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 /* if skb is set it will be used and fl6 can be NULL */
2338 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2339 const struct sk_buff *skb, struct flow_keys *flkeys)
2341 struct flow_keys hash_keys;
2344 switch (ip6_multipath_hash_policy(net)) {
2346 memset(&hash_keys, 0, sizeof(hash_keys));
2347 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2349 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2351 hash_keys.addrs.v6addrs.src = fl6->saddr;
2352 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2353 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2354 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2359 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2360 struct flow_keys keys;
2362 /* short-circuit if we already have L4 hash present */
2364 return skb_get_hash_raw(skb) >> 1;
2366 memset(&hash_keys, 0, sizeof(hash_keys));
2369 skb_flow_dissect_flow_keys(skb, &keys, flag);
2372 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2373 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2374 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2375 hash_keys.ports.src = flkeys->ports.src;
2376 hash_keys.ports.dst = flkeys->ports.dst;
2377 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2379 memset(&hash_keys, 0, sizeof(hash_keys));
2380 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2381 hash_keys.addrs.v6addrs.src = fl6->saddr;
2382 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2383 hash_keys.ports.src = fl6->fl6_sport;
2384 hash_keys.ports.dst = fl6->fl6_dport;
2385 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2389 memset(&hash_keys, 0, sizeof(hash_keys));
2390 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2392 struct flow_keys keys;
2395 skb_flow_dissect_flow_keys(skb, &keys, 0);
2399 /* Inner can be v4 or v6 */
2400 if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2401 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2402 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2403 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2404 } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2405 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2406 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2407 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2408 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2409 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2411 /* Same as case 0 */
2412 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2413 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2416 /* Same as case 0 */
2417 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2418 hash_keys.addrs.v6addrs.src = fl6->saddr;
2419 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2420 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2421 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2425 mhash = flow_hash_from_keys(&hash_keys);
2430 /* Called with rcu held */
2431 void ip6_route_input(struct sk_buff *skb)
2433 const struct ipv6hdr *iph = ipv6_hdr(skb);
2434 struct net *net = dev_net(skb->dev);
2435 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2436 struct ip_tunnel_info *tun_info;
2437 struct flowi6 fl6 = {
2438 .flowi6_iif = skb->dev->ifindex,
2439 .daddr = iph->daddr,
2440 .saddr = iph->saddr,
2441 .flowlabel = ip6_flowinfo(iph),
2442 .flowi6_mark = skb->mark,
2443 .flowi6_proto = iph->nexthdr,
2445 struct flow_keys *flkeys = NULL, _flkeys;
2447 tun_info = skb_tunnel_info(skb);
2448 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2449 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2451 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2454 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2455 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2457 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2461 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
2462 struct fib6_table *table,
2464 const struct sk_buff *skb,
2467 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2470 struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2471 const struct sock *sk,
2472 struct flowi6 *fl6, int flags)
2476 if (ipv6_addr_type(&fl6->daddr) &
2477 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2478 struct dst_entry *dst;
2480 /* This function does not take refcnt on the dst */
2481 dst = l3mdev_link_scope_lookup(net, fl6);
2486 fl6->flowi6_iif = LOOPBACK_IFINDEX;
2488 flags |= RT6_LOOKUP_F_DST_NOREF;
2489 any_src = ipv6_addr_any(&fl6->saddr);
2490 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2491 (fl6->flowi6_oif && any_src))
2492 flags |= RT6_LOOKUP_F_IFACE;
2495 flags |= RT6_LOOKUP_F_HAS_SADDR;
2497 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
2499 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2501 EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2503 struct dst_entry *ip6_route_output_flags(struct net *net,
2504 const struct sock *sk,
2508 struct dst_entry *dst;
2509 struct rt6_info *rt6;
2512 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2513 rt6 = (struct rt6_info *)dst;
2514 /* For dst cached in uncached_list, refcnt is already taken. */
2515 if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2516 dst = &net->ipv6.ip6_null_entry->dst;
2523 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2525 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2527 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2528 struct net_device *loopback_dev = net->loopback_dev;
2529 struct dst_entry *new = NULL;
2531 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
2532 DST_OBSOLETE_DEAD, 0);
2535 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2539 new->input = dst_discard;
2540 new->output = dst_discard_out;
2542 dst_copy_metrics(new, &ort->dst);
2544 rt->rt6i_idev = in6_dev_get(loopback_dev);
2545 rt->rt6i_gateway = ort->rt6i_gateway;
2546 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2548 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2549 #ifdef CONFIG_IPV6_SUBTREES
2550 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2554 dst_release(dst_orig);
2555 return new ? new : ERR_PTR(-ENOMEM);
2559 * Destination cache support functions
2562 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2566 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2569 if (fib6_check_expired(f6i))
2575 static struct dst_entry *rt6_check(struct rt6_info *rt,
2576 struct fib6_info *from,
2581 if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2582 rt_cookie != cookie)
2585 if (rt6_check_expired(rt))
2591 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2592 struct fib6_info *from,
2595 if (!__rt6_check_expired(rt) &&
2596 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2597 fib6_check(from, cookie))
2603 INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2606 struct dst_entry *dst_ret;
2607 struct fib6_info *from;
2608 struct rt6_info *rt;
2610 rt = container_of(dst, struct rt6_info, dst);
2613 return rt6_is_valid(rt) ? dst : NULL;
2617 /* All IPV6 dsts are created with ->obsolete set to the value
2618 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2619 * into this function always.
2622 from = rcu_dereference(rt->from);
2624 if (from && (rt->rt6i_flags & RTF_PCPU ||
2625 unlikely(!list_empty(&rt->rt6i_uncached))))
2626 dst_ret = rt6_dst_from_check(rt, from, cookie);
2628 dst_ret = rt6_check(rt, from, cookie);
2634 EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
2636 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2638 struct rt6_info *rt = (struct rt6_info *) dst;
2641 if (rt->rt6i_flags & RTF_CACHE) {
2643 if (rt6_check_expired(rt)) {
2644 rt6_remove_exception_rt(rt);
2656 static void ip6_link_failure(struct sk_buff *skb)
2658 struct rt6_info *rt;
2660 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2662 rt = (struct rt6_info *) skb_dst(skb);
2665 if (rt->rt6i_flags & RTF_CACHE) {
2666 rt6_remove_exception_rt(rt);
2668 struct fib6_info *from;
2669 struct fib6_node *fn;
2671 from = rcu_dereference(rt->from);
2673 fn = rcu_dereference(from->fib6_node);
2674 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2682 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2684 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2685 struct fib6_info *from;
2688 from = rcu_dereference(rt0->from);
2690 rt0->dst.expires = from->expires;
2694 dst_set_expires(&rt0->dst, timeout);
2695 rt0->rt6i_flags |= RTF_EXPIRES;
2698 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2700 struct net *net = dev_net(rt->dst.dev);
2702 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2703 rt->rt6i_flags |= RTF_MODIFIED;
2704 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2707 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2709 return !(rt->rt6i_flags & RTF_CACHE) &&
2710 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2713 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2714 const struct ipv6hdr *iph, u32 mtu,
2717 const struct in6_addr *daddr, *saddr;
2718 struct rt6_info *rt6 = (struct rt6_info *)dst;
2720 /* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2721 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2722 * [see also comment in rt6_mtu_change_route()]
2726 daddr = &iph->daddr;
2727 saddr = &iph->saddr;
2729 daddr = &sk->sk_v6_daddr;
2730 saddr = &inet6_sk(sk)->saddr;
2737 dst_confirm_neigh(dst, daddr);
2739 if (mtu < IPV6_MIN_MTU)
2741 if (mtu >= dst_mtu(dst))
2744 if (!rt6_cache_allowed_for_pmtu(rt6)) {
2745 rt6_do_update_pmtu(rt6, mtu);
2746 /* update rt6_ex->stamp for cache */
2747 if (rt6->rt6i_flags & RTF_CACHE)
2748 rt6_update_exception_stamp_rt(rt6);
2750 struct fib6_result res = {};
2751 struct rt6_info *nrt6;
2754 res.f6i = rcu_dereference(rt6->from);
2758 res.fib6_flags = res.f6i->fib6_flags;
2759 res.fib6_type = res.f6i->fib6_type;
2762 struct fib6_nh_match_arg arg = {
2764 .gw = &rt6->rt6i_gateway,
2767 nexthop_for_each_fib6_nh(res.f6i->nh,
2768 fib6_nh_find_match, &arg);
2770 /* fib6_info uses a nexthop that does not have fib6_nh
2771 * using the dst->dev + gw. Should be impossible.
2778 res.nh = res.f6i->fib6_nh;
2781 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2783 rt6_do_update_pmtu(nrt6, mtu);
2784 if (rt6_insert_exception(nrt6, &res))
2785 dst_release_immediate(&nrt6->dst);
2792 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2793 struct sk_buff *skb, u32 mtu,
2796 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2800 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2801 int oif, u32 mark, kuid_t uid)
2803 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2804 struct dst_entry *dst;
2805 struct flowi6 fl6 = {
2807 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2808 .daddr = iph->daddr,
2809 .saddr = iph->saddr,
2810 .flowlabel = ip6_flowinfo(iph),
2814 dst = ip6_route_output(net, NULL, &fl6);
2816 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
2819 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2821 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2823 int oif = sk->sk_bound_dev_if;
2824 struct dst_entry *dst;
2826 if (!oif && skb->dev)
2827 oif = l3mdev_master_ifindex(skb->dev);
2829 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
2831 dst = __sk_dst_get(sk);
2832 if (!dst || !dst->obsolete ||
2833 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2837 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2838 ip6_datagram_dst_update(sk, false);
2841 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2843 void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2844 const struct flowi6 *fl6)
2846 #ifdef CONFIG_IPV6_SUBTREES
2847 struct ipv6_pinfo *np = inet6_sk(sk);
2850 ip6_dst_store(sk, dst,
2851 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2852 &sk->sk_v6_daddr : NULL,
2853 #ifdef CONFIG_IPV6_SUBTREES
2854 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2860 static bool ip6_redirect_nh_match(const struct fib6_result *res,
2862 const struct in6_addr *gw,
2863 struct rt6_info **ret)
2865 const struct fib6_nh *nh = res->nh;
2867 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2868 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2871 /* rt_cache's gateway might be different from its 'parent'
2872 * in the case of an ip redirect.
2873 * So we keep searching in the exception table if the gateway
2876 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2877 struct rt6_info *rt_cache;
2879 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
2881 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
2890 struct fib6_nh_rd_arg {
2891 struct fib6_result *res;
2893 const struct in6_addr *gw;
2894 struct rt6_info **ret;
2897 static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
2899 struct fib6_nh_rd_arg *arg = _arg;
2902 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
2905 /* Handle redirects */
2906 struct ip6rd_flowi {
2908 struct in6_addr gateway;
2911 INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
2912 struct fib6_table *table,
2914 const struct sk_buff *skb,
2917 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
2918 struct rt6_info *ret = NULL;
2919 struct fib6_result res = {};
2920 struct fib6_nh_rd_arg arg = {
2923 .gw = &rdfl->gateway,
2926 struct fib6_info *rt;
2927 struct fib6_node *fn;
2929 /* l3mdev_update_flow overrides oif if the device is enslaved; in
2930 * this case we must match on the real ingress device, so reset it
2932 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2933 fl6->flowi6_oif = skb->dev->ifindex;
2935 /* Get the "current" route for this destination and
2936 * check if the redirect has come from appropriate router.
2938 * RFC 4861 specifies that redirects should only be
2939 * accepted if they come from the nexthop to the target.
2940 * Due to the way the routes are chosen, this notion
2941 * is a bit fuzzy and one might need to check all possible
2946 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2948 for_each_fib6_node_rt_rcu(fn) {
2950 if (fib6_check_expired(rt))
2952 if (rt->fib6_flags & RTF_REJECT)
2954 if (unlikely(rt->nh)) {
2955 if (nexthop_is_blackhole(rt->nh))
2957 /* on match, res->nh is filled in and potentially ret */
2958 if (nexthop_for_each_fib6_nh(rt->nh,
2959 fib6_nh_redirect_match,
2963 res.nh = rt->fib6_nh;
2964 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
2971 rt = net->ipv6.fib6_null_entry;
2972 else if (rt->fib6_flags & RTF_REJECT) {
2973 ret = net->ipv6.ip6_null_entry;
2977 if (rt == net->ipv6.fib6_null_entry) {
2978 fn = fib6_backtrack(fn, &fl6->saddr);
2984 res.nh = rt->fib6_nh;
2987 ip6_hold_safe(net, &ret);
2989 res.fib6_flags = res.f6i->fib6_flags;
2990 res.fib6_type = res.f6i->fib6_type;
2991 ret = ip6_create_rt_rcu(&res);
2996 trace_fib6_table_lookup(net, &res, table, fl6);
3000 static struct dst_entry *ip6_route_redirect(struct net *net,
3001 const struct flowi6 *fl6,
3002 const struct sk_buff *skb,
3003 const struct in6_addr *gateway)
3005 int flags = RT6_LOOKUP_F_HAS_SADDR;
3006 struct ip6rd_flowi rdfl;
3009 rdfl.gateway = *gateway;
3011 return fib6_rule_lookup(net, &rdfl.fl6, skb,
3012 flags, __ip6_route_redirect);
3015 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3018 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3019 struct dst_entry *dst;
3020 struct flowi6 fl6 = {
3021 .flowi6_iif = LOOPBACK_IFINDEX,
3023 .flowi6_mark = mark,
3024 .daddr = iph->daddr,
3025 .saddr = iph->saddr,
3026 .flowlabel = ip6_flowinfo(iph),
3030 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3031 rt6_do_redirect(dst, NULL, skb);
3034 EXPORT_SYMBOL_GPL(ip6_redirect);
3036 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
3038 const struct ipv6hdr *iph = ipv6_hdr(skb);
3039 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3040 struct dst_entry *dst;
3041 struct flowi6 fl6 = {
3042 .flowi6_iif = LOOPBACK_IFINDEX,
3045 .saddr = iph->daddr,
3046 .flowi6_uid = sock_net_uid(net, NULL),
3049 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3050 rt6_do_redirect(dst, NULL, skb);
3054 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3056 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
3059 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3061 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3063 struct net_device *dev = dst->dev;
3064 unsigned int mtu = dst_mtu(dst);
3065 struct net *net = dev_net(dev);
3067 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3069 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3070 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3073 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3074 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3075 * IPV6_MAXPLEN is also valid and means: "any MSS,
3076 * rely only on pmtu discovery"
3078 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3083 INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
3085 struct inet6_dev *idev;
3088 mtu = dst_metric_raw(dst, RTAX_MTU);
3095 idev = __in6_dev_get(dst->dev);
3097 mtu = idev->cnf.mtu6;
3101 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3103 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
3105 EXPORT_INDIRECT_CALLABLE(ip6_mtu);
3108 * 1. mtu on route is locked - use it
3109 * 2. mtu from nexthop exception
3110 * 3. mtu from egress device
3112 * based on ip6_dst_mtu_forward and exception logic of
3113 * rt6_find_cached_rt; called with rcu_read_lock
3115 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3116 const struct in6_addr *daddr,
3117 const struct in6_addr *saddr)
3119 const struct fib6_nh *nh = res->nh;
3120 struct fib6_info *f6i = res->f6i;
3121 struct inet6_dev *idev;
3122 struct rt6_info *rt;
3125 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3126 mtu = f6i->fib6_pmtu;
3131 rt = rt6_find_cached_rt(res, daddr, saddr);
3133 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3135 struct net_device *dev = nh->fib_nh_dev;
3138 idev = __in6_dev_get(dev);
3139 if (idev && idev->cnf.mtu6 > mtu)
3140 mtu = idev->cnf.mtu6;
3143 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3145 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3148 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3151 struct dst_entry *dst;
3152 struct rt6_info *rt;
3153 struct inet6_dev *idev = in6_dev_get(dev);
3154 struct net *net = dev_net(dev);
3156 if (unlikely(!idev))
3157 return ERR_PTR(-ENODEV);
3159 rt = ip6_dst_alloc(net, dev, 0);
3160 if (unlikely(!rt)) {
3162 dst = ERR_PTR(-ENOMEM);
3166 rt->dst.input = ip6_input;
3167 rt->dst.output = ip6_output;
3168 rt->rt6i_gateway = fl6->daddr;
3169 rt->rt6i_dst.addr = fl6->daddr;
3170 rt->rt6i_dst.plen = 128;
3171 rt->rt6i_idev = idev;
3172 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3174 /* Add this dst into uncached_list so that rt6_disable_ip() can
3175 * do proper release of the net_device
3177 rt6_uncached_list_add(rt);
3178 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
3180 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3186 static int ip6_dst_gc(struct dst_ops *ops)
3188 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3189 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3190 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3191 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3192 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3193 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3196 entries = dst_entries_get_fast(ops);
3197 if (entries > rt_max_size)
3198 entries = dst_entries_get_slow(ops);
3200 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
3201 entries <= rt_max_size)
3204 net->ipv6.ip6_rt_gc_expire++;
3205 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
3206 entries = dst_entries_get_slow(ops);
3207 if (entries < ops->gc_thresh)
3208 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
3210 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
3211 return entries > rt_max_size;
3214 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3215 const struct in6_addr *gw_addr, u32 tbid,
3216 int flags, struct fib6_result *res)
3218 struct flowi6 fl6 = {
3219 .flowi6_oif = cfg->fc_ifindex,
3221 .saddr = cfg->fc_prefsrc,
3223 struct fib6_table *table;
3226 table = fib6_get_table(net, tbid);
3230 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3231 flags |= RT6_LOOKUP_F_HAS_SADDR;
3233 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3235 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3236 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3237 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3238 cfg->fc_ifindex != 0, NULL, flags);
3243 static int ip6_route_check_nh_onlink(struct net *net,
3244 struct fib6_config *cfg,
3245 const struct net_device *dev,
3246 struct netlink_ext_ack *extack)
3248 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3249 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3250 struct fib6_result res = {};
3253 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3254 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3255 /* ignore match if it is the default route */
3256 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3257 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3258 NL_SET_ERR_MSG(extack,
3259 "Nexthop has invalid gateway or device mismatch");
3266 static int ip6_route_check_nh(struct net *net,
3267 struct fib6_config *cfg,
3268 struct net_device **_dev,
3269 struct inet6_dev **idev)
3271 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3272 struct net_device *dev = _dev ? *_dev : NULL;
3273 int flags = RT6_LOOKUP_F_IFACE;
3274 struct fib6_result res = {};
3275 int err = -EHOSTUNREACH;
3277 if (cfg->fc_table) {
3278 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3279 cfg->fc_table, flags, &res);
3280 /* gw_addr can not require a gateway or resolve to a reject
3281 * route. If a device is given, it must match the result.
3283 if (err || res.fib6_flags & RTF_REJECT ||
3284 res.nh->fib_nh_gw_family ||
3285 (dev && dev != res.nh->fib_nh_dev))
3286 err = -EHOSTUNREACH;
3290 struct flowi6 fl6 = {
3291 .flowi6_oif = cfg->fc_ifindex,
3295 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3296 if (err || res.fib6_flags & RTF_REJECT ||
3297 res.nh->fib_nh_gw_family)
3298 err = -EHOSTUNREACH;
3303 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3304 cfg->fc_ifindex != 0, NULL, flags);
3309 if (dev != res.nh->fib_nh_dev)
3310 err = -EHOSTUNREACH;
3312 *_dev = dev = res.nh->fib_nh_dev;
3314 *idev = in6_dev_get(dev);
3320 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3321 struct net_device **_dev, struct inet6_dev **idev,
3322 struct netlink_ext_ack *extack)
3324 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3325 int gwa_type = ipv6_addr_type(gw_addr);
3326 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3327 const struct net_device *dev = *_dev;
3328 bool need_addr_check = !dev;
3331 /* if gw_addr is local we will fail to detect this in case
3332 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3333 * will return already-added prefix route via interface that
3334 * prefix route was assigned to, which might be non-loopback.
3337 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3338 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3342 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3343 /* IPv6 strictly inhibits using not link-local
3344 * addresses as nexthop address.
3345 * Otherwise, router will not able to send redirects.
3346 * It is very good, but in some (rare!) circumstances
3347 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3348 * some exceptions. --ANK
3349 * We allow IPv4-mapped nexthops to support RFC4798-type
3352 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3353 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3359 if (cfg->fc_flags & RTNH_F_ONLINK)
3360 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3362 err = ip6_route_check_nh(net, cfg, _dev, idev);
3370 /* reload in case device was changed */
3375 NL_SET_ERR_MSG(extack, "Egress device not specified");
3377 } else if (dev->flags & IFF_LOOPBACK) {
3378 NL_SET_ERR_MSG(extack,
3379 "Egress device can not be loopback device for this route");
3383 /* if we did not check gw_addr above, do so now that the
3384 * egress device has been resolved.
3386 if (need_addr_check &&
3387 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3388 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3397 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3399 if ((flags & RTF_REJECT) ||
3400 (dev && (dev->flags & IFF_LOOPBACK) &&
3401 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3402 !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3408 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3409 struct fib6_config *cfg, gfp_t gfp_flags,
3410 struct netlink_ext_ack *extack)
3412 struct net_device *dev = NULL;
3413 struct inet6_dev *idev = NULL;
3417 fib6_nh->fib_nh_family = AF_INET6;
3418 #ifdef CONFIG_IPV6_ROUTER_PREF
3419 fib6_nh->last_probe = jiffies;
3421 if (cfg->fc_is_fdb) {
3422 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3423 fib6_nh->fib_nh_gw_family = AF_INET6;
3428 if (cfg->fc_ifindex) {
3429 dev = dev_get_by_index(net, cfg->fc_ifindex);
3432 idev = in6_dev_get(dev);
3437 if (cfg->fc_flags & RTNH_F_ONLINK) {
3439 NL_SET_ERR_MSG(extack,
3440 "Nexthop device required for onlink");
3444 if (!(dev->flags & IFF_UP)) {
3445 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3450 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3453 fib6_nh->fib_nh_weight = 1;
3455 /* We cannot add true routes via loopback here,
3456 * they would result in kernel looping; promote them to reject routes
3458 addr_type = ipv6_addr_type(&cfg->fc_dst);
3459 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3460 /* hold loopback dev/idev if we haven't done so. */
3461 if (dev != net->loopback_dev) {
3466 dev = net->loopback_dev;
3468 idev = in6_dev_get(dev);
3477 if (cfg->fc_flags & RTF_GATEWAY) {
3478 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3482 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3483 fib6_nh->fib_nh_gw_family = AF_INET6;
3490 if (idev->cnf.disable_ipv6) {
3491 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3496 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3497 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3502 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3503 !netif_carrier_ok(dev))
3504 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3506 err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3507 cfg->fc_encap_type, cfg, gfp_flags, extack);
3512 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3513 if (!fib6_nh->rt6i_pcpu) {
3518 fib6_nh->fib_nh_dev = dev;
3519 fib6_nh->fib_nh_oif = dev->ifindex;
3526 lwtstate_put(fib6_nh->fib_nh_lws);
3527 fib6_nh->fib_nh_lws = NULL;
3535 void fib6_nh_release(struct fib6_nh *fib6_nh)
3537 struct rt6_exception_bucket *bucket;
3541 fib6_nh_flush_exceptions(fib6_nh, NULL);
3542 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3544 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3550 if (fib6_nh->rt6i_pcpu) {
3553 for_each_possible_cpu(cpu) {
3554 struct rt6_info **ppcpu_rt;
3555 struct rt6_info *pcpu_rt;
3557 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3558 pcpu_rt = *ppcpu_rt;
3560 dst_dev_put(&pcpu_rt->dst);
3561 dst_release(&pcpu_rt->dst);
3566 free_percpu(fib6_nh->rt6i_pcpu);
3569 fib_nh_common_release(&fib6_nh->nh_common);
3572 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3574 struct netlink_ext_ack *extack)
3576 struct net *net = cfg->fc_nlinfo.nl_net;
3577 struct fib6_info *rt = NULL;
3578 struct nexthop *nh = NULL;
3579 struct fib6_table *table;
3580 struct fib6_nh *fib6_nh;
3584 /* RTF_PCPU is an internal flag; can not be set by userspace */
3585 if (cfg->fc_flags & RTF_PCPU) {
3586 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3590 /* RTF_CACHE is an internal flag; can not be set by userspace */
3591 if (cfg->fc_flags & RTF_CACHE) {
3592 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3596 if (cfg->fc_type > RTN_MAX) {
3597 NL_SET_ERR_MSG(extack, "Invalid route type");
3601 if (cfg->fc_dst_len > 128) {
3602 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3605 if (cfg->fc_src_len > 128) {
3606 NL_SET_ERR_MSG(extack, "Invalid source address length");
3609 #ifndef CONFIG_IPV6_SUBTREES
3610 if (cfg->fc_src_len) {
3611 NL_SET_ERR_MSG(extack,
3612 "Specifying source address requires IPV6_SUBTREES to be enabled");
3616 if (cfg->fc_nh_id) {
3617 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3619 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3622 err = fib6_check_nexthop(nh, cfg, extack);
3628 if (cfg->fc_nlinfo.nlh &&
3629 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3630 table = fib6_get_table(net, cfg->fc_table);
3632 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3633 table = fib6_new_table(net, cfg->fc_table);
3636 table = fib6_new_table(net, cfg->fc_table);
3643 rt = fib6_info_alloc(gfp_flags, !nh);
3647 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3649 if (IS_ERR(rt->fib6_metrics)) {
3650 err = PTR_ERR(rt->fib6_metrics);
3651 /* Do not leave garbage there. */
3652 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3656 if (cfg->fc_flags & RTF_ADDRCONF)
3657 rt->dst_nocount = true;
3659 if (cfg->fc_flags & RTF_EXPIRES)
3660 fib6_set_expires(rt, jiffies +
3661 clock_t_to_jiffies(cfg->fc_expires));
3663 fib6_clean_expires(rt);
3665 if (cfg->fc_protocol == RTPROT_UNSPEC)
3666 cfg->fc_protocol = RTPROT_BOOT;
3667 rt->fib6_protocol = cfg->fc_protocol;
3669 rt->fib6_table = table;
3670 rt->fib6_metric = cfg->fc_metric;
3671 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3672 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3674 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3675 rt->fib6_dst.plen = cfg->fc_dst_len;
3677 #ifdef CONFIG_IPV6_SUBTREES
3678 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3679 rt->fib6_src.plen = cfg->fc_src_len;
3682 if (rt->fib6_src.plen) {
3683 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3686 if (!nexthop_get(nh)) {
3687 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3691 fib6_nh = nexthop_fib6_nh(rt->nh);
3693 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3697 fib6_nh = rt->fib6_nh;
3699 /* We cannot add true routes via loopback here, they would
3700 * result in kernel looping; promote them to reject routes
3702 addr_type = ipv6_addr_type(&cfg->fc_dst);
3703 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3705 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3708 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3709 struct net_device *dev = fib6_nh->fib_nh_dev;
3711 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3712 NL_SET_ERR_MSG(extack, "Invalid source address");
3716 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3717 rt->fib6_prefsrc.plen = 128;
3719 rt->fib6_prefsrc.plen = 0;
3723 fib6_info_release(rt);
3724 return ERR_PTR(err);
3726 ip_fib_metrics_put(rt->fib6_metrics);
3728 return ERR_PTR(err);
3731 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3732 struct netlink_ext_ack *extack)
3734 struct fib6_info *rt;
3737 rt = ip6_route_info_create(cfg, gfp_flags, extack);
3741 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3742 fib6_info_release(rt);
3747 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3749 struct net *net = info->nl_net;
3750 struct fib6_table *table;
3753 if (rt == net->ipv6.fib6_null_entry) {
3758 table = rt->fib6_table;
3759 spin_lock_bh(&table->tb6_lock);
3760 err = fib6_del(rt, info);
3761 spin_unlock_bh(&table->tb6_lock);
3764 fib6_info_release(rt);
3768 int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
3770 struct nl_info info = {
3772 .skip_notify = skip_notify
3775 return __ip6_del_rt(rt, &info);
3778 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3780 struct nl_info *info = &cfg->fc_nlinfo;
3781 struct net *net = info->nl_net;
3782 struct sk_buff *skb = NULL;
3783 struct fib6_table *table;
3786 if (rt == net->ipv6.fib6_null_entry)
3788 table = rt->fib6_table;
3789 spin_lock_bh(&table->tb6_lock);
3791 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3792 struct fib6_info *sibling, *next_sibling;
3793 struct fib6_node *fn;
3795 /* prefer to send a single notification with all hops */
3796 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3798 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3800 if (rt6_fill_node(net, skb, rt, NULL,
3801 NULL, NULL, 0, RTM_DELROUTE,
3802 info->portid, seq, 0) < 0) {
3806 info->skip_notify = 1;
3809 /* 'rt' points to the first sibling route. If it is not the
3810 * leaf, then we do not need to send a notification. Otherwise,
3811 * we need to check if the last sibling has a next route or not
3812 * and emit a replace or delete notification, respectively.
3814 info->skip_notify_kernel = 1;
3815 fn = rcu_dereference_protected(rt->fib6_node,
3816 lockdep_is_held(&table->tb6_lock));
3817 if (rcu_access_pointer(fn->leaf) == rt) {
3818 struct fib6_info *last_sibling, *replace_rt;
3820 last_sibling = list_last_entry(&rt->fib6_siblings,
3823 replace_rt = rcu_dereference_protected(
3824 last_sibling->fib6_next,
3825 lockdep_is_held(&table->tb6_lock));
3827 call_fib6_entry_notifiers_replace(net,
3830 call_fib6_multipath_entry_notifiers(net,
3831 FIB_EVENT_ENTRY_DEL,
3832 rt, rt->fib6_nsiblings,
3835 list_for_each_entry_safe(sibling, next_sibling,
3838 err = fib6_del(sibling, info);
3844 err = fib6_del(rt, info);
3846 spin_unlock_bh(&table->tb6_lock);
3848 fib6_info_release(rt);
3851 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3852 info->nlh, gfp_any());
3857 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3861 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3864 if (cfg->fc_flags & RTF_GATEWAY &&
3865 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3868 rc = rt6_remove_exception_rt(rt);
3873 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3876 struct fib6_result res = {
3880 struct rt6_info *rt_cache;
3882 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3884 return __ip6_del_cached_rt(rt_cache, cfg);
3889 struct fib6_nh_del_cached_rt_arg {
3890 struct fib6_config *cfg;
3891 struct fib6_info *f6i;
3894 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
3896 struct fib6_nh_del_cached_rt_arg *arg = _arg;
3899 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
3900 return rc != -ESRCH ? rc : 0;
3903 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
3905 struct fib6_nh_del_cached_rt_arg arg = {
3910 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
3913 static int ip6_route_del(struct fib6_config *cfg,
3914 struct netlink_ext_ack *extack)
3916 struct fib6_table *table;
3917 struct fib6_info *rt;
3918 struct fib6_node *fn;
3921 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
3923 NL_SET_ERR_MSG(extack, "FIB table does not exist");
3929 fn = fib6_locate(&table->tb6_root,
3930 &cfg->fc_dst, cfg->fc_dst_len,
3931 &cfg->fc_src, cfg->fc_src_len,
3932 !(cfg->fc_flags & RTF_CACHE));
3935 for_each_fib6_node_rt_rcu(fn) {
3938 if (rt->nh && cfg->fc_nh_id &&
3939 rt->nh->id != cfg->fc_nh_id)
3942 if (cfg->fc_flags & RTF_CACHE) {
3946 rc = ip6_del_cached_rt_nh(cfg, rt);
3947 } else if (cfg->fc_nh_id) {
3951 rc = ip6_del_cached_rt(cfg, rt, nh);
3960 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
3962 if (cfg->fc_protocol &&
3963 cfg->fc_protocol != rt->fib6_protocol)
3967 if (!fib6_info_hold_safe(rt))
3971 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3977 if (cfg->fc_ifindex &&
3979 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
3981 if (cfg->fc_flags & RTF_GATEWAY &&
3982 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
3984 if (!fib6_info_hold_safe(rt))
3988 /* if gateway was specified only delete the one hop */
3989 if (cfg->fc_flags & RTF_GATEWAY)
3990 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3992 return __ip6_del_rt_siblings(rt, cfg);
4000 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
4002 struct netevent_redirect netevent;
4003 struct rt6_info *rt, *nrt = NULL;
4004 struct fib6_result res = {};
4005 struct ndisc_options ndopts;
4006 struct inet6_dev *in6_dev;
4007 struct neighbour *neigh;
4009 int optlen, on_link;
4012 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
4013 optlen -= sizeof(*msg);
4016 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4020 msg = (struct rd_msg *)icmp6_hdr(skb);
4022 if (ipv6_addr_is_multicast(&msg->dest)) {
4023 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4028 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4030 } else if (ipv6_addr_type(&msg->target) !=
4031 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4032 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4036 in6_dev = __in6_dev_get(skb->dev);
4039 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
4043 * The IP source address of the Redirect MUST be the same as the current
4044 * first-hop router for the specified ICMP Destination Address.
4047 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4048 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4053 if (ndopts.nd_opts_tgt_lladdr) {
4054 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4057 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4062 rt = (struct rt6_info *) dst;
4063 if (rt->rt6i_flags & RTF_REJECT) {
4064 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4068 /* Redirect received -> path was valid.
4069 * Look, redirects are sent only in response to data packets,
4070 * so that this nexthop apparently is reachable. --ANK
4072 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4074 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4079 * We have finally decided to accept it.
4082 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4083 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4084 NEIGH_UPDATE_F_OVERRIDE|
4085 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4086 NEIGH_UPDATE_F_ISROUTER)),
4087 NDISC_REDIRECT, &ndopts);
4090 res.f6i = rcu_dereference(rt->from);
4095 struct fib6_nh_match_arg arg = {
4097 .gw = &rt->rt6i_gateway,
4100 nexthop_for_each_fib6_nh(res.f6i->nh,
4101 fib6_nh_find_match, &arg);
4103 /* fib6_info uses a nexthop that does not have fib6_nh
4104 * using the dst->dev. Should be impossible
4110 res.nh = res.f6i->fib6_nh;
4113 res.fib6_flags = res.f6i->fib6_flags;
4114 res.fib6_type = res.f6i->fib6_type;
4115 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4119 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4121 nrt->rt6i_flags &= ~RTF_GATEWAY;
4123 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4125 /* rt6_insert_exception() will take care of duplicated exceptions */
4126 if (rt6_insert_exception(nrt, &res)) {
4127 dst_release_immediate(&nrt->dst);
4131 netevent.old = &rt->dst;
4132 netevent.new = &nrt->dst;
4133 netevent.daddr = &msg->dest;
4134 netevent.neigh = neigh;
4135 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4139 neigh_release(neigh);
4142 #ifdef CONFIG_IPV6_ROUTE_INFO
4143 static struct fib6_info *rt6_get_route_info(struct net *net,
4144 const struct in6_addr *prefix, int prefixlen,
4145 const struct in6_addr *gwaddr,
4146 struct net_device *dev)
4148 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4149 int ifindex = dev->ifindex;
4150 struct fib6_node *fn;
4151 struct fib6_info *rt = NULL;
4152 struct fib6_table *table;
4154 table = fib6_get_table(net, tb_id);
4159 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4163 for_each_fib6_node_rt_rcu(fn) {
4164 /* these routes do not use nexthops */
4167 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4169 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4170 !rt->fib6_nh->fib_nh_gw_family)
4172 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4174 if (!fib6_info_hold_safe(rt))
4183 static struct fib6_info *rt6_add_route_info(struct net *net,
4184 const struct in6_addr *prefix, int prefixlen,
4185 const struct in6_addr *gwaddr,
4186 struct net_device *dev,
4189 struct fib6_config cfg = {
4190 .fc_metric = IP6_RT_PRIO_USER,
4191 .fc_ifindex = dev->ifindex,
4192 .fc_dst_len = prefixlen,
4193 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4194 RTF_UP | RTF_PREF(pref),
4195 .fc_protocol = RTPROT_RA,
4196 .fc_type = RTN_UNICAST,
4197 .fc_nlinfo.portid = 0,
4198 .fc_nlinfo.nlh = NULL,
4199 .fc_nlinfo.nl_net = net,
4202 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4203 cfg.fc_dst = *prefix;
4204 cfg.fc_gateway = *gwaddr;
4206 /* We should treat it as a default route if prefix length is 0. */
4208 cfg.fc_flags |= RTF_DEFAULT;
4210 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4212 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4216 struct fib6_info *rt6_get_dflt_router(struct net *net,
4217 const struct in6_addr *addr,
4218 struct net_device *dev)
4220 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4221 struct fib6_info *rt;
4222 struct fib6_table *table;
4224 table = fib6_get_table(net, tb_id);
4229 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4232 /* RA routes do not use nexthops */
4237 if (dev == nh->fib_nh_dev &&
4238 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4239 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4242 if (rt && !fib6_info_hold_safe(rt))
4248 struct fib6_info *rt6_add_dflt_router(struct net *net,
4249 const struct in6_addr *gwaddr,
4250 struct net_device *dev,
4252 u32 defrtr_usr_metric)
4254 struct fib6_config cfg = {
4255 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4256 .fc_metric = defrtr_usr_metric,
4257 .fc_ifindex = dev->ifindex,
4258 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4259 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4260 .fc_protocol = RTPROT_RA,
4261 .fc_type = RTN_UNICAST,
4262 .fc_nlinfo.portid = 0,
4263 .fc_nlinfo.nlh = NULL,
4264 .fc_nlinfo.nl_net = net,
4267 cfg.fc_gateway = *gwaddr;
4269 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4270 struct fib6_table *table;
4272 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4274 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4277 return rt6_get_dflt_router(net, gwaddr, dev);
4280 static void __rt6_purge_dflt_routers(struct net *net,
4281 struct fib6_table *table)
4283 struct fib6_info *rt;
4287 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4288 struct net_device *dev = fib6_info_nh_dev(rt);
4289 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4291 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4292 (!idev || idev->cnf.accept_ra != 2) &&
4293 fib6_info_hold_safe(rt)) {
4295 ip6_del_rt(net, rt, false);
4301 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4304 void rt6_purge_dflt_routers(struct net *net)
4306 struct fib6_table *table;
4307 struct hlist_head *head;
4312 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4313 head = &net->ipv6.fib_table_hash[h];
4314 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4315 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4316 __rt6_purge_dflt_routers(net, table);
4323 static void rtmsg_to_fib6_config(struct net *net,
4324 struct in6_rtmsg *rtmsg,
4325 struct fib6_config *cfg)
4327 *cfg = (struct fib6_config){
4328 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4330 .fc_ifindex = rtmsg->rtmsg_ifindex,
4331 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
4332 .fc_expires = rtmsg->rtmsg_info,
4333 .fc_dst_len = rtmsg->rtmsg_dst_len,
4334 .fc_src_len = rtmsg->rtmsg_src_len,
4335 .fc_flags = rtmsg->rtmsg_flags,
4336 .fc_type = rtmsg->rtmsg_type,
4338 .fc_nlinfo.nl_net = net,
4340 .fc_dst = rtmsg->rtmsg_dst,
4341 .fc_src = rtmsg->rtmsg_src,
4342 .fc_gateway = rtmsg->rtmsg_gateway,
4346 int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
4348 struct fib6_config cfg;
4351 if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4353 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4356 rtmsg_to_fib6_config(net, rtmsg, &cfg);
4361 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4364 err = ip6_route_del(&cfg, NULL);
4372 * Drop the packet on the floor
4375 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4377 struct dst_entry *dst = skb_dst(skb);
4378 struct net *net = dev_net(dst->dev);
4379 struct inet6_dev *idev;
4382 if (netif_is_l3_master(skb->dev) &&
4383 dst->dev == net->loopback_dev)
4384 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4386 idev = ip6_dst_idev(dst);
4388 switch (ipstats_mib_noroutes) {
4389 case IPSTATS_MIB_INNOROUTES:
4390 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4391 if (type == IPV6_ADDR_ANY) {
4392 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4396 case IPSTATS_MIB_OUTNOROUTES:
4397 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4401 /* Start over by dropping the dst for l3mdev case */
4402 if (netif_is_l3_master(skb->dev))
4405 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4410 static int ip6_pkt_discard(struct sk_buff *skb)
4412 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4415 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4417 skb->dev = skb_dst(skb)->dev;
4418 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4421 static int ip6_pkt_prohibit(struct sk_buff *skb)
4423 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4426 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4428 skb->dev = skb_dst(skb)->dev;
4429 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4433 * Allocate a dst for local (unicast / anycast) address.
4436 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4437 struct inet6_dev *idev,
4438 const struct in6_addr *addr,
4439 bool anycast, gfp_t gfp_flags)
4441 struct fib6_config cfg = {
4442 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4443 .fc_ifindex = idev->dev->ifindex,
4444 .fc_flags = RTF_UP | RTF_NONEXTHOP,
4447 .fc_protocol = RTPROT_KERNEL,
4448 .fc_nlinfo.nl_net = net,
4449 .fc_ignore_dev_down = true,
4451 struct fib6_info *f6i;
4454 cfg.fc_type = RTN_ANYCAST;
4455 cfg.fc_flags |= RTF_ANYCAST;
4457 cfg.fc_type = RTN_LOCAL;
4458 cfg.fc_flags |= RTF_LOCAL;
4461 f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
4463 f6i->dst_nocount = true;
4467 /* remove deleted ip from prefsrc entries */
4468 struct arg_dev_net_ip {
4469 struct net_device *dev;
4471 struct in6_addr *addr;
4474 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4476 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4477 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4478 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4481 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
4482 rt != net->ipv6.fib6_null_entry &&
4483 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
4484 spin_lock_bh(&rt6_exception_lock);
4485 /* remove prefsrc entry */
4486 rt->fib6_prefsrc.plen = 0;
4487 spin_unlock_bh(&rt6_exception_lock);
4492 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4494 struct net *net = dev_net(ifp->idev->dev);
4495 struct arg_dev_net_ip adni = {
4496 .dev = ifp->idev->dev,
4500 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4503 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
4505 /* Remove routers and update dst entries when gateway turn into host. */
4506 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4508 struct in6_addr *gateway = (struct in6_addr *)arg;
4511 /* RA routes do not use nexthops */
4516 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4517 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4520 /* Further clean up cached routes in exception table.
4521 * This is needed because cached route may have a different
4522 * gateway than its 'parent' in the case of an ip redirect.
4524 fib6_nh_exceptions_clean_tohost(nh, gateway);
4529 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4531 fib6_clean_all(net, fib6_clean_tohost, gateway);
4534 struct arg_netdev_event {
4535 const struct net_device *dev;
4537 unsigned char nh_flags;
4538 unsigned long event;
4542 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4544 struct fib6_info *iter;
4545 struct fib6_node *fn;
4547 fn = rcu_dereference_protected(rt->fib6_node,
4548 lockdep_is_held(&rt->fib6_table->tb6_lock));
4549 iter = rcu_dereference_protected(fn->leaf,
4550 lockdep_is_held(&rt->fib6_table->tb6_lock));
4552 if (iter->fib6_metric == rt->fib6_metric &&
4553 rt6_qualify_for_ecmp(iter))
4555 iter = rcu_dereference_protected(iter->fib6_next,
4556 lockdep_is_held(&rt->fib6_table->tb6_lock));
4562 /* only called for fib entries with builtin fib6_nh */
4563 static bool rt6_is_dead(const struct fib6_info *rt)
4565 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4566 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4567 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4573 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4575 struct fib6_info *iter;
4578 if (!rt6_is_dead(rt))
4579 total += rt->fib6_nh->fib_nh_weight;
4581 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4582 if (!rt6_is_dead(iter))
4583 total += iter->fib6_nh->fib_nh_weight;
4589 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4591 int upper_bound = -1;
4593 if (!rt6_is_dead(rt)) {
4594 *weight += rt->fib6_nh->fib_nh_weight;
4595 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4598 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4601 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4603 struct fib6_info *iter;
4606 rt6_upper_bound_set(rt, &weight, total);
4608 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4609 rt6_upper_bound_set(iter, &weight, total);
4612 void rt6_multipath_rebalance(struct fib6_info *rt)
4614 struct fib6_info *first;
4617 /* In case the entire multipath route was marked for flushing,
4618 * then there is no need to rebalance upon the removal of every
4621 if (!rt->fib6_nsiblings || rt->should_flush)
4624 /* During lookup routes are evaluated in order, so we need to
4625 * make sure upper bounds are assigned from the first sibling
4628 first = rt6_multipath_first_sibling(rt);
4629 if (WARN_ON_ONCE(!first))
4632 total = rt6_multipath_total_weight(first);
4633 rt6_multipath_upper_bound_set(first, total);
4636 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4638 const struct arg_netdev_event *arg = p_arg;
4639 struct net *net = dev_net(arg->dev);
4641 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4642 rt->fib6_nh->fib_nh_dev == arg->dev) {
4643 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4644 fib6_update_sernum_upto_root(net, rt);
4645 rt6_multipath_rebalance(rt);
4651 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4653 struct arg_netdev_event arg = {
4656 .nh_flags = nh_flags,
4660 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4661 arg.nh_flags |= RTNH_F_LINKDOWN;
4663 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4666 /* only called for fib entries with inline fib6_nh */
4667 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4668 const struct net_device *dev)
4670 struct fib6_info *iter;
4672 if (rt->fib6_nh->fib_nh_dev == dev)
4674 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4675 if (iter->fib6_nh->fib_nh_dev == dev)
4681 static void rt6_multipath_flush(struct fib6_info *rt)
4683 struct fib6_info *iter;
4685 rt->should_flush = 1;
4686 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4687 iter->should_flush = 1;
4690 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4691 const struct net_device *down_dev)
4693 struct fib6_info *iter;
4694 unsigned int dead = 0;
4696 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4697 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4699 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4700 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4701 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4707 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4708 const struct net_device *dev,
4709 unsigned char nh_flags)
4711 struct fib6_info *iter;
4713 if (rt->fib6_nh->fib_nh_dev == dev)
4714 rt->fib6_nh->fib_nh_flags |= nh_flags;
4715 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4716 if (iter->fib6_nh->fib_nh_dev == dev)
4717 iter->fib6_nh->fib_nh_flags |= nh_flags;
4720 /* called with write lock held for table with rt */
4721 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4723 const struct arg_netdev_event *arg = p_arg;
4724 const struct net_device *dev = arg->dev;
4725 struct net *net = dev_net(dev);
4727 if (rt == net->ipv6.fib6_null_entry || rt->nh)
4730 switch (arg->event) {
4731 case NETDEV_UNREGISTER:
4732 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4734 if (rt->should_flush)
4736 if (!rt->fib6_nsiblings)
4737 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4738 if (rt6_multipath_uses_dev(rt, dev)) {
4741 count = rt6_multipath_dead_count(rt, dev);
4742 if (rt->fib6_nsiblings + 1 == count) {
4743 rt6_multipath_flush(rt);
4746 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4748 fib6_update_sernum(net, rt);
4749 rt6_multipath_rebalance(rt);
4753 if (rt->fib6_nh->fib_nh_dev != dev ||
4754 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4756 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4757 rt6_multipath_rebalance(rt);
4764 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4766 struct arg_netdev_event arg = {
4772 struct net *net = dev_net(dev);
4774 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4775 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4777 fib6_clean_all(net, fib6_ifdown, &arg);
4780 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4782 rt6_sync_down_dev(dev, event);
4783 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4784 neigh_ifdown(&nd_tbl, dev);
4787 struct rt6_mtu_change_arg {
4788 struct net_device *dev;
4790 struct fib6_info *f6i;
4793 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4795 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4796 struct fib6_info *f6i = arg->f6i;
4798 /* For administrative MTU increase, there is no way to discover
4799 * IPv6 PMTU increase, so PMTU increase should be updated here.
4800 * Since RFC 1981 doesn't include administrative MTU increase
4801 * update PMTU increase is a MUST. (i.e. jumbo frame)
4803 if (nh->fib_nh_dev == arg->dev) {
4804 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4805 u32 mtu = f6i->fib6_pmtu;
4807 if (mtu >= arg->mtu ||
4808 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4809 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4811 spin_lock_bh(&rt6_exception_lock);
4812 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4813 spin_unlock_bh(&rt6_exception_lock);
4819 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4821 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4822 struct inet6_dev *idev;
4824 /* In IPv6 pmtu discovery is not optional,
4825 so that RTAX_MTU lock cannot disable it.
4826 We still use this lock to block changes
4827 caused by addrconf/ndisc.
4830 idev = __in6_dev_get(arg->dev);
4834 if (fib6_metric_locked(f6i, RTAX_MTU))
4839 /* fib6_nh_mtu_change only returns 0, so this is safe */
4840 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4844 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4847 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4849 struct rt6_mtu_change_arg arg = {
4854 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4857 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4858 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
4859 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
4860 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
4861 [RTA_OIF] = { .type = NLA_U32 },
4862 [RTA_IIF] = { .type = NLA_U32 },
4863 [RTA_PRIORITY] = { .type = NLA_U32 },
4864 [RTA_METRICS] = { .type = NLA_NESTED },
4865 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
4866 [RTA_PREF] = { .type = NLA_U8 },
4867 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4868 [RTA_ENCAP] = { .type = NLA_NESTED },
4869 [RTA_EXPIRES] = { .type = NLA_U32 },
4870 [RTA_UID] = { .type = NLA_U32 },
4871 [RTA_MARK] = { .type = NLA_U32 },
4872 [RTA_TABLE] = { .type = NLA_U32 },
4873 [RTA_IP_PROTO] = { .type = NLA_U8 },
4874 [RTA_SPORT] = { .type = NLA_U16 },
4875 [RTA_DPORT] = { .type = NLA_U16 },
4876 [RTA_NH_ID] = { .type = NLA_U32 },
4879 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
4880 struct fib6_config *cfg,
4881 struct netlink_ext_ack *extack)
4884 struct nlattr *tb[RTA_MAX+1];
4888 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4889 rtm_ipv6_policy, extack);
4894 rtm = nlmsg_data(nlh);
4896 *cfg = (struct fib6_config){
4897 .fc_table = rtm->rtm_table,
4898 .fc_dst_len = rtm->rtm_dst_len,
4899 .fc_src_len = rtm->rtm_src_len,
4901 .fc_protocol = rtm->rtm_protocol,
4902 .fc_type = rtm->rtm_type,
4904 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4905 .fc_nlinfo.nlh = nlh,
4906 .fc_nlinfo.nl_net = sock_net(skb->sk),
4909 if (rtm->rtm_type == RTN_UNREACHABLE ||
4910 rtm->rtm_type == RTN_BLACKHOLE ||
4911 rtm->rtm_type == RTN_PROHIBIT ||
4912 rtm->rtm_type == RTN_THROW)
4913 cfg->fc_flags |= RTF_REJECT;
4915 if (rtm->rtm_type == RTN_LOCAL)
4916 cfg->fc_flags |= RTF_LOCAL;
4918 if (rtm->rtm_flags & RTM_F_CLONED)
4919 cfg->fc_flags |= RTF_CACHE;
4921 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4923 if (tb[RTA_NH_ID]) {
4924 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
4925 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
4926 NL_SET_ERR_MSG(extack,
4927 "Nexthop specification and nexthop id are mutually exclusive");
4930 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
4933 if (tb[RTA_GATEWAY]) {
4934 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
4935 cfg->fc_flags |= RTF_GATEWAY;
4938 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
4943 int plen = (rtm->rtm_dst_len + 7) >> 3;
4945 if (nla_len(tb[RTA_DST]) < plen)
4948 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
4952 int plen = (rtm->rtm_src_len + 7) >> 3;
4954 if (nla_len(tb[RTA_SRC]) < plen)
4957 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
4960 if (tb[RTA_PREFSRC])
4961 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
4964 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4966 if (tb[RTA_PRIORITY])
4967 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4969 if (tb[RTA_METRICS]) {
4970 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4971 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
4975 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4977 if (tb[RTA_MULTIPATH]) {
4978 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4979 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
4981 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
4982 cfg->fc_mp_len, extack);
4988 pref = nla_get_u8(tb[RTA_PREF]);
4989 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4990 pref != ICMPV6_ROUTER_PREF_HIGH)
4991 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4992 cfg->fc_flags |= RTF_PREF(pref);
4996 cfg->fc_encap = tb[RTA_ENCAP];
4998 if (tb[RTA_ENCAP_TYPE]) {
4999 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5001 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5006 if (tb[RTA_EXPIRES]) {
5007 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5009 if (addrconf_finite_timeout(timeout)) {
5010 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5011 cfg->fc_flags |= RTF_EXPIRES;
5021 struct fib6_info *fib6_info;
5022 struct fib6_config r_cfg;
5023 struct list_head next;
5026 static int ip6_route_info_append(struct net *net,
5027 struct list_head *rt6_nh_list,
5028 struct fib6_info *rt,
5029 struct fib6_config *r_cfg)
5034 list_for_each_entry(nh, rt6_nh_list, next) {
5035 /* check if fib6_info already exists */
5036 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5040 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5044 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5045 list_add_tail(&nh->next, rt6_nh_list);
5050 static void ip6_route_mpath_notify(struct fib6_info *rt,
5051 struct fib6_info *rt_last,
5052 struct nl_info *info,
5055 /* if this is an APPEND route, then rt points to the first route
5056 * inserted and rt_last points to last route inserted. Userspace
5057 * wants a consistent dump of the route which starts at the first
5058 * nexthop. Since sibling routes are always added at the end of
5059 * the list, find the first sibling of the last route appended
5061 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5062 rt = list_first_entry(&rt_last->fib6_siblings,
5068 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5071 static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5073 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5074 bool should_notify = false;
5075 struct fib6_info *leaf;
5076 struct fib6_node *fn;
5079 fn = rcu_dereference(rt->fib6_node);
5083 leaf = rcu_dereference(fn->leaf);
5088 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5089 rt6_qualify_for_ecmp(leaf)))
5090 should_notify = true;
5094 return should_notify;
5097 static int ip6_route_multipath_add(struct fib6_config *cfg,
5098 struct netlink_ext_ack *extack)
5100 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5101 struct nl_info *info = &cfg->fc_nlinfo;
5102 struct fib6_config r_cfg;
5103 struct rtnexthop *rtnh;
5104 struct fib6_info *rt;
5105 struct rt6_nh *err_nh;
5106 struct rt6_nh *nh, *nh_safe;
5112 int replace = (cfg->fc_nlinfo.nlh &&
5113 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5114 LIST_HEAD(rt6_nh_list);
5116 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5117 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5118 nlflags |= NLM_F_APPEND;
5120 remaining = cfg->fc_mp_len;
5121 rtnh = (struct rtnexthop *)cfg->fc_mp;
5123 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
5124 * fib6_info structs per nexthop
5126 while (rtnh_ok(rtnh, remaining)) {
5127 memcpy(&r_cfg, cfg, sizeof(*cfg));
5128 if (rtnh->rtnh_ifindex)
5129 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5131 attrlen = rtnh_attrlen(rtnh);
5133 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5135 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5137 r_cfg.fc_gateway = nla_get_in6_addr(nla);
5138 r_cfg.fc_flags |= RTF_GATEWAY;
5140 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5141 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5143 r_cfg.fc_encap_type = nla_get_u16(nla);
5146 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5147 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5153 if (!rt6_qualify_for_ecmp(rt)) {
5155 NL_SET_ERR_MSG(extack,
5156 "Device only routes can not be added for IPv6 using the multipath API.");
5157 fib6_info_release(rt);
5161 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5163 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5166 fib6_info_release(rt);
5170 rtnh = rtnh_next(rtnh, &remaining);
5173 if (list_empty(&rt6_nh_list)) {
5174 NL_SET_ERR_MSG(extack,
5175 "Invalid nexthop configuration - no valid nexthops");
5179 /* for add and replace send one notification with all nexthops.
5180 * Skip the notification in fib6_add_rt2node and send one with
5181 * the full route when done
5183 info->skip_notify = 1;
5185 /* For add and replace, send one notification with all nexthops. For
5186 * append, send one notification with all appended nexthops.
5188 info->skip_notify_kernel = 1;
5191 list_for_each_entry(nh, &rt6_nh_list, next) {
5192 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5193 fib6_info_release(nh->fib6_info);
5196 /* save reference to last route successfully inserted */
5197 rt_last = nh->fib6_info;
5199 /* save reference to first route for notification */
5201 rt_notif = nh->fib6_info;
5204 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5205 nh->fib6_info = NULL;
5208 NL_SET_ERR_MSG_MOD(extack,
5209 "multipath route replace failed (check consistency of installed routes)");
5214 /* Because each route is added like a single route we remove
5215 * these flags after the first nexthop: if there is a collision,
5216 * we have already failed to add the first nexthop:
5217 * fib6_add_rt2node() has rejected it; when replacing, old
5218 * nexthops have been replaced by first new, the rest should
5221 if (cfg->fc_nlinfo.nlh) {
5222 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5224 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5229 /* An in-kernel notification should only be sent in case the new
5230 * multipath route is added as the first route in the node, or if
5231 * it was appended to it. We pass 'rt_notif' since it is the first
5232 * sibling and might allow us to skip some checks in the replace case.
5234 if (ip6_route_mpath_should_notify(rt_notif)) {
5235 enum fib_event_type fib_event;
5237 if (rt_notif->fib6_nsiblings != nhn - 1)
5238 fib_event = FIB_EVENT_ENTRY_APPEND;
5240 fib_event = FIB_EVENT_ENTRY_REPLACE;
5242 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5243 fib_event, rt_notif,
5246 /* Delete all the siblings that were just added */
5252 /* success ... tell user about new route */
5253 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5257 /* send notification for routes that were added so that
5258 * the delete notifications sent by ip6_route_del are
5262 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5264 /* Delete routes that were already added */
5265 list_for_each_entry(nh, &rt6_nh_list, next) {
5268 ip6_route_del(&nh->r_cfg, extack);
5272 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5274 fib6_info_release(nh->fib6_info);
5275 list_del(&nh->next);
5282 static int ip6_route_multipath_del(struct fib6_config *cfg,
5283 struct netlink_ext_ack *extack)
5285 struct fib6_config r_cfg;
5286 struct rtnexthop *rtnh;
5292 remaining = cfg->fc_mp_len;
5293 rtnh = (struct rtnexthop *)cfg->fc_mp;
5295 /* Parse a Multipath Entry */
5296 while (rtnh_ok(rtnh, remaining)) {
5297 memcpy(&r_cfg, cfg, sizeof(*cfg));
5298 if (rtnh->rtnh_ifindex)
5299 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5301 attrlen = rtnh_attrlen(rtnh);
5303 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5305 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5307 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5308 r_cfg.fc_flags |= RTF_GATEWAY;
5311 err = ip6_route_del(&r_cfg, extack);
5315 rtnh = rtnh_next(rtnh, &remaining);
5321 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5322 struct netlink_ext_ack *extack)
5324 struct fib6_config cfg;
5327 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5332 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5333 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5338 return ip6_route_multipath_del(&cfg, extack);
5340 cfg.fc_delete_all_nh = 1;
5341 return ip6_route_del(&cfg, extack);
5345 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5346 struct netlink_ext_ack *extack)
5348 struct fib6_config cfg;
5351 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5355 if (cfg.fc_metric == 0)
5356 cfg.fc_metric = IP6_RT_PRIO_USER;
5359 return ip6_route_multipath_add(&cfg, extack);
5361 return ip6_route_add(&cfg, GFP_KERNEL, extack);
5364 /* add the overhead of this fib6_nh to nexthop_len */
5365 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5367 int *nexthop_len = arg;
5369 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5370 + NLA_ALIGN(sizeof(struct rtnexthop))
5371 + nla_total_size(16); /* RTA_GATEWAY */
5373 if (nh->fib_nh_lws) {
5374 /* RTA_ENCAP_TYPE */
5375 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5377 *nexthop_len += nla_total_size(2);
5383 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5388 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5389 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5392 struct fib6_nh *nh = f6i->fib6_nh;
5395 if (f6i->fib6_nsiblings) {
5396 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
5397 + NLA_ALIGN(sizeof(struct rtnexthop))
5398 + nla_total_size(16) /* RTA_GATEWAY */
5399 + lwtunnel_get_encap_size(nh->fib_nh_lws);
5401 nexthop_len *= f6i->fib6_nsiblings;
5403 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5406 return NLMSG_ALIGN(sizeof(struct rtmsg))
5407 + nla_total_size(16) /* RTA_SRC */
5408 + nla_total_size(16) /* RTA_DST */
5409 + nla_total_size(16) /* RTA_GATEWAY */
5410 + nla_total_size(16) /* RTA_PREFSRC */
5411 + nla_total_size(4) /* RTA_TABLE */
5412 + nla_total_size(4) /* RTA_IIF */
5413 + nla_total_size(4) /* RTA_OIF */
5414 + nla_total_size(4) /* RTA_PRIORITY */
5415 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5416 + nla_total_size(sizeof(struct rta_cacheinfo))
5417 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5418 + nla_total_size(1) /* RTA_PREF */
5422 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5423 unsigned char *flags)
5425 if (nexthop_is_multipath(nh)) {
5428 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5430 goto nla_put_failure;
5432 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5433 goto nla_put_failure;
5435 nla_nest_end(skb, mp);
5437 struct fib6_nh *fib6_nh;
5439 fib6_nh = nexthop_fib6_nh(nh);
5440 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5442 goto nla_put_failure;
5451 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5452 struct fib6_info *rt, struct dst_entry *dst,
5453 struct in6_addr *dest, struct in6_addr *src,
5454 int iif, int type, u32 portid, u32 seq,
5457 struct rt6_info *rt6 = (struct rt6_info *)dst;
5458 struct rt6key *rt6_dst, *rt6_src;
5459 u32 *pmetrics, table, rt6_flags;
5460 unsigned char nh_flags = 0;
5461 struct nlmsghdr *nlh;
5465 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5470 rt6_dst = &rt6->rt6i_dst;
5471 rt6_src = &rt6->rt6i_src;
5472 rt6_flags = rt6->rt6i_flags;
5474 rt6_dst = &rt->fib6_dst;
5475 rt6_src = &rt->fib6_src;
5476 rt6_flags = rt->fib6_flags;
5479 rtm = nlmsg_data(nlh);
5480 rtm->rtm_family = AF_INET6;
5481 rtm->rtm_dst_len = rt6_dst->plen;
5482 rtm->rtm_src_len = rt6_src->plen;
5485 table = rt->fib6_table->tb6_id;
5487 table = RT6_TABLE_UNSPEC;
5488 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5489 if (nla_put_u32(skb, RTA_TABLE, table))
5490 goto nla_put_failure;
5492 rtm->rtm_type = rt->fib6_type;
5494 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5495 rtm->rtm_protocol = rt->fib6_protocol;
5497 if (rt6_flags & RTF_CACHE)
5498 rtm->rtm_flags |= RTM_F_CLONED;
5501 if (nla_put_in6_addr(skb, RTA_DST, dest))
5502 goto nla_put_failure;
5503 rtm->rtm_dst_len = 128;
5504 } else if (rtm->rtm_dst_len)
5505 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5506 goto nla_put_failure;
5507 #ifdef CONFIG_IPV6_SUBTREES
5509 if (nla_put_in6_addr(skb, RTA_SRC, src))
5510 goto nla_put_failure;
5511 rtm->rtm_src_len = 128;
5512 } else if (rtm->rtm_src_len &&
5513 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5514 goto nla_put_failure;
5517 #ifdef CONFIG_IPV6_MROUTE
5518 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5519 int err = ip6mr_get_route(net, skb, rtm, portid);
5524 goto nla_put_failure;
5527 if (nla_put_u32(skb, RTA_IIF, iif))
5528 goto nla_put_failure;
5530 struct in6_addr saddr_buf;
5531 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
5532 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5533 goto nla_put_failure;
5536 if (rt->fib6_prefsrc.plen) {
5537 struct in6_addr saddr_buf;
5538 saddr_buf = rt->fib6_prefsrc.addr;
5539 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5540 goto nla_put_failure;
5543 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5544 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5545 goto nla_put_failure;
5547 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5548 goto nla_put_failure;
5550 /* For multipath routes, walk the siblings list and add
5551 * each as a nexthop within RTA_MULTIPATH.
5554 if (rt6_flags & RTF_GATEWAY &&
5555 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5556 goto nla_put_failure;
5558 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5559 goto nla_put_failure;
5561 if (dst->lwtstate &&
5562 lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5563 goto nla_put_failure;
5564 } else if (rt->fib6_nsiblings) {
5565 struct fib6_info *sibling, *next_sibling;
5568 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5570 goto nla_put_failure;
5572 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5573 rt->fib6_nh->fib_nh_weight, AF_INET6) < 0)
5574 goto nla_put_failure;
5576 list_for_each_entry_safe(sibling, next_sibling,
5577 &rt->fib6_siblings, fib6_siblings) {
5578 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5579 sibling->fib6_nh->fib_nh_weight,
5581 goto nla_put_failure;
5584 nla_nest_end(skb, mp);
5585 } else if (rt->nh) {
5586 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5587 goto nla_put_failure;
5589 if (nexthop_is_blackhole(rt->nh))
5590 rtm->rtm_type = RTN_BLACKHOLE;
5592 if (net->ipv4.sysctl_nexthop_compat_mode &&
5593 rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5594 goto nla_put_failure;
5596 rtm->rtm_flags |= nh_flags;
5598 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5599 &nh_flags, false) < 0)
5600 goto nla_put_failure;
5602 rtm->rtm_flags |= nh_flags;
5605 if (rt6_flags & RTF_EXPIRES) {
5606 expires = dst ? dst->expires : rt->expires;
5612 rtm->rtm_flags |= RTM_F_OFFLOAD;
5614 rtm->rtm_flags |= RTM_F_TRAP;
5615 if (rt->offload_failed)
5616 rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5619 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5620 goto nla_put_failure;
5622 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5623 goto nla_put_failure;
5626 nlmsg_end(skb, nlh);
5630 nlmsg_cancel(skb, nlh);
5634 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5636 const struct net_device *dev = arg;
5638 if (nh->fib_nh_dev == dev)
5644 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5645 const struct net_device *dev)
5648 struct net_device *_dev = (struct net_device *)dev;
5650 return !!nexthop_for_each_fib6_nh(f6i->nh,
5651 fib6_info_nh_uses_dev,
5655 if (f6i->fib6_nh->fib_nh_dev == dev)
5658 if (f6i->fib6_nsiblings) {
5659 struct fib6_info *sibling, *next_sibling;
5661 list_for_each_entry_safe(sibling, next_sibling,
5662 &f6i->fib6_siblings, fib6_siblings) {
5663 if (sibling->fib6_nh->fib_nh_dev == dev)
5671 struct fib6_nh_exception_dump_walker {
5672 struct rt6_rtnl_dump_arg *dump;
5673 struct fib6_info *rt;
5679 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5681 struct fib6_nh_exception_dump_walker *w = arg;
5682 struct rt6_rtnl_dump_arg *dump = w->dump;
5683 struct rt6_exception_bucket *bucket;
5684 struct rt6_exception *rt6_ex;
5687 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5691 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5692 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5698 /* Expiration of entries doesn't bump sernum, insertion
5699 * does. Removal is triggered by insertion, so we can
5700 * rely on the fact that if entries change between two
5701 * partial dumps, this node is scanned again completely,
5702 * see rt6_insert_exception() and fib6_dump_table().
5704 * Count expired entries we go through as handled
5705 * entries that we'll skip next time, in case of partial
5706 * node dump. Otherwise, if entries expire meanwhile,
5707 * we'll skip the wrong amount.
5709 if (rt6_check_expired(rt6_ex->rt6i)) {
5714 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5715 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5717 NETLINK_CB(dump->cb->skb).portid,
5718 dump->cb->nlh->nlmsg_seq, w->flags);
5730 /* Return -1 if done with node, number of handled routes on partial dump */
5731 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5733 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5734 struct fib_dump_filter *filter = &arg->filter;
5735 unsigned int flags = NLM_F_MULTI;
5736 struct net *net = arg->net;
5739 if (rt == net->ipv6.fib6_null_entry)
5742 if ((filter->flags & RTM_F_PREFIX) &&
5743 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5744 /* success since this is not a prefix route */
5747 if (filter->filter_set &&
5748 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5749 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5750 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5754 if (filter->filter_set ||
5755 !filter->dump_routes || !filter->dump_exceptions) {
5756 flags |= NLM_F_DUMP_FILTERED;
5759 if (filter->dump_routes) {
5763 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5765 NETLINK_CB(arg->cb->skb).portid,
5766 arg->cb->nlh->nlmsg_seq, flags)) {
5773 if (filter->dump_exceptions) {
5774 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5783 err = nexthop_for_each_fib6_nh(rt->nh,
5784 rt6_nh_dump_exceptions,
5787 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5792 return count += w.count;
5798 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5799 const struct nlmsghdr *nlh,
5801 struct netlink_ext_ack *extack)
5806 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5807 NL_SET_ERR_MSG_MOD(extack,
5808 "Invalid header for get route request");
5812 if (!netlink_strict_get_check(skb))
5813 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5814 rtm_ipv6_policy, extack);
5816 rtm = nlmsg_data(nlh);
5817 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5818 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5819 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5821 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5824 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5825 NL_SET_ERR_MSG_MOD(extack,
5826 "Invalid flags for get route request");
5830 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5831 rtm_ipv6_policy, extack);
5835 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5836 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5837 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5841 for (i = 0; i <= RTA_MAX; i++) {
5857 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5865 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5866 struct netlink_ext_ack *extack)
5868 struct net *net = sock_net(in_skb->sk);
5869 struct nlattr *tb[RTA_MAX+1];
5870 int err, iif = 0, oif = 0;
5871 struct fib6_info *from;
5872 struct dst_entry *dst;
5873 struct rt6_info *rt;
5874 struct sk_buff *skb;
5876 struct flowi6 fl6 = {};
5879 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
5884 rtm = nlmsg_data(nlh);
5885 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
5886 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
5889 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
5892 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
5896 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
5899 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
5903 iif = nla_get_u32(tb[RTA_IIF]);
5906 oif = nla_get_u32(tb[RTA_OIF]);
5909 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
5912 fl6.flowi6_uid = make_kuid(current_user_ns(),
5913 nla_get_u32(tb[RTA_UID]));
5915 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
5918 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
5921 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
5923 if (tb[RTA_IP_PROTO]) {
5924 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
5925 &fl6.flowi6_proto, AF_INET6,
5932 struct net_device *dev;
5937 dev = dev_get_by_index_rcu(net, iif);
5944 fl6.flowi6_iif = iif;
5946 if (!ipv6_addr_any(&fl6.saddr))
5947 flags |= RT6_LOOKUP_F_HAS_SADDR;
5949 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
5953 fl6.flowi6_oif = oif;
5955 dst = ip6_route_output(net, NULL, &fl6);
5959 rt = container_of(dst, struct rt6_info, dst);
5960 if (rt->dst.error) {
5961 err = rt->dst.error;
5966 if (rt == net->ipv6.ip6_null_entry) {
5967 err = rt->dst.error;
5972 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
5979 skb_dst_set(skb, &rt->dst);
5982 from = rcu_dereference(rt->from);
5985 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
5987 NETLINK_CB(in_skb).portid,
5990 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
5991 &fl6.saddr, iif, RTM_NEWROUTE,
5992 NETLINK_CB(in_skb).portid,
6004 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6009 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6010 unsigned int nlm_flags)
6012 struct sk_buff *skb;
6013 struct net *net = info->nl_net;
6018 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6020 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6024 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6025 event, info->portid, seq, nlm_flags);
6027 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6028 WARN_ON(err == -EMSGSIZE);
6032 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6033 info->nlh, gfp_any());
6037 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6040 void fib6_rt_update(struct net *net, struct fib6_info *rt,
6041 struct nl_info *info)
6043 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6044 struct sk_buff *skb;
6047 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6051 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6052 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6054 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6055 WARN_ON(err == -EMSGSIZE);
6059 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6060 info->nlh, gfp_any());
6064 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6067 void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6068 bool offload, bool trap, bool offload_failed)
6070 struct sk_buff *skb;
6073 if (f6i->offload == offload && f6i->trap == trap &&
6074 f6i->offload_failed == offload_failed)
6077 f6i->offload = offload;
6080 /* 2 means send notifications only if offload_failed was changed. */
6081 if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6082 f6i->offload_failed == offload_failed)
6085 f6i->offload_failed = offload_failed;
6087 if (!rcu_access_pointer(f6i->fib6_node))
6088 /* The route was removed from the tree, do not send
6093 if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6096 skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6102 err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6105 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6106 WARN_ON(err == -EMSGSIZE);
6111 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6115 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6117 EXPORT_SYMBOL(fib6_info_hw_flags_set);
6119 static int ip6_route_dev_notify(struct notifier_block *this,
6120 unsigned long event, void *ptr)
6122 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6123 struct net *net = dev_net(dev);
6125 if (!(dev->flags & IFF_LOOPBACK))
6128 if (event == NETDEV_REGISTER) {
6129 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6130 net->ipv6.ip6_null_entry->dst.dev = dev;
6131 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6132 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6133 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6134 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6135 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6136 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6138 } else if (event == NETDEV_UNREGISTER &&
6139 dev->reg_state != NETREG_UNREGISTERED) {
6140 /* NETDEV_UNREGISTER could be fired for multiple times by
6141 * netdev_wait_allrefs(). Make sure we only call this once.
6143 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6144 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6145 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6146 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6157 #ifdef CONFIG_PROC_FS
6158 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6160 struct net *net = (struct net *)seq->private;
6161 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6162 net->ipv6.rt6_stats->fib_nodes,
6163 net->ipv6.rt6_stats->fib_route_nodes,
6164 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6165 net->ipv6.rt6_stats->fib_rt_entries,
6166 net->ipv6.rt6_stats->fib_rt_cache,
6167 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6168 net->ipv6.rt6_stats->fib_discarded_routes);
6172 #endif /* CONFIG_PROC_FS */
6174 #ifdef CONFIG_SYSCTL
6176 static int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6177 void *buffer, size_t *lenp, loff_t *ppos)
6185 net = (struct net *)ctl->extra1;
6186 delay = net->ipv6.sysctl.flush_delay;
6187 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6191 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6195 static struct ctl_table ipv6_route_table_template[] = {
6197 .procname = "flush",
6198 .data = &init_net.ipv6.sysctl.flush_delay,
6199 .maxlen = sizeof(int),
6201 .proc_handler = ipv6_sysctl_rtcache_flush
6204 .procname = "gc_thresh",
6205 .data = &ip6_dst_ops_template.gc_thresh,
6206 .maxlen = sizeof(int),
6208 .proc_handler = proc_dointvec,
6211 .procname = "max_size",
6212 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
6213 .maxlen = sizeof(int),
6215 .proc_handler = proc_dointvec,
6218 .procname = "gc_min_interval",
6219 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6220 .maxlen = sizeof(int),
6222 .proc_handler = proc_dointvec_jiffies,
6225 .procname = "gc_timeout",
6226 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6227 .maxlen = sizeof(int),
6229 .proc_handler = proc_dointvec_jiffies,
6232 .procname = "gc_interval",
6233 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
6234 .maxlen = sizeof(int),
6236 .proc_handler = proc_dointvec_jiffies,
6239 .procname = "gc_elasticity",
6240 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6241 .maxlen = sizeof(int),
6243 .proc_handler = proc_dointvec,
6246 .procname = "mtu_expires",
6247 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6248 .maxlen = sizeof(int),
6250 .proc_handler = proc_dointvec_jiffies,
6253 .procname = "min_adv_mss",
6254 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
6255 .maxlen = sizeof(int),
6257 .proc_handler = proc_dointvec,
6260 .procname = "gc_min_interval_ms",
6261 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6262 .maxlen = sizeof(int),
6264 .proc_handler = proc_dointvec_ms_jiffies,
6267 .procname = "skip_notify_on_dev_down",
6268 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6269 .maxlen = sizeof(int),
6271 .proc_handler = proc_dointvec_minmax,
6272 .extra1 = SYSCTL_ZERO,
6273 .extra2 = SYSCTL_ONE,
6278 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6280 struct ctl_table *table;
6282 table = kmemdup(ipv6_route_table_template,
6283 sizeof(ipv6_route_table_template),
6287 table[0].data = &net->ipv6.sysctl.flush_delay;
6288 table[0].extra1 = net;
6289 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6290 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
6291 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6292 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6293 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6294 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6295 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6296 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6297 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6298 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6300 /* Don't export sysctls to unprivileged users */
6301 if (net->user_ns != &init_user_ns)
6302 table[0].procname = NULL;
6309 static int __net_init ip6_route_net_init(struct net *net)
6313 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6314 sizeof(net->ipv6.ip6_dst_ops));
6316 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6317 goto out_ip6_dst_ops;
6319 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6320 if (!net->ipv6.fib6_null_entry)
6321 goto out_ip6_dst_entries;
6322 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6323 sizeof(*net->ipv6.fib6_null_entry));
6325 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6326 sizeof(*net->ipv6.ip6_null_entry),
6328 if (!net->ipv6.ip6_null_entry)
6329 goto out_fib6_null_entry;
6330 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6331 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6332 ip6_template_metrics, true);
6333 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
6335 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6336 net->ipv6.fib6_has_custom_rules = false;
6337 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6338 sizeof(*net->ipv6.ip6_prohibit_entry),
6340 if (!net->ipv6.ip6_prohibit_entry)
6341 goto out_ip6_null_entry;
6342 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6343 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6344 ip6_template_metrics, true);
6345 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
6347 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6348 sizeof(*net->ipv6.ip6_blk_hole_entry),
6350 if (!net->ipv6.ip6_blk_hole_entry)
6351 goto out_ip6_prohibit_entry;
6352 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6353 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6354 ip6_template_metrics, true);
6355 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
6356 #ifdef CONFIG_IPV6_SUBTREES
6357 net->ipv6.fib6_routes_require_src = 0;
6361 net->ipv6.sysctl.flush_delay = 0;
6362 net->ipv6.sysctl.ip6_rt_max_size = 4096;
6363 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6364 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6365 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6366 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6367 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6368 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6369 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6371 net->ipv6.ip6_rt_gc_expire = 30*HZ;
6377 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6378 out_ip6_prohibit_entry:
6379 kfree(net->ipv6.ip6_prohibit_entry);
6381 kfree(net->ipv6.ip6_null_entry);
6383 out_fib6_null_entry:
6384 kfree(net->ipv6.fib6_null_entry);
6385 out_ip6_dst_entries:
6386 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6391 static void __net_exit ip6_route_net_exit(struct net *net)
6393 kfree(net->ipv6.fib6_null_entry);
6394 kfree(net->ipv6.ip6_null_entry);
6395 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6396 kfree(net->ipv6.ip6_prohibit_entry);
6397 kfree(net->ipv6.ip6_blk_hole_entry);
6399 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6402 static int __net_init ip6_route_net_init_late(struct net *net)
6404 #ifdef CONFIG_PROC_FS
6405 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6406 sizeof(struct ipv6_route_iter));
6407 proc_create_net_single("rt6_stats", 0444, net->proc_net,
6408 rt6_stats_seq_show, NULL);
6413 static void __net_exit ip6_route_net_exit_late(struct net *net)
6415 #ifdef CONFIG_PROC_FS
6416 remove_proc_entry("ipv6_route", net->proc_net);
6417 remove_proc_entry("rt6_stats", net->proc_net);
6421 static struct pernet_operations ip6_route_net_ops = {
6422 .init = ip6_route_net_init,
6423 .exit = ip6_route_net_exit,
6426 static int __net_init ipv6_inetpeer_init(struct net *net)
6428 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6432 inet_peer_base_init(bp);
6433 net->ipv6.peers = bp;
6437 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6439 struct inet_peer_base *bp = net->ipv6.peers;
6441 net->ipv6.peers = NULL;
6442 inetpeer_invalidate_tree(bp);
6446 static struct pernet_operations ipv6_inetpeer_ops = {
6447 .init = ipv6_inetpeer_init,
6448 .exit = ipv6_inetpeer_exit,
6451 static struct pernet_operations ip6_route_net_late_ops = {
6452 .init = ip6_route_net_init_late,
6453 .exit = ip6_route_net_exit_late,
6456 static struct notifier_block ip6_route_dev_notifier = {
6457 .notifier_call = ip6_route_dev_notify,
6458 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6461 void __init ip6_route_init_special_entries(void)
6463 /* Registering of the loopback is done before this portion of code,
6464 * the loopback reference in rt6_info will not be taken, do it
6465 * manually for init_net */
6466 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6467 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6468 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6469 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6470 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6471 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6472 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6473 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6477 #if IS_BUILTIN(CONFIG_IPV6)
6478 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6479 DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6481 BTF_ID_LIST(btf_fib6_info_id)
6482 BTF_ID(struct, fib6_info)
6484 static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6485 .seq_ops = &ipv6_route_seq_ops,
6486 .init_seq_private = bpf_iter_init_seq_net,
6487 .fini_seq_private = bpf_iter_fini_seq_net,
6488 .seq_priv_size = sizeof(struct ipv6_route_iter),
6491 static struct bpf_iter_reg ipv6_route_reg_info = {
6492 .target = "ipv6_route",
6493 .ctx_arg_info_size = 1,
6495 { offsetof(struct bpf_iter__ipv6_route, rt),
6496 PTR_TO_BTF_ID_OR_NULL },
6498 .seq_info = &ipv6_route_seq_info,
6501 static int __init bpf_iter_register(void)
6503 ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6504 return bpf_iter_reg_target(&ipv6_route_reg_info);
6507 static void bpf_iter_unregister(void)
6509 bpf_iter_unreg_target(&ipv6_route_reg_info);
6514 int __init ip6_route_init(void)
6520 ip6_dst_ops_template.kmem_cachep =
6521 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6522 SLAB_HWCACHE_ALIGN, NULL);
6523 if (!ip6_dst_ops_template.kmem_cachep)
6526 ret = dst_entries_init(&ip6_dst_blackhole_ops);
6528 goto out_kmem_cache;
6530 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6532 goto out_dst_entries;
6534 ret = register_pernet_subsys(&ip6_route_net_ops);
6536 goto out_register_inetpeer;
6538 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6542 goto out_register_subsys;
6548 ret = fib6_rules_init();
6552 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6554 goto fib6_rules_init;
6556 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6557 inet6_rtm_newroute, NULL, 0);
6559 goto out_register_late_subsys;
6561 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6562 inet6_rtm_delroute, NULL, 0);
6564 goto out_register_late_subsys;
6566 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6567 inet6_rtm_getroute, NULL,
6568 RTNL_FLAG_DOIT_UNLOCKED);
6570 goto out_register_late_subsys;
6572 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6574 goto out_register_late_subsys;
6576 #if IS_BUILTIN(CONFIG_IPV6)
6577 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6578 ret = bpf_iter_register();
6580 goto out_register_late_subsys;
6584 for_each_possible_cpu(cpu) {
6585 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6587 INIT_LIST_HEAD(&ul->head);
6588 spin_lock_init(&ul->lock);
6594 out_register_late_subsys:
6595 rtnl_unregister_all(PF_INET6);
6596 unregister_pernet_subsys(&ip6_route_net_late_ops);
6598 fib6_rules_cleanup();
6603 out_register_subsys:
6604 unregister_pernet_subsys(&ip6_route_net_ops);
6605 out_register_inetpeer:
6606 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6608 dst_entries_destroy(&ip6_dst_blackhole_ops);
6610 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6614 void ip6_route_cleanup(void)
6616 #if IS_BUILTIN(CONFIG_IPV6)
6617 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6618 bpf_iter_unregister();
6621 unregister_netdevice_notifier(&ip6_route_dev_notifier);
6622 unregister_pernet_subsys(&ip6_route_net_late_ops);
6623 fib6_rules_cleanup();
6626 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6627 unregister_pernet_subsys(&ip6_route_net_ops);
6628 dst_entries_destroy(&ip6_dst_blackhole_ops);
6629 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);