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>
66 #include <asm/uaccess.h>
69 #include <linux/sysctl.h>
73 RT6_NUD_FAIL_HARD = -3,
74 RT6_NUD_FAIL_PROBE = -2,
75 RT6_NUD_FAIL_DO_RR = -1,
79 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
80 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
81 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
82 static unsigned int ip6_mtu(const struct dst_entry *dst);
83 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
84 static void ip6_dst_destroy(struct dst_entry *);
85 static void ip6_dst_ifdown(struct dst_entry *,
86 struct net_device *dev, int how);
87 static int ip6_dst_gc(struct dst_ops *ops);
89 static int ip6_pkt_discard(struct sk_buff *skb);
90 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
91 static int ip6_pkt_prohibit(struct sk_buff *skb);
92 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
93 static void ip6_link_failure(struct sk_buff *skb);
94 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
95 struct sk_buff *skb, u32 mtu);
96 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
98 static void rt6_dst_from_metrics_check(struct rt6_info *rt);
99 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
101 #ifdef CONFIG_IPV6_ROUTE_INFO
102 static struct rt6_info *rt6_add_route_info(struct net_device *dev,
103 const struct in6_addr *prefix, int prefixlen,
104 const struct in6_addr *gwaddr, unsigned int pref);
105 static struct rt6_info *rt6_get_route_info(struct net_device *dev,
106 const struct in6_addr *prefix, int prefixlen,
107 const struct in6_addr *gwaddr);
110 struct uncached_list {
112 struct list_head head;
115 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
117 static void rt6_uncached_list_add(struct rt6_info *rt)
119 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
121 rt->dst.flags |= DST_NOCACHE;
122 rt->rt6i_uncached_list = ul;
124 spin_lock_bh(&ul->lock);
125 list_add_tail(&rt->rt6i_uncached, &ul->head);
126 spin_unlock_bh(&ul->lock);
129 static void rt6_uncached_list_del(struct rt6_info *rt)
131 if (!list_empty(&rt->rt6i_uncached)) {
132 struct uncached_list *ul = rt->rt6i_uncached_list;
134 spin_lock_bh(&ul->lock);
135 list_del(&rt->rt6i_uncached);
136 spin_unlock_bh(&ul->lock);
140 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
142 struct net_device *loopback_dev = net->loopback_dev;
145 if (dev == loopback_dev)
148 for_each_possible_cpu(cpu) {
149 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
152 spin_lock_bh(&ul->lock);
153 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
154 struct inet6_dev *rt_idev = rt->rt6i_idev;
155 struct net_device *rt_dev = rt->dst.dev;
157 if (rt_idev->dev == dev) {
158 rt->rt6i_idev = in6_dev_get(loopback_dev);
159 in6_dev_put(rt_idev);
163 rt->dst.dev = loopback_dev;
164 dev_hold(rt->dst.dev);
168 spin_unlock_bh(&ul->lock);
172 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
174 return dst_metrics_write_ptr(rt->dst.from);
177 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
179 struct rt6_info *rt = (struct rt6_info *)dst;
181 if (rt->rt6i_flags & RTF_PCPU)
182 return rt6_pcpu_cow_metrics(rt);
183 else if (rt->rt6i_flags & RTF_CACHE)
186 return dst_cow_metrics_generic(dst, old);
189 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
193 struct in6_addr *p = &rt->rt6i_gateway;
195 if (!ipv6_addr_any(p))
196 return (const void *) p;
198 return &ipv6_hdr(skb)->daddr;
202 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
206 struct rt6_info *rt = (struct rt6_info *) dst;
209 daddr = choose_neigh_daddr(rt, skb, daddr);
210 n = __ipv6_neigh_lookup(dst->dev, daddr);
213 return neigh_create(&nd_tbl, daddr, dst->dev);
216 static struct dst_ops ip6_dst_ops_template = {
220 .check = ip6_dst_check,
221 .default_advmss = ip6_default_advmss,
223 .cow_metrics = ipv6_cow_metrics,
224 .destroy = ip6_dst_destroy,
225 .ifdown = ip6_dst_ifdown,
226 .negative_advice = ip6_negative_advice,
227 .link_failure = ip6_link_failure,
228 .update_pmtu = ip6_rt_update_pmtu,
229 .redirect = rt6_do_redirect,
230 .local_out = __ip6_local_out,
231 .neigh_lookup = ip6_neigh_lookup,
234 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
236 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
238 return mtu ? : dst->dev->mtu;
241 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
242 struct sk_buff *skb, u32 mtu)
246 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
251 static struct dst_ops ip6_dst_blackhole_ops = {
253 .destroy = ip6_dst_destroy,
254 .check = ip6_dst_check,
255 .mtu = ip6_blackhole_mtu,
256 .default_advmss = ip6_default_advmss,
257 .update_pmtu = ip6_rt_blackhole_update_pmtu,
258 .redirect = ip6_rt_blackhole_redirect,
259 .cow_metrics = dst_cow_metrics_generic,
260 .neigh_lookup = ip6_neigh_lookup,
263 static const u32 ip6_template_metrics[RTAX_MAX] = {
264 [RTAX_HOPLIMIT - 1] = 0,
267 static const struct rt6_info ip6_null_entry_template = {
269 .__refcnt = ATOMIC_INIT(1),
271 .obsolete = DST_OBSOLETE_FORCE_CHK,
272 .error = -ENETUNREACH,
273 .input = ip6_pkt_discard,
274 .output = ip6_pkt_discard_out,
276 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
277 .rt6i_protocol = RTPROT_KERNEL,
278 .rt6i_metric = ~(u32) 0,
279 .rt6i_ref = ATOMIC_INIT(1),
282 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
284 static const struct rt6_info ip6_prohibit_entry_template = {
286 .__refcnt = ATOMIC_INIT(1),
288 .obsolete = DST_OBSOLETE_FORCE_CHK,
290 .input = ip6_pkt_prohibit,
291 .output = ip6_pkt_prohibit_out,
293 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
294 .rt6i_protocol = RTPROT_KERNEL,
295 .rt6i_metric = ~(u32) 0,
296 .rt6i_ref = ATOMIC_INIT(1),
299 static const struct rt6_info ip6_blk_hole_entry_template = {
301 .__refcnt = ATOMIC_INIT(1),
303 .obsolete = DST_OBSOLETE_FORCE_CHK,
305 .input = dst_discard,
306 .output = dst_discard_out,
308 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
309 .rt6i_protocol = RTPROT_KERNEL,
310 .rt6i_metric = ~(u32) 0,
311 .rt6i_ref = ATOMIC_INIT(1),
316 static void rt6_info_init(struct rt6_info *rt)
318 struct dst_entry *dst = &rt->dst;
320 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
321 INIT_LIST_HEAD(&rt->rt6i_siblings);
322 INIT_LIST_HEAD(&rt->rt6i_uncached);
325 /* allocate dst with ip6_dst_ops */
326 static struct rt6_info *__ip6_dst_alloc(struct net *net,
327 struct net_device *dev,
330 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
331 0, DST_OBSOLETE_FORCE_CHK, flags);
339 static struct rt6_info *ip6_dst_alloc(struct net *net,
340 struct net_device *dev,
343 struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
346 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
350 for_each_possible_cpu(cpu) {
353 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
354 /* no one shares rt */
358 dst_destroy((struct dst_entry *)rt);
366 static void ip6_dst_destroy(struct dst_entry *dst)
368 struct rt6_info *rt = (struct rt6_info *)dst;
369 struct dst_entry *from = dst->from;
370 struct inet6_dev *idev;
372 dst_destroy_metrics_generic(dst);
373 free_percpu(rt->rt6i_pcpu);
374 rt6_uncached_list_del(rt);
376 idev = rt->rt6i_idev;
378 rt->rt6i_idev = NULL;
386 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
389 struct rt6_info *rt = (struct rt6_info *)dst;
390 struct inet6_dev *idev = rt->rt6i_idev;
391 struct net_device *loopback_dev =
392 dev_net(dev)->loopback_dev;
394 if (dev != loopback_dev) {
395 if (idev && idev->dev == dev) {
396 struct inet6_dev *loopback_idev =
397 in6_dev_get(loopback_dev);
399 rt->rt6i_idev = loopback_idev;
406 static bool __rt6_check_expired(const struct rt6_info *rt)
408 if (rt->rt6i_flags & RTF_EXPIRES)
409 return time_after(jiffies, rt->dst.expires);
414 static bool rt6_check_expired(const struct rt6_info *rt)
416 if (rt->rt6i_flags & RTF_EXPIRES) {
417 if (time_after(jiffies, rt->dst.expires))
419 } else if (rt->dst.from) {
420 return rt6_check_expired((struct rt6_info *) rt->dst.from);
425 /* Multipath route selection:
426 * Hash based function using packet header and flowlabel.
427 * Adapted from fib_info_hashfn()
429 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
430 const struct flowi6 *fl6)
432 return get_hash_from_flowi6(fl6) % candidate_count;
435 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
436 struct flowi6 *fl6, int oif,
439 struct rt6_info *sibling, *next_sibling;
442 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
443 /* Don't change the route, if route_choosen == 0
444 * (siblings does not include ourself)
447 list_for_each_entry_safe(sibling, next_sibling,
448 &match->rt6i_siblings, rt6i_siblings) {
450 if (route_choosen == 0) {
451 if (rt6_score_route(sibling, oif, strict) < 0)
461 * Route lookup. Any table->tb6_lock is implied.
464 static inline struct rt6_info *rt6_device_match(struct net *net,
466 const struct in6_addr *saddr,
470 struct rt6_info *local = NULL;
471 struct rt6_info *sprt;
473 if (!oif && ipv6_addr_any(saddr))
476 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
477 struct net_device *dev = sprt->dst.dev;
480 if (dev->ifindex == oif)
482 if (dev->flags & IFF_LOOPBACK) {
483 if (!sprt->rt6i_idev ||
484 sprt->rt6i_idev->dev->ifindex != oif) {
485 if (flags & RT6_LOOKUP_F_IFACE)
488 local->rt6i_idev->dev->ifindex == oif)
494 if (ipv6_chk_addr(net, saddr, dev,
495 flags & RT6_LOOKUP_F_IFACE))
504 if (flags & RT6_LOOKUP_F_IFACE)
505 return net->ipv6.ip6_null_entry;
511 #ifdef CONFIG_IPV6_ROUTER_PREF
512 struct __rt6_probe_work {
513 struct work_struct work;
514 struct in6_addr target;
515 struct net_device *dev;
518 static void rt6_probe_deferred(struct work_struct *w)
520 struct in6_addr mcaddr;
521 struct __rt6_probe_work *work =
522 container_of(w, struct __rt6_probe_work, work);
524 addrconf_addr_solict_mult(&work->target, &mcaddr);
525 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL);
530 static void rt6_probe(struct rt6_info *rt)
532 struct __rt6_probe_work *work;
533 struct neighbour *neigh;
535 * Okay, this does not seem to be appropriate
536 * for now, however, we need to check if it
537 * is really so; aka Router Reachability Probing.
539 * Router Reachability Probe MUST be rate-limited
540 * to no more than one per minute.
542 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
545 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
547 if (neigh->nud_state & NUD_VALID)
551 write_lock(&neigh->lock);
552 if (!(neigh->nud_state & NUD_VALID) &&
555 rt->rt6i_idev->cnf.rtr_probe_interval)) {
556 work = kmalloc(sizeof(*work), GFP_ATOMIC);
558 __neigh_set_probe_once(neigh);
560 write_unlock(&neigh->lock);
562 work = kmalloc(sizeof(*work), GFP_ATOMIC);
566 INIT_WORK(&work->work, rt6_probe_deferred);
567 work->target = rt->rt6i_gateway;
568 dev_hold(rt->dst.dev);
569 work->dev = rt->dst.dev;
570 schedule_work(&work->work);
574 rcu_read_unlock_bh();
577 static inline void rt6_probe(struct rt6_info *rt)
583 * Default Router Selection (RFC 2461 6.3.6)
585 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
587 struct net_device *dev = rt->dst.dev;
588 if (!oif || dev->ifindex == oif)
590 if ((dev->flags & IFF_LOOPBACK) &&
591 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
596 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
598 struct neighbour *neigh;
599 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
601 if (rt->rt6i_flags & RTF_NONEXTHOP ||
602 !(rt->rt6i_flags & RTF_GATEWAY))
603 return RT6_NUD_SUCCEED;
606 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
608 read_lock(&neigh->lock);
609 if (neigh->nud_state & NUD_VALID)
610 ret = RT6_NUD_SUCCEED;
611 #ifdef CONFIG_IPV6_ROUTER_PREF
612 else if (!(neigh->nud_state & NUD_FAILED))
613 ret = RT6_NUD_SUCCEED;
615 ret = RT6_NUD_FAIL_PROBE;
617 read_unlock(&neigh->lock);
619 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
620 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
622 rcu_read_unlock_bh();
627 static int rt6_score_route(struct rt6_info *rt, int oif,
632 m = rt6_check_dev(rt, oif);
633 if (!m && (strict & RT6_LOOKUP_F_IFACE))
634 return RT6_NUD_FAIL_HARD;
635 #ifdef CONFIG_IPV6_ROUTER_PREF
636 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
638 if (strict & RT6_LOOKUP_F_REACHABLE) {
639 int n = rt6_check_neigh(rt);
646 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
647 int *mpri, struct rt6_info *match,
651 bool match_do_rr = false;
652 struct inet6_dev *idev = rt->rt6i_idev;
653 struct net_device *dev = rt->dst.dev;
655 if (dev && !netif_carrier_ok(dev) &&
656 idev->cnf.ignore_routes_with_linkdown)
659 if (rt6_check_expired(rt))
662 m = rt6_score_route(rt, oif, strict);
663 if (m == RT6_NUD_FAIL_DO_RR) {
665 m = 0; /* lowest valid score */
666 } else if (m == RT6_NUD_FAIL_HARD) {
670 if (strict & RT6_LOOKUP_F_REACHABLE)
673 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
675 *do_rr = match_do_rr;
683 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
684 struct rt6_info *rr_head,
685 u32 metric, int oif, int strict,
688 struct rt6_info *rt, *match, *cont;
693 for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
694 if (rt->rt6i_metric != metric) {
699 match = find_match(rt, oif, strict, &mpri, match, do_rr);
702 for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
703 if (rt->rt6i_metric != metric) {
708 match = find_match(rt, oif, strict, &mpri, match, do_rr);
714 for (rt = cont; rt; rt = rt->dst.rt6_next)
715 match = find_match(rt, oif, strict, &mpri, match, do_rr);
720 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
722 struct rt6_info *match, *rt0;
728 fn->rr_ptr = rt0 = fn->leaf;
730 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
734 struct rt6_info *next = rt0->dst.rt6_next;
736 /* no entries matched; do round-robin */
737 if (!next || next->rt6i_metric != rt0->rt6i_metric)
744 net = dev_net(rt0->dst.dev);
745 return match ? match : net->ipv6.ip6_null_entry;
748 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
750 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
753 #ifdef CONFIG_IPV6_ROUTE_INFO
754 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
755 const struct in6_addr *gwaddr)
757 struct route_info *rinfo = (struct route_info *) opt;
758 struct in6_addr prefix_buf, *prefix;
760 unsigned long lifetime;
763 if (len < sizeof(struct route_info)) {
767 /* Sanity check for prefix_len and length */
768 if (rinfo->length > 3) {
770 } else if (rinfo->prefix_len > 128) {
772 } else if (rinfo->prefix_len > 64) {
773 if (rinfo->length < 2) {
776 } else if (rinfo->prefix_len > 0) {
777 if (rinfo->length < 1) {
782 pref = rinfo->route_pref;
783 if (pref == ICMPV6_ROUTER_PREF_INVALID)
786 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
788 if (rinfo->length == 3)
789 prefix = (struct in6_addr *)rinfo->prefix;
791 /* this function is safe */
792 ipv6_addr_prefix(&prefix_buf,
793 (struct in6_addr *)rinfo->prefix,
795 prefix = &prefix_buf;
798 if (rinfo->prefix_len == 0)
799 rt = rt6_get_dflt_router(gwaddr, dev);
801 rt = rt6_get_route_info(dev, prefix, rinfo->prefix_len, gwaddr);
803 if (rt && !lifetime) {
809 rt = rt6_add_route_info(dev, prefix, rinfo->prefix_len, gwaddr, pref);
811 rt->rt6i_flags = RTF_ROUTEINFO |
812 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
815 if (!addrconf_finite_timeout(lifetime))
816 rt6_clean_expires(rt);
818 rt6_set_expires(rt, jiffies + HZ * lifetime);
826 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
827 struct in6_addr *saddr)
829 struct fib6_node *pn;
831 if (fn->fn_flags & RTN_TL_ROOT)
834 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
835 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
838 if (fn->fn_flags & RTN_RTINFO)
843 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
844 struct fib6_table *table,
845 struct flowi6 *fl6, int flags)
847 struct fib6_node *fn;
850 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
851 flags &= ~RT6_LOOKUP_F_IFACE;
853 read_lock_bh(&table->tb6_lock);
854 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
857 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
858 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
859 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
860 if (rt == net->ipv6.ip6_null_entry) {
861 fn = fib6_backtrack(fn, &fl6->saddr);
865 dst_use(&rt->dst, jiffies);
866 read_unlock_bh(&table->tb6_lock);
871 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
874 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
876 EXPORT_SYMBOL_GPL(ip6_route_lookup);
878 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
879 const struct in6_addr *saddr, int oif, int strict)
881 struct flowi6 fl6 = {
885 struct dst_entry *dst;
886 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
889 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
890 flags |= RT6_LOOKUP_F_HAS_SADDR;
893 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
895 return (struct rt6_info *) dst;
901 EXPORT_SYMBOL(rt6_lookup);
903 /* ip6_ins_rt is called with FREE table->tb6_lock.
904 It takes new route entry, the addition fails by any reason the
905 route is freed. In any case, if caller does not hold it, it may
909 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
910 struct mx6_config *mxc)
913 struct fib6_table *table;
915 table = rt->rt6i_table;
916 write_lock_bh(&table->tb6_lock);
917 err = fib6_add(&table->tb6_root, rt, info, mxc);
918 write_unlock_bh(&table->tb6_lock);
923 int ip6_ins_rt(struct rt6_info *rt)
925 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
926 struct mx6_config mxc = { .mx = NULL, };
928 return __ip6_ins_rt(rt, &info, &mxc);
931 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
932 const struct in6_addr *daddr,
933 const struct in6_addr *saddr)
941 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
942 ort = (struct rt6_info *)ort->dst.from;
944 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
949 ip6_rt_copy_init(rt, ort);
950 rt->rt6i_flags |= RTF_CACHE;
952 rt->dst.flags |= DST_HOST;
953 rt->rt6i_dst.addr = *daddr;
954 rt->rt6i_dst.plen = 128;
956 if (!rt6_is_gw_or_nonexthop(ort)) {
957 if (ort->rt6i_dst.plen != 128 &&
958 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
959 rt->rt6i_flags |= RTF_ANYCAST;
960 #ifdef CONFIG_IPV6_SUBTREES
961 if (rt->rt6i_src.plen && saddr) {
962 rt->rt6i_src.addr = *saddr;
963 rt->rt6i_src.plen = 128;
971 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
973 struct rt6_info *pcpu_rt;
975 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
976 rt->dst.dev, rt->dst.flags);
980 ip6_rt_copy_init(pcpu_rt, rt);
981 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
982 pcpu_rt->rt6i_flags |= RTF_PCPU;
986 /* It should be called with read_lock_bh(&tb6_lock) acquired */
987 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
989 struct rt6_info *pcpu_rt, **p;
991 p = this_cpu_ptr(rt->rt6i_pcpu);
995 dst_hold(&pcpu_rt->dst);
996 rt6_dst_from_metrics_check(pcpu_rt);
1001 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1003 struct fib6_table *table = rt->rt6i_table;
1004 struct rt6_info *pcpu_rt, *prev, **p;
1006 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1008 struct net *net = dev_net(rt->dst.dev);
1010 dst_hold(&net->ipv6.ip6_null_entry->dst);
1011 return net->ipv6.ip6_null_entry;
1014 read_lock_bh(&table->tb6_lock);
1015 if (rt->rt6i_pcpu) {
1016 p = this_cpu_ptr(rt->rt6i_pcpu);
1017 prev = cmpxchg(p, NULL, pcpu_rt);
1019 /* If someone did it before us, return prev instead */
1020 dst_destroy(&pcpu_rt->dst);
1024 /* rt has been removed from the fib6 tree
1025 * before we have a chance to acquire the read_lock.
1026 * In this case, don't brother to create a pcpu rt
1027 * since rt is going away anyway. The next
1028 * dst_check() will trigger a re-lookup.
1030 dst_destroy(&pcpu_rt->dst);
1033 dst_hold(&pcpu_rt->dst);
1034 rt6_dst_from_metrics_check(pcpu_rt);
1035 read_unlock_bh(&table->tb6_lock);
1039 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1040 struct flowi6 *fl6, int flags)
1042 struct fib6_node *fn, *saved_fn;
1043 struct rt6_info *rt;
1046 strict |= flags & RT6_LOOKUP_F_IFACE;
1047 if (net->ipv6.devconf_all->forwarding == 0)
1048 strict |= RT6_LOOKUP_F_REACHABLE;
1050 read_lock_bh(&table->tb6_lock);
1052 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1055 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1059 rt = rt6_select(fn, oif, strict);
1060 if (rt->rt6i_nsiblings)
1061 rt = rt6_multipath_select(rt, fl6, oif, strict);
1062 if (rt == net->ipv6.ip6_null_entry) {
1063 fn = fib6_backtrack(fn, &fl6->saddr);
1065 goto redo_rt6_select;
1066 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1067 /* also consider unreachable route */
1068 strict &= ~RT6_LOOKUP_F_REACHABLE;
1070 goto redo_rt6_select;
1075 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1076 dst_use(&rt->dst, jiffies);
1077 read_unlock_bh(&table->tb6_lock);
1079 rt6_dst_from_metrics_check(rt);
1081 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1082 !(rt->rt6i_flags & RTF_GATEWAY))) {
1083 /* Create a RTF_CACHE clone which will not be
1084 * owned by the fib6 tree. It is for the special case where
1085 * the daddr in the skb during the neighbor look-up is different
1086 * from the fl6->daddr used to look-up route here.
1089 struct rt6_info *uncached_rt;
1091 dst_use(&rt->dst, jiffies);
1092 read_unlock_bh(&table->tb6_lock);
1094 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1095 dst_release(&rt->dst);
1098 rt6_uncached_list_add(uncached_rt);
1100 uncached_rt = net->ipv6.ip6_null_entry;
1102 dst_hold(&uncached_rt->dst);
1106 /* Get a percpu copy */
1108 struct rt6_info *pcpu_rt;
1110 rt->dst.lastuse = jiffies;
1112 pcpu_rt = rt6_get_pcpu_route(rt);
1115 read_unlock_bh(&table->tb6_lock);
1117 /* We have to do the read_unlock first
1118 * because rt6_make_pcpu_route() may trigger
1119 * ip6_dst_gc() which will take the write_lock.
1122 read_unlock_bh(&table->tb6_lock);
1123 pcpu_rt = rt6_make_pcpu_route(rt);
1124 dst_release(&rt->dst);
1132 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1133 struct flowi6 *fl6, int flags)
1135 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1138 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1139 struct net_device *dev,
1140 struct flowi6 *fl6, int flags)
1142 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1143 flags |= RT6_LOOKUP_F_IFACE;
1145 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1148 void ip6_route_input(struct sk_buff *skb)
1150 const struct ipv6hdr *iph = ipv6_hdr(skb);
1151 struct net *net = dev_net(skb->dev);
1152 int flags = RT6_LOOKUP_F_HAS_SADDR;
1153 struct ip_tunnel_info *tun_info;
1154 struct flowi6 fl6 = {
1155 .flowi6_iif = l3mdev_fib_oif(skb->dev),
1156 .daddr = iph->daddr,
1157 .saddr = iph->saddr,
1158 .flowlabel = ip6_flowinfo(iph),
1159 .flowi6_mark = skb->mark,
1160 .flowi6_proto = iph->nexthdr,
1163 tun_info = skb_tunnel_info(skb);
1164 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1165 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1167 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1170 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1171 struct flowi6 *fl6, int flags)
1173 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1176 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1177 struct flowi6 *fl6, int flags)
1179 struct dst_entry *dst;
1182 dst = l3mdev_rt6_dst_by_oif(net, fl6);
1186 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1188 any_src = ipv6_addr_any(&fl6->saddr);
1189 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1190 (fl6->flowi6_oif && any_src))
1191 flags |= RT6_LOOKUP_F_IFACE;
1194 flags |= RT6_LOOKUP_F_HAS_SADDR;
1196 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1198 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1200 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1202 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1204 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1205 struct dst_entry *new = NULL;
1207 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1213 new->input = dst_discard;
1214 new->output = dst_discard_out;
1216 dst_copy_metrics(new, &ort->dst);
1217 rt->rt6i_idev = ort->rt6i_idev;
1219 in6_dev_hold(rt->rt6i_idev);
1221 rt->rt6i_gateway = ort->rt6i_gateway;
1222 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1223 rt->rt6i_metric = 0;
1225 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1226 #ifdef CONFIG_IPV6_SUBTREES
1227 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1233 dst_release(dst_orig);
1234 return new ? new : ERR_PTR(-ENOMEM);
1238 * Destination cache support functions
1241 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1244 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1245 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1248 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1252 if (!rt6_get_cookie_safe(rt, &rt_cookie) || rt_cookie != cookie)
1255 if (rt6_check_expired(rt))
1261 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1263 if (!__rt6_check_expired(rt) &&
1264 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1265 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1271 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1273 struct rt6_info *rt;
1275 rt = (struct rt6_info *) dst;
1277 /* All IPV6 dsts are created with ->obsolete set to the value
1278 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1279 * into this function always.
1282 rt6_dst_from_metrics_check(rt);
1284 if (rt->rt6i_flags & RTF_PCPU ||
1285 (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1286 return rt6_dst_from_check(rt, cookie);
1288 return rt6_check(rt, cookie);
1291 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1293 struct rt6_info *rt = (struct rt6_info *) dst;
1296 if (rt->rt6i_flags & RTF_CACHE) {
1297 if (rt6_check_expired(rt)) {
1309 static void ip6_link_failure(struct sk_buff *skb)
1311 struct rt6_info *rt;
1313 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1315 rt = (struct rt6_info *) skb_dst(skb);
1317 if (rt->rt6i_flags & RTF_CACHE) {
1321 struct fib6_node *fn;
1324 fn = rcu_dereference(rt->rt6i_node);
1325 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
1332 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1334 struct net *net = dev_net(rt->dst.dev);
1336 rt->rt6i_flags |= RTF_MODIFIED;
1337 rt->rt6i_pmtu = mtu;
1338 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1341 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1343 return !(rt->rt6i_flags & RTF_CACHE) &&
1344 (rt->rt6i_flags & RTF_PCPU ||
1345 rcu_access_pointer(rt->rt6i_node));
1348 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1349 const struct ipv6hdr *iph, u32 mtu)
1351 struct rt6_info *rt6 = (struct rt6_info *)dst;
1353 if (rt6->rt6i_flags & RTF_LOCAL)
1357 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1358 if (mtu >= dst_mtu(dst))
1361 if (!rt6_cache_allowed_for_pmtu(rt6)) {
1362 rt6_do_update_pmtu(rt6, mtu);
1364 const struct in6_addr *daddr, *saddr;
1365 struct rt6_info *nrt6;
1368 daddr = &iph->daddr;
1369 saddr = &iph->saddr;
1371 daddr = &sk->sk_v6_daddr;
1372 saddr = &inet6_sk(sk)->saddr;
1376 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1378 rt6_do_update_pmtu(nrt6, mtu);
1380 /* ip6_ins_rt(nrt6) will bump the
1381 * rt6->rt6i_node->fn_sernum
1382 * which will fail the next rt6_check() and
1383 * invalidate the sk->sk_dst_cache.
1390 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1391 struct sk_buff *skb, u32 mtu)
1393 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1396 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1397 int oif, u32 mark, kuid_t uid)
1399 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1400 struct dst_entry *dst;
1403 memset(&fl6, 0, sizeof(fl6));
1404 fl6.flowi6_oif = oif;
1405 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1406 fl6.daddr = iph->daddr;
1407 fl6.saddr = iph->saddr;
1408 fl6.flowlabel = ip6_flowinfo(iph);
1409 fl6.flowi6_uid = uid;
1411 dst = ip6_route_output(net, NULL, &fl6);
1413 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1416 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1418 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1420 int oif = sk->sk_bound_dev_if;
1422 if (!oif && skb->dev)
1423 oif = l3mdev_master_ifindex(skb->dev);
1425 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
1427 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1429 /* Handle redirects */
1430 struct ip6rd_flowi {
1432 struct in6_addr gateway;
1435 static struct rt6_info *__ip6_route_redirect(struct net *net,
1436 struct fib6_table *table,
1440 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1441 struct rt6_info *rt;
1442 struct fib6_node *fn;
1444 /* Get the "current" route for this destination and
1445 * check if the redirect has come from approriate router.
1447 * RFC 4861 specifies that redirects should only be
1448 * accepted if they come from the nexthop to the target.
1449 * Due to the way the routes are chosen, this notion
1450 * is a bit fuzzy and one might need to check all possible
1454 read_lock_bh(&table->tb6_lock);
1455 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1457 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1458 if (rt6_check_expired(rt))
1462 if (!(rt->rt6i_flags & RTF_GATEWAY))
1464 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1466 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1472 rt = net->ipv6.ip6_null_entry;
1473 else if (rt->dst.error) {
1474 rt = net->ipv6.ip6_null_entry;
1478 if (rt == net->ipv6.ip6_null_entry) {
1479 fn = fib6_backtrack(fn, &fl6->saddr);
1487 read_unlock_bh(&table->tb6_lock);
1492 static struct dst_entry *ip6_route_redirect(struct net *net,
1493 const struct flowi6 *fl6,
1494 const struct in6_addr *gateway)
1496 int flags = RT6_LOOKUP_F_HAS_SADDR;
1497 struct ip6rd_flowi rdfl;
1500 rdfl.gateway = *gateway;
1502 return fib6_rule_lookup(net, &rdfl.fl6,
1503 flags, __ip6_route_redirect);
1506 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
1509 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1510 struct dst_entry *dst;
1513 memset(&fl6, 0, sizeof(fl6));
1514 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1515 fl6.flowi6_oif = oif;
1516 fl6.flowi6_mark = mark;
1517 fl6.daddr = iph->daddr;
1518 fl6.saddr = iph->saddr;
1519 fl6.flowlabel = ip6_flowinfo(iph);
1520 fl6.flowi6_uid = uid;
1522 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1523 rt6_do_redirect(dst, NULL, skb);
1526 EXPORT_SYMBOL_GPL(ip6_redirect);
1528 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1531 const struct ipv6hdr *iph = ipv6_hdr(skb);
1532 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1533 struct dst_entry *dst;
1536 memset(&fl6, 0, sizeof(fl6));
1537 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1538 fl6.flowi6_oif = oif;
1539 fl6.flowi6_mark = mark;
1540 fl6.daddr = msg->dest;
1541 fl6.saddr = iph->daddr;
1542 fl6.flowi6_uid = sock_net_uid(net, NULL);
1544 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1545 rt6_do_redirect(dst, NULL, skb);
1549 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1551 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
1554 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1556 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1558 struct net_device *dev = dst->dev;
1559 unsigned int mtu = dst_mtu(dst);
1560 struct net *net = dev_net(dev);
1562 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1564 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1565 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1568 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1569 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1570 * IPV6_MAXPLEN is also valid and means: "any MSS,
1571 * rely only on pmtu discovery"
1573 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1578 static unsigned int ip6_mtu(const struct dst_entry *dst)
1580 const struct rt6_info *rt = (const struct rt6_info *)dst;
1581 unsigned int mtu = rt->rt6i_pmtu;
1582 struct inet6_dev *idev;
1587 mtu = dst_metric_raw(dst, RTAX_MTU);
1594 idev = __in6_dev_get(dst->dev);
1596 mtu = idev->cnf.mtu6;
1600 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1603 static struct dst_entry *icmp6_dst_gc_list;
1604 static DEFINE_SPINLOCK(icmp6_dst_lock);
1606 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1609 struct dst_entry *dst;
1610 struct rt6_info *rt;
1611 struct inet6_dev *idev = in6_dev_get(dev);
1612 struct net *net = dev_net(dev);
1614 if (unlikely(!idev))
1615 return ERR_PTR(-ENODEV);
1617 rt = ip6_dst_alloc(net, dev, 0);
1618 if (unlikely(!rt)) {
1620 dst = ERR_PTR(-ENOMEM);
1624 rt->dst.flags |= DST_HOST;
1625 rt->dst.input = ip6_input;
1626 rt->dst.output = ip6_output;
1627 atomic_set(&rt->dst.__refcnt, 1);
1628 rt->rt6i_gateway = fl6->daddr;
1629 rt->rt6i_dst.addr = fl6->daddr;
1630 rt->rt6i_dst.plen = 128;
1631 rt->rt6i_idev = idev;
1632 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1634 spin_lock_bh(&icmp6_dst_lock);
1635 rt->dst.next = icmp6_dst_gc_list;
1636 icmp6_dst_gc_list = &rt->dst;
1637 spin_unlock_bh(&icmp6_dst_lock);
1639 fib6_force_start_gc(net);
1641 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1647 int icmp6_dst_gc(void)
1649 struct dst_entry *dst, **pprev;
1652 spin_lock_bh(&icmp6_dst_lock);
1653 pprev = &icmp6_dst_gc_list;
1655 while ((dst = *pprev) != NULL) {
1656 if (!atomic_read(&dst->__refcnt)) {
1665 spin_unlock_bh(&icmp6_dst_lock);
1670 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1673 struct dst_entry *dst, **pprev;
1675 spin_lock_bh(&icmp6_dst_lock);
1676 pprev = &icmp6_dst_gc_list;
1677 while ((dst = *pprev) != NULL) {
1678 struct rt6_info *rt = (struct rt6_info *) dst;
1679 if (func(rt, arg)) {
1686 spin_unlock_bh(&icmp6_dst_lock);
1689 static int ip6_dst_gc(struct dst_ops *ops)
1691 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1692 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1693 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1694 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1695 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1696 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1699 entries = dst_entries_get_fast(ops);
1700 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1701 entries <= rt_max_size)
1704 net->ipv6.ip6_rt_gc_expire++;
1705 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1706 entries = dst_entries_get_slow(ops);
1707 if (entries < ops->gc_thresh)
1708 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1710 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1711 return entries > rt_max_size;
1714 static int ip6_convert_metrics(struct mx6_config *mxc,
1715 const struct fib6_config *cfg)
1717 bool ecn_ca = false;
1725 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1729 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1730 int type = nla_type(nla);
1735 if (unlikely(type > RTAX_MAX))
1738 if (type == RTAX_CC_ALGO) {
1739 char tmp[TCP_CA_NAME_MAX];
1741 nla_strlcpy(tmp, nla, sizeof(tmp));
1742 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1743 if (val == TCP_CA_UNSPEC)
1746 val = nla_get_u32(nla);
1748 if (type == RTAX_HOPLIMIT && val > 255)
1750 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1754 __set_bit(type - 1, mxc->mx_valid);
1758 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1759 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1769 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1771 struct net *net = cfg->fc_nlinfo.nl_net;
1772 struct rt6_info *rt = NULL;
1773 struct net_device *dev = NULL;
1774 struct inet6_dev *idev = NULL;
1775 struct fib6_table *table;
1779 /* RTF_PCPU is an internal flag; can not be set by userspace */
1780 if (cfg->fc_flags & RTF_PCPU)
1783 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1785 #ifndef CONFIG_IPV6_SUBTREES
1786 if (cfg->fc_src_len)
1789 if (cfg->fc_ifindex) {
1791 dev = dev_get_by_index(net, cfg->fc_ifindex);
1794 idev = in6_dev_get(dev);
1799 if (cfg->fc_metric == 0)
1800 cfg->fc_metric = IP6_RT_PRIO_USER;
1803 if (cfg->fc_nlinfo.nlh &&
1804 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1805 table = fib6_get_table(net, cfg->fc_table);
1807 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1808 table = fib6_new_table(net, cfg->fc_table);
1811 table = fib6_new_table(net, cfg->fc_table);
1817 rt = ip6_dst_alloc(net, NULL,
1818 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1825 if (cfg->fc_flags & RTF_EXPIRES)
1826 rt6_set_expires(rt, jiffies +
1827 clock_t_to_jiffies(cfg->fc_expires));
1829 rt6_clean_expires(rt);
1831 if (cfg->fc_protocol == RTPROT_UNSPEC)
1832 cfg->fc_protocol = RTPROT_BOOT;
1833 rt->rt6i_protocol = cfg->fc_protocol;
1835 addr_type = ipv6_addr_type(&cfg->fc_dst);
1837 if (addr_type & IPV6_ADDR_MULTICAST)
1838 rt->dst.input = ip6_mc_input;
1839 else if (cfg->fc_flags & RTF_LOCAL)
1840 rt->dst.input = ip6_input;
1842 rt->dst.input = ip6_forward;
1844 rt->dst.output = ip6_output;
1846 if (cfg->fc_encap) {
1847 struct lwtunnel_state *lwtstate;
1849 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1850 cfg->fc_encap, AF_INET6, cfg,
1854 rt->dst.lwtstate = lwtstate_get(lwtstate);
1855 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1856 rt->dst.lwtstate->orig_output = rt->dst.output;
1857 rt->dst.output = lwtunnel_output;
1859 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1860 rt->dst.lwtstate->orig_input = rt->dst.input;
1861 rt->dst.input = lwtunnel_input;
1865 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1866 rt->rt6i_dst.plen = cfg->fc_dst_len;
1867 if (rt->rt6i_dst.plen == 128)
1868 rt->dst.flags |= DST_HOST;
1870 #ifdef CONFIG_IPV6_SUBTREES
1871 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1872 rt->rt6i_src.plen = cfg->fc_src_len;
1875 rt->rt6i_metric = cfg->fc_metric;
1877 /* We cannot add true routes via loopback here,
1878 they would result in kernel looping; promote them to reject routes
1880 if ((cfg->fc_flags & RTF_REJECT) ||
1881 (dev && (dev->flags & IFF_LOOPBACK) &&
1882 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1883 !(cfg->fc_flags & RTF_LOCAL))) {
1884 /* hold loopback dev/idev if we haven't done so. */
1885 if (dev != net->loopback_dev) {
1890 dev = net->loopback_dev;
1892 idev = in6_dev_get(dev);
1898 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1899 switch (cfg->fc_type) {
1901 rt->dst.error = -EINVAL;
1902 rt->dst.output = dst_discard_out;
1903 rt->dst.input = dst_discard;
1906 rt->dst.error = -EACCES;
1907 rt->dst.output = ip6_pkt_prohibit_out;
1908 rt->dst.input = ip6_pkt_prohibit;
1911 case RTN_UNREACHABLE:
1913 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1914 : (cfg->fc_type == RTN_UNREACHABLE)
1915 ? -EHOSTUNREACH : -ENETUNREACH;
1916 rt->dst.output = ip6_pkt_discard_out;
1917 rt->dst.input = ip6_pkt_discard;
1923 if (cfg->fc_flags & RTF_GATEWAY) {
1924 const struct in6_addr *gw_addr;
1927 gw_addr = &cfg->fc_gateway;
1928 gwa_type = ipv6_addr_type(gw_addr);
1930 /* if gw_addr is local we will fail to detect this in case
1931 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1932 * will return already-added prefix route via interface that
1933 * prefix route was assigned to, which might be non-loopback.
1936 if (ipv6_chk_addr_and_flags(net, gw_addr,
1937 gwa_type & IPV6_ADDR_LINKLOCAL ?
1941 rt->rt6i_gateway = *gw_addr;
1943 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1944 struct rt6_info *grt;
1946 /* IPv6 strictly inhibits using not link-local
1947 addresses as nexthop address.
1948 Otherwise, router will not able to send redirects.
1949 It is very good, but in some (rare!) circumstances
1950 (SIT, PtP, NBMA NOARP links) it is handy to allow
1951 some exceptions. --ANK
1953 if (!(gwa_type & IPV6_ADDR_UNICAST))
1956 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1958 err = -EHOSTUNREACH;
1962 if (dev != grt->dst.dev) {
1968 idev = grt->rt6i_idev;
1970 in6_dev_hold(grt->rt6i_idev);
1972 if (!(grt->rt6i_flags & RTF_GATEWAY))
1980 if (!dev || (dev->flags & IFF_LOOPBACK))
1988 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1989 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1993 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1994 rt->rt6i_prefsrc.plen = 128;
1996 rt->rt6i_prefsrc.plen = 0;
1998 rt->rt6i_flags = cfg->fc_flags;
2002 rt->rt6i_idev = idev;
2003 rt->rt6i_table = table;
2005 cfg->fc_nlinfo.nl_net = dev_net(dev);
2016 return ERR_PTR(err);
2019 int ip6_route_add(struct fib6_config *cfg)
2021 struct mx6_config mxc = { .mx = NULL, };
2022 struct rt6_info *rt;
2025 rt = ip6_route_info_create(cfg);
2032 err = ip6_convert_metrics(&mxc, cfg);
2036 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2048 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2051 struct fib6_table *table;
2052 struct net *net = dev_net(rt->dst.dev);
2054 if (rt == net->ipv6.ip6_null_entry ||
2055 rt->dst.flags & DST_NOCACHE) {
2060 table = rt->rt6i_table;
2061 write_lock_bh(&table->tb6_lock);
2062 err = fib6_del(rt, info);
2063 write_unlock_bh(&table->tb6_lock);
2070 int ip6_del_rt(struct rt6_info *rt)
2072 struct nl_info info = {
2073 .nl_net = dev_net(rt->dst.dev),
2075 return __ip6_del_rt(rt, &info);
2078 static int ip6_route_del(struct fib6_config *cfg)
2080 struct fib6_table *table;
2081 struct fib6_node *fn;
2082 struct rt6_info *rt;
2085 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2089 read_lock_bh(&table->tb6_lock);
2091 fn = fib6_locate(&table->tb6_root,
2092 &cfg->fc_dst, cfg->fc_dst_len,
2093 &cfg->fc_src, cfg->fc_src_len);
2096 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2097 if ((rt->rt6i_flags & RTF_CACHE) &&
2098 !(cfg->fc_flags & RTF_CACHE))
2100 if (cfg->fc_ifindex &&
2102 rt->dst.dev->ifindex != cfg->fc_ifindex))
2104 if (cfg->fc_flags & RTF_GATEWAY &&
2105 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2107 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2109 if (cfg->fc_protocol && cfg->fc_protocol != rt->rt6i_protocol)
2112 read_unlock_bh(&table->tb6_lock);
2114 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2117 read_unlock_bh(&table->tb6_lock);
2122 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2124 struct netevent_redirect netevent;
2125 struct rt6_info *rt, *nrt = NULL;
2126 struct ndisc_options ndopts;
2127 struct inet6_dev *in6_dev;
2128 struct neighbour *neigh;
2130 int optlen, on_link;
2133 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2134 optlen -= sizeof(*msg);
2137 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2141 msg = (struct rd_msg *)icmp6_hdr(skb);
2143 if (ipv6_addr_is_multicast(&msg->dest)) {
2144 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2149 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2151 } else if (ipv6_addr_type(&msg->target) !=
2152 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2153 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2157 in6_dev = __in6_dev_get(skb->dev);
2160 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2164 * The IP source address of the Redirect MUST be the same as the current
2165 * first-hop router for the specified ICMP Destination Address.
2168 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2169 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2174 if (ndopts.nd_opts_tgt_lladdr) {
2175 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2178 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2183 rt = (struct rt6_info *) dst;
2184 if (rt->rt6i_flags & RTF_REJECT) {
2185 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2189 /* Redirect received -> path was valid.
2190 * Look, redirects are sent only in response to data packets,
2191 * so that this nexthop apparently is reachable. --ANK
2193 dst_confirm(&rt->dst);
2195 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2200 * We have finally decided to accept it.
2203 neigh_update(neigh, lladdr, NUD_STALE,
2204 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2205 NEIGH_UPDATE_F_OVERRIDE|
2206 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2207 NEIGH_UPDATE_F_ISROUTER))
2210 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2214 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2216 nrt->rt6i_flags &= ~RTF_GATEWAY;
2218 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2220 if (ip6_ins_rt(nrt))
2223 netevent.old = &rt->dst;
2224 netevent.new = &nrt->dst;
2225 netevent.daddr = &msg->dest;
2226 netevent.neigh = neigh;
2227 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2229 if (rt->rt6i_flags & RTF_CACHE) {
2230 rt = (struct rt6_info *) dst_clone(&rt->dst);
2235 neigh_release(neigh);
2239 * Misc support functions
2242 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2244 BUG_ON(from->dst.from);
2246 rt->rt6i_flags &= ~RTF_EXPIRES;
2247 dst_hold(&from->dst);
2248 rt->dst.from = &from->dst;
2249 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2252 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2254 rt->dst.input = ort->dst.input;
2255 rt->dst.output = ort->dst.output;
2256 rt->rt6i_dst = ort->rt6i_dst;
2257 rt->dst.error = ort->dst.error;
2258 rt->rt6i_idev = ort->rt6i_idev;
2260 in6_dev_hold(rt->rt6i_idev);
2261 rt->dst.lastuse = jiffies;
2262 rt->rt6i_gateway = ort->rt6i_gateway;
2263 rt->rt6i_flags = ort->rt6i_flags;
2264 rt6_set_from(rt, ort);
2265 rt->rt6i_metric = ort->rt6i_metric;
2266 #ifdef CONFIG_IPV6_SUBTREES
2267 rt->rt6i_src = ort->rt6i_src;
2269 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2270 rt->rt6i_table = ort->rt6i_table;
2271 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2274 #ifdef CONFIG_IPV6_ROUTE_INFO
2275 static struct rt6_info *rt6_get_route_info(struct net_device *dev,
2276 const struct in6_addr *prefix, int prefixlen,
2277 const struct in6_addr *gwaddr)
2279 struct fib6_node *fn;
2280 struct rt6_info *rt = NULL;
2281 struct fib6_table *table;
2283 table = fib6_get_table(dev_net(dev),
2284 addrconf_rt_table(dev, RT6_TABLE_INFO));
2288 read_lock_bh(&table->tb6_lock);
2289 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2293 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2294 if (rt->dst.dev->ifindex != dev->ifindex)
2296 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2298 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2304 read_unlock_bh(&table->tb6_lock);
2308 static struct rt6_info *rt6_add_route_info(struct net_device *dev,
2309 const struct in6_addr *prefix, int prefixlen,
2310 const struct in6_addr *gwaddr, unsigned int pref)
2312 struct fib6_config cfg = {
2313 .fc_metric = IP6_RT_PRIO_USER,
2314 .fc_ifindex = dev->ifindex,
2315 .fc_dst_len = prefixlen,
2316 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2317 RTF_UP | RTF_PREF(pref),
2318 .fc_nlinfo.portid = 0,
2319 .fc_nlinfo.nlh = NULL,
2320 .fc_nlinfo.nl_net = dev_net(dev),
2323 cfg.fc_table = l3mdev_fib_table_by_index(dev_net(dev), dev->ifindex) ? : addrconf_rt_table(dev, RT6_TABLE_INFO);
2324 cfg.fc_dst = *prefix;
2325 cfg.fc_gateway = *gwaddr;
2327 /* We should treat it as a default route if prefix length is 0. */
2329 cfg.fc_flags |= RTF_DEFAULT;
2331 ip6_route_add(&cfg);
2333 return rt6_get_route_info(dev, prefix, prefixlen, gwaddr);
2337 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2339 struct rt6_info *rt;
2340 struct fib6_table *table;
2342 table = fib6_get_table(dev_net(dev),
2343 addrconf_rt_table(dev, RT6_TABLE_MAIN));
2347 read_lock_bh(&table->tb6_lock);
2348 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2349 if (dev == rt->dst.dev &&
2350 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2351 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2356 read_unlock_bh(&table->tb6_lock);
2360 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2361 struct net_device *dev,
2364 struct fib6_config cfg = {
2365 .fc_table = l3mdev_fib_table(dev) ? : addrconf_rt_table(dev, RT6_TABLE_DFLT),
2366 .fc_metric = IP6_RT_PRIO_USER,
2367 .fc_ifindex = dev->ifindex,
2368 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2369 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2370 .fc_nlinfo.portid = 0,
2371 .fc_nlinfo.nlh = NULL,
2372 .fc_nlinfo.nl_net = dev_net(dev),
2375 cfg.fc_gateway = *gwaddr;
2377 ip6_route_add(&cfg);
2379 return rt6_get_dflt_router(gwaddr, dev);
2383 int rt6_addrconf_purge(struct rt6_info *rt, void *arg) {
2384 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2385 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2))
2390 void rt6_purge_dflt_routers(struct net *net)
2392 fib6_clean_all(net, rt6_addrconf_purge, NULL);
2395 static void rtmsg_to_fib6_config(struct net *net,
2396 struct in6_rtmsg *rtmsg,
2397 struct fib6_config *cfg)
2399 memset(cfg, 0, sizeof(*cfg));
2401 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2403 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2404 cfg->fc_metric = rtmsg->rtmsg_metric;
2405 cfg->fc_expires = rtmsg->rtmsg_info;
2406 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2407 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2408 cfg->fc_flags = rtmsg->rtmsg_flags;
2410 cfg->fc_nlinfo.nl_net = net;
2412 cfg->fc_dst = rtmsg->rtmsg_dst;
2413 cfg->fc_src = rtmsg->rtmsg_src;
2414 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2417 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2419 struct fib6_config cfg;
2420 struct in6_rtmsg rtmsg;
2424 case SIOCADDRT: /* Add a route */
2425 case SIOCDELRT: /* Delete a route */
2426 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2428 err = copy_from_user(&rtmsg, arg,
2429 sizeof(struct in6_rtmsg));
2433 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2438 err = ip6_route_add(&cfg);
2441 err = ip6_route_del(&cfg);
2455 * Drop the packet on the floor
2458 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2461 struct dst_entry *dst = skb_dst(skb);
2462 switch (ipstats_mib_noroutes) {
2463 case IPSTATS_MIB_INNOROUTES:
2464 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2465 if (type == IPV6_ADDR_ANY) {
2466 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2467 IPSTATS_MIB_INADDRERRORS);
2471 case IPSTATS_MIB_OUTNOROUTES:
2472 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2473 ipstats_mib_noroutes);
2476 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2481 static int ip6_pkt_discard(struct sk_buff *skb)
2483 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2486 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2488 skb->dev = skb_dst(skb)->dev;
2489 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2492 static int ip6_pkt_prohibit(struct sk_buff *skb)
2494 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2497 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2499 skb->dev = skb_dst(skb)->dev;
2500 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2504 * Allocate a dst for local (unicast / anycast) address.
2507 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2508 const struct in6_addr *addr,
2512 struct net *net = dev_net(idev->dev);
2513 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2516 return ERR_PTR(-ENOMEM);
2520 rt->dst.flags |= DST_HOST;
2521 rt->dst.input = ip6_input;
2522 rt->dst.output = ip6_output;
2523 rt->rt6i_idev = idev;
2525 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2527 rt->rt6i_flags |= RTF_ANYCAST;
2529 rt->rt6i_flags |= RTF_LOCAL;
2531 rt->rt6i_gateway = *addr;
2532 rt->rt6i_dst.addr = *addr;
2533 rt->rt6i_dst.plen = 128;
2534 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2535 rt->rt6i_table = fib6_get_table(net, tb_id);
2536 rt->dst.flags |= DST_NOCACHE;
2538 atomic_set(&rt->dst.__refcnt, 1);
2543 int ip6_route_get_saddr(struct net *net,
2544 struct rt6_info *rt,
2545 const struct in6_addr *daddr,
2547 struct in6_addr *saddr)
2549 struct inet6_dev *idev =
2550 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2552 if (rt && rt->rt6i_prefsrc.plen)
2553 *saddr = rt->rt6i_prefsrc.addr;
2555 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2556 daddr, prefs, saddr);
2560 /* remove deleted ip from prefsrc entries */
2561 struct arg_dev_net_ip {
2562 struct net_device *dev;
2564 struct in6_addr *addr;
2567 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2569 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2570 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2571 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2573 if (((void *)rt->dst.dev == dev || !dev) &&
2574 rt != net->ipv6.ip6_null_entry &&
2575 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2576 /* remove prefsrc entry */
2577 rt->rt6i_prefsrc.plen = 0;
2582 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2584 struct net *net = dev_net(ifp->idev->dev);
2585 struct arg_dev_net_ip adni = {
2586 .dev = ifp->idev->dev,
2590 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2593 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2594 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2596 /* Remove routers and update dst entries when gateway turn into host. */
2597 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2599 struct in6_addr *gateway = (struct in6_addr *)arg;
2601 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2602 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2603 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2609 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2611 fib6_clean_all(net, fib6_clean_tohost, gateway);
2614 struct arg_dev_net {
2615 struct net_device *dev;
2619 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2621 const struct arg_dev_net *adn = arg;
2622 const struct net_device *dev = adn->dev;
2624 if ((rt->dst.dev == dev || !dev) &&
2625 rt != adn->net->ipv6.ip6_null_entry)
2631 void rt6_ifdown(struct net *net, struct net_device *dev)
2633 struct arg_dev_net adn = {
2638 fib6_clean_all(net, fib6_ifdown, &adn);
2639 icmp6_clean_all(fib6_ifdown, &adn);
2641 rt6_uncached_list_flush_dev(net, dev);
2644 struct rt6_mtu_change_arg {
2645 struct net_device *dev;
2649 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2651 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2652 struct inet6_dev *idev;
2654 /* In IPv6 pmtu discovery is not optional,
2655 so that RTAX_MTU lock cannot disable it.
2656 We still use this lock to block changes
2657 caused by addrconf/ndisc.
2660 idev = __in6_dev_get(arg->dev);
2664 /* For administrative MTU increase, there is no way to discover
2665 IPv6 PMTU increase, so PMTU increase should be updated here.
2666 Since RFC 1981 doesn't include administrative MTU increase
2667 update PMTU increase is a MUST. (i.e. jumbo frame)
2670 If new MTU is less than route PMTU, this new MTU will be the
2671 lowest MTU in the path, update the route PMTU to reflect PMTU
2672 decreases; if new MTU is greater than route PMTU, and the
2673 old MTU is the lowest MTU in the path, update the route PMTU
2674 to reflect the increase. In this case if the other nodes' MTU
2675 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2678 if (rt->dst.dev == arg->dev &&
2679 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2680 if (rt->rt6i_flags & RTF_CACHE) {
2681 /* For RTF_CACHE with rt6i_pmtu == 0
2682 * (i.e. a redirected route),
2683 * the metrics of its rt->dst.from has already
2686 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2687 rt->rt6i_pmtu = arg->mtu;
2688 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2689 (dst_mtu(&rt->dst) < arg->mtu &&
2690 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2691 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2697 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2699 struct rt6_mtu_change_arg arg = {
2704 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2707 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2708 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2709 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
2710 [RTA_OIF] = { .type = NLA_U32 },
2711 [RTA_IIF] = { .type = NLA_U32 },
2712 [RTA_PRIORITY] = { .type = NLA_U32 },
2713 [RTA_METRICS] = { .type = NLA_NESTED },
2714 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2715 [RTA_PREF] = { .type = NLA_U8 },
2716 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2717 [RTA_ENCAP] = { .type = NLA_NESTED },
2718 [RTA_UID] = { .type = NLA_U32 },
2719 [RTA_TABLE] = { .type = NLA_U32 },
2722 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2723 struct fib6_config *cfg)
2726 struct nlattr *tb[RTA_MAX+1];
2730 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2735 rtm = nlmsg_data(nlh);
2736 memset(cfg, 0, sizeof(*cfg));
2738 cfg->fc_table = rtm->rtm_table;
2739 cfg->fc_dst_len = rtm->rtm_dst_len;
2740 cfg->fc_src_len = rtm->rtm_src_len;
2741 cfg->fc_flags = RTF_UP;
2742 cfg->fc_protocol = rtm->rtm_protocol;
2743 cfg->fc_type = rtm->rtm_type;
2745 if (rtm->rtm_type == RTN_UNREACHABLE ||
2746 rtm->rtm_type == RTN_BLACKHOLE ||
2747 rtm->rtm_type == RTN_PROHIBIT ||
2748 rtm->rtm_type == RTN_THROW)
2749 cfg->fc_flags |= RTF_REJECT;
2751 if (rtm->rtm_type == RTN_LOCAL)
2752 cfg->fc_flags |= RTF_LOCAL;
2754 if (rtm->rtm_flags & RTM_F_CLONED)
2755 cfg->fc_flags |= RTF_CACHE;
2757 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2758 cfg->fc_nlinfo.nlh = nlh;
2759 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2761 if (tb[RTA_GATEWAY]) {
2762 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2763 cfg->fc_flags |= RTF_GATEWAY;
2767 int plen = (rtm->rtm_dst_len + 7) >> 3;
2769 if (nla_len(tb[RTA_DST]) < plen)
2772 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2776 int plen = (rtm->rtm_src_len + 7) >> 3;
2778 if (nla_len(tb[RTA_SRC]) < plen)
2781 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2784 if (tb[RTA_PREFSRC])
2785 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2788 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2790 if (tb[RTA_PRIORITY])
2791 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2793 if (tb[RTA_METRICS]) {
2794 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2795 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2799 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2801 if (tb[RTA_MULTIPATH]) {
2802 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2803 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2807 pref = nla_get_u8(tb[RTA_PREF]);
2808 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2809 pref != ICMPV6_ROUTER_PREF_HIGH)
2810 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2811 cfg->fc_flags |= RTF_PREF(pref);
2815 cfg->fc_encap = tb[RTA_ENCAP];
2817 if (tb[RTA_ENCAP_TYPE])
2818 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2826 struct rt6_info *rt6_info;
2827 struct fib6_config r_cfg;
2828 struct mx6_config mxc;
2829 struct list_head next;
2832 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2836 list_for_each_entry(nh, rt6_nh_list, next) {
2837 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2838 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2839 nh->r_cfg.fc_ifindex);
2843 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2844 struct rt6_info *rt, struct fib6_config *r_cfg)
2849 list_for_each_entry(nh, rt6_nh_list, next) {
2850 /* check if rt6_info already exists */
2851 if (rt6_duplicate_nexthop(nh->rt6_info, rt))
2855 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2859 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2864 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2865 list_add_tail(&nh->next, rt6_nh_list);
2870 static int ip6_route_multipath_add(struct fib6_config *cfg)
2872 struct fib6_config r_cfg;
2873 struct rtnexthop *rtnh;
2874 struct rt6_info *rt;
2875 struct rt6_nh *err_nh;
2876 struct rt6_nh *nh, *nh_safe;
2881 int replace = (cfg->fc_nlinfo.nlh &&
2882 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2883 LIST_HEAD(rt6_nh_list);
2885 remaining = cfg->fc_mp_len;
2886 rtnh = (struct rtnexthop *)cfg->fc_mp;
2888 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2889 * rt6_info structs per nexthop
2891 while (rtnh_ok(rtnh, remaining)) {
2892 memcpy(&r_cfg, cfg, sizeof(*cfg));
2893 if (rtnh->rtnh_ifindex)
2894 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2896 attrlen = rtnh_attrlen(rtnh);
2898 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2900 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2902 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2903 r_cfg.fc_flags |= RTF_GATEWAY;
2905 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2906 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2908 r_cfg.fc_encap_type = nla_get_u16(nla);
2911 rt = ip6_route_info_create(&r_cfg);
2918 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2924 rtnh = rtnh_next(rtnh, &remaining);
2928 list_for_each_entry(nh, &rt6_nh_list, next) {
2929 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2930 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2931 nh->rt6_info = NULL;
2934 ip6_print_replace_route_err(&rt6_nh_list);
2939 /* Because each route is added like a single route we remove
2940 * these flags after the first nexthop: if there is a collision,
2941 * we have already failed to add the first nexthop:
2942 * fib6_add_rt2node() has rejected it; when replacing, old
2943 * nexthops have been replaced by first new, the rest should
2946 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2954 /* Delete routes that were already added */
2955 list_for_each_entry(nh, &rt6_nh_list, next) {
2958 ip6_route_del(&nh->r_cfg);
2962 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2964 dst_free(&nh->rt6_info->dst);
2966 list_del(&nh->next);
2973 static int ip6_route_multipath_del(struct fib6_config *cfg)
2975 struct fib6_config r_cfg;
2976 struct rtnexthop *rtnh;
2979 int err = 1, last_err = 0;
2981 remaining = cfg->fc_mp_len;
2982 rtnh = (struct rtnexthop *)cfg->fc_mp;
2984 /* Parse a Multipath Entry */
2985 while (rtnh_ok(rtnh, remaining)) {
2986 memcpy(&r_cfg, cfg, sizeof(*cfg));
2987 if (rtnh->rtnh_ifindex)
2988 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2990 attrlen = rtnh_attrlen(rtnh);
2992 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2994 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2996 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2997 r_cfg.fc_flags |= RTF_GATEWAY;
3000 err = ip6_route_del(&r_cfg);
3004 rtnh = rtnh_next(rtnh, &remaining);
3010 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3012 struct fib6_config cfg;
3015 err = rtm_to_fib6_config(skb, nlh, &cfg);
3020 return ip6_route_multipath_del(&cfg);
3022 return ip6_route_del(&cfg);
3025 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3027 struct fib6_config cfg;
3030 err = rtm_to_fib6_config(skb, nlh, &cfg);
3035 return ip6_route_multipath_add(&cfg);
3037 return ip6_route_add(&cfg);
3040 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3042 return NLMSG_ALIGN(sizeof(struct rtmsg))
3043 + nla_total_size(16) /* RTA_SRC */
3044 + nla_total_size(16) /* RTA_DST */
3045 + nla_total_size(16) /* RTA_GATEWAY */
3046 + nla_total_size(16) /* RTA_PREFSRC */
3047 + nla_total_size(4) /* RTA_TABLE */
3048 + nla_total_size(4) /* RTA_IIF */
3049 + nla_total_size(4) /* RTA_OIF */
3050 + nla_total_size(4) /* RTA_PRIORITY */
3051 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3052 + nla_total_size(sizeof(struct rta_cacheinfo))
3053 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3054 + nla_total_size(1) /* RTA_PREF */
3055 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3058 static int rt6_fill_node(struct net *net,
3059 struct sk_buff *skb, struct rt6_info *rt,
3060 struct in6_addr *dst, struct in6_addr *src,
3061 int iif, int type, u32 portid, u32 seq,
3062 int prefix, int nowait, unsigned int flags)
3064 u32 metrics[RTAX_MAX];
3066 struct nlmsghdr *nlh;
3070 if (prefix) { /* user wants prefix routes only */
3071 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3072 /* success since this is not a prefix route */
3077 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3081 rtm = nlmsg_data(nlh);
3082 rtm->rtm_family = AF_INET6;
3083 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3084 rtm->rtm_src_len = rt->rt6i_src.plen;
3087 table = rt->rt6i_table->tb6_id;
3089 table = RT6_TABLE_UNSPEC;
3090 rtm->rtm_table = table;
3091 if (nla_put_u32(skb, RTA_TABLE, table))
3092 goto nla_put_failure;
3093 if (rt->rt6i_flags & RTF_REJECT) {
3094 switch (rt->dst.error) {
3096 rtm->rtm_type = RTN_BLACKHOLE;
3099 rtm->rtm_type = RTN_PROHIBIT;
3102 rtm->rtm_type = RTN_THROW;
3105 rtm->rtm_type = RTN_UNREACHABLE;
3109 else if (rt->rt6i_flags & RTF_LOCAL)
3110 rtm->rtm_type = RTN_LOCAL;
3111 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3112 rtm->rtm_type = RTN_LOCAL;
3114 rtm->rtm_type = RTN_UNICAST;
3116 if (!netif_carrier_ok(rt->dst.dev)) {
3117 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3118 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3119 rtm->rtm_flags |= RTNH_F_DEAD;
3121 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3122 rtm->rtm_protocol = rt->rt6i_protocol;
3123 if (rt->rt6i_flags & RTF_DYNAMIC)
3124 rtm->rtm_protocol = RTPROT_REDIRECT;
3125 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3126 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3127 rtm->rtm_protocol = RTPROT_RA;
3129 rtm->rtm_protocol = RTPROT_KERNEL;
3132 if (rt->rt6i_flags & RTF_CACHE)
3133 rtm->rtm_flags |= RTM_F_CLONED;
3136 if (nla_put_in6_addr(skb, RTA_DST, dst))
3137 goto nla_put_failure;
3138 rtm->rtm_dst_len = 128;
3139 } else if (rtm->rtm_dst_len)
3140 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3141 goto nla_put_failure;
3142 #ifdef CONFIG_IPV6_SUBTREES
3144 if (nla_put_in6_addr(skb, RTA_SRC, src))
3145 goto nla_put_failure;
3146 rtm->rtm_src_len = 128;
3147 } else if (rtm->rtm_src_len &&
3148 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3149 goto nla_put_failure;
3152 #ifdef CONFIG_IPV6_MROUTE
3153 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3154 int err = ip6mr_get_route(net, skb, rtm, nowait,
3161 goto nla_put_failure;
3163 if (err == -EMSGSIZE)
3164 goto nla_put_failure;
3169 if (nla_put_u32(skb, RTA_IIF, iif))
3170 goto nla_put_failure;
3172 struct in6_addr saddr_buf;
3173 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3174 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3175 goto nla_put_failure;
3178 if (rt->rt6i_prefsrc.plen) {
3179 struct in6_addr saddr_buf;
3180 saddr_buf = rt->rt6i_prefsrc.addr;
3181 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3182 goto nla_put_failure;
3185 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3187 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3188 if (rtnetlink_put_metrics(skb, metrics) < 0)
3189 goto nla_put_failure;
3191 if (rt->rt6i_flags & RTF_GATEWAY) {
3192 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3193 goto nla_put_failure;
3197 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3198 goto nla_put_failure;
3199 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3200 goto nla_put_failure;
3202 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3204 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3205 goto nla_put_failure;
3207 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3208 goto nla_put_failure;
3210 if (lwtunnel_fill_encap(skb, rt->dst.lwtstate) < 0)
3211 goto nla_put_failure;
3213 nlmsg_end(skb, nlh);
3217 nlmsg_cancel(skb, nlh);
3221 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3223 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3226 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3227 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3228 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3232 return rt6_fill_node(arg->net,
3233 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3234 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3235 prefix, 0, NLM_F_MULTI);
3238 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3240 struct net *net = sock_net(in_skb->sk);
3241 struct nlattr *tb[RTA_MAX+1];
3242 struct rt6_info *rt;
3243 struct sk_buff *skb;
3246 int err, iif = 0, oif = 0;
3248 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3253 memset(&fl6, 0, sizeof(fl6));
3256 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3259 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3263 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3266 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3270 iif = nla_get_u32(tb[RTA_IIF]);
3273 oif = nla_get_u32(tb[RTA_OIF]);
3276 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3279 fl6.flowi6_uid = make_kuid(current_user_ns(),
3280 nla_get_u32(tb[RTA_UID]));
3282 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
3285 struct net_device *dev;
3288 dev = __dev_get_by_index(net, iif);
3294 fl6.flowi6_iif = iif;
3296 if (!ipv6_addr_any(&fl6.saddr))
3297 flags |= RT6_LOOKUP_F_HAS_SADDR;
3299 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3302 fl6.flowi6_oif = oif;
3304 if (netif_index_is_l3_master(net, oif)) {
3305 fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3306 FLOWI_FLAG_SKIP_NH_OIF;
3309 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3312 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3319 /* Reserve room for dummy headers, this skb can pass
3320 through good chunk of routing engine.
3322 skb_reset_mac_header(skb);
3323 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3325 skb_dst_set(skb, &rt->dst);
3327 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3328 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3329 nlh->nlmsg_seq, 0, 0, 0);
3335 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3340 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3341 unsigned int nlm_flags)
3343 struct sk_buff *skb;
3344 struct net *net = info->nl_net;
3349 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3351 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3355 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3356 event, info->portid, seq, 0, 0, nlm_flags);
3358 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3359 WARN_ON(err == -EMSGSIZE);
3363 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3364 info->nlh, gfp_any());
3368 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3371 static int ip6_route_dev_notify(struct notifier_block *this,
3372 unsigned long event, void *ptr)
3374 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3375 struct net *net = dev_net(dev);
3377 if (!(dev->flags & IFF_LOOPBACK))
3380 if (event == NETDEV_REGISTER) {
3381 net->ipv6.ip6_null_entry->dst.dev = dev;
3382 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3383 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3384 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3385 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3386 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3387 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3389 } else if (event == NETDEV_UNREGISTER &&
3390 dev->reg_state != NETREG_UNREGISTERED) {
3391 /* NETDEV_UNREGISTER could be fired for multiple times by
3392 * netdev_wait_allrefs(). Make sure we only call this once.
3394 in6_dev_put(net->ipv6.ip6_null_entry->rt6i_idev);
3395 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3396 in6_dev_put(net->ipv6.ip6_prohibit_entry->rt6i_idev);
3397 in6_dev_put(net->ipv6.ip6_blk_hole_entry->rt6i_idev);
3408 #ifdef CONFIG_PROC_FS
3410 static const struct file_operations ipv6_route_proc_fops = {
3411 .owner = THIS_MODULE,
3412 .open = ipv6_route_open,
3414 .llseek = seq_lseek,
3415 .release = seq_release_net,
3418 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3420 struct net *net = (struct net *)seq->private;
3421 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3422 net->ipv6.rt6_stats->fib_nodes,
3423 net->ipv6.rt6_stats->fib_route_nodes,
3424 net->ipv6.rt6_stats->fib_rt_alloc,
3425 net->ipv6.rt6_stats->fib_rt_entries,
3426 net->ipv6.rt6_stats->fib_rt_cache,
3427 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3428 net->ipv6.rt6_stats->fib_discarded_routes);
3433 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3435 return single_open_net(inode, file, rt6_stats_seq_show);
3438 static const struct file_operations rt6_stats_seq_fops = {
3439 .owner = THIS_MODULE,
3440 .open = rt6_stats_seq_open,
3442 .llseek = seq_lseek,
3443 .release = single_release_net,
3445 #endif /* CONFIG_PROC_FS */
3447 #ifdef CONFIG_SYSCTL
3450 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3451 void __user *buffer, size_t *lenp, loff_t *ppos)
3458 net = (struct net *)ctl->extra1;
3459 delay = net->ipv6.sysctl.flush_delay;
3460 proc_dointvec(ctl, write, buffer, lenp, ppos);
3461 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3465 struct ctl_table ipv6_route_table_template[] = {
3467 .procname = "flush",
3468 .data = &init_net.ipv6.sysctl.flush_delay,
3469 .maxlen = sizeof(int),
3471 .proc_handler = ipv6_sysctl_rtcache_flush
3474 .procname = "gc_thresh",
3475 .data = &ip6_dst_ops_template.gc_thresh,
3476 .maxlen = sizeof(int),
3478 .proc_handler = proc_dointvec,
3481 .procname = "max_size",
3482 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3483 .maxlen = sizeof(int),
3485 .proc_handler = proc_dointvec,
3488 .procname = "gc_min_interval",
3489 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3490 .maxlen = sizeof(int),
3492 .proc_handler = proc_dointvec_jiffies,
3495 .procname = "gc_timeout",
3496 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3497 .maxlen = sizeof(int),
3499 .proc_handler = proc_dointvec_jiffies,
3502 .procname = "gc_interval",
3503 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3504 .maxlen = sizeof(int),
3506 .proc_handler = proc_dointvec_jiffies,
3509 .procname = "gc_elasticity",
3510 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3511 .maxlen = sizeof(int),
3513 .proc_handler = proc_dointvec,
3516 .procname = "mtu_expires",
3517 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3518 .maxlen = sizeof(int),
3520 .proc_handler = proc_dointvec_jiffies,
3523 .procname = "min_adv_mss",
3524 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3525 .maxlen = sizeof(int),
3527 .proc_handler = proc_dointvec,
3530 .procname = "gc_min_interval_ms",
3531 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3532 .maxlen = sizeof(int),
3534 .proc_handler = proc_dointvec_ms_jiffies,
3539 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3541 struct ctl_table *table;
3543 table = kmemdup(ipv6_route_table_template,
3544 sizeof(ipv6_route_table_template),
3548 table[0].data = &net->ipv6.sysctl.flush_delay;
3549 table[0].extra1 = net;
3550 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3551 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3552 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3553 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3554 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3555 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3556 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3557 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3558 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3560 /* Don't export sysctls to unprivileged users */
3561 if (net->user_ns != &init_user_ns)
3562 table[0].procname = NULL;
3569 static int __net_init ip6_route_net_init(struct net *net)
3573 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3574 sizeof(net->ipv6.ip6_dst_ops));
3576 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3577 goto out_ip6_dst_ops;
3579 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3580 sizeof(*net->ipv6.ip6_null_entry),
3582 if (!net->ipv6.ip6_null_entry)
3583 goto out_ip6_dst_entries;
3584 net->ipv6.ip6_null_entry->dst.path =
3585 (struct dst_entry *)net->ipv6.ip6_null_entry;
3586 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3587 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3588 ip6_template_metrics, true);
3590 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3591 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3592 sizeof(*net->ipv6.ip6_prohibit_entry),
3594 if (!net->ipv6.ip6_prohibit_entry)
3595 goto out_ip6_null_entry;
3596 net->ipv6.ip6_prohibit_entry->dst.path =
3597 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3598 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3599 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3600 ip6_template_metrics, true);
3602 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3603 sizeof(*net->ipv6.ip6_blk_hole_entry),
3605 if (!net->ipv6.ip6_blk_hole_entry)
3606 goto out_ip6_prohibit_entry;
3607 net->ipv6.ip6_blk_hole_entry->dst.path =
3608 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3609 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3610 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3611 ip6_template_metrics, true);
3614 net->ipv6.sysctl.flush_delay = 0;
3615 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3616 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3617 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3618 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3619 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3620 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3621 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3623 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3629 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3630 out_ip6_prohibit_entry:
3631 kfree(net->ipv6.ip6_prohibit_entry);
3633 kfree(net->ipv6.ip6_null_entry);
3635 out_ip6_dst_entries:
3636 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3641 static void __net_exit ip6_route_net_exit(struct net *net)
3643 kfree(net->ipv6.ip6_null_entry);
3644 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3645 kfree(net->ipv6.ip6_prohibit_entry);
3646 kfree(net->ipv6.ip6_blk_hole_entry);
3648 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3651 static int __net_init ip6_route_net_init_late(struct net *net)
3653 #ifdef CONFIG_PROC_FS
3654 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3655 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3660 static void __net_exit ip6_route_net_exit_late(struct net *net)
3662 #ifdef CONFIG_PROC_FS
3663 remove_proc_entry("ipv6_route", net->proc_net);
3664 remove_proc_entry("rt6_stats", net->proc_net);
3668 static struct pernet_operations ip6_route_net_ops = {
3669 .init = ip6_route_net_init,
3670 .exit = ip6_route_net_exit,
3673 static int __net_init ipv6_inetpeer_init(struct net *net)
3675 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3679 inet_peer_base_init(bp);
3680 net->ipv6.peers = bp;
3684 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3686 struct inet_peer_base *bp = net->ipv6.peers;
3688 net->ipv6.peers = NULL;
3689 inetpeer_invalidate_tree(bp);
3693 static struct pernet_operations ipv6_inetpeer_ops = {
3694 .init = ipv6_inetpeer_init,
3695 .exit = ipv6_inetpeer_exit,
3698 static struct pernet_operations ip6_route_net_late_ops = {
3699 .init = ip6_route_net_init_late,
3700 .exit = ip6_route_net_exit_late,
3703 static struct notifier_block ip6_route_dev_notifier = {
3704 .notifier_call = ip6_route_dev_notify,
3705 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
3708 void __init ip6_route_init_special_entries(void)
3710 /* Registering of the loopback is done before this portion of code,
3711 * the loopback reference in rt6_info will not be taken, do it
3712 * manually for init_net */
3713 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3714 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3715 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3716 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3717 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3718 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3719 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3723 int __init ip6_route_init(void)
3729 ip6_dst_ops_template.kmem_cachep =
3730 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3731 SLAB_HWCACHE_ALIGN, NULL);
3732 if (!ip6_dst_ops_template.kmem_cachep)
3735 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3737 goto out_kmem_cache;
3739 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3741 goto out_dst_entries;
3743 ret = register_pernet_subsys(&ip6_route_net_ops);
3745 goto out_register_inetpeer;
3747 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3751 goto out_register_subsys;
3757 ret = fib6_rules_init();
3761 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3763 goto fib6_rules_init;
3766 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3767 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3768 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3769 goto out_register_late_subsys;
3771 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3773 goto out_register_late_subsys;
3775 for_each_possible_cpu(cpu) {
3776 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3778 INIT_LIST_HEAD(&ul->head);
3779 spin_lock_init(&ul->lock);
3785 out_register_late_subsys:
3786 unregister_pernet_subsys(&ip6_route_net_late_ops);
3788 fib6_rules_cleanup();
3793 out_register_subsys:
3794 unregister_pernet_subsys(&ip6_route_net_ops);
3795 out_register_inetpeer:
3796 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3798 dst_entries_destroy(&ip6_dst_blackhole_ops);
3800 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3804 void ip6_route_cleanup(void)
3806 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3807 unregister_pernet_subsys(&ip6_route_net_late_ops);
3808 fib6_rules_cleanup();
3811 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3812 unregister_pernet_subsys(&ip6_route_net_ops);
3813 dst_entries_destroy(&ip6_dst_blackhole_ops);
3814 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);