1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Linux INET6 implementation
7 * Pedro Roque <roque@di.fc.ul.pt>
12 * YOSHIFUJI Hideaki @USAGI
13 * reworked default router selection.
14 * - respect outgoing interface
15 * - select from (probably) reachable routers (i.e.
16 * routers in REACHABLE, STALE, DELAY or PROBE states).
17 * - always select the same router if it is (probably)
18 * reachable. otherwise, round-robin the list.
20 * Fixed routing subtrees.
23 #define pr_fmt(fmt) "IPv6: " fmt
25 #include <linux/capability.h>
26 #include <linux/errno.h>
27 #include <linux/export.h>
28 #include <linux/types.h>
29 #include <linux/times.h>
30 #include <linux/socket.h>
31 #include <linux/sockios.h>
32 #include <linux/net.h>
33 #include <linux/route.h>
34 #include <linux/netdevice.h>
35 #include <linux/in6.h>
36 #include <linux/mroute6.h>
37 #include <linux/init.h>
38 #include <linux/if_arp.h>
39 #include <linux/proc_fs.h>
40 #include <linux/seq_file.h>
41 #include <linux/nsproxy.h>
42 #include <linux/slab.h>
43 #include <linux/jhash.h>
44 #include <net/net_namespace.h>
47 #include <net/ip6_fib.h>
48 #include <net/ip6_route.h>
49 #include <net/ndisc.h>
50 #include <net/addrconf.h>
52 #include <linux/rtnetlink.h>
54 #include <net/dst_metadata.h>
56 #include <net/netevent.h>
57 #include <net/netlink.h>
59 #include <net/lwtunnel.h>
60 #include <net/ip_tunnels.h>
61 #include <net/l3mdev.h>
63 #include <linux/uaccess.h>
64 #include <linux/btf_ids.h>
67 #include <linux/sysctl.h>
70 static int ip6_rt_type_to_error(u8 fib6_type);
72 #define CREATE_TRACE_POINTS
73 #include <trace/events/fib6.h>
74 EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
75 #undef CREATE_TRACE_POINTS
78 RT6_NUD_FAIL_HARD = -3,
79 RT6_NUD_FAIL_PROBE = -2,
80 RT6_NUD_FAIL_DO_RR = -1,
84 INDIRECT_CALLABLE_SCOPE
85 struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
86 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
87 INDIRECT_CALLABLE_SCOPE
88 unsigned int ip6_mtu(const struct dst_entry *dst);
89 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
90 static void ip6_dst_destroy(struct dst_entry *);
91 static void ip6_dst_ifdown(struct dst_entry *,
92 struct net_device *dev, int how);
93 static int ip6_dst_gc(struct dst_ops *ops);
95 static int ip6_pkt_discard(struct sk_buff *skb);
96 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
97 static int ip6_pkt_prohibit(struct sk_buff *skb);
98 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
99 static void ip6_link_failure(struct sk_buff *skb);
100 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
101 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(const struct fib6_nh *nh, u32 fib6_flags, int oif,
107 static size_t rt6_nlmsg_size(struct fib6_info *f6i);
108 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
109 struct fib6_info *rt, struct dst_entry *dst,
110 struct in6_addr *dest, struct in6_addr *src,
111 int iif, int type, u32 portid, u32 seq,
113 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
114 const struct in6_addr *daddr,
115 const struct in6_addr *saddr);
117 #ifdef CONFIG_IPV6_ROUTE_INFO
118 static struct fib6_info *rt6_add_route_info(struct net *net,
119 const struct in6_addr *prefix, int prefixlen,
120 const struct in6_addr *gwaddr,
121 struct net_device *dev,
123 static struct fib6_info *rt6_get_route_info(struct net *net,
124 const struct in6_addr *prefix, int prefixlen,
125 const struct in6_addr *gwaddr,
126 struct net_device *dev);
129 struct uncached_list {
131 struct list_head head;
134 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
136 void rt6_uncached_list_add(struct rt6_info *rt)
138 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
140 rt->rt6i_uncached_list = ul;
142 spin_lock_bh(&ul->lock);
143 list_add_tail(&rt->rt6i_uncached, &ul->head);
144 spin_unlock_bh(&ul->lock);
147 void rt6_uncached_list_del(struct rt6_info *rt)
149 if (!list_empty(&rt->rt6i_uncached)) {
150 struct uncached_list *ul = rt->rt6i_uncached_list;
151 struct net *net = dev_net(rt->dst.dev);
153 spin_lock_bh(&ul->lock);
154 list_del(&rt->rt6i_uncached);
155 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
156 spin_unlock_bh(&ul->lock);
160 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
162 struct net_device *loopback_dev = net->loopback_dev;
165 if (dev == loopback_dev)
168 for_each_possible_cpu(cpu) {
169 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
172 spin_lock_bh(&ul->lock);
173 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
174 struct inet6_dev *rt_idev = rt->rt6i_idev;
175 struct net_device *rt_dev = rt->dst.dev;
177 if (rt_idev->dev == dev) {
178 rt->rt6i_idev = in6_dev_get(loopback_dev);
179 in6_dev_put(rt_idev);
183 rt->dst.dev = blackhole_netdev;
184 dev_hold(rt->dst.dev);
188 spin_unlock_bh(&ul->lock);
192 static inline const void *choose_neigh_daddr(const struct in6_addr *p,
196 if (!ipv6_addr_any(p))
197 return (const void *) p;
199 return &ipv6_hdr(skb)->daddr;
203 struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
204 struct net_device *dev,
210 daddr = choose_neigh_daddr(gw, skb, daddr);
211 n = __ipv6_neigh_lookup(dev, daddr);
215 n = neigh_create(&nd_tbl, daddr, dev);
216 return IS_ERR(n) ? NULL : n;
219 static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
223 const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
225 return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
226 dst->dev, skb, daddr);
229 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
231 struct net_device *dev = dst->dev;
232 struct rt6_info *rt = (struct rt6_info *)dst;
234 daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
237 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
239 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
241 __ipv6_confirm_neigh(dev, daddr);
244 static struct dst_ops ip6_dst_ops_template = {
248 .check = ip6_dst_check,
249 .default_advmss = ip6_default_advmss,
251 .cow_metrics = dst_cow_metrics_generic,
252 .destroy = ip6_dst_destroy,
253 .ifdown = ip6_dst_ifdown,
254 .negative_advice = ip6_negative_advice,
255 .link_failure = ip6_link_failure,
256 .update_pmtu = ip6_rt_update_pmtu,
257 .redirect = rt6_do_redirect,
258 .local_out = __ip6_local_out,
259 .neigh_lookup = ip6_dst_neigh_lookup,
260 .confirm_neigh = ip6_confirm_neigh,
263 static struct dst_ops ip6_dst_blackhole_ops = {
265 .default_advmss = ip6_default_advmss,
266 .neigh_lookup = ip6_dst_neigh_lookup,
267 .check = ip6_dst_check,
268 .destroy = ip6_dst_destroy,
269 .cow_metrics = dst_cow_metrics_generic,
270 .update_pmtu = dst_blackhole_update_pmtu,
271 .redirect = dst_blackhole_redirect,
272 .mtu = dst_blackhole_mtu,
275 static const u32 ip6_template_metrics[RTAX_MAX] = {
276 [RTAX_HOPLIMIT - 1] = 0,
279 static const struct fib6_info fib6_null_entry_template = {
280 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
281 .fib6_protocol = RTPROT_KERNEL,
282 .fib6_metric = ~(u32)0,
283 .fib6_ref = REFCOUNT_INIT(1),
284 .fib6_type = RTN_UNREACHABLE,
285 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
288 static const struct rt6_info ip6_null_entry_template = {
290 .__refcnt = ATOMIC_INIT(1),
292 .obsolete = DST_OBSOLETE_FORCE_CHK,
293 .error = -ENETUNREACH,
294 .input = ip6_pkt_discard,
295 .output = ip6_pkt_discard_out,
297 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
300 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
302 static const struct rt6_info ip6_prohibit_entry_template = {
304 .__refcnt = ATOMIC_INIT(1),
306 .obsolete = DST_OBSOLETE_FORCE_CHK,
308 .input = ip6_pkt_prohibit,
309 .output = ip6_pkt_prohibit_out,
311 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
314 static const struct rt6_info ip6_blk_hole_entry_template = {
316 .__refcnt = ATOMIC_INIT(1),
318 .obsolete = DST_OBSOLETE_FORCE_CHK,
320 .input = dst_discard,
321 .output = dst_discard_out,
323 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
328 static void rt6_info_init(struct rt6_info *rt)
330 struct dst_entry *dst = &rt->dst;
332 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
333 INIT_LIST_HEAD(&rt->rt6i_uncached);
336 /* allocate dst with ip6_dst_ops */
337 struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
340 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
341 1, DST_OBSOLETE_FORCE_CHK, flags);
345 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
350 EXPORT_SYMBOL(ip6_dst_alloc);
352 static void ip6_dst_destroy(struct dst_entry *dst)
354 struct rt6_info *rt = (struct rt6_info *)dst;
355 struct fib6_info *from;
356 struct inet6_dev *idev;
358 ip_dst_metrics_put(dst);
359 rt6_uncached_list_del(rt);
361 idev = rt->rt6i_idev;
363 rt->rt6i_idev = NULL;
367 from = xchg((__force struct fib6_info **)&rt->from, NULL);
368 fib6_info_release(from);
371 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
374 struct rt6_info *rt = (struct rt6_info *)dst;
375 struct inet6_dev *idev = rt->rt6i_idev;
376 struct net_device *loopback_dev =
377 dev_net(dev)->loopback_dev;
379 if (idev && idev->dev != loopback_dev) {
380 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
382 rt->rt6i_idev = loopback_idev;
388 static bool __rt6_check_expired(const struct rt6_info *rt)
390 if (rt->rt6i_flags & RTF_EXPIRES)
391 return time_after(jiffies, rt->dst.expires);
396 static bool rt6_check_expired(const struct rt6_info *rt)
398 struct fib6_info *from;
400 from = rcu_dereference(rt->from);
402 if (rt->rt6i_flags & RTF_EXPIRES) {
403 if (time_after(jiffies, rt->dst.expires))
406 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
407 fib6_check_expired(from);
412 void fib6_select_path(const struct net *net, struct fib6_result *res,
413 struct flowi6 *fl6, int oif, bool have_oif_match,
414 const struct sk_buff *skb, int strict)
416 struct fib6_info *sibling, *next_sibling;
417 struct fib6_info *match = res->f6i;
419 if (!match->nh && (!match->fib6_nsiblings || have_oif_match))
422 if (match->nh && have_oif_match && res->nh)
425 /* We might have already computed the hash for ICMPv6 errors. In such
426 * case it will always be non-zero. Otherwise now is the time to do it.
429 (!match->nh || nexthop_is_multipath(match->nh)))
430 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
432 if (unlikely(match->nh)) {
433 nexthop_path_fib6_result(res, fl6->mp_hash);
437 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
440 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
442 const struct fib6_nh *nh = sibling->fib6_nh;
445 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
446 if (fl6->mp_hash > nh_upper_bound)
448 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
456 res->nh = match->fib6_nh;
460 * Route lookup. rcu_read_lock() should be held.
463 static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
464 const struct in6_addr *saddr, int oif, int flags)
466 const struct net_device *dev;
468 if (nh->fib_nh_flags & RTNH_F_DEAD)
471 dev = nh->fib_nh_dev;
473 if (dev->ifindex == oif)
476 if (ipv6_chk_addr(net, saddr, dev,
477 flags & RT6_LOOKUP_F_IFACE))
484 struct fib6_nh_dm_arg {
486 const struct in6_addr *saddr;
492 static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
494 struct fib6_nh_dm_arg *arg = _arg;
497 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
501 /* returns fib6_nh from nexthop or NULL */
502 static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
503 struct fib6_result *res,
504 const struct in6_addr *saddr,
507 struct fib6_nh_dm_arg arg = {
514 if (nexthop_is_blackhole(nh))
517 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
523 static void rt6_device_match(struct net *net, struct fib6_result *res,
524 const struct in6_addr *saddr, int oif, int flags)
526 struct fib6_info *f6i = res->f6i;
527 struct fib6_info *spf6i;
530 if (!oif && ipv6_addr_any(saddr)) {
531 if (unlikely(f6i->nh)) {
532 nh = nexthop_fib6_nh(f6i->nh);
533 if (nexthop_is_blackhole(f6i->nh))
538 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
542 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
543 bool matched = false;
545 if (unlikely(spf6i->nh)) {
546 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
552 if (__rt6_device_match(net, nh, saddr, oif, flags))
561 if (oif && flags & RT6_LOOKUP_F_IFACE) {
562 res->f6i = net->ipv6.fib6_null_entry;
563 nh = res->f6i->fib6_nh;
567 if (unlikely(f6i->nh)) {
568 nh = nexthop_fib6_nh(f6i->nh);
569 if (nexthop_is_blackhole(f6i->nh))
575 if (nh->fib_nh_flags & RTNH_F_DEAD) {
576 res->f6i = net->ipv6.fib6_null_entry;
577 nh = res->f6i->fib6_nh;
581 res->fib6_type = res->f6i->fib6_type;
582 res->fib6_flags = res->f6i->fib6_flags;
586 res->fib6_flags |= RTF_REJECT;
587 res->fib6_type = RTN_BLACKHOLE;
591 #ifdef CONFIG_IPV6_ROUTER_PREF
592 struct __rt6_probe_work {
593 struct work_struct work;
594 struct in6_addr target;
595 struct net_device *dev;
598 static void rt6_probe_deferred(struct work_struct *w)
600 struct in6_addr mcaddr;
601 struct __rt6_probe_work *work =
602 container_of(w, struct __rt6_probe_work, work);
604 addrconf_addr_solict_mult(&work->target, &mcaddr);
605 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
610 static void rt6_probe(struct fib6_nh *fib6_nh)
612 struct __rt6_probe_work *work = NULL;
613 const struct in6_addr *nh_gw;
614 unsigned long last_probe;
615 struct neighbour *neigh;
616 struct net_device *dev;
617 struct inet6_dev *idev;
620 * Okay, this does not seem to be appropriate
621 * for now, however, we need to check if it
622 * is really so; aka Router Reachability Probing.
624 * Router Reachability Probe MUST be rate-limited
625 * to no more than one per minute.
627 if (!fib6_nh->fib_nh_gw_family)
630 nh_gw = &fib6_nh->fib_nh_gw6;
631 dev = fib6_nh->fib_nh_dev;
633 last_probe = READ_ONCE(fib6_nh->last_probe);
634 idev = __in6_dev_get(dev);
635 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
637 if (neigh->nud_state & NUD_VALID)
640 write_lock(&neigh->lock);
641 if (!(neigh->nud_state & NUD_VALID) &&
643 neigh->updated + idev->cnf.rtr_probe_interval)) {
644 work = kmalloc(sizeof(*work), GFP_ATOMIC);
646 __neigh_set_probe_once(neigh);
648 write_unlock(&neigh->lock);
649 } else if (time_after(jiffies, last_probe +
650 idev->cnf.rtr_probe_interval)) {
651 work = kmalloc(sizeof(*work), GFP_ATOMIC);
654 if (!work || cmpxchg(&fib6_nh->last_probe,
655 last_probe, jiffies) != last_probe) {
658 INIT_WORK(&work->work, rt6_probe_deferred);
659 work->target = *nh_gw;
662 schedule_work(&work->work);
666 rcu_read_unlock_bh();
669 static inline void rt6_probe(struct fib6_nh *fib6_nh)
675 * Default Router Selection (RFC 2461 6.3.6)
677 static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
679 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
680 struct neighbour *neigh;
683 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
684 &fib6_nh->fib_nh_gw6);
686 read_lock(&neigh->lock);
687 if (neigh->nud_state & NUD_VALID)
688 ret = RT6_NUD_SUCCEED;
689 #ifdef CONFIG_IPV6_ROUTER_PREF
690 else if (!(neigh->nud_state & NUD_FAILED))
691 ret = RT6_NUD_SUCCEED;
693 ret = RT6_NUD_FAIL_PROBE;
695 read_unlock(&neigh->lock);
697 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
698 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
700 rcu_read_unlock_bh();
705 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
710 if (!oif || nh->fib_nh_dev->ifindex == oif)
713 if (!m && (strict & RT6_LOOKUP_F_IFACE))
714 return RT6_NUD_FAIL_HARD;
715 #ifdef CONFIG_IPV6_ROUTER_PREF
716 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
718 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
719 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
720 int n = rt6_check_neigh(nh);
727 static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
728 int oif, int strict, int *mpri, bool *do_rr)
730 bool match_do_rr = false;
734 if (nh->fib_nh_flags & RTNH_F_DEAD)
737 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
738 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
739 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
742 m = rt6_score_route(nh, fib6_flags, oif, strict);
743 if (m == RT6_NUD_FAIL_DO_RR) {
745 m = 0; /* lowest valid score */
746 } else if (m == RT6_NUD_FAIL_HARD) {
750 if (strict & RT6_LOOKUP_F_REACHABLE)
753 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
755 *do_rr = match_do_rr;
763 struct fib6_nh_frl_arg {
772 static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
774 struct fib6_nh_frl_arg *arg = _arg;
777 return find_match(nh, arg->flags, arg->oif, arg->strict,
778 arg->mpri, arg->do_rr);
781 static void __find_rr_leaf(struct fib6_info *f6i_start,
782 struct fib6_info *nomatch, u32 metric,
783 struct fib6_result *res, struct fib6_info **cont,
784 int oif, int strict, bool *do_rr, int *mpri)
786 struct fib6_info *f6i;
788 for (f6i = f6i_start;
789 f6i && f6i != nomatch;
790 f6i = rcu_dereference(f6i->fib6_next)) {
791 bool matched = false;
794 if (cont && f6i->fib6_metric != metric) {
799 if (fib6_check_expired(f6i))
802 if (unlikely(f6i->nh)) {
803 struct fib6_nh_frl_arg arg = {
804 .flags = f6i->fib6_flags,
811 if (nexthop_is_blackhole(f6i->nh)) {
812 res->fib6_flags = RTF_REJECT;
813 res->fib6_type = RTN_BLACKHOLE;
815 res->nh = nexthop_fib6_nh(f6i->nh);
818 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
825 if (find_match(nh, f6i->fib6_flags, oif, strict,
832 res->fib6_flags = f6i->fib6_flags;
833 res->fib6_type = f6i->fib6_type;
838 static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
839 struct fib6_info *rr_head, int oif, int strict,
840 bool *do_rr, struct fib6_result *res)
842 u32 metric = rr_head->fib6_metric;
843 struct fib6_info *cont = NULL;
846 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
847 oif, strict, do_rr, &mpri);
849 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
850 oif, strict, do_rr, &mpri);
852 if (res->f6i || !cont)
855 __find_rr_leaf(cont, NULL, metric, res, NULL,
856 oif, strict, do_rr, &mpri);
859 static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
860 struct fib6_result *res, int strict)
862 struct fib6_info *leaf = rcu_dereference(fn->leaf);
863 struct fib6_info *rt0;
867 /* make sure this function or its helpers sets f6i */
870 if (!leaf || leaf == net->ipv6.fib6_null_entry)
873 rt0 = rcu_dereference(fn->rr_ptr);
877 /* Double check to make sure fn is not an intermediate node
878 * and fn->leaf does not points to its child's leaf
879 * (This might happen if all routes under fn are deleted from
880 * the tree and fib6_repair_tree() is called on the node.)
882 key_plen = rt0->fib6_dst.plen;
883 #ifdef CONFIG_IPV6_SUBTREES
884 if (rt0->fib6_src.plen)
885 key_plen = rt0->fib6_src.plen;
887 if (fn->fn_bit != key_plen)
890 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
892 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
894 /* no entries matched; do round-robin */
895 if (!next || next->fib6_metric != rt0->fib6_metric)
899 spin_lock_bh(&leaf->fib6_table->tb6_lock);
900 /* make sure next is not being deleted from the tree */
902 rcu_assign_pointer(fn->rr_ptr, next);
903 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
909 res->f6i = net->ipv6.fib6_null_entry;
910 res->nh = res->f6i->fib6_nh;
911 res->fib6_flags = res->f6i->fib6_flags;
912 res->fib6_type = res->f6i->fib6_type;
916 static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
918 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
919 res->nh->fib_nh_gw_family;
922 #ifdef CONFIG_IPV6_ROUTE_INFO
923 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
924 const struct in6_addr *gwaddr)
926 struct net *net = dev_net(dev);
927 struct route_info *rinfo = (struct route_info *) opt;
928 struct in6_addr prefix_buf, *prefix;
930 unsigned long lifetime;
931 struct fib6_info *rt;
933 if (len < sizeof(struct route_info)) {
937 /* Sanity check for prefix_len and length */
938 if (rinfo->length > 3) {
940 } else if (rinfo->prefix_len > 128) {
942 } else if (rinfo->prefix_len > 64) {
943 if (rinfo->length < 2) {
946 } else if (rinfo->prefix_len > 0) {
947 if (rinfo->length < 1) {
952 pref = rinfo->route_pref;
953 if (pref == ICMPV6_ROUTER_PREF_INVALID)
956 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
958 if (rinfo->length == 3)
959 prefix = (struct in6_addr *)rinfo->prefix;
961 /* this function is safe */
962 ipv6_addr_prefix(&prefix_buf,
963 (struct in6_addr *)rinfo->prefix,
965 prefix = &prefix_buf;
968 if (rinfo->prefix_len == 0)
969 rt = rt6_get_dflt_router(net, gwaddr, dev);
971 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
974 if (rt && !lifetime) {
975 ip6_del_rt(net, rt, false);
980 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
983 rt->fib6_flags = RTF_ROUTEINFO |
984 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
987 if (!addrconf_finite_timeout(lifetime))
988 fib6_clean_expires(rt);
990 fib6_set_expires(rt, jiffies + HZ * lifetime);
992 fib6_info_release(rt);
999 * Misc support functions
1002 /* called with rcu_lock held */
1003 static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1005 struct net_device *dev = res->nh->fib_nh_dev;
1007 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1008 /* for copies of local routes, dst->dev needs to be the
1009 * device if it is a master device, the master device if
1010 * device is enslaved, and the loopback as the default
1012 if (netif_is_l3_slave(dev) &&
1013 !rt6_need_strict(&res->f6i->fib6_dst.addr))
1014 dev = l3mdev_master_dev_rcu(dev);
1015 else if (!netif_is_l3_master(dev))
1016 dev = dev_net(dev)->loopback_dev;
1017 /* last case is netif_is_l3_master(dev) is true in which
1018 * case we want dev returned to be dev
1025 static const int fib6_prop[RTN_MAX + 1] = {
1029 [RTN_BROADCAST] = 0,
1031 [RTN_MULTICAST] = 0,
1032 [RTN_BLACKHOLE] = -EINVAL,
1033 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1034 [RTN_PROHIBIT] = -EACCES,
1035 [RTN_THROW] = -EAGAIN,
1036 [RTN_NAT] = -EINVAL,
1037 [RTN_XRESOLVE] = -EINVAL,
1040 static int ip6_rt_type_to_error(u8 fib6_type)
1042 return fib6_prop[fib6_type];
1045 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1047 unsigned short flags = 0;
1049 if (rt->dst_nocount)
1050 flags |= DST_NOCOUNT;
1051 if (rt->dst_nopolicy)
1052 flags |= DST_NOPOLICY;
1057 static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1059 rt->dst.error = ip6_rt_type_to_error(fib6_type);
1061 switch (fib6_type) {
1063 rt->dst.output = dst_discard_out;
1064 rt->dst.input = dst_discard;
1067 rt->dst.output = ip6_pkt_prohibit_out;
1068 rt->dst.input = ip6_pkt_prohibit;
1071 case RTN_UNREACHABLE:
1073 rt->dst.output = ip6_pkt_discard_out;
1074 rt->dst.input = ip6_pkt_discard;
1079 static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1081 struct fib6_info *f6i = res->f6i;
1083 if (res->fib6_flags & RTF_REJECT) {
1084 ip6_rt_init_dst_reject(rt, res->fib6_type);
1089 rt->dst.output = ip6_output;
1091 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1092 rt->dst.input = ip6_input;
1093 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1094 rt->dst.input = ip6_mc_input;
1096 rt->dst.input = ip6_forward;
1099 if (res->nh->fib_nh_lws) {
1100 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1101 lwtunnel_set_redirect(&rt->dst);
1104 rt->dst.lastuse = jiffies;
1107 /* Caller must already hold reference to @from */
1108 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1110 rt->rt6i_flags &= ~RTF_EXPIRES;
1111 rcu_assign_pointer(rt->from, from);
1112 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1115 /* Caller must already hold reference to f6i in result */
1116 static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1118 const struct fib6_nh *nh = res->nh;
1119 const struct net_device *dev = nh->fib_nh_dev;
1120 struct fib6_info *f6i = res->f6i;
1122 ip6_rt_init_dst(rt, res);
1124 rt->rt6i_dst = f6i->fib6_dst;
1125 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1126 rt->rt6i_flags = res->fib6_flags;
1127 if (nh->fib_nh_gw_family) {
1128 rt->rt6i_gateway = nh->fib_nh_gw6;
1129 rt->rt6i_flags |= RTF_GATEWAY;
1131 rt6_set_from(rt, f6i);
1132 #ifdef CONFIG_IPV6_SUBTREES
1133 rt->rt6i_src = f6i->fib6_src;
1137 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1138 struct in6_addr *saddr)
1140 struct fib6_node *pn, *sn;
1142 if (fn->fn_flags & RTN_TL_ROOT)
1144 pn = rcu_dereference(fn->parent);
1145 sn = FIB6_SUBTREE(pn);
1147 fn = fib6_node_lookup(sn, NULL, saddr);
1150 if (fn->fn_flags & RTN_RTINFO)
1155 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1157 struct rt6_info *rt = *prt;
1159 if (dst_hold_safe(&rt->dst))
1162 rt = net->ipv6.ip6_null_entry;
1171 /* called with rcu_lock held */
1172 static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1174 struct net_device *dev = res->nh->fib_nh_dev;
1175 struct fib6_info *f6i = res->f6i;
1176 unsigned short flags;
1177 struct rt6_info *nrt;
1179 if (!fib6_info_hold_safe(f6i))
1182 flags = fib6_info_dst_flags(f6i);
1183 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1185 fib6_info_release(f6i);
1189 ip6_rt_copy_init(nrt, res);
1193 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1194 dst_hold(&nrt->dst);
1198 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
1199 struct fib6_table *table,
1201 const struct sk_buff *skb,
1204 struct fib6_result res = {};
1205 struct fib6_node *fn;
1206 struct rt6_info *rt;
1208 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1209 flags &= ~RT6_LOOKUP_F_IFACE;
1212 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1214 res.f6i = rcu_dereference(fn->leaf);
1216 res.f6i = net->ipv6.fib6_null_entry;
1218 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1221 if (res.f6i == net->ipv6.fib6_null_entry) {
1222 fn = fib6_backtrack(fn, &fl6->saddr);
1226 rt = net->ipv6.ip6_null_entry;
1229 } else if (res.fib6_flags & RTF_REJECT) {
1233 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1234 fl6->flowi6_oif != 0, skb, flags);
1236 /* Search through exception table */
1237 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1239 if (ip6_hold_safe(net, &rt))
1240 dst_use_noref(&rt->dst, jiffies);
1243 rt = ip6_create_rt_rcu(&res);
1247 trace_fib6_table_lookup(net, &res, table, fl6);
1254 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1255 const struct sk_buff *skb, int flags)
1257 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1259 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1261 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1262 const struct in6_addr *saddr, int oif,
1263 const struct sk_buff *skb, int strict)
1265 struct flowi6 fl6 = {
1269 struct dst_entry *dst;
1270 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1273 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1274 flags |= RT6_LOOKUP_F_HAS_SADDR;
1277 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1278 if (dst->error == 0)
1279 return (struct rt6_info *) dst;
1285 EXPORT_SYMBOL(rt6_lookup);
1287 /* ip6_ins_rt is called with FREE table->tb6_lock.
1288 * It takes new route entry, the addition fails by any reason the
1289 * route is released.
1290 * Caller must hold dst before calling it.
1293 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1294 struct netlink_ext_ack *extack)
1297 struct fib6_table *table;
1299 table = rt->fib6_table;
1300 spin_lock_bh(&table->tb6_lock);
1301 err = fib6_add(&table->tb6_root, rt, info, extack);
1302 spin_unlock_bh(&table->tb6_lock);
1307 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1309 struct nl_info info = { .nl_net = net, };
1311 return __ip6_ins_rt(rt, &info, NULL);
1314 static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1315 const struct in6_addr *daddr,
1316 const struct in6_addr *saddr)
1318 struct fib6_info *f6i = res->f6i;
1319 struct net_device *dev;
1320 struct rt6_info *rt;
1326 if (!fib6_info_hold_safe(f6i))
1329 dev = ip6_rt_get_dev_rcu(res);
1330 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1332 fib6_info_release(f6i);
1336 ip6_rt_copy_init(rt, res);
1337 rt->rt6i_flags |= RTF_CACHE;
1338 rt->rt6i_dst.addr = *daddr;
1339 rt->rt6i_dst.plen = 128;
1341 if (!rt6_is_gw_or_nonexthop(res)) {
1342 if (f6i->fib6_dst.plen != 128 &&
1343 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1344 rt->rt6i_flags |= RTF_ANYCAST;
1345 #ifdef CONFIG_IPV6_SUBTREES
1346 if (rt->rt6i_src.plen && saddr) {
1347 rt->rt6i_src.addr = *saddr;
1348 rt->rt6i_src.plen = 128;
1356 static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1358 struct fib6_info *f6i = res->f6i;
1359 unsigned short flags = fib6_info_dst_flags(f6i);
1360 struct net_device *dev;
1361 struct rt6_info *pcpu_rt;
1363 if (!fib6_info_hold_safe(f6i))
1367 dev = ip6_rt_get_dev_rcu(res);
1368 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
1371 fib6_info_release(f6i);
1374 ip6_rt_copy_init(pcpu_rt, res);
1375 pcpu_rt->rt6i_flags |= RTF_PCPU;
1378 pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1383 static bool rt6_is_valid(const struct rt6_info *rt6)
1385 return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1388 /* It should be called with rcu_read_lock() acquired */
1389 static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1391 struct rt6_info *pcpu_rt;
1393 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1395 if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1396 struct rt6_info *prev, **p;
1398 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1399 prev = xchg(p, NULL);
1401 dst_dev_put(&prev->dst);
1402 dst_release(&prev->dst);
1411 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1412 const struct fib6_result *res)
1414 struct rt6_info *pcpu_rt, *prev, **p;
1416 pcpu_rt = ip6_rt_pcpu_alloc(res);
1420 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1421 prev = cmpxchg(p, NULL, pcpu_rt);
1424 if (res->f6i->fib6_destroying) {
1425 struct fib6_info *from;
1427 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1428 fib6_info_release(from);
1434 /* exception hash table implementation
1436 static DEFINE_SPINLOCK(rt6_exception_lock);
1438 /* Remove rt6_ex from hash table and free the memory
1439 * Caller must hold rt6_exception_lock
1441 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1442 struct rt6_exception *rt6_ex)
1444 struct fib6_info *from;
1447 if (!bucket || !rt6_ex)
1450 net = dev_net(rt6_ex->rt6i->dst.dev);
1451 net->ipv6.rt6_stats->fib_rt_cache--;
1453 /* purge completely the exception to allow releasing the held resources:
1454 * some [sk] cache may keep the dst around for unlimited time
1456 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
1457 fib6_info_release(from);
1458 dst_dev_put(&rt6_ex->rt6i->dst);
1460 hlist_del_rcu(&rt6_ex->hlist);
1461 dst_release(&rt6_ex->rt6i->dst);
1462 kfree_rcu(rt6_ex, rcu);
1463 WARN_ON_ONCE(!bucket->depth);
1467 /* Remove oldest rt6_ex in bucket and free the memory
1468 * Caller must hold rt6_exception_lock
1470 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1472 struct rt6_exception *rt6_ex, *oldest = NULL;
1477 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1478 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1481 rt6_remove_exception(bucket, oldest);
1484 static u32 rt6_exception_hash(const struct in6_addr *dst,
1485 const struct in6_addr *src)
1487 static u32 seed __read_mostly;
1490 net_get_random_once(&seed, sizeof(seed));
1491 val = jhash2((const u32 *)dst, sizeof(*dst)/sizeof(u32), seed);
1493 #ifdef CONFIG_IPV6_SUBTREES
1495 val = jhash2((const u32 *)src, sizeof(*src)/sizeof(u32), val);
1497 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1500 /* Helper function to find the cached rt in the hash table
1501 * and update bucket pointer to point to the bucket for this
1502 * (daddr, saddr) pair
1503 * Caller must hold rt6_exception_lock
1505 static struct rt6_exception *
1506 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1507 const struct in6_addr *daddr,
1508 const struct in6_addr *saddr)
1510 struct rt6_exception *rt6_ex;
1513 if (!(*bucket) || !daddr)
1516 hval = rt6_exception_hash(daddr, saddr);
1519 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1520 struct rt6_info *rt6 = rt6_ex->rt6i;
1521 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1523 #ifdef CONFIG_IPV6_SUBTREES
1524 if (matched && saddr)
1525 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1533 /* Helper function to find the cached rt in the hash table
1534 * and update bucket pointer to point to the bucket for this
1535 * (daddr, saddr) pair
1536 * Caller must hold rcu_read_lock()
1538 static struct rt6_exception *
1539 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1540 const struct in6_addr *daddr,
1541 const struct in6_addr *saddr)
1543 struct rt6_exception *rt6_ex;
1546 WARN_ON_ONCE(!rcu_read_lock_held());
1548 if (!(*bucket) || !daddr)
1551 hval = rt6_exception_hash(daddr, saddr);
1554 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1555 struct rt6_info *rt6 = rt6_ex->rt6i;
1556 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1558 #ifdef CONFIG_IPV6_SUBTREES
1559 if (matched && saddr)
1560 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1568 static unsigned int fib6_mtu(const struct fib6_result *res)
1570 const struct fib6_nh *nh = res->nh;
1573 if (res->f6i->fib6_pmtu) {
1574 mtu = res->f6i->fib6_pmtu;
1576 struct net_device *dev = nh->fib_nh_dev;
1577 struct inet6_dev *idev;
1580 idev = __in6_dev_get(dev);
1581 mtu = idev->cnf.mtu6;
1585 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1587 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1590 #define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1592 /* used when the flushed bit is not relevant, only access to the bucket
1593 * (ie., all bucket users except rt6_insert_exception);
1595 * called under rcu lock; sometimes called with rt6_exception_lock held
1598 struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1601 struct rt6_exception_bucket *bucket;
1604 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1605 lockdep_is_held(lock));
1607 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1609 /* remove bucket flushed bit if set */
1611 unsigned long p = (unsigned long)bucket;
1613 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1614 bucket = (struct rt6_exception_bucket *)p;
1620 static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1622 unsigned long p = (unsigned long)bucket;
1624 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1627 /* called with rt6_exception_lock held */
1628 static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1631 struct rt6_exception_bucket *bucket;
1634 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1635 lockdep_is_held(lock));
1637 p = (unsigned long)bucket;
1638 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1639 bucket = (struct rt6_exception_bucket *)p;
1640 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1643 static int rt6_insert_exception(struct rt6_info *nrt,
1644 const struct fib6_result *res)
1646 struct net *net = dev_net(nrt->dst.dev);
1647 struct rt6_exception_bucket *bucket;
1648 struct fib6_info *f6i = res->f6i;
1649 struct in6_addr *src_key = NULL;
1650 struct rt6_exception *rt6_ex;
1651 struct fib6_nh *nh = res->nh;
1654 spin_lock_bh(&rt6_exception_lock);
1656 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1657 lockdep_is_held(&rt6_exception_lock));
1659 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1665 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1666 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1671 #ifdef CONFIG_IPV6_SUBTREES
1672 /* fib6_src.plen != 0 indicates f6i is in subtree
1673 * and exception table is indexed by a hash of
1674 * both fib6_dst and fib6_src.
1675 * Otherwise, the exception table is indexed by
1676 * a hash of only fib6_dst.
1678 if (f6i->fib6_src.plen)
1679 src_key = &nrt->rt6i_src.addr;
1681 /* rt6_mtu_change() might lower mtu on f6i.
1682 * Only insert this exception route if its mtu
1683 * is less than f6i's mtu value.
1685 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1690 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1693 rt6_remove_exception(bucket, rt6_ex);
1695 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1701 rt6_ex->stamp = jiffies;
1702 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1704 net->ipv6.rt6_stats->fib_rt_cache++;
1706 if (bucket->depth > FIB6_MAX_DEPTH)
1707 rt6_exception_remove_oldest(bucket);
1710 spin_unlock_bh(&rt6_exception_lock);
1712 /* Update fn->fn_sernum to invalidate all cached dst */
1714 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1715 fib6_update_sernum(net, f6i);
1716 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1717 fib6_force_start_gc(net);
1723 static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1725 struct rt6_exception_bucket *bucket;
1726 struct rt6_exception *rt6_ex;
1727 struct hlist_node *tmp;
1730 spin_lock_bh(&rt6_exception_lock);
1732 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1736 /* Prevent rt6_insert_exception() to recreate the bucket list */
1738 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1740 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1741 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1743 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1744 rt6_remove_exception(bucket, rt6_ex);
1746 WARN_ON_ONCE(!from && bucket->depth);
1750 spin_unlock_bh(&rt6_exception_lock);
1753 static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1755 struct fib6_info *f6i = arg;
1757 fib6_nh_flush_exceptions(nh, f6i);
1762 void rt6_flush_exceptions(struct fib6_info *f6i)
1765 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1768 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1771 /* Find cached rt in the hash table inside passed in rt
1772 * Caller has to hold rcu_read_lock()
1774 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1775 const struct in6_addr *daddr,
1776 const struct in6_addr *saddr)
1778 const struct in6_addr *src_key = NULL;
1779 struct rt6_exception_bucket *bucket;
1780 struct rt6_exception *rt6_ex;
1781 struct rt6_info *ret = NULL;
1783 #ifdef CONFIG_IPV6_SUBTREES
1784 /* fib6i_src.plen != 0 indicates f6i is in subtree
1785 * and exception table is indexed by a hash of
1786 * both fib6_dst and fib6_src.
1787 * However, the src addr used to create the hash
1788 * might not be exactly the passed in saddr which
1789 * is a /128 addr from the flow.
1790 * So we need to use f6i->fib6_src to redo lookup
1791 * if the passed in saddr does not find anything.
1792 * (See the logic in ip6_rt_cache_alloc() on how
1793 * rt->rt6i_src is updated.)
1795 if (res->f6i->fib6_src.plen)
1799 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1800 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1802 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1805 #ifdef CONFIG_IPV6_SUBTREES
1806 /* Use fib6_src as src_key and redo lookup */
1807 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1808 src_key = &res->f6i->fib6_src.addr;
1816 /* Remove the passed in cached rt from the hash table that contains it */
1817 static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1818 const struct rt6_info *rt)
1820 const struct in6_addr *src_key = NULL;
1821 struct rt6_exception_bucket *bucket;
1822 struct rt6_exception *rt6_ex;
1825 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1828 spin_lock_bh(&rt6_exception_lock);
1829 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1831 #ifdef CONFIG_IPV6_SUBTREES
1832 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1833 * and exception table is indexed by a hash of
1834 * both rt6i_dst and rt6i_src.
1835 * Otherwise, the exception table is indexed by
1836 * a hash of only rt6i_dst.
1839 src_key = &rt->rt6i_src.addr;
1841 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1845 rt6_remove_exception(bucket, rt6_ex);
1851 spin_unlock_bh(&rt6_exception_lock);
1855 struct fib6_nh_excptn_arg {
1856 struct rt6_info *rt;
1860 static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1862 struct fib6_nh_excptn_arg *arg = _arg;
1865 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1872 static int rt6_remove_exception_rt(struct rt6_info *rt)
1874 struct fib6_info *from;
1876 from = rcu_dereference(rt->from);
1877 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1881 struct fib6_nh_excptn_arg arg = {
1883 .plen = from->fib6_src.plen
1887 /* rc = 1 means an entry was found */
1888 rc = nexthop_for_each_fib6_nh(from->nh,
1889 rt6_nh_remove_exception_rt,
1891 return rc ? 0 : -ENOENT;
1894 return fib6_nh_remove_exception(from->fib6_nh,
1895 from->fib6_src.plen, rt);
1898 /* Find rt6_ex which contains the passed in rt cache and
1901 static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1902 const struct rt6_info *rt)
1904 const struct in6_addr *src_key = NULL;
1905 struct rt6_exception_bucket *bucket;
1906 struct rt6_exception *rt6_ex;
1908 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1909 #ifdef CONFIG_IPV6_SUBTREES
1910 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1911 * and exception table is indexed by a hash of
1912 * both rt6i_dst and rt6i_src.
1913 * Otherwise, the exception table is indexed by
1914 * a hash of only rt6i_dst.
1917 src_key = &rt->rt6i_src.addr;
1919 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1921 rt6_ex->stamp = jiffies;
1924 struct fib6_nh_match_arg {
1925 const struct net_device *dev;
1926 const struct in6_addr *gw;
1927 struct fib6_nh *match;
1930 /* determine if fib6_nh has given device and gateway */
1931 static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1933 struct fib6_nh_match_arg *arg = _arg;
1935 if (arg->dev != nh->fib_nh_dev ||
1936 (arg->gw && !nh->fib_nh_gw_family) ||
1937 (!arg->gw && nh->fib_nh_gw_family) ||
1938 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1943 /* found a match, break the loop */
1947 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1949 struct fib6_info *from;
1950 struct fib6_nh *fib6_nh;
1954 from = rcu_dereference(rt->from);
1955 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1959 struct fib6_nh_match_arg arg = {
1961 .gw = &rt->rt6i_gateway,
1964 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1968 fib6_nh = arg.match;
1970 fib6_nh = from->fib6_nh;
1972 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1977 static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1978 struct rt6_info *rt, int mtu)
1980 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1981 * lowest MTU in the path: always allow updating the route PMTU to
1982 * reflect PMTU decreases.
1984 * If the new MTU is higher, and the route PMTU is equal to the local
1985 * MTU, this means the old MTU is the lowest in the path, so allow
1986 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1990 if (dst_mtu(&rt->dst) >= mtu)
1993 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
1999 static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
2000 const struct fib6_nh *nh, int mtu)
2002 struct rt6_exception_bucket *bucket;
2003 struct rt6_exception *rt6_ex;
2006 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2010 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2011 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2012 struct rt6_info *entry = rt6_ex->rt6i;
2014 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2015 * route), the metrics of its rt->from have already
2018 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2019 rt6_mtu_change_route_allowed(idev, entry, mtu))
2020 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2026 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2028 static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2029 const struct in6_addr *gateway)
2031 struct rt6_exception_bucket *bucket;
2032 struct rt6_exception *rt6_ex;
2033 struct hlist_node *tmp;
2036 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2039 spin_lock_bh(&rt6_exception_lock);
2040 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2042 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2043 hlist_for_each_entry_safe(rt6_ex, tmp,
2044 &bucket->chain, hlist) {
2045 struct rt6_info *entry = rt6_ex->rt6i;
2047 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2048 RTF_CACHE_GATEWAY &&
2049 ipv6_addr_equal(gateway,
2050 &entry->rt6i_gateway)) {
2051 rt6_remove_exception(bucket, rt6_ex);
2058 spin_unlock_bh(&rt6_exception_lock);
2061 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2062 struct rt6_exception *rt6_ex,
2063 struct fib6_gc_args *gc_args,
2066 struct rt6_info *rt = rt6_ex->rt6i;
2068 /* we are pruning and obsoleting aged-out and non gateway exceptions
2069 * even if others have still references to them, so that on next
2070 * dst_check() such references can be dropped.
2071 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2072 * expired, independently from their aging, as per RFC 8201 section 4
2074 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2075 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2076 RT6_TRACE("aging clone %p\n", rt);
2077 rt6_remove_exception(bucket, rt6_ex);
2080 } else if (time_after(jiffies, rt->dst.expires)) {
2081 RT6_TRACE("purging expired route %p\n", rt);
2082 rt6_remove_exception(bucket, rt6_ex);
2086 if (rt->rt6i_flags & RTF_GATEWAY) {
2087 struct neighbour *neigh;
2089 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2091 if (!(neigh && (neigh->flags & NTF_ROUTER))) {
2092 RT6_TRACE("purging route %p via non-router but gateway\n",
2094 rt6_remove_exception(bucket, rt6_ex);
2102 static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2103 struct fib6_gc_args *gc_args,
2106 struct rt6_exception_bucket *bucket;
2107 struct rt6_exception *rt6_ex;
2108 struct hlist_node *tmp;
2111 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2115 spin_lock(&rt6_exception_lock);
2116 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2118 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2119 hlist_for_each_entry_safe(rt6_ex, tmp,
2120 &bucket->chain, hlist) {
2121 rt6_age_examine_exception(bucket, rt6_ex,
2127 spin_unlock(&rt6_exception_lock);
2128 rcu_read_unlock_bh();
2131 struct fib6_nh_age_excptn_arg {
2132 struct fib6_gc_args *gc_args;
2136 static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2138 struct fib6_nh_age_excptn_arg *arg = _arg;
2140 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2144 void rt6_age_exceptions(struct fib6_info *f6i,
2145 struct fib6_gc_args *gc_args,
2149 struct fib6_nh_age_excptn_arg arg = {
2154 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2157 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2161 /* must be called with rcu lock held */
2162 int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2163 struct flowi6 *fl6, struct fib6_result *res, int strict)
2165 struct fib6_node *fn, *saved_fn;
2167 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2170 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2174 rt6_select(net, fn, oif, res, strict);
2175 if (res->f6i == net->ipv6.fib6_null_entry) {
2176 fn = fib6_backtrack(fn, &fl6->saddr);
2178 goto redo_rt6_select;
2179 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2180 /* also consider unreachable route */
2181 strict &= ~RT6_LOOKUP_F_REACHABLE;
2183 goto redo_rt6_select;
2187 trace_fib6_table_lookup(net, res, table, fl6);
2192 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2193 int oif, struct flowi6 *fl6,
2194 const struct sk_buff *skb, int flags)
2196 struct fib6_result res = {};
2197 struct rt6_info *rt = NULL;
2200 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2201 !rcu_read_lock_held());
2203 strict |= flags & RT6_LOOKUP_F_IFACE;
2204 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2205 if (net->ipv6.devconf_all->forwarding == 0)
2206 strict |= RT6_LOOKUP_F_REACHABLE;
2210 fib6_table_lookup(net, table, oif, fl6, &res, strict);
2211 if (res.f6i == net->ipv6.fib6_null_entry)
2214 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2216 /*Search through exception table */
2217 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2220 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2221 !res.nh->fib_nh_gw_family)) {
2222 /* Create a RTF_CACHE clone which will not be
2223 * owned by the fib6 tree. It is for the special case where
2224 * the daddr in the skb during the neighbor look-up is different
2225 * from the fl6->daddr used to look-up route here.
2227 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2230 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2231 * As rt6_uncached_list_add() does not consume refcnt,
2232 * this refcnt is always returned to the caller even
2233 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2235 rt6_uncached_list_add(rt);
2236 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
2242 /* Get a percpu copy */
2244 rt = rt6_get_pcpu_route(&res);
2247 rt = rt6_make_pcpu_route(net, &res);
2253 rt = net->ipv6.ip6_null_entry;
2254 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2255 ip6_hold_safe(net, &rt);
2260 EXPORT_SYMBOL_GPL(ip6_pol_route);
2262 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
2263 struct fib6_table *table,
2265 const struct sk_buff *skb,
2268 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2271 struct dst_entry *ip6_route_input_lookup(struct net *net,
2272 struct net_device *dev,
2274 const struct sk_buff *skb,
2277 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2278 flags |= RT6_LOOKUP_F_IFACE;
2280 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2282 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2284 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2285 struct flow_keys *keys,
2286 struct flow_keys *flkeys)
2288 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2289 const struct ipv6hdr *key_iph = outer_iph;
2290 struct flow_keys *_flkeys = flkeys;
2291 const struct ipv6hdr *inner_iph;
2292 const struct icmp6hdr *icmph;
2293 struct ipv6hdr _inner_iph;
2294 struct icmp6hdr _icmph;
2296 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2299 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2300 sizeof(_icmph), &_icmph);
2304 if (!icmpv6_is_err(icmph->icmp6_type))
2307 inner_iph = skb_header_pointer(skb,
2308 skb_transport_offset(skb) + sizeof(*icmph),
2309 sizeof(_inner_iph), &_inner_iph);
2313 key_iph = inner_iph;
2317 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2318 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2319 keys->tags.flow_label = _flkeys->tags.flow_label;
2320 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2322 keys->addrs.v6addrs.src = key_iph->saddr;
2323 keys->addrs.v6addrs.dst = key_iph->daddr;
2324 keys->tags.flow_label = ip6_flowlabel(key_iph);
2325 keys->basic.ip_proto = key_iph->nexthdr;
2329 static u32 rt6_multipath_custom_hash_outer(const struct net *net,
2330 const struct sk_buff *skb,
2333 u32 hash_fields = ip6_multipath_hash_fields(net);
2334 struct flow_keys keys, hash_keys;
2336 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2339 memset(&hash_keys, 0, sizeof(hash_keys));
2340 skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_ENCAP);
2342 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2343 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2344 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2345 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2346 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2347 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2348 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2349 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2350 hash_keys.tags.flow_label = keys.tags.flow_label;
2351 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2352 hash_keys.ports.src = keys.ports.src;
2353 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2354 hash_keys.ports.dst = keys.ports.dst;
2356 *p_has_inner = !!(keys.control.flags & FLOW_DIS_ENCAPSULATION);
2357 return flow_hash_from_keys(&hash_keys);
2360 static u32 rt6_multipath_custom_hash_inner(const struct net *net,
2361 const struct sk_buff *skb,
2364 u32 hash_fields = ip6_multipath_hash_fields(net);
2365 struct flow_keys keys, hash_keys;
2367 /* We assume the packet carries an encapsulation, but if none was
2368 * encountered during dissection of the outer flow, then there is no
2369 * point in calling the flow dissector again.
2374 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_MASK))
2377 memset(&hash_keys, 0, sizeof(hash_keys));
2378 skb_flow_dissect_flow_keys(skb, &keys, 0);
2380 if (!(keys.control.flags & FLOW_DIS_ENCAPSULATION))
2383 if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2384 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2385 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2386 hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
2387 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2388 hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst;
2389 } else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2390 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2391 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2392 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2393 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2394 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2395 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL)
2396 hash_keys.tags.flow_label = keys.tags.flow_label;
2399 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO)
2400 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2401 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT)
2402 hash_keys.ports.src = keys.ports.src;
2403 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
2404 hash_keys.ports.dst = keys.ports.dst;
2406 return flow_hash_from_keys(&hash_keys);
2409 static u32 rt6_multipath_custom_hash_skb(const struct net *net,
2410 const struct sk_buff *skb)
2412 u32 mhash, mhash_inner;
2413 bool has_inner = true;
2415 mhash = rt6_multipath_custom_hash_outer(net, skb, &has_inner);
2416 mhash_inner = rt6_multipath_custom_hash_inner(net, skb, has_inner);
2418 return jhash_2words(mhash, mhash_inner, 0);
2421 static u32 rt6_multipath_custom_hash_fl6(const struct net *net,
2422 const struct flowi6 *fl6)
2424 u32 hash_fields = ip6_multipath_hash_fields(net);
2425 struct flow_keys hash_keys;
2427 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2430 memset(&hash_keys, 0, sizeof(hash_keys));
2431 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2432 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2433 hash_keys.addrs.v6addrs.src = fl6->saddr;
2434 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2435 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2436 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2437 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2438 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2439 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2440 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2441 hash_keys.ports.src = fl6->fl6_sport;
2442 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2443 hash_keys.ports.dst = fl6->fl6_dport;
2445 return flow_hash_from_keys(&hash_keys);
2448 /* if skb is set it will be used and fl6 can be NULL */
2449 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2450 const struct sk_buff *skb, struct flow_keys *flkeys)
2452 struct flow_keys hash_keys;
2455 switch (ip6_multipath_hash_policy(net)) {
2457 memset(&hash_keys, 0, sizeof(hash_keys));
2458 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2460 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2462 hash_keys.addrs.v6addrs.src = fl6->saddr;
2463 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2464 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2465 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2467 mhash = flow_hash_from_keys(&hash_keys);
2471 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2472 struct flow_keys keys;
2474 /* short-circuit if we already have L4 hash present */
2476 return skb_get_hash_raw(skb) >> 1;
2478 memset(&hash_keys, 0, sizeof(hash_keys));
2481 skb_flow_dissect_flow_keys(skb, &keys, flag);
2484 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2485 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2486 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2487 hash_keys.ports.src = flkeys->ports.src;
2488 hash_keys.ports.dst = flkeys->ports.dst;
2489 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2491 memset(&hash_keys, 0, sizeof(hash_keys));
2492 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2493 hash_keys.addrs.v6addrs.src = fl6->saddr;
2494 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2495 hash_keys.ports.src = fl6->fl6_sport;
2496 hash_keys.ports.dst = fl6->fl6_dport;
2497 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2499 mhash = flow_hash_from_keys(&hash_keys);
2502 memset(&hash_keys, 0, sizeof(hash_keys));
2503 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2505 struct flow_keys keys;
2508 skb_flow_dissect_flow_keys(skb, &keys, 0);
2512 /* Inner can be v4 or v6 */
2513 if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2514 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2515 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2516 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2517 } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2518 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2519 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2520 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2521 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2522 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2524 /* Same as case 0 */
2525 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2526 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2529 /* Same as case 0 */
2530 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2531 hash_keys.addrs.v6addrs.src = fl6->saddr;
2532 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2533 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2534 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2536 mhash = flow_hash_from_keys(&hash_keys);
2540 mhash = rt6_multipath_custom_hash_skb(net, skb);
2542 mhash = rt6_multipath_custom_hash_fl6(net, fl6);
2549 /* Called with rcu held */
2550 void ip6_route_input(struct sk_buff *skb)
2552 const struct ipv6hdr *iph = ipv6_hdr(skb);
2553 struct net *net = dev_net(skb->dev);
2554 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2555 struct ip_tunnel_info *tun_info;
2556 struct flowi6 fl6 = {
2557 .flowi6_iif = skb->dev->ifindex,
2558 .daddr = iph->daddr,
2559 .saddr = iph->saddr,
2560 .flowlabel = ip6_flowinfo(iph),
2561 .flowi6_mark = skb->mark,
2562 .flowi6_proto = iph->nexthdr,
2564 struct flow_keys *flkeys = NULL, _flkeys;
2566 tun_info = skb_tunnel_info(skb);
2567 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2568 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2570 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2573 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2574 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2576 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2580 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
2581 struct fib6_table *table,
2583 const struct sk_buff *skb,
2586 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2589 struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2590 const struct sock *sk,
2591 struct flowi6 *fl6, int flags)
2595 if (ipv6_addr_type(&fl6->daddr) &
2596 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2597 struct dst_entry *dst;
2599 /* This function does not take refcnt on the dst */
2600 dst = l3mdev_link_scope_lookup(net, fl6);
2605 fl6->flowi6_iif = LOOPBACK_IFINDEX;
2607 flags |= RT6_LOOKUP_F_DST_NOREF;
2608 any_src = ipv6_addr_any(&fl6->saddr);
2609 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2610 (fl6->flowi6_oif && any_src))
2611 flags |= RT6_LOOKUP_F_IFACE;
2614 flags |= RT6_LOOKUP_F_HAS_SADDR;
2616 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
2618 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2620 EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2622 struct dst_entry *ip6_route_output_flags(struct net *net,
2623 const struct sock *sk,
2627 struct dst_entry *dst;
2628 struct rt6_info *rt6;
2631 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2632 rt6 = (struct rt6_info *)dst;
2633 /* For dst cached in uncached_list, refcnt is already taken. */
2634 if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2635 dst = &net->ipv6.ip6_null_entry->dst;
2642 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2644 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2646 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2647 struct net_device *loopback_dev = net->loopback_dev;
2648 struct dst_entry *new = NULL;
2650 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
2651 DST_OBSOLETE_DEAD, 0);
2654 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2658 new->input = dst_discard;
2659 new->output = dst_discard_out;
2661 dst_copy_metrics(new, &ort->dst);
2663 rt->rt6i_idev = in6_dev_get(loopback_dev);
2664 rt->rt6i_gateway = ort->rt6i_gateway;
2665 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2667 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2668 #ifdef CONFIG_IPV6_SUBTREES
2669 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2673 dst_release(dst_orig);
2674 return new ? new : ERR_PTR(-ENOMEM);
2678 * Destination cache support functions
2681 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2685 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2688 if (fib6_check_expired(f6i))
2694 static struct dst_entry *rt6_check(struct rt6_info *rt,
2695 struct fib6_info *from,
2700 if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2701 rt_cookie != cookie)
2704 if (rt6_check_expired(rt))
2710 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2711 struct fib6_info *from,
2714 if (!__rt6_check_expired(rt) &&
2715 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2716 fib6_check(from, cookie))
2722 INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2725 struct dst_entry *dst_ret;
2726 struct fib6_info *from;
2727 struct rt6_info *rt;
2729 rt = container_of(dst, struct rt6_info, dst);
2732 return rt6_is_valid(rt) ? dst : NULL;
2736 /* All IPV6 dsts are created with ->obsolete set to the value
2737 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2738 * into this function always.
2741 from = rcu_dereference(rt->from);
2743 if (from && (rt->rt6i_flags & RTF_PCPU ||
2744 unlikely(!list_empty(&rt->rt6i_uncached))))
2745 dst_ret = rt6_dst_from_check(rt, from, cookie);
2747 dst_ret = rt6_check(rt, from, cookie);
2753 EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
2755 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2757 struct rt6_info *rt = (struct rt6_info *) dst;
2760 if (rt->rt6i_flags & RTF_CACHE) {
2762 if (rt6_check_expired(rt)) {
2763 rt6_remove_exception_rt(rt);
2775 static void ip6_link_failure(struct sk_buff *skb)
2777 struct rt6_info *rt;
2779 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2781 rt = (struct rt6_info *) skb_dst(skb);
2784 if (rt->rt6i_flags & RTF_CACHE) {
2785 rt6_remove_exception_rt(rt);
2787 struct fib6_info *from;
2788 struct fib6_node *fn;
2790 from = rcu_dereference(rt->from);
2792 fn = rcu_dereference(from->fib6_node);
2793 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2801 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2803 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2804 struct fib6_info *from;
2807 from = rcu_dereference(rt0->from);
2809 rt0->dst.expires = from->expires;
2813 dst_set_expires(&rt0->dst, timeout);
2814 rt0->rt6i_flags |= RTF_EXPIRES;
2817 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2819 struct net *net = dev_net(rt->dst.dev);
2821 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2822 rt->rt6i_flags |= RTF_MODIFIED;
2823 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2826 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2828 return !(rt->rt6i_flags & RTF_CACHE) &&
2829 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2832 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2833 const struct ipv6hdr *iph, u32 mtu,
2836 const struct in6_addr *daddr, *saddr;
2837 struct rt6_info *rt6 = (struct rt6_info *)dst;
2839 /* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2840 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2841 * [see also comment in rt6_mtu_change_route()]
2845 daddr = &iph->daddr;
2846 saddr = &iph->saddr;
2848 daddr = &sk->sk_v6_daddr;
2849 saddr = &inet6_sk(sk)->saddr;
2856 dst_confirm_neigh(dst, daddr);
2858 if (mtu < IPV6_MIN_MTU)
2860 if (mtu >= dst_mtu(dst))
2863 if (!rt6_cache_allowed_for_pmtu(rt6)) {
2864 rt6_do_update_pmtu(rt6, mtu);
2865 /* update rt6_ex->stamp for cache */
2866 if (rt6->rt6i_flags & RTF_CACHE)
2867 rt6_update_exception_stamp_rt(rt6);
2869 struct fib6_result res = {};
2870 struct rt6_info *nrt6;
2873 res.f6i = rcu_dereference(rt6->from);
2877 res.fib6_flags = res.f6i->fib6_flags;
2878 res.fib6_type = res.f6i->fib6_type;
2881 struct fib6_nh_match_arg arg = {
2883 .gw = &rt6->rt6i_gateway,
2886 nexthop_for_each_fib6_nh(res.f6i->nh,
2887 fib6_nh_find_match, &arg);
2889 /* fib6_info uses a nexthop that does not have fib6_nh
2890 * using the dst->dev + gw. Should be impossible.
2897 res.nh = res.f6i->fib6_nh;
2900 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2902 rt6_do_update_pmtu(nrt6, mtu);
2903 if (rt6_insert_exception(nrt6, &res))
2904 dst_release_immediate(&nrt6->dst);
2911 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2912 struct sk_buff *skb, u32 mtu,
2915 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2919 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2920 int oif, u32 mark, kuid_t uid)
2922 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2923 struct dst_entry *dst;
2924 struct flowi6 fl6 = {
2926 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2927 .daddr = iph->daddr,
2928 .saddr = iph->saddr,
2929 .flowlabel = ip6_flowinfo(iph),
2933 dst = ip6_route_output(net, NULL, &fl6);
2935 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
2938 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2940 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2942 int oif = sk->sk_bound_dev_if;
2943 struct dst_entry *dst;
2945 if (!oif && skb->dev)
2946 oif = l3mdev_master_ifindex(skb->dev);
2948 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
2950 dst = __sk_dst_get(sk);
2951 if (!dst || !dst->obsolete ||
2952 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2956 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2957 ip6_datagram_dst_update(sk, false);
2960 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2962 void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2963 const struct flowi6 *fl6)
2965 #ifdef CONFIG_IPV6_SUBTREES
2966 struct ipv6_pinfo *np = inet6_sk(sk);
2969 ip6_dst_store(sk, dst,
2970 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2971 &sk->sk_v6_daddr : NULL,
2972 #ifdef CONFIG_IPV6_SUBTREES
2973 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2979 static bool ip6_redirect_nh_match(const struct fib6_result *res,
2981 const struct in6_addr *gw,
2982 struct rt6_info **ret)
2984 const struct fib6_nh *nh = res->nh;
2986 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2987 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2990 /* rt_cache's gateway might be different from its 'parent'
2991 * in the case of an ip redirect.
2992 * So we keep searching in the exception table if the gateway
2995 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2996 struct rt6_info *rt_cache;
2998 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
3000 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
3009 struct fib6_nh_rd_arg {
3010 struct fib6_result *res;
3012 const struct in6_addr *gw;
3013 struct rt6_info **ret;
3016 static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
3018 struct fib6_nh_rd_arg *arg = _arg;
3021 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
3024 /* Handle redirects */
3025 struct ip6rd_flowi {
3027 struct in6_addr gateway;
3030 INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
3031 struct fib6_table *table,
3033 const struct sk_buff *skb,
3036 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
3037 struct rt6_info *ret = NULL;
3038 struct fib6_result res = {};
3039 struct fib6_nh_rd_arg arg = {
3042 .gw = &rdfl->gateway,
3045 struct fib6_info *rt;
3046 struct fib6_node *fn;
3048 /* l3mdev_update_flow overrides oif if the device is enslaved; in
3049 * this case we must match on the real ingress device, so reset it
3051 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
3052 fl6->flowi6_oif = skb->dev->ifindex;
3054 /* Get the "current" route for this destination and
3055 * check if the redirect has come from appropriate router.
3057 * RFC 4861 specifies that redirects should only be
3058 * accepted if they come from the nexthop to the target.
3059 * Due to the way the routes are chosen, this notion
3060 * is a bit fuzzy and one might need to check all possible
3065 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
3067 for_each_fib6_node_rt_rcu(fn) {
3069 if (fib6_check_expired(rt))
3071 if (rt->fib6_flags & RTF_REJECT)
3073 if (unlikely(rt->nh)) {
3074 if (nexthop_is_blackhole(rt->nh))
3076 /* on match, res->nh is filled in and potentially ret */
3077 if (nexthop_for_each_fib6_nh(rt->nh,
3078 fib6_nh_redirect_match,
3082 res.nh = rt->fib6_nh;
3083 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
3090 rt = net->ipv6.fib6_null_entry;
3091 else if (rt->fib6_flags & RTF_REJECT) {
3092 ret = net->ipv6.ip6_null_entry;
3096 if (rt == net->ipv6.fib6_null_entry) {
3097 fn = fib6_backtrack(fn, &fl6->saddr);
3103 res.nh = rt->fib6_nh;
3106 ip6_hold_safe(net, &ret);
3108 res.fib6_flags = res.f6i->fib6_flags;
3109 res.fib6_type = res.f6i->fib6_type;
3110 ret = ip6_create_rt_rcu(&res);
3115 trace_fib6_table_lookup(net, &res, table, fl6);
3119 static struct dst_entry *ip6_route_redirect(struct net *net,
3120 const struct flowi6 *fl6,
3121 const struct sk_buff *skb,
3122 const struct in6_addr *gateway)
3124 int flags = RT6_LOOKUP_F_HAS_SADDR;
3125 struct ip6rd_flowi rdfl;
3128 rdfl.gateway = *gateway;
3130 return fib6_rule_lookup(net, &rdfl.fl6, skb,
3131 flags, __ip6_route_redirect);
3134 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3137 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3138 struct dst_entry *dst;
3139 struct flowi6 fl6 = {
3140 .flowi6_iif = LOOPBACK_IFINDEX,
3142 .flowi6_mark = mark,
3143 .daddr = iph->daddr,
3144 .saddr = iph->saddr,
3145 .flowlabel = ip6_flowinfo(iph),
3149 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3150 rt6_do_redirect(dst, NULL, skb);
3153 EXPORT_SYMBOL_GPL(ip6_redirect);
3155 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
3157 const struct ipv6hdr *iph = ipv6_hdr(skb);
3158 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3159 struct dst_entry *dst;
3160 struct flowi6 fl6 = {
3161 .flowi6_iif = LOOPBACK_IFINDEX,
3164 .saddr = iph->daddr,
3165 .flowi6_uid = sock_net_uid(net, NULL),
3168 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3169 rt6_do_redirect(dst, NULL, skb);
3173 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3175 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
3178 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3180 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3182 struct net_device *dev = dst->dev;
3183 unsigned int mtu = dst_mtu(dst);
3184 struct net *net = dev_net(dev);
3186 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3188 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3189 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3192 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3193 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3194 * IPV6_MAXPLEN is also valid and means: "any MSS,
3195 * rely only on pmtu discovery"
3197 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3202 INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
3204 return ip6_dst_mtu_maybe_forward(dst, false);
3206 EXPORT_INDIRECT_CALLABLE(ip6_mtu);
3209 * 1. mtu on route is locked - use it
3210 * 2. mtu from nexthop exception
3211 * 3. mtu from egress device
3213 * based on ip6_dst_mtu_forward and exception logic of
3214 * rt6_find_cached_rt; called with rcu_read_lock
3216 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3217 const struct in6_addr *daddr,
3218 const struct in6_addr *saddr)
3220 const struct fib6_nh *nh = res->nh;
3221 struct fib6_info *f6i = res->f6i;
3222 struct inet6_dev *idev;
3223 struct rt6_info *rt;
3226 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3227 mtu = f6i->fib6_pmtu;
3232 rt = rt6_find_cached_rt(res, daddr, saddr);
3234 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3236 struct net_device *dev = nh->fib_nh_dev;
3239 idev = __in6_dev_get(dev);
3240 if (idev && idev->cnf.mtu6 > mtu)
3241 mtu = idev->cnf.mtu6;
3244 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3246 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3249 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3252 struct dst_entry *dst;
3253 struct rt6_info *rt;
3254 struct inet6_dev *idev = in6_dev_get(dev);
3255 struct net *net = dev_net(dev);
3257 if (unlikely(!idev))
3258 return ERR_PTR(-ENODEV);
3260 rt = ip6_dst_alloc(net, dev, 0);
3261 if (unlikely(!rt)) {
3263 dst = ERR_PTR(-ENOMEM);
3267 rt->dst.input = ip6_input;
3268 rt->dst.output = ip6_output;
3269 rt->rt6i_gateway = fl6->daddr;
3270 rt->rt6i_dst.addr = fl6->daddr;
3271 rt->rt6i_dst.plen = 128;
3272 rt->rt6i_idev = idev;
3273 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3275 /* Add this dst into uncached_list so that rt6_disable_ip() can
3276 * do proper release of the net_device
3278 rt6_uncached_list_add(rt);
3279 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
3281 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3287 static int ip6_dst_gc(struct dst_ops *ops)
3289 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3290 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3291 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3292 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3293 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3294 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3297 entries = dst_entries_get_fast(ops);
3298 if (entries > rt_max_size)
3299 entries = dst_entries_get_slow(ops);
3301 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
3302 entries <= rt_max_size)
3305 net->ipv6.ip6_rt_gc_expire++;
3306 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
3307 entries = dst_entries_get_slow(ops);
3308 if (entries < ops->gc_thresh)
3309 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
3311 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
3312 return entries > rt_max_size;
3315 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3316 const struct in6_addr *gw_addr, u32 tbid,
3317 int flags, struct fib6_result *res)
3319 struct flowi6 fl6 = {
3320 .flowi6_oif = cfg->fc_ifindex,
3322 .saddr = cfg->fc_prefsrc,
3324 struct fib6_table *table;
3327 table = fib6_get_table(net, tbid);
3331 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3332 flags |= RT6_LOOKUP_F_HAS_SADDR;
3334 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3336 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3337 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3338 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3339 cfg->fc_ifindex != 0, NULL, flags);
3344 static int ip6_route_check_nh_onlink(struct net *net,
3345 struct fib6_config *cfg,
3346 const struct net_device *dev,
3347 struct netlink_ext_ack *extack)
3349 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3350 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3351 struct fib6_result res = {};
3354 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3355 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3356 /* ignore match if it is the default route */
3357 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3358 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3359 NL_SET_ERR_MSG(extack,
3360 "Nexthop has invalid gateway or device mismatch");
3367 static int ip6_route_check_nh(struct net *net,
3368 struct fib6_config *cfg,
3369 struct net_device **_dev,
3370 struct inet6_dev **idev)
3372 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3373 struct net_device *dev = _dev ? *_dev : NULL;
3374 int flags = RT6_LOOKUP_F_IFACE;
3375 struct fib6_result res = {};
3376 int err = -EHOSTUNREACH;
3378 if (cfg->fc_table) {
3379 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3380 cfg->fc_table, flags, &res);
3381 /* gw_addr can not require a gateway or resolve to a reject
3382 * route. If a device is given, it must match the result.
3384 if (err || res.fib6_flags & RTF_REJECT ||
3385 res.nh->fib_nh_gw_family ||
3386 (dev && dev != res.nh->fib_nh_dev))
3387 err = -EHOSTUNREACH;
3391 struct flowi6 fl6 = {
3392 .flowi6_oif = cfg->fc_ifindex,
3396 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3397 if (err || res.fib6_flags & RTF_REJECT ||
3398 res.nh->fib_nh_gw_family)
3399 err = -EHOSTUNREACH;
3404 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3405 cfg->fc_ifindex != 0, NULL, flags);
3410 if (dev != res.nh->fib_nh_dev)
3411 err = -EHOSTUNREACH;
3413 *_dev = dev = res.nh->fib_nh_dev;
3415 *idev = in6_dev_get(dev);
3421 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3422 struct net_device **_dev, struct inet6_dev **idev,
3423 struct netlink_ext_ack *extack)
3425 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3426 int gwa_type = ipv6_addr_type(gw_addr);
3427 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3428 const struct net_device *dev = *_dev;
3429 bool need_addr_check = !dev;
3432 /* if gw_addr is local we will fail to detect this in case
3433 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3434 * will return already-added prefix route via interface that
3435 * prefix route was assigned to, which might be non-loopback.
3438 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3439 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3443 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3444 /* IPv6 strictly inhibits using not link-local
3445 * addresses as nexthop address.
3446 * Otherwise, router will not able to send redirects.
3447 * It is very good, but in some (rare!) circumstances
3448 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3449 * some exceptions. --ANK
3450 * We allow IPv4-mapped nexthops to support RFC4798-type
3453 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3454 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3460 if (cfg->fc_flags & RTNH_F_ONLINK)
3461 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3463 err = ip6_route_check_nh(net, cfg, _dev, idev);
3471 /* reload in case device was changed */
3476 NL_SET_ERR_MSG(extack, "Egress device not specified");
3478 } else if (dev->flags & IFF_LOOPBACK) {
3479 NL_SET_ERR_MSG(extack,
3480 "Egress device can not be loopback device for this route");
3484 /* if we did not check gw_addr above, do so now that the
3485 * egress device has been resolved.
3487 if (need_addr_check &&
3488 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3489 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3498 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3500 if ((flags & RTF_REJECT) ||
3501 (dev && (dev->flags & IFF_LOOPBACK) &&
3502 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3503 !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3509 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3510 struct fib6_config *cfg, gfp_t gfp_flags,
3511 struct netlink_ext_ack *extack)
3513 struct net_device *dev = NULL;
3514 struct inet6_dev *idev = NULL;
3518 fib6_nh->fib_nh_family = AF_INET6;
3519 #ifdef CONFIG_IPV6_ROUTER_PREF
3520 fib6_nh->last_probe = jiffies;
3522 if (cfg->fc_is_fdb) {
3523 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3524 fib6_nh->fib_nh_gw_family = AF_INET6;
3529 if (cfg->fc_ifindex) {
3530 dev = dev_get_by_index(net, cfg->fc_ifindex);
3533 idev = in6_dev_get(dev);
3538 if (cfg->fc_flags & RTNH_F_ONLINK) {
3540 NL_SET_ERR_MSG(extack,
3541 "Nexthop device required for onlink");
3545 if (!(dev->flags & IFF_UP)) {
3546 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3551 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3554 fib6_nh->fib_nh_weight = 1;
3556 /* We cannot add true routes via loopback here,
3557 * they would result in kernel looping; promote them to reject routes
3559 addr_type = ipv6_addr_type(&cfg->fc_dst);
3560 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3561 /* hold loopback dev/idev if we haven't done so. */
3562 if (dev != net->loopback_dev) {
3567 dev = net->loopback_dev;
3569 idev = in6_dev_get(dev);
3578 if (cfg->fc_flags & RTF_GATEWAY) {
3579 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3583 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3584 fib6_nh->fib_nh_gw_family = AF_INET6;
3591 if (idev->cnf.disable_ipv6) {
3592 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3597 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3598 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3603 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3604 !netif_carrier_ok(dev))
3605 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3607 err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3608 cfg->fc_encap_type, cfg, gfp_flags, extack);
3613 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3614 if (!fib6_nh->rt6i_pcpu) {
3619 fib6_nh->fib_nh_dev = dev;
3620 fib6_nh->fib_nh_oif = dev->ifindex;
3627 lwtstate_put(fib6_nh->fib_nh_lws);
3628 fib6_nh->fib_nh_lws = NULL;
3635 void fib6_nh_release(struct fib6_nh *fib6_nh)
3637 struct rt6_exception_bucket *bucket;
3641 fib6_nh_flush_exceptions(fib6_nh, NULL);
3642 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3644 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3650 if (fib6_nh->rt6i_pcpu) {
3653 for_each_possible_cpu(cpu) {
3654 struct rt6_info **ppcpu_rt;
3655 struct rt6_info *pcpu_rt;
3657 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3658 pcpu_rt = *ppcpu_rt;
3660 dst_dev_put(&pcpu_rt->dst);
3661 dst_release(&pcpu_rt->dst);
3666 free_percpu(fib6_nh->rt6i_pcpu);
3669 fib_nh_common_release(&fib6_nh->nh_common);
3672 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3674 struct netlink_ext_ack *extack)
3676 struct net *net = cfg->fc_nlinfo.nl_net;
3677 struct fib6_info *rt = NULL;
3678 struct nexthop *nh = NULL;
3679 struct fib6_table *table;
3680 struct fib6_nh *fib6_nh;
3684 /* RTF_PCPU is an internal flag; can not be set by userspace */
3685 if (cfg->fc_flags & RTF_PCPU) {
3686 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3690 /* RTF_CACHE is an internal flag; can not be set by userspace */
3691 if (cfg->fc_flags & RTF_CACHE) {
3692 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3696 if (cfg->fc_type > RTN_MAX) {
3697 NL_SET_ERR_MSG(extack, "Invalid route type");
3701 if (cfg->fc_dst_len > 128) {
3702 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3705 if (cfg->fc_src_len > 128) {
3706 NL_SET_ERR_MSG(extack, "Invalid source address length");
3709 #ifndef CONFIG_IPV6_SUBTREES
3710 if (cfg->fc_src_len) {
3711 NL_SET_ERR_MSG(extack,
3712 "Specifying source address requires IPV6_SUBTREES to be enabled");
3716 if (cfg->fc_nh_id) {
3717 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3719 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3722 err = fib6_check_nexthop(nh, cfg, extack);
3728 if (cfg->fc_nlinfo.nlh &&
3729 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3730 table = fib6_get_table(net, cfg->fc_table);
3732 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3733 table = fib6_new_table(net, cfg->fc_table);
3736 table = fib6_new_table(net, cfg->fc_table);
3743 rt = fib6_info_alloc(gfp_flags, !nh);
3747 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3749 if (IS_ERR(rt->fib6_metrics)) {
3750 err = PTR_ERR(rt->fib6_metrics);
3751 /* Do not leave garbage there. */
3752 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3756 if (cfg->fc_flags & RTF_ADDRCONF)
3757 rt->dst_nocount = true;
3759 if (cfg->fc_flags & RTF_EXPIRES)
3760 fib6_set_expires(rt, jiffies +
3761 clock_t_to_jiffies(cfg->fc_expires));
3763 fib6_clean_expires(rt);
3765 if (cfg->fc_protocol == RTPROT_UNSPEC)
3766 cfg->fc_protocol = RTPROT_BOOT;
3767 rt->fib6_protocol = cfg->fc_protocol;
3769 rt->fib6_table = table;
3770 rt->fib6_metric = cfg->fc_metric;
3771 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3772 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3774 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3775 rt->fib6_dst.plen = cfg->fc_dst_len;
3777 #ifdef CONFIG_IPV6_SUBTREES
3778 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3779 rt->fib6_src.plen = cfg->fc_src_len;
3782 if (rt->fib6_src.plen) {
3783 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3786 if (!nexthop_get(nh)) {
3787 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3791 fib6_nh = nexthop_fib6_nh(rt->nh);
3793 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3797 fib6_nh = rt->fib6_nh;
3799 /* We cannot add true routes via loopback here, they would
3800 * result in kernel looping; promote them to reject routes
3802 addr_type = ipv6_addr_type(&cfg->fc_dst);
3803 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3805 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3808 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3809 struct net_device *dev = fib6_nh->fib_nh_dev;
3811 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3812 NL_SET_ERR_MSG(extack, "Invalid source address");
3816 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3817 rt->fib6_prefsrc.plen = 128;
3819 rt->fib6_prefsrc.plen = 0;
3823 fib6_info_release(rt);
3824 return ERR_PTR(err);
3826 ip_fib_metrics_put(rt->fib6_metrics);
3828 return ERR_PTR(err);
3831 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3832 struct netlink_ext_ack *extack)
3834 struct fib6_info *rt;
3837 rt = ip6_route_info_create(cfg, gfp_flags, extack);
3841 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3842 fib6_info_release(rt);
3847 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3849 struct net *net = info->nl_net;
3850 struct fib6_table *table;
3853 if (rt == net->ipv6.fib6_null_entry) {
3858 table = rt->fib6_table;
3859 spin_lock_bh(&table->tb6_lock);
3860 err = fib6_del(rt, info);
3861 spin_unlock_bh(&table->tb6_lock);
3864 fib6_info_release(rt);
3868 int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
3870 struct nl_info info = {
3872 .skip_notify = skip_notify
3875 return __ip6_del_rt(rt, &info);
3878 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3880 struct nl_info *info = &cfg->fc_nlinfo;
3881 struct net *net = info->nl_net;
3882 struct sk_buff *skb = NULL;
3883 struct fib6_table *table;
3886 if (rt == net->ipv6.fib6_null_entry)
3888 table = rt->fib6_table;
3889 spin_lock_bh(&table->tb6_lock);
3891 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3892 struct fib6_info *sibling, *next_sibling;
3893 struct fib6_node *fn;
3895 /* prefer to send a single notification with all hops */
3896 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3898 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3900 if (rt6_fill_node(net, skb, rt, NULL,
3901 NULL, NULL, 0, RTM_DELROUTE,
3902 info->portid, seq, 0) < 0) {
3906 info->skip_notify = 1;
3909 /* 'rt' points to the first sibling route. If it is not the
3910 * leaf, then we do not need to send a notification. Otherwise,
3911 * we need to check if the last sibling has a next route or not
3912 * and emit a replace or delete notification, respectively.
3914 info->skip_notify_kernel = 1;
3915 fn = rcu_dereference_protected(rt->fib6_node,
3916 lockdep_is_held(&table->tb6_lock));
3917 if (rcu_access_pointer(fn->leaf) == rt) {
3918 struct fib6_info *last_sibling, *replace_rt;
3920 last_sibling = list_last_entry(&rt->fib6_siblings,
3923 replace_rt = rcu_dereference_protected(
3924 last_sibling->fib6_next,
3925 lockdep_is_held(&table->tb6_lock));
3927 call_fib6_entry_notifiers_replace(net,
3930 call_fib6_multipath_entry_notifiers(net,
3931 FIB_EVENT_ENTRY_DEL,
3932 rt, rt->fib6_nsiblings,
3935 list_for_each_entry_safe(sibling, next_sibling,
3938 err = fib6_del(sibling, info);
3944 err = fib6_del(rt, info);
3946 spin_unlock_bh(&table->tb6_lock);
3948 fib6_info_release(rt);
3951 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3952 info->nlh, gfp_any());
3957 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3961 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3964 if (cfg->fc_flags & RTF_GATEWAY &&
3965 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3968 rc = rt6_remove_exception_rt(rt);
3973 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3976 struct fib6_result res = {
3980 struct rt6_info *rt_cache;
3982 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3984 return __ip6_del_cached_rt(rt_cache, cfg);
3989 struct fib6_nh_del_cached_rt_arg {
3990 struct fib6_config *cfg;
3991 struct fib6_info *f6i;
3994 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
3996 struct fib6_nh_del_cached_rt_arg *arg = _arg;
3999 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
4000 return rc != -ESRCH ? rc : 0;
4003 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
4005 struct fib6_nh_del_cached_rt_arg arg = {
4010 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
4013 static int ip6_route_del(struct fib6_config *cfg,
4014 struct netlink_ext_ack *extack)
4016 struct fib6_table *table;
4017 struct fib6_info *rt;
4018 struct fib6_node *fn;
4021 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
4023 NL_SET_ERR_MSG(extack, "FIB table does not exist");
4029 fn = fib6_locate(&table->tb6_root,
4030 &cfg->fc_dst, cfg->fc_dst_len,
4031 &cfg->fc_src, cfg->fc_src_len,
4032 !(cfg->fc_flags & RTF_CACHE));
4035 for_each_fib6_node_rt_rcu(fn) {
4038 if (rt->nh && cfg->fc_nh_id &&
4039 rt->nh->id != cfg->fc_nh_id)
4042 if (cfg->fc_flags & RTF_CACHE) {
4046 rc = ip6_del_cached_rt_nh(cfg, rt);
4047 } else if (cfg->fc_nh_id) {
4051 rc = ip6_del_cached_rt(cfg, rt, nh);
4060 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
4062 if (cfg->fc_protocol &&
4063 cfg->fc_protocol != rt->fib6_protocol)
4067 if (!fib6_info_hold_safe(rt))
4071 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4077 if (cfg->fc_ifindex &&
4079 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
4081 if (cfg->fc_flags & RTF_GATEWAY &&
4082 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
4084 if (!fib6_info_hold_safe(rt))
4088 /* if gateway was specified only delete the one hop */
4089 if (cfg->fc_flags & RTF_GATEWAY)
4090 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4092 return __ip6_del_rt_siblings(rt, cfg);
4100 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
4102 struct netevent_redirect netevent;
4103 struct rt6_info *rt, *nrt = NULL;
4104 struct fib6_result res = {};
4105 struct ndisc_options ndopts;
4106 struct inet6_dev *in6_dev;
4107 struct neighbour *neigh;
4109 int optlen, on_link;
4112 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
4113 optlen -= sizeof(*msg);
4116 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4120 msg = (struct rd_msg *)icmp6_hdr(skb);
4122 if (ipv6_addr_is_multicast(&msg->dest)) {
4123 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4128 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4130 } else if (ipv6_addr_type(&msg->target) !=
4131 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4132 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4136 in6_dev = __in6_dev_get(skb->dev);
4139 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
4143 * The IP source address of the Redirect MUST be the same as the current
4144 * first-hop router for the specified ICMP Destination Address.
4147 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4148 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4153 if (ndopts.nd_opts_tgt_lladdr) {
4154 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4157 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4162 rt = (struct rt6_info *) dst;
4163 if (rt->rt6i_flags & RTF_REJECT) {
4164 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4168 /* Redirect received -> path was valid.
4169 * Look, redirects are sent only in response to data packets,
4170 * so that this nexthop apparently is reachable. --ANK
4172 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4174 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4179 * We have finally decided to accept it.
4182 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4183 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4184 NEIGH_UPDATE_F_OVERRIDE|
4185 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4186 NEIGH_UPDATE_F_ISROUTER)),
4187 NDISC_REDIRECT, &ndopts);
4190 res.f6i = rcu_dereference(rt->from);
4195 struct fib6_nh_match_arg arg = {
4197 .gw = &rt->rt6i_gateway,
4200 nexthop_for_each_fib6_nh(res.f6i->nh,
4201 fib6_nh_find_match, &arg);
4203 /* fib6_info uses a nexthop that does not have fib6_nh
4204 * using the dst->dev. Should be impossible
4210 res.nh = res.f6i->fib6_nh;
4213 res.fib6_flags = res.f6i->fib6_flags;
4214 res.fib6_type = res.f6i->fib6_type;
4215 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4219 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4221 nrt->rt6i_flags &= ~RTF_GATEWAY;
4223 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4225 /* rt6_insert_exception() will take care of duplicated exceptions */
4226 if (rt6_insert_exception(nrt, &res)) {
4227 dst_release_immediate(&nrt->dst);
4231 netevent.old = &rt->dst;
4232 netevent.new = &nrt->dst;
4233 netevent.daddr = &msg->dest;
4234 netevent.neigh = neigh;
4235 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4239 neigh_release(neigh);
4242 #ifdef CONFIG_IPV6_ROUTE_INFO
4243 static struct fib6_info *rt6_get_route_info(struct net *net,
4244 const struct in6_addr *prefix, int prefixlen,
4245 const struct in6_addr *gwaddr,
4246 struct net_device *dev)
4248 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4249 int ifindex = dev->ifindex;
4250 struct fib6_node *fn;
4251 struct fib6_info *rt = NULL;
4252 struct fib6_table *table;
4254 table = fib6_get_table(net, tb_id);
4259 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4263 for_each_fib6_node_rt_rcu(fn) {
4264 /* these routes do not use nexthops */
4267 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4269 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4270 !rt->fib6_nh->fib_nh_gw_family)
4272 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4274 if (!fib6_info_hold_safe(rt))
4283 static struct fib6_info *rt6_add_route_info(struct net *net,
4284 const struct in6_addr *prefix, int prefixlen,
4285 const struct in6_addr *gwaddr,
4286 struct net_device *dev,
4289 struct fib6_config cfg = {
4290 .fc_metric = IP6_RT_PRIO_USER,
4291 .fc_ifindex = dev->ifindex,
4292 .fc_dst_len = prefixlen,
4293 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4294 RTF_UP | RTF_PREF(pref),
4295 .fc_protocol = RTPROT_RA,
4296 .fc_type = RTN_UNICAST,
4297 .fc_nlinfo.portid = 0,
4298 .fc_nlinfo.nlh = NULL,
4299 .fc_nlinfo.nl_net = net,
4302 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4303 cfg.fc_dst = *prefix;
4304 cfg.fc_gateway = *gwaddr;
4306 /* We should treat it as a default route if prefix length is 0. */
4308 cfg.fc_flags |= RTF_DEFAULT;
4310 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4312 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4316 struct fib6_info *rt6_get_dflt_router(struct net *net,
4317 const struct in6_addr *addr,
4318 struct net_device *dev)
4320 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4321 struct fib6_info *rt;
4322 struct fib6_table *table;
4324 table = fib6_get_table(net, tb_id);
4329 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4332 /* RA routes do not use nexthops */
4337 if (dev == nh->fib_nh_dev &&
4338 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4339 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4342 if (rt && !fib6_info_hold_safe(rt))
4348 struct fib6_info *rt6_add_dflt_router(struct net *net,
4349 const struct in6_addr *gwaddr,
4350 struct net_device *dev,
4352 u32 defrtr_usr_metric)
4354 struct fib6_config cfg = {
4355 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4356 .fc_metric = defrtr_usr_metric,
4357 .fc_ifindex = dev->ifindex,
4358 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4359 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4360 .fc_protocol = RTPROT_RA,
4361 .fc_type = RTN_UNICAST,
4362 .fc_nlinfo.portid = 0,
4363 .fc_nlinfo.nlh = NULL,
4364 .fc_nlinfo.nl_net = net,
4367 cfg.fc_gateway = *gwaddr;
4369 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4370 struct fib6_table *table;
4372 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4374 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4377 return rt6_get_dflt_router(net, gwaddr, dev);
4380 static void __rt6_purge_dflt_routers(struct net *net,
4381 struct fib6_table *table)
4383 struct fib6_info *rt;
4387 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4388 struct net_device *dev = fib6_info_nh_dev(rt);
4389 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4391 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4392 (!idev || idev->cnf.accept_ra != 2) &&
4393 fib6_info_hold_safe(rt)) {
4395 ip6_del_rt(net, rt, false);
4401 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4404 void rt6_purge_dflt_routers(struct net *net)
4406 struct fib6_table *table;
4407 struct hlist_head *head;
4412 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4413 head = &net->ipv6.fib_table_hash[h];
4414 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4415 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4416 __rt6_purge_dflt_routers(net, table);
4423 static void rtmsg_to_fib6_config(struct net *net,
4424 struct in6_rtmsg *rtmsg,
4425 struct fib6_config *cfg)
4427 *cfg = (struct fib6_config){
4428 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4430 .fc_ifindex = rtmsg->rtmsg_ifindex,
4431 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
4432 .fc_expires = rtmsg->rtmsg_info,
4433 .fc_dst_len = rtmsg->rtmsg_dst_len,
4434 .fc_src_len = rtmsg->rtmsg_src_len,
4435 .fc_flags = rtmsg->rtmsg_flags,
4436 .fc_type = rtmsg->rtmsg_type,
4438 .fc_nlinfo.nl_net = net,
4440 .fc_dst = rtmsg->rtmsg_dst,
4441 .fc_src = rtmsg->rtmsg_src,
4442 .fc_gateway = rtmsg->rtmsg_gateway,
4446 int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
4448 struct fib6_config cfg;
4451 if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4453 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4456 rtmsg_to_fib6_config(net, rtmsg, &cfg);
4461 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4464 err = ip6_route_del(&cfg, NULL);
4472 * Drop the packet on the floor
4475 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4477 struct dst_entry *dst = skb_dst(skb);
4478 struct net *net = dev_net(dst->dev);
4479 struct inet6_dev *idev;
4482 if (netif_is_l3_master(skb->dev) &&
4483 dst->dev == net->loopback_dev)
4484 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4486 idev = ip6_dst_idev(dst);
4488 switch (ipstats_mib_noroutes) {
4489 case IPSTATS_MIB_INNOROUTES:
4490 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4491 if (type == IPV6_ADDR_ANY) {
4492 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4496 case IPSTATS_MIB_OUTNOROUTES:
4497 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4501 /* Start over by dropping the dst for l3mdev case */
4502 if (netif_is_l3_master(skb->dev))
4505 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4510 static int ip6_pkt_discard(struct sk_buff *skb)
4512 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4515 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4517 skb->dev = skb_dst(skb)->dev;
4518 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4521 static int ip6_pkt_prohibit(struct sk_buff *skb)
4523 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4526 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4528 skb->dev = skb_dst(skb)->dev;
4529 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4533 * Allocate a dst for local (unicast / anycast) address.
4536 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4537 struct inet6_dev *idev,
4538 const struct in6_addr *addr,
4539 bool anycast, gfp_t gfp_flags)
4541 struct fib6_config cfg = {
4542 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4543 .fc_ifindex = idev->dev->ifindex,
4544 .fc_flags = RTF_UP | RTF_NONEXTHOP,
4547 .fc_protocol = RTPROT_KERNEL,
4548 .fc_nlinfo.nl_net = net,
4549 .fc_ignore_dev_down = true,
4551 struct fib6_info *f6i;
4554 cfg.fc_type = RTN_ANYCAST;
4555 cfg.fc_flags |= RTF_ANYCAST;
4557 cfg.fc_type = RTN_LOCAL;
4558 cfg.fc_flags |= RTF_LOCAL;
4561 f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
4563 f6i->dst_nocount = true;
4567 /* remove deleted ip from prefsrc entries */
4568 struct arg_dev_net_ip {
4569 struct net_device *dev;
4571 struct in6_addr *addr;
4574 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4576 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4577 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4578 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4581 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
4582 rt != net->ipv6.fib6_null_entry &&
4583 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
4584 spin_lock_bh(&rt6_exception_lock);
4585 /* remove prefsrc entry */
4586 rt->fib6_prefsrc.plen = 0;
4587 spin_unlock_bh(&rt6_exception_lock);
4592 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4594 struct net *net = dev_net(ifp->idev->dev);
4595 struct arg_dev_net_ip adni = {
4596 .dev = ifp->idev->dev,
4600 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4603 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
4605 /* Remove routers and update dst entries when gateway turn into host. */
4606 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4608 struct in6_addr *gateway = (struct in6_addr *)arg;
4611 /* RA routes do not use nexthops */
4616 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4617 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4620 /* Further clean up cached routes in exception table.
4621 * This is needed because cached route may have a different
4622 * gateway than its 'parent' in the case of an ip redirect.
4624 fib6_nh_exceptions_clean_tohost(nh, gateway);
4629 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4631 fib6_clean_all(net, fib6_clean_tohost, gateway);
4634 struct arg_netdev_event {
4635 const struct net_device *dev;
4637 unsigned char nh_flags;
4638 unsigned long event;
4642 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4644 struct fib6_info *iter;
4645 struct fib6_node *fn;
4647 fn = rcu_dereference_protected(rt->fib6_node,
4648 lockdep_is_held(&rt->fib6_table->tb6_lock));
4649 iter = rcu_dereference_protected(fn->leaf,
4650 lockdep_is_held(&rt->fib6_table->tb6_lock));
4652 if (iter->fib6_metric == rt->fib6_metric &&
4653 rt6_qualify_for_ecmp(iter))
4655 iter = rcu_dereference_protected(iter->fib6_next,
4656 lockdep_is_held(&rt->fib6_table->tb6_lock));
4662 /* only called for fib entries with builtin fib6_nh */
4663 static bool rt6_is_dead(const struct fib6_info *rt)
4665 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4666 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4667 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4673 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4675 struct fib6_info *iter;
4678 if (!rt6_is_dead(rt))
4679 total += rt->fib6_nh->fib_nh_weight;
4681 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4682 if (!rt6_is_dead(iter))
4683 total += iter->fib6_nh->fib_nh_weight;
4689 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4691 int upper_bound = -1;
4693 if (!rt6_is_dead(rt)) {
4694 *weight += rt->fib6_nh->fib_nh_weight;
4695 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4698 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4701 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4703 struct fib6_info *iter;
4706 rt6_upper_bound_set(rt, &weight, total);
4708 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4709 rt6_upper_bound_set(iter, &weight, total);
4712 void rt6_multipath_rebalance(struct fib6_info *rt)
4714 struct fib6_info *first;
4717 /* In case the entire multipath route was marked for flushing,
4718 * then there is no need to rebalance upon the removal of every
4721 if (!rt->fib6_nsiblings || rt->should_flush)
4724 /* During lookup routes are evaluated in order, so we need to
4725 * make sure upper bounds are assigned from the first sibling
4728 first = rt6_multipath_first_sibling(rt);
4729 if (WARN_ON_ONCE(!first))
4732 total = rt6_multipath_total_weight(first);
4733 rt6_multipath_upper_bound_set(first, total);
4736 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4738 const struct arg_netdev_event *arg = p_arg;
4739 struct net *net = dev_net(arg->dev);
4741 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4742 rt->fib6_nh->fib_nh_dev == arg->dev) {
4743 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4744 fib6_update_sernum_upto_root(net, rt);
4745 rt6_multipath_rebalance(rt);
4751 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4753 struct arg_netdev_event arg = {
4756 .nh_flags = nh_flags,
4760 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4761 arg.nh_flags |= RTNH_F_LINKDOWN;
4763 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4766 /* only called for fib entries with inline fib6_nh */
4767 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4768 const struct net_device *dev)
4770 struct fib6_info *iter;
4772 if (rt->fib6_nh->fib_nh_dev == dev)
4774 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4775 if (iter->fib6_nh->fib_nh_dev == dev)
4781 static void rt6_multipath_flush(struct fib6_info *rt)
4783 struct fib6_info *iter;
4785 rt->should_flush = 1;
4786 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4787 iter->should_flush = 1;
4790 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4791 const struct net_device *down_dev)
4793 struct fib6_info *iter;
4794 unsigned int dead = 0;
4796 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4797 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4799 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4800 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4801 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4807 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4808 const struct net_device *dev,
4809 unsigned char nh_flags)
4811 struct fib6_info *iter;
4813 if (rt->fib6_nh->fib_nh_dev == dev)
4814 rt->fib6_nh->fib_nh_flags |= nh_flags;
4815 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4816 if (iter->fib6_nh->fib_nh_dev == dev)
4817 iter->fib6_nh->fib_nh_flags |= nh_flags;
4820 /* called with write lock held for table with rt */
4821 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4823 const struct arg_netdev_event *arg = p_arg;
4824 const struct net_device *dev = arg->dev;
4825 struct net *net = dev_net(dev);
4827 if (rt == net->ipv6.fib6_null_entry || rt->nh)
4830 switch (arg->event) {
4831 case NETDEV_UNREGISTER:
4832 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4834 if (rt->should_flush)
4836 if (!rt->fib6_nsiblings)
4837 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4838 if (rt6_multipath_uses_dev(rt, dev)) {
4841 count = rt6_multipath_dead_count(rt, dev);
4842 if (rt->fib6_nsiblings + 1 == count) {
4843 rt6_multipath_flush(rt);
4846 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4848 fib6_update_sernum(net, rt);
4849 rt6_multipath_rebalance(rt);
4853 if (rt->fib6_nh->fib_nh_dev != dev ||
4854 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4856 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4857 rt6_multipath_rebalance(rt);
4864 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4866 struct arg_netdev_event arg = {
4872 struct net *net = dev_net(dev);
4874 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4875 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4877 fib6_clean_all(net, fib6_ifdown, &arg);
4880 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4882 rt6_sync_down_dev(dev, event);
4883 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4884 neigh_ifdown(&nd_tbl, dev);
4887 struct rt6_mtu_change_arg {
4888 struct net_device *dev;
4890 struct fib6_info *f6i;
4893 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4895 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4896 struct fib6_info *f6i = arg->f6i;
4898 /* For administrative MTU increase, there is no way to discover
4899 * IPv6 PMTU increase, so PMTU increase should be updated here.
4900 * Since RFC 1981 doesn't include administrative MTU increase
4901 * update PMTU increase is a MUST. (i.e. jumbo frame)
4903 if (nh->fib_nh_dev == arg->dev) {
4904 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4905 u32 mtu = f6i->fib6_pmtu;
4907 if (mtu >= arg->mtu ||
4908 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4909 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4911 spin_lock_bh(&rt6_exception_lock);
4912 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4913 spin_unlock_bh(&rt6_exception_lock);
4919 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4921 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4922 struct inet6_dev *idev;
4924 /* In IPv6 pmtu discovery is not optional,
4925 so that RTAX_MTU lock cannot disable it.
4926 We still use this lock to block changes
4927 caused by addrconf/ndisc.
4930 idev = __in6_dev_get(arg->dev);
4934 if (fib6_metric_locked(f6i, RTAX_MTU))
4939 /* fib6_nh_mtu_change only returns 0, so this is safe */
4940 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4944 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4947 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4949 struct rt6_mtu_change_arg arg = {
4954 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4957 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4958 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
4959 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
4960 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
4961 [RTA_OIF] = { .type = NLA_U32 },
4962 [RTA_IIF] = { .type = NLA_U32 },
4963 [RTA_PRIORITY] = { .type = NLA_U32 },
4964 [RTA_METRICS] = { .type = NLA_NESTED },
4965 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
4966 [RTA_PREF] = { .type = NLA_U8 },
4967 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4968 [RTA_ENCAP] = { .type = NLA_NESTED },
4969 [RTA_EXPIRES] = { .type = NLA_U32 },
4970 [RTA_UID] = { .type = NLA_U32 },
4971 [RTA_MARK] = { .type = NLA_U32 },
4972 [RTA_TABLE] = { .type = NLA_U32 },
4973 [RTA_IP_PROTO] = { .type = NLA_U8 },
4974 [RTA_SPORT] = { .type = NLA_U16 },
4975 [RTA_DPORT] = { .type = NLA_U16 },
4976 [RTA_NH_ID] = { .type = NLA_U32 },
4979 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
4980 struct fib6_config *cfg,
4981 struct netlink_ext_ack *extack)
4984 struct nlattr *tb[RTA_MAX+1];
4988 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4989 rtm_ipv6_policy, extack);
4994 rtm = nlmsg_data(nlh);
4996 *cfg = (struct fib6_config){
4997 .fc_table = rtm->rtm_table,
4998 .fc_dst_len = rtm->rtm_dst_len,
4999 .fc_src_len = rtm->rtm_src_len,
5001 .fc_protocol = rtm->rtm_protocol,
5002 .fc_type = rtm->rtm_type,
5004 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
5005 .fc_nlinfo.nlh = nlh,
5006 .fc_nlinfo.nl_net = sock_net(skb->sk),
5009 if (rtm->rtm_type == RTN_UNREACHABLE ||
5010 rtm->rtm_type == RTN_BLACKHOLE ||
5011 rtm->rtm_type == RTN_PROHIBIT ||
5012 rtm->rtm_type == RTN_THROW)
5013 cfg->fc_flags |= RTF_REJECT;
5015 if (rtm->rtm_type == RTN_LOCAL)
5016 cfg->fc_flags |= RTF_LOCAL;
5018 if (rtm->rtm_flags & RTM_F_CLONED)
5019 cfg->fc_flags |= RTF_CACHE;
5021 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5023 if (tb[RTA_NH_ID]) {
5024 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
5025 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5026 NL_SET_ERR_MSG(extack,
5027 "Nexthop specification and nexthop id are mutually exclusive");
5030 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5033 if (tb[RTA_GATEWAY]) {
5034 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
5035 cfg->fc_flags |= RTF_GATEWAY;
5038 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5043 int plen = (rtm->rtm_dst_len + 7) >> 3;
5045 if (nla_len(tb[RTA_DST]) < plen)
5048 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
5052 int plen = (rtm->rtm_src_len + 7) >> 3;
5054 if (nla_len(tb[RTA_SRC]) < plen)
5057 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
5060 if (tb[RTA_PREFSRC])
5061 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
5064 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5066 if (tb[RTA_PRIORITY])
5067 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5069 if (tb[RTA_METRICS]) {
5070 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5071 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
5075 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5077 if (tb[RTA_MULTIPATH]) {
5078 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5079 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
5081 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
5082 cfg->fc_mp_len, extack);
5088 pref = nla_get_u8(tb[RTA_PREF]);
5089 if (pref != ICMPV6_ROUTER_PREF_LOW &&
5090 pref != ICMPV6_ROUTER_PREF_HIGH)
5091 pref = ICMPV6_ROUTER_PREF_MEDIUM;
5092 cfg->fc_flags |= RTF_PREF(pref);
5096 cfg->fc_encap = tb[RTA_ENCAP];
5098 if (tb[RTA_ENCAP_TYPE]) {
5099 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5101 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5106 if (tb[RTA_EXPIRES]) {
5107 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5109 if (addrconf_finite_timeout(timeout)) {
5110 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5111 cfg->fc_flags |= RTF_EXPIRES;
5121 struct fib6_info *fib6_info;
5122 struct fib6_config r_cfg;
5123 struct list_head next;
5126 static int ip6_route_info_append(struct net *net,
5127 struct list_head *rt6_nh_list,
5128 struct fib6_info *rt,
5129 struct fib6_config *r_cfg)
5134 list_for_each_entry(nh, rt6_nh_list, next) {
5135 /* check if fib6_info already exists */
5136 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5140 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5144 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5145 list_add_tail(&nh->next, rt6_nh_list);
5150 static void ip6_route_mpath_notify(struct fib6_info *rt,
5151 struct fib6_info *rt_last,
5152 struct nl_info *info,
5155 /* if this is an APPEND route, then rt points to the first route
5156 * inserted and rt_last points to last route inserted. Userspace
5157 * wants a consistent dump of the route which starts at the first
5158 * nexthop. Since sibling routes are always added at the end of
5159 * the list, find the first sibling of the last route appended
5161 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5162 rt = list_first_entry(&rt_last->fib6_siblings,
5168 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5171 static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5173 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5174 bool should_notify = false;
5175 struct fib6_info *leaf;
5176 struct fib6_node *fn;
5179 fn = rcu_dereference(rt->fib6_node);
5183 leaf = rcu_dereference(fn->leaf);
5188 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5189 rt6_qualify_for_ecmp(leaf)))
5190 should_notify = true;
5194 return should_notify;
5197 static int ip6_route_multipath_add(struct fib6_config *cfg,
5198 struct netlink_ext_ack *extack)
5200 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5201 struct nl_info *info = &cfg->fc_nlinfo;
5202 struct fib6_config r_cfg;
5203 struct rtnexthop *rtnh;
5204 struct fib6_info *rt;
5205 struct rt6_nh *err_nh;
5206 struct rt6_nh *nh, *nh_safe;
5212 int replace = (cfg->fc_nlinfo.nlh &&
5213 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5214 LIST_HEAD(rt6_nh_list);
5216 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5217 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5218 nlflags |= NLM_F_APPEND;
5220 remaining = cfg->fc_mp_len;
5221 rtnh = (struct rtnexthop *)cfg->fc_mp;
5223 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
5224 * fib6_info structs per nexthop
5226 while (rtnh_ok(rtnh, remaining)) {
5227 memcpy(&r_cfg, cfg, sizeof(*cfg));
5228 if (rtnh->rtnh_ifindex)
5229 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5231 attrlen = rtnh_attrlen(rtnh);
5233 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5235 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5237 r_cfg.fc_gateway = nla_get_in6_addr(nla);
5238 r_cfg.fc_flags |= RTF_GATEWAY;
5240 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5241 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5243 r_cfg.fc_encap_type = nla_get_u16(nla);
5246 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5247 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5253 if (!rt6_qualify_for_ecmp(rt)) {
5255 NL_SET_ERR_MSG(extack,
5256 "Device only routes can not be added for IPv6 using the multipath API.");
5257 fib6_info_release(rt);
5261 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5263 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5266 fib6_info_release(rt);
5270 rtnh = rtnh_next(rtnh, &remaining);
5273 if (list_empty(&rt6_nh_list)) {
5274 NL_SET_ERR_MSG(extack,
5275 "Invalid nexthop configuration - no valid nexthops");
5279 /* for add and replace send one notification with all nexthops.
5280 * Skip the notification in fib6_add_rt2node and send one with
5281 * the full route when done
5283 info->skip_notify = 1;
5285 /* For add and replace, send one notification with all nexthops. For
5286 * append, send one notification with all appended nexthops.
5288 info->skip_notify_kernel = 1;
5291 list_for_each_entry(nh, &rt6_nh_list, next) {
5292 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5293 fib6_info_release(nh->fib6_info);
5296 /* save reference to last route successfully inserted */
5297 rt_last = nh->fib6_info;
5299 /* save reference to first route for notification */
5301 rt_notif = nh->fib6_info;
5304 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5305 nh->fib6_info = NULL;
5308 NL_SET_ERR_MSG_MOD(extack,
5309 "multipath route replace failed (check consistency of installed routes)");
5314 /* Because each route is added like a single route we remove
5315 * these flags after the first nexthop: if there is a collision,
5316 * we have already failed to add the first nexthop:
5317 * fib6_add_rt2node() has rejected it; when replacing, old
5318 * nexthops have been replaced by first new, the rest should
5321 if (cfg->fc_nlinfo.nlh) {
5322 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5324 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5329 /* An in-kernel notification should only be sent in case the new
5330 * multipath route is added as the first route in the node, or if
5331 * it was appended to it. We pass 'rt_notif' since it is the first
5332 * sibling and might allow us to skip some checks in the replace case.
5334 if (ip6_route_mpath_should_notify(rt_notif)) {
5335 enum fib_event_type fib_event;
5337 if (rt_notif->fib6_nsiblings != nhn - 1)
5338 fib_event = FIB_EVENT_ENTRY_APPEND;
5340 fib_event = FIB_EVENT_ENTRY_REPLACE;
5342 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5343 fib_event, rt_notif,
5346 /* Delete all the siblings that were just added */
5352 /* success ... tell user about new route */
5353 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5357 /* send notification for routes that were added so that
5358 * the delete notifications sent by ip6_route_del are
5362 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5364 /* Delete routes that were already added */
5365 list_for_each_entry(nh, &rt6_nh_list, next) {
5368 ip6_route_del(&nh->r_cfg, extack);
5372 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5374 fib6_info_release(nh->fib6_info);
5375 list_del(&nh->next);
5382 static int ip6_route_multipath_del(struct fib6_config *cfg,
5383 struct netlink_ext_ack *extack)
5385 struct fib6_config r_cfg;
5386 struct rtnexthop *rtnh;
5392 remaining = cfg->fc_mp_len;
5393 rtnh = (struct rtnexthop *)cfg->fc_mp;
5395 /* Parse a Multipath Entry */
5396 while (rtnh_ok(rtnh, remaining)) {
5397 memcpy(&r_cfg, cfg, sizeof(*cfg));
5398 if (rtnh->rtnh_ifindex)
5399 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5401 attrlen = rtnh_attrlen(rtnh);
5403 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5405 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5407 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5408 r_cfg.fc_flags |= RTF_GATEWAY;
5411 err = ip6_route_del(&r_cfg, extack);
5415 rtnh = rtnh_next(rtnh, &remaining);
5421 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5422 struct netlink_ext_ack *extack)
5424 struct fib6_config cfg;
5427 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5432 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5433 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5438 return ip6_route_multipath_del(&cfg, extack);
5440 cfg.fc_delete_all_nh = 1;
5441 return ip6_route_del(&cfg, extack);
5445 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5446 struct netlink_ext_ack *extack)
5448 struct fib6_config cfg;
5451 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5455 if (cfg.fc_metric == 0)
5456 cfg.fc_metric = IP6_RT_PRIO_USER;
5459 return ip6_route_multipath_add(&cfg, extack);
5461 return ip6_route_add(&cfg, GFP_KERNEL, extack);
5464 /* add the overhead of this fib6_nh to nexthop_len */
5465 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5467 int *nexthop_len = arg;
5469 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5470 + NLA_ALIGN(sizeof(struct rtnexthop))
5471 + nla_total_size(16); /* RTA_GATEWAY */
5473 if (nh->fib_nh_lws) {
5474 /* RTA_ENCAP_TYPE */
5475 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5477 *nexthop_len += nla_total_size(2);
5483 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5488 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5489 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5492 struct fib6_nh *nh = f6i->fib6_nh;
5495 if (f6i->fib6_nsiblings) {
5496 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
5497 + NLA_ALIGN(sizeof(struct rtnexthop))
5498 + nla_total_size(16) /* RTA_GATEWAY */
5499 + lwtunnel_get_encap_size(nh->fib_nh_lws);
5501 nexthop_len *= f6i->fib6_nsiblings;
5503 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5506 return NLMSG_ALIGN(sizeof(struct rtmsg))
5507 + nla_total_size(16) /* RTA_SRC */
5508 + nla_total_size(16) /* RTA_DST */
5509 + nla_total_size(16) /* RTA_GATEWAY */
5510 + nla_total_size(16) /* RTA_PREFSRC */
5511 + nla_total_size(4) /* RTA_TABLE */
5512 + nla_total_size(4) /* RTA_IIF */
5513 + nla_total_size(4) /* RTA_OIF */
5514 + nla_total_size(4) /* RTA_PRIORITY */
5515 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5516 + nla_total_size(sizeof(struct rta_cacheinfo))
5517 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5518 + nla_total_size(1) /* RTA_PREF */
5522 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5523 unsigned char *flags)
5525 if (nexthop_is_multipath(nh)) {
5528 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5530 goto nla_put_failure;
5532 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5533 goto nla_put_failure;
5535 nla_nest_end(skb, mp);
5537 struct fib6_nh *fib6_nh;
5539 fib6_nh = nexthop_fib6_nh(nh);
5540 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5542 goto nla_put_failure;
5551 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5552 struct fib6_info *rt, struct dst_entry *dst,
5553 struct in6_addr *dest, struct in6_addr *src,
5554 int iif, int type, u32 portid, u32 seq,
5557 struct rt6_info *rt6 = (struct rt6_info *)dst;
5558 struct rt6key *rt6_dst, *rt6_src;
5559 u32 *pmetrics, table, rt6_flags;
5560 unsigned char nh_flags = 0;
5561 struct nlmsghdr *nlh;
5565 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5570 rt6_dst = &rt6->rt6i_dst;
5571 rt6_src = &rt6->rt6i_src;
5572 rt6_flags = rt6->rt6i_flags;
5574 rt6_dst = &rt->fib6_dst;
5575 rt6_src = &rt->fib6_src;
5576 rt6_flags = rt->fib6_flags;
5579 rtm = nlmsg_data(nlh);
5580 rtm->rtm_family = AF_INET6;
5581 rtm->rtm_dst_len = rt6_dst->plen;
5582 rtm->rtm_src_len = rt6_src->plen;
5585 table = rt->fib6_table->tb6_id;
5587 table = RT6_TABLE_UNSPEC;
5588 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5589 if (nla_put_u32(skb, RTA_TABLE, table))
5590 goto nla_put_failure;
5592 rtm->rtm_type = rt->fib6_type;
5594 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5595 rtm->rtm_protocol = rt->fib6_protocol;
5597 if (rt6_flags & RTF_CACHE)
5598 rtm->rtm_flags |= RTM_F_CLONED;
5601 if (nla_put_in6_addr(skb, RTA_DST, dest))
5602 goto nla_put_failure;
5603 rtm->rtm_dst_len = 128;
5604 } else if (rtm->rtm_dst_len)
5605 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5606 goto nla_put_failure;
5607 #ifdef CONFIG_IPV6_SUBTREES
5609 if (nla_put_in6_addr(skb, RTA_SRC, src))
5610 goto nla_put_failure;
5611 rtm->rtm_src_len = 128;
5612 } else if (rtm->rtm_src_len &&
5613 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5614 goto nla_put_failure;
5617 #ifdef CONFIG_IPV6_MROUTE
5618 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5619 int err = ip6mr_get_route(net, skb, rtm, portid);
5624 goto nla_put_failure;
5627 if (nla_put_u32(skb, RTA_IIF, iif))
5628 goto nla_put_failure;
5630 struct in6_addr saddr_buf;
5631 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
5632 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5633 goto nla_put_failure;
5636 if (rt->fib6_prefsrc.plen) {
5637 struct in6_addr saddr_buf;
5638 saddr_buf = rt->fib6_prefsrc.addr;
5639 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5640 goto nla_put_failure;
5643 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5644 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5645 goto nla_put_failure;
5647 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5648 goto nla_put_failure;
5650 /* For multipath routes, walk the siblings list and add
5651 * each as a nexthop within RTA_MULTIPATH.
5654 if (rt6_flags & RTF_GATEWAY &&
5655 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5656 goto nla_put_failure;
5658 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5659 goto nla_put_failure;
5661 if (dst->lwtstate &&
5662 lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5663 goto nla_put_failure;
5664 } else if (rt->fib6_nsiblings) {
5665 struct fib6_info *sibling, *next_sibling;
5668 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5670 goto nla_put_failure;
5672 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5673 rt->fib6_nh->fib_nh_weight, AF_INET6) < 0)
5674 goto nla_put_failure;
5676 list_for_each_entry_safe(sibling, next_sibling,
5677 &rt->fib6_siblings, fib6_siblings) {
5678 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5679 sibling->fib6_nh->fib_nh_weight,
5681 goto nla_put_failure;
5684 nla_nest_end(skb, mp);
5685 } else if (rt->nh) {
5686 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5687 goto nla_put_failure;
5689 if (nexthop_is_blackhole(rt->nh))
5690 rtm->rtm_type = RTN_BLACKHOLE;
5692 if (net->ipv4.sysctl_nexthop_compat_mode &&
5693 rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5694 goto nla_put_failure;
5696 rtm->rtm_flags |= nh_flags;
5698 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5699 &nh_flags, false) < 0)
5700 goto nla_put_failure;
5702 rtm->rtm_flags |= nh_flags;
5705 if (rt6_flags & RTF_EXPIRES) {
5706 expires = dst ? dst->expires : rt->expires;
5712 rtm->rtm_flags |= RTM_F_OFFLOAD;
5714 rtm->rtm_flags |= RTM_F_TRAP;
5715 if (rt->offload_failed)
5716 rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5719 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5720 goto nla_put_failure;
5722 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5723 goto nla_put_failure;
5726 nlmsg_end(skb, nlh);
5730 nlmsg_cancel(skb, nlh);
5734 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5736 const struct net_device *dev = arg;
5738 if (nh->fib_nh_dev == dev)
5744 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5745 const struct net_device *dev)
5748 struct net_device *_dev = (struct net_device *)dev;
5750 return !!nexthop_for_each_fib6_nh(f6i->nh,
5751 fib6_info_nh_uses_dev,
5755 if (f6i->fib6_nh->fib_nh_dev == dev)
5758 if (f6i->fib6_nsiblings) {
5759 struct fib6_info *sibling, *next_sibling;
5761 list_for_each_entry_safe(sibling, next_sibling,
5762 &f6i->fib6_siblings, fib6_siblings) {
5763 if (sibling->fib6_nh->fib_nh_dev == dev)
5771 struct fib6_nh_exception_dump_walker {
5772 struct rt6_rtnl_dump_arg *dump;
5773 struct fib6_info *rt;
5779 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5781 struct fib6_nh_exception_dump_walker *w = arg;
5782 struct rt6_rtnl_dump_arg *dump = w->dump;
5783 struct rt6_exception_bucket *bucket;
5784 struct rt6_exception *rt6_ex;
5787 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5791 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5792 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5798 /* Expiration of entries doesn't bump sernum, insertion
5799 * does. Removal is triggered by insertion, so we can
5800 * rely on the fact that if entries change between two
5801 * partial dumps, this node is scanned again completely,
5802 * see rt6_insert_exception() and fib6_dump_table().
5804 * Count expired entries we go through as handled
5805 * entries that we'll skip next time, in case of partial
5806 * node dump. Otherwise, if entries expire meanwhile,
5807 * we'll skip the wrong amount.
5809 if (rt6_check_expired(rt6_ex->rt6i)) {
5814 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5815 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5817 NETLINK_CB(dump->cb->skb).portid,
5818 dump->cb->nlh->nlmsg_seq, w->flags);
5830 /* Return -1 if done with node, number of handled routes on partial dump */
5831 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5833 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5834 struct fib_dump_filter *filter = &arg->filter;
5835 unsigned int flags = NLM_F_MULTI;
5836 struct net *net = arg->net;
5839 if (rt == net->ipv6.fib6_null_entry)
5842 if ((filter->flags & RTM_F_PREFIX) &&
5843 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5844 /* success since this is not a prefix route */
5847 if (filter->filter_set &&
5848 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5849 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5850 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5854 if (filter->filter_set ||
5855 !filter->dump_routes || !filter->dump_exceptions) {
5856 flags |= NLM_F_DUMP_FILTERED;
5859 if (filter->dump_routes) {
5863 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5865 NETLINK_CB(arg->cb->skb).portid,
5866 arg->cb->nlh->nlmsg_seq, flags)) {
5873 if (filter->dump_exceptions) {
5874 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5883 err = nexthop_for_each_fib6_nh(rt->nh,
5884 rt6_nh_dump_exceptions,
5887 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5892 return count += w.count;
5898 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5899 const struct nlmsghdr *nlh,
5901 struct netlink_ext_ack *extack)
5906 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5907 NL_SET_ERR_MSG_MOD(extack,
5908 "Invalid header for get route request");
5912 if (!netlink_strict_get_check(skb))
5913 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5914 rtm_ipv6_policy, extack);
5916 rtm = nlmsg_data(nlh);
5917 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5918 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5919 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5921 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5924 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5925 NL_SET_ERR_MSG_MOD(extack,
5926 "Invalid flags for get route request");
5930 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5931 rtm_ipv6_policy, extack);
5935 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5936 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5937 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5941 for (i = 0; i <= RTA_MAX; i++) {
5957 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5965 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5966 struct netlink_ext_ack *extack)
5968 struct net *net = sock_net(in_skb->sk);
5969 struct nlattr *tb[RTA_MAX+1];
5970 int err, iif = 0, oif = 0;
5971 struct fib6_info *from;
5972 struct dst_entry *dst;
5973 struct rt6_info *rt;
5974 struct sk_buff *skb;
5976 struct flowi6 fl6 = {};
5979 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
5984 rtm = nlmsg_data(nlh);
5985 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
5986 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
5989 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
5992 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
5996 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
5999 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
6003 iif = nla_get_u32(tb[RTA_IIF]);
6006 oif = nla_get_u32(tb[RTA_OIF]);
6009 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6012 fl6.flowi6_uid = make_kuid(current_user_ns(),
6013 nla_get_u32(tb[RTA_UID]));
6015 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6018 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6021 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6023 if (tb[RTA_IP_PROTO]) {
6024 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
6025 &fl6.flowi6_proto, AF_INET6,
6032 struct net_device *dev;
6037 dev = dev_get_by_index_rcu(net, iif);
6044 fl6.flowi6_iif = iif;
6046 if (!ipv6_addr_any(&fl6.saddr))
6047 flags |= RT6_LOOKUP_F_HAS_SADDR;
6049 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
6053 fl6.flowi6_oif = oif;
6055 dst = ip6_route_output(net, NULL, &fl6);
6059 rt = container_of(dst, struct rt6_info, dst);
6060 if (rt->dst.error) {
6061 err = rt->dst.error;
6066 if (rt == net->ipv6.ip6_null_entry) {
6067 err = rt->dst.error;
6072 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
6079 skb_dst_set(skb, &rt->dst);
6082 from = rcu_dereference(rt->from);
6085 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6087 NETLINK_CB(in_skb).portid,
6090 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6091 &fl6.saddr, iif, RTM_NEWROUTE,
6092 NETLINK_CB(in_skb).portid,
6104 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6109 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6110 unsigned int nlm_flags)
6112 struct sk_buff *skb;
6113 struct net *net = info->nl_net;
6118 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6120 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6124 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6125 event, info->portid, seq, nlm_flags);
6127 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6128 WARN_ON(err == -EMSGSIZE);
6132 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6133 info->nlh, gfp_any());
6137 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6140 void fib6_rt_update(struct net *net, struct fib6_info *rt,
6141 struct nl_info *info)
6143 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6144 struct sk_buff *skb;
6147 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6151 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6152 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6154 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6155 WARN_ON(err == -EMSGSIZE);
6159 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6160 info->nlh, gfp_any());
6164 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6167 void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6168 bool offload, bool trap, bool offload_failed)
6170 struct sk_buff *skb;
6173 if (f6i->offload == offload && f6i->trap == trap &&
6174 f6i->offload_failed == offload_failed)
6177 f6i->offload = offload;
6180 /* 2 means send notifications only if offload_failed was changed. */
6181 if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6182 f6i->offload_failed == offload_failed)
6185 f6i->offload_failed = offload_failed;
6187 if (!rcu_access_pointer(f6i->fib6_node))
6188 /* The route was removed from the tree, do not send
6193 if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6196 skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6202 err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6205 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6206 WARN_ON(err == -EMSGSIZE);
6211 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6215 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6217 EXPORT_SYMBOL(fib6_info_hw_flags_set);
6219 static int ip6_route_dev_notify(struct notifier_block *this,
6220 unsigned long event, void *ptr)
6222 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6223 struct net *net = dev_net(dev);
6225 if (!(dev->flags & IFF_LOOPBACK))
6228 if (event == NETDEV_REGISTER) {
6229 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6230 net->ipv6.ip6_null_entry->dst.dev = dev;
6231 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6232 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6233 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6234 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6235 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6236 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6238 } else if (event == NETDEV_UNREGISTER &&
6239 dev->reg_state != NETREG_UNREGISTERED) {
6240 /* NETDEV_UNREGISTER could be fired for multiple times by
6241 * netdev_wait_allrefs(). Make sure we only call this once.
6243 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6244 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6245 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6246 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6257 #ifdef CONFIG_PROC_FS
6258 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6260 struct net *net = (struct net *)seq->private;
6261 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6262 net->ipv6.rt6_stats->fib_nodes,
6263 net->ipv6.rt6_stats->fib_route_nodes,
6264 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6265 net->ipv6.rt6_stats->fib_rt_entries,
6266 net->ipv6.rt6_stats->fib_rt_cache,
6267 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6268 net->ipv6.rt6_stats->fib_discarded_routes);
6272 #endif /* CONFIG_PROC_FS */
6274 #ifdef CONFIG_SYSCTL
6276 static int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6277 void *buffer, size_t *lenp, loff_t *ppos)
6285 net = (struct net *)ctl->extra1;
6286 delay = net->ipv6.sysctl.flush_delay;
6287 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6291 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6295 static struct ctl_table ipv6_route_table_template[] = {
6297 .procname = "flush",
6298 .data = &init_net.ipv6.sysctl.flush_delay,
6299 .maxlen = sizeof(int),
6301 .proc_handler = ipv6_sysctl_rtcache_flush
6304 .procname = "gc_thresh",
6305 .data = &ip6_dst_ops_template.gc_thresh,
6306 .maxlen = sizeof(int),
6308 .proc_handler = proc_dointvec,
6311 .procname = "max_size",
6312 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
6313 .maxlen = sizeof(int),
6315 .proc_handler = proc_dointvec,
6318 .procname = "gc_min_interval",
6319 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6320 .maxlen = sizeof(int),
6322 .proc_handler = proc_dointvec_jiffies,
6325 .procname = "gc_timeout",
6326 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6327 .maxlen = sizeof(int),
6329 .proc_handler = proc_dointvec_jiffies,
6332 .procname = "gc_interval",
6333 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
6334 .maxlen = sizeof(int),
6336 .proc_handler = proc_dointvec_jiffies,
6339 .procname = "gc_elasticity",
6340 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6341 .maxlen = sizeof(int),
6343 .proc_handler = proc_dointvec,
6346 .procname = "mtu_expires",
6347 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6348 .maxlen = sizeof(int),
6350 .proc_handler = proc_dointvec_jiffies,
6353 .procname = "min_adv_mss",
6354 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
6355 .maxlen = sizeof(int),
6357 .proc_handler = proc_dointvec,
6360 .procname = "gc_min_interval_ms",
6361 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6362 .maxlen = sizeof(int),
6364 .proc_handler = proc_dointvec_ms_jiffies,
6367 .procname = "skip_notify_on_dev_down",
6368 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6369 .maxlen = sizeof(int),
6371 .proc_handler = proc_dointvec_minmax,
6372 .extra1 = SYSCTL_ZERO,
6373 .extra2 = SYSCTL_ONE,
6378 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6380 struct ctl_table *table;
6382 table = kmemdup(ipv6_route_table_template,
6383 sizeof(ipv6_route_table_template),
6387 table[0].data = &net->ipv6.sysctl.flush_delay;
6388 table[0].extra1 = net;
6389 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6390 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
6391 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6392 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6393 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6394 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6395 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6396 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6397 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6398 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6400 /* Don't export sysctls to unprivileged users */
6401 if (net->user_ns != &init_user_ns)
6402 table[0].procname = NULL;
6409 static int __net_init ip6_route_net_init(struct net *net)
6413 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6414 sizeof(net->ipv6.ip6_dst_ops));
6416 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6417 goto out_ip6_dst_ops;
6419 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6420 if (!net->ipv6.fib6_null_entry)
6421 goto out_ip6_dst_entries;
6422 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6423 sizeof(*net->ipv6.fib6_null_entry));
6425 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6426 sizeof(*net->ipv6.ip6_null_entry),
6428 if (!net->ipv6.ip6_null_entry)
6429 goto out_fib6_null_entry;
6430 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6431 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6432 ip6_template_metrics, true);
6433 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
6435 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6436 net->ipv6.fib6_has_custom_rules = false;
6437 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6438 sizeof(*net->ipv6.ip6_prohibit_entry),
6440 if (!net->ipv6.ip6_prohibit_entry)
6441 goto out_ip6_null_entry;
6442 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6443 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6444 ip6_template_metrics, true);
6445 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
6447 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6448 sizeof(*net->ipv6.ip6_blk_hole_entry),
6450 if (!net->ipv6.ip6_blk_hole_entry)
6451 goto out_ip6_prohibit_entry;
6452 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6453 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6454 ip6_template_metrics, true);
6455 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
6456 #ifdef CONFIG_IPV6_SUBTREES
6457 net->ipv6.fib6_routes_require_src = 0;
6461 net->ipv6.sysctl.flush_delay = 0;
6462 net->ipv6.sysctl.ip6_rt_max_size = 4096;
6463 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6464 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6465 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6466 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6467 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6468 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6469 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6471 net->ipv6.ip6_rt_gc_expire = 30*HZ;
6477 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6478 out_ip6_prohibit_entry:
6479 kfree(net->ipv6.ip6_prohibit_entry);
6481 kfree(net->ipv6.ip6_null_entry);
6483 out_fib6_null_entry:
6484 kfree(net->ipv6.fib6_null_entry);
6485 out_ip6_dst_entries:
6486 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6491 static void __net_exit ip6_route_net_exit(struct net *net)
6493 kfree(net->ipv6.fib6_null_entry);
6494 kfree(net->ipv6.ip6_null_entry);
6495 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6496 kfree(net->ipv6.ip6_prohibit_entry);
6497 kfree(net->ipv6.ip6_blk_hole_entry);
6499 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6502 static int __net_init ip6_route_net_init_late(struct net *net)
6504 #ifdef CONFIG_PROC_FS
6505 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6506 sizeof(struct ipv6_route_iter));
6507 proc_create_net_single("rt6_stats", 0444, net->proc_net,
6508 rt6_stats_seq_show, NULL);
6513 static void __net_exit ip6_route_net_exit_late(struct net *net)
6515 #ifdef CONFIG_PROC_FS
6516 remove_proc_entry("ipv6_route", net->proc_net);
6517 remove_proc_entry("rt6_stats", net->proc_net);
6521 static struct pernet_operations ip6_route_net_ops = {
6522 .init = ip6_route_net_init,
6523 .exit = ip6_route_net_exit,
6526 static int __net_init ipv6_inetpeer_init(struct net *net)
6528 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6532 inet_peer_base_init(bp);
6533 net->ipv6.peers = bp;
6537 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6539 struct inet_peer_base *bp = net->ipv6.peers;
6541 net->ipv6.peers = NULL;
6542 inetpeer_invalidate_tree(bp);
6546 static struct pernet_operations ipv6_inetpeer_ops = {
6547 .init = ipv6_inetpeer_init,
6548 .exit = ipv6_inetpeer_exit,
6551 static struct pernet_operations ip6_route_net_late_ops = {
6552 .init = ip6_route_net_init_late,
6553 .exit = ip6_route_net_exit_late,
6556 static struct notifier_block ip6_route_dev_notifier = {
6557 .notifier_call = ip6_route_dev_notify,
6558 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6561 void __init ip6_route_init_special_entries(void)
6563 /* Registering of the loopback is done before this portion of code,
6564 * the loopback reference in rt6_info will not be taken, do it
6565 * manually for init_net */
6566 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6567 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6568 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6569 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6570 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6571 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6572 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6573 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6577 #if IS_BUILTIN(CONFIG_IPV6)
6578 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6579 DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6581 BTF_ID_LIST(btf_fib6_info_id)
6582 BTF_ID(struct, fib6_info)
6584 static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6585 .seq_ops = &ipv6_route_seq_ops,
6586 .init_seq_private = bpf_iter_init_seq_net,
6587 .fini_seq_private = bpf_iter_fini_seq_net,
6588 .seq_priv_size = sizeof(struct ipv6_route_iter),
6591 static struct bpf_iter_reg ipv6_route_reg_info = {
6592 .target = "ipv6_route",
6593 .ctx_arg_info_size = 1,
6595 { offsetof(struct bpf_iter__ipv6_route, rt),
6596 PTR_TO_BTF_ID_OR_NULL },
6598 .seq_info = &ipv6_route_seq_info,
6601 static int __init bpf_iter_register(void)
6603 ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6604 return bpf_iter_reg_target(&ipv6_route_reg_info);
6607 static void bpf_iter_unregister(void)
6609 bpf_iter_unreg_target(&ipv6_route_reg_info);
6614 int __init ip6_route_init(void)
6620 ip6_dst_ops_template.kmem_cachep =
6621 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6622 SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL);
6623 if (!ip6_dst_ops_template.kmem_cachep)
6626 ret = dst_entries_init(&ip6_dst_blackhole_ops);
6628 goto out_kmem_cache;
6630 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6632 goto out_dst_entries;
6634 ret = register_pernet_subsys(&ip6_route_net_ops);
6636 goto out_register_inetpeer;
6638 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6642 goto out_register_subsys;
6648 ret = fib6_rules_init();
6652 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6654 goto fib6_rules_init;
6656 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6657 inet6_rtm_newroute, NULL, 0);
6659 goto out_register_late_subsys;
6661 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6662 inet6_rtm_delroute, NULL, 0);
6664 goto out_register_late_subsys;
6666 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6667 inet6_rtm_getroute, NULL,
6668 RTNL_FLAG_DOIT_UNLOCKED);
6670 goto out_register_late_subsys;
6672 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6674 goto out_register_late_subsys;
6676 #if IS_BUILTIN(CONFIG_IPV6)
6677 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6678 ret = bpf_iter_register();
6680 goto out_register_late_subsys;
6684 for_each_possible_cpu(cpu) {
6685 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6687 INIT_LIST_HEAD(&ul->head);
6688 spin_lock_init(&ul->lock);
6694 out_register_late_subsys:
6695 rtnl_unregister_all(PF_INET6);
6696 unregister_pernet_subsys(&ip6_route_net_late_ops);
6698 fib6_rules_cleanup();
6703 out_register_subsys:
6704 unregister_pernet_subsys(&ip6_route_net_ops);
6705 out_register_inetpeer:
6706 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6708 dst_entries_destroy(&ip6_dst_blackhole_ops);
6710 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6714 void ip6_route_cleanup(void)
6716 #if IS_BUILTIN(CONFIG_IPV6)
6717 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6718 bpf_iter_unregister();
6721 unregister_netdevice_notifier(&ip6_route_dev_notifier);
6722 unregister_pernet_subsys(&ip6_route_net_late_ops);
6723 fib6_rules_cleanup();
6726 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6727 unregister_pernet_subsys(&ip6_route_net_ops);
6728 dst_entries_destroy(&ip6_dst_blackhole_ops);
6729 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);