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,
942 struct netlink_ext_ack *extack)
945 struct fib6_table *table;
947 table = rt->rt6i_table;
948 write_lock_bh(&table->tb6_lock);
949 err = fib6_add(&table->tb6_root, rt, info, mxc, extack);
950 write_unlock_bh(&table->tb6_lock);
955 int ip6_ins_rt(struct rt6_info *rt)
957 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
958 struct mx6_config mxc = { .mx = NULL, };
960 return __ip6_ins_rt(rt, &info, &mxc, NULL);
963 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
964 const struct in6_addr *daddr,
965 const struct in6_addr *saddr)
973 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
974 ort = (struct rt6_info *)ort->dst.from;
976 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
981 ip6_rt_copy_init(rt, ort);
982 rt->rt6i_flags |= RTF_CACHE;
984 rt->dst.flags |= DST_HOST;
985 rt->rt6i_dst.addr = *daddr;
986 rt->rt6i_dst.plen = 128;
988 if (!rt6_is_gw_or_nonexthop(ort)) {
989 if (ort->rt6i_dst.plen != 128 &&
990 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
991 rt->rt6i_flags |= RTF_ANYCAST;
992 #ifdef CONFIG_IPV6_SUBTREES
993 if (rt->rt6i_src.plen && saddr) {
994 rt->rt6i_src.addr = *saddr;
995 rt->rt6i_src.plen = 128;
1003 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
1005 struct rt6_info *pcpu_rt;
1007 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
1008 rt->dst.dev, rt->dst.flags);
1012 ip6_rt_copy_init(pcpu_rt, rt);
1013 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
1014 pcpu_rt->rt6i_flags |= RTF_PCPU;
1018 /* It should be called with read_lock_bh(&tb6_lock) acquired */
1019 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
1021 struct rt6_info *pcpu_rt, **p;
1023 p = this_cpu_ptr(rt->rt6i_pcpu);
1027 dst_hold(&pcpu_rt->dst);
1028 rt6_dst_from_metrics_check(pcpu_rt);
1033 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1035 struct fib6_table *table = rt->rt6i_table;
1036 struct rt6_info *pcpu_rt, *prev, **p;
1038 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1040 struct net *net = dev_net(rt->dst.dev);
1042 dst_hold(&net->ipv6.ip6_null_entry->dst);
1043 return net->ipv6.ip6_null_entry;
1046 read_lock_bh(&table->tb6_lock);
1047 if (rt->rt6i_pcpu) {
1048 p = this_cpu_ptr(rt->rt6i_pcpu);
1049 prev = cmpxchg(p, NULL, pcpu_rt);
1051 /* If someone did it before us, return prev instead */
1052 dst_destroy(&pcpu_rt->dst);
1056 /* rt has been removed from the fib6 tree
1057 * before we have a chance to acquire the read_lock.
1058 * In this case, don't brother to create a pcpu rt
1059 * since rt is going away anyway. The next
1060 * dst_check() will trigger a re-lookup.
1062 dst_destroy(&pcpu_rt->dst);
1065 dst_hold(&pcpu_rt->dst);
1066 rt6_dst_from_metrics_check(pcpu_rt);
1067 read_unlock_bh(&table->tb6_lock);
1071 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1072 int oif, struct flowi6 *fl6, int flags)
1074 struct fib6_node *fn, *saved_fn;
1075 struct rt6_info *rt;
1078 strict |= flags & RT6_LOOKUP_F_IFACE;
1079 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
1080 if (net->ipv6.devconf_all->forwarding == 0)
1081 strict |= RT6_LOOKUP_F_REACHABLE;
1083 read_lock_bh(&table->tb6_lock);
1085 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1088 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1092 rt = rt6_select(fn, oif, strict);
1093 if (rt->rt6i_nsiblings)
1094 rt = rt6_multipath_select(rt, fl6, oif, strict);
1095 if (rt == net->ipv6.ip6_null_entry) {
1096 fn = fib6_backtrack(fn, &fl6->saddr);
1098 goto redo_rt6_select;
1099 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1100 /* also consider unreachable route */
1101 strict &= ~RT6_LOOKUP_F_REACHABLE;
1103 goto redo_rt6_select;
1108 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1109 dst_use(&rt->dst, jiffies);
1110 read_unlock_bh(&table->tb6_lock);
1112 rt6_dst_from_metrics_check(rt);
1114 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1116 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1117 !(rt->rt6i_flags & RTF_GATEWAY))) {
1118 /* Create a RTF_CACHE clone which will not be
1119 * owned by the fib6 tree. It is for the special case where
1120 * the daddr in the skb during the neighbor look-up is different
1121 * from the fl6->daddr used to look-up route here.
1124 struct rt6_info *uncached_rt;
1126 dst_use(&rt->dst, jiffies);
1127 read_unlock_bh(&table->tb6_lock);
1129 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1130 dst_release(&rt->dst);
1133 rt6_uncached_list_add(uncached_rt);
1135 uncached_rt = net->ipv6.ip6_null_entry;
1137 dst_hold(&uncached_rt->dst);
1139 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1143 /* Get a percpu copy */
1145 struct rt6_info *pcpu_rt;
1147 rt->dst.lastuse = jiffies;
1149 pcpu_rt = rt6_get_pcpu_route(rt);
1152 read_unlock_bh(&table->tb6_lock);
1154 /* We have to do the read_unlock first
1155 * because rt6_make_pcpu_route() may trigger
1156 * ip6_dst_gc() which will take the write_lock.
1159 read_unlock_bh(&table->tb6_lock);
1160 pcpu_rt = rt6_make_pcpu_route(rt);
1161 dst_release(&rt->dst);
1164 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1169 EXPORT_SYMBOL_GPL(ip6_pol_route);
1171 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1172 struct flowi6 *fl6, int flags)
1174 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1177 struct dst_entry *ip6_route_input_lookup(struct net *net,
1178 struct net_device *dev,
1179 struct flowi6 *fl6, int flags)
1181 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1182 flags |= RT6_LOOKUP_F_IFACE;
1184 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1186 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
1188 void ip6_route_input(struct sk_buff *skb)
1190 const struct ipv6hdr *iph = ipv6_hdr(skb);
1191 struct net *net = dev_net(skb->dev);
1192 int flags = RT6_LOOKUP_F_HAS_SADDR;
1193 struct ip_tunnel_info *tun_info;
1194 struct flowi6 fl6 = {
1195 .flowi6_iif = skb->dev->ifindex,
1196 .daddr = iph->daddr,
1197 .saddr = iph->saddr,
1198 .flowlabel = ip6_flowinfo(iph),
1199 .flowi6_mark = skb->mark,
1200 .flowi6_proto = iph->nexthdr,
1203 tun_info = skb_tunnel_info(skb);
1204 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1205 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1207 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1210 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1211 struct flowi6 *fl6, int flags)
1213 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1216 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1217 struct flowi6 *fl6, int flags)
1221 if (rt6_need_strict(&fl6->daddr)) {
1222 struct dst_entry *dst;
1224 dst = l3mdev_link_scope_lookup(net, fl6);
1229 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1231 any_src = ipv6_addr_any(&fl6->saddr);
1232 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1233 (fl6->flowi6_oif && any_src))
1234 flags |= RT6_LOOKUP_F_IFACE;
1237 flags |= RT6_LOOKUP_F_HAS_SADDR;
1239 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1241 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1243 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1245 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1247 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1248 struct dst_entry *new = NULL;
1250 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1256 new->input = dst_discard;
1257 new->output = dst_discard_out;
1259 dst_copy_metrics(new, &ort->dst);
1260 rt->rt6i_idev = ort->rt6i_idev;
1262 in6_dev_hold(rt->rt6i_idev);
1264 rt->rt6i_gateway = ort->rt6i_gateway;
1265 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1266 rt->rt6i_metric = 0;
1268 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1269 #ifdef CONFIG_IPV6_SUBTREES
1270 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1276 dst_release(dst_orig);
1277 return new ? new : ERR_PTR(-ENOMEM);
1281 * Destination cache support functions
1284 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1287 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1288 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1291 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1293 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1296 if (rt6_check_expired(rt))
1302 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1304 if (!__rt6_check_expired(rt) &&
1305 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1306 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1312 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1314 struct rt6_info *rt;
1316 rt = (struct rt6_info *) dst;
1318 /* All IPV6 dsts are created with ->obsolete set to the value
1319 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1320 * into this function always.
1323 rt6_dst_from_metrics_check(rt);
1325 if (rt->rt6i_flags & RTF_PCPU ||
1326 (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1327 return rt6_dst_from_check(rt, cookie);
1329 return rt6_check(rt, cookie);
1332 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1334 struct rt6_info *rt = (struct rt6_info *) dst;
1337 if (rt->rt6i_flags & RTF_CACHE) {
1338 if (rt6_check_expired(rt)) {
1350 static void ip6_link_failure(struct sk_buff *skb)
1352 struct rt6_info *rt;
1354 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1356 rt = (struct rt6_info *) skb_dst(skb);
1358 if (rt->rt6i_flags & RTF_CACHE) {
1361 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1362 rt->rt6i_node->fn_sernum = -1;
1367 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1369 struct net *net = dev_net(rt->dst.dev);
1371 rt->rt6i_flags |= RTF_MODIFIED;
1372 rt->rt6i_pmtu = mtu;
1373 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1376 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1378 return !(rt->rt6i_flags & RTF_CACHE) &&
1379 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1382 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1383 const struct ipv6hdr *iph, u32 mtu)
1385 const struct in6_addr *daddr, *saddr;
1386 struct rt6_info *rt6 = (struct rt6_info *)dst;
1388 if (rt6->rt6i_flags & RTF_LOCAL)
1391 if (dst_metric_locked(dst, RTAX_MTU))
1395 daddr = &iph->daddr;
1396 saddr = &iph->saddr;
1398 daddr = &sk->sk_v6_daddr;
1399 saddr = &inet6_sk(sk)->saddr;
1404 dst_confirm_neigh(dst, daddr);
1405 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1406 if (mtu >= dst_mtu(dst))
1409 if (!rt6_cache_allowed_for_pmtu(rt6)) {
1410 rt6_do_update_pmtu(rt6, mtu);
1412 struct rt6_info *nrt6;
1414 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1416 rt6_do_update_pmtu(nrt6, mtu);
1418 /* ip6_ins_rt(nrt6) will bump the
1419 * rt6->rt6i_node->fn_sernum
1420 * which will fail the next rt6_check() and
1421 * invalidate the sk->sk_dst_cache.
1428 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1429 struct sk_buff *skb, u32 mtu)
1431 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1434 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1435 int oif, u32 mark, kuid_t uid)
1437 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1438 struct dst_entry *dst;
1441 memset(&fl6, 0, sizeof(fl6));
1442 fl6.flowi6_oif = oif;
1443 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1444 fl6.daddr = iph->daddr;
1445 fl6.saddr = iph->saddr;
1446 fl6.flowlabel = ip6_flowinfo(iph);
1447 fl6.flowi6_uid = uid;
1449 dst = ip6_route_output(net, NULL, &fl6);
1451 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1454 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1456 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1458 struct dst_entry *dst;
1460 ip6_update_pmtu(skb, sock_net(sk), mtu,
1461 sk->sk_bound_dev_if, sk->sk_mark, sk->sk_uid);
1463 dst = __sk_dst_get(sk);
1464 if (!dst || !dst->obsolete ||
1465 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1469 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1470 ip6_datagram_dst_update(sk, false);
1473 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1475 /* Handle redirects */
1476 struct ip6rd_flowi {
1478 struct in6_addr gateway;
1481 static struct rt6_info *__ip6_route_redirect(struct net *net,
1482 struct fib6_table *table,
1486 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1487 struct rt6_info *rt;
1488 struct fib6_node *fn;
1490 /* Get the "current" route for this destination and
1491 * check if the redirect has come from appropriate router.
1493 * RFC 4861 specifies that redirects should only be
1494 * accepted if they come from the nexthop to the target.
1495 * Due to the way the routes are chosen, this notion
1496 * is a bit fuzzy and one might need to check all possible
1500 read_lock_bh(&table->tb6_lock);
1501 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1503 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1504 if (rt6_check_expired(rt))
1508 if (!(rt->rt6i_flags & RTF_GATEWAY))
1510 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1512 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1518 rt = net->ipv6.ip6_null_entry;
1519 else if (rt->dst.error) {
1520 rt = net->ipv6.ip6_null_entry;
1524 if (rt == net->ipv6.ip6_null_entry) {
1525 fn = fib6_backtrack(fn, &fl6->saddr);
1533 read_unlock_bh(&table->tb6_lock);
1535 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1539 static struct dst_entry *ip6_route_redirect(struct net *net,
1540 const struct flowi6 *fl6,
1541 const struct in6_addr *gateway)
1543 int flags = RT6_LOOKUP_F_HAS_SADDR;
1544 struct ip6rd_flowi rdfl;
1547 rdfl.gateway = *gateway;
1549 return fib6_rule_lookup(net, &rdfl.fl6,
1550 flags, __ip6_route_redirect);
1553 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
1556 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1557 struct dst_entry *dst;
1560 memset(&fl6, 0, sizeof(fl6));
1561 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1562 fl6.flowi6_oif = oif;
1563 fl6.flowi6_mark = mark;
1564 fl6.daddr = iph->daddr;
1565 fl6.saddr = iph->saddr;
1566 fl6.flowlabel = ip6_flowinfo(iph);
1567 fl6.flowi6_uid = uid;
1569 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1570 rt6_do_redirect(dst, NULL, skb);
1573 EXPORT_SYMBOL_GPL(ip6_redirect);
1575 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1578 const struct ipv6hdr *iph = ipv6_hdr(skb);
1579 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1580 struct dst_entry *dst;
1583 memset(&fl6, 0, sizeof(fl6));
1584 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1585 fl6.flowi6_oif = oif;
1586 fl6.flowi6_mark = mark;
1587 fl6.daddr = msg->dest;
1588 fl6.saddr = iph->daddr;
1589 fl6.flowi6_uid = sock_net_uid(net, NULL);
1591 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1592 rt6_do_redirect(dst, NULL, skb);
1596 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1598 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
1601 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1603 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1605 struct net_device *dev = dst->dev;
1606 unsigned int mtu = dst_mtu(dst);
1607 struct net *net = dev_net(dev);
1609 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1611 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1612 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1615 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1616 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1617 * IPV6_MAXPLEN is also valid and means: "any MSS,
1618 * rely only on pmtu discovery"
1620 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1625 static unsigned int ip6_mtu(const struct dst_entry *dst)
1627 const struct rt6_info *rt = (const struct rt6_info *)dst;
1628 unsigned int mtu = rt->rt6i_pmtu;
1629 struct inet6_dev *idev;
1634 mtu = dst_metric_raw(dst, RTAX_MTU);
1641 idev = __in6_dev_get(dst->dev);
1643 mtu = idev->cnf.mtu6;
1647 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1649 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
1652 static struct dst_entry *icmp6_dst_gc_list;
1653 static DEFINE_SPINLOCK(icmp6_dst_lock);
1655 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1658 struct dst_entry *dst;
1659 struct rt6_info *rt;
1660 struct inet6_dev *idev = in6_dev_get(dev);
1661 struct net *net = dev_net(dev);
1663 if (unlikely(!idev))
1664 return ERR_PTR(-ENODEV);
1666 rt = ip6_dst_alloc(net, dev, 0);
1667 if (unlikely(!rt)) {
1669 dst = ERR_PTR(-ENOMEM);
1673 rt->dst.flags |= DST_HOST;
1674 rt->dst.output = ip6_output;
1675 atomic_set(&rt->dst.__refcnt, 1);
1676 rt->rt6i_gateway = fl6->daddr;
1677 rt->rt6i_dst.addr = fl6->daddr;
1678 rt->rt6i_dst.plen = 128;
1679 rt->rt6i_idev = idev;
1680 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1682 spin_lock_bh(&icmp6_dst_lock);
1683 rt->dst.next = icmp6_dst_gc_list;
1684 icmp6_dst_gc_list = &rt->dst;
1685 spin_unlock_bh(&icmp6_dst_lock);
1687 fib6_force_start_gc(net);
1689 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1695 int icmp6_dst_gc(void)
1697 struct dst_entry *dst, **pprev;
1700 spin_lock_bh(&icmp6_dst_lock);
1701 pprev = &icmp6_dst_gc_list;
1703 while ((dst = *pprev) != NULL) {
1704 if (!atomic_read(&dst->__refcnt)) {
1713 spin_unlock_bh(&icmp6_dst_lock);
1718 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1721 struct dst_entry *dst, **pprev;
1723 spin_lock_bh(&icmp6_dst_lock);
1724 pprev = &icmp6_dst_gc_list;
1725 while ((dst = *pprev) != NULL) {
1726 struct rt6_info *rt = (struct rt6_info *) dst;
1727 if (func(rt, arg)) {
1734 spin_unlock_bh(&icmp6_dst_lock);
1737 static int ip6_dst_gc(struct dst_ops *ops)
1739 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1740 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1741 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1742 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1743 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1744 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1747 entries = dst_entries_get_fast(ops);
1748 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1749 entries <= rt_max_size)
1752 net->ipv6.ip6_rt_gc_expire++;
1753 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1754 entries = dst_entries_get_slow(ops);
1755 if (entries < ops->gc_thresh)
1756 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1758 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1759 return entries > rt_max_size;
1762 static int ip6_convert_metrics(struct mx6_config *mxc,
1763 const struct fib6_config *cfg)
1765 bool ecn_ca = false;
1773 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1777 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1778 int type = nla_type(nla);
1783 if (unlikely(type > RTAX_MAX))
1786 if (type == RTAX_CC_ALGO) {
1787 char tmp[TCP_CA_NAME_MAX];
1789 nla_strlcpy(tmp, nla, sizeof(tmp));
1790 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1791 if (val == TCP_CA_UNSPEC)
1794 val = nla_get_u32(nla);
1796 if (type == RTAX_HOPLIMIT && val > 255)
1798 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1802 __set_bit(type - 1, mxc->mx_valid);
1806 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1807 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1817 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
1818 struct fib6_config *cfg,
1819 const struct in6_addr *gw_addr)
1821 struct flowi6 fl6 = {
1822 .flowi6_oif = cfg->fc_ifindex,
1824 .saddr = cfg->fc_prefsrc,
1826 struct fib6_table *table;
1827 struct rt6_info *rt;
1828 int flags = RT6_LOOKUP_F_IFACE | RT6_LOOKUP_F_IGNORE_LINKSTATE;
1830 table = fib6_get_table(net, cfg->fc_table);
1834 if (!ipv6_addr_any(&cfg->fc_prefsrc))
1835 flags |= RT6_LOOKUP_F_HAS_SADDR;
1837 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
1839 /* if table lookup failed, fall back to full lookup */
1840 if (rt == net->ipv6.ip6_null_entry) {
1848 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg,
1849 struct netlink_ext_ack *extack)
1851 struct net *net = cfg->fc_nlinfo.nl_net;
1852 struct rt6_info *rt = NULL;
1853 struct net_device *dev = NULL;
1854 struct inet6_dev *idev = NULL;
1855 struct fib6_table *table;
1859 /* RTF_PCPU is an internal flag; can not be set by userspace */
1860 if (cfg->fc_flags & RTF_PCPU)
1863 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1865 #ifndef CONFIG_IPV6_SUBTREES
1866 if (cfg->fc_src_len)
1869 if (cfg->fc_ifindex) {
1871 dev = dev_get_by_index(net, cfg->fc_ifindex);
1874 idev = in6_dev_get(dev);
1879 if (cfg->fc_metric == 0)
1880 cfg->fc_metric = IP6_RT_PRIO_USER;
1883 if (cfg->fc_nlinfo.nlh &&
1884 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1885 table = fib6_get_table(net, cfg->fc_table);
1887 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1888 table = fib6_new_table(net, cfg->fc_table);
1891 table = fib6_new_table(net, cfg->fc_table);
1897 rt = ip6_dst_alloc(net, NULL,
1898 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1905 if (cfg->fc_flags & RTF_EXPIRES)
1906 rt6_set_expires(rt, jiffies +
1907 clock_t_to_jiffies(cfg->fc_expires));
1909 rt6_clean_expires(rt);
1911 if (cfg->fc_protocol == RTPROT_UNSPEC)
1912 cfg->fc_protocol = RTPROT_BOOT;
1913 rt->rt6i_protocol = cfg->fc_protocol;
1915 addr_type = ipv6_addr_type(&cfg->fc_dst);
1917 if (addr_type & IPV6_ADDR_MULTICAST)
1918 rt->dst.input = ip6_mc_input;
1919 else if (cfg->fc_flags & RTF_LOCAL)
1920 rt->dst.input = ip6_input;
1922 rt->dst.input = ip6_forward;
1924 rt->dst.output = ip6_output;
1926 if (cfg->fc_encap) {
1927 struct lwtunnel_state *lwtstate;
1929 err = lwtunnel_build_state(cfg->fc_encap_type,
1930 cfg->fc_encap, AF_INET6, cfg,
1934 rt->dst.lwtstate = lwtstate_get(lwtstate);
1935 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1936 rt->dst.lwtstate->orig_output = rt->dst.output;
1937 rt->dst.output = lwtunnel_output;
1939 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1940 rt->dst.lwtstate->orig_input = rt->dst.input;
1941 rt->dst.input = lwtunnel_input;
1945 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1946 rt->rt6i_dst.plen = cfg->fc_dst_len;
1947 if (rt->rt6i_dst.plen == 128)
1948 rt->dst.flags |= DST_HOST;
1950 #ifdef CONFIG_IPV6_SUBTREES
1951 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1952 rt->rt6i_src.plen = cfg->fc_src_len;
1955 rt->rt6i_metric = cfg->fc_metric;
1957 /* We cannot add true routes via loopback here,
1958 they would result in kernel looping; promote them to reject routes
1960 if ((cfg->fc_flags & RTF_REJECT) ||
1961 (dev && (dev->flags & IFF_LOOPBACK) &&
1962 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1963 !(cfg->fc_flags & RTF_LOCAL))) {
1964 /* hold loopback dev/idev if we haven't done so. */
1965 if (dev != net->loopback_dev) {
1970 dev = net->loopback_dev;
1972 idev = in6_dev_get(dev);
1978 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1979 switch (cfg->fc_type) {
1981 rt->dst.error = -EINVAL;
1982 rt->dst.output = dst_discard_out;
1983 rt->dst.input = dst_discard;
1986 rt->dst.error = -EACCES;
1987 rt->dst.output = ip6_pkt_prohibit_out;
1988 rt->dst.input = ip6_pkt_prohibit;
1991 case RTN_UNREACHABLE:
1993 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1994 : (cfg->fc_type == RTN_UNREACHABLE)
1995 ? -EHOSTUNREACH : -ENETUNREACH;
1996 rt->dst.output = ip6_pkt_discard_out;
1997 rt->dst.input = ip6_pkt_discard;
2003 if (cfg->fc_flags & RTF_GATEWAY) {
2004 const struct in6_addr *gw_addr;
2007 gw_addr = &cfg->fc_gateway;
2008 gwa_type = ipv6_addr_type(gw_addr);
2010 /* if gw_addr is local we will fail to detect this in case
2011 * address is still TENTATIVE (DAD in progress). rt6_lookup()
2012 * will return already-added prefix route via interface that
2013 * prefix route was assigned to, which might be non-loopback.
2016 if (ipv6_chk_addr_and_flags(net, gw_addr,
2017 gwa_type & IPV6_ADDR_LINKLOCAL ?
2021 rt->rt6i_gateway = *gw_addr;
2023 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
2024 struct rt6_info *grt = NULL;
2026 /* IPv6 strictly inhibits using not link-local
2027 addresses as nexthop address.
2028 Otherwise, router will not able to send redirects.
2029 It is very good, but in some (rare!) circumstances
2030 (SIT, PtP, NBMA NOARP links) it is handy to allow
2031 some exceptions. --ANK
2032 We allow IPv4-mapped nexthops to support RFC4798-type
2035 if (!(gwa_type & (IPV6_ADDR_UNICAST |
2039 if (cfg->fc_table) {
2040 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
2043 if (grt->rt6i_flags & RTF_GATEWAY ||
2044 (dev && dev != grt->dst.dev)) {
2052 grt = rt6_lookup(net, gw_addr, NULL,
2053 cfg->fc_ifindex, 1);
2055 err = -EHOSTUNREACH;
2059 if (dev != grt->dst.dev) {
2065 idev = grt->rt6i_idev;
2067 in6_dev_hold(grt->rt6i_idev);
2069 if (!(grt->rt6i_flags & RTF_GATEWAY))
2077 if (!dev || (dev->flags & IFF_LOOPBACK))
2085 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2086 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2090 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2091 rt->rt6i_prefsrc.plen = 128;
2093 rt->rt6i_prefsrc.plen = 0;
2095 rt->rt6i_flags = cfg->fc_flags;
2099 rt->rt6i_idev = idev;
2100 rt->rt6i_table = table;
2102 cfg->fc_nlinfo.nl_net = dev_net(dev);
2113 return ERR_PTR(err);
2116 int ip6_route_add(struct fib6_config *cfg,
2117 struct netlink_ext_ack *extack)
2119 struct mx6_config mxc = { .mx = NULL, };
2120 struct rt6_info *rt;
2123 rt = ip6_route_info_create(cfg, extack);
2130 err = ip6_convert_metrics(&mxc, cfg);
2134 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc, extack);
2146 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2149 struct fib6_table *table;
2150 struct net *net = dev_net(rt->dst.dev);
2152 if (rt == net->ipv6.ip6_null_entry ||
2153 rt->dst.flags & DST_NOCACHE) {
2158 table = rt->rt6i_table;
2159 write_lock_bh(&table->tb6_lock);
2160 err = fib6_del(rt, info);
2161 write_unlock_bh(&table->tb6_lock);
2168 int ip6_del_rt(struct rt6_info *rt)
2170 struct nl_info info = {
2171 .nl_net = dev_net(rt->dst.dev),
2173 return __ip6_del_rt(rt, &info);
2176 static int __ip6_del_rt_siblings(struct rt6_info *rt, struct fib6_config *cfg)
2178 struct nl_info *info = &cfg->fc_nlinfo;
2179 struct net *net = info->nl_net;
2180 struct sk_buff *skb = NULL;
2181 struct fib6_table *table;
2184 if (rt == net->ipv6.ip6_null_entry)
2186 table = rt->rt6i_table;
2187 write_lock_bh(&table->tb6_lock);
2189 if (rt->rt6i_nsiblings && cfg->fc_delete_all_nh) {
2190 struct rt6_info *sibling, *next_sibling;
2192 /* prefer to send a single notification with all hops */
2193 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
2195 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2197 if (rt6_fill_node(net, skb, rt,
2198 NULL, NULL, 0, RTM_DELROUTE,
2199 info->portid, seq, 0) < 0) {
2203 info->skip_notify = 1;
2206 list_for_each_entry_safe(sibling, next_sibling,
2209 err = fib6_del(sibling, info);
2215 err = fib6_del(rt, info);
2217 write_unlock_bh(&table->tb6_lock);
2222 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2223 info->nlh, gfp_any());
2228 static int ip6_route_del(struct fib6_config *cfg,
2229 struct netlink_ext_ack *extack)
2231 struct fib6_table *table;
2232 struct fib6_node *fn;
2233 struct rt6_info *rt;
2236 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2240 read_lock_bh(&table->tb6_lock);
2242 fn = fib6_locate(&table->tb6_root,
2243 &cfg->fc_dst, cfg->fc_dst_len,
2244 &cfg->fc_src, cfg->fc_src_len);
2247 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2248 if ((rt->rt6i_flags & RTF_CACHE) &&
2249 !(cfg->fc_flags & RTF_CACHE))
2251 if (cfg->fc_ifindex &&
2253 rt->dst.dev->ifindex != cfg->fc_ifindex))
2255 if (cfg->fc_flags & RTF_GATEWAY &&
2256 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2258 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2260 if (cfg->fc_protocol && cfg->fc_protocol != rt->rt6i_protocol)
2263 read_unlock_bh(&table->tb6_lock);
2265 /* if gateway was specified only delete the one hop */
2266 if (cfg->fc_flags & RTF_GATEWAY)
2267 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2269 return __ip6_del_rt_siblings(rt, cfg);
2272 read_unlock_bh(&table->tb6_lock);
2277 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2279 struct netevent_redirect netevent;
2280 struct rt6_info *rt, *nrt = NULL;
2281 struct ndisc_options ndopts;
2282 struct inet6_dev *in6_dev;
2283 struct neighbour *neigh;
2285 int optlen, on_link;
2288 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2289 optlen -= sizeof(*msg);
2292 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2296 msg = (struct rd_msg *)icmp6_hdr(skb);
2298 if (ipv6_addr_is_multicast(&msg->dest)) {
2299 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2304 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2306 } else if (ipv6_addr_type(&msg->target) !=
2307 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2308 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2312 in6_dev = __in6_dev_get(skb->dev);
2315 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2319 * The IP source address of the Redirect MUST be the same as the current
2320 * first-hop router for the specified ICMP Destination Address.
2323 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
2324 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2329 if (ndopts.nd_opts_tgt_lladdr) {
2330 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2333 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2338 rt = (struct rt6_info *) dst;
2339 if (rt->rt6i_flags & RTF_REJECT) {
2340 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2344 /* Redirect received -> path was valid.
2345 * Look, redirects are sent only in response to data packets,
2346 * so that this nexthop apparently is reachable. --ANK
2348 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
2350 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2355 * We have finally decided to accept it.
2358 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
2359 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2360 NEIGH_UPDATE_F_OVERRIDE|
2361 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2362 NEIGH_UPDATE_F_ISROUTER)),
2363 NDISC_REDIRECT, &ndopts);
2365 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2369 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2371 nrt->rt6i_flags &= ~RTF_GATEWAY;
2373 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2375 if (ip6_ins_rt(nrt))
2378 netevent.old = &rt->dst;
2379 netevent.new = &nrt->dst;
2380 netevent.daddr = &msg->dest;
2381 netevent.neigh = neigh;
2382 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2384 if (rt->rt6i_flags & RTF_CACHE) {
2385 rt = (struct rt6_info *) dst_clone(&rt->dst);
2390 neigh_release(neigh);
2394 * Misc support functions
2397 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2399 BUG_ON(from->dst.from);
2401 rt->rt6i_flags &= ~RTF_EXPIRES;
2402 dst_hold(&from->dst);
2403 rt->dst.from = &from->dst;
2404 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2407 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2409 rt->dst.input = ort->dst.input;
2410 rt->dst.output = ort->dst.output;
2411 rt->rt6i_dst = ort->rt6i_dst;
2412 rt->dst.error = ort->dst.error;
2413 rt->rt6i_idev = ort->rt6i_idev;
2415 in6_dev_hold(rt->rt6i_idev);
2416 rt->dst.lastuse = jiffies;
2417 rt->rt6i_gateway = ort->rt6i_gateway;
2418 rt->rt6i_flags = ort->rt6i_flags;
2419 rt6_set_from(rt, ort);
2420 rt->rt6i_metric = ort->rt6i_metric;
2421 #ifdef CONFIG_IPV6_SUBTREES
2422 rt->rt6i_src = ort->rt6i_src;
2424 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2425 rt->rt6i_table = ort->rt6i_table;
2426 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2429 #ifdef CONFIG_IPV6_ROUTE_INFO
2430 static struct rt6_info *rt6_get_route_info(struct net *net,
2431 const struct in6_addr *prefix, int prefixlen,
2432 const struct in6_addr *gwaddr,
2433 struct net_device *dev)
2435 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
2436 int ifindex = dev->ifindex;
2437 struct fib6_node *fn;
2438 struct rt6_info *rt = NULL;
2439 struct fib6_table *table;
2441 table = fib6_get_table(net, tb_id);
2445 read_lock_bh(&table->tb6_lock);
2446 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2450 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2451 if (rt->dst.dev->ifindex != ifindex)
2453 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2455 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2461 read_unlock_bh(&table->tb6_lock);
2465 static struct rt6_info *rt6_add_route_info(struct net *net,
2466 const struct in6_addr *prefix, int prefixlen,
2467 const struct in6_addr *gwaddr,
2468 struct net_device *dev,
2471 struct fib6_config cfg = {
2472 .fc_metric = IP6_RT_PRIO_USER,
2473 .fc_ifindex = dev->ifindex,
2474 .fc_dst_len = prefixlen,
2475 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2476 RTF_UP | RTF_PREF(pref),
2477 .fc_nlinfo.portid = 0,
2478 .fc_nlinfo.nlh = NULL,
2479 .fc_nlinfo.nl_net = net,
2482 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
2483 cfg.fc_dst = *prefix;
2484 cfg.fc_gateway = *gwaddr;
2486 /* We should treat it as a default route if prefix length is 0. */
2488 cfg.fc_flags |= RTF_DEFAULT;
2490 ip6_route_add(&cfg, NULL);
2492 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
2496 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2498 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
2499 struct rt6_info *rt;
2500 struct fib6_table *table;
2502 table = fib6_get_table(dev_net(dev), tb_id);
2506 read_lock_bh(&table->tb6_lock);
2507 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2508 if (dev == rt->dst.dev &&
2509 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2510 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2515 read_unlock_bh(&table->tb6_lock);
2519 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2520 struct net_device *dev,
2523 struct fib6_config cfg = {
2524 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2525 .fc_metric = IP6_RT_PRIO_USER,
2526 .fc_ifindex = dev->ifindex,
2527 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2528 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2529 .fc_nlinfo.portid = 0,
2530 .fc_nlinfo.nlh = NULL,
2531 .fc_nlinfo.nl_net = dev_net(dev),
2534 cfg.fc_gateway = *gwaddr;
2536 if (!ip6_route_add(&cfg, NULL)) {
2537 struct fib6_table *table;
2539 table = fib6_get_table(dev_net(dev), cfg.fc_table);
2541 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
2544 return rt6_get_dflt_router(gwaddr, dev);
2547 static void __rt6_purge_dflt_routers(struct fib6_table *table)
2549 struct rt6_info *rt;
2552 read_lock_bh(&table->tb6_lock);
2553 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2554 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2555 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2557 read_unlock_bh(&table->tb6_lock);
2562 read_unlock_bh(&table->tb6_lock);
2564 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
2567 void rt6_purge_dflt_routers(struct net *net)
2569 struct fib6_table *table;
2570 struct hlist_head *head;
2575 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
2576 head = &net->ipv6.fib_table_hash[h];
2577 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
2578 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
2579 __rt6_purge_dflt_routers(table);
2586 static void rtmsg_to_fib6_config(struct net *net,
2587 struct in6_rtmsg *rtmsg,
2588 struct fib6_config *cfg)
2590 memset(cfg, 0, sizeof(*cfg));
2592 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2594 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2595 cfg->fc_metric = rtmsg->rtmsg_metric;
2596 cfg->fc_expires = rtmsg->rtmsg_info;
2597 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2598 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2599 cfg->fc_flags = rtmsg->rtmsg_flags;
2601 cfg->fc_nlinfo.nl_net = net;
2603 cfg->fc_dst = rtmsg->rtmsg_dst;
2604 cfg->fc_src = rtmsg->rtmsg_src;
2605 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2608 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2610 struct fib6_config cfg;
2611 struct in6_rtmsg rtmsg;
2615 case SIOCADDRT: /* Add a route */
2616 case SIOCDELRT: /* Delete a route */
2617 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2619 err = copy_from_user(&rtmsg, arg,
2620 sizeof(struct in6_rtmsg));
2624 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2629 err = ip6_route_add(&cfg, NULL);
2632 err = ip6_route_del(&cfg, NULL);
2646 * Drop the packet on the floor
2649 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2652 struct dst_entry *dst = skb_dst(skb);
2653 switch (ipstats_mib_noroutes) {
2654 case IPSTATS_MIB_INNOROUTES:
2655 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2656 if (type == IPV6_ADDR_ANY) {
2657 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2658 IPSTATS_MIB_INADDRERRORS);
2662 case IPSTATS_MIB_OUTNOROUTES:
2663 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2664 ipstats_mib_noroutes);
2667 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2672 static int ip6_pkt_discard(struct sk_buff *skb)
2674 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2677 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2679 skb->dev = skb_dst(skb)->dev;
2680 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2683 static int ip6_pkt_prohibit(struct sk_buff *skb)
2685 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2688 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2690 skb->dev = skb_dst(skb)->dev;
2691 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2695 * Allocate a dst for local (unicast / anycast) address.
2698 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2699 const struct in6_addr *addr,
2703 struct net *net = dev_net(idev->dev);
2704 struct net_device *dev = net->loopback_dev;
2705 struct rt6_info *rt;
2707 /* use L3 Master device as loopback for host routes if device
2708 * is enslaved and address is not link local or multicast
2710 if (!rt6_need_strict(addr))
2711 dev = l3mdev_master_dev_rcu(idev->dev) ? : dev;
2713 rt = ip6_dst_alloc(net, dev, DST_NOCOUNT);
2715 return ERR_PTR(-ENOMEM);
2719 rt->dst.flags |= DST_HOST;
2720 rt->dst.input = ip6_input;
2721 rt->dst.output = ip6_output;
2722 rt->rt6i_idev = idev;
2724 rt->rt6i_protocol = RTPROT_KERNEL;
2725 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2727 rt->rt6i_flags |= RTF_ANYCAST;
2729 rt->rt6i_flags |= RTF_LOCAL;
2731 rt->rt6i_gateway = *addr;
2732 rt->rt6i_dst.addr = *addr;
2733 rt->rt6i_dst.plen = 128;
2734 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2735 rt->rt6i_table = fib6_get_table(net, tb_id);
2736 rt->dst.flags |= DST_NOCACHE;
2738 atomic_set(&rt->dst.__refcnt, 1);
2743 /* remove deleted ip from prefsrc entries */
2744 struct arg_dev_net_ip {
2745 struct net_device *dev;
2747 struct in6_addr *addr;
2750 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2752 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2753 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2754 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2756 if (((void *)rt->dst.dev == dev || !dev) &&
2757 rt != net->ipv6.ip6_null_entry &&
2758 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2759 /* remove prefsrc entry */
2760 rt->rt6i_prefsrc.plen = 0;
2765 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2767 struct net *net = dev_net(ifp->idev->dev);
2768 struct arg_dev_net_ip adni = {
2769 .dev = ifp->idev->dev,
2773 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2776 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2777 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2779 /* Remove routers and update dst entries when gateway turn into host. */
2780 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2782 struct in6_addr *gateway = (struct in6_addr *)arg;
2784 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2785 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2786 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2792 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2794 fib6_clean_all(net, fib6_clean_tohost, gateway);
2797 struct arg_dev_net {
2798 struct net_device *dev;
2802 /* called with write lock held for table with rt */
2803 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2805 const struct arg_dev_net *adn = arg;
2806 const struct net_device *dev = adn->dev;
2808 if ((rt->dst.dev == dev || !dev) &&
2809 rt != adn->net->ipv6.ip6_null_entry &&
2810 (rt->rt6i_nsiblings == 0 ||
2811 !rt->rt6i_idev->cnf.ignore_routes_with_linkdown))
2817 void rt6_ifdown(struct net *net, struct net_device *dev)
2819 struct arg_dev_net adn = {
2824 fib6_clean_all(net, fib6_ifdown, &adn);
2825 icmp6_clean_all(fib6_ifdown, &adn);
2827 rt6_uncached_list_flush_dev(net, dev);
2830 struct rt6_mtu_change_arg {
2831 struct net_device *dev;
2835 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2837 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2838 struct inet6_dev *idev;
2840 /* In IPv6 pmtu discovery is not optional,
2841 so that RTAX_MTU lock cannot disable it.
2842 We still use this lock to block changes
2843 caused by addrconf/ndisc.
2846 idev = __in6_dev_get(arg->dev);
2850 /* For administrative MTU increase, there is no way to discover
2851 IPv6 PMTU increase, so PMTU increase should be updated here.
2852 Since RFC 1981 doesn't include administrative MTU increase
2853 update PMTU increase is a MUST. (i.e. jumbo frame)
2856 If new MTU is less than route PMTU, this new MTU will be the
2857 lowest MTU in the path, update the route PMTU to reflect PMTU
2858 decreases; if new MTU is greater than route PMTU, and the
2859 old MTU is the lowest MTU in the path, update the route PMTU
2860 to reflect the increase. In this case if the other nodes' MTU
2861 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2864 if (rt->dst.dev == arg->dev &&
2865 dst_metric_raw(&rt->dst, RTAX_MTU) &&
2866 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2867 if (rt->rt6i_flags & RTF_CACHE) {
2868 /* For RTF_CACHE with rt6i_pmtu == 0
2869 * (i.e. a redirected route),
2870 * the metrics of its rt->dst.from has already
2873 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2874 rt->rt6i_pmtu = arg->mtu;
2875 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2876 (dst_mtu(&rt->dst) < arg->mtu &&
2877 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2878 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2884 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2886 struct rt6_mtu_change_arg arg = {
2891 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2894 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2895 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2896 [RTA_OIF] = { .type = NLA_U32 },
2897 [RTA_IIF] = { .type = NLA_U32 },
2898 [RTA_PRIORITY] = { .type = NLA_U32 },
2899 [RTA_METRICS] = { .type = NLA_NESTED },
2900 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2901 [RTA_PREF] = { .type = NLA_U8 },
2902 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2903 [RTA_ENCAP] = { .type = NLA_NESTED },
2904 [RTA_EXPIRES] = { .type = NLA_U32 },
2905 [RTA_UID] = { .type = NLA_U32 },
2906 [RTA_MARK] = { .type = NLA_U32 },
2909 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2910 struct fib6_config *cfg,
2911 struct netlink_ext_ack *extack)
2914 struct nlattr *tb[RTA_MAX+1];
2918 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
2924 rtm = nlmsg_data(nlh);
2925 memset(cfg, 0, sizeof(*cfg));
2927 cfg->fc_table = rtm->rtm_table;
2928 cfg->fc_dst_len = rtm->rtm_dst_len;
2929 cfg->fc_src_len = rtm->rtm_src_len;
2930 cfg->fc_flags = RTF_UP;
2931 cfg->fc_protocol = rtm->rtm_protocol;
2932 cfg->fc_type = rtm->rtm_type;
2934 if (rtm->rtm_type == RTN_UNREACHABLE ||
2935 rtm->rtm_type == RTN_BLACKHOLE ||
2936 rtm->rtm_type == RTN_PROHIBIT ||
2937 rtm->rtm_type == RTN_THROW)
2938 cfg->fc_flags |= RTF_REJECT;
2940 if (rtm->rtm_type == RTN_LOCAL)
2941 cfg->fc_flags |= RTF_LOCAL;
2943 if (rtm->rtm_flags & RTM_F_CLONED)
2944 cfg->fc_flags |= RTF_CACHE;
2946 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2947 cfg->fc_nlinfo.nlh = nlh;
2948 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2950 if (tb[RTA_GATEWAY]) {
2951 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2952 cfg->fc_flags |= RTF_GATEWAY;
2956 int plen = (rtm->rtm_dst_len + 7) >> 3;
2958 if (nla_len(tb[RTA_DST]) < plen)
2961 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2965 int plen = (rtm->rtm_src_len + 7) >> 3;
2967 if (nla_len(tb[RTA_SRC]) < plen)
2970 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2973 if (tb[RTA_PREFSRC])
2974 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2977 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2979 if (tb[RTA_PRIORITY])
2980 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2982 if (tb[RTA_METRICS]) {
2983 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2984 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2988 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2990 if (tb[RTA_MULTIPATH]) {
2991 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2992 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2994 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
3001 pref = nla_get_u8(tb[RTA_PREF]);
3002 if (pref != ICMPV6_ROUTER_PREF_LOW &&
3003 pref != ICMPV6_ROUTER_PREF_HIGH)
3004 pref = ICMPV6_ROUTER_PREF_MEDIUM;
3005 cfg->fc_flags |= RTF_PREF(pref);
3009 cfg->fc_encap = tb[RTA_ENCAP];
3011 if (tb[RTA_ENCAP_TYPE]) {
3012 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
3014 err = lwtunnel_valid_encap_type(cfg->fc_encap_type);
3019 if (tb[RTA_EXPIRES]) {
3020 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
3022 if (addrconf_finite_timeout(timeout)) {
3023 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
3024 cfg->fc_flags |= RTF_EXPIRES;
3034 struct rt6_info *rt6_info;
3035 struct fib6_config r_cfg;
3036 struct mx6_config mxc;
3037 struct list_head next;
3040 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
3044 list_for_each_entry(nh, rt6_nh_list, next) {
3045 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6c nexthop %pI6c ifi %d\n",
3046 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
3047 nh->r_cfg.fc_ifindex);
3051 static int ip6_route_info_append(struct list_head *rt6_nh_list,
3052 struct rt6_info *rt, struct fib6_config *r_cfg)
3055 struct rt6_info *rtnh;
3058 list_for_each_entry(nh, rt6_nh_list, next) {
3059 /* check if rt6_info already exists */
3060 rtnh = nh->rt6_info;
3062 if (rtnh->dst.dev == rt->dst.dev &&
3063 rtnh->rt6i_idev == rt->rt6i_idev &&
3064 ipv6_addr_equal(&rtnh->rt6i_gateway,
3069 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
3073 err = ip6_convert_metrics(&nh->mxc, r_cfg);
3078 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
3079 list_add_tail(&nh->next, rt6_nh_list);
3084 static void ip6_route_mpath_notify(struct rt6_info *rt,
3085 struct rt6_info *rt_last,
3086 struct nl_info *info,
3089 /* if this is an APPEND route, then rt points to the first route
3090 * inserted and rt_last points to last route inserted. Userspace
3091 * wants a consistent dump of the route which starts at the first
3092 * nexthop. Since sibling routes are always added at the end of
3093 * the list, find the first sibling of the last route appended
3095 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->rt6i_nsiblings) {
3096 rt = list_first_entry(&rt_last->rt6i_siblings,
3102 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
3105 static int ip6_route_multipath_add(struct fib6_config *cfg,
3106 struct netlink_ext_ack *extack)
3108 struct rt6_info *rt_notif = NULL, *rt_last = NULL;
3109 struct nl_info *info = &cfg->fc_nlinfo;
3110 struct fib6_config r_cfg;
3111 struct rtnexthop *rtnh;
3112 struct rt6_info *rt;
3113 struct rt6_nh *err_nh;
3114 struct rt6_nh *nh, *nh_safe;
3120 int replace = (cfg->fc_nlinfo.nlh &&
3121 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
3122 LIST_HEAD(rt6_nh_list);
3124 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
3125 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
3126 nlflags |= NLM_F_APPEND;
3128 remaining = cfg->fc_mp_len;
3129 rtnh = (struct rtnexthop *)cfg->fc_mp;
3131 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
3132 * rt6_info structs per nexthop
3134 while (rtnh_ok(rtnh, remaining)) {
3135 memcpy(&r_cfg, cfg, sizeof(*cfg));
3136 if (rtnh->rtnh_ifindex)
3137 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3139 attrlen = rtnh_attrlen(rtnh);
3141 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3143 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3145 r_cfg.fc_gateway = nla_get_in6_addr(nla);
3146 r_cfg.fc_flags |= RTF_GATEWAY;
3148 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
3149 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
3151 r_cfg.fc_encap_type = nla_get_u16(nla);
3154 rt = ip6_route_info_create(&r_cfg, extack);
3161 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
3167 rtnh = rtnh_next(rtnh, &remaining);
3170 /* for add and replace send one notification with all nexthops.
3171 * Skip the notification in fib6_add_rt2node and send one with
3172 * the full route when done
3174 info->skip_notify = 1;
3177 list_for_each_entry(nh, &rt6_nh_list, next) {
3178 rt_last = nh->rt6_info;
3179 err = __ip6_ins_rt(nh->rt6_info, info, &nh->mxc, extack);
3180 /* save reference to first route for notification */
3181 if (!rt_notif && !err)
3182 rt_notif = nh->rt6_info;
3184 /* nh->rt6_info is used or freed at this point, reset to NULL*/
3185 nh->rt6_info = NULL;
3188 ip6_print_replace_route_err(&rt6_nh_list);
3193 /* Because each route is added like a single route we remove
3194 * these flags after the first nexthop: if there is a collision,
3195 * we have already failed to add the first nexthop:
3196 * fib6_add_rt2node() has rejected it; when replacing, old
3197 * nexthops have been replaced by first new, the rest should
3200 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
3205 /* success ... tell user about new route */
3206 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
3210 /* send notification for routes that were added so that
3211 * the delete notifications sent by ip6_route_del are
3215 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
3217 /* Delete routes that were already added */
3218 list_for_each_entry(nh, &rt6_nh_list, next) {
3221 ip6_route_del(&nh->r_cfg, extack);
3225 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3227 dst_free(&nh->rt6_info->dst);
3229 list_del(&nh->next);
3236 static int ip6_route_multipath_del(struct fib6_config *cfg,
3237 struct netlink_ext_ack *extack)
3239 struct fib6_config r_cfg;
3240 struct rtnexthop *rtnh;
3243 int err = 1, last_err = 0;
3245 remaining = cfg->fc_mp_len;
3246 rtnh = (struct rtnexthop *)cfg->fc_mp;
3248 /* Parse a Multipath Entry */
3249 while (rtnh_ok(rtnh, remaining)) {
3250 memcpy(&r_cfg, cfg, sizeof(*cfg));
3251 if (rtnh->rtnh_ifindex)
3252 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3254 attrlen = rtnh_attrlen(rtnh);
3256 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3258 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3260 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3261 r_cfg.fc_flags |= RTF_GATEWAY;
3264 err = ip6_route_del(&r_cfg, extack);
3268 rtnh = rtnh_next(rtnh, &remaining);
3274 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
3275 struct netlink_ext_ack *extack)
3277 struct fib6_config cfg;
3280 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
3285 return ip6_route_multipath_del(&cfg, extack);
3287 cfg.fc_delete_all_nh = 1;
3288 return ip6_route_del(&cfg, extack);
3292 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
3293 struct netlink_ext_ack *extack)
3295 struct fib6_config cfg;
3298 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
3303 return ip6_route_multipath_add(&cfg, extack);
3305 return ip6_route_add(&cfg, extack);
3308 static size_t rt6_nlmsg_size(struct rt6_info *rt)
3310 int nexthop_len = 0;
3312 if (rt->rt6i_nsiblings) {
3313 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
3314 + NLA_ALIGN(sizeof(struct rtnexthop))
3315 + nla_total_size(16) /* RTA_GATEWAY */
3316 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3318 nexthop_len *= rt->rt6i_nsiblings;
3321 return NLMSG_ALIGN(sizeof(struct rtmsg))
3322 + nla_total_size(16) /* RTA_SRC */
3323 + nla_total_size(16) /* RTA_DST */
3324 + nla_total_size(16) /* RTA_GATEWAY */
3325 + nla_total_size(16) /* RTA_PREFSRC */
3326 + nla_total_size(4) /* RTA_TABLE */
3327 + nla_total_size(4) /* RTA_IIF */
3328 + nla_total_size(4) /* RTA_OIF */
3329 + nla_total_size(4) /* RTA_PRIORITY */
3330 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3331 + nla_total_size(sizeof(struct rta_cacheinfo))
3332 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3333 + nla_total_size(1) /* RTA_PREF */
3334 + lwtunnel_get_encap_size(rt->dst.lwtstate)
3338 static int rt6_nexthop_info(struct sk_buff *skb, struct rt6_info *rt,
3339 unsigned int *flags, bool skip_oif)
3341 if (!netif_running(rt->dst.dev) || !netif_carrier_ok(rt->dst.dev)) {
3342 *flags |= RTNH_F_LINKDOWN;
3343 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3344 *flags |= RTNH_F_DEAD;
3347 if (rt->rt6i_flags & RTF_GATEWAY) {
3348 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3349 goto nla_put_failure;
3352 /* not needed for multipath encoding b/c it has a rtnexthop struct */
3353 if (!skip_oif && rt->dst.dev &&
3354 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3355 goto nla_put_failure;
3357 if (rt->dst.lwtstate &&
3358 lwtunnel_fill_encap(skb, rt->dst.lwtstate) < 0)
3359 goto nla_put_failure;
3367 /* add multipath next hop */
3368 static int rt6_add_nexthop(struct sk_buff *skb, struct rt6_info *rt)
3370 struct rtnexthop *rtnh;
3371 unsigned int flags = 0;
3373 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
3375 goto nla_put_failure;
3377 rtnh->rtnh_hops = 0;
3378 rtnh->rtnh_ifindex = rt->dst.dev ? rt->dst.dev->ifindex : 0;
3380 if (rt6_nexthop_info(skb, rt, &flags, true) < 0)
3381 goto nla_put_failure;
3383 rtnh->rtnh_flags = flags;
3385 /* length of rtnetlink header + attributes */
3386 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
3394 static int rt6_fill_node(struct net *net,
3395 struct sk_buff *skb, struct rt6_info *rt,
3396 struct in6_addr *dst, struct in6_addr *src,
3397 int iif, int type, u32 portid, u32 seq,
3400 u32 metrics[RTAX_MAX];
3402 struct nlmsghdr *nlh;
3406 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3410 rtm = nlmsg_data(nlh);
3411 rtm->rtm_family = AF_INET6;
3412 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3413 rtm->rtm_src_len = rt->rt6i_src.plen;
3416 table = rt->rt6i_table->tb6_id;
3418 table = RT6_TABLE_UNSPEC;
3419 rtm->rtm_table = table;
3420 if (nla_put_u32(skb, RTA_TABLE, table))
3421 goto nla_put_failure;
3422 if (rt->rt6i_flags & RTF_REJECT) {
3423 switch (rt->dst.error) {
3425 rtm->rtm_type = RTN_BLACKHOLE;
3428 rtm->rtm_type = RTN_PROHIBIT;
3431 rtm->rtm_type = RTN_THROW;
3434 rtm->rtm_type = RTN_UNREACHABLE;
3438 else if (rt->rt6i_flags & RTF_LOCAL)
3439 rtm->rtm_type = RTN_LOCAL;
3440 else if (rt->rt6i_flags & RTF_ANYCAST)
3441 rtm->rtm_type = RTN_ANYCAST;
3442 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3443 rtm->rtm_type = RTN_LOCAL;
3445 rtm->rtm_type = RTN_UNICAST;
3447 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3448 rtm->rtm_protocol = rt->rt6i_protocol;
3449 if (rt->rt6i_flags & RTF_DYNAMIC)
3450 rtm->rtm_protocol = RTPROT_REDIRECT;
3451 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3452 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3453 rtm->rtm_protocol = RTPROT_RA;
3455 rtm->rtm_protocol = RTPROT_KERNEL;
3458 if (rt->rt6i_flags & RTF_CACHE)
3459 rtm->rtm_flags |= RTM_F_CLONED;
3462 if (nla_put_in6_addr(skb, RTA_DST, dst))
3463 goto nla_put_failure;
3464 rtm->rtm_dst_len = 128;
3465 } else if (rtm->rtm_dst_len)
3466 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3467 goto nla_put_failure;
3468 #ifdef CONFIG_IPV6_SUBTREES
3470 if (nla_put_in6_addr(skb, RTA_SRC, src))
3471 goto nla_put_failure;
3472 rtm->rtm_src_len = 128;
3473 } else if (rtm->rtm_src_len &&
3474 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3475 goto nla_put_failure;
3478 #ifdef CONFIG_IPV6_MROUTE
3479 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3480 int err = ip6mr_get_route(net, skb, rtm, portid);
3485 goto nla_put_failure;
3488 if (nla_put_u32(skb, RTA_IIF, iif))
3489 goto nla_put_failure;
3491 struct in6_addr saddr_buf;
3492 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3493 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3494 goto nla_put_failure;
3497 if (rt->rt6i_prefsrc.plen) {
3498 struct in6_addr saddr_buf;
3499 saddr_buf = rt->rt6i_prefsrc.addr;
3500 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3501 goto nla_put_failure;
3504 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3506 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3507 if (rtnetlink_put_metrics(skb, metrics) < 0)
3508 goto nla_put_failure;
3510 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3511 goto nla_put_failure;
3513 /* For multipath routes, walk the siblings list and add
3514 * each as a nexthop within RTA_MULTIPATH.
3516 if (rt->rt6i_nsiblings) {
3517 struct rt6_info *sibling, *next_sibling;
3520 mp = nla_nest_start(skb, RTA_MULTIPATH);
3522 goto nla_put_failure;
3524 if (rt6_add_nexthop(skb, rt) < 0)
3525 goto nla_put_failure;
3527 list_for_each_entry_safe(sibling, next_sibling,
3528 &rt->rt6i_siblings, rt6i_siblings) {
3529 if (rt6_add_nexthop(skb, sibling) < 0)
3530 goto nla_put_failure;
3533 nla_nest_end(skb, mp);
3535 if (rt6_nexthop_info(skb, rt, &rtm->rtm_flags, false) < 0)
3536 goto nla_put_failure;
3539 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3541 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3542 goto nla_put_failure;
3544 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3545 goto nla_put_failure;
3548 nlmsg_end(skb, nlh);
3552 nlmsg_cancel(skb, nlh);
3556 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3558 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3559 struct net *net = arg->net;
3561 if (rt == net->ipv6.ip6_null_entry)
3564 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3565 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3567 /* user wants prefix routes only */
3568 if (rtm->rtm_flags & RTM_F_PREFIX &&
3569 !(rt->rt6i_flags & RTF_PREFIX_RT)) {
3570 /* success since this is not a prefix route */
3575 return rt6_fill_node(net,
3576 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3577 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3581 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3582 struct netlink_ext_ack *extack)
3584 struct net *net = sock_net(in_skb->sk);
3585 struct nlattr *tb[RTA_MAX+1];
3586 struct rt6_info *rt;
3587 struct sk_buff *skb;
3590 int err, iif = 0, oif = 0;
3592 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
3598 memset(&fl6, 0, sizeof(fl6));
3599 rtm = nlmsg_data(nlh);
3600 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
3603 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3606 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3610 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3613 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3617 iif = nla_get_u32(tb[RTA_IIF]);
3620 oif = nla_get_u32(tb[RTA_OIF]);
3623 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3626 fl6.flowi6_uid = make_kuid(current_user_ns(),
3627 nla_get_u32(tb[RTA_UID]));
3629 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
3632 struct net_device *dev;
3635 dev = __dev_get_by_index(net, iif);
3641 fl6.flowi6_iif = iif;
3643 if (!ipv6_addr_any(&fl6.saddr))
3644 flags |= RT6_LOOKUP_F_HAS_SADDR;
3646 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3649 fl6.flowi6_oif = oif;
3651 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3654 if (rt == net->ipv6.ip6_null_entry) {
3655 err = rt->dst.error;
3660 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3667 skb_dst_set(skb, &rt->dst);
3669 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3670 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3677 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3682 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3683 unsigned int nlm_flags)
3685 struct sk_buff *skb;
3686 struct net *net = info->nl_net;
3691 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3693 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3697 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3698 event, info->portid, seq, nlm_flags);
3700 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3701 WARN_ON(err == -EMSGSIZE);
3705 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3706 info->nlh, gfp_any());
3710 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3713 static int ip6_route_dev_notify(struct notifier_block *this,
3714 unsigned long event, void *ptr)
3716 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3717 struct net *net = dev_net(dev);
3719 if (!(dev->flags & IFF_LOOPBACK))
3722 if (event == NETDEV_REGISTER) {
3723 net->ipv6.ip6_null_entry->dst.dev = dev;
3724 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3725 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3726 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3727 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3728 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3729 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3731 } else if (event == NETDEV_UNREGISTER) {
3732 in6_dev_put(net->ipv6.ip6_null_entry->rt6i_idev);
3733 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3734 in6_dev_put(net->ipv6.ip6_prohibit_entry->rt6i_idev);
3735 in6_dev_put(net->ipv6.ip6_blk_hole_entry->rt6i_idev);
3746 #ifdef CONFIG_PROC_FS
3748 static const struct file_operations ipv6_route_proc_fops = {
3749 .owner = THIS_MODULE,
3750 .open = ipv6_route_open,
3752 .llseek = seq_lseek,
3753 .release = seq_release_net,
3756 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3758 struct net *net = (struct net *)seq->private;
3759 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3760 net->ipv6.rt6_stats->fib_nodes,
3761 net->ipv6.rt6_stats->fib_route_nodes,
3762 net->ipv6.rt6_stats->fib_rt_alloc,
3763 net->ipv6.rt6_stats->fib_rt_entries,
3764 net->ipv6.rt6_stats->fib_rt_cache,
3765 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3766 net->ipv6.rt6_stats->fib_discarded_routes);
3771 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3773 return single_open_net(inode, file, rt6_stats_seq_show);
3776 static const struct file_operations rt6_stats_seq_fops = {
3777 .owner = THIS_MODULE,
3778 .open = rt6_stats_seq_open,
3780 .llseek = seq_lseek,
3781 .release = single_release_net,
3783 #endif /* CONFIG_PROC_FS */
3785 #ifdef CONFIG_SYSCTL
3788 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3789 void __user *buffer, size_t *lenp, loff_t *ppos)
3796 net = (struct net *)ctl->extra1;
3797 delay = net->ipv6.sysctl.flush_delay;
3798 proc_dointvec(ctl, write, buffer, lenp, ppos);
3799 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3803 struct ctl_table ipv6_route_table_template[] = {
3805 .procname = "flush",
3806 .data = &init_net.ipv6.sysctl.flush_delay,
3807 .maxlen = sizeof(int),
3809 .proc_handler = ipv6_sysctl_rtcache_flush
3812 .procname = "gc_thresh",
3813 .data = &ip6_dst_ops_template.gc_thresh,
3814 .maxlen = sizeof(int),
3816 .proc_handler = proc_dointvec,
3819 .procname = "max_size",
3820 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3821 .maxlen = sizeof(int),
3823 .proc_handler = proc_dointvec,
3826 .procname = "gc_min_interval",
3827 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3828 .maxlen = sizeof(int),
3830 .proc_handler = proc_dointvec_jiffies,
3833 .procname = "gc_timeout",
3834 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3835 .maxlen = sizeof(int),
3837 .proc_handler = proc_dointvec_jiffies,
3840 .procname = "gc_interval",
3841 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3842 .maxlen = sizeof(int),
3844 .proc_handler = proc_dointvec_jiffies,
3847 .procname = "gc_elasticity",
3848 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3849 .maxlen = sizeof(int),
3851 .proc_handler = proc_dointvec,
3854 .procname = "mtu_expires",
3855 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3856 .maxlen = sizeof(int),
3858 .proc_handler = proc_dointvec_jiffies,
3861 .procname = "min_adv_mss",
3862 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3863 .maxlen = sizeof(int),
3865 .proc_handler = proc_dointvec,
3868 .procname = "gc_min_interval_ms",
3869 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3870 .maxlen = sizeof(int),
3872 .proc_handler = proc_dointvec_ms_jiffies,
3877 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3879 struct ctl_table *table;
3881 table = kmemdup(ipv6_route_table_template,
3882 sizeof(ipv6_route_table_template),
3886 table[0].data = &net->ipv6.sysctl.flush_delay;
3887 table[0].extra1 = net;
3888 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3889 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3890 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3891 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3892 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3893 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3894 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3895 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3896 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3898 /* Don't export sysctls to unprivileged users */
3899 if (net->user_ns != &init_user_ns)
3900 table[0].procname = NULL;
3907 static int __net_init ip6_route_net_init(struct net *net)
3911 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3912 sizeof(net->ipv6.ip6_dst_ops));
3914 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3915 goto out_ip6_dst_ops;
3917 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3918 sizeof(*net->ipv6.ip6_null_entry),
3920 if (!net->ipv6.ip6_null_entry)
3921 goto out_ip6_dst_entries;
3922 net->ipv6.ip6_null_entry->dst.path =
3923 (struct dst_entry *)net->ipv6.ip6_null_entry;
3924 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3925 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3926 ip6_template_metrics, true);
3928 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3929 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3930 sizeof(*net->ipv6.ip6_prohibit_entry),
3932 if (!net->ipv6.ip6_prohibit_entry)
3933 goto out_ip6_null_entry;
3934 net->ipv6.ip6_prohibit_entry->dst.path =
3935 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3936 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3937 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3938 ip6_template_metrics, true);
3940 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3941 sizeof(*net->ipv6.ip6_blk_hole_entry),
3943 if (!net->ipv6.ip6_blk_hole_entry)
3944 goto out_ip6_prohibit_entry;
3945 net->ipv6.ip6_blk_hole_entry->dst.path =
3946 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3947 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3948 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3949 ip6_template_metrics, true);
3952 net->ipv6.sysctl.flush_delay = 0;
3953 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3954 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3955 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3956 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3957 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3958 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3959 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3961 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3967 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3968 out_ip6_prohibit_entry:
3969 kfree(net->ipv6.ip6_prohibit_entry);
3971 kfree(net->ipv6.ip6_null_entry);
3973 out_ip6_dst_entries:
3974 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3979 static void __net_exit ip6_route_net_exit(struct net *net)
3981 kfree(net->ipv6.ip6_null_entry);
3982 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3983 kfree(net->ipv6.ip6_prohibit_entry);
3984 kfree(net->ipv6.ip6_blk_hole_entry);
3986 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3989 static int __net_init ip6_route_net_init_late(struct net *net)
3991 #ifdef CONFIG_PROC_FS
3992 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3993 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3998 static void __net_exit ip6_route_net_exit_late(struct net *net)
4000 #ifdef CONFIG_PROC_FS
4001 remove_proc_entry("ipv6_route", net->proc_net);
4002 remove_proc_entry("rt6_stats", net->proc_net);
4006 static struct pernet_operations ip6_route_net_ops = {
4007 .init = ip6_route_net_init,
4008 .exit = ip6_route_net_exit,
4011 static int __net_init ipv6_inetpeer_init(struct net *net)
4013 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
4017 inet_peer_base_init(bp);
4018 net->ipv6.peers = bp;
4022 static void __net_exit ipv6_inetpeer_exit(struct net *net)
4024 struct inet_peer_base *bp = net->ipv6.peers;
4026 net->ipv6.peers = NULL;
4027 inetpeer_invalidate_tree(bp);
4031 static struct pernet_operations ipv6_inetpeer_ops = {
4032 .init = ipv6_inetpeer_init,
4033 .exit = ipv6_inetpeer_exit,
4036 static struct pernet_operations ip6_route_net_late_ops = {
4037 .init = ip6_route_net_init_late,
4038 .exit = ip6_route_net_exit_late,
4041 static struct notifier_block ip6_route_dev_notifier = {
4042 .notifier_call = ip6_route_dev_notify,
4043 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
4046 void __init ip6_route_init_special_entries(void)
4048 /* Registering of the loopback is done before this portion of code,
4049 * the loopback reference in rt6_info will not be taken, do it
4050 * manually for init_net */
4051 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
4052 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4053 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4054 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
4055 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4056 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
4057 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4061 int __init ip6_route_init(void)
4067 ip6_dst_ops_template.kmem_cachep =
4068 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
4069 SLAB_HWCACHE_ALIGN, NULL);
4070 if (!ip6_dst_ops_template.kmem_cachep)
4073 ret = dst_entries_init(&ip6_dst_blackhole_ops);
4075 goto out_kmem_cache;
4077 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
4079 goto out_dst_entries;
4081 ret = register_pernet_subsys(&ip6_route_net_ops);
4083 goto out_register_inetpeer;
4085 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
4089 goto out_register_subsys;
4095 ret = fib6_rules_init();
4099 ret = register_pernet_subsys(&ip6_route_net_late_ops);
4101 goto fib6_rules_init;
4104 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
4105 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
4106 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
4107 goto out_register_late_subsys;
4109 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
4111 goto out_register_late_subsys;
4113 for_each_possible_cpu(cpu) {
4114 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
4116 INIT_LIST_HEAD(&ul->head);
4117 spin_lock_init(&ul->lock);
4123 out_register_late_subsys:
4124 unregister_pernet_subsys(&ip6_route_net_late_ops);
4126 fib6_rules_cleanup();
4131 out_register_subsys:
4132 unregister_pernet_subsys(&ip6_route_net_ops);
4133 out_register_inetpeer:
4134 unregister_pernet_subsys(&ipv6_inetpeer_ops);
4136 dst_entries_destroy(&ip6_dst_blackhole_ops);
4138 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
4142 void ip6_route_cleanup(void)
4144 unregister_netdevice_notifier(&ip6_route_dev_notifier);
4145 unregister_pernet_subsys(&ip6_route_net_late_ops);
4146 fib6_rules_cleanup();
4149 unregister_pernet_subsys(&ipv6_inetpeer_ops);
4150 unregister_pernet_subsys(&ip6_route_net_ops);
4151 dst_entries_destroy(&ip6_dst_blackhole_ops);
4152 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);