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 <linux/siphash.h>
45 #include <net/net_namespace.h>
48 #include <net/ip6_fib.h>
49 #include <net/ip6_route.h>
50 #include <net/ndisc.h>
51 #include <net/addrconf.h>
53 #include <linux/rtnetlink.h>
55 #include <net/dst_metadata.h>
57 #include <net/netevent.h>
58 #include <net/netlink.h>
60 #include <net/lwtunnel.h>
61 #include <net/ip_tunnels.h>
62 #include <net/l3mdev.h>
64 #include <linux/uaccess.h>
65 #include <linux/btf_ids.h>
68 #include <linux/sysctl.h>
71 static int ip6_rt_type_to_error(u8 fib6_type);
73 #define CREATE_TRACE_POINTS
74 #include <trace/events/fib6.h>
75 EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
76 #undef CREATE_TRACE_POINTS
79 RT6_NUD_FAIL_HARD = -3,
80 RT6_NUD_FAIL_PROBE = -2,
81 RT6_NUD_FAIL_DO_RR = -1,
85 INDIRECT_CALLABLE_SCOPE
86 struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
87 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
88 INDIRECT_CALLABLE_SCOPE
89 unsigned int ip6_mtu(const struct dst_entry *dst);
90 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
91 static void ip6_dst_destroy(struct dst_entry *);
92 static void ip6_dst_ifdown(struct dst_entry *,
93 struct net_device *dev, int how);
94 static int ip6_dst_gc(struct dst_ops *ops);
96 static int ip6_pkt_discard(struct sk_buff *skb);
97 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
98 static int ip6_pkt_prohibit(struct sk_buff *skb);
99 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
100 static void ip6_link_failure(struct sk_buff *skb);
101 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
102 struct sk_buff *skb, u32 mtu,
104 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
105 struct sk_buff *skb);
106 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
108 static size_t rt6_nlmsg_size(struct fib6_info *f6i);
109 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
110 struct fib6_info *rt, struct dst_entry *dst,
111 struct in6_addr *dest, struct in6_addr *src,
112 int iif, int type, u32 portid, u32 seq,
114 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
115 const struct in6_addr *daddr,
116 const struct in6_addr *saddr);
118 #ifdef CONFIG_IPV6_ROUTE_INFO
119 static struct fib6_info *rt6_add_route_info(struct net *net,
120 const struct in6_addr *prefix, int prefixlen,
121 const struct in6_addr *gwaddr,
122 struct net_device *dev,
124 static struct fib6_info *rt6_get_route_info(struct net *net,
125 const struct in6_addr *prefix, int prefixlen,
126 const struct in6_addr *gwaddr,
127 struct net_device *dev);
130 struct uncached_list {
132 struct list_head head;
135 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
137 void rt6_uncached_list_add(struct rt6_info *rt)
139 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
141 rt->rt6i_uncached_list = ul;
143 spin_lock_bh(&ul->lock);
144 list_add_tail(&rt->rt6i_uncached, &ul->head);
145 spin_unlock_bh(&ul->lock);
148 void rt6_uncached_list_del(struct rt6_info *rt)
150 if (!list_empty(&rt->rt6i_uncached)) {
151 struct uncached_list *ul = rt->rt6i_uncached_list;
152 struct net *net = dev_net(rt->dst.dev);
154 spin_lock_bh(&ul->lock);
155 list_del(&rt->rt6i_uncached);
156 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
157 spin_unlock_bh(&ul->lock);
161 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
163 struct net_device *loopback_dev = net->loopback_dev;
166 if (dev == loopback_dev)
169 for_each_possible_cpu(cpu) {
170 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
173 spin_lock_bh(&ul->lock);
174 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
175 struct inet6_dev *rt_idev = rt->rt6i_idev;
176 struct net_device *rt_dev = rt->dst.dev;
178 if (rt_idev->dev == dev) {
179 rt->rt6i_idev = in6_dev_get(loopback_dev);
180 in6_dev_put(rt_idev);
184 rt->dst.dev = blackhole_netdev;
185 dev_hold(rt->dst.dev);
189 spin_unlock_bh(&ul->lock);
193 static inline const void *choose_neigh_daddr(const struct in6_addr *p,
197 if (!ipv6_addr_any(p))
198 return (const void *) p;
200 return &ipv6_hdr(skb)->daddr;
204 struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
205 struct net_device *dev,
211 daddr = choose_neigh_daddr(gw, skb, daddr);
212 n = __ipv6_neigh_lookup(dev, daddr);
216 n = neigh_create(&nd_tbl, daddr, dev);
217 return IS_ERR(n) ? NULL : n;
220 static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
224 const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
226 return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
227 dst->dev, skb, daddr);
230 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
232 struct net_device *dev = dst->dev;
233 struct rt6_info *rt = (struct rt6_info *)dst;
235 daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
238 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
240 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
242 __ipv6_confirm_neigh(dev, daddr);
245 static struct dst_ops ip6_dst_ops_template = {
249 .check = ip6_dst_check,
250 .default_advmss = ip6_default_advmss,
252 .cow_metrics = dst_cow_metrics_generic,
253 .destroy = ip6_dst_destroy,
254 .ifdown = ip6_dst_ifdown,
255 .negative_advice = ip6_negative_advice,
256 .link_failure = ip6_link_failure,
257 .update_pmtu = ip6_rt_update_pmtu,
258 .redirect = rt6_do_redirect,
259 .local_out = __ip6_local_out,
260 .neigh_lookup = ip6_dst_neigh_lookup,
261 .confirm_neigh = ip6_confirm_neigh,
264 static struct dst_ops ip6_dst_blackhole_ops = {
266 .default_advmss = ip6_default_advmss,
267 .neigh_lookup = ip6_dst_neigh_lookup,
268 .check = ip6_dst_check,
269 .destroy = ip6_dst_destroy,
270 .cow_metrics = dst_cow_metrics_generic,
271 .update_pmtu = dst_blackhole_update_pmtu,
272 .redirect = dst_blackhole_redirect,
273 .mtu = dst_blackhole_mtu,
276 static const u32 ip6_template_metrics[RTAX_MAX] = {
277 [RTAX_HOPLIMIT - 1] = 0,
280 static const struct fib6_info fib6_null_entry_template = {
281 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
282 .fib6_protocol = RTPROT_KERNEL,
283 .fib6_metric = ~(u32)0,
284 .fib6_ref = REFCOUNT_INIT(1),
285 .fib6_type = RTN_UNREACHABLE,
286 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
289 static const struct rt6_info ip6_null_entry_template = {
291 .__refcnt = ATOMIC_INIT(1),
293 .obsolete = DST_OBSOLETE_FORCE_CHK,
294 .error = -ENETUNREACH,
295 .input = ip6_pkt_discard,
296 .output = ip6_pkt_discard_out,
298 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
301 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
303 static const struct rt6_info ip6_prohibit_entry_template = {
305 .__refcnt = ATOMIC_INIT(1),
307 .obsolete = DST_OBSOLETE_FORCE_CHK,
309 .input = ip6_pkt_prohibit,
310 .output = ip6_pkt_prohibit_out,
312 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
315 static const struct rt6_info ip6_blk_hole_entry_template = {
317 .__refcnt = ATOMIC_INIT(1),
319 .obsolete = DST_OBSOLETE_FORCE_CHK,
321 .input = dst_discard,
322 .output = dst_discard_out,
324 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
329 static void rt6_info_init(struct rt6_info *rt)
331 struct dst_entry *dst = &rt->dst;
333 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
334 INIT_LIST_HEAD(&rt->rt6i_uncached);
337 /* allocate dst with ip6_dst_ops */
338 struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
341 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
342 1, DST_OBSOLETE_FORCE_CHK, flags);
346 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
351 EXPORT_SYMBOL(ip6_dst_alloc);
353 static void ip6_dst_destroy(struct dst_entry *dst)
355 struct rt6_info *rt = (struct rt6_info *)dst;
356 struct fib6_info *from;
357 struct inet6_dev *idev;
359 ip_dst_metrics_put(dst);
360 rt6_uncached_list_del(rt);
362 idev = rt->rt6i_idev;
364 rt->rt6i_idev = NULL;
368 from = xchg((__force struct fib6_info **)&rt->from, NULL);
369 fib6_info_release(from);
372 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
375 struct rt6_info *rt = (struct rt6_info *)dst;
376 struct inet6_dev *idev = rt->rt6i_idev;
377 struct net_device *loopback_dev =
378 dev_net(dev)->loopback_dev;
380 if (idev && idev->dev != loopback_dev) {
381 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
383 rt->rt6i_idev = loopback_idev;
389 static bool __rt6_check_expired(const struct rt6_info *rt)
391 if (rt->rt6i_flags & RTF_EXPIRES)
392 return time_after(jiffies, rt->dst.expires);
397 static bool rt6_check_expired(const struct rt6_info *rt)
399 struct fib6_info *from;
401 from = rcu_dereference(rt->from);
403 if (rt->rt6i_flags & RTF_EXPIRES) {
404 if (time_after(jiffies, rt->dst.expires))
407 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
408 fib6_check_expired(from);
413 void fib6_select_path(const struct net *net, struct fib6_result *res,
414 struct flowi6 *fl6, int oif, bool have_oif_match,
415 const struct sk_buff *skb, int strict)
417 struct fib6_info *sibling, *next_sibling;
418 struct fib6_info *match = res->f6i;
420 if (!match->nh && (!match->fib6_nsiblings || have_oif_match))
423 if (match->nh && have_oif_match && res->nh)
426 /* We might have already computed the hash for ICMPv6 errors. In such
427 * case it will always be non-zero. Otherwise now is the time to do it.
430 (!match->nh || nexthop_is_multipath(match->nh)))
431 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
433 if (unlikely(match->nh)) {
434 nexthop_path_fib6_result(res, fl6->mp_hash);
438 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
441 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
443 const struct fib6_nh *nh = sibling->fib6_nh;
446 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
447 if (fl6->mp_hash > nh_upper_bound)
449 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
457 res->nh = match->fib6_nh;
461 * Route lookup. rcu_read_lock() should be held.
464 static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
465 const struct in6_addr *saddr, int oif, int flags)
467 const struct net_device *dev;
469 if (nh->fib_nh_flags & RTNH_F_DEAD)
472 dev = nh->fib_nh_dev;
474 if (dev->ifindex == oif)
477 if (ipv6_chk_addr(net, saddr, dev,
478 flags & RT6_LOOKUP_F_IFACE))
485 struct fib6_nh_dm_arg {
487 const struct in6_addr *saddr;
493 static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
495 struct fib6_nh_dm_arg *arg = _arg;
498 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
502 /* returns fib6_nh from nexthop or NULL */
503 static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
504 struct fib6_result *res,
505 const struct in6_addr *saddr,
508 struct fib6_nh_dm_arg arg = {
515 if (nexthop_is_blackhole(nh))
518 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
524 static void rt6_device_match(struct net *net, struct fib6_result *res,
525 const struct in6_addr *saddr, int oif, int flags)
527 struct fib6_info *f6i = res->f6i;
528 struct fib6_info *spf6i;
531 if (!oif && ipv6_addr_any(saddr)) {
532 if (unlikely(f6i->nh)) {
533 nh = nexthop_fib6_nh(f6i->nh);
534 if (nexthop_is_blackhole(f6i->nh))
539 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
543 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
544 bool matched = false;
546 if (unlikely(spf6i->nh)) {
547 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
553 if (__rt6_device_match(net, nh, saddr, oif, flags))
562 if (oif && flags & RT6_LOOKUP_F_IFACE) {
563 res->f6i = net->ipv6.fib6_null_entry;
564 nh = res->f6i->fib6_nh;
568 if (unlikely(f6i->nh)) {
569 nh = nexthop_fib6_nh(f6i->nh);
570 if (nexthop_is_blackhole(f6i->nh))
576 if (nh->fib_nh_flags & RTNH_F_DEAD) {
577 res->f6i = net->ipv6.fib6_null_entry;
578 nh = res->f6i->fib6_nh;
582 res->fib6_type = res->f6i->fib6_type;
583 res->fib6_flags = res->f6i->fib6_flags;
587 res->fib6_flags |= RTF_REJECT;
588 res->fib6_type = RTN_BLACKHOLE;
592 #ifdef CONFIG_IPV6_ROUTER_PREF
593 struct __rt6_probe_work {
594 struct work_struct work;
595 struct in6_addr target;
596 struct net_device *dev;
599 static void rt6_probe_deferred(struct work_struct *w)
601 struct in6_addr mcaddr;
602 struct __rt6_probe_work *work =
603 container_of(w, struct __rt6_probe_work, work);
605 addrconf_addr_solict_mult(&work->target, &mcaddr);
606 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
611 static void rt6_probe(struct fib6_nh *fib6_nh)
613 struct __rt6_probe_work *work = NULL;
614 const struct in6_addr *nh_gw;
615 unsigned long last_probe;
616 struct neighbour *neigh;
617 struct net_device *dev;
618 struct inet6_dev *idev;
621 * Okay, this does not seem to be appropriate
622 * for now, however, we need to check if it
623 * is really so; aka Router Reachability Probing.
625 * Router Reachability Probe MUST be rate-limited
626 * to no more than one per minute.
628 if (!fib6_nh->fib_nh_gw_family)
631 nh_gw = &fib6_nh->fib_nh_gw6;
632 dev = fib6_nh->fib_nh_dev;
634 last_probe = READ_ONCE(fib6_nh->last_probe);
635 idev = __in6_dev_get(dev);
636 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
638 if (neigh->nud_state & NUD_VALID)
641 write_lock(&neigh->lock);
642 if (!(neigh->nud_state & NUD_VALID) &&
644 neigh->updated + idev->cnf.rtr_probe_interval)) {
645 work = kmalloc(sizeof(*work), GFP_ATOMIC);
647 __neigh_set_probe_once(neigh);
649 write_unlock(&neigh->lock);
650 } else if (time_after(jiffies, last_probe +
651 idev->cnf.rtr_probe_interval)) {
652 work = kmalloc(sizeof(*work), GFP_ATOMIC);
655 if (!work || cmpxchg(&fib6_nh->last_probe,
656 last_probe, jiffies) != last_probe) {
659 INIT_WORK(&work->work, rt6_probe_deferred);
660 work->target = *nh_gw;
663 schedule_work(&work->work);
667 rcu_read_unlock_bh();
670 static inline void rt6_probe(struct fib6_nh *fib6_nh)
676 * Default Router Selection (RFC 2461 6.3.6)
678 static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
680 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
681 struct neighbour *neigh;
684 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
685 &fib6_nh->fib_nh_gw6);
687 read_lock(&neigh->lock);
688 if (neigh->nud_state & NUD_VALID)
689 ret = RT6_NUD_SUCCEED;
690 #ifdef CONFIG_IPV6_ROUTER_PREF
691 else if (!(neigh->nud_state & NUD_FAILED))
692 ret = RT6_NUD_SUCCEED;
694 ret = RT6_NUD_FAIL_PROBE;
696 read_unlock(&neigh->lock);
698 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
699 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
701 rcu_read_unlock_bh();
706 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
711 if (!oif || nh->fib_nh_dev->ifindex == oif)
714 if (!m && (strict & RT6_LOOKUP_F_IFACE))
715 return RT6_NUD_FAIL_HARD;
716 #ifdef CONFIG_IPV6_ROUTER_PREF
717 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
719 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
720 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
721 int n = rt6_check_neigh(nh);
728 static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
729 int oif, int strict, int *mpri, bool *do_rr)
731 bool match_do_rr = false;
735 if (nh->fib_nh_flags & RTNH_F_DEAD)
738 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
739 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
740 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
743 m = rt6_score_route(nh, fib6_flags, oif, strict);
744 if (m == RT6_NUD_FAIL_DO_RR) {
746 m = 0; /* lowest valid score */
747 } else if (m == RT6_NUD_FAIL_HARD) {
751 if (strict & RT6_LOOKUP_F_REACHABLE)
754 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
756 *do_rr = match_do_rr;
764 struct fib6_nh_frl_arg {
773 static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
775 struct fib6_nh_frl_arg *arg = _arg;
778 return find_match(nh, arg->flags, arg->oif, arg->strict,
779 arg->mpri, arg->do_rr);
782 static void __find_rr_leaf(struct fib6_info *f6i_start,
783 struct fib6_info *nomatch, u32 metric,
784 struct fib6_result *res, struct fib6_info **cont,
785 int oif, int strict, bool *do_rr, int *mpri)
787 struct fib6_info *f6i;
789 for (f6i = f6i_start;
790 f6i && f6i != nomatch;
791 f6i = rcu_dereference(f6i->fib6_next)) {
792 bool matched = false;
795 if (cont && f6i->fib6_metric != metric) {
800 if (fib6_check_expired(f6i))
803 if (unlikely(f6i->nh)) {
804 struct fib6_nh_frl_arg arg = {
805 .flags = f6i->fib6_flags,
812 if (nexthop_is_blackhole(f6i->nh)) {
813 res->fib6_flags = RTF_REJECT;
814 res->fib6_type = RTN_BLACKHOLE;
816 res->nh = nexthop_fib6_nh(f6i->nh);
819 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
826 if (find_match(nh, f6i->fib6_flags, oif, strict,
833 res->fib6_flags = f6i->fib6_flags;
834 res->fib6_type = f6i->fib6_type;
839 static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
840 struct fib6_info *rr_head, int oif, int strict,
841 bool *do_rr, struct fib6_result *res)
843 u32 metric = rr_head->fib6_metric;
844 struct fib6_info *cont = NULL;
847 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
848 oif, strict, do_rr, &mpri);
850 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
851 oif, strict, do_rr, &mpri);
853 if (res->f6i || !cont)
856 __find_rr_leaf(cont, NULL, metric, res, NULL,
857 oif, strict, do_rr, &mpri);
860 static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
861 struct fib6_result *res, int strict)
863 struct fib6_info *leaf = rcu_dereference(fn->leaf);
864 struct fib6_info *rt0;
868 /* make sure this function or its helpers sets f6i */
871 if (!leaf || leaf == net->ipv6.fib6_null_entry)
874 rt0 = rcu_dereference(fn->rr_ptr);
878 /* Double check to make sure fn is not an intermediate node
879 * and fn->leaf does not points to its child's leaf
880 * (This might happen if all routes under fn are deleted from
881 * the tree and fib6_repair_tree() is called on the node.)
883 key_plen = rt0->fib6_dst.plen;
884 #ifdef CONFIG_IPV6_SUBTREES
885 if (rt0->fib6_src.plen)
886 key_plen = rt0->fib6_src.plen;
888 if (fn->fn_bit != key_plen)
891 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
893 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
895 /* no entries matched; do round-robin */
896 if (!next || next->fib6_metric != rt0->fib6_metric)
900 spin_lock_bh(&leaf->fib6_table->tb6_lock);
901 /* make sure next is not being deleted from the tree */
903 rcu_assign_pointer(fn->rr_ptr, next);
904 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
910 res->f6i = net->ipv6.fib6_null_entry;
911 res->nh = res->f6i->fib6_nh;
912 res->fib6_flags = res->f6i->fib6_flags;
913 res->fib6_type = res->f6i->fib6_type;
917 static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
919 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
920 res->nh->fib_nh_gw_family;
923 #ifdef CONFIG_IPV6_ROUTE_INFO
924 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
925 const struct in6_addr *gwaddr)
927 struct net *net = dev_net(dev);
928 struct route_info *rinfo = (struct route_info *) opt;
929 struct in6_addr prefix_buf, *prefix;
931 unsigned long lifetime;
932 struct fib6_info *rt;
934 if (len < sizeof(struct route_info)) {
938 /* Sanity check for prefix_len and length */
939 if (rinfo->length > 3) {
941 } else if (rinfo->prefix_len > 128) {
943 } else if (rinfo->prefix_len > 64) {
944 if (rinfo->length < 2) {
947 } else if (rinfo->prefix_len > 0) {
948 if (rinfo->length < 1) {
953 pref = rinfo->route_pref;
954 if (pref == ICMPV6_ROUTER_PREF_INVALID)
957 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
959 if (rinfo->length == 3)
960 prefix = (struct in6_addr *)rinfo->prefix;
962 /* this function is safe */
963 ipv6_addr_prefix(&prefix_buf,
964 (struct in6_addr *)rinfo->prefix,
966 prefix = &prefix_buf;
969 if (rinfo->prefix_len == 0)
970 rt = rt6_get_dflt_router(net, gwaddr, dev);
972 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
975 if (rt && !lifetime) {
976 ip6_del_rt(net, rt, false);
981 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
984 rt->fib6_flags = RTF_ROUTEINFO |
985 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
988 if (!addrconf_finite_timeout(lifetime))
989 fib6_clean_expires(rt);
991 fib6_set_expires(rt, jiffies + HZ * lifetime);
993 fib6_info_release(rt);
1000 * Misc support functions
1003 /* called with rcu_lock held */
1004 static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1006 struct net_device *dev = res->nh->fib_nh_dev;
1008 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1009 /* for copies of local routes, dst->dev needs to be the
1010 * device if it is a master device, the master device if
1011 * device is enslaved, and the loopback as the default
1013 if (netif_is_l3_slave(dev) &&
1014 !rt6_need_strict(&res->f6i->fib6_dst.addr))
1015 dev = l3mdev_master_dev_rcu(dev);
1016 else if (!netif_is_l3_master(dev))
1017 dev = dev_net(dev)->loopback_dev;
1018 /* last case is netif_is_l3_master(dev) is true in which
1019 * case we want dev returned to be dev
1026 static const int fib6_prop[RTN_MAX + 1] = {
1030 [RTN_BROADCAST] = 0,
1032 [RTN_MULTICAST] = 0,
1033 [RTN_BLACKHOLE] = -EINVAL,
1034 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1035 [RTN_PROHIBIT] = -EACCES,
1036 [RTN_THROW] = -EAGAIN,
1037 [RTN_NAT] = -EINVAL,
1038 [RTN_XRESOLVE] = -EINVAL,
1041 static int ip6_rt_type_to_error(u8 fib6_type)
1043 return fib6_prop[fib6_type];
1046 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1048 unsigned short flags = 0;
1050 if (rt->dst_nocount)
1051 flags |= DST_NOCOUNT;
1052 if (rt->dst_nopolicy)
1053 flags |= DST_NOPOLICY;
1058 static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1060 rt->dst.error = ip6_rt_type_to_error(fib6_type);
1062 switch (fib6_type) {
1064 rt->dst.output = dst_discard_out;
1065 rt->dst.input = dst_discard;
1068 rt->dst.output = ip6_pkt_prohibit_out;
1069 rt->dst.input = ip6_pkt_prohibit;
1072 case RTN_UNREACHABLE:
1074 rt->dst.output = ip6_pkt_discard_out;
1075 rt->dst.input = ip6_pkt_discard;
1080 static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1082 struct fib6_info *f6i = res->f6i;
1084 if (res->fib6_flags & RTF_REJECT) {
1085 ip6_rt_init_dst_reject(rt, res->fib6_type);
1090 rt->dst.output = ip6_output;
1092 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1093 rt->dst.input = ip6_input;
1094 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1095 rt->dst.input = ip6_mc_input;
1097 rt->dst.input = ip6_forward;
1100 if (res->nh->fib_nh_lws) {
1101 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1102 lwtunnel_set_redirect(&rt->dst);
1105 rt->dst.lastuse = jiffies;
1108 /* Caller must already hold reference to @from */
1109 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1111 rt->rt6i_flags &= ~RTF_EXPIRES;
1112 rcu_assign_pointer(rt->from, from);
1113 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1116 /* Caller must already hold reference to f6i in result */
1117 static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1119 const struct fib6_nh *nh = res->nh;
1120 const struct net_device *dev = nh->fib_nh_dev;
1121 struct fib6_info *f6i = res->f6i;
1123 ip6_rt_init_dst(rt, res);
1125 rt->rt6i_dst = f6i->fib6_dst;
1126 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1127 rt->rt6i_flags = res->fib6_flags;
1128 if (nh->fib_nh_gw_family) {
1129 rt->rt6i_gateway = nh->fib_nh_gw6;
1130 rt->rt6i_flags |= RTF_GATEWAY;
1132 rt6_set_from(rt, f6i);
1133 #ifdef CONFIG_IPV6_SUBTREES
1134 rt->rt6i_src = f6i->fib6_src;
1138 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1139 struct in6_addr *saddr)
1141 struct fib6_node *pn, *sn;
1143 if (fn->fn_flags & RTN_TL_ROOT)
1145 pn = rcu_dereference(fn->parent);
1146 sn = FIB6_SUBTREE(pn);
1148 fn = fib6_node_lookup(sn, NULL, saddr);
1151 if (fn->fn_flags & RTN_RTINFO)
1156 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1158 struct rt6_info *rt = *prt;
1160 if (dst_hold_safe(&rt->dst))
1163 rt = net->ipv6.ip6_null_entry;
1172 /* called with rcu_lock held */
1173 static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1175 struct net_device *dev = res->nh->fib_nh_dev;
1176 struct fib6_info *f6i = res->f6i;
1177 unsigned short flags;
1178 struct rt6_info *nrt;
1180 if (!fib6_info_hold_safe(f6i))
1183 flags = fib6_info_dst_flags(f6i);
1184 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1186 fib6_info_release(f6i);
1190 ip6_rt_copy_init(nrt, res);
1194 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1195 dst_hold(&nrt->dst);
1199 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
1200 struct fib6_table *table,
1202 const struct sk_buff *skb,
1205 struct fib6_result res = {};
1206 struct fib6_node *fn;
1207 struct rt6_info *rt;
1209 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1210 flags &= ~RT6_LOOKUP_F_IFACE;
1213 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1215 res.f6i = rcu_dereference(fn->leaf);
1217 res.f6i = net->ipv6.fib6_null_entry;
1219 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1222 if (res.f6i == net->ipv6.fib6_null_entry) {
1223 fn = fib6_backtrack(fn, &fl6->saddr);
1227 rt = net->ipv6.ip6_null_entry;
1230 } else if (res.fib6_flags & RTF_REJECT) {
1234 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1235 fl6->flowi6_oif != 0, skb, flags);
1237 /* Search through exception table */
1238 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1240 if (ip6_hold_safe(net, &rt))
1241 dst_use_noref(&rt->dst, jiffies);
1244 rt = ip6_create_rt_rcu(&res);
1248 trace_fib6_table_lookup(net, &res, table, fl6);
1255 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1256 const struct sk_buff *skb, int flags)
1258 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1260 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1262 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1263 const struct in6_addr *saddr, int oif,
1264 const struct sk_buff *skb, int strict)
1266 struct flowi6 fl6 = {
1270 struct dst_entry *dst;
1271 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1274 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1275 flags |= RT6_LOOKUP_F_HAS_SADDR;
1278 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1279 if (dst->error == 0)
1280 return (struct rt6_info *) dst;
1286 EXPORT_SYMBOL(rt6_lookup);
1288 /* ip6_ins_rt is called with FREE table->tb6_lock.
1289 * It takes new route entry, the addition fails by any reason the
1290 * route is released.
1291 * Caller must hold dst before calling it.
1294 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1295 struct netlink_ext_ack *extack)
1298 struct fib6_table *table;
1300 table = rt->fib6_table;
1301 spin_lock_bh(&table->tb6_lock);
1302 err = fib6_add(&table->tb6_root, rt, info, extack);
1303 spin_unlock_bh(&table->tb6_lock);
1308 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1310 struct nl_info info = { .nl_net = net, };
1312 return __ip6_ins_rt(rt, &info, NULL);
1315 static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1316 const struct in6_addr *daddr,
1317 const struct in6_addr *saddr)
1319 struct fib6_info *f6i = res->f6i;
1320 struct net_device *dev;
1321 struct rt6_info *rt;
1327 if (!fib6_info_hold_safe(f6i))
1330 dev = ip6_rt_get_dev_rcu(res);
1331 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1333 fib6_info_release(f6i);
1337 ip6_rt_copy_init(rt, res);
1338 rt->rt6i_flags |= RTF_CACHE;
1339 rt->rt6i_dst.addr = *daddr;
1340 rt->rt6i_dst.plen = 128;
1342 if (!rt6_is_gw_or_nonexthop(res)) {
1343 if (f6i->fib6_dst.plen != 128 &&
1344 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1345 rt->rt6i_flags |= RTF_ANYCAST;
1346 #ifdef CONFIG_IPV6_SUBTREES
1347 if (rt->rt6i_src.plen && saddr) {
1348 rt->rt6i_src.addr = *saddr;
1349 rt->rt6i_src.plen = 128;
1357 static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1359 struct fib6_info *f6i = res->f6i;
1360 unsigned short flags = fib6_info_dst_flags(f6i);
1361 struct net_device *dev;
1362 struct rt6_info *pcpu_rt;
1364 if (!fib6_info_hold_safe(f6i))
1368 dev = ip6_rt_get_dev_rcu(res);
1369 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
1372 fib6_info_release(f6i);
1375 ip6_rt_copy_init(pcpu_rt, res);
1376 pcpu_rt->rt6i_flags |= RTF_PCPU;
1379 pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1384 static bool rt6_is_valid(const struct rt6_info *rt6)
1386 return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1389 /* It should be called with rcu_read_lock() acquired */
1390 static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1392 struct rt6_info *pcpu_rt;
1394 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1396 if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1397 struct rt6_info *prev, **p;
1399 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1400 prev = xchg(p, NULL);
1402 dst_dev_put(&prev->dst);
1403 dst_release(&prev->dst);
1412 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1413 const struct fib6_result *res)
1415 struct rt6_info *pcpu_rt, *prev, **p;
1417 pcpu_rt = ip6_rt_pcpu_alloc(res);
1421 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1422 prev = cmpxchg(p, NULL, pcpu_rt);
1425 if (res->f6i->fib6_destroying) {
1426 struct fib6_info *from;
1428 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1429 fib6_info_release(from);
1435 /* exception hash table implementation
1437 static DEFINE_SPINLOCK(rt6_exception_lock);
1439 /* Remove rt6_ex from hash table and free the memory
1440 * Caller must hold rt6_exception_lock
1442 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1443 struct rt6_exception *rt6_ex)
1445 struct fib6_info *from;
1448 if (!bucket || !rt6_ex)
1451 net = dev_net(rt6_ex->rt6i->dst.dev);
1452 net->ipv6.rt6_stats->fib_rt_cache--;
1454 /* purge completely the exception to allow releasing the held resources:
1455 * some [sk] cache may keep the dst around for unlimited time
1457 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
1458 fib6_info_release(from);
1459 dst_dev_put(&rt6_ex->rt6i->dst);
1461 hlist_del_rcu(&rt6_ex->hlist);
1462 dst_release(&rt6_ex->rt6i->dst);
1463 kfree_rcu(rt6_ex, rcu);
1464 WARN_ON_ONCE(!bucket->depth);
1468 /* Remove oldest rt6_ex in bucket and free the memory
1469 * Caller must hold rt6_exception_lock
1471 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1473 struct rt6_exception *rt6_ex, *oldest = NULL;
1478 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1479 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1482 rt6_remove_exception(bucket, oldest);
1485 static u32 rt6_exception_hash(const struct in6_addr *dst,
1486 const struct in6_addr *src)
1488 static siphash_key_t rt6_exception_key __read_mostly;
1490 struct in6_addr dst;
1491 struct in6_addr src;
1492 } __aligned(SIPHASH_ALIGNMENT) combined = {
1497 net_get_random_once(&rt6_exception_key, sizeof(rt6_exception_key));
1499 #ifdef CONFIG_IPV6_SUBTREES
1501 combined.src = *src;
1503 val = siphash(&combined, sizeof(combined), &rt6_exception_key);
1505 return hash_64(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1508 /* Helper function to find the cached rt in the hash table
1509 * and update bucket pointer to point to the bucket for this
1510 * (daddr, saddr) pair
1511 * Caller must hold rt6_exception_lock
1513 static struct rt6_exception *
1514 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1515 const struct in6_addr *daddr,
1516 const struct in6_addr *saddr)
1518 struct rt6_exception *rt6_ex;
1521 if (!(*bucket) || !daddr)
1524 hval = rt6_exception_hash(daddr, saddr);
1527 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1528 struct rt6_info *rt6 = rt6_ex->rt6i;
1529 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1531 #ifdef CONFIG_IPV6_SUBTREES
1532 if (matched && saddr)
1533 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1541 /* Helper function to find the cached rt in the hash table
1542 * and update bucket pointer to point to the bucket for this
1543 * (daddr, saddr) pair
1544 * Caller must hold rcu_read_lock()
1546 static struct rt6_exception *
1547 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1548 const struct in6_addr *daddr,
1549 const struct in6_addr *saddr)
1551 struct rt6_exception *rt6_ex;
1554 WARN_ON_ONCE(!rcu_read_lock_held());
1556 if (!(*bucket) || !daddr)
1559 hval = rt6_exception_hash(daddr, saddr);
1562 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1563 struct rt6_info *rt6 = rt6_ex->rt6i;
1564 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1566 #ifdef CONFIG_IPV6_SUBTREES
1567 if (matched && saddr)
1568 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1576 static unsigned int fib6_mtu(const struct fib6_result *res)
1578 const struct fib6_nh *nh = res->nh;
1581 if (res->f6i->fib6_pmtu) {
1582 mtu = res->f6i->fib6_pmtu;
1584 struct net_device *dev = nh->fib_nh_dev;
1585 struct inet6_dev *idev;
1588 idev = __in6_dev_get(dev);
1589 mtu = idev->cnf.mtu6;
1593 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1595 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1598 #define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1600 /* used when the flushed bit is not relevant, only access to the bucket
1601 * (ie., all bucket users except rt6_insert_exception);
1603 * called under rcu lock; sometimes called with rt6_exception_lock held
1606 struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1609 struct rt6_exception_bucket *bucket;
1612 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1613 lockdep_is_held(lock));
1615 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1617 /* remove bucket flushed bit if set */
1619 unsigned long p = (unsigned long)bucket;
1621 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1622 bucket = (struct rt6_exception_bucket *)p;
1628 static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1630 unsigned long p = (unsigned long)bucket;
1632 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1635 /* called with rt6_exception_lock held */
1636 static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1639 struct rt6_exception_bucket *bucket;
1642 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1643 lockdep_is_held(lock));
1645 p = (unsigned long)bucket;
1646 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1647 bucket = (struct rt6_exception_bucket *)p;
1648 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1651 static int rt6_insert_exception(struct rt6_info *nrt,
1652 const struct fib6_result *res)
1654 struct net *net = dev_net(nrt->dst.dev);
1655 struct rt6_exception_bucket *bucket;
1656 struct fib6_info *f6i = res->f6i;
1657 struct in6_addr *src_key = NULL;
1658 struct rt6_exception *rt6_ex;
1659 struct fib6_nh *nh = res->nh;
1662 spin_lock_bh(&rt6_exception_lock);
1664 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1665 lockdep_is_held(&rt6_exception_lock));
1667 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1673 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1674 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1679 #ifdef CONFIG_IPV6_SUBTREES
1680 /* fib6_src.plen != 0 indicates f6i is in subtree
1681 * and exception table is indexed by a hash of
1682 * both fib6_dst and fib6_src.
1683 * Otherwise, the exception table is indexed by
1684 * a hash of only fib6_dst.
1686 if (f6i->fib6_src.plen)
1687 src_key = &nrt->rt6i_src.addr;
1689 /* rt6_mtu_change() might lower mtu on f6i.
1690 * Only insert this exception route if its mtu
1691 * is less than f6i's mtu value.
1693 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1698 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1701 rt6_remove_exception(bucket, rt6_ex);
1703 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1709 rt6_ex->stamp = jiffies;
1710 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1712 net->ipv6.rt6_stats->fib_rt_cache++;
1714 if (bucket->depth > FIB6_MAX_DEPTH)
1715 rt6_exception_remove_oldest(bucket);
1718 spin_unlock_bh(&rt6_exception_lock);
1720 /* Update fn->fn_sernum to invalidate all cached dst */
1722 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1723 fib6_update_sernum(net, f6i);
1724 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1725 fib6_force_start_gc(net);
1731 static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1733 struct rt6_exception_bucket *bucket;
1734 struct rt6_exception *rt6_ex;
1735 struct hlist_node *tmp;
1738 spin_lock_bh(&rt6_exception_lock);
1740 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1744 /* Prevent rt6_insert_exception() to recreate the bucket list */
1746 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1748 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1749 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1751 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1752 rt6_remove_exception(bucket, rt6_ex);
1754 WARN_ON_ONCE(!from && bucket->depth);
1758 spin_unlock_bh(&rt6_exception_lock);
1761 static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1763 struct fib6_info *f6i = arg;
1765 fib6_nh_flush_exceptions(nh, f6i);
1770 void rt6_flush_exceptions(struct fib6_info *f6i)
1773 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1776 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1779 /* Find cached rt in the hash table inside passed in rt
1780 * Caller has to hold rcu_read_lock()
1782 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1783 const struct in6_addr *daddr,
1784 const struct in6_addr *saddr)
1786 const struct in6_addr *src_key = NULL;
1787 struct rt6_exception_bucket *bucket;
1788 struct rt6_exception *rt6_ex;
1789 struct rt6_info *ret = NULL;
1791 #ifdef CONFIG_IPV6_SUBTREES
1792 /* fib6i_src.plen != 0 indicates f6i is in subtree
1793 * and exception table is indexed by a hash of
1794 * both fib6_dst and fib6_src.
1795 * However, the src addr used to create the hash
1796 * might not be exactly the passed in saddr which
1797 * is a /128 addr from the flow.
1798 * So we need to use f6i->fib6_src to redo lookup
1799 * if the passed in saddr does not find anything.
1800 * (See the logic in ip6_rt_cache_alloc() on how
1801 * rt->rt6i_src is updated.)
1803 if (res->f6i->fib6_src.plen)
1807 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1808 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1810 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1813 #ifdef CONFIG_IPV6_SUBTREES
1814 /* Use fib6_src as src_key and redo lookup */
1815 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1816 src_key = &res->f6i->fib6_src.addr;
1824 /* Remove the passed in cached rt from the hash table that contains it */
1825 static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1826 const struct rt6_info *rt)
1828 const struct in6_addr *src_key = NULL;
1829 struct rt6_exception_bucket *bucket;
1830 struct rt6_exception *rt6_ex;
1833 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1836 spin_lock_bh(&rt6_exception_lock);
1837 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1839 #ifdef CONFIG_IPV6_SUBTREES
1840 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1841 * and exception table is indexed by a hash of
1842 * both rt6i_dst and rt6i_src.
1843 * Otherwise, the exception table is indexed by
1844 * a hash of only rt6i_dst.
1847 src_key = &rt->rt6i_src.addr;
1849 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1853 rt6_remove_exception(bucket, rt6_ex);
1859 spin_unlock_bh(&rt6_exception_lock);
1863 struct fib6_nh_excptn_arg {
1864 struct rt6_info *rt;
1868 static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1870 struct fib6_nh_excptn_arg *arg = _arg;
1873 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1880 static int rt6_remove_exception_rt(struct rt6_info *rt)
1882 struct fib6_info *from;
1884 from = rcu_dereference(rt->from);
1885 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1889 struct fib6_nh_excptn_arg arg = {
1891 .plen = from->fib6_src.plen
1895 /* rc = 1 means an entry was found */
1896 rc = nexthop_for_each_fib6_nh(from->nh,
1897 rt6_nh_remove_exception_rt,
1899 return rc ? 0 : -ENOENT;
1902 return fib6_nh_remove_exception(from->fib6_nh,
1903 from->fib6_src.plen, rt);
1906 /* Find rt6_ex which contains the passed in rt cache and
1909 static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1910 const struct rt6_info *rt)
1912 const struct in6_addr *src_key = NULL;
1913 struct rt6_exception_bucket *bucket;
1914 struct rt6_exception *rt6_ex;
1916 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1917 #ifdef CONFIG_IPV6_SUBTREES
1918 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1919 * and exception table is indexed by a hash of
1920 * both rt6i_dst and rt6i_src.
1921 * Otherwise, the exception table is indexed by
1922 * a hash of only rt6i_dst.
1925 src_key = &rt->rt6i_src.addr;
1927 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1929 rt6_ex->stamp = jiffies;
1932 struct fib6_nh_match_arg {
1933 const struct net_device *dev;
1934 const struct in6_addr *gw;
1935 struct fib6_nh *match;
1938 /* determine if fib6_nh has given device and gateway */
1939 static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1941 struct fib6_nh_match_arg *arg = _arg;
1943 if (arg->dev != nh->fib_nh_dev ||
1944 (arg->gw && !nh->fib_nh_gw_family) ||
1945 (!arg->gw && nh->fib_nh_gw_family) ||
1946 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1951 /* found a match, break the loop */
1955 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1957 struct fib6_info *from;
1958 struct fib6_nh *fib6_nh;
1962 from = rcu_dereference(rt->from);
1963 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1967 struct fib6_nh_match_arg arg = {
1969 .gw = &rt->rt6i_gateway,
1972 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1976 fib6_nh = arg.match;
1978 fib6_nh = from->fib6_nh;
1980 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1985 static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1986 struct rt6_info *rt, int mtu)
1988 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1989 * lowest MTU in the path: always allow updating the route PMTU to
1990 * reflect PMTU decreases.
1992 * If the new MTU is higher, and the route PMTU is equal to the local
1993 * MTU, this means the old MTU is the lowest in the path, so allow
1994 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1998 if (dst_mtu(&rt->dst) >= mtu)
2001 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
2007 static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
2008 const struct fib6_nh *nh, int mtu)
2010 struct rt6_exception_bucket *bucket;
2011 struct rt6_exception *rt6_ex;
2014 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2018 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2019 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2020 struct rt6_info *entry = rt6_ex->rt6i;
2022 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2023 * route), the metrics of its rt->from have already
2026 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2027 rt6_mtu_change_route_allowed(idev, entry, mtu))
2028 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2034 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2036 static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2037 const struct in6_addr *gateway)
2039 struct rt6_exception_bucket *bucket;
2040 struct rt6_exception *rt6_ex;
2041 struct hlist_node *tmp;
2044 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2047 spin_lock_bh(&rt6_exception_lock);
2048 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2050 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2051 hlist_for_each_entry_safe(rt6_ex, tmp,
2052 &bucket->chain, hlist) {
2053 struct rt6_info *entry = rt6_ex->rt6i;
2055 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2056 RTF_CACHE_GATEWAY &&
2057 ipv6_addr_equal(gateway,
2058 &entry->rt6i_gateway)) {
2059 rt6_remove_exception(bucket, rt6_ex);
2066 spin_unlock_bh(&rt6_exception_lock);
2069 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2070 struct rt6_exception *rt6_ex,
2071 struct fib6_gc_args *gc_args,
2074 struct rt6_info *rt = rt6_ex->rt6i;
2076 /* we are pruning and obsoleting aged-out and non gateway exceptions
2077 * even if others have still references to them, so that on next
2078 * dst_check() such references can be dropped.
2079 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2080 * expired, independently from their aging, as per RFC 8201 section 4
2082 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2083 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2084 RT6_TRACE("aging clone %p\n", rt);
2085 rt6_remove_exception(bucket, rt6_ex);
2088 } else if (time_after(jiffies, rt->dst.expires)) {
2089 RT6_TRACE("purging expired route %p\n", rt);
2090 rt6_remove_exception(bucket, rt6_ex);
2094 if (rt->rt6i_flags & RTF_GATEWAY) {
2095 struct neighbour *neigh;
2097 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2099 if (!(neigh && (neigh->flags & NTF_ROUTER))) {
2100 RT6_TRACE("purging route %p via non-router but gateway\n",
2102 rt6_remove_exception(bucket, rt6_ex);
2110 static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2111 struct fib6_gc_args *gc_args,
2114 struct rt6_exception_bucket *bucket;
2115 struct rt6_exception *rt6_ex;
2116 struct hlist_node *tmp;
2119 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2123 spin_lock(&rt6_exception_lock);
2124 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2126 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2127 hlist_for_each_entry_safe(rt6_ex, tmp,
2128 &bucket->chain, hlist) {
2129 rt6_age_examine_exception(bucket, rt6_ex,
2135 spin_unlock(&rt6_exception_lock);
2136 rcu_read_unlock_bh();
2139 struct fib6_nh_age_excptn_arg {
2140 struct fib6_gc_args *gc_args;
2144 static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2146 struct fib6_nh_age_excptn_arg *arg = _arg;
2148 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2152 void rt6_age_exceptions(struct fib6_info *f6i,
2153 struct fib6_gc_args *gc_args,
2157 struct fib6_nh_age_excptn_arg arg = {
2162 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2165 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2169 /* must be called with rcu lock held */
2170 int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2171 struct flowi6 *fl6, struct fib6_result *res, int strict)
2173 struct fib6_node *fn, *saved_fn;
2175 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2178 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2182 rt6_select(net, fn, oif, res, strict);
2183 if (res->f6i == net->ipv6.fib6_null_entry) {
2184 fn = fib6_backtrack(fn, &fl6->saddr);
2186 goto redo_rt6_select;
2187 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2188 /* also consider unreachable route */
2189 strict &= ~RT6_LOOKUP_F_REACHABLE;
2191 goto redo_rt6_select;
2195 trace_fib6_table_lookup(net, res, table, fl6);
2200 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2201 int oif, struct flowi6 *fl6,
2202 const struct sk_buff *skb, int flags)
2204 struct fib6_result res = {};
2205 struct rt6_info *rt = NULL;
2208 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2209 !rcu_read_lock_held());
2211 strict |= flags & RT6_LOOKUP_F_IFACE;
2212 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2213 if (net->ipv6.devconf_all->forwarding == 0)
2214 strict |= RT6_LOOKUP_F_REACHABLE;
2218 fib6_table_lookup(net, table, oif, fl6, &res, strict);
2219 if (res.f6i == net->ipv6.fib6_null_entry)
2222 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2224 /*Search through exception table */
2225 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2228 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2229 !res.nh->fib_nh_gw_family)) {
2230 /* Create a RTF_CACHE clone which will not be
2231 * owned by the fib6 tree. It is for the special case where
2232 * the daddr in the skb during the neighbor look-up is different
2233 * from the fl6->daddr used to look-up route here.
2235 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2238 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2239 * As rt6_uncached_list_add() does not consume refcnt,
2240 * this refcnt is always returned to the caller even
2241 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2243 rt6_uncached_list_add(rt);
2244 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
2250 /* Get a percpu copy */
2252 rt = rt6_get_pcpu_route(&res);
2255 rt = rt6_make_pcpu_route(net, &res);
2261 rt = net->ipv6.ip6_null_entry;
2262 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2263 ip6_hold_safe(net, &rt);
2268 EXPORT_SYMBOL_GPL(ip6_pol_route);
2270 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
2271 struct fib6_table *table,
2273 const struct sk_buff *skb,
2276 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2279 struct dst_entry *ip6_route_input_lookup(struct net *net,
2280 struct net_device *dev,
2282 const struct sk_buff *skb,
2285 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2286 flags |= RT6_LOOKUP_F_IFACE;
2288 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2290 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2292 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2293 struct flow_keys *keys,
2294 struct flow_keys *flkeys)
2296 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2297 const struct ipv6hdr *key_iph = outer_iph;
2298 struct flow_keys *_flkeys = flkeys;
2299 const struct ipv6hdr *inner_iph;
2300 const struct icmp6hdr *icmph;
2301 struct ipv6hdr _inner_iph;
2302 struct icmp6hdr _icmph;
2304 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2307 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2308 sizeof(_icmph), &_icmph);
2312 if (!icmpv6_is_err(icmph->icmp6_type))
2315 inner_iph = skb_header_pointer(skb,
2316 skb_transport_offset(skb) + sizeof(*icmph),
2317 sizeof(_inner_iph), &_inner_iph);
2321 key_iph = inner_iph;
2325 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2326 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2327 keys->tags.flow_label = _flkeys->tags.flow_label;
2328 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2330 keys->addrs.v6addrs.src = key_iph->saddr;
2331 keys->addrs.v6addrs.dst = key_iph->daddr;
2332 keys->tags.flow_label = ip6_flowlabel(key_iph);
2333 keys->basic.ip_proto = key_iph->nexthdr;
2337 static u32 rt6_multipath_custom_hash_outer(const struct net *net,
2338 const struct sk_buff *skb,
2341 u32 hash_fields = ip6_multipath_hash_fields(net);
2342 struct flow_keys keys, hash_keys;
2344 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2347 memset(&hash_keys, 0, sizeof(hash_keys));
2348 skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_ENCAP);
2350 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2351 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2352 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2353 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2354 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2355 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2356 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2357 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2358 hash_keys.tags.flow_label = keys.tags.flow_label;
2359 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2360 hash_keys.ports.src = keys.ports.src;
2361 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2362 hash_keys.ports.dst = keys.ports.dst;
2364 *p_has_inner = !!(keys.control.flags & FLOW_DIS_ENCAPSULATION);
2365 return flow_hash_from_keys(&hash_keys);
2368 static u32 rt6_multipath_custom_hash_inner(const struct net *net,
2369 const struct sk_buff *skb,
2372 u32 hash_fields = ip6_multipath_hash_fields(net);
2373 struct flow_keys keys, hash_keys;
2375 /* We assume the packet carries an encapsulation, but if none was
2376 * encountered during dissection of the outer flow, then there is no
2377 * point in calling the flow dissector again.
2382 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_MASK))
2385 memset(&hash_keys, 0, sizeof(hash_keys));
2386 skb_flow_dissect_flow_keys(skb, &keys, 0);
2388 if (!(keys.control.flags & FLOW_DIS_ENCAPSULATION))
2391 if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2392 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2393 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2394 hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
2395 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2396 hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst;
2397 } else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2398 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2399 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2400 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2401 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2402 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2403 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL)
2404 hash_keys.tags.flow_label = keys.tags.flow_label;
2407 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO)
2408 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2409 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT)
2410 hash_keys.ports.src = keys.ports.src;
2411 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
2412 hash_keys.ports.dst = keys.ports.dst;
2414 return flow_hash_from_keys(&hash_keys);
2417 static u32 rt6_multipath_custom_hash_skb(const struct net *net,
2418 const struct sk_buff *skb)
2420 u32 mhash, mhash_inner;
2421 bool has_inner = true;
2423 mhash = rt6_multipath_custom_hash_outer(net, skb, &has_inner);
2424 mhash_inner = rt6_multipath_custom_hash_inner(net, skb, has_inner);
2426 return jhash_2words(mhash, mhash_inner, 0);
2429 static u32 rt6_multipath_custom_hash_fl6(const struct net *net,
2430 const struct flowi6 *fl6)
2432 u32 hash_fields = ip6_multipath_hash_fields(net);
2433 struct flow_keys hash_keys;
2435 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2438 memset(&hash_keys, 0, sizeof(hash_keys));
2439 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2440 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2441 hash_keys.addrs.v6addrs.src = fl6->saddr;
2442 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2443 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2444 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2445 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2446 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2447 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2448 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2449 hash_keys.ports.src = fl6->fl6_sport;
2450 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2451 hash_keys.ports.dst = fl6->fl6_dport;
2453 return flow_hash_from_keys(&hash_keys);
2456 /* if skb is set it will be used and fl6 can be NULL */
2457 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2458 const struct sk_buff *skb, struct flow_keys *flkeys)
2460 struct flow_keys hash_keys;
2463 switch (ip6_multipath_hash_policy(net)) {
2465 memset(&hash_keys, 0, sizeof(hash_keys));
2466 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2468 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2470 hash_keys.addrs.v6addrs.src = fl6->saddr;
2471 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2472 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2473 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2475 mhash = flow_hash_from_keys(&hash_keys);
2479 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2480 struct flow_keys keys;
2482 /* short-circuit if we already have L4 hash present */
2484 return skb_get_hash_raw(skb) >> 1;
2486 memset(&hash_keys, 0, sizeof(hash_keys));
2489 skb_flow_dissect_flow_keys(skb, &keys, flag);
2492 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2493 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2494 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2495 hash_keys.ports.src = flkeys->ports.src;
2496 hash_keys.ports.dst = flkeys->ports.dst;
2497 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2499 memset(&hash_keys, 0, sizeof(hash_keys));
2500 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2501 hash_keys.addrs.v6addrs.src = fl6->saddr;
2502 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2503 hash_keys.ports.src = fl6->fl6_sport;
2504 hash_keys.ports.dst = fl6->fl6_dport;
2505 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2507 mhash = flow_hash_from_keys(&hash_keys);
2510 memset(&hash_keys, 0, sizeof(hash_keys));
2511 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2513 struct flow_keys keys;
2516 skb_flow_dissect_flow_keys(skb, &keys, 0);
2520 /* Inner can be v4 or v6 */
2521 if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2522 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2523 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2524 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2525 } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2526 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2527 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2528 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2529 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2530 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2532 /* Same as case 0 */
2533 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2534 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2537 /* Same as case 0 */
2538 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2539 hash_keys.addrs.v6addrs.src = fl6->saddr;
2540 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2541 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2542 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2544 mhash = flow_hash_from_keys(&hash_keys);
2548 mhash = rt6_multipath_custom_hash_skb(net, skb);
2550 mhash = rt6_multipath_custom_hash_fl6(net, fl6);
2557 /* Called with rcu held */
2558 void ip6_route_input(struct sk_buff *skb)
2560 const struct ipv6hdr *iph = ipv6_hdr(skb);
2561 struct net *net = dev_net(skb->dev);
2562 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2563 struct ip_tunnel_info *tun_info;
2564 struct flowi6 fl6 = {
2565 .flowi6_iif = skb->dev->ifindex,
2566 .daddr = iph->daddr,
2567 .saddr = iph->saddr,
2568 .flowlabel = ip6_flowinfo(iph),
2569 .flowi6_mark = skb->mark,
2570 .flowi6_proto = iph->nexthdr,
2572 struct flow_keys *flkeys = NULL, _flkeys;
2574 tun_info = skb_tunnel_info(skb);
2575 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2576 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2578 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2581 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2582 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2584 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2588 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
2589 struct fib6_table *table,
2591 const struct sk_buff *skb,
2594 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2597 struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2598 const struct sock *sk,
2599 struct flowi6 *fl6, int flags)
2603 if (ipv6_addr_type(&fl6->daddr) &
2604 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2605 struct dst_entry *dst;
2607 /* This function does not take refcnt on the dst */
2608 dst = l3mdev_link_scope_lookup(net, fl6);
2613 fl6->flowi6_iif = LOOPBACK_IFINDEX;
2615 flags |= RT6_LOOKUP_F_DST_NOREF;
2616 any_src = ipv6_addr_any(&fl6->saddr);
2617 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2618 (fl6->flowi6_oif && any_src))
2619 flags |= RT6_LOOKUP_F_IFACE;
2622 flags |= RT6_LOOKUP_F_HAS_SADDR;
2624 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
2626 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2628 EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2630 struct dst_entry *ip6_route_output_flags(struct net *net,
2631 const struct sock *sk,
2635 struct dst_entry *dst;
2636 struct rt6_info *rt6;
2639 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2640 rt6 = (struct rt6_info *)dst;
2641 /* For dst cached in uncached_list, refcnt is already taken. */
2642 if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2643 dst = &net->ipv6.ip6_null_entry->dst;
2650 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2652 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2654 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2655 struct net_device *loopback_dev = net->loopback_dev;
2656 struct dst_entry *new = NULL;
2658 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
2659 DST_OBSOLETE_DEAD, 0);
2662 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2666 new->input = dst_discard;
2667 new->output = dst_discard_out;
2669 dst_copy_metrics(new, &ort->dst);
2671 rt->rt6i_idev = in6_dev_get(loopback_dev);
2672 rt->rt6i_gateway = ort->rt6i_gateway;
2673 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2675 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2676 #ifdef CONFIG_IPV6_SUBTREES
2677 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2681 dst_release(dst_orig);
2682 return new ? new : ERR_PTR(-ENOMEM);
2686 * Destination cache support functions
2689 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2693 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2696 if (fib6_check_expired(f6i))
2702 static struct dst_entry *rt6_check(struct rt6_info *rt,
2703 struct fib6_info *from,
2708 if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2709 rt_cookie != cookie)
2712 if (rt6_check_expired(rt))
2718 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2719 struct fib6_info *from,
2722 if (!__rt6_check_expired(rt) &&
2723 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2724 fib6_check(from, cookie))
2730 INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2733 struct dst_entry *dst_ret;
2734 struct fib6_info *from;
2735 struct rt6_info *rt;
2737 rt = container_of(dst, struct rt6_info, dst);
2740 return rt6_is_valid(rt) ? dst : NULL;
2744 /* All IPV6 dsts are created with ->obsolete set to the value
2745 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2746 * into this function always.
2749 from = rcu_dereference(rt->from);
2751 if (from && (rt->rt6i_flags & RTF_PCPU ||
2752 unlikely(!list_empty(&rt->rt6i_uncached))))
2753 dst_ret = rt6_dst_from_check(rt, from, cookie);
2755 dst_ret = rt6_check(rt, from, cookie);
2761 EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
2763 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2765 struct rt6_info *rt = (struct rt6_info *) dst;
2768 if (rt->rt6i_flags & RTF_CACHE) {
2770 if (rt6_check_expired(rt)) {
2771 rt6_remove_exception_rt(rt);
2783 static void ip6_link_failure(struct sk_buff *skb)
2785 struct rt6_info *rt;
2787 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2789 rt = (struct rt6_info *) skb_dst(skb);
2792 if (rt->rt6i_flags & RTF_CACHE) {
2793 rt6_remove_exception_rt(rt);
2795 struct fib6_info *from;
2796 struct fib6_node *fn;
2798 from = rcu_dereference(rt->from);
2800 fn = rcu_dereference(from->fib6_node);
2801 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2809 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2811 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2812 struct fib6_info *from;
2815 from = rcu_dereference(rt0->from);
2817 rt0->dst.expires = from->expires;
2821 dst_set_expires(&rt0->dst, timeout);
2822 rt0->rt6i_flags |= RTF_EXPIRES;
2825 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2827 struct net *net = dev_net(rt->dst.dev);
2829 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2830 rt->rt6i_flags |= RTF_MODIFIED;
2831 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2834 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2836 return !(rt->rt6i_flags & RTF_CACHE) &&
2837 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2840 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2841 const struct ipv6hdr *iph, u32 mtu,
2844 const struct in6_addr *daddr, *saddr;
2845 struct rt6_info *rt6 = (struct rt6_info *)dst;
2847 /* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2848 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2849 * [see also comment in rt6_mtu_change_route()]
2853 daddr = &iph->daddr;
2854 saddr = &iph->saddr;
2856 daddr = &sk->sk_v6_daddr;
2857 saddr = &inet6_sk(sk)->saddr;
2864 dst_confirm_neigh(dst, daddr);
2866 if (mtu < IPV6_MIN_MTU)
2868 if (mtu >= dst_mtu(dst))
2871 if (!rt6_cache_allowed_for_pmtu(rt6)) {
2872 rt6_do_update_pmtu(rt6, mtu);
2873 /* update rt6_ex->stamp for cache */
2874 if (rt6->rt6i_flags & RTF_CACHE)
2875 rt6_update_exception_stamp_rt(rt6);
2877 struct fib6_result res = {};
2878 struct rt6_info *nrt6;
2881 res.f6i = rcu_dereference(rt6->from);
2885 res.fib6_flags = res.f6i->fib6_flags;
2886 res.fib6_type = res.f6i->fib6_type;
2889 struct fib6_nh_match_arg arg = {
2891 .gw = &rt6->rt6i_gateway,
2894 nexthop_for_each_fib6_nh(res.f6i->nh,
2895 fib6_nh_find_match, &arg);
2897 /* fib6_info uses a nexthop that does not have fib6_nh
2898 * using the dst->dev + gw. Should be impossible.
2905 res.nh = res.f6i->fib6_nh;
2908 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2910 rt6_do_update_pmtu(nrt6, mtu);
2911 if (rt6_insert_exception(nrt6, &res))
2912 dst_release_immediate(&nrt6->dst);
2919 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2920 struct sk_buff *skb, u32 mtu,
2923 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2927 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2928 int oif, u32 mark, kuid_t uid)
2930 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2931 struct dst_entry *dst;
2932 struct flowi6 fl6 = {
2934 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2935 .daddr = iph->daddr,
2936 .saddr = iph->saddr,
2937 .flowlabel = ip6_flowinfo(iph),
2941 dst = ip6_route_output(net, NULL, &fl6);
2943 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
2946 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2948 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2950 int oif = sk->sk_bound_dev_if;
2951 struct dst_entry *dst;
2953 if (!oif && skb->dev)
2954 oif = l3mdev_master_ifindex(skb->dev);
2956 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
2958 dst = __sk_dst_get(sk);
2959 if (!dst || !dst->obsolete ||
2960 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2964 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2965 ip6_datagram_dst_update(sk, false);
2968 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2970 void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2971 const struct flowi6 *fl6)
2973 #ifdef CONFIG_IPV6_SUBTREES
2974 struct ipv6_pinfo *np = inet6_sk(sk);
2977 ip6_dst_store(sk, dst,
2978 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2979 &sk->sk_v6_daddr : NULL,
2980 #ifdef CONFIG_IPV6_SUBTREES
2981 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2987 static bool ip6_redirect_nh_match(const struct fib6_result *res,
2989 const struct in6_addr *gw,
2990 struct rt6_info **ret)
2992 const struct fib6_nh *nh = res->nh;
2994 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2995 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2998 /* rt_cache's gateway might be different from its 'parent'
2999 * in the case of an ip redirect.
3000 * So we keep searching in the exception table if the gateway
3003 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
3004 struct rt6_info *rt_cache;
3006 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
3008 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
3017 struct fib6_nh_rd_arg {
3018 struct fib6_result *res;
3020 const struct in6_addr *gw;
3021 struct rt6_info **ret;
3024 static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
3026 struct fib6_nh_rd_arg *arg = _arg;
3029 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
3032 /* Handle redirects */
3033 struct ip6rd_flowi {
3035 struct in6_addr gateway;
3038 INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
3039 struct fib6_table *table,
3041 const struct sk_buff *skb,
3044 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
3045 struct rt6_info *ret = NULL;
3046 struct fib6_result res = {};
3047 struct fib6_nh_rd_arg arg = {
3050 .gw = &rdfl->gateway,
3053 struct fib6_info *rt;
3054 struct fib6_node *fn;
3056 /* l3mdev_update_flow overrides oif if the device is enslaved; in
3057 * this case we must match on the real ingress device, so reset it
3059 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
3060 fl6->flowi6_oif = skb->dev->ifindex;
3062 /* Get the "current" route for this destination and
3063 * check if the redirect has come from appropriate router.
3065 * RFC 4861 specifies that redirects should only be
3066 * accepted if they come from the nexthop to the target.
3067 * Due to the way the routes are chosen, this notion
3068 * is a bit fuzzy and one might need to check all possible
3073 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
3075 for_each_fib6_node_rt_rcu(fn) {
3077 if (fib6_check_expired(rt))
3079 if (rt->fib6_flags & RTF_REJECT)
3081 if (unlikely(rt->nh)) {
3082 if (nexthop_is_blackhole(rt->nh))
3084 /* on match, res->nh is filled in and potentially ret */
3085 if (nexthop_for_each_fib6_nh(rt->nh,
3086 fib6_nh_redirect_match,
3090 res.nh = rt->fib6_nh;
3091 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
3098 rt = net->ipv6.fib6_null_entry;
3099 else if (rt->fib6_flags & RTF_REJECT) {
3100 ret = net->ipv6.ip6_null_entry;
3104 if (rt == net->ipv6.fib6_null_entry) {
3105 fn = fib6_backtrack(fn, &fl6->saddr);
3111 res.nh = rt->fib6_nh;
3114 ip6_hold_safe(net, &ret);
3116 res.fib6_flags = res.f6i->fib6_flags;
3117 res.fib6_type = res.f6i->fib6_type;
3118 ret = ip6_create_rt_rcu(&res);
3123 trace_fib6_table_lookup(net, &res, table, fl6);
3127 static struct dst_entry *ip6_route_redirect(struct net *net,
3128 const struct flowi6 *fl6,
3129 const struct sk_buff *skb,
3130 const struct in6_addr *gateway)
3132 int flags = RT6_LOOKUP_F_HAS_SADDR;
3133 struct ip6rd_flowi rdfl;
3136 rdfl.gateway = *gateway;
3138 return fib6_rule_lookup(net, &rdfl.fl6, skb,
3139 flags, __ip6_route_redirect);
3142 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3145 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3146 struct dst_entry *dst;
3147 struct flowi6 fl6 = {
3148 .flowi6_iif = LOOPBACK_IFINDEX,
3150 .flowi6_mark = mark,
3151 .daddr = iph->daddr,
3152 .saddr = iph->saddr,
3153 .flowlabel = ip6_flowinfo(iph),
3157 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3158 rt6_do_redirect(dst, NULL, skb);
3161 EXPORT_SYMBOL_GPL(ip6_redirect);
3163 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
3165 const struct ipv6hdr *iph = ipv6_hdr(skb);
3166 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3167 struct dst_entry *dst;
3168 struct flowi6 fl6 = {
3169 .flowi6_iif = LOOPBACK_IFINDEX,
3172 .saddr = iph->daddr,
3173 .flowi6_uid = sock_net_uid(net, NULL),
3176 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3177 rt6_do_redirect(dst, NULL, skb);
3181 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3183 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
3186 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3188 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3190 struct net_device *dev = dst->dev;
3191 unsigned int mtu = dst_mtu(dst);
3192 struct net *net = dev_net(dev);
3194 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3196 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3197 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3200 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3201 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3202 * IPV6_MAXPLEN is also valid and means: "any MSS,
3203 * rely only on pmtu discovery"
3205 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3210 INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
3212 return ip6_dst_mtu_maybe_forward(dst, false);
3214 EXPORT_INDIRECT_CALLABLE(ip6_mtu);
3217 * 1. mtu on route is locked - use it
3218 * 2. mtu from nexthop exception
3219 * 3. mtu from egress device
3221 * based on ip6_dst_mtu_forward and exception logic of
3222 * rt6_find_cached_rt; called with rcu_read_lock
3224 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3225 const struct in6_addr *daddr,
3226 const struct in6_addr *saddr)
3228 const struct fib6_nh *nh = res->nh;
3229 struct fib6_info *f6i = res->f6i;
3230 struct inet6_dev *idev;
3231 struct rt6_info *rt;
3234 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3235 mtu = f6i->fib6_pmtu;
3240 rt = rt6_find_cached_rt(res, daddr, saddr);
3242 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3244 struct net_device *dev = nh->fib_nh_dev;
3247 idev = __in6_dev_get(dev);
3248 if (idev && idev->cnf.mtu6 > mtu)
3249 mtu = idev->cnf.mtu6;
3252 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3254 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3257 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3260 struct dst_entry *dst;
3261 struct rt6_info *rt;
3262 struct inet6_dev *idev = in6_dev_get(dev);
3263 struct net *net = dev_net(dev);
3265 if (unlikely(!idev))
3266 return ERR_PTR(-ENODEV);
3268 rt = ip6_dst_alloc(net, dev, 0);
3269 if (unlikely(!rt)) {
3271 dst = ERR_PTR(-ENOMEM);
3275 rt->dst.input = ip6_input;
3276 rt->dst.output = ip6_output;
3277 rt->rt6i_gateway = fl6->daddr;
3278 rt->rt6i_dst.addr = fl6->daddr;
3279 rt->rt6i_dst.plen = 128;
3280 rt->rt6i_idev = idev;
3281 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3283 /* Add this dst into uncached_list so that rt6_disable_ip() can
3284 * do proper release of the net_device
3286 rt6_uncached_list_add(rt);
3287 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
3289 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3295 static int ip6_dst_gc(struct dst_ops *ops)
3297 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3298 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3299 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3300 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3301 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3302 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3305 entries = dst_entries_get_fast(ops);
3306 if (entries > rt_max_size)
3307 entries = dst_entries_get_slow(ops);
3309 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
3310 entries <= rt_max_size)
3313 net->ipv6.ip6_rt_gc_expire++;
3314 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
3315 entries = dst_entries_get_slow(ops);
3316 if (entries < ops->gc_thresh)
3317 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
3319 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
3320 return entries > rt_max_size;
3323 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3324 const struct in6_addr *gw_addr, u32 tbid,
3325 int flags, struct fib6_result *res)
3327 struct flowi6 fl6 = {
3328 .flowi6_oif = cfg->fc_ifindex,
3330 .saddr = cfg->fc_prefsrc,
3332 struct fib6_table *table;
3335 table = fib6_get_table(net, tbid);
3339 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3340 flags |= RT6_LOOKUP_F_HAS_SADDR;
3342 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3344 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3345 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3346 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3347 cfg->fc_ifindex != 0, NULL, flags);
3352 static int ip6_route_check_nh_onlink(struct net *net,
3353 struct fib6_config *cfg,
3354 const struct net_device *dev,
3355 struct netlink_ext_ack *extack)
3357 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3358 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3359 struct fib6_result res = {};
3362 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3363 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3364 /* ignore match if it is the default route */
3365 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3366 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3367 NL_SET_ERR_MSG(extack,
3368 "Nexthop has invalid gateway or device mismatch");
3375 static int ip6_route_check_nh(struct net *net,
3376 struct fib6_config *cfg,
3377 struct net_device **_dev,
3378 struct inet6_dev **idev)
3380 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3381 struct net_device *dev = _dev ? *_dev : NULL;
3382 int flags = RT6_LOOKUP_F_IFACE;
3383 struct fib6_result res = {};
3384 int err = -EHOSTUNREACH;
3386 if (cfg->fc_table) {
3387 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3388 cfg->fc_table, flags, &res);
3389 /* gw_addr can not require a gateway or resolve to a reject
3390 * route. If a device is given, it must match the result.
3392 if (err || res.fib6_flags & RTF_REJECT ||
3393 res.nh->fib_nh_gw_family ||
3394 (dev && dev != res.nh->fib_nh_dev))
3395 err = -EHOSTUNREACH;
3399 struct flowi6 fl6 = {
3400 .flowi6_oif = cfg->fc_ifindex,
3404 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3405 if (err || res.fib6_flags & RTF_REJECT ||
3406 res.nh->fib_nh_gw_family)
3407 err = -EHOSTUNREACH;
3412 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3413 cfg->fc_ifindex != 0, NULL, flags);
3418 if (dev != res.nh->fib_nh_dev)
3419 err = -EHOSTUNREACH;
3421 *_dev = dev = res.nh->fib_nh_dev;
3423 *idev = in6_dev_get(dev);
3429 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3430 struct net_device **_dev, struct inet6_dev **idev,
3431 struct netlink_ext_ack *extack)
3433 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3434 int gwa_type = ipv6_addr_type(gw_addr);
3435 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3436 const struct net_device *dev = *_dev;
3437 bool need_addr_check = !dev;
3440 /* if gw_addr is local we will fail to detect this in case
3441 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3442 * will return already-added prefix route via interface that
3443 * prefix route was assigned to, which might be non-loopback.
3446 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3447 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3451 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3452 /* IPv6 strictly inhibits using not link-local
3453 * addresses as nexthop address.
3454 * Otherwise, router will not able to send redirects.
3455 * It is very good, but in some (rare!) circumstances
3456 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3457 * some exceptions. --ANK
3458 * We allow IPv4-mapped nexthops to support RFC4798-type
3461 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3462 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3468 if (cfg->fc_flags & RTNH_F_ONLINK)
3469 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3471 err = ip6_route_check_nh(net, cfg, _dev, idev);
3479 /* reload in case device was changed */
3484 NL_SET_ERR_MSG(extack, "Egress device not specified");
3486 } else if (dev->flags & IFF_LOOPBACK) {
3487 NL_SET_ERR_MSG(extack,
3488 "Egress device can not be loopback device for this route");
3492 /* if we did not check gw_addr above, do so now that the
3493 * egress device has been resolved.
3495 if (need_addr_check &&
3496 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3497 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3506 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3508 if ((flags & RTF_REJECT) ||
3509 (dev && (dev->flags & IFF_LOOPBACK) &&
3510 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3511 !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3517 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3518 struct fib6_config *cfg, gfp_t gfp_flags,
3519 struct netlink_ext_ack *extack)
3521 struct net_device *dev = NULL;
3522 struct inet6_dev *idev = NULL;
3526 fib6_nh->fib_nh_family = AF_INET6;
3527 #ifdef CONFIG_IPV6_ROUTER_PREF
3528 fib6_nh->last_probe = jiffies;
3530 if (cfg->fc_is_fdb) {
3531 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3532 fib6_nh->fib_nh_gw_family = AF_INET6;
3537 if (cfg->fc_ifindex) {
3538 dev = dev_get_by_index(net, cfg->fc_ifindex);
3541 idev = in6_dev_get(dev);
3546 if (cfg->fc_flags & RTNH_F_ONLINK) {
3548 NL_SET_ERR_MSG(extack,
3549 "Nexthop device required for onlink");
3553 if (!(dev->flags & IFF_UP)) {
3554 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3559 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3562 fib6_nh->fib_nh_weight = 1;
3564 /* We cannot add true routes via loopback here,
3565 * they would result in kernel looping; promote them to reject routes
3567 addr_type = ipv6_addr_type(&cfg->fc_dst);
3568 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3569 /* hold loopback dev/idev if we haven't done so. */
3570 if (dev != net->loopback_dev) {
3575 dev = net->loopback_dev;
3577 idev = in6_dev_get(dev);
3586 if (cfg->fc_flags & RTF_GATEWAY) {
3587 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3591 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3592 fib6_nh->fib_nh_gw_family = AF_INET6;
3599 if (idev->cnf.disable_ipv6) {
3600 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3605 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3606 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3611 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3612 !netif_carrier_ok(dev))
3613 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3615 err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3616 cfg->fc_encap_type, cfg, gfp_flags, extack);
3621 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3622 if (!fib6_nh->rt6i_pcpu) {
3627 fib6_nh->fib_nh_dev = dev;
3628 fib6_nh->fib_nh_oif = dev->ifindex;
3635 lwtstate_put(fib6_nh->fib_nh_lws);
3636 fib6_nh->fib_nh_lws = NULL;
3643 void fib6_nh_release(struct fib6_nh *fib6_nh)
3645 struct rt6_exception_bucket *bucket;
3649 fib6_nh_flush_exceptions(fib6_nh, NULL);
3650 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3652 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3658 if (fib6_nh->rt6i_pcpu) {
3661 for_each_possible_cpu(cpu) {
3662 struct rt6_info **ppcpu_rt;
3663 struct rt6_info *pcpu_rt;
3665 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3666 pcpu_rt = *ppcpu_rt;
3668 dst_dev_put(&pcpu_rt->dst);
3669 dst_release(&pcpu_rt->dst);
3674 free_percpu(fib6_nh->rt6i_pcpu);
3677 fib_nh_common_release(&fib6_nh->nh_common);
3680 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3682 struct netlink_ext_ack *extack)
3684 struct net *net = cfg->fc_nlinfo.nl_net;
3685 struct fib6_info *rt = NULL;
3686 struct nexthop *nh = NULL;
3687 struct fib6_table *table;
3688 struct fib6_nh *fib6_nh;
3692 /* RTF_PCPU is an internal flag; can not be set by userspace */
3693 if (cfg->fc_flags & RTF_PCPU) {
3694 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3698 /* RTF_CACHE is an internal flag; can not be set by userspace */
3699 if (cfg->fc_flags & RTF_CACHE) {
3700 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3704 if (cfg->fc_type > RTN_MAX) {
3705 NL_SET_ERR_MSG(extack, "Invalid route type");
3709 if (cfg->fc_dst_len > 128) {
3710 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3713 if (cfg->fc_src_len > 128) {
3714 NL_SET_ERR_MSG(extack, "Invalid source address length");
3717 #ifndef CONFIG_IPV6_SUBTREES
3718 if (cfg->fc_src_len) {
3719 NL_SET_ERR_MSG(extack,
3720 "Specifying source address requires IPV6_SUBTREES to be enabled");
3724 if (cfg->fc_nh_id) {
3725 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3727 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3730 err = fib6_check_nexthop(nh, cfg, extack);
3736 if (cfg->fc_nlinfo.nlh &&
3737 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3738 table = fib6_get_table(net, cfg->fc_table);
3740 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3741 table = fib6_new_table(net, cfg->fc_table);
3744 table = fib6_new_table(net, cfg->fc_table);
3751 rt = fib6_info_alloc(gfp_flags, !nh);
3755 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3757 if (IS_ERR(rt->fib6_metrics)) {
3758 err = PTR_ERR(rt->fib6_metrics);
3759 /* Do not leave garbage there. */
3760 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3764 if (cfg->fc_flags & RTF_ADDRCONF)
3765 rt->dst_nocount = true;
3767 if (cfg->fc_flags & RTF_EXPIRES)
3768 fib6_set_expires(rt, jiffies +
3769 clock_t_to_jiffies(cfg->fc_expires));
3771 fib6_clean_expires(rt);
3773 if (cfg->fc_protocol == RTPROT_UNSPEC)
3774 cfg->fc_protocol = RTPROT_BOOT;
3775 rt->fib6_protocol = cfg->fc_protocol;
3777 rt->fib6_table = table;
3778 rt->fib6_metric = cfg->fc_metric;
3779 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3780 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3782 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3783 rt->fib6_dst.plen = cfg->fc_dst_len;
3785 #ifdef CONFIG_IPV6_SUBTREES
3786 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3787 rt->fib6_src.plen = cfg->fc_src_len;
3790 if (rt->fib6_src.plen) {
3791 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3794 if (!nexthop_get(nh)) {
3795 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3799 fib6_nh = nexthop_fib6_nh(rt->nh);
3801 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3805 fib6_nh = rt->fib6_nh;
3807 /* We cannot add true routes via loopback here, they would
3808 * result in kernel looping; promote them to reject routes
3810 addr_type = ipv6_addr_type(&cfg->fc_dst);
3811 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3813 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3816 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3817 struct net_device *dev = fib6_nh->fib_nh_dev;
3819 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3820 NL_SET_ERR_MSG(extack, "Invalid source address");
3824 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3825 rt->fib6_prefsrc.plen = 128;
3827 rt->fib6_prefsrc.plen = 0;
3831 fib6_info_release(rt);
3832 return ERR_PTR(err);
3834 ip_fib_metrics_put(rt->fib6_metrics);
3836 return ERR_PTR(err);
3839 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3840 struct netlink_ext_ack *extack)
3842 struct fib6_info *rt;
3845 rt = ip6_route_info_create(cfg, gfp_flags, extack);
3849 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3850 fib6_info_release(rt);
3855 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3857 struct net *net = info->nl_net;
3858 struct fib6_table *table;
3861 if (rt == net->ipv6.fib6_null_entry) {
3866 table = rt->fib6_table;
3867 spin_lock_bh(&table->tb6_lock);
3868 err = fib6_del(rt, info);
3869 spin_unlock_bh(&table->tb6_lock);
3872 fib6_info_release(rt);
3876 int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
3878 struct nl_info info = {
3880 .skip_notify = skip_notify
3883 return __ip6_del_rt(rt, &info);
3886 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3888 struct nl_info *info = &cfg->fc_nlinfo;
3889 struct net *net = info->nl_net;
3890 struct sk_buff *skb = NULL;
3891 struct fib6_table *table;
3894 if (rt == net->ipv6.fib6_null_entry)
3896 table = rt->fib6_table;
3897 spin_lock_bh(&table->tb6_lock);
3899 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3900 struct fib6_info *sibling, *next_sibling;
3901 struct fib6_node *fn;
3903 /* prefer to send a single notification with all hops */
3904 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3906 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3908 if (rt6_fill_node(net, skb, rt, NULL,
3909 NULL, NULL, 0, RTM_DELROUTE,
3910 info->portid, seq, 0) < 0) {
3914 info->skip_notify = 1;
3917 /* 'rt' points to the first sibling route. If it is not the
3918 * leaf, then we do not need to send a notification. Otherwise,
3919 * we need to check if the last sibling has a next route or not
3920 * and emit a replace or delete notification, respectively.
3922 info->skip_notify_kernel = 1;
3923 fn = rcu_dereference_protected(rt->fib6_node,
3924 lockdep_is_held(&table->tb6_lock));
3925 if (rcu_access_pointer(fn->leaf) == rt) {
3926 struct fib6_info *last_sibling, *replace_rt;
3928 last_sibling = list_last_entry(&rt->fib6_siblings,
3931 replace_rt = rcu_dereference_protected(
3932 last_sibling->fib6_next,
3933 lockdep_is_held(&table->tb6_lock));
3935 call_fib6_entry_notifiers_replace(net,
3938 call_fib6_multipath_entry_notifiers(net,
3939 FIB_EVENT_ENTRY_DEL,
3940 rt, rt->fib6_nsiblings,
3943 list_for_each_entry_safe(sibling, next_sibling,
3946 err = fib6_del(sibling, info);
3952 err = fib6_del(rt, info);
3954 spin_unlock_bh(&table->tb6_lock);
3956 fib6_info_release(rt);
3959 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3960 info->nlh, gfp_any());
3965 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3969 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3972 if (cfg->fc_flags & RTF_GATEWAY &&
3973 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3976 rc = rt6_remove_exception_rt(rt);
3981 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3984 struct fib6_result res = {
3988 struct rt6_info *rt_cache;
3990 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3992 return __ip6_del_cached_rt(rt_cache, cfg);
3997 struct fib6_nh_del_cached_rt_arg {
3998 struct fib6_config *cfg;
3999 struct fib6_info *f6i;
4002 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
4004 struct fib6_nh_del_cached_rt_arg *arg = _arg;
4007 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
4008 return rc != -ESRCH ? rc : 0;
4011 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
4013 struct fib6_nh_del_cached_rt_arg arg = {
4018 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
4021 static int ip6_route_del(struct fib6_config *cfg,
4022 struct netlink_ext_ack *extack)
4024 struct fib6_table *table;
4025 struct fib6_info *rt;
4026 struct fib6_node *fn;
4029 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
4031 NL_SET_ERR_MSG(extack, "FIB table does not exist");
4037 fn = fib6_locate(&table->tb6_root,
4038 &cfg->fc_dst, cfg->fc_dst_len,
4039 &cfg->fc_src, cfg->fc_src_len,
4040 !(cfg->fc_flags & RTF_CACHE));
4043 for_each_fib6_node_rt_rcu(fn) {
4046 if (rt->nh && cfg->fc_nh_id &&
4047 rt->nh->id != cfg->fc_nh_id)
4050 if (cfg->fc_flags & RTF_CACHE) {
4054 rc = ip6_del_cached_rt_nh(cfg, rt);
4055 } else if (cfg->fc_nh_id) {
4059 rc = ip6_del_cached_rt(cfg, rt, nh);
4068 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
4070 if (cfg->fc_protocol &&
4071 cfg->fc_protocol != rt->fib6_protocol)
4075 if (!fib6_info_hold_safe(rt))
4079 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4085 if (cfg->fc_ifindex &&
4087 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
4089 if (cfg->fc_flags & RTF_GATEWAY &&
4090 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
4092 if (!fib6_info_hold_safe(rt))
4096 /* if gateway was specified only delete the one hop */
4097 if (cfg->fc_flags & RTF_GATEWAY)
4098 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4100 return __ip6_del_rt_siblings(rt, cfg);
4108 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
4110 struct netevent_redirect netevent;
4111 struct rt6_info *rt, *nrt = NULL;
4112 struct fib6_result res = {};
4113 struct ndisc_options ndopts;
4114 struct inet6_dev *in6_dev;
4115 struct neighbour *neigh;
4117 int optlen, on_link;
4120 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
4121 optlen -= sizeof(*msg);
4124 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4128 msg = (struct rd_msg *)icmp6_hdr(skb);
4130 if (ipv6_addr_is_multicast(&msg->dest)) {
4131 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4136 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4138 } else if (ipv6_addr_type(&msg->target) !=
4139 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4140 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4144 in6_dev = __in6_dev_get(skb->dev);
4147 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
4151 * The IP source address of the Redirect MUST be the same as the current
4152 * first-hop router for the specified ICMP Destination Address.
4155 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4156 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4161 if (ndopts.nd_opts_tgt_lladdr) {
4162 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4165 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4170 rt = (struct rt6_info *) dst;
4171 if (rt->rt6i_flags & RTF_REJECT) {
4172 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4176 /* Redirect received -> path was valid.
4177 * Look, redirects are sent only in response to data packets,
4178 * so that this nexthop apparently is reachable. --ANK
4180 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4182 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4187 * We have finally decided to accept it.
4190 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4191 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4192 NEIGH_UPDATE_F_OVERRIDE|
4193 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4194 NEIGH_UPDATE_F_ISROUTER)),
4195 NDISC_REDIRECT, &ndopts);
4198 res.f6i = rcu_dereference(rt->from);
4203 struct fib6_nh_match_arg arg = {
4205 .gw = &rt->rt6i_gateway,
4208 nexthop_for_each_fib6_nh(res.f6i->nh,
4209 fib6_nh_find_match, &arg);
4211 /* fib6_info uses a nexthop that does not have fib6_nh
4212 * using the dst->dev. Should be impossible
4218 res.nh = res.f6i->fib6_nh;
4221 res.fib6_flags = res.f6i->fib6_flags;
4222 res.fib6_type = res.f6i->fib6_type;
4223 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4227 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4229 nrt->rt6i_flags &= ~RTF_GATEWAY;
4231 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4233 /* rt6_insert_exception() will take care of duplicated exceptions */
4234 if (rt6_insert_exception(nrt, &res)) {
4235 dst_release_immediate(&nrt->dst);
4239 netevent.old = &rt->dst;
4240 netevent.new = &nrt->dst;
4241 netevent.daddr = &msg->dest;
4242 netevent.neigh = neigh;
4243 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4247 neigh_release(neigh);
4250 #ifdef CONFIG_IPV6_ROUTE_INFO
4251 static struct fib6_info *rt6_get_route_info(struct net *net,
4252 const struct in6_addr *prefix, int prefixlen,
4253 const struct in6_addr *gwaddr,
4254 struct net_device *dev)
4256 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4257 int ifindex = dev->ifindex;
4258 struct fib6_node *fn;
4259 struct fib6_info *rt = NULL;
4260 struct fib6_table *table;
4262 table = fib6_get_table(net, tb_id);
4267 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4271 for_each_fib6_node_rt_rcu(fn) {
4272 /* these routes do not use nexthops */
4275 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4277 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4278 !rt->fib6_nh->fib_nh_gw_family)
4280 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4282 if (!fib6_info_hold_safe(rt))
4291 static struct fib6_info *rt6_add_route_info(struct net *net,
4292 const struct in6_addr *prefix, int prefixlen,
4293 const struct in6_addr *gwaddr,
4294 struct net_device *dev,
4297 struct fib6_config cfg = {
4298 .fc_metric = IP6_RT_PRIO_USER,
4299 .fc_ifindex = dev->ifindex,
4300 .fc_dst_len = prefixlen,
4301 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4302 RTF_UP | RTF_PREF(pref),
4303 .fc_protocol = RTPROT_RA,
4304 .fc_type = RTN_UNICAST,
4305 .fc_nlinfo.portid = 0,
4306 .fc_nlinfo.nlh = NULL,
4307 .fc_nlinfo.nl_net = net,
4310 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4311 cfg.fc_dst = *prefix;
4312 cfg.fc_gateway = *gwaddr;
4314 /* We should treat it as a default route if prefix length is 0. */
4316 cfg.fc_flags |= RTF_DEFAULT;
4318 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4320 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4324 struct fib6_info *rt6_get_dflt_router(struct net *net,
4325 const struct in6_addr *addr,
4326 struct net_device *dev)
4328 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4329 struct fib6_info *rt;
4330 struct fib6_table *table;
4332 table = fib6_get_table(net, tb_id);
4337 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4340 /* RA routes do not use nexthops */
4345 if (dev == nh->fib_nh_dev &&
4346 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4347 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4350 if (rt && !fib6_info_hold_safe(rt))
4356 struct fib6_info *rt6_add_dflt_router(struct net *net,
4357 const struct in6_addr *gwaddr,
4358 struct net_device *dev,
4360 u32 defrtr_usr_metric)
4362 struct fib6_config cfg = {
4363 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4364 .fc_metric = defrtr_usr_metric,
4365 .fc_ifindex = dev->ifindex,
4366 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4367 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4368 .fc_protocol = RTPROT_RA,
4369 .fc_type = RTN_UNICAST,
4370 .fc_nlinfo.portid = 0,
4371 .fc_nlinfo.nlh = NULL,
4372 .fc_nlinfo.nl_net = net,
4375 cfg.fc_gateway = *gwaddr;
4377 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4378 struct fib6_table *table;
4380 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4382 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4385 return rt6_get_dflt_router(net, gwaddr, dev);
4388 static void __rt6_purge_dflt_routers(struct net *net,
4389 struct fib6_table *table)
4391 struct fib6_info *rt;
4395 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4396 struct net_device *dev = fib6_info_nh_dev(rt);
4397 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4399 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4400 (!idev || idev->cnf.accept_ra != 2) &&
4401 fib6_info_hold_safe(rt)) {
4403 ip6_del_rt(net, rt, false);
4409 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4412 void rt6_purge_dflt_routers(struct net *net)
4414 struct fib6_table *table;
4415 struct hlist_head *head;
4420 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4421 head = &net->ipv6.fib_table_hash[h];
4422 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4423 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4424 __rt6_purge_dflt_routers(net, table);
4431 static void rtmsg_to_fib6_config(struct net *net,
4432 struct in6_rtmsg *rtmsg,
4433 struct fib6_config *cfg)
4435 *cfg = (struct fib6_config){
4436 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4438 .fc_ifindex = rtmsg->rtmsg_ifindex,
4439 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
4440 .fc_expires = rtmsg->rtmsg_info,
4441 .fc_dst_len = rtmsg->rtmsg_dst_len,
4442 .fc_src_len = rtmsg->rtmsg_src_len,
4443 .fc_flags = rtmsg->rtmsg_flags,
4444 .fc_type = rtmsg->rtmsg_type,
4446 .fc_nlinfo.nl_net = net,
4448 .fc_dst = rtmsg->rtmsg_dst,
4449 .fc_src = rtmsg->rtmsg_src,
4450 .fc_gateway = rtmsg->rtmsg_gateway,
4454 int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
4456 struct fib6_config cfg;
4459 if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4461 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4464 rtmsg_to_fib6_config(net, rtmsg, &cfg);
4469 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4472 err = ip6_route_del(&cfg, NULL);
4480 * Drop the packet on the floor
4483 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4485 struct dst_entry *dst = skb_dst(skb);
4486 struct net *net = dev_net(dst->dev);
4487 struct inet6_dev *idev;
4490 if (netif_is_l3_master(skb->dev) &&
4491 dst->dev == net->loopback_dev)
4492 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4494 idev = ip6_dst_idev(dst);
4496 switch (ipstats_mib_noroutes) {
4497 case IPSTATS_MIB_INNOROUTES:
4498 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4499 if (type == IPV6_ADDR_ANY) {
4500 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4504 case IPSTATS_MIB_OUTNOROUTES:
4505 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4509 /* Start over by dropping the dst for l3mdev case */
4510 if (netif_is_l3_master(skb->dev))
4513 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4518 static int ip6_pkt_discard(struct sk_buff *skb)
4520 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4523 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4525 skb->dev = skb_dst(skb)->dev;
4526 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4529 static int ip6_pkt_prohibit(struct sk_buff *skb)
4531 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4534 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4536 skb->dev = skb_dst(skb)->dev;
4537 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4541 * Allocate a dst for local (unicast / anycast) address.
4544 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4545 struct inet6_dev *idev,
4546 const struct in6_addr *addr,
4547 bool anycast, gfp_t gfp_flags)
4549 struct fib6_config cfg = {
4550 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4551 .fc_ifindex = idev->dev->ifindex,
4552 .fc_flags = RTF_UP | RTF_NONEXTHOP,
4555 .fc_protocol = RTPROT_KERNEL,
4556 .fc_nlinfo.nl_net = net,
4557 .fc_ignore_dev_down = true,
4559 struct fib6_info *f6i;
4562 cfg.fc_type = RTN_ANYCAST;
4563 cfg.fc_flags |= RTF_ANYCAST;
4565 cfg.fc_type = RTN_LOCAL;
4566 cfg.fc_flags |= RTF_LOCAL;
4569 f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
4571 f6i->dst_nocount = true;
4575 /* remove deleted ip from prefsrc entries */
4576 struct arg_dev_net_ip {
4577 struct net_device *dev;
4579 struct in6_addr *addr;
4582 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4584 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4585 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4586 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4589 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
4590 rt != net->ipv6.fib6_null_entry &&
4591 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
4592 spin_lock_bh(&rt6_exception_lock);
4593 /* remove prefsrc entry */
4594 rt->fib6_prefsrc.plen = 0;
4595 spin_unlock_bh(&rt6_exception_lock);
4600 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4602 struct net *net = dev_net(ifp->idev->dev);
4603 struct arg_dev_net_ip adni = {
4604 .dev = ifp->idev->dev,
4608 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4611 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
4613 /* Remove routers and update dst entries when gateway turn into host. */
4614 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4616 struct in6_addr *gateway = (struct in6_addr *)arg;
4619 /* RA routes do not use nexthops */
4624 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4625 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4628 /* Further clean up cached routes in exception table.
4629 * This is needed because cached route may have a different
4630 * gateway than its 'parent' in the case of an ip redirect.
4632 fib6_nh_exceptions_clean_tohost(nh, gateway);
4637 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4639 fib6_clean_all(net, fib6_clean_tohost, gateway);
4642 struct arg_netdev_event {
4643 const struct net_device *dev;
4645 unsigned char nh_flags;
4646 unsigned long event;
4650 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4652 struct fib6_info *iter;
4653 struct fib6_node *fn;
4655 fn = rcu_dereference_protected(rt->fib6_node,
4656 lockdep_is_held(&rt->fib6_table->tb6_lock));
4657 iter = rcu_dereference_protected(fn->leaf,
4658 lockdep_is_held(&rt->fib6_table->tb6_lock));
4660 if (iter->fib6_metric == rt->fib6_metric &&
4661 rt6_qualify_for_ecmp(iter))
4663 iter = rcu_dereference_protected(iter->fib6_next,
4664 lockdep_is_held(&rt->fib6_table->tb6_lock));
4670 /* only called for fib entries with builtin fib6_nh */
4671 static bool rt6_is_dead(const struct fib6_info *rt)
4673 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4674 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4675 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4681 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4683 struct fib6_info *iter;
4686 if (!rt6_is_dead(rt))
4687 total += rt->fib6_nh->fib_nh_weight;
4689 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4690 if (!rt6_is_dead(iter))
4691 total += iter->fib6_nh->fib_nh_weight;
4697 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4699 int upper_bound = -1;
4701 if (!rt6_is_dead(rt)) {
4702 *weight += rt->fib6_nh->fib_nh_weight;
4703 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4706 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4709 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4711 struct fib6_info *iter;
4714 rt6_upper_bound_set(rt, &weight, total);
4716 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4717 rt6_upper_bound_set(iter, &weight, total);
4720 void rt6_multipath_rebalance(struct fib6_info *rt)
4722 struct fib6_info *first;
4725 /* In case the entire multipath route was marked for flushing,
4726 * then there is no need to rebalance upon the removal of every
4729 if (!rt->fib6_nsiblings || rt->should_flush)
4732 /* During lookup routes are evaluated in order, so we need to
4733 * make sure upper bounds are assigned from the first sibling
4736 first = rt6_multipath_first_sibling(rt);
4737 if (WARN_ON_ONCE(!first))
4740 total = rt6_multipath_total_weight(first);
4741 rt6_multipath_upper_bound_set(first, total);
4744 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4746 const struct arg_netdev_event *arg = p_arg;
4747 struct net *net = dev_net(arg->dev);
4749 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4750 rt->fib6_nh->fib_nh_dev == arg->dev) {
4751 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4752 fib6_update_sernum_upto_root(net, rt);
4753 rt6_multipath_rebalance(rt);
4759 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4761 struct arg_netdev_event arg = {
4764 .nh_flags = nh_flags,
4768 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4769 arg.nh_flags |= RTNH_F_LINKDOWN;
4771 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4774 /* only called for fib entries with inline fib6_nh */
4775 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4776 const struct net_device *dev)
4778 struct fib6_info *iter;
4780 if (rt->fib6_nh->fib_nh_dev == dev)
4782 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4783 if (iter->fib6_nh->fib_nh_dev == dev)
4789 static void rt6_multipath_flush(struct fib6_info *rt)
4791 struct fib6_info *iter;
4793 rt->should_flush = 1;
4794 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4795 iter->should_flush = 1;
4798 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4799 const struct net_device *down_dev)
4801 struct fib6_info *iter;
4802 unsigned int dead = 0;
4804 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4805 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4807 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4808 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4809 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4815 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4816 const struct net_device *dev,
4817 unsigned char nh_flags)
4819 struct fib6_info *iter;
4821 if (rt->fib6_nh->fib_nh_dev == dev)
4822 rt->fib6_nh->fib_nh_flags |= nh_flags;
4823 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4824 if (iter->fib6_nh->fib_nh_dev == dev)
4825 iter->fib6_nh->fib_nh_flags |= nh_flags;
4828 /* called with write lock held for table with rt */
4829 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4831 const struct arg_netdev_event *arg = p_arg;
4832 const struct net_device *dev = arg->dev;
4833 struct net *net = dev_net(dev);
4835 if (rt == net->ipv6.fib6_null_entry || rt->nh)
4838 switch (arg->event) {
4839 case NETDEV_UNREGISTER:
4840 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4842 if (rt->should_flush)
4844 if (!rt->fib6_nsiblings)
4845 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4846 if (rt6_multipath_uses_dev(rt, dev)) {
4849 count = rt6_multipath_dead_count(rt, dev);
4850 if (rt->fib6_nsiblings + 1 == count) {
4851 rt6_multipath_flush(rt);
4854 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4856 fib6_update_sernum(net, rt);
4857 rt6_multipath_rebalance(rt);
4861 if (rt->fib6_nh->fib_nh_dev != dev ||
4862 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4864 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4865 rt6_multipath_rebalance(rt);
4872 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4874 struct arg_netdev_event arg = {
4880 struct net *net = dev_net(dev);
4882 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4883 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4885 fib6_clean_all(net, fib6_ifdown, &arg);
4888 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4890 rt6_sync_down_dev(dev, event);
4891 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4892 neigh_ifdown(&nd_tbl, dev);
4895 struct rt6_mtu_change_arg {
4896 struct net_device *dev;
4898 struct fib6_info *f6i;
4901 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4903 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4904 struct fib6_info *f6i = arg->f6i;
4906 /* For administrative MTU increase, there is no way to discover
4907 * IPv6 PMTU increase, so PMTU increase should be updated here.
4908 * Since RFC 1981 doesn't include administrative MTU increase
4909 * update PMTU increase is a MUST. (i.e. jumbo frame)
4911 if (nh->fib_nh_dev == arg->dev) {
4912 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4913 u32 mtu = f6i->fib6_pmtu;
4915 if (mtu >= arg->mtu ||
4916 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4917 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4919 spin_lock_bh(&rt6_exception_lock);
4920 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4921 spin_unlock_bh(&rt6_exception_lock);
4927 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4929 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4930 struct inet6_dev *idev;
4932 /* In IPv6 pmtu discovery is not optional,
4933 so that RTAX_MTU lock cannot disable it.
4934 We still use this lock to block changes
4935 caused by addrconf/ndisc.
4938 idev = __in6_dev_get(arg->dev);
4942 if (fib6_metric_locked(f6i, RTAX_MTU))
4947 /* fib6_nh_mtu_change only returns 0, so this is safe */
4948 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4952 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4955 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4957 struct rt6_mtu_change_arg arg = {
4962 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4965 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4966 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
4967 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
4968 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
4969 [RTA_OIF] = { .type = NLA_U32 },
4970 [RTA_IIF] = { .type = NLA_U32 },
4971 [RTA_PRIORITY] = { .type = NLA_U32 },
4972 [RTA_METRICS] = { .type = NLA_NESTED },
4973 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
4974 [RTA_PREF] = { .type = NLA_U8 },
4975 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4976 [RTA_ENCAP] = { .type = NLA_NESTED },
4977 [RTA_EXPIRES] = { .type = NLA_U32 },
4978 [RTA_UID] = { .type = NLA_U32 },
4979 [RTA_MARK] = { .type = NLA_U32 },
4980 [RTA_TABLE] = { .type = NLA_U32 },
4981 [RTA_IP_PROTO] = { .type = NLA_U8 },
4982 [RTA_SPORT] = { .type = NLA_U16 },
4983 [RTA_DPORT] = { .type = NLA_U16 },
4984 [RTA_NH_ID] = { .type = NLA_U32 },
4987 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
4988 struct fib6_config *cfg,
4989 struct netlink_ext_ack *extack)
4992 struct nlattr *tb[RTA_MAX+1];
4996 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4997 rtm_ipv6_policy, extack);
5002 rtm = nlmsg_data(nlh);
5004 *cfg = (struct fib6_config){
5005 .fc_table = rtm->rtm_table,
5006 .fc_dst_len = rtm->rtm_dst_len,
5007 .fc_src_len = rtm->rtm_src_len,
5009 .fc_protocol = rtm->rtm_protocol,
5010 .fc_type = rtm->rtm_type,
5012 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
5013 .fc_nlinfo.nlh = nlh,
5014 .fc_nlinfo.nl_net = sock_net(skb->sk),
5017 if (rtm->rtm_type == RTN_UNREACHABLE ||
5018 rtm->rtm_type == RTN_BLACKHOLE ||
5019 rtm->rtm_type == RTN_PROHIBIT ||
5020 rtm->rtm_type == RTN_THROW)
5021 cfg->fc_flags |= RTF_REJECT;
5023 if (rtm->rtm_type == RTN_LOCAL)
5024 cfg->fc_flags |= RTF_LOCAL;
5026 if (rtm->rtm_flags & RTM_F_CLONED)
5027 cfg->fc_flags |= RTF_CACHE;
5029 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5031 if (tb[RTA_NH_ID]) {
5032 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
5033 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5034 NL_SET_ERR_MSG(extack,
5035 "Nexthop specification and nexthop id are mutually exclusive");
5038 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5041 if (tb[RTA_GATEWAY]) {
5042 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
5043 cfg->fc_flags |= RTF_GATEWAY;
5046 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5051 int plen = (rtm->rtm_dst_len + 7) >> 3;
5053 if (nla_len(tb[RTA_DST]) < plen)
5056 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
5060 int plen = (rtm->rtm_src_len + 7) >> 3;
5062 if (nla_len(tb[RTA_SRC]) < plen)
5065 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
5068 if (tb[RTA_PREFSRC])
5069 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
5072 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5074 if (tb[RTA_PRIORITY])
5075 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5077 if (tb[RTA_METRICS]) {
5078 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5079 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
5083 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5085 if (tb[RTA_MULTIPATH]) {
5086 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5087 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
5089 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
5090 cfg->fc_mp_len, extack);
5096 pref = nla_get_u8(tb[RTA_PREF]);
5097 if (pref != ICMPV6_ROUTER_PREF_LOW &&
5098 pref != ICMPV6_ROUTER_PREF_HIGH)
5099 pref = ICMPV6_ROUTER_PREF_MEDIUM;
5100 cfg->fc_flags |= RTF_PREF(pref);
5104 cfg->fc_encap = tb[RTA_ENCAP];
5106 if (tb[RTA_ENCAP_TYPE]) {
5107 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5109 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5114 if (tb[RTA_EXPIRES]) {
5115 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5117 if (addrconf_finite_timeout(timeout)) {
5118 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5119 cfg->fc_flags |= RTF_EXPIRES;
5129 struct fib6_info *fib6_info;
5130 struct fib6_config r_cfg;
5131 struct list_head next;
5134 static int ip6_route_info_append(struct net *net,
5135 struct list_head *rt6_nh_list,
5136 struct fib6_info *rt,
5137 struct fib6_config *r_cfg)
5142 list_for_each_entry(nh, rt6_nh_list, next) {
5143 /* check if fib6_info already exists */
5144 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5148 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5152 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5153 list_add_tail(&nh->next, rt6_nh_list);
5158 static void ip6_route_mpath_notify(struct fib6_info *rt,
5159 struct fib6_info *rt_last,
5160 struct nl_info *info,
5163 /* if this is an APPEND route, then rt points to the first route
5164 * inserted and rt_last points to last route inserted. Userspace
5165 * wants a consistent dump of the route which starts at the first
5166 * nexthop. Since sibling routes are always added at the end of
5167 * the list, find the first sibling of the last route appended
5169 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5170 rt = list_first_entry(&rt_last->fib6_siblings,
5176 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5179 static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5181 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5182 bool should_notify = false;
5183 struct fib6_info *leaf;
5184 struct fib6_node *fn;
5187 fn = rcu_dereference(rt->fib6_node);
5191 leaf = rcu_dereference(fn->leaf);
5196 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5197 rt6_qualify_for_ecmp(leaf)))
5198 should_notify = true;
5202 return should_notify;
5205 static int ip6_route_multipath_add(struct fib6_config *cfg,
5206 struct netlink_ext_ack *extack)
5208 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5209 struct nl_info *info = &cfg->fc_nlinfo;
5210 struct fib6_config r_cfg;
5211 struct rtnexthop *rtnh;
5212 struct fib6_info *rt;
5213 struct rt6_nh *err_nh;
5214 struct rt6_nh *nh, *nh_safe;
5220 int replace = (cfg->fc_nlinfo.nlh &&
5221 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5222 LIST_HEAD(rt6_nh_list);
5224 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5225 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5226 nlflags |= NLM_F_APPEND;
5228 remaining = cfg->fc_mp_len;
5229 rtnh = (struct rtnexthop *)cfg->fc_mp;
5231 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
5232 * fib6_info structs per nexthop
5234 while (rtnh_ok(rtnh, remaining)) {
5235 memcpy(&r_cfg, cfg, sizeof(*cfg));
5236 if (rtnh->rtnh_ifindex)
5237 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5239 attrlen = rtnh_attrlen(rtnh);
5241 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5243 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5245 r_cfg.fc_gateway = nla_get_in6_addr(nla);
5246 r_cfg.fc_flags |= RTF_GATEWAY;
5248 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5249 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5251 r_cfg.fc_encap_type = nla_get_u16(nla);
5254 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5255 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5261 if (!rt6_qualify_for_ecmp(rt)) {
5263 NL_SET_ERR_MSG(extack,
5264 "Device only routes can not be added for IPv6 using the multipath API.");
5265 fib6_info_release(rt);
5269 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5271 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5274 fib6_info_release(rt);
5278 rtnh = rtnh_next(rtnh, &remaining);
5281 if (list_empty(&rt6_nh_list)) {
5282 NL_SET_ERR_MSG(extack,
5283 "Invalid nexthop configuration - no valid nexthops");
5287 /* for add and replace send one notification with all nexthops.
5288 * Skip the notification in fib6_add_rt2node and send one with
5289 * the full route when done
5291 info->skip_notify = 1;
5293 /* For add and replace, send one notification with all nexthops. For
5294 * append, send one notification with all appended nexthops.
5296 info->skip_notify_kernel = 1;
5299 list_for_each_entry(nh, &rt6_nh_list, next) {
5300 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5301 fib6_info_release(nh->fib6_info);
5304 /* save reference to last route successfully inserted */
5305 rt_last = nh->fib6_info;
5307 /* save reference to first route for notification */
5309 rt_notif = nh->fib6_info;
5312 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5313 nh->fib6_info = NULL;
5316 NL_SET_ERR_MSG_MOD(extack,
5317 "multipath route replace failed (check consistency of installed routes)");
5322 /* Because each route is added like a single route we remove
5323 * these flags after the first nexthop: if there is a collision,
5324 * we have already failed to add the first nexthop:
5325 * fib6_add_rt2node() has rejected it; when replacing, old
5326 * nexthops have been replaced by first new, the rest should
5329 if (cfg->fc_nlinfo.nlh) {
5330 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5332 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5337 /* An in-kernel notification should only be sent in case the new
5338 * multipath route is added as the first route in the node, or if
5339 * it was appended to it. We pass 'rt_notif' since it is the first
5340 * sibling and might allow us to skip some checks in the replace case.
5342 if (ip6_route_mpath_should_notify(rt_notif)) {
5343 enum fib_event_type fib_event;
5345 if (rt_notif->fib6_nsiblings != nhn - 1)
5346 fib_event = FIB_EVENT_ENTRY_APPEND;
5348 fib_event = FIB_EVENT_ENTRY_REPLACE;
5350 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5351 fib_event, rt_notif,
5354 /* Delete all the siblings that were just added */
5360 /* success ... tell user about new route */
5361 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5365 /* send notification for routes that were added so that
5366 * the delete notifications sent by ip6_route_del are
5370 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5372 /* Delete routes that were already added */
5373 list_for_each_entry(nh, &rt6_nh_list, next) {
5376 ip6_route_del(&nh->r_cfg, extack);
5380 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5382 fib6_info_release(nh->fib6_info);
5383 list_del(&nh->next);
5390 static int ip6_route_multipath_del(struct fib6_config *cfg,
5391 struct netlink_ext_ack *extack)
5393 struct fib6_config r_cfg;
5394 struct rtnexthop *rtnh;
5400 remaining = cfg->fc_mp_len;
5401 rtnh = (struct rtnexthop *)cfg->fc_mp;
5403 /* Parse a Multipath Entry */
5404 while (rtnh_ok(rtnh, remaining)) {
5405 memcpy(&r_cfg, cfg, sizeof(*cfg));
5406 if (rtnh->rtnh_ifindex)
5407 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5409 attrlen = rtnh_attrlen(rtnh);
5411 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5413 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5415 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5416 r_cfg.fc_flags |= RTF_GATEWAY;
5419 err = ip6_route_del(&r_cfg, extack);
5423 rtnh = rtnh_next(rtnh, &remaining);
5429 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5430 struct netlink_ext_ack *extack)
5432 struct fib6_config cfg;
5435 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5440 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5441 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5446 return ip6_route_multipath_del(&cfg, extack);
5448 cfg.fc_delete_all_nh = 1;
5449 return ip6_route_del(&cfg, extack);
5453 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5454 struct netlink_ext_ack *extack)
5456 struct fib6_config cfg;
5459 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5463 if (cfg.fc_metric == 0)
5464 cfg.fc_metric = IP6_RT_PRIO_USER;
5467 return ip6_route_multipath_add(&cfg, extack);
5469 return ip6_route_add(&cfg, GFP_KERNEL, extack);
5472 /* add the overhead of this fib6_nh to nexthop_len */
5473 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5475 int *nexthop_len = arg;
5477 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5478 + NLA_ALIGN(sizeof(struct rtnexthop))
5479 + nla_total_size(16); /* RTA_GATEWAY */
5481 if (nh->fib_nh_lws) {
5482 /* RTA_ENCAP_TYPE */
5483 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5485 *nexthop_len += nla_total_size(2);
5491 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5496 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5497 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5500 struct fib6_nh *nh = f6i->fib6_nh;
5503 if (f6i->fib6_nsiblings) {
5504 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
5505 + NLA_ALIGN(sizeof(struct rtnexthop))
5506 + nla_total_size(16) /* RTA_GATEWAY */
5507 + lwtunnel_get_encap_size(nh->fib_nh_lws);
5509 nexthop_len *= f6i->fib6_nsiblings;
5511 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5514 return NLMSG_ALIGN(sizeof(struct rtmsg))
5515 + nla_total_size(16) /* RTA_SRC */
5516 + nla_total_size(16) /* RTA_DST */
5517 + nla_total_size(16) /* RTA_GATEWAY */
5518 + nla_total_size(16) /* RTA_PREFSRC */
5519 + nla_total_size(4) /* RTA_TABLE */
5520 + nla_total_size(4) /* RTA_IIF */
5521 + nla_total_size(4) /* RTA_OIF */
5522 + nla_total_size(4) /* RTA_PRIORITY */
5523 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5524 + nla_total_size(sizeof(struct rta_cacheinfo))
5525 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5526 + nla_total_size(1) /* RTA_PREF */
5530 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5531 unsigned char *flags)
5533 if (nexthop_is_multipath(nh)) {
5536 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5538 goto nla_put_failure;
5540 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5541 goto nla_put_failure;
5543 nla_nest_end(skb, mp);
5545 struct fib6_nh *fib6_nh;
5547 fib6_nh = nexthop_fib6_nh(nh);
5548 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5550 goto nla_put_failure;
5559 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5560 struct fib6_info *rt, struct dst_entry *dst,
5561 struct in6_addr *dest, struct in6_addr *src,
5562 int iif, int type, u32 portid, u32 seq,
5565 struct rt6_info *rt6 = (struct rt6_info *)dst;
5566 struct rt6key *rt6_dst, *rt6_src;
5567 u32 *pmetrics, table, rt6_flags;
5568 unsigned char nh_flags = 0;
5569 struct nlmsghdr *nlh;
5573 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5578 rt6_dst = &rt6->rt6i_dst;
5579 rt6_src = &rt6->rt6i_src;
5580 rt6_flags = rt6->rt6i_flags;
5582 rt6_dst = &rt->fib6_dst;
5583 rt6_src = &rt->fib6_src;
5584 rt6_flags = rt->fib6_flags;
5587 rtm = nlmsg_data(nlh);
5588 rtm->rtm_family = AF_INET6;
5589 rtm->rtm_dst_len = rt6_dst->plen;
5590 rtm->rtm_src_len = rt6_src->plen;
5593 table = rt->fib6_table->tb6_id;
5595 table = RT6_TABLE_UNSPEC;
5596 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5597 if (nla_put_u32(skb, RTA_TABLE, table))
5598 goto nla_put_failure;
5600 rtm->rtm_type = rt->fib6_type;
5602 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5603 rtm->rtm_protocol = rt->fib6_protocol;
5605 if (rt6_flags & RTF_CACHE)
5606 rtm->rtm_flags |= RTM_F_CLONED;
5609 if (nla_put_in6_addr(skb, RTA_DST, dest))
5610 goto nla_put_failure;
5611 rtm->rtm_dst_len = 128;
5612 } else if (rtm->rtm_dst_len)
5613 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5614 goto nla_put_failure;
5615 #ifdef CONFIG_IPV6_SUBTREES
5617 if (nla_put_in6_addr(skb, RTA_SRC, src))
5618 goto nla_put_failure;
5619 rtm->rtm_src_len = 128;
5620 } else if (rtm->rtm_src_len &&
5621 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5622 goto nla_put_failure;
5625 #ifdef CONFIG_IPV6_MROUTE
5626 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5627 int err = ip6mr_get_route(net, skb, rtm, portid);
5632 goto nla_put_failure;
5635 if (nla_put_u32(skb, RTA_IIF, iif))
5636 goto nla_put_failure;
5638 struct in6_addr saddr_buf;
5639 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
5640 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5641 goto nla_put_failure;
5644 if (rt->fib6_prefsrc.plen) {
5645 struct in6_addr saddr_buf;
5646 saddr_buf = rt->fib6_prefsrc.addr;
5647 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5648 goto nla_put_failure;
5651 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5652 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5653 goto nla_put_failure;
5655 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5656 goto nla_put_failure;
5658 /* For multipath routes, walk the siblings list and add
5659 * each as a nexthop within RTA_MULTIPATH.
5662 if (rt6_flags & RTF_GATEWAY &&
5663 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5664 goto nla_put_failure;
5666 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5667 goto nla_put_failure;
5669 if (dst->lwtstate &&
5670 lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5671 goto nla_put_failure;
5672 } else if (rt->fib6_nsiblings) {
5673 struct fib6_info *sibling, *next_sibling;
5676 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5678 goto nla_put_failure;
5680 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5681 rt->fib6_nh->fib_nh_weight, AF_INET6) < 0)
5682 goto nla_put_failure;
5684 list_for_each_entry_safe(sibling, next_sibling,
5685 &rt->fib6_siblings, fib6_siblings) {
5686 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5687 sibling->fib6_nh->fib_nh_weight,
5689 goto nla_put_failure;
5692 nla_nest_end(skb, mp);
5693 } else if (rt->nh) {
5694 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5695 goto nla_put_failure;
5697 if (nexthop_is_blackhole(rt->nh))
5698 rtm->rtm_type = RTN_BLACKHOLE;
5700 if (net->ipv4.sysctl_nexthop_compat_mode &&
5701 rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5702 goto nla_put_failure;
5704 rtm->rtm_flags |= nh_flags;
5706 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5707 &nh_flags, false) < 0)
5708 goto nla_put_failure;
5710 rtm->rtm_flags |= nh_flags;
5713 if (rt6_flags & RTF_EXPIRES) {
5714 expires = dst ? dst->expires : rt->expires;
5720 rtm->rtm_flags |= RTM_F_OFFLOAD;
5722 rtm->rtm_flags |= RTM_F_TRAP;
5723 if (rt->offload_failed)
5724 rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5727 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5728 goto nla_put_failure;
5730 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5731 goto nla_put_failure;
5734 nlmsg_end(skb, nlh);
5738 nlmsg_cancel(skb, nlh);
5742 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5744 const struct net_device *dev = arg;
5746 if (nh->fib_nh_dev == dev)
5752 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5753 const struct net_device *dev)
5756 struct net_device *_dev = (struct net_device *)dev;
5758 return !!nexthop_for_each_fib6_nh(f6i->nh,
5759 fib6_info_nh_uses_dev,
5763 if (f6i->fib6_nh->fib_nh_dev == dev)
5766 if (f6i->fib6_nsiblings) {
5767 struct fib6_info *sibling, *next_sibling;
5769 list_for_each_entry_safe(sibling, next_sibling,
5770 &f6i->fib6_siblings, fib6_siblings) {
5771 if (sibling->fib6_nh->fib_nh_dev == dev)
5779 struct fib6_nh_exception_dump_walker {
5780 struct rt6_rtnl_dump_arg *dump;
5781 struct fib6_info *rt;
5787 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5789 struct fib6_nh_exception_dump_walker *w = arg;
5790 struct rt6_rtnl_dump_arg *dump = w->dump;
5791 struct rt6_exception_bucket *bucket;
5792 struct rt6_exception *rt6_ex;
5795 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5799 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5800 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5806 /* Expiration of entries doesn't bump sernum, insertion
5807 * does. Removal is triggered by insertion, so we can
5808 * rely on the fact that if entries change between two
5809 * partial dumps, this node is scanned again completely,
5810 * see rt6_insert_exception() and fib6_dump_table().
5812 * Count expired entries we go through as handled
5813 * entries that we'll skip next time, in case of partial
5814 * node dump. Otherwise, if entries expire meanwhile,
5815 * we'll skip the wrong amount.
5817 if (rt6_check_expired(rt6_ex->rt6i)) {
5822 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5823 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5825 NETLINK_CB(dump->cb->skb).portid,
5826 dump->cb->nlh->nlmsg_seq, w->flags);
5838 /* Return -1 if done with node, number of handled routes on partial dump */
5839 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5841 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5842 struct fib_dump_filter *filter = &arg->filter;
5843 unsigned int flags = NLM_F_MULTI;
5844 struct net *net = arg->net;
5847 if (rt == net->ipv6.fib6_null_entry)
5850 if ((filter->flags & RTM_F_PREFIX) &&
5851 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5852 /* success since this is not a prefix route */
5855 if (filter->filter_set &&
5856 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5857 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5858 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5862 if (filter->filter_set ||
5863 !filter->dump_routes || !filter->dump_exceptions) {
5864 flags |= NLM_F_DUMP_FILTERED;
5867 if (filter->dump_routes) {
5871 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5873 NETLINK_CB(arg->cb->skb).portid,
5874 arg->cb->nlh->nlmsg_seq, flags)) {
5881 if (filter->dump_exceptions) {
5882 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5891 err = nexthop_for_each_fib6_nh(rt->nh,
5892 rt6_nh_dump_exceptions,
5895 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5900 return count += w.count;
5906 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5907 const struct nlmsghdr *nlh,
5909 struct netlink_ext_ack *extack)
5914 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5915 NL_SET_ERR_MSG_MOD(extack,
5916 "Invalid header for get route request");
5920 if (!netlink_strict_get_check(skb))
5921 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5922 rtm_ipv6_policy, extack);
5924 rtm = nlmsg_data(nlh);
5925 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5926 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5927 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5929 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5932 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5933 NL_SET_ERR_MSG_MOD(extack,
5934 "Invalid flags for get route request");
5938 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5939 rtm_ipv6_policy, extack);
5943 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5944 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5945 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5949 for (i = 0; i <= RTA_MAX; i++) {
5965 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5973 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5974 struct netlink_ext_ack *extack)
5976 struct net *net = sock_net(in_skb->sk);
5977 struct nlattr *tb[RTA_MAX+1];
5978 int err, iif = 0, oif = 0;
5979 struct fib6_info *from;
5980 struct dst_entry *dst;
5981 struct rt6_info *rt;
5982 struct sk_buff *skb;
5984 struct flowi6 fl6 = {};
5987 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
5992 rtm = nlmsg_data(nlh);
5993 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
5994 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
5997 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
6000 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
6004 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
6007 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
6011 iif = nla_get_u32(tb[RTA_IIF]);
6014 oif = nla_get_u32(tb[RTA_OIF]);
6017 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6020 fl6.flowi6_uid = make_kuid(current_user_ns(),
6021 nla_get_u32(tb[RTA_UID]));
6023 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6026 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6029 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6031 if (tb[RTA_IP_PROTO]) {
6032 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
6033 &fl6.flowi6_proto, AF_INET6,
6040 struct net_device *dev;
6045 dev = dev_get_by_index_rcu(net, iif);
6052 fl6.flowi6_iif = iif;
6054 if (!ipv6_addr_any(&fl6.saddr))
6055 flags |= RT6_LOOKUP_F_HAS_SADDR;
6057 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
6061 fl6.flowi6_oif = oif;
6063 dst = ip6_route_output(net, NULL, &fl6);
6067 rt = container_of(dst, struct rt6_info, dst);
6068 if (rt->dst.error) {
6069 err = rt->dst.error;
6074 if (rt == net->ipv6.ip6_null_entry) {
6075 err = rt->dst.error;
6080 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
6087 skb_dst_set(skb, &rt->dst);
6090 from = rcu_dereference(rt->from);
6093 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6095 NETLINK_CB(in_skb).portid,
6098 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6099 &fl6.saddr, iif, RTM_NEWROUTE,
6100 NETLINK_CB(in_skb).portid,
6112 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6117 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6118 unsigned int nlm_flags)
6120 struct sk_buff *skb;
6121 struct net *net = info->nl_net;
6126 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6128 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6132 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6133 event, info->portid, seq, nlm_flags);
6135 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6136 WARN_ON(err == -EMSGSIZE);
6140 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6141 info->nlh, gfp_any());
6145 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6148 void fib6_rt_update(struct net *net, struct fib6_info *rt,
6149 struct nl_info *info)
6151 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6152 struct sk_buff *skb;
6155 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6159 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6160 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6162 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6163 WARN_ON(err == -EMSGSIZE);
6167 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6168 info->nlh, gfp_any());
6172 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6175 void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6176 bool offload, bool trap, bool offload_failed)
6178 struct sk_buff *skb;
6181 if (f6i->offload == offload && f6i->trap == trap &&
6182 f6i->offload_failed == offload_failed)
6185 f6i->offload = offload;
6188 /* 2 means send notifications only if offload_failed was changed. */
6189 if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6190 f6i->offload_failed == offload_failed)
6193 f6i->offload_failed = offload_failed;
6195 if (!rcu_access_pointer(f6i->fib6_node))
6196 /* The route was removed from the tree, do not send
6201 if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6204 skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6210 err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6213 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6214 WARN_ON(err == -EMSGSIZE);
6219 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6223 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6225 EXPORT_SYMBOL(fib6_info_hw_flags_set);
6227 static int ip6_route_dev_notify(struct notifier_block *this,
6228 unsigned long event, void *ptr)
6230 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6231 struct net *net = dev_net(dev);
6233 if (!(dev->flags & IFF_LOOPBACK))
6236 if (event == NETDEV_REGISTER) {
6237 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6238 net->ipv6.ip6_null_entry->dst.dev = dev;
6239 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6240 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6241 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6242 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6243 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6244 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6246 } else if (event == NETDEV_UNREGISTER &&
6247 dev->reg_state != NETREG_UNREGISTERED) {
6248 /* NETDEV_UNREGISTER could be fired for multiple times by
6249 * netdev_wait_allrefs(). Make sure we only call this once.
6251 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6252 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6253 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6254 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6265 #ifdef CONFIG_PROC_FS
6266 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6268 struct net *net = (struct net *)seq->private;
6269 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6270 net->ipv6.rt6_stats->fib_nodes,
6271 net->ipv6.rt6_stats->fib_route_nodes,
6272 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6273 net->ipv6.rt6_stats->fib_rt_entries,
6274 net->ipv6.rt6_stats->fib_rt_cache,
6275 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6276 net->ipv6.rt6_stats->fib_discarded_routes);
6280 #endif /* CONFIG_PROC_FS */
6282 #ifdef CONFIG_SYSCTL
6284 static int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6285 void *buffer, size_t *lenp, loff_t *ppos)
6293 net = (struct net *)ctl->extra1;
6294 delay = net->ipv6.sysctl.flush_delay;
6295 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6299 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6303 static struct ctl_table ipv6_route_table_template[] = {
6305 .procname = "flush",
6306 .data = &init_net.ipv6.sysctl.flush_delay,
6307 .maxlen = sizeof(int),
6309 .proc_handler = ipv6_sysctl_rtcache_flush
6312 .procname = "gc_thresh",
6313 .data = &ip6_dst_ops_template.gc_thresh,
6314 .maxlen = sizeof(int),
6316 .proc_handler = proc_dointvec,
6319 .procname = "max_size",
6320 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
6321 .maxlen = sizeof(int),
6323 .proc_handler = proc_dointvec,
6326 .procname = "gc_min_interval",
6327 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6328 .maxlen = sizeof(int),
6330 .proc_handler = proc_dointvec_jiffies,
6333 .procname = "gc_timeout",
6334 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6335 .maxlen = sizeof(int),
6337 .proc_handler = proc_dointvec_jiffies,
6340 .procname = "gc_interval",
6341 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
6342 .maxlen = sizeof(int),
6344 .proc_handler = proc_dointvec_jiffies,
6347 .procname = "gc_elasticity",
6348 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6349 .maxlen = sizeof(int),
6351 .proc_handler = proc_dointvec,
6354 .procname = "mtu_expires",
6355 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6356 .maxlen = sizeof(int),
6358 .proc_handler = proc_dointvec_jiffies,
6361 .procname = "min_adv_mss",
6362 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
6363 .maxlen = sizeof(int),
6365 .proc_handler = proc_dointvec,
6368 .procname = "gc_min_interval_ms",
6369 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6370 .maxlen = sizeof(int),
6372 .proc_handler = proc_dointvec_ms_jiffies,
6375 .procname = "skip_notify_on_dev_down",
6376 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6377 .maxlen = sizeof(int),
6379 .proc_handler = proc_dointvec_minmax,
6380 .extra1 = SYSCTL_ZERO,
6381 .extra2 = SYSCTL_ONE,
6386 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6388 struct ctl_table *table;
6390 table = kmemdup(ipv6_route_table_template,
6391 sizeof(ipv6_route_table_template),
6395 table[0].data = &net->ipv6.sysctl.flush_delay;
6396 table[0].extra1 = net;
6397 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6398 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
6399 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6400 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6401 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6402 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6403 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6404 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6405 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6406 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6408 /* Don't export sysctls to unprivileged users */
6409 if (net->user_ns != &init_user_ns)
6410 table[0].procname = NULL;
6417 static int __net_init ip6_route_net_init(struct net *net)
6421 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6422 sizeof(net->ipv6.ip6_dst_ops));
6424 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6425 goto out_ip6_dst_ops;
6427 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6428 if (!net->ipv6.fib6_null_entry)
6429 goto out_ip6_dst_entries;
6430 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6431 sizeof(*net->ipv6.fib6_null_entry));
6433 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6434 sizeof(*net->ipv6.ip6_null_entry),
6436 if (!net->ipv6.ip6_null_entry)
6437 goto out_fib6_null_entry;
6438 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6439 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6440 ip6_template_metrics, true);
6441 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
6443 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6444 net->ipv6.fib6_has_custom_rules = false;
6445 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6446 sizeof(*net->ipv6.ip6_prohibit_entry),
6448 if (!net->ipv6.ip6_prohibit_entry)
6449 goto out_ip6_null_entry;
6450 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6451 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6452 ip6_template_metrics, true);
6453 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
6455 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6456 sizeof(*net->ipv6.ip6_blk_hole_entry),
6458 if (!net->ipv6.ip6_blk_hole_entry)
6459 goto out_ip6_prohibit_entry;
6460 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6461 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6462 ip6_template_metrics, true);
6463 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
6464 #ifdef CONFIG_IPV6_SUBTREES
6465 net->ipv6.fib6_routes_require_src = 0;
6469 net->ipv6.sysctl.flush_delay = 0;
6470 net->ipv6.sysctl.ip6_rt_max_size = 4096;
6471 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6472 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6473 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6474 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6475 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6476 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6477 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6479 net->ipv6.ip6_rt_gc_expire = 30*HZ;
6485 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6486 out_ip6_prohibit_entry:
6487 kfree(net->ipv6.ip6_prohibit_entry);
6489 kfree(net->ipv6.ip6_null_entry);
6491 out_fib6_null_entry:
6492 kfree(net->ipv6.fib6_null_entry);
6493 out_ip6_dst_entries:
6494 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6499 static void __net_exit ip6_route_net_exit(struct net *net)
6501 kfree(net->ipv6.fib6_null_entry);
6502 kfree(net->ipv6.ip6_null_entry);
6503 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6504 kfree(net->ipv6.ip6_prohibit_entry);
6505 kfree(net->ipv6.ip6_blk_hole_entry);
6507 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6510 static int __net_init ip6_route_net_init_late(struct net *net)
6512 #ifdef CONFIG_PROC_FS
6513 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6514 sizeof(struct ipv6_route_iter));
6515 proc_create_net_single("rt6_stats", 0444, net->proc_net,
6516 rt6_stats_seq_show, NULL);
6521 static void __net_exit ip6_route_net_exit_late(struct net *net)
6523 #ifdef CONFIG_PROC_FS
6524 remove_proc_entry("ipv6_route", net->proc_net);
6525 remove_proc_entry("rt6_stats", net->proc_net);
6529 static struct pernet_operations ip6_route_net_ops = {
6530 .init = ip6_route_net_init,
6531 .exit = ip6_route_net_exit,
6534 static int __net_init ipv6_inetpeer_init(struct net *net)
6536 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6540 inet_peer_base_init(bp);
6541 net->ipv6.peers = bp;
6545 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6547 struct inet_peer_base *bp = net->ipv6.peers;
6549 net->ipv6.peers = NULL;
6550 inetpeer_invalidate_tree(bp);
6554 static struct pernet_operations ipv6_inetpeer_ops = {
6555 .init = ipv6_inetpeer_init,
6556 .exit = ipv6_inetpeer_exit,
6559 static struct pernet_operations ip6_route_net_late_ops = {
6560 .init = ip6_route_net_init_late,
6561 .exit = ip6_route_net_exit_late,
6564 static struct notifier_block ip6_route_dev_notifier = {
6565 .notifier_call = ip6_route_dev_notify,
6566 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6569 void __init ip6_route_init_special_entries(void)
6571 /* Registering of the loopback is done before this portion of code,
6572 * the loopback reference in rt6_info will not be taken, do it
6573 * manually for init_net */
6574 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6575 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6576 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6577 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6578 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6579 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6580 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6581 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6585 #if IS_BUILTIN(CONFIG_IPV6)
6586 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6587 DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6589 BTF_ID_LIST(btf_fib6_info_id)
6590 BTF_ID(struct, fib6_info)
6592 static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6593 .seq_ops = &ipv6_route_seq_ops,
6594 .init_seq_private = bpf_iter_init_seq_net,
6595 .fini_seq_private = bpf_iter_fini_seq_net,
6596 .seq_priv_size = sizeof(struct ipv6_route_iter),
6599 static struct bpf_iter_reg ipv6_route_reg_info = {
6600 .target = "ipv6_route",
6601 .ctx_arg_info_size = 1,
6603 { offsetof(struct bpf_iter__ipv6_route, rt),
6604 PTR_TO_BTF_ID_OR_NULL },
6606 .seq_info = &ipv6_route_seq_info,
6609 static int __init bpf_iter_register(void)
6611 ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6612 return bpf_iter_reg_target(&ipv6_route_reg_info);
6615 static void bpf_iter_unregister(void)
6617 bpf_iter_unreg_target(&ipv6_route_reg_info);
6622 int __init ip6_route_init(void)
6628 ip6_dst_ops_template.kmem_cachep =
6629 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6630 SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL);
6631 if (!ip6_dst_ops_template.kmem_cachep)
6634 ret = dst_entries_init(&ip6_dst_blackhole_ops);
6636 goto out_kmem_cache;
6638 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6640 goto out_dst_entries;
6642 ret = register_pernet_subsys(&ip6_route_net_ops);
6644 goto out_register_inetpeer;
6646 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6650 goto out_register_subsys;
6656 ret = fib6_rules_init();
6660 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6662 goto fib6_rules_init;
6664 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6665 inet6_rtm_newroute, NULL, 0);
6667 goto out_register_late_subsys;
6669 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6670 inet6_rtm_delroute, NULL, 0);
6672 goto out_register_late_subsys;
6674 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6675 inet6_rtm_getroute, NULL,
6676 RTNL_FLAG_DOIT_UNLOCKED);
6678 goto out_register_late_subsys;
6680 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6682 goto out_register_late_subsys;
6684 #if IS_BUILTIN(CONFIG_IPV6)
6685 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6686 ret = bpf_iter_register();
6688 goto out_register_late_subsys;
6692 for_each_possible_cpu(cpu) {
6693 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6695 INIT_LIST_HEAD(&ul->head);
6696 spin_lock_init(&ul->lock);
6702 out_register_late_subsys:
6703 rtnl_unregister_all(PF_INET6);
6704 unregister_pernet_subsys(&ip6_route_net_late_ops);
6706 fib6_rules_cleanup();
6711 out_register_subsys:
6712 unregister_pernet_subsys(&ip6_route_net_ops);
6713 out_register_inetpeer:
6714 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6716 dst_entries_destroy(&ip6_dst_blackhole_ops);
6718 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6722 void ip6_route_cleanup(void)
6724 #if IS_BUILTIN(CONFIG_IPV6)
6725 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6726 bpf_iter_unregister();
6729 unregister_netdevice_notifier(&ip6_route_dev_notifier);
6730 unregister_pernet_subsys(&ip6_route_net_late_ops);
6731 fib6_rules_cleanup();
6734 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6735 unregister_pernet_subsys(&ip6_route_net_ops);
6736 dst_entries_destroy(&ip6_dst_blackhole_ops);
6737 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);