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 read_lock_bh(&table->tb6_lock);
851 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
854 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
855 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
856 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
857 if (rt == net->ipv6.ip6_null_entry) {
858 fn = fib6_backtrack(fn, &fl6->saddr);
862 dst_use(&rt->dst, jiffies);
863 read_unlock_bh(&table->tb6_lock);
868 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
871 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
873 EXPORT_SYMBOL_GPL(ip6_route_lookup);
875 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
876 const struct in6_addr *saddr, int oif, int strict)
878 struct flowi6 fl6 = {
882 struct dst_entry *dst;
883 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
886 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
887 flags |= RT6_LOOKUP_F_HAS_SADDR;
890 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
892 return (struct rt6_info *) dst;
898 EXPORT_SYMBOL(rt6_lookup);
900 /* ip6_ins_rt is called with FREE table->tb6_lock.
901 It takes new route entry, the addition fails by any reason the
902 route is freed. In any case, if caller does not hold it, it may
906 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
907 struct mx6_config *mxc)
910 struct fib6_table *table;
912 table = rt->rt6i_table;
913 write_lock_bh(&table->tb6_lock);
914 err = fib6_add(&table->tb6_root, rt, info, mxc);
915 write_unlock_bh(&table->tb6_lock);
920 int ip6_ins_rt(struct rt6_info *rt)
922 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
923 struct mx6_config mxc = { .mx = NULL, };
925 return __ip6_ins_rt(rt, &info, &mxc);
928 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
929 const struct in6_addr *daddr,
930 const struct in6_addr *saddr)
938 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
939 ort = (struct rt6_info *)ort->dst.from;
941 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
946 ip6_rt_copy_init(rt, ort);
947 rt->rt6i_flags |= RTF_CACHE;
949 rt->dst.flags |= DST_HOST;
950 rt->rt6i_dst.addr = *daddr;
951 rt->rt6i_dst.plen = 128;
953 if (!rt6_is_gw_or_nonexthop(ort)) {
954 if (ort->rt6i_dst.plen != 128 &&
955 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
956 rt->rt6i_flags |= RTF_ANYCAST;
957 #ifdef CONFIG_IPV6_SUBTREES
958 if (rt->rt6i_src.plen && saddr) {
959 rt->rt6i_src.addr = *saddr;
960 rt->rt6i_src.plen = 128;
968 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
970 struct rt6_info *pcpu_rt;
972 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
973 rt->dst.dev, rt->dst.flags);
977 ip6_rt_copy_init(pcpu_rt, rt);
978 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
979 pcpu_rt->rt6i_flags |= RTF_PCPU;
983 /* It should be called with read_lock_bh(&tb6_lock) acquired */
984 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
986 struct rt6_info *pcpu_rt, **p;
988 p = this_cpu_ptr(rt->rt6i_pcpu);
992 dst_hold(&pcpu_rt->dst);
993 rt6_dst_from_metrics_check(pcpu_rt);
998 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1000 struct fib6_table *table = rt->rt6i_table;
1001 struct rt6_info *pcpu_rt, *prev, **p;
1003 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1005 struct net *net = dev_net(rt->dst.dev);
1007 dst_hold(&net->ipv6.ip6_null_entry->dst);
1008 return net->ipv6.ip6_null_entry;
1011 read_lock_bh(&table->tb6_lock);
1012 if (rt->rt6i_pcpu) {
1013 p = this_cpu_ptr(rt->rt6i_pcpu);
1014 prev = cmpxchg(p, NULL, pcpu_rt);
1016 /* If someone did it before us, return prev instead */
1017 dst_destroy(&pcpu_rt->dst);
1021 /* rt has been removed from the fib6 tree
1022 * before we have a chance to acquire the read_lock.
1023 * In this case, don't brother to create a pcpu rt
1024 * since rt is going away anyway. The next
1025 * dst_check() will trigger a re-lookup.
1027 dst_destroy(&pcpu_rt->dst);
1030 dst_hold(&pcpu_rt->dst);
1031 rt6_dst_from_metrics_check(pcpu_rt);
1032 read_unlock_bh(&table->tb6_lock);
1036 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1037 struct flowi6 *fl6, int flags)
1039 struct fib6_node *fn, *saved_fn;
1040 struct rt6_info *rt;
1043 strict |= flags & RT6_LOOKUP_F_IFACE;
1044 if (net->ipv6.devconf_all->forwarding == 0)
1045 strict |= RT6_LOOKUP_F_REACHABLE;
1047 read_lock_bh(&table->tb6_lock);
1049 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1052 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1056 rt = rt6_select(fn, oif, strict);
1057 if (rt->rt6i_nsiblings)
1058 rt = rt6_multipath_select(rt, fl6, oif, strict);
1059 if (rt == net->ipv6.ip6_null_entry) {
1060 fn = fib6_backtrack(fn, &fl6->saddr);
1062 goto redo_rt6_select;
1063 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1064 /* also consider unreachable route */
1065 strict &= ~RT6_LOOKUP_F_REACHABLE;
1067 goto redo_rt6_select;
1072 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1073 dst_use(&rt->dst, jiffies);
1074 read_unlock_bh(&table->tb6_lock);
1076 rt6_dst_from_metrics_check(rt);
1078 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1079 !(rt->rt6i_flags & RTF_GATEWAY))) {
1080 /* Create a RTF_CACHE clone which will not be
1081 * owned by the fib6 tree. It is for the special case where
1082 * the daddr in the skb during the neighbor look-up is different
1083 * from the fl6->daddr used to look-up route here.
1086 struct rt6_info *uncached_rt;
1088 dst_use(&rt->dst, jiffies);
1089 read_unlock_bh(&table->tb6_lock);
1091 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1092 dst_release(&rt->dst);
1095 rt6_uncached_list_add(uncached_rt);
1097 uncached_rt = net->ipv6.ip6_null_entry;
1099 dst_hold(&uncached_rt->dst);
1103 /* Get a percpu copy */
1105 struct rt6_info *pcpu_rt;
1107 rt->dst.lastuse = jiffies;
1109 pcpu_rt = rt6_get_pcpu_route(rt);
1112 read_unlock_bh(&table->tb6_lock);
1114 /* We have to do the read_unlock first
1115 * because rt6_make_pcpu_route() may trigger
1116 * ip6_dst_gc() which will take the write_lock.
1119 read_unlock_bh(&table->tb6_lock);
1120 pcpu_rt = rt6_make_pcpu_route(rt);
1121 dst_release(&rt->dst);
1129 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1130 struct flowi6 *fl6, int flags)
1132 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1135 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1136 struct net_device *dev,
1137 struct flowi6 *fl6, int flags)
1139 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1140 flags |= RT6_LOOKUP_F_IFACE;
1142 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1145 void ip6_route_input(struct sk_buff *skb)
1147 const struct ipv6hdr *iph = ipv6_hdr(skb);
1148 struct net *net = dev_net(skb->dev);
1149 int flags = RT6_LOOKUP_F_HAS_SADDR;
1150 struct ip_tunnel_info *tun_info;
1151 struct flowi6 fl6 = {
1152 .flowi6_iif = l3mdev_fib_oif(skb->dev),
1153 .daddr = iph->daddr,
1154 .saddr = iph->saddr,
1155 .flowlabel = ip6_flowinfo(iph),
1156 .flowi6_mark = skb->mark,
1157 .flowi6_proto = iph->nexthdr,
1160 tun_info = skb_tunnel_info(skb);
1161 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1162 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1164 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1167 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1168 struct flowi6 *fl6, int flags)
1170 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1173 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1174 struct flowi6 *fl6, int flags)
1176 struct dst_entry *dst;
1179 dst = l3mdev_rt6_dst_by_oif(net, fl6);
1183 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1185 any_src = ipv6_addr_any(&fl6->saddr);
1186 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1187 (fl6->flowi6_oif && any_src))
1188 flags |= RT6_LOOKUP_F_IFACE;
1191 flags |= RT6_LOOKUP_F_HAS_SADDR;
1193 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1195 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1197 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1199 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1201 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1202 struct dst_entry *new = NULL;
1204 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1210 new->input = dst_discard;
1211 new->output = dst_discard_out;
1213 dst_copy_metrics(new, &ort->dst);
1214 rt->rt6i_idev = ort->rt6i_idev;
1216 in6_dev_hold(rt->rt6i_idev);
1218 rt->rt6i_gateway = ort->rt6i_gateway;
1219 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1220 rt->rt6i_metric = 0;
1222 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1223 #ifdef CONFIG_IPV6_SUBTREES
1224 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1230 dst_release(dst_orig);
1231 return new ? new : ERR_PTR(-ENOMEM);
1235 * Destination cache support functions
1238 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1241 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1242 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1245 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1247 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1250 if (rt6_check_expired(rt))
1256 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1258 if (!__rt6_check_expired(rt) &&
1259 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1260 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1266 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1268 struct rt6_info *rt;
1270 rt = (struct rt6_info *) dst;
1272 /* All IPV6 dsts are created with ->obsolete set to the value
1273 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1274 * into this function always.
1277 rt6_dst_from_metrics_check(rt);
1279 if (rt->rt6i_flags & RTF_PCPU ||
1280 (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1281 return rt6_dst_from_check(rt, cookie);
1283 return rt6_check(rt, cookie);
1286 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1288 struct rt6_info *rt = (struct rt6_info *) dst;
1291 if (rt->rt6i_flags & RTF_CACHE) {
1292 if (rt6_check_expired(rt)) {
1304 static void ip6_link_failure(struct sk_buff *skb)
1306 struct rt6_info *rt;
1308 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1310 rt = (struct rt6_info *) skb_dst(skb);
1312 if (rt->rt6i_flags & RTF_CACHE) {
1315 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1316 rt->rt6i_node->fn_sernum = -1;
1321 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1323 struct net *net = dev_net(rt->dst.dev);
1325 rt->rt6i_flags |= RTF_MODIFIED;
1326 rt->rt6i_pmtu = mtu;
1327 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1330 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1332 return !(rt->rt6i_flags & RTF_CACHE) &&
1333 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1336 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1337 const struct ipv6hdr *iph, u32 mtu)
1339 struct rt6_info *rt6 = (struct rt6_info *)dst;
1341 if (rt6->rt6i_flags & RTF_LOCAL)
1345 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1346 if (mtu >= dst_mtu(dst))
1349 if (!rt6_cache_allowed_for_pmtu(rt6)) {
1350 rt6_do_update_pmtu(rt6, mtu);
1352 const struct in6_addr *daddr, *saddr;
1353 struct rt6_info *nrt6;
1356 daddr = &iph->daddr;
1357 saddr = &iph->saddr;
1359 daddr = &sk->sk_v6_daddr;
1360 saddr = &inet6_sk(sk)->saddr;
1364 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1366 rt6_do_update_pmtu(nrt6, mtu);
1368 /* ip6_ins_rt(nrt6) will bump the
1369 * rt6->rt6i_node->fn_sernum
1370 * which will fail the next rt6_check() and
1371 * invalidate the sk->sk_dst_cache.
1378 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1379 struct sk_buff *skb, u32 mtu)
1381 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1384 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1385 int oif, u32 mark, kuid_t uid)
1387 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1388 struct dst_entry *dst;
1391 memset(&fl6, 0, sizeof(fl6));
1392 fl6.flowi6_oif = oif;
1393 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1394 fl6.daddr = iph->daddr;
1395 fl6.saddr = iph->saddr;
1396 fl6.flowlabel = ip6_flowinfo(iph);
1397 fl6.flowi6_uid = uid;
1399 dst = ip6_route_output(net, NULL, &fl6);
1401 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1404 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1406 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1408 ip6_update_pmtu(skb, sock_net(sk), mtu,
1409 sk->sk_bound_dev_if, sk->sk_mark, sk->sk_uid);
1411 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1413 /* Handle redirects */
1414 struct ip6rd_flowi {
1416 struct in6_addr gateway;
1419 static struct rt6_info *__ip6_route_redirect(struct net *net,
1420 struct fib6_table *table,
1424 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1425 struct rt6_info *rt;
1426 struct fib6_node *fn;
1428 /* Get the "current" route for this destination and
1429 * check if the redirect has come from approriate router.
1431 * RFC 4861 specifies that redirects should only be
1432 * accepted if they come from the nexthop to the target.
1433 * Due to the way the routes are chosen, this notion
1434 * is a bit fuzzy and one might need to check all possible
1438 read_lock_bh(&table->tb6_lock);
1439 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1441 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1442 if (rt6_check_expired(rt))
1446 if (!(rt->rt6i_flags & RTF_GATEWAY))
1448 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1450 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1456 rt = net->ipv6.ip6_null_entry;
1457 else if (rt->dst.error) {
1458 rt = net->ipv6.ip6_null_entry;
1462 if (rt == net->ipv6.ip6_null_entry) {
1463 fn = fib6_backtrack(fn, &fl6->saddr);
1471 read_unlock_bh(&table->tb6_lock);
1476 static struct dst_entry *ip6_route_redirect(struct net *net,
1477 const struct flowi6 *fl6,
1478 const struct in6_addr *gateway)
1480 int flags = RT6_LOOKUP_F_HAS_SADDR;
1481 struct ip6rd_flowi rdfl;
1484 rdfl.gateway = *gateway;
1486 return fib6_rule_lookup(net, &rdfl.fl6,
1487 flags, __ip6_route_redirect);
1490 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
1493 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1494 struct dst_entry *dst;
1497 memset(&fl6, 0, sizeof(fl6));
1498 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1499 fl6.flowi6_oif = oif;
1500 fl6.flowi6_mark = mark;
1501 fl6.daddr = iph->daddr;
1502 fl6.saddr = iph->saddr;
1503 fl6.flowlabel = ip6_flowinfo(iph);
1504 fl6.flowi6_uid = uid;
1506 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1507 rt6_do_redirect(dst, NULL, skb);
1510 EXPORT_SYMBOL_GPL(ip6_redirect);
1512 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1515 const struct ipv6hdr *iph = ipv6_hdr(skb);
1516 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1517 struct dst_entry *dst;
1520 memset(&fl6, 0, sizeof(fl6));
1521 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1522 fl6.flowi6_oif = oif;
1523 fl6.flowi6_mark = mark;
1524 fl6.daddr = msg->dest;
1525 fl6.saddr = iph->daddr;
1526 fl6.flowi6_uid = sock_net_uid(net, NULL);
1528 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1529 rt6_do_redirect(dst, NULL, skb);
1533 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1535 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
1538 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1540 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1542 struct net_device *dev = dst->dev;
1543 unsigned int mtu = dst_mtu(dst);
1544 struct net *net = dev_net(dev);
1546 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1548 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1549 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1552 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1553 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1554 * IPV6_MAXPLEN is also valid and means: "any MSS,
1555 * rely only on pmtu discovery"
1557 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1562 static unsigned int ip6_mtu(const struct dst_entry *dst)
1564 const struct rt6_info *rt = (const struct rt6_info *)dst;
1565 unsigned int mtu = rt->rt6i_pmtu;
1566 struct inet6_dev *idev;
1571 mtu = dst_metric_raw(dst, RTAX_MTU);
1578 idev = __in6_dev_get(dst->dev);
1580 mtu = idev->cnf.mtu6;
1584 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1587 static struct dst_entry *icmp6_dst_gc_list;
1588 static DEFINE_SPINLOCK(icmp6_dst_lock);
1590 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1593 struct dst_entry *dst;
1594 struct rt6_info *rt;
1595 struct inet6_dev *idev = in6_dev_get(dev);
1596 struct net *net = dev_net(dev);
1598 if (unlikely(!idev))
1599 return ERR_PTR(-ENODEV);
1601 rt = ip6_dst_alloc(net, dev, 0);
1602 if (unlikely(!rt)) {
1604 dst = ERR_PTR(-ENOMEM);
1608 rt->dst.flags |= DST_HOST;
1609 rt->dst.output = ip6_output;
1610 atomic_set(&rt->dst.__refcnt, 1);
1611 rt->rt6i_gateway = fl6->daddr;
1612 rt->rt6i_dst.addr = fl6->daddr;
1613 rt->rt6i_dst.plen = 128;
1614 rt->rt6i_idev = idev;
1615 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1617 spin_lock_bh(&icmp6_dst_lock);
1618 rt->dst.next = icmp6_dst_gc_list;
1619 icmp6_dst_gc_list = &rt->dst;
1620 spin_unlock_bh(&icmp6_dst_lock);
1622 fib6_force_start_gc(net);
1624 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1630 int icmp6_dst_gc(void)
1632 struct dst_entry *dst, **pprev;
1635 spin_lock_bh(&icmp6_dst_lock);
1636 pprev = &icmp6_dst_gc_list;
1638 while ((dst = *pprev) != NULL) {
1639 if (!atomic_read(&dst->__refcnt)) {
1648 spin_unlock_bh(&icmp6_dst_lock);
1653 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1656 struct dst_entry *dst, **pprev;
1658 spin_lock_bh(&icmp6_dst_lock);
1659 pprev = &icmp6_dst_gc_list;
1660 while ((dst = *pprev) != NULL) {
1661 struct rt6_info *rt = (struct rt6_info *) dst;
1662 if (func(rt, arg)) {
1669 spin_unlock_bh(&icmp6_dst_lock);
1672 static int ip6_dst_gc(struct dst_ops *ops)
1674 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1675 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1676 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1677 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1678 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1679 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1682 entries = dst_entries_get_fast(ops);
1683 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1684 entries <= rt_max_size)
1687 net->ipv6.ip6_rt_gc_expire++;
1688 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1689 entries = dst_entries_get_slow(ops);
1690 if (entries < ops->gc_thresh)
1691 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1693 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1694 return entries > rt_max_size;
1697 static int ip6_convert_metrics(struct mx6_config *mxc,
1698 const struct fib6_config *cfg)
1700 bool ecn_ca = false;
1708 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1712 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1713 int type = nla_type(nla);
1718 if (unlikely(type > RTAX_MAX))
1721 if (type == RTAX_CC_ALGO) {
1722 char tmp[TCP_CA_NAME_MAX];
1724 nla_strlcpy(tmp, nla, sizeof(tmp));
1725 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1726 if (val == TCP_CA_UNSPEC)
1729 val = nla_get_u32(nla);
1731 if (type == RTAX_HOPLIMIT && val > 255)
1733 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1737 __set_bit(type - 1, mxc->mx_valid);
1741 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1742 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1752 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1754 struct net *net = cfg->fc_nlinfo.nl_net;
1755 struct rt6_info *rt = NULL;
1756 struct net_device *dev = NULL;
1757 struct inet6_dev *idev = NULL;
1758 struct fib6_table *table;
1762 /* RTF_PCPU is an internal flag; can not be set by userspace */
1763 if (cfg->fc_flags & RTF_PCPU)
1766 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1768 #ifndef CONFIG_IPV6_SUBTREES
1769 if (cfg->fc_src_len)
1772 if (cfg->fc_ifindex) {
1774 dev = dev_get_by_index(net, cfg->fc_ifindex);
1777 idev = in6_dev_get(dev);
1782 if (cfg->fc_metric == 0)
1783 cfg->fc_metric = IP6_RT_PRIO_USER;
1786 if (cfg->fc_nlinfo.nlh &&
1787 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1788 table = fib6_get_table(net, cfg->fc_table);
1790 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1791 table = fib6_new_table(net, cfg->fc_table);
1794 table = fib6_new_table(net, cfg->fc_table);
1800 rt = ip6_dst_alloc(net, NULL,
1801 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1808 if (cfg->fc_flags & RTF_EXPIRES)
1809 rt6_set_expires(rt, jiffies +
1810 clock_t_to_jiffies(cfg->fc_expires));
1812 rt6_clean_expires(rt);
1814 if (cfg->fc_protocol == RTPROT_UNSPEC)
1815 cfg->fc_protocol = RTPROT_BOOT;
1816 rt->rt6i_protocol = cfg->fc_protocol;
1818 addr_type = ipv6_addr_type(&cfg->fc_dst);
1820 if (addr_type & IPV6_ADDR_MULTICAST)
1821 rt->dst.input = ip6_mc_input;
1822 else if (cfg->fc_flags & RTF_LOCAL)
1823 rt->dst.input = ip6_input;
1825 rt->dst.input = ip6_forward;
1827 rt->dst.output = ip6_output;
1829 if (cfg->fc_encap) {
1830 struct lwtunnel_state *lwtstate;
1832 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1833 cfg->fc_encap, AF_INET6, cfg,
1837 rt->dst.lwtstate = lwtstate_get(lwtstate);
1838 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1839 rt->dst.lwtstate->orig_output = rt->dst.output;
1840 rt->dst.output = lwtunnel_output;
1842 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1843 rt->dst.lwtstate->orig_input = rt->dst.input;
1844 rt->dst.input = lwtunnel_input;
1848 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1849 rt->rt6i_dst.plen = cfg->fc_dst_len;
1850 if (rt->rt6i_dst.plen == 128)
1851 rt->dst.flags |= DST_HOST;
1853 #ifdef CONFIG_IPV6_SUBTREES
1854 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1855 rt->rt6i_src.plen = cfg->fc_src_len;
1858 rt->rt6i_metric = cfg->fc_metric;
1860 /* We cannot add true routes via loopback here,
1861 they would result in kernel looping; promote them to reject routes
1863 if ((cfg->fc_flags & RTF_REJECT) ||
1864 (dev && (dev->flags & IFF_LOOPBACK) &&
1865 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1866 !(cfg->fc_flags & RTF_LOCAL))) {
1867 /* hold loopback dev/idev if we haven't done so. */
1868 if (dev != net->loopback_dev) {
1873 dev = net->loopback_dev;
1875 idev = in6_dev_get(dev);
1881 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1882 switch (cfg->fc_type) {
1884 rt->dst.error = -EINVAL;
1885 rt->dst.output = dst_discard_out;
1886 rt->dst.input = dst_discard;
1889 rt->dst.error = -EACCES;
1890 rt->dst.output = ip6_pkt_prohibit_out;
1891 rt->dst.input = ip6_pkt_prohibit;
1894 case RTN_UNREACHABLE:
1896 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1897 : (cfg->fc_type == RTN_UNREACHABLE)
1898 ? -EHOSTUNREACH : -ENETUNREACH;
1899 rt->dst.output = ip6_pkt_discard_out;
1900 rt->dst.input = ip6_pkt_discard;
1906 if (cfg->fc_flags & RTF_GATEWAY) {
1907 const struct in6_addr *gw_addr;
1910 gw_addr = &cfg->fc_gateway;
1911 gwa_type = ipv6_addr_type(gw_addr);
1913 /* if gw_addr is local we will fail to detect this in case
1914 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1915 * will return already-added prefix route via interface that
1916 * prefix route was assigned to, which might be non-loopback.
1919 if (ipv6_chk_addr_and_flags(net, gw_addr,
1920 gwa_type & IPV6_ADDR_LINKLOCAL ?
1924 rt->rt6i_gateway = *gw_addr;
1926 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1927 struct rt6_info *grt;
1929 /* IPv6 strictly inhibits using not link-local
1930 addresses as nexthop address.
1931 Otherwise, router will not able to send redirects.
1932 It is very good, but in some (rare!) circumstances
1933 (SIT, PtP, NBMA NOARP links) it is handy to allow
1934 some exceptions. --ANK
1936 if (!(gwa_type & IPV6_ADDR_UNICAST))
1939 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1941 err = -EHOSTUNREACH;
1945 if (dev != grt->dst.dev) {
1951 idev = grt->rt6i_idev;
1953 in6_dev_hold(grt->rt6i_idev);
1955 if (!(grt->rt6i_flags & RTF_GATEWAY))
1963 if (!dev || (dev->flags & IFF_LOOPBACK))
1971 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1972 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1976 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1977 rt->rt6i_prefsrc.plen = 128;
1979 rt->rt6i_prefsrc.plen = 0;
1981 rt->rt6i_flags = cfg->fc_flags;
1985 rt->rt6i_idev = idev;
1986 rt->rt6i_table = table;
1988 cfg->fc_nlinfo.nl_net = dev_net(dev);
1999 return ERR_PTR(err);
2002 int ip6_route_add(struct fib6_config *cfg)
2004 struct mx6_config mxc = { .mx = NULL, };
2005 struct rt6_info *rt;
2008 rt = ip6_route_info_create(cfg);
2015 err = ip6_convert_metrics(&mxc, cfg);
2019 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2031 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2034 struct fib6_table *table;
2035 struct net *net = dev_net(rt->dst.dev);
2037 if (rt == net->ipv6.ip6_null_entry ||
2038 rt->dst.flags & DST_NOCACHE) {
2043 table = rt->rt6i_table;
2044 write_lock_bh(&table->tb6_lock);
2045 err = fib6_del(rt, info);
2046 write_unlock_bh(&table->tb6_lock);
2053 int ip6_del_rt(struct rt6_info *rt)
2055 struct nl_info info = {
2056 .nl_net = dev_net(rt->dst.dev),
2058 return __ip6_del_rt(rt, &info);
2061 static int ip6_route_del(struct fib6_config *cfg)
2063 struct fib6_table *table;
2064 struct fib6_node *fn;
2065 struct rt6_info *rt;
2068 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2072 read_lock_bh(&table->tb6_lock);
2074 fn = fib6_locate(&table->tb6_root,
2075 &cfg->fc_dst, cfg->fc_dst_len,
2076 &cfg->fc_src, cfg->fc_src_len);
2079 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2080 if ((rt->rt6i_flags & RTF_CACHE) &&
2081 !(cfg->fc_flags & RTF_CACHE))
2083 if (cfg->fc_ifindex &&
2085 rt->dst.dev->ifindex != cfg->fc_ifindex))
2087 if (cfg->fc_flags & RTF_GATEWAY &&
2088 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2090 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2092 if (cfg->fc_protocol && cfg->fc_protocol != rt->rt6i_protocol)
2095 read_unlock_bh(&table->tb6_lock);
2097 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2100 read_unlock_bh(&table->tb6_lock);
2105 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2107 struct netevent_redirect netevent;
2108 struct rt6_info *rt, *nrt = NULL;
2109 struct ndisc_options ndopts;
2110 struct inet6_dev *in6_dev;
2111 struct neighbour *neigh;
2113 int optlen, on_link;
2116 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2117 optlen -= sizeof(*msg);
2120 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2124 msg = (struct rd_msg *)icmp6_hdr(skb);
2126 if (ipv6_addr_is_multicast(&msg->dest)) {
2127 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2132 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2134 } else if (ipv6_addr_type(&msg->target) !=
2135 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2136 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2140 in6_dev = __in6_dev_get(skb->dev);
2143 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2147 * The IP source address of the Redirect MUST be the same as the current
2148 * first-hop router for the specified ICMP Destination Address.
2151 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2152 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2157 if (ndopts.nd_opts_tgt_lladdr) {
2158 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2161 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2166 rt = (struct rt6_info *) dst;
2167 if (rt->rt6i_flags & RTF_REJECT) {
2168 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2172 /* Redirect received -> path was valid.
2173 * Look, redirects are sent only in response to data packets,
2174 * so that this nexthop apparently is reachable. --ANK
2176 dst_confirm(&rt->dst);
2178 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2183 * We have finally decided to accept it.
2186 neigh_update(neigh, lladdr, NUD_STALE,
2187 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2188 NEIGH_UPDATE_F_OVERRIDE|
2189 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2190 NEIGH_UPDATE_F_ISROUTER))
2193 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2197 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2199 nrt->rt6i_flags &= ~RTF_GATEWAY;
2201 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2203 if (ip6_ins_rt(nrt))
2206 netevent.old = &rt->dst;
2207 netevent.new = &nrt->dst;
2208 netevent.daddr = &msg->dest;
2209 netevent.neigh = neigh;
2210 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2212 if (rt->rt6i_flags & RTF_CACHE) {
2213 rt = (struct rt6_info *) dst_clone(&rt->dst);
2218 neigh_release(neigh);
2222 * Misc support functions
2225 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2227 BUG_ON(from->dst.from);
2229 rt->rt6i_flags &= ~RTF_EXPIRES;
2230 dst_hold(&from->dst);
2231 rt->dst.from = &from->dst;
2232 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2235 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2237 rt->dst.input = ort->dst.input;
2238 rt->dst.output = ort->dst.output;
2239 rt->rt6i_dst = ort->rt6i_dst;
2240 rt->dst.error = ort->dst.error;
2241 rt->rt6i_idev = ort->rt6i_idev;
2243 in6_dev_hold(rt->rt6i_idev);
2244 rt->dst.lastuse = jiffies;
2245 rt->rt6i_gateway = ort->rt6i_gateway;
2246 rt->rt6i_flags = ort->rt6i_flags;
2247 rt6_set_from(rt, ort);
2248 rt->rt6i_metric = ort->rt6i_metric;
2249 #ifdef CONFIG_IPV6_SUBTREES
2250 rt->rt6i_src = ort->rt6i_src;
2252 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2253 rt->rt6i_table = ort->rt6i_table;
2254 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2257 #ifdef CONFIG_IPV6_ROUTE_INFO
2258 static struct rt6_info *rt6_get_route_info(struct net_device *dev,
2259 const struct in6_addr *prefix, int prefixlen,
2260 const struct in6_addr *gwaddr)
2262 struct fib6_node *fn;
2263 struct rt6_info *rt = NULL;
2264 struct fib6_table *table;
2266 table = fib6_get_table(dev_net(dev),
2267 addrconf_rt_table(dev, RT6_TABLE_INFO));
2271 read_lock_bh(&table->tb6_lock);
2272 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2276 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2277 if (rt->dst.dev->ifindex != dev->ifindex)
2279 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2281 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2287 read_unlock_bh(&table->tb6_lock);
2291 static struct rt6_info *rt6_add_route_info(struct net_device *dev,
2292 const struct in6_addr *prefix, int prefixlen,
2293 const struct in6_addr *gwaddr, unsigned int pref)
2295 struct fib6_config cfg = {
2296 .fc_metric = IP6_RT_PRIO_USER,
2297 .fc_ifindex = dev->ifindex,
2298 .fc_dst_len = prefixlen,
2299 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2300 RTF_UP | RTF_PREF(pref),
2301 .fc_nlinfo.portid = 0,
2302 .fc_nlinfo.nlh = NULL,
2303 .fc_nlinfo.nl_net = dev_net(dev),
2306 cfg.fc_table = l3mdev_fib_table_by_index(dev_net(dev), dev->ifindex) ? : addrconf_rt_table(dev, RT6_TABLE_INFO);
2307 cfg.fc_dst = *prefix;
2308 cfg.fc_gateway = *gwaddr;
2310 /* We should treat it as a default route if prefix length is 0. */
2312 cfg.fc_flags |= RTF_DEFAULT;
2314 ip6_route_add(&cfg);
2316 return rt6_get_route_info(dev, prefix, prefixlen, gwaddr);
2320 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2322 struct rt6_info *rt;
2323 struct fib6_table *table;
2325 table = fib6_get_table(dev_net(dev),
2326 addrconf_rt_table(dev, RT6_TABLE_MAIN));
2330 read_lock_bh(&table->tb6_lock);
2331 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2332 if (dev == rt->dst.dev &&
2333 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2334 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2339 read_unlock_bh(&table->tb6_lock);
2343 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2344 struct net_device *dev,
2347 struct fib6_config cfg = {
2348 .fc_table = l3mdev_fib_table(dev) ? : addrconf_rt_table(dev, RT6_TABLE_DFLT),
2349 .fc_metric = IP6_RT_PRIO_USER,
2350 .fc_ifindex = dev->ifindex,
2351 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2352 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2353 .fc_nlinfo.portid = 0,
2354 .fc_nlinfo.nlh = NULL,
2355 .fc_nlinfo.nl_net = dev_net(dev),
2358 cfg.fc_gateway = *gwaddr;
2360 ip6_route_add(&cfg);
2362 return rt6_get_dflt_router(gwaddr, dev);
2366 int rt6_addrconf_purge(struct rt6_info *rt, void *arg) {
2367 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2368 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2))
2373 void rt6_purge_dflt_routers(struct net *net)
2375 fib6_clean_all(net, rt6_addrconf_purge, NULL);
2378 static void rtmsg_to_fib6_config(struct net *net,
2379 struct in6_rtmsg *rtmsg,
2380 struct fib6_config *cfg)
2382 memset(cfg, 0, sizeof(*cfg));
2384 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2386 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2387 cfg->fc_metric = rtmsg->rtmsg_metric;
2388 cfg->fc_expires = rtmsg->rtmsg_info;
2389 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2390 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2391 cfg->fc_flags = rtmsg->rtmsg_flags;
2393 cfg->fc_nlinfo.nl_net = net;
2395 cfg->fc_dst = rtmsg->rtmsg_dst;
2396 cfg->fc_src = rtmsg->rtmsg_src;
2397 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2400 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2402 struct fib6_config cfg;
2403 struct in6_rtmsg rtmsg;
2407 case SIOCADDRT: /* Add a route */
2408 case SIOCDELRT: /* Delete a route */
2409 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2411 err = copy_from_user(&rtmsg, arg,
2412 sizeof(struct in6_rtmsg));
2416 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2421 err = ip6_route_add(&cfg);
2424 err = ip6_route_del(&cfg);
2438 * Drop the packet on the floor
2441 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2444 struct dst_entry *dst = skb_dst(skb);
2445 switch (ipstats_mib_noroutes) {
2446 case IPSTATS_MIB_INNOROUTES:
2447 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2448 if (type == IPV6_ADDR_ANY) {
2449 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2450 IPSTATS_MIB_INADDRERRORS);
2454 case IPSTATS_MIB_OUTNOROUTES:
2455 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2456 ipstats_mib_noroutes);
2459 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2464 static int ip6_pkt_discard(struct sk_buff *skb)
2466 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2469 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2471 skb->dev = skb_dst(skb)->dev;
2472 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2475 static int ip6_pkt_prohibit(struct sk_buff *skb)
2477 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2480 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2482 skb->dev = skb_dst(skb)->dev;
2483 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2487 * Allocate a dst for local (unicast / anycast) address.
2490 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2491 const struct in6_addr *addr,
2495 struct net *net = dev_net(idev->dev);
2496 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2499 return ERR_PTR(-ENOMEM);
2503 rt->dst.flags |= DST_HOST;
2504 rt->dst.input = ip6_input;
2505 rt->dst.output = ip6_output;
2506 rt->rt6i_idev = idev;
2508 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2510 rt->rt6i_flags |= RTF_ANYCAST;
2512 rt->rt6i_flags |= RTF_LOCAL;
2514 rt->rt6i_gateway = *addr;
2515 rt->rt6i_dst.addr = *addr;
2516 rt->rt6i_dst.plen = 128;
2517 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2518 rt->rt6i_table = fib6_get_table(net, tb_id);
2519 rt->dst.flags |= DST_NOCACHE;
2521 atomic_set(&rt->dst.__refcnt, 1);
2526 int ip6_route_get_saddr(struct net *net,
2527 struct rt6_info *rt,
2528 const struct in6_addr *daddr,
2530 struct in6_addr *saddr)
2532 struct inet6_dev *idev =
2533 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2535 if (rt && rt->rt6i_prefsrc.plen)
2536 *saddr = rt->rt6i_prefsrc.addr;
2538 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2539 daddr, prefs, saddr);
2543 /* remove deleted ip from prefsrc entries */
2544 struct arg_dev_net_ip {
2545 struct net_device *dev;
2547 struct in6_addr *addr;
2550 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2552 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2553 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2554 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2556 if (((void *)rt->dst.dev == dev || !dev) &&
2557 rt != net->ipv6.ip6_null_entry &&
2558 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2559 /* remove prefsrc entry */
2560 rt->rt6i_prefsrc.plen = 0;
2565 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2567 struct net *net = dev_net(ifp->idev->dev);
2568 struct arg_dev_net_ip adni = {
2569 .dev = ifp->idev->dev,
2573 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2576 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2577 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2579 /* Remove routers and update dst entries when gateway turn into host. */
2580 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2582 struct in6_addr *gateway = (struct in6_addr *)arg;
2584 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2585 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2586 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2592 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2594 fib6_clean_all(net, fib6_clean_tohost, gateway);
2597 struct arg_dev_net {
2598 struct net_device *dev;
2602 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2604 const struct arg_dev_net *adn = arg;
2605 const struct net_device *dev = adn->dev;
2607 if ((rt->dst.dev == dev || !dev) &&
2608 rt != adn->net->ipv6.ip6_null_entry)
2614 void rt6_ifdown(struct net *net, struct net_device *dev)
2616 struct arg_dev_net adn = {
2621 fib6_clean_all(net, fib6_ifdown, &adn);
2622 icmp6_clean_all(fib6_ifdown, &adn);
2624 rt6_uncached_list_flush_dev(net, dev);
2627 struct rt6_mtu_change_arg {
2628 struct net_device *dev;
2632 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2634 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2635 struct inet6_dev *idev;
2637 /* In IPv6 pmtu discovery is not optional,
2638 so that RTAX_MTU lock cannot disable it.
2639 We still use this lock to block changes
2640 caused by addrconf/ndisc.
2643 idev = __in6_dev_get(arg->dev);
2647 /* For administrative MTU increase, there is no way to discover
2648 IPv6 PMTU increase, so PMTU increase should be updated here.
2649 Since RFC 1981 doesn't include administrative MTU increase
2650 update PMTU increase is a MUST. (i.e. jumbo frame)
2653 If new MTU is less than route PMTU, this new MTU will be the
2654 lowest MTU in the path, update the route PMTU to reflect PMTU
2655 decreases; if new MTU is greater than route PMTU, and the
2656 old MTU is the lowest MTU in the path, update the route PMTU
2657 to reflect the increase. In this case if the other nodes' MTU
2658 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2661 if (rt->dst.dev == arg->dev &&
2662 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2663 if (rt->rt6i_flags & RTF_CACHE) {
2664 /* For RTF_CACHE with rt6i_pmtu == 0
2665 * (i.e. a redirected route),
2666 * the metrics of its rt->dst.from has already
2669 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2670 rt->rt6i_pmtu = arg->mtu;
2671 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2672 (dst_mtu(&rt->dst) < arg->mtu &&
2673 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2674 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2680 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2682 struct rt6_mtu_change_arg arg = {
2687 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2690 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2691 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2692 [RTA_OIF] = { .type = NLA_U32 },
2693 [RTA_IIF] = { .type = NLA_U32 },
2694 [RTA_PRIORITY] = { .type = NLA_U32 },
2695 [RTA_METRICS] = { .type = NLA_NESTED },
2696 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2697 [RTA_PREF] = { .type = NLA_U8 },
2698 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2699 [RTA_ENCAP] = { .type = NLA_NESTED },
2700 [RTA_UID] = { .type = NLA_U32 },
2703 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2704 struct fib6_config *cfg)
2707 struct nlattr *tb[RTA_MAX+1];
2711 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2716 rtm = nlmsg_data(nlh);
2717 memset(cfg, 0, sizeof(*cfg));
2719 cfg->fc_table = rtm->rtm_table;
2720 cfg->fc_dst_len = rtm->rtm_dst_len;
2721 cfg->fc_src_len = rtm->rtm_src_len;
2722 cfg->fc_flags = RTF_UP;
2723 cfg->fc_protocol = rtm->rtm_protocol;
2724 cfg->fc_type = rtm->rtm_type;
2726 if (rtm->rtm_type == RTN_UNREACHABLE ||
2727 rtm->rtm_type == RTN_BLACKHOLE ||
2728 rtm->rtm_type == RTN_PROHIBIT ||
2729 rtm->rtm_type == RTN_THROW)
2730 cfg->fc_flags |= RTF_REJECT;
2732 if (rtm->rtm_type == RTN_LOCAL)
2733 cfg->fc_flags |= RTF_LOCAL;
2735 if (rtm->rtm_flags & RTM_F_CLONED)
2736 cfg->fc_flags |= RTF_CACHE;
2738 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2739 cfg->fc_nlinfo.nlh = nlh;
2740 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2742 if (tb[RTA_GATEWAY]) {
2743 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2744 cfg->fc_flags |= RTF_GATEWAY;
2748 int plen = (rtm->rtm_dst_len + 7) >> 3;
2750 if (nla_len(tb[RTA_DST]) < plen)
2753 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2757 int plen = (rtm->rtm_src_len + 7) >> 3;
2759 if (nla_len(tb[RTA_SRC]) < plen)
2762 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2765 if (tb[RTA_PREFSRC])
2766 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2769 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2771 if (tb[RTA_PRIORITY])
2772 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2774 if (tb[RTA_METRICS]) {
2775 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2776 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2780 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2782 if (tb[RTA_MULTIPATH]) {
2783 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2784 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2788 pref = nla_get_u8(tb[RTA_PREF]);
2789 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2790 pref != ICMPV6_ROUTER_PREF_HIGH)
2791 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2792 cfg->fc_flags |= RTF_PREF(pref);
2796 cfg->fc_encap = tb[RTA_ENCAP];
2798 if (tb[RTA_ENCAP_TYPE])
2799 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2807 struct rt6_info *rt6_info;
2808 struct fib6_config r_cfg;
2809 struct mx6_config mxc;
2810 struct list_head next;
2813 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2817 list_for_each_entry(nh, rt6_nh_list, next) {
2818 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2819 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2820 nh->r_cfg.fc_ifindex);
2824 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2825 struct rt6_info *rt, struct fib6_config *r_cfg)
2828 struct rt6_info *rtnh;
2831 list_for_each_entry(nh, rt6_nh_list, next) {
2832 /* check if rt6_info already exists */
2833 rtnh = nh->rt6_info;
2835 if (rtnh->dst.dev == rt->dst.dev &&
2836 rtnh->rt6i_idev == rt->rt6i_idev &&
2837 ipv6_addr_equal(&rtnh->rt6i_gateway,
2842 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2846 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2851 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2852 list_add_tail(&nh->next, rt6_nh_list);
2857 static int ip6_route_multipath_add(struct fib6_config *cfg)
2859 struct fib6_config r_cfg;
2860 struct rtnexthop *rtnh;
2861 struct rt6_info *rt;
2862 struct rt6_nh *err_nh;
2863 struct rt6_nh *nh, *nh_safe;
2868 int replace = (cfg->fc_nlinfo.nlh &&
2869 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2870 LIST_HEAD(rt6_nh_list);
2872 remaining = cfg->fc_mp_len;
2873 rtnh = (struct rtnexthop *)cfg->fc_mp;
2875 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2876 * rt6_info structs per nexthop
2878 while (rtnh_ok(rtnh, remaining)) {
2879 memcpy(&r_cfg, cfg, sizeof(*cfg));
2880 if (rtnh->rtnh_ifindex)
2881 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2883 attrlen = rtnh_attrlen(rtnh);
2885 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2887 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2889 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2890 r_cfg.fc_flags |= RTF_GATEWAY;
2892 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2893 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2895 r_cfg.fc_encap_type = nla_get_u16(nla);
2898 rt = ip6_route_info_create(&r_cfg);
2905 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2911 rtnh = rtnh_next(rtnh, &remaining);
2915 list_for_each_entry(nh, &rt6_nh_list, next) {
2916 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2917 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2918 nh->rt6_info = NULL;
2921 ip6_print_replace_route_err(&rt6_nh_list);
2926 /* Because each route is added like a single route we remove
2927 * these flags after the first nexthop: if there is a collision,
2928 * we have already failed to add the first nexthop:
2929 * fib6_add_rt2node() has rejected it; when replacing, old
2930 * nexthops have been replaced by first new, the rest should
2933 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2941 /* Delete routes that were already added */
2942 list_for_each_entry(nh, &rt6_nh_list, next) {
2945 ip6_route_del(&nh->r_cfg);
2949 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2951 dst_free(&nh->rt6_info->dst);
2953 list_del(&nh->next);
2960 static int ip6_route_multipath_del(struct fib6_config *cfg)
2962 struct fib6_config r_cfg;
2963 struct rtnexthop *rtnh;
2966 int err = 1, last_err = 0;
2968 remaining = cfg->fc_mp_len;
2969 rtnh = (struct rtnexthop *)cfg->fc_mp;
2971 /* Parse a Multipath Entry */
2972 while (rtnh_ok(rtnh, remaining)) {
2973 memcpy(&r_cfg, cfg, sizeof(*cfg));
2974 if (rtnh->rtnh_ifindex)
2975 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2977 attrlen = rtnh_attrlen(rtnh);
2979 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2981 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2983 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2984 r_cfg.fc_flags |= RTF_GATEWAY;
2987 err = ip6_route_del(&r_cfg);
2991 rtnh = rtnh_next(rtnh, &remaining);
2997 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2999 struct fib6_config cfg;
3002 err = rtm_to_fib6_config(skb, nlh, &cfg);
3007 return ip6_route_multipath_del(&cfg);
3009 return ip6_route_del(&cfg);
3012 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3014 struct fib6_config cfg;
3017 err = rtm_to_fib6_config(skb, nlh, &cfg);
3022 return ip6_route_multipath_add(&cfg);
3024 return ip6_route_add(&cfg);
3027 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3029 return NLMSG_ALIGN(sizeof(struct rtmsg))
3030 + nla_total_size(16) /* RTA_SRC */
3031 + nla_total_size(16) /* RTA_DST */
3032 + nla_total_size(16) /* RTA_GATEWAY */
3033 + nla_total_size(16) /* RTA_PREFSRC */
3034 + nla_total_size(4) /* RTA_TABLE */
3035 + nla_total_size(4) /* RTA_IIF */
3036 + nla_total_size(4) /* RTA_OIF */
3037 + nla_total_size(4) /* RTA_PRIORITY */
3038 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3039 + nla_total_size(sizeof(struct rta_cacheinfo))
3040 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3041 + nla_total_size(1) /* RTA_PREF */
3042 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3045 static int rt6_fill_node(struct net *net,
3046 struct sk_buff *skb, struct rt6_info *rt,
3047 struct in6_addr *dst, struct in6_addr *src,
3048 int iif, int type, u32 portid, u32 seq,
3049 int prefix, int nowait, unsigned int flags)
3051 u32 metrics[RTAX_MAX];
3053 struct nlmsghdr *nlh;
3057 if (prefix) { /* user wants prefix routes only */
3058 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3059 /* success since this is not a prefix route */
3064 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3068 rtm = nlmsg_data(nlh);
3069 rtm->rtm_family = AF_INET6;
3070 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3071 rtm->rtm_src_len = rt->rt6i_src.plen;
3074 table = rt->rt6i_table->tb6_id;
3076 table = RT6_TABLE_UNSPEC;
3077 rtm->rtm_table = table;
3078 if (nla_put_u32(skb, RTA_TABLE, table))
3079 goto nla_put_failure;
3080 if (rt->rt6i_flags & RTF_REJECT) {
3081 switch (rt->dst.error) {
3083 rtm->rtm_type = RTN_BLACKHOLE;
3086 rtm->rtm_type = RTN_PROHIBIT;
3089 rtm->rtm_type = RTN_THROW;
3092 rtm->rtm_type = RTN_UNREACHABLE;
3096 else if (rt->rt6i_flags & RTF_LOCAL)
3097 rtm->rtm_type = RTN_LOCAL;
3098 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3099 rtm->rtm_type = RTN_LOCAL;
3101 rtm->rtm_type = RTN_UNICAST;
3103 if (!netif_carrier_ok(rt->dst.dev)) {
3104 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3105 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3106 rtm->rtm_flags |= RTNH_F_DEAD;
3108 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3109 rtm->rtm_protocol = rt->rt6i_protocol;
3110 if (rt->rt6i_flags & RTF_DYNAMIC)
3111 rtm->rtm_protocol = RTPROT_REDIRECT;
3112 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3113 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3114 rtm->rtm_protocol = RTPROT_RA;
3116 rtm->rtm_protocol = RTPROT_KERNEL;
3119 if (rt->rt6i_flags & RTF_CACHE)
3120 rtm->rtm_flags |= RTM_F_CLONED;
3123 if (nla_put_in6_addr(skb, RTA_DST, dst))
3124 goto nla_put_failure;
3125 rtm->rtm_dst_len = 128;
3126 } else if (rtm->rtm_dst_len)
3127 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3128 goto nla_put_failure;
3129 #ifdef CONFIG_IPV6_SUBTREES
3131 if (nla_put_in6_addr(skb, RTA_SRC, src))
3132 goto nla_put_failure;
3133 rtm->rtm_src_len = 128;
3134 } else if (rtm->rtm_src_len &&
3135 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3136 goto nla_put_failure;
3139 #ifdef CONFIG_IPV6_MROUTE
3140 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3141 int err = ip6mr_get_route(net, skb, rtm, nowait,
3148 goto nla_put_failure;
3150 if (err == -EMSGSIZE)
3151 goto nla_put_failure;
3156 if (nla_put_u32(skb, RTA_IIF, iif))
3157 goto nla_put_failure;
3159 struct in6_addr saddr_buf;
3160 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3161 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3162 goto nla_put_failure;
3165 if (rt->rt6i_prefsrc.plen) {
3166 struct in6_addr saddr_buf;
3167 saddr_buf = rt->rt6i_prefsrc.addr;
3168 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3169 goto nla_put_failure;
3172 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3174 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3175 if (rtnetlink_put_metrics(skb, metrics) < 0)
3176 goto nla_put_failure;
3178 if (rt->rt6i_flags & RTF_GATEWAY) {
3179 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3180 goto nla_put_failure;
3184 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3185 goto nla_put_failure;
3186 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3187 goto nla_put_failure;
3189 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3191 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3192 goto nla_put_failure;
3194 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3195 goto nla_put_failure;
3197 if (lwtunnel_fill_encap(skb, rt->dst.lwtstate) < 0)
3198 goto nla_put_failure;
3200 nlmsg_end(skb, nlh);
3204 nlmsg_cancel(skb, nlh);
3208 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3210 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3213 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3214 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3215 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3219 return rt6_fill_node(arg->net,
3220 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3221 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3222 prefix, 0, NLM_F_MULTI);
3225 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3227 struct net *net = sock_net(in_skb->sk);
3228 struct nlattr *tb[RTA_MAX+1];
3229 struct rt6_info *rt;
3230 struct sk_buff *skb;
3233 int err, iif = 0, oif = 0;
3235 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3240 memset(&fl6, 0, sizeof(fl6));
3243 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3246 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3250 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3253 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3257 iif = nla_get_u32(tb[RTA_IIF]);
3260 oif = nla_get_u32(tb[RTA_OIF]);
3263 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3266 fl6.flowi6_uid = make_kuid(current_user_ns(),
3267 nla_get_u32(tb[RTA_UID]));
3269 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
3272 struct net_device *dev;
3275 dev = __dev_get_by_index(net, iif);
3281 fl6.flowi6_iif = iif;
3283 if (!ipv6_addr_any(&fl6.saddr))
3284 flags |= RT6_LOOKUP_F_HAS_SADDR;
3286 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3289 fl6.flowi6_oif = oif;
3291 if (netif_index_is_l3_master(net, oif)) {
3292 fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3293 FLOWI_FLAG_SKIP_NH_OIF;
3296 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3299 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3306 /* Reserve room for dummy headers, this skb can pass
3307 through good chunk of routing engine.
3309 skb_reset_mac_header(skb);
3310 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3312 skb_dst_set(skb, &rt->dst);
3314 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3315 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3316 nlh->nlmsg_seq, 0, 0, 0);
3322 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3327 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3328 unsigned int nlm_flags)
3330 struct sk_buff *skb;
3331 struct net *net = info->nl_net;
3336 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3338 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3342 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3343 event, info->portid, seq, 0, 0, nlm_flags);
3345 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3346 WARN_ON(err == -EMSGSIZE);
3350 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3351 info->nlh, gfp_any());
3355 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3358 static int ip6_route_dev_notify(struct notifier_block *this,
3359 unsigned long event, void *ptr)
3361 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3362 struct net *net = dev_net(dev);
3364 if (!(dev->flags & IFF_LOOPBACK))
3367 if (event == NETDEV_REGISTER) {
3368 net->ipv6.ip6_null_entry->dst.dev = dev;
3369 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3370 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3371 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3372 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3373 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3374 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3376 } else if (event == NETDEV_UNREGISTER) {
3377 in6_dev_put(net->ipv6.ip6_null_entry->rt6i_idev);
3378 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3379 in6_dev_put(net->ipv6.ip6_prohibit_entry->rt6i_idev);
3380 in6_dev_put(net->ipv6.ip6_blk_hole_entry->rt6i_idev);
3391 #ifdef CONFIG_PROC_FS
3393 static const struct file_operations ipv6_route_proc_fops = {
3394 .owner = THIS_MODULE,
3395 .open = ipv6_route_open,
3397 .llseek = seq_lseek,
3398 .release = seq_release_net,
3401 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3403 struct net *net = (struct net *)seq->private;
3404 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3405 net->ipv6.rt6_stats->fib_nodes,
3406 net->ipv6.rt6_stats->fib_route_nodes,
3407 net->ipv6.rt6_stats->fib_rt_alloc,
3408 net->ipv6.rt6_stats->fib_rt_entries,
3409 net->ipv6.rt6_stats->fib_rt_cache,
3410 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3411 net->ipv6.rt6_stats->fib_discarded_routes);
3416 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3418 return single_open_net(inode, file, rt6_stats_seq_show);
3421 static const struct file_operations rt6_stats_seq_fops = {
3422 .owner = THIS_MODULE,
3423 .open = rt6_stats_seq_open,
3425 .llseek = seq_lseek,
3426 .release = single_release_net,
3428 #endif /* CONFIG_PROC_FS */
3430 #ifdef CONFIG_SYSCTL
3433 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3434 void __user *buffer, size_t *lenp, loff_t *ppos)
3441 net = (struct net *)ctl->extra1;
3442 delay = net->ipv6.sysctl.flush_delay;
3443 proc_dointvec(ctl, write, buffer, lenp, ppos);
3444 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3448 struct ctl_table ipv6_route_table_template[] = {
3450 .procname = "flush",
3451 .data = &init_net.ipv6.sysctl.flush_delay,
3452 .maxlen = sizeof(int),
3454 .proc_handler = ipv6_sysctl_rtcache_flush
3457 .procname = "gc_thresh",
3458 .data = &ip6_dst_ops_template.gc_thresh,
3459 .maxlen = sizeof(int),
3461 .proc_handler = proc_dointvec,
3464 .procname = "max_size",
3465 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3466 .maxlen = sizeof(int),
3468 .proc_handler = proc_dointvec,
3471 .procname = "gc_min_interval",
3472 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3473 .maxlen = sizeof(int),
3475 .proc_handler = proc_dointvec_jiffies,
3478 .procname = "gc_timeout",
3479 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3480 .maxlen = sizeof(int),
3482 .proc_handler = proc_dointvec_jiffies,
3485 .procname = "gc_interval",
3486 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3487 .maxlen = sizeof(int),
3489 .proc_handler = proc_dointvec_jiffies,
3492 .procname = "gc_elasticity",
3493 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3494 .maxlen = sizeof(int),
3496 .proc_handler = proc_dointvec,
3499 .procname = "mtu_expires",
3500 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3501 .maxlen = sizeof(int),
3503 .proc_handler = proc_dointvec_jiffies,
3506 .procname = "min_adv_mss",
3507 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3508 .maxlen = sizeof(int),
3510 .proc_handler = proc_dointvec,
3513 .procname = "gc_min_interval_ms",
3514 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3515 .maxlen = sizeof(int),
3517 .proc_handler = proc_dointvec_ms_jiffies,
3522 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3524 struct ctl_table *table;
3526 table = kmemdup(ipv6_route_table_template,
3527 sizeof(ipv6_route_table_template),
3531 table[0].data = &net->ipv6.sysctl.flush_delay;
3532 table[0].extra1 = net;
3533 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3534 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3535 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3536 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3537 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3538 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3539 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3540 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3541 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3543 /* Don't export sysctls to unprivileged users */
3544 if (net->user_ns != &init_user_ns)
3545 table[0].procname = NULL;
3552 static int __net_init ip6_route_net_init(struct net *net)
3556 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3557 sizeof(net->ipv6.ip6_dst_ops));
3559 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3560 goto out_ip6_dst_ops;
3562 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3563 sizeof(*net->ipv6.ip6_null_entry),
3565 if (!net->ipv6.ip6_null_entry)
3566 goto out_ip6_dst_entries;
3567 net->ipv6.ip6_null_entry->dst.path =
3568 (struct dst_entry *)net->ipv6.ip6_null_entry;
3569 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3570 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3571 ip6_template_metrics, true);
3573 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3574 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3575 sizeof(*net->ipv6.ip6_prohibit_entry),
3577 if (!net->ipv6.ip6_prohibit_entry)
3578 goto out_ip6_null_entry;
3579 net->ipv6.ip6_prohibit_entry->dst.path =
3580 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3581 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3582 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3583 ip6_template_metrics, true);
3585 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3586 sizeof(*net->ipv6.ip6_blk_hole_entry),
3588 if (!net->ipv6.ip6_blk_hole_entry)
3589 goto out_ip6_prohibit_entry;
3590 net->ipv6.ip6_blk_hole_entry->dst.path =
3591 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3592 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3593 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3594 ip6_template_metrics, true);
3597 net->ipv6.sysctl.flush_delay = 0;
3598 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3599 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3600 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3601 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3602 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3603 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3604 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3606 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3612 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3613 out_ip6_prohibit_entry:
3614 kfree(net->ipv6.ip6_prohibit_entry);
3616 kfree(net->ipv6.ip6_null_entry);
3618 out_ip6_dst_entries:
3619 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3624 static void __net_exit ip6_route_net_exit(struct net *net)
3626 kfree(net->ipv6.ip6_null_entry);
3627 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3628 kfree(net->ipv6.ip6_prohibit_entry);
3629 kfree(net->ipv6.ip6_blk_hole_entry);
3631 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3634 static int __net_init ip6_route_net_init_late(struct net *net)
3636 #ifdef CONFIG_PROC_FS
3637 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3638 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3643 static void __net_exit ip6_route_net_exit_late(struct net *net)
3645 #ifdef CONFIG_PROC_FS
3646 remove_proc_entry("ipv6_route", net->proc_net);
3647 remove_proc_entry("rt6_stats", net->proc_net);
3651 static struct pernet_operations ip6_route_net_ops = {
3652 .init = ip6_route_net_init,
3653 .exit = ip6_route_net_exit,
3656 static int __net_init ipv6_inetpeer_init(struct net *net)
3658 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3662 inet_peer_base_init(bp);
3663 net->ipv6.peers = bp;
3667 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3669 struct inet_peer_base *bp = net->ipv6.peers;
3671 net->ipv6.peers = NULL;
3672 inetpeer_invalidate_tree(bp);
3676 static struct pernet_operations ipv6_inetpeer_ops = {
3677 .init = ipv6_inetpeer_init,
3678 .exit = ipv6_inetpeer_exit,
3681 static struct pernet_operations ip6_route_net_late_ops = {
3682 .init = ip6_route_net_init_late,
3683 .exit = ip6_route_net_exit_late,
3686 static struct notifier_block ip6_route_dev_notifier = {
3687 .notifier_call = ip6_route_dev_notify,
3688 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
3691 void __init ip6_route_init_special_entries(void)
3693 /* Registering of the loopback is done before this portion of code,
3694 * the loopback reference in rt6_info will not be taken, do it
3695 * manually for init_net */
3696 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3697 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3698 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3699 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3700 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3701 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3702 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3706 int __init ip6_route_init(void)
3712 ip6_dst_ops_template.kmem_cachep =
3713 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3714 SLAB_HWCACHE_ALIGN, NULL);
3715 if (!ip6_dst_ops_template.kmem_cachep)
3718 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3720 goto out_kmem_cache;
3722 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3724 goto out_dst_entries;
3726 ret = register_pernet_subsys(&ip6_route_net_ops);
3728 goto out_register_inetpeer;
3730 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3734 goto out_register_subsys;
3740 ret = fib6_rules_init();
3744 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3746 goto fib6_rules_init;
3749 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3750 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3751 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3752 goto out_register_late_subsys;
3754 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3756 goto out_register_late_subsys;
3758 for_each_possible_cpu(cpu) {
3759 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3761 INIT_LIST_HEAD(&ul->head);
3762 spin_lock_init(&ul->lock);
3768 out_register_late_subsys:
3769 unregister_pernet_subsys(&ip6_route_net_late_ops);
3771 fib6_rules_cleanup();
3776 out_register_subsys:
3777 unregister_pernet_subsys(&ip6_route_net_ops);
3778 out_register_inetpeer:
3779 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3781 dst_entries_destroy(&ip6_dst_blackhole_ops);
3783 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3787 void ip6_route_cleanup(void)
3789 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3790 unregister_pernet_subsys(&ip6_route_net_late_ops);
3791 fib6_rules_cleanup();
3794 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3795 unregister_pernet_subsys(&ip6_route_net_ops);
3796 dst_entries_destroy(&ip6_dst_blackhole_ops);
3797 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);