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 for_each_possible_cpu(cpu) {
146 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
149 spin_lock_bh(&ul->lock);
150 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
151 struct inet6_dev *rt_idev = rt->rt6i_idev;
152 struct net_device *rt_dev = rt->dst.dev;
154 if (rt_idev && (rt_idev->dev == dev || !dev) &&
155 rt_idev->dev != loopback_dev) {
156 rt->rt6i_idev = in6_dev_get(loopback_dev);
157 in6_dev_put(rt_idev);
160 if (rt_dev && (rt_dev == dev || !dev) &&
161 rt_dev != loopback_dev) {
162 rt->dst.dev = loopback_dev;
163 dev_hold(rt->dst.dev);
167 spin_unlock_bh(&ul->lock);
171 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
173 return dst_metrics_write_ptr(rt->dst.from);
176 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
178 struct rt6_info *rt = (struct rt6_info *)dst;
180 if (rt->rt6i_flags & RTF_PCPU)
181 return rt6_pcpu_cow_metrics(rt);
182 else if (rt->rt6i_flags & RTF_CACHE)
185 return dst_cow_metrics_generic(dst, old);
188 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
192 struct in6_addr *p = &rt->rt6i_gateway;
194 if (!ipv6_addr_any(p))
195 return (const void *) p;
197 return &ipv6_hdr(skb)->daddr;
201 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
205 struct rt6_info *rt = (struct rt6_info *) dst;
208 daddr = choose_neigh_daddr(rt, skb, daddr);
209 n = __ipv6_neigh_lookup(dst->dev, daddr);
212 return neigh_create(&nd_tbl, daddr, dst->dev);
215 static struct dst_ops ip6_dst_ops_template = {
219 .check = ip6_dst_check,
220 .default_advmss = ip6_default_advmss,
222 .cow_metrics = ipv6_cow_metrics,
223 .destroy = ip6_dst_destroy,
224 .ifdown = ip6_dst_ifdown,
225 .negative_advice = ip6_negative_advice,
226 .link_failure = ip6_link_failure,
227 .update_pmtu = ip6_rt_update_pmtu,
228 .redirect = rt6_do_redirect,
229 .local_out = __ip6_local_out,
230 .neigh_lookup = ip6_neigh_lookup,
233 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
235 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
237 return mtu ? : dst->dev->mtu;
240 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
241 struct sk_buff *skb, u32 mtu)
245 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
250 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
256 static struct dst_ops ip6_dst_blackhole_ops = {
258 .destroy = ip6_dst_destroy,
259 .check = ip6_dst_check,
260 .mtu = ip6_blackhole_mtu,
261 .default_advmss = ip6_default_advmss,
262 .update_pmtu = ip6_rt_blackhole_update_pmtu,
263 .redirect = ip6_rt_blackhole_redirect,
264 .cow_metrics = ip6_rt_blackhole_cow_metrics,
265 .neigh_lookup = ip6_neigh_lookup,
268 static const u32 ip6_template_metrics[RTAX_MAX] = {
269 [RTAX_HOPLIMIT - 1] = 0,
272 static const struct rt6_info ip6_null_entry_template = {
274 .__refcnt = ATOMIC_INIT(1),
276 .obsolete = DST_OBSOLETE_FORCE_CHK,
277 .error = -ENETUNREACH,
278 .input = ip6_pkt_discard,
279 .output = ip6_pkt_discard_out,
281 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
282 .rt6i_protocol = RTPROT_KERNEL,
283 .rt6i_metric = ~(u32) 0,
284 .rt6i_ref = ATOMIC_INIT(1),
287 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
289 static const struct rt6_info ip6_prohibit_entry_template = {
291 .__refcnt = ATOMIC_INIT(1),
293 .obsolete = DST_OBSOLETE_FORCE_CHK,
295 .input = ip6_pkt_prohibit,
296 .output = ip6_pkt_prohibit_out,
298 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
299 .rt6i_protocol = RTPROT_KERNEL,
300 .rt6i_metric = ~(u32) 0,
301 .rt6i_ref = ATOMIC_INIT(1),
304 static const struct rt6_info ip6_blk_hole_entry_template = {
306 .__refcnt = ATOMIC_INIT(1),
308 .obsolete = DST_OBSOLETE_FORCE_CHK,
310 .input = dst_discard,
311 .output = dst_discard_sk,
313 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
314 .rt6i_protocol = RTPROT_KERNEL,
315 .rt6i_metric = ~(u32) 0,
316 .rt6i_ref = ATOMIC_INIT(1),
321 /* allocate dst with ip6_dst_ops */
322 static struct rt6_info *__ip6_dst_alloc(struct net *net,
323 struct net_device *dev,
326 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
327 0, DST_OBSOLETE_FORCE_CHK, flags);
330 struct dst_entry *dst = &rt->dst;
332 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
333 INIT_LIST_HEAD(&rt->rt6i_siblings);
334 INIT_LIST_HEAD(&rt->rt6i_uncached);
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);
1072 rt = rt6_select(fn, oif, strict);
1073 if (rt->rt6i_nsiblings)
1074 rt = rt6_multipath_select(rt, fl6, oif, strict);
1075 if (rt == net->ipv6.ip6_null_entry) {
1076 fn = fib6_backtrack(fn, &fl6->saddr);
1078 goto redo_rt6_select;
1079 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1080 /* also consider unreachable route */
1081 strict &= ~RT6_LOOKUP_F_REACHABLE;
1083 goto redo_rt6_select;
1088 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1089 dst_use(&rt->dst, jiffies);
1090 read_unlock_bh(&table->tb6_lock);
1092 rt6_dst_from_metrics_check(rt);
1094 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1095 !(rt->rt6i_flags & RTF_GATEWAY))) {
1096 /* Create a RTF_CACHE clone which will not be
1097 * owned by the fib6 tree. It is for the special case where
1098 * the daddr in the skb during the neighbor look-up is different
1099 * from the fl6->daddr used to look-up route here.
1102 struct rt6_info *uncached_rt;
1104 dst_use(&rt->dst, jiffies);
1105 read_unlock_bh(&table->tb6_lock);
1107 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1108 dst_release(&rt->dst);
1111 rt6_uncached_list_add(uncached_rt);
1113 uncached_rt = net->ipv6.ip6_null_entry;
1115 dst_hold(&uncached_rt->dst);
1119 /* Get a percpu copy */
1121 struct rt6_info *pcpu_rt;
1123 rt->dst.lastuse = jiffies;
1125 pcpu_rt = rt6_get_pcpu_route(rt);
1128 read_unlock_bh(&table->tb6_lock);
1130 /* We have to do the read_unlock first
1131 * because rt6_make_pcpu_route() may trigger
1132 * ip6_dst_gc() which will take the write_lock.
1135 read_unlock_bh(&table->tb6_lock);
1136 pcpu_rt = rt6_make_pcpu_route(rt);
1137 dst_release(&rt->dst);
1145 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1146 struct flowi6 *fl6, int flags)
1148 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1151 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1152 struct net_device *dev,
1153 struct flowi6 *fl6, int flags)
1155 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1156 flags |= RT6_LOOKUP_F_IFACE;
1158 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1161 void ip6_route_input(struct sk_buff *skb)
1163 const struct ipv6hdr *iph = ipv6_hdr(skb);
1164 struct net *net = dev_net(skb->dev);
1165 int flags = RT6_LOOKUP_F_HAS_SADDR;
1166 struct ip_tunnel_info *tun_info;
1167 struct flowi6 fl6 = {
1168 .flowi6_iif = skb->dev->ifindex,
1169 .daddr = iph->daddr,
1170 .saddr = iph->saddr,
1171 .flowlabel = ip6_flowinfo(iph),
1172 .flowi6_mark = skb->mark,
1173 .flowi6_proto = iph->nexthdr,
1176 tun_info = skb_tunnel_info(skb);
1177 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1178 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1180 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1183 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1184 struct flowi6 *fl6, int flags)
1186 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1189 struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
1194 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1196 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1197 flags |= RT6_LOOKUP_F_IFACE;
1199 if (!ipv6_addr_any(&fl6->saddr))
1200 flags |= RT6_LOOKUP_F_HAS_SADDR;
1202 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1204 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1206 EXPORT_SYMBOL(ip6_route_output);
1208 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1210 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1211 struct dst_entry *new = NULL;
1213 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1217 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1220 new->input = dst_discard;
1221 new->output = dst_discard_sk;
1223 if (dst_metrics_read_only(&ort->dst))
1224 new->_metrics = ort->dst._metrics;
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;
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) {
1327 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1328 rt->rt6i_node->fn_sernum = -1;
1333 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1335 struct net *net = dev_net(rt->dst.dev);
1337 rt->rt6i_flags |= RTF_MODIFIED;
1338 rt->rt6i_pmtu = mtu;
1339 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1342 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1343 const struct ipv6hdr *iph, u32 mtu)
1345 struct rt6_info *rt6 = (struct rt6_info *)dst;
1347 if (rt6->rt6i_flags & RTF_LOCAL)
1351 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1352 if (mtu >= dst_mtu(dst))
1355 if (rt6->rt6i_flags & RTF_CACHE) {
1356 rt6_do_update_pmtu(rt6, mtu);
1358 const struct in6_addr *daddr, *saddr;
1359 struct rt6_info *nrt6;
1362 daddr = &iph->daddr;
1363 saddr = &iph->saddr;
1365 daddr = &sk->sk_v6_daddr;
1366 saddr = &inet6_sk(sk)->saddr;
1370 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1372 rt6_do_update_pmtu(nrt6, mtu);
1374 /* ip6_ins_rt(nrt6) will bump the
1375 * rt6->rt6i_node->fn_sernum
1376 * which will fail the next rt6_check() and
1377 * invalidate the sk->sk_dst_cache.
1384 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1385 struct sk_buff *skb, u32 mtu)
1387 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1390 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1393 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1394 struct dst_entry *dst;
1397 memset(&fl6, 0, sizeof(fl6));
1398 fl6.flowi6_oif = oif;
1399 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1400 fl6.daddr = iph->daddr;
1401 fl6.saddr = iph->saddr;
1402 fl6.flowlabel = ip6_flowinfo(iph);
1404 dst = ip6_route_output(net, NULL, &fl6);
1406 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1409 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1411 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1413 ip6_update_pmtu(skb, sock_net(sk), mtu,
1414 sk->sk_bound_dev_if, sk->sk_mark);
1416 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1418 /* Handle redirects */
1419 struct ip6rd_flowi {
1421 struct in6_addr gateway;
1424 static struct rt6_info *__ip6_route_redirect(struct net *net,
1425 struct fib6_table *table,
1429 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1430 struct rt6_info *rt;
1431 struct fib6_node *fn;
1433 /* Get the "current" route for this destination and
1434 * check if the redirect has come from approriate router.
1436 * RFC 4861 specifies that redirects should only be
1437 * accepted if they come from the nexthop to the target.
1438 * Due to the way the routes are chosen, this notion
1439 * is a bit fuzzy and one might need to check all possible
1443 read_lock_bh(&table->tb6_lock);
1444 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1446 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1447 if (rt6_check_expired(rt))
1451 if (!(rt->rt6i_flags & RTF_GATEWAY))
1453 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1455 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1461 rt = net->ipv6.ip6_null_entry;
1462 else if (rt->dst.error) {
1463 rt = net->ipv6.ip6_null_entry;
1467 if (rt == net->ipv6.ip6_null_entry) {
1468 fn = fib6_backtrack(fn, &fl6->saddr);
1476 read_unlock_bh(&table->tb6_lock);
1481 static struct dst_entry *ip6_route_redirect(struct net *net,
1482 const struct flowi6 *fl6,
1483 const struct in6_addr *gateway)
1485 int flags = RT6_LOOKUP_F_HAS_SADDR;
1486 struct ip6rd_flowi rdfl;
1489 rdfl.gateway = *gateway;
1491 return fib6_rule_lookup(net, &rdfl.fl6,
1492 flags, __ip6_route_redirect);
1495 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1497 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1498 struct dst_entry *dst;
1501 memset(&fl6, 0, sizeof(fl6));
1502 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1503 fl6.flowi6_oif = oif;
1504 fl6.flowi6_mark = mark;
1505 fl6.daddr = iph->daddr;
1506 fl6.saddr = iph->saddr;
1507 fl6.flowlabel = ip6_flowinfo(iph);
1509 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1510 rt6_do_redirect(dst, NULL, skb);
1513 EXPORT_SYMBOL_GPL(ip6_redirect);
1515 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1518 const struct ipv6hdr *iph = ipv6_hdr(skb);
1519 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1520 struct dst_entry *dst;
1523 memset(&fl6, 0, sizeof(fl6));
1524 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1525 fl6.flowi6_oif = oif;
1526 fl6.flowi6_mark = mark;
1527 fl6.daddr = msg->dest;
1528 fl6.saddr = iph->daddr;
1530 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1531 rt6_do_redirect(dst, NULL, skb);
1535 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1537 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1539 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1541 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1543 struct net_device *dev = dst->dev;
1544 unsigned int mtu = dst_mtu(dst);
1545 struct net *net = dev_net(dev);
1547 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1549 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1550 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1553 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1554 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1555 * IPV6_MAXPLEN is also valid and means: "any MSS,
1556 * rely only on pmtu discovery"
1558 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1563 static unsigned int ip6_mtu(const struct dst_entry *dst)
1565 const struct rt6_info *rt = (const struct rt6_info *)dst;
1566 unsigned int mtu = rt->rt6i_pmtu;
1567 struct inet6_dev *idev;
1572 mtu = dst_metric_raw(dst, RTAX_MTU);
1579 idev = __in6_dev_get(dst->dev);
1581 mtu = idev->cnf.mtu6;
1585 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1588 static struct dst_entry *icmp6_dst_gc_list;
1589 static DEFINE_SPINLOCK(icmp6_dst_lock);
1591 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1594 struct dst_entry *dst;
1595 struct rt6_info *rt;
1596 struct inet6_dev *idev = in6_dev_get(dev);
1597 struct net *net = dev_net(dev);
1599 if (unlikely(!idev))
1600 return ERR_PTR(-ENODEV);
1602 rt = ip6_dst_alloc(net, dev, 0);
1603 if (unlikely(!rt)) {
1605 dst = ERR_PTR(-ENOMEM);
1609 rt->dst.flags |= DST_HOST;
1610 rt->dst.output = ip6_output;
1611 atomic_set(&rt->dst.__refcnt, 1);
1612 rt->rt6i_gateway = fl6->daddr;
1613 rt->rt6i_dst.addr = fl6->daddr;
1614 rt->rt6i_dst.plen = 128;
1615 rt->rt6i_idev = idev;
1616 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1618 spin_lock_bh(&icmp6_dst_lock);
1619 rt->dst.next = icmp6_dst_gc_list;
1620 icmp6_dst_gc_list = &rt->dst;
1621 spin_unlock_bh(&icmp6_dst_lock);
1623 fib6_force_start_gc(net);
1625 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1631 int icmp6_dst_gc(void)
1633 struct dst_entry *dst, **pprev;
1636 spin_lock_bh(&icmp6_dst_lock);
1637 pprev = &icmp6_dst_gc_list;
1639 while ((dst = *pprev) != NULL) {
1640 if (!atomic_read(&dst->__refcnt)) {
1649 spin_unlock_bh(&icmp6_dst_lock);
1654 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1657 struct dst_entry *dst, **pprev;
1659 spin_lock_bh(&icmp6_dst_lock);
1660 pprev = &icmp6_dst_gc_list;
1661 while ((dst = *pprev) != NULL) {
1662 struct rt6_info *rt = (struct rt6_info *) dst;
1663 if (func(rt, arg)) {
1670 spin_unlock_bh(&icmp6_dst_lock);
1673 static int ip6_dst_gc(struct dst_ops *ops)
1675 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1676 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1677 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1678 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1679 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1680 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1683 entries = dst_entries_get_fast(ops);
1684 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1685 entries <= rt_max_size)
1688 net->ipv6.ip6_rt_gc_expire++;
1689 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1690 entries = dst_entries_get_slow(ops);
1691 if (entries < ops->gc_thresh)
1692 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1694 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1695 return entries > rt_max_size;
1698 static int ip6_convert_metrics(struct mx6_config *mxc,
1699 const struct fib6_config *cfg)
1701 bool ecn_ca = false;
1709 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1713 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1714 int type = nla_type(nla);
1719 if (unlikely(type > RTAX_MAX))
1722 if (type == RTAX_CC_ALGO) {
1723 char tmp[TCP_CA_NAME_MAX];
1725 nla_strlcpy(tmp, nla, sizeof(tmp));
1726 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1727 if (val == TCP_CA_UNSPEC)
1730 val = nla_get_u32(nla);
1732 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1736 __set_bit(type - 1, mxc->mx_valid);
1740 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1741 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1751 int ip6_route_info_create(struct fib6_config *cfg, struct rt6_info **rt_ret)
1754 struct net *net = cfg->fc_nlinfo.nl_net;
1755 struct rt6_info *rt = NULL;
1756 struct net_device *dev = NULL;
1757 struct inet6_dev *idev = NULL;
1758 struct fib6_table *table;
1761 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1763 #ifndef CONFIG_IPV6_SUBTREES
1764 if (cfg->fc_src_len)
1767 if (cfg->fc_ifindex) {
1769 dev = dev_get_by_index(net, cfg->fc_ifindex);
1772 idev = in6_dev_get(dev);
1777 if (cfg->fc_metric == 0)
1778 cfg->fc_metric = IP6_RT_PRIO_USER;
1781 if (cfg->fc_nlinfo.nlh &&
1782 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1783 table = fib6_get_table(net, cfg->fc_table);
1785 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1786 table = fib6_new_table(net, cfg->fc_table);
1789 table = fib6_new_table(net, cfg->fc_table);
1795 rt = ip6_dst_alloc(net, NULL,
1796 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1803 if (cfg->fc_flags & RTF_EXPIRES)
1804 rt6_set_expires(rt, jiffies +
1805 clock_t_to_jiffies(cfg->fc_expires));
1807 rt6_clean_expires(rt);
1809 if (cfg->fc_protocol == RTPROT_UNSPEC)
1810 cfg->fc_protocol = RTPROT_BOOT;
1811 rt->rt6i_protocol = cfg->fc_protocol;
1813 addr_type = ipv6_addr_type(&cfg->fc_dst);
1815 if (addr_type & IPV6_ADDR_MULTICAST)
1816 rt->dst.input = ip6_mc_input;
1817 else if (cfg->fc_flags & RTF_LOCAL)
1818 rt->dst.input = ip6_input;
1820 rt->dst.input = ip6_forward;
1822 rt->dst.output = ip6_output;
1824 if (cfg->fc_encap) {
1825 struct lwtunnel_state *lwtstate;
1827 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1828 cfg->fc_encap, AF_INET6, cfg,
1832 rt->dst.lwtstate = lwtstate_get(lwtstate);
1833 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1834 rt->dst.lwtstate->orig_output = rt->dst.output;
1835 rt->dst.output = lwtunnel_output;
1837 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1838 rt->dst.lwtstate->orig_input = rt->dst.input;
1839 rt->dst.input = lwtunnel_input;
1843 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1844 rt->rt6i_dst.plen = cfg->fc_dst_len;
1845 if (rt->rt6i_dst.plen == 128)
1846 rt->dst.flags |= DST_HOST;
1848 #ifdef CONFIG_IPV6_SUBTREES
1849 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1850 rt->rt6i_src.plen = cfg->fc_src_len;
1853 rt->rt6i_metric = cfg->fc_metric;
1855 /* We cannot add true routes via loopback here,
1856 they would result in kernel looping; promote them to reject routes
1858 if ((cfg->fc_flags & RTF_REJECT) ||
1859 (dev && (dev->flags & IFF_LOOPBACK) &&
1860 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1861 !(cfg->fc_flags & RTF_LOCAL))) {
1862 /* hold loopback dev/idev if we haven't done so. */
1863 if (dev != net->loopback_dev) {
1868 dev = net->loopback_dev;
1870 idev = in6_dev_get(dev);
1876 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1877 switch (cfg->fc_type) {
1879 rt->dst.error = -EINVAL;
1880 rt->dst.output = dst_discard_sk;
1881 rt->dst.input = dst_discard;
1884 rt->dst.error = -EACCES;
1885 rt->dst.output = ip6_pkt_prohibit_out;
1886 rt->dst.input = ip6_pkt_prohibit;
1890 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1892 rt->dst.output = ip6_pkt_discard_out;
1893 rt->dst.input = ip6_pkt_discard;
1899 if (cfg->fc_flags & RTF_GATEWAY) {
1900 const struct in6_addr *gw_addr;
1903 gw_addr = &cfg->fc_gateway;
1904 gwa_type = ipv6_addr_type(gw_addr);
1906 /* if gw_addr is local we will fail to detect this in case
1907 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1908 * will return already-added prefix route via interface that
1909 * prefix route was assigned to, which might be non-loopback.
1912 if (ipv6_chk_addr_and_flags(net, gw_addr,
1913 gwa_type & IPV6_ADDR_LINKLOCAL ?
1917 rt->rt6i_gateway = *gw_addr;
1919 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1920 struct rt6_info *grt;
1922 /* IPv6 strictly inhibits using not link-local
1923 addresses as nexthop address.
1924 Otherwise, router will not able to send redirects.
1925 It is very good, but in some (rare!) circumstances
1926 (SIT, PtP, NBMA NOARP links) it is handy to allow
1927 some exceptions. --ANK
1929 if (!(gwa_type & IPV6_ADDR_UNICAST))
1932 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1934 err = -EHOSTUNREACH;
1938 if (dev != grt->dst.dev) {
1944 idev = grt->rt6i_idev;
1946 in6_dev_hold(grt->rt6i_idev);
1948 if (!(grt->rt6i_flags & RTF_GATEWAY))
1956 if (!dev || (dev->flags & IFF_LOOPBACK))
1964 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1965 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1969 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1970 rt->rt6i_prefsrc.plen = 128;
1972 rt->rt6i_prefsrc.plen = 0;
1974 rt->rt6i_flags = cfg->fc_flags;
1978 rt->rt6i_idev = idev;
1979 rt->rt6i_table = table;
1981 cfg->fc_nlinfo.nl_net = dev_net(dev);
1999 int ip6_route_add(struct fib6_config *cfg)
2001 struct mx6_config mxc = { .mx = NULL, };
2002 struct rt6_info *rt = NULL;
2005 err = ip6_route_info_create(cfg, &rt);
2009 err = ip6_convert_metrics(&mxc, cfg);
2013 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2025 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2028 struct fib6_table *table;
2029 struct net *net = dev_net(rt->dst.dev);
2031 if (rt == net->ipv6.ip6_null_entry) {
2036 table = rt->rt6i_table;
2037 write_lock_bh(&table->tb6_lock);
2038 err = fib6_del(rt, info);
2039 write_unlock_bh(&table->tb6_lock);
2046 int ip6_del_rt(struct rt6_info *rt)
2048 struct nl_info info = {
2049 .nl_net = dev_net(rt->dst.dev),
2051 return __ip6_del_rt(rt, &info);
2054 static int ip6_route_del(struct fib6_config *cfg)
2056 struct fib6_table *table;
2057 struct fib6_node *fn;
2058 struct rt6_info *rt;
2061 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2065 read_lock_bh(&table->tb6_lock);
2067 fn = fib6_locate(&table->tb6_root,
2068 &cfg->fc_dst, cfg->fc_dst_len,
2069 &cfg->fc_src, cfg->fc_src_len);
2072 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2073 if ((rt->rt6i_flags & RTF_CACHE) &&
2074 !(cfg->fc_flags & RTF_CACHE))
2076 if (cfg->fc_ifindex &&
2078 rt->dst.dev->ifindex != cfg->fc_ifindex))
2080 if (cfg->fc_flags & RTF_GATEWAY &&
2081 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2083 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2086 read_unlock_bh(&table->tb6_lock);
2088 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2091 read_unlock_bh(&table->tb6_lock);
2096 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2098 struct net *net = dev_net(skb->dev);
2099 struct netevent_redirect netevent;
2100 struct rt6_info *rt, *nrt = NULL;
2101 struct ndisc_options ndopts;
2102 struct inet6_dev *in6_dev;
2103 struct neighbour *neigh;
2105 int optlen, on_link;
2108 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2109 optlen -= sizeof(*msg);
2112 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2116 msg = (struct rd_msg *)icmp6_hdr(skb);
2118 if (ipv6_addr_is_multicast(&msg->dest)) {
2119 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2124 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2126 } else if (ipv6_addr_type(&msg->target) !=
2127 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2128 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2132 in6_dev = __in6_dev_get(skb->dev);
2135 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2139 * The IP source address of the Redirect MUST be the same as the current
2140 * first-hop router for the specified ICMP Destination Address.
2143 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2144 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2149 if (ndopts.nd_opts_tgt_lladdr) {
2150 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2153 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2158 rt = (struct rt6_info *) dst;
2159 if (rt == net->ipv6.ip6_null_entry) {
2160 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2164 /* Redirect received -> path was valid.
2165 * Look, redirects are sent only in response to data packets,
2166 * so that this nexthop apparently is reachable. --ANK
2168 dst_confirm(&rt->dst);
2170 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2175 * We have finally decided to accept it.
2178 neigh_update(neigh, lladdr, NUD_STALE,
2179 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2180 NEIGH_UPDATE_F_OVERRIDE|
2181 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2182 NEIGH_UPDATE_F_ISROUTER))
2185 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2189 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2191 nrt->rt6i_flags &= ~RTF_GATEWAY;
2193 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2195 if (ip6_ins_rt(nrt))
2198 netevent.old = &rt->dst;
2199 netevent.new = &nrt->dst;
2200 netevent.daddr = &msg->dest;
2201 netevent.neigh = neigh;
2202 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2204 if (rt->rt6i_flags & RTF_CACHE) {
2205 rt = (struct rt6_info *) dst_clone(&rt->dst);
2210 neigh_release(neigh);
2214 * Misc support functions
2217 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2219 BUG_ON(from->dst.from);
2221 rt->rt6i_flags &= ~RTF_EXPIRES;
2222 dst_hold(&from->dst);
2223 rt->dst.from = &from->dst;
2224 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2227 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2229 rt->dst.input = ort->dst.input;
2230 rt->dst.output = ort->dst.output;
2231 rt->rt6i_dst = ort->rt6i_dst;
2232 rt->dst.error = ort->dst.error;
2233 rt->rt6i_idev = ort->rt6i_idev;
2235 in6_dev_hold(rt->rt6i_idev);
2236 rt->dst.lastuse = jiffies;
2237 rt->rt6i_gateway = ort->rt6i_gateway;
2238 rt->rt6i_flags = ort->rt6i_flags;
2239 rt6_set_from(rt, ort);
2240 rt->rt6i_metric = ort->rt6i_metric;
2241 #ifdef CONFIG_IPV6_SUBTREES
2242 rt->rt6i_src = ort->rt6i_src;
2244 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2245 rt->rt6i_table = ort->rt6i_table;
2246 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2249 #ifdef CONFIG_IPV6_ROUTE_INFO
2250 static struct rt6_info *rt6_get_route_info(struct net *net,
2251 const struct in6_addr *prefix, int prefixlen,
2252 const struct in6_addr *gwaddr, int ifindex)
2254 struct fib6_node *fn;
2255 struct rt6_info *rt = NULL;
2256 struct fib6_table *table;
2258 table = fib6_get_table(net, RT6_TABLE_INFO);
2262 read_lock_bh(&table->tb6_lock);
2263 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2267 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2268 if (rt->dst.dev->ifindex != ifindex)
2270 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2272 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2278 read_unlock_bh(&table->tb6_lock);
2282 static struct rt6_info *rt6_add_route_info(struct net *net,
2283 const struct in6_addr *prefix, int prefixlen,
2284 const struct in6_addr *gwaddr, int ifindex,
2287 struct fib6_config cfg = {
2288 .fc_table = RT6_TABLE_INFO,
2289 .fc_metric = IP6_RT_PRIO_USER,
2290 .fc_ifindex = ifindex,
2291 .fc_dst_len = prefixlen,
2292 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2293 RTF_UP | RTF_PREF(pref),
2294 .fc_nlinfo.portid = 0,
2295 .fc_nlinfo.nlh = NULL,
2296 .fc_nlinfo.nl_net = net,
2299 cfg.fc_dst = *prefix;
2300 cfg.fc_gateway = *gwaddr;
2302 /* We should treat it as a default route if prefix length is 0. */
2304 cfg.fc_flags |= RTF_DEFAULT;
2306 ip6_route_add(&cfg);
2308 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2312 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2314 struct rt6_info *rt;
2315 struct fib6_table *table;
2317 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2321 read_lock_bh(&table->tb6_lock);
2322 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2323 if (dev == rt->dst.dev &&
2324 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2325 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2330 read_unlock_bh(&table->tb6_lock);
2334 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2335 struct net_device *dev,
2338 struct fib6_config cfg = {
2339 .fc_table = RT6_TABLE_DFLT,
2340 .fc_metric = IP6_RT_PRIO_USER,
2341 .fc_ifindex = dev->ifindex,
2342 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2343 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2344 .fc_nlinfo.portid = 0,
2345 .fc_nlinfo.nlh = NULL,
2346 .fc_nlinfo.nl_net = dev_net(dev),
2349 cfg.fc_gateway = *gwaddr;
2351 ip6_route_add(&cfg);
2353 return rt6_get_dflt_router(gwaddr, dev);
2356 void rt6_purge_dflt_routers(struct net *net)
2358 struct rt6_info *rt;
2359 struct fib6_table *table;
2361 /* NOTE: Keep consistent with rt6_get_dflt_router */
2362 table = fib6_get_table(net, RT6_TABLE_DFLT);
2367 read_lock_bh(&table->tb6_lock);
2368 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2369 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2370 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2372 read_unlock_bh(&table->tb6_lock);
2377 read_unlock_bh(&table->tb6_lock);
2380 static void rtmsg_to_fib6_config(struct net *net,
2381 struct in6_rtmsg *rtmsg,
2382 struct fib6_config *cfg)
2384 memset(cfg, 0, sizeof(*cfg));
2386 cfg->fc_table = RT6_TABLE_MAIN;
2387 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2388 cfg->fc_metric = rtmsg->rtmsg_metric;
2389 cfg->fc_expires = rtmsg->rtmsg_info;
2390 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2391 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2392 cfg->fc_flags = rtmsg->rtmsg_flags;
2394 cfg->fc_nlinfo.nl_net = net;
2396 cfg->fc_dst = rtmsg->rtmsg_dst;
2397 cfg->fc_src = rtmsg->rtmsg_src;
2398 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2401 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2403 struct fib6_config cfg;
2404 struct in6_rtmsg rtmsg;
2408 case SIOCADDRT: /* Add a route */
2409 case SIOCDELRT: /* Delete a route */
2410 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2412 err = copy_from_user(&rtmsg, arg,
2413 sizeof(struct in6_rtmsg));
2417 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2422 err = ip6_route_add(&cfg);
2425 err = ip6_route_del(&cfg);
2439 * Drop the packet on the floor
2442 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2445 struct dst_entry *dst = skb_dst(skb);
2446 switch (ipstats_mib_noroutes) {
2447 case IPSTATS_MIB_INNOROUTES:
2448 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2449 if (type == IPV6_ADDR_ANY) {
2450 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2451 IPSTATS_MIB_INADDRERRORS);
2455 case IPSTATS_MIB_OUTNOROUTES:
2456 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2457 ipstats_mib_noroutes);
2460 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2465 static int ip6_pkt_discard(struct sk_buff *skb)
2467 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2470 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2472 skb->dev = skb_dst(skb)->dev;
2473 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2476 static int ip6_pkt_prohibit(struct sk_buff *skb)
2478 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2481 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2483 skb->dev = skb_dst(skb)->dev;
2484 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2488 * Allocate a dst for local (unicast / anycast) address.
2491 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2492 const struct in6_addr *addr,
2495 struct net *net = dev_net(idev->dev);
2496 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2499 return ERR_PTR(-ENOMEM);
2503 rt->dst.flags |= DST_HOST;
2504 rt->dst.input = ip6_input;
2505 rt->dst.output = ip6_output;
2506 rt->rt6i_idev = idev;
2508 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2510 rt->rt6i_flags |= RTF_ANYCAST;
2512 rt->rt6i_flags |= RTF_LOCAL;
2514 rt->rt6i_gateway = *addr;
2515 rt->rt6i_dst.addr = *addr;
2516 rt->rt6i_dst.plen = 128;
2517 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2519 atomic_set(&rt->dst.__refcnt, 1);
2524 int ip6_route_get_saddr(struct net *net,
2525 struct rt6_info *rt,
2526 const struct in6_addr *daddr,
2528 struct in6_addr *saddr)
2530 struct inet6_dev *idev =
2531 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2533 if (rt && rt->rt6i_prefsrc.plen)
2534 *saddr = rt->rt6i_prefsrc.addr;
2536 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2537 daddr, prefs, saddr);
2541 /* remove deleted ip from prefsrc entries */
2542 struct arg_dev_net_ip {
2543 struct net_device *dev;
2545 struct in6_addr *addr;
2548 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2550 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2551 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2552 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2554 if (((void *)rt->dst.dev == dev || !dev) &&
2555 rt != net->ipv6.ip6_null_entry &&
2556 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2557 /* remove prefsrc entry */
2558 rt->rt6i_prefsrc.plen = 0;
2563 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2565 struct net *net = dev_net(ifp->idev->dev);
2566 struct arg_dev_net_ip adni = {
2567 .dev = ifp->idev->dev,
2571 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2574 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2575 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2577 /* Remove routers and update dst entries when gateway turn into host. */
2578 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2580 struct in6_addr *gateway = (struct in6_addr *)arg;
2582 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2583 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2584 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2590 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2592 fib6_clean_all(net, fib6_clean_tohost, gateway);
2595 struct arg_dev_net {
2596 struct net_device *dev;
2600 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2602 const struct arg_dev_net *adn = arg;
2603 const struct net_device *dev = adn->dev;
2605 if ((rt->dst.dev == dev || !dev) &&
2606 rt != adn->net->ipv6.ip6_null_entry)
2612 void rt6_ifdown(struct net *net, struct net_device *dev)
2614 struct arg_dev_net adn = {
2619 fib6_clean_all(net, fib6_ifdown, &adn);
2620 icmp6_clean_all(fib6_ifdown, &adn);
2621 rt6_uncached_list_flush_dev(net, dev);
2624 struct rt6_mtu_change_arg {
2625 struct net_device *dev;
2629 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2631 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2632 struct inet6_dev *idev;
2634 /* In IPv6 pmtu discovery is not optional,
2635 so that RTAX_MTU lock cannot disable it.
2636 We still use this lock to block changes
2637 caused by addrconf/ndisc.
2640 idev = __in6_dev_get(arg->dev);
2644 /* For administrative MTU increase, there is no way to discover
2645 IPv6 PMTU increase, so PMTU increase should be updated here.
2646 Since RFC 1981 doesn't include administrative MTU increase
2647 update PMTU increase is a MUST. (i.e. jumbo frame)
2650 If new MTU is less than route PMTU, this new MTU will be the
2651 lowest MTU in the path, update the route PMTU to reflect PMTU
2652 decreases; if new MTU is greater than route PMTU, and the
2653 old MTU is the lowest MTU in the path, update the route PMTU
2654 to reflect the increase. In this case if the other nodes' MTU
2655 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2658 if (rt->dst.dev == arg->dev &&
2659 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2660 if (rt->rt6i_flags & RTF_CACHE) {
2661 /* For RTF_CACHE with rt6i_pmtu == 0
2662 * (i.e. a redirected route),
2663 * the metrics of its rt->dst.from has already
2666 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2667 rt->rt6i_pmtu = arg->mtu;
2668 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2669 (dst_mtu(&rt->dst) < arg->mtu &&
2670 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2671 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2677 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2679 struct rt6_mtu_change_arg arg = {
2684 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2687 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2688 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2689 [RTA_OIF] = { .type = NLA_U32 },
2690 [RTA_IIF] = { .type = NLA_U32 },
2691 [RTA_PRIORITY] = { .type = NLA_U32 },
2692 [RTA_METRICS] = { .type = NLA_NESTED },
2693 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2694 [RTA_PREF] = { .type = NLA_U8 },
2695 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2696 [RTA_ENCAP] = { .type = NLA_NESTED },
2699 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2700 struct fib6_config *cfg)
2703 struct nlattr *tb[RTA_MAX+1];
2707 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2712 rtm = nlmsg_data(nlh);
2713 memset(cfg, 0, sizeof(*cfg));
2715 cfg->fc_table = rtm->rtm_table;
2716 cfg->fc_dst_len = rtm->rtm_dst_len;
2717 cfg->fc_src_len = rtm->rtm_src_len;
2718 cfg->fc_flags = RTF_UP;
2719 cfg->fc_protocol = rtm->rtm_protocol;
2720 cfg->fc_type = rtm->rtm_type;
2722 if (rtm->rtm_type == RTN_UNREACHABLE ||
2723 rtm->rtm_type == RTN_BLACKHOLE ||
2724 rtm->rtm_type == RTN_PROHIBIT ||
2725 rtm->rtm_type == RTN_THROW)
2726 cfg->fc_flags |= RTF_REJECT;
2728 if (rtm->rtm_type == RTN_LOCAL)
2729 cfg->fc_flags |= RTF_LOCAL;
2731 if (rtm->rtm_flags & RTM_F_CLONED)
2732 cfg->fc_flags |= RTF_CACHE;
2734 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2735 cfg->fc_nlinfo.nlh = nlh;
2736 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2738 if (tb[RTA_GATEWAY]) {
2739 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2740 cfg->fc_flags |= RTF_GATEWAY;
2744 int plen = (rtm->rtm_dst_len + 7) >> 3;
2746 if (nla_len(tb[RTA_DST]) < plen)
2749 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2753 int plen = (rtm->rtm_src_len + 7) >> 3;
2755 if (nla_len(tb[RTA_SRC]) < plen)
2758 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2761 if (tb[RTA_PREFSRC])
2762 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2765 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2767 if (tb[RTA_PRIORITY])
2768 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2770 if (tb[RTA_METRICS]) {
2771 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2772 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2776 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2778 if (tb[RTA_MULTIPATH]) {
2779 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2780 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2784 pref = nla_get_u8(tb[RTA_PREF]);
2785 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2786 pref != ICMPV6_ROUTER_PREF_HIGH)
2787 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2788 cfg->fc_flags |= RTF_PREF(pref);
2792 cfg->fc_encap = tb[RTA_ENCAP];
2794 if (tb[RTA_ENCAP_TYPE])
2795 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2803 struct rt6_info *rt6_info;
2804 struct fib6_config r_cfg;
2805 struct mx6_config mxc;
2806 struct list_head next;
2809 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2813 list_for_each_entry(nh, rt6_nh_list, next) {
2814 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2815 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2816 nh->r_cfg.fc_ifindex);
2820 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2821 struct rt6_info *rt, struct fib6_config *r_cfg)
2824 struct rt6_info *rtnh;
2827 list_for_each_entry(nh, rt6_nh_list, next) {
2828 /* check if rt6_info already exists */
2829 rtnh = nh->rt6_info;
2831 if (rtnh->dst.dev == rt->dst.dev &&
2832 rtnh->rt6i_idev == rt->rt6i_idev &&
2833 ipv6_addr_equal(&rtnh->rt6i_gateway,
2838 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2842 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2847 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2848 list_add_tail(&nh->next, rt6_nh_list);
2853 static int ip6_route_multipath_add(struct fib6_config *cfg)
2855 struct fib6_config r_cfg;
2856 struct rtnexthop *rtnh;
2857 struct rt6_info *rt;
2858 struct rt6_nh *err_nh;
2859 struct rt6_nh *nh, *nh_safe;
2864 int replace = (cfg->fc_nlinfo.nlh &&
2865 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2866 LIST_HEAD(rt6_nh_list);
2868 remaining = cfg->fc_mp_len;
2869 rtnh = (struct rtnexthop *)cfg->fc_mp;
2871 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2872 * rt6_info structs per nexthop
2874 while (rtnh_ok(rtnh, remaining)) {
2875 memcpy(&r_cfg, cfg, sizeof(*cfg));
2876 if (rtnh->rtnh_ifindex)
2877 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2879 attrlen = rtnh_attrlen(rtnh);
2881 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2883 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2885 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2886 r_cfg.fc_flags |= RTF_GATEWAY;
2888 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2889 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2891 r_cfg.fc_encap_type = nla_get_u16(nla);
2894 err = ip6_route_info_create(&r_cfg, &rt);
2898 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2904 rtnh = rtnh_next(rtnh, &remaining);
2908 list_for_each_entry(nh, &rt6_nh_list, next) {
2909 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2910 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2911 nh->rt6_info = NULL;
2914 ip6_print_replace_route_err(&rt6_nh_list);
2919 /* Because each route is added like a single route we remove
2920 * these flags after the first nexthop: if there is a collision,
2921 * we have already failed to add the first nexthop:
2922 * fib6_add_rt2node() has rejected it; when replacing, old
2923 * nexthops have been replaced by first new, the rest should
2926 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2934 /* Delete routes that were already added */
2935 list_for_each_entry(nh, &rt6_nh_list, next) {
2938 ip6_route_del(&nh->r_cfg);
2942 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2944 dst_free(&nh->rt6_info->dst);
2946 list_del(&nh->next);
2953 static int ip6_route_multipath_del(struct fib6_config *cfg)
2955 struct fib6_config r_cfg;
2956 struct rtnexthop *rtnh;
2959 int err = 1, last_err = 0;
2961 remaining = cfg->fc_mp_len;
2962 rtnh = (struct rtnexthop *)cfg->fc_mp;
2964 /* Parse a Multipath Entry */
2965 while (rtnh_ok(rtnh, remaining)) {
2966 memcpy(&r_cfg, cfg, sizeof(*cfg));
2967 if (rtnh->rtnh_ifindex)
2968 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2970 attrlen = rtnh_attrlen(rtnh);
2972 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2974 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2976 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2977 r_cfg.fc_flags |= RTF_GATEWAY;
2980 err = ip6_route_del(&r_cfg);
2984 rtnh = rtnh_next(rtnh, &remaining);
2990 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2992 struct fib6_config cfg;
2995 err = rtm_to_fib6_config(skb, nlh, &cfg);
3000 return ip6_route_multipath_del(&cfg);
3002 return ip6_route_del(&cfg);
3005 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3007 struct fib6_config cfg;
3010 err = rtm_to_fib6_config(skb, nlh, &cfg);
3015 return ip6_route_multipath_add(&cfg);
3017 return ip6_route_add(&cfg);
3020 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3022 return NLMSG_ALIGN(sizeof(struct rtmsg))
3023 + nla_total_size(16) /* RTA_SRC */
3024 + nla_total_size(16) /* RTA_DST */
3025 + nla_total_size(16) /* RTA_GATEWAY */
3026 + nla_total_size(16) /* RTA_PREFSRC */
3027 + nla_total_size(4) /* RTA_TABLE */
3028 + nla_total_size(4) /* RTA_IIF */
3029 + nla_total_size(4) /* RTA_OIF */
3030 + nla_total_size(4) /* RTA_PRIORITY */
3031 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3032 + nla_total_size(sizeof(struct rta_cacheinfo))
3033 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3034 + nla_total_size(1) /* RTA_PREF */
3035 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3038 static int rt6_fill_node(struct net *net,
3039 struct sk_buff *skb, struct rt6_info *rt,
3040 struct in6_addr *dst, struct in6_addr *src,
3041 int iif, int type, u32 portid, u32 seq,
3042 int prefix, int nowait, unsigned int flags)
3044 u32 metrics[RTAX_MAX];
3046 struct nlmsghdr *nlh;
3050 if (prefix) { /* user wants prefix routes only */
3051 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3052 /* success since this is not a prefix route */
3057 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3061 rtm = nlmsg_data(nlh);
3062 rtm->rtm_family = AF_INET6;
3063 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3064 rtm->rtm_src_len = rt->rt6i_src.plen;
3067 table = rt->rt6i_table->tb6_id;
3069 table = RT6_TABLE_UNSPEC;
3070 rtm->rtm_table = table;
3071 if (nla_put_u32(skb, RTA_TABLE, table))
3072 goto nla_put_failure;
3073 if (rt->rt6i_flags & RTF_REJECT) {
3074 switch (rt->dst.error) {
3076 rtm->rtm_type = RTN_BLACKHOLE;
3079 rtm->rtm_type = RTN_PROHIBIT;
3082 rtm->rtm_type = RTN_THROW;
3085 rtm->rtm_type = RTN_UNREACHABLE;
3089 else if (rt->rt6i_flags & RTF_LOCAL)
3090 rtm->rtm_type = RTN_LOCAL;
3091 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3092 rtm->rtm_type = RTN_LOCAL;
3094 rtm->rtm_type = RTN_UNICAST;
3096 if (!netif_carrier_ok(rt->dst.dev)) {
3097 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3098 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3099 rtm->rtm_flags |= RTNH_F_DEAD;
3101 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3102 rtm->rtm_protocol = rt->rt6i_protocol;
3103 if (rt->rt6i_flags & RTF_DYNAMIC)
3104 rtm->rtm_protocol = RTPROT_REDIRECT;
3105 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3106 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3107 rtm->rtm_protocol = RTPROT_RA;
3109 rtm->rtm_protocol = RTPROT_KERNEL;
3112 if (rt->rt6i_flags & RTF_CACHE)
3113 rtm->rtm_flags |= RTM_F_CLONED;
3116 if (nla_put_in6_addr(skb, RTA_DST, dst))
3117 goto nla_put_failure;
3118 rtm->rtm_dst_len = 128;
3119 } else if (rtm->rtm_dst_len)
3120 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3121 goto nla_put_failure;
3122 #ifdef CONFIG_IPV6_SUBTREES
3124 if (nla_put_in6_addr(skb, RTA_SRC, src))
3125 goto nla_put_failure;
3126 rtm->rtm_src_len = 128;
3127 } else if (rtm->rtm_src_len &&
3128 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3129 goto nla_put_failure;
3132 #ifdef CONFIG_IPV6_MROUTE
3133 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3134 int err = ip6mr_get_route(net, skb, rtm, nowait);
3139 goto nla_put_failure;
3141 if (err == -EMSGSIZE)
3142 goto nla_put_failure;
3147 if (nla_put_u32(skb, RTA_IIF, iif))
3148 goto nla_put_failure;
3150 struct in6_addr saddr_buf;
3151 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3152 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3153 goto nla_put_failure;
3156 if (rt->rt6i_prefsrc.plen) {
3157 struct in6_addr saddr_buf;
3158 saddr_buf = rt->rt6i_prefsrc.addr;
3159 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3160 goto nla_put_failure;
3163 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3165 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3166 if (rtnetlink_put_metrics(skb, metrics) < 0)
3167 goto nla_put_failure;
3169 if (rt->rt6i_flags & RTF_GATEWAY) {
3170 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3171 goto nla_put_failure;
3175 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3176 goto nla_put_failure;
3177 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3178 goto nla_put_failure;
3180 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3182 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3183 goto nla_put_failure;
3185 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3186 goto nla_put_failure;
3188 lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3190 nlmsg_end(skb, nlh);
3194 nlmsg_cancel(skb, nlh);
3198 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3200 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3203 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3204 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3205 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3209 return rt6_fill_node(arg->net,
3210 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3211 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3212 prefix, 0, NLM_F_MULTI);
3215 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3217 struct net *net = sock_net(in_skb->sk);
3218 struct nlattr *tb[RTA_MAX+1];
3219 struct rt6_info *rt;
3220 struct sk_buff *skb;
3223 int err, iif = 0, oif = 0;
3225 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3230 memset(&fl6, 0, sizeof(fl6));
3233 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3236 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3240 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3243 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3247 iif = nla_get_u32(tb[RTA_IIF]);
3250 oif = nla_get_u32(tb[RTA_OIF]);
3253 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3256 struct net_device *dev;
3259 dev = __dev_get_by_index(net, iif);
3265 fl6.flowi6_iif = iif;
3267 if (!ipv6_addr_any(&fl6.saddr))
3268 flags |= RT6_LOOKUP_F_HAS_SADDR;
3270 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3273 fl6.flowi6_oif = oif;
3275 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3278 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3285 /* Reserve room for dummy headers, this skb can pass
3286 through good chunk of routing engine.
3288 skb_reset_mac_header(skb);
3289 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3291 skb_dst_set(skb, &rt->dst);
3293 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3294 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3295 nlh->nlmsg_seq, 0, 0, 0);
3301 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3306 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
3308 struct sk_buff *skb;
3309 struct net *net = info->nl_net;
3314 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3316 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3320 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3321 event, info->portid, seq, 0, 0, 0);
3323 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3324 WARN_ON(err == -EMSGSIZE);
3328 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3329 info->nlh, gfp_any());
3333 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3336 static int ip6_route_dev_notify(struct notifier_block *this,
3337 unsigned long event, void *ptr)
3339 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3340 struct net *net = dev_net(dev);
3342 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3343 net->ipv6.ip6_null_entry->dst.dev = dev;
3344 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3345 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3346 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3347 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3348 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3349 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3360 #ifdef CONFIG_PROC_FS
3362 static const struct file_operations ipv6_route_proc_fops = {
3363 .owner = THIS_MODULE,
3364 .open = ipv6_route_open,
3366 .llseek = seq_lseek,
3367 .release = seq_release_net,
3370 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3372 struct net *net = (struct net *)seq->private;
3373 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3374 net->ipv6.rt6_stats->fib_nodes,
3375 net->ipv6.rt6_stats->fib_route_nodes,
3376 net->ipv6.rt6_stats->fib_rt_alloc,
3377 net->ipv6.rt6_stats->fib_rt_entries,
3378 net->ipv6.rt6_stats->fib_rt_cache,
3379 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3380 net->ipv6.rt6_stats->fib_discarded_routes);
3385 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3387 return single_open_net(inode, file, rt6_stats_seq_show);
3390 static const struct file_operations rt6_stats_seq_fops = {
3391 .owner = THIS_MODULE,
3392 .open = rt6_stats_seq_open,
3394 .llseek = seq_lseek,
3395 .release = single_release_net,
3397 #endif /* CONFIG_PROC_FS */
3399 #ifdef CONFIG_SYSCTL
3402 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3403 void __user *buffer, size_t *lenp, loff_t *ppos)
3410 net = (struct net *)ctl->extra1;
3411 delay = net->ipv6.sysctl.flush_delay;
3412 proc_dointvec(ctl, write, buffer, lenp, ppos);
3413 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3417 struct ctl_table ipv6_route_table_template[] = {
3419 .procname = "flush",
3420 .data = &init_net.ipv6.sysctl.flush_delay,
3421 .maxlen = sizeof(int),
3423 .proc_handler = ipv6_sysctl_rtcache_flush
3426 .procname = "gc_thresh",
3427 .data = &ip6_dst_ops_template.gc_thresh,
3428 .maxlen = sizeof(int),
3430 .proc_handler = proc_dointvec,
3433 .procname = "max_size",
3434 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3435 .maxlen = sizeof(int),
3437 .proc_handler = proc_dointvec,
3440 .procname = "gc_min_interval",
3441 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3442 .maxlen = sizeof(int),
3444 .proc_handler = proc_dointvec_jiffies,
3447 .procname = "gc_timeout",
3448 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3449 .maxlen = sizeof(int),
3451 .proc_handler = proc_dointvec_jiffies,
3454 .procname = "gc_interval",
3455 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3456 .maxlen = sizeof(int),
3458 .proc_handler = proc_dointvec_jiffies,
3461 .procname = "gc_elasticity",
3462 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3463 .maxlen = sizeof(int),
3465 .proc_handler = proc_dointvec,
3468 .procname = "mtu_expires",
3469 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3470 .maxlen = sizeof(int),
3472 .proc_handler = proc_dointvec_jiffies,
3475 .procname = "min_adv_mss",
3476 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3477 .maxlen = sizeof(int),
3479 .proc_handler = proc_dointvec,
3482 .procname = "gc_min_interval_ms",
3483 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3484 .maxlen = sizeof(int),
3486 .proc_handler = proc_dointvec_ms_jiffies,
3491 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3493 struct ctl_table *table;
3495 table = kmemdup(ipv6_route_table_template,
3496 sizeof(ipv6_route_table_template),
3500 table[0].data = &net->ipv6.sysctl.flush_delay;
3501 table[0].extra1 = net;
3502 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3503 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3504 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3505 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3506 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3507 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3508 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3509 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3510 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3512 /* Don't export sysctls to unprivileged users */
3513 if (net->user_ns != &init_user_ns)
3514 table[0].procname = NULL;
3521 static int __net_init ip6_route_net_init(struct net *net)
3525 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3526 sizeof(net->ipv6.ip6_dst_ops));
3528 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3529 goto out_ip6_dst_ops;
3531 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3532 sizeof(*net->ipv6.ip6_null_entry),
3534 if (!net->ipv6.ip6_null_entry)
3535 goto out_ip6_dst_entries;
3536 net->ipv6.ip6_null_entry->dst.path =
3537 (struct dst_entry *)net->ipv6.ip6_null_entry;
3538 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3539 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3540 ip6_template_metrics, true);
3542 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3543 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3544 sizeof(*net->ipv6.ip6_prohibit_entry),
3546 if (!net->ipv6.ip6_prohibit_entry)
3547 goto out_ip6_null_entry;
3548 net->ipv6.ip6_prohibit_entry->dst.path =
3549 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3550 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3551 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3552 ip6_template_metrics, true);
3554 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3555 sizeof(*net->ipv6.ip6_blk_hole_entry),
3557 if (!net->ipv6.ip6_blk_hole_entry)
3558 goto out_ip6_prohibit_entry;
3559 net->ipv6.ip6_blk_hole_entry->dst.path =
3560 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3561 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3562 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3563 ip6_template_metrics, true);
3566 net->ipv6.sysctl.flush_delay = 0;
3567 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3568 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3569 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3570 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3571 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3572 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3573 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3575 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3581 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3582 out_ip6_prohibit_entry:
3583 kfree(net->ipv6.ip6_prohibit_entry);
3585 kfree(net->ipv6.ip6_null_entry);
3587 out_ip6_dst_entries:
3588 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3593 static void __net_exit ip6_route_net_exit(struct net *net)
3595 kfree(net->ipv6.ip6_null_entry);
3596 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3597 kfree(net->ipv6.ip6_prohibit_entry);
3598 kfree(net->ipv6.ip6_blk_hole_entry);
3600 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3603 static int __net_init ip6_route_net_init_late(struct net *net)
3605 #ifdef CONFIG_PROC_FS
3606 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3607 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3612 static void __net_exit ip6_route_net_exit_late(struct net *net)
3614 #ifdef CONFIG_PROC_FS
3615 remove_proc_entry("ipv6_route", net->proc_net);
3616 remove_proc_entry("rt6_stats", net->proc_net);
3620 static struct pernet_operations ip6_route_net_ops = {
3621 .init = ip6_route_net_init,
3622 .exit = ip6_route_net_exit,
3625 static int __net_init ipv6_inetpeer_init(struct net *net)
3627 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3631 inet_peer_base_init(bp);
3632 net->ipv6.peers = bp;
3636 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3638 struct inet_peer_base *bp = net->ipv6.peers;
3640 net->ipv6.peers = NULL;
3641 inetpeer_invalidate_tree(bp);
3645 static struct pernet_operations ipv6_inetpeer_ops = {
3646 .init = ipv6_inetpeer_init,
3647 .exit = ipv6_inetpeer_exit,
3650 static struct pernet_operations ip6_route_net_late_ops = {
3651 .init = ip6_route_net_init_late,
3652 .exit = ip6_route_net_exit_late,
3655 static struct notifier_block ip6_route_dev_notifier = {
3656 .notifier_call = ip6_route_dev_notify,
3660 int __init ip6_route_init(void)
3666 ip6_dst_ops_template.kmem_cachep =
3667 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3668 SLAB_HWCACHE_ALIGN, NULL);
3669 if (!ip6_dst_ops_template.kmem_cachep)
3672 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3674 goto out_kmem_cache;
3676 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3678 goto out_dst_entries;
3680 ret = register_pernet_subsys(&ip6_route_net_ops);
3682 goto out_register_inetpeer;
3684 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3686 /* Registering of the loopback is done before this portion of code,
3687 * the loopback reference in rt6_info will not be taken, do it
3688 * manually for init_net */
3689 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3690 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3691 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3692 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3693 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3694 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3695 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3699 goto out_register_subsys;
3705 ret = fib6_rules_init();
3709 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3711 goto fib6_rules_init;
3714 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3715 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3716 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3717 goto out_register_late_subsys;
3719 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3721 goto out_register_late_subsys;
3723 for_each_possible_cpu(cpu) {
3724 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3726 INIT_LIST_HEAD(&ul->head);
3727 spin_lock_init(&ul->lock);
3733 out_register_late_subsys:
3734 unregister_pernet_subsys(&ip6_route_net_late_ops);
3736 fib6_rules_cleanup();
3741 out_register_subsys:
3742 unregister_pernet_subsys(&ip6_route_net_ops);
3743 out_register_inetpeer:
3744 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3746 dst_entries_destroy(&ip6_dst_blackhole_ops);
3748 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3752 void ip6_route_cleanup(void)
3754 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3755 unregister_pernet_subsys(&ip6_route_net_late_ops);
3756 fib6_rules_cleanup();
3759 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3760 unregister_pernet_subsys(&ip6_route_net_ops);
3761 dst_entries_destroy(&ip6_dst_blackhole_ops);
3762 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);