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 <linux/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);
101 static size_t rt6_nlmsg_size(struct rt6_info *rt);
102 static int rt6_fill_node(struct net *net,
103 struct sk_buff *skb, struct rt6_info *rt,
104 struct in6_addr *dst, struct in6_addr *src,
105 int iif, int type, u32 portid, u32 seq,
108 #ifdef CONFIG_IPV6_ROUTE_INFO
109 static struct rt6_info *rt6_add_route_info(struct net *net,
110 const struct in6_addr *prefix, int prefixlen,
111 const struct in6_addr *gwaddr,
112 struct net_device *dev,
114 static struct rt6_info *rt6_get_route_info(struct net *net,
115 const struct in6_addr *prefix, int prefixlen,
116 const struct in6_addr *gwaddr,
117 struct net_device *dev);
120 struct uncached_list {
122 struct list_head head;
125 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
127 static void rt6_uncached_list_add(struct rt6_info *rt)
129 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
131 rt->dst.flags |= DST_NOCACHE;
132 rt->rt6i_uncached_list = ul;
134 spin_lock_bh(&ul->lock);
135 list_add_tail(&rt->rt6i_uncached, &ul->head);
136 spin_unlock_bh(&ul->lock);
139 static void rt6_uncached_list_del(struct rt6_info *rt)
141 if (!list_empty(&rt->rt6i_uncached)) {
142 struct uncached_list *ul = rt->rt6i_uncached_list;
144 spin_lock_bh(&ul->lock);
145 list_del(&rt->rt6i_uncached);
146 spin_unlock_bh(&ul->lock);
150 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
152 struct net_device *loopback_dev = net->loopback_dev;
155 if (dev == loopback_dev)
158 for_each_possible_cpu(cpu) {
159 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
162 spin_lock_bh(&ul->lock);
163 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
164 struct inet6_dev *rt_idev = rt->rt6i_idev;
165 struct net_device *rt_dev = rt->dst.dev;
167 if (rt_idev->dev == dev) {
168 rt->rt6i_idev = in6_dev_get(loopback_dev);
169 in6_dev_put(rt_idev);
173 rt->dst.dev = loopback_dev;
174 dev_hold(rt->dst.dev);
178 spin_unlock_bh(&ul->lock);
182 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
184 return dst_metrics_write_ptr(rt->dst.from);
187 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
189 struct rt6_info *rt = (struct rt6_info *)dst;
191 if (rt->rt6i_flags & RTF_PCPU)
192 return rt6_pcpu_cow_metrics(rt);
193 else if (rt->rt6i_flags & RTF_CACHE)
196 return dst_cow_metrics_generic(dst, old);
199 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
203 struct in6_addr *p = &rt->rt6i_gateway;
205 if (!ipv6_addr_any(p))
206 return (const void *) p;
208 return &ipv6_hdr(skb)->daddr;
212 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
216 struct rt6_info *rt = (struct rt6_info *) dst;
219 daddr = choose_neigh_daddr(rt, skb, daddr);
220 n = __ipv6_neigh_lookup(dst->dev, daddr);
223 return neigh_create(&nd_tbl, daddr, dst->dev);
226 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
228 struct net_device *dev = dst->dev;
229 struct rt6_info *rt = (struct rt6_info *)dst;
231 daddr = choose_neigh_daddr(rt, NULL, daddr);
234 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
236 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
238 __ipv6_confirm_neigh(dev, daddr);
241 static struct dst_ops ip6_dst_ops_template = {
245 .check = ip6_dst_check,
246 .default_advmss = ip6_default_advmss,
248 .cow_metrics = ipv6_cow_metrics,
249 .destroy = ip6_dst_destroy,
250 .ifdown = ip6_dst_ifdown,
251 .negative_advice = ip6_negative_advice,
252 .link_failure = ip6_link_failure,
253 .update_pmtu = ip6_rt_update_pmtu,
254 .redirect = rt6_do_redirect,
255 .local_out = __ip6_local_out,
256 .neigh_lookup = ip6_neigh_lookup,
257 .confirm_neigh = ip6_confirm_neigh,
260 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
262 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
264 return mtu ? : dst->dev->mtu;
267 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
268 struct sk_buff *skb, u32 mtu)
272 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
277 static struct dst_ops ip6_dst_blackhole_ops = {
279 .destroy = ip6_dst_destroy,
280 .check = ip6_dst_check,
281 .mtu = ip6_blackhole_mtu,
282 .default_advmss = ip6_default_advmss,
283 .update_pmtu = ip6_rt_blackhole_update_pmtu,
284 .redirect = ip6_rt_blackhole_redirect,
285 .cow_metrics = dst_cow_metrics_generic,
286 .neigh_lookup = ip6_neigh_lookup,
289 static const u32 ip6_template_metrics[RTAX_MAX] = {
290 [RTAX_HOPLIMIT - 1] = 0,
293 static const struct rt6_info ip6_null_entry_template = {
295 .__refcnt = ATOMIC_INIT(1),
297 .obsolete = DST_OBSOLETE_FORCE_CHK,
298 .error = -ENETUNREACH,
299 .input = ip6_pkt_discard,
300 .output = ip6_pkt_discard_out,
302 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
303 .rt6i_protocol = RTPROT_KERNEL,
304 .rt6i_metric = ~(u32) 0,
305 .rt6i_ref = ATOMIC_INIT(1),
308 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
310 static const struct rt6_info ip6_prohibit_entry_template = {
312 .__refcnt = ATOMIC_INIT(1),
314 .obsolete = DST_OBSOLETE_FORCE_CHK,
316 .input = ip6_pkt_prohibit,
317 .output = ip6_pkt_prohibit_out,
319 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
320 .rt6i_protocol = RTPROT_KERNEL,
321 .rt6i_metric = ~(u32) 0,
322 .rt6i_ref = ATOMIC_INIT(1),
325 static const struct rt6_info ip6_blk_hole_entry_template = {
327 .__refcnt = ATOMIC_INIT(1),
329 .obsolete = DST_OBSOLETE_FORCE_CHK,
331 .input = dst_discard,
332 .output = dst_discard_out,
334 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
335 .rt6i_protocol = RTPROT_KERNEL,
336 .rt6i_metric = ~(u32) 0,
337 .rt6i_ref = ATOMIC_INIT(1),
342 static void rt6_info_init(struct rt6_info *rt)
344 struct dst_entry *dst = &rt->dst;
346 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
347 INIT_LIST_HEAD(&rt->rt6i_siblings);
348 INIT_LIST_HEAD(&rt->rt6i_uncached);
351 /* allocate dst with ip6_dst_ops */
352 static struct rt6_info *__ip6_dst_alloc(struct net *net,
353 struct net_device *dev,
356 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
357 0, DST_OBSOLETE_FORCE_CHK, flags);
365 struct rt6_info *ip6_dst_alloc(struct net *net,
366 struct net_device *dev,
369 struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
372 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
376 for_each_possible_cpu(cpu) {
379 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
380 /* no one shares rt */
384 dst_destroy((struct dst_entry *)rt);
391 EXPORT_SYMBOL(ip6_dst_alloc);
393 static void ip6_dst_destroy(struct dst_entry *dst)
395 struct rt6_info *rt = (struct rt6_info *)dst;
396 struct dst_entry *from = dst->from;
397 struct inet6_dev *idev;
399 dst_destroy_metrics_generic(dst);
400 free_percpu(rt->rt6i_pcpu);
401 rt6_uncached_list_del(rt);
403 idev = rt->rt6i_idev;
405 rt->rt6i_idev = NULL;
413 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
416 struct rt6_info *rt = (struct rt6_info *)dst;
417 struct inet6_dev *idev = rt->rt6i_idev;
418 struct net_device *loopback_dev =
419 dev_net(dev)->loopback_dev;
421 if (dev != loopback_dev) {
422 if (idev && idev->dev == dev) {
423 struct inet6_dev *loopback_idev =
424 in6_dev_get(loopback_dev);
426 rt->rt6i_idev = loopback_idev;
433 static bool __rt6_check_expired(const struct rt6_info *rt)
435 if (rt->rt6i_flags & RTF_EXPIRES)
436 return time_after(jiffies, rt->dst.expires);
441 static bool rt6_check_expired(const struct rt6_info *rt)
443 if (rt->rt6i_flags & RTF_EXPIRES) {
444 if (time_after(jiffies, rt->dst.expires))
446 } else if (rt->dst.from) {
447 return rt6_check_expired((struct rt6_info *) rt->dst.from);
452 /* Multipath route selection:
453 * Hash based function using packet header and flowlabel.
454 * Adapted from fib_info_hashfn()
456 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
457 const struct flowi6 *fl6)
459 return get_hash_from_flowi6(fl6) % candidate_count;
462 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
463 struct flowi6 *fl6, int oif,
466 struct rt6_info *sibling, *next_sibling;
469 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
470 /* Don't change the route, if route_choosen == 0
471 * (siblings does not include ourself)
474 list_for_each_entry_safe(sibling, next_sibling,
475 &match->rt6i_siblings, rt6i_siblings) {
477 if (route_choosen == 0) {
478 if (rt6_score_route(sibling, oif, strict) < 0)
488 * Route lookup. Any table->tb6_lock is implied.
491 static inline struct rt6_info *rt6_device_match(struct net *net,
493 const struct in6_addr *saddr,
497 struct rt6_info *local = NULL;
498 struct rt6_info *sprt;
500 if (!oif && ipv6_addr_any(saddr))
503 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
504 struct net_device *dev = sprt->dst.dev;
507 if (dev->ifindex == oif)
509 if (dev->flags & IFF_LOOPBACK) {
510 if (!sprt->rt6i_idev ||
511 sprt->rt6i_idev->dev->ifindex != oif) {
512 if (flags & RT6_LOOKUP_F_IFACE)
515 local->rt6i_idev->dev->ifindex == oif)
521 if (ipv6_chk_addr(net, saddr, dev,
522 flags & RT6_LOOKUP_F_IFACE))
531 if (flags & RT6_LOOKUP_F_IFACE)
532 return net->ipv6.ip6_null_entry;
538 #ifdef CONFIG_IPV6_ROUTER_PREF
539 struct __rt6_probe_work {
540 struct work_struct work;
541 struct in6_addr target;
542 struct net_device *dev;
545 static void rt6_probe_deferred(struct work_struct *w)
547 struct in6_addr mcaddr;
548 struct __rt6_probe_work *work =
549 container_of(w, struct __rt6_probe_work, work);
551 addrconf_addr_solict_mult(&work->target, &mcaddr);
552 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
557 static void rt6_probe(struct rt6_info *rt)
559 struct __rt6_probe_work *work;
560 struct neighbour *neigh;
562 * Okay, this does not seem to be appropriate
563 * for now, however, we need to check if it
564 * is really so; aka Router Reachability Probing.
566 * Router Reachability Probe MUST be rate-limited
567 * to no more than one per minute.
569 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
572 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
574 if (neigh->nud_state & NUD_VALID)
578 write_lock(&neigh->lock);
579 if (!(neigh->nud_state & NUD_VALID) &&
582 rt->rt6i_idev->cnf.rtr_probe_interval)) {
583 work = kmalloc(sizeof(*work), GFP_ATOMIC);
585 __neigh_set_probe_once(neigh);
587 write_unlock(&neigh->lock);
589 work = kmalloc(sizeof(*work), GFP_ATOMIC);
593 INIT_WORK(&work->work, rt6_probe_deferred);
594 work->target = rt->rt6i_gateway;
595 dev_hold(rt->dst.dev);
596 work->dev = rt->dst.dev;
597 schedule_work(&work->work);
601 rcu_read_unlock_bh();
604 static inline void rt6_probe(struct rt6_info *rt)
610 * Default Router Selection (RFC 2461 6.3.6)
612 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
614 struct net_device *dev = rt->dst.dev;
615 if (!oif || dev->ifindex == oif)
617 if ((dev->flags & IFF_LOOPBACK) &&
618 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
623 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
625 struct neighbour *neigh;
626 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
628 if (rt->rt6i_flags & RTF_NONEXTHOP ||
629 !(rt->rt6i_flags & RTF_GATEWAY))
630 return RT6_NUD_SUCCEED;
633 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
635 read_lock(&neigh->lock);
636 if (neigh->nud_state & NUD_VALID)
637 ret = RT6_NUD_SUCCEED;
638 #ifdef CONFIG_IPV6_ROUTER_PREF
639 else if (!(neigh->nud_state & NUD_FAILED))
640 ret = RT6_NUD_SUCCEED;
642 ret = RT6_NUD_FAIL_PROBE;
644 read_unlock(&neigh->lock);
646 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
647 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
649 rcu_read_unlock_bh();
654 static int rt6_score_route(struct rt6_info *rt, int oif,
659 m = rt6_check_dev(rt, oif);
660 if (!m && (strict & RT6_LOOKUP_F_IFACE))
661 return RT6_NUD_FAIL_HARD;
662 #ifdef CONFIG_IPV6_ROUTER_PREF
663 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
665 if (strict & RT6_LOOKUP_F_REACHABLE) {
666 int n = rt6_check_neigh(rt);
673 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
674 int *mpri, struct rt6_info *match,
678 bool match_do_rr = false;
679 struct inet6_dev *idev = rt->rt6i_idev;
680 struct net_device *dev = rt->dst.dev;
682 if (dev && !netif_carrier_ok(dev) &&
683 idev->cnf.ignore_routes_with_linkdown &&
684 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
687 if (rt6_check_expired(rt))
690 m = rt6_score_route(rt, oif, strict);
691 if (m == RT6_NUD_FAIL_DO_RR) {
693 m = 0; /* lowest valid score */
694 } else if (m == RT6_NUD_FAIL_HARD) {
698 if (strict & RT6_LOOKUP_F_REACHABLE)
701 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
703 *do_rr = match_do_rr;
711 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
712 struct rt6_info *rr_head,
713 u32 metric, int oif, int strict,
716 struct rt6_info *rt, *match, *cont;
721 for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
722 if (rt->rt6i_metric != metric) {
727 match = find_match(rt, oif, strict, &mpri, match, do_rr);
730 for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
731 if (rt->rt6i_metric != metric) {
736 match = find_match(rt, oif, strict, &mpri, match, do_rr);
742 for (rt = cont; rt; rt = rt->dst.rt6_next)
743 match = find_match(rt, oif, strict, &mpri, match, do_rr);
748 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
750 struct rt6_info *match, *rt0;
756 fn->rr_ptr = rt0 = fn->leaf;
758 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
762 struct rt6_info *next = rt0->dst.rt6_next;
764 /* no entries matched; do round-robin */
765 if (!next || next->rt6i_metric != rt0->rt6i_metric)
772 net = dev_net(rt0->dst.dev);
773 return match ? match : net->ipv6.ip6_null_entry;
776 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
778 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
781 #ifdef CONFIG_IPV6_ROUTE_INFO
782 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
783 const struct in6_addr *gwaddr)
785 struct net *net = dev_net(dev);
786 struct route_info *rinfo = (struct route_info *) opt;
787 struct in6_addr prefix_buf, *prefix;
789 unsigned long lifetime;
792 if (len < sizeof(struct route_info)) {
796 /* Sanity check for prefix_len and length */
797 if (rinfo->length > 3) {
799 } else if (rinfo->prefix_len > 128) {
801 } else if (rinfo->prefix_len > 64) {
802 if (rinfo->length < 2) {
805 } else if (rinfo->prefix_len > 0) {
806 if (rinfo->length < 1) {
811 pref = rinfo->route_pref;
812 if (pref == ICMPV6_ROUTER_PREF_INVALID)
815 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
817 if (rinfo->length == 3)
818 prefix = (struct in6_addr *)rinfo->prefix;
820 /* this function is safe */
821 ipv6_addr_prefix(&prefix_buf,
822 (struct in6_addr *)rinfo->prefix,
824 prefix = &prefix_buf;
827 if (rinfo->prefix_len == 0)
828 rt = rt6_get_dflt_router(gwaddr, dev);
830 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
833 if (rt && !lifetime) {
839 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
842 rt->rt6i_flags = RTF_ROUTEINFO |
843 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
846 if (!addrconf_finite_timeout(lifetime))
847 rt6_clean_expires(rt);
849 rt6_set_expires(rt, jiffies + HZ * lifetime);
857 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
858 struct in6_addr *saddr)
860 struct fib6_node *pn;
862 if (fn->fn_flags & RTN_TL_ROOT)
865 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
866 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
869 if (fn->fn_flags & RTN_RTINFO)
874 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
875 struct fib6_table *table,
876 struct flowi6 *fl6, int flags)
878 struct fib6_node *fn;
881 read_lock_bh(&table->tb6_lock);
882 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
885 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
886 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
887 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
888 if (rt == net->ipv6.ip6_null_entry) {
889 fn = fib6_backtrack(fn, &fl6->saddr);
893 dst_use(&rt->dst, jiffies);
894 read_unlock_bh(&table->tb6_lock);
896 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
902 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
905 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
907 EXPORT_SYMBOL_GPL(ip6_route_lookup);
909 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
910 const struct in6_addr *saddr, int oif, int strict)
912 struct flowi6 fl6 = {
916 struct dst_entry *dst;
917 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
920 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
921 flags |= RT6_LOOKUP_F_HAS_SADDR;
924 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
926 return (struct rt6_info *) dst;
932 EXPORT_SYMBOL(rt6_lookup);
934 /* ip6_ins_rt is called with FREE table->tb6_lock.
935 It takes new route entry, the addition fails by any reason the
936 route is freed. In any case, if caller does not hold it, it may
940 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
941 struct mx6_config *mxc)
944 struct fib6_table *table;
946 table = rt->rt6i_table;
947 write_lock_bh(&table->tb6_lock);
948 err = fib6_add(&table->tb6_root, rt, info, mxc);
949 write_unlock_bh(&table->tb6_lock);
954 int ip6_ins_rt(struct rt6_info *rt)
956 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
957 struct mx6_config mxc = { .mx = NULL, };
959 return __ip6_ins_rt(rt, &info, &mxc);
962 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
963 const struct in6_addr *daddr,
964 const struct in6_addr *saddr)
972 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
973 ort = (struct rt6_info *)ort->dst.from;
975 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
980 ip6_rt_copy_init(rt, ort);
981 rt->rt6i_flags |= RTF_CACHE;
983 rt->dst.flags |= DST_HOST;
984 rt->rt6i_dst.addr = *daddr;
985 rt->rt6i_dst.plen = 128;
987 if (!rt6_is_gw_or_nonexthop(ort)) {
988 if (ort->rt6i_dst.plen != 128 &&
989 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
990 rt->rt6i_flags |= RTF_ANYCAST;
991 #ifdef CONFIG_IPV6_SUBTREES
992 if (rt->rt6i_src.plen && saddr) {
993 rt->rt6i_src.addr = *saddr;
994 rt->rt6i_src.plen = 128;
1002 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
1004 struct rt6_info *pcpu_rt;
1006 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
1007 rt->dst.dev, rt->dst.flags);
1011 ip6_rt_copy_init(pcpu_rt, rt);
1012 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
1013 pcpu_rt->rt6i_flags |= RTF_PCPU;
1017 /* It should be called with read_lock_bh(&tb6_lock) acquired */
1018 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
1020 struct rt6_info *pcpu_rt, **p;
1022 p = this_cpu_ptr(rt->rt6i_pcpu);
1026 dst_hold(&pcpu_rt->dst);
1027 rt6_dst_from_metrics_check(pcpu_rt);
1032 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1034 struct fib6_table *table = rt->rt6i_table;
1035 struct rt6_info *pcpu_rt, *prev, **p;
1037 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1039 struct net *net = dev_net(rt->dst.dev);
1041 dst_hold(&net->ipv6.ip6_null_entry->dst);
1042 return net->ipv6.ip6_null_entry;
1045 read_lock_bh(&table->tb6_lock);
1046 if (rt->rt6i_pcpu) {
1047 p = this_cpu_ptr(rt->rt6i_pcpu);
1048 prev = cmpxchg(p, NULL, pcpu_rt);
1050 /* If someone did it before us, return prev instead */
1051 dst_destroy(&pcpu_rt->dst);
1055 /* rt has been removed from the fib6 tree
1056 * before we have a chance to acquire the read_lock.
1057 * In this case, don't brother to create a pcpu rt
1058 * since rt is going away anyway. The next
1059 * dst_check() will trigger a re-lookup.
1061 dst_destroy(&pcpu_rt->dst);
1064 dst_hold(&pcpu_rt->dst);
1065 rt6_dst_from_metrics_check(pcpu_rt);
1066 read_unlock_bh(&table->tb6_lock);
1070 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1071 int oif, struct flowi6 *fl6, int flags)
1073 struct fib6_node *fn, *saved_fn;
1074 struct rt6_info *rt;
1077 strict |= flags & RT6_LOOKUP_F_IFACE;
1078 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
1079 if (net->ipv6.devconf_all->forwarding == 0)
1080 strict |= RT6_LOOKUP_F_REACHABLE;
1082 read_lock_bh(&table->tb6_lock);
1084 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1087 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1091 rt = rt6_select(fn, oif, strict);
1092 if (rt->rt6i_nsiblings)
1093 rt = rt6_multipath_select(rt, fl6, oif, strict);
1094 if (rt == net->ipv6.ip6_null_entry) {
1095 fn = fib6_backtrack(fn, &fl6->saddr);
1097 goto redo_rt6_select;
1098 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1099 /* also consider unreachable route */
1100 strict &= ~RT6_LOOKUP_F_REACHABLE;
1102 goto redo_rt6_select;
1107 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1108 dst_use(&rt->dst, jiffies);
1109 read_unlock_bh(&table->tb6_lock);
1111 rt6_dst_from_metrics_check(rt);
1113 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1115 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1116 !(rt->rt6i_flags & RTF_GATEWAY))) {
1117 /* Create a RTF_CACHE clone which will not be
1118 * owned by the fib6 tree. It is for the special case where
1119 * the daddr in the skb during the neighbor look-up is different
1120 * from the fl6->daddr used to look-up route here.
1123 struct rt6_info *uncached_rt;
1125 dst_use(&rt->dst, jiffies);
1126 read_unlock_bh(&table->tb6_lock);
1128 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1129 dst_release(&rt->dst);
1132 rt6_uncached_list_add(uncached_rt);
1134 uncached_rt = net->ipv6.ip6_null_entry;
1136 dst_hold(&uncached_rt->dst);
1138 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1142 /* Get a percpu copy */
1144 struct rt6_info *pcpu_rt;
1146 rt->dst.lastuse = jiffies;
1148 pcpu_rt = rt6_get_pcpu_route(rt);
1151 read_unlock_bh(&table->tb6_lock);
1153 /* We have to do the read_unlock first
1154 * because rt6_make_pcpu_route() may trigger
1155 * ip6_dst_gc() which will take the write_lock.
1158 read_unlock_bh(&table->tb6_lock);
1159 pcpu_rt = rt6_make_pcpu_route(rt);
1160 dst_release(&rt->dst);
1163 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1168 EXPORT_SYMBOL_GPL(ip6_pol_route);
1170 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1171 struct flowi6 *fl6, int flags)
1173 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1176 struct dst_entry *ip6_route_input_lookup(struct net *net,
1177 struct net_device *dev,
1178 struct flowi6 *fl6, int flags)
1180 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1181 flags |= RT6_LOOKUP_F_IFACE;
1183 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1185 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
1187 void ip6_route_input(struct sk_buff *skb)
1189 const struct ipv6hdr *iph = ipv6_hdr(skb);
1190 struct net *net = dev_net(skb->dev);
1191 int flags = RT6_LOOKUP_F_HAS_SADDR;
1192 struct ip_tunnel_info *tun_info;
1193 struct flowi6 fl6 = {
1194 .flowi6_iif = skb->dev->ifindex,
1195 .daddr = iph->daddr,
1196 .saddr = iph->saddr,
1197 .flowlabel = ip6_flowinfo(iph),
1198 .flowi6_mark = skb->mark,
1199 .flowi6_proto = iph->nexthdr,
1202 tun_info = skb_tunnel_info(skb);
1203 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1204 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1206 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1209 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1210 struct flowi6 *fl6, int flags)
1212 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1215 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1216 struct flowi6 *fl6, int flags)
1220 if (rt6_need_strict(&fl6->daddr)) {
1221 struct dst_entry *dst;
1223 dst = l3mdev_link_scope_lookup(net, fl6);
1228 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1230 any_src = ipv6_addr_any(&fl6->saddr);
1231 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1232 (fl6->flowi6_oif && any_src))
1233 flags |= RT6_LOOKUP_F_IFACE;
1236 flags |= RT6_LOOKUP_F_HAS_SADDR;
1238 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1240 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1242 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1244 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1246 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1247 struct dst_entry *new = NULL;
1249 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1255 new->input = dst_discard;
1256 new->output = dst_discard_out;
1258 dst_copy_metrics(new, &ort->dst);
1259 rt->rt6i_idev = ort->rt6i_idev;
1261 in6_dev_hold(rt->rt6i_idev);
1263 rt->rt6i_gateway = ort->rt6i_gateway;
1264 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1265 rt->rt6i_metric = 0;
1267 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1268 #ifdef CONFIG_IPV6_SUBTREES
1269 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1275 dst_release(dst_orig);
1276 return new ? new : ERR_PTR(-ENOMEM);
1280 * Destination cache support functions
1283 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1286 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1287 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1290 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1292 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1295 if (rt6_check_expired(rt))
1301 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1303 if (!__rt6_check_expired(rt) &&
1304 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1305 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1311 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1313 struct rt6_info *rt;
1315 rt = (struct rt6_info *) dst;
1317 /* All IPV6 dsts are created with ->obsolete set to the value
1318 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1319 * into this function always.
1322 rt6_dst_from_metrics_check(rt);
1324 if (rt->rt6i_flags & RTF_PCPU ||
1325 (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1326 return rt6_dst_from_check(rt, cookie);
1328 return rt6_check(rt, cookie);
1331 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1333 struct rt6_info *rt = (struct rt6_info *) dst;
1336 if (rt->rt6i_flags & RTF_CACHE) {
1337 if (rt6_check_expired(rt)) {
1349 static void ip6_link_failure(struct sk_buff *skb)
1351 struct rt6_info *rt;
1353 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1355 rt = (struct rt6_info *) skb_dst(skb);
1357 if (rt->rt6i_flags & RTF_CACHE) {
1360 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1361 rt->rt6i_node->fn_sernum = -1;
1366 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1368 struct net *net = dev_net(rt->dst.dev);
1370 rt->rt6i_flags |= RTF_MODIFIED;
1371 rt->rt6i_pmtu = mtu;
1372 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1375 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1377 return !(rt->rt6i_flags & RTF_CACHE) &&
1378 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1381 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1382 const struct ipv6hdr *iph, u32 mtu)
1384 const struct in6_addr *daddr, *saddr;
1385 struct rt6_info *rt6 = (struct rt6_info *)dst;
1387 if (rt6->rt6i_flags & RTF_LOCAL)
1390 if (dst_metric_locked(dst, RTAX_MTU))
1394 daddr = &iph->daddr;
1395 saddr = &iph->saddr;
1397 daddr = &sk->sk_v6_daddr;
1398 saddr = &inet6_sk(sk)->saddr;
1403 dst_confirm_neigh(dst, daddr);
1404 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1405 if (mtu >= dst_mtu(dst))
1408 if (!rt6_cache_allowed_for_pmtu(rt6)) {
1409 rt6_do_update_pmtu(rt6, mtu);
1411 struct rt6_info *nrt6;
1413 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1415 rt6_do_update_pmtu(nrt6, mtu);
1417 /* ip6_ins_rt(nrt6) will bump the
1418 * rt6->rt6i_node->fn_sernum
1419 * which will fail the next rt6_check() and
1420 * invalidate the sk->sk_dst_cache.
1427 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1428 struct sk_buff *skb, u32 mtu)
1430 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1433 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1434 int oif, u32 mark, kuid_t uid)
1436 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1437 struct dst_entry *dst;
1440 memset(&fl6, 0, sizeof(fl6));
1441 fl6.flowi6_oif = oif;
1442 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1443 fl6.daddr = iph->daddr;
1444 fl6.saddr = iph->saddr;
1445 fl6.flowlabel = ip6_flowinfo(iph);
1446 fl6.flowi6_uid = uid;
1448 dst = ip6_route_output(net, NULL, &fl6);
1450 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1453 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1455 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1457 struct dst_entry *dst;
1459 ip6_update_pmtu(skb, sock_net(sk), mtu,
1460 sk->sk_bound_dev_if, sk->sk_mark, sk->sk_uid);
1462 dst = __sk_dst_get(sk);
1463 if (!dst || !dst->obsolete ||
1464 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1468 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1469 ip6_datagram_dst_update(sk, false);
1472 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1474 /* Handle redirects */
1475 struct ip6rd_flowi {
1477 struct in6_addr gateway;
1480 static struct rt6_info *__ip6_route_redirect(struct net *net,
1481 struct fib6_table *table,
1485 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1486 struct rt6_info *rt;
1487 struct fib6_node *fn;
1489 /* Get the "current" route for this destination and
1490 * check if the redirect has come from appropriate router.
1492 * RFC 4861 specifies that redirects should only be
1493 * accepted if they come from the nexthop to the target.
1494 * Due to the way the routes are chosen, this notion
1495 * is a bit fuzzy and one might need to check all possible
1499 read_lock_bh(&table->tb6_lock);
1500 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1502 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1503 if (rt6_check_expired(rt))
1507 if (!(rt->rt6i_flags & RTF_GATEWAY))
1509 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1511 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1517 rt = net->ipv6.ip6_null_entry;
1518 else if (rt->dst.error) {
1519 rt = net->ipv6.ip6_null_entry;
1523 if (rt == net->ipv6.ip6_null_entry) {
1524 fn = fib6_backtrack(fn, &fl6->saddr);
1532 read_unlock_bh(&table->tb6_lock);
1534 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1538 static struct dst_entry *ip6_route_redirect(struct net *net,
1539 const struct flowi6 *fl6,
1540 const struct in6_addr *gateway)
1542 int flags = RT6_LOOKUP_F_HAS_SADDR;
1543 struct ip6rd_flowi rdfl;
1546 rdfl.gateway = *gateway;
1548 return fib6_rule_lookup(net, &rdfl.fl6,
1549 flags, __ip6_route_redirect);
1552 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
1555 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1556 struct dst_entry *dst;
1559 memset(&fl6, 0, sizeof(fl6));
1560 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1561 fl6.flowi6_oif = oif;
1562 fl6.flowi6_mark = mark;
1563 fl6.daddr = iph->daddr;
1564 fl6.saddr = iph->saddr;
1565 fl6.flowlabel = ip6_flowinfo(iph);
1566 fl6.flowi6_uid = uid;
1568 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1569 rt6_do_redirect(dst, NULL, skb);
1572 EXPORT_SYMBOL_GPL(ip6_redirect);
1574 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1577 const struct ipv6hdr *iph = ipv6_hdr(skb);
1578 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1579 struct dst_entry *dst;
1582 memset(&fl6, 0, sizeof(fl6));
1583 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1584 fl6.flowi6_oif = oif;
1585 fl6.flowi6_mark = mark;
1586 fl6.daddr = msg->dest;
1587 fl6.saddr = iph->daddr;
1588 fl6.flowi6_uid = sock_net_uid(net, NULL);
1590 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1591 rt6_do_redirect(dst, NULL, skb);
1595 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1597 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
1600 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1602 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1604 struct net_device *dev = dst->dev;
1605 unsigned int mtu = dst_mtu(dst);
1606 struct net *net = dev_net(dev);
1608 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1610 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1611 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1614 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1615 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1616 * IPV6_MAXPLEN is also valid and means: "any MSS,
1617 * rely only on pmtu discovery"
1619 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1624 static unsigned int ip6_mtu(const struct dst_entry *dst)
1626 const struct rt6_info *rt = (const struct rt6_info *)dst;
1627 unsigned int mtu = rt->rt6i_pmtu;
1628 struct inet6_dev *idev;
1633 mtu = dst_metric_raw(dst, RTAX_MTU);
1640 idev = __in6_dev_get(dst->dev);
1642 mtu = idev->cnf.mtu6;
1646 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1648 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
1651 static struct dst_entry *icmp6_dst_gc_list;
1652 static DEFINE_SPINLOCK(icmp6_dst_lock);
1654 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1657 struct dst_entry *dst;
1658 struct rt6_info *rt;
1659 struct inet6_dev *idev = in6_dev_get(dev);
1660 struct net *net = dev_net(dev);
1662 if (unlikely(!idev))
1663 return ERR_PTR(-ENODEV);
1665 rt = ip6_dst_alloc(net, dev, 0);
1666 if (unlikely(!rt)) {
1668 dst = ERR_PTR(-ENOMEM);
1672 rt->dst.flags |= DST_HOST;
1673 rt->dst.output = ip6_output;
1674 atomic_set(&rt->dst.__refcnt, 1);
1675 rt->rt6i_gateway = fl6->daddr;
1676 rt->rt6i_dst.addr = fl6->daddr;
1677 rt->rt6i_dst.plen = 128;
1678 rt->rt6i_idev = idev;
1679 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1681 spin_lock_bh(&icmp6_dst_lock);
1682 rt->dst.next = icmp6_dst_gc_list;
1683 icmp6_dst_gc_list = &rt->dst;
1684 spin_unlock_bh(&icmp6_dst_lock);
1686 fib6_force_start_gc(net);
1688 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1694 int icmp6_dst_gc(void)
1696 struct dst_entry *dst, **pprev;
1699 spin_lock_bh(&icmp6_dst_lock);
1700 pprev = &icmp6_dst_gc_list;
1702 while ((dst = *pprev) != NULL) {
1703 if (!atomic_read(&dst->__refcnt)) {
1712 spin_unlock_bh(&icmp6_dst_lock);
1717 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1720 struct dst_entry *dst, **pprev;
1722 spin_lock_bh(&icmp6_dst_lock);
1723 pprev = &icmp6_dst_gc_list;
1724 while ((dst = *pprev) != NULL) {
1725 struct rt6_info *rt = (struct rt6_info *) dst;
1726 if (func(rt, arg)) {
1733 spin_unlock_bh(&icmp6_dst_lock);
1736 static int ip6_dst_gc(struct dst_ops *ops)
1738 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1739 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1740 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1741 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1742 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1743 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1746 entries = dst_entries_get_fast(ops);
1747 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1748 entries <= rt_max_size)
1751 net->ipv6.ip6_rt_gc_expire++;
1752 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1753 entries = dst_entries_get_slow(ops);
1754 if (entries < ops->gc_thresh)
1755 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1757 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1758 return entries > rt_max_size;
1761 static int ip6_convert_metrics(struct mx6_config *mxc,
1762 const struct fib6_config *cfg)
1764 bool ecn_ca = false;
1772 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1776 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1777 int type = nla_type(nla);
1782 if (unlikely(type > RTAX_MAX))
1785 if (type == RTAX_CC_ALGO) {
1786 char tmp[TCP_CA_NAME_MAX];
1788 nla_strlcpy(tmp, nla, sizeof(tmp));
1789 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1790 if (val == TCP_CA_UNSPEC)
1793 val = nla_get_u32(nla);
1795 if (type == RTAX_HOPLIMIT && val > 255)
1797 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1801 __set_bit(type - 1, mxc->mx_valid);
1805 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1806 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1816 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
1817 struct fib6_config *cfg,
1818 const struct in6_addr *gw_addr)
1820 struct flowi6 fl6 = {
1821 .flowi6_oif = cfg->fc_ifindex,
1823 .saddr = cfg->fc_prefsrc,
1825 struct fib6_table *table;
1826 struct rt6_info *rt;
1827 int flags = RT6_LOOKUP_F_IFACE | RT6_LOOKUP_F_IGNORE_LINKSTATE;
1829 table = fib6_get_table(net, cfg->fc_table);
1833 if (!ipv6_addr_any(&cfg->fc_prefsrc))
1834 flags |= RT6_LOOKUP_F_HAS_SADDR;
1836 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
1838 /* if table lookup failed, fall back to full lookup */
1839 if (rt == net->ipv6.ip6_null_entry) {
1847 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1849 struct net *net = cfg->fc_nlinfo.nl_net;
1850 struct rt6_info *rt = NULL;
1851 struct net_device *dev = NULL;
1852 struct inet6_dev *idev = NULL;
1853 struct fib6_table *table;
1857 /* RTF_PCPU is an internal flag; can not be set by userspace */
1858 if (cfg->fc_flags & RTF_PCPU)
1861 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1863 #ifndef CONFIG_IPV6_SUBTREES
1864 if (cfg->fc_src_len)
1867 if (cfg->fc_ifindex) {
1869 dev = dev_get_by_index(net, cfg->fc_ifindex);
1872 idev = in6_dev_get(dev);
1877 if (cfg->fc_metric == 0)
1878 cfg->fc_metric = IP6_RT_PRIO_USER;
1881 if (cfg->fc_nlinfo.nlh &&
1882 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1883 table = fib6_get_table(net, cfg->fc_table);
1885 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1886 table = fib6_new_table(net, cfg->fc_table);
1889 table = fib6_new_table(net, cfg->fc_table);
1895 rt = ip6_dst_alloc(net, NULL,
1896 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1903 if (cfg->fc_flags & RTF_EXPIRES)
1904 rt6_set_expires(rt, jiffies +
1905 clock_t_to_jiffies(cfg->fc_expires));
1907 rt6_clean_expires(rt);
1909 if (cfg->fc_protocol == RTPROT_UNSPEC)
1910 cfg->fc_protocol = RTPROT_BOOT;
1911 rt->rt6i_protocol = cfg->fc_protocol;
1913 addr_type = ipv6_addr_type(&cfg->fc_dst);
1915 if (addr_type & IPV6_ADDR_MULTICAST)
1916 rt->dst.input = ip6_mc_input;
1917 else if (cfg->fc_flags & RTF_LOCAL)
1918 rt->dst.input = ip6_input;
1920 rt->dst.input = ip6_forward;
1922 rt->dst.output = ip6_output;
1924 if (cfg->fc_encap) {
1925 struct lwtunnel_state *lwtstate;
1927 err = lwtunnel_build_state(cfg->fc_encap_type,
1928 cfg->fc_encap, AF_INET6, cfg,
1932 rt->dst.lwtstate = lwtstate_get(lwtstate);
1933 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1934 rt->dst.lwtstate->orig_output = rt->dst.output;
1935 rt->dst.output = lwtunnel_output;
1937 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1938 rt->dst.lwtstate->orig_input = rt->dst.input;
1939 rt->dst.input = lwtunnel_input;
1943 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1944 rt->rt6i_dst.plen = cfg->fc_dst_len;
1945 if (rt->rt6i_dst.plen == 128)
1946 rt->dst.flags |= DST_HOST;
1948 #ifdef CONFIG_IPV6_SUBTREES
1949 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1950 rt->rt6i_src.plen = cfg->fc_src_len;
1953 rt->rt6i_metric = cfg->fc_metric;
1955 /* We cannot add true routes via loopback here,
1956 they would result in kernel looping; promote them to reject routes
1958 if ((cfg->fc_flags & RTF_REJECT) ||
1959 (dev && (dev->flags & IFF_LOOPBACK) &&
1960 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1961 !(cfg->fc_flags & RTF_LOCAL))) {
1962 /* hold loopback dev/idev if we haven't done so. */
1963 if (dev != net->loopback_dev) {
1968 dev = net->loopback_dev;
1970 idev = in6_dev_get(dev);
1976 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1977 switch (cfg->fc_type) {
1979 rt->dst.error = -EINVAL;
1980 rt->dst.output = dst_discard_out;
1981 rt->dst.input = dst_discard;
1984 rt->dst.error = -EACCES;
1985 rt->dst.output = ip6_pkt_prohibit_out;
1986 rt->dst.input = ip6_pkt_prohibit;
1989 case RTN_UNREACHABLE:
1991 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1992 : (cfg->fc_type == RTN_UNREACHABLE)
1993 ? -EHOSTUNREACH : -ENETUNREACH;
1994 rt->dst.output = ip6_pkt_discard_out;
1995 rt->dst.input = ip6_pkt_discard;
2001 if (cfg->fc_flags & RTF_GATEWAY) {
2002 const struct in6_addr *gw_addr;
2005 gw_addr = &cfg->fc_gateway;
2006 gwa_type = ipv6_addr_type(gw_addr);
2008 /* if gw_addr is local we will fail to detect this in case
2009 * address is still TENTATIVE (DAD in progress). rt6_lookup()
2010 * will return already-added prefix route via interface that
2011 * prefix route was assigned to, which might be non-loopback.
2014 if (ipv6_chk_addr_and_flags(net, gw_addr,
2015 gwa_type & IPV6_ADDR_LINKLOCAL ?
2019 rt->rt6i_gateway = *gw_addr;
2021 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
2022 struct rt6_info *grt = NULL;
2024 /* IPv6 strictly inhibits using not link-local
2025 addresses as nexthop address.
2026 Otherwise, router will not able to send redirects.
2027 It is very good, but in some (rare!) circumstances
2028 (SIT, PtP, NBMA NOARP links) it is handy to allow
2029 some exceptions. --ANK
2030 We allow IPv4-mapped nexthops to support RFC4798-type
2033 if (!(gwa_type & (IPV6_ADDR_UNICAST |
2037 if (cfg->fc_table) {
2038 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
2041 if (grt->rt6i_flags & RTF_GATEWAY ||
2042 (dev && dev != grt->dst.dev)) {
2050 grt = rt6_lookup(net, gw_addr, NULL,
2051 cfg->fc_ifindex, 1);
2053 err = -EHOSTUNREACH;
2057 if (dev != grt->dst.dev) {
2063 idev = grt->rt6i_idev;
2065 in6_dev_hold(grt->rt6i_idev);
2067 if (!(grt->rt6i_flags & RTF_GATEWAY))
2075 if (!dev || (dev->flags & IFF_LOOPBACK))
2083 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2084 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2088 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2089 rt->rt6i_prefsrc.plen = 128;
2091 rt->rt6i_prefsrc.plen = 0;
2093 rt->rt6i_flags = cfg->fc_flags;
2097 rt->rt6i_idev = idev;
2098 rt->rt6i_table = table;
2100 cfg->fc_nlinfo.nl_net = dev_net(dev);
2111 return ERR_PTR(err);
2114 int ip6_route_add(struct fib6_config *cfg)
2116 struct mx6_config mxc = { .mx = NULL, };
2117 struct rt6_info *rt;
2120 rt = ip6_route_info_create(cfg);
2127 err = ip6_convert_metrics(&mxc, cfg);
2131 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2143 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2146 struct fib6_table *table;
2147 struct net *net = dev_net(rt->dst.dev);
2149 if (rt == net->ipv6.ip6_null_entry ||
2150 rt->dst.flags & DST_NOCACHE) {
2155 table = rt->rt6i_table;
2156 write_lock_bh(&table->tb6_lock);
2157 err = fib6_del(rt, info);
2158 write_unlock_bh(&table->tb6_lock);
2165 int ip6_del_rt(struct rt6_info *rt)
2167 struct nl_info info = {
2168 .nl_net = dev_net(rt->dst.dev),
2170 return __ip6_del_rt(rt, &info);
2173 static int __ip6_del_rt_siblings(struct rt6_info *rt, struct fib6_config *cfg)
2175 struct nl_info *info = &cfg->fc_nlinfo;
2176 struct net *net = info->nl_net;
2177 struct sk_buff *skb = NULL;
2178 struct fib6_table *table;
2181 if (rt == net->ipv6.ip6_null_entry)
2183 table = rt->rt6i_table;
2184 write_lock_bh(&table->tb6_lock);
2186 if (rt->rt6i_nsiblings && cfg->fc_delete_all_nh) {
2187 struct rt6_info *sibling, *next_sibling;
2189 /* prefer to send a single notification with all hops */
2190 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
2192 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2194 if (rt6_fill_node(net, skb, rt,
2195 NULL, NULL, 0, RTM_DELROUTE,
2196 info->portid, seq, 0) < 0) {
2200 info->skip_notify = 1;
2203 list_for_each_entry_safe(sibling, next_sibling,
2206 err = fib6_del(sibling, info);
2212 err = fib6_del(rt, info);
2214 write_unlock_bh(&table->tb6_lock);
2219 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2220 info->nlh, gfp_any());
2225 static int ip6_route_del(struct fib6_config *cfg)
2227 struct fib6_table *table;
2228 struct fib6_node *fn;
2229 struct rt6_info *rt;
2232 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2236 read_lock_bh(&table->tb6_lock);
2238 fn = fib6_locate(&table->tb6_root,
2239 &cfg->fc_dst, cfg->fc_dst_len,
2240 &cfg->fc_src, cfg->fc_src_len);
2243 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2244 if ((rt->rt6i_flags & RTF_CACHE) &&
2245 !(cfg->fc_flags & RTF_CACHE))
2247 if (cfg->fc_ifindex &&
2249 rt->dst.dev->ifindex != cfg->fc_ifindex))
2251 if (cfg->fc_flags & RTF_GATEWAY &&
2252 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2254 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2256 if (cfg->fc_protocol && cfg->fc_protocol != rt->rt6i_protocol)
2259 read_unlock_bh(&table->tb6_lock);
2261 /* if gateway was specified only delete the one hop */
2262 if (cfg->fc_flags & RTF_GATEWAY)
2263 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2265 return __ip6_del_rt_siblings(rt, cfg);
2268 read_unlock_bh(&table->tb6_lock);
2273 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2275 struct netevent_redirect netevent;
2276 struct rt6_info *rt, *nrt = NULL;
2277 struct ndisc_options ndopts;
2278 struct inet6_dev *in6_dev;
2279 struct neighbour *neigh;
2281 int optlen, on_link;
2284 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2285 optlen -= sizeof(*msg);
2288 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2292 msg = (struct rd_msg *)icmp6_hdr(skb);
2294 if (ipv6_addr_is_multicast(&msg->dest)) {
2295 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2300 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2302 } else if (ipv6_addr_type(&msg->target) !=
2303 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2304 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2308 in6_dev = __in6_dev_get(skb->dev);
2311 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2315 * The IP source address of the Redirect MUST be the same as the current
2316 * first-hop router for the specified ICMP Destination Address.
2319 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
2320 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2325 if (ndopts.nd_opts_tgt_lladdr) {
2326 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2329 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2334 rt = (struct rt6_info *) dst;
2335 if (rt->rt6i_flags & RTF_REJECT) {
2336 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2340 /* Redirect received -> path was valid.
2341 * Look, redirects are sent only in response to data packets,
2342 * so that this nexthop apparently is reachable. --ANK
2344 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
2346 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2351 * We have finally decided to accept it.
2354 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
2355 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2356 NEIGH_UPDATE_F_OVERRIDE|
2357 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2358 NEIGH_UPDATE_F_ISROUTER)),
2359 NDISC_REDIRECT, &ndopts);
2361 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2365 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2367 nrt->rt6i_flags &= ~RTF_GATEWAY;
2369 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2371 if (ip6_ins_rt(nrt))
2374 netevent.old = &rt->dst;
2375 netevent.new = &nrt->dst;
2376 netevent.daddr = &msg->dest;
2377 netevent.neigh = neigh;
2378 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2380 if (rt->rt6i_flags & RTF_CACHE) {
2381 rt = (struct rt6_info *) dst_clone(&rt->dst);
2386 neigh_release(neigh);
2390 * Misc support functions
2393 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2395 BUG_ON(from->dst.from);
2397 rt->rt6i_flags &= ~RTF_EXPIRES;
2398 dst_hold(&from->dst);
2399 rt->dst.from = &from->dst;
2400 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2403 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2405 rt->dst.input = ort->dst.input;
2406 rt->dst.output = ort->dst.output;
2407 rt->rt6i_dst = ort->rt6i_dst;
2408 rt->dst.error = ort->dst.error;
2409 rt->rt6i_idev = ort->rt6i_idev;
2411 in6_dev_hold(rt->rt6i_idev);
2412 rt->dst.lastuse = jiffies;
2413 rt->rt6i_gateway = ort->rt6i_gateway;
2414 rt->rt6i_flags = ort->rt6i_flags;
2415 rt6_set_from(rt, ort);
2416 rt->rt6i_metric = ort->rt6i_metric;
2417 #ifdef CONFIG_IPV6_SUBTREES
2418 rt->rt6i_src = ort->rt6i_src;
2420 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2421 rt->rt6i_table = ort->rt6i_table;
2422 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2425 #ifdef CONFIG_IPV6_ROUTE_INFO
2426 static struct rt6_info *rt6_get_route_info(struct net *net,
2427 const struct in6_addr *prefix, int prefixlen,
2428 const struct in6_addr *gwaddr,
2429 struct net_device *dev)
2431 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
2432 int ifindex = dev->ifindex;
2433 struct fib6_node *fn;
2434 struct rt6_info *rt = NULL;
2435 struct fib6_table *table;
2437 table = fib6_get_table(net, tb_id);
2441 read_lock_bh(&table->tb6_lock);
2442 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2446 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2447 if (rt->dst.dev->ifindex != ifindex)
2449 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2451 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2457 read_unlock_bh(&table->tb6_lock);
2461 static struct rt6_info *rt6_add_route_info(struct net *net,
2462 const struct in6_addr *prefix, int prefixlen,
2463 const struct in6_addr *gwaddr,
2464 struct net_device *dev,
2467 struct fib6_config cfg = {
2468 .fc_metric = IP6_RT_PRIO_USER,
2469 .fc_ifindex = dev->ifindex,
2470 .fc_dst_len = prefixlen,
2471 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2472 RTF_UP | RTF_PREF(pref),
2473 .fc_nlinfo.portid = 0,
2474 .fc_nlinfo.nlh = NULL,
2475 .fc_nlinfo.nl_net = net,
2478 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
2479 cfg.fc_dst = *prefix;
2480 cfg.fc_gateway = *gwaddr;
2482 /* We should treat it as a default route if prefix length is 0. */
2484 cfg.fc_flags |= RTF_DEFAULT;
2486 ip6_route_add(&cfg);
2488 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
2492 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2494 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
2495 struct rt6_info *rt;
2496 struct fib6_table *table;
2498 table = fib6_get_table(dev_net(dev), tb_id);
2502 read_lock_bh(&table->tb6_lock);
2503 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2504 if (dev == rt->dst.dev &&
2505 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2506 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2511 read_unlock_bh(&table->tb6_lock);
2515 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2516 struct net_device *dev,
2519 struct fib6_config cfg = {
2520 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2521 .fc_metric = IP6_RT_PRIO_USER,
2522 .fc_ifindex = dev->ifindex,
2523 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2524 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2525 .fc_nlinfo.portid = 0,
2526 .fc_nlinfo.nlh = NULL,
2527 .fc_nlinfo.nl_net = dev_net(dev),
2530 cfg.fc_gateway = *gwaddr;
2532 if (!ip6_route_add(&cfg)) {
2533 struct fib6_table *table;
2535 table = fib6_get_table(dev_net(dev), cfg.fc_table);
2537 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
2540 return rt6_get_dflt_router(gwaddr, dev);
2543 static void __rt6_purge_dflt_routers(struct fib6_table *table)
2545 struct rt6_info *rt;
2548 read_lock_bh(&table->tb6_lock);
2549 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2550 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2551 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2553 read_unlock_bh(&table->tb6_lock);
2558 read_unlock_bh(&table->tb6_lock);
2560 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
2563 void rt6_purge_dflt_routers(struct net *net)
2565 struct fib6_table *table;
2566 struct hlist_head *head;
2571 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
2572 head = &net->ipv6.fib_table_hash[h];
2573 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
2574 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
2575 __rt6_purge_dflt_routers(table);
2582 static void rtmsg_to_fib6_config(struct net *net,
2583 struct in6_rtmsg *rtmsg,
2584 struct fib6_config *cfg)
2586 memset(cfg, 0, sizeof(*cfg));
2588 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2590 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2591 cfg->fc_metric = rtmsg->rtmsg_metric;
2592 cfg->fc_expires = rtmsg->rtmsg_info;
2593 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2594 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2595 cfg->fc_flags = rtmsg->rtmsg_flags;
2597 cfg->fc_nlinfo.nl_net = net;
2599 cfg->fc_dst = rtmsg->rtmsg_dst;
2600 cfg->fc_src = rtmsg->rtmsg_src;
2601 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2604 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2606 struct fib6_config cfg;
2607 struct in6_rtmsg rtmsg;
2611 case SIOCADDRT: /* Add a route */
2612 case SIOCDELRT: /* Delete a route */
2613 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2615 err = copy_from_user(&rtmsg, arg,
2616 sizeof(struct in6_rtmsg));
2620 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2625 err = ip6_route_add(&cfg);
2628 err = ip6_route_del(&cfg);
2642 * Drop the packet on the floor
2645 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2648 struct dst_entry *dst = skb_dst(skb);
2649 switch (ipstats_mib_noroutes) {
2650 case IPSTATS_MIB_INNOROUTES:
2651 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2652 if (type == IPV6_ADDR_ANY) {
2653 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2654 IPSTATS_MIB_INADDRERRORS);
2658 case IPSTATS_MIB_OUTNOROUTES:
2659 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2660 ipstats_mib_noroutes);
2663 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2668 static int ip6_pkt_discard(struct sk_buff *skb)
2670 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2673 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2675 skb->dev = skb_dst(skb)->dev;
2676 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2679 static int ip6_pkt_prohibit(struct sk_buff *skb)
2681 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2684 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2686 skb->dev = skb_dst(skb)->dev;
2687 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2691 * Allocate a dst for local (unicast / anycast) address.
2694 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2695 const struct in6_addr *addr,
2699 struct net *net = dev_net(idev->dev);
2700 struct net_device *dev = net->loopback_dev;
2701 struct rt6_info *rt;
2703 /* use L3 Master device as loopback for host routes if device
2704 * is enslaved and address is not link local or multicast
2706 if (!rt6_need_strict(addr))
2707 dev = l3mdev_master_dev_rcu(idev->dev) ? : dev;
2709 rt = ip6_dst_alloc(net, dev, DST_NOCOUNT);
2711 return ERR_PTR(-ENOMEM);
2715 rt->dst.flags |= DST_HOST;
2716 rt->dst.input = ip6_input;
2717 rt->dst.output = ip6_output;
2718 rt->rt6i_idev = idev;
2720 rt->rt6i_protocol = RTPROT_KERNEL;
2721 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2723 rt->rt6i_flags |= RTF_ANYCAST;
2725 rt->rt6i_flags |= RTF_LOCAL;
2727 rt->rt6i_gateway = *addr;
2728 rt->rt6i_dst.addr = *addr;
2729 rt->rt6i_dst.plen = 128;
2730 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2731 rt->rt6i_table = fib6_get_table(net, tb_id);
2732 rt->dst.flags |= DST_NOCACHE;
2734 atomic_set(&rt->dst.__refcnt, 1);
2739 /* remove deleted ip from prefsrc entries */
2740 struct arg_dev_net_ip {
2741 struct net_device *dev;
2743 struct in6_addr *addr;
2746 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2748 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2749 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2750 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2752 if (((void *)rt->dst.dev == dev || !dev) &&
2753 rt != net->ipv6.ip6_null_entry &&
2754 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2755 /* remove prefsrc entry */
2756 rt->rt6i_prefsrc.plen = 0;
2761 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2763 struct net *net = dev_net(ifp->idev->dev);
2764 struct arg_dev_net_ip adni = {
2765 .dev = ifp->idev->dev,
2769 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2772 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2773 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2775 /* Remove routers and update dst entries when gateway turn into host. */
2776 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2778 struct in6_addr *gateway = (struct in6_addr *)arg;
2780 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2781 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2782 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2788 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2790 fib6_clean_all(net, fib6_clean_tohost, gateway);
2793 struct arg_dev_net {
2794 struct net_device *dev;
2798 /* called with write lock held for table with rt */
2799 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2801 const struct arg_dev_net *adn = arg;
2802 const struct net_device *dev = adn->dev;
2804 if ((rt->dst.dev == dev || !dev) &&
2805 rt != adn->net->ipv6.ip6_null_entry &&
2806 (rt->rt6i_nsiblings == 0 ||
2807 (dev && netdev_unregistering(dev)) ||
2808 !rt->rt6i_idev->cnf.ignore_routes_with_linkdown))
2814 void rt6_ifdown(struct net *net, struct net_device *dev)
2816 struct arg_dev_net adn = {
2821 fib6_clean_all(net, fib6_ifdown, &adn);
2822 icmp6_clean_all(fib6_ifdown, &adn);
2824 rt6_uncached_list_flush_dev(net, dev);
2827 struct rt6_mtu_change_arg {
2828 struct net_device *dev;
2832 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2834 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2835 struct inet6_dev *idev;
2837 /* In IPv6 pmtu discovery is not optional,
2838 so that RTAX_MTU lock cannot disable it.
2839 We still use this lock to block changes
2840 caused by addrconf/ndisc.
2843 idev = __in6_dev_get(arg->dev);
2847 /* For administrative MTU increase, there is no way to discover
2848 IPv6 PMTU increase, so PMTU increase should be updated here.
2849 Since RFC 1981 doesn't include administrative MTU increase
2850 update PMTU increase is a MUST. (i.e. jumbo frame)
2853 If new MTU is less than route PMTU, this new MTU will be the
2854 lowest MTU in the path, update the route PMTU to reflect PMTU
2855 decreases; if new MTU is greater than route PMTU, and the
2856 old MTU is the lowest MTU in the path, update the route PMTU
2857 to reflect the increase. In this case if the other nodes' MTU
2858 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2861 if (rt->dst.dev == arg->dev &&
2862 dst_metric_raw(&rt->dst, RTAX_MTU) &&
2863 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2864 if (rt->rt6i_flags & RTF_CACHE) {
2865 /* For RTF_CACHE with rt6i_pmtu == 0
2866 * (i.e. a redirected route),
2867 * the metrics of its rt->dst.from has already
2870 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2871 rt->rt6i_pmtu = arg->mtu;
2872 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2873 (dst_mtu(&rt->dst) < arg->mtu &&
2874 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2875 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2881 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2883 struct rt6_mtu_change_arg arg = {
2888 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2891 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2892 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2893 [RTA_OIF] = { .type = NLA_U32 },
2894 [RTA_IIF] = { .type = NLA_U32 },
2895 [RTA_PRIORITY] = { .type = NLA_U32 },
2896 [RTA_METRICS] = { .type = NLA_NESTED },
2897 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2898 [RTA_PREF] = { .type = NLA_U8 },
2899 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2900 [RTA_ENCAP] = { .type = NLA_NESTED },
2901 [RTA_EXPIRES] = { .type = NLA_U32 },
2902 [RTA_UID] = { .type = NLA_U32 },
2903 [RTA_MARK] = { .type = NLA_U32 },
2906 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2907 struct fib6_config *cfg)
2910 struct nlattr *tb[RTA_MAX+1];
2914 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
2920 rtm = nlmsg_data(nlh);
2921 memset(cfg, 0, sizeof(*cfg));
2923 cfg->fc_table = rtm->rtm_table;
2924 cfg->fc_dst_len = rtm->rtm_dst_len;
2925 cfg->fc_src_len = rtm->rtm_src_len;
2926 cfg->fc_flags = RTF_UP;
2927 cfg->fc_protocol = rtm->rtm_protocol;
2928 cfg->fc_type = rtm->rtm_type;
2930 if (rtm->rtm_type == RTN_UNREACHABLE ||
2931 rtm->rtm_type == RTN_BLACKHOLE ||
2932 rtm->rtm_type == RTN_PROHIBIT ||
2933 rtm->rtm_type == RTN_THROW)
2934 cfg->fc_flags |= RTF_REJECT;
2936 if (rtm->rtm_type == RTN_LOCAL)
2937 cfg->fc_flags |= RTF_LOCAL;
2939 if (rtm->rtm_flags & RTM_F_CLONED)
2940 cfg->fc_flags |= RTF_CACHE;
2942 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2943 cfg->fc_nlinfo.nlh = nlh;
2944 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2946 if (tb[RTA_GATEWAY]) {
2947 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2948 cfg->fc_flags |= RTF_GATEWAY;
2952 int plen = (rtm->rtm_dst_len + 7) >> 3;
2954 if (nla_len(tb[RTA_DST]) < plen)
2957 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2961 int plen = (rtm->rtm_src_len + 7) >> 3;
2963 if (nla_len(tb[RTA_SRC]) < plen)
2966 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2969 if (tb[RTA_PREFSRC])
2970 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2973 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2975 if (tb[RTA_PRIORITY])
2976 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2978 if (tb[RTA_METRICS]) {
2979 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2980 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2984 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2986 if (tb[RTA_MULTIPATH]) {
2987 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2988 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2990 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
2997 pref = nla_get_u8(tb[RTA_PREF]);
2998 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2999 pref != ICMPV6_ROUTER_PREF_HIGH)
3000 pref = ICMPV6_ROUTER_PREF_MEDIUM;
3001 cfg->fc_flags |= RTF_PREF(pref);
3005 cfg->fc_encap = tb[RTA_ENCAP];
3007 if (tb[RTA_ENCAP_TYPE]) {
3008 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
3010 err = lwtunnel_valid_encap_type(cfg->fc_encap_type);
3015 if (tb[RTA_EXPIRES]) {
3016 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
3018 if (addrconf_finite_timeout(timeout)) {
3019 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
3020 cfg->fc_flags |= RTF_EXPIRES;
3030 struct rt6_info *rt6_info;
3031 struct fib6_config r_cfg;
3032 struct mx6_config mxc;
3033 struct list_head next;
3036 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
3040 list_for_each_entry(nh, rt6_nh_list, next) {
3041 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6c nexthop %pI6c ifi %d\n",
3042 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
3043 nh->r_cfg.fc_ifindex);
3047 static int ip6_route_info_append(struct list_head *rt6_nh_list,
3048 struct rt6_info *rt, struct fib6_config *r_cfg)
3051 struct rt6_info *rtnh;
3054 list_for_each_entry(nh, rt6_nh_list, next) {
3055 /* check if rt6_info already exists */
3056 rtnh = nh->rt6_info;
3058 if (rtnh->dst.dev == rt->dst.dev &&
3059 rtnh->rt6i_idev == rt->rt6i_idev &&
3060 ipv6_addr_equal(&rtnh->rt6i_gateway,
3065 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
3069 err = ip6_convert_metrics(&nh->mxc, r_cfg);
3074 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
3075 list_add_tail(&nh->next, rt6_nh_list);
3080 static void ip6_route_mpath_notify(struct rt6_info *rt,
3081 struct rt6_info *rt_last,
3082 struct nl_info *info,
3085 /* if this is an APPEND route, then rt points to the first route
3086 * inserted and rt_last points to last route inserted. Userspace
3087 * wants a consistent dump of the route which starts at the first
3088 * nexthop. Since sibling routes are always added at the end of
3089 * the list, find the first sibling of the last route appended
3091 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->rt6i_nsiblings) {
3092 rt = list_first_entry(&rt_last->rt6i_siblings,
3098 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
3101 static int ip6_route_multipath_add(struct fib6_config *cfg)
3103 struct rt6_info *rt_notif = NULL, *rt_last = NULL;
3104 struct nl_info *info = &cfg->fc_nlinfo;
3105 struct fib6_config r_cfg;
3106 struct rtnexthop *rtnh;
3107 struct rt6_info *rt;
3108 struct rt6_nh *err_nh;
3109 struct rt6_nh *nh, *nh_safe;
3115 int replace = (cfg->fc_nlinfo.nlh &&
3116 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
3117 LIST_HEAD(rt6_nh_list);
3119 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
3120 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
3121 nlflags |= NLM_F_APPEND;
3123 remaining = cfg->fc_mp_len;
3124 rtnh = (struct rtnexthop *)cfg->fc_mp;
3126 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
3127 * rt6_info structs per nexthop
3129 while (rtnh_ok(rtnh, remaining)) {
3130 memcpy(&r_cfg, cfg, sizeof(*cfg));
3131 if (rtnh->rtnh_ifindex)
3132 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3134 attrlen = rtnh_attrlen(rtnh);
3136 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3138 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3140 r_cfg.fc_gateway = nla_get_in6_addr(nla);
3141 r_cfg.fc_flags |= RTF_GATEWAY;
3143 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
3144 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
3146 r_cfg.fc_encap_type = nla_get_u16(nla);
3149 rt = ip6_route_info_create(&r_cfg);
3156 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
3162 rtnh = rtnh_next(rtnh, &remaining);
3165 /* for add and replace send one notification with all nexthops.
3166 * Skip the notification in fib6_add_rt2node and send one with
3167 * the full route when done
3169 info->skip_notify = 1;
3172 list_for_each_entry(nh, &rt6_nh_list, next) {
3173 rt_last = nh->rt6_info;
3174 err = __ip6_ins_rt(nh->rt6_info, info, &nh->mxc);
3175 /* save reference to first route for notification */
3176 if (!rt_notif && !err)
3177 rt_notif = nh->rt6_info;
3179 /* nh->rt6_info is used or freed at this point, reset to NULL*/
3180 nh->rt6_info = NULL;
3183 ip6_print_replace_route_err(&rt6_nh_list);
3188 /* Because each route is added like a single route we remove
3189 * these flags after the first nexthop: if there is a collision,
3190 * we have already failed to add the first nexthop:
3191 * fib6_add_rt2node() has rejected it; when replacing, old
3192 * nexthops have been replaced by first new, the rest should
3195 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
3200 /* success ... tell user about new route */
3201 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
3205 /* send notification for routes that were added so that
3206 * the delete notifications sent by ip6_route_del are
3210 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
3212 /* Delete routes that were already added */
3213 list_for_each_entry(nh, &rt6_nh_list, next) {
3216 ip6_route_del(&nh->r_cfg);
3220 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3222 dst_free(&nh->rt6_info->dst);
3224 list_del(&nh->next);
3231 static int ip6_route_multipath_del(struct fib6_config *cfg)
3233 struct fib6_config r_cfg;
3234 struct rtnexthop *rtnh;
3237 int err = 1, last_err = 0;
3239 remaining = cfg->fc_mp_len;
3240 rtnh = (struct rtnexthop *)cfg->fc_mp;
3242 /* Parse a Multipath Entry */
3243 while (rtnh_ok(rtnh, remaining)) {
3244 memcpy(&r_cfg, cfg, sizeof(*cfg));
3245 if (rtnh->rtnh_ifindex)
3246 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3248 attrlen = rtnh_attrlen(rtnh);
3250 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3252 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3254 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3255 r_cfg.fc_flags |= RTF_GATEWAY;
3258 err = ip6_route_del(&r_cfg);
3262 rtnh = rtnh_next(rtnh, &remaining);
3268 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
3269 struct netlink_ext_ack *extack)
3271 struct fib6_config cfg;
3274 err = rtm_to_fib6_config(skb, nlh, &cfg);
3279 return ip6_route_multipath_del(&cfg);
3281 cfg.fc_delete_all_nh = 1;
3282 return ip6_route_del(&cfg);
3286 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
3287 struct netlink_ext_ack *extack)
3289 struct fib6_config cfg;
3292 err = rtm_to_fib6_config(skb, nlh, &cfg);
3297 return ip6_route_multipath_add(&cfg);
3299 return ip6_route_add(&cfg);
3302 static size_t rt6_nlmsg_size(struct rt6_info *rt)
3304 int nexthop_len = 0;
3306 if (rt->rt6i_nsiblings) {
3307 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
3308 + NLA_ALIGN(sizeof(struct rtnexthop))
3309 + nla_total_size(16) /* RTA_GATEWAY */
3310 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3312 nexthop_len *= rt->rt6i_nsiblings;
3315 return NLMSG_ALIGN(sizeof(struct rtmsg))
3316 + nla_total_size(16) /* RTA_SRC */
3317 + nla_total_size(16) /* RTA_DST */
3318 + nla_total_size(16) /* RTA_GATEWAY */
3319 + nla_total_size(16) /* RTA_PREFSRC */
3320 + nla_total_size(4) /* RTA_TABLE */
3321 + nla_total_size(4) /* RTA_IIF */
3322 + nla_total_size(4) /* RTA_OIF */
3323 + nla_total_size(4) /* RTA_PRIORITY */
3324 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3325 + nla_total_size(sizeof(struct rta_cacheinfo))
3326 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3327 + nla_total_size(1) /* RTA_PREF */
3328 + lwtunnel_get_encap_size(rt->dst.lwtstate)
3332 static int rt6_nexthop_info(struct sk_buff *skb, struct rt6_info *rt,
3333 unsigned int *flags, bool skip_oif)
3335 if (!netif_running(rt->dst.dev) || !netif_carrier_ok(rt->dst.dev)) {
3336 *flags |= RTNH_F_LINKDOWN;
3337 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3338 *flags |= RTNH_F_DEAD;
3341 if (rt->rt6i_flags & RTF_GATEWAY) {
3342 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3343 goto nla_put_failure;
3346 /* not needed for multipath encoding b/c it has a rtnexthop struct */
3347 if (!skip_oif && rt->dst.dev &&
3348 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3349 goto nla_put_failure;
3351 if (rt->dst.lwtstate &&
3352 lwtunnel_fill_encap(skb, rt->dst.lwtstate) < 0)
3353 goto nla_put_failure;
3361 /* add multipath next hop */
3362 static int rt6_add_nexthop(struct sk_buff *skb, struct rt6_info *rt)
3364 struct rtnexthop *rtnh;
3365 unsigned int flags = 0;
3367 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
3369 goto nla_put_failure;
3371 rtnh->rtnh_hops = 0;
3372 rtnh->rtnh_ifindex = rt->dst.dev ? rt->dst.dev->ifindex : 0;
3374 if (rt6_nexthop_info(skb, rt, &flags, true) < 0)
3375 goto nla_put_failure;
3377 rtnh->rtnh_flags = flags;
3379 /* length of rtnetlink header + attributes */
3380 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
3388 static int rt6_fill_node(struct net *net,
3389 struct sk_buff *skb, struct rt6_info *rt,
3390 struct in6_addr *dst, struct in6_addr *src,
3391 int iif, int type, u32 portid, u32 seq,
3394 u32 metrics[RTAX_MAX];
3396 struct nlmsghdr *nlh;
3400 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3404 rtm = nlmsg_data(nlh);
3405 rtm->rtm_family = AF_INET6;
3406 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3407 rtm->rtm_src_len = rt->rt6i_src.plen;
3410 table = rt->rt6i_table->tb6_id;
3412 table = RT6_TABLE_UNSPEC;
3413 rtm->rtm_table = table;
3414 if (nla_put_u32(skb, RTA_TABLE, table))
3415 goto nla_put_failure;
3416 if (rt->rt6i_flags & RTF_REJECT) {
3417 switch (rt->dst.error) {
3419 rtm->rtm_type = RTN_BLACKHOLE;
3422 rtm->rtm_type = RTN_PROHIBIT;
3425 rtm->rtm_type = RTN_THROW;
3428 rtm->rtm_type = RTN_UNREACHABLE;
3432 else if (rt->rt6i_flags & RTF_LOCAL)
3433 rtm->rtm_type = RTN_LOCAL;
3434 else if (rt->rt6i_flags & RTF_ANYCAST)
3435 rtm->rtm_type = RTN_ANYCAST;
3436 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3437 rtm->rtm_type = RTN_LOCAL;
3439 rtm->rtm_type = RTN_UNICAST;
3441 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3442 rtm->rtm_protocol = rt->rt6i_protocol;
3443 if (rt->rt6i_flags & RTF_DYNAMIC)
3444 rtm->rtm_protocol = RTPROT_REDIRECT;
3445 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3446 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3447 rtm->rtm_protocol = RTPROT_RA;
3449 rtm->rtm_protocol = RTPROT_KERNEL;
3452 if (rt->rt6i_flags & RTF_CACHE)
3453 rtm->rtm_flags |= RTM_F_CLONED;
3456 if (nla_put_in6_addr(skb, RTA_DST, dst))
3457 goto nla_put_failure;
3458 rtm->rtm_dst_len = 128;
3459 } else if (rtm->rtm_dst_len)
3460 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3461 goto nla_put_failure;
3462 #ifdef CONFIG_IPV6_SUBTREES
3464 if (nla_put_in6_addr(skb, RTA_SRC, src))
3465 goto nla_put_failure;
3466 rtm->rtm_src_len = 128;
3467 } else if (rtm->rtm_src_len &&
3468 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3469 goto nla_put_failure;
3472 #ifdef CONFIG_IPV6_MROUTE
3473 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3474 int err = ip6mr_get_route(net, skb, rtm, portid);
3479 goto nla_put_failure;
3482 if (nla_put_u32(skb, RTA_IIF, iif))
3483 goto nla_put_failure;
3485 struct in6_addr saddr_buf;
3486 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3487 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3488 goto nla_put_failure;
3491 if (rt->rt6i_prefsrc.plen) {
3492 struct in6_addr saddr_buf;
3493 saddr_buf = rt->rt6i_prefsrc.addr;
3494 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3495 goto nla_put_failure;
3498 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3500 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3501 if (rtnetlink_put_metrics(skb, metrics) < 0)
3502 goto nla_put_failure;
3504 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3505 goto nla_put_failure;
3507 /* For multipath routes, walk the siblings list and add
3508 * each as a nexthop within RTA_MULTIPATH.
3510 if (rt->rt6i_nsiblings) {
3511 struct rt6_info *sibling, *next_sibling;
3514 mp = nla_nest_start(skb, RTA_MULTIPATH);
3516 goto nla_put_failure;
3518 if (rt6_add_nexthop(skb, rt) < 0)
3519 goto nla_put_failure;
3521 list_for_each_entry_safe(sibling, next_sibling,
3522 &rt->rt6i_siblings, rt6i_siblings) {
3523 if (rt6_add_nexthop(skb, sibling) < 0)
3524 goto nla_put_failure;
3527 nla_nest_end(skb, mp);
3529 if (rt6_nexthop_info(skb, rt, &rtm->rtm_flags, false) < 0)
3530 goto nla_put_failure;
3533 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3535 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3536 goto nla_put_failure;
3538 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3539 goto nla_put_failure;
3542 nlmsg_end(skb, nlh);
3546 nlmsg_cancel(skb, nlh);
3550 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3552 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3553 struct net *net = arg->net;
3555 if (rt == net->ipv6.ip6_null_entry)
3558 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3559 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3561 /* user wants prefix routes only */
3562 if (rtm->rtm_flags & RTM_F_PREFIX &&
3563 !(rt->rt6i_flags & RTF_PREFIX_RT)) {
3564 /* success since this is not a prefix route */
3569 return rt6_fill_node(net,
3570 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3571 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3575 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3576 struct netlink_ext_ack *extack)
3578 struct net *net = sock_net(in_skb->sk);
3579 struct nlattr *tb[RTA_MAX+1];
3580 struct rt6_info *rt;
3581 struct sk_buff *skb;
3584 int err, iif = 0, oif = 0;
3586 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
3592 memset(&fl6, 0, sizeof(fl6));
3593 rtm = nlmsg_data(nlh);
3594 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
3597 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3600 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3604 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3607 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3611 iif = nla_get_u32(tb[RTA_IIF]);
3614 oif = nla_get_u32(tb[RTA_OIF]);
3617 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3620 fl6.flowi6_uid = make_kuid(current_user_ns(),
3621 nla_get_u32(tb[RTA_UID]));
3623 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
3626 struct net_device *dev;
3629 dev = __dev_get_by_index(net, iif);
3635 fl6.flowi6_iif = iif;
3637 if (!ipv6_addr_any(&fl6.saddr))
3638 flags |= RT6_LOOKUP_F_HAS_SADDR;
3640 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3643 fl6.flowi6_oif = oif;
3645 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3648 if (rt == net->ipv6.ip6_null_entry) {
3649 err = rt->dst.error;
3654 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3661 skb_dst_set(skb, &rt->dst);
3663 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3664 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3671 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3676 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3677 unsigned int nlm_flags)
3679 struct sk_buff *skb;
3680 struct net *net = info->nl_net;
3685 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3687 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3691 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3692 event, info->portid, seq, nlm_flags);
3694 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3695 WARN_ON(err == -EMSGSIZE);
3699 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3700 info->nlh, gfp_any());
3704 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3707 static int ip6_route_dev_notify(struct notifier_block *this,
3708 unsigned long event, void *ptr)
3710 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3711 struct net *net = dev_net(dev);
3713 if (!(dev->flags & IFF_LOOPBACK))
3716 if (event == NETDEV_REGISTER) {
3717 net->ipv6.ip6_null_entry->dst.dev = dev;
3718 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3719 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3720 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3721 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3722 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3723 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3725 } else if (event == NETDEV_UNREGISTER) {
3726 in6_dev_put(net->ipv6.ip6_null_entry->rt6i_idev);
3727 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3728 in6_dev_put(net->ipv6.ip6_prohibit_entry->rt6i_idev);
3729 in6_dev_put(net->ipv6.ip6_blk_hole_entry->rt6i_idev);
3740 #ifdef CONFIG_PROC_FS
3742 static const struct file_operations ipv6_route_proc_fops = {
3743 .owner = THIS_MODULE,
3744 .open = ipv6_route_open,
3746 .llseek = seq_lseek,
3747 .release = seq_release_net,
3750 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3752 struct net *net = (struct net *)seq->private;
3753 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3754 net->ipv6.rt6_stats->fib_nodes,
3755 net->ipv6.rt6_stats->fib_route_nodes,
3756 net->ipv6.rt6_stats->fib_rt_alloc,
3757 net->ipv6.rt6_stats->fib_rt_entries,
3758 net->ipv6.rt6_stats->fib_rt_cache,
3759 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3760 net->ipv6.rt6_stats->fib_discarded_routes);
3765 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3767 return single_open_net(inode, file, rt6_stats_seq_show);
3770 static const struct file_operations rt6_stats_seq_fops = {
3771 .owner = THIS_MODULE,
3772 .open = rt6_stats_seq_open,
3774 .llseek = seq_lseek,
3775 .release = single_release_net,
3777 #endif /* CONFIG_PROC_FS */
3779 #ifdef CONFIG_SYSCTL
3782 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3783 void __user *buffer, size_t *lenp, loff_t *ppos)
3790 net = (struct net *)ctl->extra1;
3791 delay = net->ipv6.sysctl.flush_delay;
3792 proc_dointvec(ctl, write, buffer, lenp, ppos);
3793 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3797 struct ctl_table ipv6_route_table_template[] = {
3799 .procname = "flush",
3800 .data = &init_net.ipv6.sysctl.flush_delay,
3801 .maxlen = sizeof(int),
3803 .proc_handler = ipv6_sysctl_rtcache_flush
3806 .procname = "gc_thresh",
3807 .data = &ip6_dst_ops_template.gc_thresh,
3808 .maxlen = sizeof(int),
3810 .proc_handler = proc_dointvec,
3813 .procname = "max_size",
3814 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3815 .maxlen = sizeof(int),
3817 .proc_handler = proc_dointvec,
3820 .procname = "gc_min_interval",
3821 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3822 .maxlen = sizeof(int),
3824 .proc_handler = proc_dointvec_jiffies,
3827 .procname = "gc_timeout",
3828 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3829 .maxlen = sizeof(int),
3831 .proc_handler = proc_dointvec_jiffies,
3834 .procname = "gc_interval",
3835 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3836 .maxlen = sizeof(int),
3838 .proc_handler = proc_dointvec_jiffies,
3841 .procname = "gc_elasticity",
3842 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3843 .maxlen = sizeof(int),
3845 .proc_handler = proc_dointvec,
3848 .procname = "mtu_expires",
3849 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3850 .maxlen = sizeof(int),
3852 .proc_handler = proc_dointvec_jiffies,
3855 .procname = "min_adv_mss",
3856 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3857 .maxlen = sizeof(int),
3859 .proc_handler = proc_dointvec,
3862 .procname = "gc_min_interval_ms",
3863 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3864 .maxlen = sizeof(int),
3866 .proc_handler = proc_dointvec_ms_jiffies,
3871 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3873 struct ctl_table *table;
3875 table = kmemdup(ipv6_route_table_template,
3876 sizeof(ipv6_route_table_template),
3880 table[0].data = &net->ipv6.sysctl.flush_delay;
3881 table[0].extra1 = net;
3882 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3883 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3884 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3885 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3886 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3887 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3888 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3889 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3890 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3892 /* Don't export sysctls to unprivileged users */
3893 if (net->user_ns != &init_user_ns)
3894 table[0].procname = NULL;
3901 static int __net_init ip6_route_net_init(struct net *net)
3905 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3906 sizeof(net->ipv6.ip6_dst_ops));
3908 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3909 goto out_ip6_dst_ops;
3911 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3912 sizeof(*net->ipv6.ip6_null_entry),
3914 if (!net->ipv6.ip6_null_entry)
3915 goto out_ip6_dst_entries;
3916 net->ipv6.ip6_null_entry->dst.path =
3917 (struct dst_entry *)net->ipv6.ip6_null_entry;
3918 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3919 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3920 ip6_template_metrics, true);
3922 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3923 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3924 sizeof(*net->ipv6.ip6_prohibit_entry),
3926 if (!net->ipv6.ip6_prohibit_entry)
3927 goto out_ip6_null_entry;
3928 net->ipv6.ip6_prohibit_entry->dst.path =
3929 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3930 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3931 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3932 ip6_template_metrics, true);
3934 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3935 sizeof(*net->ipv6.ip6_blk_hole_entry),
3937 if (!net->ipv6.ip6_blk_hole_entry)
3938 goto out_ip6_prohibit_entry;
3939 net->ipv6.ip6_blk_hole_entry->dst.path =
3940 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3941 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3942 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3943 ip6_template_metrics, true);
3946 net->ipv6.sysctl.flush_delay = 0;
3947 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3948 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3949 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3950 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3951 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3952 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3953 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3955 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3961 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3962 out_ip6_prohibit_entry:
3963 kfree(net->ipv6.ip6_prohibit_entry);
3965 kfree(net->ipv6.ip6_null_entry);
3967 out_ip6_dst_entries:
3968 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3973 static void __net_exit ip6_route_net_exit(struct net *net)
3975 kfree(net->ipv6.ip6_null_entry);
3976 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3977 kfree(net->ipv6.ip6_prohibit_entry);
3978 kfree(net->ipv6.ip6_blk_hole_entry);
3980 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3983 static int __net_init ip6_route_net_init_late(struct net *net)
3985 #ifdef CONFIG_PROC_FS
3986 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3987 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3992 static void __net_exit ip6_route_net_exit_late(struct net *net)
3994 #ifdef CONFIG_PROC_FS
3995 remove_proc_entry("ipv6_route", net->proc_net);
3996 remove_proc_entry("rt6_stats", net->proc_net);
4000 static struct pernet_operations ip6_route_net_ops = {
4001 .init = ip6_route_net_init,
4002 .exit = ip6_route_net_exit,
4005 static int __net_init ipv6_inetpeer_init(struct net *net)
4007 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
4011 inet_peer_base_init(bp);
4012 net->ipv6.peers = bp;
4016 static void __net_exit ipv6_inetpeer_exit(struct net *net)
4018 struct inet_peer_base *bp = net->ipv6.peers;
4020 net->ipv6.peers = NULL;
4021 inetpeer_invalidate_tree(bp);
4025 static struct pernet_operations ipv6_inetpeer_ops = {
4026 .init = ipv6_inetpeer_init,
4027 .exit = ipv6_inetpeer_exit,
4030 static struct pernet_operations ip6_route_net_late_ops = {
4031 .init = ip6_route_net_init_late,
4032 .exit = ip6_route_net_exit_late,
4035 static struct notifier_block ip6_route_dev_notifier = {
4036 .notifier_call = ip6_route_dev_notify,
4037 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
4040 void __init ip6_route_init_special_entries(void)
4042 /* Registering of the loopback is done before this portion of code,
4043 * the loopback reference in rt6_info will not be taken, do it
4044 * manually for init_net */
4045 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
4046 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4047 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4048 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
4049 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4050 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
4051 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4055 int __init ip6_route_init(void)
4061 ip6_dst_ops_template.kmem_cachep =
4062 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
4063 SLAB_HWCACHE_ALIGN, NULL);
4064 if (!ip6_dst_ops_template.kmem_cachep)
4067 ret = dst_entries_init(&ip6_dst_blackhole_ops);
4069 goto out_kmem_cache;
4071 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
4073 goto out_dst_entries;
4075 ret = register_pernet_subsys(&ip6_route_net_ops);
4077 goto out_register_inetpeer;
4079 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
4083 goto out_register_subsys;
4089 ret = fib6_rules_init();
4093 ret = register_pernet_subsys(&ip6_route_net_late_ops);
4095 goto fib6_rules_init;
4098 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
4099 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
4100 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
4101 goto out_register_late_subsys;
4103 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
4105 goto out_register_late_subsys;
4107 for_each_possible_cpu(cpu) {
4108 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
4110 INIT_LIST_HEAD(&ul->head);
4111 spin_lock_init(&ul->lock);
4117 out_register_late_subsys:
4118 unregister_pernet_subsys(&ip6_route_net_late_ops);
4120 fib6_rules_cleanup();
4125 out_register_subsys:
4126 unregister_pernet_subsys(&ip6_route_net_ops);
4127 out_register_inetpeer:
4128 unregister_pernet_subsys(&ipv6_inetpeer_ops);
4130 dst_entries_destroy(&ip6_dst_blackhole_ops);
4132 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
4136 void ip6_route_cleanup(void)
4138 unregister_netdevice_notifier(&ip6_route_dev_notifier);
4139 unregister_pernet_subsys(&ip6_route_net_late_ops);
4140 fib6_rules_cleanup();
4143 unregister_pernet_subsys(&ipv6_inetpeer_ops);
4144 unregister_pernet_subsys(&ip6_route_net_ops);
4145 dst_entries_destroy(&ip6_dst_blackhole_ops);
4146 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);