3;
2819 if (nla_len(tb[RTA_DST]) < plen)
2822 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2826 int plen = (rtm->rtm_src_len + 7) >> 3;
2828 if (nla_len(tb[RTA_SRC]) < plen)
2831 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2834 if (tb[RTA_PREFSRC])
2835 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2838 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2840 if (tb[RTA_PRIORITY])
2841 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2843 if (tb[RTA_METRICS]) {
2844 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2845 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2849 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2851 if (tb[RTA_MULTIPATH]) {
2852 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2853 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2857 pref = nla_get_u8(tb[RTA_PREF]);
2858 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2859 pref != ICMPV6_ROUTER_PREF_HIGH)
2860 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2861 cfg->fc_flags |= RTF_PREF(pref);
2865 cfg->fc_encap = tb[RTA_ENCAP];
2867 if (tb[RTA_ENCAP_TYPE])
2868 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2870 if (tb[RTA_EXPIRES]) {
2871 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
2873 if (addrconf_finite_timeout(timeout)) {
2874 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
2875 cfg->fc_flags |= RTF_EXPIRES;
2885 struct rt6_info *rt6_info;
2886 struct fib6_config r_cfg;
2887 struct mx6_config mxc;
2888 struct list_head next;
2891 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2895 list_for_each_entry(nh, rt6_nh_list, next) {
2896 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2897 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2898 nh->r_cfg.fc_ifindex);
2902 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2903 struct rt6_info *rt, struct fib6_config *r_cfg)
2906 struct rt6_info *rtnh;
2909 list_for_each_entry(nh, rt6_nh_list, next) {
2910 /* check if rt6_info already exists */
2911 rtnh = nh->rt6_info;
2913 if (rtnh->dst.dev == rt->dst.dev &&
2914 rtnh->rt6i_idev == rt->rt6i_idev &&
2915 ipv6_addr_equal(&rtnh->rt6i_gateway,
2920 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2924 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2929 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2930 list_add_tail(&nh->next, rt6_nh_list);
2935 static int ip6_route_multipath_add(struct fib6_config *cfg)
2937 struct fib6_config r_cfg;
2938 struct rtnexthop *rtnh;
2939 struct rt6_info *rt;
2940 struct rt6_nh *err_nh;
2941 struct rt6_nh *nh, *nh_safe;
2946 int replace = (cfg->fc_nlinfo.nlh &&
2947 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2948 LIST_HEAD(rt6_nh_list);
2950 remaining = cfg->fc_mp_len;
2951 rtnh = (struct rtnexthop *)cfg->fc_mp;
2953 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2954 * rt6_info structs per nexthop
2956 while (rtnh_ok(rtnh, remaining)) {
2957 memcpy(&r_cfg, cfg, sizeof(*cfg));
2958 if (rtnh->rtnh_ifindex)
2959 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2961 attrlen = rtnh_attrlen(rtnh);
2963 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2965 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2967 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2968 r_cfg.fc_flags |= RTF_GATEWAY;
2970 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2971 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2973 r_cfg.fc_encap_type = nla_get_u16(nla);
2976 rt = ip6_route_info_create(&r_cfg);
2983 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2989 rtnh = rtnh_next(rtnh, &remaining);
2993 list_for_each_entry(nh, &rt6_nh_list, next) {
2994 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2995 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2996 nh->rt6_info = NULL;
2999 ip6_print_replace_route_err(&rt6_nh_list);
3004 /* Because each route is added like a single route we remove
3005 * these flags after the first nexthop: if there is a collision,
3006 * we have already failed to add the first nexthop:
3007 * fib6_add_rt2node() has rejected it; when replacing, old
3008 * nexthops have been replaced by first new, the rest should
3011 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
3019 /* Delete routes that were already added */
3020 list_for_each_entry(nh, &rt6_nh_list, next) {
3023 ip6_route_del(&nh->r_cfg);
3027 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3029 dst_free(&nh->rt6_info->dst);
3031 list_del(&nh->next);
3038 static int ip6_route_multipath_del(struct fib6_config *cfg)
3040 struct fib6_config r_cfg;
3041 struct rtnexthop *rtnh;
3044 int err = 1, last_err = 0;
3046 remaining = cfg->fc_mp_len;
3047 rtnh = (struct rtnexthop *)cfg->fc_mp;
3049 /* Parse a Multipath Entry */
3050 while (rtnh_ok(rtnh, remaining)) {
3051 memcpy(&r_cfg, cfg, sizeof(*cfg));
3052 if (rtnh->rtnh_ifindex)
3053 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3055 attrlen = rtnh_attrlen(rtnh);
3057 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3059 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3061 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3062 r_cfg.fc_flags |= RTF_GATEWAY;
3065 err = ip6_route_del(&r_cfg);
3069 rtnh = rtnh_next(rtnh, &remaining);
3075 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3077 struct fib6_config cfg;
3080 err = rtm_to_fib6_config(skb, nlh, &cfg);
3085 return ip6_route_multipath_del(&cfg);
3087 return ip6_route_del(&cfg);
3090 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3092 struct fib6_config cfg;
3095 err = rtm_to_fib6_config(skb, nlh, &cfg);
3100 return ip6_route_multipath_add(&cfg);
3102 return ip6_route_add(&cfg);
3105 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3107 return NLMSG_ALIGN(sizeof(struct rtmsg))
3108 + nla_total_size(16) /* RTA_SRC */
3109 + nla_total_size(16) /* RTA_DST */
3110 + nla_total_size(16) /* RTA_GATEWAY */
3111 + nla_total_size(16) /* RTA_PREFSRC */
3112 + nla_total_size(4) /* RTA_TABLE */
3113 + nla_total_size(4) /* RTA_IIF */
3114 + nla_total_size(4) /* RTA_OIF */
3115 + nla_total_size(4) /* RTA_PRIORITY */
3116 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3117 + nla_total_size(sizeof(struct rta_cacheinfo))
3118 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3119 + nla_total_size(1) /* RTA_PREF */
3120 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3123 static int rt6_fill_node(struct net *net,
3124 struct sk_buff *skb, struct rt6_info *rt,
3125 struct in6_addr *dst, struct in6_addr *src,
3126 int iif, int type, u32 portid, u32 seq,
3127 int prefix, int nowait, unsigned int flags)
3129 u32 metrics[RTAX_MAX];
3131 struct nlmsghdr *nlh;
3135 if (prefix) { /* user wants prefix routes only */
3136 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3137 /* success since this is not a prefix route */
3142 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3146 rtm = nlmsg_data(nlh);
3147 rtm->rtm_family = AF_INET6;
3148 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3149 rtm->rtm_src_len = rt->rt6i_src.plen;
3152 table = rt->rt6i_table->tb6_id;
3154 table = RT6_TABLE_UNSPEC;
3155 rtm->rtm_table = table;
3156 if (nla_put_u32(skb, RTA_TABLE, table))
3157 goto nla_put_failure;
3158 if (rt->rt6i_flags & RTF_REJECT) {
3159 switch (rt->dst.error) {
3161 rtm->rtm_type = RTN_BLACKHOLE;
3164 rtm->rtm_type = RTN_PROHIBIT;
3167 rtm->rtm_type = RTN_THROW;
3170 rtm->rtm_type = RTN_UNREACHABLE;
3174 else if (rt->rt6i_flags & RTF_LOCAL)
3175 rtm->rtm_type = RTN_LOCAL;
3176 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3177 rtm->rtm_type = RTN_LOCAL;
3179 rtm->rtm_type = RTN_UNICAST;
3181 if (!netif_carrier_ok(rt->dst.dev)) {
3182 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3183 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3184 rtm->rtm_flags |= RTNH_F_DEAD;
3186 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3187 rtm->rtm_protocol = rt->rt6i_protocol;
3188 if (rt->rt6i_flags & RTF_DYNAMIC)
3189 rtm->rtm_protocol = RTPROT_REDIRECT;
3190 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3191 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3192 rtm->rtm_protocol = RTPROT_RA;
3194 rtm->rtm_protocol = RTPROT_KERNEL;
3197 if (rt->rt6i_flags & RTF_CACHE)
3198 rtm->rtm_flags |= RTM_F_CLONED;
3201 if (nla_put_in6_addr(skb, RTA_DST, dst))
3202 goto nla_put_failure;
3203 rtm->rtm_dst_len = 128;
3204 } else if (rtm->rtm_dst_len)
3205 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3206 goto nla_put_failure;
3207 #ifdef CONFIG_IPV6_SUBTREES
3209 if (nla_put_in6_addr(skb, RTA_SRC, src))
3210 goto nla_put_failure;
3211 rtm->rtm_src_len = 128;
3212 } else if (rtm->rtm_src_len &&
3213 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3214 goto nla_put_failure;
3217 #ifdef CONFIG_IPV6_MROUTE
3218 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3219 int err = ip6mr_get_route(net, skb, rtm, nowait,
3226 goto nla_put_failure;
3228 if (err == -EMSGSIZE)
3229 goto nla_put_failure;
3234 if (nla_put_u32(skb, RTA_IIF, iif))
3235 goto nla_put_failure;
3237 struct in6_addr saddr_buf;
3238 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3239 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3240 goto nla_put_failure;
3243 if (rt->rt6i_prefsrc.plen) {
3244 struct in6_addr saddr_buf;
3245 saddr_buf = rt->rt6i_prefsrc.addr;
3246 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3247 goto nla_put_failure;
3250 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3252 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3253 if (rtnetlink_put_metrics(skb, metrics) < 0)
3254 goto nla_put_failure;
3256 if (rt->rt6i_flags & RTF_GATEWAY) {
3257 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3258 goto nla_put_failure;
3262 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3263 goto nla_put_failure;
3264 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3265 goto nla_put_failure;
3267 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3269 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3270 goto nla_put_failure;
3272 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3273 goto nla_put_failure;
3275 lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3277 nlmsg_end(skb, nlh);
3281 nlmsg_cancel(skb, nlh);
3285 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3287 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3290 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3291 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3292 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3296 return rt6_fill_node(arg->net,
3297 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3298 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3299 prefix, 0, NLM_F_MULTI);
3302 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3304 struct net *net = sock_net(in_skb->sk);
3305 struct nlattr *tb[RTA_MAX+1];
3306 struct rt6_info *rt;
3307 struct sk_buff *skb;
3310 int err, iif = 0, oif = 0;
3312 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3317 memset(&fl6, 0, sizeof(fl6));
3318 rtm = nlmsg_data(nlh);
3319 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
3322 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3325 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3329 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3332 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3336 iif = nla_get_u32(tb[RTA_IIF]);
3339 oif = nla_get_u32(tb[RTA_OIF]);
3342 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3345 struct net_device *dev;
3348 dev = __dev_get_by_index(net, iif);
3354 fl6.flowi6_iif = iif;
3356 if (!ipv6_addr_any(&fl6.saddr))
3357 flags |= RT6_LOOKUP_F_HAS_SADDR;
3359 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3362 fl6.flowi6_oif = oif;
3364 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3367 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3374 /* Reserve room for dummy headers, this skb can pass
3375 through good chunk of routing engine.
3377 skb_reset_mac_header(skb);
3378 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3380 skb_dst_set(skb, &rt->dst);
3382 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3383 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3384 nlh->nlmsg_seq, 0, 0, 0);
3390 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3395 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3396 unsigned int nlm_flags)
3398 struct sk_buff *skb;
3399 struct net *net = info->nl_net;
3404 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3406 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3410 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3411 event, info->portid, seq, 0, 0, nlm_flags);
3413 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3414 WARN_ON(err == -EMSGSIZE);
3418 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3419 info->nlh, gfp_any());
3423 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3426 static int ip6_route_dev_notify(struct notifier_block *this,
3427 unsigned long event, void *ptr)
3429 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3430 struct net *net = dev_net(dev);
3432 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3433 net->ipv6.ip6_null_entry->dst.dev = dev;
3434 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3435 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3436 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3437 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3438 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3439 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3450 #ifdef CONFIG_PROC_FS
3452 static const struct file_operations ipv6_route_proc_fops = {
3453 .owner = THIS_MODULE,
3454 .open = ipv6_route_open,
3456 .llseek = seq_lseek,
3457 .release = seq_release_net,
3460 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3462 struct net *net = (struct net *)seq->private;
3463 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3464 net->ipv6.rt6_stats->fib_nodes,
3465 net->ipv6.rt6_stats->fib_route_nodes,
3466 net->ipv6.rt6_stats->fib_rt_alloc,
3467 net->ipv6.rt6_stats->fib_rt_entries,
3468 net->ipv6.rt6_stats->fib_rt_cache,
3469 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3470 net->ipv6.rt6_stats->fib_discarded_routes);
3475 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3477 return single_open_net(inode, file, rt6_stats_seq_show);
3480 static const struct file_operations rt6_stats_seq_fops = {
3481 .owner = THIS_MODULE,
3482 .open = rt6_stats_seq_open,
3484 .llseek = seq_lseek,
3485 .release = single_release_net,
3487 #endif /* CONFIG_PROC_FS */
3489 #ifdef CONFIG_SYSCTL
3492 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3493 void __user *buffer, size_t *lenp, loff_t *ppos)
3500 net = (struct net *)ctl->extra1;
3501 delay = net->ipv6.sysctl.flush_delay;
3502 proc_dointvec(ctl, write, buffer, lenp, ppos);
3503 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3507 struct ctl_table ipv6_route_table_template[] = {
3509 .procname = "flush",
3510 .data = &init_net.ipv6.sysctl.flush_delay,
3511 .maxlen = sizeof(int),
3513 .proc_handler = ipv6_sysctl_rtcache_flush
3516 .procname = "gc_thresh",
3517 .data = &ip6_dst_ops_template.gc_thresh,
3518 .maxlen = sizeof(int),
3520 .proc_handler = proc_dointvec,
3523 .procname = "max_size",
3524 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3525 .maxlen = sizeof(int),
3527 .proc_handler = proc_dointvec,
3530 .procname = "gc_min_interval",
3531 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3532 .maxlen = sizeof(int),
3534 .proc_handler = proc_dointvec_jiffies,
3537 .procname = "gc_timeout",
3538 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3539 .maxlen = sizeof(int),
3541 .proc_handler = proc_dointvec_jiffies,
3544 .procname = "gc_interval",
3545 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3546 .maxlen = sizeof(int),
3548 .proc_handler = proc_dointvec_jiffies,
3551 .procname = "gc_elasticity",
3552 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3553 .maxlen = sizeof(int),
3555 .proc_handler = proc_dointvec,
3558 .procname = "mtu_expires",
3559 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3560 .maxlen = sizeof(int),
3562 .proc_handler = proc_dointvec_jiffies,
3565 .procname = "min_adv_mss",
3566 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3567 .maxlen = sizeof(int),
3569 .proc_handler = proc_dointvec,
3572 .procname = "gc_min_interval_ms",
3573 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3574 .maxlen = sizeof(int),
3576 .proc_handler = proc_dointvec_ms_jiffies,
3581 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3583 struct ctl_table *table;
3585 table = kmemdup(ipv6_route_table_template,
3586 sizeof(ipv6_route_table_template),
3590 table[0].data = &net->ipv6.sysctl.flush_delay;
3591 table[0].extra1 = net;
3592 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3593 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3594 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3595 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3596 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3597 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3598 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3599 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3600 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3602 /* Don't export sysctls to unprivileged users */
3603 if (net->user_ns != &init_user_ns)
3604 table[0].procname = NULL;
3611 static int __net_init ip6_route_net_init(struct net *net)
3615 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3616 sizeof(net->ipv6.ip6_dst_ops));
3618 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3619 goto out_ip6_dst_ops;
3621 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3622 sizeof(*net->ipv6.ip6_null_entry),
3624 if (!net->ipv6.ip6_null_entry)
3625 goto out_ip6_dst_entries;
3626 net->ipv6.ip6_null_entry->dst.path =
3627 (struct dst_entry *)net->ipv6.ip6_null_entry;
3628 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3629 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3630 ip6_template_metrics, true);
3632 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3633 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3634 sizeof(*net->ipv6.ip6_prohibit_entry),
3636 if (!net->ipv6.ip6_prohibit_entry)
3637 goto out_ip6_null_entry;
3638 net->ipv6.ip6_prohibit_entry->dst.path =
3639 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3640 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3641 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3642 ip6_template_metrics, true);
3644 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3645 sizeof(*net->ipv6.ip6_blk_hole_entry),
3647 if (!net->ipv6.ip6_blk_hole_entry)
3648 goto out_ip6_prohibit_entry;
3649 net->ipv6.ip6_blk_hole_entry->dst.path =
3650 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3651 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3652 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3653 ip6_template_metrics, true);
3656 net->ipv6.sysctl.flush_delay = 0;
3657 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3658 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3659 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3660 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3661 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3662 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3663 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3665 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3671 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3672 out_ip6_prohibit_entry:
3673 kfree(net->ipv6.ip6_prohibit_entry);
3675 kfree(net->ipv6.ip6_null_entry);
3677 out_ip6_dst_entries:
3678 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3683 static void __net_exit ip6_route_net_exit(struct net *net)
3685 kfree(net->ipv6.ip6_null_entry);
3686 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3687 kfree(net->ipv6.ip6_prohibit_entry);
3688 kfree(net->ipv6.ip6_blk_hole_entry);
3690 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3693 static int __net_init ip6_route_net_init_late(struct net *net)
3695 #ifdef CONFIG_PROC_FS
3696 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3697 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3702 static void __net_exit ip6_route_net_exit_late(struct net *net)
3704 #ifdef CONFIG_PROC_FS
3705 remove_proc_entry("ipv6_route", net->proc_net);
3706 remove_proc_entry("rt6_stats", net->proc_net);
3710 static struct pernet_operations ip6_route_net_ops = {
3711 .init = ip6_route_net_init,
3712 .exit = ip6_route_net_exit,
3715 static int __net_init ipv6_inetpeer_init(struct net *net)
3717 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3721 inet_peer_base_init(bp);
3722 net->ipv6.peers = bp;
3726 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3728 struct inet_peer_base *bp = net->ipv6.peers;
3730 net->ipv6.peers = NULL;
3731 inetpeer_invalidate_tree(bp);
3735 static struct pernet_operations ipv6_inetpeer_ops = {
3736 .init = ipv6_inetpeer_init,
3737 .exit = ipv6_inetpeer_exit,
3740 static struct pernet_operations ip6_route_net_late_ops = {
3741 .init = ip6_route_net_init_late,
3742 .exit = ip6_route_net_exit_late,
3745 static struct notifier_block ip6_route_dev_notifier = {
3746 .notifier_call = ip6_route_dev_notify,
3750 int __init ip6_route_init(void)
3756 ip6_dst_ops_template.kmem_cachep =
3757 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3758 SLAB_HWCACHE_ALIGN, NULL);
3759 if (!ip6_dst_ops_template.kmem_cachep)
3762 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3764 goto out_kmem_cache;
3766 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3768 goto out_dst_entries;
3770 ret = register_pernet_subsys(&ip6_route_net_ops);
3772 goto out_register_inetpeer;
3774 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3776 /* Registering of the loopback is done before this portion of code,
3777 * the loopback reference in rt6_info will not be taken, do it
3778 * manually for init_net */
3779 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3780 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3781 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3782 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3783 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3784 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3785 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3789 goto out_register_subsys;
3795 ret = fib6_rules_init();
3799 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3801 goto fib6_rules_init;
3804 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3805 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3806 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3807 goto out_register_late_subsys;
3809 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3811 goto out_register_late_subsys;
3813 for_each_possible_cpu(cpu) {
3814 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3816 INIT_LIST_HEAD(&ul->head);
3817 spin_lock_init(&ul->lock);
3823 out_register_late_subsys:
3824 unregister_pernet_subsys(&ip6_route_net_late_ops);
3826 fib6_rules_cleanup();
3831 out_register_subsys:
3832 unregister_pernet_subsys(&ip6_route_net_ops);
3833 out_register_inetpeer:
3834 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3836 dst_entries_destroy(&ip6_dst_blackhole_ops);
3838 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3842 void ip6_route_cleanup(void)
3844 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3845 unregister_pernet_subsys(&ip6_route_net_late_ops);
3846 fib6_rules_cleanup();
3849 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3850 unregister_pernet_subsys(&ip6_route_net_ops);
3851 dst_entries_destroy(&ip6_dst_blackhole_ops);
3852 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);