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 void ip6_dst_gc(struct dst_ops *ops);
96 static int ip6_pkt_discard(struct sk_buff *skb);
97 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
98 static int ip6_pkt_prohibit(struct sk_buff *skb);
99 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
100 static void ip6_link_failure(struct sk_buff *skb);
101 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
102 struct sk_buff *skb, u32 mtu,
104 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
105 struct sk_buff *skb);
106 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
108 static size_t rt6_nlmsg_size(struct fib6_info *f6i);
109 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
110 struct fib6_info *rt, struct dst_entry *dst,
111 struct in6_addr *dest, struct in6_addr *src,
112 int iif, int type, u32 portid, u32 seq,
114 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
115 const struct in6_addr *daddr,
116 const struct in6_addr *saddr);
118 #ifdef CONFIG_IPV6_ROUTE_INFO
119 static struct fib6_info *rt6_add_route_info(struct net *net,
120 const struct in6_addr *prefix, int prefixlen,
121 const struct in6_addr *gwaddr,
122 struct net_device *dev,
124 static struct fib6_info *rt6_get_route_info(struct net *net,
125 const struct in6_addr *prefix, int prefixlen,
126 const struct in6_addr *gwaddr,
127 struct net_device *dev);
130 struct uncached_list {
132 struct list_head head;
133 struct list_head quarantine;
136 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
138 void rt6_uncached_list_add(struct rt6_info *rt)
140 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
142 rt->dst.rt_uncached_list = ul;
144 spin_lock_bh(&ul->lock);
145 list_add_tail(&rt->dst.rt_uncached, &ul->head);
146 spin_unlock_bh(&ul->lock);
149 void rt6_uncached_list_del(struct rt6_info *rt)
151 if (!list_empty(&rt->dst.rt_uncached)) {
152 struct uncached_list *ul = rt->dst.rt_uncached_list;
154 spin_lock_bh(&ul->lock);
155 list_del_init(&rt->dst.rt_uncached);
156 spin_unlock_bh(&ul->lock);
160 static void rt6_uncached_list_flush_dev(struct net_device *dev)
164 for_each_possible_cpu(cpu) {
165 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
166 struct rt6_info *rt, *safe;
168 if (list_empty(&ul->head))
171 spin_lock_bh(&ul->lock);
172 list_for_each_entry_safe(rt, safe, &ul->head, dst.rt_uncached) {
173 struct inet6_dev *rt_idev = rt->rt6i_idev;
174 struct net_device *rt_dev = rt->dst.dev;
175 bool handled = false;
177 if (rt_idev->dev == dev) {
178 rt->rt6i_idev = in6_dev_get(blackhole_netdev);
179 in6_dev_put(rt_idev);
184 rt->dst.dev = blackhole_netdev;
185 netdev_ref_replace(rt_dev, blackhole_netdev,
186 &rt->dst.dev_tracker,
191 list_move(&rt->dst.rt_uncached,
194 spin_unlock_bh(&ul->lock);
198 static inline const void *choose_neigh_daddr(const struct in6_addr *p,
202 if (!ipv6_addr_any(p))
203 return (const void *) p;
205 return &ipv6_hdr(skb)->daddr;
209 struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
210 struct net_device *dev,
216 daddr = choose_neigh_daddr(gw, skb, daddr);
217 n = __ipv6_neigh_lookup(dev, daddr);
221 n = neigh_create(&nd_tbl, daddr, dev);
222 return IS_ERR(n) ? NULL : n;
225 static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
229 const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
231 return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
232 dst->dev, skb, daddr);
235 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
237 struct net_device *dev = dst->dev;
238 struct rt6_info *rt = (struct rt6_info *)dst;
240 daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
243 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
245 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
247 __ipv6_confirm_neigh(dev, daddr);
250 static struct dst_ops ip6_dst_ops_template = {
254 .check = ip6_dst_check,
255 .default_advmss = ip6_default_advmss,
257 .cow_metrics = dst_cow_metrics_generic,
258 .destroy = ip6_dst_destroy,
259 .ifdown = ip6_dst_ifdown,
260 .negative_advice = ip6_negative_advice,
261 .link_failure = ip6_link_failure,
262 .update_pmtu = ip6_rt_update_pmtu,
263 .redirect = rt6_do_redirect,
264 .local_out = __ip6_local_out,
265 .neigh_lookup = ip6_dst_neigh_lookup,
266 .confirm_neigh = ip6_confirm_neigh,
269 static struct dst_ops ip6_dst_blackhole_ops = {
271 .default_advmss = ip6_default_advmss,
272 .neigh_lookup = ip6_dst_neigh_lookup,
273 .check = ip6_dst_check,
274 .destroy = ip6_dst_destroy,
275 .cow_metrics = dst_cow_metrics_generic,
276 .update_pmtu = dst_blackhole_update_pmtu,
277 .redirect = dst_blackhole_redirect,
278 .mtu = dst_blackhole_mtu,
281 static const u32 ip6_template_metrics[RTAX_MAX] = {
282 [RTAX_HOPLIMIT - 1] = 0,
285 static const struct fib6_info fib6_null_entry_template = {
286 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
287 .fib6_protocol = RTPROT_KERNEL,
288 .fib6_metric = ~(u32)0,
289 .fib6_ref = REFCOUNT_INIT(1),
290 .fib6_type = RTN_UNREACHABLE,
291 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
294 static const struct rt6_info ip6_null_entry_template = {
296 .__rcuref = RCUREF_INIT(1),
298 .obsolete = DST_OBSOLETE_FORCE_CHK,
299 .error = -ENETUNREACH,
300 .input = ip6_pkt_discard,
301 .output = ip6_pkt_discard_out,
303 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
306 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
308 static const struct rt6_info ip6_prohibit_entry_template = {
310 .__rcuref = RCUREF_INIT(1),
312 .obsolete = DST_OBSOLETE_FORCE_CHK,
314 .input = ip6_pkt_prohibit,
315 .output = ip6_pkt_prohibit_out,
317 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
320 static const struct rt6_info ip6_blk_hole_entry_template = {
322 .__rcuref = RCUREF_INIT(1),
324 .obsolete = DST_OBSOLETE_FORCE_CHK,
326 .input = dst_discard,
327 .output = dst_discard_out,
329 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
334 static void rt6_info_init(struct rt6_info *rt)
336 memset_after(rt, 0, dst);
339 /* allocate dst with ip6_dst_ops */
340 struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
343 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
344 1, DST_OBSOLETE_FORCE_CHK, flags);
348 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
353 EXPORT_SYMBOL(ip6_dst_alloc);
355 static void ip6_dst_destroy(struct dst_entry *dst)
357 struct rt6_info *rt = (struct rt6_info *)dst;
358 struct fib6_info *from;
359 struct inet6_dev *idev;
361 ip_dst_metrics_put(dst);
362 rt6_uncached_list_del(rt);
364 idev = rt->rt6i_idev;
366 rt->rt6i_idev = NULL;
370 from = xchg((__force struct fib6_info **)&rt->from, NULL);
371 fib6_info_release(from);
374 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
377 struct rt6_info *rt = (struct rt6_info *)dst;
378 struct inet6_dev *idev = rt->rt6i_idev;
380 if (idev && idev->dev != blackhole_netdev) {
381 struct inet6_dev *blackhole_idev = in6_dev_get(blackhole_netdev);
383 if (blackhole_idev) {
384 rt->rt6i_idev = blackhole_idev;
390 static bool __rt6_check_expired(const struct rt6_info *rt)
392 if (rt->rt6i_flags & RTF_EXPIRES)
393 return time_after(jiffies, rt->dst.expires);
398 static bool rt6_check_expired(const struct rt6_info *rt)
400 struct fib6_info *from;
402 from = rcu_dereference(rt->from);
404 if (rt->rt6i_flags & RTF_EXPIRES) {
405 if (time_after(jiffies, rt->dst.expires))
408 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
409 fib6_check_expired(from);
414 void fib6_select_path(const struct net *net, struct fib6_result *res,
415 struct flowi6 *fl6, int oif, bool have_oif_match,
416 const struct sk_buff *skb, int strict)
418 struct fib6_info *sibling, *next_sibling;
419 struct fib6_info *match = res->f6i;
421 if (!match->nh && (!match->fib6_nsiblings || have_oif_match))
424 if (match->nh && have_oif_match && res->nh)
427 /* We might have already computed the hash for ICMPv6 errors. In such
428 * case it will always be non-zero. Otherwise now is the time to do it.
431 (!match->nh || nexthop_is_multipath(match->nh)))
432 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
434 if (unlikely(match->nh)) {
435 nexthop_path_fib6_result(res, fl6->mp_hash);
439 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
442 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
444 const struct fib6_nh *nh = sibling->fib6_nh;
447 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
448 if (fl6->mp_hash > nh_upper_bound)
450 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
458 res->nh = match->fib6_nh;
462 * Route lookup. rcu_read_lock() should be held.
465 static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
466 const struct in6_addr *saddr, int oif, int flags)
468 const struct net_device *dev;
470 if (nh->fib_nh_flags & RTNH_F_DEAD)
473 dev = nh->fib_nh_dev;
475 if (dev->ifindex == oif)
478 if (ipv6_chk_addr(net, saddr, dev,
479 flags & RT6_LOOKUP_F_IFACE))
486 struct fib6_nh_dm_arg {
488 const struct in6_addr *saddr;
494 static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
496 struct fib6_nh_dm_arg *arg = _arg;
499 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
503 /* returns fib6_nh from nexthop or NULL */
504 static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
505 struct fib6_result *res,
506 const struct in6_addr *saddr,
509 struct fib6_nh_dm_arg arg = {
516 if (nexthop_is_blackhole(nh))
519 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
525 static void rt6_device_match(struct net *net, struct fib6_result *res,
526 const struct in6_addr *saddr, int oif, int flags)
528 struct fib6_info *f6i = res->f6i;
529 struct fib6_info *spf6i;
532 if (!oif && ipv6_addr_any(saddr)) {
533 if (unlikely(f6i->nh)) {
534 nh = nexthop_fib6_nh(f6i->nh);
535 if (nexthop_is_blackhole(f6i->nh))
540 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
544 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
545 bool matched = false;
547 if (unlikely(spf6i->nh)) {
548 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
554 if (__rt6_device_match(net, nh, saddr, oif, flags))
563 if (oif && flags & RT6_LOOKUP_F_IFACE) {
564 res->f6i = net->ipv6.fib6_null_entry;
565 nh = res->f6i->fib6_nh;
569 if (unlikely(f6i->nh)) {
570 nh = nexthop_fib6_nh(f6i->nh);
571 if (nexthop_is_blackhole(f6i->nh))
577 if (nh->fib_nh_flags & RTNH_F_DEAD) {
578 res->f6i = net->ipv6.fib6_null_entry;
579 nh = res->f6i->fib6_nh;
583 res->fib6_type = res->f6i->fib6_type;
584 res->fib6_flags = res->f6i->fib6_flags;
588 res->fib6_flags |= RTF_REJECT;
589 res->fib6_type = RTN_BLACKHOLE;
593 #ifdef CONFIG_IPV6_ROUTER_PREF
594 struct __rt6_probe_work {
595 struct work_struct work;
596 struct in6_addr target;
597 struct net_device *dev;
598 netdevice_tracker dev_tracker;
601 static void rt6_probe_deferred(struct work_struct *w)
603 struct in6_addr mcaddr;
604 struct __rt6_probe_work *work =
605 container_of(w, struct __rt6_probe_work, work);
607 addrconf_addr_solict_mult(&work->target, &mcaddr);
608 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
609 netdev_put(work->dev, &work->dev_tracker);
613 static void rt6_probe(struct fib6_nh *fib6_nh)
615 struct __rt6_probe_work *work = NULL;
616 const struct in6_addr *nh_gw;
617 unsigned long last_probe;
618 struct neighbour *neigh;
619 struct net_device *dev;
620 struct inet6_dev *idev;
623 * Okay, this does not seem to be appropriate
624 * for now, however, we need to check if it
625 * is really so; aka Router Reachability Probing.
627 * Router Reachability Probe MUST be rate-limited
628 * to no more than one per minute.
630 if (!fib6_nh->fib_nh_gw_family)
633 nh_gw = &fib6_nh->fib_nh_gw6;
634 dev = fib6_nh->fib_nh_dev;
636 last_probe = READ_ONCE(fib6_nh->last_probe);
637 idev = __in6_dev_get(dev);
638 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
640 if (READ_ONCE(neigh->nud_state) & NUD_VALID)
643 write_lock_bh(&neigh->lock);
644 if (!(neigh->nud_state & NUD_VALID) &&
646 neigh->updated + idev->cnf.rtr_probe_interval)) {
647 work = kmalloc(sizeof(*work), GFP_ATOMIC);
649 __neigh_set_probe_once(neigh);
651 write_unlock_bh(&neigh->lock);
652 } else if (time_after(jiffies, last_probe +
653 idev->cnf.rtr_probe_interval)) {
654 work = kmalloc(sizeof(*work), GFP_ATOMIC);
657 if (!work || cmpxchg(&fib6_nh->last_probe,
658 last_probe, jiffies) != last_probe) {
661 INIT_WORK(&work->work, rt6_probe_deferred);
662 work->target = *nh_gw;
663 netdev_hold(dev, &work->dev_tracker, GFP_ATOMIC);
665 schedule_work(&work->work);
672 static inline void rt6_probe(struct fib6_nh *fib6_nh)
678 * Default Router Selection (RFC 2461 6.3.6)
680 static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
682 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
683 struct neighbour *neigh;
686 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
687 &fib6_nh->fib_nh_gw6);
689 u8 nud_state = READ_ONCE(neigh->nud_state);
691 if (nud_state & NUD_VALID)
692 ret = RT6_NUD_SUCCEED;
693 #ifdef CONFIG_IPV6_ROUTER_PREF
694 else if (!(nud_state & NUD_FAILED))
695 ret = RT6_NUD_SUCCEED;
697 ret = RT6_NUD_FAIL_PROBE;
700 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
701 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
708 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
713 if (!oif || nh->fib_nh_dev->ifindex == oif)
716 if (!m && (strict & RT6_LOOKUP_F_IFACE))
717 return RT6_NUD_FAIL_HARD;
718 #ifdef CONFIG_IPV6_ROUTER_PREF
719 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
721 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
722 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
723 int n = rt6_check_neigh(nh);
730 static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
731 int oif, int strict, int *mpri, bool *do_rr)
733 bool match_do_rr = false;
737 if (nh->fib_nh_flags & RTNH_F_DEAD)
740 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
741 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
742 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
745 m = rt6_score_route(nh, fib6_flags, oif, strict);
746 if (m == RT6_NUD_FAIL_DO_RR) {
748 m = 0; /* lowest valid score */
749 } else if (m == RT6_NUD_FAIL_HARD) {
753 if (strict & RT6_LOOKUP_F_REACHABLE)
756 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
758 *do_rr = match_do_rr;
766 struct fib6_nh_frl_arg {
775 static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
777 struct fib6_nh_frl_arg *arg = _arg;
780 return find_match(nh, arg->flags, arg->oif, arg->strict,
781 arg->mpri, arg->do_rr);
784 static void __find_rr_leaf(struct fib6_info *f6i_start,
785 struct fib6_info *nomatch, u32 metric,
786 struct fib6_result *res, struct fib6_info **cont,
787 int oif, int strict, bool *do_rr, int *mpri)
789 struct fib6_info *f6i;
791 for (f6i = f6i_start;
792 f6i && f6i != nomatch;
793 f6i = rcu_dereference(f6i->fib6_next)) {
794 bool matched = false;
797 if (cont && f6i->fib6_metric != metric) {
802 if (fib6_check_expired(f6i))
805 if (unlikely(f6i->nh)) {
806 struct fib6_nh_frl_arg arg = {
807 .flags = f6i->fib6_flags,
814 if (nexthop_is_blackhole(f6i->nh)) {
815 res->fib6_flags = RTF_REJECT;
816 res->fib6_type = RTN_BLACKHOLE;
818 res->nh = nexthop_fib6_nh(f6i->nh);
821 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
828 if (find_match(nh, f6i->fib6_flags, oif, strict,
835 res->fib6_flags = f6i->fib6_flags;
836 res->fib6_type = f6i->fib6_type;
841 static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
842 struct fib6_info *rr_head, int oif, int strict,
843 bool *do_rr, struct fib6_result *res)
845 u32 metric = rr_head->fib6_metric;
846 struct fib6_info *cont = NULL;
849 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
850 oif, strict, do_rr, &mpri);
852 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
853 oif, strict, do_rr, &mpri);
855 if (res->f6i || !cont)
858 __find_rr_leaf(cont, NULL, metric, res, NULL,
859 oif, strict, do_rr, &mpri);
862 static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
863 struct fib6_result *res, int strict)
865 struct fib6_info *leaf = rcu_dereference(fn->leaf);
866 struct fib6_info *rt0;
870 /* make sure this function or its helpers sets f6i */
873 if (!leaf || leaf == net->ipv6.fib6_null_entry)
876 rt0 = rcu_dereference(fn->rr_ptr);
880 /* Double check to make sure fn is not an intermediate node
881 * and fn->leaf does not points to its child's leaf
882 * (This might happen if all routes under fn are deleted from
883 * the tree and fib6_repair_tree() is called on the node.)
885 key_plen = rt0->fib6_dst.plen;
886 #ifdef CONFIG_IPV6_SUBTREES
887 if (rt0->fib6_src.plen)
888 key_plen = rt0->fib6_src.plen;
890 if (fn->fn_bit != key_plen)
893 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
895 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
897 /* no entries matched; do round-robin */
898 if (!next || next->fib6_metric != rt0->fib6_metric)
902 spin_lock_bh(&leaf->fib6_table->tb6_lock);
903 /* make sure next is not being deleted from the tree */
905 rcu_assign_pointer(fn->rr_ptr, next);
906 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
912 res->f6i = net->ipv6.fib6_null_entry;
913 res->nh = res->f6i->fib6_nh;
914 res->fib6_flags = res->f6i->fib6_flags;
915 res->fib6_type = res->f6i->fib6_type;
919 static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
921 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
922 res->nh->fib_nh_gw_family;
925 #ifdef CONFIG_IPV6_ROUTE_INFO
926 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
927 const struct in6_addr *gwaddr)
929 struct net *net = dev_net(dev);
930 struct route_info *rinfo = (struct route_info *) opt;
931 struct in6_addr prefix_buf, *prefix;
933 unsigned long lifetime;
934 struct fib6_info *rt;
936 if (len < sizeof(struct route_info)) {
940 /* Sanity check for prefix_len and length */
941 if (rinfo->length > 3) {
943 } else if (rinfo->prefix_len > 128) {
945 } else if (rinfo->prefix_len > 64) {
946 if (rinfo->length < 2) {
949 } else if (rinfo->prefix_len > 0) {
950 if (rinfo->length < 1) {
955 pref = rinfo->route_pref;
956 if (pref == ICMPV6_ROUTER_PREF_INVALID)
959 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
961 if (rinfo->length == 3)
962 prefix = (struct in6_addr *)rinfo->prefix;
964 /* this function is safe */
965 ipv6_addr_prefix(&prefix_buf,
966 (struct in6_addr *)rinfo->prefix,
968 prefix = &prefix_buf;
971 if (rinfo->prefix_len == 0)
972 rt = rt6_get_dflt_router(net, gwaddr, dev);
974 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
977 if (rt && !lifetime) {
978 ip6_del_rt(net, rt, false);
983 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
986 rt->fib6_flags = RTF_ROUTEINFO |
987 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
990 if (!addrconf_finite_timeout(lifetime))
991 fib6_clean_expires(rt);
993 fib6_set_expires(rt, jiffies + HZ * lifetime);
995 fib6_info_release(rt);
1002 * Misc support functions
1005 /* called with rcu_lock held */
1006 static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1008 struct net_device *dev = res->nh->fib_nh_dev;
1010 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1011 /* for copies of local routes, dst->dev needs to be the
1012 * device if it is a master device, the master device if
1013 * device is enslaved, and the loopback as the default
1015 if (netif_is_l3_slave(dev) &&
1016 !rt6_need_strict(&res->f6i->fib6_dst.addr))
1017 dev = l3mdev_master_dev_rcu(dev);
1018 else if (!netif_is_l3_master(dev))
1019 dev = dev_net(dev)->loopback_dev;
1020 /* last case is netif_is_l3_master(dev) is true in which
1021 * case we want dev returned to be dev
1028 static const int fib6_prop[RTN_MAX + 1] = {
1032 [RTN_BROADCAST] = 0,
1034 [RTN_MULTICAST] = 0,
1035 [RTN_BLACKHOLE] = -EINVAL,
1036 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1037 [RTN_PROHIBIT] = -EACCES,
1038 [RTN_THROW] = -EAGAIN,
1039 [RTN_NAT] = -EINVAL,
1040 [RTN_XRESOLVE] = -EINVAL,
1043 static int ip6_rt_type_to_error(u8 fib6_type)
1045 return fib6_prop[fib6_type];
1048 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1050 unsigned short flags = 0;
1052 if (rt->dst_nocount)
1053 flags |= DST_NOCOUNT;
1054 if (rt->dst_nopolicy)
1055 flags |= DST_NOPOLICY;
1060 static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1062 rt->dst.error = ip6_rt_type_to_error(fib6_type);
1064 switch (fib6_type) {
1066 rt->dst.output = dst_discard_out;
1067 rt->dst.input = dst_discard;
1070 rt->dst.output = ip6_pkt_prohibit_out;
1071 rt->dst.input = ip6_pkt_prohibit;
1074 case RTN_UNREACHABLE:
1076 rt->dst.output = ip6_pkt_discard_out;
1077 rt->dst.input = ip6_pkt_discard;
1082 static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1084 struct fib6_info *f6i = res->f6i;
1086 if (res->fib6_flags & RTF_REJECT) {
1087 ip6_rt_init_dst_reject(rt, res->fib6_type);
1092 rt->dst.output = ip6_output;
1094 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1095 rt->dst.input = ip6_input;
1096 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1097 rt->dst.input = ip6_mc_input;
1099 rt->dst.input = ip6_forward;
1102 if (res->nh->fib_nh_lws) {
1103 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1104 lwtunnel_set_redirect(&rt->dst);
1107 rt->dst.lastuse = jiffies;
1110 /* Caller must already hold reference to @from */
1111 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1113 rt->rt6i_flags &= ~RTF_EXPIRES;
1114 rcu_assign_pointer(rt->from, from);
1115 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1118 /* Caller must already hold reference to f6i in result */
1119 static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1121 const struct fib6_nh *nh = res->nh;
1122 const struct net_device *dev = nh->fib_nh_dev;
1123 struct fib6_info *f6i = res->f6i;
1125 ip6_rt_init_dst(rt, res);
1127 rt->rt6i_dst = f6i->fib6_dst;
1128 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1129 rt->rt6i_flags = res->fib6_flags;
1130 if (nh->fib_nh_gw_family) {
1131 rt->rt6i_gateway = nh->fib_nh_gw6;
1132 rt->rt6i_flags |= RTF_GATEWAY;
1134 rt6_set_from(rt, f6i);
1135 #ifdef CONFIG_IPV6_SUBTREES
1136 rt->rt6i_src = f6i->fib6_src;
1140 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1141 struct in6_addr *saddr)
1143 struct fib6_node *pn, *sn;
1145 if (fn->fn_flags & RTN_TL_ROOT)
1147 pn = rcu_dereference(fn->parent);
1148 sn = FIB6_SUBTREE(pn);
1150 fn = fib6_node_lookup(sn, NULL, saddr);
1153 if (fn->fn_flags & RTN_RTINFO)
1158 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1160 struct rt6_info *rt = *prt;
1162 if (dst_hold_safe(&rt->dst))
1165 rt = net->ipv6.ip6_null_entry;
1174 /* called with rcu_lock held */
1175 static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1177 struct net_device *dev = res->nh->fib_nh_dev;
1178 struct fib6_info *f6i = res->f6i;
1179 unsigned short flags;
1180 struct rt6_info *nrt;
1182 if (!fib6_info_hold_safe(f6i))
1185 flags = fib6_info_dst_flags(f6i);
1186 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1188 fib6_info_release(f6i);
1192 ip6_rt_copy_init(nrt, res);
1196 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1197 dst_hold(&nrt->dst);
1201 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
1202 struct fib6_table *table,
1204 const struct sk_buff *skb,
1207 struct fib6_result res = {};
1208 struct fib6_node *fn;
1209 struct rt6_info *rt;
1212 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1214 res.f6i = rcu_dereference(fn->leaf);
1216 res.f6i = net->ipv6.fib6_null_entry;
1218 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1221 if (res.f6i == net->ipv6.fib6_null_entry) {
1222 fn = fib6_backtrack(fn, &fl6->saddr);
1226 rt = net->ipv6.ip6_null_entry;
1229 } else if (res.fib6_flags & RTF_REJECT) {
1233 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1234 fl6->flowi6_oif != 0, skb, flags);
1236 /* Search through exception table */
1237 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1239 if (ip6_hold_safe(net, &rt))
1240 dst_use_noref(&rt->dst, jiffies);
1243 rt = ip6_create_rt_rcu(&res);
1247 trace_fib6_table_lookup(net, &res, table, fl6);
1254 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1255 const struct sk_buff *skb, int flags)
1257 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1259 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1261 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1262 const struct in6_addr *saddr, int oif,
1263 const struct sk_buff *skb, int strict)
1265 struct flowi6 fl6 = {
1269 struct dst_entry *dst;
1270 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1273 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1274 flags |= RT6_LOOKUP_F_HAS_SADDR;
1277 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1278 if (dst->error == 0)
1279 return (struct rt6_info *) dst;
1285 EXPORT_SYMBOL(rt6_lookup);
1287 /* ip6_ins_rt is called with FREE table->tb6_lock.
1288 * It takes new route entry, the addition fails by any reason the
1289 * route is released.
1290 * Caller must hold dst before calling it.
1293 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1294 struct netlink_ext_ack *extack)
1297 struct fib6_table *table;
1299 table = rt->fib6_table;
1300 spin_lock_bh(&table->tb6_lock);
1301 err = fib6_add(&table->tb6_root, rt, info, extack);
1302 spin_unlock_bh(&table->tb6_lock);
1307 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1309 struct nl_info info = { .nl_net = net, };
1311 return __ip6_ins_rt(rt, &info, NULL);
1314 static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1315 const struct in6_addr *daddr,
1316 const struct in6_addr *saddr)
1318 struct fib6_info *f6i = res->f6i;
1319 struct net_device *dev;
1320 struct rt6_info *rt;
1326 if (!fib6_info_hold_safe(f6i))
1329 dev = ip6_rt_get_dev_rcu(res);
1330 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1332 fib6_info_release(f6i);
1336 ip6_rt_copy_init(rt, res);
1337 rt->rt6i_flags |= RTF_CACHE;
1338 rt->rt6i_dst.addr = *daddr;
1339 rt->rt6i_dst.plen = 128;
1341 if (!rt6_is_gw_or_nonexthop(res)) {
1342 if (f6i->fib6_dst.plen != 128 &&
1343 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1344 rt->rt6i_flags |= RTF_ANYCAST;
1345 #ifdef CONFIG_IPV6_SUBTREES
1346 if (rt->rt6i_src.plen && saddr) {
1347 rt->rt6i_src.addr = *saddr;
1348 rt->rt6i_src.plen = 128;
1356 static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1358 struct fib6_info *f6i = res->f6i;
1359 unsigned short flags = fib6_info_dst_flags(f6i);
1360 struct net_device *dev;
1361 struct rt6_info *pcpu_rt;
1363 if (!fib6_info_hold_safe(f6i))
1367 dev = ip6_rt_get_dev_rcu(res);
1368 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
1371 fib6_info_release(f6i);
1374 ip6_rt_copy_init(pcpu_rt, res);
1375 pcpu_rt->rt6i_flags |= RTF_PCPU;
1378 pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1383 static bool rt6_is_valid(const struct rt6_info *rt6)
1385 return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1388 /* It should be called with rcu_read_lock() acquired */
1389 static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1391 struct rt6_info *pcpu_rt;
1393 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1395 if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1396 struct rt6_info *prev, **p;
1398 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1399 prev = xchg(p, NULL);
1401 dst_dev_put(&prev->dst);
1402 dst_release(&prev->dst);
1411 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1412 const struct fib6_result *res)
1414 struct rt6_info *pcpu_rt, *prev, **p;
1416 pcpu_rt = ip6_rt_pcpu_alloc(res);
1420 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1421 prev = cmpxchg(p, NULL, pcpu_rt);
1424 if (res->f6i->fib6_destroying) {
1425 struct fib6_info *from;
1427 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1428 fib6_info_release(from);
1434 /* exception hash table implementation
1436 static DEFINE_SPINLOCK(rt6_exception_lock);
1438 /* Remove rt6_ex from hash table and free the memory
1439 * Caller must hold rt6_exception_lock
1441 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1442 struct rt6_exception *rt6_ex)
1444 struct fib6_info *from;
1447 if (!bucket || !rt6_ex)
1450 net = dev_net(rt6_ex->rt6i->dst.dev);
1451 net->ipv6.rt6_stats->fib_rt_cache--;
1453 /* purge completely the exception to allow releasing the held resources:
1454 * some [sk] cache may keep the dst around for unlimited time
1456 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
1457 fib6_info_release(from);
1458 dst_dev_put(&rt6_ex->rt6i->dst);
1460 hlist_del_rcu(&rt6_ex->hlist);
1461 dst_release(&rt6_ex->rt6i->dst);
1462 kfree_rcu(rt6_ex, rcu);
1463 WARN_ON_ONCE(!bucket->depth);
1467 /* Remove oldest rt6_ex in bucket and free the memory
1468 * Caller must hold rt6_exception_lock
1470 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1472 struct rt6_exception *rt6_ex, *oldest = NULL;
1477 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1478 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1481 rt6_remove_exception(bucket, oldest);
1484 static u32 rt6_exception_hash(const struct in6_addr *dst,
1485 const struct in6_addr *src)
1487 static siphash_aligned_key_t rt6_exception_key;
1489 struct in6_addr dst;
1490 struct in6_addr src;
1491 } __aligned(SIPHASH_ALIGNMENT) combined = {
1496 net_get_random_once(&rt6_exception_key, sizeof(rt6_exception_key));
1498 #ifdef CONFIG_IPV6_SUBTREES
1500 combined.src = *src;
1502 val = siphash(&combined, sizeof(combined), &rt6_exception_key);
1504 return hash_64(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1507 /* Helper function to find the cached rt in the hash table
1508 * and update bucket pointer to point to the bucket for this
1509 * (daddr, saddr) pair
1510 * Caller must hold rt6_exception_lock
1512 static struct rt6_exception *
1513 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1514 const struct in6_addr *daddr,
1515 const struct in6_addr *saddr)
1517 struct rt6_exception *rt6_ex;
1520 if (!(*bucket) || !daddr)
1523 hval = rt6_exception_hash(daddr, saddr);
1526 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1527 struct rt6_info *rt6 = rt6_ex->rt6i;
1528 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1530 #ifdef CONFIG_IPV6_SUBTREES
1531 if (matched && saddr)
1532 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1540 /* Helper function to find the cached rt in the hash table
1541 * and update bucket pointer to point to the bucket for this
1542 * (daddr, saddr) pair
1543 * Caller must hold rcu_read_lock()
1545 static struct rt6_exception *
1546 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1547 const struct in6_addr *daddr,
1548 const struct in6_addr *saddr)
1550 struct rt6_exception *rt6_ex;
1553 WARN_ON_ONCE(!rcu_read_lock_held());
1555 if (!(*bucket) || !daddr)
1558 hval = rt6_exception_hash(daddr, saddr);
1561 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1562 struct rt6_info *rt6 = rt6_ex->rt6i;
1563 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1565 #ifdef CONFIG_IPV6_SUBTREES
1566 if (matched && saddr)
1567 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1575 static unsigned int fib6_mtu(const struct fib6_result *res)
1577 const struct fib6_nh *nh = res->nh;
1580 if (res->f6i->fib6_pmtu) {
1581 mtu = res->f6i->fib6_pmtu;
1583 struct net_device *dev = nh->fib_nh_dev;
1584 struct inet6_dev *idev;
1587 idev = __in6_dev_get(dev);
1588 mtu = idev->cnf.mtu6;
1592 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1594 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1597 #define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1599 /* used when the flushed bit is not relevant, only access to the bucket
1600 * (ie., all bucket users except rt6_insert_exception);
1602 * called under rcu lock; sometimes called with rt6_exception_lock held
1605 struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1608 struct rt6_exception_bucket *bucket;
1611 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1612 lockdep_is_held(lock));
1614 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1616 /* remove bucket flushed bit if set */
1618 unsigned long p = (unsigned long)bucket;
1620 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1621 bucket = (struct rt6_exception_bucket *)p;
1627 static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1629 unsigned long p = (unsigned long)bucket;
1631 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1634 /* called with rt6_exception_lock held */
1635 static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1638 struct rt6_exception_bucket *bucket;
1641 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1642 lockdep_is_held(lock));
1644 p = (unsigned long)bucket;
1645 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1646 bucket = (struct rt6_exception_bucket *)p;
1647 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1650 static int rt6_insert_exception(struct rt6_info *nrt,
1651 const struct fib6_result *res)
1653 struct net *net = dev_net(nrt->dst.dev);
1654 struct rt6_exception_bucket *bucket;
1655 struct fib6_info *f6i = res->f6i;
1656 struct in6_addr *src_key = NULL;
1657 struct rt6_exception *rt6_ex;
1658 struct fib6_nh *nh = res->nh;
1662 spin_lock_bh(&rt6_exception_lock);
1664 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1665 lockdep_is_held(&rt6_exception_lock));
1667 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1673 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1674 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1679 #ifdef CONFIG_IPV6_SUBTREES
1680 /* fib6_src.plen != 0 indicates f6i is in subtree
1681 * and exception table is indexed by a hash of
1682 * both fib6_dst and fib6_src.
1683 * Otherwise, the exception table is indexed by
1684 * a hash of only fib6_dst.
1686 if (f6i->fib6_src.plen)
1687 src_key = &nrt->rt6i_src.addr;
1689 /* rt6_mtu_change() might lower mtu on f6i.
1690 * Only insert this exception route if its mtu
1691 * is less than f6i's mtu value.
1693 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1698 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1701 rt6_remove_exception(bucket, rt6_ex);
1703 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1709 rt6_ex->stamp = jiffies;
1710 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1712 net->ipv6.rt6_stats->fib_rt_cache++;
1714 /* Randomize max depth to avoid some side channels attacks. */
1715 max_depth = FIB6_MAX_DEPTH + get_random_u32_below(FIB6_MAX_DEPTH);
1716 while (bucket->depth > max_depth)
1717 rt6_exception_remove_oldest(bucket);
1720 spin_unlock_bh(&rt6_exception_lock);
1722 /* Update fn->fn_sernum to invalidate all cached dst */
1724 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1725 fib6_update_sernum(net, f6i);
1726 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1727 fib6_force_start_gc(net);
1733 static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1735 struct rt6_exception_bucket *bucket;
1736 struct rt6_exception *rt6_ex;
1737 struct hlist_node *tmp;
1740 spin_lock_bh(&rt6_exception_lock);
1742 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1746 /* Prevent rt6_insert_exception() to recreate the bucket list */
1748 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1750 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1751 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1753 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1754 rt6_remove_exception(bucket, rt6_ex);
1756 WARN_ON_ONCE(!from && bucket->depth);
1760 spin_unlock_bh(&rt6_exception_lock);
1763 static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1765 struct fib6_info *f6i = arg;
1767 fib6_nh_flush_exceptions(nh, f6i);
1772 void rt6_flush_exceptions(struct fib6_info *f6i)
1775 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1778 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1781 /* Find cached rt in the hash table inside passed in rt
1782 * Caller has to hold rcu_read_lock()
1784 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1785 const struct in6_addr *daddr,
1786 const struct in6_addr *saddr)
1788 const struct in6_addr *src_key = NULL;
1789 struct rt6_exception_bucket *bucket;
1790 struct rt6_exception *rt6_ex;
1791 struct rt6_info *ret = NULL;
1793 #ifdef CONFIG_IPV6_SUBTREES
1794 /* fib6i_src.plen != 0 indicates f6i is in subtree
1795 * and exception table is indexed by a hash of
1796 * both fib6_dst and fib6_src.
1797 * However, the src addr used to create the hash
1798 * might not be exactly the passed in saddr which
1799 * is a /128 addr from the flow.
1800 * So we need to use f6i->fib6_src to redo lookup
1801 * if the passed in saddr does not find anything.
1802 * (See the logic in ip6_rt_cache_alloc() on how
1803 * rt->rt6i_src is updated.)
1805 if (res->f6i->fib6_src.plen)
1809 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1810 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1812 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1815 #ifdef CONFIG_IPV6_SUBTREES
1816 /* Use fib6_src as src_key and redo lookup */
1817 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1818 src_key = &res->f6i->fib6_src.addr;
1826 /* Remove the passed in cached rt from the hash table that contains it */
1827 static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1828 const struct rt6_info *rt)
1830 const struct in6_addr *src_key = NULL;
1831 struct rt6_exception_bucket *bucket;
1832 struct rt6_exception *rt6_ex;
1835 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1838 spin_lock_bh(&rt6_exception_lock);
1839 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1841 #ifdef CONFIG_IPV6_SUBTREES
1842 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1843 * and exception table is indexed by a hash of
1844 * both rt6i_dst and rt6i_src.
1845 * Otherwise, the exception table is indexed by
1846 * a hash of only rt6i_dst.
1849 src_key = &rt->rt6i_src.addr;
1851 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1855 rt6_remove_exception(bucket, rt6_ex);
1861 spin_unlock_bh(&rt6_exception_lock);
1865 struct fib6_nh_excptn_arg {
1866 struct rt6_info *rt;
1870 static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1872 struct fib6_nh_excptn_arg *arg = _arg;
1875 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1882 static int rt6_remove_exception_rt(struct rt6_info *rt)
1884 struct fib6_info *from;
1886 from = rcu_dereference(rt->from);
1887 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1891 struct fib6_nh_excptn_arg arg = {
1893 .plen = from->fib6_src.plen
1897 /* rc = 1 means an entry was found */
1898 rc = nexthop_for_each_fib6_nh(from->nh,
1899 rt6_nh_remove_exception_rt,
1901 return rc ? 0 : -ENOENT;
1904 return fib6_nh_remove_exception(from->fib6_nh,
1905 from->fib6_src.plen, rt);
1908 /* Find rt6_ex which contains the passed in rt cache and
1911 static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1912 const struct rt6_info *rt)
1914 const struct in6_addr *src_key = NULL;
1915 struct rt6_exception_bucket *bucket;
1916 struct rt6_exception *rt6_ex;
1918 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1919 #ifdef CONFIG_IPV6_SUBTREES
1920 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1921 * and exception table is indexed by a hash of
1922 * both rt6i_dst and rt6i_src.
1923 * Otherwise, the exception table is indexed by
1924 * a hash of only rt6i_dst.
1927 src_key = &rt->rt6i_src.addr;
1929 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1931 rt6_ex->stamp = jiffies;
1934 struct fib6_nh_match_arg {
1935 const struct net_device *dev;
1936 const struct in6_addr *gw;
1937 struct fib6_nh *match;
1940 /* determine if fib6_nh has given device and gateway */
1941 static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1943 struct fib6_nh_match_arg *arg = _arg;
1945 if (arg->dev != nh->fib_nh_dev ||
1946 (arg->gw && !nh->fib_nh_gw_family) ||
1947 (!arg->gw && nh->fib_nh_gw_family) ||
1948 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1953 /* found a match, break the loop */
1957 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1959 struct fib6_info *from;
1960 struct fib6_nh *fib6_nh;
1964 from = rcu_dereference(rt->from);
1965 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1969 struct fib6_nh_match_arg arg = {
1971 .gw = &rt->rt6i_gateway,
1974 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1978 fib6_nh = arg.match;
1980 fib6_nh = from->fib6_nh;
1982 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1987 static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1988 struct rt6_info *rt, int mtu)
1990 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1991 * lowest MTU in the path: always allow updating the route PMTU to
1992 * reflect PMTU decreases.
1994 * If the new MTU is higher, and the route PMTU is equal to the local
1995 * MTU, this means the old MTU is the lowest in the path, so allow
1996 * updating it: if other nodes now have lower MTUs, PMTU discovery will
2000 if (dst_mtu(&rt->dst) >= mtu)
2003 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
2009 static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
2010 const struct fib6_nh *nh, int mtu)
2012 struct rt6_exception_bucket *bucket;
2013 struct rt6_exception *rt6_ex;
2016 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2020 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2021 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2022 struct rt6_info *entry = rt6_ex->rt6i;
2024 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2025 * route), the metrics of its rt->from have already
2028 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2029 rt6_mtu_change_route_allowed(idev, entry, mtu))
2030 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2036 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2038 static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2039 const struct in6_addr *gateway)
2041 struct rt6_exception_bucket *bucket;
2042 struct rt6_exception *rt6_ex;
2043 struct hlist_node *tmp;
2046 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2049 spin_lock_bh(&rt6_exception_lock);
2050 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2052 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2053 hlist_for_each_entry_safe(rt6_ex, tmp,
2054 &bucket->chain, hlist) {
2055 struct rt6_info *entry = rt6_ex->rt6i;
2057 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2058 RTF_CACHE_GATEWAY &&
2059 ipv6_addr_equal(gateway,
2060 &entry->rt6i_gateway)) {
2061 rt6_remove_exception(bucket, rt6_ex);
2068 spin_unlock_bh(&rt6_exception_lock);
2071 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2072 struct rt6_exception *rt6_ex,
2073 struct fib6_gc_args *gc_args,
2076 struct rt6_info *rt = rt6_ex->rt6i;
2078 /* we are pruning and obsoleting aged-out and non gateway exceptions
2079 * even if others have still references to them, so that on next
2080 * dst_check() such references can be dropped.
2081 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2082 * expired, independently from their aging, as per RFC 8201 section 4
2084 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2085 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2086 RT6_TRACE("aging clone %p\n", rt);
2087 rt6_remove_exception(bucket, rt6_ex);
2090 } else if (time_after(jiffies, rt->dst.expires)) {
2091 RT6_TRACE("purging expired route %p\n", rt);
2092 rt6_remove_exception(bucket, rt6_ex);
2096 if (rt->rt6i_flags & RTF_GATEWAY) {
2097 struct neighbour *neigh;
2099 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2101 if (!(neigh && (neigh->flags & NTF_ROUTER))) {
2102 RT6_TRACE("purging route %p via non-router but gateway\n",
2104 rt6_remove_exception(bucket, rt6_ex);
2112 static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2113 struct fib6_gc_args *gc_args,
2116 struct rt6_exception_bucket *bucket;
2117 struct rt6_exception *rt6_ex;
2118 struct hlist_node *tmp;
2121 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2125 spin_lock(&rt6_exception_lock);
2126 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2128 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2129 hlist_for_each_entry_safe(rt6_ex, tmp,
2130 &bucket->chain, hlist) {
2131 rt6_age_examine_exception(bucket, rt6_ex,
2137 spin_unlock(&rt6_exception_lock);
2138 rcu_read_unlock_bh();
2141 struct fib6_nh_age_excptn_arg {
2142 struct fib6_gc_args *gc_args;
2146 static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2148 struct fib6_nh_age_excptn_arg *arg = _arg;
2150 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2154 void rt6_age_exceptions(struct fib6_info *f6i,
2155 struct fib6_gc_args *gc_args,
2159 struct fib6_nh_age_excptn_arg arg = {
2164 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2167 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2171 /* must be called with rcu lock held */
2172 int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2173 struct flowi6 *fl6, struct fib6_result *res, int strict)
2175 struct fib6_node *fn, *saved_fn;
2177 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2181 rt6_select(net, fn, oif, res, strict);
2182 if (res->f6i == net->ipv6.fib6_null_entry) {
2183 fn = fib6_backtrack(fn, &fl6->saddr);
2185 goto redo_rt6_select;
2186 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2187 /* also consider unreachable route */
2188 strict &= ~RT6_LOOKUP_F_REACHABLE;
2190 goto redo_rt6_select;
2194 trace_fib6_table_lookup(net, res, table, fl6);
2199 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2200 int oif, struct flowi6 *fl6,
2201 const struct sk_buff *skb, int flags)
2203 struct fib6_result res = {};
2204 struct rt6_info *rt = NULL;
2207 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2208 !rcu_read_lock_held());
2210 strict |= flags & RT6_LOOKUP_F_IFACE;
2211 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2212 if (net->ipv6.devconf_all->forwarding == 0)
2213 strict |= RT6_LOOKUP_F_REACHABLE;
2217 fib6_table_lookup(net, table, oif, fl6, &res, strict);
2218 if (res.f6i == net->ipv6.fib6_null_entry)
2221 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2223 /*Search through exception table */
2224 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2227 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2228 !res.nh->fib_nh_gw_family)) {
2229 /* Create a RTF_CACHE clone which will not be
2230 * owned by the fib6 tree. It is for the special case where
2231 * the daddr in the skb during the neighbor look-up is different
2232 * from the fl6->daddr used to look-up route here.
2234 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2237 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2238 * As rt6_uncached_list_add() does not consume refcnt,
2239 * this refcnt is always returned to the caller even
2240 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2242 rt6_uncached_list_add(rt);
2248 /* Get a percpu copy */
2250 rt = rt6_get_pcpu_route(&res);
2253 rt = rt6_make_pcpu_route(net, &res);
2259 rt = net->ipv6.ip6_null_entry;
2260 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2261 ip6_hold_safe(net, &rt);
2266 EXPORT_SYMBOL_GPL(ip6_pol_route);
2268 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
2269 struct fib6_table *table,
2271 const struct sk_buff *skb,
2274 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2277 struct dst_entry *ip6_route_input_lookup(struct net *net,
2278 struct net_device *dev,
2280 const struct sk_buff *skb,
2283 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2284 flags |= RT6_LOOKUP_F_IFACE;
2286 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2288 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2290 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2291 struct flow_keys *keys,
2292 struct flow_keys *flkeys)
2294 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2295 const struct ipv6hdr *key_iph = outer_iph;
2296 struct flow_keys *_flkeys = flkeys;
2297 const struct ipv6hdr *inner_iph;
2298 const struct icmp6hdr *icmph;
2299 struct ipv6hdr _inner_iph;
2300 struct icmp6hdr _icmph;
2302 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2305 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2306 sizeof(_icmph), &_icmph);
2310 if (!icmpv6_is_err(icmph->icmp6_type))
2313 inner_iph = skb_header_pointer(skb,
2314 skb_transport_offset(skb) + sizeof(*icmph),
2315 sizeof(_inner_iph), &_inner_iph);
2319 key_iph = inner_iph;
2323 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2324 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2325 keys->tags.flow_label = _flkeys->tags.flow_label;
2326 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2328 keys->addrs.v6addrs.src = key_iph->saddr;
2329 keys->addrs.v6addrs.dst = key_iph->daddr;
2330 keys->tags.flow_label = ip6_flowlabel(key_iph);
2331 keys->basic.ip_proto = key_iph->nexthdr;
2335 static u32 rt6_multipath_custom_hash_outer(const struct net *net,
2336 const struct sk_buff *skb,
2339 u32 hash_fields = ip6_multipath_hash_fields(net);
2340 struct flow_keys keys, hash_keys;
2342 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2345 memset(&hash_keys, 0, sizeof(hash_keys));
2346 skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_ENCAP);
2348 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2349 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2350 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2351 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2352 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2353 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2354 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2355 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2356 hash_keys.tags.flow_label = keys.tags.flow_label;
2357 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2358 hash_keys.ports.src = keys.ports.src;
2359 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2360 hash_keys.ports.dst = keys.ports.dst;
2362 *p_has_inner = !!(keys.control.flags & FLOW_DIS_ENCAPSULATION);
2363 return flow_hash_from_keys(&hash_keys);
2366 static u32 rt6_multipath_custom_hash_inner(const struct net *net,
2367 const struct sk_buff *skb,
2370 u32 hash_fields = ip6_multipath_hash_fields(net);
2371 struct flow_keys keys, hash_keys;
2373 /* We assume the packet carries an encapsulation, but if none was
2374 * encountered during dissection of the outer flow, then there is no
2375 * point in calling the flow dissector again.
2380 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_MASK))
2383 memset(&hash_keys, 0, sizeof(hash_keys));
2384 skb_flow_dissect_flow_keys(skb, &keys, 0);
2386 if (!(keys.control.flags & FLOW_DIS_ENCAPSULATION))
2389 if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2390 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2391 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2392 hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
2393 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2394 hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst;
2395 } else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2396 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2397 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2398 hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2399 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2400 hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2401 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL)
2402 hash_keys.tags.flow_label = keys.tags.flow_label;
2405 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO)
2406 hash_keys.basic.ip_proto = keys.basic.ip_proto;
2407 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT)
2408 hash_keys.ports.src = keys.ports.src;
2409 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
2410 hash_keys.ports.dst = keys.ports.dst;
2412 return flow_hash_from_keys(&hash_keys);
2415 static u32 rt6_multipath_custom_hash_skb(const struct net *net,
2416 const struct sk_buff *skb)
2418 u32 mhash, mhash_inner;
2419 bool has_inner = true;
2421 mhash = rt6_multipath_custom_hash_outer(net, skb, &has_inner);
2422 mhash_inner = rt6_multipath_custom_hash_inner(net, skb, has_inner);
2424 return jhash_2words(mhash, mhash_inner, 0);
2427 static u32 rt6_multipath_custom_hash_fl6(const struct net *net,
2428 const struct flowi6 *fl6)
2430 u32 hash_fields = ip6_multipath_hash_fields(net);
2431 struct flow_keys hash_keys;
2433 if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2436 memset(&hash_keys, 0, sizeof(hash_keys));
2437 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2438 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2439 hash_keys.addrs.v6addrs.src = fl6->saddr;
2440 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2441 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2442 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2443 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2444 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2445 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2446 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2447 hash_keys.ports.src = fl6->fl6_sport;
2448 if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2449 hash_keys.ports.dst = fl6->fl6_dport;
2451 return flow_hash_from_keys(&hash_keys);
2454 /* if skb is set it will be used and fl6 can be NULL */
2455 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2456 const struct sk_buff *skb, struct flow_keys *flkeys)
2458 struct flow_keys hash_keys;
2461 switch (ip6_multipath_hash_policy(net)) {
2463 memset(&hash_keys, 0, sizeof(hash_keys));
2464 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2466 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2468 hash_keys.addrs.v6addrs.src = fl6->saddr;
2469 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2470 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2471 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2473 mhash = flow_hash_from_keys(&hash_keys);
2477 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2478 struct flow_keys keys;
2480 /* short-circuit if we already have L4 hash present */
2482 return skb_get_hash_raw(skb) >> 1;
2484 memset(&hash_keys, 0, sizeof(hash_keys));
2487 skb_flow_dissect_flow_keys(skb, &keys, flag);
2490 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2491 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2492 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2493 hash_keys.ports.src = flkeys->ports.src;
2494 hash_keys.ports.dst = flkeys->ports.dst;
2495 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2497 memset(&hash_keys, 0, sizeof(hash_keys));
2498 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2499 hash_keys.addrs.v6addrs.src = fl6->saddr;
2500 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2501 hash_keys.ports.src = fl6->fl6_sport;
2502 hash_keys.ports.dst = fl6->fl6_dport;
2503 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2505 mhash = flow_hash_from_keys(&hash_keys);
2508 memset(&hash_keys, 0, sizeof(hash_keys));
2509 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2511 struct flow_keys keys;
2514 skb_flow_dissect_flow_keys(skb, &keys, 0);
2518 /* Inner can be v4 or v6 */
2519 if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2520 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2521 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2522 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2523 } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2524 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2525 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2526 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2527 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2528 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2530 /* Same as case 0 */
2531 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2532 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2535 /* Same as case 0 */
2536 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2537 hash_keys.addrs.v6addrs.src = fl6->saddr;
2538 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2539 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2540 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2542 mhash = flow_hash_from_keys(&hash_keys);
2546 mhash = rt6_multipath_custom_hash_skb(net, skb);
2548 mhash = rt6_multipath_custom_hash_fl6(net, fl6);
2555 /* Called with rcu held */
2556 void ip6_route_input(struct sk_buff *skb)
2558 const struct ipv6hdr *iph = ipv6_hdr(skb);
2559 struct net *net = dev_net(skb->dev);
2560 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2561 struct ip_tunnel_info *tun_info;
2562 struct flowi6 fl6 = {
2563 .flowi6_iif = skb->dev->ifindex,
2564 .daddr = iph->daddr,
2565 .saddr = iph->saddr,
2566 .flowlabel = ip6_flowinfo(iph),
2567 .flowi6_mark = skb->mark,
2568 .flowi6_proto = iph->nexthdr,
2570 struct flow_keys *flkeys = NULL, _flkeys;
2572 tun_info = skb_tunnel_info(skb);
2573 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2574 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2576 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2579 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2580 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2582 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2586 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
2587 struct fib6_table *table,
2589 const struct sk_buff *skb,
2592 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2595 static struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2596 const struct sock *sk,
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);
2628 struct dst_entry *ip6_route_output_flags(struct net *net,
2629 const struct sock *sk,
2633 struct dst_entry *dst;
2634 struct rt6_info *rt6;
2637 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2638 rt6 = (struct rt6_info *)dst;
2639 /* For dst cached in uncached_list, refcnt is already taken. */
2640 if (list_empty(&rt6->dst.rt_uncached) && !dst_hold_safe(dst)) {
2641 dst = &net->ipv6.ip6_null_entry->dst;
2648 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2650 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2652 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2653 struct net_device *loopback_dev = net->loopback_dev;
2654 struct dst_entry *new = NULL;
2656 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
2657 DST_OBSOLETE_DEAD, 0);
2660 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2664 new->input = dst_discard;
2665 new->output = dst_discard_out;
2667 dst_copy_metrics(new, &ort->dst);
2669 rt->rt6i_idev = in6_dev_get(loopback_dev);
2670 rt->rt6i_gateway = ort->rt6i_gateway;
2671 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2673 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2674 #ifdef CONFIG_IPV6_SUBTREES
2675 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2679 dst_release(dst_orig);
2680 return new ? new : ERR_PTR(-ENOMEM);
2684 * Destination cache support functions
2687 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2691 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2694 if (fib6_check_expired(f6i))
2700 static struct dst_entry *rt6_check(struct rt6_info *rt,
2701 struct fib6_info *from,
2706 if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2707 rt_cookie != cookie)
2710 if (rt6_check_expired(rt))
2716 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2717 struct fib6_info *from,
2720 if (!__rt6_check_expired(rt) &&
2721 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2722 fib6_check(from, cookie))
2728 INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2731 struct dst_entry *dst_ret;
2732 struct fib6_info *from;
2733 struct rt6_info *rt;
2735 rt = container_of(dst, struct rt6_info, dst);
2738 return rt6_is_valid(rt) ? dst : NULL;
2742 /* All IPV6 dsts are created with ->obsolete set to the value
2743 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2744 * into this function always.
2747 from = rcu_dereference(rt->from);
2749 if (from && (rt->rt6i_flags & RTF_PCPU ||
2750 unlikely(!list_empty(&rt->dst.rt_uncached))))
2751 dst_ret = rt6_dst_from_check(rt, from, cookie);
2753 dst_ret = rt6_check(rt, from, cookie);
2759 EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
2761 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2763 struct rt6_info *rt = (struct rt6_info *) dst;
2766 if (rt->rt6i_flags & RTF_CACHE) {
2768 if (rt6_check_expired(rt)) {
2769 rt6_remove_exception_rt(rt);
2781 static void ip6_link_failure(struct sk_buff *skb)
2783 struct rt6_info *rt;
2785 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2787 rt = (struct rt6_info *) skb_dst(skb);
2790 if (rt->rt6i_flags & RTF_CACHE) {
2791 rt6_remove_exception_rt(rt);
2793 struct fib6_info *from;
2794 struct fib6_node *fn;
2796 from = rcu_dereference(rt->from);
2798 fn = rcu_dereference(from->fib6_node);
2799 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2800 WRITE_ONCE(fn->fn_sernum, -1);
2807 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2809 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2810 struct fib6_info *from;
2813 from = rcu_dereference(rt0->from);
2815 rt0->dst.expires = from->expires;
2819 dst_set_expires(&rt0->dst, timeout);
2820 rt0->rt6i_flags |= RTF_EXPIRES;
2823 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2825 struct net *net = dev_net(rt->dst.dev);
2827 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2828 rt->rt6i_flags |= RTF_MODIFIED;
2829 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2832 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2834 return !(rt->rt6i_flags & RTF_CACHE) &&
2835 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2838 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2839 const struct ipv6hdr *iph, u32 mtu,
2842 const struct in6_addr *daddr, *saddr;
2843 struct rt6_info *rt6 = (struct rt6_info *)dst;
2845 /* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2846 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2847 * [see also comment in rt6_mtu_change_route()]
2851 daddr = &iph->daddr;
2852 saddr = &iph->saddr;
2854 daddr = &sk->sk_v6_daddr;
2855 saddr = &inet6_sk(sk)->saddr;
2862 dst_confirm_neigh(dst, daddr);
2864 if (mtu < IPV6_MIN_MTU)
2866 if (mtu >= dst_mtu(dst))
2869 if (!rt6_cache_allowed_for_pmtu(rt6)) {
2870 rt6_do_update_pmtu(rt6, mtu);
2871 /* update rt6_ex->stamp for cache */
2872 if (rt6->rt6i_flags & RTF_CACHE)
2873 rt6_update_exception_stamp_rt(rt6);
2875 struct fib6_result res = {};
2876 struct rt6_info *nrt6;
2879 res.f6i = rcu_dereference(rt6->from);
2883 res.fib6_flags = res.f6i->fib6_flags;
2884 res.fib6_type = res.f6i->fib6_type;
2887 struct fib6_nh_match_arg arg = {
2889 .gw = &rt6->rt6i_gateway,
2892 nexthop_for_each_fib6_nh(res.f6i->nh,
2893 fib6_nh_find_match, &arg);
2895 /* fib6_info uses a nexthop that does not have fib6_nh
2896 * using the dst->dev + gw. Should be impossible.
2903 res.nh = res.f6i->fib6_nh;
2906 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2908 rt6_do_update_pmtu(nrt6, mtu);
2909 if (rt6_insert_exception(nrt6, &res))
2910 dst_release_immediate(&nrt6->dst);
2917 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2918 struct sk_buff *skb, u32 mtu,
2921 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2925 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2926 int oif, u32 mark, kuid_t uid)
2928 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2929 struct dst_entry *dst;
2930 struct flowi6 fl6 = {
2932 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2933 .daddr = iph->daddr,
2934 .saddr = iph->saddr,
2935 .flowlabel = ip6_flowinfo(iph),
2939 dst = ip6_route_output(net, NULL, &fl6);
2941 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
2944 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2946 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2948 int oif = sk->sk_bound_dev_if;
2949 struct dst_entry *dst;
2951 if (!oif && skb->dev)
2952 oif = l3mdev_master_ifindex(skb->dev);
2954 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, READ_ONCE(sk->sk_mark),
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,
3177 READ_ONCE(sk->sk_mark), sk->sk_uid);
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 void 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_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3292 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3293 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3297 if (time_after(rt_last_gc + rt_min_interval, jiffies))
3300 fib6_run_gc(atomic_inc_return(&net->ipv6.ip6_rt_gc_expire), net, true);
3301 entries = dst_entries_get_slow(ops);
3302 if (entries < ops->gc_thresh)
3303 atomic_set(&net->ipv6.ip6_rt_gc_expire, rt_gc_timeout >> 1);
3305 val = atomic_read(&net->ipv6.ip6_rt_gc_expire);
3306 atomic_set(&net->ipv6.ip6_rt_gc_expire, val - (val >> rt_elasticity));
3309 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3310 const struct in6_addr *gw_addr, u32 tbid,
3311 int flags, struct fib6_result *res)
3313 struct flowi6 fl6 = {
3314 .flowi6_oif = cfg->fc_ifindex,
3316 .saddr = cfg->fc_prefsrc,
3318 struct fib6_table *table;
3321 table = fib6_get_table(net, tbid);
3325 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3326 flags |= RT6_LOOKUP_F_HAS_SADDR;
3328 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3330 err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3331 if (!err && res->f6i != net->ipv6.fib6_null_entry)
3332 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3333 cfg->fc_ifindex != 0, NULL, flags);
3338 static int ip6_route_check_nh_onlink(struct net *net,
3339 struct fib6_config *cfg,
3340 const struct net_device *dev,
3341 struct netlink_ext_ack *extack)
3343 u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3344 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3345 struct fib6_result res = {};
3348 err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3349 if (!err && !(res.fib6_flags & RTF_REJECT) &&
3350 /* ignore match if it is the default route */
3351 !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3352 (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3353 NL_SET_ERR_MSG(extack,
3354 "Nexthop has invalid gateway or device mismatch");
3361 static int ip6_route_check_nh(struct net *net,
3362 struct fib6_config *cfg,
3363 struct net_device **_dev,
3364 netdevice_tracker *dev_tracker,
3365 struct inet6_dev **idev)
3367 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3368 struct net_device *dev = _dev ? *_dev : NULL;
3369 int flags = RT6_LOOKUP_F_IFACE;
3370 struct fib6_result res = {};
3371 int err = -EHOSTUNREACH;
3373 if (cfg->fc_table) {
3374 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3375 cfg->fc_table, flags, &res);
3376 /* gw_addr can not require a gateway or resolve to a reject
3377 * route. If a device is given, it must match the result.
3379 if (err || res.fib6_flags & RTF_REJECT ||
3380 res.nh->fib_nh_gw_family ||
3381 (dev && dev != res.nh->fib_nh_dev))
3382 err = -EHOSTUNREACH;
3386 struct flowi6 fl6 = {
3387 .flowi6_oif = cfg->fc_ifindex,
3391 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3392 if (err || res.fib6_flags & RTF_REJECT ||
3393 res.nh->fib_nh_gw_family)
3394 err = -EHOSTUNREACH;
3399 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3400 cfg->fc_ifindex != 0, NULL, flags);
3405 if (dev != res.nh->fib_nh_dev)
3406 err = -EHOSTUNREACH;
3408 *_dev = dev = res.nh->fib_nh_dev;
3409 netdev_hold(dev, dev_tracker, GFP_ATOMIC);
3410 *idev = in6_dev_get(dev);
3416 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3417 struct net_device **_dev,
3418 netdevice_tracker *dev_tracker,
3419 struct inet6_dev **idev,
3420 struct netlink_ext_ack *extack)
3422 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3423 int gwa_type = ipv6_addr_type(gw_addr);
3424 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3425 const struct net_device *dev = *_dev;
3426 bool need_addr_check = !dev;
3429 /* if gw_addr is local we will fail to detect this in case
3430 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3431 * will return already-added prefix route via interface that
3432 * prefix route was assigned to, which might be non-loopback.
3435 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3436 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3440 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3441 /* IPv6 strictly inhibits using not link-local
3442 * addresses as nexthop address.
3443 * Otherwise, router will not able to send redirects.
3444 * It is very good, but in some (rare!) circumstances
3445 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3446 * some exceptions. --ANK
3447 * We allow IPv4-mapped nexthops to support RFC4798-type
3450 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3451 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3457 if (cfg->fc_flags & RTNH_F_ONLINK)
3458 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3460 err = ip6_route_check_nh(net, cfg, _dev, dev_tracker,
3469 /* reload in case device was changed */
3474 NL_SET_ERR_MSG(extack, "Egress device not specified");
3476 } else if (dev->flags & IFF_LOOPBACK) {
3477 NL_SET_ERR_MSG(extack,
3478 "Egress device can not be loopback device for this route");
3482 /* if we did not check gw_addr above, do so now that the
3483 * egress device has been resolved.
3485 if (need_addr_check &&
3486 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3487 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3496 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3498 if ((flags & RTF_REJECT) ||
3499 (dev && (dev->flags & IFF_LOOPBACK) &&
3500 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3501 !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3507 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3508 struct fib6_config *cfg, gfp_t gfp_flags,
3509 struct netlink_ext_ack *extack)
3511 netdevice_tracker *dev_tracker = &fib6_nh->fib_nh_dev_tracker;
3512 struct net_device *dev = NULL;
3513 struct inet6_dev *idev = NULL;
3517 fib6_nh->fib_nh_family = AF_INET6;
3518 #ifdef CONFIG_IPV6_ROUTER_PREF
3519 fib6_nh->last_probe = jiffies;
3521 if (cfg->fc_is_fdb) {
3522 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3523 fib6_nh->fib_nh_gw_family = AF_INET6;
3528 if (cfg->fc_ifindex) {
3529 dev = netdev_get_by_index(net, cfg->fc_ifindex,
3530 dev_tracker, gfp_flags);
3533 idev = in6_dev_get(dev);
3538 if (cfg->fc_flags & RTNH_F_ONLINK) {
3540 NL_SET_ERR_MSG(extack,
3541 "Nexthop device required for onlink");
3545 if (!(dev->flags & IFF_UP)) {
3546 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3551 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3554 fib6_nh->fib_nh_weight = 1;
3556 /* We cannot add true routes via loopback here,
3557 * they would result in kernel looping; promote them to reject routes
3559 addr_type = ipv6_addr_type(&cfg->fc_dst);
3560 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3561 /* hold loopback dev/idev if we haven't done so. */
3562 if (dev != net->loopback_dev) {
3564 netdev_put(dev, dev_tracker);
3567 dev = net->loopback_dev;
3568 netdev_hold(dev, dev_tracker, gfp_flags);
3569 idev = in6_dev_get(dev);
3578 if (cfg->fc_flags & RTF_GATEWAY) {
3579 err = ip6_validate_gw(net, cfg, &dev, dev_tracker,
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 fib6_nh->fib_nh_oif = dev->ifindex;
3628 lwtstate_put(fib6_nh->fib_nh_lws);
3629 fib6_nh->fib_nh_lws = NULL;
3630 netdev_put(dev, dev_tracker);
3636 void fib6_nh_release(struct fib6_nh *fib6_nh)
3638 struct rt6_exception_bucket *bucket;
3642 fib6_nh_flush_exceptions(fib6_nh, NULL);
3643 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3645 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3651 fib6_nh_release_dsts(fib6_nh);
3652 free_percpu(fib6_nh->rt6i_pcpu);
3654 fib_nh_common_release(&fib6_nh->nh_common);
3657 void fib6_nh_release_dsts(struct fib6_nh *fib6_nh)
3661 if (!fib6_nh->rt6i_pcpu)
3664 for_each_possible_cpu(cpu) {
3665 struct rt6_info *pcpu_rt, **ppcpu_rt;
3667 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3668 pcpu_rt = xchg(ppcpu_rt, NULL);
3670 dst_dev_put(&pcpu_rt->dst);
3671 dst_release(&pcpu_rt->dst);
3676 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3678 struct netlink_ext_ack *extack)
3680 struct net *net = cfg->fc_nlinfo.nl_net;
3681 struct fib6_info *rt = NULL;
3682 struct nexthop *nh = NULL;
3683 struct fib6_table *table;
3684 struct fib6_nh *fib6_nh;
3688 /* RTF_PCPU is an internal flag; can not be set by userspace */
3689 if (cfg->fc_flags & RTF_PCPU) {
3690 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3694 /* RTF_CACHE is an internal flag; can not be set by userspace */
3695 if (cfg->fc_flags & RTF_CACHE) {
3696 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3700 if (cfg->fc_type > RTN_MAX) {
3701 NL_SET_ERR_MSG(extack, "Invalid route type");
3705 if (cfg->fc_dst_len > 128) {
3706 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3709 if (cfg->fc_src_len > 128) {
3710 NL_SET_ERR_MSG(extack, "Invalid source address length");
3713 #ifndef CONFIG_IPV6_SUBTREES
3714 if (cfg->fc_src_len) {
3715 NL_SET_ERR_MSG(extack,
3716 "Specifying source address requires IPV6_SUBTREES to be enabled");
3720 if (cfg->fc_nh_id) {
3721 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3723 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3726 err = fib6_check_nexthop(nh, cfg, extack);
3732 if (cfg->fc_nlinfo.nlh &&
3733 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3734 table = fib6_get_table(net, cfg->fc_table);
3736 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3737 table = fib6_new_table(net, cfg->fc_table);
3740 table = fib6_new_table(net, cfg->fc_table);
3747 rt = fib6_info_alloc(gfp_flags, !nh);
3751 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3753 if (IS_ERR(rt->fib6_metrics)) {
3754 err = PTR_ERR(rt->fib6_metrics);
3755 /* Do not leave garbage there. */
3756 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3760 if (cfg->fc_flags & RTF_ADDRCONF)
3761 rt->dst_nocount = true;
3763 if (cfg->fc_flags & RTF_EXPIRES)
3764 fib6_set_expires(rt, jiffies +
3765 clock_t_to_jiffies(cfg->fc_expires));
3767 fib6_clean_expires(rt);
3769 if (cfg->fc_protocol == RTPROT_UNSPEC)
3770 cfg->fc_protocol = RTPROT_BOOT;
3771 rt->fib6_protocol = cfg->fc_protocol;
3773 rt->fib6_table = table;
3774 rt->fib6_metric = cfg->fc_metric;
3775 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3776 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3778 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3779 rt->fib6_dst.plen = cfg->fc_dst_len;
3781 #ifdef CONFIG_IPV6_SUBTREES
3782 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3783 rt->fib6_src.plen = cfg->fc_src_len;
3786 if (rt->fib6_src.plen) {
3787 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3790 if (!nexthop_get(nh)) {
3791 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3795 fib6_nh = nexthop_fib6_nh(rt->nh);
3797 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3801 fib6_nh = rt->fib6_nh;
3803 /* We cannot add true routes via loopback here, they would
3804 * result in kernel looping; promote them to reject routes
3806 addr_type = ipv6_addr_type(&cfg->fc_dst);
3807 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3809 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3812 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3813 struct net_device *dev = fib6_nh->fib_nh_dev;
3815 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3816 NL_SET_ERR_MSG(extack, "Invalid source address");
3820 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3821 rt->fib6_prefsrc.plen = 128;
3823 rt->fib6_prefsrc.plen = 0;
3827 fib6_info_release(rt);
3828 return ERR_PTR(err);
3830 ip_fib_metrics_put(rt->fib6_metrics);
3832 return ERR_PTR(err);
3835 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3836 struct netlink_ext_ack *extack)
3838 struct fib6_info *rt;
3841 rt = ip6_route_info_create(cfg, gfp_flags, extack);
3845 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3846 fib6_info_release(rt);
3851 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3853 struct net *net = info->nl_net;
3854 struct fib6_table *table;
3857 if (rt == net->ipv6.fib6_null_entry) {
3862 table = rt->fib6_table;
3863 spin_lock_bh(&table->tb6_lock);
3864 err = fib6_del(rt, info);
3865 spin_unlock_bh(&table->tb6_lock);
3868 fib6_info_release(rt);
3872 int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
3874 struct nl_info info = {
3876 .skip_notify = skip_notify
3879 return __ip6_del_rt(rt, &info);
3882 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3884 struct nl_info *info = &cfg->fc_nlinfo;
3885 struct net *net = info->nl_net;
3886 struct sk_buff *skb = NULL;
3887 struct fib6_table *table;
3890 if (rt == net->ipv6.fib6_null_entry)
3892 table = rt->fib6_table;
3893 spin_lock_bh(&table->tb6_lock);
3895 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3896 struct fib6_info *sibling, *next_sibling;
3897 struct fib6_node *fn;
3899 /* prefer to send a single notification with all hops */
3900 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3902 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3904 if (rt6_fill_node(net, skb, rt, NULL,
3905 NULL, NULL, 0, RTM_DELROUTE,
3906 info->portid, seq, 0) < 0) {
3910 info->skip_notify = 1;
3913 /* 'rt' points to the first sibling route. If it is not the
3914 * leaf, then we do not need to send a notification. Otherwise,
3915 * we need to check if the last sibling has a next route or not
3916 * and emit a replace or delete notification, respectively.
3918 info->skip_notify_kernel = 1;
3919 fn = rcu_dereference_protected(rt->fib6_node,
3920 lockdep_is_held(&table->tb6_lock));
3921 if (rcu_access_pointer(fn->leaf) == rt) {
3922 struct fib6_info *last_sibling, *replace_rt;
3924 last_sibling = list_last_entry(&rt->fib6_siblings,
3927 replace_rt = rcu_dereference_protected(
3928 last_sibling->fib6_next,
3929 lockdep_is_held(&table->tb6_lock));
3931 call_fib6_entry_notifiers_replace(net,
3934 call_fib6_multipath_entry_notifiers(net,
3935 FIB_EVENT_ENTRY_DEL,
3936 rt, rt->fib6_nsiblings,
3939 list_for_each_entry_safe(sibling, next_sibling,
3942 err = fib6_del(sibling, info);
3948 err = fib6_del(rt, info);
3950 spin_unlock_bh(&table->tb6_lock);
3952 fib6_info_release(rt);
3955 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3956 info->nlh, gfp_any());
3961 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3965 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3968 if (cfg->fc_flags & RTF_GATEWAY &&
3969 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3972 rc = rt6_remove_exception_rt(rt);
3977 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3980 struct fib6_result res = {
3984 struct rt6_info *rt_cache;
3986 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3988 return __ip6_del_cached_rt(rt_cache, cfg);
3993 struct fib6_nh_del_cached_rt_arg {
3994 struct fib6_config *cfg;
3995 struct fib6_info *f6i;
3998 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
4000 struct fib6_nh_del_cached_rt_arg *arg = _arg;
4003 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
4004 return rc != -ESRCH ? rc : 0;
4007 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
4009 struct fib6_nh_del_cached_rt_arg arg = {
4014 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
4017 static int ip6_route_del(struct fib6_config *cfg,
4018 struct netlink_ext_ack *extack)
4020 struct fib6_table *table;
4021 struct fib6_info *rt;
4022 struct fib6_node *fn;
4025 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
4027 NL_SET_ERR_MSG(extack, "FIB table does not exist");
4033 fn = fib6_locate(&table->tb6_root,
4034 &cfg->fc_dst, cfg->fc_dst_len,
4035 &cfg->fc_src, cfg->fc_src_len,
4036 !(cfg->fc_flags & RTF_CACHE));
4039 for_each_fib6_node_rt_rcu(fn) {
4042 if (rt->nh && cfg->fc_nh_id &&
4043 rt->nh->id != cfg->fc_nh_id)
4046 if (cfg->fc_flags & RTF_CACHE) {
4050 rc = ip6_del_cached_rt_nh(cfg, rt);
4051 } else if (cfg->fc_nh_id) {
4055 rc = ip6_del_cached_rt(cfg, rt, nh);
4064 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
4066 if (cfg->fc_protocol &&
4067 cfg->fc_protocol != rt->fib6_protocol)
4071 if (!fib6_info_hold_safe(rt))
4075 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4081 if (cfg->fc_ifindex &&
4083 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
4085 if (cfg->fc_flags & RTF_GATEWAY &&
4086 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
4088 if (!fib6_info_hold_safe(rt))
4092 /* if gateway was specified only delete the one hop */
4093 if (cfg->fc_flags & RTF_GATEWAY)
4094 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4096 return __ip6_del_rt_siblings(rt, cfg);
4104 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
4106 struct netevent_redirect netevent;
4107 struct rt6_info *rt, *nrt = NULL;
4108 struct fib6_result res = {};
4109 struct ndisc_options ndopts;
4110 struct inet6_dev *in6_dev;
4111 struct neighbour *neigh;
4113 int optlen, on_link;
4116 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
4117 optlen -= sizeof(*msg);
4120 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4124 msg = (struct rd_msg *)icmp6_hdr(skb);
4126 if (ipv6_addr_is_multicast(&msg->dest)) {
4127 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4132 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4134 } else if (ipv6_addr_type(&msg->target) !=
4135 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4136 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4140 in6_dev = __in6_dev_get(skb->dev);
4143 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
4147 * The IP source address of the Redirect MUST be the same as the current
4148 * first-hop router for the specified ICMP Destination Address.
4151 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4152 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4157 if (ndopts.nd_opts_tgt_lladdr) {
4158 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4161 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4166 rt = (struct rt6_info *) dst;
4167 if (rt->rt6i_flags & RTF_REJECT) {
4168 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4172 /* Redirect received -> path was valid.
4173 * Look, redirects are sent only in response to data packets,
4174 * so that this nexthop apparently is reachable. --ANK
4176 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4178 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4183 * We have finally decided to accept it.
4186 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4187 NEIGH_UPDATE_F_WEAK_OVERRIDE|
4188 NEIGH_UPDATE_F_OVERRIDE|
4189 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4190 NEIGH_UPDATE_F_ISROUTER)),
4191 NDISC_REDIRECT, &ndopts);
4194 res.f6i = rcu_dereference(rt->from);
4199 struct fib6_nh_match_arg arg = {
4201 .gw = &rt->rt6i_gateway,
4204 nexthop_for_each_fib6_nh(res.f6i->nh,
4205 fib6_nh_find_match, &arg);
4207 /* fib6_info uses a nexthop that does not have fib6_nh
4208 * using the dst->dev. Should be impossible
4214 res.nh = res.f6i->fib6_nh;
4217 res.fib6_flags = res.f6i->fib6_flags;
4218 res.fib6_type = res.f6i->fib6_type;
4219 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4223 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4225 nrt->rt6i_flags &= ~RTF_GATEWAY;
4227 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4229 /* rt6_insert_exception() will take care of duplicated exceptions */
4230 if (rt6_insert_exception(nrt, &res)) {
4231 dst_release_immediate(&nrt->dst);
4235 netevent.old = &rt->dst;
4236 netevent.new = &nrt->dst;
4237 netevent.daddr = &msg->dest;
4238 netevent.neigh = neigh;
4239 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4243 neigh_release(neigh);
4246 #ifdef CONFIG_IPV6_ROUTE_INFO
4247 static struct fib6_info *rt6_get_route_info(struct net *net,
4248 const struct in6_addr *prefix, int prefixlen,
4249 const struct in6_addr *gwaddr,
4250 struct net_device *dev)
4252 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4253 int ifindex = dev->ifindex;
4254 struct fib6_node *fn;
4255 struct fib6_info *rt = NULL;
4256 struct fib6_table *table;
4258 table = fib6_get_table(net, tb_id);
4263 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4267 for_each_fib6_node_rt_rcu(fn) {
4268 /* these routes do not use nexthops */
4271 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4273 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4274 !rt->fib6_nh->fib_nh_gw_family)
4276 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4278 if (!fib6_info_hold_safe(rt))
4287 static struct fib6_info *rt6_add_route_info(struct net *net,
4288 const struct in6_addr *prefix, int prefixlen,
4289 const struct in6_addr *gwaddr,
4290 struct net_device *dev,
4293 struct fib6_config cfg = {
4294 .fc_metric = IP6_RT_PRIO_USER,
4295 .fc_ifindex = dev->ifindex,
4296 .fc_dst_len = prefixlen,
4297 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4298 RTF_UP | RTF_PREF(pref),
4299 .fc_protocol = RTPROT_RA,
4300 .fc_type = RTN_UNICAST,
4301 .fc_nlinfo.portid = 0,
4302 .fc_nlinfo.nlh = NULL,
4303 .fc_nlinfo.nl_net = net,
4306 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4307 cfg.fc_dst = *prefix;
4308 cfg.fc_gateway = *gwaddr;
4310 /* We should treat it as a default route if prefix length is 0. */
4312 cfg.fc_flags |= RTF_DEFAULT;
4314 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4316 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4320 struct fib6_info *rt6_get_dflt_router(struct net *net,
4321 const struct in6_addr *addr,
4322 struct net_device *dev)
4324 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4325 struct fib6_info *rt;
4326 struct fib6_table *table;
4328 table = fib6_get_table(net, tb_id);
4333 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4336 /* RA routes do not use nexthops */
4341 if (dev == nh->fib_nh_dev &&
4342 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4343 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4346 if (rt && !fib6_info_hold_safe(rt))
4352 struct fib6_info *rt6_add_dflt_router(struct net *net,
4353 const struct in6_addr *gwaddr,
4354 struct net_device *dev,
4356 u32 defrtr_usr_metric)
4358 struct fib6_config cfg = {
4359 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4360 .fc_metric = defrtr_usr_metric,
4361 .fc_ifindex = dev->ifindex,
4362 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4363 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4364 .fc_protocol = RTPROT_RA,
4365 .fc_type = RTN_UNICAST,
4366 .fc_nlinfo.portid = 0,
4367 .fc_nlinfo.nlh = NULL,
4368 .fc_nlinfo.nl_net = net,
4371 cfg.fc_gateway = *gwaddr;
4373 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4374 struct fib6_table *table;
4376 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4378 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4381 return rt6_get_dflt_router(net, gwaddr, dev);
4384 static void __rt6_purge_dflt_routers(struct net *net,
4385 struct fib6_table *table)
4387 struct fib6_info *rt;
4391 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4392 struct net_device *dev = fib6_info_nh_dev(rt);
4393 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4395 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4396 (!idev || idev->cnf.accept_ra != 2) &&
4397 fib6_info_hold_safe(rt)) {
4399 ip6_del_rt(net, rt, false);
4405 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4408 void rt6_purge_dflt_routers(struct net *net)
4410 struct fib6_table *table;
4411 struct hlist_head *head;
4416 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4417 head = &net->ipv6.fib_table_hash[h];
4418 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4419 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4420 __rt6_purge_dflt_routers(net, table);
4427 static void rtmsg_to_fib6_config(struct net *net,
4428 struct in6_rtmsg *rtmsg,
4429 struct fib6_config *cfg)
4431 *cfg = (struct fib6_config){
4432 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4434 .fc_ifindex = rtmsg->rtmsg_ifindex,
4435 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
4436 .fc_expires = rtmsg->rtmsg_info,
4437 .fc_dst_len = rtmsg->rtmsg_dst_len,
4438 .fc_src_len = rtmsg->rtmsg_src_len,
4439 .fc_flags = rtmsg->rtmsg_flags,
4440 .fc_type = rtmsg->rtmsg_type,
4442 .fc_nlinfo.nl_net = net,
4444 .fc_dst = rtmsg->rtmsg_dst,
4445 .fc_src = rtmsg->rtmsg_src,
4446 .fc_gateway = rtmsg->rtmsg_gateway,
4450 int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
4452 struct fib6_config cfg;
4455 if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4457 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4460 rtmsg_to_fib6_config(net, rtmsg, &cfg);
4465 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4468 err = ip6_route_del(&cfg, NULL);
4476 * Drop the packet on the floor
4479 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4481 struct dst_entry *dst = skb_dst(skb);
4482 struct net *net = dev_net(dst->dev);
4483 struct inet6_dev *idev;
4487 if (netif_is_l3_master(skb->dev) ||
4488 dst->dev == net->loopback_dev)
4489 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4491 idev = ip6_dst_idev(dst);
4493 switch (ipstats_mib_noroutes) {
4494 case IPSTATS_MIB_INNOROUTES:
4495 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4496 if (type == IPV6_ADDR_ANY) {
4497 SKB_DR_SET(reason, IP_INADDRERRORS);
4498 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4501 SKB_DR_SET(reason, IP_INNOROUTES);
4503 case IPSTATS_MIB_OUTNOROUTES:
4504 SKB_DR_OR(reason, IP_OUTNOROUTES);
4505 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4509 /* Start over by dropping the dst for l3mdev case */
4510 if (netif_is_l3_master(skb->dev))
4513 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4514 kfree_skb_reason(skb, reason);
4518 static int ip6_pkt_discard(struct sk_buff *skb)
4520 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4523 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4525 skb->dev = skb_dst(skb)->dev;
4526 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4529 static int ip6_pkt_prohibit(struct sk_buff *skb)
4531 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4534 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4536 skb->dev = skb_dst(skb)->dev;
4537 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4541 * Allocate a dst for local (unicast / anycast) address.
4544 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4545 struct inet6_dev *idev,
4546 const struct in6_addr *addr,
4547 bool anycast, gfp_t gfp_flags)
4549 struct fib6_config cfg = {
4550 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4551 .fc_ifindex = idev->dev->ifindex,
4552 .fc_flags = RTF_UP | RTF_NONEXTHOP,
4555 .fc_protocol = RTPROT_KERNEL,
4556 .fc_nlinfo.nl_net = net,
4557 .fc_ignore_dev_down = true,
4559 struct fib6_info *f6i;
4562 cfg.fc_type = RTN_ANYCAST;
4563 cfg.fc_flags |= RTF_ANYCAST;
4565 cfg.fc_type = RTN_LOCAL;
4566 cfg.fc_flags |= RTF_LOCAL;
4569 f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
4571 f6i->dst_nocount = true;
4574 (net->ipv6.devconf_all->disable_policy ||
4575 idev->cnf.disable_policy))
4576 f6i->dst_nopolicy = true;
4582 /* remove deleted ip from prefsrc entries */
4583 struct arg_dev_net_ip {
4584 struct net_device *dev;
4586 struct in6_addr *addr;
4589 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4591 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4592 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4593 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4596 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
4597 rt != net->ipv6.fib6_null_entry &&
4598 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
4599 spin_lock_bh(&rt6_exception_lock);
4600 /* remove prefsrc entry */
4601 rt->fib6_prefsrc.plen = 0;
4602 spin_unlock_bh(&rt6_exception_lock);
4607 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4609 struct net *net = dev_net(ifp->idev->dev);
4610 struct arg_dev_net_ip adni = {
4611 .dev = ifp->idev->dev,
4615 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4618 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
4620 /* Remove routers and update dst entries when gateway turn into host. */
4621 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4623 struct in6_addr *gateway = (struct in6_addr *)arg;
4626 /* RA routes do not use nexthops */
4631 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4632 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4635 /* Further clean up cached routes in exception table.
4636 * This is needed because cached route may have a different
4637 * gateway than its 'parent' in the case of an ip redirect.
4639 fib6_nh_exceptions_clean_tohost(nh, gateway);
4644 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4646 fib6_clean_all(net, fib6_clean_tohost, gateway);
4649 struct arg_netdev_event {
4650 const struct net_device *dev;
4652 unsigned char nh_flags;
4653 unsigned long event;
4657 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4659 struct fib6_info *iter;
4660 struct fib6_node *fn;
4662 fn = rcu_dereference_protected(rt->fib6_node,
4663 lockdep_is_held(&rt->fib6_table->tb6_lock));
4664 iter = rcu_dereference_protected(fn->leaf,
4665 lockdep_is_held(&rt->fib6_table->tb6_lock));
4667 if (iter->fib6_metric == rt->fib6_metric &&
4668 rt6_qualify_for_ecmp(iter))
4670 iter = rcu_dereference_protected(iter->fib6_next,
4671 lockdep_is_held(&rt->fib6_table->tb6_lock));
4677 /* only called for fib entries with builtin fib6_nh */
4678 static bool rt6_is_dead(const struct fib6_info *rt)
4680 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4681 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4682 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4688 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4690 struct fib6_info *iter;
4693 if (!rt6_is_dead(rt))
4694 total += rt->fib6_nh->fib_nh_weight;
4696 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4697 if (!rt6_is_dead(iter))
4698 total += iter->fib6_nh->fib_nh_weight;
4704 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4706 int upper_bound = -1;
4708 if (!rt6_is_dead(rt)) {
4709 *weight += rt->fib6_nh->fib_nh_weight;
4710 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4713 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4716 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4718 struct fib6_info *iter;
4721 rt6_upper_bound_set(rt, &weight, total);
4723 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4724 rt6_upper_bound_set(iter, &weight, total);
4727 void rt6_multipath_rebalance(struct fib6_info *rt)
4729 struct fib6_info *first;
4732 /* In case the entire multipath route was marked for flushing,
4733 * then there is no need to rebalance upon the removal of every
4736 if (!rt->fib6_nsiblings || rt->should_flush)
4739 /* During lookup routes are evaluated in order, so we need to
4740 * make sure upper bounds are assigned from the first sibling
4743 first = rt6_multipath_first_sibling(rt);
4744 if (WARN_ON_ONCE(!first))
4747 total = rt6_multipath_total_weight(first);
4748 rt6_multipath_upper_bound_set(first, total);
4751 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4753 const struct arg_netdev_event *arg = p_arg;
4754 struct net *net = dev_net(arg->dev);
4756 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4757 rt->fib6_nh->fib_nh_dev == arg->dev) {
4758 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4759 fib6_update_sernum_upto_root(net, rt);
4760 rt6_multipath_rebalance(rt);
4766 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4768 struct arg_netdev_event arg = {
4771 .nh_flags = nh_flags,
4775 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4776 arg.nh_flags |= RTNH_F_LINKDOWN;
4778 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4781 /* only called for fib entries with inline fib6_nh */
4782 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4783 const struct net_device *dev)
4785 struct fib6_info *iter;
4787 if (rt->fib6_nh->fib_nh_dev == dev)
4789 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4790 if (iter->fib6_nh->fib_nh_dev == dev)
4796 static void rt6_multipath_flush(struct fib6_info *rt)
4798 struct fib6_info *iter;
4800 rt->should_flush = 1;
4801 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4802 iter->should_flush = 1;
4805 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4806 const struct net_device *down_dev)
4808 struct fib6_info *iter;
4809 unsigned int dead = 0;
4811 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4812 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4814 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4815 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4816 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4822 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4823 const struct net_device *dev,
4824 unsigned char nh_flags)
4826 struct fib6_info *iter;
4828 if (rt->fib6_nh->fib_nh_dev == dev)
4829 rt->fib6_nh->fib_nh_flags |= nh_flags;
4830 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4831 if (iter->fib6_nh->fib_nh_dev == dev)
4832 iter->fib6_nh->fib_nh_flags |= nh_flags;
4835 /* called with write lock held for table with rt */
4836 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4838 const struct arg_netdev_event *arg = p_arg;
4839 const struct net_device *dev = arg->dev;
4840 struct net *net = dev_net(dev);
4842 if (rt == net->ipv6.fib6_null_entry || rt->nh)
4845 switch (arg->event) {
4846 case NETDEV_UNREGISTER:
4847 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4849 if (rt->should_flush)
4851 if (!rt->fib6_nsiblings)
4852 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4853 if (rt6_multipath_uses_dev(rt, dev)) {
4856 count = rt6_multipath_dead_count(rt, dev);
4857 if (rt->fib6_nsiblings + 1 == count) {
4858 rt6_multipath_flush(rt);
4861 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4863 fib6_update_sernum(net, rt);
4864 rt6_multipath_rebalance(rt);
4868 if (rt->fib6_nh->fib_nh_dev != dev ||
4869 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4871 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4872 rt6_multipath_rebalance(rt);
4879 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4881 struct arg_netdev_event arg = {
4887 struct net *net = dev_net(dev);
4889 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4890 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4892 fib6_clean_all(net, fib6_ifdown, &arg);
4895 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4897 rt6_sync_down_dev(dev, event);
4898 rt6_uncached_list_flush_dev(dev);
4899 neigh_ifdown(&nd_tbl, dev);
4902 struct rt6_mtu_change_arg {
4903 struct net_device *dev;
4905 struct fib6_info *f6i;
4908 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4910 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4911 struct fib6_info *f6i = arg->f6i;
4913 /* For administrative MTU increase, there is no way to discover
4914 * IPv6 PMTU increase, so PMTU increase should be updated here.
4915 * Since RFC 1981 doesn't include administrative MTU increase
4916 * update PMTU increase is a MUST. (i.e. jumbo frame)
4918 if (nh->fib_nh_dev == arg->dev) {
4919 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4920 u32 mtu = f6i->fib6_pmtu;
4922 if (mtu >= arg->mtu ||
4923 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4924 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4926 spin_lock_bh(&rt6_exception_lock);
4927 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4928 spin_unlock_bh(&rt6_exception_lock);
4934 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4936 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4937 struct inet6_dev *idev;
4939 /* In IPv6 pmtu discovery is not optional,
4940 so that RTAX_MTU lock cannot disable it.
4941 We still use this lock to block changes
4942 caused by addrconf/ndisc.
4945 idev = __in6_dev_get(arg->dev);
4949 if (fib6_metric_locked(f6i, RTAX_MTU))
4954 /* fib6_nh_mtu_change only returns 0, so this is safe */
4955 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4959 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4962 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4964 struct rt6_mtu_change_arg arg = {
4969 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4972 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4973 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
4974 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
4975 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
4976 [RTA_OIF] = { .type = NLA_U32 },
4977 [RTA_IIF] = { .type = NLA_U32 },
4978 [RTA_PRIORITY] = { .type = NLA_U32 },
4979 [RTA_METRICS] = { .type = NLA_NESTED },
4980 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
4981 [RTA_PREF] = { .type = NLA_U8 },
4982 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4983 [RTA_ENCAP] = { .type = NLA_NESTED },
4984 [RTA_EXPIRES] = { .type = NLA_U32 },
4985 [RTA_UID] = { .type = NLA_U32 },
4986 [RTA_MARK] = { .type = NLA_U32 },
4987 [RTA_TABLE] = { .type = NLA_U32 },
4988 [RTA_IP_PROTO] = { .type = NLA_U8 },
4989 [RTA_SPORT] = { .type = NLA_U16 },
4990 [RTA_DPORT] = { .type = NLA_U16 },
4991 [RTA_NH_ID] = { .type = NLA_U32 },
4994 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
4995 struct fib6_config *cfg,
4996 struct netlink_ext_ack *extack)
4999 struct nlattr *tb[RTA_MAX+1];
5003 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5004 rtm_ipv6_policy, extack);
5009 rtm = nlmsg_data(nlh);
5012 NL_SET_ERR_MSG(extack,
5013 "Invalid dsfield (tos): option not available for IPv6");
5017 *cfg = (struct fib6_config){
5018 .fc_table = rtm->rtm_table,
5019 .fc_dst_len = rtm->rtm_dst_len,
5020 .fc_src_len = rtm->rtm_src_len,
5022 .fc_protocol = rtm->rtm_protocol,
5023 .fc_type = rtm->rtm_type,
5025 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
5026 .fc_nlinfo.nlh = nlh,
5027 .fc_nlinfo.nl_net = sock_net(skb->sk),
5030 if (rtm->rtm_type == RTN_UNREACHABLE ||
5031 rtm->rtm_type == RTN_BLACKHOLE ||
5032 rtm->rtm_type == RTN_PROHIBIT ||
5033 rtm->rtm_type == RTN_THROW)
5034 cfg->fc_flags |= RTF_REJECT;
5036 if (rtm->rtm_type == RTN_LOCAL)
5037 cfg->fc_flags |= RTF_LOCAL;
5039 if (rtm->rtm_flags & RTM_F_CLONED)
5040 cfg->fc_flags |= RTF_CACHE;
5042 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5044 if (tb[RTA_NH_ID]) {
5045 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
5046 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5047 NL_SET_ERR_MSG(extack,
5048 "Nexthop specification and nexthop id are mutually exclusive");
5051 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5054 if (tb[RTA_GATEWAY]) {
5055 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
5056 cfg->fc_flags |= RTF_GATEWAY;
5059 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5064 int plen = (rtm->rtm_dst_len + 7) >> 3;
5066 if (nla_len(tb[RTA_DST]) < plen)
5069 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
5073 int plen = (rtm->rtm_src_len + 7) >> 3;
5075 if (nla_len(tb[RTA_SRC]) < plen)
5078 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
5081 if (tb[RTA_PREFSRC])
5082 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
5085 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5087 if (tb[RTA_PRIORITY])
5088 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5090 if (tb[RTA_METRICS]) {
5091 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5092 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
5096 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5098 if (tb[RTA_MULTIPATH]) {
5099 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5100 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
5102 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
5103 cfg->fc_mp_len, extack);
5109 pref = nla_get_u8(tb[RTA_PREF]);
5110 if (pref != ICMPV6_ROUTER_PREF_LOW &&
5111 pref != ICMPV6_ROUTER_PREF_HIGH)
5112 pref = ICMPV6_ROUTER_PREF_MEDIUM;
5113 cfg->fc_flags |= RTF_PREF(pref);
5117 cfg->fc_encap = tb[RTA_ENCAP];
5119 if (tb[RTA_ENCAP_TYPE]) {
5120 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5122 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5127 if (tb[RTA_EXPIRES]) {
5128 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5130 if (addrconf_finite_timeout(timeout)) {
5131 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5132 cfg->fc_flags |= RTF_EXPIRES;
5142 struct fib6_info *fib6_info;
5143 struct fib6_config r_cfg;
5144 struct list_head next;
5147 static int ip6_route_info_append(struct net *net,
5148 struct list_head *rt6_nh_list,
5149 struct fib6_info *rt,
5150 struct fib6_config *r_cfg)
5155 list_for_each_entry(nh, rt6_nh_list, next) {
5156 /* check if fib6_info already exists */
5157 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5161 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5165 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5166 list_add_tail(&nh->next, rt6_nh_list);
5171 static void ip6_route_mpath_notify(struct fib6_info *rt,
5172 struct fib6_info *rt_last,
5173 struct nl_info *info,
5176 /* if this is an APPEND route, then rt points to the first route
5177 * inserted and rt_last points to last route inserted. Userspace
5178 * wants a consistent dump of the route which starts at the first
5179 * nexthop. Since sibling routes are always added at the end of
5180 * the list, find the first sibling of the last route appended
5182 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5183 rt = list_first_entry(&rt_last->fib6_siblings,
5189 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5192 static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5194 bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5195 bool should_notify = false;
5196 struct fib6_info *leaf;
5197 struct fib6_node *fn;
5200 fn = rcu_dereference(rt->fib6_node);
5204 leaf = rcu_dereference(fn->leaf);
5209 (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5210 rt6_qualify_for_ecmp(leaf)))
5211 should_notify = true;
5215 return should_notify;
5218 static int fib6_gw_from_attr(struct in6_addr *gw, struct nlattr *nla,
5219 struct netlink_ext_ack *extack)
5221 if (nla_len(nla) < sizeof(*gw)) {
5222 NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_GATEWAY");
5226 *gw = nla_get_in6_addr(nla);
5231 static int ip6_route_multipath_add(struct fib6_config *cfg,
5232 struct netlink_ext_ack *extack)
5234 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5235 struct nl_info *info = &cfg->fc_nlinfo;
5236 struct fib6_config r_cfg;
5237 struct rtnexthop *rtnh;
5238 struct fib6_info *rt;
5239 struct rt6_nh *err_nh;
5240 struct rt6_nh *nh, *nh_safe;
5246 int replace = (cfg->fc_nlinfo.nlh &&
5247 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5248 LIST_HEAD(rt6_nh_list);
5250 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5251 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5252 nlflags |= NLM_F_APPEND;
5254 remaining = cfg->fc_mp_len;
5255 rtnh = (struct rtnexthop *)cfg->fc_mp;
5257 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
5258 * fib6_info structs per nexthop
5260 while (rtnh_ok(rtnh, remaining)) {
5261 memcpy(&r_cfg, cfg, sizeof(*cfg));
5262 if (rtnh->rtnh_ifindex)
5263 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5265 attrlen = rtnh_attrlen(rtnh);
5267 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5269 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5271 err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5276 r_cfg.fc_flags |= RTF_GATEWAY;
5278 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5280 /* RTA_ENCAP_TYPE length checked in
5281 * lwtunnel_valid_encap_type_attr
5283 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5285 r_cfg.fc_encap_type = nla_get_u16(nla);
5288 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5289 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5295 if (!rt6_qualify_for_ecmp(rt)) {
5297 NL_SET_ERR_MSG(extack,
5298 "Device only routes can not be added for IPv6 using the multipath API.");
5299 fib6_info_release(rt);
5303 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5305 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5308 fib6_info_release(rt);
5312 rtnh = rtnh_next(rtnh, &remaining);
5315 if (list_empty(&rt6_nh_list)) {
5316 NL_SET_ERR_MSG(extack,
5317 "Invalid nexthop configuration - no valid nexthops");
5321 /* for add and replace send one notification with all nexthops.
5322 * Skip the notification in fib6_add_rt2node and send one with
5323 * the full route when done
5325 info->skip_notify = 1;
5327 /* For add and replace, send one notification with all nexthops. For
5328 * append, send one notification with all appended nexthops.
5330 info->skip_notify_kernel = 1;
5333 list_for_each_entry(nh, &rt6_nh_list, next) {
5334 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5335 fib6_info_release(nh->fib6_info);
5338 /* save reference to last route successfully inserted */
5339 rt_last = nh->fib6_info;
5341 /* save reference to first route for notification */
5343 rt_notif = nh->fib6_info;
5346 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5347 nh->fib6_info = NULL;
5350 NL_SET_ERR_MSG_MOD(extack,
5351 "multipath route replace failed (check consistency of installed routes)");
5356 /* Because each route is added like a single route we remove
5357 * these flags after the first nexthop: if there is a collision,
5358 * we have already failed to add the first nexthop:
5359 * fib6_add_rt2node() has rejected it; when replacing, old
5360 * nexthops have been replaced by first new, the rest should
5363 if (cfg->fc_nlinfo.nlh) {
5364 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5366 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5371 /* An in-kernel notification should only be sent in case the new
5372 * multipath route is added as the first route in the node, or if
5373 * it was appended to it. We pass 'rt_notif' since it is the first
5374 * sibling and might allow us to skip some checks in the replace case.
5376 if (ip6_route_mpath_should_notify(rt_notif)) {
5377 enum fib_event_type fib_event;
5379 if (rt_notif->fib6_nsiblings != nhn - 1)
5380 fib_event = FIB_EVENT_ENTRY_APPEND;
5382 fib_event = FIB_EVENT_ENTRY_REPLACE;
5384 err = call_fib6_multipath_entry_notifiers(info->nl_net,
5385 fib_event, rt_notif,
5388 /* Delete all the siblings that were just added */
5394 /* success ... tell user about new route */
5395 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5399 /* send notification for routes that were added so that
5400 * the delete notifications sent by ip6_route_del are
5404 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5406 /* Delete routes that were already added */
5407 list_for_each_entry(nh, &rt6_nh_list, next) {
5410 ip6_route_del(&nh->r_cfg, extack);
5414 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5416 fib6_info_release(nh->fib6_info);
5417 list_del(&nh->next);
5424 static int ip6_route_multipath_del(struct fib6_config *cfg,
5425 struct netlink_ext_ack *extack)
5427 struct fib6_config r_cfg;
5428 struct rtnexthop *rtnh;
5434 remaining = cfg->fc_mp_len;
5435 rtnh = (struct rtnexthop *)cfg->fc_mp;
5437 /* Parse a Multipath Entry */
5438 while (rtnh_ok(rtnh, remaining)) {
5439 memcpy(&r_cfg, cfg, sizeof(*cfg));
5440 if (rtnh->rtnh_ifindex)
5441 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5443 attrlen = rtnh_attrlen(rtnh);
5445 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5447 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5449 err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5456 r_cfg.fc_flags |= RTF_GATEWAY;
5459 err = ip6_route_del(&r_cfg, extack);
5464 rtnh = rtnh_next(rtnh, &remaining);
5470 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5471 struct netlink_ext_ack *extack)
5473 struct fib6_config cfg;
5476 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5481 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5482 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5487 return ip6_route_multipath_del(&cfg, extack);
5489 cfg.fc_delete_all_nh = 1;
5490 return ip6_route_del(&cfg, extack);
5494 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5495 struct netlink_ext_ack *extack)
5497 struct fib6_config cfg;
5500 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5504 if (cfg.fc_metric == 0)
5505 cfg.fc_metric = IP6_RT_PRIO_USER;
5508 return ip6_route_multipath_add(&cfg, extack);
5510 return ip6_route_add(&cfg, GFP_KERNEL, extack);
5513 /* add the overhead of this fib6_nh to nexthop_len */
5514 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5516 int *nexthop_len = arg;
5518 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5519 + NLA_ALIGN(sizeof(struct rtnexthop))
5520 + nla_total_size(16); /* RTA_GATEWAY */
5522 if (nh->fib_nh_lws) {
5523 /* RTA_ENCAP_TYPE */
5524 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5526 *nexthop_len += nla_total_size(2);
5532 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5537 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5538 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5541 struct fib6_info *sibling, *next_sibling;
5542 struct fib6_nh *nh = f6i->fib6_nh;
5545 if (f6i->fib6_nsiblings) {
5546 rt6_nh_nlmsg_size(nh, &nexthop_len);
5548 list_for_each_entry_safe(sibling, next_sibling,
5549 &f6i->fib6_siblings, fib6_siblings) {
5550 rt6_nh_nlmsg_size(sibling->fib6_nh, &nexthop_len);
5553 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5556 return NLMSG_ALIGN(sizeof(struct rtmsg))
5557 + nla_total_size(16) /* RTA_SRC */
5558 + nla_total_size(16) /* RTA_DST */
5559 + nla_total_size(16) /* RTA_GATEWAY */
5560 + nla_total_size(16) /* RTA_PREFSRC */
5561 + nla_total_size(4) /* RTA_TABLE */
5562 + nla_total_size(4) /* RTA_IIF */
5563 + nla_total_size(4) /* RTA_OIF */
5564 + nla_total_size(4) /* RTA_PRIORITY */
5565 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5566 + nla_total_size(sizeof(struct rta_cacheinfo))
5567 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5568 + nla_total_size(1) /* RTA_PREF */
5572 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5573 unsigned char *flags)
5575 if (nexthop_is_multipath(nh)) {
5578 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5580 goto nla_put_failure;
5582 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5583 goto nla_put_failure;
5585 nla_nest_end(skb, mp);
5587 struct fib6_nh *fib6_nh;
5589 fib6_nh = nexthop_fib6_nh(nh);
5590 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5592 goto nla_put_failure;
5601 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5602 struct fib6_info *rt, struct dst_entry *dst,
5603 struct in6_addr *dest, struct in6_addr *src,
5604 int iif, int type, u32 portid, u32 seq,
5607 struct rt6_info *rt6 = (struct rt6_info *)dst;
5608 struct rt6key *rt6_dst, *rt6_src;
5609 u32 *pmetrics, table, rt6_flags;
5610 unsigned char nh_flags = 0;
5611 struct nlmsghdr *nlh;
5615 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5620 rt6_dst = &rt6->rt6i_dst;
5621 rt6_src = &rt6->rt6i_src;
5622 rt6_flags = rt6->rt6i_flags;
5624 rt6_dst = &rt->fib6_dst;
5625 rt6_src = &rt->fib6_src;
5626 rt6_flags = rt->fib6_flags;
5629 rtm = nlmsg_data(nlh);
5630 rtm->rtm_family = AF_INET6;
5631 rtm->rtm_dst_len = rt6_dst->plen;
5632 rtm->rtm_src_len = rt6_src->plen;
5635 table = rt->fib6_table->tb6_id;
5637 table = RT6_TABLE_UNSPEC;
5638 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5639 if (nla_put_u32(skb, RTA_TABLE, table))
5640 goto nla_put_failure;
5642 rtm->rtm_type = rt->fib6_type;
5644 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5645 rtm->rtm_protocol = rt->fib6_protocol;
5647 if (rt6_flags & RTF_CACHE)
5648 rtm->rtm_flags |= RTM_F_CLONED;
5651 if (nla_put_in6_addr(skb, RTA_DST, dest))
5652 goto nla_put_failure;
5653 rtm->rtm_dst_len = 128;
5654 } else if (rtm->rtm_dst_len)
5655 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5656 goto nla_put_failure;
5657 #ifdef CONFIG_IPV6_SUBTREES
5659 if (nla_put_in6_addr(skb, RTA_SRC, src))
5660 goto nla_put_failure;
5661 rtm->rtm_src_len = 128;
5662 } else if (rtm->rtm_src_len &&
5663 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5664 goto nla_put_failure;
5667 #ifdef CONFIG_IPV6_MROUTE
5668 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5669 int err = ip6mr_get_route(net, skb, rtm, portid);
5674 goto nla_put_failure;
5677 if (nla_put_u32(skb, RTA_IIF, iif))
5678 goto nla_put_failure;
5680 struct in6_addr saddr_buf;
5681 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
5682 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5683 goto nla_put_failure;
5686 if (rt->fib6_prefsrc.plen) {
5687 struct in6_addr saddr_buf;
5688 saddr_buf = rt->fib6_prefsrc.addr;
5689 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5690 goto nla_put_failure;
5693 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5694 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5695 goto nla_put_failure;
5697 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5698 goto nla_put_failure;
5700 /* For multipath routes, walk the siblings list and add
5701 * each as a nexthop within RTA_MULTIPATH.
5704 if (rt6_flags & RTF_GATEWAY &&
5705 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5706 goto nla_put_failure;
5708 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5709 goto nla_put_failure;
5711 if (dst->lwtstate &&
5712 lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5713 goto nla_put_failure;
5714 } else if (rt->fib6_nsiblings) {
5715 struct fib6_info *sibling, *next_sibling;
5718 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5720 goto nla_put_failure;
5722 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5723 rt->fib6_nh->fib_nh_weight, AF_INET6,
5725 goto nla_put_failure;
5727 list_for_each_entry_safe(sibling, next_sibling,
5728 &rt->fib6_siblings, fib6_siblings) {
5729 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5730 sibling->fib6_nh->fib_nh_weight,
5732 goto nla_put_failure;
5735 nla_nest_end(skb, mp);
5736 } else if (rt->nh) {
5737 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5738 goto nla_put_failure;
5740 if (nexthop_is_blackhole(rt->nh))
5741 rtm->rtm_type = RTN_BLACKHOLE;
5743 if (READ_ONCE(net->ipv4.sysctl_nexthop_compat_mode) &&
5744 rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5745 goto nla_put_failure;
5747 rtm->rtm_flags |= nh_flags;
5749 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5750 &nh_flags, false) < 0)
5751 goto nla_put_failure;
5753 rtm->rtm_flags |= nh_flags;
5756 if (rt6_flags & RTF_EXPIRES) {
5757 expires = dst ? dst->expires : rt->expires;
5762 if (READ_ONCE(rt->offload))
5763 rtm->rtm_flags |= RTM_F_OFFLOAD;
5764 if (READ_ONCE(rt->trap))
5765 rtm->rtm_flags |= RTM_F_TRAP;
5766 if (READ_ONCE(rt->offload_failed))
5767 rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5770 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5771 goto nla_put_failure;
5773 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5774 goto nla_put_failure;
5777 nlmsg_end(skb, nlh);
5781 nlmsg_cancel(skb, nlh);
5785 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5787 const struct net_device *dev = arg;
5789 if (nh->fib_nh_dev == dev)
5795 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5796 const struct net_device *dev)
5799 struct net_device *_dev = (struct net_device *)dev;
5801 return !!nexthop_for_each_fib6_nh(f6i->nh,
5802 fib6_info_nh_uses_dev,
5806 if (f6i->fib6_nh->fib_nh_dev == dev)
5809 if (f6i->fib6_nsiblings) {
5810 struct fib6_info *sibling, *next_sibling;
5812 list_for_each_entry_safe(sibling, next_sibling,
5813 &f6i->fib6_siblings, fib6_siblings) {
5814 if (sibling->fib6_nh->fib_nh_dev == dev)
5822 struct fib6_nh_exception_dump_walker {
5823 struct rt6_rtnl_dump_arg *dump;
5824 struct fib6_info *rt;
5830 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5832 struct fib6_nh_exception_dump_walker *w = arg;
5833 struct rt6_rtnl_dump_arg *dump = w->dump;
5834 struct rt6_exception_bucket *bucket;
5835 struct rt6_exception *rt6_ex;
5838 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5842 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5843 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5849 /* Expiration of entries doesn't bump sernum, insertion
5850 * does. Removal is triggered by insertion, so we can
5851 * rely on the fact that if entries change between two
5852 * partial dumps, this node is scanned again completely,
5853 * see rt6_insert_exception() and fib6_dump_table().
5855 * Count expired entries we go through as handled
5856 * entries that we'll skip next time, in case of partial
5857 * node dump. Otherwise, if entries expire meanwhile,
5858 * we'll skip the wrong amount.
5860 if (rt6_check_expired(rt6_ex->rt6i)) {
5865 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5866 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5868 NETLINK_CB(dump->cb->skb).portid,
5869 dump->cb->nlh->nlmsg_seq, w->flags);
5881 /* Return -1 if done with node, number of handled routes on partial dump */
5882 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5884 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5885 struct fib_dump_filter *filter = &arg->filter;
5886 unsigned int flags = NLM_F_MULTI;
5887 struct net *net = arg->net;
5890 if (rt == net->ipv6.fib6_null_entry)
5893 if ((filter->flags & RTM_F_PREFIX) &&
5894 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5895 /* success since this is not a prefix route */
5898 if (filter->filter_set &&
5899 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5900 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5901 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5905 if (filter->filter_set ||
5906 !filter->dump_routes || !filter->dump_exceptions) {
5907 flags |= NLM_F_DUMP_FILTERED;
5910 if (filter->dump_routes) {
5914 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5916 NETLINK_CB(arg->cb->skb).portid,
5917 arg->cb->nlh->nlmsg_seq, flags)) {
5924 if (filter->dump_exceptions) {
5925 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5934 err = nexthop_for_each_fib6_nh(rt->nh,
5935 rt6_nh_dump_exceptions,
5938 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5943 return count + w.count;
5949 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5950 const struct nlmsghdr *nlh,
5952 struct netlink_ext_ack *extack)
5957 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5958 NL_SET_ERR_MSG_MOD(extack,
5959 "Invalid header for get route request");
5963 if (!netlink_strict_get_check(skb))
5964 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5965 rtm_ipv6_policy, extack);
5967 rtm = nlmsg_data(nlh);
5968 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5969 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5970 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5972 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5975 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5976 NL_SET_ERR_MSG_MOD(extack,
5977 "Invalid flags for get route request");
5981 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5982 rtm_ipv6_policy, extack);
5986 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5987 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5988 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5992 for (i = 0; i <= RTA_MAX; i++) {
6008 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
6016 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6017 struct netlink_ext_ack *extack)
6019 struct net *net = sock_net(in_skb->sk);
6020 struct nlattr *tb[RTA_MAX+1];
6021 int err, iif = 0, oif = 0;
6022 struct fib6_info *from;
6023 struct dst_entry *dst;
6024 struct rt6_info *rt;
6025 struct sk_buff *skb;
6027 struct flowi6 fl6 = {};
6030 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
6035 rtm = nlmsg_data(nlh);
6036 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
6037 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
6040 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
6043 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
6047 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
6050 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
6054 iif = nla_get_u32(tb[RTA_IIF]);
6057 oif = nla_get_u32(tb[RTA_OIF]);
6060 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6063 fl6.flowi6_uid = make_kuid(current_user_ns(),
6064 nla_get_u32(tb[RTA_UID]));
6066 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6069 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6072 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6074 if (tb[RTA_IP_PROTO]) {
6075 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
6076 &fl6.flowi6_proto, AF_INET6,
6083 struct net_device *dev;
6088 dev = dev_get_by_index_rcu(net, iif);
6095 fl6.flowi6_iif = iif;
6097 if (!ipv6_addr_any(&fl6.saddr))
6098 flags |= RT6_LOOKUP_F_HAS_SADDR;
6100 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
6104 fl6.flowi6_oif = oif;
6106 dst = ip6_route_output(net, NULL, &fl6);
6110 rt = container_of(dst, struct rt6_info, dst);
6111 if (rt->dst.error) {
6112 err = rt->dst.error;
6117 if (rt == net->ipv6.ip6_null_entry) {
6118 err = rt->dst.error;
6123 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
6130 skb_dst_set(skb, &rt->dst);
6133 from = rcu_dereference(rt->from);
6136 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6138 NETLINK_CB(in_skb).portid,
6141 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6142 &fl6.saddr, iif, RTM_NEWROUTE,
6143 NETLINK_CB(in_skb).portid,
6155 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6160 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6161 unsigned int nlm_flags)
6163 struct sk_buff *skb;
6164 struct net *net = info->nl_net;
6169 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6171 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6175 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6176 event, info->portid, seq, nlm_flags);
6178 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6179 WARN_ON(err == -EMSGSIZE);
6183 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6184 info->nlh, gfp_any());
6188 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6191 void fib6_rt_update(struct net *net, struct fib6_info *rt,
6192 struct nl_info *info)
6194 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6195 struct sk_buff *skb;
6198 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6202 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6203 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6205 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6206 WARN_ON(err == -EMSGSIZE);
6210 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6211 info->nlh, gfp_any());
6215 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6218 void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6219 bool offload, bool trap, bool offload_failed)
6221 struct sk_buff *skb;
6224 if (READ_ONCE(f6i->offload) == offload &&
6225 READ_ONCE(f6i->trap) == trap &&
6226 READ_ONCE(f6i->offload_failed) == offload_failed)
6229 WRITE_ONCE(f6i->offload, offload);
6230 WRITE_ONCE(f6i->trap, trap);
6232 /* 2 means send notifications only if offload_failed was changed. */
6233 if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6234 READ_ONCE(f6i->offload_failed) == offload_failed)
6237 WRITE_ONCE(f6i->offload_failed, offload_failed);
6239 if (!rcu_access_pointer(f6i->fib6_node))
6240 /* The route was removed from the tree, do not send
6245 if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6248 skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6254 err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6257 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6258 WARN_ON(err == -EMSGSIZE);
6263 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6267 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6269 EXPORT_SYMBOL(fib6_info_hw_flags_set);
6271 static int ip6_route_dev_notify(struct notifier_block *this,
6272 unsigned long event, void *ptr)
6274 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6275 struct net *net = dev_net(dev);
6277 if (!(dev->flags & IFF_LOOPBACK))
6280 if (event == NETDEV_REGISTER) {
6281 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6282 net->ipv6.ip6_null_entry->dst.dev = dev;
6283 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6284 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6285 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6286 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6287 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6288 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6290 } else if (event == NETDEV_UNREGISTER &&
6291 dev->reg_state != NETREG_UNREGISTERED) {
6292 /* NETDEV_UNREGISTER could be fired for multiple times by
6293 * netdev_wait_allrefs(). Make sure we only call this once.
6295 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6296 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6297 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6298 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6309 #ifdef CONFIG_PROC_FS
6310 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6312 struct net *net = (struct net *)seq->private;
6313 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6314 net->ipv6.rt6_stats->fib_nodes,
6315 net->ipv6.rt6_stats->fib_route_nodes,
6316 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6317 net->ipv6.rt6_stats->fib_rt_entries,
6318 net->ipv6.rt6_stats->fib_rt_cache,
6319 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6320 net->ipv6.rt6_stats->fib_discarded_routes);
6324 #endif /* CONFIG_PROC_FS */
6326 #ifdef CONFIG_SYSCTL
6328 static int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6329 void *buffer, size_t *lenp, loff_t *ppos)
6337 net = (struct net *)ctl->extra1;
6338 delay = net->ipv6.sysctl.flush_delay;
6339 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6343 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6347 static struct ctl_table ipv6_route_table_template[] = {
6349 .procname = "max_size",
6350 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
6351 .maxlen = sizeof(int),
6353 .proc_handler = proc_dointvec,
6356 .procname = "gc_thresh",
6357 .data = &ip6_dst_ops_template.gc_thresh,
6358 .maxlen = sizeof(int),
6360 .proc_handler = proc_dointvec,
6363 .procname = "flush",
6364 .data = &init_net.ipv6.sysctl.flush_delay,
6365 .maxlen = sizeof(int),
6367 .proc_handler = ipv6_sysctl_rtcache_flush
6370 .procname = "gc_min_interval",
6371 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6372 .maxlen = sizeof(int),
6374 .proc_handler = proc_dointvec_jiffies,
6377 .procname = "gc_timeout",
6378 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6379 .maxlen = sizeof(int),
6381 .proc_handler = proc_dointvec_jiffies,
6384 .procname = "gc_interval",
6385 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
6386 .maxlen = sizeof(int),
6388 .proc_handler = proc_dointvec_jiffies,
6391 .procname = "gc_elasticity",
6392 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6393 .maxlen = sizeof(int),
6395 .proc_handler = proc_dointvec,
6398 .procname = "mtu_expires",
6399 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6400 .maxlen = sizeof(int),
6402 .proc_handler = proc_dointvec_jiffies,
6405 .procname = "min_adv_mss",
6406 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
6407 .maxlen = sizeof(int),
6409 .proc_handler = proc_dointvec,
6412 .procname = "gc_min_interval_ms",
6413 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6414 .maxlen = sizeof(int),
6416 .proc_handler = proc_dointvec_ms_jiffies,
6419 .procname = "skip_notify_on_dev_down",
6420 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6421 .maxlen = sizeof(u8),
6423 .proc_handler = proc_dou8vec_minmax,
6424 .extra1 = SYSCTL_ZERO,
6425 .extra2 = SYSCTL_ONE,
6430 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6432 struct ctl_table *table;
6434 table = kmemdup(ipv6_route_table_template,
6435 sizeof(ipv6_route_table_template),
6439 table[0].data = &net->ipv6.sysctl.ip6_rt_max_size;
6440 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6441 table[2].data = &net->ipv6.sysctl.flush_delay;
6442 table[2].extra1 = net;
6443 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6444 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6445 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6446 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6447 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6448 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6449 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6450 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6452 /* Don't export sysctls to unprivileged users */
6453 if (net->user_ns != &init_user_ns)
6454 table[1].procname = NULL;
6461 static int __net_init ip6_route_net_init(struct net *net)
6465 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6466 sizeof(net->ipv6.ip6_dst_ops));
6468 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6469 goto out_ip6_dst_ops;
6471 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6472 if (!net->ipv6.fib6_null_entry)
6473 goto out_ip6_dst_entries;
6474 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6475 sizeof(*net->ipv6.fib6_null_entry));
6477 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6478 sizeof(*net->ipv6.ip6_null_entry),
6480 if (!net->ipv6.ip6_null_entry)
6481 goto out_fib6_null_entry;
6482 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6483 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6484 ip6_template_metrics, true);
6485 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->dst.rt_uncached);
6487 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6488 net->ipv6.fib6_has_custom_rules = false;
6489 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6490 sizeof(*net->ipv6.ip6_prohibit_entry),
6492 if (!net->ipv6.ip6_prohibit_entry)
6493 goto out_ip6_null_entry;
6494 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6495 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6496 ip6_template_metrics, true);
6497 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->dst.rt_uncached);
6499 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6500 sizeof(*net->ipv6.ip6_blk_hole_entry),
6502 if (!net->ipv6.ip6_blk_hole_entry)
6503 goto out_ip6_prohibit_entry;
6504 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6505 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6506 ip6_template_metrics, true);
6507 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->dst.rt_uncached);
6508 #ifdef CONFIG_IPV6_SUBTREES
6509 net->ipv6.fib6_routes_require_src = 0;
6513 net->ipv6.sysctl.flush_delay = 0;
6514 net->ipv6.sysctl.ip6_rt_max_size = INT_MAX;
6515 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6516 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6517 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6518 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6519 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6520 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6521 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6523 atomic_set(&net->ipv6.ip6_rt_gc_expire, 30*HZ);
6529 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6530 out_ip6_prohibit_entry:
6531 kfree(net->ipv6.ip6_prohibit_entry);
6533 kfree(net->ipv6.ip6_null_entry);
6535 out_fib6_null_entry:
6536 kfree(net->ipv6.fib6_null_entry);
6537 out_ip6_dst_entries:
6538 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6543 static void __net_exit ip6_route_net_exit(struct net *net)
6545 kfree(net->ipv6.fib6_null_entry);
6546 kfree(net->ipv6.ip6_null_entry);
6547 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6548 kfree(net->ipv6.ip6_prohibit_entry);
6549 kfree(net->ipv6.ip6_blk_hole_entry);
6551 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6554 static int __net_init ip6_route_net_init_late(struct net *net)
6556 #ifdef CONFIG_PROC_FS
6557 if (!proc_create_net("ipv6_route", 0, net->proc_net,
6558 &ipv6_route_seq_ops,
6559 sizeof(struct ipv6_route_iter)))
6562 if (!proc_create_net_single("rt6_stats", 0444, net->proc_net,
6563 rt6_stats_seq_show, NULL)) {
6564 remove_proc_entry("ipv6_route", net->proc_net);
6571 static void __net_exit ip6_route_net_exit_late(struct net *net)
6573 #ifdef CONFIG_PROC_FS
6574 remove_proc_entry("ipv6_route", net->proc_net);
6575 remove_proc_entry("rt6_stats", net->proc_net);
6579 static struct pernet_operations ip6_route_net_ops = {
6580 .init = ip6_route_net_init,
6581 .exit = ip6_route_net_exit,
6584 static int __net_init ipv6_inetpeer_init(struct net *net)
6586 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6590 inet_peer_base_init(bp);
6591 net->ipv6.peers = bp;
6595 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6597 struct inet_peer_base *bp = net->ipv6.peers;
6599 net->ipv6.peers = NULL;
6600 inetpeer_invalidate_tree(bp);
6604 static struct pernet_operations ipv6_inetpeer_ops = {
6605 .init = ipv6_inetpeer_init,
6606 .exit = ipv6_inetpeer_exit,
6609 static struct pernet_operations ip6_route_net_late_ops = {
6610 .init = ip6_route_net_init_late,
6611 .exit = ip6_route_net_exit_late,
6614 static struct notifier_block ip6_route_dev_notifier = {
6615 .notifier_call = ip6_route_dev_notify,
6616 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6619 void __init ip6_route_init_special_entries(void)
6621 /* Registering of the loopback is done before this portion of code,
6622 * the loopback reference in rt6_info will not be taken, do it
6623 * manually for init_net */
6624 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6625 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6626 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6627 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6628 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6629 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6630 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6631 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6635 #if IS_BUILTIN(CONFIG_IPV6)
6636 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6637 DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6639 BTF_ID_LIST(btf_fib6_info_id)
6640 BTF_ID(struct, fib6_info)
6642 static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6643 .seq_ops = &ipv6_route_seq_ops,
6644 .init_seq_private = bpf_iter_init_seq_net,
6645 .fini_seq_private = bpf_iter_fini_seq_net,
6646 .seq_priv_size = sizeof(struct ipv6_route_iter),
6649 static struct bpf_iter_reg ipv6_route_reg_info = {
6650 .target = "ipv6_route",
6651 .ctx_arg_info_size = 1,
6653 { offsetof(struct bpf_iter__ipv6_route, rt),
6654 PTR_TO_BTF_ID_OR_NULL },
6656 .seq_info = &ipv6_route_seq_info,
6659 static int __init bpf_iter_register(void)
6661 ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6662 return bpf_iter_reg_target(&ipv6_route_reg_info);
6665 static void bpf_iter_unregister(void)
6667 bpf_iter_unreg_target(&ipv6_route_reg_info);
6672 int __init ip6_route_init(void)
6678 ip6_dst_ops_template.kmem_cachep =
6679 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6680 SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL);
6681 if (!ip6_dst_ops_template.kmem_cachep)
6684 ret = dst_entries_init(&ip6_dst_blackhole_ops);
6686 goto out_kmem_cache;
6688 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6690 goto out_dst_entries;
6692 ret = register_pernet_subsys(&ip6_route_net_ops);
6694 goto out_register_inetpeer;
6696 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6700 goto out_register_subsys;
6706 ret = fib6_rules_init();
6710 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6712 goto fib6_rules_init;
6714 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6715 inet6_rtm_newroute, NULL, 0);
6717 goto out_register_late_subsys;
6719 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6720 inet6_rtm_delroute, NULL, 0);
6722 goto out_register_late_subsys;
6724 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6725 inet6_rtm_getroute, NULL,
6726 RTNL_FLAG_DOIT_UNLOCKED);
6728 goto out_register_late_subsys;
6730 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6732 goto out_register_late_subsys;
6734 #if IS_BUILTIN(CONFIG_IPV6)
6735 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6736 ret = bpf_iter_register();
6738 goto out_register_late_subsys;
6742 for_each_possible_cpu(cpu) {
6743 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6745 INIT_LIST_HEAD(&ul->head);
6746 INIT_LIST_HEAD(&ul->quarantine);
6747 spin_lock_init(&ul->lock);
6753 out_register_late_subsys:
6754 rtnl_unregister_all(PF_INET6);
6755 unregister_pernet_subsys(&ip6_route_net_late_ops);
6757 fib6_rules_cleanup();
6762 out_register_subsys:
6763 unregister_pernet_subsys(&ip6_route_net_ops);
6764 out_register_inetpeer:
6765 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6767 dst_entries_destroy(&ip6_dst_blackhole_ops);
6769 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6773 void ip6_route_cleanup(void)
6775 #if IS_BUILTIN(CONFIG_IPV6)
6776 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6777 bpf_iter_unregister();
6780 unregister_netdevice_notifier(&ip6_route_dev_notifier);
6781 unregister_pernet_subsys(&ip6_route_net_late_ops);
6782 fib6_rules_cleanup();
6785 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6786 unregister_pernet_subsys(&ip6_route_net_ops);
6787 dst_entries_destroy(&ip6_dst_blackhole_ops);
6788 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);