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 <linux/jhash.h>
48 #include <net/net_namespace.h>
51 #include <net/ip6_fib.h>
52 #include <net/ip6_route.h>
53 #include <net/ndisc.h>
54 #include <net/addrconf.h>
56 #include <linux/rtnetlink.h>
58 #include <net/dst_metadata.h>
60 #include <net/netevent.h>
61 #include <net/netlink.h>
62 #include <net/nexthop.h>
63 #include <net/lwtunnel.h>
64 #include <net/ip_tunnels.h>
65 #include <net/l3mdev.h>
67 #include <linux/uaccess.h>
70 #include <linux/sysctl.h>
73 static int ip6_rt_type_to_error(u8 fib6_type);
75 #define CREATE_TRACE_POINTS
76 #include <trace/events/fib6.h>
77 EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
78 #undef CREATE_TRACE_POINTS
81 RT6_NUD_FAIL_HARD = -3,
82 RT6_NUD_FAIL_PROBE = -2,
83 RT6_NUD_FAIL_DO_RR = -1,
87 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
88 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
89 static unsigned int ip6_mtu(const struct dst_entry *dst);
90 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
91 static void ip6_dst_destroy(struct dst_entry *);
92 static void ip6_dst_ifdown(struct dst_entry *,
93 struct net_device *dev, int how);
94 static int ip6_dst_gc(struct dst_ops *ops);
96 static int ip6_pkt_discard(struct sk_buff *skb);
97 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
98 static int ip6_pkt_prohibit(struct sk_buff *skb);
99 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
100 static void ip6_link_failure(struct sk_buff *skb);
101 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
102 struct sk_buff *skb, u32 mtu);
103 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
104 struct sk_buff *skb);
105 static int rt6_score_route(struct fib6_info *rt, int oif, int strict);
106 static size_t rt6_nlmsg_size(struct fib6_info *rt);
107 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
108 struct fib6_info *rt, struct dst_entry *dst,
109 struct in6_addr *dest, struct in6_addr *src,
110 int iif, int type, u32 portid, u32 seq,
112 static struct rt6_info *rt6_find_cached_rt(struct fib6_info *rt,
113 struct in6_addr *daddr,
114 struct in6_addr *saddr);
116 #ifdef CONFIG_IPV6_ROUTE_INFO
117 static struct fib6_info *rt6_add_route_info(struct net *net,
118 const struct in6_addr *prefix, int prefixlen,
119 const struct in6_addr *gwaddr,
120 struct net_device *dev,
122 static struct fib6_info *rt6_get_route_info(struct net *net,
123 const struct in6_addr *prefix, int prefixlen,
124 const struct in6_addr *gwaddr,
125 struct net_device *dev);
128 struct uncached_list {
130 struct list_head head;
133 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
135 void rt6_uncached_list_add(struct rt6_info *rt)
137 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
139 rt->rt6i_uncached_list = ul;
141 spin_lock_bh(&ul->lock);
142 list_add_tail(&rt->rt6i_uncached, &ul->head);
143 spin_unlock_bh(&ul->lock);
146 void rt6_uncached_list_del(struct rt6_info *rt)
148 if (!list_empty(&rt->rt6i_uncached)) {
149 struct uncached_list *ul = rt->rt6i_uncached_list;
150 struct net *net = dev_net(rt->dst.dev);
152 spin_lock_bh(&ul->lock);
153 list_del(&rt->rt6i_uncached);
154 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
155 spin_unlock_bh(&ul->lock);
159 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
161 struct net_device *loopback_dev = net->loopback_dev;
164 if (dev == loopback_dev)
167 for_each_possible_cpu(cpu) {
168 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
171 spin_lock_bh(&ul->lock);
172 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
173 struct inet6_dev *rt_idev = rt->rt6i_idev;
174 struct net_device *rt_dev = rt->dst.dev;
176 if (rt_idev->dev == dev) {
177 rt->rt6i_idev = in6_dev_get(loopback_dev);
178 in6_dev_put(rt_idev);
182 rt->dst.dev = loopback_dev;
183 dev_hold(rt->dst.dev);
187 spin_unlock_bh(&ul->lock);
191 static inline const void *choose_neigh_daddr(const struct in6_addr *p,
195 if (!ipv6_addr_any(p))
196 return (const void *) p;
198 return &ipv6_hdr(skb)->daddr;
202 struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
203 struct net_device *dev,
209 daddr = choose_neigh_daddr(gw, skb, daddr);
210 n = __ipv6_neigh_lookup(dev, daddr);
213 return neigh_create(&nd_tbl, daddr, dev);
216 static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
220 const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
222 return ip6_neigh_lookup(&rt->rt6i_gateway, dst->dev, skb, daddr);
225 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
227 struct net_device *dev = dst->dev;
228 struct rt6_info *rt = (struct rt6_info *)dst;
230 daddr = choose_neigh_daddr(&rt->rt6i_gateway, NULL, daddr);
233 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
235 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
237 __ipv6_confirm_neigh(dev, daddr);
240 static struct dst_ops ip6_dst_ops_template = {
244 .check = ip6_dst_check,
245 .default_advmss = ip6_default_advmss,
247 .cow_metrics = dst_cow_metrics_generic,
248 .destroy = ip6_dst_destroy,
249 .ifdown = ip6_dst_ifdown,
250 .negative_advice = ip6_negative_advice,
251 .link_failure = ip6_link_failure,
252 .update_pmtu = ip6_rt_update_pmtu,
253 .redirect = rt6_do_redirect,
254 .local_out = __ip6_local_out,
255 .neigh_lookup = ip6_dst_neigh_lookup,
256 .confirm_neigh = ip6_confirm_neigh,
259 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
261 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
263 return mtu ? : dst->dev->mtu;
266 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
267 struct sk_buff *skb, u32 mtu)
271 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
276 static struct dst_ops ip6_dst_blackhole_ops = {
278 .destroy = ip6_dst_destroy,
279 .check = ip6_dst_check,
280 .mtu = ip6_blackhole_mtu,
281 .default_advmss = ip6_default_advmss,
282 .update_pmtu = ip6_rt_blackhole_update_pmtu,
283 .redirect = ip6_rt_blackhole_redirect,
284 .cow_metrics = dst_cow_metrics_generic,
285 .neigh_lookup = ip6_dst_neigh_lookup,
288 static const u32 ip6_template_metrics[RTAX_MAX] = {
289 [RTAX_HOPLIMIT - 1] = 0,
292 static const struct fib6_info fib6_null_entry_template = {
293 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
294 .fib6_protocol = RTPROT_KERNEL,
295 .fib6_metric = ~(u32)0,
296 .fib6_ref = ATOMIC_INIT(1),
297 .fib6_type = RTN_UNREACHABLE,
298 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
301 static const struct rt6_info ip6_null_entry_template = {
303 .__refcnt = ATOMIC_INIT(1),
305 .obsolete = DST_OBSOLETE_FORCE_CHK,
306 .error = -ENETUNREACH,
307 .input = ip6_pkt_discard,
308 .output = ip6_pkt_discard_out,
310 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
313 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
315 static const struct rt6_info ip6_prohibit_entry_template = {
317 .__refcnt = ATOMIC_INIT(1),
319 .obsolete = DST_OBSOLETE_FORCE_CHK,
321 .input = ip6_pkt_prohibit,
322 .output = ip6_pkt_prohibit_out,
324 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
327 static const struct rt6_info ip6_blk_hole_entry_template = {
329 .__refcnt = ATOMIC_INIT(1),
331 .obsolete = DST_OBSOLETE_FORCE_CHK,
333 .input = dst_discard,
334 .output = dst_discard_out,
336 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
341 static void rt6_info_init(struct rt6_info *rt)
343 struct dst_entry *dst = &rt->dst;
345 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
346 INIT_LIST_HEAD(&rt->rt6i_uncached);
349 /* allocate dst with ip6_dst_ops */
350 struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
353 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
354 1, DST_OBSOLETE_FORCE_CHK, flags);
358 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
363 EXPORT_SYMBOL(ip6_dst_alloc);
365 static void ip6_dst_destroy(struct dst_entry *dst)
367 struct rt6_info *rt = (struct rt6_info *)dst;
368 struct fib6_info *from;
369 struct inet6_dev *idev;
371 ip_dst_metrics_put(dst);
372 rt6_uncached_list_del(rt);
374 idev = rt->rt6i_idev;
376 rt->rt6i_idev = NULL;
381 from = rcu_dereference(rt->from);
382 rcu_assign_pointer(rt->from, NULL);
383 fib6_info_release(from);
387 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
390 struct rt6_info *rt = (struct rt6_info *)dst;
391 struct inet6_dev *idev = rt->rt6i_idev;
392 struct net_device *loopback_dev =
393 dev_net(dev)->loopback_dev;
395 if (idev && idev->dev != loopback_dev) {
396 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
398 rt->rt6i_idev = loopback_idev;
404 static bool __rt6_check_expired(const struct rt6_info *rt)
406 if (rt->rt6i_flags & RTF_EXPIRES)
407 return time_after(jiffies, rt->dst.expires);
412 static bool rt6_check_expired(const struct rt6_info *rt)
414 struct fib6_info *from;
416 from = rcu_dereference(rt->from);
418 if (rt->rt6i_flags & RTF_EXPIRES) {
419 if (time_after(jiffies, rt->dst.expires))
422 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
423 fib6_check_expired(from);
428 struct fib6_info *fib6_multipath_select(const struct net *net,
429 struct fib6_info *match,
430 struct flowi6 *fl6, int oif,
431 const struct sk_buff *skb,
434 struct fib6_info *sibling, *next_sibling;
436 /* We might have already computed the hash for ICMPv6 errors. In such
437 * case it will always be non-zero. Otherwise now is the time to do it.
440 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
442 if (fl6->mp_hash <= atomic_read(&match->fib6_nh.nh_upper_bound))
445 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
449 nh_upper_bound = atomic_read(&sibling->fib6_nh.nh_upper_bound);
450 if (fl6->mp_hash > nh_upper_bound)
452 if (rt6_score_route(sibling, oif, strict) < 0)
462 * Route lookup. rcu_read_lock() should be held.
465 static inline struct fib6_info *rt6_device_match(struct net *net,
466 struct fib6_info *rt,
467 const struct in6_addr *saddr,
471 struct fib6_info *sprt;
473 if (!oif && ipv6_addr_any(saddr) &&
474 !(rt->fib6_nh.nh_flags & RTNH_F_DEAD))
477 for (sprt = rt; sprt; sprt = rcu_dereference(sprt->fib6_next)) {
478 const struct net_device *dev = sprt->fib6_nh.nh_dev;
480 if (sprt->fib6_nh.nh_flags & RTNH_F_DEAD)
484 if (dev->ifindex == oif)
487 if (ipv6_chk_addr(net, saddr, dev,
488 flags & RT6_LOOKUP_F_IFACE))
493 if (oif && flags & RT6_LOOKUP_F_IFACE)
494 return net->ipv6.fib6_null_entry;
496 return rt->fib6_nh.nh_flags & RTNH_F_DEAD ? net->ipv6.fib6_null_entry : rt;
499 #ifdef CONFIG_IPV6_ROUTER_PREF
500 struct __rt6_probe_work {
501 struct work_struct work;
502 struct in6_addr target;
503 struct net_device *dev;
506 static void rt6_probe_deferred(struct work_struct *w)
508 struct in6_addr mcaddr;
509 struct __rt6_probe_work *work =
510 container_of(w, struct __rt6_probe_work, work);
512 addrconf_addr_solict_mult(&work->target, &mcaddr);
513 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
518 static void rt6_probe(struct fib6_info *rt)
520 struct __rt6_probe_work *work = NULL;
521 const struct in6_addr *nh_gw;
522 struct neighbour *neigh;
523 struct net_device *dev;
524 struct inet6_dev *idev;
527 * Okay, this does not seem to be appropriate
528 * for now, however, we need to check if it
529 * is really so; aka Router Reachability Probing.
531 * Router Reachability Probe MUST be rate-limited
532 * to no more than one per minute.
534 if (!rt || !(rt->fib6_flags & RTF_GATEWAY))
537 nh_gw = &rt->fib6_nh.nh_gw;
538 dev = rt->fib6_nh.nh_dev;
540 idev = __in6_dev_get(dev);
541 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
543 if (neigh->nud_state & NUD_VALID)
546 write_lock(&neigh->lock);
547 if (!(neigh->nud_state & NUD_VALID) &&
549 neigh->updated + idev->cnf.rtr_probe_interval)) {
550 work = kmalloc(sizeof(*work), GFP_ATOMIC);
552 __neigh_set_probe_once(neigh);
554 write_unlock(&neigh->lock);
555 } else if (time_after(jiffies, rt->last_probe +
556 idev->cnf.rtr_probe_interval)) {
557 work = kmalloc(sizeof(*work), GFP_ATOMIC);
561 rt->last_probe = jiffies;
562 INIT_WORK(&work->work, rt6_probe_deferred);
563 work->target = *nh_gw;
566 schedule_work(&work->work);
570 rcu_read_unlock_bh();
573 static inline void rt6_probe(struct fib6_info *rt)
579 * Default Router Selection (RFC 2461 6.3.6)
581 static inline int rt6_check_dev(struct fib6_info *rt, int oif)
583 const struct net_device *dev = rt->fib6_nh.nh_dev;
585 if (!oif || dev->ifindex == oif)
590 static inline enum rt6_nud_state rt6_check_neigh(struct fib6_info *rt)
592 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
593 struct neighbour *neigh;
595 if (rt->fib6_flags & RTF_NONEXTHOP ||
596 !(rt->fib6_flags & RTF_GATEWAY))
597 return RT6_NUD_SUCCEED;
600 neigh = __ipv6_neigh_lookup_noref(rt->fib6_nh.nh_dev,
603 read_lock(&neigh->lock);
604 if (neigh->nud_state & NUD_VALID)
605 ret = RT6_NUD_SUCCEED;
606 #ifdef CONFIG_IPV6_ROUTER_PREF
607 else if (!(neigh->nud_state & NUD_FAILED))
608 ret = RT6_NUD_SUCCEED;
610 ret = RT6_NUD_FAIL_PROBE;
612 read_unlock(&neigh->lock);
614 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
615 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
617 rcu_read_unlock_bh();
622 static int rt6_score_route(struct fib6_info *rt, int oif, int strict)
626 m = rt6_check_dev(rt, oif);
627 if (!m && (strict & RT6_LOOKUP_F_IFACE))
628 return RT6_NUD_FAIL_HARD;
629 #ifdef CONFIG_IPV6_ROUTER_PREF
630 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->fib6_flags)) << 2;
632 if (strict & RT6_LOOKUP_F_REACHABLE) {
633 int n = rt6_check_neigh(rt);
640 /* called with rc_read_lock held */
641 static inline bool fib6_ignore_linkdown(const struct fib6_info *f6i)
643 const struct net_device *dev = fib6_info_nh_dev(f6i);
647 const struct inet6_dev *idev = __in6_dev_get(dev);
649 rc = !!idev->cnf.ignore_routes_with_linkdown;
655 static struct fib6_info *find_match(struct fib6_info *rt, int oif, int strict,
656 int *mpri, struct fib6_info *match,
660 bool match_do_rr = false;
662 if (rt->fib6_nh.nh_flags & RTNH_F_DEAD)
665 if (fib6_ignore_linkdown(rt) &&
666 rt->fib6_nh.nh_flags & RTNH_F_LINKDOWN &&
667 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
670 if (fib6_check_expired(rt))
673 m = rt6_score_route(rt, oif, strict);
674 if (m == RT6_NUD_FAIL_DO_RR) {
676 m = 0; /* lowest valid score */
677 } else if (m == RT6_NUD_FAIL_HARD) {
681 if (strict & RT6_LOOKUP_F_REACHABLE)
684 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
686 *do_rr = match_do_rr;
694 static struct fib6_info *find_rr_leaf(struct fib6_node *fn,
695 struct fib6_info *leaf,
696 struct fib6_info *rr_head,
697 u32 metric, int oif, int strict,
700 struct fib6_info *rt, *match, *cont;
705 for (rt = rr_head; rt; rt = rcu_dereference(rt->fib6_next)) {
706 if (rt->fib6_metric != metric) {
711 match = find_match(rt, oif, strict, &mpri, match, do_rr);
714 for (rt = leaf; rt && rt != rr_head;
715 rt = rcu_dereference(rt->fib6_next)) {
716 if (rt->fib6_metric != metric) {
721 match = find_match(rt, oif, strict, &mpri, match, do_rr);
727 for (rt = cont; rt; rt = rcu_dereference(rt->fib6_next))
728 match = find_match(rt, oif, strict, &mpri, match, do_rr);
733 static struct fib6_info *rt6_select(struct net *net, struct fib6_node *fn,
736 struct fib6_info *leaf = rcu_dereference(fn->leaf);
737 struct fib6_info *match, *rt0;
741 if (!leaf || leaf == net->ipv6.fib6_null_entry)
742 return net->ipv6.fib6_null_entry;
744 rt0 = rcu_dereference(fn->rr_ptr);
748 /* Double check to make sure fn is not an intermediate node
749 * and fn->leaf does not points to its child's leaf
750 * (This might happen if all routes under fn are deleted from
751 * the tree and fib6_repair_tree() is called on the node.)
753 key_plen = rt0->fib6_dst.plen;
754 #ifdef CONFIG_IPV6_SUBTREES
755 if (rt0->fib6_src.plen)
756 key_plen = rt0->fib6_src.plen;
758 if (fn->fn_bit != key_plen)
759 return net->ipv6.fib6_null_entry;
761 match = find_rr_leaf(fn, leaf, rt0, rt0->fib6_metric, oif, strict,
765 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
767 /* no entries matched; do round-robin */
768 if (!next || next->fib6_metric != rt0->fib6_metric)
772 spin_lock_bh(&leaf->fib6_table->tb6_lock);
773 /* make sure next is not being deleted from the tree */
775 rcu_assign_pointer(fn->rr_ptr, next);
776 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
780 return match ? match : net->ipv6.fib6_null_entry;
783 static bool rt6_is_gw_or_nonexthop(const struct fib6_info *rt)
785 return (rt->fib6_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
788 #ifdef CONFIG_IPV6_ROUTE_INFO
789 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
790 const struct in6_addr *gwaddr)
792 struct net *net = dev_net(dev);
793 struct route_info *rinfo = (struct route_info *) opt;
794 struct in6_addr prefix_buf, *prefix;
796 unsigned long lifetime;
797 struct fib6_info *rt;
799 if (len < sizeof(struct route_info)) {
803 /* Sanity check for prefix_len and length */
804 if (rinfo->length > 3) {
806 } else if (rinfo->prefix_len > 128) {
808 } else if (rinfo->prefix_len > 64) {
809 if (rinfo->length < 2) {
812 } else if (rinfo->prefix_len > 0) {
813 if (rinfo->length < 1) {
818 pref = rinfo->route_pref;
819 if (pref == ICMPV6_ROUTER_PREF_INVALID)
822 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
824 if (rinfo->length == 3)
825 prefix = (struct in6_addr *)rinfo->prefix;
827 /* this function is safe */
828 ipv6_addr_prefix(&prefix_buf,
829 (struct in6_addr *)rinfo->prefix,
831 prefix = &prefix_buf;
834 if (rinfo->prefix_len == 0)
835 rt = rt6_get_dflt_router(net, gwaddr, dev);
837 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
840 if (rt && !lifetime) {
846 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
849 rt->fib6_flags = RTF_ROUTEINFO |
850 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
853 if (!addrconf_finite_timeout(lifetime))
854 fib6_clean_expires(rt);
856 fib6_set_expires(rt, jiffies + HZ * lifetime);
858 fib6_info_release(rt);
865 * Misc support functions
868 /* called with rcu_lock held */
869 static struct net_device *ip6_rt_get_dev_rcu(struct fib6_info *rt)
871 struct net_device *dev = rt->fib6_nh.nh_dev;
873 if (rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
874 /* for copies of local routes, dst->dev needs to be the
875 * device if it is a master device, the master device if
876 * device is enslaved, and the loopback as the default
878 if (netif_is_l3_slave(dev) &&
879 !rt6_need_strict(&rt->fib6_dst.addr))
880 dev = l3mdev_master_dev_rcu(dev);
881 else if (!netif_is_l3_master(dev))
882 dev = dev_net(dev)->loopback_dev;
883 /* last case is netif_is_l3_master(dev) is true in which
884 * case we want dev returned to be dev
891 static const int fib6_prop[RTN_MAX + 1] = {
898 [RTN_BLACKHOLE] = -EINVAL,
899 [RTN_UNREACHABLE] = -EHOSTUNREACH,
900 [RTN_PROHIBIT] = -EACCES,
901 [RTN_THROW] = -EAGAIN,
903 [RTN_XRESOLVE] = -EINVAL,
906 static int ip6_rt_type_to_error(u8 fib6_type)
908 return fib6_prop[fib6_type];
911 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
913 unsigned short flags = 0;
916 flags |= DST_NOCOUNT;
917 if (rt->dst_nopolicy)
918 flags |= DST_NOPOLICY;
925 static void ip6_rt_init_dst_reject(struct rt6_info *rt, struct fib6_info *ort)
927 rt->dst.error = ip6_rt_type_to_error(ort->fib6_type);
929 switch (ort->fib6_type) {
931 rt->dst.output = dst_discard_out;
932 rt->dst.input = dst_discard;
935 rt->dst.output = ip6_pkt_prohibit_out;
936 rt->dst.input = ip6_pkt_prohibit;
939 case RTN_UNREACHABLE:
941 rt->dst.output = ip6_pkt_discard_out;
942 rt->dst.input = ip6_pkt_discard;
947 static void ip6_rt_init_dst(struct rt6_info *rt, struct fib6_info *ort)
949 if (ort->fib6_flags & RTF_REJECT) {
950 ip6_rt_init_dst_reject(rt, ort);
955 rt->dst.output = ip6_output;
957 if (ort->fib6_type == RTN_LOCAL || ort->fib6_type == RTN_ANYCAST) {
958 rt->dst.input = ip6_input;
959 } else if (ipv6_addr_type(&ort->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
960 rt->dst.input = ip6_mc_input;
962 rt->dst.input = ip6_forward;
965 if (ort->fib6_nh.nh_lwtstate) {
966 rt->dst.lwtstate = lwtstate_get(ort->fib6_nh.nh_lwtstate);
967 lwtunnel_set_redirect(&rt->dst);
970 rt->dst.lastuse = jiffies;
973 /* Caller must already hold reference to @from */
974 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
976 rt->rt6i_flags &= ~RTF_EXPIRES;
977 rcu_assign_pointer(rt->from, from);
978 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
981 /* Caller must already hold reference to @ort */
982 static void ip6_rt_copy_init(struct rt6_info *rt, struct fib6_info *ort)
984 struct net_device *dev = fib6_info_nh_dev(ort);
986 ip6_rt_init_dst(rt, ort);
988 rt->rt6i_dst = ort->fib6_dst;
989 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
990 rt->rt6i_gateway = ort->fib6_nh.nh_gw;
991 rt->rt6i_flags = ort->fib6_flags;
992 rt6_set_from(rt, ort);
993 #ifdef CONFIG_IPV6_SUBTREES
994 rt->rt6i_src = ort->fib6_src;
998 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
999 struct in6_addr *saddr)
1001 struct fib6_node *pn, *sn;
1003 if (fn->fn_flags & RTN_TL_ROOT)
1005 pn = rcu_dereference(fn->parent);
1006 sn = FIB6_SUBTREE(pn);
1008 fn = fib6_node_lookup(sn, NULL, saddr);
1011 if (fn->fn_flags & RTN_RTINFO)
1016 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt,
1019 struct rt6_info *rt = *prt;
1021 if (dst_hold_safe(&rt->dst))
1023 if (null_fallback) {
1024 rt = net->ipv6.ip6_null_entry;
1033 /* called with rcu_lock held */
1034 static struct rt6_info *ip6_create_rt_rcu(struct fib6_info *rt)
1036 unsigned short flags = fib6_info_dst_flags(rt);
1037 struct net_device *dev = rt->fib6_nh.nh_dev;
1038 struct rt6_info *nrt;
1040 if (!fib6_info_hold_safe(rt))
1043 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1045 ip6_rt_copy_init(nrt, rt);
1047 fib6_info_release(rt);
1052 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
1053 struct fib6_table *table,
1055 const struct sk_buff *skb,
1058 struct fib6_info *f6i;
1059 struct fib6_node *fn;
1060 struct rt6_info *rt;
1062 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1063 flags &= ~RT6_LOOKUP_F_IFACE;
1066 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1068 f6i = rcu_dereference(fn->leaf);
1070 f6i = net->ipv6.fib6_null_entry;
1072 f6i = rt6_device_match(net, f6i, &fl6->saddr,
1073 fl6->flowi6_oif, flags);
1074 if (f6i->fib6_nsiblings && fl6->flowi6_oif == 0)
1075 f6i = fib6_multipath_select(net, f6i, fl6,
1076 fl6->flowi6_oif, skb,
1079 if (f6i == net->ipv6.fib6_null_entry) {
1080 fn = fib6_backtrack(fn, &fl6->saddr);
1085 trace_fib6_table_lookup(net, f6i, table, fl6);
1087 /* Search through exception table */
1088 rt = rt6_find_cached_rt(f6i, &fl6->daddr, &fl6->saddr);
1090 if (ip6_hold_safe(net, &rt, true))
1091 dst_use_noref(&rt->dst, jiffies);
1092 } else if (f6i == net->ipv6.fib6_null_entry) {
1093 rt = net->ipv6.ip6_null_entry;
1096 rt = ip6_create_rt_rcu(f6i);
1098 rt = net->ipv6.ip6_null_entry;
1108 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1109 const struct sk_buff *skb, int flags)
1111 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1113 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1115 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1116 const struct in6_addr *saddr, int oif,
1117 const struct sk_buff *skb, int strict)
1119 struct flowi6 fl6 = {
1123 struct dst_entry *dst;
1124 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1127 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1128 flags |= RT6_LOOKUP_F_HAS_SADDR;
1131 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1132 if (dst->error == 0)
1133 return (struct rt6_info *) dst;
1139 EXPORT_SYMBOL(rt6_lookup);
1141 /* ip6_ins_rt is called with FREE table->tb6_lock.
1142 * It takes new route entry, the addition fails by any reason the
1143 * route is released.
1144 * Caller must hold dst before calling it.
1147 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1148 struct netlink_ext_ack *extack)
1151 struct fib6_table *table;
1153 table = rt->fib6_table;
1154 spin_lock_bh(&table->tb6_lock);
1155 err = fib6_add(&table->tb6_root, rt, info, extack);
1156 spin_unlock_bh(&table->tb6_lock);
1161 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1163 struct nl_info info = { .nl_net = net, };
1165 return __ip6_ins_rt(rt, &info, NULL);
1168 static struct rt6_info *ip6_rt_cache_alloc(struct fib6_info *ort,
1169 const struct in6_addr *daddr,
1170 const struct in6_addr *saddr)
1172 struct net_device *dev;
1173 struct rt6_info *rt;
1179 if (!fib6_info_hold_safe(ort))
1182 dev = ip6_rt_get_dev_rcu(ort);
1183 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1185 fib6_info_release(ort);
1189 ip6_rt_copy_init(rt, ort);
1190 rt->rt6i_flags |= RTF_CACHE;
1191 rt->dst.flags |= DST_HOST;
1192 rt->rt6i_dst.addr = *daddr;
1193 rt->rt6i_dst.plen = 128;
1195 if (!rt6_is_gw_or_nonexthop(ort)) {
1196 if (ort->fib6_dst.plen != 128 &&
1197 ipv6_addr_equal(&ort->fib6_dst.addr, daddr))
1198 rt->rt6i_flags |= RTF_ANYCAST;
1199 #ifdef CONFIG_IPV6_SUBTREES
1200 if (rt->rt6i_src.plen && saddr) {
1201 rt->rt6i_src.addr = *saddr;
1202 rt->rt6i_src.plen = 128;
1210 static struct rt6_info *ip6_rt_pcpu_alloc(struct fib6_info *rt)
1212 unsigned short flags = fib6_info_dst_flags(rt);
1213 struct net_device *dev;
1214 struct rt6_info *pcpu_rt;
1216 if (!fib6_info_hold_safe(rt))
1220 dev = ip6_rt_get_dev_rcu(rt);
1221 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags);
1224 fib6_info_release(rt);
1227 ip6_rt_copy_init(pcpu_rt, rt);
1228 pcpu_rt->rt6i_flags |= RTF_PCPU;
1232 /* It should be called with rcu_read_lock() acquired */
1233 static struct rt6_info *rt6_get_pcpu_route(struct fib6_info *rt)
1235 struct rt6_info *pcpu_rt, **p;
1237 p = this_cpu_ptr(rt->rt6i_pcpu);
1241 ip6_hold_safe(NULL, &pcpu_rt, false);
1246 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1247 struct fib6_info *rt)
1249 struct rt6_info *pcpu_rt, *prev, **p;
1251 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1253 dst_hold(&net->ipv6.ip6_null_entry->dst);
1254 return net->ipv6.ip6_null_entry;
1257 dst_hold(&pcpu_rt->dst);
1258 p = this_cpu_ptr(rt->rt6i_pcpu);
1259 prev = cmpxchg(p, NULL, pcpu_rt);
1265 /* exception hash table implementation
1267 static DEFINE_SPINLOCK(rt6_exception_lock);
1269 /* Remove rt6_ex from hash table and free the memory
1270 * Caller must hold rt6_exception_lock
1272 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1273 struct rt6_exception *rt6_ex)
1277 if (!bucket || !rt6_ex)
1280 net = dev_net(rt6_ex->rt6i->dst.dev);
1281 hlist_del_rcu(&rt6_ex->hlist);
1282 dst_release(&rt6_ex->rt6i->dst);
1283 kfree_rcu(rt6_ex, rcu);
1284 WARN_ON_ONCE(!bucket->depth);
1286 net->ipv6.rt6_stats->fib_rt_cache--;
1289 /* Remove oldest rt6_ex in bucket and free the memory
1290 * Caller must hold rt6_exception_lock
1292 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1294 struct rt6_exception *rt6_ex, *oldest = NULL;
1299 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1300 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1303 rt6_remove_exception(bucket, oldest);
1306 static u32 rt6_exception_hash(const struct in6_addr *dst,
1307 const struct in6_addr *src)
1309 static u32 seed __read_mostly;
1312 net_get_random_once(&seed, sizeof(seed));
1313 val = jhash(dst, sizeof(*dst), seed);
1315 #ifdef CONFIG_IPV6_SUBTREES
1317 val = jhash(src, sizeof(*src), val);
1319 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1322 /* Helper function to find the cached rt in the hash table
1323 * and update bucket pointer to point to the bucket for this
1324 * (daddr, saddr) pair
1325 * Caller must hold rt6_exception_lock
1327 static struct rt6_exception *
1328 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1329 const struct in6_addr *daddr,
1330 const struct in6_addr *saddr)
1332 struct rt6_exception *rt6_ex;
1335 if (!(*bucket) || !daddr)
1338 hval = rt6_exception_hash(daddr, saddr);
1341 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1342 struct rt6_info *rt6 = rt6_ex->rt6i;
1343 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1345 #ifdef CONFIG_IPV6_SUBTREES
1346 if (matched && saddr)
1347 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1355 /* Helper function to find the cached rt in the hash table
1356 * and update bucket pointer to point to the bucket for this
1357 * (daddr, saddr) pair
1358 * Caller must hold rcu_read_lock()
1360 static struct rt6_exception *
1361 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1362 const struct in6_addr *daddr,
1363 const struct in6_addr *saddr)
1365 struct rt6_exception *rt6_ex;
1368 WARN_ON_ONCE(!rcu_read_lock_held());
1370 if (!(*bucket) || !daddr)
1373 hval = rt6_exception_hash(daddr, saddr);
1376 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1377 struct rt6_info *rt6 = rt6_ex->rt6i;
1378 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1380 #ifdef CONFIG_IPV6_SUBTREES
1381 if (matched && saddr)
1382 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1390 static unsigned int fib6_mtu(const struct fib6_info *rt)
1394 if (rt->fib6_pmtu) {
1395 mtu = rt->fib6_pmtu;
1397 struct net_device *dev = fib6_info_nh_dev(rt);
1398 struct inet6_dev *idev;
1401 idev = __in6_dev_get(dev);
1402 mtu = idev->cnf.mtu6;
1406 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1408 return mtu - lwtunnel_headroom(rt->fib6_nh.nh_lwtstate, mtu);
1411 static int rt6_insert_exception(struct rt6_info *nrt,
1412 struct fib6_info *ort)
1414 struct net *net = dev_net(nrt->dst.dev);
1415 struct rt6_exception_bucket *bucket;
1416 struct in6_addr *src_key = NULL;
1417 struct rt6_exception *rt6_ex;
1420 spin_lock_bh(&rt6_exception_lock);
1422 if (ort->exception_bucket_flushed) {
1427 bucket = rcu_dereference_protected(ort->rt6i_exception_bucket,
1428 lockdep_is_held(&rt6_exception_lock));
1430 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1436 rcu_assign_pointer(ort->rt6i_exception_bucket, bucket);
1439 #ifdef CONFIG_IPV6_SUBTREES
1440 /* rt6i_src.plen != 0 indicates ort is in subtree
1441 * and exception table is indexed by a hash of
1442 * both rt6i_dst and rt6i_src.
1443 * Otherwise, the exception table is indexed by
1444 * a hash of only rt6i_dst.
1446 if (ort->fib6_src.plen)
1447 src_key = &nrt->rt6i_src.addr;
1449 /* rt6_mtu_change() might lower mtu on ort.
1450 * Only insert this exception route if its mtu
1451 * is less than ort's mtu value.
1453 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(ort)) {
1458 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1461 rt6_remove_exception(bucket, rt6_ex);
1463 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1469 rt6_ex->stamp = jiffies;
1470 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1472 net->ipv6.rt6_stats->fib_rt_cache++;
1474 if (bucket->depth > FIB6_MAX_DEPTH)
1475 rt6_exception_remove_oldest(bucket);
1478 spin_unlock_bh(&rt6_exception_lock);
1480 /* Update fn->fn_sernum to invalidate all cached dst */
1482 spin_lock_bh(&ort->fib6_table->tb6_lock);
1483 fib6_update_sernum(net, ort);
1484 spin_unlock_bh(&ort->fib6_table->tb6_lock);
1485 fib6_force_start_gc(net);
1491 void rt6_flush_exceptions(struct fib6_info *rt)
1493 struct rt6_exception_bucket *bucket;
1494 struct rt6_exception *rt6_ex;
1495 struct hlist_node *tmp;
1498 spin_lock_bh(&rt6_exception_lock);
1499 /* Prevent rt6_insert_exception() to recreate the bucket list */
1500 rt->exception_bucket_flushed = 1;
1502 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1503 lockdep_is_held(&rt6_exception_lock));
1507 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1508 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist)
1509 rt6_remove_exception(bucket, rt6_ex);
1510 WARN_ON_ONCE(bucket->depth);
1515 spin_unlock_bh(&rt6_exception_lock);
1518 /* Find cached rt in the hash table inside passed in rt
1519 * Caller has to hold rcu_read_lock()
1521 static struct rt6_info *rt6_find_cached_rt(struct fib6_info *rt,
1522 struct in6_addr *daddr,
1523 struct in6_addr *saddr)
1525 struct rt6_exception_bucket *bucket;
1526 struct in6_addr *src_key = NULL;
1527 struct rt6_exception *rt6_ex;
1528 struct rt6_info *res = NULL;
1530 bucket = rcu_dereference(rt->rt6i_exception_bucket);
1532 #ifdef CONFIG_IPV6_SUBTREES
1533 /* rt6i_src.plen != 0 indicates rt is in subtree
1534 * and exception table is indexed by a hash of
1535 * both rt6i_dst and rt6i_src.
1536 * Otherwise, the exception table is indexed by
1537 * a hash of only rt6i_dst.
1539 if (rt->fib6_src.plen)
1542 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1544 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1550 /* Remove the passed in cached rt from the hash table that contains it */
1551 static int rt6_remove_exception_rt(struct rt6_info *rt)
1553 struct rt6_exception_bucket *bucket;
1554 struct in6_addr *src_key = NULL;
1555 struct rt6_exception *rt6_ex;
1556 struct fib6_info *from;
1559 from = rcu_dereference(rt->from);
1561 !(rt->rt6i_flags & RTF_CACHE))
1564 if (!rcu_access_pointer(from->rt6i_exception_bucket))
1567 spin_lock_bh(&rt6_exception_lock);
1568 bucket = rcu_dereference_protected(from->rt6i_exception_bucket,
1569 lockdep_is_held(&rt6_exception_lock));
1570 #ifdef CONFIG_IPV6_SUBTREES
1571 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1572 * and exception table is indexed by a hash of
1573 * both rt6i_dst and rt6i_src.
1574 * Otherwise, the exception table is indexed by
1575 * a hash of only rt6i_dst.
1577 if (from->fib6_src.plen)
1578 src_key = &rt->rt6i_src.addr;
1580 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1584 rt6_remove_exception(bucket, rt6_ex);
1590 spin_unlock_bh(&rt6_exception_lock);
1594 /* Find rt6_ex which contains the passed in rt cache and
1597 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1599 struct rt6_exception_bucket *bucket;
1600 struct fib6_info *from = rt->from;
1601 struct in6_addr *src_key = NULL;
1602 struct rt6_exception *rt6_ex;
1605 !(rt->rt6i_flags & RTF_CACHE))
1609 bucket = rcu_dereference(from->rt6i_exception_bucket);
1611 #ifdef CONFIG_IPV6_SUBTREES
1612 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1613 * and exception table is indexed by a hash of
1614 * both rt6i_dst and rt6i_src.
1615 * Otherwise, the exception table is indexed by
1616 * a hash of only rt6i_dst.
1618 if (from->fib6_src.plen)
1619 src_key = &rt->rt6i_src.addr;
1621 rt6_ex = __rt6_find_exception_rcu(&bucket,
1625 rt6_ex->stamp = jiffies;
1630 static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1631 struct rt6_info *rt, int mtu)
1633 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1634 * lowest MTU in the path: always allow updating the route PMTU to
1635 * reflect PMTU decreases.
1637 * If the new MTU is higher, and the route PMTU is equal to the local
1638 * MTU, this means the old MTU is the lowest in the path, so allow
1639 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1643 if (dst_mtu(&rt->dst) >= mtu)
1646 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
1652 static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
1653 struct fib6_info *rt, int mtu)
1655 struct rt6_exception_bucket *bucket;
1656 struct rt6_exception *rt6_ex;
1659 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1660 lockdep_is_held(&rt6_exception_lock));
1665 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1666 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1667 struct rt6_info *entry = rt6_ex->rt6i;
1669 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
1670 * route), the metrics of its rt->from have already
1673 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
1674 rt6_mtu_change_route_allowed(idev, entry, mtu))
1675 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
1681 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
1683 static void rt6_exceptions_clean_tohost(struct fib6_info *rt,
1684 struct in6_addr *gateway)
1686 struct rt6_exception_bucket *bucket;
1687 struct rt6_exception *rt6_ex;
1688 struct hlist_node *tmp;
1691 if (!rcu_access_pointer(rt->rt6i_exception_bucket))
1694 spin_lock_bh(&rt6_exception_lock);
1695 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1696 lockdep_is_held(&rt6_exception_lock));
1699 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1700 hlist_for_each_entry_safe(rt6_ex, tmp,
1701 &bucket->chain, hlist) {
1702 struct rt6_info *entry = rt6_ex->rt6i;
1704 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
1705 RTF_CACHE_GATEWAY &&
1706 ipv6_addr_equal(gateway,
1707 &entry->rt6i_gateway)) {
1708 rt6_remove_exception(bucket, rt6_ex);
1715 spin_unlock_bh(&rt6_exception_lock);
1718 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
1719 struct rt6_exception *rt6_ex,
1720 struct fib6_gc_args *gc_args,
1723 struct rt6_info *rt = rt6_ex->rt6i;
1725 /* we are pruning and obsoleting aged-out and non gateway exceptions
1726 * even if others have still references to them, so that on next
1727 * dst_check() such references can be dropped.
1728 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
1729 * expired, independently from their aging, as per RFC 8201 section 4
1731 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
1732 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
1733 RT6_TRACE("aging clone %p\n", rt);
1734 rt6_remove_exception(bucket, rt6_ex);
1737 } else if (time_after(jiffies, rt->dst.expires)) {
1738 RT6_TRACE("purging expired route %p\n", rt);
1739 rt6_remove_exception(bucket, rt6_ex);
1743 if (rt->rt6i_flags & RTF_GATEWAY) {
1744 struct neighbour *neigh;
1745 __u8 neigh_flags = 0;
1747 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
1749 neigh_flags = neigh->flags;
1751 if (!(neigh_flags & NTF_ROUTER)) {
1752 RT6_TRACE("purging route %p via non-router but gateway\n",
1754 rt6_remove_exception(bucket, rt6_ex);
1762 void rt6_age_exceptions(struct fib6_info *rt,
1763 struct fib6_gc_args *gc_args,
1766 struct rt6_exception_bucket *bucket;
1767 struct rt6_exception *rt6_ex;
1768 struct hlist_node *tmp;
1771 if (!rcu_access_pointer(rt->rt6i_exception_bucket))
1775 spin_lock(&rt6_exception_lock);
1776 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1777 lockdep_is_held(&rt6_exception_lock));
1780 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1781 hlist_for_each_entry_safe(rt6_ex, tmp,
1782 &bucket->chain, hlist) {
1783 rt6_age_examine_exception(bucket, rt6_ex,
1789 spin_unlock(&rt6_exception_lock);
1790 rcu_read_unlock_bh();
1793 /* must be called with rcu lock held */
1794 struct fib6_info *fib6_table_lookup(struct net *net, struct fib6_table *table,
1795 int oif, struct flowi6 *fl6, int strict)
1797 struct fib6_node *fn, *saved_fn;
1798 struct fib6_info *f6i;
1800 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1803 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1807 f6i = rt6_select(net, fn, oif, strict);
1808 if (f6i == net->ipv6.fib6_null_entry) {
1809 fn = fib6_backtrack(fn, &fl6->saddr);
1811 goto redo_rt6_select;
1812 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1813 /* also consider unreachable route */
1814 strict &= ~RT6_LOOKUP_F_REACHABLE;
1816 goto redo_rt6_select;
1820 trace_fib6_table_lookup(net, f6i, table, fl6);
1825 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1826 int oif, struct flowi6 *fl6,
1827 const struct sk_buff *skb, int flags)
1829 struct fib6_info *f6i;
1830 struct rt6_info *rt;
1833 strict |= flags & RT6_LOOKUP_F_IFACE;
1834 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
1835 if (net->ipv6.devconf_all->forwarding == 0)
1836 strict |= RT6_LOOKUP_F_REACHABLE;
1840 f6i = fib6_table_lookup(net, table, oif, fl6, strict);
1841 if (f6i->fib6_nsiblings)
1842 f6i = fib6_multipath_select(net, f6i, fl6, oif, skb, strict);
1844 if (f6i == net->ipv6.fib6_null_entry) {
1845 rt = net->ipv6.ip6_null_entry;
1851 /*Search through exception table */
1852 rt = rt6_find_cached_rt(f6i, &fl6->daddr, &fl6->saddr);
1854 if (ip6_hold_safe(net, &rt, true))
1855 dst_use_noref(&rt->dst, jiffies);
1859 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1860 !(f6i->fib6_flags & RTF_GATEWAY))) {
1861 /* Create a RTF_CACHE clone which will not be
1862 * owned by the fib6 tree. It is for the special case where
1863 * the daddr in the skb during the neighbor look-up is different
1864 * from the fl6->daddr used to look-up route here.
1866 struct rt6_info *uncached_rt;
1868 uncached_rt = ip6_rt_cache_alloc(f6i, &fl6->daddr, NULL);
1873 /* Uncached_rt's refcnt is taken during ip6_rt_cache_alloc()
1874 * No need for another dst_hold()
1876 rt6_uncached_list_add(uncached_rt);
1877 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
1879 uncached_rt = net->ipv6.ip6_null_entry;
1880 dst_hold(&uncached_rt->dst);
1885 /* Get a percpu copy */
1887 struct rt6_info *pcpu_rt;
1890 pcpu_rt = rt6_get_pcpu_route(f6i);
1893 pcpu_rt = rt6_make_pcpu_route(net, f6i);
1901 EXPORT_SYMBOL_GPL(ip6_pol_route);
1903 static struct rt6_info *ip6_pol_route_input(struct net *net,
1904 struct fib6_table *table,
1906 const struct sk_buff *skb,
1909 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
1912 struct dst_entry *ip6_route_input_lookup(struct net *net,
1913 struct net_device *dev,
1915 const struct sk_buff *skb,
1918 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1919 flags |= RT6_LOOKUP_F_IFACE;
1921 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
1923 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
1925 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
1926 struct flow_keys *keys,
1927 struct flow_keys *flkeys)
1929 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
1930 const struct ipv6hdr *key_iph = outer_iph;
1931 struct flow_keys *_flkeys = flkeys;
1932 const struct ipv6hdr *inner_iph;
1933 const struct icmp6hdr *icmph;
1934 struct ipv6hdr _inner_iph;
1935 struct icmp6hdr _icmph;
1937 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
1940 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
1941 sizeof(_icmph), &_icmph);
1945 if (icmph->icmp6_type != ICMPV6_DEST_UNREACH &&
1946 icmph->icmp6_type != ICMPV6_PKT_TOOBIG &&
1947 icmph->icmp6_type != ICMPV6_TIME_EXCEED &&
1948 icmph->icmp6_type != ICMPV6_PARAMPROB)
1951 inner_iph = skb_header_pointer(skb,
1952 skb_transport_offset(skb) + sizeof(*icmph),
1953 sizeof(_inner_iph), &_inner_iph);
1957 key_iph = inner_iph;
1961 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
1962 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
1963 keys->tags.flow_label = _flkeys->tags.flow_label;
1964 keys->basic.ip_proto = _flkeys->basic.ip_proto;
1966 keys->addrs.v6addrs.src = key_iph->saddr;
1967 keys->addrs.v6addrs.dst = key_iph->daddr;
1968 keys->tags.flow_label = ip6_flowlabel(key_iph);
1969 keys->basic.ip_proto = key_iph->nexthdr;
1973 /* if skb is set it will be used and fl6 can be NULL */
1974 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
1975 const struct sk_buff *skb, struct flow_keys *flkeys)
1977 struct flow_keys hash_keys;
1980 switch (ip6_multipath_hash_policy(net)) {
1982 memset(&hash_keys, 0, sizeof(hash_keys));
1983 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1985 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
1987 hash_keys.addrs.v6addrs.src = fl6->saddr;
1988 hash_keys.addrs.v6addrs.dst = fl6->daddr;
1989 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1990 hash_keys.basic.ip_proto = fl6->flowi6_proto;
1995 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
1996 struct flow_keys keys;
1998 /* short-circuit if we already have L4 hash present */
2000 return skb_get_hash_raw(skb) >> 1;
2002 memset(&hash_keys, 0, sizeof(hash_keys));
2005 skb_flow_dissect_flow_keys(skb, &keys, flag);
2008 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2009 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2010 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2011 hash_keys.ports.src = flkeys->ports.src;
2012 hash_keys.ports.dst = flkeys->ports.dst;
2013 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2015 memset(&hash_keys, 0, sizeof(hash_keys));
2016 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2017 hash_keys.addrs.v6addrs.src = fl6->saddr;
2018 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2019 hash_keys.ports.src = fl6->fl6_sport;
2020 hash_keys.ports.dst = fl6->fl6_dport;
2021 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2025 mhash = flow_hash_from_keys(&hash_keys);
2030 void ip6_route_input(struct sk_buff *skb)
2032 const struct ipv6hdr *iph = ipv6_hdr(skb);
2033 struct net *net = dev_net(skb->dev);
2034 int flags = RT6_LOOKUP_F_HAS_SADDR;
2035 struct ip_tunnel_info *tun_info;
2036 struct flowi6 fl6 = {
2037 .flowi6_iif = skb->dev->ifindex,
2038 .daddr = iph->daddr,
2039 .saddr = iph->saddr,
2040 .flowlabel = ip6_flowinfo(iph),
2041 .flowi6_mark = skb->mark,
2042 .flowi6_proto = iph->nexthdr,
2044 struct flow_keys *flkeys = NULL, _flkeys;
2046 tun_info = skb_tunnel_info(skb);
2047 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2048 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2050 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2053 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2054 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2057 ip6_route_input_lookup(net, skb->dev, &fl6, skb, flags));
2060 static struct rt6_info *ip6_pol_route_output(struct net *net,
2061 struct fib6_table *table,
2063 const struct sk_buff *skb,
2066 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2069 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
2070 struct flowi6 *fl6, int flags)
2074 if (ipv6_addr_type(&fl6->daddr) &
2075 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2076 struct dst_entry *dst;
2078 dst = l3mdev_link_scope_lookup(net, fl6);
2083 fl6->flowi6_iif = LOOPBACK_IFINDEX;
2085 any_src = ipv6_addr_any(&fl6->saddr);
2086 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2087 (fl6->flowi6_oif && any_src))
2088 flags |= RT6_LOOKUP_F_IFACE;
2091 flags |= RT6_LOOKUP_F_HAS_SADDR;
2093 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
2095 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2097 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2099 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2101 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2102 struct net_device *loopback_dev = net->loopback_dev;
2103 struct dst_entry *new = NULL;
2105 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
2106 DST_OBSOLETE_DEAD, 0);
2109 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2113 new->input = dst_discard;
2114 new->output = dst_discard_out;
2116 dst_copy_metrics(new, &ort->dst);
2118 rt->rt6i_idev = in6_dev_get(loopback_dev);
2119 rt->rt6i_gateway = ort->rt6i_gateway;
2120 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2122 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2123 #ifdef CONFIG_IPV6_SUBTREES
2124 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2128 dst_release(dst_orig);
2129 return new ? new : ERR_PTR(-ENOMEM);
2133 * Destination cache support functions
2136 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2140 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2143 if (fib6_check_expired(f6i))
2149 static struct dst_entry *rt6_check(struct rt6_info *rt,
2150 struct fib6_info *from,
2155 if ((from && !fib6_get_cookie_safe(from, &rt_cookie)) ||
2156 rt_cookie != cookie)
2159 if (rt6_check_expired(rt))
2165 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2166 struct fib6_info *from,
2169 if (!__rt6_check_expired(rt) &&
2170 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2171 fib6_check(from, cookie))
2177 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
2179 struct dst_entry *dst_ret;
2180 struct fib6_info *from;
2181 struct rt6_info *rt;
2183 rt = container_of(dst, struct rt6_info, dst);
2187 /* All IPV6 dsts are created with ->obsolete set to the value
2188 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2189 * into this function always.
2192 from = rcu_dereference(rt->from);
2194 if (from && (rt->rt6i_flags & RTF_PCPU ||
2195 unlikely(!list_empty(&rt->rt6i_uncached))))
2196 dst_ret = rt6_dst_from_check(rt, from, cookie);
2198 dst_ret = rt6_check(rt, from, cookie);
2205 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2207 struct rt6_info *rt = (struct rt6_info *) dst;
2210 if (rt->rt6i_flags & RTF_CACHE) {
2212 if (rt6_check_expired(rt)) {
2213 rt6_remove_exception_rt(rt);
2225 static void ip6_link_failure(struct sk_buff *skb)
2227 struct rt6_info *rt;
2229 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2231 rt = (struct rt6_info *) skb_dst(skb);
2234 if (rt->rt6i_flags & RTF_CACHE) {
2235 rt6_remove_exception_rt(rt);
2237 struct fib6_info *from;
2238 struct fib6_node *fn;
2240 from = rcu_dereference(rt->from);
2242 fn = rcu_dereference(from->fib6_node);
2243 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2251 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2253 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2254 struct fib6_info *from;
2257 from = rcu_dereference(rt0->from);
2259 rt0->dst.expires = from->expires;
2263 dst_set_expires(&rt0->dst, timeout);
2264 rt0->rt6i_flags |= RTF_EXPIRES;
2267 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2269 struct net *net = dev_net(rt->dst.dev);
2271 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2272 rt->rt6i_flags |= RTF_MODIFIED;
2273 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2276 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2281 from_set = !!rcu_dereference(rt->from);
2284 return !(rt->rt6i_flags & RTF_CACHE) &&
2285 (rt->rt6i_flags & RTF_PCPU || from_set);
2288 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2289 const struct ipv6hdr *iph, u32 mtu)
2291 const struct in6_addr *daddr, *saddr;
2292 struct rt6_info *rt6 = (struct rt6_info *)dst;
2294 if (dst_metric_locked(dst, RTAX_MTU))
2298 daddr = &iph->daddr;
2299 saddr = &iph->saddr;
2301 daddr = &sk->sk_v6_daddr;
2302 saddr = &inet6_sk(sk)->saddr;
2307 dst_confirm_neigh(dst, daddr);
2308 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2309 if (mtu >= dst_mtu(dst))
2312 if (!rt6_cache_allowed_for_pmtu(rt6)) {
2313 rt6_do_update_pmtu(rt6, mtu);
2314 /* update rt6_ex->stamp for cache */
2315 if (rt6->rt6i_flags & RTF_CACHE)
2316 rt6_update_exception_stamp_rt(rt6);
2318 struct fib6_info *from;
2319 struct rt6_info *nrt6;
2322 from = rcu_dereference(rt6->from);
2323 nrt6 = ip6_rt_cache_alloc(from, daddr, saddr);
2325 rt6_do_update_pmtu(nrt6, mtu);
2326 if (rt6_insert_exception(nrt6, from))
2327 dst_release_immediate(&nrt6->dst);
2333 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2334 struct sk_buff *skb, u32 mtu)
2336 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
2339 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2340 int oif, u32 mark, kuid_t uid)
2342 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2343 struct dst_entry *dst;
2344 struct flowi6 fl6 = {
2346 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2347 .daddr = iph->daddr,
2348 .saddr = iph->saddr,
2349 .flowlabel = ip6_flowinfo(iph),
2353 dst = ip6_route_output(net, NULL, &fl6);
2355 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
2358 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2360 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2362 int oif = sk->sk_bound_dev_if;
2363 struct dst_entry *dst;
2365 if (!oif && skb->dev)
2366 oif = l3mdev_master_ifindex(skb->dev);
2368 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
2370 dst = __sk_dst_get(sk);
2371 if (!dst || !dst->obsolete ||
2372 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2376 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2377 ip6_datagram_dst_update(sk, false);
2380 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2382 void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2383 const struct flowi6 *fl6)
2385 #ifdef CONFIG_IPV6_SUBTREES
2386 struct ipv6_pinfo *np = inet6_sk(sk);
2389 ip6_dst_store(sk, dst,
2390 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2391 &sk->sk_v6_daddr : NULL,
2392 #ifdef CONFIG_IPV6_SUBTREES
2393 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2399 /* Handle redirects */
2400 struct ip6rd_flowi {
2402 struct in6_addr gateway;
2405 static struct rt6_info *__ip6_route_redirect(struct net *net,
2406 struct fib6_table *table,
2408 const struct sk_buff *skb,
2411 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
2412 struct rt6_info *ret = NULL, *rt_cache;
2413 struct fib6_info *rt;
2414 struct fib6_node *fn;
2416 /* Get the "current" route for this destination and
2417 * check if the redirect has come from appropriate router.
2419 * RFC 4861 specifies that redirects should only be
2420 * accepted if they come from the nexthop to the target.
2421 * Due to the way the routes are chosen, this notion
2422 * is a bit fuzzy and one might need to check all possible
2427 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2429 for_each_fib6_node_rt_rcu(fn) {
2430 if (rt->fib6_nh.nh_flags & RTNH_F_DEAD)
2432 if (fib6_check_expired(rt))
2434 if (rt->fib6_flags & RTF_REJECT)
2436 if (!(rt->fib6_flags & RTF_GATEWAY))
2438 if (fl6->flowi6_oif != rt->fib6_nh.nh_dev->ifindex)
2440 /* rt_cache's gateway might be different from its 'parent'
2441 * in the case of an ip redirect.
2442 * So we keep searching in the exception table if the gateway
2445 if (!ipv6_addr_equal(&rdfl->gateway, &rt->fib6_nh.nh_gw)) {
2446 rt_cache = rt6_find_cached_rt(rt,
2450 ipv6_addr_equal(&rdfl->gateway,
2451 &rt_cache->rt6i_gateway)) {
2461 rt = net->ipv6.fib6_null_entry;
2462 else if (rt->fib6_flags & RTF_REJECT) {
2463 ret = net->ipv6.ip6_null_entry;
2467 if (rt == net->ipv6.fib6_null_entry) {
2468 fn = fib6_backtrack(fn, &fl6->saddr);
2475 ip6_hold_safe(net, &ret, true);
2477 ret = ip6_create_rt_rcu(rt);
2481 trace_fib6_table_lookup(net, rt, table, fl6);
2485 static struct dst_entry *ip6_route_redirect(struct net *net,
2486 const struct flowi6 *fl6,
2487 const struct sk_buff *skb,
2488 const struct in6_addr *gateway)
2490 int flags = RT6_LOOKUP_F_HAS_SADDR;
2491 struct ip6rd_flowi rdfl;
2494 rdfl.gateway = *gateway;
2496 return fib6_rule_lookup(net, &rdfl.fl6, skb,
2497 flags, __ip6_route_redirect);
2500 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
2503 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2504 struct dst_entry *dst;
2505 struct flowi6 fl6 = {
2506 .flowi6_iif = LOOPBACK_IFINDEX,
2508 .flowi6_mark = mark,
2509 .daddr = iph->daddr,
2510 .saddr = iph->saddr,
2511 .flowlabel = ip6_flowinfo(iph),
2515 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
2516 rt6_do_redirect(dst, NULL, skb);
2519 EXPORT_SYMBOL_GPL(ip6_redirect);
2521 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
2523 const struct ipv6hdr *iph = ipv6_hdr(skb);
2524 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
2525 struct dst_entry *dst;
2526 struct flowi6 fl6 = {
2527 .flowi6_iif = LOOPBACK_IFINDEX,
2530 .saddr = iph->daddr,
2531 .flowi6_uid = sock_net_uid(net, NULL),
2534 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
2535 rt6_do_redirect(dst, NULL, skb);
2539 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
2541 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
2544 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
2546 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
2548 struct net_device *dev = dst->dev;
2549 unsigned int mtu = dst_mtu(dst);
2550 struct net *net = dev_net(dev);
2552 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
2554 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
2555 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
2558 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
2559 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
2560 * IPV6_MAXPLEN is also valid and means: "any MSS,
2561 * rely only on pmtu discovery"
2563 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
2568 static unsigned int ip6_mtu(const struct dst_entry *dst)
2570 struct inet6_dev *idev;
2573 mtu = dst_metric_raw(dst, RTAX_MTU);
2580 idev = __in6_dev_get(dst->dev);
2582 mtu = idev->cnf.mtu6;
2586 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
2588 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
2592 * 1. mtu on route is locked - use it
2593 * 2. mtu from nexthop exception
2594 * 3. mtu from egress device
2596 * based on ip6_dst_mtu_forward and exception logic of
2597 * rt6_find_cached_rt; called with rcu_read_lock
2599 u32 ip6_mtu_from_fib6(struct fib6_info *f6i, struct in6_addr *daddr,
2600 struct in6_addr *saddr)
2602 struct rt6_exception_bucket *bucket;
2603 struct rt6_exception *rt6_ex;
2604 struct in6_addr *src_key;
2605 struct inet6_dev *idev;
2608 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
2609 mtu = f6i->fib6_pmtu;
2615 #ifdef CONFIG_IPV6_SUBTREES
2616 if (f6i->fib6_src.plen)
2620 bucket = rcu_dereference(f6i->rt6i_exception_bucket);
2621 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
2622 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
2623 mtu = dst_metric_raw(&rt6_ex->rt6i->dst, RTAX_MTU);
2626 struct net_device *dev = fib6_info_nh_dev(f6i);
2629 idev = __in6_dev_get(dev);
2630 if (idev && idev->cnf.mtu6 > mtu)
2631 mtu = idev->cnf.mtu6;
2634 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
2636 return mtu - lwtunnel_headroom(fib6_info_nh_lwt(f6i), mtu);
2639 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
2642 struct dst_entry *dst;
2643 struct rt6_info *rt;
2644 struct inet6_dev *idev = in6_dev_get(dev);
2645 struct net *net = dev_net(dev);
2647 if (unlikely(!idev))
2648 return ERR_PTR(-ENODEV);
2650 rt = ip6_dst_alloc(net, dev, 0);
2651 if (unlikely(!rt)) {
2653 dst = ERR_PTR(-ENOMEM);
2657 rt->dst.flags |= DST_HOST;
2658 rt->dst.input = ip6_input;
2659 rt->dst.output = ip6_output;
2660 rt->rt6i_gateway = fl6->daddr;
2661 rt->rt6i_dst.addr = fl6->daddr;
2662 rt->rt6i_dst.plen = 128;
2663 rt->rt6i_idev = idev;
2664 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
2666 /* Add this dst into uncached_list so that rt6_disable_ip() can
2667 * do proper release of the net_device
2669 rt6_uncached_list_add(rt);
2670 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
2672 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
2678 static int ip6_dst_gc(struct dst_ops *ops)
2680 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
2681 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
2682 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
2683 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
2684 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
2685 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
2688 entries = dst_entries_get_fast(ops);
2689 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
2690 entries <= rt_max_size)
2693 net->ipv6.ip6_rt_gc_expire++;
2694 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
2695 entries = dst_entries_get_slow(ops);
2696 if (entries < ops->gc_thresh)
2697 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
2699 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
2700 return entries > rt_max_size;
2703 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
2704 struct fib6_config *cfg,
2705 const struct in6_addr *gw_addr,
2706 u32 tbid, int flags)
2708 struct flowi6 fl6 = {
2709 .flowi6_oif = cfg->fc_ifindex,
2711 .saddr = cfg->fc_prefsrc,
2713 struct fib6_table *table;
2714 struct rt6_info *rt;
2716 table = fib6_get_table(net, tbid);
2720 if (!ipv6_addr_any(&cfg->fc_prefsrc))
2721 flags |= RT6_LOOKUP_F_HAS_SADDR;
2723 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
2724 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, NULL, flags);
2726 /* if table lookup failed, fall back to full lookup */
2727 if (rt == net->ipv6.ip6_null_entry) {
2735 static int ip6_route_check_nh_onlink(struct net *net,
2736 struct fib6_config *cfg,
2737 const struct net_device *dev,
2738 struct netlink_ext_ack *extack)
2740 u32 tbid = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN;
2741 const struct in6_addr *gw_addr = &cfg->fc_gateway;
2742 u32 flags = RTF_LOCAL | RTF_ANYCAST | RTF_REJECT;
2743 struct rt6_info *grt;
2747 grt = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0);
2749 if (!grt->dst.error &&
2750 /* ignore match if it is the default route */
2751 grt->from && !ipv6_addr_any(&grt->from->fib6_dst.addr) &&
2752 (grt->rt6i_flags & flags || dev != grt->dst.dev)) {
2753 NL_SET_ERR_MSG(extack,
2754 "Nexthop has invalid gateway or device mismatch");
2764 static int ip6_route_check_nh(struct net *net,
2765 struct fib6_config *cfg,
2766 struct net_device **_dev,
2767 struct inet6_dev **idev)
2769 const struct in6_addr *gw_addr = &cfg->fc_gateway;
2770 struct net_device *dev = _dev ? *_dev : NULL;
2771 struct rt6_info *grt = NULL;
2772 int err = -EHOSTUNREACH;
2774 if (cfg->fc_table) {
2775 int flags = RT6_LOOKUP_F_IFACE;
2777 grt = ip6_nh_lookup_table(net, cfg, gw_addr,
2778 cfg->fc_table, flags);
2780 if (grt->rt6i_flags & RTF_GATEWAY ||
2781 (dev && dev != grt->dst.dev)) {
2789 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, NULL, 1);
2795 if (dev != grt->dst.dev) {
2800 *_dev = dev = grt->dst.dev;
2801 *idev = grt->rt6i_idev;
2803 in6_dev_hold(grt->rt6i_idev);
2806 if (!(grt->rt6i_flags & RTF_GATEWAY))
2815 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
2816 struct net_device **_dev, struct inet6_dev **idev,
2817 struct netlink_ext_ack *extack)
2819 const struct in6_addr *gw_addr = &cfg->fc_gateway;
2820 int gwa_type = ipv6_addr_type(gw_addr);
2821 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
2822 const struct net_device *dev = *_dev;
2823 bool need_addr_check = !dev;
2826 /* if gw_addr is local we will fail to detect this in case
2827 * address is still TENTATIVE (DAD in progress). rt6_lookup()
2828 * will return already-added prefix route via interface that
2829 * prefix route was assigned to, which might be non-loopback.
2832 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
2833 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
2837 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
2838 /* IPv6 strictly inhibits using not link-local
2839 * addresses as nexthop address.
2840 * Otherwise, router will not able to send redirects.
2841 * It is very good, but in some (rare!) circumstances
2842 * (SIT, PtP, NBMA NOARP links) it is handy to allow
2843 * some exceptions. --ANK
2844 * We allow IPv4-mapped nexthops to support RFC4798-type
2847 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
2848 NL_SET_ERR_MSG(extack, "Invalid gateway address");
2852 if (cfg->fc_flags & RTNH_F_ONLINK)
2853 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
2855 err = ip6_route_check_nh(net, cfg, _dev, idev);
2861 /* reload in case device was changed */
2866 NL_SET_ERR_MSG(extack, "Egress device not specified");
2868 } else if (dev->flags & IFF_LOOPBACK) {
2869 NL_SET_ERR_MSG(extack,
2870 "Egress device can not be loopback device for this route");
2874 /* if we did not check gw_addr above, do so now that the
2875 * egress device has been resolved.
2877 if (need_addr_check &&
2878 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
2879 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
2888 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
2890 struct netlink_ext_ack *extack)
2892 struct net *net = cfg->fc_nlinfo.nl_net;
2893 struct fib6_info *rt = NULL;
2894 struct net_device *dev = NULL;
2895 struct inet6_dev *idev = NULL;
2896 struct fib6_table *table;
2900 /* RTF_PCPU is an internal flag; can not be set by userspace */
2901 if (cfg->fc_flags & RTF_PCPU) {
2902 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
2906 /* RTF_CACHE is an internal flag; can not be set by userspace */
2907 if (cfg->fc_flags & RTF_CACHE) {
2908 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
2912 if (cfg->fc_type > RTN_MAX) {
2913 NL_SET_ERR_MSG(extack, "Invalid route type");
2917 if (cfg->fc_dst_len > 128) {
2918 NL_SET_ERR_MSG(extack, "Invalid prefix length");
2921 if (cfg->fc_src_len > 128) {
2922 NL_SET_ERR_MSG(extack, "Invalid source address length");
2925 #ifndef CONFIG_IPV6_SUBTREES
2926 if (cfg->fc_src_len) {
2927 NL_SET_ERR_MSG(extack,
2928 "Specifying source address requires IPV6_SUBTREES to be enabled");
2932 if (cfg->fc_ifindex) {
2934 dev = dev_get_by_index(net, cfg->fc_ifindex);
2937 idev = in6_dev_get(dev);
2942 if (cfg->fc_metric == 0)
2943 cfg->fc_metric = IP6_RT_PRIO_USER;
2945 if (cfg->fc_flags & RTNH_F_ONLINK) {
2947 NL_SET_ERR_MSG(extack,
2948 "Nexthop device required for onlink");
2953 if (!(dev->flags & IFF_UP)) {
2954 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
2961 if (cfg->fc_nlinfo.nlh &&
2962 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
2963 table = fib6_get_table(net, cfg->fc_table);
2965 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
2966 table = fib6_new_table(net, cfg->fc_table);
2969 table = fib6_new_table(net, cfg->fc_table);
2976 rt = fib6_info_alloc(gfp_flags);
2980 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
2982 if (IS_ERR(rt->fib6_metrics)) {
2983 err = PTR_ERR(rt->fib6_metrics);
2984 /* Do not leave garbage there. */
2985 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
2989 if (cfg->fc_flags & RTF_ADDRCONF)
2990 rt->dst_nocount = true;
2992 if (cfg->fc_flags & RTF_EXPIRES)
2993 fib6_set_expires(rt, jiffies +
2994 clock_t_to_jiffies(cfg->fc_expires));
2996 fib6_clean_expires(rt);
2998 if (cfg->fc_protocol == RTPROT_UNSPEC)
2999 cfg->fc_protocol = RTPROT_BOOT;
3000 rt->fib6_protocol = cfg->fc_protocol;
3002 addr_type = ipv6_addr_type(&cfg->fc_dst);
3004 if (cfg->fc_encap) {
3005 struct lwtunnel_state *lwtstate;
3007 err = lwtunnel_build_state(cfg->fc_encap_type,
3008 cfg->fc_encap, AF_INET6, cfg,
3012 rt->fib6_nh.nh_lwtstate = lwtstate_get(lwtstate);
3015 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3016 rt->fib6_dst.plen = cfg->fc_dst_len;
3017 if (rt->fib6_dst.plen == 128)
3018 rt->dst_host = true;
3020 #ifdef CONFIG_IPV6_SUBTREES
3021 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3022 rt->fib6_src.plen = cfg->fc_src_len;
3025 rt->fib6_metric = cfg->fc_metric;
3026 rt->fib6_nh.nh_weight = 1;
3028 rt->fib6_type = cfg->fc_type;
3030 /* We cannot add true routes via loopback here,
3031 they would result in kernel looping; promote them to reject routes
3033 if ((cfg->fc_flags & RTF_REJECT) ||
3034 (dev && (dev->flags & IFF_LOOPBACK) &&
3035 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3036 !(cfg->fc_flags & RTF_LOCAL))) {
3037 /* hold loopback dev/idev if we haven't done so. */
3038 if (dev != net->loopback_dev) {
3043 dev = net->loopback_dev;
3045 idev = in6_dev_get(dev);
3051 rt->fib6_flags = RTF_REJECT|RTF_NONEXTHOP;
3055 if (cfg->fc_flags & RTF_GATEWAY) {
3056 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3060 rt->fib6_nh.nh_gw = cfg->fc_gateway;
3067 if (idev->cnf.disable_ipv6) {
3068 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3073 if (!(dev->flags & IFF_UP)) {
3074 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3079 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3080 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3081 NL_SET_ERR_MSG(extack, "Invalid source address");
3085 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3086 rt->fib6_prefsrc.plen = 128;
3088 rt->fib6_prefsrc.plen = 0;
3090 rt->fib6_flags = cfg->fc_flags;
3093 if (!(rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3094 !netif_carrier_ok(dev))
3095 rt->fib6_nh.nh_flags |= RTNH_F_LINKDOWN;
3096 rt->fib6_nh.nh_flags |= (cfg->fc_flags & RTNH_F_ONLINK);
3097 rt->fib6_nh.nh_dev = dev;
3098 rt->fib6_table = table;
3110 fib6_info_release(rt);
3111 return ERR_PTR(err);
3114 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3115 struct netlink_ext_ack *extack)
3117 struct fib6_info *rt;
3120 rt = ip6_route_info_create(cfg, gfp_flags, extack);
3124 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3125 fib6_info_release(rt);
3130 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3132 struct net *net = info->nl_net;
3133 struct fib6_table *table;
3136 if (rt == net->ipv6.fib6_null_entry) {
3141 table = rt->fib6_table;
3142 spin_lock_bh(&table->tb6_lock);
3143 err = fib6_del(rt, info);
3144 spin_unlock_bh(&table->tb6_lock);
3147 fib6_info_release(rt);
3151 int ip6_del_rt(struct net *net, struct fib6_info *rt)
3153 struct nl_info info = { .nl_net = net };
3155 return __ip6_del_rt(rt, &info);
3158 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3160 struct nl_info *info = &cfg->fc_nlinfo;
3161 struct net *net = info->nl_net;
3162 struct sk_buff *skb = NULL;
3163 struct fib6_table *table;
3166 if (rt == net->ipv6.fib6_null_entry)
3168 table = rt->fib6_table;
3169 spin_lock_bh(&table->tb6_lock);
3171 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3172 struct fib6_info *sibling, *next_sibling;
3174 /* prefer to send a single notification with all hops */
3175 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3177 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3179 if (rt6_fill_node(net, skb, rt, NULL,
3180 NULL, NULL, 0, RTM_DELROUTE,
3181 info->portid, seq, 0) < 0) {
3185 info->skip_notify = 1;
3188 list_for_each_entry_safe(sibling, next_sibling,
3191 err = fib6_del(sibling, info);
3197 err = fib6_del(rt, info);
3199 spin_unlock_bh(&table->tb6_lock);
3201 fib6_info_release(rt);
3204 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3205 info->nlh, gfp_any());
3210 static int ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3214 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3217 if (cfg->fc_flags & RTF_GATEWAY &&
3218 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3221 rc = rt6_remove_exception_rt(rt);
3226 static int ip6_route_del(struct fib6_config *cfg,
3227 struct netlink_ext_ack *extack)
3229 struct rt6_info *rt_cache;
3230 struct fib6_table *table;
3231 struct fib6_info *rt;
3232 struct fib6_node *fn;
3235 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
3237 NL_SET_ERR_MSG(extack, "FIB table does not exist");
3243 fn = fib6_locate(&table->tb6_root,
3244 &cfg->fc_dst, cfg->fc_dst_len,
3245 &cfg->fc_src, cfg->fc_src_len,
3246 !(cfg->fc_flags & RTF_CACHE));
3249 for_each_fib6_node_rt_rcu(fn) {
3250 if (cfg->fc_flags & RTF_CACHE) {
3253 rt_cache = rt6_find_cached_rt(rt, &cfg->fc_dst,
3256 rc = ip6_del_cached_rt(rt_cache, cfg);
3264 if (cfg->fc_ifindex &&
3265 (!rt->fib6_nh.nh_dev ||
3266 rt->fib6_nh.nh_dev->ifindex != cfg->fc_ifindex))
3268 if (cfg->fc_flags & RTF_GATEWAY &&
3269 !ipv6_addr_equal(&cfg->fc_gateway, &rt->fib6_nh.nh_gw))
3271 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
3273 if (cfg->fc_protocol && cfg->fc_protocol != rt->fib6_protocol)
3275 if (!fib6_info_hold_safe(rt))
3279 /* if gateway was specified only delete the one hop */
3280 if (cfg->fc_flags & RTF_GATEWAY)
3281 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3283 return __ip6_del_rt_siblings(rt, cfg);
3291 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
3293 struct netevent_redirect netevent;
3294 struct rt6_info *rt, *nrt = NULL;
3295 struct ndisc_options ndopts;
3296 struct inet6_dev *in6_dev;
3297 struct neighbour *neigh;
3298 struct fib6_info *from;
3300 int optlen, on_link;
3303 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
3304 optlen -= sizeof(*msg);
3307 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
3311 msg = (struct rd_msg *)icmp6_hdr(skb);
3313 if (ipv6_addr_is_multicast(&msg->dest)) {
3314 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
3319 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
3321 } else if (ipv6_addr_type(&msg->target) !=
3322 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
3323 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
3327 in6_dev = __in6_dev_get(skb->dev);
3330 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
3334 * The IP source address of the Redirect MUST be the same as the current
3335 * first-hop router for the specified ICMP Destination Address.
3338 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
3339 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
3344 if (ndopts.nd_opts_tgt_lladdr) {
3345 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
3348 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
3353 rt = (struct rt6_info *) dst;
3354 if (rt->rt6i_flags & RTF_REJECT) {
3355 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
3359 /* Redirect received -> path was valid.
3360 * Look, redirects are sent only in response to data packets,
3361 * so that this nexthop apparently is reachable. --ANK
3363 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
3365 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
3370 * We have finally decided to accept it.
3373 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
3374 NEIGH_UPDATE_F_WEAK_OVERRIDE|
3375 NEIGH_UPDATE_F_OVERRIDE|
3376 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
3377 NEIGH_UPDATE_F_ISROUTER)),
3378 NDISC_REDIRECT, &ndopts);
3381 from = rcu_dereference(rt->from);
3382 /* This fib6_info_hold() is safe here because we hold reference to rt
3383 * and rt already holds reference to fib6_info.
3385 fib6_info_hold(from);
3388 nrt = ip6_rt_cache_alloc(from, &msg->dest, NULL);
3392 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
3394 nrt->rt6i_flags &= ~RTF_GATEWAY;
3396 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
3398 /* No need to remove rt from the exception table if rt is
3399 * a cached route because rt6_insert_exception() will
3402 if (rt6_insert_exception(nrt, from)) {
3403 dst_release_immediate(&nrt->dst);
3407 netevent.old = &rt->dst;
3408 netevent.new = &nrt->dst;
3409 netevent.daddr = &msg->dest;
3410 netevent.neigh = neigh;
3411 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
3414 fib6_info_release(from);
3415 neigh_release(neigh);
3418 #ifdef CONFIG_IPV6_ROUTE_INFO
3419 static struct fib6_info *rt6_get_route_info(struct net *net,
3420 const struct in6_addr *prefix, int prefixlen,
3421 const struct in6_addr *gwaddr,
3422 struct net_device *dev)
3424 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
3425 int ifindex = dev->ifindex;
3426 struct fib6_node *fn;
3427 struct fib6_info *rt = NULL;
3428 struct fib6_table *table;
3430 table = fib6_get_table(net, tb_id);
3435 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
3439 for_each_fib6_node_rt_rcu(fn) {
3440 if (rt->fib6_nh.nh_dev->ifindex != ifindex)
3442 if ((rt->fib6_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
3444 if (!ipv6_addr_equal(&rt->fib6_nh.nh_gw, gwaddr))
3446 if (!fib6_info_hold_safe(rt))
3455 static struct fib6_info *rt6_add_route_info(struct net *net,
3456 const struct in6_addr *prefix, int prefixlen,
3457 const struct in6_addr *gwaddr,
3458 struct net_device *dev,
3461 struct fib6_config cfg = {
3462 .fc_metric = IP6_RT_PRIO_USER,
3463 .fc_ifindex = dev->ifindex,
3464 .fc_dst_len = prefixlen,
3465 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
3466 RTF_UP | RTF_PREF(pref),
3467 .fc_protocol = RTPROT_RA,
3468 .fc_type = RTN_UNICAST,
3469 .fc_nlinfo.portid = 0,
3470 .fc_nlinfo.nlh = NULL,
3471 .fc_nlinfo.nl_net = net,
3474 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
3475 cfg.fc_dst = *prefix;
3476 cfg.fc_gateway = *gwaddr;
3478 /* We should treat it as a default route if prefix length is 0. */
3480 cfg.fc_flags |= RTF_DEFAULT;
3482 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
3484 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
3488 struct fib6_info *rt6_get_dflt_router(struct net *net,
3489 const struct in6_addr *addr,
3490 struct net_device *dev)
3492 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
3493 struct fib6_info *rt;
3494 struct fib6_table *table;
3496 table = fib6_get_table(net, tb_id);
3501 for_each_fib6_node_rt_rcu(&table->tb6_root) {
3502 if (dev == rt->fib6_nh.nh_dev &&
3503 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
3504 ipv6_addr_equal(&rt->fib6_nh.nh_gw, addr))
3507 if (rt && !fib6_info_hold_safe(rt))
3513 struct fib6_info *rt6_add_dflt_router(struct net *net,
3514 const struct in6_addr *gwaddr,
3515 struct net_device *dev,
3518 struct fib6_config cfg = {
3519 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
3520 .fc_metric = IP6_RT_PRIO_USER,
3521 .fc_ifindex = dev->ifindex,
3522 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
3523 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
3524 .fc_protocol = RTPROT_RA,
3525 .fc_type = RTN_UNICAST,
3526 .fc_nlinfo.portid = 0,
3527 .fc_nlinfo.nlh = NULL,
3528 .fc_nlinfo.nl_net = net,
3531 cfg.fc_gateway = *gwaddr;
3533 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
3534 struct fib6_table *table;
3536 table = fib6_get_table(dev_net(dev), cfg.fc_table);
3538 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
3541 return rt6_get_dflt_router(net, gwaddr, dev);
3544 static void __rt6_purge_dflt_routers(struct net *net,
3545 struct fib6_table *table)
3547 struct fib6_info *rt;
3551 for_each_fib6_node_rt_rcu(&table->tb6_root) {
3552 struct net_device *dev = fib6_info_nh_dev(rt);
3553 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
3555 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
3556 (!idev || idev->cnf.accept_ra != 2) &&
3557 fib6_info_hold_safe(rt)) {
3559 ip6_del_rt(net, rt);
3565 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
3568 void rt6_purge_dflt_routers(struct net *net)
3570 struct fib6_table *table;
3571 struct hlist_head *head;
3576 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
3577 head = &net->ipv6.fib_table_hash[h];
3578 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
3579 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
3580 __rt6_purge_dflt_routers(net, table);
3587 static void rtmsg_to_fib6_config(struct net *net,
3588 struct in6_rtmsg *rtmsg,
3589 struct fib6_config *cfg)
3591 *cfg = (struct fib6_config){
3592 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
3594 .fc_ifindex = rtmsg->rtmsg_ifindex,
3595 .fc_metric = rtmsg->rtmsg_metric,
3596 .fc_expires = rtmsg->rtmsg_info,
3597 .fc_dst_len = rtmsg->rtmsg_dst_len,
3598 .fc_src_len = rtmsg->rtmsg_src_len,
3599 .fc_flags = rtmsg->rtmsg_flags,
3600 .fc_type = rtmsg->rtmsg_type,
3602 .fc_nlinfo.nl_net = net,
3604 .fc_dst = rtmsg->rtmsg_dst,
3605 .fc_src = rtmsg->rtmsg_src,
3606 .fc_gateway = rtmsg->rtmsg_gateway,
3610 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
3612 struct fib6_config cfg;
3613 struct in6_rtmsg rtmsg;
3617 case SIOCADDRT: /* Add a route */
3618 case SIOCDELRT: /* Delete a route */
3619 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3621 err = copy_from_user(&rtmsg, arg,
3622 sizeof(struct in6_rtmsg));
3626 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
3631 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
3634 err = ip6_route_del(&cfg, NULL);
3648 * Drop the packet on the floor
3651 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
3654 struct dst_entry *dst = skb_dst(skb);
3655 switch (ipstats_mib_noroutes) {
3656 case IPSTATS_MIB_INNOROUTES:
3657 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
3658 if (type == IPV6_ADDR_ANY) {
3659 IP6_INC_STATS(dev_net(dst->dev),
3660 __in6_dev_get_safely(skb->dev),
3661 IPSTATS_MIB_INADDRERRORS);
3665 case IPSTATS_MIB_OUTNOROUTES:
3666 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
3667 ipstats_mib_noroutes);
3670 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
3675 static int ip6_pkt_discard(struct sk_buff *skb)
3677 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
3680 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
3682 skb->dev = skb_dst(skb)->dev;
3683 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
3686 static int ip6_pkt_prohibit(struct sk_buff *skb)
3688 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
3691 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
3693 skb->dev = skb_dst(skb)->dev;
3694 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
3698 * Allocate a dst for local (unicast / anycast) address.
3701 struct fib6_info *addrconf_f6i_alloc(struct net *net,
3702 struct inet6_dev *idev,
3703 const struct in6_addr *addr,
3704 bool anycast, gfp_t gfp_flags)
3707 struct net_device *dev = idev->dev;
3708 struct fib6_info *f6i;
3710 f6i = fib6_info_alloc(gfp_flags);
3712 return ERR_PTR(-ENOMEM);
3714 f6i->fib6_metrics = ip_fib_metrics_init(net, NULL, 0, NULL);
3715 f6i->dst_nocount = true;
3716 f6i->dst_host = true;
3717 f6i->fib6_protocol = RTPROT_KERNEL;
3718 f6i->fib6_flags = RTF_UP | RTF_NONEXTHOP;
3720 f6i->fib6_type = RTN_ANYCAST;
3721 f6i->fib6_flags |= RTF_ANYCAST;
3723 f6i->fib6_type = RTN_LOCAL;
3724 f6i->fib6_flags |= RTF_LOCAL;
3727 f6i->fib6_nh.nh_gw = *addr;
3729 f6i->fib6_nh.nh_dev = dev;
3730 f6i->fib6_dst.addr = *addr;
3731 f6i->fib6_dst.plen = 128;
3732 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
3733 f6i->fib6_table = fib6_get_table(net, tb_id);
3738 /* remove deleted ip from prefsrc entries */
3739 struct arg_dev_net_ip {
3740 struct net_device *dev;
3742 struct in6_addr *addr;
3745 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
3747 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
3748 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
3749 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
3751 if (((void *)rt->fib6_nh.nh_dev == dev || !dev) &&
3752 rt != net->ipv6.fib6_null_entry &&
3753 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
3754 spin_lock_bh(&rt6_exception_lock);
3755 /* remove prefsrc entry */
3756 rt->fib6_prefsrc.plen = 0;
3757 spin_unlock_bh(&rt6_exception_lock);
3762 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
3764 struct net *net = dev_net(ifp->idev->dev);
3765 struct arg_dev_net_ip adni = {
3766 .dev = ifp->idev->dev,
3770 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
3773 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
3775 /* Remove routers and update dst entries when gateway turn into host. */
3776 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
3778 struct in6_addr *gateway = (struct in6_addr *)arg;
3780 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
3781 ipv6_addr_equal(gateway, &rt->fib6_nh.nh_gw)) {
3785 /* Further clean up cached routes in exception table.
3786 * This is needed because cached route may have a different
3787 * gateway than its 'parent' in the case of an ip redirect.
3789 rt6_exceptions_clean_tohost(rt, gateway);
3794 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
3796 fib6_clean_all(net, fib6_clean_tohost, gateway);
3799 struct arg_netdev_event {
3800 const struct net_device *dev;
3802 unsigned int nh_flags;
3803 unsigned long event;
3807 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
3809 struct fib6_info *iter;
3810 struct fib6_node *fn;
3812 fn = rcu_dereference_protected(rt->fib6_node,
3813 lockdep_is_held(&rt->fib6_table->tb6_lock));
3814 iter = rcu_dereference_protected(fn->leaf,
3815 lockdep_is_held(&rt->fib6_table->tb6_lock));
3817 if (iter->fib6_metric == rt->fib6_metric &&
3818 rt6_qualify_for_ecmp(iter))
3820 iter = rcu_dereference_protected(iter->fib6_next,
3821 lockdep_is_held(&rt->fib6_table->tb6_lock));
3827 static bool rt6_is_dead(const struct fib6_info *rt)
3829 if (rt->fib6_nh.nh_flags & RTNH_F_DEAD ||
3830 (rt->fib6_nh.nh_flags & RTNH_F_LINKDOWN &&
3831 fib6_ignore_linkdown(rt)))
3837 static int rt6_multipath_total_weight(const struct fib6_info *rt)
3839 struct fib6_info *iter;
3842 if (!rt6_is_dead(rt))
3843 total += rt->fib6_nh.nh_weight;
3845 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
3846 if (!rt6_is_dead(iter))
3847 total += iter->fib6_nh.nh_weight;
3853 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
3855 int upper_bound = -1;
3857 if (!rt6_is_dead(rt)) {
3858 *weight += rt->fib6_nh.nh_weight;
3859 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
3862 atomic_set(&rt->fib6_nh.nh_upper_bound, upper_bound);
3865 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
3867 struct fib6_info *iter;
3870 rt6_upper_bound_set(rt, &weight, total);
3872 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
3873 rt6_upper_bound_set(iter, &weight, total);
3876 void rt6_multipath_rebalance(struct fib6_info *rt)
3878 struct fib6_info *first;
3881 /* In case the entire multipath route was marked for flushing,
3882 * then there is no need to rebalance upon the removal of every
3885 if (!rt->fib6_nsiblings || rt->should_flush)
3888 /* During lookup routes are evaluated in order, so we need to
3889 * make sure upper bounds are assigned from the first sibling
3892 first = rt6_multipath_first_sibling(rt);
3893 if (WARN_ON_ONCE(!first))
3896 total = rt6_multipath_total_weight(first);
3897 rt6_multipath_upper_bound_set(first, total);
3900 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
3902 const struct arg_netdev_event *arg = p_arg;
3903 struct net *net = dev_net(arg->dev);
3905 if (rt != net->ipv6.fib6_null_entry && rt->fib6_nh.nh_dev == arg->dev) {
3906 rt->fib6_nh.nh_flags &= ~arg->nh_flags;
3907 fib6_update_sernum_upto_root(net, rt);
3908 rt6_multipath_rebalance(rt);
3914 void rt6_sync_up(struct net_device *dev, unsigned int nh_flags)
3916 struct arg_netdev_event arg = {
3919 .nh_flags = nh_flags,
3923 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
3924 arg.nh_flags |= RTNH_F_LINKDOWN;
3926 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
3929 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
3930 const struct net_device *dev)
3932 struct fib6_info *iter;
3934 if (rt->fib6_nh.nh_dev == dev)
3936 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
3937 if (iter->fib6_nh.nh_dev == dev)
3943 static void rt6_multipath_flush(struct fib6_info *rt)
3945 struct fib6_info *iter;
3947 rt->should_flush = 1;
3948 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
3949 iter->should_flush = 1;
3952 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
3953 const struct net_device *down_dev)
3955 struct fib6_info *iter;
3956 unsigned int dead = 0;
3958 if (rt->fib6_nh.nh_dev == down_dev ||
3959 rt->fib6_nh.nh_flags & RTNH_F_DEAD)
3961 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
3962 if (iter->fib6_nh.nh_dev == down_dev ||
3963 iter->fib6_nh.nh_flags & RTNH_F_DEAD)
3969 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
3970 const struct net_device *dev,
3971 unsigned int nh_flags)
3973 struct fib6_info *iter;
3975 if (rt->fib6_nh.nh_dev == dev)
3976 rt->fib6_nh.nh_flags |= nh_flags;
3977 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
3978 if (iter->fib6_nh.nh_dev == dev)
3979 iter->fib6_nh.nh_flags |= nh_flags;
3982 /* called with write lock held for table with rt */
3983 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
3985 const struct arg_netdev_event *arg = p_arg;
3986 const struct net_device *dev = arg->dev;
3987 struct net *net = dev_net(dev);
3989 if (rt == net->ipv6.fib6_null_entry)
3992 switch (arg->event) {
3993 case NETDEV_UNREGISTER:
3994 return rt->fib6_nh.nh_dev == dev ? -1 : 0;
3996 if (rt->should_flush)
3998 if (!rt->fib6_nsiblings)
3999 return rt->fib6_nh.nh_dev == dev ? -1 : 0;
4000 if (rt6_multipath_uses_dev(rt, dev)) {
4003 count = rt6_multipath_dead_count(rt, dev);
4004 if (rt->fib6_nsiblings + 1 == count) {
4005 rt6_multipath_flush(rt);
4008 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4010 fib6_update_sernum(net, rt);
4011 rt6_multipath_rebalance(rt);
4015 if (rt->fib6_nh.nh_dev != dev ||
4016 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4018 rt->fib6_nh.nh_flags |= RTNH_F_LINKDOWN;
4019 rt6_multipath_rebalance(rt);
4026 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4028 struct arg_netdev_event arg = {
4034 struct net *net = dev_net(dev);
4036 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4037 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4039 fib6_clean_all(net, fib6_ifdown, &arg);
4042 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4044 rt6_sync_down_dev(dev, event);
4045 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4046 neigh_ifdown(&nd_tbl, dev);
4049 struct rt6_mtu_change_arg {
4050 struct net_device *dev;
4054 static int rt6_mtu_change_route(struct fib6_info *rt, void *p_arg)
4056 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4057 struct inet6_dev *idev;
4059 /* In IPv6 pmtu discovery is not optional,
4060 so that RTAX_MTU lock cannot disable it.
4061 We still use this lock to block changes
4062 caused by addrconf/ndisc.
4065 idev = __in6_dev_get(arg->dev);
4069 /* For administrative MTU increase, there is no way to discover
4070 IPv6 PMTU increase, so PMTU increase should be updated here.
4071 Since RFC 1981 doesn't include administrative MTU increase
4072 update PMTU increase is a MUST. (i.e. jumbo frame)
4074 if (rt->fib6_nh.nh_dev == arg->dev &&
4075 !fib6_metric_locked(rt, RTAX_MTU)) {
4076 u32 mtu = rt->fib6_pmtu;
4078 if (mtu >= arg->mtu ||
4079 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4080 fib6_metric_set(rt, RTAX_MTU, arg->mtu);
4082 spin_lock_bh(&rt6_exception_lock);
4083 rt6_exceptions_update_pmtu(idev, rt, arg->mtu);
4084 spin_unlock_bh(&rt6_exception_lock);
4089 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4091 struct rt6_mtu_change_arg arg = {
4096 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4099 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4100 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
4101 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
4102 [RTA_OIF] = { .type = NLA_U32 },
4103 [RTA_IIF] = { .type = NLA_U32 },
4104 [RTA_PRIORITY] = { .type = NLA_U32 },
4105 [RTA_METRICS] = { .type = NLA_NESTED },
4106 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
4107 [RTA_PREF] = { .type = NLA_U8 },
4108 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4109 [RTA_ENCAP] = { .type = NLA_NESTED },
4110 [RTA_EXPIRES] = { .type = NLA_U32 },
4111 [RTA_UID] = { .type = NLA_U32 },
4112 [RTA_MARK] = { .type = NLA_U32 },
4113 [RTA_TABLE] = { .type = NLA_U32 },
4114 [RTA_IP_PROTO] = { .type = NLA_U8 },
4115 [RTA_SPORT] = { .type = NLA_U16 },
4116 [RTA_DPORT] = { .type = NLA_U16 },
4119 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
4120 struct fib6_config *cfg,
4121 struct netlink_ext_ack *extack)
4124 struct nlattr *tb[RTA_MAX+1];
4128 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
4134 rtm = nlmsg_data(nlh);
4136 *cfg = (struct fib6_config){
4137 .fc_table = rtm->rtm_table,
4138 .fc_dst_len = rtm->rtm_dst_len,
4139 .fc_src_len = rtm->rtm_src_len,
4141 .fc_protocol = rtm->rtm_protocol,
4142 .fc_type = rtm->rtm_type,
4144 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4145 .fc_nlinfo.nlh = nlh,
4146 .fc_nlinfo.nl_net = sock_net(skb->sk),
4149 if (rtm->rtm_type == RTN_UNREACHABLE ||
4150 rtm->rtm_type == RTN_BLACKHOLE ||
4151 rtm->rtm_type == RTN_PROHIBIT ||
4152 rtm->rtm_type == RTN_THROW)
4153 cfg->fc_flags |= RTF_REJECT;
4155 if (rtm->rtm_type == RTN_LOCAL)
4156 cfg->fc_flags |= RTF_LOCAL;
4158 if (rtm->rtm_flags & RTM_F_CLONED)
4159 cfg->fc_flags |= RTF_CACHE;
4161 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4163 if (tb[RTA_GATEWAY]) {
4164 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
4165 cfg->fc_flags |= RTF_GATEWAY;
4169 int plen = (rtm->rtm_dst_len + 7) >> 3;
4171 if (nla_len(tb[RTA_DST]) < plen)
4174 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
4178 int plen = (rtm->rtm_src_len + 7) >> 3;
4180 if (nla_len(tb[RTA_SRC]) < plen)
4183 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
4186 if (tb[RTA_PREFSRC])
4187 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
4190 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4192 if (tb[RTA_PRIORITY])
4193 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4195 if (tb[RTA_METRICS]) {
4196 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4197 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
4201 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4203 if (tb[RTA_MULTIPATH]) {
4204 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4205 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
4207 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
4208 cfg->fc_mp_len, extack);
4214 pref = nla_get_u8(tb[RTA_PREF]);
4215 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4216 pref != ICMPV6_ROUTER_PREF_HIGH)
4217 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4218 cfg->fc_flags |= RTF_PREF(pref);
4222 cfg->fc_encap = tb[RTA_ENCAP];
4224 if (tb[RTA_ENCAP_TYPE]) {
4225 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4227 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
4232 if (tb[RTA_EXPIRES]) {
4233 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
4235 if (addrconf_finite_timeout(timeout)) {
4236 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
4237 cfg->fc_flags |= RTF_EXPIRES;
4247 struct fib6_info *fib6_info;
4248 struct fib6_config r_cfg;
4249 struct list_head next;
4252 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
4256 list_for_each_entry(nh, rt6_nh_list, next) {
4257 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6c nexthop %pI6c ifi %d\n",
4258 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
4259 nh->r_cfg.fc_ifindex);
4263 static int ip6_route_info_append(struct net *net,
4264 struct list_head *rt6_nh_list,
4265 struct fib6_info *rt,
4266 struct fib6_config *r_cfg)
4271 list_for_each_entry(nh, rt6_nh_list, next) {
4272 /* check if fib6_info already exists */
4273 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
4277 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
4281 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
4282 list_add_tail(&nh->next, rt6_nh_list);
4287 static void ip6_route_mpath_notify(struct fib6_info *rt,
4288 struct fib6_info *rt_last,
4289 struct nl_info *info,
4292 /* if this is an APPEND route, then rt points to the first route
4293 * inserted and rt_last points to last route inserted. Userspace
4294 * wants a consistent dump of the route which starts at the first
4295 * nexthop. Since sibling routes are always added at the end of
4296 * the list, find the first sibling of the last route appended
4298 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
4299 rt = list_first_entry(&rt_last->fib6_siblings,
4305 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
4308 static int ip6_route_multipath_add(struct fib6_config *cfg,
4309 struct netlink_ext_ack *extack)
4311 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
4312 struct nl_info *info = &cfg->fc_nlinfo;
4313 struct fib6_config r_cfg;
4314 struct rtnexthop *rtnh;
4315 struct fib6_info *rt;
4316 struct rt6_nh *err_nh;
4317 struct rt6_nh *nh, *nh_safe;
4323 int replace = (cfg->fc_nlinfo.nlh &&
4324 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
4325 LIST_HEAD(rt6_nh_list);
4327 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
4328 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
4329 nlflags |= NLM_F_APPEND;
4331 remaining = cfg->fc_mp_len;
4332 rtnh = (struct rtnexthop *)cfg->fc_mp;
4334 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
4335 * fib6_info structs per nexthop
4337 while (rtnh_ok(rtnh, remaining)) {
4338 memcpy(&r_cfg, cfg, sizeof(*cfg));
4339 if (rtnh->rtnh_ifindex)
4340 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
4342 attrlen = rtnh_attrlen(rtnh);
4344 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
4346 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
4348 r_cfg.fc_gateway = nla_get_in6_addr(nla);
4349 r_cfg.fc_flags |= RTF_GATEWAY;
4351 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
4352 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
4354 r_cfg.fc_encap_type = nla_get_u16(nla);
4357 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
4358 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
4364 if (!rt6_qualify_for_ecmp(rt)) {
4366 NL_SET_ERR_MSG(extack,
4367 "Device only routes can not be added for IPv6 using the multipath API.");
4368 fib6_info_release(rt);
4372 rt->fib6_nh.nh_weight = rtnh->rtnh_hops + 1;
4374 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
4377 fib6_info_release(rt);
4381 rtnh = rtnh_next(rtnh, &remaining);
4384 /* for add and replace send one notification with all nexthops.
4385 * Skip the notification in fib6_add_rt2node and send one with
4386 * the full route when done
4388 info->skip_notify = 1;
4391 list_for_each_entry(nh, &rt6_nh_list, next) {
4392 err = __ip6_ins_rt(nh->fib6_info, info, extack);
4393 fib6_info_release(nh->fib6_info);
4396 /* save reference to last route successfully inserted */
4397 rt_last = nh->fib6_info;
4399 /* save reference to first route for notification */
4401 rt_notif = nh->fib6_info;
4404 /* nh->fib6_info is used or freed at this point, reset to NULL*/
4405 nh->fib6_info = NULL;
4408 ip6_print_replace_route_err(&rt6_nh_list);
4413 /* Because each route is added like a single route we remove
4414 * these flags after the first nexthop: if there is a collision,
4415 * we have already failed to add the first nexthop:
4416 * fib6_add_rt2node() has rejected it; when replacing, old
4417 * nexthops have been replaced by first new, the rest should
4420 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
4425 /* success ... tell user about new route */
4426 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
4430 /* send notification for routes that were added so that
4431 * the delete notifications sent by ip6_route_del are
4435 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
4437 /* Delete routes that were already added */
4438 list_for_each_entry(nh, &rt6_nh_list, next) {
4441 ip6_route_del(&nh->r_cfg, extack);
4445 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
4447 fib6_info_release(nh->fib6_info);
4448 list_del(&nh->next);
4455 static int ip6_route_multipath_del(struct fib6_config *cfg,
4456 struct netlink_ext_ack *extack)
4458 struct fib6_config r_cfg;
4459 struct rtnexthop *rtnh;
4462 int err = 1, last_err = 0;
4464 remaining = cfg->fc_mp_len;
4465 rtnh = (struct rtnexthop *)cfg->fc_mp;
4467 /* Parse a Multipath Entry */
4468 while (rtnh_ok(rtnh, remaining)) {
4469 memcpy(&r_cfg, cfg, sizeof(*cfg));
4470 if (rtnh->rtnh_ifindex)
4471 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
4473 attrlen = rtnh_attrlen(rtnh);
4475 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
4477 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
4479 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
4480 r_cfg.fc_flags |= RTF_GATEWAY;
4483 err = ip6_route_del(&r_cfg, extack);
4487 rtnh = rtnh_next(rtnh, &remaining);
4493 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
4494 struct netlink_ext_ack *extack)
4496 struct fib6_config cfg;
4499 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
4504 return ip6_route_multipath_del(&cfg, extack);
4506 cfg.fc_delete_all_nh = 1;
4507 return ip6_route_del(&cfg, extack);
4511 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
4512 struct netlink_ext_ack *extack)
4514 struct fib6_config cfg;
4517 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
4522 return ip6_route_multipath_add(&cfg, extack);
4524 return ip6_route_add(&cfg, GFP_KERNEL, extack);
4527 static size_t rt6_nlmsg_size(struct fib6_info *rt)
4529 int nexthop_len = 0;
4531 if (rt->fib6_nsiblings) {
4532 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
4533 + NLA_ALIGN(sizeof(struct rtnexthop))
4534 + nla_total_size(16) /* RTA_GATEWAY */
4535 + lwtunnel_get_encap_size(rt->fib6_nh.nh_lwtstate);
4537 nexthop_len *= rt->fib6_nsiblings;
4540 return NLMSG_ALIGN(sizeof(struct rtmsg))
4541 + nla_total_size(16) /* RTA_SRC */
4542 + nla_total_size(16) /* RTA_DST */
4543 + nla_total_size(16) /* RTA_GATEWAY */
4544 + nla_total_size(16) /* RTA_PREFSRC */
4545 + nla_total_size(4) /* RTA_TABLE */
4546 + nla_total_size(4) /* RTA_IIF */
4547 + nla_total_size(4) /* RTA_OIF */
4548 + nla_total_size(4) /* RTA_PRIORITY */
4549 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
4550 + nla_total_size(sizeof(struct rta_cacheinfo))
4551 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
4552 + nla_total_size(1) /* RTA_PREF */
4553 + lwtunnel_get_encap_size(rt->fib6_nh.nh_lwtstate)
4557 static int rt6_nexthop_info(struct sk_buff *skb, struct fib6_info *rt,
4558 unsigned int *flags, bool skip_oif)
4560 if (rt->fib6_nh.nh_flags & RTNH_F_DEAD)
4561 *flags |= RTNH_F_DEAD;
4563 if (rt->fib6_nh.nh_flags & RTNH_F_LINKDOWN) {
4564 *flags |= RTNH_F_LINKDOWN;
4567 if (fib6_ignore_linkdown(rt))
4568 *flags |= RTNH_F_DEAD;
4572 if (rt->fib6_flags & RTF_GATEWAY) {
4573 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->fib6_nh.nh_gw) < 0)
4574 goto nla_put_failure;
4577 *flags |= (rt->fib6_nh.nh_flags & RTNH_F_ONLINK);
4578 if (rt->fib6_nh.nh_flags & RTNH_F_OFFLOAD)
4579 *flags |= RTNH_F_OFFLOAD;
4581 /* not needed for multipath encoding b/c it has a rtnexthop struct */
4582 if (!skip_oif && rt->fib6_nh.nh_dev &&
4583 nla_put_u32(skb, RTA_OIF, rt->fib6_nh.nh_dev->ifindex))
4584 goto nla_put_failure;
4586 if (rt->fib6_nh.nh_lwtstate &&
4587 lwtunnel_fill_encap(skb, rt->fib6_nh.nh_lwtstate) < 0)
4588 goto nla_put_failure;
4596 /* add multipath next hop */
4597 static int rt6_add_nexthop(struct sk_buff *skb, struct fib6_info *rt)
4599 const struct net_device *dev = rt->fib6_nh.nh_dev;
4600 struct rtnexthop *rtnh;
4601 unsigned int flags = 0;
4603 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
4605 goto nla_put_failure;
4607 rtnh->rtnh_hops = rt->fib6_nh.nh_weight - 1;
4608 rtnh->rtnh_ifindex = dev ? dev->ifindex : 0;
4610 if (rt6_nexthop_info(skb, rt, &flags, true) < 0)
4611 goto nla_put_failure;
4613 rtnh->rtnh_flags = flags;
4615 /* length of rtnetlink header + attributes */
4616 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
4624 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
4625 struct fib6_info *rt, struct dst_entry *dst,
4626 struct in6_addr *dest, struct in6_addr *src,
4627 int iif, int type, u32 portid, u32 seq,
4630 struct rt6_info *rt6 = (struct rt6_info *)dst;
4631 struct rt6key *rt6_dst, *rt6_src;
4632 u32 *pmetrics, table, rt6_flags;
4633 struct nlmsghdr *nlh;
4637 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
4642 rt6_dst = &rt6->rt6i_dst;
4643 rt6_src = &rt6->rt6i_src;
4644 rt6_flags = rt6->rt6i_flags;
4646 rt6_dst = &rt->fib6_dst;
4647 rt6_src = &rt->fib6_src;
4648 rt6_flags = rt->fib6_flags;
4651 rtm = nlmsg_data(nlh);
4652 rtm->rtm_family = AF_INET6;
4653 rtm->rtm_dst_len = rt6_dst->plen;
4654 rtm->rtm_src_len = rt6_src->plen;
4657 table = rt->fib6_table->tb6_id;
4659 table = RT6_TABLE_UNSPEC;
4660 rtm->rtm_table = table;
4661 if (nla_put_u32(skb, RTA_TABLE, table))
4662 goto nla_put_failure;
4664 rtm->rtm_type = rt->fib6_type;
4666 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
4667 rtm->rtm_protocol = rt->fib6_protocol;
4669 if (rt6_flags & RTF_CACHE)
4670 rtm->rtm_flags |= RTM_F_CLONED;
4673 if (nla_put_in6_addr(skb, RTA_DST, dest))
4674 goto nla_put_failure;
4675 rtm->rtm_dst_len = 128;
4676 } else if (rtm->rtm_dst_len)
4677 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
4678 goto nla_put_failure;
4679 #ifdef CONFIG_IPV6_SUBTREES
4681 if (nla_put_in6_addr(skb, RTA_SRC, src))
4682 goto nla_put_failure;
4683 rtm->rtm_src_len = 128;
4684 } else if (rtm->rtm_src_len &&
4685 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
4686 goto nla_put_failure;
4689 #ifdef CONFIG_IPV6_MROUTE
4690 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
4691 int err = ip6mr_get_route(net, skb, rtm, portid);
4696 goto nla_put_failure;
4699 if (nla_put_u32(skb, RTA_IIF, iif))
4700 goto nla_put_failure;
4702 struct in6_addr saddr_buf;
4703 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
4704 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
4705 goto nla_put_failure;
4708 if (rt->fib6_prefsrc.plen) {
4709 struct in6_addr saddr_buf;
4710 saddr_buf = rt->fib6_prefsrc.addr;
4711 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
4712 goto nla_put_failure;
4715 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
4716 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
4717 goto nla_put_failure;
4719 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
4720 goto nla_put_failure;
4722 /* For multipath routes, walk the siblings list and add
4723 * each as a nexthop within RTA_MULTIPATH.
4726 if (rt6_flags & RTF_GATEWAY &&
4727 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
4728 goto nla_put_failure;
4730 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
4731 goto nla_put_failure;
4732 } else if (rt->fib6_nsiblings) {
4733 struct fib6_info *sibling, *next_sibling;
4736 mp = nla_nest_start(skb, RTA_MULTIPATH);
4738 goto nla_put_failure;
4740 if (rt6_add_nexthop(skb, rt) < 0)
4741 goto nla_put_failure;
4743 list_for_each_entry_safe(sibling, next_sibling,
4744 &rt->fib6_siblings, fib6_siblings) {
4745 if (rt6_add_nexthop(skb, sibling) < 0)
4746 goto nla_put_failure;
4749 nla_nest_end(skb, mp);
4751 if (rt6_nexthop_info(skb, rt, &rtm->rtm_flags, false) < 0)
4752 goto nla_put_failure;
4755 if (rt6_flags & RTF_EXPIRES) {
4756 expires = dst ? dst->expires : rt->expires;
4760 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
4761 goto nla_put_failure;
4763 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
4764 goto nla_put_failure;
4767 nlmsg_end(skb, nlh);
4771 nlmsg_cancel(skb, nlh);
4775 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
4776 const struct net_device *dev)
4778 if (f6i->fib6_nh.nh_dev == dev)
4781 if (f6i->fib6_nsiblings) {
4782 struct fib6_info *sibling, *next_sibling;
4784 list_for_each_entry_safe(sibling, next_sibling,
4785 &f6i->fib6_siblings, fib6_siblings) {
4786 if (sibling->fib6_nh.nh_dev == dev)
4794 int rt6_dump_route(struct fib6_info *rt, void *p_arg)
4796 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
4797 struct fib_dump_filter *filter = &arg->filter;
4798 unsigned int flags = NLM_F_MULTI;
4799 struct net *net = arg->net;
4801 if (rt == net->ipv6.fib6_null_entry)
4804 if ((filter->flags & RTM_F_PREFIX) &&
4805 !(rt->fib6_flags & RTF_PREFIX_RT)) {
4806 /* success since this is not a prefix route */
4809 if (filter->filter_set) {
4810 if ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
4811 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
4812 (filter->protocol && rt->fib6_protocol != filter->protocol)) {
4815 flags |= NLM_F_DUMP_FILTERED;
4818 return rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL, 0,
4819 RTM_NEWROUTE, NETLINK_CB(arg->cb->skb).portid,
4820 arg->cb->nlh->nlmsg_seq, flags);
4823 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4824 struct netlink_ext_ack *extack)
4826 struct net *net = sock_net(in_skb->sk);
4827 struct nlattr *tb[RTA_MAX+1];
4828 int err, iif = 0, oif = 0;
4829 struct fib6_info *from;
4830 struct dst_entry *dst;
4831 struct rt6_info *rt;
4832 struct sk_buff *skb;
4834 struct flowi6 fl6 = {};
4837 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
4843 rtm = nlmsg_data(nlh);
4844 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
4845 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
4848 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
4851 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
4855 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
4858 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
4862 iif = nla_get_u32(tb[RTA_IIF]);
4865 oif = nla_get_u32(tb[RTA_OIF]);
4868 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
4871 fl6.flowi6_uid = make_kuid(current_user_ns(),
4872 nla_get_u32(tb[RTA_UID]));
4874 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
4877 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
4880 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
4882 if (tb[RTA_IP_PROTO]) {
4883 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
4884 &fl6.flowi6_proto, extack);
4890 struct net_device *dev;
4895 dev = dev_get_by_index_rcu(net, iif);
4902 fl6.flowi6_iif = iif;
4904 if (!ipv6_addr_any(&fl6.saddr))
4905 flags |= RT6_LOOKUP_F_HAS_SADDR;
4907 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
4911 fl6.flowi6_oif = oif;
4913 dst = ip6_route_output(net, NULL, &fl6);
4917 rt = container_of(dst, struct rt6_info, dst);
4918 if (rt->dst.error) {
4919 err = rt->dst.error;
4924 if (rt == net->ipv6.ip6_null_entry) {
4925 err = rt->dst.error;
4930 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
4937 skb_dst_set(skb, &rt->dst);
4940 from = rcu_dereference(rt->from);
4943 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL, iif,
4944 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
4947 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
4948 &fl6.saddr, iif, RTM_NEWROUTE,
4949 NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
4958 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4963 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
4964 unsigned int nlm_flags)
4966 struct sk_buff *skb;
4967 struct net *net = info->nl_net;
4972 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
4974 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
4978 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
4979 event, info->portid, seq, nlm_flags);
4981 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
4982 WARN_ON(err == -EMSGSIZE);
4986 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
4987 info->nlh, gfp_any());
4991 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
4994 static int ip6_route_dev_notify(struct notifier_block *this,
4995 unsigned long event, void *ptr)
4997 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4998 struct net *net = dev_net(dev);
5000 if (!(dev->flags & IFF_LOOPBACK))
5003 if (event == NETDEV_REGISTER) {
5004 net->ipv6.fib6_null_entry->fib6_nh.nh_dev = dev;
5005 net->ipv6.ip6_null_entry->dst.dev = dev;
5006 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
5007 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5008 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
5009 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
5010 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
5011 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
5013 } else if (event == NETDEV_UNREGISTER &&
5014 dev->reg_state != NETREG_UNREGISTERED) {
5015 /* NETDEV_UNREGISTER could be fired for multiple times by
5016 * netdev_wait_allrefs(). Make sure we only call this once.
5018 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
5019 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5020 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
5021 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
5032 #ifdef CONFIG_PROC_FS
5033 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
5035 struct net *net = (struct net *)seq->private;
5036 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
5037 net->ipv6.rt6_stats->fib_nodes,
5038 net->ipv6.rt6_stats->fib_route_nodes,
5039 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
5040 net->ipv6.rt6_stats->fib_rt_entries,
5041 net->ipv6.rt6_stats->fib_rt_cache,
5042 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
5043 net->ipv6.rt6_stats->fib_discarded_routes);
5047 #endif /* CONFIG_PROC_FS */
5049 #ifdef CONFIG_SYSCTL
5052 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
5053 void __user *buffer, size_t *lenp, loff_t *ppos)
5060 net = (struct net *)ctl->extra1;
5061 delay = net->ipv6.sysctl.flush_delay;
5062 proc_dointvec(ctl, write, buffer, lenp, ppos);
5063 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
5070 static struct ctl_table ipv6_route_table_template[] = {
5072 .procname = "flush",
5073 .data = &init_net.ipv6.sysctl.flush_delay,
5074 .maxlen = sizeof(int),
5076 .proc_handler = ipv6_sysctl_rtcache_flush
5079 .procname = "gc_thresh",
5080 .data = &ip6_dst_ops_template.gc_thresh,
5081 .maxlen = sizeof(int),
5083 .proc_handler = proc_dointvec,
5086 .procname = "max_size",
5087 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
5088 .maxlen = sizeof(int),
5090 .proc_handler = proc_dointvec,
5093 .procname = "gc_min_interval",
5094 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
5095 .maxlen = sizeof(int),
5097 .proc_handler = proc_dointvec_jiffies,
5100 .procname = "gc_timeout",
5101 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
5102 .maxlen = sizeof(int),
5104 .proc_handler = proc_dointvec_jiffies,
5107 .procname = "gc_interval",
5108 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
5109 .maxlen = sizeof(int),
5111 .proc_handler = proc_dointvec_jiffies,
5114 .procname = "gc_elasticity",
5115 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
5116 .maxlen = sizeof(int),
5118 .proc_handler = proc_dointvec,
5121 .procname = "mtu_expires",
5122 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
5123 .maxlen = sizeof(int),
5125 .proc_handler = proc_dointvec_jiffies,
5128 .procname = "min_adv_mss",
5129 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
5130 .maxlen = sizeof(int),
5132 .proc_handler = proc_dointvec,
5135 .procname = "gc_min_interval_ms",
5136 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
5137 .maxlen = sizeof(int),
5139 .proc_handler = proc_dointvec_ms_jiffies,
5142 .procname = "skip_notify_on_dev_down",
5143 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
5144 .maxlen = sizeof(int),
5146 .proc_handler = proc_dointvec,
5153 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
5155 struct ctl_table *table;
5157 table = kmemdup(ipv6_route_table_template,
5158 sizeof(ipv6_route_table_template),
5162 table[0].data = &net->ipv6.sysctl.flush_delay;
5163 table[0].extra1 = net;
5164 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
5165 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
5166 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
5167 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
5168 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
5169 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
5170 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
5171 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
5172 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
5173 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
5175 /* Don't export sysctls to unprivileged users */
5176 if (net->user_ns != &init_user_ns)
5177 table[0].procname = NULL;
5184 static int __net_init ip6_route_net_init(struct net *net)
5188 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
5189 sizeof(net->ipv6.ip6_dst_ops));
5191 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
5192 goto out_ip6_dst_ops;
5194 net->ipv6.fib6_null_entry = kmemdup(&fib6_null_entry_template,
5195 sizeof(*net->ipv6.fib6_null_entry),
5197 if (!net->ipv6.fib6_null_entry)
5198 goto out_ip6_dst_entries;
5200 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
5201 sizeof(*net->ipv6.ip6_null_entry),
5203 if (!net->ipv6.ip6_null_entry)
5204 goto out_fib6_null_entry;
5205 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
5206 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
5207 ip6_template_metrics, true);
5209 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5210 net->ipv6.fib6_has_custom_rules = false;
5211 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
5212 sizeof(*net->ipv6.ip6_prohibit_entry),
5214 if (!net->ipv6.ip6_prohibit_entry)
5215 goto out_ip6_null_entry;
5216 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
5217 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
5218 ip6_template_metrics, true);
5220 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
5221 sizeof(*net->ipv6.ip6_blk_hole_entry),
5223 if (!net->ipv6.ip6_blk_hole_entry)
5224 goto out_ip6_prohibit_entry;
5225 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
5226 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
5227 ip6_template_metrics, true);
5230 net->ipv6.sysctl.flush_delay = 0;
5231 net->ipv6.sysctl.ip6_rt_max_size = 4096;
5232 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
5233 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
5234 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
5235 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
5236 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
5237 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
5238 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
5240 net->ipv6.ip6_rt_gc_expire = 30*HZ;
5246 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5247 out_ip6_prohibit_entry:
5248 kfree(net->ipv6.ip6_prohibit_entry);
5250 kfree(net->ipv6.ip6_null_entry);
5252 out_fib6_null_entry:
5253 kfree(net->ipv6.fib6_null_entry);
5254 out_ip6_dst_entries:
5255 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
5260 static void __net_exit ip6_route_net_exit(struct net *net)
5262 kfree(net->ipv6.fib6_null_entry);
5263 kfree(net->ipv6.ip6_null_entry);
5264 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5265 kfree(net->ipv6.ip6_prohibit_entry);
5266 kfree(net->ipv6.ip6_blk_hole_entry);
5268 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
5271 static int __net_init ip6_route_net_init_late(struct net *net)
5273 #ifdef CONFIG_PROC_FS
5274 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
5275 sizeof(struct ipv6_route_iter));
5276 proc_create_net_single("rt6_stats", 0444, net->proc_net,
5277 rt6_stats_seq_show, NULL);
5282 static void __net_exit ip6_route_net_exit_late(struct net *net)
5284 #ifdef CONFIG_PROC_FS
5285 remove_proc_entry("ipv6_route", net->proc_net);
5286 remove_proc_entry("rt6_stats", net->proc_net);
5290 static struct pernet_operations ip6_route_net_ops = {
5291 .init = ip6_route_net_init,
5292 .exit = ip6_route_net_exit,
5295 static int __net_init ipv6_inetpeer_init(struct net *net)
5297 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
5301 inet_peer_base_init(bp);
5302 net->ipv6.peers = bp;
5306 static void __net_exit ipv6_inetpeer_exit(struct net *net)
5308 struct inet_peer_base *bp = net->ipv6.peers;
5310 net->ipv6.peers = NULL;
5311 inetpeer_invalidate_tree(bp);
5315 static struct pernet_operations ipv6_inetpeer_ops = {
5316 .init = ipv6_inetpeer_init,
5317 .exit = ipv6_inetpeer_exit,
5320 static struct pernet_operations ip6_route_net_late_ops = {
5321 .init = ip6_route_net_init_late,
5322 .exit = ip6_route_net_exit_late,
5325 static struct notifier_block ip6_route_dev_notifier = {
5326 .notifier_call = ip6_route_dev_notify,
5327 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
5330 void __init ip6_route_init_special_entries(void)
5332 /* Registering of the loopback is done before this portion of code,
5333 * the loopback reference in rt6_info will not be taken, do it
5334 * manually for init_net */
5335 init_net.ipv6.fib6_null_entry->fib6_nh.nh_dev = init_net.loopback_dev;
5336 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
5337 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5338 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5339 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
5340 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5341 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
5342 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
5346 int __init ip6_route_init(void)
5352 ip6_dst_ops_template.kmem_cachep =
5353 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
5354 SLAB_HWCACHE_ALIGN, NULL);
5355 if (!ip6_dst_ops_template.kmem_cachep)
5358 ret = dst_entries_init(&ip6_dst_blackhole_ops);
5360 goto out_kmem_cache;
5362 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
5364 goto out_dst_entries;
5366 ret = register_pernet_subsys(&ip6_route_net_ops);
5368 goto out_register_inetpeer;
5370 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
5374 goto out_register_subsys;
5380 ret = fib6_rules_init();
5384 ret = register_pernet_subsys(&ip6_route_net_late_ops);
5386 goto fib6_rules_init;
5388 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
5389 inet6_rtm_newroute, NULL, 0);
5391 goto out_register_late_subsys;
5393 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
5394 inet6_rtm_delroute, NULL, 0);
5396 goto out_register_late_subsys;
5398 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
5399 inet6_rtm_getroute, NULL,
5400 RTNL_FLAG_DOIT_UNLOCKED);
5402 goto out_register_late_subsys;
5404 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
5406 goto out_register_late_subsys;
5408 for_each_possible_cpu(cpu) {
5409 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
5411 INIT_LIST_HEAD(&ul->head);
5412 spin_lock_init(&ul->lock);
5418 out_register_late_subsys:
5419 rtnl_unregister_all(PF_INET6);
5420 unregister_pernet_subsys(&ip6_route_net_late_ops);
5422 fib6_rules_cleanup();
5427 out_register_subsys:
5428 unregister_pernet_subsys(&ip6_route_net_ops);
5429 out_register_inetpeer:
5430 unregister_pernet_subsys(&ipv6_inetpeer_ops);
5432 dst_entries_destroy(&ip6_dst_blackhole_ops);
5434 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
5438 void ip6_route_cleanup(void)
5440 unregister_netdevice_notifier(&ip6_route_dev_notifier);
5441 unregister_pernet_subsys(&ip6_route_net_late_ops);
5442 fib6_rules_cleanup();
5445 unregister_pernet_subsys(&ipv6_inetpeer_ops);
5446 unregister_pernet_subsys(&ip6_route_net_ops);
5447 dst_entries_destroy(&ip6_dst_blackhole_ops);
5448 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);