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(struct net *net, const struct sock *sk,
1176 struct dst_entry *dst;
1180 dst = l3mdev_rt6_dst_by_oif(net, fl6);
1184 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1186 any_src = ipv6_addr_any(&fl6->saddr);
1187 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1188 (fl6->flowi6_oif && any_src))
1189 flags |= RT6_LOOKUP_F_IFACE;
1192 flags |= RT6_LOOKUP_F_HAS_SADDR;
1194 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1196 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1198 EXPORT_SYMBOL(ip6_route_output);
1200 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1202 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1203 struct dst_entry *new = NULL;
1205 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1211 new->input = dst_discard;
1212 new->output = dst_discard_out;
1214 dst_copy_metrics(new, &ort->dst);
1215 rt->rt6i_idev = ort->rt6i_idev;
1217 in6_dev_hold(rt->rt6i_idev);
1219 rt->rt6i_gateway = ort->rt6i_gateway;
1220 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1221 rt->rt6i_metric = 0;
1223 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1224 #ifdef CONFIG_IPV6_SUBTREES
1225 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1231 dst_release(dst_orig);
1232 return new ? new : ERR_PTR(-ENOMEM);
1236 * Destination cache support functions
1239 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1242 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1243 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1246 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1248 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1251 if (rt6_check_expired(rt))
1257 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1259 if (!__rt6_check_expired(rt) &&
1260 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1261 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1267 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1269 struct rt6_info *rt;
1271 rt = (struct rt6_info *) dst;
1273 /* All IPV6 dsts are created with ->obsolete set to the value
1274 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1275 * into this function always.
1278 rt6_dst_from_metrics_check(rt);
1280 if (rt->rt6i_flags & RTF_PCPU ||
1281 (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1282 return rt6_dst_from_check(rt, cookie);
1284 return rt6_check(rt, cookie);
1287 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1289 struct rt6_info *rt = (struct rt6_info *) dst;
1292 if (rt->rt6i_flags & RTF_CACHE) {
1293 if (rt6_check_expired(rt)) {
1305 static void ip6_link_failure(struct sk_buff *skb)
1307 struct rt6_info *rt;
1309 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1311 rt = (struct rt6_info *) skb_dst(skb);
1313 if (rt->rt6i_flags & RTF_CACHE) {
1316 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1317 rt->rt6i_node->fn_sernum = -1;
1322 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1324 struct net *net = dev_net(rt->dst.dev);
1326 rt->rt6i_flags |= RTF_MODIFIED;
1327 rt->rt6i_pmtu = mtu;
1328 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1331 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1333 return !(rt->rt6i_flags & RTF_CACHE) &&
1334 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1337 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1338 const struct ipv6hdr *iph, u32 mtu)
1340 struct rt6_info *rt6 = (struct rt6_info *)dst;
1342 if (rt6->rt6i_flags & RTF_LOCAL)
1346 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1347 if (mtu >= dst_mtu(dst))
1350 if (!rt6_cache_allowed_for_pmtu(rt6)) {
1351 rt6_do_update_pmtu(rt6, mtu);
1353 const struct in6_addr *daddr, *saddr;
1354 struct rt6_info *nrt6;
1357 daddr = &iph->daddr;
1358 saddr = &iph->saddr;
1360 daddr = &sk->sk_v6_daddr;
1361 saddr = &inet6_sk(sk)->saddr;
1365 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1367 rt6_do_update_pmtu(nrt6, mtu);
1369 /* ip6_ins_rt(nrt6) will bump the
1370 * rt6->rt6i_node->fn_sernum
1371 * which will fail the next rt6_check() and
1372 * invalidate the sk->sk_dst_cache.
1379 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1380 struct sk_buff *skb, u32 mtu)
1382 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1385 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1388 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1389 struct dst_entry *dst;
1392 memset(&fl6, 0, sizeof(fl6));
1393 fl6.flowi6_oif = oif;
1394 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1395 fl6.daddr = iph->daddr;
1396 fl6.saddr = iph->saddr;
1397 fl6.flowlabel = ip6_flowinfo(iph);
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);
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)
1492 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1493 struct dst_entry *dst;
1496 memset(&fl6, 0, sizeof(fl6));
1497 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1498 fl6.flowi6_oif = oif;
1499 fl6.flowi6_mark = mark;
1500 fl6.daddr = iph->daddr;
1501 fl6.saddr = iph->saddr;
1502 fl6.flowlabel = ip6_flowinfo(iph);
1504 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1505 rt6_do_redirect(dst, NULL, skb);
1508 EXPORT_SYMBOL_GPL(ip6_redirect);
1510 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1513 const struct ipv6hdr *iph = ipv6_hdr(skb);
1514 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1515 struct dst_entry *dst;
1518 memset(&fl6, 0, sizeof(fl6));
1519 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1520 fl6.flowi6_oif = oif;
1521 fl6.flowi6_mark = mark;
1522 fl6.daddr = msg->dest;
1523 fl6.saddr = iph->daddr;
1525 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1526 rt6_do_redirect(dst, NULL, skb);
1530 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1532 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1534 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1536 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1538 struct net_device *dev = dst->dev;
1539 unsigned int mtu = dst_mtu(dst);
1540 struct net *net = dev_net(dev);
1542 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1544 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1545 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1548 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1549 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1550 * IPV6_MAXPLEN is also valid and means: "any MSS,
1551 * rely only on pmtu discovery"
1553 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1558 static unsigned int ip6_mtu(const struct dst_entry *dst)
1560 const struct rt6_info *rt = (const struct rt6_info *)dst;
1561 unsigned int mtu = rt->rt6i_pmtu;
1562 struct inet6_dev *idev;
1567 mtu = dst_metric_raw(dst, RTAX_MTU);
1574 idev = __in6_dev_get(dst->dev);
1576 mtu = idev->cnf.mtu6;
1580 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1583 static struct dst_entry *icmp6_dst_gc_list;
1584 static DEFINE_SPINLOCK(icmp6_dst_lock);
1586 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1589 struct dst_entry *dst;
1590 struct rt6_info *rt;
1591 struct inet6_dev *idev = in6_dev_get(dev);
1592 struct net *net = dev_net(dev);
1594 if (unlikely(!idev))
1595 return ERR_PTR(-ENODEV);
1597 rt = ip6_dst_alloc(net, dev, 0);
1598 if (unlikely(!rt)) {
1600 dst = ERR_PTR(-ENOMEM);
1604 rt->dst.flags |= DST_HOST;
1605 rt->dst.output = ip6_output;
1606 atomic_set(&rt->dst.__refcnt, 1);
1607 rt->rt6i_gateway = fl6->daddr;
1608 rt->rt6i_dst.addr = fl6->daddr;
1609 rt->rt6i_dst.plen = 128;
1610 rt->rt6i_idev = idev;
1611 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1613 spin_lock_bh(&icmp6_dst_lock);
1614 rt->dst.next = icmp6_dst_gc_list;
1615 icmp6_dst_gc_list = &rt->dst;
1616 spin_unlock_bh(&icmp6_dst_lock);
1618 fib6_force_start_gc(net);
1620 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1626 int icmp6_dst_gc(void)
1628 struct dst_entry *dst, **pprev;
1631 spin_lock_bh(&icmp6_dst_lock);
1632 pprev = &icmp6_dst_gc_list;
1634 while ((dst = *pprev) != NULL) {
1635 if (!atomic_read(&dst->__refcnt)) {
1644 spin_unlock_bh(&icmp6_dst_lock);
1649 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1652 struct dst_entry *dst, **pprev;
1654 spin_lock_bh(&icmp6_dst_lock);
1655 pprev = &icmp6_dst_gc_list;
1656 while ((dst = *pprev) != NULL) {
1657 struct rt6_info *rt = (struct rt6_info *) dst;
1658 if (func(rt, arg)) {
1665 spin_unlock_bh(&icmp6_dst_lock);
1668 static int ip6_dst_gc(struct dst_ops *ops)
1670 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1671 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1672 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1673 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1674 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1675 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1678 entries = dst_entries_get_fast(ops);
1679 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1680 entries <= rt_max_size)
1683 net->ipv6.ip6_rt_gc_expire++;
1684 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1685 entries = dst_entries_get_slow(ops);
1686 if (entries < ops->gc_thresh)
1687 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1689 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1690 return entries > rt_max_size;
1693 static int ip6_convert_metrics(struct mx6_config *mxc,
1694 const struct fib6_config *cfg)
1696 bool ecn_ca = false;
1704 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1708 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1709 int type = nla_type(nla);
1714 if (unlikely(type > RTAX_MAX))
1717 if (type == RTAX_CC_ALGO) {
1718 char tmp[TCP_CA_NAME_MAX];
1720 nla_strlcpy(tmp, nla, sizeof(tmp));
1721 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1722 if (val == TCP_CA_UNSPEC)
1725 val = nla_get_u32(nla);
1727 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1731 __set_bit(type - 1, mxc->mx_valid);
1735 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1736 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1746 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1748 struct net *net = cfg->fc_nlinfo.nl_net;
1749 struct rt6_info *rt = NULL;
1750 struct net_device *dev = NULL;
1751 struct inet6_dev *idev = NULL;
1752 struct fib6_table *table;
1756 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1758 #ifndef CONFIG_IPV6_SUBTREES
1759 if (cfg->fc_src_len)
1762 if (cfg->fc_ifindex) {
1764 dev = dev_get_by_index(net, cfg->fc_ifindex);
1767 idev = in6_dev_get(dev);
1772 if (cfg->fc_metric == 0)
1773 cfg->fc_metric = IP6_RT_PRIO_USER;
1776 if (cfg->fc_nlinfo.nlh &&
1777 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1778 table = fib6_get_table(net, cfg->fc_table);
1780 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1781 table = fib6_new_table(net, cfg->fc_table);
1784 table = fib6_new_table(net, cfg->fc_table);
1790 rt = ip6_dst_alloc(net, NULL,
1791 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1798 if (cfg->fc_flags & RTF_EXPIRES)
1799 rt6_set_expires(rt, jiffies +
1800 clock_t_to_jiffies(cfg->fc_expires));
1802 rt6_clean_expires(rt);
1804 if (cfg->fc_protocol == RTPROT_UNSPEC)
1805 cfg->fc_protocol = RTPROT_BOOT;
1806 rt->rt6i_protocol = cfg->fc_protocol;
1808 addr_type = ipv6_addr_type(&cfg->fc_dst);
1810 if (addr_type & IPV6_ADDR_MULTICAST)
1811 rt->dst.input = ip6_mc_input;
1812 else if (cfg->fc_flags & RTF_LOCAL)
1813 rt->dst.input = ip6_input;
1815 rt->dst.input = ip6_forward;
1817 rt->dst.output = ip6_output;
1819 if (cfg->fc_encap) {
1820 struct lwtunnel_state *lwtstate;
1822 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1823 cfg->fc_encap, AF_INET6, cfg,
1827 rt->dst.lwtstate = lwtstate_get(lwtstate);
1828 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1829 rt->dst.lwtstate->orig_output = rt->dst.output;
1830 rt->dst.output = lwtunnel_output;
1832 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1833 rt->dst.lwtstate->orig_input = rt->dst.input;
1834 rt->dst.input = lwtunnel_input;
1838 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1839 rt->rt6i_dst.plen = cfg->fc_dst_len;
1840 if (rt->rt6i_dst.plen == 128)
1841 rt->dst.flags |= DST_HOST;
1843 #ifdef CONFIG_IPV6_SUBTREES
1844 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1845 rt->rt6i_src.plen = cfg->fc_src_len;
1848 rt->rt6i_metric = cfg->fc_metric;
1850 /* We cannot add true routes via loopback here,
1851 they would result in kernel looping; promote them to reject routes
1853 if ((cfg->fc_flags & RTF_REJECT) ||
1854 (dev && (dev->flags & IFF_LOOPBACK) &&
1855 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1856 !(cfg->fc_flags & RTF_LOCAL))) {
1857 /* hold loopback dev/idev if we haven't done so. */
1858 if (dev != net->loopback_dev) {
1863 dev = net->loopback_dev;
1865 idev = in6_dev_get(dev);
1871 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1872 switch (cfg->fc_type) {
1874 rt->dst.error = -EINVAL;
1875 rt->dst.output = dst_discard_out;
1876 rt->dst.input = dst_discard;
1879 rt->dst.error = -EACCES;
1880 rt->dst.output = ip6_pkt_prohibit_out;
1881 rt->dst.input = ip6_pkt_prohibit;
1884 case RTN_UNREACHABLE:
1886 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1887 : (cfg->fc_type == RTN_UNREACHABLE)
1888 ? -EHOSTUNREACH : -ENETUNREACH;
1889 rt->dst.output = ip6_pkt_discard_out;
1890 rt->dst.input = ip6_pkt_discard;
1896 if (cfg->fc_flags & RTF_GATEWAY) {
1897 const struct in6_addr *gw_addr;
1900 gw_addr = &cfg->fc_gateway;
1901 gwa_type = ipv6_addr_type(gw_addr);
1903 /* if gw_addr is local we will fail to detect this in case
1904 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1905 * will return already-added prefix route via interface that
1906 * prefix route was assigned to, which might be non-loopback.
1909 if (ipv6_chk_addr_and_flags(net, gw_addr,
1910 gwa_type & IPV6_ADDR_LINKLOCAL ?
1914 rt->rt6i_gateway = *gw_addr;
1916 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1917 struct rt6_info *grt;
1919 /* IPv6 strictly inhibits using not link-local
1920 addresses as nexthop address.
1921 Otherwise, router will not able to send redirects.
1922 It is very good, but in some (rare!) circumstances
1923 (SIT, PtP, NBMA NOARP links) it is handy to allow
1924 some exceptions. --ANK
1926 if (!(gwa_type & IPV6_ADDR_UNICAST))
1929 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1931 err = -EHOSTUNREACH;
1935 if (dev != grt->dst.dev) {
1941 idev = grt->rt6i_idev;
1943 in6_dev_hold(grt->rt6i_idev);
1945 if (!(grt->rt6i_flags & RTF_GATEWAY))
1953 if (!dev || (dev->flags & IFF_LOOPBACK))
1961 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1962 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1966 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1967 rt->rt6i_prefsrc.plen = 128;
1969 rt->rt6i_prefsrc.plen = 0;
1971 rt->rt6i_flags = cfg->fc_flags;
1975 rt->rt6i_idev = idev;
1976 rt->rt6i_table = table;
1978 cfg->fc_nlinfo.nl_net = dev_net(dev);
1989 return ERR_PTR(err);
1992 int ip6_route_add(struct fib6_config *cfg)
1994 struct mx6_config mxc = { .mx = NULL, };
1995 struct rt6_info *rt;
1998 rt = ip6_route_info_create(cfg);
2005 err = ip6_convert_metrics(&mxc, cfg);
2009 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2021 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2024 struct fib6_table *table;
2025 struct net *net = dev_net(rt->dst.dev);
2027 if (rt == net->ipv6.ip6_null_entry ||
2028 rt->dst.flags & DST_NOCACHE) {
2033 table = rt->rt6i_table;
2034 write_lock_bh(&table->tb6_lock);
2035 err = fib6_del(rt, info);
2036 write_unlock_bh(&table->tb6_lock);
2043 int ip6_del_rt(struct rt6_info *rt)
2045 struct nl_info info = {
2046 .nl_net = dev_net(rt->dst.dev),
2048 return __ip6_del_rt(rt, &info);
2051 static int ip6_route_del(struct fib6_config *cfg)
2053 struct fib6_table *table;
2054 struct fib6_node *fn;
2055 struct rt6_info *rt;
2058 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2062 read_lock_bh(&table->tb6_lock);
2064 fn = fib6_locate(&table->tb6_root,
2065 &cfg->fc_dst, cfg->fc_dst_len,
2066 &cfg->fc_src, cfg->fc_src_len);
2069 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2070 if ((rt->rt6i_flags & RTF_CACHE) &&
2071 !(cfg->fc_flags & RTF_CACHE))
2073 if (cfg->fc_ifindex &&
2075 rt->dst.dev->ifindex != cfg->fc_ifindex))
2077 if (cfg->fc_flags & RTF_GATEWAY &&
2078 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2080 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2083 read_unlock_bh(&table->tb6_lock);
2085 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2088 read_unlock_bh(&table->tb6_lock);
2093 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2095 struct netevent_redirect netevent;
2096 struct rt6_info *rt, *nrt = NULL;
2097 struct ndisc_options ndopts;
2098 struct inet6_dev *in6_dev;
2099 struct neighbour *neigh;
2101 int optlen, on_link;
2104 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2105 optlen -= sizeof(*msg);
2108 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2112 msg = (struct rd_msg *)icmp6_hdr(skb);
2114 if (ipv6_addr_is_multicast(&msg->dest)) {
2115 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2120 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2122 } else if (ipv6_addr_type(&msg->target) !=
2123 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2124 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2128 in6_dev = __in6_dev_get(skb->dev);
2131 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2135 * The IP source address of the Redirect MUST be the same as the current
2136 * first-hop router for the specified ICMP Destination Address.
2139 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2140 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2145 if (ndopts.nd_opts_tgt_lladdr) {
2146 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2149 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2154 rt = (struct rt6_info *) dst;
2155 if (rt->rt6i_flags & RTF_REJECT) {
2156 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2160 /* Redirect received -> path was valid.
2161 * Look, redirects are sent only in response to data packets,
2162 * so that this nexthop apparently is reachable. --ANK
2164 dst_confirm(&rt->dst);
2166 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2171 * We have finally decided to accept it.
2174 neigh_update(neigh, lladdr, NUD_STALE,
2175 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2176 NEIGH_UPDATE_F_OVERRIDE|
2177 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2178 NEIGH_UPDATE_F_ISROUTER))
2181 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2185 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2187 nrt->rt6i_flags &= ~RTF_GATEWAY;
2189 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2191 if (ip6_ins_rt(nrt))
2194 netevent.old = &rt->dst;
2195 netevent.new = &nrt->dst;
2196 netevent.daddr = &msg->dest;
2197 netevent.neigh = neigh;
2198 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2200 if (rt->rt6i_flags & RTF_CACHE) {
2201 rt = (struct rt6_info *) dst_clone(&rt->dst);
2206 neigh_release(neigh);
2210 * Misc support functions
2213 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2215 BUG_ON(from->dst.from);
2217 rt->rt6i_flags &= ~RTF_EXPIRES;
2218 dst_hold(&from->dst);
2219 rt->dst.from = &from->dst;
2220 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2223 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2225 rt->dst.input = ort->dst.input;
2226 rt->dst.output = ort->dst.output;
2227 rt->rt6i_dst = ort->rt6i_dst;
2228 rt->dst.error = ort->dst.error;
2229 rt->rt6i_idev = ort->rt6i_idev;
2231 in6_dev_hold(rt->rt6i_idev);
2232 rt->dst.lastuse = jiffies;
2233 rt->rt6i_gateway = ort->rt6i_gateway;
2234 rt->rt6i_flags = ort->rt6i_flags;
2235 rt6_set_from(rt, ort);
2236 rt->rt6i_metric = ort->rt6i_metric;
2237 #ifdef CONFIG_IPV6_SUBTREES
2238 rt->rt6i_src = ort->rt6i_src;
2240 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2241 rt->rt6i_table = ort->rt6i_table;
2242 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2245 #ifdef CONFIG_IPV6_ROUTE_INFO
2246 static struct rt6_info *rt6_get_route_info(struct net_device *dev,
2247 const struct in6_addr *prefix, int prefixlen,
2248 const struct in6_addr *gwaddr)
2250 struct fib6_node *fn;
2251 struct rt6_info *rt = NULL;
2252 struct fib6_table *table;
2254 table = fib6_get_table(dev_net(dev),
2255 addrconf_rt_table(dev, RT6_TABLE_INFO));
2259 read_lock_bh(&table->tb6_lock);
2260 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2264 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2265 if (rt->dst.dev->ifindex != dev->ifindex)
2267 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2269 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2275 read_unlock_bh(&table->tb6_lock);
2279 static struct rt6_info *rt6_add_route_info(struct net_device *dev,
2280 const struct in6_addr *prefix, int prefixlen,
2281 const struct in6_addr *gwaddr, unsigned int pref)
2283 struct fib6_config cfg = {
2284 .fc_metric = IP6_RT_PRIO_USER,
2285 .fc_ifindex = dev->ifindex,
2286 .fc_dst_len = prefixlen,
2287 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2288 RTF_UP | RTF_PREF(pref),
2289 .fc_nlinfo.portid = 0,
2290 .fc_nlinfo.nlh = NULL,
2291 .fc_nlinfo.nl_net = dev_net(dev),
2294 cfg.fc_table = l3mdev_fib_table_by_index(dev_net(dev), dev->ifindex) ? : addrconf_rt_table(dev, RT6_TABLE_INFO);
2295 cfg.fc_dst = *prefix;
2296 cfg.fc_gateway = *gwaddr;
2298 /* We should treat it as a default route if prefix length is 0. */
2300 cfg.fc_flags |= RTF_DEFAULT;
2302 ip6_route_add(&cfg);
2304 return rt6_get_route_info(dev, prefix, prefixlen, gwaddr);
2308 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2310 struct rt6_info *rt;
2311 struct fib6_table *table;
2313 table = fib6_get_table(dev_net(dev),
2314 addrconf_rt_table(dev, RT6_TABLE_MAIN));
2318 read_lock_bh(&table->tb6_lock);
2319 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2320 if (dev == rt->dst.dev &&
2321 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2322 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2327 read_unlock_bh(&table->tb6_lock);
2331 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2332 struct net_device *dev,
2335 struct fib6_config cfg = {
2336 .fc_table = l3mdev_fib_table(dev) ? : addrconf_rt_table(dev, RT6_TABLE_DFLT),
2337 .fc_metric = IP6_RT_PRIO_USER,
2338 .fc_ifindex = dev->ifindex,
2339 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2340 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2341 .fc_nlinfo.portid = 0,
2342 .fc_nlinfo.nlh = NULL,
2343 .fc_nlinfo.nl_net = dev_net(dev),
2346 cfg.fc_gateway = *gwaddr;
2348 ip6_route_add(&cfg);
2350 return rt6_get_dflt_router(gwaddr, dev);
2354 int rt6_addrconf_purge(struct rt6_info *rt, void *arg) {
2355 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2356 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2))
2361 void rt6_purge_dflt_routers(struct net *net)
2363 fib6_clean_all(net, rt6_addrconf_purge, NULL);
2366 static void rtmsg_to_fib6_config(struct net *net,
2367 struct in6_rtmsg *rtmsg,
2368 struct fib6_config *cfg)
2370 memset(cfg, 0, sizeof(*cfg));
2372 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2374 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2375 cfg->fc_metric = rtmsg->rtmsg_metric;
2376 cfg->fc_expires = rtmsg->rtmsg_info;
2377 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2378 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2379 cfg->fc_flags = rtmsg->rtmsg_flags;
2381 cfg->fc_nlinfo.nl_net = net;
2383 cfg->fc_dst = rtmsg->rtmsg_dst;
2384 cfg->fc_src = rtmsg->rtmsg_src;
2385 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2388 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2390 struct fib6_config cfg;
2391 struct in6_rtmsg rtmsg;
2395 case SIOCADDRT: /* Add a route */
2396 case SIOCDELRT: /* Delete a route */
2397 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2399 err = copy_from_user(&rtmsg, arg,
2400 sizeof(struct in6_rtmsg));
2404 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2409 err = ip6_route_add(&cfg);
2412 err = ip6_route_del(&cfg);
2426 * Drop the packet on the floor
2429 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2432 struct dst_entry *dst = skb_dst(skb);
2433 switch (ipstats_mib_noroutes) {
2434 case IPSTATS_MIB_INNOROUTES:
2435 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2436 if (type == IPV6_ADDR_ANY) {
2437 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2438 IPSTATS_MIB_INADDRERRORS);
2442 case IPSTATS_MIB_OUTNOROUTES:
2443 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2444 ipstats_mib_noroutes);
2447 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2452 static int ip6_pkt_discard(struct sk_buff *skb)
2454 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2457 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2459 skb->dev = skb_dst(skb)->dev;
2460 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2463 static int ip6_pkt_prohibit(struct sk_buff *skb)
2465 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2468 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2470 skb->dev = skb_dst(skb)->dev;
2471 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2475 * Allocate a dst for local (unicast / anycast) address.
2478 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2479 const struct in6_addr *addr,
2483 struct net *net = dev_net(idev->dev);
2484 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2487 return ERR_PTR(-ENOMEM);
2491 rt->dst.flags |= DST_HOST;
2492 rt->dst.input = ip6_input;
2493 rt->dst.output = ip6_output;
2494 rt->rt6i_idev = idev;
2496 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2498 rt->rt6i_flags |= RTF_ANYCAST;
2500 rt->rt6i_flags |= RTF_LOCAL;
2502 rt->rt6i_gateway = *addr;
2503 rt->rt6i_dst.addr = *addr;
2504 rt->rt6i_dst.plen = 128;
2505 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2506 rt->rt6i_table = fib6_get_table(net, tb_id);
2507 rt->dst.flags |= DST_NOCACHE;
2509 atomic_set(&rt->dst.__refcnt, 1);
2514 int ip6_route_get_saddr(struct net *net,
2515 struct rt6_info *rt,
2516 const struct in6_addr *daddr,
2518 struct in6_addr *saddr)
2520 struct inet6_dev *idev =
2521 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2523 if (rt && rt->rt6i_prefsrc.plen)
2524 *saddr = rt->rt6i_prefsrc.addr;
2526 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2527 daddr, prefs, saddr);
2531 /* remove deleted ip from prefsrc entries */
2532 struct arg_dev_net_ip {
2533 struct net_device *dev;
2535 struct in6_addr *addr;
2538 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2540 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2541 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2542 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2544 if (((void *)rt->dst.dev == dev || !dev) &&
2545 rt != net->ipv6.ip6_null_entry &&
2546 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2547 /* remove prefsrc entry */
2548 rt->rt6i_prefsrc.plen = 0;
2553 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2555 struct net *net = dev_net(ifp->idev->dev);
2556 struct arg_dev_net_ip adni = {
2557 .dev = ifp->idev->dev,
2561 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2564 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2565 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2567 /* Remove routers and update dst entries when gateway turn into host. */
2568 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2570 struct in6_addr *gateway = (struct in6_addr *)arg;
2572 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2573 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2574 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2580 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2582 fib6_clean_all(net, fib6_clean_tohost, gateway);
2585 struct arg_dev_net {
2586 struct net_device *dev;
2590 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2592 const struct arg_dev_net *adn = arg;
2593 const struct net_device *dev = adn->dev;
2595 if ((rt->dst.dev == dev || !dev) &&
2596 rt != adn->net->ipv6.ip6_null_entry)
2602 void rt6_ifdown(struct net *net, struct net_device *dev)
2604 struct arg_dev_net adn = {
2609 fib6_clean_all(net, fib6_ifdown, &adn);
2610 icmp6_clean_all(fib6_ifdown, &adn);
2612 rt6_uncached_list_flush_dev(net, dev);
2615 struct rt6_mtu_change_arg {
2616 struct net_device *dev;
2620 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2622 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2623 struct inet6_dev *idev;
2625 /* In IPv6 pmtu discovery is not optional,
2626 so that RTAX_MTU lock cannot disable it.
2627 We still use this lock to block changes
2628 caused by addrconf/ndisc.
2631 idev = __in6_dev_get(arg->dev);
2635 /* For administrative MTU increase, there is no way to discover
2636 IPv6 PMTU increase, so PMTU increase should be updated here.
2637 Since RFC 1981 doesn't include administrative MTU increase
2638 update PMTU increase is a MUST. (i.e. jumbo frame)
2641 If new MTU is less than route PMTU, this new MTU will be the
2642 lowest MTU in the path, update the route PMTU to reflect PMTU
2643 decreases; if new MTU is greater than route PMTU, and the
2644 old MTU is the lowest MTU in the path, update the route PMTU
2645 to reflect the increase. In this case if the other nodes' MTU
2646 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2649 if (rt->dst.dev == arg->dev &&
2650 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2651 if (rt->rt6i_flags & RTF_CACHE) {
2652 /* For RTF_CACHE with rt6i_pmtu == 0
2653 * (i.e. a redirected route),
2654 * the metrics of its rt->dst.from has already
2657 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2658 rt->rt6i_pmtu = arg->mtu;
2659 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2660 (dst_mtu(&rt->dst) < arg->mtu &&
2661 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2662 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2668 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2670 struct rt6_mtu_change_arg arg = {
2675 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2678 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2679 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2680 [RTA_OIF] = { .type = NLA_U32 },
2681 [RTA_IIF] = { .type = NLA_U32 },
2682 [RTA_PRIORITY] = { .type = NLA_U32 },
2683 [RTA_METRICS] = { .type = NLA_NESTED },
2684 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2685 [RTA_PREF] = { .type = NLA_U8 },
2686 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2687 [RTA_ENCAP] = { .type = NLA_NESTED },
2690 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2691 struct fib6_config *cfg)
2694 struct nlattr *tb[RTA_MAX+1];
2698 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2703 rtm = nlmsg_data(nlh);
2704 memset(cfg, 0, sizeof(*cfg));
2706 cfg->fc_table = rtm->rtm_table;
2707 cfg->fc_dst_len = rtm->rtm_dst_len;
2708 cfg->fc_src_len = rtm->rtm_src_len;
2709 cfg->fc_flags = RTF_UP;
2710 cfg->fc_protocol = rtm->rtm_protocol;
2711 cfg->fc_type = rtm->rtm_type;
2713 if (rtm->rtm_type == RTN_UNREACHABLE ||
2714 rtm->rtm_type == RTN_BLACKHOLE ||
2715 rtm->rtm_type == RTN_PROHIBIT ||
2716 rtm->rtm_type == RTN_THROW)
2717 cfg->fc_flags |= RTF_REJECT;
2719 if (rtm->rtm_type == RTN_LOCAL)
2720 cfg->fc_flags |= RTF_LOCAL;
2722 if (rtm->rtm_flags & RTM_F_CLONED)
2723 cfg->fc_flags |= RTF_CACHE;
2725 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2726 cfg->fc_nlinfo.nlh = nlh;
2727 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2729 if (tb[RTA_GATEWAY]) {
2730 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2731 cfg->fc_flags |= RTF_GATEWAY;
2735 int plen = (rtm->rtm_dst_len + 7) >> 3;
2737 if (nla_len(tb[RTA_DST]) < plen)
2740 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2744 int plen = (rtm->rtm_src_len + 7) >> 3;
2746 if (nla_len(tb[RTA_SRC]) < plen)
2749 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2752 if (tb[RTA_PREFSRC])
2753 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2756 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2758 if (tb[RTA_PRIORITY])
2759 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2761 if (tb[RTA_METRICS]) {
2762 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2763 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2767 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2769 if (tb[RTA_MULTIPATH]) {
2770 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2771 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2775 pref = nla_get_u8(tb[RTA_PREF]);
2776 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2777 pref != ICMPV6_ROUTER_PREF_HIGH)
2778 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2779 cfg->fc_flags |= RTF_PREF(pref);
2783 cfg->fc_encap = tb[RTA_ENCAP];
2785 if (tb[RTA_ENCAP_TYPE])
2786 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2794 struct rt6_info *rt6_info;
2795 struct fib6_config r_cfg;
2796 struct mx6_config mxc;
2797 struct list_head next;
2800 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2804 list_for_each_entry(nh, rt6_nh_list, next) {
2805 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2806 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2807 nh->r_cfg.fc_ifindex);
2811 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2812 struct rt6_info *rt, struct fib6_config *r_cfg)
2815 struct rt6_info *rtnh;
2818 list_for_each_entry(nh, rt6_nh_list, next) {
2819 /* check if rt6_info already exists */
2820 rtnh = nh->rt6_info;
2822 if (rtnh->dst.dev == rt->dst.dev &&
2823 rtnh->rt6i_idev == rt->rt6i_idev &&
2824 ipv6_addr_equal(&rtnh->rt6i_gateway,
2829 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2833 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2838 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2839 list_add_tail(&nh->next, rt6_nh_list);
2844 static int ip6_route_multipath_add(struct fib6_config *cfg)
2846 struct fib6_config r_cfg;
2847 struct rtnexthop *rtnh;
2848 struct rt6_info *rt;
2849 struct rt6_nh *err_nh;
2850 struct rt6_nh *nh, *nh_safe;
2855 int replace = (cfg->fc_nlinfo.nlh &&
2856 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2857 LIST_HEAD(rt6_nh_list);
2859 remaining = cfg->fc_mp_len;
2860 rtnh = (struct rtnexthop *)cfg->fc_mp;
2862 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2863 * rt6_info structs per nexthop
2865 while (rtnh_ok(rtnh, remaining)) {
2866 memcpy(&r_cfg, cfg, sizeof(*cfg));
2867 if (rtnh->rtnh_ifindex)
2868 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2870 attrlen = rtnh_attrlen(rtnh);
2872 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2874 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2876 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2877 r_cfg.fc_flags |= RTF_GATEWAY;
2879 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2880 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2882 r_cfg.fc_encap_type = nla_get_u16(nla);
2885 rt = ip6_route_info_create(&r_cfg);
2892 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2898 rtnh = rtnh_next(rtnh, &remaining);
2902 list_for_each_entry(nh, &rt6_nh_list, next) {
2903 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2904 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2905 nh->rt6_info = NULL;
2908 ip6_print_replace_route_err(&rt6_nh_list);
2913 /* Because each route is added like a single route we remove
2914 * these flags after the first nexthop: if there is a collision,
2915 * we have already failed to add the first nexthop:
2916 * fib6_add_rt2node() has rejected it; when replacing, old
2917 * nexthops have been replaced by first new, the rest should
2920 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2928 /* Delete routes that were already added */
2929 list_for_each_entry(nh, &rt6_nh_list, next) {
2932 ip6_route_del(&nh->r_cfg);
2936 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2938 dst_free(&nh->rt6_info->dst);
2940 list_del(&nh->next);
2947 static int ip6_route_multipath_del(struct fib6_config *cfg)
2949 struct fib6_config r_cfg;
2950 struct rtnexthop *rtnh;
2953 int err = 1, last_err = 0;
2955 remaining = cfg->fc_mp_len;
2956 rtnh = (struct rtnexthop *)cfg->fc_mp;
2958 /* Parse a Multipath Entry */
2959 while (rtnh_ok(rtnh, remaining)) {
2960 memcpy(&r_cfg, cfg, sizeof(*cfg));
2961 if (rtnh->rtnh_ifindex)
2962 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2964 attrlen = rtnh_attrlen(rtnh);
2966 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2968 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2970 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2971 r_cfg.fc_flags |= RTF_GATEWAY;
2974 err = ip6_route_del(&r_cfg);
2978 rtnh = rtnh_next(rtnh, &remaining);
2984 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2986 struct fib6_config cfg;
2989 err = rtm_to_fib6_config(skb, nlh, &cfg);
2994 return ip6_route_multipath_del(&cfg);
2996 return ip6_route_del(&cfg);
2999 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3001 struct fib6_config cfg;
3004 err = rtm_to_fib6_config(skb, nlh, &cfg);
3009 return ip6_route_multipath_add(&cfg);
3011 return ip6_route_add(&cfg);
3014 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3016 return NLMSG_ALIGN(sizeof(struct rtmsg))
3017 + nla_total_size(16) /* RTA_SRC */
3018 + nla_total_size(16) /* RTA_DST */
3019 + nla_total_size(16) /* RTA_GATEWAY */
3020 + nla_total_size(16) /* RTA_PREFSRC */
3021 + nla_total_size(4) /* RTA_TABLE */
3022 + nla_total_size(4) /* RTA_IIF */
3023 + nla_total_size(4) /* RTA_OIF */
3024 + nla_total_size(4) /* RTA_PRIORITY */
3025 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3026 + nla_total_size(sizeof(struct rta_cacheinfo))
3027 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3028 + nla_total_size(1) /* RTA_PREF */
3029 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3032 static int rt6_fill_node(struct net *net,
3033 struct sk_buff *skb, struct rt6_info *rt,
3034 struct in6_addr *dst, struct in6_addr *src,
3035 int iif, int type, u32 portid, u32 seq,
3036 int prefix, int nowait, unsigned int flags)
3038 u32 metrics[RTAX_MAX];
3040 struct nlmsghdr *nlh;
3044 if (prefix) { /* user wants prefix routes only */
3045 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3046 /* success since this is not a prefix route */
3051 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3055 rtm = nlmsg_data(nlh);
3056 rtm->rtm_family = AF_INET6;
3057 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3058 rtm->rtm_src_len = rt->rt6i_src.plen;
3061 table = rt->rt6i_table->tb6_id;
3063 table = RT6_TABLE_UNSPEC;
3064 rtm->rtm_table = table;
3065 if (nla_put_u32(skb, RTA_TABLE, table))
3066 goto nla_put_failure;
3067 if (rt->rt6i_flags & RTF_REJECT) {
3068 switch (rt->dst.error) {
3070 rtm->rtm_type = RTN_BLACKHOLE;
3073 rtm->rtm_type = RTN_PROHIBIT;
3076 rtm->rtm_type = RTN_THROW;
3079 rtm->rtm_type = RTN_UNREACHABLE;
3083 else if (rt->rt6i_flags & RTF_LOCAL)
3084 rtm->rtm_type = RTN_LOCAL;
3085 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3086 rtm->rtm_type = RTN_LOCAL;
3088 rtm->rtm_type = RTN_UNICAST;
3090 if (!netif_carrier_ok(rt->dst.dev)) {
3091 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3092 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3093 rtm->rtm_flags |= RTNH_F_DEAD;
3095 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3096 rtm->rtm_protocol = rt->rt6i_protocol;
3097 if (rt->rt6i_flags & RTF_DYNAMIC)
3098 rtm->rtm_protocol = RTPROT_REDIRECT;
3099 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3100 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3101 rtm->rtm_protocol = RTPROT_RA;
3103 rtm->rtm_protocol = RTPROT_KERNEL;
3106 if (rt->rt6i_flags & RTF_CACHE)
3107 rtm->rtm_flags |= RTM_F_CLONED;
3110 if (nla_put_in6_addr(skb, RTA_DST, dst))
3111 goto nla_put_failure;
3112 rtm->rtm_dst_len = 128;
3113 } else if (rtm->rtm_dst_len)
3114 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3115 goto nla_put_failure;
3116 #ifdef CONFIG_IPV6_SUBTREES
3118 if (nla_put_in6_addr(skb, RTA_SRC, src))
3119 goto nla_put_failure;
3120 rtm->rtm_src_len = 128;
3121 } else if (rtm->rtm_src_len &&
3122 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3123 goto nla_put_failure;
3126 #ifdef CONFIG_IPV6_MROUTE
3127 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3128 int err = ip6mr_get_route(net, skb, rtm, nowait);
3133 goto nla_put_failure;
3135 if (err == -EMSGSIZE)
3136 goto nla_put_failure;
3141 if (nla_put_u32(skb, RTA_IIF, iif))
3142 goto nla_put_failure;
3144 struct in6_addr saddr_buf;
3145 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3146 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3147 goto nla_put_failure;
3150 if (rt->rt6i_prefsrc.plen) {
3151 struct in6_addr saddr_buf;
3152 saddr_buf = rt->rt6i_prefsrc.addr;
3153 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3154 goto nla_put_failure;
3157 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3159 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3160 if (rtnetlink_put_metrics(skb, metrics) < 0)
3161 goto nla_put_failure;
3163 if (rt->rt6i_flags & RTF_GATEWAY) {
3164 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3165 goto nla_put_failure;
3169 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3170 goto nla_put_failure;
3171 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3172 goto nla_put_failure;
3174 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3176 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3177 goto nla_put_failure;
3179 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3180 goto nla_put_failure;
3182 lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3184 nlmsg_end(skb, nlh);
3188 nlmsg_cancel(skb, nlh);
3192 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3194 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3197 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3198 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3199 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3203 return rt6_fill_node(arg->net,
3204 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3205 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3206 prefix, 0, NLM_F_MULTI);
3209 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3211 struct net *net = sock_net(in_skb->sk);
3212 struct nlattr *tb[RTA_MAX+1];
3213 struct rt6_info *rt;
3214 struct sk_buff *skb;
3217 int err, iif = 0, oif = 0;
3219 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3224 memset(&fl6, 0, sizeof(fl6));
3227 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3230 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3234 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3237 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3241 iif = nla_get_u32(tb[RTA_IIF]);
3244 oif = nla_get_u32(tb[RTA_OIF]);
3247 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3250 struct net_device *dev;
3253 dev = __dev_get_by_index(net, iif);
3259 fl6.flowi6_iif = iif;
3261 if (!ipv6_addr_any(&fl6.saddr))
3262 flags |= RT6_LOOKUP_F_HAS_SADDR;
3264 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3267 fl6.flowi6_oif = oif;
3269 if (netif_index_is_l3_master(net, oif)) {
3270 fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3271 FLOWI_FLAG_SKIP_NH_OIF;
3274 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3277 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3284 /* Reserve room for dummy headers, this skb can pass
3285 through good chunk of routing engine.
3287 skb_reset_mac_header(skb);
3288 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3290 skb_dst_set(skb, &rt->dst);
3292 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3293 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3294 nlh->nlmsg_seq, 0, 0, 0);
3300 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3305 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3306 unsigned int nlm_flags)
3308 struct sk_buff *skb;
3309 struct net *net = info->nl_net;
3314 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3316 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3320 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3321 event, info->portid, seq, 0, 0, nlm_flags);
3323 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3324 WARN_ON(err == -EMSGSIZE);
3328 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3329 info->nlh, gfp_any());
3333 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3336 static int ip6_route_dev_notify(struct notifier_block *this,
3337 unsigned long event, void *ptr)
3339 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3340 struct net *net = dev_net(dev);
3342 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3343 net->ipv6.ip6_null_entry->dst.dev = dev;
3344 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3345 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3346 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3347 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3348 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3349 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3360 #ifdef CONFIG_PROC_FS
3362 static const struct file_operations ipv6_route_proc_fops = {
3363 .owner = THIS_MODULE,
3364 .open = ipv6_route_open,
3366 .llseek = seq_lseek,
3367 .release = seq_release_net,
3370 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3372 struct net *net = (struct net *)seq->private;
3373 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3374 net->ipv6.rt6_stats->fib_nodes,
3375 net->ipv6.rt6_stats->fib_route_nodes,
3376 net->ipv6.rt6_stats->fib_rt_alloc,
3377 net->ipv6.rt6_stats->fib_rt_entries,
3378 net->ipv6.rt6_stats->fib_rt_cache,
3379 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3380 net->ipv6.rt6_stats->fib_discarded_routes);
3385 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3387 return single_open_net(inode, file, rt6_stats_seq_show);
3390 static const struct file_operations rt6_stats_seq_fops = {
3391 .owner = THIS_MODULE,
3392 .open = rt6_stats_seq_open,
3394 .llseek = seq_lseek,
3395 .release = single_release_net,
3397 #endif /* CONFIG_PROC_FS */
3399 #ifdef CONFIG_SYSCTL
3402 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3403 void __user *buffer, size_t *lenp, loff_t *ppos)
3410 net = (struct net *)ctl->extra1;
3411 delay = net->ipv6.sysctl.flush_delay;
3412 proc_dointvec(ctl, write, buffer, lenp, ppos);
3413 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3417 struct ctl_table ipv6_route_table_template[] = {
3419 .procname = "flush",
3420 .data = &init_net.ipv6.sysctl.flush_delay,
3421 .maxlen = sizeof(int),
3423 .proc_handler = ipv6_sysctl_rtcache_flush
3426 .procname = "gc_thresh",
3427 .data = &ip6_dst_ops_template.gc_thresh,
3428 .maxlen = sizeof(int),
3430 .proc_handler = proc_dointvec,
3433 .procname = "max_size",
3434 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3435 .maxlen = sizeof(int),
3437 .proc_handler = proc_dointvec,
3440 .procname = "gc_min_interval",
3441 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3442 .maxlen = sizeof(int),
3444 .proc_handler = proc_dointvec_jiffies,
3447 .procname = "gc_timeout",
3448 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3449 .maxlen = sizeof(int),
3451 .proc_handler = proc_dointvec_jiffies,
3454 .procname = "gc_interval",
3455 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3456 .maxlen = sizeof(int),
3458 .proc_handler = proc_dointvec_jiffies,
3461 .procname = "gc_elasticity",
3462 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3463 .maxlen = sizeof(int),
3465 .proc_handler = proc_dointvec,
3468 .procname = "mtu_expires",
3469 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3470 .maxlen = sizeof(int),
3472 .proc_handler = proc_dointvec_jiffies,
3475 .procname = "min_adv_mss",
3476 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3477 .maxlen = sizeof(int),
3479 .proc_handler = proc_dointvec,
3482 .procname = "gc_min_interval_ms",
3483 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3484 .maxlen = sizeof(int),
3486 .proc_handler = proc_dointvec_ms_jiffies,
3491 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3493 struct ctl_table *table;
3495 table = kmemdup(ipv6_route_table_template,
3496 sizeof(ipv6_route_table_template),
3500 table[0].data = &net->ipv6.sysctl.flush_delay;
3501 table[0].extra1 = net;
3502 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3503 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3504 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3505 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3506 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3507 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3508 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3509 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3510 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3512 /* Don't export sysctls to unprivileged users */
3513 if (net->user_ns != &init_user_ns)
3514 table[0].procname = NULL;
3521 static int __net_init ip6_route_net_init(struct net *net)
3525 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3526 sizeof(net->ipv6.ip6_dst_ops));
3528 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3529 goto out_ip6_dst_ops;
3531 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3532 sizeof(*net->ipv6.ip6_null_entry),
3534 if (!net->ipv6.ip6_null_entry)
3535 goto out_ip6_dst_entries;
3536 net->ipv6.ip6_null_entry->dst.path =
3537 (struct dst_entry *)net->ipv6.ip6_null_entry;
3538 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3539 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3540 ip6_template_metrics, true);
3542 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3543 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3544 sizeof(*net->ipv6.ip6_prohibit_entry),
3546 if (!net->ipv6.ip6_prohibit_entry)
3547 goto out_ip6_null_entry;
3548 net->ipv6.ip6_prohibit_entry->dst.path =
3549 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3550 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3551 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3552 ip6_template_metrics, true);
3554 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3555 sizeof(*net->ipv6.ip6_blk_hole_entry),
3557 if (!net->ipv6.ip6_blk_hole_entry)
3558 goto out_ip6_prohibit_entry;
3559 net->ipv6.ip6_blk_hole_entry->dst.path =
3560 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3561 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3562 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3563 ip6_template_metrics, true);
3566 net->ipv6.sysctl.flush_delay = 0;
3567 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3568 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3569 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3570 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3571 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3572 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3573 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3575 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3581 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3582 out_ip6_prohibit_entry:
3583 kfree(net->ipv6.ip6_prohibit_entry);
3585 kfree(net->ipv6.ip6_null_entry);
3587 out_ip6_dst_entries:
3588 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3593 static void __net_exit ip6_route_net_exit(struct net *net)
3595 kfree(net->ipv6.ip6_null_entry);
3596 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3597 kfree(net->ipv6.ip6_prohibit_entry);
3598 kfree(net->ipv6.ip6_blk_hole_entry);
3600 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3603 static int __net_init ip6_route_net_init_late(struct net *net)
3605 #ifdef CONFIG_PROC_FS
3606 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3607 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3612 static void __net_exit ip6_route_net_exit_late(struct net *net)
3614 #ifdef CONFIG_PROC_FS
3615 remove_proc_entry("ipv6_route", net->proc_net);
3616 remove_proc_entry("rt6_stats", net->proc_net);
3620 static struct pernet_operations ip6_route_net_ops = {
3621 .init = ip6_route_net_init,
3622 .exit = ip6_route_net_exit,
3625 static int __net_init ipv6_inetpeer_init(struct net *net)
3627 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3631 inet_peer_base_init(bp);
3632 net->ipv6.peers = bp;
3636 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3638 struct inet_peer_base *bp = net->ipv6.peers;
3640 net->ipv6.peers = NULL;
3641 inetpeer_invalidate_tree(bp);
3645 static struct pernet_operations ipv6_inetpeer_ops = {
3646 .init = ipv6_inetpeer_init,
3647 .exit = ipv6_inetpeer_exit,
3650 static struct pernet_operations ip6_route_net_late_ops = {
3651 .init = ip6_route_net_init_late,
3652 .exit = ip6_route_net_exit_late,
3655 static struct notifier_block ip6_route_dev_notifier = {
3656 .notifier_call = ip6_route_dev_notify,
3660 int __init ip6_route_init(void)
3666 ip6_dst_ops_template.kmem_cachep =
3667 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3668 SLAB_HWCACHE_ALIGN, NULL);
3669 if (!ip6_dst_ops_template.kmem_cachep)
3672 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3674 goto out_kmem_cache;
3676 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3678 goto out_dst_entries;
3680 ret = register_pernet_subsys(&ip6_route_net_ops);
3682 goto out_register_inetpeer;
3684 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3686 /* Registering of the loopback is done before this portion of code,
3687 * the loopback reference in rt6_info will not be taken, do it
3688 * manually for init_net */
3689 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3690 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3691 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3692 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3693 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3694 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3695 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3699 goto out_register_subsys;
3705 ret = fib6_rules_init();
3709 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3711 goto fib6_rules_init;
3714 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3715 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3716 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3717 goto out_register_late_subsys;
3719 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3721 goto out_register_late_subsys;
3723 for_each_possible_cpu(cpu) {
3724 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3726 INIT_LIST_HEAD(&ul->head);
3727 spin_lock_init(&ul->lock);
3733 out_register_late_subsys:
3734 unregister_pernet_subsys(&ip6_route_net_late_ops);
3736 fib6_rules_cleanup();
3741 out_register_subsys:
3742 unregister_pernet_subsys(&ip6_route_net_ops);
3743 out_register_inetpeer:
3744 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3746 dst_entries_destroy(&ip6_dst_blackhole_ops);
3748 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3752 void ip6_route_cleanup(void)
3754 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3755 unregister_pernet_subsys(&ip6_route_net_late_ops);
3756 fib6_rules_cleanup();
3759 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3760 unregister_pernet_subsys(&ip6_route_net_ops);
3761 dst_entries_destroy(&ip6_dst_blackhole_ops);
3762 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);