1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Linux INET6 implementation
7 * Pedro Roque <roque@di.fc.ul.pt>
12 * YOSHIFUJI Hideaki @USAGI
13 * reworked default router selection.
14 * - respect outgoing interface
15 * - select from (probably) reachable routers (i.e.
16 * routers in REACHABLE, STALE, DELAY or PROBE states).
17 * - always select the same router if it is (probably)
18 * reachable. otherwise, round-robin the list.
20 * Fixed routing subtrees.
23 #define pr_fmt(fmt) "IPv6: " fmt
25 #include <linux/capability.h>
26 #include <linux/errno.h>
27 #include <linux/export.h>
28 #include <linux/types.h>
29 #include <linux/times.h>
30 #include <linux/socket.h>
31 #include <linux/sockios.h>
32 #include <linux/net.h>
33 #include <linux/route.h>
34 #include <linux/netdevice.h>
35 #include <linux/in6.h>
36 #include <linux/mroute6.h>
37 #include <linux/init.h>
38 #include <linux/if_arp.h>
39 #include <linux/proc_fs.h>
40 #include <linux/seq_file.h>
41 #include <linux/nsproxy.h>
42 #include <linux/slab.h>
43 #include <linux/jhash.h>
44 #include <linux/siphash.h>
45 #include <net/net_namespace.h>
48 #include <net/ip6_fib.h>
49 #include <net/ip6_route.h>
50 #include <net/ndisc.h>
51 #include <net/addrconf.h>
53 #include <linux/rtnetlink.h>
55 #include <net/dst_metadata.h>
57 #include <net/netevent.h>
58 #include <net/netlink.h>
60 #include <net/lwtunnel.h>
61 #include <net/ip_tunnels.h>
62 #include <net/l3mdev.h>
64 #include <linux/uaccess.h>
65 #include <linux/btf_ids.h>
68 #include <linux/sysctl.h>
71 static int ip6_rt_type_to_error(u8 fib6_type);
73 #define CREATE_TRACE_POINTS
74 #include <trace/events/fib6.h>
75 EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
76 #undef CREATE_TRACE_POINTS
79 RT6_NUD_FAIL_HARD = -3,
80 RT6_NUD_FAIL_PROBE = -2,
81 RT6_NUD_FAIL_DO_RR = -1,
85 INDIRECT_CALLABLE_SCOPE
86 struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
87 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
88 INDIRECT_CALLABLE_SCOPE
89 unsigned int ip6_mtu(const struct dst_entry *dst);
90 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
91 static void ip6_dst_destroy(struct dst_entry *);
92 static void ip6_dst_ifdown(struct dst_entry *,
93 struct net_device *dev, int how);
94 static int ip6_dst_gc(struct dst_ops *ops);
96 static int ip6_pkt_discard(struct sk_buff *skb);
97 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
98 static int ip6_pkt_prohibit(struct sk_buff *skb);
99 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
100 static void ip6_link_failure(struct sk_buff *skb);
101 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
102 struct sk_buff *skb, u32 mtu,
104 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
105 struct sk_buff *skb);
106 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
108 static size_t rt6_nlmsg_size(struct fib6_info *f6i);
109 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
110 struct fib6_info *rt, struct dst_entry *dst,
111 struct in6_addr *dest, struct in6_addr *src,
112 int iif, int type, u32 portid, u32 seq,
114 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
115 const struct in6_addr *daddr,
116 const struct in6_addr *saddr);
118 #ifdef CONFIG_IPV6_ROUTE_INFO
119 static struct fib6_info *rt6_add_route_info(struct net *net,
120 const struct in6_addr *prefix, int prefixlen,
121 const struct in6_addr *gwaddr,
122 struct net_device *dev,
124 static struct fib6_info *rt6_get_route_info(struct net *net,
125 const struct in6_addr *prefix, int prefixlen,
126 const struct in6_addr *gwaddr,
127 struct net_device *dev);
130 struct uncached_list {
132 struct list_head head;
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->rt6i_uncached_list = ul;
144 spin_lock_bh(&ul->lock);
145 list_add_tail(&rt->rt6i_uncached, &ul->head);
146 spin_unlock_bh(&ul->lock);
149 void rt6_uncached_list_del(struct rt6_info *rt)
151 if (!list_empty(&rt->rt6i_uncached)) {
152 struct uncached_list *ul = rt->rt6i_uncached_list;
154 spin_lock_bh(&ul->lock);
155 list_del_init(&rt->rt6i_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, rt6i_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 dev_replace_track(rt_dev, blackhole_netdev,
186 &rt->dst.dev_tracker,
191 list_move(&rt->rt6i_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 .__refcnt = ATOMIC_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 .__refcnt = ATOMIC_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 .__refcnt = ATOMIC_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);
337 INIT_LIST_HEAD(&rt->rt6i_uncached);
340 /* allocate dst with ip6_dst_ops */
341 struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
344 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
345 1, DST_OBSOLETE_FORCE_CHK, flags);
349 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
354 EXPORT_SYMBOL(ip6_dst_alloc);
356 static void ip6_dst_destroy(struct dst_entry *dst)
358 struct rt6_info *rt = (struct rt6_info *)dst;
359 struct fib6_info *from;
360 struct inet6_dev *idev;
362 ip_dst_metrics_put(dst);
363 rt6_uncached_list_del(rt);
365 idev = rt->rt6i_idev;
367 rt->rt6i_idev = NULL;
371 from = xchg((__force struct fib6_info **)&rt->from, NULL);
372 fib6_info_release(from);
375 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
378 struct rt6_info *rt = (struct rt6_info *)dst;
379 struct inet6_dev *idev = rt->rt6i_idev;
381 if (idev && idev->dev != blackhole_netdev) {
382 struct inet6_dev *blackhole_idev = in6_dev_get(blackhole_netdev);
384 if (blackhole_idev) {
385 rt->rt6i_idev = blackhole_idev;
391 static bool __rt6_check_expired(const struct rt6_info *rt)
393 if (rt->rt6i_flags & RTF_EXPIRES)
394 return time_after(jiffies, rt->dst.expires);
399 static bool rt6_check_expired(const struct rt6_info *rt)
401 struct fib6_info *from;
403 from = rcu_dereference(rt->from);
405 if (rt->rt6i_flags & RTF_EXPIRES) {
406 if (time_after(jiffies, rt->dst.expires))
409 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
410 fib6_check_expired(from);
415 void fib6_select_path(const struct net *net, struct fib6_result *res,
416 struct flowi6 *fl6, int oif, bool have_oif_match,
417 const struct sk_buff *skb, int strict)
419 struct fib6_info *sibling, *next_sibling;
420 struct fib6_info *match = res->f6i;
422 if (!match->nh && (!match->fib6_nsiblings || have_oif_match))
425 if (match->nh && have_oif_match && res->nh)
428 /* We might have already computed the hash for ICMPv6 errors. In such
429 * case it will always be non-zero. Otherwise now is the time to do it.
432 (!match->nh || nexthop_is_multipath(match->nh)))
433 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
435 if (unlikely(match->nh)) {
436 nexthop_path_fib6_result(res, fl6->mp_hash);
440 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
443 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
445 const struct fib6_nh *nh = sibling->fib6_nh;
448 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
449 if (fl6->mp_hash > nh_upper_bound)
451 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
459 res->nh = match->fib6_nh;
463 * Route lookup. rcu_read_lock() should be held.
466 static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
467 const struct in6_addr *saddr, int oif, int flags)
469 const struct net_device *dev;
471 if (nh->fib_nh_flags & RTNH_F_DEAD)
474 dev = nh->fib_nh_dev;
476 if (dev->ifindex == oif)
479 if (ipv6_chk_addr(net, saddr, dev,
480 flags & RT6_LOOKUP_F_IFACE))
487 struct fib6_nh_dm_arg {
489 const struct in6_addr *saddr;
495 static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
497 struct fib6_nh_dm_arg *arg = _arg;
500 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
504 /* returns fib6_nh from nexthop or NULL */
505 static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
506 struct fib6_result *res,
507 const struct in6_addr *saddr,
510 struct fib6_nh_dm_arg arg = {
517 if (nexthop_is_blackhole(nh))
520 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
526 static void rt6_device_match(struct net *net, struct fib6_result *res,
527 const struct in6_addr *saddr, int oif, int flags)
529 struct fib6_info *f6i = res->f6i;
530 struct fib6_info *spf6i;
533 if (!oif && ipv6_addr_any(saddr)) {
534 if (unlikely(f6i->nh)) {
535 nh = nexthop_fib6_nh(f6i->nh);
536 if (nexthop_is_blackhole(f6i->nh))
541 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
545 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
546 bool matched = false;
548 if (unlikely(spf6i->nh)) {
549 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
555 if (__rt6_device_match(net, nh, saddr, oif, flags))
564 if (oif && flags & RT6_LOOKUP_F_IFACE) {
565 res->f6i = net->ipv6.fib6_null_entry;
566 nh = res->f6i->fib6_nh;
570 if (unlikely(f6i->nh)) {
571 nh = nexthop_fib6_nh(f6i->nh);
572 if (nexthop_is_blackhole(f6i->nh))
578 if (nh->fib_nh_flags & RTNH_F_DEAD) {
579 res->f6i = net->ipv6.fib6_null_entry;
580 nh = res->f6i->fib6_nh;
584 res->fib6_type = res->f6i->fib6_type;
585 res->fib6_flags = res->f6i->fib6_flags;
589 res->fib6_flags |= RTF_REJECT;
590 res->fib6_type = RTN_BLACKHOLE;
594 #ifdef CONFIG_IPV6_ROUTER_PREF
595 struct __rt6_probe_work {
596 struct work_struct work;
597 struct in6_addr target;
598 struct net_device *dev;
599 netdevice_tracker dev_tracker;
602 static void rt6_probe_deferred(struct work_struct *w)
604 struct in6_addr mcaddr;
605 struct __rt6_probe_work *work =
606 container_of(w, struct __rt6_probe_work, work);
608 addrconf_addr_solict_mult(&work->target, &mcaddr);
609 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
610 dev_put_track(work->dev, &work->dev_tracker);
614 static void rt6_probe(struct fib6_nh *fib6_nh)
616 struct __rt6_probe_work *work = NULL;
617 const struct in6_addr *nh_gw;
618 unsigned long last_probe;
619 struct neighbour *neigh;
620 struct net_device *dev;
621 struct inet6_dev *idev;
624 * Okay, this does not seem to be appropriate
625 * for now, however, we need to check if it
626 * is really so; aka Router Reachability Probing.
628 * Router Reachability Probe MUST be rate-limited
629 * to no more than one per minute.
631 if (!fib6_nh->fib_nh_gw_family)
634 nh_gw = &fib6_nh->fib_nh_gw6;
635 dev = fib6_nh->fib_nh_dev;
637 last_probe = READ_ONCE(fib6_nh->last_probe);
638 idev = __in6_dev_get(dev);
639 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
641 if (neigh->nud_state & NUD_VALID)
644 write_lock(&neigh->lock);
645 if (!(neigh->nud_state & NUD_VALID) &&
647 neigh->updated + idev->cnf.rtr_probe_interval)) {
648 work = kmalloc(sizeof(*work), GFP_ATOMIC);
650 __neigh_set_probe_once(neigh);
652 write_unlock(&neigh->lock);
653 } else if (time_after(jiffies, last_probe +
654 idev->cnf.rtr_probe_interval)) {
655 work = kmalloc(sizeof(*work), GFP_ATOMIC);
658 if (!work || cmpxchg(&fib6_nh->last_probe,
659 last_probe, jiffies) != last_probe) {
662 INIT_WORK(&work->work, rt6_probe_deferred);
663 work->target = *nh_gw;
664 dev_hold_track(dev, &work->dev_tracker, GFP_ATOMIC);
666 schedule_work(&work->work);
670 rcu_read_unlock_bh();
673 static inline void rt6_probe(struct fib6_nh *fib6_nh)
679 * Default Router Selection (RFC 2461 6.3.6)
681 static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
683 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
684 struct neighbour *neigh;
687 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
688 &fib6_nh->fib_nh_gw6);
690 read_lock(&neigh->lock);
691 if (neigh->nud_state & NUD_VALID)
692 ret = RT6_NUD_SUCCEED;
693 #ifdef CONFIG_IPV6_ROUTER_PREF
694 else if (!(neigh->nud_state & NUD_FAILED))
695 ret = RT6_NUD_SUCCEED;
697 ret = RT6_NUD_FAIL_PROBE;
699 read_unlock(&neigh->lock);
701 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
702 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
704 rcu_read_unlock_bh();
709 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
714 if (!oif || nh->fib_nh_dev->ifindex == oif)
717 if (!m && (strict & RT6_LOOKUP_F_IFACE))
718 return RT6_NUD_FAIL_HARD;
719 #ifdef CONFIG_IPV6_ROUTER_PREF
720 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
722 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
723 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
724 int n = rt6_check_neigh(nh);
731 static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
732 int oif, int strict, int *mpri, bool *do_rr)
734 bool match_do_rr = false;
738 if (nh->fib_nh_flags & RTNH_F_DEAD)
741 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
742 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
743 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
746 m = rt6_score_route(nh, fib6_flags, oif, strict);
747 if (m == RT6_NUD_FAIL_DO_RR) {
749 m = 0; /* lowest valid score */
750 } else if (m == RT6_NUD_FAIL_HARD) {
754 if (strict & RT6_LOOKUP_F_REACHABLE)
757 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
759 *do_rr = match_do_rr;
767 struct fib6_nh_frl_arg {
776 static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
778 struct fib6_nh_frl_arg *arg = _arg;
781 return find_match(nh, arg->flags, arg->oif, arg->strict,
782 arg->mpri, arg->do_rr);
785 static void __find_rr_leaf(struct fib6_info *f6i_start,
786 struct fib6_info *nomatch, u32 metric,
787 struct fib6_result *res, struct fib6_info **cont,
788 int oif, int strict, bool *do_rr, int *mpri)
790 struct fib6_info *f6i;
792 for (f6i = f6i_start;
793 f6i && f6i != nomatch;
794 f6i = rcu_dereference(f6i->fib6_next)) {
795 bool matched = false;
798 if (cont && f6i->fib6_metric != metric) {
803 if (fib6_check_expired(f6i))
806 if (unlikely(f6i->nh)) {
807 struct fib6_nh_frl_arg arg = {
808 .flags = f6i->fib6_flags,
815 if (nexthop_is_blackhole(f6i->nh)) {
816 res->fib6_flags = RTF_REJECT;
817 res->fib6_type = RTN_BLACKHOLE;
819 res->nh = nexthop_fib6_nh(f6i->nh);
822 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
829 if (find_match(nh, f6i->fib6_flags, oif, strict,
836 res->fib6_flags = f6i->fib6_flags;
837 res->fib6_type = f6i->fib6_type;
842 static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
843 struct fib6_info *rr_head, int oif, int strict,
844 bool *do_rr, struct fib6_result *res)
846 u32 metric = rr_head->fib6_metric;
847 struct fib6_info *cont = NULL;
850 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
851 oif, strict, do_rr, &mpri);
853 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
854 oif, strict, do_rr, &mpri);
856 if (res->f6i || !cont)
859 __find_rr_leaf(cont, NULL, metric, res, NULL,
860 oif, strict, do_rr, &mpri);
863 static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
864 struct fib6_result *res, int strict)
866 struct fib6_info *leaf = rcu_dereference(fn->leaf);
867 struct fib6_info *rt0;
871 /* make sure this function or its helpers sets f6i */
874 if (!leaf || leaf == net->ipv6.fib6_null_entry)
877 rt0 = rcu_dereference(fn->rr_ptr);
881 /* Double check to make sure fn is not an intermediate node
882 * and fn->leaf does not points to its child's leaf
883 * (This might happen if all routes under fn are deleted from
884 * the tree and fib6_repair_tree() is called on the node.)
886 key_plen = rt0->fib6_dst.plen;
887 #ifdef CONFIG_IPV6_SUBTREES
888 if (rt0->fib6_src.plen)
889 key_plen = rt0->fib6_src.plen;
891 if (fn->fn_bit != key_plen)
894 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
896 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
898 /* no entries matched; do round-robin */
899 if (!next || next->fib6_metric != rt0->fib6_metric)
903 spin_lock_bh(&leaf->fib6_table->tb6_lock);
904 /* make sure next is not being deleted from the tree */
906 rcu_assign_pointer(fn->rr_ptr, next);
907 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
913 res->f6i = net->ipv6.fib6_null_entry;
914 res->nh = res->f6i->fib6_nh;
915 res->fib6_flags = res->f6i->fib6_flags;
916 res->fib6_type = res->f6i->fib6_type;
920 static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
922 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
923 res->nh->fib_nh_gw_family;
926 #ifdef CONFIG_IPV6_ROUTE_INFO
927 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
928 const struct in6_addr *gwaddr)
930 struct net *net = dev_net(dev);
931 struct route_info *rinfo = (struct route_info *) opt;
932 struct in6_addr prefix_buf, *prefix;
934 unsigned long lifetime;
935 struct fib6_info *rt;
937 if (len < sizeof(struct route_info)) {
941 /* Sanity check for prefix_len and length */
942 if (rinfo->length > 3) {
944 } else if (rinfo->prefix_len > 128) {
946 } else if (rinfo->prefix_len > 64) {
947 if (rinfo->length < 2) {
950 } else if (rinfo->prefix_len > 0) {
951 if (rinfo->length < 1) {
956 pref = rinfo->route_pref;
957 if (pref == ICMPV6_ROUTER_PREF_INVALID)
960 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
962 if (rinfo->length == 3)
963 prefix = (struct in6_addr *)rinfo->prefix;
965 /* this function is safe */
966 ipv6_addr_prefix(&prefix_buf,
967 (struct in6_addr *)rinfo->prefix,
969 prefix = &prefix_buf;
972 if (rinfo->prefix_len == 0)
973 rt = rt6_get_dflt_router(net, gwaddr, dev);
975 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
978 if (rt && !lifetime) {
979 ip6_del_rt(net, rt, false);
984 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
987 rt->fib6_flags = RTF_ROUTEINFO |
988 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
991 if (!addrconf_finite_timeout(lifetime))
992 fib6_clean_expires(rt);
994 fib6_set_expires(rt, jiffies + HZ * lifetime);
996 fib6_info_release(rt);
1003 * Misc support functions
1006 /* called with rcu_lock held */
1007 static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1009 struct net_device *dev = res->nh->fib_nh_dev;
1011 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1012 /* for copies of local routes, dst->dev needs to be the
1013 * device if it is a master device, the master device if
1014 * device is enslaved, and the loopback as the default
1016 if (netif_is_l3_slave(dev) &&
1017 !rt6_need_strict(&res->f6i->fib6_dst.addr))
1018 dev = l3mdev_master_dev_rcu(dev);
1019 else if (!netif_is_l3_master(dev))
1020 dev = dev_net(dev)->loopback_dev;
1021 /* last case is netif_is_l3_master(dev) is true in which
1022 * case we want dev returned to be dev
1029 static const int fib6_prop[RTN_MAX + 1] = {
1033 [RTN_BROADCAST] = 0,
1035 [RTN_MULTICAST] = 0,
1036 [RTN_BLACKHOLE] = -EINVAL,
1037 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1038 [RTN_PROHIBIT] = -EACCES,
1039 [RTN_THROW] = -EAGAIN,
1040 [RTN_NAT] = -EINVAL,
1041 [RTN_XRESOLVE] = -EINVAL,
1044 static int ip6_rt_type_to_error(u8 fib6_type)
1046 return fib6_prop[fib6_type];
1049 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1051 unsigned short flags = 0;
1053 if (rt->dst_nocount)
1054 flags |= DST_NOCOUNT;
1055 if (rt->dst_nopolicy)
1056 flags |= DST_NOPOLICY;
1061 static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1063 rt->dst.error = ip6_rt_type_to_error(fib6_type);
1065 switch (fib6_type) {
1067 rt->dst.output = dst_discard_out;
1068 rt->dst.input = dst_discard;
1071 rt->dst.output = ip6_pkt_prohibit_out;
1072 rt->dst.input = ip6_pkt_prohibit;
1075 case RTN_UNREACHABLE:
1077 rt->dst.output = ip6_pkt_discard_out;
1078 rt->dst.input = ip6_pkt_discard;
1083 static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1085 struct fib6_info *f6i = res->f6i;
1087 if (res->fib6_flags & RTF_REJECT) {
1088 ip6_rt_init_dst_reject(rt, res->fib6_type);
1093 rt->dst.output = ip6_output;
1095 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1096 rt->dst.input = ip6_input;
1097 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1098 rt->dst.input = ip6_mc_input;
1100 rt->dst.input = ip6_forward;
1103 if (res->nh->fib_nh_lws) {
1104 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1105 lwtunnel_set_redirect(&rt->dst);
1108 rt->dst.lastuse = jiffies;
1111 /* Caller must already hold reference to @from */
1112 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1114 rt->rt6i_flags &= ~RTF_EXPIRES;
1115 rcu_assign_pointer(rt->from, from);
1116 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1119 /* Caller must already hold reference to f6i in result */
1120 static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1122 const struct fib6_nh *nh = res->nh;
1123 const struct net_device *dev = nh->fib_nh_dev;
1124 struct fib6_info *f6i = res->f6i;
1126 ip6_rt_init_dst(rt, res);
1128 rt->rt6i_dst = f6i->fib6_dst;
1129 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1130 rt->rt6i_flags = res->fib6_flags;
1131 if (nh->fib_nh_gw_family) {
1132 rt->rt6i_gateway = nh->fib_nh_gw6;
1133 rt->rt6i_flags |= RTF_GATEWAY;
1135 rt6_set_from(rt, f6i);
1136 #ifdef CONFIG_IPV6_SUBTREES
1137 rt->rt6i_src = f6i->fib6_src;
1141 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1142 struct in6_addr *saddr)
1144 struct fib6_node *pn, *sn;
1146 if (fn->fn_flags & RTN_TL_ROOT)
1148 pn = rcu_dereference(fn->parent);
1149 sn = FIB6_SUBTREE(pn);
1151 fn = fib6_node_lookup(sn, NULL, saddr);
1154 if (fn->fn_flags & RTN_RTINFO)
1159 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1161 struct rt6_info *rt = *prt;
1163 if (dst_hold_safe(&rt->dst))
1166 rt = net->ipv6.ip6_null_entry;
1175 /* called with rcu_lock held */
1176 static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1178 struct net_device *dev = res->nh->fib_nh_dev;
1179 struct fib6_info *f6i = res->f6i;
1180 unsigned short flags;
1181 struct rt6_info *nrt;
1183 if (!fib6_info_hold_safe(f6i))
1186 flags = fib6_info_dst_flags(f6i);
1187 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1189 fib6_info_release(f6i);
1193 ip6_rt_copy_init(nrt, res);
1197 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1198 dst_hold(&nrt->dst);
1202 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
1203 struct fib6_table *table,
1205 const struct sk_buff *skb,
1208 struct fib6_result res = {};
1209 struct fib6_node *fn;
1210 struct rt6_info *rt;
1213 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1215 res.f6i = rcu_dereference(fn->leaf);
1217 res.f6i = net->ipv6.fib6_null_entry;
1219 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1222 if (res.f6i == net->ipv6.fib6_null_entry) {
1223 fn = fib6_backtrack(fn, &fl6->saddr);
1227 rt = net->ipv6.ip6_null_entry;
1230 } else if (res.fib6_flags & RTF_REJECT) {
1234 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1235 fl6->flowi6_oif != 0, skb, flags);
1237 /* Search through exception table */
1238 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1240 if (ip6_hold_safe(net, &rt))
1241 dst_use_noref(&rt->dst, jiffies);
1244 rt = ip6_create_rt_rcu(&res);
1248 trace_fib6_table_lookup(net, &res, table, fl6);
1255 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1256 const struct sk_buff *skb, int flags)
1258 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1260 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1262 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1263 const struct in6_addr *saddr, int oif,
1264 const struct sk_buff *skb, int strict)
1266 struct flowi6 fl6 = {
1270 struct dst_entry *dst;
1271 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1274 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1275 flags |= RT6_LOOKUP_F_HAS_SADDR;
1278 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1279 if (dst->error == 0)
1280 return (struct rt6_info *) dst;
1286 EXPORT_SYMBOL(rt6_lookup);
1288 /* ip6_ins_rt is called with FREE table->tb6_lock.
1289 * It takes new route entry, the addition fails by any reason the
1290 * route is released.
1291 * Caller must hold dst before calling it.
1294 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1295 struct netlink_ext_ack *extack)
1298 struct fib6_table *table;
1300 table = rt->fib6_table;
1301 spin_lock_bh(&table->tb6_lock);
1302 err = fib6_add(&table->tb6_root, rt, info, extack);
1303 spin_unlock_bh(&table->tb6_lock);
1308 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1310 struct nl_info info = { .nl_net = net, };
1312 return __ip6_ins_rt(rt, &info, NULL);
1315 static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1316 const struct in6_addr *daddr,
1317 const struct in6_addr *saddr)
1319 struct fib6_info *f6i = res->f6i;
1320 struct net_device *dev;
1321 struct rt6_info *rt;
1327 if (!fib6_info_hold_safe(f6i))
1330 dev = ip6_rt_get_dev_rcu(res);
1331 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1333 fib6_info_release(f6i);
1337 ip6_rt_copy_init(rt, res);
1338 rt->rt6i_flags |= RTF_CACHE;
1339 rt->rt6i_dst.addr = *daddr;
1340 rt->rt6i_dst.plen = 128;
1342 if (!rt6_is_gw_or_nonexthop(res)) {
1343 if (f6i->fib6_dst.plen != 128 &&
1344 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1345 rt->rt6i_flags |= RTF_ANYCAST;
1346 #ifdef CONFIG_IPV6_SUBTREES
1347 if (rt->rt6i_src.plen && saddr) {
1348 rt->rt6i_src.addr = *saddr;
1349 rt->rt6i_src.plen = 128;
1357 static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1359 struct fib6_info *f6i = res->f6i;
1360 unsigned short flags = fib6_info_dst_flags(f6i);
1361 struct net_device *dev;
1362 struct rt6_info *pcpu_rt;
1364 if (!fib6_info_hold_safe(f6i))
1368 dev = ip6_rt_get_dev_rcu(res);
1369 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
1372 fib6_info_release(f6i);
1375 ip6_rt_copy_init(pcpu_rt, res);
1376 pcpu_rt->rt6i_flags |= RTF_PCPU;
1379 pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1384 static bool rt6_is_valid(const struct rt6_info *rt6)
1386 return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1389 /* It should be called with rcu_read_lock() acquired */
1390 static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1392 struct rt6_info *pcpu_rt;
1394 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1396 if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1397 struct rt6_info *prev, **p;
1399 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1400 prev = xchg(p, NULL);
1402 dst_dev_put(&prev->dst);
1403 dst_release(&prev->dst);
1412 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1413 const struct fib6_result *res)
1415 struct rt6_info *pcpu_rt, *prev, **p;
1417 pcpu_rt = ip6_rt_pcpu_alloc(res);
1421 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1422 prev = cmpxchg(p, NULL, pcpu_rt);
1425 if (res->f6i->fib6_destroying) {
1426 struct fib6_info *from;
1428 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1429 fib6_info_release(from);
1435 /* exception hash table implementation
1437 static DEFINE_SPINLOCK(rt6_exception_lock);
1439 /* Remove rt6_ex from hash table and free the memory
1440 * Caller must hold rt6_exception_lock
1442 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1443 struct rt6_exception *rt6_ex)
1445 struct fib6_info *from;
1448 if (!bucket || !rt6_ex)
1451 net = dev_net(rt6_ex->rt6i->dst.dev);
1452 net->ipv6.rt6_stats->fib_rt_cache--;
1454 /* purge completely the exception to allow releasing the held resources:
1455 * some [sk] cache may keep the dst around for unlimited time
1457 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
1458 fib6_info_release(from);
1459 dst_dev_put(&rt6_ex->rt6i->dst);
1461 hlist_del_rcu(&rt6_ex->hlist);
1462 dst_release(&rt6_ex->rt6i->dst);
1463 kfree_rcu(rt6_ex, rcu);
1464 WARN_ON_ONCE(!bucket->depth);
1468 /* Remove oldest rt6_ex in bucket and free the memory
1469 * Caller must hold rt6_exception_lock
1471 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1473 struct rt6_exception *rt6_ex, *oldest = NULL;
1478 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1479 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1482 rt6_remove_exception(bucket, oldest);
1485 static u32 rt6_exception_hash(const struct in6_addr *dst,
1486 const struct in6_addr *src)
1488 static siphash_aligned_key_t rt6_exception_key;
1490 struct in6_addr dst;
1491 struct in6_addr src;
1492 } __aligned(SIPHASH_ALIGNMENT) combined = {
1497 net_get_random_once(&rt6_exception_key, sizeof(rt6_exception_key));
1499 #ifdef CONFIG_IPV6_SUBTREES
1501 combined.src = *src;
1503 val = siphash(&combined, sizeof(combined), &rt6_exception_key);
1505 return hash_64(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1508 /* Helper function to find the cached rt in the hash table
1509 * and update bucket pointer to point to the bucket for this
1510 * (daddr, saddr) pair
1511 * Caller must hold rt6_exception_lock
1513 static struct rt6_exception *
1514 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1515 const struct in6_addr *daddr,
1516 const struct in6_addr *saddr)
1518 struct rt6_exception *rt6_ex;
1521 if (!(*bucket) || !daddr)
1524 hval = rt6_exception_hash(daddr, saddr);
1527 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1528 struct rt6_info *rt6 = rt6_ex->rt6i;
1529 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1531 #ifdef CONFIG_IPV6_SUBTREES
1532 if (matched && saddr)
1533 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1541 /* Helper function to find the cached rt in the hash table
1542 * and update bucket pointer to point to the bucket for this
1543 * (daddr, saddr) pair
1544 * Caller must hold rcu_read_lock()
1546 static struct rt6_exception *
1547 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1548 const struct in6_addr *daddr,
1549 const struct in6_addr *saddr)
1551 struct rt6_exception *rt6_ex;
1554 WARN_ON_ONCE(!rcu_read_lock_held());
1556 if (!(*bucket) || !daddr)
1559 hval = rt6_exception_hash(daddr, saddr);
1562 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1563 struct rt6_info *rt6 = rt6_ex->rt6i;
1564 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1566 #ifdef CONFIG_IPV6_SUBTREES
1567 if (matched && saddr)
1568 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1576 static unsigned int fib6_mtu(const struct fib6_result *res)
1578 const struct fib6_nh *nh = res->nh;
1581 if (res->f6i->fib6_pmtu) {
1582 mtu = res->f6i->fib6_pmtu;
1584 struct net_device *dev = nh->fib_nh_dev;
1585 struct inet6_dev *idev;
1588 idev = __in6_dev_get(dev);
1589 mtu = idev->cnf.mtu6;
1593 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1595 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1598 #define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1600 /* used when the flushed bit is not relevant, only access to the bucket
1601 * (ie., all bucket users except rt6_insert_exception);
1603 * called under rcu lock; sometimes called with rt6_exception_lock held
1606 struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1609 struct rt6_exception_bucket *bucket;
1612 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1613 lockdep_is_held(lock));
1615 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1617 /* remove bucket flushed bit if set */
1619 unsigned long p = (unsigned long)bucket;
1621 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1622 bucket = (struct rt6_exception_bucket *)p;
1628 static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1630 unsigned long p = (unsigned long)bucket;
1632 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1635 /* called with rt6_exception_lock held */
1636 static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1639 struct rt6_exception_bucket *bucket;
1642 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1643 lockdep_is_held(lock));
1645 p = (unsigned long)bucket;
1646 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1647 bucket = (struct rt6_exception_bucket *)p;
1648 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1651 static int rt6_insert_exception(struct rt6_info *nrt,
1652 const struct fib6_result *res)
1654 struct net *net = dev_net(nrt->dst.dev);
1655 struct rt6_exception_bucket *bucket;
1656 struct fib6_info *f6i = res->f6i;
1657 struct in6_addr *src_key = NULL;
1658 struct rt6_exception *rt6_ex;
1659 struct fib6_nh *nh = res->nh;
1663 spin_lock_bh(&rt6_exception_lock);
1665 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1666 lockdep_is_held(&rt6_exception_lock));
1668 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1674 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1675 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1680 #ifdef CONFIG_IPV6_SUBTREES
1681 /* fib6_src.plen != 0 indicates f6i is in subtree
1682 * and exception table is indexed by a hash of
1683 * both fib6_dst and fib6_src.
1684 * Otherwise, the exception table is indexed by
1685 * a hash of only fib6_dst.
1687 if (f6i->fib6_src.plen)
1688 src_key = &nrt->rt6i_src.addr;
1690 /* rt6_mtu_change() might lower mtu on f6i.
1691 * Only insert this exception route if its mtu
1692 * is less than f6i's mtu value.
1694 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1699 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1702 rt6_remove_exception(bucket, rt6_ex);
1704 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1710 rt6_ex->stamp = jiffies;
1711 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1713 net->ipv6.rt6_stats->fib_rt_cache++;
1715 /* Randomize max depth to avoid some side channels attacks. */
1716 max_depth = FIB6_MAX_DEPTH + prandom_u32_max(FIB6_MAX_DEPTH);
1717 while (bucket->depth > max_depth)
1718 rt6_exception_remove_oldest(bucket);
1721 spin_unlock_bh(&rt6_exception_lock);
1723 /* Update fn->fn_sernum to invalidate all cached dst */
1725 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1726 fib6_update_sernum(net, f6i);
1727 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1728 fib6_force_start_gc(net);
1734 static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1736 struct rt6_exception_bucket *bucket;
1737 struct rt6_exception *rt6_ex;
1738 struct hlist_node *tmp;
1741 spin_lock_bh(&rt6_exception_lock);
1743 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1747 /* Prevent rt6_insert_exception() to recreate the bucket list */
1749 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1751 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1752 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1754 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1755 rt6_remove_exception(bucket, rt6_ex);
1757 WARN_ON_ONCE(!from && bucket->depth);
1761 spin_unlock_bh(&rt6_exception_lock);
1764 static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1766 struct fib6_info *f6i = arg;
1768 fib6_nh_flush_exceptions(nh, f6i);
1773 void rt6_flush_exceptions(struct fib6_info *f6i)
1776 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1779 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1782 /* Find cached rt in the hash table inside passed in rt
1783 * Caller has to hold rcu_read_lock()
1785 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1786 const struct in6_addr *daddr,
1787 const struct in6_addr *saddr)
1789 const struct in6_addr *src_key = NULL;
1790 struct rt6_exception_bucket *bucket;
1791 struct rt6_exception *rt6_ex;
1792 struct rt6_info *ret = NULL;
1794 #ifdef CONFIG_IPV6_SUBTREES
1795 /* fib6i_src.plen != 0 indicates f6i is in subtree
1796 * and exception table is indexed by a hash of
1797 * both fib6_dst and fib6_src.
1798 * However, the src addr used to create the hash
1799 * might not be exactly the passed in saddr which
1800 * is a /128 addr from the flow.
1801 * So we need to use f6i->fib6_src to redo lookup
1802 * if the passed in saddr does not find anything.
1803 * (See the logic in ip6_rt_cache_alloc() on how
1804 * rt->rt6i_src is updated.)
1806 if (res->f6i->fib6_src.plen)
1810 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1811 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1813 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1816 #ifdef CONFIG_IPV6_SUBTREES
1817 /* Use fib6_src as src_key and redo lookup */
1818 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1819 src_key = &res->f6i->fib6_src.addr;
1827 /* Remove the passed in cached rt from the hash table that contains it */
1828 static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1829 const struct rt6_info *rt)
1831 const struct in6_addr *src_key = NULL;
1832 struct rt6_exception_bucket *bucket;
1833 struct rt6_exception *rt6_ex;
1836 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1839 spin_lock_bh(&rt6_exception_lock);
1840 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1842 #ifdef CONFIG_IPV6_SUBTREES
1843 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1844 * and exception table is indexed by a hash of
1845 * both rt6i_dst and rt6i_src.
1846 * Otherwise, the exception table is indexed by
1847 * a hash of only rt6i_dst.
1850 src_key = &rt->rt6i_src.addr;
1852 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1856 rt6_remove_exception(bucket, rt6_ex);
1862 spin_unlock_bh(&rt6_exception_lock);
1866 struct fib6_nh_excptn_arg {
1867 struct rt6_info *rt;
1871 static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1873 struct fib6_nh_excptn_arg *arg = _arg;
1876 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1883 static int rt6_remove_exception_rt(struct rt6_info *rt)
1885 struct fib6_info *from;
1887 from = rcu_dereference(rt->from);
1888 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1892 struct fib6_nh_excptn_arg arg = {
1894 .plen = from->fib6_src.plen
1898 /* rc = 1 means an entry was found */
1899 rc = nexthop_for_each_fib6_nh(from->nh,
1900 rt6_nh_remove_exception_rt,
1902 return rc ? 0 : -ENOENT;
1905 return fib6_nh_remove_exception(from->fib6_nh,
1906 from->fib6_src.plen, rt);
1909 /* Find rt6_ex which contains the passed in rt cache and
1912 static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1913 const struct rt6_info *rt)
1915 const struct in6_addr *src_key = NULL;
1916 struct rt6_exception_bucket *bucket;
1917 struct rt6_exception *rt6_ex;
1919 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1920 #ifdef CONFIG_IPV6_SUBTREES
1921 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1922 * and exception table is indexed by a hash of
1923 * both rt6i_dst and rt6i_src.
1924 * Otherwise, the exception table is indexed by
1925 * a hash of only rt6i_dst.
1928 src_key = &rt->rt6i_src.addr;
1930 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1932 rt6_ex->stamp = jiffies;
1935 struct fib6_nh_match_arg {
1936 const struct net_device *dev;
1937 const struct in6_addr *gw;
1938 struct fib6_nh *match;
1941 /* determine if fib6_nh has given device and gateway */
1942 static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1944 struct fib6_nh_match_arg *arg = _arg;
1946 if (arg->dev != nh->fib_nh_dev ||
1947 (arg->gw && !nh->fib_nh_gw_family) ||
1948 (!arg->gw && nh->fib_nh_gw_family) ||
1949 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1954 /* found a match, break the loop */
1958 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1960 struct fib6_info *from;
1961 struct fib6_nh *fib6_nh;
1965 from = rcu_dereference(rt->from);
1966 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1970 struct fib6_nh_match_arg arg = {
1972 .gw = &rt->rt6i_gateway,
1975 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1979 fib6_nh = arg.match;
1981 fib6_nh = from->fib6_nh;
1983 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1988 static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1989 struct rt6_info *rt, int mtu)
1991 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1992 * lowest MTU in the path: always allow updating the route PMTU to
1993 * reflect PMTU decreases.
1995 * If the new MTU is higher, and the route PMTU is equal to the local
1996 * MTU, this means the old MTU is the lowest in the path, so allow
1997 * updating it: if other nodes now have lower MTUs, PMTU discovery will
2001 if (dst_mtu(&rt->dst) >= mtu)
2004 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
2010 static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
2011 const struct fib6_nh *nh, int mtu)
2013 struct rt6_exception_bucket *bucket;
2014 struct rt6_exception *rt6_ex;
2017 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2021 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2022 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2023 struct rt6_info *entry = rt6_ex->rt6i;
2025 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2026 * route), the metrics of its rt->from have already
2029 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2030 rt6_mtu_change_route_allowed(idev, entry, mtu))
2031 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2037 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2039 static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2040 const struct in6_addr *gateway)
2042 struct rt6_exception_bucket *bucket;
2043 struct rt6_exception *rt6_ex;
2044 struct hlist_node *tmp;
2047 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2050 spin_lock_bh(&rt6_exception_lock);
2051 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2053 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2054 hlist_for_each_entry_safe(rt6_ex, tmp,
2055 &bucket->chain, hlist) {
2056 struct rt6_info *entry = rt6_ex->rt6i;
2058 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2059 RTF_CACHE_GATEWAY &&
2060 ipv6_addr_equal(gateway,
2061 &entry->rt6i_gateway)) {
2062 rt6_remove_exception(bucket, rt6_ex);
2069 spin_unlock_bh(&rt6_exception_lock);
2072 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2073 struct rt6_exception *rt6_ex,
2074 struct fib6_gc_args *gc_args,
2077 struct rt6_info *rt = rt6_ex->rt6i;
2079 /* we are pruning and obsoleting aged-out and non gateway exceptions
2080 * even if others have still references to them, so that on next
2081 * dst_check() such references can be dropped.
2082 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2083 * expired, independently from their aging, as per RFC 8201 section 4
2085 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2086 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2087 RT6_TRACE("aging clone %p\n", rt);
2088 rt6_remove_exception(bucket, rt6_ex);
2091 } else if (time_after(jiffies, rt->dst.expires)) {
2092 RT6_TRACE("purging expired route %p\n", rt);
2093 rt6_remove_exception(bucket, rt6_ex);
2097 if (rt->rt6i_flags & RTF_GATEWAY) {
2098 struct neighbour *neigh;
2100 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2102 if (!(neigh && (neigh->flags & NTF_ROUTER))) {
2103 RT6_TRACE("purging route %p via non-router but gateway\n",
2105 rt6_remove_exception(bucket, rt6_ex);
2113 static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2114 struct fib6_gc_args *gc_args,
2117 struct rt6_exception_bucket *bucket;
2118 struct rt6_exception *rt6_ex;
2119 struct hlist_node *tmp;
2122 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2126 spin_lock(&rt6_exception_lock);
2127 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2129 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2130 hlist_for_each_entry_safe(rt6_ex, tmp,
2131 &bucket->chain, hlist) {
2132 rt6_age_examine_exception(bucket, rt6_ex,
2138 spin_unlock(&rt6_exception_lock);
2139 rcu_read_unlock_bh();
2142 struct fib6_nh_age_excptn_arg {
2143 struct fib6_gc_args *gc_args;
2147 static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2149 struct fib6_nh_age_excptn_arg *arg = _arg;
2151 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2155 void rt6_age_exceptions(struct fib6_info *f6i,
2156 struct fib6_gc_args *gc_args,
2160 struct fib6_nh_age_excptn_arg arg = {
2165 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2168 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2172 /* must be called with rcu lock held */
2173 int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2174 struct flowi6 *fl6, struct fib6_result *res, int strict)
2176 struct fib6_node *fn, *saved_fn;
2178 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2182 rt6_select(net, fn, oif, res, strict);
2183 if (res->f6i == net->ipv6.fib6_null_entry) {
2184 fn = fib6_backtrack(fn, &fl6->saddr);
2186 goto redo_rt6_select;
2187 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2188 /* also consider unreachable route */
2189 strict &= ~RT6_LOOKUP_F_REACHABLE;
2191 goto redo_rt6_select;
2195 trace_fib6_table_lookup(net, res, table, fl6);
2200 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2201 int oif, struct flowi6 *fl6,
2202 const struct sk_buff *skb, int flags)
2204 struct fib6_result res = {};
2205 struct rt6_info *rt = NULL;
2208 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2209 !rcu_read_lock_held());
2211 strict |= flags & RT6_LOOKUP_F_IFACE;
2212 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2213 if (net->ipv6.devconf_all->forwarding == 0)
2214 strict |= RT6_LOOKUP_F_REACHABLE;
2218 fib6_table_lookup(net, table, oif, fl6, &res, strict);
2219 if (res.f6i == net->ipv6.fib6_null_entry)
2222 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2224 /*Search through exception table */
2225 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2228 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2229 !res.nh->fib_nh_gw_family)) {
2230 /* Create a RTF_CACHE clone which will not be
2231 * owned by the fib6 tree. It is for the special case where
2232 * the daddr in the skb during the neighbor look-up is different
2233 * from the fl6->daddr used to look-up route here.
2235 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2238 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2239 * As rt6_uncached_list_add() does not consume refcnt,
2240 * this refcnt is always returned to the caller even
2241 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2243 rt6_uncached_list_add(rt);
2249 /* Get a percpu copy */
2251 rt = rt6_get_pcpu_route(&res);
2254 rt = rt6_make_pcpu_route(net, &res);
2260 rt = net->ipv6.ip6_null_entry;
2261 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2262 ip6_hold_safe(net, &rt);
2267 EXPORT_SYMBOL_GPL(ip6_pol_route);
2269 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
2270 struct fib6_table *table,
2272 const struct sk_buff *skb,
2275 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2278 struct dst_entry *ip6_route_input_lookup(struct net *net,
2279 struct net_device *dev,
2281 const struct sk_buff *skb,
2284 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2285 flags |= RT6_LOOKUP_F_IFACE;
2287 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2289 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2291 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2292 struct flow_keys *keys,
2293 struct flow_keys *flkeys)
2295 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2296 const struct ipv6hdr *key_iph = outer_iph;
2297 struct flow_keys *_flkeys = flkeys;
2298 const struct ipv6hdr *inner_iph;
2299 const struct icmp6hdr *icmph;
2300 struct ipv6hdr _inner_iph;
2301 struct icmp6hdr _icmph;
2303 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2306 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2307 sizeof(_icmph), &_icmph);
2311 if (!icmpv6_is_err(icmph->icmp6_type))
2314 inner_iph = skb_header_pointer(skb,
2315 skb_transport_offset(skb) + sizeof(*icmph),
2316 sizeof(_inner_iph), &_inner_iph);
2320 key_iph = inner_iph;
2324 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2325 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2326 keys->tags.flow_label = _flkeys->tags.flow_label;
2327 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2329 keys->addrs.v6addrs.src = key_iph->saddr;
2330 keys->addrs.v6addrs.dst = key_iph->daddr;
2331 keys->tags.flow_label = ip6_flowlabel(key_iph);
2332 keys->basic.ip_proto = key_iph->nexthdr;
2336 static u32 rt6_multipath_custom_hash_outer(const struct net *net,
2337 const struct sk_buff *skb,
2340 u32 hash_fields = ip6_multipath_hash_fields(net);
2341 struct flow_keys keys, hash_keys;
2343 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2346 memset(&hash_keys, 0, sizeof(hash_keys));
2347 skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_ENCAP);
2349 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2350 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2351 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2352 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2353 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2354 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2355 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2356 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2357 hash_keys.tags.flow_label = keys.tags.flow_label;
2358 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2359 hash_keys.ports.src = keys.ports.src;
2360 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2361 hash_keys.ports.dst = keys.ports.dst;
2363 *p_has_inner = !!(keys.control.flags & FLOW_DIS_ENCAPSULATION);
2364 return flow_hash_from_keys(&hash_keys);
2367 static u32 rt6_multipath_custom_hash_inner(const struct net *net,
2368 const struct sk_buff *skb,
2371 u32 hash_fields = ip6_multipath_hash_fields(net);
2372 struct flow_keys keys, hash_keys;
2374 /* We assume the packet carries an encapsulation, but if none was
2375 * encountered during dissection of the outer flow, then there is no
2376 * point in calling the flow dissector again.
2381 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_MASK))
2384 memset(&hash_keys, 0, sizeof(hash_keys));
2385 skb_flow_dissect_flow_keys(skb, &keys, 0);
2387 if (!(keys.control.flags & FLOW_DIS_ENCAPSULATION))
2390 if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2391 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2392 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2393 hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
2394 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2395 hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst;
2396 } else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2397 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2398 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2399 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2400 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2401 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2402 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL)
2403 hash_keys.tags.flow_label = keys.tags.flow_label;
2406 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO)
2407 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2408 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT)
2409 hash_keys.ports.src = keys.ports.src;
2410 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
2411 hash_keys.ports.dst = keys.ports.dst;
2413 return flow_hash_from_keys(&hash_keys);
2416 static u32 rt6_multipath_custom_hash_skb(const struct net *net,
2417 const struct sk_buff *skb)
2419 u32 mhash, mhash_inner;
2420 bool has_inner = true;
2422 mhash = rt6_multipath_custom_hash_outer(net, skb, &has_inner);
2423 mhash_inner = rt6_multipath_custom_hash_inner(net, skb, has_inner);
2425 return jhash_2words(mhash, mhash_inner, 0);
2428 static u32 rt6_multipath_custom_hash_fl6(const struct net *net,
2429 const struct flowi6 *fl6)
2431 u32 hash_fields = ip6_multipath_hash_fields(net);
2432 struct flow_keys hash_keys;
2434 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2437 memset(&hash_keys, 0, sizeof(hash_keys));
2438 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2439 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2440 hash_keys.addrs.v6addrs.src = fl6->saddr;
2441 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2442 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2443 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2444 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2445 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2446 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2447 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2448 hash_keys.ports.src = fl6->fl6_sport;
2449 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2450 hash_keys.ports.dst = fl6->fl6_dport;
2452 return flow_hash_from_keys(&hash_keys);
2455 /* if skb is set it will be used and fl6 can be NULL */
2456 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2457 const struct sk_buff *skb, struct flow_keys *flkeys)
2459 struct flow_keys hash_keys;
2462 switch (ip6_multipath_hash_policy(net)) {
2464 memset(&hash_keys, 0, sizeof(hash_keys));
2465 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2467 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2469 hash_keys.addrs.v6addrs.src = fl6->saddr;
2470 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2471 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2472 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2474 mhash = flow_hash_from_keys(&hash_keys);
2478 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2479 struct flow_keys keys;
2481 /* short-circuit if we already have L4 hash present */
2483 return skb_get_hash_raw(skb) >> 1;
2485 memset(&hash_keys, 0, sizeof(hash_keys));
2488 skb_flow_dissect_flow_keys(skb, &keys, flag);
2491 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2492 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2493 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2494 hash_keys.ports.src = flkeys->ports.src;
2495 hash_keys.ports.dst = flkeys->ports.dst;
2496 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2498 memset(&hash_keys, 0, sizeof(hash_keys));
2499 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2500 hash_keys.addrs.v6addrs.src = fl6->saddr;
2501 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2502 hash_keys.ports.src = fl6->fl6_sport;
2503 hash_keys.ports.dst = fl6->fl6_dport;
2504 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2506 mhash = flow_hash_from_keys(&hash_keys);
2509 memset(&hash_keys, 0, sizeof(hash_keys));
2510 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2512 struct flow_keys keys;
2515 skb_flow_dissect_flow_keys(skb, &keys, 0);
2519 /* Inner can be v4 or v6 */
2520 if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2521 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2522 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2523 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2524 } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2525 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2526 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2527 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2528 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2529 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2531 /* Same as case 0 */
2532 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2533 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2536 /* Same as case 0 */
2537 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2538 hash_keys.addrs.v6addrs.src = fl6->saddr;
2539 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2540 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2541 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2543 mhash = flow_hash_from_keys(&hash_keys);
2547 mhash = rt6_multipath_custom_hash_skb(net, skb);
2549 mhash = rt6_multipath_custom_hash_fl6(net, fl6);
2556 /* Called with rcu held */
2557 void ip6_route_input(struct sk_buff *skb)
2559 const struct ipv6hdr *iph = ipv6_hdr(skb);
2560 struct net *net = dev_net(skb->dev);
2561 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2562 struct ip_tunnel_info *tun_info;
2563 struct flowi6 fl6 = {
2564 .flowi6_iif = skb->dev->ifindex,
2565 .daddr = iph->daddr,
2566 .saddr = iph->saddr,
2567 .flowlabel = ip6_flowinfo(iph),
2568 .flowi6_mark = skb->mark,
2569 .flowi6_proto = iph->nexthdr,
2571 struct flow_keys *flkeys = NULL, _flkeys;
2573 tun_info = skb_tunnel_info(skb);
2574 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2575 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2577 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2580 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2581 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2583 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2587 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
2588 struct fib6_table *table,
2590 const struct sk_buff *skb,
2593 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2596 struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2597 const struct sock *sk,
2598 struct flowi6 *fl6, int flags)
2602 if (ipv6_addr_type(&fl6->daddr) &
2603 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2604 struct dst_entry *dst;
2606 /* This function does not take refcnt on the dst */
2607 dst = l3mdev_link_scope_lookup(net, fl6);
2612 fl6->flowi6_iif = LOOPBACK_IFINDEX;
2614 flags |= RT6_LOOKUP_F_DST_NOREF;
2615 any_src = ipv6_addr_any(&fl6->saddr);
2616 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2617 (fl6->flowi6_oif && any_src))
2618 flags |= RT6_LOOKUP_F_IFACE;
2621 flags |= RT6_LOOKUP_F_HAS_SADDR;
2623 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
2625 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2627 EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2629 struct dst_entry *ip6_route_output_flags(struct net *net,
2630 const struct sock *sk,
2634 struct dst_entry *dst;
2635 struct rt6_info *rt6;
2638 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2639 rt6 = (struct rt6_info *)dst;
2640 /* For dst cached in uncached_list, refcnt is already taken. */
2641 if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2642 dst = &net->ipv6.ip6_null_entry->dst;
2649 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2651 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2653 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2654 struct net_device *loopback_dev = net->loopback_dev;
2655 struct dst_entry *new = NULL;
2657 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
2658 DST_OBSOLETE_DEAD, 0);
2661 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2665 new->input = dst_discard;
2666 new->output = dst_discard_out;
2668 dst_copy_metrics(new, &ort->dst);
2670 rt->rt6i_idev = in6_dev_get(loopback_dev);
2671 rt->rt6i_gateway = ort->rt6i_gateway;
2672 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2674 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2675 #ifdef CONFIG_IPV6_SUBTREES
2676 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2680 dst_release(dst_orig);
2681 return new ? new : ERR_PTR(-ENOMEM);
2685 * Destination cache support functions
2688 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2692 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2695 if (fib6_check_expired(f6i))
2701 static struct dst_entry *rt6_check(struct rt6_info *rt,
2702 struct fib6_info *from,
2707 if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2708 rt_cookie != cookie)
2711 if (rt6_check_expired(rt))
2717 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2718 struct fib6_info *from,
2721 if (!__rt6_check_expired(rt) &&
2722 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2723 fib6_check(from, cookie))
2729 INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2732 struct dst_entry *dst_ret;
2733 struct fib6_info *from;
2734 struct rt6_info *rt;
2736 rt = container_of(dst, struct rt6_info, dst);
2739 return rt6_is_valid(rt) ? dst : NULL;
2743 /* All IPV6 dsts are created with ->obsolete set to the value
2744 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2745 * into this function always.
2748 from = rcu_dereference(rt->from);
2750 if (from && (rt->rt6i_flags & RTF_PCPU ||
2751 unlikely(!list_empty(&rt->rt6i_uncached))))
2752 dst_ret = rt6_dst_from_check(rt, from, cookie);
2754 dst_ret = rt6_check(rt, from, cookie);
2760 EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
2762 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2764 struct rt6_info *rt = (struct rt6_info *) dst;
2767 if (rt->rt6i_flags & RTF_CACHE) {
2769 if (rt6_check_expired(rt)) {
2770 rt6_remove_exception_rt(rt);
2782 static void ip6_link_failure(struct sk_buff *skb)
2784 struct rt6_info *rt;
2786 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2788 rt = (struct rt6_info *) skb_dst(skb);
2791 if (rt->rt6i_flags & RTF_CACHE) {
2792 rt6_remove_exception_rt(rt);
2794 struct fib6_info *from;
2795 struct fib6_node *fn;
2797 from = rcu_dereference(rt->from);
2799 fn = rcu_dereference(from->fib6_node);
2800 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2801 WRITE_ONCE(fn->fn_sernum, -1);
2808 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2810 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2811 struct fib6_info *from;
2814 from = rcu_dereference(rt0->from);
2816 rt0->dst.expires = from->expires;
2820 dst_set_expires(&rt0->dst, timeout);
2821 rt0->rt6i_flags |= RTF_EXPIRES;
2824 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2826 struct net *net = dev_net(rt->dst.dev);
2828 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2829 rt->rt6i_flags |= RTF_MODIFIED;
2830 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2833 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2835 return !(rt->rt6i_flags & RTF_CACHE) &&
2836 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2839 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2840 const struct ipv6hdr *iph, u32 mtu,
2843 const struct in6_addr *daddr, *saddr;
2844 struct rt6_info *rt6 = (struct rt6_info *)dst;
2846 /* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2847 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2848 * [see also comment in rt6_mtu_change_route()]
2852 daddr = &iph->daddr;
2853 saddr = &iph->saddr;
2855 daddr = &sk->sk_v6_daddr;
2856 saddr = &inet6_sk(sk)->saddr;
2863 dst_confirm_neigh(dst, daddr);
2865 if (mtu < IPV6_MIN_MTU)
2867 if (mtu >= dst_mtu(dst))
2870 if (!rt6_cache_allowed_for_pmtu(rt6)) {
2871 rt6_do_update_pmtu(rt6, mtu);
2872 /* update rt6_ex->stamp for cache */
2873 if (rt6->rt6i_flags & RTF_CACHE)
2874 rt6_update_exception_stamp_rt(rt6);
2876 struct fib6_result res = {};
2877 struct rt6_info *nrt6;
2880 res.f6i = rcu_dereference(rt6->from);
2884 res.fib6_flags = res.f6i->fib6_flags;
2885 res.fib6_type = res.f6i->fib6_type;
2888 struct fib6_nh_match_arg arg = {
2890 .gw = &rt6->rt6i_gateway,
2893 nexthop_for_each_fib6_nh(res.f6i->nh,
2894 fib6_nh_find_match, &arg);
2896 /* fib6_info uses a nexthop that does not have fib6_nh
2897 * using the dst->dev + gw. Should be impossible.
2904 res.nh = res.f6i->fib6_nh;
2907 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2909 rt6_do_update_pmtu(nrt6, mtu);
2910 if (rt6_insert_exception(nrt6, &res))
2911 dst_release_immediate(&nrt6->dst);
2918 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2919 struct sk_buff *skb, u32 mtu,
2922 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2926 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2927 int oif, u32 mark, kuid_t uid)
2929 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2930 struct dst_entry *dst;
2931 struct flowi6 fl6 = {
2933 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2934 .daddr = iph->daddr,
2935 .saddr = iph->saddr,
2936 .flowlabel = ip6_flowinfo(iph),
2940 dst = ip6_route_output(net, NULL, &fl6);
2942 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
2945 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2947 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2949 int oif = sk->sk_bound_dev_if;
2950 struct dst_entry *dst;
2952 if (!oif && skb->dev)
2953 oif = l3mdev_master_ifindex(skb->dev);
2955 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
2957 dst = __sk_dst_get(sk);
2958 if (!dst || !dst->obsolete ||
2959 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2963 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2964 ip6_datagram_dst_update(sk, false);
2967 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2969 void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2970 const struct flowi6 *fl6)
2972 #ifdef CONFIG_IPV6_SUBTREES
2973 struct ipv6_pinfo *np = inet6_sk(sk);
2976 ip6_dst_store(sk, dst,
2977 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2978 &sk->sk_v6_daddr : NULL,
2979 #ifdef CONFIG_IPV6_SUBTREES
2980 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2986 static bool ip6_redirect_nh_match(const struct fib6_result *res,
2988 const struct in6_addr *gw,
2989 struct rt6_info **ret)
2991 const struct fib6_nh *nh = res->nh;
2993 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2994 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2997 /* rt_cache's gateway might be different from its 'parent'
2998 * in the case of an ip redirect.
2999 * So we keep searching in the exception table if the gateway
3002 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
3003 struct rt6_info *rt_cache;
3005 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
3007 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
3016 struct fib6_nh_rd_arg {
3017 struct fib6_result *res;
3019 const struct in6_addr *gw;
3020 struct rt6_info **ret;
3023 static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
3025 struct fib6_nh_rd_arg *arg = _arg;
3028 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
3031 /* Handle redirects */
3032 struct ip6rd_flowi {
3034 struct in6_addr gateway;
3037 INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
3038 struct fib6_table *table,
3040 const struct sk_buff *skb,
3043 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
3044 struct rt6_info *ret = NULL;
3045 struct fib6_result res = {};
3046 struct fib6_nh_rd_arg arg = {
3049 .gw = &rdfl->gateway,
3052 struct fib6_info *rt;
3053 struct fib6_node *fn;
3055 /* Get the "current" route for this destination and
3056 * check if the redirect has come from appropriate router.
3058 * RFC 4861 specifies that redirects should only be
3059 * accepted if they come from the nexthop to the target.
3060 * Due to the way the routes are chosen, this notion
3061 * is a bit fuzzy and one might need to check all possible
3066 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
3068 for_each_fib6_node_rt_rcu(fn) {
3070 if (fib6_check_expired(rt))
3072 if (rt->fib6_flags & RTF_REJECT)
3074 if (unlikely(rt->nh)) {
3075 if (nexthop_is_blackhole(rt->nh))
3077 /* on match, res->nh is filled in and potentially ret */
3078 if (nexthop_for_each_fib6_nh(rt->nh,
3079 fib6_nh_redirect_match,
3083 res.nh = rt->fib6_nh;
3084 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
3091 rt = net->ipv6.fib6_null_entry;
3092 else if (rt->fib6_flags & RTF_REJECT) {
3093 ret = net->ipv6.ip6_null_entry;
3097 if (rt == net->ipv6.fib6_null_entry) {
3098 fn = fib6_backtrack(fn, &fl6->saddr);
3104 res.nh = rt->fib6_nh;
3107 ip6_hold_safe(net, &ret);
3109 res.fib6_flags = res.f6i->fib6_flags;
3110 res.fib6_type = res.f6i->fib6_type;
3111 ret = ip6_create_rt_rcu(&res);
3116 trace_fib6_table_lookup(net, &res, table, fl6);
3120 static struct dst_entry *ip6_route_redirect(struct net *net,
3121 const struct flowi6 *fl6,
3122 const struct sk_buff *skb,
3123 const struct in6_addr *gateway)
3125 int flags = RT6_LOOKUP_F_HAS_SADDR;
3126 struct ip6rd_flowi rdfl;
3129 rdfl.gateway = *gateway;
3131 return fib6_rule_lookup(net, &rdfl.fl6, skb,
3132 flags, __ip6_route_redirect);
3135 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3138 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3139 struct dst_entry *dst;
3140 struct flowi6 fl6 = {
3141 .flowi6_iif = LOOPBACK_IFINDEX,
3143 .flowi6_mark = mark,
3144 .daddr = iph->daddr,
3145 .saddr = iph->saddr,
3146 .flowlabel = ip6_flowinfo(iph),
3150 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3151 rt6_do_redirect(dst, NULL, skb);
3154 EXPORT_SYMBOL_GPL(ip6_redirect);
3156 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
3158 const struct ipv6hdr *iph = ipv6_hdr(skb);
3159 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3160 struct dst_entry *dst;
3161 struct flowi6 fl6 = {
3162 .flowi6_iif = LOOPBACK_IFINDEX,
3165 .saddr = iph->daddr,
3166 .flowi6_uid = sock_net_uid(net, NULL),
3169 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3170 rt6_do_redirect(dst, NULL, skb);
3174 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3176 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
3179 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3181 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3183 struct net_device *dev = dst->dev;
3184 unsigned int mtu = dst_mtu(dst);
3185 struct net *net = dev_net(dev);
3187 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3189 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3190 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3193 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3194 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3195 * IPV6_MAXPLEN is also valid and means: "any MSS,
3196 * rely only on pmtu discovery"
3198 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3203 INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
3205 return ip6_dst_mtu_maybe_forward(dst, false);
3207 EXPORT_INDIRECT_CALLABLE(ip6_mtu);
3210 * 1. mtu on route is locked - use it
3211 * 2. mtu from nexthop exception
3212 * 3. mtu from egress device
3214 * based on ip6_dst_mtu_forward and exception logic of
3215 * rt6_find_cached_rt; called with rcu_read_lock
3217 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3218 const struct in6_addr *daddr,
3219 const struct in6_addr *saddr)
3221 const struct fib6_nh *nh = res->nh;
3222 struct fib6_info *f6i = res->f6i;
3223 struct inet6_dev *idev;
3224 struct rt6_info *rt;
3227 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3228 mtu = f6i->fib6_pmtu;
3233 rt = rt6_find_cached_rt(res, daddr, saddr);
3235 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3237 struct net_device *dev = nh->fib_nh_dev;
3240 idev = __in6_dev_get(dev);
3241 if (idev && idev->cnf.mtu6 > mtu)
3242 mtu = idev->cnf.mtu6;
3245 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3247 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3250 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3253 struct dst_entry *dst;
3254 struct rt6_info *rt;
3255 struct inet6_dev *idev = in6_dev_get(dev);
3256 struct net *net = dev_net(dev);
3258 if (unlikely(!idev))
3259 return ERR_PTR(-ENODEV);
3261 rt = ip6_dst_alloc(net, dev, 0);
3262 if (unlikely(!rt)) {
3264 dst = ERR_PTR(-ENOMEM);
3268 rt->dst.input = ip6_input;
3269 rt->dst.output = ip6_output;
3270 rt->rt6i_gateway = fl6->daddr;
3271 rt->rt6i_dst.addr = fl6->daddr;
3272 rt->rt6i_dst.plen = 128;
3273 rt->rt6i_idev = idev;
3274 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3276 /* Add this dst into uncached_list so that rt6_disable_ip() can
3277 * do proper release of the net_device
3279 rt6_uncached_list_add(rt);
3281 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3287 static int ip6_dst_gc(struct dst_ops *ops)
3289 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3290 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3291 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3292 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3293 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3294 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3298 entries = dst_entries_get_fast(ops);
3299 if (entries > rt_max_size)
3300 entries = dst_entries_get_slow(ops);
3302 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
3303 entries <= rt_max_size)
3306 fib6_run_gc(atomic_inc_return(&net->ipv6.ip6_rt_gc_expire), net, true);
3307 entries = dst_entries_get_slow(ops);
3308 if (entries < ops->gc_thresh)
3309 atomic_set(&net->ipv6.ip6_rt_gc_expire, rt_gc_timeout >> 1);
3311 val = atomic_read(&net->ipv6.ip6_rt_gc_expire);
3312 atomic_set(&net->ipv6.ip6_rt_gc_expire, val - (val >> rt_elasticity));
3313 return entries > rt_max_size;
3316 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3317 const struct in6_addr *gw_addr, u32 tbid,
3318 int flags, struct fib6_result *res)
3320 struct flowi6 fl6 = {
3321 .flowi6_oif = cfg->fc_ifindex,
3323 .saddr = cfg->fc_prefsrc,
3325 struct fib6_table *table;
3328 table = fib6_get_table(net, tbid);
3332 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3333 flags |= RT6_LOOKUP_F_HAS_SADDR;
3335 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3337 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3338 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3339 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3340 cfg->fc_ifindex != 0, NULL, flags);
3345 static int ip6_route_check_nh_onlink(struct net *net,
3346 struct fib6_config *cfg,
3347 const struct net_device *dev,
3348 struct netlink_ext_ack *extack)
3350 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3351 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3352 struct fib6_result res = {};
3355 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3356 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3357 /* ignore match if it is the default route */
3358 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3359 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3360 NL_SET_ERR_MSG(extack,
3361 "Nexthop has invalid gateway or device mismatch");
3368 static int ip6_route_check_nh(struct net *net,
3369 struct fib6_config *cfg,
3370 struct net_device **_dev,
3371 struct inet6_dev **idev)
3373 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3374 struct net_device *dev = _dev ? *_dev : NULL;
3375 int flags = RT6_LOOKUP_F_IFACE;
3376 struct fib6_result res = {};
3377 int err = -EHOSTUNREACH;
3379 if (cfg->fc_table) {
3380 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3381 cfg->fc_table, flags, &res);
3382 /* gw_addr can not require a gateway or resolve to a reject
3383 * route. If a device is given, it must match the result.
3385 if (err || res.fib6_flags & RTF_REJECT ||
3386 res.nh->fib_nh_gw_family ||
3387 (dev && dev != res.nh->fib_nh_dev))
3388 err = -EHOSTUNREACH;
3392 struct flowi6 fl6 = {
3393 .flowi6_oif = cfg->fc_ifindex,
3397 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3398 if (err || res.fib6_flags & RTF_REJECT ||
3399 res.nh->fib_nh_gw_family)
3400 err = -EHOSTUNREACH;
3405 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3406 cfg->fc_ifindex != 0, NULL, flags);
3411 if (dev != res.nh->fib_nh_dev)
3412 err = -EHOSTUNREACH;
3414 *_dev = dev = res.nh->fib_nh_dev;
3416 *idev = in6_dev_get(dev);
3422 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3423 struct net_device **_dev, struct inet6_dev **idev,
3424 struct netlink_ext_ack *extack)
3426 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3427 int gwa_type = ipv6_addr_type(gw_addr);
3428 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3429 const struct net_device *dev = *_dev;
3430 bool need_addr_check = !dev;
3433 /* if gw_addr is local we will fail to detect this in case
3434 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3435 * will return already-added prefix route via interface that
3436 * prefix route was assigned to, which might be non-loopback.
3439 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3440 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3444 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3445 /* IPv6 strictly inhibits using not link-local
3446 * addresses as nexthop address.
3447 * Otherwise, router will not able to send redirects.
3448 * It is very good, but in some (rare!) circumstances
3449 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3450 * some exceptions. --ANK
3451 * We allow IPv4-mapped nexthops to support RFC4798-type
3454 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3455 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3461 if (cfg->fc_flags & RTNH_F_ONLINK)
3462 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3464 err = ip6_route_check_nh(net, cfg, _dev, idev);
3472 /* reload in case device was changed */
3477 NL_SET_ERR_MSG(extack, "Egress device not specified");
3479 } else if (dev->flags & IFF_LOOPBACK) {
3480 NL_SET_ERR_MSG(extack,
3481 "Egress device can not be loopback device for this route");
3485 /* if we did not check gw_addr above, do so now that the
3486 * egress device has been resolved.
3488 if (need_addr_check &&
3489 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3490 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3499 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3501 if ((flags & RTF_REJECT) ||
3502 (dev && (dev->flags & IFF_LOOPBACK) &&
3503 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3504 !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3510 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3511 struct fib6_config *cfg, gfp_t gfp_flags,
3512 struct netlink_ext_ack *extack)
3514 struct net_device *dev = NULL;
3515 struct inet6_dev *idev = NULL;
3519 fib6_nh->fib_nh_family = AF_INET6;
3520 #ifdef CONFIG_IPV6_ROUTER_PREF
3521 fib6_nh->last_probe = jiffies;
3523 if (cfg->fc_is_fdb) {
3524 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3525 fib6_nh->fib_nh_gw_family = AF_INET6;
3530 if (cfg->fc_ifindex) {
3531 dev = dev_get_by_index(net, cfg->fc_ifindex);
3534 idev = in6_dev_get(dev);
3539 if (cfg->fc_flags & RTNH_F_ONLINK) {
3541 NL_SET_ERR_MSG(extack,
3542 "Nexthop device required for onlink");
3546 if (!(dev->flags & IFF_UP)) {
3547 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3552 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3555 fib6_nh->fib_nh_weight = 1;
3557 /* We cannot add true routes via loopback here,
3558 * they would result in kernel looping; promote them to reject routes
3560 addr_type = ipv6_addr_type(&cfg->fc_dst);
3561 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3562 /* hold loopback dev/idev if we haven't done so. */
3563 if (dev != net->loopback_dev) {
3568 dev = net->loopback_dev;
3570 idev = in6_dev_get(dev);
3579 if (cfg->fc_flags & RTF_GATEWAY) {
3580 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3584 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3585 fib6_nh->fib_nh_gw_family = AF_INET6;
3592 if (idev->cnf.disable_ipv6) {
3593 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3598 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3599 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3604 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3605 !netif_carrier_ok(dev))
3606 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3608 err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3609 cfg->fc_encap_type, cfg, gfp_flags, extack);
3614 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3615 if (!fib6_nh->rt6i_pcpu) {
3620 fib6_nh->fib_nh_dev = dev;
3621 netdev_tracker_alloc(dev, &fib6_nh->fib_nh_dev_tracker, gfp_flags);
3623 fib6_nh->fib_nh_oif = dev->ifindex;
3630 lwtstate_put(fib6_nh->fib_nh_lws);
3631 fib6_nh->fib_nh_lws = NULL;
3638 void fib6_nh_release(struct fib6_nh *fib6_nh)
3640 struct rt6_exception_bucket *bucket;
3644 fib6_nh_flush_exceptions(fib6_nh, NULL);
3645 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3647 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3653 fib6_nh_release_dsts(fib6_nh);
3654 free_percpu(fib6_nh->rt6i_pcpu);
3656 fib_nh_common_release(&fib6_nh->nh_common);
3659 void fib6_nh_release_dsts(struct fib6_nh *fib6_nh)
3663 if (!fib6_nh->rt6i_pcpu)
3666 for_each_possible_cpu(cpu) {
3667 struct rt6_info *pcpu_rt, **ppcpu_rt;
3669 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3670 pcpu_rt = xchg(ppcpu_rt, NULL);
3672 dst_dev_put(&pcpu_rt->dst);
3673 dst_release(&pcpu_rt->dst);
3678 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3680 struct netlink_ext_ack *extack)
3682 struct net *net = cfg->fc_nlinfo.nl_net;
3683 struct fib6_info *rt = NULL;
3684 struct nexthop *nh = NULL;
3685 struct fib6_table *table;
3686 struct fib6_nh *fib6_nh;
3690 /* RTF_PCPU is an internal flag; can not be set by userspace */
3691 if (cfg->fc_flags & RTF_PCPU) {
3692 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3696 /* RTF_CACHE is an internal flag; can not be set by userspace */
3697 if (cfg->fc_flags & RTF_CACHE) {
3698 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3702 if (cfg->fc_type > RTN_MAX) {
3703 NL_SET_ERR_MSG(extack, "Invalid route type");
3707 if (cfg->fc_dst_len > 128) {
3708 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3711 if (cfg->fc_src_len > 128) {
3712 NL_SET_ERR_MSG(extack, "Invalid source address length");
3715 #ifndef CONFIG_IPV6_SUBTREES
3716 if (cfg->fc_src_len) {
3717 NL_SET_ERR_MSG(extack,
3718 "Specifying source address requires IPV6_SUBTREES to be enabled");
3722 if (cfg->fc_nh_id) {
3723 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3725 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3728 err = fib6_check_nexthop(nh, cfg, extack);
3734 if (cfg->fc_nlinfo.nlh &&
3735 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3736 table = fib6_get_table(net, cfg->fc_table);
3738 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3739 table = fib6_new_table(net, cfg->fc_table);
3742 table = fib6_new_table(net, cfg->fc_table);
3749 rt = fib6_info_alloc(gfp_flags, !nh);
3753 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3755 if (IS_ERR(rt->fib6_metrics)) {
3756 err = PTR_ERR(rt->fib6_metrics);
3757 /* Do not leave garbage there. */
3758 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3762 if (cfg->fc_flags & RTF_ADDRCONF)
3763 rt->dst_nocount = true;
3765 if (cfg->fc_flags & RTF_EXPIRES)
3766 fib6_set_expires(rt, jiffies +
3767 clock_t_to_jiffies(cfg->fc_expires));
3769 fib6_clean_expires(rt);
3771 if (cfg->fc_protocol == RTPROT_UNSPEC)
3772 cfg->fc_protocol = RTPROT_BOOT;
3773 rt->fib6_protocol = cfg->fc_protocol;
3775 rt->fib6_table = table;
3776 rt->fib6_metric = cfg->fc_metric;
3777 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3778 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3780 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3781 rt->fib6_dst.plen = cfg->fc_dst_len;
3783 #ifdef CONFIG_IPV6_SUBTREES
3784 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3785 rt->fib6_src.plen = cfg->fc_src_len;
3788 if (rt->fib6_src.plen) {
3789 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3792 if (!nexthop_get(nh)) {
3793 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3797 fib6_nh = nexthop_fib6_nh(rt->nh);
3799 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3803 fib6_nh = rt->fib6_nh;
3805 /* We cannot add true routes via loopback here, they would
3806 * result in kernel looping; promote them to reject routes
3808 addr_type = ipv6_addr_type(&cfg->fc_dst);
3809 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3811 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3814 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3815 struct net_device *dev = fib6_nh->fib_nh_dev;
3817 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3818 NL_SET_ERR_MSG(extack, "Invalid source address");
3822 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3823 rt->fib6_prefsrc.plen = 128;
3825 rt->fib6_prefsrc.plen = 0;
3829 fib6_info_release(rt);
3830 return ERR_PTR(err);
3832 ip_fib_metrics_put(rt->fib6_metrics);
3834 return ERR_PTR(err);
3837 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3838 struct netlink_ext_ack *extack)
3840 struct fib6_info *rt;
3843 rt = ip6_route_info_create(cfg, gfp_flags, extack);
3847 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3848 fib6_info_release(rt);
3853 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3855 struct net *net = info->nl_net;
3856 struct fib6_table *table;
3859 if (rt == net->ipv6.fib6_null_entry) {
3864 table = rt->fib6_table;
3865 spin_lock_bh(&table->tb6_lock);
3866 err = fib6_del(rt, info);
3867 spin_unlock_bh(&table->tb6_lock);
3870 fib6_info_release(rt);
3874 int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
3876 struct nl_info info = {
3878 .skip_notify = skip_notify
3881 return __ip6_del_rt(rt, &info);
3884 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3886 struct nl_info *info = &cfg->fc_nlinfo;
3887 struct net *net = info->nl_net;
3888 struct sk_buff *skb = NULL;
3889 struct fib6_table *table;
3892 if (rt == net->ipv6.fib6_null_entry)
3894 table = rt->fib6_table;
3895 spin_lock_bh(&table->tb6_lock);
3897 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3898 struct fib6_info *sibling, *next_sibling;
3899 struct fib6_node *fn;
3901 /* prefer to send a single notification with all hops */
3902 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3904 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3906 if (rt6_fill_node(net, skb, rt, NULL,
3907 NULL, NULL, 0, RTM_DELROUTE,
3908 info->portid, seq, 0) < 0) {
3912 info->skip_notify = 1;
3915 /* 'rt' points to the first sibling route. If it is not the
3916 * leaf, then we do not need to send a notification. Otherwise,
3917 * we need to check if the last sibling has a next route or not
3918 * and emit a replace or delete notification, respectively.
3920 info->skip_notify_kernel = 1;
3921 fn = rcu_dereference_protected(rt->fib6_node,
3922 lockdep_is_held(&table->tb6_lock));
3923 if (rcu_access_pointer(fn->leaf) == rt) {
3924 struct fib6_info *last_sibling, *replace_rt;
3926 last_sibling = list_last_entry(&rt->fib6_siblings,
3929 replace_rt = rcu_dereference_protected(
3930 last_sibling->fib6_next,
3931 lockdep_is_held(&table->tb6_lock));
3933 call_fib6_entry_notifiers_replace(net,
3936 call_fib6_multipath_entry_notifiers(net,
3937 FIB_EVENT_ENTRY_DEL,
3938 rt, rt->fib6_nsiblings,
3941 list_for_each_entry_safe(sibling, next_sibling,
3944 err = fib6_del(sibling, info);
3950 err = fib6_del(rt, info);
3952 spin_unlock_bh(&table->tb6_lock);
3954 fib6_info_release(rt);
3957 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3958 info->nlh, gfp_any());
3963 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3967 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3970 if (cfg->fc_flags & RTF_GATEWAY &&
3971 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3974 rc = rt6_remove_exception_rt(rt);
3979 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3982 struct fib6_result res = {
3986 struct rt6_info *rt_cache;
3988 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3990 return __ip6_del_cached_rt(rt_cache, cfg);
3995 struct fib6_nh_del_cached_rt_arg {
3996 struct fib6_config *cfg;
3997 struct fib6_info *f6i;
4000 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
4002 struct fib6_nh_del_cached_rt_arg *arg = _arg;
4005 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
4006 return rc != -ESRCH ? rc : 0;
4009 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
4011 struct fib6_nh_del_cached_rt_arg arg = {
4016 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
4019 static int ip6_route_del(struct fib6_config *cfg,
4020 struct netlink_ext_ack *extack)
4022 struct fib6_table *table;
4023 struct fib6_info *rt;
4024 struct fib6_node *fn;
4027 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
4029 NL_SET_ERR_MSG(extack, "FIB table does not exist");
4035 fn = fib6_locate(&table->tb6_root,
4036 &cfg->fc_dst, cfg->fc_dst_len,
4037 &cfg->fc_src, cfg->fc_src_len,
4038 !(cfg->fc_flags & RTF_CACHE));
4041 for_each_fib6_node_rt_rcu(fn) {
4044 if (rt->nh && cfg->fc_nh_id &&
4045 rt->nh->id != cfg->fc_nh_id)
4048 if (cfg->fc_flags & RTF_CACHE) {
4052 rc = ip6_del_cached_rt_nh(cfg, rt);
4053 } else if (cfg->fc_nh_id) {
4057 rc = ip6_del_cached_rt(cfg, rt, nh);
4066 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
4068 if (cfg->fc_protocol &&
4069 cfg->fc_protocol != rt->fib6_protocol)
4073 if (!fib6_info_hold_safe(rt))
4077 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4083 if (cfg->fc_ifindex &&
4085 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
4087 if (cfg->fc_flags & RTF_GATEWAY &&
4088 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
4090 if (!fib6_info_hold_safe(rt))
4094 /* if gateway was specified only delete the one hop */
4095 if (cfg->fc_flags & RTF_GATEWAY)
4096 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4098 return __ip6_del_rt_siblings(rt, cfg);
4106 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
4108 struct netevent_redirect netevent;
4109 struct rt6_info *rt, *nrt = NULL;
4110 struct fib6_result res = {};
4111 struct ndisc_options ndopts;
4112 struct inet6_dev *in6_dev;
4113 struct neighbour *neigh;
4115 int optlen, on_link;
4118 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
4119 optlen -= sizeof(*msg);
4122 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4126 msg = (struct rd_msg *)icmp6_hdr(skb);
4128 if (ipv6_addr_is_multicast(&msg->dest)) {
4129 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4134 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4136 } else if (ipv6_addr_type(&msg->target) !=
4137 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4138 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4142 in6_dev = __in6_dev_get(skb->dev);
4145 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
4149 * The IP source address of the Redirect MUST be the same as the current
4150 * first-hop router for the specified ICMP Destination Address.
4153 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4154 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4159 if (ndopts.nd_opts_tgt_lladdr) {
4160 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4163 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4168 rt = (struct rt6_info *) dst;
4169 if (rt->rt6i_flags & RTF_REJECT) {
4170 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4174 /* Redirect received -> path was valid.
4175 * Look, redirects are sent only in response to data packets,
4176 * so that this nexthop apparently is reachable. --ANK
4178 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4180 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4185 * We have finally decided to accept it.
4188 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4189 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4190 NEIGH_UPDATE_F_OVERRIDE|
4191 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4192 NEIGH_UPDATE_F_ISROUTER)),
4193 NDISC_REDIRECT, &ndopts);
4196 res.f6i = rcu_dereference(rt->from);
4201 struct fib6_nh_match_arg arg = {
4203 .gw = &rt->rt6i_gateway,
4206 nexthop_for_each_fib6_nh(res.f6i->nh,
4207 fib6_nh_find_match, &arg);
4209 /* fib6_info uses a nexthop that does not have fib6_nh
4210 * using the dst->dev. Should be impossible
4216 res.nh = res.f6i->fib6_nh;
4219 res.fib6_flags = res.f6i->fib6_flags;
4220 res.fib6_type = res.f6i->fib6_type;
4221 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4225 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4227 nrt->rt6i_flags &= ~RTF_GATEWAY;
4229 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4231 /* rt6_insert_exception() will take care of duplicated exceptions */
4232 if (rt6_insert_exception(nrt, &res)) {
4233 dst_release_immediate(&nrt->dst);
4237 netevent.old = &rt->dst;
4238 netevent.new = &nrt->dst;
4239 netevent.daddr = &msg->dest;
4240 netevent.neigh = neigh;
4241 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4245 neigh_release(neigh);
4248 #ifdef CONFIG_IPV6_ROUTE_INFO
4249 static struct fib6_info *rt6_get_route_info(struct net *net,
4250 const struct in6_addr *prefix, int prefixlen,
4251 const struct in6_addr *gwaddr,
4252 struct net_device *dev)
4254 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4255 int ifindex = dev->ifindex;
4256 struct fib6_node *fn;
4257 struct fib6_info *rt = NULL;
4258 struct fib6_table *table;
4260 table = fib6_get_table(net, tb_id);
4265 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4269 for_each_fib6_node_rt_rcu(fn) {
4270 /* these routes do not use nexthops */
4273 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4275 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4276 !rt->fib6_nh->fib_nh_gw_family)
4278 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4280 if (!fib6_info_hold_safe(rt))
4289 static struct fib6_info *rt6_add_route_info(struct net *net,
4290 const struct in6_addr *prefix, int prefixlen,
4291 const struct in6_addr *gwaddr,
4292 struct net_device *dev,
4295 struct fib6_config cfg = {
4296 .fc_metric = IP6_RT_PRIO_USER,
4297 .fc_ifindex = dev->ifindex,
4298 .fc_dst_len = prefixlen,
4299 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4300 RTF_UP | RTF_PREF(pref),
4301 .fc_protocol = RTPROT_RA,
4302 .fc_type = RTN_UNICAST,
4303 .fc_nlinfo.portid = 0,
4304 .fc_nlinfo.nlh = NULL,
4305 .fc_nlinfo.nl_net = net,
4308 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4309 cfg.fc_dst = *prefix;
4310 cfg.fc_gateway = *gwaddr;
4312 /* We should treat it as a default route if prefix length is 0. */
4314 cfg.fc_flags |= RTF_DEFAULT;
4316 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4318 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4322 struct fib6_info *rt6_get_dflt_router(struct net *net,
4323 const struct in6_addr *addr,
4324 struct net_device *dev)
4326 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4327 struct fib6_info *rt;
4328 struct fib6_table *table;
4330 table = fib6_get_table(net, tb_id);
4335 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4338 /* RA routes do not use nexthops */
4343 if (dev == nh->fib_nh_dev &&
4344 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4345 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4348 if (rt && !fib6_info_hold_safe(rt))
4354 struct fib6_info *rt6_add_dflt_router(struct net *net,
4355 const struct in6_addr *gwaddr,
4356 struct net_device *dev,
4358 u32 defrtr_usr_metric)
4360 struct fib6_config cfg = {
4361 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4362 .fc_metric = defrtr_usr_metric,
4363 .fc_ifindex = dev->ifindex,
4364 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4365 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4366 .fc_protocol = RTPROT_RA,
4367 .fc_type = RTN_UNICAST,
4368 .fc_nlinfo.portid = 0,
4369 .fc_nlinfo.nlh = NULL,
4370 .fc_nlinfo.nl_net = net,
4373 cfg.fc_gateway = *gwaddr;
4375 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4376 struct fib6_table *table;
4378 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4380 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4383 return rt6_get_dflt_router(net, gwaddr, dev);
4386 static void __rt6_purge_dflt_routers(struct net *net,
4387 struct fib6_table *table)
4389 struct fib6_info *rt;
4393 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4394 struct net_device *dev = fib6_info_nh_dev(rt);
4395 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4397 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4398 (!idev || idev->cnf.accept_ra != 2) &&
4399 fib6_info_hold_safe(rt)) {
4401 ip6_del_rt(net, rt, false);
4407 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4410 void rt6_purge_dflt_routers(struct net *net)
4412 struct fib6_table *table;
4413 struct hlist_head *head;
4418 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4419 head = &net->ipv6.fib_table_hash[h];
4420 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4421 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4422 __rt6_purge_dflt_routers(net, table);
4429 static void rtmsg_to_fib6_config(struct net *net,
4430 struct in6_rtmsg *rtmsg,
4431 struct fib6_config *cfg)
4433 *cfg = (struct fib6_config){
4434 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4436 .fc_ifindex = rtmsg->rtmsg_ifindex,
4437 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
4438 .fc_expires = rtmsg->rtmsg_info,
4439 .fc_dst_len = rtmsg->rtmsg_dst_len,
4440 .fc_src_len = rtmsg->rtmsg_src_len,
4441 .fc_flags = rtmsg->rtmsg_flags,
4442 .fc_type = rtmsg->rtmsg_type,
4444 .fc_nlinfo.nl_net = net,
4446 .fc_dst = rtmsg->rtmsg_dst,
4447 .fc_src = rtmsg->rtmsg_src,
4448 .fc_gateway = rtmsg->rtmsg_gateway,
4452 int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
4454 struct fib6_config cfg;
4457 if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4459 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4462 rtmsg_to_fib6_config(net, rtmsg, &cfg);
4467 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4470 err = ip6_route_del(&cfg, NULL);
4478 * Drop the packet on the floor
4481 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4483 struct dst_entry *dst = skb_dst(skb);
4484 struct net *net = dev_net(dst->dev);
4485 struct inet6_dev *idev;
4489 if (netif_is_l3_master(skb->dev) ||
4490 dst->dev == net->loopback_dev)
4491 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4493 idev = ip6_dst_idev(dst);
4495 switch (ipstats_mib_noroutes) {
4496 case IPSTATS_MIB_INNOROUTES:
4497 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4498 if (type == IPV6_ADDR_ANY) {
4499 SKB_DR_SET(reason, IP_INADDRERRORS);
4500 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4503 SKB_DR_SET(reason, IP_INNOROUTES);
4505 case IPSTATS_MIB_OUTNOROUTES:
4506 SKB_DR_OR(reason, IP_OUTNOROUTES);
4507 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4511 /* Start over by dropping the dst for l3mdev case */
4512 if (netif_is_l3_master(skb->dev))
4515 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4516 kfree_skb_reason(skb, reason);
4520 static int ip6_pkt_discard(struct sk_buff *skb)
4522 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4525 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4527 skb->dev = skb_dst(skb)->dev;
4528 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4531 static int ip6_pkt_prohibit(struct sk_buff *skb)
4533 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4536 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4538 skb->dev = skb_dst(skb)->dev;
4539 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4543 * Allocate a dst for local (unicast / anycast) address.
4546 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4547 struct inet6_dev *idev,
4548 const struct in6_addr *addr,
4549 bool anycast, gfp_t gfp_flags)
4551 struct fib6_config cfg = {
4552 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4553 .fc_ifindex = idev->dev->ifindex,
4554 .fc_flags = RTF_UP | RTF_NONEXTHOP,
4557 .fc_protocol = RTPROT_KERNEL,
4558 .fc_nlinfo.nl_net = net,
4559 .fc_ignore_dev_down = true,
4561 struct fib6_info *f6i;
4564 cfg.fc_type = RTN_ANYCAST;
4565 cfg.fc_flags |= RTF_ANYCAST;
4567 cfg.fc_type = RTN_LOCAL;
4568 cfg.fc_flags |= RTF_LOCAL;
4571 f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
4573 f6i->dst_nocount = true;
4576 (net->ipv6.devconf_all->disable_policy ||
4577 idev->cnf.disable_policy))
4578 f6i->dst_nopolicy = true;
4584 /* remove deleted ip from prefsrc entries */
4585 struct arg_dev_net_ip {
4586 struct net_device *dev;
4588 struct in6_addr *addr;
4591 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4593 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4594 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4595 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4598 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
4599 rt != net->ipv6.fib6_null_entry &&
4600 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
4601 spin_lock_bh(&rt6_exception_lock);
4602 /* remove prefsrc entry */
4603 rt->fib6_prefsrc.plen = 0;
4604 spin_unlock_bh(&rt6_exception_lock);
4609 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4611 struct net *net = dev_net(ifp->idev->dev);
4612 struct arg_dev_net_ip adni = {
4613 .dev = ifp->idev->dev,
4617 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4620 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
4622 /* Remove routers and update dst entries when gateway turn into host. */
4623 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4625 struct in6_addr *gateway = (struct in6_addr *)arg;
4628 /* RA routes do not use nexthops */
4633 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4634 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4637 /* Further clean up cached routes in exception table.
4638 * This is needed because cached route may have a different
4639 * gateway than its 'parent' in the case of an ip redirect.
4641 fib6_nh_exceptions_clean_tohost(nh, gateway);
4646 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4648 fib6_clean_all(net, fib6_clean_tohost, gateway);
4651 struct arg_netdev_event {
4652 const struct net_device *dev;
4654 unsigned char nh_flags;
4655 unsigned long event;
4659 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4661 struct fib6_info *iter;
4662 struct fib6_node *fn;
4664 fn = rcu_dereference_protected(rt->fib6_node,
4665 lockdep_is_held(&rt->fib6_table->tb6_lock));
4666 iter = rcu_dereference_protected(fn->leaf,
4667 lockdep_is_held(&rt->fib6_table->tb6_lock));
4669 if (iter->fib6_metric == rt->fib6_metric &&
4670 rt6_qualify_for_ecmp(iter))
4672 iter = rcu_dereference_protected(iter->fib6_next,
4673 lockdep_is_held(&rt->fib6_table->tb6_lock));
4679 /* only called for fib entries with builtin fib6_nh */
4680 static bool rt6_is_dead(const struct fib6_info *rt)
4682 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4683 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4684 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4690 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4692 struct fib6_info *iter;
4695 if (!rt6_is_dead(rt))
4696 total += rt->fib6_nh->fib_nh_weight;
4698 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4699 if (!rt6_is_dead(iter))
4700 total += iter->fib6_nh->fib_nh_weight;
4706 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4708 int upper_bound = -1;
4710 if (!rt6_is_dead(rt)) {
4711 *weight += rt->fib6_nh->fib_nh_weight;
4712 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4715 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4718 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4720 struct fib6_info *iter;
4723 rt6_upper_bound_set(rt, &weight, total);
4725 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4726 rt6_upper_bound_set(iter, &weight, total);
4729 void rt6_multipath_rebalance(struct fib6_info *rt)
4731 struct fib6_info *first;
4734 /* In case the entire multipath route was marked for flushing,
4735 * then there is no need to rebalance upon the removal of every
4738 if (!rt->fib6_nsiblings || rt->should_flush)
4741 /* During lookup routes are evaluated in order, so we need to
4742 * make sure upper bounds are assigned from the first sibling
4745 first = rt6_multipath_first_sibling(rt);
4746 if (WARN_ON_ONCE(!first))
4749 total = rt6_multipath_total_weight(first);
4750 rt6_multipath_upper_bound_set(first, total);
4753 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4755 const struct arg_netdev_event *arg = p_arg;
4756 struct net *net = dev_net(arg->dev);
4758 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4759 rt->fib6_nh->fib_nh_dev == arg->dev) {
4760 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4761 fib6_update_sernum_upto_root(net, rt);
4762 rt6_multipath_rebalance(rt);
4768 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4770 struct arg_netdev_event arg = {
4773 .nh_flags = nh_flags,
4777 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4778 arg.nh_flags |= RTNH_F_LINKDOWN;
4780 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4783 /* only called for fib entries with inline fib6_nh */
4784 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4785 const struct net_device *dev)
4787 struct fib6_info *iter;
4789 if (rt->fib6_nh->fib_nh_dev == dev)
4791 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4792 if (iter->fib6_nh->fib_nh_dev == dev)
4798 static void rt6_multipath_flush(struct fib6_info *rt)
4800 struct fib6_info *iter;
4802 rt->should_flush = 1;
4803 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4804 iter->should_flush = 1;
4807 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4808 const struct net_device *down_dev)
4810 struct fib6_info *iter;
4811 unsigned int dead = 0;
4813 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4814 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4816 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4817 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4818 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4824 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4825 const struct net_device *dev,
4826 unsigned char nh_flags)
4828 struct fib6_info *iter;
4830 if (rt->fib6_nh->fib_nh_dev == dev)
4831 rt->fib6_nh->fib_nh_flags |= nh_flags;
4832 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4833 if (iter->fib6_nh->fib_nh_dev == dev)
4834 iter->fib6_nh->fib_nh_flags |= nh_flags;
4837 /* called with write lock held for table with rt */
4838 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4840 const struct arg_netdev_event *arg = p_arg;
4841 const struct net_device *dev = arg->dev;
4842 struct net *net = dev_net(dev);
4844 if (rt == net->ipv6.fib6_null_entry || rt->nh)
4847 switch (arg->event) {
4848 case NETDEV_UNREGISTER:
4849 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4851 if (rt->should_flush)
4853 if (!rt->fib6_nsiblings)
4854 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4855 if (rt6_multipath_uses_dev(rt, dev)) {
4858 count = rt6_multipath_dead_count(rt, dev);
4859 if (rt->fib6_nsiblings + 1 == count) {
4860 rt6_multipath_flush(rt);
4863 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4865 fib6_update_sernum(net, rt);
4866 rt6_multipath_rebalance(rt);
4870 if (rt->fib6_nh->fib_nh_dev != dev ||
4871 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4873 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4874 rt6_multipath_rebalance(rt);
4881 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4883 struct arg_netdev_event arg = {
4889 struct net *net = dev_net(dev);
4891 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4892 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4894 fib6_clean_all(net, fib6_ifdown, &arg);
4897 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4899 rt6_sync_down_dev(dev, event);
4900 rt6_uncached_list_flush_dev(dev);
4901 neigh_ifdown(&nd_tbl, dev);
4904 struct rt6_mtu_change_arg {
4905 struct net_device *dev;
4907 struct fib6_info *f6i;
4910 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4912 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4913 struct fib6_info *f6i = arg->f6i;
4915 /* For administrative MTU increase, there is no way to discover
4916 * IPv6 PMTU increase, so PMTU increase should be updated here.
4917 * Since RFC 1981 doesn't include administrative MTU increase
4918 * update PMTU increase is a MUST. (i.e. jumbo frame)
4920 if (nh->fib_nh_dev == arg->dev) {
4921 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4922 u32 mtu = f6i->fib6_pmtu;
4924 if (mtu >= arg->mtu ||
4925 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4926 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4928 spin_lock_bh(&rt6_exception_lock);
4929 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4930 spin_unlock_bh(&rt6_exception_lock);
4936 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4938 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4939 struct inet6_dev *idev;
4941 /* In IPv6 pmtu discovery is not optional,
4942 so that RTAX_MTU lock cannot disable it.
4943 We still use this lock to block changes
4944 caused by addrconf/ndisc.
4947 idev = __in6_dev_get(arg->dev);
4951 if (fib6_metric_locked(f6i, RTAX_MTU))
4956 /* fib6_nh_mtu_change only returns 0, so this is safe */
4957 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4961 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4964 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4966 struct rt6_mtu_change_arg arg = {
4971 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4974 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4975 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
4976 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
4977 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
4978 [RTA_OIF] = { .type = NLA_U32 },
4979 [RTA_IIF] = { .type = NLA_U32 },
4980 [RTA_PRIORITY] = { .type = NLA_U32 },
4981 [RTA_METRICS] = { .type = NLA_NESTED },
4982 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
4983 [RTA_PREF] = { .type = NLA_U8 },
4984 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4985 [RTA_ENCAP] = { .type = NLA_NESTED },
4986 [RTA_EXPIRES] = { .type = NLA_U32 },
4987 [RTA_UID] = { .type = NLA_U32 },
4988 [RTA_MARK] = { .type = NLA_U32 },
4989 [RTA_TABLE] = { .type = NLA_U32 },
4990 [RTA_IP_PROTO] = { .type = NLA_U8 },
4991 [RTA_SPORT] = { .type = NLA_U16 },
4992 [RTA_DPORT] = { .type = NLA_U16 },
4993 [RTA_NH_ID] = { .type = NLA_U32 },
4996 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
4997 struct fib6_config *cfg,
4998 struct netlink_ext_ack *extack)
5001 struct nlattr *tb[RTA_MAX+1];
5005 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5006 rtm_ipv6_policy, extack);
5011 rtm = nlmsg_data(nlh);
5014 NL_SET_ERR_MSG(extack,
5015 "Invalid dsfield (tos): option not available for IPv6");
5019 *cfg = (struct fib6_config){
5020 .fc_table = rtm->rtm_table,
5021 .fc_dst_len = rtm->rtm_dst_len,
5022 .fc_src_len = rtm->rtm_src_len,
5024 .fc_protocol = rtm->rtm_protocol,
5025 .fc_type = rtm->rtm_type,
5027 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
5028 .fc_nlinfo.nlh = nlh,
5029 .fc_nlinfo.nl_net = sock_net(skb->sk),
5032 if (rtm->rtm_type == RTN_UNREACHABLE ||
5033 rtm->rtm_type == RTN_BLACKHOLE ||
5034 rtm->rtm_type == RTN_PROHIBIT ||
5035 rtm->rtm_type == RTN_THROW)
5036 cfg->fc_flags |= RTF_REJECT;
5038 if (rtm->rtm_type == RTN_LOCAL)
5039 cfg->fc_flags |= RTF_LOCAL;
5041 if (rtm->rtm_flags & RTM_F_CLONED)
5042 cfg->fc_flags |= RTF_CACHE;
5044 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5046 if (tb[RTA_NH_ID]) {
5047 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
5048 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5049 NL_SET_ERR_MSG(extack,
5050 "Nexthop specification and nexthop id are mutually exclusive");
5053 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5056 if (tb[RTA_GATEWAY]) {
5057 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
5058 cfg->fc_flags |= RTF_GATEWAY;
5061 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5066 int plen = (rtm->rtm_dst_len + 7) >> 3;
5068 if (nla_len(tb[RTA_DST]) < plen)
5071 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
5075 int plen = (rtm->rtm_src_len + 7) >> 3;
5077 if (nla_len(tb[RTA_SRC]) < plen)
5080 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
5083 if (tb[RTA_PREFSRC])
5084 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
5087 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5089 if (tb[RTA_PRIORITY])
5090 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5092 if (tb[RTA_METRICS]) {
5093 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5094 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
5098 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5100 if (tb[RTA_MULTIPATH]) {
5101 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5102 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
5104 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
5105 cfg->fc_mp_len, extack);
5111 pref = nla_get_u8(tb[RTA_PREF]);
5112 if (pref != ICMPV6_ROUTER_PREF_LOW &&
5113 pref != ICMPV6_ROUTER_PREF_HIGH)
5114 pref = ICMPV6_ROUTER_PREF_MEDIUM;
5115 cfg->fc_flags |= RTF_PREF(pref);
5119 cfg->fc_encap = tb[RTA_ENCAP];
5121 if (tb[RTA_ENCAP_TYPE]) {
5122 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5124 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5129 if (tb[RTA_EXPIRES]) {
5130 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5132 if (addrconf_finite_timeout(timeout)) {
5133 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5134 cfg->fc_flags |= RTF_EXPIRES;
5144 struct fib6_info *fib6_info;
5145 struct fib6_config r_cfg;
5146 struct list_head next;
5149 static int ip6_route_info_append(struct net *net,
5150 struct list_head *rt6_nh_list,
5151 struct fib6_info *rt,
5152 struct fib6_config *r_cfg)
5157 list_for_each_entry(nh, rt6_nh_list, next) {
5158 /* check if fib6_info already exists */
5159 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5163 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5167 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5168 list_add_tail(&nh->next, rt6_nh_list);
5173 static void ip6_route_mpath_notify(struct fib6_info *rt,
5174 struct fib6_info *rt_last,
5175 struct nl_info *info,
5178 /* if this is an APPEND route, then rt points to the first route
5179 * inserted and rt_last points to last route inserted. Userspace
5180 * wants a consistent dump of the route which starts at the first
5181 * nexthop. Since sibling routes are always added at the end of
5182 * the list, find the first sibling of the last route appended
5184 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5185 rt = list_first_entry(&rt_last->fib6_siblings,
5191 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5194 static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5196 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5197 bool should_notify = false;
5198 struct fib6_info *leaf;
5199 struct fib6_node *fn;
5202 fn = rcu_dereference(rt->fib6_node);
5206 leaf = rcu_dereference(fn->leaf);
5211 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5212 rt6_qualify_for_ecmp(leaf)))
5213 should_notify = true;
5217 return should_notify;
5220 static int fib6_gw_from_attr(struct in6_addr *gw, struct nlattr *nla,
5221 struct netlink_ext_ack *extack)
5223 if (nla_len(nla) < sizeof(*gw)) {
5224 NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_GATEWAY");
5228 *gw = nla_get_in6_addr(nla);
5233 static int ip6_route_multipath_add(struct fib6_config *cfg,
5234 struct netlink_ext_ack *extack)
5236 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5237 struct nl_info *info = &cfg->fc_nlinfo;
5238 struct fib6_config r_cfg;
5239 struct rtnexthop *rtnh;
5240 struct fib6_info *rt;
5241 struct rt6_nh *err_nh;
5242 struct rt6_nh *nh, *nh_safe;
5248 int replace = (cfg->fc_nlinfo.nlh &&
5249 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5250 LIST_HEAD(rt6_nh_list);
5252 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5253 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5254 nlflags |= NLM_F_APPEND;
5256 remaining = cfg->fc_mp_len;
5257 rtnh = (struct rtnexthop *)cfg->fc_mp;
5259 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
5260 * fib6_info structs per nexthop
5262 while (rtnh_ok(rtnh, remaining)) {
5263 memcpy(&r_cfg, cfg, sizeof(*cfg));
5264 if (rtnh->rtnh_ifindex)
5265 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5267 attrlen = rtnh_attrlen(rtnh);
5269 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5271 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5273 err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5278 r_cfg.fc_flags |= RTF_GATEWAY;
5280 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5282 /* RTA_ENCAP_TYPE length checked in
5283 * lwtunnel_valid_encap_type_attr
5285 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5287 r_cfg.fc_encap_type = nla_get_u16(nla);
5290 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5291 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5297 if (!rt6_qualify_for_ecmp(rt)) {
5299 NL_SET_ERR_MSG(extack,
5300 "Device only routes can not be added for IPv6 using the multipath API.");
5301 fib6_info_release(rt);
5305 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5307 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5310 fib6_info_release(rt);
5314 rtnh = rtnh_next(rtnh, &remaining);
5317 if (list_empty(&rt6_nh_list)) {
5318 NL_SET_ERR_MSG(extack,
5319 "Invalid nexthop configuration - no valid nexthops");
5323 /* for add and replace send one notification with all nexthops.
5324 * Skip the notification in fib6_add_rt2node and send one with
5325 * the full route when done
5327 info->skip_notify = 1;
5329 /* For add and replace, send one notification with all nexthops. For
5330 * append, send one notification with all appended nexthops.
5332 info->skip_notify_kernel = 1;
5335 list_for_each_entry(nh, &rt6_nh_list, next) {
5336 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5337 fib6_info_release(nh->fib6_info);
5340 /* save reference to last route successfully inserted */
5341 rt_last = nh->fib6_info;
5343 /* save reference to first route for notification */
5345 rt_notif = nh->fib6_info;
5348 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5349 nh->fib6_info = NULL;
5352 NL_SET_ERR_MSG_MOD(extack,
5353 "multipath route replace failed (check consistency of installed routes)");
5358 /* Because each route is added like a single route we remove
5359 * these flags after the first nexthop: if there is a collision,
5360 * we have already failed to add the first nexthop:
5361 * fib6_add_rt2node() has rejected it; when replacing, old
5362 * nexthops have been replaced by first new, the rest should
5365 if (cfg->fc_nlinfo.nlh) {
5366 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5368 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5373 /* An in-kernel notification should only be sent in case the new
5374 * multipath route is added as the first route in the node, or if
5375 * it was appended to it. We pass 'rt_notif' since it is the first
5376 * sibling and might allow us to skip some checks in the replace case.
5378 if (ip6_route_mpath_should_notify(rt_notif)) {
5379 enum fib_event_type fib_event;
5381 if (rt_notif->fib6_nsiblings != nhn - 1)
5382 fib_event = FIB_EVENT_ENTRY_APPEND;
5384 fib_event = FIB_EVENT_ENTRY_REPLACE;
5386 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5387 fib_event, rt_notif,
5390 /* Delete all the siblings that were just added */
5396 /* success ... tell user about new route */
5397 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5401 /* send notification for routes that were added so that
5402 * the delete notifications sent by ip6_route_del are
5406 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5408 /* Delete routes that were already added */
5409 list_for_each_entry(nh, &rt6_nh_list, next) {
5412 ip6_route_del(&nh->r_cfg, extack);
5416 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5418 fib6_info_release(nh->fib6_info);
5419 list_del(&nh->next);
5426 static int ip6_route_multipath_del(struct fib6_config *cfg,
5427 struct netlink_ext_ack *extack)
5429 struct fib6_config r_cfg;
5430 struct rtnexthop *rtnh;
5436 remaining = cfg->fc_mp_len;
5437 rtnh = (struct rtnexthop *)cfg->fc_mp;
5439 /* Parse a Multipath Entry */
5440 while (rtnh_ok(rtnh, remaining)) {
5441 memcpy(&r_cfg, cfg, sizeof(*cfg));
5442 if (rtnh->rtnh_ifindex)
5443 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5445 attrlen = rtnh_attrlen(rtnh);
5447 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5449 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5451 err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5458 r_cfg.fc_flags |= RTF_GATEWAY;
5461 err = ip6_route_del(&r_cfg, extack);
5466 rtnh = rtnh_next(rtnh, &remaining);
5472 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5473 struct netlink_ext_ack *extack)
5475 struct fib6_config cfg;
5478 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5483 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5484 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5489 return ip6_route_multipath_del(&cfg, extack);
5491 cfg.fc_delete_all_nh = 1;
5492 return ip6_route_del(&cfg, extack);
5496 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5497 struct netlink_ext_ack *extack)
5499 struct fib6_config cfg;
5502 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5506 if (cfg.fc_metric == 0)
5507 cfg.fc_metric = IP6_RT_PRIO_USER;
5510 return ip6_route_multipath_add(&cfg, extack);
5512 return ip6_route_add(&cfg, GFP_KERNEL, extack);
5515 /* add the overhead of this fib6_nh to nexthop_len */
5516 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5518 int *nexthop_len = arg;
5520 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5521 + NLA_ALIGN(sizeof(struct rtnexthop))
5522 + nla_total_size(16); /* RTA_GATEWAY */
5524 if (nh->fib_nh_lws) {
5525 /* RTA_ENCAP_TYPE */
5526 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5528 *nexthop_len += nla_total_size(2);
5534 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5539 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5540 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5543 struct fib6_nh *nh = f6i->fib6_nh;
5546 if (f6i->fib6_nsiblings) {
5547 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
5548 + NLA_ALIGN(sizeof(struct rtnexthop))
5549 + nla_total_size(16) /* RTA_GATEWAY */
5550 + lwtunnel_get_encap_size(nh->fib_nh_lws);
5552 nexthop_len *= f6i->fib6_nsiblings;
5554 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5557 return NLMSG_ALIGN(sizeof(struct rtmsg))
5558 + nla_total_size(16) /* RTA_SRC */
5559 + nla_total_size(16) /* RTA_DST */
5560 + nla_total_size(16) /* RTA_GATEWAY */
5561 + nla_total_size(16) /* RTA_PREFSRC */
5562 + nla_total_size(4) /* RTA_TABLE */
5563 + nla_total_size(4) /* RTA_IIF */
5564 + nla_total_size(4) /* RTA_OIF */
5565 + nla_total_size(4) /* RTA_PRIORITY */
5566 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5567 + nla_total_size(sizeof(struct rta_cacheinfo))
5568 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5569 + nla_total_size(1) /* RTA_PREF */
5573 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5574 unsigned char *flags)
5576 if (nexthop_is_multipath(nh)) {
5579 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5581 goto nla_put_failure;
5583 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5584 goto nla_put_failure;
5586 nla_nest_end(skb, mp);
5588 struct fib6_nh *fib6_nh;
5590 fib6_nh = nexthop_fib6_nh(nh);
5591 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5593 goto nla_put_failure;
5602 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5603 struct fib6_info *rt, struct dst_entry *dst,
5604 struct in6_addr *dest, struct in6_addr *src,
5605 int iif, int type, u32 portid, u32 seq,
5608 struct rt6_info *rt6 = (struct rt6_info *)dst;
5609 struct rt6key *rt6_dst, *rt6_src;
5610 u32 *pmetrics, table, rt6_flags;
5611 unsigned char nh_flags = 0;
5612 struct nlmsghdr *nlh;
5616 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5621 rt6_dst = &rt6->rt6i_dst;
5622 rt6_src = &rt6->rt6i_src;
5623 rt6_flags = rt6->rt6i_flags;
5625 rt6_dst = &rt->fib6_dst;
5626 rt6_src = &rt->fib6_src;
5627 rt6_flags = rt->fib6_flags;
5630 rtm = nlmsg_data(nlh);
5631 rtm->rtm_family = AF_INET6;
5632 rtm->rtm_dst_len = rt6_dst->plen;
5633 rtm->rtm_src_len = rt6_src->plen;
5636 table = rt->fib6_table->tb6_id;
5638 table = RT6_TABLE_UNSPEC;
5639 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5640 if (nla_put_u32(skb, RTA_TABLE, table))
5641 goto nla_put_failure;
5643 rtm->rtm_type = rt->fib6_type;
5645 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5646 rtm->rtm_protocol = rt->fib6_protocol;
5648 if (rt6_flags & RTF_CACHE)
5649 rtm->rtm_flags |= RTM_F_CLONED;
5652 if (nla_put_in6_addr(skb, RTA_DST, dest))
5653 goto nla_put_failure;
5654 rtm->rtm_dst_len = 128;
5655 } else if (rtm->rtm_dst_len)
5656 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5657 goto nla_put_failure;
5658 #ifdef CONFIG_IPV6_SUBTREES
5660 if (nla_put_in6_addr(skb, RTA_SRC, src))
5661 goto nla_put_failure;
5662 rtm->rtm_src_len = 128;
5663 } else if (rtm->rtm_src_len &&
5664 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5665 goto nla_put_failure;
5668 #ifdef CONFIG_IPV6_MROUTE
5669 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5670 int err = ip6mr_get_route(net, skb, rtm, portid);
5675 goto nla_put_failure;
5678 if (nla_put_u32(skb, RTA_IIF, iif))
5679 goto nla_put_failure;
5681 struct in6_addr saddr_buf;
5682 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
5683 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5684 goto nla_put_failure;
5687 if (rt->fib6_prefsrc.plen) {
5688 struct in6_addr saddr_buf;
5689 saddr_buf = rt->fib6_prefsrc.addr;
5690 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5691 goto nla_put_failure;
5694 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5695 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5696 goto nla_put_failure;
5698 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5699 goto nla_put_failure;
5701 /* For multipath routes, walk the siblings list and add
5702 * each as a nexthop within RTA_MULTIPATH.
5705 if (rt6_flags & RTF_GATEWAY &&
5706 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5707 goto nla_put_failure;
5709 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5710 goto nla_put_failure;
5712 if (dst->lwtstate &&
5713 lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5714 goto nla_put_failure;
5715 } else if (rt->fib6_nsiblings) {
5716 struct fib6_info *sibling, *next_sibling;
5719 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5721 goto nla_put_failure;
5723 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5724 rt->fib6_nh->fib_nh_weight, AF_INET6,
5726 goto nla_put_failure;
5728 list_for_each_entry_safe(sibling, next_sibling,
5729 &rt->fib6_siblings, fib6_siblings) {
5730 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5731 sibling->fib6_nh->fib_nh_weight,
5733 goto nla_put_failure;
5736 nla_nest_end(skb, mp);
5737 } else if (rt->nh) {
5738 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5739 goto nla_put_failure;
5741 if (nexthop_is_blackhole(rt->nh))
5742 rtm->rtm_type = RTN_BLACKHOLE;
5744 if (READ_ONCE(net->ipv4.sysctl_nexthop_compat_mode) &&
5745 rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5746 goto nla_put_failure;
5748 rtm->rtm_flags |= nh_flags;
5750 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5751 &nh_flags, false) < 0)
5752 goto nla_put_failure;
5754 rtm->rtm_flags |= nh_flags;
5757 if (rt6_flags & RTF_EXPIRES) {
5758 expires = dst ? dst->expires : rt->expires;
5763 if (READ_ONCE(rt->offload))
5764 rtm->rtm_flags |= RTM_F_OFFLOAD;
5765 if (READ_ONCE(rt->trap))
5766 rtm->rtm_flags |= RTM_F_TRAP;
5767 if (READ_ONCE(rt->offload_failed))
5768 rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5771 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5772 goto nla_put_failure;
5774 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5775 goto nla_put_failure;
5778 nlmsg_end(skb, nlh);
5782 nlmsg_cancel(skb, nlh);
5786 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5788 const struct net_device *dev = arg;
5790 if (nh->fib_nh_dev == dev)
5796 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5797 const struct net_device *dev)
5800 struct net_device *_dev = (struct net_device *)dev;
5802 return !!nexthop_for_each_fib6_nh(f6i->nh,
5803 fib6_info_nh_uses_dev,
5807 if (f6i->fib6_nh->fib_nh_dev == dev)
5810 if (f6i->fib6_nsiblings) {
5811 struct fib6_info *sibling, *next_sibling;
5813 list_for_each_entry_safe(sibling, next_sibling,
5814 &f6i->fib6_siblings, fib6_siblings) {
5815 if (sibling->fib6_nh->fib_nh_dev == dev)
5823 struct fib6_nh_exception_dump_walker {
5824 struct rt6_rtnl_dump_arg *dump;
5825 struct fib6_info *rt;
5831 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5833 struct fib6_nh_exception_dump_walker *w = arg;
5834 struct rt6_rtnl_dump_arg *dump = w->dump;
5835 struct rt6_exception_bucket *bucket;
5836 struct rt6_exception *rt6_ex;
5839 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5843 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5844 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5850 /* Expiration of entries doesn't bump sernum, insertion
5851 * does. Removal is triggered by insertion, so we can
5852 * rely on the fact that if entries change between two
5853 * partial dumps, this node is scanned again completely,
5854 * see rt6_insert_exception() and fib6_dump_table().
5856 * Count expired entries we go through as handled
5857 * entries that we'll skip next time, in case of partial
5858 * node dump. Otherwise, if entries expire meanwhile,
5859 * we'll skip the wrong amount.
5861 if (rt6_check_expired(rt6_ex->rt6i)) {
5866 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5867 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5869 NETLINK_CB(dump->cb->skb).portid,
5870 dump->cb->nlh->nlmsg_seq, w->flags);
5882 /* Return -1 if done with node, number of handled routes on partial dump */
5883 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5885 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5886 struct fib_dump_filter *filter = &arg->filter;
5887 unsigned int flags = NLM_F_MULTI;
5888 struct net *net = arg->net;
5891 if (rt == net->ipv6.fib6_null_entry)
5894 if ((filter->flags & RTM_F_PREFIX) &&
5895 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5896 /* success since this is not a prefix route */
5899 if (filter->filter_set &&
5900 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5901 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5902 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5906 if (filter->filter_set ||
5907 !filter->dump_routes || !filter->dump_exceptions) {
5908 flags |= NLM_F_DUMP_FILTERED;
5911 if (filter->dump_routes) {
5915 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5917 NETLINK_CB(arg->cb->skb).portid,
5918 arg->cb->nlh->nlmsg_seq, flags)) {
5925 if (filter->dump_exceptions) {
5926 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5935 err = nexthop_for_each_fib6_nh(rt->nh,
5936 rt6_nh_dump_exceptions,
5939 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5944 return count += w.count;
5950 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5951 const struct nlmsghdr *nlh,
5953 struct netlink_ext_ack *extack)
5958 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5959 NL_SET_ERR_MSG_MOD(extack,
5960 "Invalid header for get route request");
5964 if (!netlink_strict_get_check(skb))
5965 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5966 rtm_ipv6_policy, extack);
5968 rtm = nlmsg_data(nlh);
5969 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5970 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5971 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5973 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5976 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5977 NL_SET_ERR_MSG_MOD(extack,
5978 "Invalid flags for get route request");
5982 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5983 rtm_ipv6_policy, extack);
5987 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5988 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5989 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5993 for (i = 0; i <= RTA_MAX; i++) {
6009 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
6017 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6018 struct netlink_ext_ack *extack)
6020 struct net *net = sock_net(in_skb->sk);
6021 struct nlattr *tb[RTA_MAX+1];
6022 int err, iif = 0, oif = 0;
6023 struct fib6_info *from;
6024 struct dst_entry *dst;
6025 struct rt6_info *rt;
6026 struct sk_buff *skb;
6028 struct flowi6 fl6 = {};
6031 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
6036 rtm = nlmsg_data(nlh);
6037 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
6038 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
6041 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
6044 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
6048 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
6051 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
6055 iif = nla_get_u32(tb[RTA_IIF]);
6058 oif = nla_get_u32(tb[RTA_OIF]);
6061 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6064 fl6.flowi6_uid = make_kuid(current_user_ns(),
6065 nla_get_u32(tb[RTA_UID]));
6067 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6070 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6073 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6075 if (tb[RTA_IP_PROTO]) {
6076 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
6077 &fl6.flowi6_proto, AF_INET6,
6084 struct net_device *dev;
6089 dev = dev_get_by_index_rcu(net, iif);
6096 fl6.flowi6_iif = iif;
6098 if (!ipv6_addr_any(&fl6.saddr))
6099 flags |= RT6_LOOKUP_F_HAS_SADDR;
6101 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
6105 fl6.flowi6_oif = oif;
6107 dst = ip6_route_output(net, NULL, &fl6);
6111 rt = container_of(dst, struct rt6_info, dst);
6112 if (rt->dst.error) {
6113 err = rt->dst.error;
6118 if (rt == net->ipv6.ip6_null_entry) {
6119 err = rt->dst.error;
6124 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
6131 skb_dst_set(skb, &rt->dst);
6134 from = rcu_dereference(rt->from);
6137 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6139 NETLINK_CB(in_skb).portid,
6142 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6143 &fl6.saddr, iif, RTM_NEWROUTE,
6144 NETLINK_CB(in_skb).portid,
6156 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6161 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6162 unsigned int nlm_flags)
6164 struct sk_buff *skb;
6165 struct net *net = info->nl_net;
6170 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6172 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6176 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6177 event, info->portid, seq, nlm_flags);
6179 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6180 WARN_ON(err == -EMSGSIZE);
6184 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6185 info->nlh, gfp_any());
6189 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6192 void fib6_rt_update(struct net *net, struct fib6_info *rt,
6193 struct nl_info *info)
6195 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6196 struct sk_buff *skb;
6199 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6203 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6204 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6206 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6207 WARN_ON(err == -EMSGSIZE);
6211 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6212 info->nlh, gfp_any());
6216 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6219 void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6220 bool offload, bool trap, bool offload_failed)
6222 struct sk_buff *skb;
6225 if (READ_ONCE(f6i->offload) == offload &&
6226 READ_ONCE(f6i->trap) == trap &&
6227 READ_ONCE(f6i->offload_failed) == offload_failed)
6230 WRITE_ONCE(f6i->offload, offload);
6231 WRITE_ONCE(f6i->trap, trap);
6233 /* 2 means send notifications only if offload_failed was changed. */
6234 if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6235 READ_ONCE(f6i->offload_failed) == offload_failed)
6238 WRITE_ONCE(f6i->offload_failed, offload_failed);
6240 if (!rcu_access_pointer(f6i->fib6_node))
6241 /* The route was removed from the tree, do not send
6246 if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6249 skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6255 err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6258 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6259 WARN_ON(err == -EMSGSIZE);
6264 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6268 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6270 EXPORT_SYMBOL(fib6_info_hw_flags_set);
6272 static int ip6_route_dev_notify(struct notifier_block *this,
6273 unsigned long event, void *ptr)
6275 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6276 struct net *net = dev_net(dev);
6278 if (!(dev->flags & IFF_LOOPBACK))
6281 if (event == NETDEV_REGISTER) {
6282 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6283 net->ipv6.ip6_null_entry->dst.dev = dev;
6284 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6285 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6286 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6287 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6288 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6289 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6291 } else if (event == NETDEV_UNREGISTER &&
6292 dev->reg_state != NETREG_UNREGISTERED) {
6293 /* NETDEV_UNREGISTER could be fired for multiple times by
6294 * netdev_wait_allrefs(). Make sure we only call this once.
6296 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6297 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6298 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6299 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6310 #ifdef CONFIG_PROC_FS
6311 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6313 struct net *net = (struct net *)seq->private;
6314 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6315 net->ipv6.rt6_stats->fib_nodes,
6316 net->ipv6.rt6_stats->fib_route_nodes,
6317 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6318 net->ipv6.rt6_stats->fib_rt_entries,
6319 net->ipv6.rt6_stats->fib_rt_cache,
6320 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6321 net->ipv6.rt6_stats->fib_discarded_routes);
6325 #endif /* CONFIG_PROC_FS */
6327 #ifdef CONFIG_SYSCTL
6329 static int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6330 void *buffer, size_t *lenp, loff_t *ppos)
6338 net = (struct net *)ctl->extra1;
6339 delay = net->ipv6.sysctl.flush_delay;
6340 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6344 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6348 static struct ctl_table ipv6_route_table_template[] = {
6350 .procname = "max_size",
6351 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
6352 .maxlen = sizeof(int),
6354 .proc_handler = proc_dointvec,
6357 .procname = "gc_thresh",
6358 .data = &ip6_dst_ops_template.gc_thresh,
6359 .maxlen = sizeof(int),
6361 .proc_handler = proc_dointvec,
6364 .procname = "flush",
6365 .data = &init_net.ipv6.sysctl.flush_delay,
6366 .maxlen = sizeof(int),
6368 .proc_handler = ipv6_sysctl_rtcache_flush
6371 .procname = "gc_min_interval",
6372 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6373 .maxlen = sizeof(int),
6375 .proc_handler = proc_dointvec_jiffies,
6378 .procname = "gc_timeout",
6379 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6380 .maxlen = sizeof(int),
6382 .proc_handler = proc_dointvec_jiffies,
6385 .procname = "gc_interval",
6386 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
6387 .maxlen = sizeof(int),
6389 .proc_handler = proc_dointvec_jiffies,
6392 .procname = "gc_elasticity",
6393 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6394 .maxlen = sizeof(int),
6396 .proc_handler = proc_dointvec,
6399 .procname = "mtu_expires",
6400 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6401 .maxlen = sizeof(int),
6403 .proc_handler = proc_dointvec_jiffies,
6406 .procname = "min_adv_mss",
6407 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
6408 .maxlen = sizeof(int),
6410 .proc_handler = proc_dointvec,
6413 .procname = "gc_min_interval_ms",
6414 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6415 .maxlen = sizeof(int),
6417 .proc_handler = proc_dointvec_ms_jiffies,
6420 .procname = "skip_notify_on_dev_down",
6421 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6422 .maxlen = sizeof(int),
6424 .proc_handler = proc_dointvec_minmax,
6425 .extra1 = SYSCTL_ZERO,
6426 .extra2 = SYSCTL_ONE,
6431 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6433 struct ctl_table *table;
6435 table = kmemdup(ipv6_route_table_template,
6436 sizeof(ipv6_route_table_template),
6440 table[0].data = &net->ipv6.sysctl.ip6_rt_max_size;
6441 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6442 table[2].data = &net->ipv6.sysctl.flush_delay;
6443 table[2].extra1 = net;
6444 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6445 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6446 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6447 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6448 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6449 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6450 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6451 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6453 /* Don't export sysctls to unprivileged users */
6454 if (net->user_ns != &init_user_ns)
6455 table[1].procname = NULL;
6462 static int __net_init ip6_route_net_init(struct net *net)
6466 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6467 sizeof(net->ipv6.ip6_dst_ops));
6469 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6470 goto out_ip6_dst_ops;
6472 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6473 if (!net->ipv6.fib6_null_entry)
6474 goto out_ip6_dst_entries;
6475 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6476 sizeof(*net->ipv6.fib6_null_entry));
6478 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6479 sizeof(*net->ipv6.ip6_null_entry),
6481 if (!net->ipv6.ip6_null_entry)
6482 goto out_fib6_null_entry;
6483 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6484 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6485 ip6_template_metrics, true);
6486 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
6488 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6489 net->ipv6.fib6_has_custom_rules = false;
6490 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6491 sizeof(*net->ipv6.ip6_prohibit_entry),
6493 if (!net->ipv6.ip6_prohibit_entry)
6494 goto out_ip6_null_entry;
6495 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6496 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6497 ip6_template_metrics, true);
6498 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
6500 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6501 sizeof(*net->ipv6.ip6_blk_hole_entry),
6503 if (!net->ipv6.ip6_blk_hole_entry)
6504 goto out_ip6_prohibit_entry;
6505 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6506 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6507 ip6_template_metrics, true);
6508 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
6509 #ifdef CONFIG_IPV6_SUBTREES
6510 net->ipv6.fib6_routes_require_src = 0;
6514 net->ipv6.sysctl.flush_delay = 0;
6515 net->ipv6.sysctl.ip6_rt_max_size = 4096;
6516 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6517 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6518 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6519 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6520 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6521 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6522 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6524 atomic_set(&net->ipv6.ip6_rt_gc_expire, 30*HZ);
6530 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6531 out_ip6_prohibit_entry:
6532 kfree(net->ipv6.ip6_prohibit_entry);
6534 kfree(net->ipv6.ip6_null_entry);
6536 out_fib6_null_entry:
6537 kfree(net->ipv6.fib6_null_entry);
6538 out_ip6_dst_entries:
6539 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6544 static void __net_exit ip6_route_net_exit(struct net *net)
6546 kfree(net->ipv6.fib6_null_entry);
6547 kfree(net->ipv6.ip6_null_entry);
6548 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6549 kfree(net->ipv6.ip6_prohibit_entry);
6550 kfree(net->ipv6.ip6_blk_hole_entry);
6552 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6555 static int __net_init ip6_route_net_init_late(struct net *net)
6557 #ifdef CONFIG_PROC_FS
6558 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6559 sizeof(struct ipv6_route_iter));
6560 proc_create_net_single("rt6_stats", 0444, net->proc_net,
6561 rt6_stats_seq_show, NULL);
6566 static void __net_exit ip6_route_net_exit_late(struct net *net)
6568 #ifdef CONFIG_PROC_FS
6569 remove_proc_entry("ipv6_route", net->proc_net);
6570 remove_proc_entry("rt6_stats", net->proc_net);
6574 static struct pernet_operations ip6_route_net_ops = {
6575 .init = ip6_route_net_init,
6576 .exit = ip6_route_net_exit,
6579 static int __net_init ipv6_inetpeer_init(struct net *net)
6581 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6585 inet_peer_base_init(bp);
6586 net->ipv6.peers = bp;
6590 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6592 struct inet_peer_base *bp = net->ipv6.peers;
6594 net->ipv6.peers = NULL;
6595 inetpeer_invalidate_tree(bp);
6599 static struct pernet_operations ipv6_inetpeer_ops = {
6600 .init = ipv6_inetpeer_init,
6601 .exit = ipv6_inetpeer_exit,
6604 static struct pernet_operations ip6_route_net_late_ops = {
6605 .init = ip6_route_net_init_late,
6606 .exit = ip6_route_net_exit_late,
6609 static struct notifier_block ip6_route_dev_notifier = {
6610 .notifier_call = ip6_route_dev_notify,
6611 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6614 void __init ip6_route_init_special_entries(void)
6616 /* Registering of the loopback is done before this portion of code,
6617 * the loopback reference in rt6_info will not be taken, do it
6618 * manually for init_net */
6619 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6620 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6621 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6622 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6623 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6624 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6625 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6626 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6630 #if IS_BUILTIN(CONFIG_IPV6)
6631 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6632 DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6634 BTF_ID_LIST(btf_fib6_info_id)
6635 BTF_ID(struct, fib6_info)
6637 static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6638 .seq_ops = &ipv6_route_seq_ops,
6639 .init_seq_private = bpf_iter_init_seq_net,
6640 .fini_seq_private = bpf_iter_fini_seq_net,
6641 .seq_priv_size = sizeof(struct ipv6_route_iter),
6644 static struct bpf_iter_reg ipv6_route_reg_info = {
6645 .target = "ipv6_route",
6646 .ctx_arg_info_size = 1,
6648 { offsetof(struct bpf_iter__ipv6_route, rt),
6649 PTR_TO_BTF_ID_OR_NULL },
6651 .seq_info = &ipv6_route_seq_info,
6654 static int __init bpf_iter_register(void)
6656 ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6657 return bpf_iter_reg_target(&ipv6_route_reg_info);
6660 static void bpf_iter_unregister(void)
6662 bpf_iter_unreg_target(&ipv6_route_reg_info);
6667 int __init ip6_route_init(void)
6673 ip6_dst_ops_template.kmem_cachep =
6674 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6675 SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL);
6676 if (!ip6_dst_ops_template.kmem_cachep)
6679 ret = dst_entries_init(&ip6_dst_blackhole_ops);
6681 goto out_kmem_cache;
6683 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6685 goto out_dst_entries;
6687 ret = register_pernet_subsys(&ip6_route_net_ops);
6689 goto out_register_inetpeer;
6691 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6695 goto out_register_subsys;
6701 ret = fib6_rules_init();
6705 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6707 goto fib6_rules_init;
6709 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6710 inet6_rtm_newroute, NULL, 0);
6712 goto out_register_late_subsys;
6714 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6715 inet6_rtm_delroute, NULL, 0);
6717 goto out_register_late_subsys;
6719 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6720 inet6_rtm_getroute, NULL,
6721 RTNL_FLAG_DOIT_UNLOCKED);
6723 goto out_register_late_subsys;
6725 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6727 goto out_register_late_subsys;
6729 #if IS_BUILTIN(CONFIG_IPV6)
6730 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6731 ret = bpf_iter_register();
6733 goto out_register_late_subsys;
6737 for_each_possible_cpu(cpu) {
6738 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6740 INIT_LIST_HEAD(&ul->head);
6741 INIT_LIST_HEAD(&ul->quarantine);
6742 spin_lock_init(&ul->lock);
6748 out_register_late_subsys:
6749 rtnl_unregister_all(PF_INET6);
6750 unregister_pernet_subsys(&ip6_route_net_late_ops);
6752 fib6_rules_cleanup();
6757 out_register_subsys:
6758 unregister_pernet_subsys(&ip6_route_net_ops);
6759 out_register_inetpeer:
6760 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6762 dst_entries_destroy(&ip6_dst_blackhole_ops);
6764 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6768 void ip6_route_cleanup(void)
6770 #if IS_BUILTIN(CONFIG_IPV6)
6771 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6772 bpf_iter_unregister();
6775 unregister_netdevice_notifier(&ip6_route_dev_notifier);
6776 unregister_pernet_subsys(&ip6_route_net_late_ops);
6777 fib6_rules_cleanup();
6780 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6781 unregister_pernet_subsys(&ip6_route_net_ops);
6782 dst_entries_destroy(&ip6_dst_blackhole_ops);
6783 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);