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);
94 static void 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;
133 struct list_head quarantine;
136 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
138 void rt6_uncached_list_add(struct rt6_info *rt)
140 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
142 rt->dst.rt_uncached_list = ul;
144 spin_lock_bh(&ul->lock);
145 list_add_tail(&rt->dst.rt_uncached, &ul->head);
146 spin_unlock_bh(&ul->lock);
149 void rt6_uncached_list_del(struct rt6_info *rt)
151 if (!list_empty(&rt->dst.rt_uncached)) {
152 struct uncached_list *ul = rt->dst.rt_uncached_list;
154 spin_lock_bh(&ul->lock);
155 list_del_init(&rt->dst.rt_uncached);
156 spin_unlock_bh(&ul->lock);
160 static void rt6_uncached_list_flush_dev(struct net_device *dev)
164 for_each_possible_cpu(cpu) {
165 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
166 struct rt6_info *rt, *safe;
168 if (list_empty(&ul->head))
171 spin_lock_bh(&ul->lock);
172 list_for_each_entry_safe(rt, safe, &ul->head, dst.rt_uncached) {
173 struct inet6_dev *rt_idev = rt->rt6i_idev;
174 struct net_device *rt_dev = rt->dst.dev;
175 bool handled = false;
177 if (rt_idev->dev == dev) {
178 rt->rt6i_idev = in6_dev_get(blackhole_netdev);
179 in6_dev_put(rt_idev);
184 rt->dst.dev = blackhole_netdev;
185 netdev_ref_replace(rt_dev, blackhole_netdev,
186 &rt->dst.dev_tracker,
191 list_move(&rt->dst.rt_uncached,
194 spin_unlock_bh(&ul->lock);
198 static inline const void *choose_neigh_daddr(const struct in6_addr *p,
202 if (!ipv6_addr_any(p))
203 return (const void *) p;
205 return &ipv6_hdr(skb)->daddr;
209 struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
210 struct net_device *dev,
216 daddr = choose_neigh_daddr(gw, skb, daddr);
217 n = __ipv6_neigh_lookup(dev, daddr);
221 n = neigh_create(&nd_tbl, daddr, dev);
222 return IS_ERR(n) ? NULL : n;
225 static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
229 const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
231 return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
232 dst->dev, skb, daddr);
235 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
237 struct net_device *dev = dst->dev;
238 struct rt6_info *rt = (struct rt6_info *)dst;
240 daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
243 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
245 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
247 __ipv6_confirm_neigh(dev, daddr);
250 static struct dst_ops ip6_dst_ops_template = {
254 .check = ip6_dst_check,
255 .default_advmss = ip6_default_advmss,
257 .cow_metrics = dst_cow_metrics_generic,
258 .destroy = ip6_dst_destroy,
259 .ifdown = ip6_dst_ifdown,
260 .negative_advice = ip6_negative_advice,
261 .link_failure = ip6_link_failure,
262 .update_pmtu = ip6_rt_update_pmtu,
263 .redirect = rt6_do_redirect,
264 .local_out = __ip6_local_out,
265 .neigh_lookup = ip6_dst_neigh_lookup,
266 .confirm_neigh = ip6_confirm_neigh,
269 static struct dst_ops ip6_dst_blackhole_ops = {
271 .default_advmss = ip6_default_advmss,
272 .neigh_lookup = ip6_dst_neigh_lookup,
273 .check = ip6_dst_check,
274 .destroy = ip6_dst_destroy,
275 .cow_metrics = dst_cow_metrics_generic,
276 .update_pmtu = dst_blackhole_update_pmtu,
277 .redirect = dst_blackhole_redirect,
278 .mtu = dst_blackhole_mtu,
281 static const u32 ip6_template_metrics[RTAX_MAX] = {
282 [RTAX_HOPLIMIT - 1] = 0,
285 static const struct fib6_info fib6_null_entry_template = {
286 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
287 .fib6_protocol = RTPROT_KERNEL,
288 .fib6_metric = ~(u32)0,
289 .fib6_ref = REFCOUNT_INIT(1),
290 .fib6_type = RTN_UNREACHABLE,
291 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
294 static const struct rt6_info ip6_null_entry_template = {
296 .__rcuref = RCUREF_INIT(1),
298 .obsolete = DST_OBSOLETE_FORCE_CHK,
299 .error = -ENETUNREACH,
300 .input = ip6_pkt_discard,
301 .output = ip6_pkt_discard_out,
303 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
306 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
308 static const struct rt6_info ip6_prohibit_entry_template = {
310 .__rcuref = RCUREF_INIT(1),
312 .obsolete = DST_OBSOLETE_FORCE_CHK,
314 .input = ip6_pkt_prohibit,
315 .output = ip6_pkt_prohibit_out,
317 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
320 static const struct rt6_info ip6_blk_hole_entry_template = {
322 .__rcuref = RCUREF_INIT(1),
324 .obsolete = DST_OBSOLETE_FORCE_CHK,
326 .input = dst_discard,
327 .output = dst_discard_out,
329 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
334 static void rt6_info_init(struct rt6_info *rt)
336 memset_after(rt, 0, dst);
339 /* allocate dst with ip6_dst_ops */
340 struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
343 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
344 DST_OBSOLETE_FORCE_CHK, flags);
348 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
353 EXPORT_SYMBOL(ip6_dst_alloc);
355 static void ip6_dst_destroy(struct dst_entry *dst)
357 struct rt6_info *rt = (struct rt6_info *)dst;
358 struct fib6_info *from;
359 struct inet6_dev *idev;
361 ip_dst_metrics_put(dst);
362 rt6_uncached_list_del(rt);
364 idev = rt->rt6i_idev;
366 rt->rt6i_idev = NULL;
370 from = xchg((__force struct fib6_info **)&rt->from, NULL);
371 fib6_info_release(from);
374 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
376 struct rt6_info *rt = (struct rt6_info *)dst;
377 struct inet6_dev *idev = rt->rt6i_idev;
379 if (idev && idev->dev != blackhole_netdev) {
380 struct inet6_dev *blackhole_idev = in6_dev_get(blackhole_netdev);
382 if (blackhole_idev) {
383 rt->rt6i_idev = blackhole_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)
427 IP6CB(skb)->flags |= IP6SKB_MULTIPATH;
429 /* We might have already computed the hash for ICMPv6 errors. In such
430 * case it will always be non-zero. Otherwise now is the time to do it.
433 (!match->nh || nexthop_is_multipath(match->nh)))
434 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
436 if (unlikely(match->nh)) {
437 nexthop_path_fib6_result(res, fl6->mp_hash);
441 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
444 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
446 const struct fib6_nh *nh = sibling->fib6_nh;
449 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
450 if (fl6->mp_hash > nh_upper_bound)
452 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
460 res->nh = match->fib6_nh;
464 * Route lookup. rcu_read_lock() should be held.
467 static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
468 const struct in6_addr *saddr, int oif, int flags)
470 const struct net_device *dev;
472 if (nh->fib_nh_flags & RTNH_F_DEAD)
475 dev = nh->fib_nh_dev;
477 if (dev->ifindex == oif)
480 if (ipv6_chk_addr(net, saddr, dev,
481 flags & RT6_LOOKUP_F_IFACE))
488 struct fib6_nh_dm_arg {
490 const struct in6_addr *saddr;
496 static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
498 struct fib6_nh_dm_arg *arg = _arg;
501 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
505 /* returns fib6_nh from nexthop or NULL */
506 static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
507 struct fib6_result *res,
508 const struct in6_addr *saddr,
511 struct fib6_nh_dm_arg arg = {
518 if (nexthop_is_blackhole(nh))
521 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
527 static void rt6_device_match(struct net *net, struct fib6_result *res,
528 const struct in6_addr *saddr, int oif, int flags)
530 struct fib6_info *f6i = res->f6i;
531 struct fib6_info *spf6i;
534 if (!oif && ipv6_addr_any(saddr)) {
535 if (unlikely(f6i->nh)) {
536 nh = nexthop_fib6_nh(f6i->nh);
537 if (nexthop_is_blackhole(f6i->nh))
542 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
546 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
547 bool matched = false;
549 if (unlikely(spf6i->nh)) {
550 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
556 if (__rt6_device_match(net, nh, saddr, oif, flags))
565 if (oif && flags & RT6_LOOKUP_F_IFACE) {
566 res->f6i = net->ipv6.fib6_null_entry;
567 nh = res->f6i->fib6_nh;
571 if (unlikely(f6i->nh)) {
572 nh = nexthop_fib6_nh(f6i->nh);
573 if (nexthop_is_blackhole(f6i->nh))
579 if (nh->fib_nh_flags & RTNH_F_DEAD) {
580 res->f6i = net->ipv6.fib6_null_entry;
581 nh = res->f6i->fib6_nh;
585 res->fib6_type = res->f6i->fib6_type;
586 res->fib6_flags = res->f6i->fib6_flags;
590 res->fib6_flags |= RTF_REJECT;
591 res->fib6_type = RTN_BLACKHOLE;
595 #ifdef CONFIG_IPV6_ROUTER_PREF
596 struct __rt6_probe_work {
597 struct work_struct work;
598 struct in6_addr target;
599 struct net_device *dev;
600 netdevice_tracker dev_tracker;
603 static void rt6_probe_deferred(struct work_struct *w)
605 struct in6_addr mcaddr;
606 struct __rt6_probe_work *work =
607 container_of(w, struct __rt6_probe_work, work);
609 addrconf_addr_solict_mult(&work->target, &mcaddr);
610 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
611 netdev_put(work->dev, &work->dev_tracker);
615 static void rt6_probe(struct fib6_nh *fib6_nh)
617 struct __rt6_probe_work *work = NULL;
618 const struct in6_addr *nh_gw;
619 unsigned long last_probe;
620 struct neighbour *neigh;
621 struct net_device *dev;
622 struct inet6_dev *idev;
625 * Okay, this does not seem to be appropriate
626 * for now, however, we need to check if it
627 * is really so; aka Router Reachability Probing.
629 * Router Reachability Probe MUST be rate-limited
630 * to no more than one per minute.
632 if (!fib6_nh->fib_nh_gw_family)
635 nh_gw = &fib6_nh->fib_nh_gw6;
636 dev = fib6_nh->fib_nh_dev;
638 last_probe = READ_ONCE(fib6_nh->last_probe);
639 idev = __in6_dev_get(dev);
640 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
642 if (READ_ONCE(neigh->nud_state) & NUD_VALID)
645 write_lock_bh(&neigh->lock);
646 if (!(neigh->nud_state & NUD_VALID) &&
649 READ_ONCE(idev->cnf.rtr_probe_interval))) {
650 work = kmalloc(sizeof(*work), GFP_ATOMIC);
652 __neigh_set_probe_once(neigh);
654 write_unlock_bh(&neigh->lock);
655 } else if (time_after(jiffies, last_probe +
656 READ_ONCE(idev->cnf.rtr_probe_interval))) {
657 work = kmalloc(sizeof(*work), GFP_ATOMIC);
660 if (!work || cmpxchg(&fib6_nh->last_probe,
661 last_probe, jiffies) != last_probe) {
664 INIT_WORK(&work->work, rt6_probe_deferred);
665 work->target = *nh_gw;
666 netdev_hold(dev, &work->dev_tracker, GFP_ATOMIC);
668 schedule_work(&work->work);
675 static inline void rt6_probe(struct fib6_nh *fib6_nh)
681 * Default Router Selection (RFC 2461 6.3.6)
683 static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
685 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
686 struct neighbour *neigh;
689 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
690 &fib6_nh->fib_nh_gw6);
692 u8 nud_state = READ_ONCE(neigh->nud_state);
694 if (nud_state & NUD_VALID)
695 ret = RT6_NUD_SUCCEED;
696 #ifdef CONFIG_IPV6_ROUTER_PREF
697 else if (!(nud_state & NUD_FAILED))
698 ret = RT6_NUD_SUCCEED;
700 ret = RT6_NUD_FAIL_PROBE;
703 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
704 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
711 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
716 if (!oif || nh->fib_nh_dev->ifindex == oif)
719 if (!m && (strict & RT6_LOOKUP_F_IFACE))
720 return RT6_NUD_FAIL_HARD;
721 #ifdef CONFIG_IPV6_ROUTER_PREF
722 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
724 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
725 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
726 int n = rt6_check_neigh(nh);
733 static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
734 int oif, int strict, int *mpri, bool *do_rr)
736 bool match_do_rr = false;
740 if (nh->fib_nh_flags & RTNH_F_DEAD)
743 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
744 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
745 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
748 m = rt6_score_route(nh, fib6_flags, oif, strict);
749 if (m == RT6_NUD_FAIL_DO_RR) {
751 m = 0; /* lowest valid score */
752 } else if (m == RT6_NUD_FAIL_HARD) {
756 if (strict & RT6_LOOKUP_F_REACHABLE)
759 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
761 *do_rr = match_do_rr;
769 struct fib6_nh_frl_arg {
778 static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
780 struct fib6_nh_frl_arg *arg = _arg;
783 return find_match(nh, arg->flags, arg->oif, arg->strict,
784 arg->mpri, arg->do_rr);
787 static void __find_rr_leaf(struct fib6_info *f6i_start,
788 struct fib6_info *nomatch, u32 metric,
789 struct fib6_result *res, struct fib6_info **cont,
790 int oif, int strict, bool *do_rr, int *mpri)
792 struct fib6_info *f6i;
794 for (f6i = f6i_start;
795 f6i && f6i != nomatch;
796 f6i = rcu_dereference(f6i->fib6_next)) {
797 bool matched = false;
800 if (cont && f6i->fib6_metric != metric) {
805 if (fib6_check_expired(f6i))
808 if (unlikely(f6i->nh)) {
809 struct fib6_nh_frl_arg arg = {
810 .flags = f6i->fib6_flags,
817 if (nexthop_is_blackhole(f6i->nh)) {
818 res->fib6_flags = RTF_REJECT;
819 res->fib6_type = RTN_BLACKHOLE;
821 res->nh = nexthop_fib6_nh(f6i->nh);
824 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
831 if (find_match(nh, f6i->fib6_flags, oif, strict,
838 res->fib6_flags = f6i->fib6_flags;
839 res->fib6_type = f6i->fib6_type;
844 static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
845 struct fib6_info *rr_head, int oif, int strict,
846 bool *do_rr, struct fib6_result *res)
848 u32 metric = rr_head->fib6_metric;
849 struct fib6_info *cont = NULL;
852 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
853 oif, strict, do_rr, &mpri);
855 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
856 oif, strict, do_rr, &mpri);
858 if (res->f6i || !cont)
861 __find_rr_leaf(cont, NULL, metric, res, NULL,
862 oif, strict, do_rr, &mpri);
865 static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
866 struct fib6_result *res, int strict)
868 struct fib6_info *leaf = rcu_dereference(fn->leaf);
869 struct fib6_info *rt0;
873 /* make sure this function or its helpers sets f6i */
876 if (!leaf || leaf == net->ipv6.fib6_null_entry)
879 rt0 = rcu_dereference(fn->rr_ptr);
883 /* Double check to make sure fn is not an intermediate node
884 * and fn->leaf does not points to its child's leaf
885 * (This might happen if all routes under fn are deleted from
886 * the tree and fib6_repair_tree() is called on the node.)
888 key_plen = rt0->fib6_dst.plen;
889 #ifdef CONFIG_IPV6_SUBTREES
890 if (rt0->fib6_src.plen)
891 key_plen = rt0->fib6_src.plen;
893 if (fn->fn_bit != key_plen)
896 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
898 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
900 /* no entries matched; do round-robin */
901 if (!next || next->fib6_metric != rt0->fib6_metric)
905 spin_lock_bh(&leaf->fib6_table->tb6_lock);
906 /* make sure next is not being deleted from the tree */
908 rcu_assign_pointer(fn->rr_ptr, next);
909 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
915 res->f6i = net->ipv6.fib6_null_entry;
916 res->nh = res->f6i->fib6_nh;
917 res->fib6_flags = res->f6i->fib6_flags;
918 res->fib6_type = res->f6i->fib6_type;
922 static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
924 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
925 res->nh->fib_nh_gw_family;
928 #ifdef CONFIG_IPV6_ROUTE_INFO
929 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
930 const struct in6_addr *gwaddr)
932 struct net *net = dev_net(dev);
933 struct route_info *rinfo = (struct route_info *) opt;
934 struct in6_addr prefix_buf, *prefix;
935 struct fib6_table *table;
937 unsigned long lifetime;
938 struct fib6_info *rt;
940 if (len < sizeof(struct route_info)) {
944 /* Sanity check for prefix_len and length */
945 if (rinfo->length > 3) {
947 } else if (rinfo->prefix_len > 128) {
949 } else if (rinfo->prefix_len > 64) {
950 if (rinfo->length < 2) {
953 } else if (rinfo->prefix_len > 0) {
954 if (rinfo->length < 1) {
959 pref = rinfo->route_pref;
960 if (pref == ICMPV6_ROUTER_PREF_INVALID)
963 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
965 if (rinfo->length == 3)
966 prefix = (struct in6_addr *)rinfo->prefix;
968 /* this function is safe */
969 ipv6_addr_prefix(&prefix_buf,
970 (struct in6_addr *)rinfo->prefix,
972 prefix = &prefix_buf;
975 if (rinfo->prefix_len == 0)
976 rt = rt6_get_dflt_router(net, gwaddr, dev);
978 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
981 if (rt && !lifetime) {
982 ip6_del_rt(net, rt, false);
987 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
990 rt->fib6_flags = RTF_ROUTEINFO |
991 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
994 table = rt->fib6_table;
995 spin_lock_bh(&table->tb6_lock);
997 if (!addrconf_finite_timeout(lifetime)) {
998 fib6_clean_expires(rt);
999 fib6_remove_gc_list(rt);
1001 fib6_set_expires(rt, jiffies + HZ * lifetime);
1002 fib6_add_gc_list(rt);
1005 spin_unlock_bh(&table->tb6_lock);
1007 fib6_info_release(rt);
1014 * Misc support functions
1017 /* called with rcu_lock held */
1018 static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1020 struct net_device *dev = res->nh->fib_nh_dev;
1022 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1023 /* for copies of local routes, dst->dev needs to be the
1024 * device if it is a master device, the master device if
1025 * device is enslaved, and the loopback as the default
1027 if (netif_is_l3_slave(dev) &&
1028 !rt6_need_strict(&res->f6i->fib6_dst.addr))
1029 dev = l3mdev_master_dev_rcu(dev);
1030 else if (!netif_is_l3_master(dev))
1031 dev = dev_net(dev)->loopback_dev;
1032 /* last case is netif_is_l3_master(dev) is true in which
1033 * case we want dev returned to be dev
1040 static const int fib6_prop[RTN_MAX + 1] = {
1044 [RTN_BROADCAST] = 0,
1046 [RTN_MULTICAST] = 0,
1047 [RTN_BLACKHOLE] = -EINVAL,
1048 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1049 [RTN_PROHIBIT] = -EACCES,
1050 [RTN_THROW] = -EAGAIN,
1051 [RTN_NAT] = -EINVAL,
1052 [RTN_XRESOLVE] = -EINVAL,
1055 static int ip6_rt_type_to_error(u8 fib6_type)
1057 return fib6_prop[fib6_type];
1060 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1062 unsigned short flags = 0;
1064 if (rt->dst_nocount)
1065 flags |= DST_NOCOUNT;
1066 if (rt->dst_nopolicy)
1067 flags |= DST_NOPOLICY;
1072 static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1074 rt->dst.error = ip6_rt_type_to_error(fib6_type);
1076 switch (fib6_type) {
1078 rt->dst.output = dst_discard_out;
1079 rt->dst.input = dst_discard;
1082 rt->dst.output = ip6_pkt_prohibit_out;
1083 rt->dst.input = ip6_pkt_prohibit;
1086 case RTN_UNREACHABLE:
1088 rt->dst.output = ip6_pkt_discard_out;
1089 rt->dst.input = ip6_pkt_discard;
1094 static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1096 struct fib6_info *f6i = res->f6i;
1098 if (res->fib6_flags & RTF_REJECT) {
1099 ip6_rt_init_dst_reject(rt, res->fib6_type);
1104 rt->dst.output = ip6_output;
1106 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1107 rt->dst.input = ip6_input;
1108 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1109 rt->dst.input = ip6_mc_input;
1111 rt->dst.input = ip6_forward;
1114 if (res->nh->fib_nh_lws) {
1115 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1116 lwtunnel_set_redirect(&rt->dst);
1119 rt->dst.lastuse = jiffies;
1122 /* Caller must already hold reference to @from */
1123 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1125 rt->rt6i_flags &= ~RTF_EXPIRES;
1126 rcu_assign_pointer(rt->from, from);
1127 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1130 /* Caller must already hold reference to f6i in result */
1131 static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1133 const struct fib6_nh *nh = res->nh;
1134 const struct net_device *dev = nh->fib_nh_dev;
1135 struct fib6_info *f6i = res->f6i;
1137 ip6_rt_init_dst(rt, res);
1139 rt->rt6i_dst = f6i->fib6_dst;
1140 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1141 rt->rt6i_flags = res->fib6_flags;
1142 if (nh->fib_nh_gw_family) {
1143 rt->rt6i_gateway = nh->fib_nh_gw6;
1144 rt->rt6i_flags |= RTF_GATEWAY;
1146 rt6_set_from(rt, f6i);
1147 #ifdef CONFIG_IPV6_SUBTREES
1148 rt->rt6i_src = f6i->fib6_src;
1152 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1153 struct in6_addr *saddr)
1155 struct fib6_node *pn, *sn;
1157 if (fn->fn_flags & RTN_TL_ROOT)
1159 pn = rcu_dereference(fn->parent);
1160 sn = FIB6_SUBTREE(pn);
1162 fn = fib6_node_lookup(sn, NULL, saddr);
1165 if (fn->fn_flags & RTN_RTINFO)
1170 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1172 struct rt6_info *rt = *prt;
1174 if (dst_hold_safe(&rt->dst))
1177 rt = net->ipv6.ip6_null_entry;
1186 /* called with rcu_lock held */
1187 static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1189 struct net_device *dev = res->nh->fib_nh_dev;
1190 struct fib6_info *f6i = res->f6i;
1191 unsigned short flags;
1192 struct rt6_info *nrt;
1194 if (!fib6_info_hold_safe(f6i))
1197 flags = fib6_info_dst_flags(f6i);
1198 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1200 fib6_info_release(f6i);
1204 ip6_rt_copy_init(nrt, res);
1208 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1209 dst_hold(&nrt->dst);
1213 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
1214 struct fib6_table *table,
1216 const struct sk_buff *skb,
1219 struct fib6_result res = {};
1220 struct fib6_node *fn;
1221 struct rt6_info *rt;
1224 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1226 res.f6i = rcu_dereference(fn->leaf);
1228 res.f6i = net->ipv6.fib6_null_entry;
1230 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1233 if (res.f6i == net->ipv6.fib6_null_entry) {
1234 fn = fib6_backtrack(fn, &fl6->saddr);
1238 rt = net->ipv6.ip6_null_entry;
1241 } else if (res.fib6_flags & RTF_REJECT) {
1245 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1246 fl6->flowi6_oif != 0, skb, flags);
1248 /* Search through exception table */
1249 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1251 if (ip6_hold_safe(net, &rt))
1252 dst_use_noref(&rt->dst, jiffies);
1255 rt = ip6_create_rt_rcu(&res);
1259 trace_fib6_table_lookup(net, &res, table, fl6);
1266 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1267 const struct sk_buff *skb, int flags)
1269 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1271 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1273 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1274 const struct in6_addr *saddr, int oif,
1275 const struct sk_buff *skb, int strict)
1277 struct flowi6 fl6 = {
1281 struct dst_entry *dst;
1282 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1285 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1286 flags |= RT6_LOOKUP_F_HAS_SADDR;
1289 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1290 if (dst->error == 0)
1291 return (struct rt6_info *) dst;
1297 EXPORT_SYMBOL(rt6_lookup);
1299 /* ip6_ins_rt is called with FREE table->tb6_lock.
1300 * It takes new route entry, the addition fails by any reason the
1301 * route is released.
1302 * Caller must hold dst before calling it.
1305 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1306 struct netlink_ext_ack *extack)
1309 struct fib6_table *table;
1311 table = rt->fib6_table;
1312 spin_lock_bh(&table->tb6_lock);
1313 err = fib6_add(&table->tb6_root, rt, info, extack);
1314 spin_unlock_bh(&table->tb6_lock);
1319 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1321 struct nl_info info = { .nl_net = net, };
1323 return __ip6_ins_rt(rt, &info, NULL);
1326 static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1327 const struct in6_addr *daddr,
1328 const struct in6_addr *saddr)
1330 struct fib6_info *f6i = res->f6i;
1331 struct net_device *dev;
1332 struct rt6_info *rt;
1338 if (!fib6_info_hold_safe(f6i))
1341 dev = ip6_rt_get_dev_rcu(res);
1342 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1344 fib6_info_release(f6i);
1348 ip6_rt_copy_init(rt, res);
1349 rt->rt6i_flags |= RTF_CACHE;
1350 rt->rt6i_dst.addr = *daddr;
1351 rt->rt6i_dst.plen = 128;
1353 if (!rt6_is_gw_or_nonexthop(res)) {
1354 if (f6i->fib6_dst.plen != 128 &&
1355 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1356 rt->rt6i_flags |= RTF_ANYCAST;
1357 #ifdef CONFIG_IPV6_SUBTREES
1358 if (rt->rt6i_src.plen && saddr) {
1359 rt->rt6i_src.addr = *saddr;
1360 rt->rt6i_src.plen = 128;
1368 static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1370 struct fib6_info *f6i = res->f6i;
1371 unsigned short flags = fib6_info_dst_flags(f6i);
1372 struct net_device *dev;
1373 struct rt6_info *pcpu_rt;
1375 if (!fib6_info_hold_safe(f6i))
1379 dev = ip6_rt_get_dev_rcu(res);
1380 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
1383 fib6_info_release(f6i);
1386 ip6_rt_copy_init(pcpu_rt, res);
1387 pcpu_rt->rt6i_flags |= RTF_PCPU;
1390 pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1395 static bool rt6_is_valid(const struct rt6_info *rt6)
1397 return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1400 /* It should be called with rcu_read_lock() acquired */
1401 static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1403 struct rt6_info *pcpu_rt;
1405 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1407 if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1408 struct rt6_info *prev, **p;
1410 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1411 prev = xchg(p, NULL);
1413 dst_dev_put(&prev->dst);
1414 dst_release(&prev->dst);
1423 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1424 const struct fib6_result *res)
1426 struct rt6_info *pcpu_rt, *prev, **p;
1428 pcpu_rt = ip6_rt_pcpu_alloc(res);
1432 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1433 prev = cmpxchg(p, NULL, pcpu_rt);
1436 if (res->f6i->fib6_destroying) {
1437 struct fib6_info *from;
1439 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1440 fib6_info_release(from);
1446 /* exception hash table implementation
1448 static DEFINE_SPINLOCK(rt6_exception_lock);
1450 /* Remove rt6_ex from hash table and free the memory
1451 * Caller must hold rt6_exception_lock
1453 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1454 struct rt6_exception *rt6_ex)
1456 struct fib6_info *from;
1459 if (!bucket || !rt6_ex)
1462 net = dev_net(rt6_ex->rt6i->dst.dev);
1463 net->ipv6.rt6_stats->fib_rt_cache--;
1465 /* purge completely the exception to allow releasing the held resources:
1466 * some [sk] cache may keep the dst around for unlimited time
1468 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
1469 fib6_info_release(from);
1470 dst_dev_put(&rt6_ex->rt6i->dst);
1472 hlist_del_rcu(&rt6_ex->hlist);
1473 dst_release(&rt6_ex->rt6i->dst);
1474 kfree_rcu(rt6_ex, rcu);
1475 WARN_ON_ONCE(!bucket->depth);
1479 /* Remove oldest rt6_ex in bucket and free the memory
1480 * Caller must hold rt6_exception_lock
1482 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1484 struct rt6_exception *rt6_ex, *oldest = NULL;
1489 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1490 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1493 rt6_remove_exception(bucket, oldest);
1496 static u32 rt6_exception_hash(const struct in6_addr *dst,
1497 const struct in6_addr *src)
1499 static siphash_aligned_key_t rt6_exception_key;
1501 struct in6_addr dst;
1502 struct in6_addr src;
1503 } __aligned(SIPHASH_ALIGNMENT) combined = {
1508 net_get_random_once(&rt6_exception_key, sizeof(rt6_exception_key));
1510 #ifdef CONFIG_IPV6_SUBTREES
1512 combined.src = *src;
1514 val = siphash(&combined, sizeof(combined), &rt6_exception_key);
1516 return hash_64(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1519 /* Helper function to find the cached rt in the hash table
1520 * and update bucket pointer to point to the bucket for this
1521 * (daddr, saddr) pair
1522 * Caller must hold rt6_exception_lock
1524 static struct rt6_exception *
1525 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1526 const struct in6_addr *daddr,
1527 const struct in6_addr *saddr)
1529 struct rt6_exception *rt6_ex;
1532 if (!(*bucket) || !daddr)
1535 hval = rt6_exception_hash(daddr, saddr);
1538 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1539 struct rt6_info *rt6 = rt6_ex->rt6i;
1540 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1542 #ifdef CONFIG_IPV6_SUBTREES
1543 if (matched && saddr)
1544 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1552 /* Helper function to find the cached rt in the hash table
1553 * and update bucket pointer to point to the bucket for this
1554 * (daddr, saddr) pair
1555 * Caller must hold rcu_read_lock()
1557 static struct rt6_exception *
1558 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1559 const struct in6_addr *daddr,
1560 const struct in6_addr *saddr)
1562 struct rt6_exception *rt6_ex;
1565 WARN_ON_ONCE(!rcu_read_lock_held());
1567 if (!(*bucket) || !daddr)
1570 hval = rt6_exception_hash(daddr, saddr);
1573 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1574 struct rt6_info *rt6 = rt6_ex->rt6i;
1575 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1577 #ifdef CONFIG_IPV6_SUBTREES
1578 if (matched && saddr)
1579 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1587 static unsigned int fib6_mtu(const struct fib6_result *res)
1589 const struct fib6_nh *nh = res->nh;
1592 if (res->f6i->fib6_pmtu) {
1593 mtu = res->f6i->fib6_pmtu;
1595 struct net_device *dev = nh->fib_nh_dev;
1596 struct inet6_dev *idev;
1599 idev = __in6_dev_get(dev);
1600 mtu = READ_ONCE(idev->cnf.mtu6);
1604 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1606 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1609 #define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1611 /* used when the flushed bit is not relevant, only access to the bucket
1612 * (ie., all bucket users except rt6_insert_exception);
1614 * called under rcu lock; sometimes called with rt6_exception_lock held
1617 struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1620 struct rt6_exception_bucket *bucket;
1623 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1624 lockdep_is_held(lock));
1626 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1628 /* remove bucket flushed bit if set */
1630 unsigned long p = (unsigned long)bucket;
1632 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1633 bucket = (struct rt6_exception_bucket *)p;
1639 static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1641 unsigned long p = (unsigned long)bucket;
1643 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1646 /* called with rt6_exception_lock held */
1647 static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1650 struct rt6_exception_bucket *bucket;
1653 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1654 lockdep_is_held(lock));
1656 p = (unsigned long)bucket;
1657 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1658 bucket = (struct rt6_exception_bucket *)p;
1659 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1662 static int rt6_insert_exception(struct rt6_info *nrt,
1663 const struct fib6_result *res)
1665 struct net *net = dev_net(nrt->dst.dev);
1666 struct rt6_exception_bucket *bucket;
1667 struct fib6_info *f6i = res->f6i;
1668 struct in6_addr *src_key = NULL;
1669 struct rt6_exception *rt6_ex;
1670 struct fib6_nh *nh = res->nh;
1674 spin_lock_bh(&rt6_exception_lock);
1676 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1677 lockdep_is_held(&rt6_exception_lock));
1679 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1685 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1686 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1691 #ifdef CONFIG_IPV6_SUBTREES
1692 /* fib6_src.plen != 0 indicates f6i is in subtree
1693 * and exception table is indexed by a hash of
1694 * both fib6_dst and fib6_src.
1695 * Otherwise, the exception table is indexed by
1696 * a hash of only fib6_dst.
1698 if (f6i->fib6_src.plen)
1699 src_key = &nrt->rt6i_src.addr;
1701 /* rt6_mtu_change() might lower mtu on f6i.
1702 * Only insert this exception route if its mtu
1703 * is less than f6i's mtu value.
1705 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1710 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1713 rt6_remove_exception(bucket, rt6_ex);
1715 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1721 rt6_ex->stamp = jiffies;
1722 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1724 net->ipv6.rt6_stats->fib_rt_cache++;
1726 /* Randomize max depth to avoid some side channels attacks. */
1727 max_depth = FIB6_MAX_DEPTH + get_random_u32_below(FIB6_MAX_DEPTH);
1728 while (bucket->depth > max_depth)
1729 rt6_exception_remove_oldest(bucket);
1732 spin_unlock_bh(&rt6_exception_lock);
1734 /* Update fn->fn_sernum to invalidate all cached dst */
1736 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1737 fib6_update_sernum(net, f6i);
1738 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1739 fib6_force_start_gc(net);
1745 static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1747 struct rt6_exception_bucket *bucket;
1748 struct rt6_exception *rt6_ex;
1749 struct hlist_node *tmp;
1752 spin_lock_bh(&rt6_exception_lock);
1754 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1758 /* Prevent rt6_insert_exception() to recreate the bucket list */
1760 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1762 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1763 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1765 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1766 rt6_remove_exception(bucket, rt6_ex);
1768 WARN_ON_ONCE(!from && bucket->depth);
1772 spin_unlock_bh(&rt6_exception_lock);
1775 static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1777 struct fib6_info *f6i = arg;
1779 fib6_nh_flush_exceptions(nh, f6i);
1784 void rt6_flush_exceptions(struct fib6_info *f6i)
1787 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1790 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1793 /* Find cached rt in the hash table inside passed in rt
1794 * Caller has to hold rcu_read_lock()
1796 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1797 const struct in6_addr *daddr,
1798 const struct in6_addr *saddr)
1800 const struct in6_addr *src_key = NULL;
1801 struct rt6_exception_bucket *bucket;
1802 struct rt6_exception *rt6_ex;
1803 struct rt6_info *ret = NULL;
1805 #ifdef CONFIG_IPV6_SUBTREES
1806 /* fib6i_src.plen != 0 indicates f6i is in subtree
1807 * and exception table is indexed by a hash of
1808 * both fib6_dst and fib6_src.
1809 * However, the src addr used to create the hash
1810 * might not be exactly the passed in saddr which
1811 * is a /128 addr from the flow.
1812 * So we need to use f6i->fib6_src to redo lookup
1813 * if the passed in saddr does not find anything.
1814 * (See the logic in ip6_rt_cache_alloc() on how
1815 * rt->rt6i_src is updated.)
1817 if (res->f6i->fib6_src.plen)
1821 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1822 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1824 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1827 #ifdef CONFIG_IPV6_SUBTREES
1828 /* Use fib6_src as src_key and redo lookup */
1829 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1830 src_key = &res->f6i->fib6_src.addr;
1838 /* Remove the passed in cached rt from the hash table that contains it */
1839 static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1840 const struct rt6_info *rt)
1842 const struct in6_addr *src_key = NULL;
1843 struct rt6_exception_bucket *bucket;
1844 struct rt6_exception *rt6_ex;
1847 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1850 spin_lock_bh(&rt6_exception_lock);
1851 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1853 #ifdef CONFIG_IPV6_SUBTREES
1854 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1855 * and exception table is indexed by a hash of
1856 * both rt6i_dst and rt6i_src.
1857 * Otherwise, the exception table is indexed by
1858 * a hash of only rt6i_dst.
1861 src_key = &rt->rt6i_src.addr;
1863 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1867 rt6_remove_exception(bucket, rt6_ex);
1873 spin_unlock_bh(&rt6_exception_lock);
1877 struct fib6_nh_excptn_arg {
1878 struct rt6_info *rt;
1882 static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1884 struct fib6_nh_excptn_arg *arg = _arg;
1887 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1894 static int rt6_remove_exception_rt(struct rt6_info *rt)
1896 struct fib6_info *from;
1898 from = rcu_dereference(rt->from);
1899 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1903 struct fib6_nh_excptn_arg arg = {
1905 .plen = from->fib6_src.plen
1909 /* rc = 1 means an entry was found */
1910 rc = nexthop_for_each_fib6_nh(from->nh,
1911 rt6_nh_remove_exception_rt,
1913 return rc ? 0 : -ENOENT;
1916 return fib6_nh_remove_exception(from->fib6_nh,
1917 from->fib6_src.plen, rt);
1920 /* Find rt6_ex which contains the passed in rt cache and
1923 static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1924 const struct rt6_info *rt)
1926 const struct in6_addr *src_key = NULL;
1927 struct rt6_exception_bucket *bucket;
1928 struct rt6_exception *rt6_ex;
1930 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1931 #ifdef CONFIG_IPV6_SUBTREES
1932 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1933 * and exception table is indexed by a hash of
1934 * both rt6i_dst and rt6i_src.
1935 * Otherwise, the exception table is indexed by
1936 * a hash of only rt6i_dst.
1939 src_key = &rt->rt6i_src.addr;
1941 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1943 rt6_ex->stamp = jiffies;
1946 struct fib6_nh_match_arg {
1947 const struct net_device *dev;
1948 const struct in6_addr *gw;
1949 struct fib6_nh *match;
1952 /* determine if fib6_nh has given device and gateway */
1953 static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1955 struct fib6_nh_match_arg *arg = _arg;
1957 if (arg->dev != nh->fib_nh_dev ||
1958 (arg->gw && !nh->fib_nh_gw_family) ||
1959 (!arg->gw && nh->fib_nh_gw_family) ||
1960 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1965 /* found a match, break the loop */
1969 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1971 struct fib6_info *from;
1972 struct fib6_nh *fib6_nh;
1976 from = rcu_dereference(rt->from);
1977 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1981 struct fib6_nh_match_arg arg = {
1983 .gw = &rt->rt6i_gateway,
1986 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1990 fib6_nh = arg.match;
1992 fib6_nh = from->fib6_nh;
1994 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1999 static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
2000 struct rt6_info *rt, int mtu)
2002 /* If the new MTU is lower than the route PMTU, this new MTU will be the
2003 * lowest MTU in the path: always allow updating the route PMTU to
2004 * reflect PMTU decreases.
2006 * If the new MTU is higher, and the route PMTU is equal to the local
2007 * MTU, this means the old MTU is the lowest in the path, so allow
2008 * updating it: if other nodes now have lower MTUs, PMTU discovery will
2012 if (dst_mtu(&rt->dst) >= mtu)
2015 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
2021 static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
2022 const struct fib6_nh *nh, int mtu)
2024 struct rt6_exception_bucket *bucket;
2025 struct rt6_exception *rt6_ex;
2028 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2032 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2033 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2034 struct rt6_info *entry = rt6_ex->rt6i;
2036 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2037 * route), the metrics of its rt->from have already
2040 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2041 rt6_mtu_change_route_allowed(idev, entry, mtu))
2042 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2048 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2050 static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2051 const struct in6_addr *gateway)
2053 struct rt6_exception_bucket *bucket;
2054 struct rt6_exception *rt6_ex;
2055 struct hlist_node *tmp;
2058 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2061 spin_lock_bh(&rt6_exception_lock);
2062 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2064 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2065 hlist_for_each_entry_safe(rt6_ex, tmp,
2066 &bucket->chain, hlist) {
2067 struct rt6_info *entry = rt6_ex->rt6i;
2069 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2070 RTF_CACHE_GATEWAY &&
2071 ipv6_addr_equal(gateway,
2072 &entry->rt6i_gateway)) {
2073 rt6_remove_exception(bucket, rt6_ex);
2080 spin_unlock_bh(&rt6_exception_lock);
2083 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2084 struct rt6_exception *rt6_ex,
2085 struct fib6_gc_args *gc_args,
2088 struct rt6_info *rt = rt6_ex->rt6i;
2090 /* we are pruning and obsoleting aged-out and non gateway exceptions
2091 * even if others have still references to them, so that on next
2092 * dst_check() such references can be dropped.
2093 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2094 * expired, independently from their aging, as per RFC 8201 section 4
2096 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2097 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2098 pr_debug("aging clone %p\n", rt);
2099 rt6_remove_exception(bucket, rt6_ex);
2102 } else if (time_after(jiffies, rt->dst.expires)) {
2103 pr_debug("purging expired route %p\n", rt);
2104 rt6_remove_exception(bucket, rt6_ex);
2108 if (rt->rt6i_flags & RTF_GATEWAY) {
2109 struct neighbour *neigh;
2111 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2113 if (!(neigh && (neigh->flags & NTF_ROUTER))) {
2114 pr_debug("purging route %p via non-router but gateway\n",
2116 rt6_remove_exception(bucket, rt6_ex);
2124 static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2125 struct fib6_gc_args *gc_args,
2128 struct rt6_exception_bucket *bucket;
2129 struct rt6_exception *rt6_ex;
2130 struct hlist_node *tmp;
2133 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2137 spin_lock(&rt6_exception_lock);
2138 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2140 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2141 hlist_for_each_entry_safe(rt6_ex, tmp,
2142 &bucket->chain, hlist) {
2143 rt6_age_examine_exception(bucket, rt6_ex,
2149 spin_unlock(&rt6_exception_lock);
2150 rcu_read_unlock_bh();
2153 struct fib6_nh_age_excptn_arg {
2154 struct fib6_gc_args *gc_args;
2158 static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2160 struct fib6_nh_age_excptn_arg *arg = _arg;
2162 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2166 void rt6_age_exceptions(struct fib6_info *f6i,
2167 struct fib6_gc_args *gc_args,
2171 struct fib6_nh_age_excptn_arg arg = {
2176 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2179 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2183 /* must be called with rcu lock held */
2184 int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2185 struct flowi6 *fl6, struct fib6_result *res, int strict)
2187 struct fib6_node *fn, *saved_fn;
2189 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2193 rt6_select(net, fn, oif, res, strict);
2194 if (res->f6i == net->ipv6.fib6_null_entry) {
2195 fn = fib6_backtrack(fn, &fl6->saddr);
2197 goto redo_rt6_select;
2198 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2199 /* also consider unreachable route */
2200 strict &= ~RT6_LOOKUP_F_REACHABLE;
2202 goto redo_rt6_select;
2206 trace_fib6_table_lookup(net, res, table, fl6);
2211 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2212 int oif, struct flowi6 *fl6,
2213 const struct sk_buff *skb, int flags)
2215 struct fib6_result res = {};
2216 struct rt6_info *rt = NULL;
2219 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2220 !rcu_read_lock_held());
2222 strict |= flags & RT6_LOOKUP_F_IFACE;
2223 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2224 if (READ_ONCE(net->ipv6.devconf_all->forwarding) == 0)
2225 strict |= RT6_LOOKUP_F_REACHABLE;
2229 fib6_table_lookup(net, table, oif, fl6, &res, strict);
2230 if (res.f6i == net->ipv6.fib6_null_entry)
2233 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2235 /*Search through exception table */
2236 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2239 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2240 !res.nh->fib_nh_gw_family)) {
2241 /* Create a RTF_CACHE clone which will not be
2242 * owned by the fib6 tree. It is for the special case where
2243 * the daddr in the skb during the neighbor look-up is different
2244 * from the fl6->daddr used to look-up route here.
2246 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2249 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2250 * As rt6_uncached_list_add() does not consume refcnt,
2251 * this refcnt is always returned to the caller even
2252 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2254 rt6_uncached_list_add(rt);
2260 /* Get a percpu copy */
2262 rt = rt6_get_pcpu_route(&res);
2265 rt = rt6_make_pcpu_route(net, &res);
2271 rt = net->ipv6.ip6_null_entry;
2272 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2273 ip6_hold_safe(net, &rt);
2278 EXPORT_SYMBOL_GPL(ip6_pol_route);
2280 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
2281 struct fib6_table *table,
2283 const struct sk_buff *skb,
2286 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2289 struct dst_entry *ip6_route_input_lookup(struct net *net,
2290 struct net_device *dev,
2292 const struct sk_buff *skb,
2295 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2296 flags |= RT6_LOOKUP_F_IFACE;
2298 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2300 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2302 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2303 struct flow_keys *keys,
2304 struct flow_keys *flkeys)
2306 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2307 const struct ipv6hdr *key_iph = outer_iph;
2308 struct flow_keys *_flkeys = flkeys;
2309 const struct ipv6hdr *inner_iph;
2310 const struct icmp6hdr *icmph;
2311 struct ipv6hdr _inner_iph;
2312 struct icmp6hdr _icmph;
2314 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2317 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2318 sizeof(_icmph), &_icmph);
2322 if (!icmpv6_is_err(icmph->icmp6_type))
2325 inner_iph = skb_header_pointer(skb,
2326 skb_transport_offset(skb) + sizeof(*icmph),
2327 sizeof(_inner_iph), &_inner_iph);
2331 key_iph = inner_iph;
2335 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2336 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2337 keys->tags.flow_label = _flkeys->tags.flow_label;
2338 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2340 keys->addrs.v6addrs.src = key_iph->saddr;
2341 keys->addrs.v6addrs.dst = key_iph->daddr;
2342 keys->tags.flow_label = ip6_flowlabel(key_iph);
2343 keys->basic.ip_proto = key_iph->nexthdr;
2347 static u32 rt6_multipath_custom_hash_outer(const struct net *net,
2348 const struct sk_buff *skb,
2351 u32 hash_fields = ip6_multipath_hash_fields(net);
2352 struct flow_keys keys, hash_keys;
2354 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2357 memset(&hash_keys, 0, sizeof(hash_keys));
2358 skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_ENCAP);
2360 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2361 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2362 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2363 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2364 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2365 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2366 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2367 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2368 hash_keys.tags.flow_label = keys.tags.flow_label;
2369 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2370 hash_keys.ports.src = keys.ports.src;
2371 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2372 hash_keys.ports.dst = keys.ports.dst;
2374 *p_has_inner = !!(keys.control.flags & FLOW_DIS_ENCAPSULATION);
2375 return flow_hash_from_keys(&hash_keys);
2378 static u32 rt6_multipath_custom_hash_inner(const struct net *net,
2379 const struct sk_buff *skb,
2382 u32 hash_fields = ip6_multipath_hash_fields(net);
2383 struct flow_keys keys, hash_keys;
2385 /* We assume the packet carries an encapsulation, but if none was
2386 * encountered during dissection of the outer flow, then there is no
2387 * point in calling the flow dissector again.
2392 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_MASK))
2395 memset(&hash_keys, 0, sizeof(hash_keys));
2396 skb_flow_dissect_flow_keys(skb, &keys, 0);
2398 if (!(keys.control.flags & FLOW_DIS_ENCAPSULATION))
2401 if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2402 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2403 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2404 hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
2405 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2406 hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst;
2407 } else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2408 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2409 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2410 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2411 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2412 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2413 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL)
2414 hash_keys.tags.flow_label = keys.tags.flow_label;
2417 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO)
2418 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2419 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT)
2420 hash_keys.ports.src = keys.ports.src;
2421 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
2422 hash_keys.ports.dst = keys.ports.dst;
2424 return flow_hash_from_keys(&hash_keys);
2427 static u32 rt6_multipath_custom_hash_skb(const struct net *net,
2428 const struct sk_buff *skb)
2430 u32 mhash, mhash_inner;
2431 bool has_inner = true;
2433 mhash = rt6_multipath_custom_hash_outer(net, skb, &has_inner);
2434 mhash_inner = rt6_multipath_custom_hash_inner(net, skb, has_inner);
2436 return jhash_2words(mhash, mhash_inner, 0);
2439 static u32 rt6_multipath_custom_hash_fl6(const struct net *net,
2440 const struct flowi6 *fl6)
2442 u32 hash_fields = ip6_multipath_hash_fields(net);
2443 struct flow_keys hash_keys;
2445 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2448 memset(&hash_keys, 0, sizeof(hash_keys));
2449 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2450 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2451 hash_keys.addrs.v6addrs.src = fl6->saddr;
2452 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2453 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2454 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2455 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2456 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2457 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2458 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2459 hash_keys.ports.src = fl6->fl6_sport;
2460 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2461 hash_keys.ports.dst = fl6->fl6_dport;
2463 return flow_hash_from_keys(&hash_keys);
2466 /* if skb is set it will be used and fl6 can be NULL */
2467 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2468 const struct sk_buff *skb, struct flow_keys *flkeys)
2470 struct flow_keys hash_keys;
2473 switch (ip6_multipath_hash_policy(net)) {
2475 memset(&hash_keys, 0, sizeof(hash_keys));
2476 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2478 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2480 hash_keys.addrs.v6addrs.src = fl6->saddr;
2481 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2482 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2483 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2485 mhash = flow_hash_from_keys(&hash_keys);
2489 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2490 struct flow_keys keys;
2492 /* short-circuit if we already have L4 hash present */
2494 return skb_get_hash_raw(skb) >> 1;
2496 memset(&hash_keys, 0, sizeof(hash_keys));
2499 skb_flow_dissect_flow_keys(skb, &keys, flag);
2502 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2503 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2504 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2505 hash_keys.ports.src = flkeys->ports.src;
2506 hash_keys.ports.dst = flkeys->ports.dst;
2507 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2509 memset(&hash_keys, 0, sizeof(hash_keys));
2510 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2511 hash_keys.addrs.v6addrs.src = fl6->saddr;
2512 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2513 hash_keys.ports.src = fl6->fl6_sport;
2514 hash_keys.ports.dst = fl6->fl6_dport;
2515 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2517 mhash = flow_hash_from_keys(&hash_keys);
2520 memset(&hash_keys, 0, sizeof(hash_keys));
2521 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2523 struct flow_keys keys;
2526 skb_flow_dissect_flow_keys(skb, &keys, 0);
2530 /* Inner can be v4 or v6 */
2531 if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2532 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2533 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2534 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2535 } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2536 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2537 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2538 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2539 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2540 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2542 /* Same as case 0 */
2543 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2544 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2547 /* Same as case 0 */
2548 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2549 hash_keys.addrs.v6addrs.src = fl6->saddr;
2550 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2551 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2552 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2554 mhash = flow_hash_from_keys(&hash_keys);
2558 mhash = rt6_multipath_custom_hash_skb(net, skb);
2560 mhash = rt6_multipath_custom_hash_fl6(net, fl6);
2567 /* Called with rcu held */
2568 void ip6_route_input(struct sk_buff *skb)
2570 const struct ipv6hdr *iph = ipv6_hdr(skb);
2571 struct net *net = dev_net(skb->dev);
2572 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2573 struct ip_tunnel_info *tun_info;
2574 struct flowi6 fl6 = {
2575 .flowi6_iif = skb->dev->ifindex,
2576 .daddr = iph->daddr,
2577 .saddr = iph->saddr,
2578 .flowlabel = ip6_flowinfo(iph),
2579 .flowi6_mark = skb->mark,
2580 .flowi6_proto = iph->nexthdr,
2582 struct flow_keys *flkeys = NULL, _flkeys;
2584 tun_info = skb_tunnel_info(skb);
2585 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2586 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2588 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2591 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2592 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2594 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2598 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
2599 struct fib6_table *table,
2601 const struct sk_buff *skb,
2604 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2607 static struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2608 const struct sock *sk,
2614 if (ipv6_addr_type(&fl6->daddr) &
2615 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2616 struct dst_entry *dst;
2618 /* This function does not take refcnt on the dst */
2619 dst = l3mdev_link_scope_lookup(net, fl6);
2624 fl6->flowi6_iif = LOOPBACK_IFINDEX;
2626 flags |= RT6_LOOKUP_F_DST_NOREF;
2627 any_src = ipv6_addr_any(&fl6->saddr);
2628 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2629 (fl6->flowi6_oif && any_src))
2630 flags |= RT6_LOOKUP_F_IFACE;
2633 flags |= RT6_LOOKUP_F_HAS_SADDR;
2635 flags |= rt6_srcprefs2flags(READ_ONCE(inet6_sk(sk)->srcprefs));
2637 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2640 struct dst_entry *ip6_route_output_flags(struct net *net,
2641 const struct sock *sk,
2645 struct dst_entry *dst;
2646 struct rt6_info *rt6;
2649 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2650 rt6 = (struct rt6_info *)dst;
2651 /* For dst cached in uncached_list, refcnt is already taken. */
2652 if (list_empty(&rt6->dst.rt_uncached) && !dst_hold_safe(dst)) {
2653 dst = &net->ipv6.ip6_null_entry->dst;
2660 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2662 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2664 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2665 struct net_device *loopback_dev = net->loopback_dev;
2666 struct dst_entry *new = NULL;
2668 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev,
2669 DST_OBSOLETE_DEAD, 0);
2672 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2676 new->input = dst_discard;
2677 new->output = dst_discard_out;
2679 dst_copy_metrics(new, &ort->dst);
2681 rt->rt6i_idev = in6_dev_get(loopback_dev);
2682 rt->rt6i_gateway = ort->rt6i_gateway;
2683 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2685 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2686 #ifdef CONFIG_IPV6_SUBTREES
2687 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2691 dst_release(dst_orig);
2692 return new ? new : ERR_PTR(-ENOMEM);
2696 * Destination cache support functions
2699 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2703 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2706 if (fib6_check_expired(f6i))
2712 static struct dst_entry *rt6_check(struct rt6_info *rt,
2713 struct fib6_info *from,
2718 if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2719 rt_cookie != cookie)
2722 if (rt6_check_expired(rt))
2728 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2729 struct fib6_info *from,
2732 if (!__rt6_check_expired(rt) &&
2733 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2734 fib6_check(from, cookie))
2740 INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2743 struct dst_entry *dst_ret;
2744 struct fib6_info *from;
2745 struct rt6_info *rt;
2747 rt = container_of(dst, struct rt6_info, dst);
2750 return rt6_is_valid(rt) ? dst : NULL;
2754 /* All IPV6 dsts are created with ->obsolete set to the value
2755 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2756 * into this function always.
2759 from = rcu_dereference(rt->from);
2761 if (from && (rt->rt6i_flags & RTF_PCPU ||
2762 unlikely(!list_empty(&rt->dst.rt_uncached))))
2763 dst_ret = rt6_dst_from_check(rt, from, cookie);
2765 dst_ret = rt6_check(rt, from, cookie);
2771 EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
2773 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2775 struct rt6_info *rt = (struct rt6_info *) dst;
2778 if (rt->rt6i_flags & RTF_CACHE) {
2780 if (rt6_check_expired(rt)) {
2781 rt6_remove_exception_rt(rt);
2793 static void ip6_link_failure(struct sk_buff *skb)
2795 struct rt6_info *rt;
2797 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2799 rt = (struct rt6_info *) skb_dst(skb);
2802 if (rt->rt6i_flags & RTF_CACHE) {
2803 rt6_remove_exception_rt(rt);
2805 struct fib6_info *from;
2806 struct fib6_node *fn;
2808 from = rcu_dereference(rt->from);
2810 fn = rcu_dereference(from->fib6_node);
2811 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2812 WRITE_ONCE(fn->fn_sernum, -1);
2819 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2821 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2822 struct fib6_info *from;
2825 from = rcu_dereference(rt0->from);
2827 rt0->dst.expires = from->expires;
2831 dst_set_expires(&rt0->dst, timeout);
2832 rt0->rt6i_flags |= RTF_EXPIRES;
2835 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2837 struct net *net = dev_net(rt->dst.dev);
2839 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2840 rt->rt6i_flags |= RTF_MODIFIED;
2841 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2844 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2846 return !(rt->rt6i_flags & RTF_CACHE) &&
2847 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2850 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2851 const struct ipv6hdr *iph, u32 mtu,
2854 const struct in6_addr *daddr, *saddr;
2855 struct rt6_info *rt6 = (struct rt6_info *)dst;
2857 /* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2858 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2859 * [see also comment in rt6_mtu_change_route()]
2863 daddr = &iph->daddr;
2864 saddr = &iph->saddr;
2866 daddr = &sk->sk_v6_daddr;
2867 saddr = &inet6_sk(sk)->saddr;
2874 dst_confirm_neigh(dst, daddr);
2876 if (mtu < IPV6_MIN_MTU)
2878 if (mtu >= dst_mtu(dst))
2881 if (!rt6_cache_allowed_for_pmtu(rt6)) {
2882 rt6_do_update_pmtu(rt6, mtu);
2883 /* update rt6_ex->stamp for cache */
2884 if (rt6->rt6i_flags & RTF_CACHE)
2885 rt6_update_exception_stamp_rt(rt6);
2887 struct fib6_result res = {};
2888 struct rt6_info *nrt6;
2891 res.f6i = rcu_dereference(rt6->from);
2895 res.fib6_flags = res.f6i->fib6_flags;
2896 res.fib6_type = res.f6i->fib6_type;
2899 struct fib6_nh_match_arg arg = {
2901 .gw = &rt6->rt6i_gateway,
2904 nexthop_for_each_fib6_nh(res.f6i->nh,
2905 fib6_nh_find_match, &arg);
2907 /* fib6_info uses a nexthop that does not have fib6_nh
2908 * using the dst->dev + gw. Should be impossible.
2915 res.nh = res.f6i->fib6_nh;
2918 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2920 rt6_do_update_pmtu(nrt6, mtu);
2921 if (rt6_insert_exception(nrt6, &res))
2922 dst_release_immediate(&nrt6->dst);
2929 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2930 struct sk_buff *skb, u32 mtu,
2933 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2937 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2938 int oif, u32 mark, kuid_t uid)
2940 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2941 struct dst_entry *dst;
2942 struct flowi6 fl6 = {
2944 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2945 .daddr = iph->daddr,
2946 .saddr = iph->saddr,
2947 .flowlabel = ip6_flowinfo(iph),
2951 dst = ip6_route_output(net, NULL, &fl6);
2953 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
2956 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2958 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2960 int oif = sk->sk_bound_dev_if;
2961 struct dst_entry *dst;
2963 if (!oif && skb->dev)
2964 oif = l3mdev_master_ifindex(skb->dev);
2966 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, READ_ONCE(sk->sk_mark),
2969 dst = __sk_dst_get(sk);
2970 if (!dst || !dst->obsolete ||
2971 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2975 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2976 ip6_datagram_dst_update(sk, false);
2979 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2981 void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2982 const struct flowi6 *fl6)
2984 #ifdef CONFIG_IPV6_SUBTREES
2985 struct ipv6_pinfo *np = inet6_sk(sk);
2988 ip6_dst_store(sk, dst,
2989 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2990 &sk->sk_v6_daddr : NULL,
2991 #ifdef CONFIG_IPV6_SUBTREES
2992 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2998 static bool ip6_redirect_nh_match(const struct fib6_result *res,
3000 const struct in6_addr *gw,
3001 struct rt6_info **ret)
3003 const struct fib6_nh *nh = res->nh;
3005 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
3006 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
3009 /* rt_cache's gateway might be different from its 'parent'
3010 * in the case of an ip redirect.
3011 * So we keep searching in the exception table if the gateway
3014 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
3015 struct rt6_info *rt_cache;
3017 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
3019 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
3028 struct fib6_nh_rd_arg {
3029 struct fib6_result *res;
3031 const struct in6_addr *gw;
3032 struct rt6_info **ret;
3035 static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
3037 struct fib6_nh_rd_arg *arg = _arg;
3040 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
3043 /* Handle redirects */
3044 struct ip6rd_flowi {
3046 struct in6_addr gateway;
3049 INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
3050 struct fib6_table *table,
3052 const struct sk_buff *skb,
3055 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
3056 struct rt6_info *ret = NULL;
3057 struct fib6_result res = {};
3058 struct fib6_nh_rd_arg arg = {
3061 .gw = &rdfl->gateway,
3064 struct fib6_info *rt;
3065 struct fib6_node *fn;
3067 /* Get the "current" route for this destination and
3068 * check if the redirect has come from appropriate router.
3070 * RFC 4861 specifies that redirects should only be
3071 * accepted if they come from the nexthop to the target.
3072 * Due to the way the routes are chosen, this notion
3073 * is a bit fuzzy and one might need to check all possible
3078 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
3080 for_each_fib6_node_rt_rcu(fn) {
3082 if (fib6_check_expired(rt))
3084 if (rt->fib6_flags & RTF_REJECT)
3086 if (unlikely(rt->nh)) {
3087 if (nexthop_is_blackhole(rt->nh))
3089 /* on match, res->nh is filled in and potentially ret */
3090 if (nexthop_for_each_fib6_nh(rt->nh,
3091 fib6_nh_redirect_match,
3095 res.nh = rt->fib6_nh;
3096 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
3103 rt = net->ipv6.fib6_null_entry;
3104 else if (rt->fib6_flags & RTF_REJECT) {
3105 ret = net->ipv6.ip6_null_entry;
3109 if (rt == net->ipv6.fib6_null_entry) {
3110 fn = fib6_backtrack(fn, &fl6->saddr);
3116 res.nh = rt->fib6_nh;
3119 ip6_hold_safe(net, &ret);
3121 res.fib6_flags = res.f6i->fib6_flags;
3122 res.fib6_type = res.f6i->fib6_type;
3123 ret = ip6_create_rt_rcu(&res);
3128 trace_fib6_table_lookup(net, &res, table, fl6);
3132 static struct dst_entry *ip6_route_redirect(struct net *net,
3133 const struct flowi6 *fl6,
3134 const struct sk_buff *skb,
3135 const struct in6_addr *gateway)
3137 int flags = RT6_LOOKUP_F_HAS_SADDR;
3138 struct ip6rd_flowi rdfl;
3141 rdfl.gateway = *gateway;
3143 return fib6_rule_lookup(net, &rdfl.fl6, skb,
3144 flags, __ip6_route_redirect);
3147 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3150 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3151 struct dst_entry *dst;
3152 struct flowi6 fl6 = {
3153 .flowi6_iif = LOOPBACK_IFINDEX,
3155 .flowi6_mark = mark,
3156 .daddr = iph->daddr,
3157 .saddr = iph->saddr,
3158 .flowlabel = ip6_flowinfo(iph),
3162 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3163 rt6_do_redirect(dst, NULL, skb);
3166 EXPORT_SYMBOL_GPL(ip6_redirect);
3168 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
3170 const struct ipv6hdr *iph = ipv6_hdr(skb);
3171 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3172 struct dst_entry *dst;
3173 struct flowi6 fl6 = {
3174 .flowi6_iif = LOOPBACK_IFINDEX,
3177 .saddr = iph->daddr,
3178 .flowi6_uid = sock_net_uid(net, NULL),
3181 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3182 rt6_do_redirect(dst, NULL, skb);
3186 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3188 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if,
3189 READ_ONCE(sk->sk_mark), sk->sk_uid);
3191 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3193 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3195 struct net_device *dev = dst->dev;
3196 unsigned int mtu = dst_mtu(dst);
3197 struct net *net = dev_net(dev);
3199 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3201 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3202 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3205 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3206 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3207 * IPV6_MAXPLEN is also valid and means: "any MSS,
3208 * rely only on pmtu discovery"
3210 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3215 INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
3217 return ip6_dst_mtu_maybe_forward(dst, false);
3219 EXPORT_INDIRECT_CALLABLE(ip6_mtu);
3222 * 1. mtu on route is locked - use it
3223 * 2. mtu from nexthop exception
3224 * 3. mtu from egress device
3226 * based on ip6_dst_mtu_forward and exception logic of
3227 * rt6_find_cached_rt; called with rcu_read_lock
3229 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3230 const struct in6_addr *daddr,
3231 const struct in6_addr *saddr)
3233 const struct fib6_nh *nh = res->nh;
3234 struct fib6_info *f6i = res->f6i;
3235 struct inet6_dev *idev;
3236 struct rt6_info *rt;
3239 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3240 mtu = f6i->fib6_pmtu;
3245 rt = rt6_find_cached_rt(res, daddr, saddr);
3247 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3249 struct net_device *dev = nh->fib_nh_dev;
3252 idev = __in6_dev_get(dev);
3254 mtu = max_t(u32, mtu, READ_ONCE(idev->cnf.mtu6));
3257 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3259 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3262 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3265 struct dst_entry *dst;
3266 struct rt6_info *rt;
3267 struct inet6_dev *idev = in6_dev_get(dev);
3268 struct net *net = dev_net(dev);
3270 if (unlikely(!idev))
3271 return ERR_PTR(-ENODEV);
3273 rt = ip6_dst_alloc(net, dev, 0);
3274 if (unlikely(!rt)) {
3276 dst = ERR_PTR(-ENOMEM);
3280 rt->dst.input = ip6_input;
3281 rt->dst.output = ip6_output;
3282 rt->rt6i_gateway = fl6->daddr;
3283 rt->rt6i_dst.addr = fl6->daddr;
3284 rt->rt6i_dst.plen = 128;
3285 rt->rt6i_idev = idev;
3286 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3288 /* Add this dst into uncached_list so that rt6_disable_ip() can
3289 * do proper release of the net_device
3291 rt6_uncached_list_add(rt);
3293 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3299 static void ip6_dst_gc(struct dst_ops *ops)
3301 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3302 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3303 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3304 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3305 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3309 if (time_after(rt_last_gc + rt_min_interval, jiffies))
3312 fib6_run_gc(atomic_inc_return(&net->ipv6.ip6_rt_gc_expire), net, true);
3313 entries = dst_entries_get_slow(ops);
3314 if (entries < ops->gc_thresh)
3315 atomic_set(&net->ipv6.ip6_rt_gc_expire, rt_gc_timeout >> 1);
3317 val = atomic_read(&net->ipv6.ip6_rt_gc_expire);
3318 atomic_set(&net->ipv6.ip6_rt_gc_expire, val - (val >> rt_elasticity));
3321 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3322 const struct in6_addr *gw_addr, u32 tbid,
3323 int flags, struct fib6_result *res)
3325 struct flowi6 fl6 = {
3326 .flowi6_oif = cfg->fc_ifindex,
3328 .saddr = cfg->fc_prefsrc,
3330 struct fib6_table *table;
3333 table = fib6_get_table(net, tbid);
3337 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3338 flags |= RT6_LOOKUP_F_HAS_SADDR;
3340 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3342 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3343 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3344 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3345 cfg->fc_ifindex != 0, NULL, flags);
3350 static int ip6_route_check_nh_onlink(struct net *net,
3351 struct fib6_config *cfg,
3352 const struct net_device *dev,
3353 struct netlink_ext_ack *extack)
3355 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3356 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3357 struct fib6_result res = {};
3360 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3361 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3362 /* ignore match if it is the default route */
3363 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3364 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3365 NL_SET_ERR_MSG(extack,
3366 "Nexthop has invalid gateway or device mismatch");
3373 static int ip6_route_check_nh(struct net *net,
3374 struct fib6_config *cfg,
3375 struct net_device **_dev,
3376 netdevice_tracker *dev_tracker,
3377 struct inet6_dev **idev)
3379 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3380 struct net_device *dev = _dev ? *_dev : NULL;
3381 int flags = RT6_LOOKUP_F_IFACE;
3382 struct fib6_result res = {};
3383 int err = -EHOSTUNREACH;
3385 if (cfg->fc_table) {
3386 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3387 cfg->fc_table, flags, &res);
3388 /* gw_addr can not require a gateway or resolve to a reject
3389 * route. If a device is given, it must match the result.
3391 if (err || res.fib6_flags & RTF_REJECT ||
3392 res.nh->fib_nh_gw_family ||
3393 (dev && dev != res.nh->fib_nh_dev))
3394 err = -EHOSTUNREACH;
3398 struct flowi6 fl6 = {
3399 .flowi6_oif = cfg->fc_ifindex,
3403 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3404 if (err || res.fib6_flags & RTF_REJECT ||
3405 res.nh->fib_nh_gw_family)
3406 err = -EHOSTUNREACH;
3411 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3412 cfg->fc_ifindex != 0, NULL, flags);
3417 if (dev != res.nh->fib_nh_dev)
3418 err = -EHOSTUNREACH;
3420 *_dev = dev = res.nh->fib_nh_dev;
3421 netdev_hold(dev, dev_tracker, GFP_ATOMIC);
3422 *idev = in6_dev_get(dev);
3428 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3429 struct net_device **_dev,
3430 netdevice_tracker *dev_tracker,
3431 struct inet6_dev **idev,
3432 struct netlink_ext_ack *extack)
3434 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3435 int gwa_type = ipv6_addr_type(gw_addr);
3436 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3437 const struct net_device *dev = *_dev;
3438 bool need_addr_check = !dev;
3441 /* if gw_addr is local we will fail to detect this in case
3442 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3443 * will return already-added prefix route via interface that
3444 * prefix route was assigned to, which might be non-loopback.
3447 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3448 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3452 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3453 /* IPv6 strictly inhibits using not link-local
3454 * addresses as nexthop address.
3455 * Otherwise, router will not able to send redirects.
3456 * It is very good, but in some (rare!) circumstances
3457 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3458 * some exceptions. --ANK
3459 * We allow IPv4-mapped nexthops to support RFC4798-type
3462 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3463 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3469 if (cfg->fc_flags & RTNH_F_ONLINK)
3470 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3472 err = ip6_route_check_nh(net, cfg, _dev, dev_tracker,
3481 /* reload in case device was changed */
3486 NL_SET_ERR_MSG(extack, "Egress device not specified");
3488 } else if (dev->flags & IFF_LOOPBACK) {
3489 NL_SET_ERR_MSG(extack,
3490 "Egress device can not be loopback device for this route");
3494 /* if we did not check gw_addr above, do so now that the
3495 * egress device has been resolved.
3497 if (need_addr_check &&
3498 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3499 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3508 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3510 if ((flags & RTF_REJECT) ||
3511 (dev && (dev->flags & IFF_LOOPBACK) &&
3512 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3513 !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3519 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3520 struct fib6_config *cfg, gfp_t gfp_flags,
3521 struct netlink_ext_ack *extack)
3523 netdevice_tracker *dev_tracker = &fib6_nh->fib_nh_dev_tracker;
3524 struct net_device *dev = NULL;
3525 struct inet6_dev *idev = NULL;
3529 fib6_nh->fib_nh_family = AF_INET6;
3530 #ifdef CONFIG_IPV6_ROUTER_PREF
3531 fib6_nh->last_probe = jiffies;
3533 if (cfg->fc_is_fdb) {
3534 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3535 fib6_nh->fib_nh_gw_family = AF_INET6;
3540 if (cfg->fc_ifindex) {
3541 dev = netdev_get_by_index(net, cfg->fc_ifindex,
3542 dev_tracker, gfp_flags);
3545 idev = in6_dev_get(dev);
3550 if (cfg->fc_flags & RTNH_F_ONLINK) {
3552 NL_SET_ERR_MSG(extack,
3553 "Nexthop device required for onlink");
3557 if (!(dev->flags & IFF_UP)) {
3558 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3563 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3566 fib6_nh->fib_nh_weight = 1;
3568 /* We cannot add true routes via loopback here,
3569 * they would result in kernel looping; promote them to reject routes
3571 addr_type = ipv6_addr_type(&cfg->fc_dst);
3572 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3573 /* hold loopback dev/idev if we haven't done so. */
3574 if (dev != net->loopback_dev) {
3576 netdev_put(dev, dev_tracker);
3579 dev = net->loopback_dev;
3580 netdev_hold(dev, dev_tracker, gfp_flags);
3581 idev = in6_dev_get(dev);
3590 if (cfg->fc_flags & RTF_GATEWAY) {
3591 err = ip6_validate_gw(net, cfg, &dev, dev_tracker,
3596 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3597 fib6_nh->fib_nh_gw_family = AF_INET6;
3604 if (idev->cnf.disable_ipv6) {
3605 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3610 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3611 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3616 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3617 !netif_carrier_ok(dev))
3618 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3620 err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3621 cfg->fc_encap_type, cfg, gfp_flags, extack);
3626 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3627 if (!fib6_nh->rt6i_pcpu) {
3632 fib6_nh->fib_nh_dev = dev;
3633 fib6_nh->fib_nh_oif = dev->ifindex;
3640 lwtstate_put(fib6_nh->fib_nh_lws);
3641 fib6_nh->fib_nh_lws = NULL;
3642 netdev_put(dev, dev_tracker);
3648 void fib6_nh_release(struct fib6_nh *fib6_nh)
3650 struct rt6_exception_bucket *bucket;
3654 fib6_nh_flush_exceptions(fib6_nh, NULL);
3655 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3657 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3663 fib6_nh_release_dsts(fib6_nh);
3664 free_percpu(fib6_nh->rt6i_pcpu);
3666 fib_nh_common_release(&fib6_nh->nh_common);
3669 void fib6_nh_release_dsts(struct fib6_nh *fib6_nh)
3673 if (!fib6_nh->rt6i_pcpu)
3676 for_each_possible_cpu(cpu) {
3677 struct rt6_info *pcpu_rt, **ppcpu_rt;
3679 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3680 pcpu_rt = xchg(ppcpu_rt, NULL);
3682 dst_dev_put(&pcpu_rt->dst);
3683 dst_release(&pcpu_rt->dst);
3688 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3690 struct netlink_ext_ack *extack)
3692 struct net *net = cfg->fc_nlinfo.nl_net;
3693 struct fib6_info *rt = NULL;
3694 struct nexthop *nh = NULL;
3695 struct fib6_table *table;
3696 struct fib6_nh *fib6_nh;
3700 /* RTF_PCPU is an internal flag; can not be set by userspace */
3701 if (cfg->fc_flags & RTF_PCPU) {
3702 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3706 /* RTF_CACHE is an internal flag; can not be set by userspace */
3707 if (cfg->fc_flags & RTF_CACHE) {
3708 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3712 if (cfg->fc_type > RTN_MAX) {
3713 NL_SET_ERR_MSG(extack, "Invalid route type");
3717 if (cfg->fc_dst_len > 128) {
3718 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3721 if (cfg->fc_src_len > 128) {
3722 NL_SET_ERR_MSG(extack, "Invalid source address length");
3725 #ifndef CONFIG_IPV6_SUBTREES
3726 if (cfg->fc_src_len) {
3727 NL_SET_ERR_MSG(extack,
3728 "Specifying source address requires IPV6_SUBTREES to be enabled");
3732 if (cfg->fc_nh_id) {
3733 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3735 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3738 err = fib6_check_nexthop(nh, cfg, extack);
3744 if (cfg->fc_nlinfo.nlh &&
3745 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3746 table = fib6_get_table(net, cfg->fc_table);
3748 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3749 table = fib6_new_table(net, cfg->fc_table);
3752 table = fib6_new_table(net, cfg->fc_table);
3759 rt = fib6_info_alloc(gfp_flags, !nh);
3763 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3765 if (IS_ERR(rt->fib6_metrics)) {
3766 err = PTR_ERR(rt->fib6_metrics);
3767 /* Do not leave garbage there. */
3768 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3772 if (cfg->fc_flags & RTF_ADDRCONF)
3773 rt->dst_nocount = true;
3775 if (cfg->fc_flags & RTF_EXPIRES)
3776 fib6_set_expires(rt, jiffies +
3777 clock_t_to_jiffies(cfg->fc_expires));
3779 if (cfg->fc_protocol == RTPROT_UNSPEC)
3780 cfg->fc_protocol = RTPROT_BOOT;
3781 rt->fib6_protocol = cfg->fc_protocol;
3783 rt->fib6_table = table;
3784 rt->fib6_metric = cfg->fc_metric;
3785 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3786 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3788 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3789 rt->fib6_dst.plen = cfg->fc_dst_len;
3791 #ifdef CONFIG_IPV6_SUBTREES
3792 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3793 rt->fib6_src.plen = cfg->fc_src_len;
3796 if (rt->fib6_src.plen) {
3797 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3800 if (!nexthop_get(nh)) {
3801 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3805 fib6_nh = nexthop_fib6_nh(rt->nh);
3807 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3811 fib6_nh = rt->fib6_nh;
3813 /* We cannot add true routes via loopback here, they would
3814 * result in kernel looping; promote them to reject routes
3816 addr_type = ipv6_addr_type(&cfg->fc_dst);
3817 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3819 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3822 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3823 struct net_device *dev = fib6_nh->fib_nh_dev;
3825 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3826 NL_SET_ERR_MSG(extack, "Invalid source address");
3830 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3831 rt->fib6_prefsrc.plen = 128;
3833 rt->fib6_prefsrc.plen = 0;
3837 fib6_info_release(rt);
3838 return ERR_PTR(err);
3840 ip_fib_metrics_put(rt->fib6_metrics);
3842 return ERR_PTR(err);
3845 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3846 struct netlink_ext_ack *extack)
3848 struct fib6_info *rt;
3851 rt = ip6_route_info_create(cfg, gfp_flags, extack);
3855 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3856 fib6_info_release(rt);
3861 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3863 struct net *net = info->nl_net;
3864 struct fib6_table *table;
3867 if (rt == net->ipv6.fib6_null_entry) {
3872 table = rt->fib6_table;
3873 spin_lock_bh(&table->tb6_lock);
3874 err = fib6_del(rt, info);
3875 spin_unlock_bh(&table->tb6_lock);
3878 fib6_info_release(rt);
3882 int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
3884 struct nl_info info = {
3886 .skip_notify = skip_notify
3889 return __ip6_del_rt(rt, &info);
3892 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3894 struct nl_info *info = &cfg->fc_nlinfo;
3895 struct net *net = info->nl_net;
3896 struct sk_buff *skb = NULL;
3897 struct fib6_table *table;
3900 if (rt == net->ipv6.fib6_null_entry)
3902 table = rt->fib6_table;
3903 spin_lock_bh(&table->tb6_lock);
3905 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3906 struct fib6_info *sibling, *next_sibling;
3907 struct fib6_node *fn;
3909 /* prefer to send a single notification with all hops */
3910 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3912 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3914 if (rt6_fill_node(net, skb, rt, NULL,
3915 NULL, NULL, 0, RTM_DELROUTE,
3916 info->portid, seq, 0) < 0) {
3920 info->skip_notify = 1;
3923 /* 'rt' points to the first sibling route. If it is not the
3924 * leaf, then we do not need to send a notification. Otherwise,
3925 * we need to check if the last sibling has a next route or not
3926 * and emit a replace or delete notification, respectively.
3928 info->skip_notify_kernel = 1;
3929 fn = rcu_dereference_protected(rt->fib6_node,
3930 lockdep_is_held(&table->tb6_lock));
3931 if (rcu_access_pointer(fn->leaf) == rt) {
3932 struct fib6_info *last_sibling, *replace_rt;
3934 last_sibling = list_last_entry(&rt->fib6_siblings,
3937 replace_rt = rcu_dereference_protected(
3938 last_sibling->fib6_next,
3939 lockdep_is_held(&table->tb6_lock));
3941 call_fib6_entry_notifiers_replace(net,
3944 call_fib6_multipath_entry_notifiers(net,
3945 FIB_EVENT_ENTRY_DEL,
3946 rt, rt->fib6_nsiblings,
3949 list_for_each_entry_safe(sibling, next_sibling,
3952 err = fib6_del(sibling, info);
3958 err = fib6_del(rt, info);
3960 spin_unlock_bh(&table->tb6_lock);
3962 fib6_info_release(rt);
3965 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3966 info->nlh, gfp_any());
3971 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3975 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3978 if (cfg->fc_flags & RTF_GATEWAY &&
3979 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3982 rc = rt6_remove_exception_rt(rt);
3987 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3990 struct fib6_result res = {
3994 struct rt6_info *rt_cache;
3996 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3998 return __ip6_del_cached_rt(rt_cache, cfg);
4003 struct fib6_nh_del_cached_rt_arg {
4004 struct fib6_config *cfg;
4005 struct fib6_info *f6i;
4008 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
4010 struct fib6_nh_del_cached_rt_arg *arg = _arg;
4013 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
4014 return rc != -ESRCH ? rc : 0;
4017 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
4019 struct fib6_nh_del_cached_rt_arg arg = {
4024 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
4027 static int ip6_route_del(struct fib6_config *cfg,
4028 struct netlink_ext_ack *extack)
4030 struct fib6_table *table;
4031 struct fib6_info *rt;
4032 struct fib6_node *fn;
4035 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
4037 NL_SET_ERR_MSG(extack, "FIB table does not exist");
4043 fn = fib6_locate(&table->tb6_root,
4044 &cfg->fc_dst, cfg->fc_dst_len,
4045 &cfg->fc_src, cfg->fc_src_len,
4046 !(cfg->fc_flags & RTF_CACHE));
4049 for_each_fib6_node_rt_rcu(fn) {
4052 if (rt->nh && cfg->fc_nh_id &&
4053 rt->nh->id != cfg->fc_nh_id)
4056 if (cfg->fc_flags & RTF_CACHE) {
4060 rc = ip6_del_cached_rt_nh(cfg, rt);
4061 } else if (cfg->fc_nh_id) {
4065 rc = ip6_del_cached_rt(cfg, rt, nh);
4074 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
4076 if (cfg->fc_protocol &&
4077 cfg->fc_protocol != rt->fib6_protocol)
4081 if (!fib6_info_hold_safe(rt))
4085 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4091 if (cfg->fc_ifindex &&
4093 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
4095 if (cfg->fc_flags & RTF_GATEWAY &&
4096 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
4098 if (!fib6_info_hold_safe(rt))
4102 /* if gateway was specified only delete the one hop */
4103 if (cfg->fc_flags & RTF_GATEWAY)
4104 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4106 return __ip6_del_rt_siblings(rt, cfg);
4114 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
4116 struct netevent_redirect netevent;
4117 struct rt6_info *rt, *nrt = NULL;
4118 struct fib6_result res = {};
4119 struct ndisc_options ndopts;
4120 struct inet6_dev *in6_dev;
4121 struct neighbour *neigh;
4123 int optlen, on_link;
4126 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
4127 optlen -= sizeof(*msg);
4130 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4134 msg = (struct rd_msg *)icmp6_hdr(skb);
4136 if (ipv6_addr_is_multicast(&msg->dest)) {
4137 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4142 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4144 } else if (ipv6_addr_type(&msg->target) !=
4145 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4146 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4150 in6_dev = __in6_dev_get(skb->dev);
4153 if (READ_ONCE(in6_dev->cnf.forwarding) ||
4154 !READ_ONCE(in6_dev->cnf.accept_redirects))
4158 * The IP source address of the Redirect MUST be the same as the current
4159 * first-hop router for the specified ICMP Destination Address.
4162 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4163 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4168 if (ndopts.nd_opts_tgt_lladdr) {
4169 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4172 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4177 rt = (struct rt6_info *) dst;
4178 if (rt->rt6i_flags & RTF_REJECT) {
4179 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4183 /* Redirect received -> path was valid.
4184 * Look, redirects are sent only in response to data packets,
4185 * so that this nexthop apparently is reachable. --ANK
4187 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4189 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4194 * We have finally decided to accept it.
4197 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4198 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4199 NEIGH_UPDATE_F_OVERRIDE|
4200 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4201 NEIGH_UPDATE_F_ISROUTER)),
4202 NDISC_REDIRECT, &ndopts);
4205 res.f6i = rcu_dereference(rt->from);
4210 struct fib6_nh_match_arg arg = {
4212 .gw = &rt->rt6i_gateway,
4215 nexthop_for_each_fib6_nh(res.f6i->nh,
4216 fib6_nh_find_match, &arg);
4218 /* fib6_info uses a nexthop that does not have fib6_nh
4219 * using the dst->dev. Should be impossible
4225 res.nh = res.f6i->fib6_nh;
4228 res.fib6_flags = res.f6i->fib6_flags;
4229 res.fib6_type = res.f6i->fib6_type;
4230 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4234 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4236 nrt->rt6i_flags &= ~RTF_GATEWAY;
4238 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4240 /* rt6_insert_exception() will take care of duplicated exceptions */
4241 if (rt6_insert_exception(nrt, &res)) {
4242 dst_release_immediate(&nrt->dst);
4246 netevent.old = &rt->dst;
4247 netevent.new = &nrt->dst;
4248 netevent.daddr = &msg->dest;
4249 netevent.neigh = neigh;
4250 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4254 neigh_release(neigh);
4257 #ifdef CONFIG_IPV6_ROUTE_INFO
4258 static struct fib6_info *rt6_get_route_info(struct net *net,
4259 const struct in6_addr *prefix, int prefixlen,
4260 const struct in6_addr *gwaddr,
4261 struct net_device *dev)
4263 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4264 int ifindex = dev->ifindex;
4265 struct fib6_node *fn;
4266 struct fib6_info *rt = NULL;
4267 struct fib6_table *table;
4269 table = fib6_get_table(net, tb_id);
4274 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4278 for_each_fib6_node_rt_rcu(fn) {
4279 /* these routes do not use nexthops */
4282 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4284 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4285 !rt->fib6_nh->fib_nh_gw_family)
4287 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4289 if (!fib6_info_hold_safe(rt))
4298 static struct fib6_info *rt6_add_route_info(struct net *net,
4299 const struct in6_addr *prefix, int prefixlen,
4300 const struct in6_addr *gwaddr,
4301 struct net_device *dev,
4304 struct fib6_config cfg = {
4305 .fc_metric = IP6_RT_PRIO_USER,
4306 .fc_ifindex = dev->ifindex,
4307 .fc_dst_len = prefixlen,
4308 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4309 RTF_UP | RTF_PREF(pref),
4310 .fc_protocol = RTPROT_RA,
4311 .fc_type = RTN_UNICAST,
4312 .fc_nlinfo.portid = 0,
4313 .fc_nlinfo.nlh = NULL,
4314 .fc_nlinfo.nl_net = net,
4317 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4318 cfg.fc_dst = *prefix;
4319 cfg.fc_gateway = *gwaddr;
4321 /* We should treat it as a default route if prefix length is 0. */
4323 cfg.fc_flags |= RTF_DEFAULT;
4325 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4327 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4331 struct fib6_info *rt6_get_dflt_router(struct net *net,
4332 const struct in6_addr *addr,
4333 struct net_device *dev)
4335 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4336 struct fib6_info *rt;
4337 struct fib6_table *table;
4339 table = fib6_get_table(net, tb_id);
4344 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4347 /* RA routes do not use nexthops */
4352 if (dev == nh->fib_nh_dev &&
4353 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4354 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4357 if (rt && !fib6_info_hold_safe(rt))
4363 struct fib6_info *rt6_add_dflt_router(struct net *net,
4364 const struct in6_addr *gwaddr,
4365 struct net_device *dev,
4367 u32 defrtr_usr_metric,
4370 struct fib6_config cfg = {
4371 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4372 .fc_metric = defrtr_usr_metric,
4373 .fc_ifindex = dev->ifindex,
4374 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4375 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4376 .fc_protocol = RTPROT_RA,
4377 .fc_type = RTN_UNICAST,
4378 .fc_nlinfo.portid = 0,
4379 .fc_nlinfo.nlh = NULL,
4380 .fc_nlinfo.nl_net = net,
4381 .fc_expires = jiffies_to_clock_t(lifetime * HZ),
4384 cfg.fc_gateway = *gwaddr;
4386 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4387 struct fib6_table *table;
4389 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4391 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4394 return rt6_get_dflt_router(net, gwaddr, dev);
4397 static void __rt6_purge_dflt_routers(struct net *net,
4398 struct fib6_table *table)
4400 struct fib6_info *rt;
4404 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4405 struct net_device *dev = fib6_info_nh_dev(rt);
4406 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4408 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4409 (!idev || idev->cnf.accept_ra != 2) &&
4410 fib6_info_hold_safe(rt)) {
4412 ip6_del_rt(net, rt, false);
4418 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4421 void rt6_purge_dflt_routers(struct net *net)
4423 struct fib6_table *table;
4424 struct hlist_head *head;
4429 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4430 head = &net->ipv6.fib_table_hash[h];
4431 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4432 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4433 __rt6_purge_dflt_routers(net, table);
4440 static void rtmsg_to_fib6_config(struct net *net,
4441 struct in6_rtmsg *rtmsg,
4442 struct fib6_config *cfg)
4444 *cfg = (struct fib6_config){
4445 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4447 .fc_ifindex = rtmsg->rtmsg_ifindex,
4448 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
4449 .fc_expires = rtmsg->rtmsg_info,
4450 .fc_dst_len = rtmsg->rtmsg_dst_len,
4451 .fc_src_len = rtmsg->rtmsg_src_len,
4452 .fc_flags = rtmsg->rtmsg_flags,
4453 .fc_type = rtmsg->rtmsg_type,
4455 .fc_nlinfo.nl_net = net,
4457 .fc_dst = rtmsg->rtmsg_dst,
4458 .fc_src = rtmsg->rtmsg_src,
4459 .fc_gateway = rtmsg->rtmsg_gateway,
4463 int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
4465 struct fib6_config cfg;
4468 if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4470 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4473 rtmsg_to_fib6_config(net, rtmsg, &cfg);
4478 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4481 err = ip6_route_del(&cfg, NULL);
4489 * Drop the packet on the floor
4492 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4494 struct dst_entry *dst = skb_dst(skb);
4495 struct net *net = dev_net(dst->dev);
4496 struct inet6_dev *idev;
4500 if (netif_is_l3_master(skb->dev) ||
4501 dst->dev == net->loopback_dev)
4502 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4504 idev = ip6_dst_idev(dst);
4506 switch (ipstats_mib_noroutes) {
4507 case IPSTATS_MIB_INNOROUTES:
4508 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4509 if (type == IPV6_ADDR_ANY) {
4510 SKB_DR_SET(reason, IP_INADDRERRORS);
4511 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4514 SKB_DR_SET(reason, IP_INNOROUTES);
4516 case IPSTATS_MIB_OUTNOROUTES:
4517 SKB_DR_OR(reason, IP_OUTNOROUTES);
4518 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4522 /* Start over by dropping the dst for l3mdev case */
4523 if (netif_is_l3_master(skb->dev))
4526 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4527 kfree_skb_reason(skb, reason);
4531 static int ip6_pkt_discard(struct sk_buff *skb)
4533 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4536 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4538 skb->dev = skb_dst(skb)->dev;
4539 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4542 static int ip6_pkt_prohibit(struct sk_buff *skb)
4544 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4547 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4549 skb->dev = skb_dst(skb)->dev;
4550 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4554 * Allocate a dst for local (unicast / anycast) address.
4557 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4558 struct inet6_dev *idev,
4559 const struct in6_addr *addr,
4560 bool anycast, gfp_t gfp_flags,
4561 struct netlink_ext_ack *extack)
4563 struct fib6_config cfg = {
4564 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4565 .fc_ifindex = idev->dev->ifindex,
4566 .fc_flags = RTF_UP | RTF_NONEXTHOP,
4569 .fc_protocol = RTPROT_KERNEL,
4570 .fc_nlinfo.nl_net = net,
4571 .fc_ignore_dev_down = true,
4573 struct fib6_info *f6i;
4576 cfg.fc_type = RTN_ANYCAST;
4577 cfg.fc_flags |= RTF_ANYCAST;
4579 cfg.fc_type = RTN_LOCAL;
4580 cfg.fc_flags |= RTF_LOCAL;
4583 f6i = ip6_route_info_create(&cfg, gfp_flags, extack);
4585 f6i->dst_nocount = true;
4588 (READ_ONCE(net->ipv6.devconf_all->disable_policy) ||
4589 READ_ONCE(idev->cnf.disable_policy)))
4590 f6i->dst_nopolicy = true;
4596 /* remove deleted ip from prefsrc entries */
4597 struct arg_dev_net_ip {
4599 struct in6_addr *addr;
4602 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4604 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4605 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4608 rt != net->ipv6.fib6_null_entry &&
4609 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr) &&
4610 !ipv6_chk_addr(net, addr, rt->fib6_nh->fib_nh_dev, 0)) {
4611 spin_lock_bh(&rt6_exception_lock);
4612 /* remove prefsrc entry */
4613 rt->fib6_prefsrc.plen = 0;
4614 spin_unlock_bh(&rt6_exception_lock);
4619 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4621 struct net *net = dev_net(ifp->idev->dev);
4622 struct arg_dev_net_ip adni = {
4626 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4629 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
4631 /* Remove routers and update dst entries when gateway turn into host. */
4632 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4634 struct in6_addr *gateway = (struct in6_addr *)arg;
4637 /* RA routes do not use nexthops */
4642 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4643 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4646 /* Further clean up cached routes in exception table.
4647 * This is needed because cached route may have a different
4648 * gateway than its 'parent' in the case of an ip redirect.
4650 fib6_nh_exceptions_clean_tohost(nh, gateway);
4655 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4657 fib6_clean_all(net, fib6_clean_tohost, gateway);
4660 struct arg_netdev_event {
4661 const struct net_device *dev;
4663 unsigned char nh_flags;
4664 unsigned long event;
4668 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4670 struct fib6_info *iter;
4671 struct fib6_node *fn;
4673 fn = rcu_dereference_protected(rt->fib6_node,
4674 lockdep_is_held(&rt->fib6_table->tb6_lock));
4675 iter = rcu_dereference_protected(fn->leaf,
4676 lockdep_is_held(&rt->fib6_table->tb6_lock));
4678 if (iter->fib6_metric == rt->fib6_metric &&
4679 rt6_qualify_for_ecmp(iter))
4681 iter = rcu_dereference_protected(iter->fib6_next,
4682 lockdep_is_held(&rt->fib6_table->tb6_lock));
4688 /* only called for fib entries with builtin fib6_nh */
4689 static bool rt6_is_dead(const struct fib6_info *rt)
4691 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4692 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4693 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4699 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4701 struct fib6_info *iter;
4704 if (!rt6_is_dead(rt))
4705 total += rt->fib6_nh->fib_nh_weight;
4707 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4708 if (!rt6_is_dead(iter))
4709 total += iter->fib6_nh->fib_nh_weight;
4715 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4717 int upper_bound = -1;
4719 if (!rt6_is_dead(rt)) {
4720 *weight += rt->fib6_nh->fib_nh_weight;
4721 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4724 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4727 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4729 struct fib6_info *iter;
4732 rt6_upper_bound_set(rt, &weight, total);
4734 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4735 rt6_upper_bound_set(iter, &weight, total);
4738 void rt6_multipath_rebalance(struct fib6_info *rt)
4740 struct fib6_info *first;
4743 /* In case the entire multipath route was marked for flushing,
4744 * then there is no need to rebalance upon the removal of every
4747 if (!rt->fib6_nsiblings || rt->should_flush)
4750 /* During lookup routes are evaluated in order, so we need to
4751 * make sure upper bounds are assigned from the first sibling
4754 first = rt6_multipath_first_sibling(rt);
4755 if (WARN_ON_ONCE(!first))
4758 total = rt6_multipath_total_weight(first);
4759 rt6_multipath_upper_bound_set(first, total);
4762 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4764 const struct arg_netdev_event *arg = p_arg;
4765 struct net *net = dev_net(arg->dev);
4767 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4768 rt->fib6_nh->fib_nh_dev == arg->dev) {
4769 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4770 fib6_update_sernum_upto_root(net, rt);
4771 rt6_multipath_rebalance(rt);
4777 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4779 struct arg_netdev_event arg = {
4782 .nh_flags = nh_flags,
4786 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4787 arg.nh_flags |= RTNH_F_LINKDOWN;
4789 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4792 /* only called for fib entries with inline fib6_nh */
4793 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4794 const struct net_device *dev)
4796 struct fib6_info *iter;
4798 if (rt->fib6_nh->fib_nh_dev == dev)
4800 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4801 if (iter->fib6_nh->fib_nh_dev == dev)
4807 static void rt6_multipath_flush(struct fib6_info *rt)
4809 struct fib6_info *iter;
4811 rt->should_flush = 1;
4812 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4813 iter->should_flush = 1;
4816 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4817 const struct net_device *down_dev)
4819 struct fib6_info *iter;
4820 unsigned int dead = 0;
4822 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4823 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4825 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4826 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4827 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4833 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4834 const struct net_device *dev,
4835 unsigned char nh_flags)
4837 struct fib6_info *iter;
4839 if (rt->fib6_nh->fib_nh_dev == dev)
4840 rt->fib6_nh->fib_nh_flags |= nh_flags;
4841 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4842 if (iter->fib6_nh->fib_nh_dev == dev)
4843 iter->fib6_nh->fib_nh_flags |= nh_flags;
4846 /* called with write lock held for table with rt */
4847 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4849 const struct arg_netdev_event *arg = p_arg;
4850 const struct net_device *dev = arg->dev;
4851 struct net *net = dev_net(dev);
4853 if (rt == net->ipv6.fib6_null_entry || rt->nh)
4856 switch (arg->event) {
4857 case NETDEV_UNREGISTER:
4858 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4860 if (rt->should_flush)
4862 if (!rt->fib6_nsiblings)
4863 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4864 if (rt6_multipath_uses_dev(rt, dev)) {
4867 count = rt6_multipath_dead_count(rt, dev);
4868 if (rt->fib6_nsiblings + 1 == count) {
4869 rt6_multipath_flush(rt);
4872 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4874 fib6_update_sernum(net, rt);
4875 rt6_multipath_rebalance(rt);
4879 if (rt->fib6_nh->fib_nh_dev != dev ||
4880 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4882 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4883 rt6_multipath_rebalance(rt);
4890 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4892 struct arg_netdev_event arg = {
4898 struct net *net = dev_net(dev);
4900 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4901 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4903 fib6_clean_all(net, fib6_ifdown, &arg);
4906 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4908 rt6_sync_down_dev(dev, event);
4909 rt6_uncached_list_flush_dev(dev);
4910 neigh_ifdown(&nd_tbl, dev);
4913 struct rt6_mtu_change_arg {
4914 struct net_device *dev;
4916 struct fib6_info *f6i;
4919 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4921 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4922 struct fib6_info *f6i = arg->f6i;
4924 /* For administrative MTU increase, there is no way to discover
4925 * IPv6 PMTU increase, so PMTU increase should be updated here.
4926 * Since RFC 1981 doesn't include administrative MTU increase
4927 * update PMTU increase is a MUST. (i.e. jumbo frame)
4929 if (nh->fib_nh_dev == arg->dev) {
4930 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4931 u32 mtu = f6i->fib6_pmtu;
4933 if (mtu >= arg->mtu ||
4934 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4935 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4937 spin_lock_bh(&rt6_exception_lock);
4938 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4939 spin_unlock_bh(&rt6_exception_lock);
4945 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4947 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4948 struct inet6_dev *idev;
4950 /* In IPv6 pmtu discovery is not optional,
4951 so that RTAX_MTU lock cannot disable it.
4952 We still use this lock to block changes
4953 caused by addrconf/ndisc.
4956 idev = __in6_dev_get(arg->dev);
4960 if (fib6_metric_locked(f6i, RTAX_MTU))
4965 /* fib6_nh_mtu_change only returns 0, so this is safe */
4966 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4970 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4973 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4975 struct rt6_mtu_change_arg arg = {
4980 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4983 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4984 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
4985 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
4986 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
4987 [RTA_OIF] = { .type = NLA_U32 },
4988 [RTA_IIF] = { .type = NLA_U32 },
4989 [RTA_PRIORITY] = { .type = NLA_U32 },
4990 [RTA_METRICS] = { .type = NLA_NESTED },
4991 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
4992 [RTA_PREF] = { .type = NLA_U8 },
4993 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4994 [RTA_ENCAP] = { .type = NLA_NESTED },
4995 [RTA_EXPIRES] = { .type = NLA_U32 },
4996 [RTA_UID] = { .type = NLA_U32 },
4997 [RTA_MARK] = { .type = NLA_U32 },
4998 [RTA_TABLE] = { .type = NLA_U32 },
4999 [RTA_IP_PROTO] = { .type = NLA_U8 },
5000 [RTA_SPORT] = { .type = NLA_U16 },
5001 [RTA_DPORT] = { .type = NLA_U16 },
5002 [RTA_NH_ID] = { .type = NLA_U32 },
5005 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
5006 struct fib6_config *cfg,
5007 struct netlink_ext_ack *extack)
5010 struct nlattr *tb[RTA_MAX+1];
5014 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5015 rtm_ipv6_policy, extack);
5020 rtm = nlmsg_data(nlh);
5023 NL_SET_ERR_MSG(extack,
5024 "Invalid dsfield (tos): option not available for IPv6");
5028 *cfg = (struct fib6_config){
5029 .fc_table = rtm->rtm_table,
5030 .fc_dst_len = rtm->rtm_dst_len,
5031 .fc_src_len = rtm->rtm_src_len,
5033 .fc_protocol = rtm->rtm_protocol,
5034 .fc_type = rtm->rtm_type,
5036 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
5037 .fc_nlinfo.nlh = nlh,
5038 .fc_nlinfo.nl_net = sock_net(skb->sk),
5041 if (rtm->rtm_type == RTN_UNREACHABLE ||
5042 rtm->rtm_type == RTN_BLACKHOLE ||
5043 rtm->rtm_type == RTN_PROHIBIT ||
5044 rtm->rtm_type == RTN_THROW)
5045 cfg->fc_flags |= RTF_REJECT;
5047 if (rtm->rtm_type == RTN_LOCAL)
5048 cfg->fc_flags |= RTF_LOCAL;
5050 if (rtm->rtm_flags & RTM_F_CLONED)
5051 cfg->fc_flags |= RTF_CACHE;
5053 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5055 if (tb[RTA_NH_ID]) {
5056 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
5057 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5058 NL_SET_ERR_MSG(extack,
5059 "Nexthop specification and nexthop id are mutually exclusive");
5062 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5065 if (tb[RTA_GATEWAY]) {
5066 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
5067 cfg->fc_flags |= RTF_GATEWAY;
5070 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5075 int plen = (rtm->rtm_dst_len + 7) >> 3;
5077 if (nla_len(tb[RTA_DST]) < plen)
5080 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
5084 int plen = (rtm->rtm_src_len + 7) >> 3;
5086 if (nla_len(tb[RTA_SRC]) < plen)
5089 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
5092 if (tb[RTA_PREFSRC])
5093 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
5096 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5098 if (tb[RTA_PRIORITY])
5099 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5101 if (tb[RTA_METRICS]) {
5102 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5103 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
5107 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5109 if (tb[RTA_MULTIPATH]) {
5110 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5111 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
5113 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
5114 cfg->fc_mp_len, extack);
5120 pref = nla_get_u8(tb[RTA_PREF]);
5121 if (pref != ICMPV6_ROUTER_PREF_LOW &&
5122 pref != ICMPV6_ROUTER_PREF_HIGH)
5123 pref = ICMPV6_ROUTER_PREF_MEDIUM;
5124 cfg->fc_flags |= RTF_PREF(pref);
5128 cfg->fc_encap = tb[RTA_ENCAP];
5130 if (tb[RTA_ENCAP_TYPE]) {
5131 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5133 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5138 if (tb[RTA_EXPIRES]) {
5139 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5141 if (addrconf_finite_timeout(timeout)) {
5142 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5143 cfg->fc_flags |= RTF_EXPIRES;
5153 struct fib6_info *fib6_info;
5154 struct fib6_config r_cfg;
5155 struct list_head next;
5158 static int ip6_route_info_append(struct net *net,
5159 struct list_head *rt6_nh_list,
5160 struct fib6_info *rt,
5161 struct fib6_config *r_cfg)
5166 list_for_each_entry(nh, rt6_nh_list, next) {
5167 /* check if fib6_info already exists */
5168 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5172 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5176 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5177 list_add_tail(&nh->next, rt6_nh_list);
5182 static void ip6_route_mpath_notify(struct fib6_info *rt,
5183 struct fib6_info *rt_last,
5184 struct nl_info *info,
5187 /* if this is an APPEND route, then rt points to the first route
5188 * inserted and rt_last points to last route inserted. Userspace
5189 * wants a consistent dump of the route which starts at the first
5190 * nexthop. Since sibling routes are always added at the end of
5191 * the list, find the first sibling of the last route appended
5193 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5194 rt = list_first_entry(&rt_last->fib6_siblings,
5200 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5203 static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5205 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5206 bool should_notify = false;
5207 struct fib6_info *leaf;
5208 struct fib6_node *fn;
5211 fn = rcu_dereference(rt->fib6_node);
5215 leaf = rcu_dereference(fn->leaf);
5220 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5221 rt6_qualify_for_ecmp(leaf)))
5222 should_notify = true;
5226 return should_notify;
5229 static int fib6_gw_from_attr(struct in6_addr *gw, struct nlattr *nla,
5230 struct netlink_ext_ack *extack)
5232 if (nla_len(nla) < sizeof(*gw)) {
5233 NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_GATEWAY");
5237 *gw = nla_get_in6_addr(nla);
5242 static int ip6_route_multipath_add(struct fib6_config *cfg,
5243 struct netlink_ext_ack *extack)
5245 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5246 struct nl_info *info = &cfg->fc_nlinfo;
5247 struct fib6_config r_cfg;
5248 struct rtnexthop *rtnh;
5249 struct fib6_info *rt;
5250 struct rt6_nh *err_nh;
5251 struct rt6_nh *nh, *nh_safe;
5257 int replace = (cfg->fc_nlinfo.nlh &&
5258 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5259 LIST_HEAD(rt6_nh_list);
5261 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5262 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5263 nlflags |= NLM_F_APPEND;
5265 remaining = cfg->fc_mp_len;
5266 rtnh = (struct rtnexthop *)cfg->fc_mp;
5268 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
5269 * fib6_info structs per nexthop
5271 while (rtnh_ok(rtnh, remaining)) {
5272 memcpy(&r_cfg, cfg, sizeof(*cfg));
5273 if (rtnh->rtnh_ifindex)
5274 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5276 attrlen = rtnh_attrlen(rtnh);
5278 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5280 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5282 err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5287 r_cfg.fc_flags |= RTF_GATEWAY;
5289 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5291 /* RTA_ENCAP_TYPE length checked in
5292 * lwtunnel_valid_encap_type_attr
5294 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5296 r_cfg.fc_encap_type = nla_get_u16(nla);
5299 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5300 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5306 if (!rt6_qualify_for_ecmp(rt)) {
5308 NL_SET_ERR_MSG(extack,
5309 "Device only routes can not be added for IPv6 using the multipath API.");
5310 fib6_info_release(rt);
5314 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5316 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5319 fib6_info_release(rt);
5323 rtnh = rtnh_next(rtnh, &remaining);
5326 if (list_empty(&rt6_nh_list)) {
5327 NL_SET_ERR_MSG(extack,
5328 "Invalid nexthop configuration - no valid nexthops");
5332 /* for add and replace send one notification with all nexthops.
5333 * Skip the notification in fib6_add_rt2node and send one with
5334 * the full route when done
5336 info->skip_notify = 1;
5338 /* For add and replace, send one notification with all nexthops. For
5339 * append, send one notification with all appended nexthops.
5341 info->skip_notify_kernel = 1;
5344 list_for_each_entry(nh, &rt6_nh_list, next) {
5345 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5349 NL_SET_ERR_MSG_MOD(extack,
5350 "multipath route replace failed (check consistency of installed routes)");
5354 /* save reference to last route successfully inserted */
5355 rt_last = nh->fib6_info;
5357 /* save reference to first route for notification */
5359 rt_notif = nh->fib6_info;
5361 /* Because each route is added like a single route we remove
5362 * these flags after the first nexthop: if there is a collision,
5363 * we have already failed to add the first nexthop:
5364 * fib6_add_rt2node() has rejected it; when replacing, old
5365 * nexthops have been replaced by first new, the rest should
5368 if (cfg->fc_nlinfo.nlh) {
5369 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5371 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5376 /* An in-kernel notification should only be sent in case the new
5377 * multipath route is added as the first route in the node, or if
5378 * it was appended to it. We pass 'rt_notif' since it is the first
5379 * sibling and might allow us to skip some checks in the replace case.
5381 if (ip6_route_mpath_should_notify(rt_notif)) {
5382 enum fib_event_type fib_event;
5384 if (rt_notif->fib6_nsiblings != nhn - 1)
5385 fib_event = FIB_EVENT_ENTRY_APPEND;
5387 fib_event = FIB_EVENT_ENTRY_REPLACE;
5389 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5390 fib_event, rt_notif,
5393 /* Delete all the siblings that were just added */
5399 /* success ... tell user about new route */
5400 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5404 /* send notification for routes that were added so that
5405 * the delete notifications sent by ip6_route_del are
5409 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5411 /* Delete routes that were already added */
5412 list_for_each_entry(nh, &rt6_nh_list, next) {
5415 ip6_route_del(&nh->r_cfg, extack);
5419 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5420 fib6_info_release(nh->fib6_info);
5421 list_del(&nh->next);
5428 static int ip6_route_multipath_del(struct fib6_config *cfg,
5429 struct netlink_ext_ack *extack)
5431 struct fib6_config r_cfg;
5432 struct rtnexthop *rtnh;
5438 remaining = cfg->fc_mp_len;
5439 rtnh = (struct rtnexthop *)cfg->fc_mp;
5441 /* Parse a Multipath Entry */
5442 while (rtnh_ok(rtnh, remaining)) {
5443 memcpy(&r_cfg, cfg, sizeof(*cfg));
5444 if (rtnh->rtnh_ifindex)
5445 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5447 attrlen = rtnh_attrlen(rtnh);
5449 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5451 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5453 err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5460 r_cfg.fc_flags |= RTF_GATEWAY;
5463 err = ip6_route_del(&r_cfg, extack);
5468 rtnh = rtnh_next(rtnh, &remaining);
5474 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5475 struct netlink_ext_ack *extack)
5477 struct fib6_config cfg;
5480 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5485 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5486 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5491 return ip6_route_multipath_del(&cfg, extack);
5493 cfg.fc_delete_all_nh = 1;
5494 return ip6_route_del(&cfg, extack);
5498 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5499 struct netlink_ext_ack *extack)
5501 struct fib6_config cfg;
5504 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5508 if (cfg.fc_metric == 0)
5509 cfg.fc_metric = IP6_RT_PRIO_USER;
5512 return ip6_route_multipath_add(&cfg, extack);
5514 return ip6_route_add(&cfg, GFP_KERNEL, extack);
5517 /* add the overhead of this fib6_nh to nexthop_len */
5518 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5520 int *nexthop_len = arg;
5522 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5523 + NLA_ALIGN(sizeof(struct rtnexthop))
5524 + nla_total_size(16); /* RTA_GATEWAY */
5526 if (nh->fib_nh_lws) {
5527 /* RTA_ENCAP_TYPE */
5528 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5530 *nexthop_len += nla_total_size(2);
5536 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5541 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5542 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5545 struct fib6_info *sibling, *next_sibling;
5546 struct fib6_nh *nh = f6i->fib6_nh;
5549 if (f6i->fib6_nsiblings) {
5550 rt6_nh_nlmsg_size(nh, &nexthop_len);
5552 list_for_each_entry_safe(sibling, next_sibling,
5553 &f6i->fib6_siblings, fib6_siblings) {
5554 rt6_nh_nlmsg_size(sibling->fib6_nh, &nexthop_len);
5557 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5560 return NLMSG_ALIGN(sizeof(struct rtmsg))
5561 + nla_total_size(16) /* RTA_SRC */
5562 + nla_total_size(16) /* RTA_DST */
5563 + nla_total_size(16) /* RTA_GATEWAY */
5564 + nla_total_size(16) /* RTA_PREFSRC */
5565 + nla_total_size(4) /* RTA_TABLE */
5566 + nla_total_size(4) /* RTA_IIF */
5567 + nla_total_size(4) /* RTA_OIF */
5568 + nla_total_size(4) /* RTA_PRIORITY */
5569 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5570 + nla_total_size(sizeof(struct rta_cacheinfo))
5571 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5572 + nla_total_size(1) /* RTA_PREF */
5576 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5577 unsigned char *flags)
5579 if (nexthop_is_multipath(nh)) {
5582 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5584 goto nla_put_failure;
5586 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5587 goto nla_put_failure;
5589 nla_nest_end(skb, mp);
5591 struct fib6_nh *fib6_nh;
5593 fib6_nh = nexthop_fib6_nh(nh);
5594 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5596 goto nla_put_failure;
5605 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5606 struct fib6_info *rt, struct dst_entry *dst,
5607 struct in6_addr *dest, struct in6_addr *src,
5608 int iif, int type, u32 portid, u32 seq,
5611 struct rt6_info *rt6 = (struct rt6_info *)dst;
5612 struct rt6key *rt6_dst, *rt6_src;
5613 u32 *pmetrics, table, rt6_flags;
5614 unsigned char nh_flags = 0;
5615 struct nlmsghdr *nlh;
5619 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5624 rt6_dst = &rt6->rt6i_dst;
5625 rt6_src = &rt6->rt6i_src;
5626 rt6_flags = rt6->rt6i_flags;
5628 rt6_dst = &rt->fib6_dst;
5629 rt6_src = &rt->fib6_src;
5630 rt6_flags = rt->fib6_flags;
5633 rtm = nlmsg_data(nlh);
5634 rtm->rtm_family = AF_INET6;
5635 rtm->rtm_dst_len = rt6_dst->plen;
5636 rtm->rtm_src_len = rt6_src->plen;
5639 table = rt->fib6_table->tb6_id;
5641 table = RT6_TABLE_UNSPEC;
5642 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5643 if (nla_put_u32(skb, RTA_TABLE, table))
5644 goto nla_put_failure;
5646 rtm->rtm_type = rt->fib6_type;
5648 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5649 rtm->rtm_protocol = rt->fib6_protocol;
5651 if (rt6_flags & RTF_CACHE)
5652 rtm->rtm_flags |= RTM_F_CLONED;
5655 if (nla_put_in6_addr(skb, RTA_DST, dest))
5656 goto nla_put_failure;
5657 rtm->rtm_dst_len = 128;
5658 } else if (rtm->rtm_dst_len)
5659 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5660 goto nla_put_failure;
5661 #ifdef CONFIG_IPV6_SUBTREES
5663 if (nla_put_in6_addr(skb, RTA_SRC, src))
5664 goto nla_put_failure;
5665 rtm->rtm_src_len = 128;
5666 } else if (rtm->rtm_src_len &&
5667 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5668 goto nla_put_failure;
5671 #ifdef CONFIG_IPV6_MROUTE
5672 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5673 int err = ip6mr_get_route(net, skb, rtm, portid);
5678 goto nla_put_failure;
5681 if (nla_put_u32(skb, RTA_IIF, iif))
5682 goto nla_put_failure;
5684 struct in6_addr saddr_buf;
5685 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
5686 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5687 goto nla_put_failure;
5690 if (rt->fib6_prefsrc.plen) {
5691 struct in6_addr saddr_buf;
5692 saddr_buf = rt->fib6_prefsrc.addr;
5693 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5694 goto nla_put_failure;
5697 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5698 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5699 goto nla_put_failure;
5701 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5702 goto nla_put_failure;
5704 /* For multipath routes, walk the siblings list and add
5705 * each as a nexthop within RTA_MULTIPATH.
5708 if (rt6_flags & RTF_GATEWAY &&
5709 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5710 goto nla_put_failure;
5712 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5713 goto nla_put_failure;
5715 if (dst->lwtstate &&
5716 lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5717 goto nla_put_failure;
5718 } else if (rt->fib6_nsiblings) {
5719 struct fib6_info *sibling, *next_sibling;
5722 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5724 goto nla_put_failure;
5726 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5727 rt->fib6_nh->fib_nh_weight, AF_INET6,
5729 goto nla_put_failure;
5731 list_for_each_entry_safe(sibling, next_sibling,
5732 &rt->fib6_siblings, fib6_siblings) {
5733 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5734 sibling->fib6_nh->fib_nh_weight,
5736 goto nla_put_failure;
5739 nla_nest_end(skb, mp);
5740 } else if (rt->nh) {
5741 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5742 goto nla_put_failure;
5744 if (nexthop_is_blackhole(rt->nh))
5745 rtm->rtm_type = RTN_BLACKHOLE;
5747 if (READ_ONCE(net->ipv4.sysctl_nexthop_compat_mode) &&
5748 rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5749 goto nla_put_failure;
5751 rtm->rtm_flags |= nh_flags;
5753 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5754 &nh_flags, false) < 0)
5755 goto nla_put_failure;
5757 rtm->rtm_flags |= nh_flags;
5760 if (rt6_flags & RTF_EXPIRES) {
5761 expires = dst ? dst->expires : rt->expires;
5766 if (READ_ONCE(rt->offload))
5767 rtm->rtm_flags |= RTM_F_OFFLOAD;
5768 if (READ_ONCE(rt->trap))
5769 rtm->rtm_flags |= RTM_F_TRAP;
5770 if (READ_ONCE(rt->offload_failed))
5771 rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5774 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5775 goto nla_put_failure;
5777 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5778 goto nla_put_failure;
5781 nlmsg_end(skb, nlh);
5785 nlmsg_cancel(skb, nlh);
5789 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5791 const struct net_device *dev = arg;
5793 if (nh->fib_nh_dev == dev)
5799 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5800 const struct net_device *dev)
5803 struct net_device *_dev = (struct net_device *)dev;
5805 return !!nexthop_for_each_fib6_nh(f6i->nh,
5806 fib6_info_nh_uses_dev,
5810 if (f6i->fib6_nh->fib_nh_dev == dev)
5813 if (f6i->fib6_nsiblings) {
5814 struct fib6_info *sibling, *next_sibling;
5816 list_for_each_entry_safe(sibling, next_sibling,
5817 &f6i->fib6_siblings, fib6_siblings) {
5818 if (sibling->fib6_nh->fib_nh_dev == dev)
5826 struct fib6_nh_exception_dump_walker {
5827 struct rt6_rtnl_dump_arg *dump;
5828 struct fib6_info *rt;
5834 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5836 struct fib6_nh_exception_dump_walker *w = arg;
5837 struct rt6_rtnl_dump_arg *dump = w->dump;
5838 struct rt6_exception_bucket *bucket;
5839 struct rt6_exception *rt6_ex;
5842 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5846 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5847 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5853 /* Expiration of entries doesn't bump sernum, insertion
5854 * does. Removal is triggered by insertion, so we can
5855 * rely on the fact that if entries change between two
5856 * partial dumps, this node is scanned again completely,
5857 * see rt6_insert_exception() and fib6_dump_table().
5859 * Count expired entries we go through as handled
5860 * entries that we'll skip next time, in case of partial
5861 * node dump. Otherwise, if entries expire meanwhile,
5862 * we'll skip the wrong amount.
5864 if (rt6_check_expired(rt6_ex->rt6i)) {
5869 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5870 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5872 NETLINK_CB(dump->cb->skb).portid,
5873 dump->cb->nlh->nlmsg_seq, w->flags);
5885 /* Return -1 if done with node, number of handled routes on partial dump */
5886 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5888 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5889 struct fib_dump_filter *filter = &arg->filter;
5890 unsigned int flags = NLM_F_MULTI;
5891 struct net *net = arg->net;
5894 if (rt == net->ipv6.fib6_null_entry)
5897 if ((filter->flags & RTM_F_PREFIX) &&
5898 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5899 /* success since this is not a prefix route */
5902 if (filter->filter_set &&
5903 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5904 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5905 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5909 if (filter->filter_set ||
5910 !filter->dump_routes || !filter->dump_exceptions) {
5911 flags |= NLM_F_DUMP_FILTERED;
5914 if (filter->dump_routes) {
5918 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5920 NETLINK_CB(arg->cb->skb).portid,
5921 arg->cb->nlh->nlmsg_seq, flags)) {
5928 if (filter->dump_exceptions) {
5929 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5938 err = nexthop_for_each_fib6_nh(rt->nh,
5939 rt6_nh_dump_exceptions,
5942 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5947 return count + w.count;
5953 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5954 const struct nlmsghdr *nlh,
5956 struct netlink_ext_ack *extack)
5961 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5962 NL_SET_ERR_MSG_MOD(extack,
5963 "Invalid header for get route request");
5967 if (!netlink_strict_get_check(skb))
5968 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5969 rtm_ipv6_policy, extack);
5971 rtm = nlmsg_data(nlh);
5972 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5973 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5974 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5976 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5979 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5980 NL_SET_ERR_MSG_MOD(extack,
5981 "Invalid flags for get route request");
5985 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5986 rtm_ipv6_policy, extack);
5990 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5991 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5992 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5996 for (i = 0; i <= RTA_MAX; i++) {
6012 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
6020 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6021 struct netlink_ext_ack *extack)
6023 struct net *net = sock_net(in_skb->sk);
6024 struct nlattr *tb[RTA_MAX+1];
6025 int err, iif = 0, oif = 0;
6026 struct fib6_info *from;
6027 struct dst_entry *dst;
6028 struct rt6_info *rt;
6029 struct sk_buff *skb;
6031 struct flowi6 fl6 = {};
6034 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
6039 rtm = nlmsg_data(nlh);
6040 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
6041 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
6044 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
6047 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
6051 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
6054 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
6058 iif = nla_get_u32(tb[RTA_IIF]);
6061 oif = nla_get_u32(tb[RTA_OIF]);
6064 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6067 fl6.flowi6_uid = make_kuid(current_user_ns(),
6068 nla_get_u32(tb[RTA_UID]));
6070 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6073 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6076 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6078 if (tb[RTA_IP_PROTO]) {
6079 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
6080 &fl6.flowi6_proto, AF_INET6,
6087 struct net_device *dev;
6092 dev = dev_get_by_index_rcu(net, iif);
6099 fl6.flowi6_iif = iif;
6101 if (!ipv6_addr_any(&fl6.saddr))
6102 flags |= RT6_LOOKUP_F_HAS_SADDR;
6104 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
6108 fl6.flowi6_oif = oif;
6110 dst = ip6_route_output(net, NULL, &fl6);
6114 rt = container_of(dst, struct rt6_info, dst);
6115 if (rt->dst.error) {
6116 err = rt->dst.error;
6121 if (rt == net->ipv6.ip6_null_entry) {
6122 err = rt->dst.error;
6127 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
6134 skb_dst_set(skb, &rt->dst);
6137 from = rcu_dereference(rt->from);
6140 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6142 NETLINK_CB(in_skb).portid,
6145 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6146 &fl6.saddr, iif, RTM_NEWROUTE,
6147 NETLINK_CB(in_skb).portid,
6159 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6164 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6165 unsigned int nlm_flags)
6167 struct sk_buff *skb;
6168 struct net *net = info->nl_net;
6173 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6175 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6179 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6180 event, info->portid, seq, nlm_flags);
6182 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6183 WARN_ON(err == -EMSGSIZE);
6187 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6188 info->nlh, gfp_any());
6192 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6195 void fib6_rt_update(struct net *net, struct fib6_info *rt,
6196 struct nl_info *info)
6198 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6199 struct sk_buff *skb;
6202 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6206 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6207 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6209 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6210 WARN_ON(err == -EMSGSIZE);
6214 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6215 info->nlh, gfp_any());
6219 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6222 void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6223 bool offload, bool trap, bool offload_failed)
6225 struct sk_buff *skb;
6228 if (READ_ONCE(f6i->offload) == offload &&
6229 READ_ONCE(f6i->trap) == trap &&
6230 READ_ONCE(f6i->offload_failed) == offload_failed)
6233 WRITE_ONCE(f6i->offload, offload);
6234 WRITE_ONCE(f6i->trap, trap);
6236 /* 2 means send notifications only if offload_failed was changed. */
6237 if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6238 READ_ONCE(f6i->offload_failed) == offload_failed)
6241 WRITE_ONCE(f6i->offload_failed, offload_failed);
6243 if (!rcu_access_pointer(f6i->fib6_node))
6244 /* The route was removed from the tree, do not send
6249 if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6252 skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6258 err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6261 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6262 WARN_ON(err == -EMSGSIZE);
6267 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6271 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6273 EXPORT_SYMBOL(fib6_info_hw_flags_set);
6275 static int ip6_route_dev_notify(struct notifier_block *this,
6276 unsigned long event, void *ptr)
6278 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6279 struct net *net = dev_net(dev);
6281 if (!(dev->flags & IFF_LOOPBACK))
6284 if (event == NETDEV_REGISTER) {
6285 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6286 net->ipv6.ip6_null_entry->dst.dev = dev;
6287 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6288 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6289 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6290 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6291 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6292 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6294 } else if (event == NETDEV_UNREGISTER &&
6295 dev->reg_state != NETREG_UNREGISTERED) {
6296 /* NETDEV_UNREGISTER could be fired for multiple times by
6297 * netdev_wait_allrefs(). Make sure we only call this once.
6299 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6300 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6301 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6302 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6313 #ifdef CONFIG_PROC_FS
6314 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6316 struct net *net = (struct net *)seq->private;
6317 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6318 net->ipv6.rt6_stats->fib_nodes,
6319 net->ipv6.rt6_stats->fib_route_nodes,
6320 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6321 net->ipv6.rt6_stats->fib_rt_entries,
6322 net->ipv6.rt6_stats->fib_rt_cache,
6323 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6324 net->ipv6.rt6_stats->fib_discarded_routes);
6328 #endif /* CONFIG_PROC_FS */
6330 #ifdef CONFIG_SYSCTL
6332 static int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6333 void *buffer, size_t *lenp, loff_t *ppos)
6341 net = (struct net *)ctl->extra1;
6342 delay = net->ipv6.sysctl.flush_delay;
6343 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6347 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6351 static struct ctl_table ipv6_route_table_template[] = {
6353 .procname = "max_size",
6354 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
6355 .maxlen = sizeof(int),
6357 .proc_handler = proc_dointvec,
6360 .procname = "gc_thresh",
6361 .data = &ip6_dst_ops_template.gc_thresh,
6362 .maxlen = sizeof(int),
6364 .proc_handler = proc_dointvec,
6367 .procname = "flush",
6368 .data = &init_net.ipv6.sysctl.flush_delay,
6369 .maxlen = sizeof(int),
6371 .proc_handler = ipv6_sysctl_rtcache_flush
6374 .procname = "gc_min_interval",
6375 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6376 .maxlen = sizeof(int),
6378 .proc_handler = proc_dointvec_jiffies,
6381 .procname = "gc_timeout",
6382 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6383 .maxlen = sizeof(int),
6385 .proc_handler = proc_dointvec_jiffies,
6388 .procname = "gc_interval",
6389 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
6390 .maxlen = sizeof(int),
6392 .proc_handler = proc_dointvec_jiffies,
6395 .procname = "gc_elasticity",
6396 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6397 .maxlen = sizeof(int),
6399 .proc_handler = proc_dointvec,
6402 .procname = "mtu_expires",
6403 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6404 .maxlen = sizeof(int),
6406 .proc_handler = proc_dointvec_jiffies,
6409 .procname = "min_adv_mss",
6410 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
6411 .maxlen = sizeof(int),
6413 .proc_handler = proc_dointvec,
6416 .procname = "gc_min_interval_ms",
6417 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6418 .maxlen = sizeof(int),
6420 .proc_handler = proc_dointvec_ms_jiffies,
6423 .procname = "skip_notify_on_dev_down",
6424 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6425 .maxlen = sizeof(u8),
6427 .proc_handler = proc_dou8vec_minmax,
6428 .extra1 = SYSCTL_ZERO,
6429 .extra2 = SYSCTL_ONE,
6434 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6436 struct ctl_table *table;
6438 table = kmemdup(ipv6_route_table_template,
6439 sizeof(ipv6_route_table_template),
6443 table[0].data = &net->ipv6.sysctl.ip6_rt_max_size;
6444 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6445 table[2].data = &net->ipv6.sysctl.flush_delay;
6446 table[2].extra1 = net;
6447 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6448 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6449 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6450 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6451 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6452 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6453 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6454 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6456 /* Don't export sysctls to unprivileged users */
6457 if (net->user_ns != &init_user_ns)
6458 table[1].procname = NULL;
6464 size_t ipv6_route_sysctl_table_size(struct net *net)
6466 /* Don't export sysctls to unprivileged users */
6467 if (net->user_ns != &init_user_ns)
6470 return ARRAY_SIZE(ipv6_route_table_template);
6474 static int __net_init ip6_route_net_init(struct net *net)
6478 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6479 sizeof(net->ipv6.ip6_dst_ops));
6481 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6482 goto out_ip6_dst_ops;
6484 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6485 if (!net->ipv6.fib6_null_entry)
6486 goto out_ip6_dst_entries;
6487 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6488 sizeof(*net->ipv6.fib6_null_entry));
6490 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6491 sizeof(*net->ipv6.ip6_null_entry),
6493 if (!net->ipv6.ip6_null_entry)
6494 goto out_fib6_null_entry;
6495 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6496 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6497 ip6_template_metrics, true);
6498 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->dst.rt_uncached);
6500 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6501 net->ipv6.fib6_has_custom_rules = false;
6502 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6503 sizeof(*net->ipv6.ip6_prohibit_entry),
6505 if (!net->ipv6.ip6_prohibit_entry)
6506 goto out_ip6_null_entry;
6507 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6508 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6509 ip6_template_metrics, true);
6510 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->dst.rt_uncached);
6512 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6513 sizeof(*net->ipv6.ip6_blk_hole_entry),
6515 if (!net->ipv6.ip6_blk_hole_entry)
6516 goto out_ip6_prohibit_entry;
6517 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6518 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6519 ip6_template_metrics, true);
6520 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->dst.rt_uncached);
6521 #ifdef CONFIG_IPV6_SUBTREES
6522 net->ipv6.fib6_routes_require_src = 0;
6526 net->ipv6.sysctl.flush_delay = 0;
6527 net->ipv6.sysctl.ip6_rt_max_size = INT_MAX;
6528 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6529 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6530 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6531 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6532 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6533 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6534 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6536 atomic_set(&net->ipv6.ip6_rt_gc_expire, 30*HZ);
6542 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6543 out_ip6_prohibit_entry:
6544 kfree(net->ipv6.ip6_prohibit_entry);
6546 kfree(net->ipv6.ip6_null_entry);
6548 out_fib6_null_entry:
6549 kfree(net->ipv6.fib6_null_entry);
6550 out_ip6_dst_entries:
6551 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6556 static void __net_exit ip6_route_net_exit(struct net *net)
6558 kfree(net->ipv6.fib6_null_entry);
6559 kfree(net->ipv6.ip6_null_entry);
6560 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6561 kfree(net->ipv6.ip6_prohibit_entry);
6562 kfree(net->ipv6.ip6_blk_hole_entry);
6564 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6567 static int __net_init ip6_route_net_init_late(struct net *net)
6569 #ifdef CONFIG_PROC_FS
6570 if (!proc_create_net("ipv6_route", 0, net->proc_net,
6571 &ipv6_route_seq_ops,
6572 sizeof(struct ipv6_route_iter)))
6575 if (!proc_create_net_single("rt6_stats", 0444, net->proc_net,
6576 rt6_stats_seq_show, NULL)) {
6577 remove_proc_entry("ipv6_route", net->proc_net);
6584 static void __net_exit ip6_route_net_exit_late(struct net *net)
6586 #ifdef CONFIG_PROC_FS
6587 remove_proc_entry("ipv6_route", net->proc_net);
6588 remove_proc_entry("rt6_stats", net->proc_net);
6592 static struct pernet_operations ip6_route_net_ops = {
6593 .init = ip6_route_net_init,
6594 .exit = ip6_route_net_exit,
6597 static int __net_init ipv6_inetpeer_init(struct net *net)
6599 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6603 inet_peer_base_init(bp);
6604 net->ipv6.peers = bp;
6608 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6610 struct inet_peer_base *bp = net->ipv6.peers;
6612 net->ipv6.peers = NULL;
6613 inetpeer_invalidate_tree(bp);
6617 static struct pernet_operations ipv6_inetpeer_ops = {
6618 .init = ipv6_inetpeer_init,
6619 .exit = ipv6_inetpeer_exit,
6622 static struct pernet_operations ip6_route_net_late_ops = {
6623 .init = ip6_route_net_init_late,
6624 .exit = ip6_route_net_exit_late,
6627 static struct notifier_block ip6_route_dev_notifier = {
6628 .notifier_call = ip6_route_dev_notify,
6629 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6632 void __init ip6_route_init_special_entries(void)
6634 /* Registering of the loopback is done before this portion of code,
6635 * the loopback reference in rt6_info will not be taken, do it
6636 * manually for init_net */
6637 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6638 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6639 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6640 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6641 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6642 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6643 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6644 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6648 #if IS_BUILTIN(CONFIG_IPV6)
6649 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6650 DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6652 BTF_ID_LIST(btf_fib6_info_id)
6653 BTF_ID(struct, fib6_info)
6655 static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6656 .seq_ops = &ipv6_route_seq_ops,
6657 .init_seq_private = bpf_iter_init_seq_net,
6658 .fini_seq_private = bpf_iter_fini_seq_net,
6659 .seq_priv_size = sizeof(struct ipv6_route_iter),
6662 static struct bpf_iter_reg ipv6_route_reg_info = {
6663 .target = "ipv6_route",
6664 .ctx_arg_info_size = 1,
6666 { offsetof(struct bpf_iter__ipv6_route, rt),
6667 PTR_TO_BTF_ID_OR_NULL },
6669 .seq_info = &ipv6_route_seq_info,
6672 static int __init bpf_iter_register(void)
6674 ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6675 return bpf_iter_reg_target(&ipv6_route_reg_info);
6678 static void bpf_iter_unregister(void)
6680 bpf_iter_unreg_target(&ipv6_route_reg_info);
6685 int __init ip6_route_init(void)
6691 ip6_dst_ops_template.kmem_cachep =
6692 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6693 SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL);
6694 if (!ip6_dst_ops_template.kmem_cachep)
6697 ret = dst_entries_init(&ip6_dst_blackhole_ops);
6699 goto out_kmem_cache;
6701 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6703 goto out_dst_entries;
6705 ret = register_pernet_subsys(&ip6_route_net_ops);
6707 goto out_register_inetpeer;
6709 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6713 goto out_register_subsys;
6719 ret = fib6_rules_init();
6723 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6725 goto fib6_rules_init;
6727 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6728 inet6_rtm_newroute, NULL, 0);
6730 goto out_register_late_subsys;
6732 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6733 inet6_rtm_delroute, NULL, 0);
6735 goto out_register_late_subsys;
6737 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6738 inet6_rtm_getroute, NULL,
6739 RTNL_FLAG_DOIT_UNLOCKED);
6741 goto out_register_late_subsys;
6743 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6745 goto out_register_late_subsys;
6747 #if IS_BUILTIN(CONFIG_IPV6)
6748 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6749 ret = bpf_iter_register();
6751 goto out_register_late_subsys;
6755 for_each_possible_cpu(cpu) {
6756 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6758 INIT_LIST_HEAD(&ul->head);
6759 INIT_LIST_HEAD(&ul->quarantine);
6760 spin_lock_init(&ul->lock);
6766 out_register_late_subsys:
6767 rtnl_unregister_all(PF_INET6);
6768 unregister_pernet_subsys(&ip6_route_net_late_ops);
6770 fib6_rules_cleanup();
6775 out_register_subsys:
6776 unregister_pernet_subsys(&ip6_route_net_ops);
6777 out_register_inetpeer:
6778 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6780 dst_entries_destroy(&ip6_dst_blackhole_ops);
6782 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6786 void ip6_route_cleanup(void)
6788 #if IS_BUILTIN(CONFIG_IPV6)
6789 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6790 bpf_iter_unregister();
6793 unregister_netdevice_notifier(&ip6_route_dev_notifier);
6794 unregister_pernet_subsys(&ip6_route_net_late_ops);
6795 fib6_rules_cleanup();
6798 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6799 unregister_pernet_subsys(&ip6_route_net_ops);
6800 dst_entries_destroy(&ip6_dst_blackhole_ops);
6801 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);