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) &&
648 neigh->updated + idev->cnf.rtr_probe_interval)) {
649 work = kmalloc(sizeof(*work), GFP_ATOMIC);
651 __neigh_set_probe_once(neigh);
653 write_unlock_bh(&neigh->lock);
654 } else if (time_after(jiffies, last_probe +
655 idev->cnf.rtr_probe_interval)) {
656 work = kmalloc(sizeof(*work), GFP_ATOMIC);
659 if (!work || cmpxchg(&fib6_nh->last_probe,
660 last_probe, jiffies) != last_probe) {
663 INIT_WORK(&work->work, rt6_probe_deferred);
664 work->target = *nh_gw;
665 netdev_hold(dev, &work->dev_tracker, GFP_ATOMIC);
667 schedule_work(&work->work);
674 static inline void rt6_probe(struct fib6_nh *fib6_nh)
680 * Default Router Selection (RFC 2461 6.3.6)
682 static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
684 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
685 struct neighbour *neigh;
688 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
689 &fib6_nh->fib_nh_gw6);
691 u8 nud_state = READ_ONCE(neigh->nud_state);
693 if (nud_state & NUD_VALID)
694 ret = RT6_NUD_SUCCEED;
695 #ifdef CONFIG_IPV6_ROUTER_PREF
696 else if (!(nud_state & NUD_FAILED))
697 ret = RT6_NUD_SUCCEED;
699 ret = RT6_NUD_FAIL_PROBE;
702 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
703 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
710 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
715 if (!oif || nh->fib_nh_dev->ifindex == oif)
718 if (!m && (strict & RT6_LOOKUP_F_IFACE))
719 return RT6_NUD_FAIL_HARD;
720 #ifdef CONFIG_IPV6_ROUTER_PREF
721 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
723 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
724 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
725 int n = rt6_check_neigh(nh);
732 static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
733 int oif, int strict, int *mpri, bool *do_rr)
735 bool match_do_rr = false;
739 if (nh->fib_nh_flags & RTNH_F_DEAD)
742 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
743 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
744 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
747 m = rt6_score_route(nh, fib6_flags, oif, strict);
748 if (m == RT6_NUD_FAIL_DO_RR) {
750 m = 0; /* lowest valid score */
751 } else if (m == RT6_NUD_FAIL_HARD) {
755 if (strict & RT6_LOOKUP_F_REACHABLE)
758 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
760 *do_rr = match_do_rr;
768 struct fib6_nh_frl_arg {
777 static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
779 struct fib6_nh_frl_arg *arg = _arg;
782 return find_match(nh, arg->flags, arg->oif, arg->strict,
783 arg->mpri, arg->do_rr);
786 static void __find_rr_leaf(struct fib6_info *f6i_start,
787 struct fib6_info *nomatch, u32 metric,
788 struct fib6_result *res, struct fib6_info **cont,
789 int oif, int strict, bool *do_rr, int *mpri)
791 struct fib6_info *f6i;
793 for (f6i = f6i_start;
794 f6i && f6i != nomatch;
795 f6i = rcu_dereference(f6i->fib6_next)) {
796 bool matched = false;
799 if (cont && f6i->fib6_metric != metric) {
804 if (fib6_check_expired(f6i))
807 if (unlikely(f6i->nh)) {
808 struct fib6_nh_frl_arg arg = {
809 .flags = f6i->fib6_flags,
816 if (nexthop_is_blackhole(f6i->nh)) {
817 res->fib6_flags = RTF_REJECT;
818 res->fib6_type = RTN_BLACKHOLE;
820 res->nh = nexthop_fib6_nh(f6i->nh);
823 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
830 if (find_match(nh, f6i->fib6_flags, oif, strict,
837 res->fib6_flags = f6i->fib6_flags;
838 res->fib6_type = f6i->fib6_type;
843 static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
844 struct fib6_info *rr_head, int oif, int strict,
845 bool *do_rr, struct fib6_result *res)
847 u32 metric = rr_head->fib6_metric;
848 struct fib6_info *cont = NULL;
851 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
852 oif, strict, do_rr, &mpri);
854 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
855 oif, strict, do_rr, &mpri);
857 if (res->f6i || !cont)
860 __find_rr_leaf(cont, NULL, metric, res, NULL,
861 oif, strict, do_rr, &mpri);
864 static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
865 struct fib6_result *res, int strict)
867 struct fib6_info *leaf = rcu_dereference(fn->leaf);
868 struct fib6_info *rt0;
872 /* make sure this function or its helpers sets f6i */
875 if (!leaf || leaf == net->ipv6.fib6_null_entry)
878 rt0 = rcu_dereference(fn->rr_ptr);
882 /* Double check to make sure fn is not an intermediate node
883 * and fn->leaf does not points to its child's leaf
884 * (This might happen if all routes under fn are deleted from
885 * the tree and fib6_repair_tree() is called on the node.)
887 key_plen = rt0->fib6_dst.plen;
888 #ifdef CONFIG_IPV6_SUBTREES
889 if (rt0->fib6_src.plen)
890 key_plen = rt0->fib6_src.plen;
892 if (fn->fn_bit != key_plen)
895 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
897 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
899 /* no entries matched; do round-robin */
900 if (!next || next->fib6_metric != rt0->fib6_metric)
904 spin_lock_bh(&leaf->fib6_table->tb6_lock);
905 /* make sure next is not being deleted from the tree */
907 rcu_assign_pointer(fn->rr_ptr, next);
908 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
914 res->f6i = net->ipv6.fib6_null_entry;
915 res->nh = res->f6i->fib6_nh;
916 res->fib6_flags = res->f6i->fib6_flags;
917 res->fib6_type = res->f6i->fib6_type;
921 static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
923 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
924 res->nh->fib_nh_gw_family;
927 #ifdef CONFIG_IPV6_ROUTE_INFO
928 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
929 const struct in6_addr *gwaddr)
931 struct net *net = dev_net(dev);
932 struct route_info *rinfo = (struct route_info *) opt;
933 struct in6_addr prefix_buf, *prefix;
934 struct fib6_table *table;
936 unsigned long lifetime;
937 struct fib6_info *rt;
939 if (len < sizeof(struct route_info)) {
943 /* Sanity check for prefix_len and length */
944 if (rinfo->length > 3) {
946 } else if (rinfo->prefix_len > 128) {
948 } else if (rinfo->prefix_len > 64) {
949 if (rinfo->length < 2) {
952 } else if (rinfo->prefix_len > 0) {
953 if (rinfo->length < 1) {
958 pref = rinfo->route_pref;
959 if (pref == ICMPV6_ROUTER_PREF_INVALID)
962 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
964 if (rinfo->length == 3)
965 prefix = (struct in6_addr *)rinfo->prefix;
967 /* this function is safe */
968 ipv6_addr_prefix(&prefix_buf,
969 (struct in6_addr *)rinfo->prefix,
971 prefix = &prefix_buf;
974 if (rinfo->prefix_len == 0)
975 rt = rt6_get_dflt_router(net, gwaddr, dev);
977 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
980 if (rt && !lifetime) {
981 ip6_del_rt(net, rt, false);
986 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
989 rt->fib6_flags = RTF_ROUTEINFO |
990 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
993 table = rt->fib6_table;
994 spin_lock_bh(&table->tb6_lock);
996 if (!addrconf_finite_timeout(lifetime)) {
997 fib6_clean_expires(rt);
998 fib6_remove_gc_list(rt);
1000 fib6_set_expires(rt, jiffies + HZ * lifetime);
1001 fib6_add_gc_list(rt);
1004 spin_unlock_bh(&table->tb6_lock);
1006 fib6_info_release(rt);
1013 * Misc support functions
1016 /* called with rcu_lock held */
1017 static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1019 struct net_device *dev = res->nh->fib_nh_dev;
1021 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1022 /* for copies of local routes, dst->dev needs to be the
1023 * device if it is a master device, the master device if
1024 * device is enslaved, and the loopback as the default
1026 if (netif_is_l3_slave(dev) &&
1027 !rt6_need_strict(&res->f6i->fib6_dst.addr))
1028 dev = l3mdev_master_dev_rcu(dev);
1029 else if (!netif_is_l3_master(dev))
1030 dev = dev_net(dev)->loopback_dev;
1031 /* last case is netif_is_l3_master(dev) is true in which
1032 * case we want dev returned to be dev
1039 static const int fib6_prop[RTN_MAX + 1] = {
1043 [RTN_BROADCAST] = 0,
1045 [RTN_MULTICAST] = 0,
1046 [RTN_BLACKHOLE] = -EINVAL,
1047 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1048 [RTN_PROHIBIT] = -EACCES,
1049 [RTN_THROW] = -EAGAIN,
1050 [RTN_NAT] = -EINVAL,
1051 [RTN_XRESOLVE] = -EINVAL,
1054 static int ip6_rt_type_to_error(u8 fib6_type)
1056 return fib6_prop[fib6_type];
1059 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1061 unsigned short flags = 0;
1063 if (rt->dst_nocount)
1064 flags |= DST_NOCOUNT;
1065 if (rt->dst_nopolicy)
1066 flags |= DST_NOPOLICY;
1071 static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1073 rt->dst.error = ip6_rt_type_to_error(fib6_type);
1075 switch (fib6_type) {
1077 rt->dst.output = dst_discard_out;
1078 rt->dst.input = dst_discard;
1081 rt->dst.output = ip6_pkt_prohibit_out;
1082 rt->dst.input = ip6_pkt_prohibit;
1085 case RTN_UNREACHABLE:
1087 rt->dst.output = ip6_pkt_discard_out;
1088 rt->dst.input = ip6_pkt_discard;
1093 static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1095 struct fib6_info *f6i = res->f6i;
1097 if (res->fib6_flags & RTF_REJECT) {
1098 ip6_rt_init_dst_reject(rt, res->fib6_type);
1103 rt->dst.output = ip6_output;
1105 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1106 rt->dst.input = ip6_input;
1107 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1108 rt->dst.input = ip6_mc_input;
1110 rt->dst.input = ip6_forward;
1113 if (res->nh->fib_nh_lws) {
1114 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1115 lwtunnel_set_redirect(&rt->dst);
1118 rt->dst.lastuse = jiffies;
1121 /* Caller must already hold reference to @from */
1122 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1124 rt->rt6i_flags &= ~RTF_EXPIRES;
1125 rcu_assign_pointer(rt->from, from);
1126 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1129 /* Caller must already hold reference to f6i in result */
1130 static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1132 const struct fib6_nh *nh = res->nh;
1133 const struct net_device *dev = nh->fib_nh_dev;
1134 struct fib6_info *f6i = res->f6i;
1136 ip6_rt_init_dst(rt, res);
1138 rt->rt6i_dst = f6i->fib6_dst;
1139 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1140 rt->rt6i_flags = res->fib6_flags;
1141 if (nh->fib_nh_gw_family) {
1142 rt->rt6i_gateway = nh->fib_nh_gw6;
1143 rt->rt6i_flags |= RTF_GATEWAY;
1145 rt6_set_from(rt, f6i);
1146 #ifdef CONFIG_IPV6_SUBTREES
1147 rt->rt6i_src = f6i->fib6_src;
1151 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1152 struct in6_addr *saddr)
1154 struct fib6_node *pn, *sn;
1156 if (fn->fn_flags & RTN_TL_ROOT)
1158 pn = rcu_dereference(fn->parent);
1159 sn = FIB6_SUBTREE(pn);
1161 fn = fib6_node_lookup(sn, NULL, saddr);
1164 if (fn->fn_flags & RTN_RTINFO)
1169 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1171 struct rt6_info *rt = *prt;
1173 if (dst_hold_safe(&rt->dst))
1176 rt = net->ipv6.ip6_null_entry;
1185 /* called with rcu_lock held */
1186 static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1188 struct net_device *dev = res->nh->fib_nh_dev;
1189 struct fib6_info *f6i = res->f6i;
1190 unsigned short flags;
1191 struct rt6_info *nrt;
1193 if (!fib6_info_hold_safe(f6i))
1196 flags = fib6_info_dst_flags(f6i);
1197 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1199 fib6_info_release(f6i);
1203 ip6_rt_copy_init(nrt, res);
1207 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1208 dst_hold(&nrt->dst);
1212 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
1213 struct fib6_table *table,
1215 const struct sk_buff *skb,
1218 struct fib6_result res = {};
1219 struct fib6_node *fn;
1220 struct rt6_info *rt;
1223 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1225 res.f6i = rcu_dereference(fn->leaf);
1227 res.f6i = net->ipv6.fib6_null_entry;
1229 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1232 if (res.f6i == net->ipv6.fib6_null_entry) {
1233 fn = fib6_backtrack(fn, &fl6->saddr);
1237 rt = net->ipv6.ip6_null_entry;
1240 } else if (res.fib6_flags & RTF_REJECT) {
1244 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1245 fl6->flowi6_oif != 0, skb, flags);
1247 /* Search through exception table */
1248 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1250 if (ip6_hold_safe(net, &rt))
1251 dst_use_noref(&rt->dst, jiffies);
1254 rt = ip6_create_rt_rcu(&res);
1258 trace_fib6_table_lookup(net, &res, table, fl6);
1265 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1266 const struct sk_buff *skb, int flags)
1268 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1270 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1272 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1273 const struct in6_addr *saddr, int oif,
1274 const struct sk_buff *skb, int strict)
1276 struct flowi6 fl6 = {
1280 struct dst_entry *dst;
1281 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1284 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1285 flags |= RT6_LOOKUP_F_HAS_SADDR;
1288 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1289 if (dst->error == 0)
1290 return (struct rt6_info *) dst;
1296 EXPORT_SYMBOL(rt6_lookup);
1298 /* ip6_ins_rt is called with FREE table->tb6_lock.
1299 * It takes new route entry, the addition fails by any reason the
1300 * route is released.
1301 * Caller must hold dst before calling it.
1304 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1305 struct netlink_ext_ack *extack)
1308 struct fib6_table *table;
1310 table = rt->fib6_table;
1311 spin_lock_bh(&table->tb6_lock);
1312 err = fib6_add(&table->tb6_root, rt, info, extack);
1313 spin_unlock_bh(&table->tb6_lock);
1318 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1320 struct nl_info info = { .nl_net = net, };
1322 return __ip6_ins_rt(rt, &info, NULL);
1325 static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1326 const struct in6_addr *daddr,
1327 const struct in6_addr *saddr)
1329 struct fib6_info *f6i = res->f6i;
1330 struct net_device *dev;
1331 struct rt6_info *rt;
1337 if (!fib6_info_hold_safe(f6i))
1340 dev = ip6_rt_get_dev_rcu(res);
1341 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1343 fib6_info_release(f6i);
1347 ip6_rt_copy_init(rt, res);
1348 rt->rt6i_flags |= RTF_CACHE;
1349 rt->rt6i_dst.addr = *daddr;
1350 rt->rt6i_dst.plen = 128;
1352 if (!rt6_is_gw_or_nonexthop(res)) {
1353 if (f6i->fib6_dst.plen != 128 &&
1354 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1355 rt->rt6i_flags |= RTF_ANYCAST;
1356 #ifdef CONFIG_IPV6_SUBTREES
1357 if (rt->rt6i_src.plen && saddr) {
1358 rt->rt6i_src.addr = *saddr;
1359 rt->rt6i_src.plen = 128;
1367 static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1369 struct fib6_info *f6i = res->f6i;
1370 unsigned short flags = fib6_info_dst_flags(f6i);
1371 struct net_device *dev;
1372 struct rt6_info *pcpu_rt;
1374 if (!fib6_info_hold_safe(f6i))
1378 dev = ip6_rt_get_dev_rcu(res);
1379 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
1382 fib6_info_release(f6i);
1385 ip6_rt_copy_init(pcpu_rt, res);
1386 pcpu_rt->rt6i_flags |= RTF_PCPU;
1389 pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1394 static bool rt6_is_valid(const struct rt6_info *rt6)
1396 return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1399 /* It should be called with rcu_read_lock() acquired */
1400 static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1402 struct rt6_info *pcpu_rt;
1404 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1406 if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1407 struct rt6_info *prev, **p;
1409 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1410 prev = xchg(p, NULL);
1412 dst_dev_put(&prev->dst);
1413 dst_release(&prev->dst);
1422 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1423 const struct fib6_result *res)
1425 struct rt6_info *pcpu_rt, *prev, **p;
1427 pcpu_rt = ip6_rt_pcpu_alloc(res);
1431 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1432 prev = cmpxchg(p, NULL, pcpu_rt);
1435 if (res->f6i->fib6_destroying) {
1436 struct fib6_info *from;
1438 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1439 fib6_info_release(from);
1445 /* exception hash table implementation
1447 static DEFINE_SPINLOCK(rt6_exception_lock);
1449 /* Remove rt6_ex from hash table and free the memory
1450 * Caller must hold rt6_exception_lock
1452 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1453 struct rt6_exception *rt6_ex)
1455 struct fib6_info *from;
1458 if (!bucket || !rt6_ex)
1461 net = dev_net(rt6_ex->rt6i->dst.dev);
1462 net->ipv6.rt6_stats->fib_rt_cache--;
1464 /* purge completely the exception to allow releasing the held resources:
1465 * some [sk] cache may keep the dst around for unlimited time
1467 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
1468 fib6_info_release(from);
1469 dst_dev_put(&rt6_ex->rt6i->dst);
1471 hlist_del_rcu(&rt6_ex->hlist);
1472 dst_release(&rt6_ex->rt6i->dst);
1473 kfree_rcu(rt6_ex, rcu);
1474 WARN_ON_ONCE(!bucket->depth);
1478 /* Remove oldest rt6_ex in bucket and free the memory
1479 * Caller must hold rt6_exception_lock
1481 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1483 struct rt6_exception *rt6_ex, *oldest = NULL;
1488 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1489 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1492 rt6_remove_exception(bucket, oldest);
1495 static u32 rt6_exception_hash(const struct in6_addr *dst,
1496 const struct in6_addr *src)
1498 static siphash_aligned_key_t rt6_exception_key;
1500 struct in6_addr dst;
1501 struct in6_addr src;
1502 } __aligned(SIPHASH_ALIGNMENT) combined = {
1507 net_get_random_once(&rt6_exception_key, sizeof(rt6_exception_key));
1509 #ifdef CONFIG_IPV6_SUBTREES
1511 combined.src = *src;
1513 val = siphash(&combined, sizeof(combined), &rt6_exception_key);
1515 return hash_64(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1518 /* Helper function to find the cached rt in the hash table
1519 * and update bucket pointer to point to the bucket for this
1520 * (daddr, saddr) pair
1521 * Caller must hold rt6_exception_lock
1523 static struct rt6_exception *
1524 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1525 const struct in6_addr *daddr,
1526 const struct in6_addr *saddr)
1528 struct rt6_exception *rt6_ex;
1531 if (!(*bucket) || !daddr)
1534 hval = rt6_exception_hash(daddr, saddr);
1537 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1538 struct rt6_info *rt6 = rt6_ex->rt6i;
1539 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1541 #ifdef CONFIG_IPV6_SUBTREES
1542 if (matched && saddr)
1543 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1551 /* Helper function to find the cached rt in the hash table
1552 * and update bucket pointer to point to the bucket for this
1553 * (daddr, saddr) pair
1554 * Caller must hold rcu_read_lock()
1556 static struct rt6_exception *
1557 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1558 const struct in6_addr *daddr,
1559 const struct in6_addr *saddr)
1561 struct rt6_exception *rt6_ex;
1564 WARN_ON_ONCE(!rcu_read_lock_held());
1566 if (!(*bucket) || !daddr)
1569 hval = rt6_exception_hash(daddr, saddr);
1572 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1573 struct rt6_info *rt6 = rt6_ex->rt6i;
1574 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1576 #ifdef CONFIG_IPV6_SUBTREES
1577 if (matched && saddr)
1578 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1586 static unsigned int fib6_mtu(const struct fib6_result *res)
1588 const struct fib6_nh *nh = res->nh;
1591 if (res->f6i->fib6_pmtu) {
1592 mtu = res->f6i->fib6_pmtu;
1594 struct net_device *dev = nh->fib_nh_dev;
1595 struct inet6_dev *idev;
1598 idev = __in6_dev_get(dev);
1599 mtu = idev->cnf.mtu6;
1603 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1605 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1608 #define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1610 /* used when the flushed bit is not relevant, only access to the bucket
1611 * (ie., all bucket users except rt6_insert_exception);
1613 * called under rcu lock; sometimes called with rt6_exception_lock held
1616 struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1619 struct rt6_exception_bucket *bucket;
1622 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1623 lockdep_is_held(lock));
1625 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1627 /* remove bucket flushed bit if set */
1629 unsigned long p = (unsigned long)bucket;
1631 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1632 bucket = (struct rt6_exception_bucket *)p;
1638 static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1640 unsigned long p = (unsigned long)bucket;
1642 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1645 /* called with rt6_exception_lock held */
1646 static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1649 struct rt6_exception_bucket *bucket;
1652 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1653 lockdep_is_held(lock));
1655 p = (unsigned long)bucket;
1656 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1657 bucket = (struct rt6_exception_bucket *)p;
1658 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1661 static int rt6_insert_exception(struct rt6_info *nrt,
1662 const struct fib6_result *res)
1664 struct net *net = dev_net(nrt->dst.dev);
1665 struct rt6_exception_bucket *bucket;
1666 struct fib6_info *f6i = res->f6i;
1667 struct in6_addr *src_key = NULL;
1668 struct rt6_exception *rt6_ex;
1669 struct fib6_nh *nh = res->nh;
1673 spin_lock_bh(&rt6_exception_lock);
1675 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1676 lockdep_is_held(&rt6_exception_lock));
1678 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1684 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1685 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1690 #ifdef CONFIG_IPV6_SUBTREES
1691 /* fib6_src.plen != 0 indicates f6i is in subtree
1692 * and exception table is indexed by a hash of
1693 * both fib6_dst and fib6_src.
1694 * Otherwise, the exception table is indexed by
1695 * a hash of only fib6_dst.
1697 if (f6i->fib6_src.plen)
1698 src_key = &nrt->rt6i_src.addr;
1700 /* rt6_mtu_change() might lower mtu on f6i.
1701 * Only insert this exception route if its mtu
1702 * is less than f6i's mtu value.
1704 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1709 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1712 rt6_remove_exception(bucket, rt6_ex);
1714 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1720 rt6_ex->stamp = jiffies;
1721 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1723 net->ipv6.rt6_stats->fib_rt_cache++;
1725 /* Randomize max depth to avoid some side channels attacks. */
1726 max_depth = FIB6_MAX_DEPTH + get_random_u32_below(FIB6_MAX_DEPTH);
1727 while (bucket->depth > max_depth)
1728 rt6_exception_remove_oldest(bucket);
1731 spin_unlock_bh(&rt6_exception_lock);
1733 /* Update fn->fn_sernum to invalidate all cached dst */
1735 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1736 fib6_update_sernum(net, f6i);
1737 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1738 fib6_force_start_gc(net);
1744 static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1746 struct rt6_exception_bucket *bucket;
1747 struct rt6_exception *rt6_ex;
1748 struct hlist_node *tmp;
1751 spin_lock_bh(&rt6_exception_lock);
1753 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1757 /* Prevent rt6_insert_exception() to recreate the bucket list */
1759 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1761 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1762 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1764 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1765 rt6_remove_exception(bucket, rt6_ex);
1767 WARN_ON_ONCE(!from && bucket->depth);
1771 spin_unlock_bh(&rt6_exception_lock);
1774 static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1776 struct fib6_info *f6i = arg;
1778 fib6_nh_flush_exceptions(nh, f6i);
1783 void rt6_flush_exceptions(struct fib6_info *f6i)
1786 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1789 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1792 /* Find cached rt in the hash table inside passed in rt
1793 * Caller has to hold rcu_read_lock()
1795 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1796 const struct in6_addr *daddr,
1797 const struct in6_addr *saddr)
1799 const struct in6_addr *src_key = NULL;
1800 struct rt6_exception_bucket *bucket;
1801 struct rt6_exception *rt6_ex;
1802 struct rt6_info *ret = NULL;
1804 #ifdef CONFIG_IPV6_SUBTREES
1805 /* fib6i_src.plen != 0 indicates f6i is in subtree
1806 * and exception table is indexed by a hash of
1807 * both fib6_dst and fib6_src.
1808 * However, the src addr used to create the hash
1809 * might not be exactly the passed in saddr which
1810 * is a /128 addr from the flow.
1811 * So we need to use f6i->fib6_src to redo lookup
1812 * if the passed in saddr does not find anything.
1813 * (See the logic in ip6_rt_cache_alloc() on how
1814 * rt->rt6i_src is updated.)
1816 if (res->f6i->fib6_src.plen)
1820 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1821 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1823 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1826 #ifdef CONFIG_IPV6_SUBTREES
1827 /* Use fib6_src as src_key and redo lookup */
1828 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1829 src_key = &res->f6i->fib6_src.addr;
1837 /* Remove the passed in cached rt from the hash table that contains it */
1838 static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1839 const struct rt6_info *rt)
1841 const struct in6_addr *src_key = NULL;
1842 struct rt6_exception_bucket *bucket;
1843 struct rt6_exception *rt6_ex;
1846 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1849 spin_lock_bh(&rt6_exception_lock);
1850 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1852 #ifdef CONFIG_IPV6_SUBTREES
1853 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1854 * and exception table is indexed by a hash of
1855 * both rt6i_dst and rt6i_src.
1856 * Otherwise, the exception table is indexed by
1857 * a hash of only rt6i_dst.
1860 src_key = &rt->rt6i_src.addr;
1862 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1866 rt6_remove_exception(bucket, rt6_ex);
1872 spin_unlock_bh(&rt6_exception_lock);
1876 struct fib6_nh_excptn_arg {
1877 struct rt6_info *rt;
1881 static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1883 struct fib6_nh_excptn_arg *arg = _arg;
1886 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1893 static int rt6_remove_exception_rt(struct rt6_info *rt)
1895 struct fib6_info *from;
1897 from = rcu_dereference(rt->from);
1898 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1902 struct fib6_nh_excptn_arg arg = {
1904 .plen = from->fib6_src.plen
1908 /* rc = 1 means an entry was found */
1909 rc = nexthop_for_each_fib6_nh(from->nh,
1910 rt6_nh_remove_exception_rt,
1912 return rc ? 0 : -ENOENT;
1915 return fib6_nh_remove_exception(from->fib6_nh,
1916 from->fib6_src.plen, rt);
1919 /* Find rt6_ex which contains the passed in rt cache and
1922 static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1923 const struct rt6_info *rt)
1925 const struct in6_addr *src_key = NULL;
1926 struct rt6_exception_bucket *bucket;
1927 struct rt6_exception *rt6_ex;
1929 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1930 #ifdef CONFIG_IPV6_SUBTREES
1931 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1932 * and exception table is indexed by a hash of
1933 * both rt6i_dst and rt6i_src.
1934 * Otherwise, the exception table is indexed by
1935 * a hash of only rt6i_dst.
1938 src_key = &rt->rt6i_src.addr;
1940 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1942 rt6_ex->stamp = jiffies;
1945 struct fib6_nh_match_arg {
1946 const struct net_device *dev;
1947 const struct in6_addr *gw;
1948 struct fib6_nh *match;
1951 /* determine if fib6_nh has given device and gateway */
1952 static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1954 struct fib6_nh_match_arg *arg = _arg;
1956 if (arg->dev != nh->fib_nh_dev ||
1957 (arg->gw && !nh->fib_nh_gw_family) ||
1958 (!arg->gw && nh->fib_nh_gw_family) ||
1959 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1964 /* found a match, break the loop */
1968 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1970 struct fib6_info *from;
1971 struct fib6_nh *fib6_nh;
1975 from = rcu_dereference(rt->from);
1976 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1980 struct fib6_nh_match_arg arg = {
1982 .gw = &rt->rt6i_gateway,
1985 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1989 fib6_nh = arg.match;
1991 fib6_nh = from->fib6_nh;
1993 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1998 static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1999 struct rt6_info *rt, int mtu)
2001 /* If the new MTU is lower than the route PMTU, this new MTU will be the
2002 * lowest MTU in the path: always allow updating the route PMTU to
2003 * reflect PMTU decreases.
2005 * If the new MTU is higher, and the route PMTU is equal to the local
2006 * MTU, this means the old MTU is the lowest in the path, so allow
2007 * updating it: if other nodes now have lower MTUs, PMTU discovery will
2011 if (dst_mtu(&rt->dst) >= mtu)
2014 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
2020 static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
2021 const struct fib6_nh *nh, int mtu)
2023 struct rt6_exception_bucket *bucket;
2024 struct rt6_exception *rt6_ex;
2027 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2031 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2032 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2033 struct rt6_info *entry = rt6_ex->rt6i;
2035 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2036 * route), the metrics of its rt->from have already
2039 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2040 rt6_mtu_change_route_allowed(idev, entry, mtu))
2041 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2047 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2049 static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2050 const struct in6_addr *gateway)
2052 struct rt6_exception_bucket *bucket;
2053 struct rt6_exception *rt6_ex;
2054 struct hlist_node *tmp;
2057 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2060 spin_lock_bh(&rt6_exception_lock);
2061 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2063 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2064 hlist_for_each_entry_safe(rt6_ex, tmp,
2065 &bucket->chain, hlist) {
2066 struct rt6_info *entry = rt6_ex->rt6i;
2068 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2069 RTF_CACHE_GATEWAY &&
2070 ipv6_addr_equal(gateway,
2071 &entry->rt6i_gateway)) {
2072 rt6_remove_exception(bucket, rt6_ex);
2079 spin_unlock_bh(&rt6_exception_lock);
2082 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2083 struct rt6_exception *rt6_ex,
2084 struct fib6_gc_args *gc_args,
2087 struct rt6_info *rt = rt6_ex->rt6i;
2089 /* we are pruning and obsoleting aged-out and non gateway exceptions
2090 * even if others have still references to them, so that on next
2091 * dst_check() such references can be dropped.
2092 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2093 * expired, independently from their aging, as per RFC 8201 section 4
2095 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2096 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2097 pr_debug("aging clone %p\n", rt);
2098 rt6_remove_exception(bucket, rt6_ex);
2101 } else if (time_after(jiffies, rt->dst.expires)) {
2102 pr_debug("purging expired route %p\n", rt);
2103 rt6_remove_exception(bucket, rt6_ex);
2107 if (rt->rt6i_flags & RTF_GATEWAY) {
2108 struct neighbour *neigh;
2110 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2112 if (!(neigh && (neigh->flags & NTF_ROUTER))) {
2113 pr_debug("purging route %p via non-router but gateway\n",
2115 rt6_remove_exception(bucket, rt6_ex);
2123 static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2124 struct fib6_gc_args *gc_args,
2127 struct rt6_exception_bucket *bucket;
2128 struct rt6_exception *rt6_ex;
2129 struct hlist_node *tmp;
2132 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2136 spin_lock(&rt6_exception_lock);
2137 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2139 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2140 hlist_for_each_entry_safe(rt6_ex, tmp,
2141 &bucket->chain, hlist) {
2142 rt6_age_examine_exception(bucket, rt6_ex,
2148 spin_unlock(&rt6_exception_lock);
2149 rcu_read_unlock_bh();
2152 struct fib6_nh_age_excptn_arg {
2153 struct fib6_gc_args *gc_args;
2157 static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2159 struct fib6_nh_age_excptn_arg *arg = _arg;
2161 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2165 void rt6_age_exceptions(struct fib6_info *f6i,
2166 struct fib6_gc_args *gc_args,
2170 struct fib6_nh_age_excptn_arg arg = {
2175 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2178 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2182 /* must be called with rcu lock held */
2183 int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2184 struct flowi6 *fl6, struct fib6_result *res, int strict)
2186 struct fib6_node *fn, *saved_fn;
2188 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2192 rt6_select(net, fn, oif, res, strict);
2193 if (res->f6i == net->ipv6.fib6_null_entry) {
2194 fn = fib6_backtrack(fn, &fl6->saddr);
2196 goto redo_rt6_select;
2197 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2198 /* also consider unreachable route */
2199 strict &= ~RT6_LOOKUP_F_REACHABLE;
2201 goto redo_rt6_select;
2205 trace_fib6_table_lookup(net, res, table, fl6);
2210 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2211 int oif, struct flowi6 *fl6,
2212 const struct sk_buff *skb, int flags)
2214 struct fib6_result res = {};
2215 struct rt6_info *rt = NULL;
2218 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2219 !rcu_read_lock_held());
2221 strict |= flags & RT6_LOOKUP_F_IFACE;
2222 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2223 if (net->ipv6.devconf_all->forwarding == 0)
2224 strict |= RT6_LOOKUP_F_REACHABLE;
2228 fib6_table_lookup(net, table, oif, fl6, &res, strict);
2229 if (res.f6i == net->ipv6.fib6_null_entry)
2232 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2234 /*Search through exception table */
2235 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2238 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2239 !res.nh->fib_nh_gw_family)) {
2240 /* Create a RTF_CACHE clone which will not be
2241 * owned by the fib6 tree. It is for the special case where
2242 * the daddr in the skb during the neighbor look-up is different
2243 * from the fl6->daddr used to look-up route here.
2245 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2248 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2249 * As rt6_uncached_list_add() does not consume refcnt,
2250 * this refcnt is always returned to the caller even
2251 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2253 rt6_uncached_list_add(rt);
2259 /* Get a percpu copy */
2261 rt = rt6_get_pcpu_route(&res);
2264 rt = rt6_make_pcpu_route(net, &res);
2270 rt = net->ipv6.ip6_null_entry;
2271 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2272 ip6_hold_safe(net, &rt);
2277 EXPORT_SYMBOL_GPL(ip6_pol_route);
2279 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
2280 struct fib6_table *table,
2282 const struct sk_buff *skb,
2285 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2288 struct dst_entry *ip6_route_input_lookup(struct net *net,
2289 struct net_device *dev,
2291 const struct sk_buff *skb,
2294 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2295 flags |= RT6_LOOKUP_F_IFACE;
2297 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2299 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2301 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2302 struct flow_keys *keys,
2303 struct flow_keys *flkeys)
2305 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2306 const struct ipv6hdr *key_iph = outer_iph;
2307 struct flow_keys *_flkeys = flkeys;
2308 const struct ipv6hdr *inner_iph;
2309 const struct icmp6hdr *icmph;
2310 struct ipv6hdr _inner_iph;
2311 struct icmp6hdr _icmph;
2313 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2316 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2317 sizeof(_icmph), &_icmph);
2321 if (!icmpv6_is_err(icmph->icmp6_type))
2324 inner_iph = skb_header_pointer(skb,
2325 skb_transport_offset(skb) + sizeof(*icmph),
2326 sizeof(_inner_iph), &_inner_iph);
2330 key_iph = inner_iph;
2334 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2335 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2336 keys->tags.flow_label = _flkeys->tags.flow_label;
2337 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2339 keys->addrs.v6addrs.src = key_iph->saddr;
2340 keys->addrs.v6addrs.dst = key_iph->daddr;
2341 keys->tags.flow_label = ip6_flowlabel(key_iph);
2342 keys->basic.ip_proto = key_iph->nexthdr;
2346 static u32 rt6_multipath_custom_hash_outer(const struct net *net,
2347 const struct sk_buff *skb,
2350 u32 hash_fields = ip6_multipath_hash_fields(net);
2351 struct flow_keys keys, hash_keys;
2353 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2356 memset(&hash_keys, 0, sizeof(hash_keys));
2357 skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_ENCAP);
2359 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2360 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2361 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2362 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2363 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2364 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2365 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2366 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2367 hash_keys.tags.flow_label = keys.tags.flow_label;
2368 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2369 hash_keys.ports.src = keys.ports.src;
2370 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2371 hash_keys.ports.dst = keys.ports.dst;
2373 *p_has_inner = !!(keys.control.flags & FLOW_DIS_ENCAPSULATION);
2374 return flow_hash_from_keys(&hash_keys);
2377 static u32 rt6_multipath_custom_hash_inner(const struct net *net,
2378 const struct sk_buff *skb,
2381 u32 hash_fields = ip6_multipath_hash_fields(net);
2382 struct flow_keys keys, hash_keys;
2384 /* We assume the packet carries an encapsulation, but if none was
2385 * encountered during dissection of the outer flow, then there is no
2386 * point in calling the flow dissector again.
2391 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_MASK))
2394 memset(&hash_keys, 0, sizeof(hash_keys));
2395 skb_flow_dissect_flow_keys(skb, &keys, 0);
2397 if (!(keys.control.flags & FLOW_DIS_ENCAPSULATION))
2400 if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2401 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2402 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2403 hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
2404 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2405 hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst;
2406 } else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2407 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2408 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2409 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2410 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2411 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2412 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL)
2413 hash_keys.tags.flow_label = keys.tags.flow_label;
2416 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO)
2417 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2418 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT)
2419 hash_keys.ports.src = keys.ports.src;
2420 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
2421 hash_keys.ports.dst = keys.ports.dst;
2423 return flow_hash_from_keys(&hash_keys);
2426 static u32 rt6_multipath_custom_hash_skb(const struct net *net,
2427 const struct sk_buff *skb)
2429 u32 mhash, mhash_inner;
2430 bool has_inner = true;
2432 mhash = rt6_multipath_custom_hash_outer(net, skb, &has_inner);
2433 mhash_inner = rt6_multipath_custom_hash_inner(net, skb, has_inner);
2435 return jhash_2words(mhash, mhash_inner, 0);
2438 static u32 rt6_multipath_custom_hash_fl6(const struct net *net,
2439 const struct flowi6 *fl6)
2441 u32 hash_fields = ip6_multipath_hash_fields(net);
2442 struct flow_keys hash_keys;
2444 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2447 memset(&hash_keys, 0, sizeof(hash_keys));
2448 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2449 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2450 hash_keys.addrs.v6addrs.src = fl6->saddr;
2451 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2452 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2453 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2454 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2455 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2456 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2457 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2458 hash_keys.ports.src = fl6->fl6_sport;
2459 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2460 hash_keys.ports.dst = fl6->fl6_dport;
2462 return flow_hash_from_keys(&hash_keys);
2465 /* if skb is set it will be used and fl6 can be NULL */
2466 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2467 const struct sk_buff *skb, struct flow_keys *flkeys)
2469 struct flow_keys hash_keys;
2472 switch (ip6_multipath_hash_policy(net)) {
2474 memset(&hash_keys, 0, sizeof(hash_keys));
2475 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2477 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2479 hash_keys.addrs.v6addrs.src = fl6->saddr;
2480 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2481 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2482 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2484 mhash = flow_hash_from_keys(&hash_keys);
2488 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2489 struct flow_keys keys;
2491 /* short-circuit if we already have L4 hash present */
2493 return skb_get_hash_raw(skb) >> 1;
2495 memset(&hash_keys, 0, sizeof(hash_keys));
2498 skb_flow_dissect_flow_keys(skb, &keys, flag);
2501 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2502 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2503 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2504 hash_keys.ports.src = flkeys->ports.src;
2505 hash_keys.ports.dst = flkeys->ports.dst;
2506 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2508 memset(&hash_keys, 0, sizeof(hash_keys));
2509 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2510 hash_keys.addrs.v6addrs.src = fl6->saddr;
2511 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2512 hash_keys.ports.src = fl6->fl6_sport;
2513 hash_keys.ports.dst = fl6->fl6_dport;
2514 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2516 mhash = flow_hash_from_keys(&hash_keys);
2519 memset(&hash_keys, 0, sizeof(hash_keys));
2520 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2522 struct flow_keys keys;
2525 skb_flow_dissect_flow_keys(skb, &keys, 0);
2529 /* Inner can be v4 or v6 */
2530 if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2531 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2532 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2533 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2534 } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2535 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2536 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2537 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2538 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2539 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2541 /* Same as case 0 */
2542 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2543 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2546 /* Same as case 0 */
2547 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2548 hash_keys.addrs.v6addrs.src = fl6->saddr;
2549 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2550 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2551 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2553 mhash = flow_hash_from_keys(&hash_keys);
2557 mhash = rt6_multipath_custom_hash_skb(net, skb);
2559 mhash = rt6_multipath_custom_hash_fl6(net, fl6);
2566 /* Called with rcu held */
2567 void ip6_route_input(struct sk_buff *skb)
2569 const struct ipv6hdr *iph = ipv6_hdr(skb);
2570 struct net *net = dev_net(skb->dev);
2571 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2572 struct ip_tunnel_info *tun_info;
2573 struct flowi6 fl6 = {
2574 .flowi6_iif = skb->dev->ifindex,
2575 .daddr = iph->daddr,
2576 .saddr = iph->saddr,
2577 .flowlabel = ip6_flowinfo(iph),
2578 .flowi6_mark = skb->mark,
2579 .flowi6_proto = iph->nexthdr,
2581 struct flow_keys *flkeys = NULL, _flkeys;
2583 tun_info = skb_tunnel_info(skb);
2584 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2585 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2587 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2590 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2591 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2593 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2597 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
2598 struct fib6_table *table,
2600 const struct sk_buff *skb,
2603 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2606 static struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2607 const struct sock *sk,
2613 if (ipv6_addr_type(&fl6->daddr) &
2614 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2615 struct dst_entry *dst;
2617 /* This function does not take refcnt on the dst */
2618 dst = l3mdev_link_scope_lookup(net, fl6);
2623 fl6->flowi6_iif = LOOPBACK_IFINDEX;
2625 flags |= RT6_LOOKUP_F_DST_NOREF;
2626 any_src = ipv6_addr_any(&fl6->saddr);
2627 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2628 (fl6->flowi6_oif && any_src))
2629 flags |= RT6_LOOKUP_F_IFACE;
2632 flags |= RT6_LOOKUP_F_HAS_SADDR;
2634 flags |= rt6_srcprefs2flags(READ_ONCE(inet6_sk(sk)->srcprefs));
2636 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2639 struct dst_entry *ip6_route_output_flags(struct net *net,
2640 const struct sock *sk,
2644 struct dst_entry *dst;
2645 struct rt6_info *rt6;
2648 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2649 rt6 = (struct rt6_info *)dst;
2650 /* For dst cached in uncached_list, refcnt is already taken. */
2651 if (list_empty(&rt6->dst.rt_uncached) && !dst_hold_safe(dst)) {
2652 dst = &net->ipv6.ip6_null_entry->dst;
2659 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2661 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2663 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2664 struct net_device *loopback_dev = net->loopback_dev;
2665 struct dst_entry *new = NULL;
2667 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev,
2668 DST_OBSOLETE_DEAD, 0);
2671 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2675 new->input = dst_discard;
2676 new->output = dst_discard_out;
2678 dst_copy_metrics(new, &ort->dst);
2680 rt->rt6i_idev = in6_dev_get(loopback_dev);
2681 rt->rt6i_gateway = ort->rt6i_gateway;
2682 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2684 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2685 #ifdef CONFIG_IPV6_SUBTREES
2686 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2690 dst_release(dst_orig);
2691 return new ? new : ERR_PTR(-ENOMEM);
2695 * Destination cache support functions
2698 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2702 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2705 if (fib6_check_expired(f6i))
2711 static struct dst_entry *rt6_check(struct rt6_info *rt,
2712 struct fib6_info *from,
2717 if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2718 rt_cookie != cookie)
2721 if (rt6_check_expired(rt))
2727 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2728 struct fib6_info *from,
2731 if (!__rt6_check_expired(rt) &&
2732 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2733 fib6_check(from, cookie))
2739 INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2742 struct dst_entry *dst_ret;
2743 struct fib6_info *from;
2744 struct rt6_info *rt;
2746 rt = container_of(dst, struct rt6_info, dst);
2749 return rt6_is_valid(rt) ? dst : NULL;
2753 /* All IPV6 dsts are created with ->obsolete set to the value
2754 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2755 * into this function always.
2758 from = rcu_dereference(rt->from);
2760 if (from && (rt->rt6i_flags & RTF_PCPU ||
2761 unlikely(!list_empty(&rt->dst.rt_uncached))))
2762 dst_ret = rt6_dst_from_check(rt, from, cookie);
2764 dst_ret = rt6_check(rt, from, cookie);
2770 EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
2772 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2774 struct rt6_info *rt = (struct rt6_info *) dst;
2777 if (rt->rt6i_flags & RTF_CACHE) {
2779 if (rt6_check_expired(rt)) {
2780 rt6_remove_exception_rt(rt);
2792 static void ip6_link_failure(struct sk_buff *skb)
2794 struct rt6_info *rt;
2796 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2798 rt = (struct rt6_info *) skb_dst(skb);
2801 if (rt->rt6i_flags & RTF_CACHE) {
2802 rt6_remove_exception_rt(rt);
2804 struct fib6_info *from;
2805 struct fib6_node *fn;
2807 from = rcu_dereference(rt->from);
2809 fn = rcu_dereference(from->fib6_node);
2810 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2811 WRITE_ONCE(fn->fn_sernum, -1);
2818 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2820 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2821 struct fib6_info *from;
2824 from = rcu_dereference(rt0->from);
2826 rt0->dst.expires = from->expires;
2830 dst_set_expires(&rt0->dst, timeout);
2831 rt0->rt6i_flags |= RTF_EXPIRES;
2834 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2836 struct net *net = dev_net(rt->dst.dev);
2838 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2839 rt->rt6i_flags |= RTF_MODIFIED;
2840 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2843 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2845 return !(rt->rt6i_flags & RTF_CACHE) &&
2846 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2849 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2850 const struct ipv6hdr *iph, u32 mtu,
2853 const struct in6_addr *daddr, *saddr;
2854 struct rt6_info *rt6 = (struct rt6_info *)dst;
2856 /* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2857 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2858 * [see also comment in rt6_mtu_change_route()]
2862 daddr = &iph->daddr;
2863 saddr = &iph->saddr;
2865 daddr = &sk->sk_v6_daddr;
2866 saddr = &inet6_sk(sk)->saddr;
2873 dst_confirm_neigh(dst, daddr);
2875 if (mtu < IPV6_MIN_MTU)
2877 if (mtu >= dst_mtu(dst))
2880 if (!rt6_cache_allowed_for_pmtu(rt6)) {
2881 rt6_do_update_pmtu(rt6, mtu);
2882 /* update rt6_ex->stamp for cache */
2883 if (rt6->rt6i_flags & RTF_CACHE)
2884 rt6_update_exception_stamp_rt(rt6);
2886 struct fib6_result res = {};
2887 struct rt6_info *nrt6;
2890 res.f6i = rcu_dereference(rt6->from);
2894 res.fib6_flags = res.f6i->fib6_flags;
2895 res.fib6_type = res.f6i->fib6_type;
2898 struct fib6_nh_match_arg arg = {
2900 .gw = &rt6->rt6i_gateway,
2903 nexthop_for_each_fib6_nh(res.f6i->nh,
2904 fib6_nh_find_match, &arg);
2906 /* fib6_info uses a nexthop that does not have fib6_nh
2907 * using the dst->dev + gw. Should be impossible.
2914 res.nh = res.f6i->fib6_nh;
2917 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2919 rt6_do_update_pmtu(nrt6, mtu);
2920 if (rt6_insert_exception(nrt6, &res))
2921 dst_release_immediate(&nrt6->dst);
2928 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2929 struct sk_buff *skb, u32 mtu,
2932 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2936 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2937 int oif, u32 mark, kuid_t uid)
2939 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2940 struct dst_entry *dst;
2941 struct flowi6 fl6 = {
2943 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2944 .daddr = iph->daddr,
2945 .saddr = iph->saddr,
2946 .flowlabel = ip6_flowinfo(iph),
2950 dst = ip6_route_output(net, NULL, &fl6);
2952 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
2955 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2957 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2959 int oif = sk->sk_bound_dev_if;
2960 struct dst_entry *dst;
2962 if (!oif && skb->dev)
2963 oif = l3mdev_master_ifindex(skb->dev);
2965 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, READ_ONCE(sk->sk_mark),
2968 dst = __sk_dst_get(sk);
2969 if (!dst || !dst->obsolete ||
2970 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2974 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2975 ip6_datagram_dst_update(sk, false);
2978 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2980 void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2981 const struct flowi6 *fl6)
2983 #ifdef CONFIG_IPV6_SUBTREES
2984 struct ipv6_pinfo *np = inet6_sk(sk);
2987 ip6_dst_store(sk, dst,
2988 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2989 &sk->sk_v6_daddr : NULL,
2990 #ifdef CONFIG_IPV6_SUBTREES
2991 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2997 static bool ip6_redirect_nh_match(const struct fib6_result *res,
2999 const struct in6_addr *gw,
3000 struct rt6_info **ret)
3002 const struct fib6_nh *nh = res->nh;
3004 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
3005 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
3008 /* rt_cache's gateway might be different from its 'parent'
3009 * in the case of an ip redirect.
3010 * So we keep searching in the exception table if the gateway
3013 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
3014 struct rt6_info *rt_cache;
3016 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
3018 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
3027 struct fib6_nh_rd_arg {
3028 struct fib6_result *res;
3030 const struct in6_addr *gw;
3031 struct rt6_info **ret;
3034 static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
3036 struct fib6_nh_rd_arg *arg = _arg;
3039 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
3042 /* Handle redirects */
3043 struct ip6rd_flowi {
3045 struct in6_addr gateway;
3048 INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
3049 struct fib6_table *table,
3051 const struct sk_buff *skb,
3054 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
3055 struct rt6_info *ret = NULL;
3056 struct fib6_result res = {};
3057 struct fib6_nh_rd_arg arg = {
3060 .gw = &rdfl->gateway,
3063 struct fib6_info *rt;
3064 struct fib6_node *fn;
3066 /* Get the "current" route for this destination and
3067 * check if the redirect has come from appropriate router.
3069 * RFC 4861 specifies that redirects should only be
3070 * accepted if they come from the nexthop to the target.
3071 * Due to the way the routes are chosen, this notion
3072 * is a bit fuzzy and one might need to check all possible
3077 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
3079 for_each_fib6_node_rt_rcu(fn) {
3081 if (fib6_check_expired(rt))
3083 if (rt->fib6_flags & RTF_REJECT)
3085 if (unlikely(rt->nh)) {
3086 if (nexthop_is_blackhole(rt->nh))
3088 /* on match, res->nh is filled in and potentially ret */
3089 if (nexthop_for_each_fib6_nh(rt->nh,
3090 fib6_nh_redirect_match,
3094 res.nh = rt->fib6_nh;
3095 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
3102 rt = net->ipv6.fib6_null_entry;
3103 else if (rt->fib6_flags & RTF_REJECT) {
3104 ret = net->ipv6.ip6_null_entry;
3108 if (rt == net->ipv6.fib6_null_entry) {
3109 fn = fib6_backtrack(fn, &fl6->saddr);
3115 res.nh = rt->fib6_nh;
3118 ip6_hold_safe(net, &ret);
3120 res.fib6_flags = res.f6i->fib6_flags;
3121 res.fib6_type = res.f6i->fib6_type;
3122 ret = ip6_create_rt_rcu(&res);
3127 trace_fib6_table_lookup(net, &res, table, fl6);
3131 static struct dst_entry *ip6_route_redirect(struct net *net,
3132 const struct flowi6 *fl6,
3133 const struct sk_buff *skb,
3134 const struct in6_addr *gateway)
3136 int flags = RT6_LOOKUP_F_HAS_SADDR;
3137 struct ip6rd_flowi rdfl;
3140 rdfl.gateway = *gateway;
3142 return fib6_rule_lookup(net, &rdfl.fl6, skb,
3143 flags, __ip6_route_redirect);
3146 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3149 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3150 struct dst_entry *dst;
3151 struct flowi6 fl6 = {
3152 .flowi6_iif = LOOPBACK_IFINDEX,
3154 .flowi6_mark = mark,
3155 .daddr = iph->daddr,
3156 .saddr = iph->saddr,
3157 .flowlabel = ip6_flowinfo(iph),
3161 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3162 rt6_do_redirect(dst, NULL, skb);
3165 EXPORT_SYMBOL_GPL(ip6_redirect);
3167 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
3169 const struct ipv6hdr *iph = ipv6_hdr(skb);
3170 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3171 struct dst_entry *dst;
3172 struct flowi6 fl6 = {
3173 .flowi6_iif = LOOPBACK_IFINDEX,
3176 .saddr = iph->daddr,
3177 .flowi6_uid = sock_net_uid(net, NULL),
3180 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3181 rt6_do_redirect(dst, NULL, skb);
3185 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3187 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if,
3188 READ_ONCE(sk->sk_mark), sk->sk_uid);
3190 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3192 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3194 struct net_device *dev = dst->dev;
3195 unsigned int mtu = dst_mtu(dst);
3196 struct net *net = dev_net(dev);
3198 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3200 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3201 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3204 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3205 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3206 * IPV6_MAXPLEN is also valid and means: "any MSS,
3207 * rely only on pmtu discovery"
3209 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3214 INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
3216 return ip6_dst_mtu_maybe_forward(dst, false);
3218 EXPORT_INDIRECT_CALLABLE(ip6_mtu);
3221 * 1. mtu on route is locked - use it
3222 * 2. mtu from nexthop exception
3223 * 3. mtu from egress device
3225 * based on ip6_dst_mtu_forward and exception logic of
3226 * rt6_find_cached_rt; called with rcu_read_lock
3228 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3229 const struct in6_addr *daddr,
3230 const struct in6_addr *saddr)
3232 const struct fib6_nh *nh = res->nh;
3233 struct fib6_info *f6i = res->f6i;
3234 struct inet6_dev *idev;
3235 struct rt6_info *rt;
3238 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3239 mtu = f6i->fib6_pmtu;
3244 rt = rt6_find_cached_rt(res, daddr, saddr);
3246 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3248 struct net_device *dev = nh->fib_nh_dev;
3251 idev = __in6_dev_get(dev);
3252 if (idev && idev->cnf.mtu6 > mtu)
3253 mtu = idev->cnf.mtu6;
3256 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3258 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3261 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3264 struct dst_entry *dst;
3265 struct rt6_info *rt;
3266 struct inet6_dev *idev = in6_dev_get(dev);
3267 struct net *net = dev_net(dev);
3269 if (unlikely(!idev))
3270 return ERR_PTR(-ENODEV);
3272 rt = ip6_dst_alloc(net, dev, 0);
3273 if (unlikely(!rt)) {
3275 dst = ERR_PTR(-ENOMEM);
3279 rt->dst.input = ip6_input;
3280 rt->dst.output = ip6_output;
3281 rt->rt6i_gateway = fl6->daddr;
3282 rt->rt6i_dst.addr = fl6->daddr;
3283 rt->rt6i_dst.plen = 128;
3284 rt->rt6i_idev = idev;
3285 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3287 /* Add this dst into uncached_list so that rt6_disable_ip() can
3288 * do proper release of the net_device
3290 rt6_uncached_list_add(rt);
3292 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3298 static void ip6_dst_gc(struct dst_ops *ops)
3300 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3301 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3302 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3303 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3304 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3308 if (time_after(rt_last_gc + rt_min_interval, jiffies))
3311 fib6_run_gc(atomic_inc_return(&net->ipv6.ip6_rt_gc_expire), net, true);
3312 entries = dst_entries_get_slow(ops);
3313 if (entries < ops->gc_thresh)
3314 atomic_set(&net->ipv6.ip6_rt_gc_expire, rt_gc_timeout >> 1);
3316 val = atomic_read(&net->ipv6.ip6_rt_gc_expire);
3317 atomic_set(&net->ipv6.ip6_rt_gc_expire, val - (val >> rt_elasticity));
3320 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3321 const struct in6_addr *gw_addr, u32 tbid,
3322 int flags, struct fib6_result *res)
3324 struct flowi6 fl6 = {
3325 .flowi6_oif = cfg->fc_ifindex,
3327 .saddr = cfg->fc_prefsrc,
3329 struct fib6_table *table;
3332 table = fib6_get_table(net, tbid);
3336 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3337 flags |= RT6_LOOKUP_F_HAS_SADDR;
3339 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3341 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3342 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3343 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3344 cfg->fc_ifindex != 0, NULL, flags);
3349 static int ip6_route_check_nh_onlink(struct net *net,
3350 struct fib6_config *cfg,
3351 const struct net_device *dev,
3352 struct netlink_ext_ack *extack)
3354 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3355 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3356 struct fib6_result res = {};
3359 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3360 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3361 /* ignore match if it is the default route */
3362 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3363 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3364 NL_SET_ERR_MSG(extack,
3365 "Nexthop has invalid gateway or device mismatch");
3372 static int ip6_route_check_nh(struct net *net,
3373 struct fib6_config *cfg,
3374 struct net_device **_dev,
3375 netdevice_tracker *dev_tracker,
3376 struct inet6_dev **idev)
3378 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3379 struct net_device *dev = _dev ? *_dev : NULL;
3380 int flags = RT6_LOOKUP_F_IFACE;
3381 struct fib6_result res = {};
3382 int err = -EHOSTUNREACH;
3384 if (cfg->fc_table) {
3385 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3386 cfg->fc_table, flags, &res);
3387 /* gw_addr can not require a gateway or resolve to a reject
3388 * route. If a device is given, it must match the result.
3390 if (err || res.fib6_flags & RTF_REJECT ||
3391 res.nh->fib_nh_gw_family ||
3392 (dev && dev != res.nh->fib_nh_dev))
3393 err = -EHOSTUNREACH;
3397 struct flowi6 fl6 = {
3398 .flowi6_oif = cfg->fc_ifindex,
3402 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3403 if (err || res.fib6_flags & RTF_REJECT ||
3404 res.nh->fib_nh_gw_family)
3405 err = -EHOSTUNREACH;
3410 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3411 cfg->fc_ifindex != 0, NULL, flags);
3416 if (dev != res.nh->fib_nh_dev)
3417 err = -EHOSTUNREACH;
3419 *_dev = dev = res.nh->fib_nh_dev;
3420 netdev_hold(dev, dev_tracker, GFP_ATOMIC);
3421 *idev = in6_dev_get(dev);
3427 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3428 struct net_device **_dev,
3429 netdevice_tracker *dev_tracker,
3430 struct inet6_dev **idev,
3431 struct netlink_ext_ack *extack)
3433 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3434 int gwa_type = ipv6_addr_type(gw_addr);
3435 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3436 const struct net_device *dev = *_dev;
3437 bool need_addr_check = !dev;
3440 /* if gw_addr is local we will fail to detect this in case
3441 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3442 * will return already-added prefix route via interface that
3443 * prefix route was assigned to, which might be non-loopback.
3446 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3447 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3451 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3452 /* IPv6 strictly inhibits using not link-local
3453 * addresses as nexthop address.
3454 * Otherwise, router will not able to send redirects.
3455 * It is very good, but in some (rare!) circumstances
3456 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3457 * some exceptions. --ANK
3458 * We allow IPv4-mapped nexthops to support RFC4798-type
3461 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3462 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3468 if (cfg->fc_flags & RTNH_F_ONLINK)
3469 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3471 err = ip6_route_check_nh(net, cfg, _dev, dev_tracker,
3480 /* reload in case device was changed */
3485 NL_SET_ERR_MSG(extack, "Egress device not specified");
3487 } else if (dev->flags & IFF_LOOPBACK) {
3488 NL_SET_ERR_MSG(extack,
3489 "Egress device can not be loopback device for this route");
3493 /* if we did not check gw_addr above, do so now that the
3494 * egress device has been resolved.
3496 if (need_addr_check &&
3497 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3498 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3507 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3509 if ((flags & RTF_REJECT) ||
3510 (dev && (dev->flags & IFF_LOOPBACK) &&
3511 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3512 !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3518 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3519 struct fib6_config *cfg, gfp_t gfp_flags,
3520 struct netlink_ext_ack *extack)
3522 netdevice_tracker *dev_tracker = &fib6_nh->fib_nh_dev_tracker;
3523 struct net_device *dev = NULL;
3524 struct inet6_dev *idev = NULL;
3528 fib6_nh->fib_nh_family = AF_INET6;
3529 #ifdef CONFIG_IPV6_ROUTER_PREF
3530 fib6_nh->last_probe = jiffies;
3532 if (cfg->fc_is_fdb) {
3533 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3534 fib6_nh->fib_nh_gw_family = AF_INET6;
3539 if (cfg->fc_ifindex) {
3540 dev = netdev_get_by_index(net, cfg->fc_ifindex,
3541 dev_tracker, gfp_flags);
3544 idev = in6_dev_get(dev);
3549 if (cfg->fc_flags & RTNH_F_ONLINK) {
3551 NL_SET_ERR_MSG(extack,
3552 "Nexthop device required for onlink");
3556 if (!(dev->flags & IFF_UP)) {
3557 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3562 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3565 fib6_nh->fib_nh_weight = 1;
3567 /* We cannot add true routes via loopback here,
3568 * they would result in kernel looping; promote them to reject routes
3570 addr_type = ipv6_addr_type(&cfg->fc_dst);
3571 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3572 /* hold loopback dev/idev if we haven't done so. */
3573 if (dev != net->loopback_dev) {
3575 netdev_put(dev, dev_tracker);
3578 dev = net->loopback_dev;
3579 netdev_hold(dev, dev_tracker, gfp_flags);
3580 idev = in6_dev_get(dev);
3589 if (cfg->fc_flags & RTF_GATEWAY) {
3590 err = ip6_validate_gw(net, cfg, &dev, dev_tracker,
3595 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3596 fib6_nh->fib_nh_gw_family = AF_INET6;
3603 if (idev->cnf.disable_ipv6) {
3604 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3609 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3610 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3615 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3616 !netif_carrier_ok(dev))
3617 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3619 err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3620 cfg->fc_encap_type, cfg, gfp_flags, extack);
3625 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3626 if (!fib6_nh->rt6i_pcpu) {
3631 fib6_nh->fib_nh_dev = dev;
3632 fib6_nh->fib_nh_oif = dev->ifindex;
3639 lwtstate_put(fib6_nh->fib_nh_lws);
3640 fib6_nh->fib_nh_lws = NULL;
3641 netdev_put(dev, dev_tracker);
3647 void fib6_nh_release(struct fib6_nh *fib6_nh)
3649 struct rt6_exception_bucket *bucket;
3653 fib6_nh_flush_exceptions(fib6_nh, NULL);
3654 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3656 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3662 fib6_nh_release_dsts(fib6_nh);
3663 free_percpu(fib6_nh->rt6i_pcpu);
3665 fib_nh_common_release(&fib6_nh->nh_common);
3668 void fib6_nh_release_dsts(struct fib6_nh *fib6_nh)
3672 if (!fib6_nh->rt6i_pcpu)
3675 for_each_possible_cpu(cpu) {
3676 struct rt6_info *pcpu_rt, **ppcpu_rt;
3678 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3679 pcpu_rt = xchg(ppcpu_rt, NULL);
3681 dst_dev_put(&pcpu_rt->dst);
3682 dst_release(&pcpu_rt->dst);
3687 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3689 struct netlink_ext_ack *extack)
3691 struct net *net = cfg->fc_nlinfo.nl_net;
3692 struct fib6_info *rt = NULL;
3693 struct nexthop *nh = NULL;
3694 struct fib6_table *table;
3695 struct fib6_nh *fib6_nh;
3699 /* RTF_PCPU is an internal flag; can not be set by userspace */
3700 if (cfg->fc_flags & RTF_PCPU) {
3701 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3705 /* RTF_CACHE is an internal flag; can not be set by userspace */
3706 if (cfg->fc_flags & RTF_CACHE) {
3707 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3711 if (cfg->fc_type > RTN_MAX) {
3712 NL_SET_ERR_MSG(extack, "Invalid route type");
3716 if (cfg->fc_dst_len > 128) {
3717 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3720 if (cfg->fc_src_len > 128) {
3721 NL_SET_ERR_MSG(extack, "Invalid source address length");
3724 #ifndef CONFIG_IPV6_SUBTREES
3725 if (cfg->fc_src_len) {
3726 NL_SET_ERR_MSG(extack,
3727 "Specifying source address requires IPV6_SUBTREES to be enabled");
3731 if (cfg->fc_nh_id) {
3732 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3734 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3737 err = fib6_check_nexthop(nh, cfg, extack);
3743 if (cfg->fc_nlinfo.nlh &&
3744 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3745 table = fib6_get_table(net, cfg->fc_table);
3747 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3748 table = fib6_new_table(net, cfg->fc_table);
3751 table = fib6_new_table(net, cfg->fc_table);
3758 rt = fib6_info_alloc(gfp_flags, !nh);
3762 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3764 if (IS_ERR(rt->fib6_metrics)) {
3765 err = PTR_ERR(rt->fib6_metrics);
3766 /* Do not leave garbage there. */
3767 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3771 if (cfg->fc_flags & RTF_ADDRCONF)
3772 rt->dst_nocount = true;
3774 if (cfg->fc_flags & RTF_EXPIRES)
3775 fib6_set_expires(rt, jiffies +
3776 clock_t_to_jiffies(cfg->fc_expires));
3778 if (cfg->fc_protocol == RTPROT_UNSPEC)
3779 cfg->fc_protocol = RTPROT_BOOT;
3780 rt->fib6_protocol = cfg->fc_protocol;
3782 rt->fib6_table = table;
3783 rt->fib6_metric = cfg->fc_metric;
3784 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3785 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3787 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3788 rt->fib6_dst.plen = cfg->fc_dst_len;
3790 #ifdef CONFIG_IPV6_SUBTREES
3791 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3792 rt->fib6_src.plen = cfg->fc_src_len;
3795 if (rt->fib6_src.plen) {
3796 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3799 if (!nexthop_get(nh)) {
3800 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3804 fib6_nh = nexthop_fib6_nh(rt->nh);
3806 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3810 fib6_nh = rt->fib6_nh;
3812 /* We cannot add true routes via loopback here, they would
3813 * result in kernel looping; promote them to reject routes
3815 addr_type = ipv6_addr_type(&cfg->fc_dst);
3816 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3818 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3821 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3822 struct net_device *dev = fib6_nh->fib_nh_dev;
3824 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3825 NL_SET_ERR_MSG(extack, "Invalid source address");
3829 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3830 rt->fib6_prefsrc.plen = 128;
3832 rt->fib6_prefsrc.plen = 0;
3836 fib6_info_release(rt);
3837 return ERR_PTR(err);
3839 ip_fib_metrics_put(rt->fib6_metrics);
3841 return ERR_PTR(err);
3844 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3845 struct netlink_ext_ack *extack)
3847 struct fib6_info *rt;
3850 rt = ip6_route_info_create(cfg, gfp_flags, extack);
3854 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3855 fib6_info_release(rt);
3860 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3862 struct net *net = info->nl_net;
3863 struct fib6_table *table;
3866 if (rt == net->ipv6.fib6_null_entry) {
3871 table = rt->fib6_table;
3872 spin_lock_bh(&table->tb6_lock);
3873 err = fib6_del(rt, info);
3874 spin_unlock_bh(&table->tb6_lock);
3877 fib6_info_release(rt);
3881 int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
3883 struct nl_info info = {
3885 .skip_notify = skip_notify
3888 return __ip6_del_rt(rt, &info);
3891 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3893 struct nl_info *info = &cfg->fc_nlinfo;
3894 struct net *net = info->nl_net;
3895 struct sk_buff *skb = NULL;
3896 struct fib6_table *table;
3899 if (rt == net->ipv6.fib6_null_entry)
3901 table = rt->fib6_table;
3902 spin_lock_bh(&table->tb6_lock);
3904 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3905 struct fib6_info *sibling, *next_sibling;
3906 struct fib6_node *fn;
3908 /* prefer to send a single notification with all hops */
3909 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3911 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3913 if (rt6_fill_node(net, skb, rt, NULL,
3914 NULL, NULL, 0, RTM_DELROUTE,
3915 info->portid, seq, 0) < 0) {
3919 info->skip_notify = 1;
3922 /* 'rt' points to the first sibling route. If it is not the
3923 * leaf, then we do not need to send a notification. Otherwise,
3924 * we need to check if the last sibling has a next route or not
3925 * and emit a replace or delete notification, respectively.
3927 info->skip_notify_kernel = 1;
3928 fn = rcu_dereference_protected(rt->fib6_node,
3929 lockdep_is_held(&table->tb6_lock));
3930 if (rcu_access_pointer(fn->leaf) == rt) {
3931 struct fib6_info *last_sibling, *replace_rt;
3933 last_sibling = list_last_entry(&rt->fib6_siblings,
3936 replace_rt = rcu_dereference_protected(
3937 last_sibling->fib6_next,
3938 lockdep_is_held(&table->tb6_lock));
3940 call_fib6_entry_notifiers_replace(net,
3943 call_fib6_multipath_entry_notifiers(net,
3944 FIB_EVENT_ENTRY_DEL,
3945 rt, rt->fib6_nsiblings,
3948 list_for_each_entry_safe(sibling, next_sibling,
3951 err = fib6_del(sibling, info);
3957 err = fib6_del(rt, info);
3959 spin_unlock_bh(&table->tb6_lock);
3961 fib6_info_release(rt);
3964 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3965 info->nlh, gfp_any());
3970 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3974 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3977 if (cfg->fc_flags & RTF_GATEWAY &&
3978 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3981 rc = rt6_remove_exception_rt(rt);
3986 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3989 struct fib6_result res = {
3993 struct rt6_info *rt_cache;
3995 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3997 return __ip6_del_cached_rt(rt_cache, cfg);
4002 struct fib6_nh_del_cached_rt_arg {
4003 struct fib6_config *cfg;
4004 struct fib6_info *f6i;
4007 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
4009 struct fib6_nh_del_cached_rt_arg *arg = _arg;
4012 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
4013 return rc != -ESRCH ? rc : 0;
4016 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
4018 struct fib6_nh_del_cached_rt_arg arg = {
4023 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
4026 static int ip6_route_del(struct fib6_config *cfg,
4027 struct netlink_ext_ack *extack)
4029 struct fib6_table *table;
4030 struct fib6_info *rt;
4031 struct fib6_node *fn;
4034 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
4036 NL_SET_ERR_MSG(extack, "FIB table does not exist");
4042 fn = fib6_locate(&table->tb6_root,
4043 &cfg->fc_dst, cfg->fc_dst_len,
4044 &cfg->fc_src, cfg->fc_src_len,
4045 !(cfg->fc_flags & RTF_CACHE));
4048 for_each_fib6_node_rt_rcu(fn) {
4051 if (rt->nh && cfg->fc_nh_id &&
4052 rt->nh->id != cfg->fc_nh_id)
4055 if (cfg->fc_flags & RTF_CACHE) {
4059 rc = ip6_del_cached_rt_nh(cfg, rt);
4060 } else if (cfg->fc_nh_id) {
4064 rc = ip6_del_cached_rt(cfg, rt, nh);
4073 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
4075 if (cfg->fc_protocol &&
4076 cfg->fc_protocol != rt->fib6_protocol)
4080 if (!fib6_info_hold_safe(rt))
4084 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4090 if (cfg->fc_ifindex &&
4092 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
4094 if (cfg->fc_flags & RTF_GATEWAY &&
4095 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
4097 if (!fib6_info_hold_safe(rt))
4101 /* if gateway was specified only delete the one hop */
4102 if (cfg->fc_flags & RTF_GATEWAY)
4103 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4105 return __ip6_del_rt_siblings(rt, cfg);
4113 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
4115 struct netevent_redirect netevent;
4116 struct rt6_info *rt, *nrt = NULL;
4117 struct fib6_result res = {};
4118 struct ndisc_options ndopts;
4119 struct inet6_dev *in6_dev;
4120 struct neighbour *neigh;
4122 int optlen, on_link;
4125 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
4126 optlen -= sizeof(*msg);
4129 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4133 msg = (struct rd_msg *)icmp6_hdr(skb);
4135 if (ipv6_addr_is_multicast(&msg->dest)) {
4136 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4141 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4143 } else if (ipv6_addr_type(&msg->target) !=
4144 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4145 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4149 in6_dev = __in6_dev_get(skb->dev);
4152 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
4156 * The IP source address of the Redirect MUST be the same as the current
4157 * first-hop router for the specified ICMP Destination Address.
4160 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4161 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4166 if (ndopts.nd_opts_tgt_lladdr) {
4167 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4170 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4175 rt = (struct rt6_info *) dst;
4176 if (rt->rt6i_flags & RTF_REJECT) {
4177 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4181 /* Redirect received -> path was valid.
4182 * Look, redirects are sent only in response to data packets,
4183 * so that this nexthop apparently is reachable. --ANK
4185 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4187 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4192 * We have finally decided to accept it.
4195 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4196 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4197 NEIGH_UPDATE_F_OVERRIDE|
4198 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4199 NEIGH_UPDATE_F_ISROUTER)),
4200 NDISC_REDIRECT, &ndopts);
4203 res.f6i = rcu_dereference(rt->from);
4208 struct fib6_nh_match_arg arg = {
4210 .gw = &rt->rt6i_gateway,
4213 nexthop_for_each_fib6_nh(res.f6i->nh,
4214 fib6_nh_find_match, &arg);
4216 /* fib6_info uses a nexthop that does not have fib6_nh
4217 * using the dst->dev. Should be impossible
4223 res.nh = res.f6i->fib6_nh;
4226 res.fib6_flags = res.f6i->fib6_flags;
4227 res.fib6_type = res.f6i->fib6_type;
4228 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4232 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4234 nrt->rt6i_flags &= ~RTF_GATEWAY;
4236 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4238 /* rt6_insert_exception() will take care of duplicated exceptions */
4239 if (rt6_insert_exception(nrt, &res)) {
4240 dst_release_immediate(&nrt->dst);
4244 netevent.old = &rt->dst;
4245 netevent.new = &nrt->dst;
4246 netevent.daddr = &msg->dest;
4247 netevent.neigh = neigh;
4248 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4252 neigh_release(neigh);
4255 #ifdef CONFIG_IPV6_ROUTE_INFO
4256 static struct fib6_info *rt6_get_route_info(struct net *net,
4257 const struct in6_addr *prefix, int prefixlen,
4258 const struct in6_addr *gwaddr,
4259 struct net_device *dev)
4261 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4262 int ifindex = dev->ifindex;
4263 struct fib6_node *fn;
4264 struct fib6_info *rt = NULL;
4265 struct fib6_table *table;
4267 table = fib6_get_table(net, tb_id);
4272 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4276 for_each_fib6_node_rt_rcu(fn) {
4277 /* these routes do not use nexthops */
4280 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4282 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4283 !rt->fib6_nh->fib_nh_gw_family)
4285 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4287 if (!fib6_info_hold_safe(rt))
4296 static struct fib6_info *rt6_add_route_info(struct net *net,
4297 const struct in6_addr *prefix, int prefixlen,
4298 const struct in6_addr *gwaddr,
4299 struct net_device *dev,
4302 struct fib6_config cfg = {
4303 .fc_metric = IP6_RT_PRIO_USER,
4304 .fc_ifindex = dev->ifindex,
4305 .fc_dst_len = prefixlen,
4306 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4307 RTF_UP | RTF_PREF(pref),
4308 .fc_protocol = RTPROT_RA,
4309 .fc_type = RTN_UNICAST,
4310 .fc_nlinfo.portid = 0,
4311 .fc_nlinfo.nlh = NULL,
4312 .fc_nlinfo.nl_net = net,
4315 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4316 cfg.fc_dst = *prefix;
4317 cfg.fc_gateway = *gwaddr;
4319 /* We should treat it as a default route if prefix length is 0. */
4321 cfg.fc_flags |= RTF_DEFAULT;
4323 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4325 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4329 struct fib6_info *rt6_get_dflt_router(struct net *net,
4330 const struct in6_addr *addr,
4331 struct net_device *dev)
4333 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4334 struct fib6_info *rt;
4335 struct fib6_table *table;
4337 table = fib6_get_table(net, tb_id);
4342 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4345 /* RA routes do not use nexthops */
4350 if (dev == nh->fib_nh_dev &&
4351 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4352 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4355 if (rt && !fib6_info_hold_safe(rt))
4361 struct fib6_info *rt6_add_dflt_router(struct net *net,
4362 const struct in6_addr *gwaddr,
4363 struct net_device *dev,
4365 u32 defrtr_usr_metric,
4368 struct fib6_config cfg = {
4369 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4370 .fc_metric = defrtr_usr_metric,
4371 .fc_ifindex = dev->ifindex,
4372 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4373 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4374 .fc_protocol = RTPROT_RA,
4375 .fc_type = RTN_UNICAST,
4376 .fc_nlinfo.portid = 0,
4377 .fc_nlinfo.nlh = NULL,
4378 .fc_nlinfo.nl_net = net,
4379 .fc_expires = jiffies_to_clock_t(lifetime * HZ),
4382 cfg.fc_gateway = *gwaddr;
4384 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4385 struct fib6_table *table;
4387 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4389 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4392 return rt6_get_dflt_router(net, gwaddr, dev);
4395 static void __rt6_purge_dflt_routers(struct net *net,
4396 struct fib6_table *table)
4398 struct fib6_info *rt;
4402 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4403 struct net_device *dev = fib6_info_nh_dev(rt);
4404 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4406 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4407 (!idev || idev->cnf.accept_ra != 2) &&
4408 fib6_info_hold_safe(rt)) {
4410 ip6_del_rt(net, rt, false);
4416 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4419 void rt6_purge_dflt_routers(struct net *net)
4421 struct fib6_table *table;
4422 struct hlist_head *head;
4427 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4428 head = &net->ipv6.fib_table_hash[h];
4429 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4430 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4431 __rt6_purge_dflt_routers(net, table);
4438 static void rtmsg_to_fib6_config(struct net *net,
4439 struct in6_rtmsg *rtmsg,
4440 struct fib6_config *cfg)
4442 *cfg = (struct fib6_config){
4443 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4445 .fc_ifindex = rtmsg->rtmsg_ifindex,
4446 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
4447 .fc_expires = rtmsg->rtmsg_info,
4448 .fc_dst_len = rtmsg->rtmsg_dst_len,
4449 .fc_src_len = rtmsg->rtmsg_src_len,
4450 .fc_flags = rtmsg->rtmsg_flags,
4451 .fc_type = rtmsg->rtmsg_type,
4453 .fc_nlinfo.nl_net = net,
4455 .fc_dst = rtmsg->rtmsg_dst,
4456 .fc_src = rtmsg->rtmsg_src,
4457 .fc_gateway = rtmsg->rtmsg_gateway,
4461 int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
4463 struct fib6_config cfg;
4466 if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4468 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4471 rtmsg_to_fib6_config(net, rtmsg, &cfg);
4476 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4479 err = ip6_route_del(&cfg, NULL);
4487 * Drop the packet on the floor
4490 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4492 struct dst_entry *dst = skb_dst(skb);
4493 struct net *net = dev_net(dst->dev);
4494 struct inet6_dev *idev;
4498 if (netif_is_l3_master(skb->dev) ||
4499 dst->dev == net->loopback_dev)
4500 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4502 idev = ip6_dst_idev(dst);
4504 switch (ipstats_mib_noroutes) {
4505 case IPSTATS_MIB_INNOROUTES:
4506 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4507 if (type == IPV6_ADDR_ANY) {
4508 SKB_DR_SET(reason, IP_INADDRERRORS);
4509 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4512 SKB_DR_SET(reason, IP_INNOROUTES);
4514 case IPSTATS_MIB_OUTNOROUTES:
4515 SKB_DR_OR(reason, IP_OUTNOROUTES);
4516 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4520 /* Start over by dropping the dst for l3mdev case */
4521 if (netif_is_l3_master(skb->dev))
4524 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4525 kfree_skb_reason(skb, reason);
4529 static int ip6_pkt_discard(struct sk_buff *skb)
4531 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4534 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4536 skb->dev = skb_dst(skb)->dev;
4537 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4540 static int ip6_pkt_prohibit(struct sk_buff *skb)
4542 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4545 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4547 skb->dev = skb_dst(skb)->dev;
4548 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4552 * Allocate a dst for local (unicast / anycast) address.
4555 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4556 struct inet6_dev *idev,
4557 const struct in6_addr *addr,
4558 bool anycast, gfp_t gfp_flags,
4559 struct netlink_ext_ack *extack)
4561 struct fib6_config cfg = {
4562 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4563 .fc_ifindex = idev->dev->ifindex,
4564 .fc_flags = RTF_UP | RTF_NONEXTHOP,
4567 .fc_protocol = RTPROT_KERNEL,
4568 .fc_nlinfo.nl_net = net,
4569 .fc_ignore_dev_down = true,
4571 struct fib6_info *f6i;
4574 cfg.fc_type = RTN_ANYCAST;
4575 cfg.fc_flags |= RTF_ANYCAST;
4577 cfg.fc_type = RTN_LOCAL;
4578 cfg.fc_flags |= RTF_LOCAL;
4581 f6i = ip6_route_info_create(&cfg, gfp_flags, extack);
4583 f6i->dst_nocount = true;
4586 (net->ipv6.devconf_all->disable_policy ||
4587 idev->cnf.disable_policy))
4588 f6i->dst_nopolicy = true;
4594 /* remove deleted ip from prefsrc entries */
4595 struct arg_dev_net_ip {
4597 struct in6_addr *addr;
4600 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4602 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4603 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4606 rt != net->ipv6.fib6_null_entry &&
4607 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr) &&
4608 !ipv6_chk_addr(net, addr, rt->fib6_nh->fib_nh_dev, 0)) {
4609 spin_lock_bh(&rt6_exception_lock);
4610 /* remove prefsrc entry */
4611 rt->fib6_prefsrc.plen = 0;
4612 spin_unlock_bh(&rt6_exception_lock);
4617 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4619 struct net *net = dev_net(ifp->idev->dev);
4620 struct arg_dev_net_ip adni = {
4624 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4627 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
4629 /* Remove routers and update dst entries when gateway turn into host. */
4630 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4632 struct in6_addr *gateway = (struct in6_addr *)arg;
4635 /* RA routes do not use nexthops */
4640 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4641 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4644 /* Further clean up cached routes in exception table.
4645 * This is needed because cached route may have a different
4646 * gateway than its 'parent' in the case of an ip redirect.
4648 fib6_nh_exceptions_clean_tohost(nh, gateway);
4653 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4655 fib6_clean_all(net, fib6_clean_tohost, gateway);
4658 struct arg_netdev_event {
4659 const struct net_device *dev;
4661 unsigned char nh_flags;
4662 unsigned long event;
4666 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4668 struct fib6_info *iter;
4669 struct fib6_node *fn;
4671 fn = rcu_dereference_protected(rt->fib6_node,
4672 lockdep_is_held(&rt->fib6_table->tb6_lock));
4673 iter = rcu_dereference_protected(fn->leaf,
4674 lockdep_is_held(&rt->fib6_table->tb6_lock));
4676 if (iter->fib6_metric == rt->fib6_metric &&
4677 rt6_qualify_for_ecmp(iter))
4679 iter = rcu_dereference_protected(iter->fib6_next,
4680 lockdep_is_held(&rt->fib6_table->tb6_lock));
4686 /* only called for fib entries with builtin fib6_nh */
4687 static bool rt6_is_dead(const struct fib6_info *rt)
4689 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4690 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4691 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4697 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4699 struct fib6_info *iter;
4702 if (!rt6_is_dead(rt))
4703 total += rt->fib6_nh->fib_nh_weight;
4705 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4706 if (!rt6_is_dead(iter))
4707 total += iter->fib6_nh->fib_nh_weight;
4713 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4715 int upper_bound = -1;
4717 if (!rt6_is_dead(rt)) {
4718 *weight += rt->fib6_nh->fib_nh_weight;
4719 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4722 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4725 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4727 struct fib6_info *iter;
4730 rt6_upper_bound_set(rt, &weight, total);
4732 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4733 rt6_upper_bound_set(iter, &weight, total);
4736 void rt6_multipath_rebalance(struct fib6_info *rt)
4738 struct fib6_info *first;
4741 /* In case the entire multipath route was marked for flushing,
4742 * then there is no need to rebalance upon the removal of every
4745 if (!rt->fib6_nsiblings || rt->should_flush)
4748 /* During lookup routes are evaluated in order, so we need to
4749 * make sure upper bounds are assigned from the first sibling
4752 first = rt6_multipath_first_sibling(rt);
4753 if (WARN_ON_ONCE(!first))
4756 total = rt6_multipath_total_weight(first);
4757 rt6_multipath_upper_bound_set(first, total);
4760 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4762 const struct arg_netdev_event *arg = p_arg;
4763 struct net *net = dev_net(arg->dev);
4765 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4766 rt->fib6_nh->fib_nh_dev == arg->dev) {
4767 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4768 fib6_update_sernum_upto_root(net, rt);
4769 rt6_multipath_rebalance(rt);
4775 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4777 struct arg_netdev_event arg = {
4780 .nh_flags = nh_flags,
4784 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4785 arg.nh_flags |= RTNH_F_LINKDOWN;
4787 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4790 /* only called for fib entries with inline fib6_nh */
4791 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4792 const struct net_device *dev)
4794 struct fib6_info *iter;
4796 if (rt->fib6_nh->fib_nh_dev == dev)
4798 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4799 if (iter->fib6_nh->fib_nh_dev == dev)
4805 static void rt6_multipath_flush(struct fib6_info *rt)
4807 struct fib6_info *iter;
4809 rt->should_flush = 1;
4810 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4811 iter->should_flush = 1;
4814 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4815 const struct net_device *down_dev)
4817 struct fib6_info *iter;
4818 unsigned int dead = 0;
4820 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4821 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4823 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4824 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4825 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4831 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4832 const struct net_device *dev,
4833 unsigned char nh_flags)
4835 struct fib6_info *iter;
4837 if (rt->fib6_nh->fib_nh_dev == dev)
4838 rt->fib6_nh->fib_nh_flags |= nh_flags;
4839 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4840 if (iter->fib6_nh->fib_nh_dev == dev)
4841 iter->fib6_nh->fib_nh_flags |= nh_flags;
4844 /* called with write lock held for table with rt */
4845 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4847 const struct arg_netdev_event *arg = p_arg;
4848 const struct net_device *dev = arg->dev;
4849 struct net *net = dev_net(dev);
4851 if (rt == net->ipv6.fib6_null_entry || rt->nh)
4854 switch (arg->event) {
4855 case NETDEV_UNREGISTER:
4856 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4858 if (rt->should_flush)
4860 if (!rt->fib6_nsiblings)
4861 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4862 if (rt6_multipath_uses_dev(rt, dev)) {
4865 count = rt6_multipath_dead_count(rt, dev);
4866 if (rt->fib6_nsiblings + 1 == count) {
4867 rt6_multipath_flush(rt);
4870 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4872 fib6_update_sernum(net, rt);
4873 rt6_multipath_rebalance(rt);
4877 if (rt->fib6_nh->fib_nh_dev != dev ||
4878 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4880 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4881 rt6_multipath_rebalance(rt);
4888 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4890 struct arg_netdev_event arg = {
4896 struct net *net = dev_net(dev);
4898 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4899 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4901 fib6_clean_all(net, fib6_ifdown, &arg);
4904 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4906 rt6_sync_down_dev(dev, event);
4907 rt6_uncached_list_flush_dev(dev);
4908 neigh_ifdown(&nd_tbl, dev);
4911 struct rt6_mtu_change_arg {
4912 struct net_device *dev;
4914 struct fib6_info *f6i;
4917 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4919 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4920 struct fib6_info *f6i = arg->f6i;
4922 /* For administrative MTU increase, there is no way to discover
4923 * IPv6 PMTU increase, so PMTU increase should be updated here.
4924 * Since RFC 1981 doesn't include administrative MTU increase
4925 * update PMTU increase is a MUST. (i.e. jumbo frame)
4927 if (nh->fib_nh_dev == arg->dev) {
4928 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4929 u32 mtu = f6i->fib6_pmtu;
4931 if (mtu >= arg->mtu ||
4932 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4933 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4935 spin_lock_bh(&rt6_exception_lock);
4936 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4937 spin_unlock_bh(&rt6_exception_lock);
4943 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4945 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4946 struct inet6_dev *idev;
4948 /* In IPv6 pmtu discovery is not optional,
4949 so that RTAX_MTU lock cannot disable it.
4950 We still use this lock to block changes
4951 caused by addrconf/ndisc.
4954 idev = __in6_dev_get(arg->dev);
4958 if (fib6_metric_locked(f6i, RTAX_MTU))
4963 /* fib6_nh_mtu_change only returns 0, so this is safe */
4964 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4968 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4971 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4973 struct rt6_mtu_change_arg arg = {
4978 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4981 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4982 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
4983 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
4984 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
4985 [RTA_OIF] = { .type = NLA_U32 },
4986 [RTA_IIF] = { .type = NLA_U32 },
4987 [RTA_PRIORITY] = { .type = NLA_U32 },
4988 [RTA_METRICS] = { .type = NLA_NESTED },
4989 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
4990 [RTA_PREF] = { .type = NLA_U8 },
4991 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4992 [RTA_ENCAP] = { .type = NLA_NESTED },
4993 [RTA_EXPIRES] = { .type = NLA_U32 },
4994 [RTA_UID] = { .type = NLA_U32 },
4995 [RTA_MARK] = { .type = NLA_U32 },
4996 [RTA_TABLE] = { .type = NLA_U32 },
4997 [RTA_IP_PROTO] = { .type = NLA_U8 },
4998 [RTA_SPORT] = { .type = NLA_U16 },
4999 [RTA_DPORT] = { .type = NLA_U16 },
5000 [RTA_NH_ID] = { .type = NLA_U32 },
5003 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
5004 struct fib6_config *cfg,
5005 struct netlink_ext_ack *extack)
5008 struct nlattr *tb[RTA_MAX+1];
5012 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5013 rtm_ipv6_policy, extack);
5018 rtm = nlmsg_data(nlh);
5021 NL_SET_ERR_MSG(extack,
5022 "Invalid dsfield (tos): option not available for IPv6");
5026 *cfg = (struct fib6_config){
5027 .fc_table = rtm->rtm_table,
5028 .fc_dst_len = rtm->rtm_dst_len,
5029 .fc_src_len = rtm->rtm_src_len,
5031 .fc_protocol = rtm->rtm_protocol,
5032 .fc_type = rtm->rtm_type,
5034 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
5035 .fc_nlinfo.nlh = nlh,
5036 .fc_nlinfo.nl_net = sock_net(skb->sk),
5039 if (rtm->rtm_type == RTN_UNREACHABLE ||
5040 rtm->rtm_type == RTN_BLACKHOLE ||
5041 rtm->rtm_type == RTN_PROHIBIT ||
5042 rtm->rtm_type == RTN_THROW)
5043 cfg->fc_flags |= RTF_REJECT;
5045 if (rtm->rtm_type == RTN_LOCAL)
5046 cfg->fc_flags |= RTF_LOCAL;
5048 if (rtm->rtm_flags & RTM_F_CLONED)
5049 cfg->fc_flags |= RTF_CACHE;
5051 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5053 if (tb[RTA_NH_ID]) {
5054 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
5055 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5056 NL_SET_ERR_MSG(extack,
5057 "Nexthop specification and nexthop id are mutually exclusive");
5060 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5063 if (tb[RTA_GATEWAY]) {
5064 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
5065 cfg->fc_flags |= RTF_GATEWAY;
5068 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5073 int plen = (rtm->rtm_dst_len + 7) >> 3;
5075 if (nla_len(tb[RTA_DST]) < plen)
5078 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
5082 int plen = (rtm->rtm_src_len + 7) >> 3;
5084 if (nla_len(tb[RTA_SRC]) < plen)
5087 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
5090 if (tb[RTA_PREFSRC])
5091 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
5094 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5096 if (tb[RTA_PRIORITY])
5097 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5099 if (tb[RTA_METRICS]) {
5100 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5101 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
5105 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5107 if (tb[RTA_MULTIPATH]) {
5108 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5109 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
5111 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
5112 cfg->fc_mp_len, extack);
5118 pref = nla_get_u8(tb[RTA_PREF]);
5119 if (pref != ICMPV6_ROUTER_PREF_LOW &&
5120 pref != ICMPV6_ROUTER_PREF_HIGH)
5121 pref = ICMPV6_ROUTER_PREF_MEDIUM;
5122 cfg->fc_flags |= RTF_PREF(pref);
5126 cfg->fc_encap = tb[RTA_ENCAP];
5128 if (tb[RTA_ENCAP_TYPE]) {
5129 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5131 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5136 if (tb[RTA_EXPIRES]) {
5137 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5139 if (addrconf_finite_timeout(timeout)) {
5140 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5141 cfg->fc_flags |= RTF_EXPIRES;
5151 struct fib6_info *fib6_info;
5152 struct fib6_config r_cfg;
5153 struct list_head next;
5156 static int ip6_route_info_append(struct net *net,
5157 struct list_head *rt6_nh_list,
5158 struct fib6_info *rt,
5159 struct fib6_config *r_cfg)
5164 list_for_each_entry(nh, rt6_nh_list, next) {
5165 /* check if fib6_info already exists */
5166 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5170 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5174 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5175 list_add_tail(&nh->next, rt6_nh_list);
5180 static void ip6_route_mpath_notify(struct fib6_info *rt,
5181 struct fib6_info *rt_last,
5182 struct nl_info *info,
5185 /* if this is an APPEND route, then rt points to the first route
5186 * inserted and rt_last points to last route inserted. Userspace
5187 * wants a consistent dump of the route which starts at the first
5188 * nexthop. Since sibling routes are always added at the end of
5189 * the list, find the first sibling of the last route appended
5191 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5192 rt = list_first_entry(&rt_last->fib6_siblings,
5198 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5201 static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5203 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5204 bool should_notify = false;
5205 struct fib6_info *leaf;
5206 struct fib6_node *fn;
5209 fn = rcu_dereference(rt->fib6_node);
5213 leaf = rcu_dereference(fn->leaf);
5218 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5219 rt6_qualify_for_ecmp(leaf)))
5220 should_notify = true;
5224 return should_notify;
5227 static int fib6_gw_from_attr(struct in6_addr *gw, struct nlattr *nla,
5228 struct netlink_ext_ack *extack)
5230 if (nla_len(nla) < sizeof(*gw)) {
5231 NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_GATEWAY");
5235 *gw = nla_get_in6_addr(nla);
5240 static int ip6_route_multipath_add(struct fib6_config *cfg,
5241 struct netlink_ext_ack *extack)
5243 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5244 struct nl_info *info = &cfg->fc_nlinfo;
5245 struct fib6_config r_cfg;
5246 struct rtnexthop *rtnh;
5247 struct fib6_info *rt;
5248 struct rt6_nh *err_nh;
5249 struct rt6_nh *nh, *nh_safe;
5255 int replace = (cfg->fc_nlinfo.nlh &&
5256 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5257 LIST_HEAD(rt6_nh_list);
5259 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5260 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5261 nlflags |= NLM_F_APPEND;
5263 remaining = cfg->fc_mp_len;
5264 rtnh = (struct rtnexthop *)cfg->fc_mp;
5266 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
5267 * fib6_info structs per nexthop
5269 while (rtnh_ok(rtnh, remaining)) {
5270 memcpy(&r_cfg, cfg, sizeof(*cfg));
5271 if (rtnh->rtnh_ifindex)
5272 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5274 attrlen = rtnh_attrlen(rtnh);
5276 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5278 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5280 err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5285 r_cfg.fc_flags |= RTF_GATEWAY;
5287 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5289 /* RTA_ENCAP_TYPE length checked in
5290 * lwtunnel_valid_encap_type_attr
5292 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5294 r_cfg.fc_encap_type = nla_get_u16(nla);
5297 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5298 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5304 if (!rt6_qualify_for_ecmp(rt)) {
5306 NL_SET_ERR_MSG(extack,
5307 "Device only routes can not be added for IPv6 using the multipath API.");
5308 fib6_info_release(rt);
5312 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5314 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5317 fib6_info_release(rt);
5321 rtnh = rtnh_next(rtnh, &remaining);
5324 if (list_empty(&rt6_nh_list)) {
5325 NL_SET_ERR_MSG(extack,
5326 "Invalid nexthop configuration - no valid nexthops");
5330 /* for add and replace send one notification with all nexthops.
5331 * Skip the notification in fib6_add_rt2node and send one with
5332 * the full route when done
5334 info->skip_notify = 1;
5336 /* For add and replace, send one notification with all nexthops. For
5337 * append, send one notification with all appended nexthops.
5339 info->skip_notify_kernel = 1;
5342 list_for_each_entry(nh, &rt6_nh_list, next) {
5343 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5344 fib6_info_release(nh->fib6_info);
5347 /* save reference to last route successfully inserted */
5348 rt_last = nh->fib6_info;
5350 /* save reference to first route for notification */
5352 rt_notif = nh->fib6_info;
5355 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5356 nh->fib6_info = NULL;
5359 NL_SET_ERR_MSG_MOD(extack,
5360 "multipath route replace failed (check consistency of installed routes)");
5365 /* Because each route is added like a single route we remove
5366 * these flags after the first nexthop: if there is a collision,
5367 * we have already failed to add the first nexthop:
5368 * fib6_add_rt2node() has rejected it; when replacing, old
5369 * nexthops have been replaced by first new, the rest should
5372 if (cfg->fc_nlinfo.nlh) {
5373 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5375 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5380 /* An in-kernel notification should only be sent in case the new
5381 * multipath route is added as the first route in the node, or if
5382 * it was appended to it. We pass 'rt_notif' since it is the first
5383 * sibling and might allow us to skip some checks in the replace case.
5385 if (ip6_route_mpath_should_notify(rt_notif)) {
5386 enum fib_event_type fib_event;
5388 if (rt_notif->fib6_nsiblings != nhn - 1)
5389 fib_event = FIB_EVENT_ENTRY_APPEND;
5391 fib_event = FIB_EVENT_ENTRY_REPLACE;
5393 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5394 fib_event, rt_notif,
5397 /* Delete all the siblings that were just added */
5403 /* success ... tell user about new route */
5404 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5408 /* send notification for routes that were added so that
5409 * the delete notifications sent by ip6_route_del are
5413 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5415 /* Delete routes that were already added */
5416 list_for_each_entry(nh, &rt6_nh_list, next) {
5419 ip6_route_del(&nh->r_cfg, extack);
5423 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5425 fib6_info_release(nh->fib6_info);
5426 list_del(&nh->next);
5433 static int ip6_route_multipath_del(struct fib6_config *cfg,
5434 struct netlink_ext_ack *extack)
5436 struct fib6_config r_cfg;
5437 struct rtnexthop *rtnh;
5443 remaining = cfg->fc_mp_len;
5444 rtnh = (struct rtnexthop *)cfg->fc_mp;
5446 /* Parse a Multipath Entry */
5447 while (rtnh_ok(rtnh, remaining)) {
5448 memcpy(&r_cfg, cfg, sizeof(*cfg));
5449 if (rtnh->rtnh_ifindex)
5450 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5452 attrlen = rtnh_attrlen(rtnh);
5454 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5456 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5458 err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5465 r_cfg.fc_flags |= RTF_GATEWAY;
5468 err = ip6_route_del(&r_cfg, extack);
5473 rtnh = rtnh_next(rtnh, &remaining);
5479 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5480 struct netlink_ext_ack *extack)
5482 struct fib6_config cfg;
5485 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5490 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5491 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5496 return ip6_route_multipath_del(&cfg, extack);
5498 cfg.fc_delete_all_nh = 1;
5499 return ip6_route_del(&cfg, extack);
5503 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5504 struct netlink_ext_ack *extack)
5506 struct fib6_config cfg;
5509 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5513 if (cfg.fc_metric == 0)
5514 cfg.fc_metric = IP6_RT_PRIO_USER;
5517 return ip6_route_multipath_add(&cfg, extack);
5519 return ip6_route_add(&cfg, GFP_KERNEL, extack);
5522 /* add the overhead of this fib6_nh to nexthop_len */
5523 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5525 int *nexthop_len = arg;
5527 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5528 + NLA_ALIGN(sizeof(struct rtnexthop))
5529 + nla_total_size(16); /* RTA_GATEWAY */
5531 if (nh->fib_nh_lws) {
5532 /* RTA_ENCAP_TYPE */
5533 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5535 *nexthop_len += nla_total_size(2);
5541 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5546 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5547 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5550 struct fib6_info *sibling, *next_sibling;
5551 struct fib6_nh *nh = f6i->fib6_nh;
5554 if (f6i->fib6_nsiblings) {
5555 rt6_nh_nlmsg_size(nh, &nexthop_len);
5557 list_for_each_entry_safe(sibling, next_sibling,
5558 &f6i->fib6_siblings, fib6_siblings) {
5559 rt6_nh_nlmsg_size(sibling->fib6_nh, &nexthop_len);
5562 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5565 return NLMSG_ALIGN(sizeof(struct rtmsg))
5566 + nla_total_size(16) /* RTA_SRC */
5567 + nla_total_size(16) /* RTA_DST */
5568 + nla_total_size(16) /* RTA_GATEWAY */
5569 + nla_total_size(16) /* RTA_PREFSRC */
5570 + nla_total_size(4) /* RTA_TABLE */
5571 + nla_total_size(4) /* RTA_IIF */
5572 + nla_total_size(4) /* RTA_OIF */
5573 + nla_total_size(4) /* RTA_PRIORITY */
5574 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5575 + nla_total_size(sizeof(struct rta_cacheinfo))
5576 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5577 + nla_total_size(1) /* RTA_PREF */
5581 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5582 unsigned char *flags)
5584 if (nexthop_is_multipath(nh)) {
5587 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5589 goto nla_put_failure;
5591 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5592 goto nla_put_failure;
5594 nla_nest_end(skb, mp);
5596 struct fib6_nh *fib6_nh;
5598 fib6_nh = nexthop_fib6_nh(nh);
5599 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5601 goto nla_put_failure;
5610 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5611 struct fib6_info *rt, struct dst_entry *dst,
5612 struct in6_addr *dest, struct in6_addr *src,
5613 int iif, int type, u32 portid, u32 seq,
5616 struct rt6_info *rt6 = (struct rt6_info *)dst;
5617 struct rt6key *rt6_dst, *rt6_src;
5618 u32 *pmetrics, table, rt6_flags;
5619 unsigned char nh_flags = 0;
5620 struct nlmsghdr *nlh;
5624 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5629 rt6_dst = &rt6->rt6i_dst;
5630 rt6_src = &rt6->rt6i_src;
5631 rt6_flags = rt6->rt6i_flags;
5633 rt6_dst = &rt->fib6_dst;
5634 rt6_src = &rt->fib6_src;
5635 rt6_flags = rt->fib6_flags;
5638 rtm = nlmsg_data(nlh);
5639 rtm->rtm_family = AF_INET6;
5640 rtm->rtm_dst_len = rt6_dst->plen;
5641 rtm->rtm_src_len = rt6_src->plen;
5644 table = rt->fib6_table->tb6_id;
5646 table = RT6_TABLE_UNSPEC;
5647 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5648 if (nla_put_u32(skb, RTA_TABLE, table))
5649 goto nla_put_failure;
5651 rtm->rtm_type = rt->fib6_type;
5653 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5654 rtm->rtm_protocol = rt->fib6_protocol;
5656 if (rt6_flags & RTF_CACHE)
5657 rtm->rtm_flags |= RTM_F_CLONED;
5660 if (nla_put_in6_addr(skb, RTA_DST, dest))
5661 goto nla_put_failure;
5662 rtm->rtm_dst_len = 128;
5663 } else if (rtm->rtm_dst_len)
5664 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5665 goto nla_put_failure;
5666 #ifdef CONFIG_IPV6_SUBTREES
5668 if (nla_put_in6_addr(skb, RTA_SRC, src))
5669 goto nla_put_failure;
5670 rtm->rtm_src_len = 128;
5671 } else if (rtm->rtm_src_len &&
5672 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5673 goto nla_put_failure;
5676 #ifdef CONFIG_IPV6_MROUTE
5677 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5678 int err = ip6mr_get_route(net, skb, rtm, portid);
5683 goto nla_put_failure;
5686 if (nla_put_u32(skb, RTA_IIF, iif))
5687 goto nla_put_failure;
5689 struct in6_addr saddr_buf;
5690 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
5691 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5692 goto nla_put_failure;
5695 if (rt->fib6_prefsrc.plen) {
5696 struct in6_addr saddr_buf;
5697 saddr_buf = rt->fib6_prefsrc.addr;
5698 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5699 goto nla_put_failure;
5702 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5703 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5704 goto nla_put_failure;
5706 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5707 goto nla_put_failure;
5709 /* For multipath routes, walk the siblings list and add
5710 * each as a nexthop within RTA_MULTIPATH.
5713 if (rt6_flags & RTF_GATEWAY &&
5714 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5715 goto nla_put_failure;
5717 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5718 goto nla_put_failure;
5720 if (dst->lwtstate &&
5721 lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5722 goto nla_put_failure;
5723 } else if (rt->fib6_nsiblings) {
5724 struct fib6_info *sibling, *next_sibling;
5727 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5729 goto nla_put_failure;
5731 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5732 rt->fib6_nh->fib_nh_weight, AF_INET6,
5734 goto nla_put_failure;
5736 list_for_each_entry_safe(sibling, next_sibling,
5737 &rt->fib6_siblings, fib6_siblings) {
5738 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5739 sibling->fib6_nh->fib_nh_weight,
5741 goto nla_put_failure;
5744 nla_nest_end(skb, mp);
5745 } else if (rt->nh) {
5746 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5747 goto nla_put_failure;
5749 if (nexthop_is_blackhole(rt->nh))
5750 rtm->rtm_type = RTN_BLACKHOLE;
5752 if (READ_ONCE(net->ipv4.sysctl_nexthop_compat_mode) &&
5753 rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5754 goto nla_put_failure;
5756 rtm->rtm_flags |= nh_flags;
5758 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5759 &nh_flags, false) < 0)
5760 goto nla_put_failure;
5762 rtm->rtm_flags |= nh_flags;
5765 if (rt6_flags & RTF_EXPIRES) {
5766 expires = dst ? dst->expires : rt->expires;
5771 if (READ_ONCE(rt->offload))
5772 rtm->rtm_flags |= RTM_F_OFFLOAD;
5773 if (READ_ONCE(rt->trap))
5774 rtm->rtm_flags |= RTM_F_TRAP;
5775 if (READ_ONCE(rt->offload_failed))
5776 rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5779 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5780 goto nla_put_failure;
5782 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5783 goto nla_put_failure;
5786 nlmsg_end(skb, nlh);
5790 nlmsg_cancel(skb, nlh);
5794 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5796 const struct net_device *dev = arg;
5798 if (nh->fib_nh_dev == dev)
5804 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5805 const struct net_device *dev)
5808 struct net_device *_dev = (struct net_device *)dev;
5810 return !!nexthop_for_each_fib6_nh(f6i->nh,
5811 fib6_info_nh_uses_dev,
5815 if (f6i->fib6_nh->fib_nh_dev == dev)
5818 if (f6i->fib6_nsiblings) {
5819 struct fib6_info *sibling, *next_sibling;
5821 list_for_each_entry_safe(sibling, next_sibling,
5822 &f6i->fib6_siblings, fib6_siblings) {
5823 if (sibling->fib6_nh->fib_nh_dev == dev)
5831 struct fib6_nh_exception_dump_walker {
5832 struct rt6_rtnl_dump_arg *dump;
5833 struct fib6_info *rt;
5839 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5841 struct fib6_nh_exception_dump_walker *w = arg;
5842 struct rt6_rtnl_dump_arg *dump = w->dump;
5843 struct rt6_exception_bucket *bucket;
5844 struct rt6_exception *rt6_ex;
5847 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5851 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5852 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5858 /* Expiration of entries doesn't bump sernum, insertion
5859 * does. Removal is triggered by insertion, so we can
5860 * rely on the fact that if entries change between two
5861 * partial dumps, this node is scanned again completely,
5862 * see rt6_insert_exception() and fib6_dump_table().
5864 * Count expired entries we go through as handled
5865 * entries that we'll skip next time, in case of partial
5866 * node dump. Otherwise, if entries expire meanwhile,
5867 * we'll skip the wrong amount.
5869 if (rt6_check_expired(rt6_ex->rt6i)) {
5874 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5875 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5877 NETLINK_CB(dump->cb->skb).portid,
5878 dump->cb->nlh->nlmsg_seq, w->flags);
5890 /* Return -1 if done with node, number of handled routes on partial dump */
5891 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5893 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5894 struct fib_dump_filter *filter = &arg->filter;
5895 unsigned int flags = NLM_F_MULTI;
5896 struct net *net = arg->net;
5899 if (rt == net->ipv6.fib6_null_entry)
5902 if ((filter->flags & RTM_F_PREFIX) &&
5903 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5904 /* success since this is not a prefix route */
5907 if (filter->filter_set &&
5908 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5909 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5910 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5914 if (filter->filter_set ||
5915 !filter->dump_routes || !filter->dump_exceptions) {
5916 flags |= NLM_F_DUMP_FILTERED;
5919 if (filter->dump_routes) {
5923 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5925 NETLINK_CB(arg->cb->skb).portid,
5926 arg->cb->nlh->nlmsg_seq, flags)) {
5933 if (filter->dump_exceptions) {
5934 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5943 err = nexthop_for_each_fib6_nh(rt->nh,
5944 rt6_nh_dump_exceptions,
5947 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5952 return count + w.count;
5958 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5959 const struct nlmsghdr *nlh,
5961 struct netlink_ext_ack *extack)
5966 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5967 NL_SET_ERR_MSG_MOD(extack,
5968 "Invalid header for get route request");
5972 if (!netlink_strict_get_check(skb))
5973 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5974 rtm_ipv6_policy, extack);
5976 rtm = nlmsg_data(nlh);
5977 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5978 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5979 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5981 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5984 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5985 NL_SET_ERR_MSG_MOD(extack,
5986 "Invalid flags for get route request");
5990 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5991 rtm_ipv6_policy, extack);
5995 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5996 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5997 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
6001 for (i = 0; i <= RTA_MAX; i++) {
6017 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
6025 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6026 struct netlink_ext_ack *extack)
6028 struct net *net = sock_net(in_skb->sk);
6029 struct nlattr *tb[RTA_MAX+1];
6030 int err, iif = 0, oif = 0;
6031 struct fib6_info *from;
6032 struct dst_entry *dst;
6033 struct rt6_info *rt;
6034 struct sk_buff *skb;
6036 struct flowi6 fl6 = {};
6039 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
6044 rtm = nlmsg_data(nlh);
6045 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
6046 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
6049 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
6052 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
6056 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
6059 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
6063 iif = nla_get_u32(tb[RTA_IIF]);
6066 oif = nla_get_u32(tb[RTA_OIF]);
6069 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6072 fl6.flowi6_uid = make_kuid(current_user_ns(),
6073 nla_get_u32(tb[RTA_UID]));
6075 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6078 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6081 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6083 if (tb[RTA_IP_PROTO]) {
6084 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
6085 &fl6.flowi6_proto, AF_INET6,
6092 struct net_device *dev;
6097 dev = dev_get_by_index_rcu(net, iif);
6104 fl6.flowi6_iif = iif;
6106 if (!ipv6_addr_any(&fl6.saddr))
6107 flags |= RT6_LOOKUP_F_HAS_SADDR;
6109 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
6113 fl6.flowi6_oif = oif;
6115 dst = ip6_route_output(net, NULL, &fl6);
6119 rt = container_of(dst, struct rt6_info, dst);
6120 if (rt->dst.error) {
6121 err = rt->dst.error;
6126 if (rt == net->ipv6.ip6_null_entry) {
6127 err = rt->dst.error;
6132 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
6139 skb_dst_set(skb, &rt->dst);
6142 from = rcu_dereference(rt->from);
6145 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6147 NETLINK_CB(in_skb).portid,
6150 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6151 &fl6.saddr, iif, RTM_NEWROUTE,
6152 NETLINK_CB(in_skb).portid,
6164 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6169 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6170 unsigned int nlm_flags)
6172 struct sk_buff *skb;
6173 struct net *net = info->nl_net;
6178 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6180 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6184 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6185 event, info->portid, seq, nlm_flags);
6187 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6188 WARN_ON(err == -EMSGSIZE);
6192 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6193 info->nlh, gfp_any());
6197 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6200 void fib6_rt_update(struct net *net, struct fib6_info *rt,
6201 struct nl_info *info)
6203 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6204 struct sk_buff *skb;
6207 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6211 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6212 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6214 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6215 WARN_ON(err == -EMSGSIZE);
6219 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6220 info->nlh, gfp_any());
6224 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6227 void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6228 bool offload, bool trap, bool offload_failed)
6230 struct sk_buff *skb;
6233 if (READ_ONCE(f6i->offload) == offload &&
6234 READ_ONCE(f6i->trap) == trap &&
6235 READ_ONCE(f6i->offload_failed) == offload_failed)
6238 WRITE_ONCE(f6i->offload, offload);
6239 WRITE_ONCE(f6i->trap, trap);
6241 /* 2 means send notifications only if offload_failed was changed. */
6242 if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6243 READ_ONCE(f6i->offload_failed) == offload_failed)
6246 WRITE_ONCE(f6i->offload_failed, offload_failed);
6248 if (!rcu_access_pointer(f6i->fib6_node))
6249 /* The route was removed from the tree, do not send
6254 if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6257 skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6263 err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6266 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6267 WARN_ON(err == -EMSGSIZE);
6272 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6276 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6278 EXPORT_SYMBOL(fib6_info_hw_flags_set);
6280 static int ip6_route_dev_notify(struct notifier_block *this,
6281 unsigned long event, void *ptr)
6283 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6284 struct net *net = dev_net(dev);
6286 if (!(dev->flags & IFF_LOOPBACK))
6289 if (event == NETDEV_REGISTER) {
6290 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6291 net->ipv6.ip6_null_entry->dst.dev = dev;
6292 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6293 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6294 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6295 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6296 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6297 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6299 } else if (event == NETDEV_UNREGISTER &&
6300 dev->reg_state != NETREG_UNREGISTERED) {
6301 /* NETDEV_UNREGISTER could be fired for multiple times by
6302 * netdev_wait_allrefs(). Make sure we only call this once.
6304 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6305 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6306 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6307 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6318 #ifdef CONFIG_PROC_FS
6319 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6321 struct net *net = (struct net *)seq->private;
6322 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6323 net->ipv6.rt6_stats->fib_nodes,
6324 net->ipv6.rt6_stats->fib_route_nodes,
6325 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6326 net->ipv6.rt6_stats->fib_rt_entries,
6327 net->ipv6.rt6_stats->fib_rt_cache,
6328 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6329 net->ipv6.rt6_stats->fib_discarded_routes);
6333 #endif /* CONFIG_PROC_FS */
6335 #ifdef CONFIG_SYSCTL
6337 static int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6338 void *buffer, size_t *lenp, loff_t *ppos)
6346 net = (struct net *)ctl->extra1;
6347 delay = net->ipv6.sysctl.flush_delay;
6348 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6352 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6356 static struct ctl_table ipv6_route_table_template[] = {
6358 .procname = "max_size",
6359 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
6360 .maxlen = sizeof(int),
6362 .proc_handler = proc_dointvec,
6365 .procname = "gc_thresh",
6366 .data = &ip6_dst_ops_template.gc_thresh,
6367 .maxlen = sizeof(int),
6369 .proc_handler = proc_dointvec,
6372 .procname = "flush",
6373 .data = &init_net.ipv6.sysctl.flush_delay,
6374 .maxlen = sizeof(int),
6376 .proc_handler = ipv6_sysctl_rtcache_flush
6379 .procname = "gc_min_interval",
6380 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6381 .maxlen = sizeof(int),
6383 .proc_handler = proc_dointvec_jiffies,
6386 .procname = "gc_timeout",
6387 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6388 .maxlen = sizeof(int),
6390 .proc_handler = proc_dointvec_jiffies,
6393 .procname = "gc_interval",
6394 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
6395 .maxlen = sizeof(int),
6397 .proc_handler = proc_dointvec_jiffies,
6400 .procname = "gc_elasticity",
6401 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6402 .maxlen = sizeof(int),
6404 .proc_handler = proc_dointvec,
6407 .procname = "mtu_expires",
6408 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6409 .maxlen = sizeof(int),
6411 .proc_handler = proc_dointvec_jiffies,
6414 .procname = "min_adv_mss",
6415 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
6416 .maxlen = sizeof(int),
6418 .proc_handler = proc_dointvec,
6421 .procname = "gc_min_interval_ms",
6422 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6423 .maxlen = sizeof(int),
6425 .proc_handler = proc_dointvec_ms_jiffies,
6428 .procname = "skip_notify_on_dev_down",
6429 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6430 .maxlen = sizeof(u8),
6432 .proc_handler = proc_dou8vec_minmax,
6433 .extra1 = SYSCTL_ZERO,
6434 .extra2 = SYSCTL_ONE,
6439 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6441 struct ctl_table *table;
6443 table = kmemdup(ipv6_route_table_template,
6444 sizeof(ipv6_route_table_template),
6448 table[0].data = &net->ipv6.sysctl.ip6_rt_max_size;
6449 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6450 table[2].data = &net->ipv6.sysctl.flush_delay;
6451 table[2].extra1 = net;
6452 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6453 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6454 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6455 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6456 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6457 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6458 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6459 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6461 /* Don't export sysctls to unprivileged users */
6462 if (net->user_ns != &init_user_ns)
6463 table[1].procname = NULL;
6469 size_t ipv6_route_sysctl_table_size(struct net *net)
6471 /* Don't export sysctls to unprivileged users */
6472 if (net->user_ns != &init_user_ns)
6475 return ARRAY_SIZE(ipv6_route_table_template);
6479 static int __net_init ip6_route_net_init(struct net *net)
6483 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6484 sizeof(net->ipv6.ip6_dst_ops));
6486 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6487 goto out_ip6_dst_ops;
6489 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6490 if (!net->ipv6.fib6_null_entry)
6491 goto out_ip6_dst_entries;
6492 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6493 sizeof(*net->ipv6.fib6_null_entry));
6495 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6496 sizeof(*net->ipv6.ip6_null_entry),
6498 if (!net->ipv6.ip6_null_entry)
6499 goto out_fib6_null_entry;
6500 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6501 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6502 ip6_template_metrics, true);
6503 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->dst.rt_uncached);
6505 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6506 net->ipv6.fib6_has_custom_rules = false;
6507 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6508 sizeof(*net->ipv6.ip6_prohibit_entry),
6510 if (!net->ipv6.ip6_prohibit_entry)
6511 goto out_ip6_null_entry;
6512 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6513 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6514 ip6_template_metrics, true);
6515 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->dst.rt_uncached);
6517 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6518 sizeof(*net->ipv6.ip6_blk_hole_entry),
6520 if (!net->ipv6.ip6_blk_hole_entry)
6521 goto out_ip6_prohibit_entry;
6522 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6523 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6524 ip6_template_metrics, true);
6525 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->dst.rt_uncached);
6526 #ifdef CONFIG_IPV6_SUBTREES
6527 net->ipv6.fib6_routes_require_src = 0;
6531 net->ipv6.sysctl.flush_delay = 0;
6532 net->ipv6.sysctl.ip6_rt_max_size = INT_MAX;
6533 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6534 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6535 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6536 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6537 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6538 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6539 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6541 atomic_set(&net->ipv6.ip6_rt_gc_expire, 30*HZ);
6547 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6548 out_ip6_prohibit_entry:
6549 kfree(net->ipv6.ip6_prohibit_entry);
6551 kfree(net->ipv6.ip6_null_entry);
6553 out_fib6_null_entry:
6554 kfree(net->ipv6.fib6_null_entry);
6555 out_ip6_dst_entries:
6556 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6561 static void __net_exit ip6_route_net_exit(struct net *net)
6563 kfree(net->ipv6.fib6_null_entry);
6564 kfree(net->ipv6.ip6_null_entry);
6565 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6566 kfree(net->ipv6.ip6_prohibit_entry);
6567 kfree(net->ipv6.ip6_blk_hole_entry);
6569 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6572 static int __net_init ip6_route_net_init_late(struct net *net)
6574 #ifdef CONFIG_PROC_FS
6575 if (!proc_create_net("ipv6_route", 0, net->proc_net,
6576 &ipv6_route_seq_ops,
6577 sizeof(struct ipv6_route_iter)))
6580 if (!proc_create_net_single("rt6_stats", 0444, net->proc_net,
6581 rt6_stats_seq_show, NULL)) {
6582 remove_proc_entry("ipv6_route", net->proc_net);
6589 static void __net_exit ip6_route_net_exit_late(struct net *net)
6591 #ifdef CONFIG_PROC_FS
6592 remove_proc_entry("ipv6_route", net->proc_net);
6593 remove_proc_entry("rt6_stats", net->proc_net);
6597 static struct pernet_operations ip6_route_net_ops = {
6598 .init = ip6_route_net_init,
6599 .exit = ip6_route_net_exit,
6602 static int __net_init ipv6_inetpeer_init(struct net *net)
6604 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6608 inet_peer_base_init(bp);
6609 net->ipv6.peers = bp;
6613 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6615 struct inet_peer_base *bp = net->ipv6.peers;
6617 net->ipv6.peers = NULL;
6618 inetpeer_invalidate_tree(bp);
6622 static struct pernet_operations ipv6_inetpeer_ops = {
6623 .init = ipv6_inetpeer_init,
6624 .exit = ipv6_inetpeer_exit,
6627 static struct pernet_operations ip6_route_net_late_ops = {
6628 .init = ip6_route_net_init_late,
6629 .exit = ip6_route_net_exit_late,
6632 static struct notifier_block ip6_route_dev_notifier = {
6633 .notifier_call = ip6_route_dev_notify,
6634 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6637 void __init ip6_route_init_special_entries(void)
6639 /* Registering of the loopback is done before this portion of code,
6640 * the loopback reference in rt6_info will not be taken, do it
6641 * manually for init_net */
6642 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6643 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6644 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6645 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6646 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6647 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6648 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6649 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6653 #if IS_BUILTIN(CONFIG_IPV6)
6654 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6655 DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6657 BTF_ID_LIST(btf_fib6_info_id)
6658 BTF_ID(struct, fib6_info)
6660 static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6661 .seq_ops = &ipv6_route_seq_ops,
6662 .init_seq_private = bpf_iter_init_seq_net,
6663 .fini_seq_private = bpf_iter_fini_seq_net,
6664 .seq_priv_size = sizeof(struct ipv6_route_iter),
6667 static struct bpf_iter_reg ipv6_route_reg_info = {
6668 .target = "ipv6_route",
6669 .ctx_arg_info_size = 1,
6671 { offsetof(struct bpf_iter__ipv6_route, rt),
6672 PTR_TO_BTF_ID_OR_NULL },
6674 .seq_info = &ipv6_route_seq_info,
6677 static int __init bpf_iter_register(void)
6679 ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6680 return bpf_iter_reg_target(&ipv6_route_reg_info);
6683 static void bpf_iter_unregister(void)
6685 bpf_iter_unreg_target(&ipv6_route_reg_info);
6690 int __init ip6_route_init(void)
6696 ip6_dst_ops_template.kmem_cachep =
6697 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6698 SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL);
6699 if (!ip6_dst_ops_template.kmem_cachep)
6702 ret = dst_entries_init(&ip6_dst_blackhole_ops);
6704 goto out_kmem_cache;
6706 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6708 goto out_dst_entries;
6710 ret = register_pernet_subsys(&ip6_route_net_ops);
6712 goto out_register_inetpeer;
6714 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6718 goto out_register_subsys;
6724 ret = fib6_rules_init();
6728 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6730 goto fib6_rules_init;
6732 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6733 inet6_rtm_newroute, NULL, 0);
6735 goto out_register_late_subsys;
6737 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6738 inet6_rtm_delroute, NULL, 0);
6740 goto out_register_late_subsys;
6742 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6743 inet6_rtm_getroute, NULL,
6744 RTNL_FLAG_DOIT_UNLOCKED);
6746 goto out_register_late_subsys;
6748 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6750 goto out_register_late_subsys;
6752 #if IS_BUILTIN(CONFIG_IPV6)
6753 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6754 ret = bpf_iter_register();
6756 goto out_register_late_subsys;
6760 for_each_possible_cpu(cpu) {
6761 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6763 INIT_LIST_HEAD(&ul->head);
6764 INIT_LIST_HEAD(&ul->quarantine);
6765 spin_lock_init(&ul->lock);
6771 out_register_late_subsys:
6772 rtnl_unregister_all(PF_INET6);
6773 unregister_pernet_subsys(&ip6_route_net_late_ops);
6775 fib6_rules_cleanup();
6780 out_register_subsys:
6781 unregister_pernet_subsys(&ip6_route_net_ops);
6782 out_register_inetpeer:
6783 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6785 dst_entries_destroy(&ip6_dst_blackhole_ops);
6787 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6791 void ip6_route_cleanup(void)
6793 #if IS_BUILTIN(CONFIG_IPV6)
6794 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6795 bpf_iter_unregister();
6798 unregister_netdevice_notifier(&ip6_route_dev_notifier);
6799 unregister_pernet_subsys(&ip6_route_net_late_ops);
6800 fib6_rules_cleanup();
6803 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6804 unregister_pernet_subsys(&ip6_route_net_ops);
6805 dst_entries_destroy(&ip6_dst_blackhole_ops);
6806 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);