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>
65 #include <asm/uaccess.h>
68 #include <linux/sysctl.h>
72 RT6_NUD_FAIL_HARD = -3,
73 RT6_NUD_FAIL_PROBE = -2,
74 RT6_NUD_FAIL_DO_RR = -1,
78 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
79 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
80 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
81 static unsigned int ip6_mtu(const struct dst_entry *dst);
82 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
83 static void ip6_dst_destroy(struct dst_entry *);
84 static void ip6_dst_ifdown(struct dst_entry *,
85 struct net_device *dev, int how);
86 static int ip6_dst_gc(struct dst_ops *ops);
88 static int ip6_pkt_discard(struct sk_buff *skb);
89 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb);
90 static int ip6_pkt_prohibit(struct sk_buff *skb);
91 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb);
92 static void ip6_link_failure(struct sk_buff *skb);
93 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
94 struct sk_buff *skb, u32 mtu);
95 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
97 static void rt6_dst_from_metrics_check(struct rt6_info *rt);
98 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
100 #ifdef CONFIG_IPV6_ROUTE_INFO
101 static struct rt6_info *rt6_add_route_info(struct net *net,
102 const struct in6_addr *prefix, int prefixlen,
103 const struct in6_addr *gwaddr, int ifindex,
105 static struct rt6_info *rt6_get_route_info(struct net *net,
106 const struct in6_addr *prefix, int prefixlen,
107 const struct in6_addr *gwaddr, int ifindex);
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_sk,
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 if (time_after(jiffies, rt->dst.expires))
411 } else if (rt->dst.from) {
412 return rt6_check_expired((struct rt6_info *) rt->dst.from);
417 /* Multipath route selection:
418 * Hash based function using packet header and flowlabel.
419 * Adapted from fib_info_hashfn()
421 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
422 const struct flowi6 *fl6)
424 unsigned int val = fl6->flowi6_proto;
426 val ^= ipv6_addr_hash(&fl6->daddr);
427 val ^= ipv6_addr_hash(&fl6->saddr);
429 /* Work only if this not encapsulated */
430 switch (fl6->flowi6_proto) {
434 val ^= (__force u16)fl6->fl6_sport;
435 val ^= (__force u16)fl6->fl6_dport;
439 val ^= (__force u16)fl6->fl6_icmp_type;
440 val ^= (__force u16)fl6->fl6_icmp_code;
443 /* RFC6438 recommands to use flowlabel */
444 val ^= (__force u32)fl6->flowlabel;
446 /* Perhaps, we need to tune, this function? */
447 val = val ^ (val >> 7) ^ (val >> 12);
448 return val % candidate_count;
451 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
452 struct flowi6 *fl6, int oif,
455 struct rt6_info *sibling, *next_sibling;
458 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
459 /* Don't change the route, if route_choosen == 0
460 * (siblings does not include ourself)
463 list_for_each_entry_safe(sibling, next_sibling,
464 &match->rt6i_siblings, rt6i_siblings) {
466 if (route_choosen == 0) {
467 if (rt6_score_route(sibling, oif, strict) < 0)
477 * Route lookup. Any table->tb6_lock is implied.
480 static inline struct rt6_info *rt6_device_match(struct net *net,
482 const struct in6_addr *saddr,
486 struct rt6_info *local = NULL;
487 struct rt6_info *sprt;
489 if (!oif && ipv6_addr_any(saddr))
492 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
493 struct net_device *dev = sprt->dst.dev;
496 if (dev->ifindex == oif)
498 if (dev->flags & IFF_LOOPBACK) {
499 if (!sprt->rt6i_idev ||
500 sprt->rt6i_idev->dev->ifindex != oif) {
501 if (flags & RT6_LOOKUP_F_IFACE && oif)
503 if (local && (!oif ||
504 local->rt6i_idev->dev->ifindex == oif))
510 if (ipv6_chk_addr(net, saddr, dev,
511 flags & RT6_LOOKUP_F_IFACE))
520 if (flags & RT6_LOOKUP_F_IFACE)
521 return net->ipv6.ip6_null_entry;
527 #ifdef CONFIG_IPV6_ROUTER_PREF
528 struct __rt6_probe_work {
529 struct work_struct work;
530 struct in6_addr target;
531 struct net_device *dev;
534 static void rt6_probe_deferred(struct work_struct *w)
536 struct in6_addr mcaddr;
537 struct __rt6_probe_work *work =
538 container_of(w, struct __rt6_probe_work, work);
540 addrconf_addr_solict_mult(&work->target, &mcaddr);
541 ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL, NULL);
546 static void rt6_probe(struct rt6_info *rt)
548 struct __rt6_probe_work *work;
549 struct neighbour *neigh;
551 * Okay, this does not seem to be appropriate
552 * for now, however, we need to check if it
553 * is really so; aka Router Reachability Probing.
555 * Router Reachability Probe MUST be rate-limited
556 * to no more than one per minute.
558 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
561 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
563 if (neigh->nud_state & NUD_VALID)
567 write_lock(&neigh->lock);
568 if (!(neigh->nud_state & NUD_VALID) &&
571 rt->rt6i_idev->cnf.rtr_probe_interval)) {
572 work = kmalloc(sizeof(*work), GFP_ATOMIC);
574 __neigh_set_probe_once(neigh);
576 write_unlock(&neigh->lock);
578 work = kmalloc(sizeof(*work), GFP_ATOMIC);
582 INIT_WORK(&work->work, rt6_probe_deferred);
583 work->target = rt->rt6i_gateway;
584 dev_hold(rt->dst.dev);
585 work->dev = rt->dst.dev;
586 schedule_work(&work->work);
590 rcu_read_unlock_bh();
593 static inline void rt6_probe(struct rt6_info *rt)
599 * Default Router Selection (RFC 2461 6.3.6)
601 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
603 struct net_device *dev = rt->dst.dev;
604 if (!oif || dev->ifindex == oif)
606 if ((dev->flags & IFF_LOOPBACK) &&
607 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
612 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
614 struct neighbour *neigh;
615 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
617 if (rt->rt6i_flags & RTF_NONEXTHOP ||
618 !(rt->rt6i_flags & RTF_GATEWAY))
619 return RT6_NUD_SUCCEED;
622 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
624 read_lock(&neigh->lock);
625 if (neigh->nud_state & NUD_VALID)
626 ret = RT6_NUD_SUCCEED;
627 #ifdef CONFIG_IPV6_ROUTER_PREF
628 else if (!(neigh->nud_state & NUD_FAILED))
629 ret = RT6_NUD_SUCCEED;
631 ret = RT6_NUD_FAIL_PROBE;
633 read_unlock(&neigh->lock);
635 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
636 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
638 rcu_read_unlock_bh();
643 static int rt6_score_route(struct rt6_info *rt, int oif,
648 m = rt6_check_dev(rt, oif);
649 if (!m && (strict & RT6_LOOKUP_F_IFACE))
650 return RT6_NUD_FAIL_HARD;
651 #ifdef CONFIG_IPV6_ROUTER_PREF
652 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
654 if (strict & RT6_LOOKUP_F_REACHABLE) {
655 int n = rt6_check_neigh(rt);
662 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
663 int *mpri, struct rt6_info *match,
667 bool match_do_rr = false;
668 struct inet6_dev *idev = rt->rt6i_idev;
669 struct net_device *dev = rt->dst.dev;
671 if (dev && !netif_carrier_ok(dev) &&
672 idev->cnf.ignore_routes_with_linkdown)
675 if (rt6_check_expired(rt))
678 m = rt6_score_route(rt, oif, strict);
679 if (m == RT6_NUD_FAIL_DO_RR) {
681 m = 0; /* lowest valid score */
682 } else if (m == RT6_NUD_FAIL_HARD) {
686 if (strict & RT6_LOOKUP_F_REACHABLE)
689 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
691 *do_rr = match_do_rr;
699 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
700 struct rt6_info *rr_head,
701 u32 metric, int oif, int strict,
704 struct rt6_info *rt, *match, *cont;
709 for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
710 if (rt->rt6i_metric != metric) {
715 match = find_match(rt, oif, strict, &mpri, match, do_rr);
718 for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
719 if (rt->rt6i_metric != metric) {
724 match = find_match(rt, oif, strict, &mpri, match, do_rr);
730 for (rt = cont; rt; rt = rt->dst.rt6_next)
731 match = find_match(rt, oif, strict, &mpri, match, do_rr);
736 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
738 struct rt6_info *match, *rt0;
744 fn->rr_ptr = rt0 = fn->leaf;
746 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
750 struct rt6_info *next = rt0->dst.rt6_next;
752 /* no entries matched; do round-robin */
753 if (!next || next->rt6i_metric != rt0->rt6i_metric)
760 net = dev_net(rt0->dst.dev);
761 return match ? match : net->ipv6.ip6_null_entry;
764 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
766 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
769 #ifdef CONFIG_IPV6_ROUTE_INFO
770 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
771 const struct in6_addr *gwaddr)
773 struct net *net = dev_net(dev);
774 struct route_info *rinfo = (struct route_info *) opt;
775 struct in6_addr prefix_buf, *prefix;
777 unsigned long lifetime;
780 if (len < sizeof(struct route_info)) {
784 /* Sanity check for prefix_len and length */
785 if (rinfo->length > 3) {
787 } else if (rinfo->prefix_len > 128) {
789 } else if (rinfo->prefix_len > 64) {
790 if (rinfo->length < 2) {
793 } else if (rinfo->prefix_len > 0) {
794 if (rinfo->length < 1) {
799 pref = rinfo->route_pref;
800 if (pref == ICMPV6_ROUTER_PREF_INVALID)
803 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
805 if (rinfo->length == 3)
806 prefix = (struct in6_addr *)rinfo->prefix;
808 /* this function is safe */
809 ipv6_addr_prefix(&prefix_buf,
810 (struct in6_addr *)rinfo->prefix,
812 prefix = &prefix_buf;
815 if (rinfo->prefix_len == 0)
816 rt = rt6_get_dflt_router(gwaddr, dev);
818 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
819 gwaddr, dev->ifindex);
821 if (rt && !lifetime) {
827 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
830 rt->rt6i_flags = RTF_ROUTEINFO |
831 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
834 if (!addrconf_finite_timeout(lifetime))
835 rt6_clean_expires(rt);
837 rt6_set_expires(rt, jiffies + HZ * lifetime);
845 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
846 struct in6_addr *saddr)
848 struct fib6_node *pn;
850 if (fn->fn_flags & RTN_TL_ROOT)
853 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
854 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
857 if (fn->fn_flags & RTN_RTINFO)
862 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
863 struct fib6_table *table,
864 struct flowi6 *fl6, int flags)
866 struct fib6_node *fn;
869 read_lock_bh(&table->tb6_lock);
870 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
873 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
874 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
875 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
876 if (rt == net->ipv6.ip6_null_entry) {
877 fn = fib6_backtrack(fn, &fl6->saddr);
881 dst_use(&rt->dst, jiffies);
882 read_unlock_bh(&table->tb6_lock);
887 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
890 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
892 EXPORT_SYMBOL_GPL(ip6_route_lookup);
894 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
895 const struct in6_addr *saddr, int oif, int strict)
897 struct flowi6 fl6 = {
901 struct dst_entry *dst;
902 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
905 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
906 flags |= RT6_LOOKUP_F_HAS_SADDR;
909 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
911 return (struct rt6_info *) dst;
917 EXPORT_SYMBOL(rt6_lookup);
919 /* ip6_ins_rt is called with FREE table->tb6_lock.
920 It takes new route entry, the addition fails by any reason the
921 route is freed. In any case, if caller does not hold it, it may
925 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
926 struct mx6_config *mxc)
929 struct fib6_table *table;
931 table = rt->rt6i_table;
932 write_lock_bh(&table->tb6_lock);
933 err = fib6_add(&table->tb6_root, rt, info, mxc);
934 write_unlock_bh(&table->tb6_lock);
939 int ip6_ins_rt(struct rt6_info *rt)
941 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
942 struct mx6_config mxc = { .mx = NULL, };
944 return __ip6_ins_rt(rt, &info, &mxc);
947 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
948 const struct in6_addr *daddr,
949 const struct in6_addr *saddr)
957 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
958 ort = (struct rt6_info *)ort->dst.from;
960 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
965 ip6_rt_copy_init(rt, ort);
966 rt->rt6i_flags |= RTF_CACHE;
968 rt->dst.flags |= DST_HOST;
969 rt->rt6i_dst.addr = *daddr;
970 rt->rt6i_dst.plen = 128;
972 if (!rt6_is_gw_or_nonexthop(ort)) {
973 if (ort->rt6i_dst.plen != 128 &&
974 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
975 rt->rt6i_flags |= RTF_ANYCAST;
976 #ifdef CONFIG_IPV6_SUBTREES
977 if (rt->rt6i_src.plen && saddr) {
978 rt->rt6i_src.addr = *saddr;
979 rt->rt6i_src.plen = 128;
987 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
989 struct rt6_info *pcpu_rt;
991 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
992 rt->dst.dev, rt->dst.flags);
996 ip6_rt_copy_init(pcpu_rt, rt);
997 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
998 pcpu_rt->rt6i_flags |= RTF_PCPU;
1002 /* It should be called with read_lock_bh(&tb6_lock) acquired */
1003 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
1005 struct rt6_info *pcpu_rt, **p;
1007 p = this_cpu_ptr(rt->rt6i_pcpu);
1011 dst_hold(&pcpu_rt->dst);
1012 rt6_dst_from_metrics_check(pcpu_rt);
1017 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1019 struct fib6_table *table = rt->rt6i_table;
1020 struct rt6_info *pcpu_rt, *prev, **p;
1022 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1024 struct net *net = dev_net(rt->dst.dev);
1026 dst_hold(&net->ipv6.ip6_null_entry->dst);
1027 return net->ipv6.ip6_null_entry;
1030 read_lock_bh(&table->tb6_lock);
1031 if (rt->rt6i_pcpu) {
1032 p = this_cpu_ptr(rt->rt6i_pcpu);
1033 prev = cmpxchg(p, NULL, pcpu_rt);
1035 /* If someone did it before us, return prev instead */
1036 dst_destroy(&pcpu_rt->dst);
1040 /* rt has been removed from the fib6 tree
1041 * before we have a chance to acquire the read_lock.
1042 * In this case, don't brother to create a pcpu rt
1043 * since rt is going away anyway. The next
1044 * dst_check() will trigger a re-lookup.
1046 dst_destroy(&pcpu_rt->dst);
1049 dst_hold(&pcpu_rt->dst);
1050 rt6_dst_from_metrics_check(pcpu_rt);
1051 read_unlock_bh(&table->tb6_lock);
1055 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1056 struct flowi6 *fl6, int flags)
1058 struct fib6_node *fn, *saved_fn;
1059 struct rt6_info *rt;
1062 strict |= flags & RT6_LOOKUP_F_IFACE;
1063 if (net->ipv6.devconf_all->forwarding == 0)
1064 strict |= RT6_LOOKUP_F_REACHABLE;
1066 read_lock_bh(&table->tb6_lock);
1068 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1071 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1075 rt = rt6_select(fn, oif, strict);
1076 if (rt->rt6i_nsiblings)
1077 rt = rt6_multipath_select(rt, fl6, oif, strict);
1078 if (rt == net->ipv6.ip6_null_entry) {
1079 fn = fib6_backtrack(fn, &fl6->saddr);
1081 goto redo_rt6_select;
1082 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1083 /* also consider unreachable route */
1084 strict &= ~RT6_LOOKUP_F_REACHABLE;
1086 goto redo_rt6_select;
1091 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1092 dst_use(&rt->dst, jiffies);
1093 read_unlock_bh(&table->tb6_lock);
1095 rt6_dst_from_metrics_check(rt);
1097 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1098 !(rt->rt6i_flags & RTF_GATEWAY))) {
1099 /* Create a RTF_CACHE clone which will not be
1100 * owned by the fib6 tree. It is for the special case where
1101 * the daddr in the skb during the neighbor look-up is different
1102 * from the fl6->daddr used to look-up route here.
1105 struct rt6_info *uncached_rt;
1107 dst_use(&rt->dst, jiffies);
1108 read_unlock_bh(&table->tb6_lock);
1110 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1111 dst_release(&rt->dst);
1114 rt6_uncached_list_add(uncached_rt);
1116 uncached_rt = net->ipv6.ip6_null_entry;
1118 dst_hold(&uncached_rt->dst);
1122 /* Get a percpu copy */
1124 struct rt6_info *pcpu_rt;
1126 rt->dst.lastuse = jiffies;
1128 pcpu_rt = rt6_get_pcpu_route(rt);
1131 read_unlock_bh(&table->tb6_lock);
1133 /* We have to do the read_unlock first
1134 * because rt6_make_pcpu_route() may trigger
1135 * ip6_dst_gc() which will take the write_lock.
1138 read_unlock_bh(&table->tb6_lock);
1139 pcpu_rt = rt6_make_pcpu_route(rt);
1140 dst_release(&rt->dst);
1148 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1149 struct flowi6 *fl6, int flags)
1151 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1154 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1155 struct net_device *dev,
1156 struct flowi6 *fl6, int flags)
1158 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1159 flags |= RT6_LOOKUP_F_IFACE;
1161 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1164 void ip6_route_input(struct sk_buff *skb)
1166 const struct ipv6hdr *iph = ipv6_hdr(skb);
1167 struct net *net = dev_net(skb->dev);
1168 int flags = RT6_LOOKUP_F_HAS_SADDR;
1169 struct ip_tunnel_info *tun_info;
1170 struct flowi6 fl6 = {
1171 .flowi6_iif = skb->dev->ifindex,
1172 .daddr = iph->daddr,
1173 .saddr = iph->saddr,
1174 .flowlabel = ip6_flowinfo(iph),
1175 .flowi6_mark = skb->mark,
1176 .flowi6_proto = iph->nexthdr,
1179 tun_info = skb_tunnel_info(skb);
1180 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1181 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1183 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1186 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1187 struct flowi6 *fl6, int flags)
1189 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1192 struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
1197 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1199 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1201 flags |= RT6_LOOKUP_F_IFACE;
1203 if (!ipv6_addr_any(&fl6->saddr))
1204 flags |= RT6_LOOKUP_F_HAS_SADDR;
1206 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1208 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1210 EXPORT_SYMBOL(ip6_route_output);
1212 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1214 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1215 struct dst_entry *new = NULL;
1217 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1223 new->input = dst_discard;
1224 new->output = dst_discard_sk;
1226 dst_copy_metrics(new, &ort->dst);
1227 rt->rt6i_idev = ort->rt6i_idev;
1229 in6_dev_hold(rt->rt6i_idev);
1231 rt->rt6i_gateway = ort->rt6i_gateway;
1232 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1233 rt->rt6i_metric = 0;
1235 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1236 #ifdef CONFIG_IPV6_SUBTREES
1237 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1243 dst_release(dst_orig);
1244 return new ? new : ERR_PTR(-ENOMEM);
1248 * Destination cache support functions
1251 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1254 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1255 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1258 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1260 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1263 if (rt6_check_expired(rt))
1269 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1271 if (rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1272 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1278 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1280 struct rt6_info *rt;
1282 rt = (struct rt6_info *) dst;
1284 /* All IPV6 dsts are created with ->obsolete set to the value
1285 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1286 * into this function always.
1289 rt6_dst_from_metrics_check(rt);
1291 if ((rt->rt6i_flags & RTF_PCPU) || unlikely(dst->flags & DST_NOCACHE))
1292 return rt6_dst_from_check(rt, cookie);
1294 return rt6_check(rt, cookie);
1297 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1299 struct rt6_info *rt = (struct rt6_info *) dst;
1302 if (rt->rt6i_flags & RTF_CACHE) {
1303 if (rt6_check_expired(rt)) {
1315 static void ip6_link_failure(struct sk_buff *skb)
1317 struct rt6_info *rt;
1319 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1321 rt = (struct rt6_info *) skb_dst(skb);
1323 if (rt->rt6i_flags & RTF_CACHE) {
1326 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1327 rt->rt6i_node->fn_sernum = -1;
1332 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1334 struct net *net = dev_net(rt->dst.dev);
1336 rt->rt6i_flags |= RTF_MODIFIED;
1337 rt->rt6i_pmtu = mtu;
1338 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1341 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1342 const struct ipv6hdr *iph, u32 mtu)
1344 struct rt6_info *rt6 = (struct rt6_info *)dst;
1346 if (rt6->rt6i_flags & RTF_LOCAL)
1350 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1351 if (mtu >= dst_mtu(dst))
1354 if (rt6->rt6i_flags & RTF_CACHE) {
1355 rt6_do_update_pmtu(rt6, mtu);
1357 const struct in6_addr *daddr, *saddr;
1358 struct rt6_info *nrt6;
1361 daddr = &iph->daddr;
1362 saddr = &iph->saddr;
1364 daddr = &sk->sk_v6_daddr;
1365 saddr = &inet6_sk(sk)->saddr;
1369 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1371 rt6_do_update_pmtu(nrt6, mtu);
1373 /* ip6_ins_rt(nrt6) will bump the
1374 * rt6->rt6i_node->fn_sernum
1375 * which will fail the next rt6_check() and
1376 * invalidate the sk->sk_dst_cache.
1383 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1384 struct sk_buff *skb, u32 mtu)
1386 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1389 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1392 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1393 struct dst_entry *dst;
1396 memset(&fl6, 0, sizeof(fl6));
1397 fl6.flowi6_oif = oif;
1398 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1399 fl6.daddr = iph->daddr;
1400 fl6.saddr = iph->saddr;
1401 fl6.flowlabel = ip6_flowinfo(iph);
1403 dst = ip6_route_output(net, NULL, &fl6);
1405 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1408 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1410 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1412 ip6_update_pmtu(skb, sock_net(sk), mtu,
1413 sk->sk_bound_dev_if, sk->sk_mark);
1415 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1417 /* Handle redirects */
1418 struct ip6rd_flowi {
1420 struct in6_addr gateway;
1423 static struct rt6_info *__ip6_route_redirect(struct net *net,
1424 struct fib6_table *table,
1428 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1429 struct rt6_info *rt;
1430 struct fib6_node *fn;
1432 /* Get the "current" route for this destination and
1433 * check if the redirect has come from approriate router.
1435 * RFC 4861 specifies that redirects should only be
1436 * accepted if they come from the nexthop to the target.
1437 * Due to the way the routes are chosen, this notion
1438 * is a bit fuzzy and one might need to check all possible
1442 read_lock_bh(&table->tb6_lock);
1443 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1445 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1446 if (rt6_check_expired(rt))
1450 if (!(rt->rt6i_flags & RTF_GATEWAY))
1452 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1454 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1460 rt = net->ipv6.ip6_null_entry;
1461 else if (rt->dst.error) {
1462 rt = net->ipv6.ip6_null_entry;
1466 if (rt == net->ipv6.ip6_null_entry) {
1467 fn = fib6_backtrack(fn, &fl6->saddr);
1475 read_unlock_bh(&table->tb6_lock);
1480 static struct dst_entry *ip6_route_redirect(struct net *net,
1481 const struct flowi6 *fl6,
1482 const struct in6_addr *gateway)
1484 int flags = RT6_LOOKUP_F_HAS_SADDR;
1485 struct ip6rd_flowi rdfl;
1488 rdfl.gateway = *gateway;
1490 return fib6_rule_lookup(net, &rdfl.fl6,
1491 flags, __ip6_route_redirect);
1494 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1496 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1497 struct dst_entry *dst;
1500 memset(&fl6, 0, sizeof(fl6));
1501 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1502 fl6.flowi6_oif = oif;
1503 fl6.flowi6_mark = mark;
1504 fl6.daddr = iph->daddr;
1505 fl6.saddr = iph->saddr;
1506 fl6.flowlabel = ip6_flowinfo(iph);
1508 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1509 rt6_do_redirect(dst, NULL, skb);
1512 EXPORT_SYMBOL_GPL(ip6_redirect);
1514 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1517 const struct ipv6hdr *iph = ipv6_hdr(skb);
1518 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1519 struct dst_entry *dst;
1522 memset(&fl6, 0, sizeof(fl6));
1523 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1524 fl6.flowi6_oif = oif;
1525 fl6.flowi6_mark = mark;
1526 fl6.daddr = msg->dest;
1527 fl6.saddr = iph->daddr;
1529 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1530 rt6_do_redirect(dst, NULL, skb);
1534 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1536 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_FEATURES && (val & ~RTAX_FEATURE_MASK))
1735 __set_bit(type - 1, mxc->mx_valid);
1739 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1740 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1750 int ip6_route_info_create(struct fib6_config *cfg, struct rt6_info **rt_ret)
1753 struct net *net = cfg->fc_nlinfo.nl_net;
1754 struct rt6_info *rt = NULL;
1755 struct net_device *dev = NULL;
1756 struct inet6_dev *idev = NULL;
1757 struct fib6_table *table;
1760 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1762 #ifndef CONFIG_IPV6_SUBTREES
1763 if (cfg->fc_src_len)
1766 if (cfg->fc_ifindex) {
1768 dev = dev_get_by_index(net, cfg->fc_ifindex);
1771 idev = in6_dev_get(dev);
1776 if (cfg->fc_metric == 0)
1777 cfg->fc_metric = IP6_RT_PRIO_USER;
1780 if (cfg->fc_nlinfo.nlh &&
1781 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1782 table = fib6_get_table(net, cfg->fc_table);
1784 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1785 table = fib6_new_table(net, cfg->fc_table);
1788 table = fib6_new_table(net, cfg->fc_table);
1794 rt = ip6_dst_alloc(net, NULL,
1795 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1802 if (cfg->fc_flags & RTF_EXPIRES)
1803 rt6_set_expires(rt, jiffies +
1804 clock_t_to_jiffies(cfg->fc_expires));
1806 rt6_clean_expires(rt);
1808 if (cfg->fc_protocol == RTPROT_UNSPEC)
1809 cfg->fc_protocol = RTPROT_BOOT;
1810 rt->rt6i_protocol = cfg->fc_protocol;
1812 addr_type = ipv6_addr_type(&cfg->fc_dst);
1814 if (addr_type & IPV6_ADDR_MULTICAST)
1815 rt->dst.input = ip6_mc_input;
1816 else if (cfg->fc_flags & RTF_LOCAL)
1817 rt->dst.input = ip6_input;
1819 rt->dst.input = ip6_forward;
1821 rt->dst.output = ip6_output;
1823 if (cfg->fc_encap) {
1824 struct lwtunnel_state *lwtstate;
1826 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1827 cfg->fc_encap, AF_INET6, cfg,
1831 rt->dst.lwtstate = lwtstate_get(lwtstate);
1832 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1833 rt->dst.lwtstate->orig_output = rt->dst.output;
1834 rt->dst.output = lwtunnel_output;
1836 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1837 rt->dst.lwtstate->orig_input = rt->dst.input;
1838 rt->dst.input = lwtunnel_input;
1842 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1843 rt->rt6i_dst.plen = cfg->fc_dst_len;
1844 if (rt->rt6i_dst.plen == 128)
1845 rt->dst.flags |= DST_HOST;
1847 #ifdef CONFIG_IPV6_SUBTREES
1848 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1849 rt->rt6i_src.plen = cfg->fc_src_len;
1852 rt->rt6i_metric = cfg->fc_metric;
1854 /* We cannot add true routes via loopback here,
1855 they would result in kernel looping; promote them to reject routes
1857 if ((cfg->fc_flags & RTF_REJECT) ||
1858 (dev && (dev->flags & IFF_LOOPBACK) &&
1859 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1860 !(cfg->fc_flags & RTF_LOCAL))) {
1861 /* hold loopback dev/idev if we haven't done so. */
1862 if (dev != net->loopback_dev) {
1867 dev = net->loopback_dev;
1869 idev = in6_dev_get(dev);
1875 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1876 switch (cfg->fc_type) {
1878 rt->dst.error = -EINVAL;
1879 rt->dst.output = dst_discard_sk;
1880 rt->dst.input = dst_discard;
1883 rt->dst.error = -EACCES;
1884 rt->dst.output = ip6_pkt_prohibit_out;
1885 rt->dst.input = ip6_pkt_prohibit;
1888 case RTN_UNREACHABLE:
1890 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1891 : (cfg->fc_type == RTN_UNREACHABLE)
1892 ? -EHOSTUNREACH : -ENETUNREACH;
1893 rt->dst.output = ip6_pkt_discard_out;
1894 rt->dst.input = ip6_pkt_discard;
1900 if (cfg->fc_flags & RTF_GATEWAY) {
1901 const struct in6_addr *gw_addr;
1904 gw_addr = &cfg->fc_gateway;
1905 gwa_type = ipv6_addr_type(gw_addr);
1907 /* if gw_addr is local we will fail to detect this in case
1908 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1909 * will return already-added prefix route via interface that
1910 * prefix route was assigned to, which might be non-loopback.
1913 if (ipv6_chk_addr_and_flags(net, gw_addr,
1914 gwa_type & IPV6_ADDR_LINKLOCAL ?
1918 rt->rt6i_gateway = *gw_addr;
1920 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1921 struct rt6_info *grt;
1923 /* IPv6 strictly inhibits using not link-local
1924 addresses as nexthop address.
1925 Otherwise, router will not able to send redirects.
1926 It is very good, but in some (rare!) circumstances
1927 (SIT, PtP, NBMA NOARP links) it is handy to allow
1928 some exceptions. --ANK
1930 if (!(gwa_type & IPV6_ADDR_UNICAST))
1933 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1935 err = -EHOSTUNREACH;
1939 if (dev != grt->dst.dev) {
1945 idev = grt->rt6i_idev;
1947 in6_dev_hold(grt->rt6i_idev);
1949 if (!(grt->rt6i_flags & RTF_GATEWAY))
1957 if (!dev || (dev->flags & IFF_LOOPBACK))
1965 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1966 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1970 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1971 rt->rt6i_prefsrc.plen = 128;
1973 rt->rt6i_prefsrc.plen = 0;
1975 rt->rt6i_flags = cfg->fc_flags;
1979 rt->rt6i_idev = idev;
1980 rt->rt6i_table = table;
1982 cfg->fc_nlinfo.nl_net = dev_net(dev);
2000 int ip6_route_add(struct fib6_config *cfg)
2002 struct mx6_config mxc = { .mx = NULL, };
2003 struct rt6_info *rt = NULL;
2006 err = ip6_route_info_create(cfg, &rt);
2010 err = ip6_convert_metrics(&mxc, cfg);
2014 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2026 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2029 struct fib6_table *table;
2030 struct net *net = dev_net(rt->dst.dev);
2032 if (rt == net->ipv6.ip6_null_entry ||
2033 rt->dst.flags & DST_NOCACHE) {
2038 table = rt->rt6i_table;
2039 write_lock_bh(&table->tb6_lock);
2040 err = fib6_del(rt, info);
2041 write_unlock_bh(&table->tb6_lock);
2048 int ip6_del_rt(struct rt6_info *rt)
2050 struct nl_info info = {
2051 .nl_net = dev_net(rt->dst.dev),
2053 return __ip6_del_rt(rt, &info);
2056 static int ip6_route_del(struct fib6_config *cfg)
2058 struct fib6_table *table;
2059 struct fib6_node *fn;
2060 struct rt6_info *rt;
2063 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2067 read_lock_bh(&table->tb6_lock);
2069 fn = fib6_locate(&table->tb6_root,
2070 &cfg->fc_dst, cfg->fc_dst_len,
2071 &cfg->fc_src, cfg->fc_src_len);
2074 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2075 if ((rt->rt6i_flags & RTF_CACHE) &&
2076 !(cfg->fc_flags & RTF_CACHE))
2078 if (cfg->fc_ifindex &&
2080 rt->dst.dev->ifindex != cfg->fc_ifindex))
2082 if (cfg->fc_flags & RTF_GATEWAY &&
2083 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2085 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2088 read_unlock_bh(&table->tb6_lock);
2090 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2093 read_unlock_bh(&table->tb6_lock);
2098 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2100 struct net *net = dev_net(skb->dev);
2101 struct netevent_redirect netevent;
2102 struct rt6_info *rt, *nrt = NULL;
2103 struct ndisc_options ndopts;
2104 struct inet6_dev *in6_dev;
2105 struct neighbour *neigh;
2107 int optlen, on_link;
2110 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2111 optlen -= sizeof(*msg);
2114 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2118 msg = (struct rd_msg *)icmp6_hdr(skb);
2120 if (ipv6_addr_is_multicast(&msg->dest)) {
2121 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2126 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2128 } else if (ipv6_addr_type(&msg->target) !=
2129 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2130 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2134 in6_dev = __in6_dev_get(skb->dev);
2137 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2141 * The IP source address of the Redirect MUST be the same as the current
2142 * first-hop router for the specified ICMP Destination Address.
2145 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2146 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2151 if (ndopts.nd_opts_tgt_lladdr) {
2152 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2155 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2160 rt = (struct rt6_info *) dst;
2161 if (rt == net->ipv6.ip6_null_entry) {
2162 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2166 /* Redirect received -> path was valid.
2167 * Look, redirects are sent only in response to data packets,
2168 * so that this nexthop apparently is reachable. --ANK
2170 dst_confirm(&rt->dst);
2172 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2177 * We have finally decided to accept it.
2180 neigh_update(neigh, lladdr, NUD_STALE,
2181 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2182 NEIGH_UPDATE_F_OVERRIDE|
2183 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2184 NEIGH_UPDATE_F_ISROUTER))
2187 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2191 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2193 nrt->rt6i_flags &= ~RTF_GATEWAY;
2195 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2197 if (ip6_ins_rt(nrt))
2200 netevent.old = &rt->dst;
2201 netevent.new = &nrt->dst;
2202 netevent.daddr = &msg->dest;
2203 netevent.neigh = neigh;
2204 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2206 if (rt->rt6i_flags & RTF_CACHE) {
2207 rt = (struct rt6_info *) dst_clone(&rt->dst);
2212 neigh_release(neigh);
2216 * Misc support functions
2219 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2221 BUG_ON(from->dst.from);
2223 rt->rt6i_flags &= ~RTF_EXPIRES;
2224 dst_hold(&from->dst);
2225 rt->dst.from = &from->dst;
2226 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2229 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2231 rt->dst.input = ort->dst.input;
2232 rt->dst.output = ort->dst.output;
2233 rt->rt6i_dst = ort->rt6i_dst;
2234 rt->dst.error = ort->dst.error;
2235 rt->rt6i_idev = ort->rt6i_idev;
2237 in6_dev_hold(rt->rt6i_idev);
2238 rt->dst.lastuse = jiffies;
2239 rt->rt6i_gateway = ort->rt6i_gateway;
2240 rt->rt6i_flags = ort->rt6i_flags;
2241 rt6_set_from(rt, ort);
2242 rt->rt6i_metric = ort->rt6i_metric;
2243 #ifdef CONFIG_IPV6_SUBTREES
2244 rt->rt6i_src = ort->rt6i_src;
2246 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2247 rt->rt6i_table = ort->rt6i_table;
2248 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2251 #ifdef CONFIG_IPV6_ROUTE_INFO
2252 static struct rt6_info *rt6_get_route_info(struct net *net,
2253 const struct in6_addr *prefix, int prefixlen,
2254 const struct in6_addr *gwaddr, int ifindex)
2256 struct fib6_node *fn;
2257 struct rt6_info *rt = NULL;
2258 struct fib6_table *table;
2260 table = fib6_get_table(net, RT6_TABLE_INFO);
2264 read_lock_bh(&table->tb6_lock);
2265 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2269 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2270 if (rt->dst.dev->ifindex != ifindex)
2272 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2274 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2280 read_unlock_bh(&table->tb6_lock);
2284 static struct rt6_info *rt6_add_route_info(struct net *net,
2285 const struct in6_addr *prefix, int prefixlen,
2286 const struct in6_addr *gwaddr, int ifindex,
2289 struct fib6_config cfg = {
2290 .fc_table = RT6_TABLE_INFO,
2291 .fc_metric = IP6_RT_PRIO_USER,
2292 .fc_ifindex = ifindex,
2293 .fc_dst_len = prefixlen,
2294 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2295 RTF_UP | RTF_PREF(pref),
2296 .fc_nlinfo.portid = 0,
2297 .fc_nlinfo.nlh = NULL,
2298 .fc_nlinfo.nl_net = net,
2301 cfg.fc_dst = *prefix;
2302 cfg.fc_gateway = *gwaddr;
2304 /* We should treat it as a default route if prefix length is 0. */
2306 cfg.fc_flags |= RTF_DEFAULT;
2308 ip6_route_add(&cfg);
2310 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2314 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2316 struct rt6_info *rt;
2317 struct fib6_table *table;
2319 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2323 read_lock_bh(&table->tb6_lock);
2324 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2325 if (dev == rt->dst.dev &&
2326 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2327 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2332 read_unlock_bh(&table->tb6_lock);
2336 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2337 struct net_device *dev,
2340 struct fib6_config cfg = {
2341 .fc_table = RT6_TABLE_DFLT,
2342 .fc_metric = IP6_RT_PRIO_USER,
2343 .fc_ifindex = dev->ifindex,
2344 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2345 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2346 .fc_nlinfo.portid = 0,
2347 .fc_nlinfo.nlh = NULL,
2348 .fc_nlinfo.nl_net = dev_net(dev),
2351 cfg.fc_gateway = *gwaddr;
2353 ip6_route_add(&cfg);
2355 return rt6_get_dflt_router(gwaddr, dev);
2358 void rt6_purge_dflt_routers(struct net *net)
2360 struct rt6_info *rt;
2361 struct fib6_table *table;
2363 /* NOTE: Keep consistent with rt6_get_dflt_router */
2364 table = fib6_get_table(net, RT6_TABLE_DFLT);
2369 read_lock_bh(&table->tb6_lock);
2370 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2371 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2372 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2374 read_unlock_bh(&table->tb6_lock);
2379 read_unlock_bh(&table->tb6_lock);
2382 static void rtmsg_to_fib6_config(struct net *net,
2383 struct in6_rtmsg *rtmsg,
2384 struct fib6_config *cfg)
2386 memset(cfg, 0, sizeof(*cfg));
2388 cfg->fc_table = RT6_TABLE_MAIN;
2389 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2390 cfg->fc_metric = rtmsg->rtmsg_metric;
2391 cfg->fc_expires = rtmsg->rtmsg_info;
2392 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2393 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2394 cfg->fc_flags = rtmsg->rtmsg_flags;
2396 cfg->fc_nlinfo.nl_net = net;
2398 cfg->fc_dst = rtmsg->rtmsg_dst;
2399 cfg->fc_src = rtmsg->rtmsg_src;
2400 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2403 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2405 struct fib6_config cfg;
2406 struct in6_rtmsg rtmsg;
2410 case SIOCADDRT: /* Add a route */
2411 case SIOCDELRT: /* Delete a route */
2412 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2414 err = copy_from_user(&rtmsg, arg,
2415 sizeof(struct in6_rtmsg));
2419 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2424 err = ip6_route_add(&cfg);
2427 err = ip6_route_del(&cfg);
2441 * Drop the packet on the floor
2444 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2447 struct dst_entry *dst = skb_dst(skb);
2448 switch (ipstats_mib_noroutes) {
2449 case IPSTATS_MIB_INNOROUTES:
2450 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2451 if (type == IPV6_ADDR_ANY) {
2452 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2453 IPSTATS_MIB_INADDRERRORS);
2457 case IPSTATS_MIB_OUTNOROUTES:
2458 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2459 ipstats_mib_noroutes);
2462 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2467 static int ip6_pkt_discard(struct sk_buff *skb)
2469 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2472 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2474 skb->dev = skb_dst(skb)->dev;
2475 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2478 static int ip6_pkt_prohibit(struct sk_buff *skb)
2480 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2483 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2485 skb->dev = skb_dst(skb)->dev;
2486 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2490 * Allocate a dst for local (unicast / anycast) address.
2493 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2494 const struct in6_addr *addr,
2497 struct net *net = dev_net(idev->dev);
2498 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2501 return ERR_PTR(-ENOMEM);
2505 rt->dst.flags |= DST_HOST;
2506 rt->dst.input = ip6_input;
2507 rt->dst.output = ip6_output;
2508 rt->rt6i_idev = idev;
2510 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2512 rt->rt6i_flags |= RTF_ANYCAST;
2514 rt->rt6i_flags |= RTF_LOCAL;
2516 rt->rt6i_gateway = *addr;
2517 rt->rt6i_dst.addr = *addr;
2518 rt->rt6i_dst.plen = 128;
2519 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2520 rt->dst.flags |= DST_NOCACHE;
2522 atomic_set(&rt->dst.__refcnt, 1);
2527 int ip6_route_get_saddr(struct net *net,
2528 struct rt6_info *rt,
2529 const struct in6_addr *daddr,
2531 struct in6_addr *saddr)
2533 struct inet6_dev *idev =
2534 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2536 if (rt && rt->rt6i_prefsrc.plen)
2537 *saddr = rt->rt6i_prefsrc.addr;
2539 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2540 daddr, prefs, saddr);
2544 /* remove deleted ip from prefsrc entries */
2545 struct arg_dev_net_ip {
2546 struct net_device *dev;
2548 struct in6_addr *addr;
2551 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2553 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2554 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2555 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2557 if (((void *)rt->dst.dev == dev || !dev) &&
2558 rt != net->ipv6.ip6_null_entry &&
2559 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2560 /* remove prefsrc entry */
2561 rt->rt6i_prefsrc.plen = 0;
2566 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2568 struct net *net = dev_net(ifp->idev->dev);
2569 struct arg_dev_net_ip adni = {
2570 .dev = ifp->idev->dev,
2574 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2577 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2578 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2580 /* Remove routers and update dst entries when gateway turn into host. */
2581 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2583 struct in6_addr *gateway = (struct in6_addr *)arg;
2585 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2586 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2587 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2593 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2595 fib6_clean_all(net, fib6_clean_tohost, gateway);
2598 struct arg_dev_net {
2599 struct net_device *dev;
2603 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2605 const struct arg_dev_net *adn = arg;
2606 const struct net_device *dev = adn->dev;
2608 if ((rt->dst.dev == dev || !dev) &&
2609 rt != adn->net->ipv6.ip6_null_entry)
2615 void rt6_ifdown(struct net *net, struct net_device *dev)
2617 struct arg_dev_net adn = {
2622 fib6_clean_all(net, fib6_ifdown, &adn);
2623 icmp6_clean_all(fib6_ifdown, &adn);
2625 rt6_uncached_list_flush_dev(net, dev);
2628 struct rt6_mtu_change_arg {
2629 struct net_device *dev;
2633 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2635 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2636 struct inet6_dev *idev;
2638 /* In IPv6 pmtu discovery is not optional,
2639 so that RTAX_MTU lock cannot disable it.
2640 We still use this lock to block changes
2641 caused by addrconf/ndisc.
2644 idev = __in6_dev_get(arg->dev);
2648 /* For administrative MTU increase, there is no way to discover
2649 IPv6 PMTU increase, so PMTU increase should be updated here.
2650 Since RFC 1981 doesn't include administrative MTU increase
2651 update PMTU increase is a MUST. (i.e. jumbo frame)
2654 If new MTU is less than route PMTU, this new MTU will be the
2655 lowest MTU in the path, update the route PMTU to reflect PMTU
2656 decreases; if new MTU is greater than route PMTU, and the
2657 old MTU is the lowest MTU in the path, update the route PMTU
2658 to reflect the increase. In this case if the other nodes' MTU
2659 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2662 if (rt->dst.dev == arg->dev &&
2663 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2664 if (rt->rt6i_flags & RTF_CACHE) {
2665 /* For RTF_CACHE with rt6i_pmtu == 0
2666 * (i.e. a redirected route),
2667 * the metrics of its rt->dst.from has already
2670 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2671 rt->rt6i_pmtu = arg->mtu;
2672 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2673 (dst_mtu(&rt->dst) < arg->mtu &&
2674 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2675 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2681 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2683 struct rt6_mtu_change_arg arg = {
2688 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2691 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2692 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2693 [RTA_OIF] = { .type = NLA_U32 },
2694 [RTA_IIF] = { .type = NLA_U32 },
2695 [RTA_PRIORITY] = { .type = NLA_U32 },
2696 [RTA_METRICS] = { .type = NLA_NESTED },
2697 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2698 [RTA_PREF] = { .type = NLA_U8 },
2699 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2700 [RTA_ENCAP] = { .type = NLA_NESTED },
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 err = ip6_route_info_create(&r_cfg, &rt);
2902 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2908 rtnh = rtnh_next(rtnh, &remaining);
2912 list_for_each_entry(nh, &rt6_nh_list, next) {
2913 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2914 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2915 nh->rt6_info = NULL;
2918 ip6_print_replace_route_err(&rt6_nh_list);
2923 /* Because each route is added like a single route we remove
2924 * these flags after the first nexthop: if there is a collision,
2925 * we have already failed to add the first nexthop:
2926 * fib6_add_rt2node() has rejected it; when replacing, old
2927 * nexthops have been replaced by first new, the rest should
2930 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2938 /* Delete routes that were already added */
2939 list_for_each_entry(nh, &rt6_nh_list, next) {
2942 ip6_route_del(&nh->r_cfg);
2946 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2948 dst_free(&nh->rt6_info->dst);
2950 list_del(&nh->next);
2957 static int ip6_route_multipath_del(struct fib6_config *cfg)
2959 struct fib6_config r_cfg;
2960 struct rtnexthop *rtnh;
2963 int err = 1, last_err = 0;
2965 remaining = cfg->fc_mp_len;
2966 rtnh = (struct rtnexthop *)cfg->fc_mp;
2968 /* Parse a Multipath Entry */
2969 while (rtnh_ok(rtnh, remaining)) {
2970 memcpy(&r_cfg, cfg, sizeof(*cfg));
2971 if (rtnh->rtnh_ifindex)
2972 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2974 attrlen = rtnh_attrlen(rtnh);
2976 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2978 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2980 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2981 r_cfg.fc_flags |= RTF_GATEWAY;
2984 err = ip6_route_del(&r_cfg);
2988 rtnh = rtnh_next(rtnh, &remaining);
2994 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2996 struct fib6_config cfg;
2999 err = rtm_to_fib6_config(skb, nlh, &cfg);
3004 return ip6_route_multipath_del(&cfg);
3006 return ip6_route_del(&cfg);
3009 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3011 struct fib6_config cfg;
3014 err = rtm_to_fib6_config(skb, nlh, &cfg);
3019 return ip6_route_multipath_add(&cfg);
3021 return ip6_route_add(&cfg);
3024 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3026 return NLMSG_ALIGN(sizeof(struct rtmsg))
3027 + nla_total_size(16) /* RTA_SRC */
3028 + nla_total_size(16) /* RTA_DST */
3029 + nla_total_size(16) /* RTA_GATEWAY */
3030 + nla_total_size(16) /* RTA_PREFSRC */
3031 + nla_total_size(4) /* RTA_TABLE */
3032 + nla_total_size(4) /* RTA_IIF */
3033 + nla_total_size(4) /* RTA_OIF */
3034 + nla_total_size(4) /* RTA_PRIORITY */
3035 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3036 + nla_total_size(sizeof(struct rta_cacheinfo))
3037 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3038 + nla_total_size(1) /* RTA_PREF */
3039 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3042 static int rt6_fill_node(struct net *net,
3043 struct sk_buff *skb, struct rt6_info *rt,
3044 struct in6_addr *dst, struct in6_addr *src,
3045 int iif, int type, u32 portid, u32 seq,
3046 int prefix, int nowait, unsigned int flags)
3048 u32 metrics[RTAX_MAX];
3050 struct nlmsghdr *nlh;
3054 if (prefix) { /* user wants prefix routes only */
3055 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3056 /* success since this is not a prefix route */
3061 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3065 rtm = nlmsg_data(nlh);
3066 rtm->rtm_family = AF_INET6;
3067 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3068 rtm->rtm_src_len = rt->rt6i_src.plen;
3071 table = rt->rt6i_table->tb6_id;
3073 table = RT6_TABLE_UNSPEC;
3074 rtm->rtm_table = table;
3075 if (nla_put_u32(skb, RTA_TABLE, table))
3076 goto nla_put_failure;
3077 if (rt->rt6i_flags & RTF_REJECT) {
3078 switch (rt->dst.error) {
3080 rtm->rtm_type = RTN_BLACKHOLE;
3083 rtm->rtm_type = RTN_PROHIBIT;
3086 rtm->rtm_type = RTN_THROW;
3089 rtm->rtm_type = RTN_UNREACHABLE;
3093 else if (rt->rt6i_flags & RTF_LOCAL)
3094 rtm->rtm_type = RTN_LOCAL;
3095 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3096 rtm->rtm_type = RTN_LOCAL;
3098 rtm->rtm_type = RTN_UNICAST;
3100 if (!netif_carrier_ok(rt->dst.dev)) {
3101 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3102 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3103 rtm->rtm_flags |= RTNH_F_DEAD;
3105 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3106 rtm->rtm_protocol = rt->rt6i_protocol;
3107 if (rt->rt6i_flags & RTF_DYNAMIC)
3108 rtm->rtm_protocol = RTPROT_REDIRECT;
3109 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3110 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3111 rtm->rtm_protocol = RTPROT_RA;
3113 rtm->rtm_protocol = RTPROT_KERNEL;
3116 if (rt->rt6i_flags & RTF_CACHE)
3117 rtm->rtm_flags |= RTM_F_CLONED;
3120 if (nla_put_in6_addr(skb, RTA_DST, dst))
3121 goto nla_put_failure;
3122 rtm->rtm_dst_len = 128;
3123 } else if (rtm->rtm_dst_len)
3124 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3125 goto nla_put_failure;
3126 #ifdef CONFIG_IPV6_SUBTREES
3128 if (nla_put_in6_addr(skb, RTA_SRC, src))
3129 goto nla_put_failure;
3130 rtm->rtm_src_len = 128;
3131 } else if (rtm->rtm_src_len &&
3132 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3133 goto nla_put_failure;
3136 #ifdef CONFIG_IPV6_MROUTE
3137 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3138 int err = ip6mr_get_route(net, skb, rtm, nowait);
3143 goto nla_put_failure;
3145 if (err == -EMSGSIZE)
3146 goto nla_put_failure;
3151 if (nla_put_u32(skb, RTA_IIF, iif))
3152 goto nla_put_failure;
3154 struct in6_addr saddr_buf;
3155 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3156 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3157 goto nla_put_failure;
3160 if (rt->rt6i_prefsrc.plen) {
3161 struct in6_addr saddr_buf;
3162 saddr_buf = rt->rt6i_prefsrc.addr;
3163 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3164 goto nla_put_failure;
3167 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3169 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3170 if (rtnetlink_put_metrics(skb, metrics) < 0)
3171 goto nla_put_failure;
3173 if (rt->rt6i_flags & RTF_GATEWAY) {
3174 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3175 goto nla_put_failure;
3179 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3180 goto nla_put_failure;
3181 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3182 goto nla_put_failure;
3184 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3186 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3187 goto nla_put_failure;
3189 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3190 goto nla_put_failure;
3192 lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3194 nlmsg_end(skb, nlh);
3198 nlmsg_cancel(skb, nlh);
3202 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3204 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3207 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3208 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3209 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3213 return rt6_fill_node(arg->net,
3214 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3215 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3216 prefix, 0, NLM_F_MULTI);
3219 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3221 struct net *net = sock_net(in_skb->sk);
3222 struct nlattr *tb[RTA_MAX+1];
3223 struct rt6_info *rt;
3224 struct sk_buff *skb;
3227 int err, iif = 0, oif = 0;
3229 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3234 memset(&fl6, 0, sizeof(fl6));
3237 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3240 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3244 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3247 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3251 iif = nla_get_u32(tb[RTA_IIF]);
3254 oif = nla_get_u32(tb[RTA_OIF]);
3257 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3260 struct net_device *dev;
3263 dev = __dev_get_by_index(net, iif);
3269 fl6.flowi6_iif = iif;
3271 if (!ipv6_addr_any(&fl6.saddr))
3272 flags |= RT6_LOOKUP_F_HAS_SADDR;
3274 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3277 fl6.flowi6_oif = oif;
3279 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3282 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3289 /* Reserve room for dummy headers, this skb can pass
3290 through good chunk of routing engine.
3292 skb_reset_mac_header(skb);
3293 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3295 skb_dst_set(skb, &rt->dst);
3297 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3298 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3299 nlh->nlmsg_seq, 0, 0, 0);
3305 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3310 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3311 unsigned int nlm_flags)
3313 struct sk_buff *skb;
3314 struct net *net = info->nl_net;
3319 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3321 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3325 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3326 event, info->portid, seq, 0, 0, nlm_flags);
3328 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3329 WARN_ON(err == -EMSGSIZE);
3333 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3334 info->nlh, gfp_any());
3338 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3341 static int ip6_route_dev_notify(struct notifier_block *this,
3342 unsigned long event, void *ptr)
3344 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3345 struct net *net = dev_net(dev);
3347 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3348 net->ipv6.ip6_null_entry->dst.dev = dev;
3349 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3350 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3351 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3352 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3353 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3354 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3365 #ifdef CONFIG_PROC_FS
3367 static const struct file_operations ipv6_route_proc_fops = {
3368 .owner = THIS_MODULE,
3369 .open = ipv6_route_open,
3371 .llseek = seq_lseek,
3372 .release = seq_release_net,
3375 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3377 struct net *net = (struct net *)seq->private;
3378 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3379 net->ipv6.rt6_stats->fib_nodes,
3380 net->ipv6.rt6_stats->fib_route_nodes,
3381 net->ipv6.rt6_stats->fib_rt_alloc,
3382 net->ipv6.rt6_stats->fib_rt_entries,
3383 net->ipv6.rt6_stats->fib_rt_cache,
3384 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3385 net->ipv6.rt6_stats->fib_discarded_routes);
3390 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3392 return single_open_net(inode, file, rt6_stats_seq_show);
3395 static const struct file_operations rt6_stats_seq_fops = {
3396 .owner = THIS_MODULE,
3397 .open = rt6_stats_seq_open,
3399 .llseek = seq_lseek,
3400 .release = single_release_net,
3402 #endif /* CONFIG_PROC_FS */
3404 #ifdef CONFIG_SYSCTL
3407 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3408 void __user *buffer, size_t *lenp, loff_t *ppos)
3415 net = (struct net *)ctl->extra1;
3416 delay = net->ipv6.sysctl.flush_delay;
3417 proc_dointvec(ctl, write, buffer, lenp, ppos);
3418 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3422 struct ctl_table ipv6_route_table_template[] = {
3424 .procname = "flush",
3425 .data = &init_net.ipv6.sysctl.flush_delay,
3426 .maxlen = sizeof(int),
3428 .proc_handler = ipv6_sysctl_rtcache_flush
3431 .procname = "gc_thresh",
3432 .data = &ip6_dst_ops_template.gc_thresh,
3433 .maxlen = sizeof(int),
3435 .proc_handler = proc_dointvec,
3438 .procname = "max_size",
3439 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3440 .maxlen = sizeof(int),
3442 .proc_handler = proc_dointvec,
3445 .procname = "gc_min_interval",
3446 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3447 .maxlen = sizeof(int),
3449 .proc_handler = proc_dointvec_jiffies,
3452 .procname = "gc_timeout",
3453 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3454 .maxlen = sizeof(int),
3456 .proc_handler = proc_dointvec_jiffies,
3459 .procname = "gc_interval",
3460 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3461 .maxlen = sizeof(int),
3463 .proc_handler = proc_dointvec_jiffies,
3466 .procname = "gc_elasticity",
3467 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3468 .maxlen = sizeof(int),
3470 .proc_handler = proc_dointvec,
3473 .procname = "mtu_expires",
3474 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3475 .maxlen = sizeof(int),
3477 .proc_handler = proc_dointvec_jiffies,
3480 .procname = "min_adv_mss",
3481 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3482 .maxlen = sizeof(int),
3484 .proc_handler = proc_dointvec,
3487 .procname = "gc_min_interval_ms",
3488 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3489 .maxlen = sizeof(int),
3491 .proc_handler = proc_dointvec_ms_jiffies,
3496 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3498 struct ctl_table *table;
3500 table = kmemdup(ipv6_route_table_template,
3501 sizeof(ipv6_route_table_template),
3505 table[0].data = &net->ipv6.sysctl.flush_delay;
3506 table[0].extra1 = net;
3507 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3508 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3509 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3510 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3511 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3512 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3513 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3514 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3515 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3517 /* Don't export sysctls to unprivileged users */
3518 if (net->user_ns != &init_user_ns)
3519 table[0].procname = NULL;
3526 static int __net_init ip6_route_net_init(struct net *net)
3530 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3531 sizeof(net->ipv6.ip6_dst_ops));
3533 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3534 goto out_ip6_dst_ops;
3536 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3537 sizeof(*net->ipv6.ip6_null_entry),
3539 if (!net->ipv6.ip6_null_entry)
3540 goto out_ip6_dst_entries;
3541 net->ipv6.ip6_null_entry->dst.path =
3542 (struct dst_entry *)net->ipv6.ip6_null_entry;
3543 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3544 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3545 ip6_template_metrics, true);
3547 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3548 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3549 sizeof(*net->ipv6.ip6_prohibit_entry),
3551 if (!net->ipv6.ip6_prohibit_entry)
3552 goto out_ip6_null_entry;
3553 net->ipv6.ip6_prohibit_entry->dst.path =
3554 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3555 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3556 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3557 ip6_template_metrics, true);
3559 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3560 sizeof(*net->ipv6.ip6_blk_hole_entry),
3562 if (!net->ipv6.ip6_blk_hole_entry)
3563 goto out_ip6_prohibit_entry;
3564 net->ipv6.ip6_blk_hole_entry->dst.path =
3565 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3566 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3567 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3568 ip6_template_metrics, true);
3571 net->ipv6.sysctl.flush_delay = 0;
3572 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3573 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3574 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3575 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3576 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3577 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3578 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3580 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3586 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3587 out_ip6_prohibit_entry:
3588 kfree(net->ipv6.ip6_prohibit_entry);
3590 kfree(net->ipv6.ip6_null_entry);
3592 out_ip6_dst_entries:
3593 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3598 static void __net_exit ip6_route_net_exit(struct net *net)
3600 kfree(net->ipv6.ip6_null_entry);
3601 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3602 kfree(net->ipv6.ip6_prohibit_entry);
3603 kfree(net->ipv6.ip6_blk_hole_entry);
3605 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3608 static int __net_init ip6_route_net_init_late(struct net *net)
3610 #ifdef CONFIG_PROC_FS
3611 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3612 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3617 static void __net_exit ip6_route_net_exit_late(struct net *net)
3619 #ifdef CONFIG_PROC_FS
3620 remove_proc_entry("ipv6_route", net->proc_net);
3621 remove_proc_entry("rt6_stats", net->proc_net);
3625 static struct pernet_operations ip6_route_net_ops = {
3626 .init = ip6_route_net_init,
3627 .exit = ip6_route_net_exit,
3630 static int __net_init ipv6_inetpeer_init(struct net *net)
3632 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3636 inet_peer_base_init(bp);
3637 net->ipv6.peers = bp;
3641 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3643 struct inet_peer_base *bp = net->ipv6.peers;
3645 net->ipv6.peers = NULL;
3646 inetpeer_invalidate_tree(bp);
3650 static struct pernet_operations ipv6_inetpeer_ops = {
3651 .init = ipv6_inetpeer_init,
3652 .exit = ipv6_inetpeer_exit,
3655 static struct pernet_operations ip6_route_net_late_ops = {
3656 .init = ip6_route_net_init_late,
3657 .exit = ip6_route_net_exit_late,
3660 static struct notifier_block ip6_route_dev_notifier = {
3661 .notifier_call = ip6_route_dev_notify,
3665 int __init ip6_route_init(void)
3671 ip6_dst_ops_template.kmem_cachep =
3672 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3673 SLAB_HWCACHE_ALIGN, NULL);
3674 if (!ip6_dst_ops_template.kmem_cachep)
3677 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3679 goto out_kmem_cache;
3681 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3683 goto out_dst_entries;
3685 ret = register_pernet_subsys(&ip6_route_net_ops);
3687 goto out_register_inetpeer;
3689 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3691 /* Registering of the loopback is done before this portion of code,
3692 * the loopback reference in rt6_info will not be taken, do it
3693 * manually for init_net */
3694 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3695 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3696 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3697 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3698 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3699 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3700 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3704 goto out_register_subsys;
3710 ret = fib6_rules_init();
3714 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3716 goto fib6_rules_init;
3719 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3720 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3721 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3722 goto out_register_late_subsys;
3724 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3726 goto out_register_late_subsys;
3728 for_each_possible_cpu(cpu) {
3729 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3731 INIT_LIST_HEAD(&ul->head);
3732 spin_lock_init(&ul->lock);
3738 out_register_late_subsys:
3739 unregister_pernet_subsys(&ip6_route_net_late_ops);
3741 fib6_rules_cleanup();
3746 out_register_subsys:
3747 unregister_pernet_subsys(&ip6_route_net_ops);
3748 out_register_inetpeer:
3749 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3751 dst_entries_destroy(&ip6_dst_blackhole_ops);
3753 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3757 void ip6_route_cleanup(void)
3759 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3760 unregister_pernet_subsys(&ip6_route_net_late_ops);
3761 fib6_rules_cleanup();
3764 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3765 unregister_pernet_subsys(&ip6_route_net_ops);
3766 dst_entries_destroy(&ip6_dst_blackhole_ops);
3767 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);