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,
1198 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1200 any_src = ipv6_addr_any(&fl6->saddr);
1201 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1202 (fl6->flowi6_oif && any_src))
1203 flags |= RT6_LOOKUP_F_IFACE;
1206 flags |= RT6_LOOKUP_F_HAS_SADDR;
1208 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1210 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1212 EXPORT_SYMBOL(ip6_route_output);
1214 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1216 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1217 struct dst_entry *new = NULL;
1219 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1225 new->input = dst_discard;
1226 new->output = dst_discard_sk;
1228 dst_copy_metrics(new, &ort->dst);
1229 rt->rt6i_idev = ort->rt6i_idev;
1231 in6_dev_hold(rt->rt6i_idev);
1233 rt->rt6i_gateway = ort->rt6i_gateway;
1234 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1235 rt->rt6i_metric = 0;
1237 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1238 #ifdef CONFIG_IPV6_SUBTREES
1239 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1245 dst_release(dst_orig);
1246 return new ? new : ERR_PTR(-ENOMEM);
1250 * Destination cache support functions
1253 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1256 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1257 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1260 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1262 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1265 if (rt6_check_expired(rt))
1271 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1273 if (rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1274 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1280 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1282 struct rt6_info *rt;
1284 rt = (struct rt6_info *) dst;
1286 /* All IPV6 dsts are created with ->obsolete set to the value
1287 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1288 * into this function always.
1291 rt6_dst_from_metrics_check(rt);
1293 if ((rt->rt6i_flags & RTF_PCPU) || unlikely(dst->flags & DST_NOCACHE))
1294 return rt6_dst_from_check(rt, cookie);
1296 return rt6_check(rt, cookie);
1299 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1301 struct rt6_info *rt = (struct rt6_info *) dst;
1304 if (rt->rt6i_flags & RTF_CACHE) {
1305 if (rt6_check_expired(rt)) {
1317 static void ip6_link_failure(struct sk_buff *skb)
1319 struct rt6_info *rt;
1321 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1323 rt = (struct rt6_info *) skb_dst(skb);
1325 if (rt->rt6i_flags & RTF_CACHE) {
1328 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1329 rt->rt6i_node->fn_sernum = -1;
1334 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1336 struct net *net = dev_net(rt->dst.dev);
1338 rt->rt6i_flags |= RTF_MODIFIED;
1339 rt->rt6i_pmtu = mtu;
1340 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1343 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1344 const struct ipv6hdr *iph, u32 mtu)
1346 struct rt6_info *rt6 = (struct rt6_info *)dst;
1348 if (rt6->rt6i_flags & RTF_LOCAL)
1352 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1353 if (mtu >= dst_mtu(dst))
1356 if (rt6->rt6i_flags & RTF_CACHE) {
1357 rt6_do_update_pmtu(rt6, mtu);
1359 const struct in6_addr *daddr, *saddr;
1360 struct rt6_info *nrt6;
1363 daddr = &iph->daddr;
1364 saddr = &iph->saddr;
1366 daddr = &sk->sk_v6_daddr;
1367 saddr = &inet6_sk(sk)->saddr;
1371 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1373 rt6_do_update_pmtu(nrt6, mtu);
1375 /* ip6_ins_rt(nrt6) will bump the
1376 * rt6->rt6i_node->fn_sernum
1377 * which will fail the next rt6_check() and
1378 * invalidate the sk->sk_dst_cache.
1385 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1386 struct sk_buff *skb, u32 mtu)
1388 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1391 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1394 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1395 struct dst_entry *dst;
1398 memset(&fl6, 0, sizeof(fl6));
1399 fl6.flowi6_oif = oif;
1400 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1401 fl6.daddr = iph->daddr;
1402 fl6.saddr = iph->saddr;
1403 fl6.flowlabel = ip6_flowinfo(iph);
1405 dst = ip6_route_output(net, NULL, &fl6);
1407 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1410 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1412 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1414 ip6_update_pmtu(skb, sock_net(sk), mtu,
1415 sk->sk_bound_dev_if, sk->sk_mark);
1417 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1419 /* Handle redirects */
1420 struct ip6rd_flowi {
1422 struct in6_addr gateway;
1425 static struct rt6_info *__ip6_route_redirect(struct net *net,
1426 struct fib6_table *table,
1430 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1431 struct rt6_info *rt;
1432 struct fib6_node *fn;
1434 /* Get the "current" route for this destination and
1435 * check if the redirect has come from approriate router.
1437 * RFC 4861 specifies that redirects should only be
1438 * accepted if they come from the nexthop to the target.
1439 * Due to the way the routes are chosen, this notion
1440 * is a bit fuzzy and one might need to check all possible
1444 read_lock_bh(&table->tb6_lock);
1445 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1447 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1448 if (rt6_check_expired(rt))
1452 if (!(rt->rt6i_flags & RTF_GATEWAY))
1454 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1456 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1462 rt = net->ipv6.ip6_null_entry;
1463 else if (rt->dst.error) {
1464 rt = net->ipv6.ip6_null_entry;
1468 if (rt == net->ipv6.ip6_null_entry) {
1469 fn = fib6_backtrack(fn, &fl6->saddr);
1477 read_unlock_bh(&table->tb6_lock);
1482 static struct dst_entry *ip6_route_redirect(struct net *net,
1483 const struct flowi6 *fl6,
1484 const struct in6_addr *gateway)
1486 int flags = RT6_LOOKUP_F_HAS_SADDR;
1487 struct ip6rd_flowi rdfl;
1490 rdfl.gateway = *gateway;
1492 return fib6_rule_lookup(net, &rdfl.fl6,
1493 flags, __ip6_route_redirect);
1496 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1498 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1499 struct dst_entry *dst;
1502 memset(&fl6, 0, sizeof(fl6));
1503 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1504 fl6.flowi6_oif = oif;
1505 fl6.flowi6_mark = mark;
1506 fl6.daddr = iph->daddr;
1507 fl6.saddr = iph->saddr;
1508 fl6.flowlabel = ip6_flowinfo(iph);
1510 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1511 rt6_do_redirect(dst, NULL, skb);
1514 EXPORT_SYMBOL_GPL(ip6_redirect);
1516 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1519 const struct ipv6hdr *iph = ipv6_hdr(skb);
1520 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1521 struct dst_entry *dst;
1524 memset(&fl6, 0, sizeof(fl6));
1525 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1526 fl6.flowi6_oif = oif;
1527 fl6.flowi6_mark = mark;
1528 fl6.daddr = msg->dest;
1529 fl6.saddr = iph->daddr;
1531 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1532 rt6_do_redirect(dst, NULL, skb);
1536 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1538 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1540 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1542 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1544 struct net_device *dev = dst->dev;
1545 unsigned int mtu = dst_mtu(dst);
1546 struct net *net = dev_net(dev);
1548 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1550 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1551 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1554 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1555 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1556 * IPV6_MAXPLEN is also valid and means: "any MSS,
1557 * rely only on pmtu discovery"
1559 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1564 static unsigned int ip6_mtu(const struct dst_entry *dst)
1566 const struct rt6_info *rt = (const struct rt6_info *)dst;
1567 unsigned int mtu = rt->rt6i_pmtu;
1568 struct inet6_dev *idev;
1573 mtu = dst_metric_raw(dst, RTAX_MTU);
1580 idev = __in6_dev_get(dst->dev);
1582 mtu = idev->cnf.mtu6;
1586 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1589 static struct dst_entry *icmp6_dst_gc_list;
1590 static DEFINE_SPINLOCK(icmp6_dst_lock);
1592 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1595 struct dst_entry *dst;
1596 struct rt6_info *rt;
1597 struct inet6_dev *idev = in6_dev_get(dev);
1598 struct net *net = dev_net(dev);
1600 if (unlikely(!idev))
1601 return ERR_PTR(-ENODEV);
1603 rt = ip6_dst_alloc(net, dev, 0);
1604 if (unlikely(!rt)) {
1606 dst = ERR_PTR(-ENOMEM);
1610 rt->dst.flags |= DST_HOST;
1611 rt->dst.output = ip6_output;
1612 atomic_set(&rt->dst.__refcnt, 1);
1613 rt->rt6i_gateway = fl6->daddr;
1614 rt->rt6i_dst.addr = fl6->daddr;
1615 rt->rt6i_dst.plen = 128;
1616 rt->rt6i_idev = idev;
1617 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1619 spin_lock_bh(&icmp6_dst_lock);
1620 rt->dst.next = icmp6_dst_gc_list;
1621 icmp6_dst_gc_list = &rt->dst;
1622 spin_unlock_bh(&icmp6_dst_lock);
1624 fib6_force_start_gc(net);
1626 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1632 int icmp6_dst_gc(void)
1634 struct dst_entry *dst, **pprev;
1637 spin_lock_bh(&icmp6_dst_lock);
1638 pprev = &icmp6_dst_gc_list;
1640 while ((dst = *pprev) != NULL) {
1641 if (!atomic_read(&dst->__refcnt)) {
1650 spin_unlock_bh(&icmp6_dst_lock);
1655 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1658 struct dst_entry *dst, **pprev;
1660 spin_lock_bh(&icmp6_dst_lock);
1661 pprev = &icmp6_dst_gc_list;
1662 while ((dst = *pprev) != NULL) {
1663 struct rt6_info *rt = (struct rt6_info *) dst;
1664 if (func(rt, arg)) {
1671 spin_unlock_bh(&icmp6_dst_lock);
1674 static int ip6_dst_gc(struct dst_ops *ops)
1676 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1677 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1678 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1679 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1680 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1681 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1684 entries = dst_entries_get_fast(ops);
1685 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1686 entries <= rt_max_size)
1689 net->ipv6.ip6_rt_gc_expire++;
1690 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1691 entries = dst_entries_get_slow(ops);
1692 if (entries < ops->gc_thresh)
1693 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1695 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1696 return entries > rt_max_size;
1699 static int ip6_convert_metrics(struct mx6_config *mxc,
1700 const struct fib6_config *cfg)
1702 bool ecn_ca = false;
1710 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1714 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1715 int type = nla_type(nla);
1720 if (unlikely(type > RTAX_MAX))
1723 if (type == RTAX_CC_ALGO) {
1724 char tmp[TCP_CA_NAME_MAX];
1726 nla_strlcpy(tmp, nla, sizeof(tmp));
1727 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1728 if (val == TCP_CA_UNSPEC)
1731 val = nla_get_u32(nla);
1733 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1737 __set_bit(type - 1, mxc->mx_valid);
1741 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1742 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1752 int ip6_route_info_create(struct fib6_config *cfg, struct rt6_info **rt_ret)
1755 struct net *net = cfg->fc_nlinfo.nl_net;
1756 struct rt6_info *rt = NULL;
1757 struct net_device *dev = NULL;
1758 struct inet6_dev *idev = NULL;
1759 struct fib6_table *table;
1762 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1764 #ifndef CONFIG_IPV6_SUBTREES
1765 if (cfg->fc_src_len)
1768 if (cfg->fc_ifindex) {
1770 dev = dev_get_by_index(net, cfg->fc_ifindex);
1773 idev = in6_dev_get(dev);
1778 if (cfg->fc_metric == 0)
1779 cfg->fc_metric = IP6_RT_PRIO_USER;
1782 if (cfg->fc_nlinfo.nlh &&
1783 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1784 table = fib6_get_table(net, cfg->fc_table);
1786 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1787 table = fib6_new_table(net, cfg->fc_table);
1790 table = fib6_new_table(net, cfg->fc_table);
1796 rt = ip6_dst_alloc(net, NULL,
1797 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1804 if (cfg->fc_flags & RTF_EXPIRES)
1805 rt6_set_expires(rt, jiffies +
1806 clock_t_to_jiffies(cfg->fc_expires));
1808 rt6_clean_expires(rt);
1810 if (cfg->fc_protocol == RTPROT_UNSPEC)
1811 cfg->fc_protocol = RTPROT_BOOT;
1812 rt->rt6i_protocol = cfg->fc_protocol;
1814 addr_type = ipv6_addr_type(&cfg->fc_dst);
1816 if (addr_type & IPV6_ADDR_MULTICAST)
1817 rt->dst.input = ip6_mc_input;
1818 else if (cfg->fc_flags & RTF_LOCAL)
1819 rt->dst.input = ip6_input;
1821 rt->dst.input = ip6_forward;
1823 rt->dst.output = ip6_output;
1825 if (cfg->fc_encap) {
1826 struct lwtunnel_state *lwtstate;
1828 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1829 cfg->fc_encap, AF_INET6, cfg,
1833 rt->dst.lwtstate = lwtstate_get(lwtstate);
1834 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1835 rt->dst.lwtstate->orig_output = rt->dst.output;
1836 rt->dst.output = lwtunnel_output;
1838 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1839 rt->dst.lwtstate->orig_input = rt->dst.input;
1840 rt->dst.input = lwtunnel_input;
1844 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1845 rt->rt6i_dst.plen = cfg->fc_dst_len;
1846 if (rt->rt6i_dst.plen == 128)
1847 rt->dst.flags |= DST_HOST;
1849 #ifdef CONFIG_IPV6_SUBTREES
1850 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1851 rt->rt6i_src.plen = cfg->fc_src_len;
1854 rt->rt6i_metric = cfg->fc_metric;
1856 /* We cannot add true routes via loopback here,
1857 they would result in kernel looping; promote them to reject routes
1859 if ((cfg->fc_flags & RTF_REJECT) ||
1860 (dev && (dev->flags & IFF_LOOPBACK) &&
1861 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1862 !(cfg->fc_flags & RTF_LOCAL))) {
1863 /* hold loopback dev/idev if we haven't done so. */
1864 if (dev != net->loopback_dev) {
1869 dev = net->loopback_dev;
1871 idev = in6_dev_get(dev);
1877 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1878 switch (cfg->fc_type) {
1880 rt->dst.error = -EINVAL;
1881 rt->dst.output = dst_discard_sk;
1882 rt->dst.input = dst_discard;
1885 rt->dst.error = -EACCES;
1886 rt->dst.output = ip6_pkt_prohibit_out;
1887 rt->dst.input = ip6_pkt_prohibit;
1890 case RTN_UNREACHABLE:
1892 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1893 : (cfg->fc_type == RTN_UNREACHABLE)
1894 ? -EHOSTUNREACH : -ENETUNREACH;
1895 rt->dst.output = ip6_pkt_discard_out;
1896 rt->dst.input = ip6_pkt_discard;
1902 if (cfg->fc_flags & RTF_GATEWAY) {
1903 const struct in6_addr *gw_addr;
1906 gw_addr = &cfg->fc_gateway;
1907 gwa_type = ipv6_addr_type(gw_addr);
1909 /* if gw_addr is local we will fail to detect this in case
1910 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1911 * will return already-added prefix route via interface that
1912 * prefix route was assigned to, which might be non-loopback.
1915 if (ipv6_chk_addr_and_flags(net, gw_addr,
1916 gwa_type & IPV6_ADDR_LINKLOCAL ?
1920 rt->rt6i_gateway = *gw_addr;
1922 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1923 struct rt6_info *grt;
1925 /* IPv6 strictly inhibits using not link-local
1926 addresses as nexthop address.
1927 Otherwise, router will not able to send redirects.
1928 It is very good, but in some (rare!) circumstances
1929 (SIT, PtP, NBMA NOARP links) it is handy to allow
1930 some exceptions. --ANK
1932 if (!(gwa_type & IPV6_ADDR_UNICAST))
1935 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1937 err = -EHOSTUNREACH;
1941 if (dev != grt->dst.dev) {
1947 idev = grt->rt6i_idev;
1949 in6_dev_hold(grt->rt6i_idev);
1951 if (!(grt->rt6i_flags & RTF_GATEWAY))
1959 if (!dev || (dev->flags & IFF_LOOPBACK))
1967 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1968 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1972 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1973 rt->rt6i_prefsrc.plen = 128;
1975 rt->rt6i_prefsrc.plen = 0;
1977 rt->rt6i_flags = cfg->fc_flags;
1981 rt->rt6i_idev = idev;
1982 rt->rt6i_table = table;
1984 cfg->fc_nlinfo.nl_net = dev_net(dev);
2002 int ip6_route_add(struct fib6_config *cfg)
2004 struct mx6_config mxc = { .mx = NULL, };
2005 struct rt6_info *rt = NULL;
2008 err = ip6_route_info_create(cfg, &rt);
2012 err = ip6_convert_metrics(&mxc, cfg);
2016 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2028 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2031 struct fib6_table *table;
2032 struct net *net = dev_net(rt->dst.dev);
2034 if (rt == net->ipv6.ip6_null_entry ||
2035 rt->dst.flags & DST_NOCACHE) {
2040 table = rt->rt6i_table;
2041 write_lock_bh(&table->tb6_lock);
2042 err = fib6_del(rt, info);
2043 write_unlock_bh(&table->tb6_lock);
2050 int ip6_del_rt(struct rt6_info *rt)
2052 struct nl_info info = {
2053 .nl_net = dev_net(rt->dst.dev),
2055 return __ip6_del_rt(rt, &info);
2058 static int ip6_route_del(struct fib6_config *cfg)
2060 struct fib6_table *table;
2061 struct fib6_node *fn;
2062 struct rt6_info *rt;
2065 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2069 read_lock_bh(&table->tb6_lock);
2071 fn = fib6_locate(&table->tb6_root,
2072 &cfg->fc_dst, cfg->fc_dst_len,
2073 &cfg->fc_src, cfg->fc_src_len);
2076 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2077 if ((rt->rt6i_flags & RTF_CACHE) &&
2078 !(cfg->fc_flags & RTF_CACHE))
2080 if (cfg->fc_ifindex &&
2082 rt->dst.dev->ifindex != cfg->fc_ifindex))
2084 if (cfg->fc_flags & RTF_GATEWAY &&
2085 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2087 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2090 read_unlock_bh(&table->tb6_lock);
2092 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2095 read_unlock_bh(&table->tb6_lock);
2100 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2102 struct net *net = dev_net(skb->dev);
2103 struct netevent_redirect netevent;
2104 struct rt6_info *rt, *nrt = NULL;
2105 struct ndisc_options ndopts;
2106 struct inet6_dev *in6_dev;
2107 struct neighbour *neigh;
2109 int optlen, on_link;
2112 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2113 optlen -= sizeof(*msg);
2116 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2120 msg = (struct rd_msg *)icmp6_hdr(skb);
2122 if (ipv6_addr_is_multicast(&msg->dest)) {
2123 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2128 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2130 } else if (ipv6_addr_type(&msg->target) !=
2131 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2132 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2136 in6_dev = __in6_dev_get(skb->dev);
2139 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2143 * The IP source address of the Redirect MUST be the same as the current
2144 * first-hop router for the specified ICMP Destination Address.
2147 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2148 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2153 if (ndopts.nd_opts_tgt_lladdr) {
2154 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2157 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2162 rt = (struct rt6_info *) dst;
2163 if (rt == net->ipv6.ip6_null_entry) {
2164 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2168 /* Redirect received -> path was valid.
2169 * Look, redirects are sent only in response to data packets,
2170 * so that this nexthop apparently is reachable. --ANK
2172 dst_confirm(&rt->dst);
2174 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2179 * We have finally decided to accept it.
2182 neigh_update(neigh, lladdr, NUD_STALE,
2183 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2184 NEIGH_UPDATE_F_OVERRIDE|
2185 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2186 NEIGH_UPDATE_F_ISROUTER))
2189 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2193 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2195 nrt->rt6i_flags &= ~RTF_GATEWAY;
2197 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2199 if (ip6_ins_rt(nrt))
2202 netevent.old = &rt->dst;
2203 netevent.new = &nrt->dst;
2204 netevent.daddr = &msg->dest;
2205 netevent.neigh = neigh;
2206 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2208 if (rt->rt6i_flags & RTF_CACHE) {
2209 rt = (struct rt6_info *) dst_clone(&rt->dst);
2214 neigh_release(neigh);
2218 * Misc support functions
2221 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2223 BUG_ON(from->dst.from);
2225 rt->rt6i_flags &= ~RTF_EXPIRES;
2226 dst_hold(&from->dst);
2227 rt->dst.from = &from->dst;
2228 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2231 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2233 rt->dst.input = ort->dst.input;
2234 rt->dst.output = ort->dst.output;
2235 rt->rt6i_dst = ort->rt6i_dst;
2236 rt->dst.error = ort->dst.error;
2237 rt->rt6i_idev = ort->rt6i_idev;
2239 in6_dev_hold(rt->rt6i_idev);
2240 rt->dst.lastuse = jiffies;
2241 rt->rt6i_gateway = ort->rt6i_gateway;
2242 rt->rt6i_flags = ort->rt6i_flags;
2243 rt6_set_from(rt, ort);
2244 rt->rt6i_metric = ort->rt6i_metric;
2245 #ifdef CONFIG_IPV6_SUBTREES
2246 rt->rt6i_src = ort->rt6i_src;
2248 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2249 rt->rt6i_table = ort->rt6i_table;
2250 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2253 #ifdef CONFIG_IPV6_ROUTE_INFO
2254 static struct rt6_info *rt6_get_route_info(struct net *net,
2255 const struct in6_addr *prefix, int prefixlen,
2256 const struct in6_addr *gwaddr, int ifindex)
2258 struct fib6_node *fn;
2259 struct rt6_info *rt = NULL;
2260 struct fib6_table *table;
2262 table = fib6_get_table(net, RT6_TABLE_INFO);
2266 read_lock_bh(&table->tb6_lock);
2267 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2271 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2272 if (rt->dst.dev->ifindex != ifindex)
2274 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2276 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2282 read_unlock_bh(&table->tb6_lock);
2286 static struct rt6_info *rt6_add_route_info(struct net *net,
2287 const struct in6_addr *prefix, int prefixlen,
2288 const struct in6_addr *gwaddr, int ifindex,
2291 struct fib6_config cfg = {
2292 .fc_table = RT6_TABLE_INFO,
2293 .fc_metric = IP6_RT_PRIO_USER,
2294 .fc_ifindex = ifindex,
2295 .fc_dst_len = prefixlen,
2296 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2297 RTF_UP | RTF_PREF(pref),
2298 .fc_nlinfo.portid = 0,
2299 .fc_nlinfo.nlh = NULL,
2300 .fc_nlinfo.nl_net = net,
2303 cfg.fc_dst = *prefix;
2304 cfg.fc_gateway = *gwaddr;
2306 /* We should treat it as a default route if prefix length is 0. */
2308 cfg.fc_flags |= RTF_DEFAULT;
2310 ip6_route_add(&cfg);
2312 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2316 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2318 struct rt6_info *rt;
2319 struct fib6_table *table;
2321 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2325 read_lock_bh(&table->tb6_lock);
2326 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2327 if (dev == rt->dst.dev &&
2328 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2329 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2334 read_unlock_bh(&table->tb6_lock);
2338 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2339 struct net_device *dev,
2342 struct fib6_config cfg = {
2343 .fc_table = RT6_TABLE_DFLT,
2344 .fc_metric = IP6_RT_PRIO_USER,
2345 .fc_ifindex = dev->ifindex,
2346 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2347 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2348 .fc_nlinfo.portid = 0,
2349 .fc_nlinfo.nlh = NULL,
2350 .fc_nlinfo.nl_net = dev_net(dev),
2353 cfg.fc_gateway = *gwaddr;
2355 ip6_route_add(&cfg);
2357 return rt6_get_dflt_router(gwaddr, dev);
2360 void rt6_purge_dflt_routers(struct net *net)
2362 struct rt6_info *rt;
2363 struct fib6_table *table;
2365 /* NOTE: Keep consistent with rt6_get_dflt_router */
2366 table = fib6_get_table(net, RT6_TABLE_DFLT);
2371 read_lock_bh(&table->tb6_lock);
2372 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2373 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2374 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2376 read_unlock_bh(&table->tb6_lock);
2381 read_unlock_bh(&table->tb6_lock);
2384 static void rtmsg_to_fib6_config(struct net *net,
2385 struct in6_rtmsg *rtmsg,
2386 struct fib6_config *cfg)
2388 memset(cfg, 0, sizeof(*cfg));
2390 cfg->fc_table = RT6_TABLE_MAIN;
2391 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2392 cfg->fc_metric = rtmsg->rtmsg_metric;
2393 cfg->fc_expires = rtmsg->rtmsg_info;
2394 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2395 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2396 cfg->fc_flags = rtmsg->rtmsg_flags;
2398 cfg->fc_nlinfo.nl_net = net;
2400 cfg->fc_dst = rtmsg->rtmsg_dst;
2401 cfg->fc_src = rtmsg->rtmsg_src;
2402 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2405 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2407 struct fib6_config cfg;
2408 struct in6_rtmsg rtmsg;
2412 case SIOCADDRT: /* Add a route */
2413 case SIOCDELRT: /* Delete a route */
2414 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2416 err = copy_from_user(&rtmsg, arg,
2417 sizeof(struct in6_rtmsg));
2421 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2426 err = ip6_route_add(&cfg);
2429 err = ip6_route_del(&cfg);
2443 * Drop the packet on the floor
2446 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2449 struct dst_entry *dst = skb_dst(skb);
2450 switch (ipstats_mib_noroutes) {
2451 case IPSTATS_MIB_INNOROUTES:
2452 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2453 if (type == IPV6_ADDR_ANY) {
2454 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2455 IPSTATS_MIB_INADDRERRORS);
2459 case IPSTATS_MIB_OUTNOROUTES:
2460 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2461 ipstats_mib_noroutes);
2464 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2469 static int ip6_pkt_discard(struct sk_buff *skb)
2471 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2474 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2476 skb->dev = skb_dst(skb)->dev;
2477 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2480 static int ip6_pkt_prohibit(struct sk_buff *skb)
2482 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2485 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2487 skb->dev = skb_dst(skb)->dev;
2488 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2492 * Allocate a dst for local (unicast / anycast) address.
2495 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2496 const struct in6_addr *addr,
2499 struct net *net = dev_net(idev->dev);
2500 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2503 return ERR_PTR(-ENOMEM);
2507 rt->dst.flags |= DST_HOST;
2508 rt->dst.input = ip6_input;
2509 rt->dst.output = ip6_output;
2510 rt->rt6i_idev = idev;
2512 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2514 rt->rt6i_flags |= RTF_ANYCAST;
2516 rt->rt6i_flags |= RTF_LOCAL;
2518 rt->rt6i_gateway = *addr;
2519 rt->rt6i_dst.addr = *addr;
2520 rt->rt6i_dst.plen = 128;
2521 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2522 rt->dst.flags |= DST_NOCACHE;
2524 atomic_set(&rt->dst.__refcnt, 1);
2529 int ip6_route_get_saddr(struct net *net,
2530 struct rt6_info *rt,
2531 const struct in6_addr *daddr,
2533 struct in6_addr *saddr)
2535 struct inet6_dev *idev =
2536 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2538 if (rt && rt->rt6i_prefsrc.plen)
2539 *saddr = rt->rt6i_prefsrc.addr;
2541 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2542 daddr, prefs, saddr);
2546 /* remove deleted ip from prefsrc entries */
2547 struct arg_dev_net_ip {
2548 struct net_device *dev;
2550 struct in6_addr *addr;
2553 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2555 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2556 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2557 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2559 if (((void *)rt->dst.dev == dev || !dev) &&
2560 rt != net->ipv6.ip6_null_entry &&
2561 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2562 /* remove prefsrc entry */
2563 rt->rt6i_prefsrc.plen = 0;
2568 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2570 struct net *net = dev_net(ifp->idev->dev);
2571 struct arg_dev_net_ip adni = {
2572 .dev = ifp->idev->dev,
2576 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2579 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2580 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2582 /* Remove routers and update dst entries when gateway turn into host. */
2583 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2585 struct in6_addr *gateway = (struct in6_addr *)arg;
2587 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2588 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2589 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2595 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2597 fib6_clean_all(net, fib6_clean_tohost, gateway);
2600 struct arg_dev_net {
2601 struct net_device *dev;
2605 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2607 const struct arg_dev_net *adn = arg;
2608 const struct net_device *dev = adn->dev;
2610 if ((rt->dst.dev == dev || !dev) &&
2611 rt != adn->net->ipv6.ip6_null_entry)
2617 void rt6_ifdown(struct net *net, struct net_device *dev)
2619 struct arg_dev_net adn = {
2624 fib6_clean_all(net, fib6_ifdown, &adn);
2625 icmp6_clean_all(fib6_ifdown, &adn);
2627 rt6_uncached_list_flush_dev(net, dev);
2630 struct rt6_mtu_change_arg {
2631 struct net_device *dev;
2635 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2637 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2638 struct inet6_dev *idev;
2640 /* In IPv6 pmtu discovery is not optional,
2641 so that RTAX_MTU lock cannot disable it.
2642 We still use this lock to block changes
2643 caused by addrconf/ndisc.
2646 idev = __in6_dev_get(arg->dev);
2650 /* For administrative MTU increase, there is no way to discover
2651 IPv6 PMTU increase, so PMTU increase should be updated here.
2652 Since RFC 1981 doesn't include administrative MTU increase
2653 update PMTU increase is a MUST. (i.e. jumbo frame)
2656 If new MTU is less than route PMTU, this new MTU will be the
2657 lowest MTU in the path, update the route PMTU to reflect PMTU
2658 decreases; if new MTU is greater than route PMTU, and the
2659 old MTU is the lowest MTU in the path, update the route PMTU
2660 to reflect the increase. In this case if the other nodes' MTU
2661 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2664 if (rt->dst.dev == arg->dev &&
2665 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2666 if (rt->rt6i_flags & RTF_CACHE) {
2667 /* For RTF_CACHE with rt6i_pmtu == 0
2668 * (i.e. a redirected route),
2669 * the metrics of its rt->dst.from has already
2672 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2673 rt->rt6i_pmtu = arg->mtu;
2674 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2675 (dst_mtu(&rt->dst) < arg->mtu &&
2676 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2677 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2683 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2685 struct rt6_mtu_change_arg arg = {
2690 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2693 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2694 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2695 [RTA_OIF] = { .type = NLA_U32 },
2696 [RTA_IIF] = { .type = NLA_U32 },
2697 [RTA_PRIORITY] = { .type = NLA_U32 },
2698 [RTA_METRICS] = { .type = NLA_NESTED },
2699 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2700 [RTA_PREF] = { .type = NLA_U8 },
2701 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2702 [RTA_ENCAP] = { .type = NLA_NESTED },
2705 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2706 struct fib6_config *cfg)
2709 struct nlattr *tb[RTA_MAX+1];
2713 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2718 rtm = nlmsg_data(nlh);
2719 memset(cfg, 0, sizeof(*cfg));
2721 cfg->fc_table = rtm->rtm_table;
2722 cfg->fc_dst_len = rtm->rtm_dst_len;
2723 cfg->fc_src_len = rtm->rtm_src_len;
2724 cfg->fc_flags = RTF_UP;
2725 cfg->fc_protocol = rtm->rtm_protocol;
2726 cfg->fc_type = rtm->rtm_type;
2728 if (rtm->rtm_type == RTN_UNREACHABLE ||
2729 rtm->rtm_type == RTN_BLACKHOLE ||
2730 rtm->rtm_type == RTN_PROHIBIT ||
2731 rtm->rtm_type == RTN_THROW)
2732 cfg->fc_flags |= RTF_REJECT;
2734 if (rtm->rtm_type == RTN_LOCAL)
2735 cfg->fc_flags |= RTF_LOCAL;
2737 if (rtm->rtm_flags & RTM_F_CLONED)
2738 cfg->fc_flags |= RTF_CACHE;
2740 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2741 cfg->fc_nlinfo.nlh = nlh;
2742 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2744 if (tb[RTA_GATEWAY]) {
2745 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2746 cfg->fc_flags |= RTF_GATEWAY;
2750 int plen = (rtm->rtm_dst_len + 7) >> 3;
2752 if (nla_len(tb[RTA_DST]) < plen)
2755 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2759 int plen = (rtm->rtm_src_len + 7) >> 3;
2761 if (nla_len(tb[RTA_SRC]) < plen)
2764 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2767 if (tb[RTA_PREFSRC])
2768 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2771 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2773 if (tb[RTA_PRIORITY])
2774 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2776 if (tb[RTA_METRICS]) {
2777 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2778 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2782 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2784 if (tb[RTA_MULTIPATH]) {
2785 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2786 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2790 pref = nla_get_u8(tb[RTA_PREF]);
2791 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2792 pref != ICMPV6_ROUTER_PREF_HIGH)
2793 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2794 cfg->fc_flags |= RTF_PREF(pref);
2798 cfg->fc_encap = tb[RTA_ENCAP];
2800 if (tb[RTA_ENCAP_TYPE])
2801 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2809 struct rt6_info *rt6_info;
2810 struct fib6_config r_cfg;
2811 struct mx6_config mxc;
2812 struct list_head next;
2815 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2819 list_for_each_entry(nh, rt6_nh_list, next) {
2820 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2821 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2822 nh->r_cfg.fc_ifindex);
2826 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2827 struct rt6_info *rt, struct fib6_config *r_cfg)
2830 struct rt6_info *rtnh;
2833 list_for_each_entry(nh, rt6_nh_list, next) {
2834 /* check if rt6_info already exists */
2835 rtnh = nh->rt6_info;
2837 if (rtnh->dst.dev == rt->dst.dev &&
2838 rtnh->rt6i_idev == rt->rt6i_idev &&
2839 ipv6_addr_equal(&rtnh->rt6i_gateway,
2844 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2848 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2853 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2854 list_add_tail(&nh->next, rt6_nh_list);
2859 static int ip6_route_multipath_add(struct fib6_config *cfg)
2861 struct fib6_config r_cfg;
2862 struct rtnexthop *rtnh;
2863 struct rt6_info *rt;
2864 struct rt6_nh *err_nh;
2865 struct rt6_nh *nh, *nh_safe;
2870 int replace = (cfg->fc_nlinfo.nlh &&
2871 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2872 LIST_HEAD(rt6_nh_list);
2874 remaining = cfg->fc_mp_len;
2875 rtnh = (struct rtnexthop *)cfg->fc_mp;
2877 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2878 * rt6_info structs per nexthop
2880 while (rtnh_ok(rtnh, remaining)) {
2881 memcpy(&r_cfg, cfg, sizeof(*cfg));
2882 if (rtnh->rtnh_ifindex)
2883 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2885 attrlen = rtnh_attrlen(rtnh);
2887 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2889 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2891 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2892 r_cfg.fc_flags |= RTF_GATEWAY;
2894 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2895 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2897 r_cfg.fc_encap_type = nla_get_u16(nla);
2900 err = ip6_route_info_create(&r_cfg, &rt);
2904 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2910 rtnh = rtnh_next(rtnh, &remaining);
2914 list_for_each_entry(nh, &rt6_nh_list, next) {
2915 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2916 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2917 nh->rt6_info = NULL;
2920 ip6_print_replace_route_err(&rt6_nh_list);
2925 /* Because each route is added like a single route we remove
2926 * these flags after the first nexthop: if there is a collision,
2927 * we have already failed to add the first nexthop:
2928 * fib6_add_rt2node() has rejected it; when replacing, old
2929 * nexthops have been replaced by first new, the rest should
2932 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2940 /* Delete routes that were already added */
2941 list_for_each_entry(nh, &rt6_nh_list, next) {
2944 ip6_route_del(&nh->r_cfg);
2948 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2950 dst_free(&nh->rt6_info->dst);
2952 list_del(&nh->next);
2959 static int ip6_route_multipath_del(struct fib6_config *cfg)
2961 struct fib6_config r_cfg;
2962 struct rtnexthop *rtnh;
2965 int err = 1, last_err = 0;
2967 remaining = cfg->fc_mp_len;
2968 rtnh = (struct rtnexthop *)cfg->fc_mp;
2970 /* Parse a Multipath Entry */
2971 while (rtnh_ok(rtnh, remaining)) {
2972 memcpy(&r_cfg, cfg, sizeof(*cfg));
2973 if (rtnh->rtnh_ifindex)
2974 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2976 attrlen = rtnh_attrlen(rtnh);
2978 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2980 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2982 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2983 r_cfg.fc_flags |= RTF_GATEWAY;
2986 err = ip6_route_del(&r_cfg);
2990 rtnh = rtnh_next(rtnh, &remaining);
2996 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2998 struct fib6_config cfg;
3001 err = rtm_to_fib6_config(skb, nlh, &cfg);
3006 return ip6_route_multipath_del(&cfg);
3008 return ip6_route_del(&cfg);
3011 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3013 struct fib6_config cfg;
3016 err = rtm_to_fib6_config(skb, nlh, &cfg);
3021 return ip6_route_multipath_add(&cfg);
3023 return ip6_route_add(&cfg);
3026 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3028 return NLMSG_ALIGN(sizeof(struct rtmsg))
3029 + nla_total_size(16) /* RTA_SRC */
3030 + nla_total_size(16) /* RTA_DST */
3031 + nla_total_size(16) /* RTA_GATEWAY */
3032 + nla_total_size(16) /* RTA_PREFSRC */
3033 + nla_total_size(4) /* RTA_TABLE */
3034 + nla_total_size(4) /* RTA_IIF */
3035 + nla_total_size(4) /* RTA_OIF */
3036 + nla_total_size(4) /* RTA_PRIORITY */
3037 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3038 + nla_total_size(sizeof(struct rta_cacheinfo))
3039 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3040 + nla_total_size(1) /* RTA_PREF */
3041 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3044 static int rt6_fill_node(struct net *net,
3045 struct sk_buff *skb, struct rt6_info *rt,
3046 struct in6_addr *dst, struct in6_addr *src,
3047 int iif, int type, u32 portid, u32 seq,
3048 int prefix, int nowait, unsigned int flags)
3050 u32 metrics[RTAX_MAX];
3052 struct nlmsghdr *nlh;
3056 if (prefix) { /* user wants prefix routes only */
3057 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3058 /* success since this is not a prefix route */
3063 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3067 rtm = nlmsg_data(nlh);
3068 rtm->rtm_family = AF_INET6;
3069 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3070 rtm->rtm_src_len = rt->rt6i_src.plen;
3073 table = rt->rt6i_table->tb6_id;
3075 table = RT6_TABLE_UNSPEC;
3076 rtm->rtm_table = table;
3077 if (nla_put_u32(skb, RTA_TABLE, table))
3078 goto nla_put_failure;
3079 if (rt->rt6i_flags & RTF_REJECT) {
3080 switch (rt->dst.error) {
3082 rtm->rtm_type = RTN_BLACKHOLE;
3085 rtm->rtm_type = RTN_PROHIBIT;
3088 rtm->rtm_type = RTN_THROW;
3091 rtm->rtm_type = RTN_UNREACHABLE;
3095 else if (rt->rt6i_flags & RTF_LOCAL)
3096 rtm->rtm_type = RTN_LOCAL;
3097 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3098 rtm->rtm_type = RTN_LOCAL;
3100 rtm->rtm_type = RTN_UNICAST;
3102 if (!netif_carrier_ok(rt->dst.dev)) {
3103 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3104 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3105 rtm->rtm_flags |= RTNH_F_DEAD;
3107 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3108 rtm->rtm_protocol = rt->rt6i_protocol;
3109 if (rt->rt6i_flags & RTF_DYNAMIC)
3110 rtm->rtm_protocol = RTPROT_REDIRECT;
3111 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3112 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3113 rtm->rtm_protocol = RTPROT_RA;
3115 rtm->rtm_protocol = RTPROT_KERNEL;
3118 if (rt->rt6i_flags & RTF_CACHE)
3119 rtm->rtm_flags |= RTM_F_CLONED;
3122 if (nla_put_in6_addr(skb, RTA_DST, dst))
3123 goto nla_put_failure;
3124 rtm->rtm_dst_len = 128;
3125 } else if (rtm->rtm_dst_len)
3126 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3127 goto nla_put_failure;
3128 #ifdef CONFIG_IPV6_SUBTREES
3130 if (nla_put_in6_addr(skb, RTA_SRC, src))
3131 goto nla_put_failure;
3132 rtm->rtm_src_len = 128;
3133 } else if (rtm->rtm_src_len &&
3134 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3135 goto nla_put_failure;
3138 #ifdef CONFIG_IPV6_MROUTE
3139 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3140 int err = ip6mr_get_route(net, skb, rtm, nowait);
3145 goto nla_put_failure;
3147 if (err == -EMSGSIZE)
3148 goto nla_put_failure;
3153 if (nla_put_u32(skb, RTA_IIF, iif))
3154 goto nla_put_failure;
3156 struct in6_addr saddr_buf;
3157 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3158 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3159 goto nla_put_failure;
3162 if (rt->rt6i_prefsrc.plen) {
3163 struct in6_addr saddr_buf;
3164 saddr_buf = rt->rt6i_prefsrc.addr;
3165 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3166 goto nla_put_failure;
3169 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3171 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3172 if (rtnetlink_put_metrics(skb, metrics) < 0)
3173 goto nla_put_failure;
3175 if (rt->rt6i_flags & RTF_GATEWAY) {
3176 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3177 goto nla_put_failure;
3181 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3182 goto nla_put_failure;
3183 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3184 goto nla_put_failure;
3186 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3188 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3189 goto nla_put_failure;
3191 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3192 goto nla_put_failure;
3194 lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3196 nlmsg_end(skb, nlh);
3200 nlmsg_cancel(skb, nlh);
3204 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3206 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3209 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3210 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3211 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3215 return rt6_fill_node(arg->net,
3216 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3217 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3218 prefix, 0, NLM_F_MULTI);
3221 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3223 struct net *net = sock_net(in_skb->sk);
3224 struct nlattr *tb[RTA_MAX+1];
3225 struct rt6_info *rt;
3226 struct sk_buff *skb;
3229 int err, iif = 0, oif = 0;
3231 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3236 memset(&fl6, 0, sizeof(fl6));
3239 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3242 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3246 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3249 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3253 iif = nla_get_u32(tb[RTA_IIF]);
3256 oif = nla_get_u32(tb[RTA_OIF]);
3259 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3262 struct net_device *dev;
3265 dev = __dev_get_by_index(net, iif);
3271 fl6.flowi6_iif = iif;
3273 if (!ipv6_addr_any(&fl6.saddr))
3274 flags |= RT6_LOOKUP_F_HAS_SADDR;
3276 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3279 fl6.flowi6_oif = oif;
3281 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3284 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3291 /* Reserve room for dummy headers, this skb can pass
3292 through good chunk of routing engine.
3294 skb_reset_mac_header(skb);
3295 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3297 skb_dst_set(skb, &rt->dst);
3299 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3300 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3301 nlh->nlmsg_seq, 0, 0, 0);
3307 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3312 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3313 unsigned int nlm_flags)
3315 struct sk_buff *skb;
3316 struct net *net = info->nl_net;
3321 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3323 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3327 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3328 event, info->portid, seq, 0, 0, nlm_flags);
3330 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3331 WARN_ON(err == -EMSGSIZE);
3335 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3336 info->nlh, gfp_any());
3340 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3343 static int ip6_route_dev_notify(struct notifier_block *this,
3344 unsigned long event, void *ptr)
3346 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3347 struct net *net = dev_net(dev);
3349 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3350 net->ipv6.ip6_null_entry->dst.dev = dev;
3351 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3352 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3353 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3354 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3355 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3356 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3367 #ifdef CONFIG_PROC_FS
3369 static const struct file_operations ipv6_route_proc_fops = {
3370 .owner = THIS_MODULE,
3371 .open = ipv6_route_open,
3373 .llseek = seq_lseek,
3374 .release = seq_release_net,
3377 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3379 struct net *net = (struct net *)seq->private;
3380 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3381 net->ipv6.rt6_stats->fib_nodes,
3382 net->ipv6.rt6_stats->fib_route_nodes,
3383 net->ipv6.rt6_stats->fib_rt_alloc,
3384 net->ipv6.rt6_stats->fib_rt_entries,
3385 net->ipv6.rt6_stats->fib_rt_cache,
3386 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3387 net->ipv6.rt6_stats->fib_discarded_routes);
3392 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3394 return single_open_net(inode, file, rt6_stats_seq_show);
3397 static const struct file_operations rt6_stats_seq_fops = {
3398 .owner = THIS_MODULE,
3399 .open = rt6_stats_seq_open,
3401 .llseek = seq_lseek,
3402 .release = single_release_net,
3404 #endif /* CONFIG_PROC_FS */
3406 #ifdef CONFIG_SYSCTL
3409 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3410 void __user *buffer, size_t *lenp, loff_t *ppos)
3417 net = (struct net *)ctl->extra1;
3418 delay = net->ipv6.sysctl.flush_delay;
3419 proc_dointvec(ctl, write, buffer, lenp, ppos);
3420 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3424 struct ctl_table ipv6_route_table_template[] = {
3426 .procname = "flush",
3427 .data = &init_net.ipv6.sysctl.flush_delay,
3428 .maxlen = sizeof(int),
3430 .proc_handler = ipv6_sysctl_rtcache_flush
3433 .procname = "gc_thresh",
3434 .data = &ip6_dst_ops_template.gc_thresh,
3435 .maxlen = sizeof(int),
3437 .proc_handler = proc_dointvec,
3440 .procname = "max_size",
3441 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3442 .maxlen = sizeof(int),
3444 .proc_handler = proc_dointvec,
3447 .procname = "gc_min_interval",
3448 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3449 .maxlen = sizeof(int),
3451 .proc_handler = proc_dointvec_jiffies,
3454 .procname = "gc_timeout",
3455 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3456 .maxlen = sizeof(int),
3458 .proc_handler = proc_dointvec_jiffies,
3461 .procname = "gc_interval",
3462 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3463 .maxlen = sizeof(int),
3465 .proc_handler = proc_dointvec_jiffies,
3468 .procname = "gc_elasticity",
3469 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3470 .maxlen = sizeof(int),
3472 .proc_handler = proc_dointvec,
3475 .procname = "mtu_expires",
3476 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3477 .maxlen = sizeof(int),
3479 .proc_handler = proc_dointvec_jiffies,
3482 .procname = "min_adv_mss",
3483 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3484 .maxlen = sizeof(int),
3486 .proc_handler = proc_dointvec,
3489 .procname = "gc_min_interval_ms",
3490 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3491 .maxlen = sizeof(int),
3493 .proc_handler = proc_dointvec_ms_jiffies,
3498 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3500 struct ctl_table *table;
3502 table = kmemdup(ipv6_route_table_template,
3503 sizeof(ipv6_route_table_template),
3507 table[0].data = &net->ipv6.sysctl.flush_delay;
3508 table[0].extra1 = net;
3509 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3510 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3511 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3512 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3513 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3514 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3515 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3516 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3517 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3519 /* Don't export sysctls to unprivileged users */
3520 if (net->user_ns != &init_user_ns)
3521 table[0].procname = NULL;
3528 static int __net_init ip6_route_net_init(struct net *net)
3532 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3533 sizeof(net->ipv6.ip6_dst_ops));
3535 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3536 goto out_ip6_dst_ops;
3538 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3539 sizeof(*net->ipv6.ip6_null_entry),
3541 if (!net->ipv6.ip6_null_entry)
3542 goto out_ip6_dst_entries;
3543 net->ipv6.ip6_null_entry->dst.path =
3544 (struct dst_entry *)net->ipv6.ip6_null_entry;
3545 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3546 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3547 ip6_template_metrics, true);
3549 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3550 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3551 sizeof(*net->ipv6.ip6_prohibit_entry),
3553 if (!net->ipv6.ip6_prohibit_entry)
3554 goto out_ip6_null_entry;
3555 net->ipv6.ip6_prohibit_entry->dst.path =
3556 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3557 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3558 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3559 ip6_template_metrics, true);
3561 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3562 sizeof(*net->ipv6.ip6_blk_hole_entry),
3564 if (!net->ipv6.ip6_blk_hole_entry)
3565 goto out_ip6_prohibit_entry;
3566 net->ipv6.ip6_blk_hole_entry->dst.path =
3567 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3568 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3569 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3570 ip6_template_metrics, true);
3573 net->ipv6.sysctl.flush_delay = 0;
3574 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3575 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3576 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3577 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3578 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3579 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3580 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3582 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3588 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3589 out_ip6_prohibit_entry:
3590 kfree(net->ipv6.ip6_prohibit_entry);
3592 kfree(net->ipv6.ip6_null_entry);
3594 out_ip6_dst_entries:
3595 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3600 static void __net_exit ip6_route_net_exit(struct net *net)
3602 kfree(net->ipv6.ip6_null_entry);
3603 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3604 kfree(net->ipv6.ip6_prohibit_entry);
3605 kfree(net->ipv6.ip6_blk_hole_entry);
3607 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3610 static int __net_init ip6_route_net_init_late(struct net *net)
3612 #ifdef CONFIG_PROC_FS
3613 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3614 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3619 static void __net_exit ip6_route_net_exit_late(struct net *net)
3621 #ifdef CONFIG_PROC_FS
3622 remove_proc_entry("ipv6_route", net->proc_net);
3623 remove_proc_entry("rt6_stats", net->proc_net);
3627 static struct pernet_operations ip6_route_net_ops = {
3628 .init = ip6_route_net_init,
3629 .exit = ip6_route_net_exit,
3632 static int __net_init ipv6_inetpeer_init(struct net *net)
3634 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3638 inet_peer_base_init(bp);
3639 net->ipv6.peers = bp;
3643 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3645 struct inet_peer_base *bp = net->ipv6.peers;
3647 net->ipv6.peers = NULL;
3648 inetpeer_invalidate_tree(bp);
3652 static struct pernet_operations ipv6_inetpeer_ops = {
3653 .init = ipv6_inetpeer_init,
3654 .exit = ipv6_inetpeer_exit,
3657 static struct pernet_operations ip6_route_net_late_ops = {
3658 .init = ip6_route_net_init_late,
3659 .exit = ip6_route_net_exit_late,
3662 static struct notifier_block ip6_route_dev_notifier = {
3663 .notifier_call = ip6_route_dev_notify,
3667 int __init ip6_route_init(void)
3673 ip6_dst_ops_template.kmem_cachep =
3674 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3675 SLAB_HWCACHE_ALIGN, NULL);
3676 if (!ip6_dst_ops_template.kmem_cachep)
3679 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3681 goto out_kmem_cache;
3683 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3685 goto out_dst_entries;
3687 ret = register_pernet_subsys(&ip6_route_net_ops);
3689 goto out_register_inetpeer;
3691 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3693 /* Registering of the loopback is done before this portion of code,
3694 * the loopback reference in rt6_info will not be taken, do it
3695 * manually for init_net */
3696 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3697 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3698 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3699 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3700 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3701 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3702 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3706 goto out_register_subsys;
3712 ret = fib6_rules_init();
3716 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3718 goto fib6_rules_init;
3721 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3722 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3723 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3724 goto out_register_late_subsys;
3726 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3728 goto out_register_late_subsys;
3730 for_each_possible_cpu(cpu) {
3731 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3733 INIT_LIST_HEAD(&ul->head);
3734 spin_lock_init(&ul->lock);
3740 out_register_late_subsys:
3741 unregister_pernet_subsys(&ip6_route_net_late_ops);
3743 fib6_rules_cleanup();
3748 out_register_subsys:
3749 unregister_pernet_subsys(&ip6_route_net_ops);
3750 out_register_inetpeer:
3751 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3753 dst_entries_destroy(&ip6_dst_blackhole_ops);
3755 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3759 void ip6_route_cleanup(void)
3761 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3762 unregister_pernet_subsys(&ip6_route_net_late_ops);
3763 fib6_rules_cleanup();
3766 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3767 unregister_pernet_subsys(&ip6_route_net_ops);
3768 dst_entries_destroy(&ip6_dst_blackhole_ops);
3769 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);