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 <net/net_namespace.h>
47 #include <net/ip6_fib.h>
48 #include <net/ip6_route.h>
49 #include <net/ndisc.h>
50 #include <net/addrconf.h>
52 #include <linux/rtnetlink.h>
54 #include <net/dst_metadata.h>
56 #include <net/netevent.h>
57 #include <net/netlink.h>
59 #include <net/lwtunnel.h>
60 #include <net/ip_tunnels.h>
61 #include <net/l3mdev.h>
63 #include <linux/uaccess.h>
66 #include <linux/sysctl.h>
69 static int ip6_rt_type_to_error(u8 fib6_type);
71 #define CREATE_TRACE_POINTS
72 #include <trace/events/fib6.h>
73 EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
74 #undef CREATE_TRACE_POINTS
77 RT6_NUD_FAIL_HARD = -3,
78 RT6_NUD_FAIL_PROBE = -2,
79 RT6_NUD_FAIL_DO_RR = -1,
83 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
84 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
85 static unsigned int ip6_mtu(const struct dst_entry *dst);
86 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
87 static void ip6_dst_destroy(struct dst_entry *);
88 static void ip6_dst_ifdown(struct dst_entry *,
89 struct net_device *dev, int how);
90 static int ip6_dst_gc(struct dst_ops *ops);
92 static int ip6_pkt_discard(struct sk_buff *skb);
93 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
94 static int ip6_pkt_prohibit(struct sk_buff *skb);
95 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
96 static void ip6_link_failure(struct sk_buff *skb);
97 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
98 struct sk_buff *skb, u32 mtu);
99 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
100 struct sk_buff *skb);
101 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
103 static size_t rt6_nlmsg_size(struct fib6_info *f6i);
104 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
105 struct fib6_info *rt, struct dst_entry *dst,
106 struct in6_addr *dest, struct in6_addr *src,
107 int iif, int type, u32 portid, u32 seq,
109 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
110 const struct in6_addr *daddr,
111 const struct in6_addr *saddr);
113 #ifdef CONFIG_IPV6_ROUTE_INFO
114 static struct fib6_info *rt6_add_route_info(struct net *net,
115 const struct in6_addr *prefix, int prefixlen,
116 const struct in6_addr *gwaddr,
117 struct net_device *dev,
119 static struct fib6_info *rt6_get_route_info(struct net *net,
120 const struct in6_addr *prefix, int prefixlen,
121 const struct in6_addr *gwaddr,
122 struct net_device *dev);
125 struct uncached_list {
127 struct list_head head;
130 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
132 void rt6_uncached_list_add(struct rt6_info *rt)
134 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
136 rt->rt6i_uncached_list = ul;
138 spin_lock_bh(&ul->lock);
139 list_add_tail(&rt->rt6i_uncached, &ul->head);
140 spin_unlock_bh(&ul->lock);
143 void rt6_uncached_list_del(struct rt6_info *rt)
145 if (!list_empty(&rt->rt6i_uncached)) {
146 struct uncached_list *ul = rt->rt6i_uncached_list;
147 struct net *net = dev_net(rt->dst.dev);
149 spin_lock_bh(&ul->lock);
150 list_del(&rt->rt6i_uncached);
151 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
152 spin_unlock_bh(&ul->lock);
156 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
158 struct net_device *loopback_dev = net->loopback_dev;
161 if (dev == loopback_dev)
164 for_each_possible_cpu(cpu) {
165 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
168 spin_lock_bh(&ul->lock);
169 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
170 struct inet6_dev *rt_idev = rt->rt6i_idev;
171 struct net_device *rt_dev = rt->dst.dev;
173 if (rt_idev->dev == dev) {
174 rt->rt6i_idev = in6_dev_get(loopback_dev);
175 in6_dev_put(rt_idev);
179 rt->dst.dev = loopback_dev;
180 dev_hold(rt->dst.dev);
184 spin_unlock_bh(&ul->lock);
188 static inline const void *choose_neigh_daddr(const struct in6_addr *p,
192 if (!ipv6_addr_any(p))
193 return (const void *) p;
195 return &ipv6_hdr(skb)->daddr;
199 struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
200 struct net_device *dev,
206 daddr = choose_neigh_daddr(gw, skb, daddr);
207 n = __ipv6_neigh_lookup(dev, daddr);
211 n = neigh_create(&nd_tbl, daddr, dev);
212 return IS_ERR(n) ? NULL : n;
215 static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
219 const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
221 return ip6_neigh_lookup(&rt->rt6i_gateway, dst->dev, skb, daddr);
224 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
226 struct net_device *dev = dst->dev;
227 struct rt6_info *rt = (struct rt6_info *)dst;
229 daddr = choose_neigh_daddr(&rt->rt6i_gateway, NULL, daddr);
232 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
234 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
236 __ipv6_confirm_neigh(dev, daddr);
239 static struct dst_ops ip6_dst_ops_template = {
243 .check = ip6_dst_check,
244 .default_advmss = ip6_default_advmss,
246 .cow_metrics = dst_cow_metrics_generic,
247 .destroy = ip6_dst_destroy,
248 .ifdown = ip6_dst_ifdown,
249 .negative_advice = ip6_negative_advice,
250 .link_failure = ip6_link_failure,
251 .update_pmtu = ip6_rt_update_pmtu,
252 .redirect = rt6_do_redirect,
253 .local_out = __ip6_local_out,
254 .neigh_lookup = ip6_dst_neigh_lookup,
255 .confirm_neigh = ip6_confirm_neigh,
258 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
260 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
262 return mtu ? : dst->dev->mtu;
265 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
266 struct sk_buff *skb, u32 mtu)
270 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
275 static struct dst_ops ip6_dst_blackhole_ops = {
277 .destroy = ip6_dst_destroy,
278 .check = ip6_dst_check,
279 .mtu = ip6_blackhole_mtu,
280 .default_advmss = ip6_default_advmss,
281 .update_pmtu = ip6_rt_blackhole_update_pmtu,
282 .redirect = ip6_rt_blackhole_redirect,
283 .cow_metrics = dst_cow_metrics_generic,
284 .neigh_lookup = ip6_dst_neigh_lookup,
287 static const u32 ip6_template_metrics[RTAX_MAX] = {
288 [RTAX_HOPLIMIT - 1] = 0,
291 static const struct fib6_info fib6_null_entry_template = {
292 .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
293 .fib6_protocol = RTPROT_KERNEL,
294 .fib6_metric = ~(u32)0,
295 .fib6_ref = REFCOUNT_INIT(1),
296 .fib6_type = RTN_UNREACHABLE,
297 .fib6_metrics = (struct dst_metrics *)&dst_default_metrics,
300 static const struct rt6_info ip6_null_entry_template = {
302 .__refcnt = ATOMIC_INIT(1),
304 .obsolete = DST_OBSOLETE_FORCE_CHK,
305 .error = -ENETUNREACH,
306 .input = ip6_pkt_discard,
307 .output = ip6_pkt_discard_out,
309 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
312 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
314 static const struct rt6_info ip6_prohibit_entry_template = {
316 .__refcnt = ATOMIC_INIT(1),
318 .obsolete = DST_OBSOLETE_FORCE_CHK,
320 .input = ip6_pkt_prohibit,
321 .output = ip6_pkt_prohibit_out,
323 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
326 static const struct rt6_info ip6_blk_hole_entry_template = {
328 .__refcnt = ATOMIC_INIT(1),
330 .obsolete = DST_OBSOLETE_FORCE_CHK,
332 .input = dst_discard,
333 .output = dst_discard_out,
335 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
340 static void rt6_info_init(struct rt6_info *rt)
342 struct dst_entry *dst = &rt->dst;
344 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
345 INIT_LIST_HEAD(&rt->rt6i_uncached);
348 /* allocate dst with ip6_dst_ops */
349 struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
352 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
353 1, DST_OBSOLETE_FORCE_CHK, flags);
357 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
362 EXPORT_SYMBOL(ip6_dst_alloc);
364 static void ip6_dst_destroy(struct dst_entry *dst)
366 struct rt6_info *rt = (struct rt6_info *)dst;
367 struct fib6_info *from;
368 struct inet6_dev *idev;
370 ip_dst_metrics_put(dst);
371 rt6_uncached_list_del(rt);
373 idev = rt->rt6i_idev;
375 rt->rt6i_idev = NULL;
379 from = xchg((__force struct fib6_info **)&rt->from, NULL);
380 fib6_info_release(from);
383 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
386 struct rt6_info *rt = (struct rt6_info *)dst;
387 struct inet6_dev *idev = rt->rt6i_idev;
388 struct net_device *loopback_dev =
389 dev_net(dev)->loopback_dev;
391 if (idev && idev->dev != loopback_dev) {
392 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
394 rt->rt6i_idev = loopback_idev;
400 static bool __rt6_check_expired(const struct rt6_info *rt)
402 if (rt->rt6i_flags & RTF_EXPIRES)
403 return time_after(jiffies, rt->dst.expires);
408 static bool rt6_check_expired(const struct rt6_info *rt)
410 struct fib6_info *from;
412 from = rcu_dereference(rt->from);
414 if (rt->rt6i_flags & RTF_EXPIRES) {
415 if (time_after(jiffies, rt->dst.expires))
418 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
419 fib6_check_expired(from);
424 void fib6_select_path(const struct net *net, struct fib6_result *res,
425 struct flowi6 *fl6, int oif, bool have_oif_match,
426 const struct sk_buff *skb, int strict)
428 struct fib6_info *sibling, *next_sibling;
429 struct fib6_info *match = res->f6i;
431 if ((!match->fib6_nsiblings && !match->nh) || have_oif_match)
434 /* We might have already computed the hash for ICMPv6 errors. In such
435 * case it will always be non-zero. Otherwise now is the time to do it.
438 (!match->nh || nexthop_is_multipath(match->nh)))
439 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
441 if (unlikely(match->nh)) {
442 nexthop_path_fib6_result(res, fl6->mp_hash);
446 if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
449 list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
451 const struct fib6_nh *nh = sibling->fib6_nh;
454 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
455 if (fl6->mp_hash > nh_upper_bound)
457 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
465 res->nh = match->fib6_nh;
469 * Route lookup. rcu_read_lock() should be held.
472 static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
473 const struct in6_addr *saddr, int oif, int flags)
475 const struct net_device *dev;
477 if (nh->fib_nh_flags & RTNH_F_DEAD)
480 dev = nh->fib_nh_dev;
482 if (dev->ifindex == oif)
485 if (ipv6_chk_addr(net, saddr, dev,
486 flags & RT6_LOOKUP_F_IFACE))
493 struct fib6_nh_dm_arg {
495 const struct in6_addr *saddr;
501 static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
503 struct fib6_nh_dm_arg *arg = _arg;
506 return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
510 /* returns fib6_nh from nexthop or NULL */
511 static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
512 struct fib6_result *res,
513 const struct in6_addr *saddr,
516 struct fib6_nh_dm_arg arg = {
523 if (nexthop_is_blackhole(nh))
526 if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
532 static void rt6_device_match(struct net *net, struct fib6_result *res,
533 const struct in6_addr *saddr, int oif, int flags)
535 struct fib6_info *f6i = res->f6i;
536 struct fib6_info *spf6i;
539 if (!oif && ipv6_addr_any(saddr)) {
540 if (unlikely(f6i->nh)) {
541 nh = nexthop_fib6_nh(f6i->nh);
542 if (nexthop_is_blackhole(f6i->nh))
547 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
551 for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
552 bool matched = false;
554 if (unlikely(spf6i->nh)) {
555 nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
561 if (__rt6_device_match(net, nh, saddr, oif, flags))
570 if (oif && flags & RT6_LOOKUP_F_IFACE) {
571 res->f6i = net->ipv6.fib6_null_entry;
572 nh = res->f6i->fib6_nh;
576 if (unlikely(f6i->nh)) {
577 nh = nexthop_fib6_nh(f6i->nh);
578 if (nexthop_is_blackhole(f6i->nh))
584 if (nh->fib_nh_flags & RTNH_F_DEAD) {
585 res->f6i = net->ipv6.fib6_null_entry;
586 nh = res->f6i->fib6_nh;
590 res->fib6_type = res->f6i->fib6_type;
591 res->fib6_flags = res->f6i->fib6_flags;
595 res->fib6_flags |= RTF_REJECT;
596 res->fib6_type = RTN_BLACKHOLE;
600 #ifdef CONFIG_IPV6_ROUTER_PREF
601 struct __rt6_probe_work {
602 struct work_struct work;
603 struct in6_addr target;
604 struct net_device *dev;
607 static void rt6_probe_deferred(struct work_struct *w)
609 struct in6_addr mcaddr;
610 struct __rt6_probe_work *work =
611 container_of(w, struct __rt6_probe_work, work);
613 addrconf_addr_solict_mult(&work->target, &mcaddr);
614 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
619 static void rt6_probe(struct fib6_nh *fib6_nh)
621 struct __rt6_probe_work *work = NULL;
622 const struct in6_addr *nh_gw;
623 struct neighbour *neigh;
624 struct net_device *dev;
625 struct inet6_dev *idev;
628 * Okay, this does not seem to be appropriate
629 * for now, however, we need to check if it
630 * is really so; aka Router Reachability Probing.
632 * Router Reachability Probe MUST be rate-limited
633 * to no more than one per minute.
635 if (fib6_nh->fib_nh_gw_family)
638 nh_gw = &fib6_nh->fib_nh_gw6;
639 dev = fib6_nh->fib_nh_dev;
641 idev = __in6_dev_get(dev);
642 neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
644 if (neigh->nud_state & NUD_VALID)
647 write_lock(&neigh->lock);
648 if (!(neigh->nud_state & NUD_VALID) &&
650 neigh->updated + idev->cnf.rtr_probe_interval)) {
651 work = kmalloc(sizeof(*work), GFP_ATOMIC);
653 __neigh_set_probe_once(neigh);
655 write_unlock(&neigh->lock);
656 } else if (time_after(jiffies, fib6_nh->last_probe +
657 idev->cnf.rtr_probe_interval)) {
658 work = kmalloc(sizeof(*work), GFP_ATOMIC);
662 fib6_nh->last_probe = jiffies;
663 INIT_WORK(&work->work, rt6_probe_deferred);
664 work->target = *nh_gw;
667 schedule_work(&work->work);
671 rcu_read_unlock_bh();
674 static inline void rt6_probe(struct fib6_nh *fib6_nh)
680 * Default Router Selection (RFC 2461 6.3.6)
682 static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
684 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
685 struct neighbour *neigh;
688 neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
689 &fib6_nh->fib_nh_gw6);
691 read_lock(&neigh->lock);
692 if (neigh->nud_state & NUD_VALID)
693 ret = RT6_NUD_SUCCEED;
694 #ifdef CONFIG_IPV6_ROUTER_PREF
695 else if (!(neigh->nud_state & NUD_FAILED))
696 ret = RT6_NUD_SUCCEED;
698 ret = RT6_NUD_FAIL_PROBE;
700 read_unlock(&neigh->lock);
702 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
703 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
705 rcu_read_unlock_bh();
710 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
715 if (!oif || nh->fib_nh_dev->ifindex == oif)
718 if (!m && (strict & RT6_LOOKUP_F_IFACE))
719 return RT6_NUD_FAIL_HARD;
720 #ifdef CONFIG_IPV6_ROUTER_PREF
721 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
723 if ((strict & RT6_LOOKUP_F_REACHABLE) &&
724 !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
725 int n = rt6_check_neigh(nh);
732 static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
733 int oif, int strict, int *mpri, bool *do_rr)
735 bool match_do_rr = false;
739 if (nh->fib_nh_flags & RTNH_F_DEAD)
742 if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
743 nh->fib_nh_flags & RTNH_F_LINKDOWN &&
744 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
747 m = rt6_score_route(nh, fib6_flags, oif, strict);
748 if (m == RT6_NUD_FAIL_DO_RR) {
750 m = 0; /* lowest valid score */
751 } else if (m == RT6_NUD_FAIL_HARD) {
755 if (strict & RT6_LOOKUP_F_REACHABLE)
758 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
760 *do_rr = match_do_rr;
768 struct fib6_nh_frl_arg {
777 static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
779 struct fib6_nh_frl_arg *arg = _arg;
782 return find_match(nh, arg->flags, arg->oif, arg->strict,
783 arg->mpri, arg->do_rr);
786 static void __find_rr_leaf(struct fib6_info *f6i_start,
787 struct fib6_info *nomatch, u32 metric,
788 struct fib6_result *res, struct fib6_info **cont,
789 int oif, int strict, bool *do_rr, int *mpri)
791 struct fib6_info *f6i;
793 for (f6i = f6i_start;
794 f6i && f6i != nomatch;
795 f6i = rcu_dereference(f6i->fib6_next)) {
796 bool matched = false;
799 if (cont && f6i->fib6_metric != metric) {
804 if (fib6_check_expired(f6i))
807 if (unlikely(f6i->nh)) {
808 struct fib6_nh_frl_arg arg = {
809 .flags = f6i->fib6_flags,
816 if (nexthop_is_blackhole(f6i->nh)) {
817 res->fib6_flags = RTF_REJECT;
818 res->fib6_type = RTN_BLACKHOLE;
820 res->nh = nexthop_fib6_nh(f6i->nh);
823 if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
830 if (find_match(nh, f6i->fib6_flags, oif, strict,
837 res->fib6_flags = f6i->fib6_flags;
838 res->fib6_type = f6i->fib6_type;
843 static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
844 struct fib6_info *rr_head, int oif, int strict,
845 bool *do_rr, struct fib6_result *res)
847 u32 metric = rr_head->fib6_metric;
848 struct fib6_info *cont = NULL;
851 __find_rr_leaf(rr_head, NULL, metric, res, &cont,
852 oif, strict, do_rr, &mpri);
854 __find_rr_leaf(leaf, rr_head, metric, res, &cont,
855 oif, strict, do_rr, &mpri);
857 if (res->f6i || !cont)
860 __find_rr_leaf(cont, NULL, metric, res, NULL,
861 oif, strict, do_rr, &mpri);
864 static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
865 struct fib6_result *res, int strict)
867 struct fib6_info *leaf = rcu_dereference(fn->leaf);
868 struct fib6_info *rt0;
872 /* make sure this function or its helpers sets f6i */
875 if (!leaf || leaf == net->ipv6.fib6_null_entry)
878 rt0 = rcu_dereference(fn->rr_ptr);
882 /* Double check to make sure fn is not an intermediate node
883 * and fn->leaf does not points to its child's leaf
884 * (This might happen if all routes under fn are deleted from
885 * the tree and fib6_repair_tree() is called on the node.)
887 key_plen = rt0->fib6_dst.plen;
888 #ifdef CONFIG_IPV6_SUBTREES
889 if (rt0->fib6_src.plen)
890 key_plen = rt0->fib6_src.plen;
892 if (fn->fn_bit != key_plen)
895 find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
897 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
899 /* no entries matched; do round-robin */
900 if (!next || next->fib6_metric != rt0->fib6_metric)
904 spin_lock_bh(&leaf->fib6_table->tb6_lock);
905 /* make sure next is not being deleted from the tree */
907 rcu_assign_pointer(fn->rr_ptr, next);
908 spin_unlock_bh(&leaf->fib6_table->tb6_lock);
914 res->f6i = net->ipv6.fib6_null_entry;
915 res->nh = res->f6i->fib6_nh;
916 res->fib6_flags = res->f6i->fib6_flags;
917 res->fib6_type = res->f6i->fib6_type;
921 static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
923 return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
924 res->nh->fib_nh_gw_family;
927 #ifdef CONFIG_IPV6_ROUTE_INFO
928 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
929 const struct in6_addr *gwaddr)
931 struct net *net = dev_net(dev);
932 struct route_info *rinfo = (struct route_info *) opt;
933 struct in6_addr prefix_buf, *prefix;
935 unsigned long lifetime;
936 struct fib6_info *rt;
938 if (len < sizeof(struct route_info)) {
942 /* Sanity check for prefix_len and length */
943 if (rinfo->length > 3) {
945 } else if (rinfo->prefix_len > 128) {
947 } else if (rinfo->prefix_len > 64) {
948 if (rinfo->length < 2) {
951 } else if (rinfo->prefix_len > 0) {
952 if (rinfo->length < 1) {
957 pref = rinfo->route_pref;
958 if (pref == ICMPV6_ROUTER_PREF_INVALID)
961 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
963 if (rinfo->length == 3)
964 prefix = (struct in6_addr *)rinfo->prefix;
966 /* this function is safe */
967 ipv6_addr_prefix(&prefix_buf,
968 (struct in6_addr *)rinfo->prefix,
970 prefix = &prefix_buf;
973 if (rinfo->prefix_len == 0)
974 rt = rt6_get_dflt_router(net, gwaddr, dev);
976 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
979 if (rt && !lifetime) {
985 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
988 rt->fib6_flags = RTF_ROUTEINFO |
989 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
992 if (!addrconf_finite_timeout(lifetime))
993 fib6_clean_expires(rt);
995 fib6_set_expires(rt, jiffies + HZ * lifetime);
997 fib6_info_release(rt);
1004 * Misc support functions
1007 /* called with rcu_lock held */
1008 static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1010 struct net_device *dev = res->nh->fib_nh_dev;
1012 if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1013 /* for copies of local routes, dst->dev needs to be the
1014 * device if it is a master device, the master device if
1015 * device is enslaved, and the loopback as the default
1017 if (netif_is_l3_slave(dev) &&
1018 !rt6_need_strict(&res->f6i->fib6_dst.addr))
1019 dev = l3mdev_master_dev_rcu(dev);
1020 else if (!netif_is_l3_master(dev))
1021 dev = dev_net(dev)->loopback_dev;
1022 /* last case is netif_is_l3_master(dev) is true in which
1023 * case we want dev returned to be dev
1030 static const int fib6_prop[RTN_MAX + 1] = {
1034 [RTN_BROADCAST] = 0,
1036 [RTN_MULTICAST] = 0,
1037 [RTN_BLACKHOLE] = -EINVAL,
1038 [RTN_UNREACHABLE] = -EHOSTUNREACH,
1039 [RTN_PROHIBIT] = -EACCES,
1040 [RTN_THROW] = -EAGAIN,
1041 [RTN_NAT] = -EINVAL,
1042 [RTN_XRESOLVE] = -EINVAL,
1045 static int ip6_rt_type_to_error(u8 fib6_type)
1047 return fib6_prop[fib6_type];
1050 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1052 unsigned short flags = 0;
1054 if (rt->dst_nocount)
1055 flags |= DST_NOCOUNT;
1056 if (rt->dst_nopolicy)
1057 flags |= DST_NOPOLICY;
1064 static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1066 rt->dst.error = ip6_rt_type_to_error(fib6_type);
1068 switch (fib6_type) {
1070 rt->dst.output = dst_discard_out;
1071 rt->dst.input = dst_discard;
1074 rt->dst.output = ip6_pkt_prohibit_out;
1075 rt->dst.input = ip6_pkt_prohibit;
1078 case RTN_UNREACHABLE:
1080 rt->dst.output = ip6_pkt_discard_out;
1081 rt->dst.input = ip6_pkt_discard;
1086 static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1088 struct fib6_info *f6i = res->f6i;
1090 if (res->fib6_flags & RTF_REJECT) {
1091 ip6_rt_init_dst_reject(rt, res->fib6_type);
1096 rt->dst.output = ip6_output;
1098 if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1099 rt->dst.input = ip6_input;
1100 } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1101 rt->dst.input = ip6_mc_input;
1103 rt->dst.input = ip6_forward;
1106 if (res->nh->fib_nh_lws) {
1107 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1108 lwtunnel_set_redirect(&rt->dst);
1111 rt->dst.lastuse = jiffies;
1114 /* Caller must already hold reference to @from */
1115 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1117 rt->rt6i_flags &= ~RTF_EXPIRES;
1118 rcu_assign_pointer(rt->from, from);
1119 ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1122 /* Caller must already hold reference to f6i in result */
1123 static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1125 const struct fib6_nh *nh = res->nh;
1126 const struct net_device *dev = nh->fib_nh_dev;
1127 struct fib6_info *f6i = res->f6i;
1129 ip6_rt_init_dst(rt, res);
1131 rt->rt6i_dst = f6i->fib6_dst;
1132 rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1133 rt->rt6i_flags = res->fib6_flags;
1134 if (nh->fib_nh_gw_family) {
1135 rt->rt6i_gateway = nh->fib_nh_gw6;
1136 rt->rt6i_flags |= RTF_GATEWAY;
1138 rt6_set_from(rt, f6i);
1139 #ifdef CONFIG_IPV6_SUBTREES
1140 rt->rt6i_src = f6i->fib6_src;
1144 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1145 struct in6_addr *saddr)
1147 struct fib6_node *pn, *sn;
1149 if (fn->fn_flags & RTN_TL_ROOT)
1151 pn = rcu_dereference(fn->parent);
1152 sn = FIB6_SUBTREE(pn);
1154 fn = fib6_node_lookup(sn, NULL, saddr);
1157 if (fn->fn_flags & RTN_RTINFO)
1162 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1164 struct rt6_info *rt = *prt;
1166 if (dst_hold_safe(&rt->dst))
1169 rt = net->ipv6.ip6_null_entry;
1178 /* called with rcu_lock held */
1179 static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1181 struct net_device *dev = res->nh->fib_nh_dev;
1182 struct fib6_info *f6i = res->f6i;
1183 unsigned short flags;
1184 struct rt6_info *nrt;
1186 if (!fib6_info_hold_safe(f6i))
1189 flags = fib6_info_dst_flags(f6i);
1190 nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1192 fib6_info_release(f6i);
1196 ip6_rt_copy_init(nrt, res);
1200 nrt = dev_net(dev)->ipv6.ip6_null_entry;
1201 dst_hold(&nrt->dst);
1205 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
1206 struct fib6_table *table,
1208 const struct sk_buff *skb,
1211 struct fib6_result res = {};
1212 struct fib6_node *fn;
1213 struct rt6_info *rt;
1215 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1216 flags &= ~RT6_LOOKUP_F_IFACE;
1219 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1221 res.f6i = rcu_dereference(fn->leaf);
1223 res.f6i = net->ipv6.fib6_null_entry;
1225 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1228 if (res.f6i == net->ipv6.fib6_null_entry) {
1229 fn = fib6_backtrack(fn, &fl6->saddr);
1233 rt = net->ipv6.ip6_null_entry;
1236 } else if (res.fib6_flags & RTF_REJECT) {
1240 fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1241 fl6->flowi6_oif != 0, skb, flags);
1243 /* Search through exception table */
1244 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1246 if (ip6_hold_safe(net, &rt))
1247 dst_use_noref(&rt->dst, jiffies);
1250 rt = ip6_create_rt_rcu(&res);
1254 trace_fib6_table_lookup(net, &res, table, fl6);
1261 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1262 const struct sk_buff *skb, int flags)
1264 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1266 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1268 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1269 const struct in6_addr *saddr, int oif,
1270 const struct sk_buff *skb, int strict)
1272 struct flowi6 fl6 = {
1276 struct dst_entry *dst;
1277 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1280 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1281 flags |= RT6_LOOKUP_F_HAS_SADDR;
1284 dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1285 if (dst->error == 0)
1286 return (struct rt6_info *) dst;
1292 EXPORT_SYMBOL(rt6_lookup);
1294 /* ip6_ins_rt is called with FREE table->tb6_lock.
1295 * It takes new route entry, the addition fails by any reason the
1296 * route is released.
1297 * Caller must hold dst before calling it.
1300 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1301 struct netlink_ext_ack *extack)
1304 struct fib6_table *table;
1306 table = rt->fib6_table;
1307 spin_lock_bh(&table->tb6_lock);
1308 err = fib6_add(&table->tb6_root, rt, info, extack);
1309 spin_unlock_bh(&table->tb6_lock);
1314 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1316 struct nl_info info = { .nl_net = net, };
1318 return __ip6_ins_rt(rt, &info, NULL);
1321 static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1322 const struct in6_addr *daddr,
1323 const struct in6_addr *saddr)
1325 struct fib6_info *f6i = res->f6i;
1326 struct net_device *dev;
1327 struct rt6_info *rt;
1333 if (!fib6_info_hold_safe(f6i))
1336 dev = ip6_rt_get_dev_rcu(res);
1337 rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1339 fib6_info_release(f6i);
1343 ip6_rt_copy_init(rt, res);
1344 rt->rt6i_flags |= RTF_CACHE;
1345 rt->dst.flags |= DST_HOST;
1346 rt->rt6i_dst.addr = *daddr;
1347 rt->rt6i_dst.plen = 128;
1349 if (!rt6_is_gw_or_nonexthop(res)) {
1350 if (f6i->fib6_dst.plen != 128 &&
1351 ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1352 rt->rt6i_flags |= RTF_ANYCAST;
1353 #ifdef CONFIG_IPV6_SUBTREES
1354 if (rt->rt6i_src.plen && saddr) {
1355 rt->rt6i_src.addr = *saddr;
1356 rt->rt6i_src.plen = 128;
1364 static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1366 struct fib6_info *f6i = res->f6i;
1367 unsigned short flags = fib6_info_dst_flags(f6i);
1368 struct net_device *dev;
1369 struct rt6_info *pcpu_rt;
1371 if (!fib6_info_hold_safe(f6i))
1375 dev = ip6_rt_get_dev_rcu(res);
1376 pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags);
1379 fib6_info_release(f6i);
1382 ip6_rt_copy_init(pcpu_rt, res);
1383 pcpu_rt->rt6i_flags |= RTF_PCPU;
1387 /* It should be called with rcu_read_lock() acquired */
1388 static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1390 struct rt6_info *pcpu_rt;
1392 pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1397 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1398 const struct fib6_result *res)
1400 struct rt6_info *pcpu_rt, *prev, **p;
1402 pcpu_rt = ip6_rt_pcpu_alloc(res);
1406 p = this_cpu_ptr(res->nh->rt6i_pcpu);
1407 prev = cmpxchg(p, NULL, pcpu_rt);
1410 if (res->f6i->fib6_destroying) {
1411 struct fib6_info *from;
1413 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1414 fib6_info_release(from);
1420 /* exception hash table implementation
1422 static DEFINE_SPINLOCK(rt6_exception_lock);
1424 /* Remove rt6_ex from hash table and free the memory
1425 * Caller must hold rt6_exception_lock
1427 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1428 struct rt6_exception *rt6_ex)
1430 struct fib6_info *from;
1433 if (!bucket || !rt6_ex)
1436 net = dev_net(rt6_ex->rt6i->dst.dev);
1437 net->ipv6.rt6_stats->fib_rt_cache--;
1439 /* purge completely the exception to allow releasing the held resources:
1440 * some [sk] cache may keep the dst around for unlimited time
1442 from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
1443 fib6_info_release(from);
1444 dst_dev_put(&rt6_ex->rt6i->dst);
1446 hlist_del_rcu(&rt6_ex->hlist);
1447 dst_release(&rt6_ex->rt6i->dst);
1448 kfree_rcu(rt6_ex, rcu);
1449 WARN_ON_ONCE(!bucket->depth);
1453 /* Remove oldest rt6_ex in bucket and free the memory
1454 * Caller must hold rt6_exception_lock
1456 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1458 struct rt6_exception *rt6_ex, *oldest = NULL;
1463 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1464 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1467 rt6_remove_exception(bucket, oldest);
1470 static u32 rt6_exception_hash(const struct in6_addr *dst,
1471 const struct in6_addr *src)
1473 static u32 seed __read_mostly;
1476 net_get_random_once(&seed, sizeof(seed));
1477 val = jhash(dst, sizeof(*dst), seed);
1479 #ifdef CONFIG_IPV6_SUBTREES
1481 val = jhash(src, sizeof(*src), val);
1483 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1486 /* Helper function to find the cached rt in the hash table
1487 * and update bucket pointer to point to the bucket for this
1488 * (daddr, saddr) pair
1489 * Caller must hold rt6_exception_lock
1491 static struct rt6_exception *
1492 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1493 const struct in6_addr *daddr,
1494 const struct in6_addr *saddr)
1496 struct rt6_exception *rt6_ex;
1499 if (!(*bucket) || !daddr)
1502 hval = rt6_exception_hash(daddr, saddr);
1505 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1506 struct rt6_info *rt6 = rt6_ex->rt6i;
1507 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1509 #ifdef CONFIG_IPV6_SUBTREES
1510 if (matched && saddr)
1511 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1519 /* Helper function to find the cached rt in the hash table
1520 * and update bucket pointer to point to the bucket for this
1521 * (daddr, saddr) pair
1522 * Caller must hold rcu_read_lock()
1524 static struct rt6_exception *
1525 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1526 const struct in6_addr *daddr,
1527 const struct in6_addr *saddr)
1529 struct rt6_exception *rt6_ex;
1532 WARN_ON_ONCE(!rcu_read_lock_held());
1534 if (!(*bucket) || !daddr)
1537 hval = rt6_exception_hash(daddr, saddr);
1540 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1541 struct rt6_info *rt6 = rt6_ex->rt6i;
1542 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1544 #ifdef CONFIG_IPV6_SUBTREES
1545 if (matched && saddr)
1546 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1554 static unsigned int fib6_mtu(const struct fib6_result *res)
1556 const struct fib6_nh *nh = res->nh;
1559 if (res->f6i->fib6_pmtu) {
1560 mtu = res->f6i->fib6_pmtu;
1562 struct net_device *dev = nh->fib_nh_dev;
1563 struct inet6_dev *idev;
1566 idev = __in6_dev_get(dev);
1567 mtu = idev->cnf.mtu6;
1571 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1573 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1576 #define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL
1578 /* used when the flushed bit is not relevant, only access to the bucket
1579 * (ie., all bucket users except rt6_insert_exception);
1581 * called under rcu lock; sometimes called with rt6_exception_lock held
1584 struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1587 struct rt6_exception_bucket *bucket;
1590 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1591 lockdep_is_held(lock));
1593 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1595 /* remove bucket flushed bit if set */
1597 unsigned long p = (unsigned long)bucket;
1599 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1600 bucket = (struct rt6_exception_bucket *)p;
1606 static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1608 unsigned long p = (unsigned long)bucket;
1610 return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1613 /* called with rt6_exception_lock held */
1614 static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1617 struct rt6_exception_bucket *bucket;
1620 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1621 lockdep_is_held(lock));
1623 p = (unsigned long)bucket;
1624 p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1625 bucket = (struct rt6_exception_bucket *)p;
1626 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1629 static int rt6_insert_exception(struct rt6_info *nrt,
1630 const struct fib6_result *res)
1632 struct net *net = dev_net(nrt->dst.dev);
1633 struct rt6_exception_bucket *bucket;
1634 struct fib6_info *f6i = res->f6i;
1635 struct in6_addr *src_key = NULL;
1636 struct rt6_exception *rt6_ex;
1637 struct fib6_nh *nh = res->nh;
1640 spin_lock_bh(&rt6_exception_lock);
1642 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1643 lockdep_is_held(&rt6_exception_lock));
1645 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1651 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1652 } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1657 #ifdef CONFIG_IPV6_SUBTREES
1658 /* fib6_src.plen != 0 indicates f6i is in subtree
1659 * and exception table is indexed by a hash of
1660 * both fib6_dst and fib6_src.
1661 * Otherwise, the exception table is indexed by
1662 * a hash of only fib6_dst.
1664 if (f6i->fib6_src.plen)
1665 src_key = &nrt->rt6i_src.addr;
1667 /* rt6_mtu_change() might lower mtu on f6i.
1668 * Only insert this exception route if its mtu
1669 * is less than f6i's mtu value.
1671 if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1676 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1679 rt6_remove_exception(bucket, rt6_ex);
1681 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1687 rt6_ex->stamp = jiffies;
1688 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1690 net->ipv6.rt6_stats->fib_rt_cache++;
1692 if (bucket->depth > FIB6_MAX_DEPTH)
1693 rt6_exception_remove_oldest(bucket);
1696 spin_unlock_bh(&rt6_exception_lock);
1698 /* Update fn->fn_sernum to invalidate all cached dst */
1700 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1701 fib6_update_sernum(net, f6i);
1702 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1703 fib6_force_start_gc(net);
1709 static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1711 struct rt6_exception_bucket *bucket;
1712 struct rt6_exception *rt6_ex;
1713 struct hlist_node *tmp;
1716 spin_lock_bh(&rt6_exception_lock);
1718 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1722 /* Prevent rt6_insert_exception() to recreate the bucket list */
1724 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1726 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1727 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1729 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1730 rt6_remove_exception(bucket, rt6_ex);
1732 WARN_ON_ONCE(!from && bucket->depth);
1736 spin_unlock_bh(&rt6_exception_lock);
1739 static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1741 struct fib6_info *f6i = arg;
1743 fib6_nh_flush_exceptions(nh, f6i);
1748 void rt6_flush_exceptions(struct fib6_info *f6i)
1751 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1754 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1757 /* Find cached rt in the hash table inside passed in rt
1758 * Caller has to hold rcu_read_lock()
1760 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1761 const struct in6_addr *daddr,
1762 const struct in6_addr *saddr)
1764 const struct in6_addr *src_key = NULL;
1765 struct rt6_exception_bucket *bucket;
1766 struct rt6_exception *rt6_ex;
1767 struct rt6_info *ret = NULL;
1769 #ifdef CONFIG_IPV6_SUBTREES
1770 /* fib6i_src.plen != 0 indicates f6i is in subtree
1771 * and exception table is indexed by a hash of
1772 * both fib6_dst and fib6_src.
1773 * However, the src addr used to create the hash
1774 * might not be exactly the passed in saddr which
1775 * is a /128 addr from the flow.
1776 * So we need to use f6i->fib6_src to redo lookup
1777 * if the passed in saddr does not find anything.
1778 * (See the logic in ip6_rt_cache_alloc() on how
1779 * rt->rt6i_src is updated.)
1781 if (res->f6i->fib6_src.plen)
1785 bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1786 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1788 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1791 #ifdef CONFIG_IPV6_SUBTREES
1792 /* Use fib6_src as src_key and redo lookup */
1793 if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1794 src_key = &res->f6i->fib6_src.addr;
1802 /* Remove the passed in cached rt from the hash table that contains it */
1803 static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1804 const struct rt6_info *rt)
1806 const struct in6_addr *src_key = NULL;
1807 struct rt6_exception_bucket *bucket;
1808 struct rt6_exception *rt6_ex;
1811 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1814 spin_lock_bh(&rt6_exception_lock);
1815 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1817 #ifdef CONFIG_IPV6_SUBTREES
1818 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1819 * and exception table is indexed by a hash of
1820 * both rt6i_dst and rt6i_src.
1821 * Otherwise, the exception table is indexed by
1822 * a hash of only rt6i_dst.
1825 src_key = &rt->rt6i_src.addr;
1827 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1831 rt6_remove_exception(bucket, rt6_ex);
1837 spin_unlock_bh(&rt6_exception_lock);
1841 struct fib6_nh_excptn_arg {
1842 struct rt6_info *rt;
1846 static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1848 struct fib6_nh_excptn_arg *arg = _arg;
1851 err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1858 static int rt6_remove_exception_rt(struct rt6_info *rt)
1860 struct fib6_info *from;
1862 from = rcu_dereference(rt->from);
1863 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1867 struct fib6_nh_excptn_arg arg = {
1869 .plen = from->fib6_src.plen
1873 /* rc = 1 means an entry was found */
1874 rc = nexthop_for_each_fib6_nh(from->nh,
1875 rt6_nh_remove_exception_rt,
1877 return rc ? 0 : -ENOENT;
1880 return fib6_nh_remove_exception(from->fib6_nh,
1881 from->fib6_src.plen, rt);
1884 /* Find rt6_ex which contains the passed in rt cache and
1887 static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1888 const struct rt6_info *rt)
1890 const struct in6_addr *src_key = NULL;
1891 struct rt6_exception_bucket *bucket;
1892 struct rt6_exception *rt6_ex;
1894 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1895 #ifdef CONFIG_IPV6_SUBTREES
1896 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1897 * and exception table is indexed by a hash of
1898 * both rt6i_dst and rt6i_src.
1899 * Otherwise, the exception table is indexed by
1900 * a hash of only rt6i_dst.
1903 src_key = &rt->rt6i_src.addr;
1905 rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1907 rt6_ex->stamp = jiffies;
1910 struct fib6_nh_match_arg {
1911 const struct net_device *dev;
1912 const struct in6_addr *gw;
1913 struct fib6_nh *match;
1916 /* determine if fib6_nh has given device and gateway */
1917 static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1919 struct fib6_nh_match_arg *arg = _arg;
1921 if (arg->dev != nh->fib_nh_dev ||
1922 (arg->gw && !nh->fib_nh_gw_family) ||
1923 (!arg->gw && nh->fib_nh_gw_family) ||
1924 (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1929 /* found a match, break the loop */
1933 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1935 struct fib6_info *from;
1936 struct fib6_nh *fib6_nh;
1940 from = rcu_dereference(rt->from);
1941 if (!from || !(rt->rt6i_flags & RTF_CACHE))
1945 struct fib6_nh_match_arg arg = {
1947 .gw = &rt->rt6i_gateway,
1950 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1954 fib6_nh = arg.match;
1956 fib6_nh = from->fib6_nh;
1958 fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1963 static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1964 struct rt6_info *rt, int mtu)
1966 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1967 * lowest MTU in the path: always allow updating the route PMTU to
1968 * reflect PMTU decreases.
1970 * If the new MTU is higher, and the route PMTU is equal to the local
1971 * MTU, this means the old MTU is the lowest in the path, so allow
1972 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1976 if (dst_mtu(&rt->dst) >= mtu)
1979 if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
1985 static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
1986 const struct fib6_nh *nh, int mtu)
1988 struct rt6_exception_bucket *bucket;
1989 struct rt6_exception *rt6_ex;
1992 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1996 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1997 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1998 struct rt6_info *entry = rt6_ex->rt6i;
2000 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2001 * route), the metrics of its rt->from have already
2004 if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2005 rt6_mtu_change_route_allowed(idev, entry, mtu))
2006 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2012 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2014 static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2015 const struct in6_addr *gateway)
2017 struct rt6_exception_bucket *bucket;
2018 struct rt6_exception *rt6_ex;
2019 struct hlist_node *tmp;
2022 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2025 spin_lock_bh(&rt6_exception_lock);
2026 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2028 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2029 hlist_for_each_entry_safe(rt6_ex, tmp,
2030 &bucket->chain, hlist) {
2031 struct rt6_info *entry = rt6_ex->rt6i;
2033 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2034 RTF_CACHE_GATEWAY &&
2035 ipv6_addr_equal(gateway,
2036 &entry->rt6i_gateway)) {
2037 rt6_remove_exception(bucket, rt6_ex);
2044 spin_unlock_bh(&rt6_exception_lock);
2047 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2048 struct rt6_exception *rt6_ex,
2049 struct fib6_gc_args *gc_args,
2052 struct rt6_info *rt = rt6_ex->rt6i;
2054 /* we are pruning and obsoleting aged-out and non gateway exceptions
2055 * even if others have still references to them, so that on next
2056 * dst_check() such references can be dropped.
2057 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2058 * expired, independently from their aging, as per RFC 8201 section 4
2060 if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2061 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2062 RT6_TRACE("aging clone %p\n", rt);
2063 rt6_remove_exception(bucket, rt6_ex);
2066 } else if (time_after(jiffies, rt->dst.expires)) {
2067 RT6_TRACE("purging expired route %p\n", rt);
2068 rt6_remove_exception(bucket, rt6_ex);
2072 if (rt->rt6i_flags & RTF_GATEWAY) {
2073 struct neighbour *neigh;
2074 __u8 neigh_flags = 0;
2076 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2078 neigh_flags = neigh->flags;
2080 if (!(neigh_flags & NTF_ROUTER)) {
2081 RT6_TRACE("purging route %p via non-router but gateway\n",
2083 rt6_remove_exception(bucket, rt6_ex);
2091 static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2092 struct fib6_gc_args *gc_args,
2095 struct rt6_exception_bucket *bucket;
2096 struct rt6_exception *rt6_ex;
2097 struct hlist_node *tmp;
2100 if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2104 spin_lock(&rt6_exception_lock);
2105 bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2107 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2108 hlist_for_each_entry_safe(rt6_ex, tmp,
2109 &bucket->chain, hlist) {
2110 rt6_age_examine_exception(bucket, rt6_ex,
2116 spin_unlock(&rt6_exception_lock);
2117 rcu_read_unlock_bh();
2120 struct fib6_nh_age_excptn_arg {
2121 struct fib6_gc_args *gc_args;
2125 static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2127 struct fib6_nh_age_excptn_arg *arg = _arg;
2129 fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2133 void rt6_age_exceptions(struct fib6_info *f6i,
2134 struct fib6_gc_args *gc_args,
2138 struct fib6_nh_age_excptn_arg arg = {
2143 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2146 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2150 /* must be called with rcu lock held */
2151 int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2152 struct flowi6 *fl6, struct fib6_result *res, int strict)
2154 struct fib6_node *fn, *saved_fn;
2156 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2159 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2163 rt6_select(net, fn, oif, res, strict);
2164 if (res->f6i == net->ipv6.fib6_null_entry) {
2165 fn = fib6_backtrack(fn, &fl6->saddr);
2167 goto redo_rt6_select;
2168 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2169 /* also consider unreachable route */
2170 strict &= ~RT6_LOOKUP_F_REACHABLE;
2172 goto redo_rt6_select;
2176 trace_fib6_table_lookup(net, res, table, fl6);
2181 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2182 int oif, struct flowi6 *fl6,
2183 const struct sk_buff *skb, int flags)
2185 struct fib6_result res = {};
2186 struct rt6_info *rt = NULL;
2189 WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2190 !rcu_read_lock_held());
2192 strict |= flags & RT6_LOOKUP_F_IFACE;
2193 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2194 if (net->ipv6.devconf_all->forwarding == 0)
2195 strict |= RT6_LOOKUP_F_REACHABLE;
2199 fib6_table_lookup(net, table, oif, fl6, &res, strict);
2200 if (res.f6i == net->ipv6.fib6_null_entry)
2203 fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2205 /*Search through exception table */
2206 rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2209 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2210 !res.nh->fib_nh_gw_family)) {
2211 /* Create a RTF_CACHE clone which will not be
2212 * owned by the fib6 tree. It is for the special case where
2213 * the daddr in the skb during the neighbor look-up is different
2214 * from the fl6->daddr used to look-up route here.
2216 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2219 /* 1 refcnt is taken during ip6_rt_cache_alloc().
2220 * As rt6_uncached_list_add() does not consume refcnt,
2221 * this refcnt is always returned to the caller even
2222 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2224 rt6_uncached_list_add(rt);
2225 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
2231 /* Get a percpu copy */
2233 rt = rt6_get_pcpu_route(&res);
2236 rt = rt6_make_pcpu_route(net, &res);
2242 rt = net->ipv6.ip6_null_entry;
2243 if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2244 ip6_hold_safe(net, &rt);
2249 EXPORT_SYMBOL_GPL(ip6_pol_route);
2251 static struct rt6_info *ip6_pol_route_input(struct net *net,
2252 struct fib6_table *table,
2254 const struct sk_buff *skb,
2257 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2260 struct dst_entry *ip6_route_input_lookup(struct net *net,
2261 struct net_device *dev,
2263 const struct sk_buff *skb,
2266 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2267 flags |= RT6_LOOKUP_F_IFACE;
2269 return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2271 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2273 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2274 struct flow_keys *keys,
2275 struct flow_keys *flkeys)
2277 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2278 const struct ipv6hdr *key_iph = outer_iph;
2279 struct flow_keys *_flkeys = flkeys;
2280 const struct ipv6hdr *inner_iph;
2281 const struct icmp6hdr *icmph;
2282 struct ipv6hdr _inner_iph;
2283 struct icmp6hdr _icmph;
2285 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2288 icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2289 sizeof(_icmph), &_icmph);
2293 if (icmph->icmp6_type != ICMPV6_DEST_UNREACH &&
2294 icmph->icmp6_type != ICMPV6_PKT_TOOBIG &&
2295 icmph->icmp6_type != ICMPV6_TIME_EXCEED &&
2296 icmph->icmp6_type != ICMPV6_PARAMPROB)
2299 inner_iph = skb_header_pointer(skb,
2300 skb_transport_offset(skb) + sizeof(*icmph),
2301 sizeof(_inner_iph), &_inner_iph);
2305 key_iph = inner_iph;
2309 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2310 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2311 keys->tags.flow_label = _flkeys->tags.flow_label;
2312 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2314 keys->addrs.v6addrs.src = key_iph->saddr;
2315 keys->addrs.v6addrs.dst = key_iph->daddr;
2316 keys->tags.flow_label = ip6_flowlabel(key_iph);
2317 keys->basic.ip_proto = key_iph->nexthdr;
2321 /* if skb is set it will be used and fl6 can be NULL */
2322 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2323 const struct sk_buff *skb, struct flow_keys *flkeys)
2325 struct flow_keys hash_keys;
2328 switch (ip6_multipath_hash_policy(net)) {
2330 memset(&hash_keys, 0, sizeof(hash_keys));
2331 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2333 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2335 hash_keys.addrs.v6addrs.src = fl6->saddr;
2336 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2337 hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2338 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2343 unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2344 struct flow_keys keys;
2346 /* short-circuit if we already have L4 hash present */
2348 return skb_get_hash_raw(skb) >> 1;
2350 memset(&hash_keys, 0, sizeof(hash_keys));
2353 skb_flow_dissect_flow_keys(skb, &keys, flag);
2356 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2357 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2358 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2359 hash_keys.ports.src = flkeys->ports.src;
2360 hash_keys.ports.dst = flkeys->ports.dst;
2361 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2363 memset(&hash_keys, 0, sizeof(hash_keys));
2364 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2365 hash_keys.addrs.v6addrs.src = fl6->saddr;
2366 hash_keys.addrs.v6addrs.dst = fl6->daddr;
2367 hash_keys.ports.src = fl6->fl6_sport;
2368 hash_keys.ports.dst = fl6->fl6_dport;
2369 hash_keys.basic.ip_proto = fl6->flowi6_proto;
2373 mhash = flow_hash_from_keys(&hash_keys);
2378 /* Called with rcu held */
2379 void ip6_route_input(struct sk_buff *skb)
2381 const struct ipv6hdr *iph = ipv6_hdr(skb);
2382 struct net *net = dev_net(skb->dev);
2383 int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2384 struct ip_tunnel_info *tun_info;
2385 struct flowi6 fl6 = {
2386 .flowi6_iif = skb->dev->ifindex,
2387 .daddr = iph->daddr,
2388 .saddr = iph->saddr,
2389 .flowlabel = ip6_flowinfo(iph),
2390 .flowi6_mark = skb->mark,
2391 .flowi6_proto = iph->nexthdr,
2393 struct flow_keys *flkeys = NULL, _flkeys;
2395 tun_info = skb_tunnel_info(skb);
2396 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2397 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2399 if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2402 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2403 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2405 skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2409 static struct rt6_info *ip6_pol_route_output(struct net *net,
2410 struct fib6_table *table,
2412 const struct sk_buff *skb,
2415 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2418 struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2419 const struct sock *sk,
2420 struct flowi6 *fl6, int flags)
2424 if (ipv6_addr_type(&fl6->daddr) &
2425 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2426 struct dst_entry *dst;
2428 /* This function does not take refcnt on the dst */
2429 dst = l3mdev_link_scope_lookup(net, fl6);
2434 fl6->flowi6_iif = LOOPBACK_IFINDEX;
2436 flags |= RT6_LOOKUP_F_DST_NOREF;
2437 any_src = ipv6_addr_any(&fl6->saddr);
2438 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2439 (fl6->flowi6_oif && any_src))
2440 flags |= RT6_LOOKUP_F_IFACE;
2443 flags |= RT6_LOOKUP_F_HAS_SADDR;
2445 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
2447 return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2449 EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2451 struct dst_entry *ip6_route_output_flags(struct net *net,
2452 const struct sock *sk,
2456 struct dst_entry *dst;
2457 struct rt6_info *rt6;
2460 dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2461 rt6 = (struct rt6_info *)dst;
2462 /* For dst cached in uncached_list, refcnt is already taken. */
2463 if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2464 dst = &net->ipv6.ip6_null_entry->dst;
2471 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2473 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2475 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2476 struct net_device *loopback_dev = net->loopback_dev;
2477 struct dst_entry *new = NULL;
2479 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
2480 DST_OBSOLETE_DEAD, 0);
2483 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2487 new->input = dst_discard;
2488 new->output = dst_discard_out;
2490 dst_copy_metrics(new, &ort->dst);
2492 rt->rt6i_idev = in6_dev_get(loopback_dev);
2493 rt->rt6i_gateway = ort->rt6i_gateway;
2494 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2496 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2497 #ifdef CONFIG_IPV6_SUBTREES
2498 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2502 dst_release(dst_orig);
2503 return new ? new : ERR_PTR(-ENOMEM);
2507 * Destination cache support functions
2510 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2514 if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2517 if (fib6_check_expired(f6i))
2523 static struct dst_entry *rt6_check(struct rt6_info *rt,
2524 struct fib6_info *from,
2529 if ((from && !fib6_get_cookie_safe(from, &rt_cookie)) ||
2530 rt_cookie != cookie)
2533 if (rt6_check_expired(rt))
2539 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2540 struct fib6_info *from,
2543 if (!__rt6_check_expired(rt) &&
2544 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2545 fib6_check(from, cookie))
2551 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
2553 struct dst_entry *dst_ret;
2554 struct fib6_info *from;
2555 struct rt6_info *rt;
2557 rt = container_of(dst, struct rt6_info, dst);
2561 /* All IPV6 dsts are created with ->obsolete set to the value
2562 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2563 * into this function always.
2566 from = rcu_dereference(rt->from);
2568 if (from && (rt->rt6i_flags & RTF_PCPU ||
2569 unlikely(!list_empty(&rt->rt6i_uncached))))
2570 dst_ret = rt6_dst_from_check(rt, from, cookie);
2572 dst_ret = rt6_check(rt, from, cookie);
2579 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2581 struct rt6_info *rt = (struct rt6_info *) dst;
2584 if (rt->rt6i_flags & RTF_CACHE) {
2586 if (rt6_check_expired(rt)) {
2587 rt6_remove_exception_rt(rt);
2599 static void ip6_link_failure(struct sk_buff *skb)
2601 struct rt6_info *rt;
2603 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2605 rt = (struct rt6_info *) skb_dst(skb);
2608 if (rt->rt6i_flags & RTF_CACHE) {
2609 rt6_remove_exception_rt(rt);
2611 struct fib6_info *from;
2612 struct fib6_node *fn;
2614 from = rcu_dereference(rt->from);
2616 fn = rcu_dereference(from->fib6_node);
2617 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2625 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2627 if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2628 struct fib6_info *from;
2631 from = rcu_dereference(rt0->from);
2633 rt0->dst.expires = from->expires;
2637 dst_set_expires(&rt0->dst, timeout);
2638 rt0->rt6i_flags |= RTF_EXPIRES;
2641 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2643 struct net *net = dev_net(rt->dst.dev);
2645 dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2646 rt->rt6i_flags |= RTF_MODIFIED;
2647 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2650 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2652 return !(rt->rt6i_flags & RTF_CACHE) &&
2653 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2656 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2657 const struct ipv6hdr *iph, u32 mtu)
2659 const struct in6_addr *daddr, *saddr;
2660 struct rt6_info *rt6 = (struct rt6_info *)dst;
2662 if (dst_metric_locked(dst, RTAX_MTU))
2666 daddr = &iph->daddr;
2667 saddr = &iph->saddr;
2669 daddr = &sk->sk_v6_daddr;
2670 saddr = &inet6_sk(sk)->saddr;
2675 dst_confirm_neigh(dst, daddr);
2676 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2677 if (mtu >= dst_mtu(dst))
2680 if (!rt6_cache_allowed_for_pmtu(rt6)) {
2681 rt6_do_update_pmtu(rt6, mtu);
2682 /* update rt6_ex->stamp for cache */
2683 if (rt6->rt6i_flags & RTF_CACHE)
2684 rt6_update_exception_stamp_rt(rt6);
2686 struct fib6_result res = {};
2687 struct rt6_info *nrt6;
2690 res.f6i = rcu_dereference(rt6->from);
2695 res.fib6_flags = res.f6i->fib6_flags;
2696 res.fib6_type = res.f6i->fib6_type;
2699 struct fib6_nh_match_arg arg = {
2701 .gw = &rt6->rt6i_gateway,
2704 nexthop_for_each_fib6_nh(res.f6i->nh,
2705 fib6_nh_find_match, &arg);
2707 /* fib6_info uses a nexthop that does not have fib6_nh
2708 * using the dst->dev + gw. Should be impossible.
2717 res.nh = res.f6i->fib6_nh;
2720 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2722 rt6_do_update_pmtu(nrt6, mtu);
2723 if (rt6_insert_exception(nrt6, &res))
2724 dst_release_immediate(&nrt6->dst);
2730 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2731 struct sk_buff *skb, u32 mtu)
2733 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
2736 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2737 int oif, u32 mark, kuid_t uid)
2739 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2740 struct dst_entry *dst;
2741 struct flowi6 fl6 = {
2743 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2744 .daddr = iph->daddr,
2745 .saddr = iph->saddr,
2746 .flowlabel = ip6_flowinfo(iph),
2750 dst = ip6_route_output(net, NULL, &fl6);
2752 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
2755 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2757 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2759 int oif = sk->sk_bound_dev_if;
2760 struct dst_entry *dst;
2762 if (!oif && skb->dev)
2763 oif = l3mdev_master_ifindex(skb->dev);
2765 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
2767 dst = __sk_dst_get(sk);
2768 if (!dst || !dst->obsolete ||
2769 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2773 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2774 ip6_datagram_dst_update(sk, false);
2777 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2779 void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2780 const struct flowi6 *fl6)
2782 #ifdef CONFIG_IPV6_SUBTREES
2783 struct ipv6_pinfo *np = inet6_sk(sk);
2786 ip6_dst_store(sk, dst,
2787 ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2788 &sk->sk_v6_daddr : NULL,
2789 #ifdef CONFIG_IPV6_SUBTREES
2790 ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2796 static bool ip6_redirect_nh_match(const struct fib6_result *res,
2798 const struct in6_addr *gw,
2799 struct rt6_info **ret)
2801 const struct fib6_nh *nh = res->nh;
2803 if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2804 fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2807 /* rt_cache's gateway might be different from its 'parent'
2808 * in the case of an ip redirect.
2809 * So we keep searching in the exception table if the gateway
2812 if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2813 struct rt6_info *rt_cache;
2815 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
2817 ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
2826 struct fib6_nh_rd_arg {
2827 struct fib6_result *res;
2829 const struct in6_addr *gw;
2830 struct rt6_info **ret;
2833 static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
2835 struct fib6_nh_rd_arg *arg = _arg;
2838 return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
2841 /* Handle redirects */
2842 struct ip6rd_flowi {
2844 struct in6_addr gateway;
2847 static struct rt6_info *__ip6_route_redirect(struct net *net,
2848 struct fib6_table *table,
2850 const struct sk_buff *skb,
2853 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
2854 struct rt6_info *ret = NULL;
2855 struct fib6_result res = {};
2856 struct fib6_nh_rd_arg arg = {
2859 .gw = &rdfl->gateway,
2862 struct fib6_info *rt;
2863 struct fib6_node *fn;
2865 /* l3mdev_update_flow overrides oif if the device is enslaved; in
2866 * this case we must match on the real ingress device, so reset it
2868 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2869 fl6->flowi6_oif = skb->dev->ifindex;
2871 /* Get the "current" route for this destination and
2872 * check if the redirect has come from appropriate router.
2874 * RFC 4861 specifies that redirects should only be
2875 * accepted if they come from the nexthop to the target.
2876 * Due to the way the routes are chosen, this notion
2877 * is a bit fuzzy and one might need to check all possible
2882 fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2884 for_each_fib6_node_rt_rcu(fn) {
2886 if (fib6_check_expired(rt))
2888 if (rt->fib6_flags & RTF_REJECT)
2890 if (unlikely(rt->nh)) {
2891 if (nexthop_is_blackhole(rt->nh))
2893 /* on match, res->nh is filled in and potentially ret */
2894 if (nexthop_for_each_fib6_nh(rt->nh,
2895 fib6_nh_redirect_match,
2899 res.nh = rt->fib6_nh;
2900 if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
2907 rt = net->ipv6.fib6_null_entry;
2908 else if (rt->fib6_flags & RTF_REJECT) {
2909 ret = net->ipv6.ip6_null_entry;
2913 if (rt == net->ipv6.fib6_null_entry) {
2914 fn = fib6_backtrack(fn, &fl6->saddr);
2920 res.nh = rt->fib6_nh;
2923 ip6_hold_safe(net, &ret);
2925 res.fib6_flags = res.f6i->fib6_flags;
2926 res.fib6_type = res.f6i->fib6_type;
2927 ret = ip6_create_rt_rcu(&res);
2932 trace_fib6_table_lookup(net, &res, table, fl6);
2936 static struct dst_entry *ip6_route_redirect(struct net *net,
2937 const struct flowi6 *fl6,
2938 const struct sk_buff *skb,
2939 const struct in6_addr *gateway)
2941 int flags = RT6_LOOKUP_F_HAS_SADDR;
2942 struct ip6rd_flowi rdfl;
2945 rdfl.gateway = *gateway;
2947 return fib6_rule_lookup(net, &rdfl.fl6, skb,
2948 flags, __ip6_route_redirect);
2951 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
2954 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2955 struct dst_entry *dst;
2956 struct flowi6 fl6 = {
2957 .flowi6_iif = LOOPBACK_IFINDEX,
2959 .flowi6_mark = mark,
2960 .daddr = iph->daddr,
2961 .saddr = iph->saddr,
2962 .flowlabel = ip6_flowinfo(iph),
2966 dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
2967 rt6_do_redirect(dst, NULL, skb);
2970 EXPORT_SYMBOL_GPL(ip6_redirect);
2972 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
2974 const struct ipv6hdr *iph = ipv6_hdr(skb);
2975 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
2976 struct dst_entry *dst;
2977 struct flowi6 fl6 = {
2978 .flowi6_iif = LOOPBACK_IFINDEX,
2981 .saddr = iph->daddr,
2982 .flowi6_uid = sock_net_uid(net, NULL),
2985 dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
2986 rt6_do_redirect(dst, NULL, skb);
2990 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
2992 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
2995 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
2997 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
2999 struct net_device *dev = dst->dev;
3000 unsigned int mtu = dst_mtu(dst);
3001 struct net *net = dev_net(dev);
3003 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3005 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3006 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3009 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3010 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3011 * IPV6_MAXPLEN is also valid and means: "any MSS,
3012 * rely only on pmtu discovery"
3014 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3019 static unsigned int ip6_mtu(const struct dst_entry *dst)
3021 struct inet6_dev *idev;
3024 mtu = dst_metric_raw(dst, RTAX_MTU);
3031 idev = __in6_dev_get(dst->dev);
3033 mtu = idev->cnf.mtu6;
3037 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3039 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
3043 * 1. mtu on route is locked - use it
3044 * 2. mtu from nexthop exception
3045 * 3. mtu from egress device
3047 * based on ip6_dst_mtu_forward and exception logic of
3048 * rt6_find_cached_rt; called with rcu_read_lock
3050 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3051 const struct in6_addr *daddr,
3052 const struct in6_addr *saddr)
3054 const struct fib6_nh *nh = res->nh;
3055 struct fib6_info *f6i = res->f6i;
3056 struct inet6_dev *idev;
3057 struct rt6_info *rt;
3060 if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3061 mtu = f6i->fib6_pmtu;
3066 rt = rt6_find_cached_rt(res, daddr, saddr);
3068 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3070 struct net_device *dev = nh->fib_nh_dev;
3073 idev = __in6_dev_get(dev);
3074 if (idev && idev->cnf.mtu6 > mtu)
3075 mtu = idev->cnf.mtu6;
3078 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3080 return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3083 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3086 struct dst_entry *dst;
3087 struct rt6_info *rt;
3088 struct inet6_dev *idev = in6_dev_get(dev);
3089 struct net *net = dev_net(dev);
3091 if (unlikely(!idev))
3092 return ERR_PTR(-ENODEV);
3094 rt = ip6_dst_alloc(net, dev, 0);
3095 if (unlikely(!rt)) {
3097 dst = ERR_PTR(-ENOMEM);
3101 rt->dst.flags |= DST_HOST;
3102 rt->dst.input = ip6_input;
3103 rt->dst.output = ip6_output;
3104 rt->rt6i_gateway = fl6->daddr;
3105 rt->rt6i_dst.addr = fl6->daddr;
3106 rt->rt6i_dst.plen = 128;
3107 rt->rt6i_idev = idev;
3108 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3110 /* Add this dst into uncached_list so that rt6_disable_ip() can
3111 * do proper release of the net_device
3113 rt6_uncached_list_add(rt);
3114 atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
3116 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3122 static int ip6_dst_gc(struct dst_ops *ops)
3124 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3125 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3126 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3127 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3128 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3129 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3132 entries = dst_entries_get_fast(ops);
3133 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
3134 entries <= rt_max_size)
3137 net->ipv6.ip6_rt_gc_expire++;
3138 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
3139 entries = dst_entries_get_slow(ops);
3140 if (entries < ops->gc_thresh)
3141 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
3143 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
3144 return entries > rt_max_size;
3147 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
3148 struct fib6_config *cfg,
3149 const struct in6_addr *gw_addr,
3150 u32 tbid, int flags)
3152 struct flowi6 fl6 = {
3153 .flowi6_oif = cfg->fc_ifindex,
3155 .saddr = cfg->fc_prefsrc,
3157 struct fib6_table *table;
3158 struct rt6_info *rt;
3160 table = fib6_get_table(net, tbid);
3164 if (!ipv6_addr_any(&cfg->fc_prefsrc))
3165 flags |= RT6_LOOKUP_F_HAS_SADDR;
3167 flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3168 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, NULL, flags);
3170 /* if table lookup failed, fall back to full lookup */
3171 if (rt == net->ipv6.ip6_null_entry) {
3179 static int ip6_route_check_nh_onlink(struct net *net,
3180 struct fib6_config *cfg,
3181 const struct net_device *dev,
3182 struct netlink_ext_ack *extack)
3184 u32 tbid = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN;
3185 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3186 u32 flags = RTF_LOCAL | RTF_ANYCAST | RTF_REJECT;
3187 struct fib6_info *from;
3188 struct rt6_info *grt;
3192 grt = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0);
3195 from = rcu_dereference(grt->from);
3196 if (!grt->dst.error &&
3197 /* ignore match if it is the default route */
3198 from && !ipv6_addr_any(&from->fib6_dst.addr) &&
3199 (grt->rt6i_flags & flags || dev != grt->dst.dev)) {
3200 NL_SET_ERR_MSG(extack,
3201 "Nexthop has invalid gateway or device mismatch");
3212 static int ip6_route_check_nh(struct net *net,
3213 struct fib6_config *cfg,
3214 struct net_device **_dev,
3215 struct inet6_dev **idev)
3217 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3218 struct net_device *dev = _dev ? *_dev : NULL;
3219 struct rt6_info *grt = NULL;
3220 int err = -EHOSTUNREACH;
3222 if (cfg->fc_table) {
3223 int flags = RT6_LOOKUP_F_IFACE;
3225 grt = ip6_nh_lookup_table(net, cfg, gw_addr,
3226 cfg->fc_table, flags);
3228 if (grt->rt6i_flags & RTF_GATEWAY ||
3229 (dev && dev != grt->dst.dev)) {
3237 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, NULL, 1);
3243 if (dev != grt->dst.dev) {
3248 *_dev = dev = grt->dst.dev;
3249 *idev = grt->rt6i_idev;
3251 in6_dev_hold(grt->rt6i_idev);
3254 if (!(grt->rt6i_flags & RTF_GATEWAY))
3263 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3264 struct net_device **_dev, struct inet6_dev **idev,
3265 struct netlink_ext_ack *extack)
3267 const struct in6_addr *gw_addr = &cfg->fc_gateway;
3268 int gwa_type = ipv6_addr_type(gw_addr);
3269 bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3270 const struct net_device *dev = *_dev;
3271 bool need_addr_check = !dev;
3274 /* if gw_addr is local we will fail to detect this in case
3275 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3276 * will return already-added prefix route via interface that
3277 * prefix route was assigned to, which might be non-loopback.
3280 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3281 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3285 if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3286 /* IPv6 strictly inhibits using not link-local
3287 * addresses as nexthop address.
3288 * Otherwise, router will not able to send redirects.
3289 * It is very good, but in some (rare!) circumstances
3290 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3291 * some exceptions. --ANK
3292 * We allow IPv4-mapped nexthops to support RFC4798-type
3295 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3296 NL_SET_ERR_MSG(extack, "Invalid gateway address");
3300 if (cfg->fc_flags & RTNH_F_ONLINK)
3301 err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3303 err = ip6_route_check_nh(net, cfg, _dev, idev);
3309 /* reload in case device was changed */
3314 NL_SET_ERR_MSG(extack, "Egress device not specified");
3316 } else if (dev->flags & IFF_LOOPBACK) {
3317 NL_SET_ERR_MSG(extack,
3318 "Egress device can not be loopback device for this route");
3322 /* if we did not check gw_addr above, do so now that the
3323 * egress device has been resolved.
3325 if (need_addr_check &&
3326 ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3327 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3336 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3338 if ((flags & RTF_REJECT) ||
3339 (dev && (dev->flags & IFF_LOOPBACK) &&
3340 !(addr_type & IPV6_ADDR_LOOPBACK) &&
3341 !(flags & RTF_LOCAL)))
3347 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3348 struct fib6_config *cfg, gfp_t gfp_flags,
3349 struct netlink_ext_ack *extack)
3351 struct net_device *dev = NULL;
3352 struct inet6_dev *idev = NULL;
3356 fib6_nh->fib_nh_family = AF_INET6;
3359 if (cfg->fc_ifindex) {
3360 dev = dev_get_by_index(net, cfg->fc_ifindex);
3363 idev = in6_dev_get(dev);
3368 if (cfg->fc_flags & RTNH_F_ONLINK) {
3370 NL_SET_ERR_MSG(extack,
3371 "Nexthop device required for onlink");
3375 if (!(dev->flags & IFF_UP)) {
3376 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3381 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3384 fib6_nh->fib_nh_weight = 1;
3386 /* We cannot add true routes via loopback here,
3387 * they would result in kernel looping; promote them to reject routes
3389 addr_type = ipv6_addr_type(&cfg->fc_dst);
3390 if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3391 /* hold loopback dev/idev if we haven't done so. */
3392 if (dev != net->loopback_dev) {
3397 dev = net->loopback_dev;
3399 idev = in6_dev_get(dev);
3408 if (cfg->fc_flags & RTF_GATEWAY) {
3409 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3413 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3414 fib6_nh->fib_nh_gw_family = AF_INET6;
3421 if (idev->cnf.disable_ipv6) {
3422 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3427 if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3428 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3433 if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3434 !netif_carrier_ok(dev))
3435 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3437 err = fib_nh_common_init(&fib6_nh->nh_common, cfg->fc_encap,
3438 cfg->fc_encap_type, cfg, gfp_flags, extack);
3443 fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3444 if (!fib6_nh->rt6i_pcpu) {
3449 fib6_nh->fib_nh_dev = dev;
3450 fib6_nh->fib_nh_oif = dev->ifindex;
3457 lwtstate_put(fib6_nh->fib_nh_lws);
3458 fib6_nh->fib_nh_lws = NULL;
3466 void fib6_nh_release(struct fib6_nh *fib6_nh)
3468 struct rt6_exception_bucket *bucket;
3472 fib6_nh_flush_exceptions(fib6_nh, NULL);
3473 bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3475 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3481 if (fib6_nh->rt6i_pcpu) {
3484 for_each_possible_cpu(cpu) {
3485 struct rt6_info **ppcpu_rt;
3486 struct rt6_info *pcpu_rt;
3488 ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3489 pcpu_rt = *ppcpu_rt;
3491 dst_dev_put(&pcpu_rt->dst);
3492 dst_release(&pcpu_rt->dst);
3497 free_percpu(fib6_nh->rt6i_pcpu);
3500 fib_nh_common_release(&fib6_nh->nh_common);
3503 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3505 struct netlink_ext_ack *extack)
3507 struct net *net = cfg->fc_nlinfo.nl_net;
3508 struct fib6_info *rt = NULL;
3509 struct nexthop *nh = NULL;
3510 struct fib6_table *table;
3511 struct fib6_nh *fib6_nh;
3515 /* RTF_PCPU is an internal flag; can not be set by userspace */
3516 if (cfg->fc_flags & RTF_PCPU) {
3517 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3521 /* RTF_CACHE is an internal flag; can not be set by userspace */
3522 if (cfg->fc_flags & RTF_CACHE) {
3523 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3527 if (cfg->fc_type > RTN_MAX) {
3528 NL_SET_ERR_MSG(extack, "Invalid route type");
3532 if (cfg->fc_dst_len > 128) {
3533 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3536 if (cfg->fc_src_len > 128) {
3537 NL_SET_ERR_MSG(extack, "Invalid source address length");
3540 #ifndef CONFIG_IPV6_SUBTREES
3541 if (cfg->fc_src_len) {
3542 NL_SET_ERR_MSG(extack,
3543 "Specifying source address requires IPV6_SUBTREES to be enabled");
3547 if (cfg->fc_nh_id) {
3548 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3550 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3553 err = fib6_check_nexthop(nh, cfg, extack);
3559 if (cfg->fc_nlinfo.nlh &&
3560 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3561 table = fib6_get_table(net, cfg->fc_table);
3563 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3564 table = fib6_new_table(net, cfg->fc_table);
3567 table = fib6_new_table(net, cfg->fc_table);
3574 rt = fib6_info_alloc(gfp_flags, !nh);
3578 rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3580 if (IS_ERR(rt->fib6_metrics)) {
3581 err = PTR_ERR(rt->fib6_metrics);
3582 /* Do not leave garbage there. */
3583 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3587 if (cfg->fc_flags & RTF_ADDRCONF)
3588 rt->dst_nocount = true;
3590 if (cfg->fc_flags & RTF_EXPIRES)
3591 fib6_set_expires(rt, jiffies +
3592 clock_t_to_jiffies(cfg->fc_expires));
3594 fib6_clean_expires(rt);
3596 if (cfg->fc_protocol == RTPROT_UNSPEC)
3597 cfg->fc_protocol = RTPROT_BOOT;
3598 rt->fib6_protocol = cfg->fc_protocol;
3600 rt->fib6_table = table;
3601 rt->fib6_metric = cfg->fc_metric;
3602 rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3603 rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3605 ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3606 rt->fib6_dst.plen = cfg->fc_dst_len;
3607 if (rt->fib6_dst.plen == 128)
3608 rt->dst_host = true;
3610 #ifdef CONFIG_IPV6_SUBTREES
3611 ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3612 rt->fib6_src.plen = cfg->fc_src_len;
3615 if (!nexthop_get(nh)) {
3616 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3619 if (rt->fib6_src.plen) {
3620 NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3624 fib6_nh = nexthop_fib6_nh(rt->nh);
3626 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3630 fib6_nh = rt->fib6_nh;
3632 /* We cannot add true routes via loopback here, they would
3633 * result in kernel looping; promote them to reject routes
3635 addr_type = ipv6_addr_type(&cfg->fc_dst);
3636 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3638 rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3641 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3642 struct net_device *dev = fib6_nh->fib_nh_dev;
3644 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3645 NL_SET_ERR_MSG(extack, "Invalid source address");
3649 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3650 rt->fib6_prefsrc.plen = 128;
3652 rt->fib6_prefsrc.plen = 0;
3656 fib6_info_release(rt);
3657 return ERR_PTR(err);
3660 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3661 struct netlink_ext_ack *extack)
3663 struct fib6_info *rt;
3666 rt = ip6_route_info_create(cfg, gfp_flags, extack);
3670 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3671 fib6_info_release(rt);
3676 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3678 struct net *net = info->nl_net;
3679 struct fib6_table *table;
3682 if (rt == net->ipv6.fib6_null_entry) {
3687 table = rt->fib6_table;
3688 spin_lock_bh(&table->tb6_lock);
3689 err = fib6_del(rt, info);
3690 spin_unlock_bh(&table->tb6_lock);
3693 fib6_info_release(rt);
3697 int ip6_del_rt(struct net *net, struct fib6_info *rt)
3699 struct nl_info info = { .nl_net = net };
3701 return __ip6_del_rt(rt, &info);
3704 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3706 struct nl_info *info = &cfg->fc_nlinfo;
3707 struct net *net = info->nl_net;
3708 struct sk_buff *skb = NULL;
3709 struct fib6_table *table;
3712 if (rt == net->ipv6.fib6_null_entry)
3714 table = rt->fib6_table;
3715 spin_lock_bh(&table->tb6_lock);
3717 if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3718 struct fib6_info *sibling, *next_sibling;
3720 /* prefer to send a single notification with all hops */
3721 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3723 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3725 if (rt6_fill_node(net, skb, rt, NULL,
3726 NULL, NULL, 0, RTM_DELROUTE,
3727 info->portid, seq, 0) < 0) {
3731 info->skip_notify = 1;
3734 info->skip_notify_kernel = 1;
3735 call_fib6_multipath_entry_notifiers(net,
3736 FIB_EVENT_ENTRY_DEL,
3740 list_for_each_entry_safe(sibling, next_sibling,
3743 err = fib6_del(sibling, info);
3749 err = fib6_del(rt, info);
3751 spin_unlock_bh(&table->tb6_lock);
3753 fib6_info_release(rt);
3756 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3757 info->nlh, gfp_any());
3762 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3766 if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3769 if (cfg->fc_flags & RTF_GATEWAY &&
3770 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3773 rc = rt6_remove_exception_rt(rt);
3778 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3781 struct fib6_result res = {
3785 struct rt6_info *rt_cache;
3787 rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3789 return __ip6_del_cached_rt(rt_cache, cfg);
3794 struct fib6_nh_del_cached_rt_arg {
3795 struct fib6_config *cfg;
3796 struct fib6_info *f6i;
3799 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
3801 struct fib6_nh_del_cached_rt_arg *arg = _arg;
3804 rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
3805 return rc != -ESRCH ? rc : 0;
3808 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
3810 struct fib6_nh_del_cached_rt_arg arg = {
3815 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
3818 static int ip6_route_del(struct fib6_config *cfg,
3819 struct netlink_ext_ack *extack)
3821 struct fib6_table *table;
3822 struct fib6_info *rt;
3823 struct fib6_node *fn;
3826 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
3828 NL_SET_ERR_MSG(extack, "FIB table does not exist");
3834 fn = fib6_locate(&table->tb6_root,
3835 &cfg->fc_dst, cfg->fc_dst_len,
3836 &cfg->fc_src, cfg->fc_src_len,
3837 !(cfg->fc_flags & RTF_CACHE));
3840 for_each_fib6_node_rt_rcu(fn) {
3843 if (rt->nh && cfg->fc_nh_id &&
3844 rt->nh->id != cfg->fc_nh_id)
3847 if (cfg->fc_flags & RTF_CACHE) {
3851 rc = ip6_del_cached_rt_nh(cfg, rt);
3852 } else if (cfg->fc_nh_id) {
3856 rc = ip6_del_cached_rt(cfg, rt, nh);
3865 if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
3867 if (cfg->fc_protocol &&
3868 cfg->fc_protocol != rt->fib6_protocol)
3872 if (!fib6_info_hold_safe(rt))
3876 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3882 if (cfg->fc_ifindex &&
3884 nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
3886 if (cfg->fc_flags & RTF_GATEWAY &&
3887 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
3889 if (!fib6_info_hold_safe(rt))
3893 /* if gateway was specified only delete the one hop */
3894 if (cfg->fc_flags & RTF_GATEWAY)
3895 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3897 return __ip6_del_rt_siblings(rt, cfg);
3905 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
3907 struct netevent_redirect netevent;
3908 struct rt6_info *rt, *nrt = NULL;
3909 struct fib6_result res = {};
3910 struct ndisc_options ndopts;
3911 struct inet6_dev *in6_dev;
3912 struct neighbour *neigh;
3914 int optlen, on_link;
3917 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
3918 optlen -= sizeof(*msg);
3921 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
3925 msg = (struct rd_msg *)icmp6_hdr(skb);
3927 if (ipv6_addr_is_multicast(&msg->dest)) {
3928 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
3933 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
3935 } else if (ipv6_addr_type(&msg->target) !=
3936 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
3937 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
3941 in6_dev = __in6_dev_get(skb->dev);
3944 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
3948 * The IP source address of the Redirect MUST be the same as the current
3949 * first-hop router for the specified ICMP Destination Address.
3952 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
3953 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
3958 if (ndopts.nd_opts_tgt_lladdr) {
3959 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
3962 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
3967 rt = (struct rt6_info *) dst;
3968 if (rt->rt6i_flags & RTF_REJECT) {
3969 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
3973 /* Redirect received -> path was valid.
3974 * Look, redirects are sent only in response to data packets,
3975 * so that this nexthop apparently is reachable. --ANK
3977 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
3979 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
3984 * We have finally decided to accept it.
3987 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
3988 NEIGH_UPDATE_F_WEAK_OVERRIDE|
3989 NEIGH_UPDATE_F_OVERRIDE|
3990 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
3991 NEIGH_UPDATE_F_ISROUTER)),
3992 NDISC_REDIRECT, &ndopts);
3995 res.f6i = rcu_dereference(rt->from);
4000 struct fib6_nh_match_arg arg = {
4002 .gw = &rt->rt6i_gateway,
4005 nexthop_for_each_fib6_nh(res.f6i->nh,
4006 fib6_nh_find_match, &arg);
4008 /* fib6_info uses a nexthop that does not have fib6_nh
4009 * using the dst->dev. Should be impossible
4015 res.nh = res.f6i->fib6_nh;
4018 res.fib6_flags = res.f6i->fib6_flags;
4019 res.fib6_type = res.f6i->fib6_type;
4020 nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4024 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4026 nrt->rt6i_flags &= ~RTF_GATEWAY;
4028 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4030 /* rt6_insert_exception() will take care of duplicated exceptions */
4031 if (rt6_insert_exception(nrt, &res)) {
4032 dst_release_immediate(&nrt->dst);
4036 netevent.old = &rt->dst;
4037 netevent.new = &nrt->dst;
4038 netevent.daddr = &msg->dest;
4039 netevent.neigh = neigh;
4040 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4044 neigh_release(neigh);
4047 #ifdef CONFIG_IPV6_ROUTE_INFO
4048 static struct fib6_info *rt6_get_route_info(struct net *net,
4049 const struct in6_addr *prefix, int prefixlen,
4050 const struct in6_addr *gwaddr,
4051 struct net_device *dev)
4053 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4054 int ifindex = dev->ifindex;
4055 struct fib6_node *fn;
4056 struct fib6_info *rt = NULL;
4057 struct fib6_table *table;
4059 table = fib6_get_table(net, tb_id);
4064 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4068 for_each_fib6_node_rt_rcu(fn) {
4069 /* these routes do not use nexthops */
4072 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4074 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4075 !rt->fib6_nh->fib_nh_gw_family)
4077 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4079 if (!fib6_info_hold_safe(rt))
4088 static struct fib6_info *rt6_add_route_info(struct net *net,
4089 const struct in6_addr *prefix, int prefixlen,
4090 const struct in6_addr *gwaddr,
4091 struct net_device *dev,
4094 struct fib6_config cfg = {
4095 .fc_metric = IP6_RT_PRIO_USER,
4096 .fc_ifindex = dev->ifindex,
4097 .fc_dst_len = prefixlen,
4098 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4099 RTF_UP | RTF_PREF(pref),
4100 .fc_protocol = RTPROT_RA,
4101 .fc_type = RTN_UNICAST,
4102 .fc_nlinfo.portid = 0,
4103 .fc_nlinfo.nlh = NULL,
4104 .fc_nlinfo.nl_net = net,
4107 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4108 cfg.fc_dst = *prefix;
4109 cfg.fc_gateway = *gwaddr;
4111 /* We should treat it as a default route if prefix length is 0. */
4113 cfg.fc_flags |= RTF_DEFAULT;
4115 ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4117 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4121 struct fib6_info *rt6_get_dflt_router(struct net *net,
4122 const struct in6_addr *addr,
4123 struct net_device *dev)
4125 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4126 struct fib6_info *rt;
4127 struct fib6_table *table;
4129 table = fib6_get_table(net, tb_id);
4134 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4137 /* RA routes do not use nexthops */
4142 if (dev == nh->fib_nh_dev &&
4143 ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4144 ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4147 if (rt && !fib6_info_hold_safe(rt))
4153 struct fib6_info *rt6_add_dflt_router(struct net *net,
4154 const struct in6_addr *gwaddr,
4155 struct net_device *dev,
4158 struct fib6_config cfg = {
4159 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4160 .fc_metric = IP6_RT_PRIO_USER,
4161 .fc_ifindex = dev->ifindex,
4162 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4163 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4164 .fc_protocol = RTPROT_RA,
4165 .fc_type = RTN_UNICAST,
4166 .fc_nlinfo.portid = 0,
4167 .fc_nlinfo.nlh = NULL,
4168 .fc_nlinfo.nl_net = net,
4171 cfg.fc_gateway = *gwaddr;
4173 if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4174 struct fib6_table *table;
4176 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4178 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4181 return rt6_get_dflt_router(net, gwaddr, dev);
4184 static void __rt6_purge_dflt_routers(struct net *net,
4185 struct fib6_table *table)
4187 struct fib6_info *rt;
4191 for_each_fib6_node_rt_rcu(&table->tb6_root) {
4192 struct net_device *dev = fib6_info_nh_dev(rt);
4193 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4195 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4196 (!idev || idev->cnf.accept_ra != 2) &&
4197 fib6_info_hold_safe(rt)) {
4199 ip6_del_rt(net, rt);
4205 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4208 void rt6_purge_dflt_routers(struct net *net)
4210 struct fib6_table *table;
4211 struct hlist_head *head;
4216 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4217 head = &net->ipv6.fib_table_hash[h];
4218 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4219 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4220 __rt6_purge_dflt_routers(net, table);
4227 static void rtmsg_to_fib6_config(struct net *net,
4228 struct in6_rtmsg *rtmsg,
4229 struct fib6_config *cfg)
4231 *cfg = (struct fib6_config){
4232 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4234 .fc_ifindex = rtmsg->rtmsg_ifindex,
4235 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
4236 .fc_expires = rtmsg->rtmsg_info,
4237 .fc_dst_len = rtmsg->rtmsg_dst_len,
4238 .fc_src_len = rtmsg->rtmsg_src_len,
4239 .fc_flags = rtmsg->rtmsg_flags,
4240 .fc_type = rtmsg->rtmsg_type,
4242 .fc_nlinfo.nl_net = net,
4244 .fc_dst = rtmsg->rtmsg_dst,
4245 .fc_src = rtmsg->rtmsg_src,
4246 .fc_gateway = rtmsg->rtmsg_gateway,
4250 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
4252 struct fib6_config cfg;
4253 struct in6_rtmsg rtmsg;
4257 case SIOCADDRT: /* Add a route */
4258 case SIOCDELRT: /* Delete a route */
4259 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4261 err = copy_from_user(&rtmsg, arg,
4262 sizeof(struct in6_rtmsg));
4266 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
4271 err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4274 err = ip6_route_del(&cfg, NULL);
4288 * Drop the packet on the floor
4291 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4293 struct dst_entry *dst = skb_dst(skb);
4294 struct net *net = dev_net(dst->dev);
4295 struct inet6_dev *idev;
4298 if (netif_is_l3_master(skb->dev) &&
4299 dst->dev == net->loopback_dev)
4300 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4302 idev = ip6_dst_idev(dst);
4304 switch (ipstats_mib_noroutes) {
4305 case IPSTATS_MIB_INNOROUTES:
4306 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4307 if (type == IPV6_ADDR_ANY) {
4308 IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4312 case IPSTATS_MIB_OUTNOROUTES:
4313 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4317 /* Start over by dropping the dst for l3mdev case */
4318 if (netif_is_l3_master(skb->dev))
4321 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4326 static int ip6_pkt_discard(struct sk_buff *skb)
4328 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4331 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4333 skb->dev = skb_dst(skb)->dev;
4334 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4337 static int ip6_pkt_prohibit(struct sk_buff *skb)
4339 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4342 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4344 skb->dev = skb_dst(skb)->dev;
4345 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4349 * Allocate a dst for local (unicast / anycast) address.
4352 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4353 struct inet6_dev *idev,
4354 const struct in6_addr *addr,
4355 bool anycast, gfp_t gfp_flags)
4357 struct fib6_config cfg = {
4358 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4359 .fc_ifindex = idev->dev->ifindex,
4360 .fc_flags = RTF_UP | RTF_ADDRCONF | RTF_NONEXTHOP,
4363 .fc_protocol = RTPROT_KERNEL,
4364 .fc_nlinfo.nl_net = net,
4365 .fc_ignore_dev_down = true,
4369 cfg.fc_type = RTN_ANYCAST;
4370 cfg.fc_flags |= RTF_ANYCAST;
4372 cfg.fc_type = RTN_LOCAL;
4373 cfg.fc_flags |= RTF_LOCAL;
4376 return ip6_route_info_create(&cfg, gfp_flags, NULL);
4379 /* remove deleted ip from prefsrc entries */
4380 struct arg_dev_net_ip {
4381 struct net_device *dev;
4383 struct in6_addr *addr;
4386 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4388 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4389 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4390 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4393 ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
4394 rt != net->ipv6.fib6_null_entry &&
4395 ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
4396 spin_lock_bh(&rt6_exception_lock);
4397 /* remove prefsrc entry */
4398 rt->fib6_prefsrc.plen = 0;
4399 spin_unlock_bh(&rt6_exception_lock);
4404 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4406 struct net *net = dev_net(ifp->idev->dev);
4407 struct arg_dev_net_ip adni = {
4408 .dev = ifp->idev->dev,
4412 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4415 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT)
4417 /* Remove routers and update dst entries when gateway turn into host. */
4418 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4420 struct in6_addr *gateway = (struct in6_addr *)arg;
4423 /* RA routes do not use nexthops */
4428 if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4429 nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4432 /* Further clean up cached routes in exception table.
4433 * This is needed because cached route may have a different
4434 * gateway than its 'parent' in the case of an ip redirect.
4436 fib6_nh_exceptions_clean_tohost(nh, gateway);
4441 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4443 fib6_clean_all(net, fib6_clean_tohost, gateway);
4446 struct arg_netdev_event {
4447 const struct net_device *dev;
4449 unsigned char nh_flags;
4450 unsigned long event;
4454 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4456 struct fib6_info *iter;
4457 struct fib6_node *fn;
4459 fn = rcu_dereference_protected(rt->fib6_node,
4460 lockdep_is_held(&rt->fib6_table->tb6_lock));
4461 iter = rcu_dereference_protected(fn->leaf,
4462 lockdep_is_held(&rt->fib6_table->tb6_lock));
4464 if (iter->fib6_metric == rt->fib6_metric &&
4465 rt6_qualify_for_ecmp(iter))
4467 iter = rcu_dereference_protected(iter->fib6_next,
4468 lockdep_is_held(&rt->fib6_table->tb6_lock));
4474 /* only called for fib entries with builtin fib6_nh */
4475 static bool rt6_is_dead(const struct fib6_info *rt)
4477 if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4478 (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4479 ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4485 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4487 struct fib6_info *iter;
4490 if (!rt6_is_dead(rt))
4491 total += rt->fib6_nh->fib_nh_weight;
4493 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4494 if (!rt6_is_dead(iter))
4495 total += iter->fib6_nh->fib_nh_weight;
4501 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4503 int upper_bound = -1;
4505 if (!rt6_is_dead(rt)) {
4506 *weight += rt->fib6_nh->fib_nh_weight;
4507 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4510 atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4513 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4515 struct fib6_info *iter;
4518 rt6_upper_bound_set(rt, &weight, total);
4520 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4521 rt6_upper_bound_set(iter, &weight, total);
4524 void rt6_multipath_rebalance(struct fib6_info *rt)
4526 struct fib6_info *first;
4529 /* In case the entire multipath route was marked for flushing,
4530 * then there is no need to rebalance upon the removal of every
4533 if (!rt->fib6_nsiblings || rt->should_flush)
4536 /* During lookup routes are evaluated in order, so we need to
4537 * make sure upper bounds are assigned from the first sibling
4540 first = rt6_multipath_first_sibling(rt);
4541 if (WARN_ON_ONCE(!first))
4544 total = rt6_multipath_total_weight(first);
4545 rt6_multipath_upper_bound_set(first, total);
4548 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4550 const struct arg_netdev_event *arg = p_arg;
4551 struct net *net = dev_net(arg->dev);
4553 if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4554 rt->fib6_nh->fib_nh_dev == arg->dev) {
4555 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4556 fib6_update_sernum_upto_root(net, rt);
4557 rt6_multipath_rebalance(rt);
4563 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4565 struct arg_netdev_event arg = {
4568 .nh_flags = nh_flags,
4572 if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4573 arg.nh_flags |= RTNH_F_LINKDOWN;
4575 fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4578 /* only called for fib entries with inline fib6_nh */
4579 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4580 const struct net_device *dev)
4582 struct fib6_info *iter;
4584 if (rt->fib6_nh->fib_nh_dev == dev)
4586 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4587 if (iter->fib6_nh->fib_nh_dev == dev)
4593 static void rt6_multipath_flush(struct fib6_info *rt)
4595 struct fib6_info *iter;
4597 rt->should_flush = 1;
4598 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4599 iter->should_flush = 1;
4602 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4603 const struct net_device *down_dev)
4605 struct fib6_info *iter;
4606 unsigned int dead = 0;
4608 if (rt->fib6_nh->fib_nh_dev == down_dev ||
4609 rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4611 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4612 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4613 iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4619 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4620 const struct net_device *dev,
4621 unsigned char nh_flags)
4623 struct fib6_info *iter;
4625 if (rt->fib6_nh->fib_nh_dev == dev)
4626 rt->fib6_nh->fib_nh_flags |= nh_flags;
4627 list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4628 if (iter->fib6_nh->fib_nh_dev == dev)
4629 iter->fib6_nh->fib_nh_flags |= nh_flags;
4632 /* called with write lock held for table with rt */
4633 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4635 const struct arg_netdev_event *arg = p_arg;
4636 const struct net_device *dev = arg->dev;
4637 struct net *net = dev_net(dev);
4639 if (rt == net->ipv6.fib6_null_entry || rt->nh)
4642 switch (arg->event) {
4643 case NETDEV_UNREGISTER:
4644 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4646 if (rt->should_flush)
4648 if (!rt->fib6_nsiblings)
4649 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4650 if (rt6_multipath_uses_dev(rt, dev)) {
4653 count = rt6_multipath_dead_count(rt, dev);
4654 if (rt->fib6_nsiblings + 1 == count) {
4655 rt6_multipath_flush(rt);
4658 rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4660 fib6_update_sernum(net, rt);
4661 rt6_multipath_rebalance(rt);
4665 if (rt->fib6_nh->fib_nh_dev != dev ||
4666 rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4668 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4669 rt6_multipath_rebalance(rt);
4676 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4678 struct arg_netdev_event arg = {
4684 struct net *net = dev_net(dev);
4686 if (net->ipv6.sysctl.skip_notify_on_dev_down)
4687 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4689 fib6_clean_all(net, fib6_ifdown, &arg);
4692 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4694 rt6_sync_down_dev(dev, event);
4695 rt6_uncached_list_flush_dev(dev_net(dev), dev);
4696 neigh_ifdown(&nd_tbl, dev);
4699 struct rt6_mtu_change_arg {
4700 struct net_device *dev;
4702 struct fib6_info *f6i;
4705 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4707 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4708 struct fib6_info *f6i = arg->f6i;
4710 /* For administrative MTU increase, there is no way to discover
4711 * IPv6 PMTU increase, so PMTU increase should be updated here.
4712 * Since RFC 1981 doesn't include administrative MTU increase
4713 * update PMTU increase is a MUST. (i.e. jumbo frame)
4715 if (nh->fib_nh_dev == arg->dev) {
4716 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4717 u32 mtu = f6i->fib6_pmtu;
4719 if (mtu >= arg->mtu ||
4720 (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4721 fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4723 spin_lock_bh(&rt6_exception_lock);
4724 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4725 spin_unlock_bh(&rt6_exception_lock);
4731 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4733 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4734 struct inet6_dev *idev;
4736 /* In IPv6 pmtu discovery is not optional,
4737 so that RTAX_MTU lock cannot disable it.
4738 We still use this lock to block changes
4739 caused by addrconf/ndisc.
4742 idev = __in6_dev_get(arg->dev);
4746 if (fib6_metric_locked(f6i, RTAX_MTU))
4751 /* fib6_nh_mtu_change only returns 0, so this is safe */
4752 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4756 return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4759 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4761 struct rt6_mtu_change_arg arg = {
4766 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4769 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4770 [RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 },
4771 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
4772 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
4773 [RTA_OIF] = { .type = NLA_U32 },
4774 [RTA_IIF] = { .type = NLA_U32 },
4775 [RTA_PRIORITY] = { .type = NLA_U32 },
4776 [RTA_METRICS] = { .type = NLA_NESTED },
4777 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
4778 [RTA_PREF] = { .type = NLA_U8 },
4779 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
4780 [RTA_ENCAP] = { .type = NLA_NESTED },
4781 [RTA_EXPIRES] = { .type = NLA_U32 },
4782 [RTA_UID] = { .type = NLA_U32 },
4783 [RTA_MARK] = { .type = NLA_U32 },
4784 [RTA_TABLE] = { .type = NLA_U32 },
4785 [RTA_IP_PROTO] = { .type = NLA_U8 },
4786 [RTA_SPORT] = { .type = NLA_U16 },
4787 [RTA_DPORT] = { .type = NLA_U16 },
4788 [RTA_NH_ID] = { .type = NLA_U32 },
4791 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
4792 struct fib6_config *cfg,
4793 struct netlink_ext_ack *extack)
4796 struct nlattr *tb[RTA_MAX+1];
4800 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4801 rtm_ipv6_policy, extack);
4806 rtm = nlmsg_data(nlh);
4808 *cfg = (struct fib6_config){
4809 .fc_table = rtm->rtm_table,
4810 .fc_dst_len = rtm->rtm_dst_len,
4811 .fc_src_len = rtm->rtm_src_len,
4813 .fc_protocol = rtm->rtm_protocol,
4814 .fc_type = rtm->rtm_type,
4816 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4817 .fc_nlinfo.nlh = nlh,
4818 .fc_nlinfo.nl_net = sock_net(skb->sk),
4821 if (rtm->rtm_type == RTN_UNREACHABLE ||
4822 rtm->rtm_type == RTN_BLACKHOLE ||
4823 rtm->rtm_type == RTN_PROHIBIT ||
4824 rtm->rtm_type == RTN_THROW)
4825 cfg->fc_flags |= RTF_REJECT;
4827 if (rtm->rtm_type == RTN_LOCAL)
4828 cfg->fc_flags |= RTF_LOCAL;
4830 if (rtm->rtm_flags & RTM_F_CLONED)
4831 cfg->fc_flags |= RTF_CACHE;
4833 cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4835 if (tb[RTA_NH_ID]) {
4836 if (tb[RTA_GATEWAY] || tb[RTA_OIF] ||
4837 tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
4838 NL_SET_ERR_MSG(extack,
4839 "Nexthop specification and nexthop id are mutually exclusive");
4842 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
4845 if (tb[RTA_GATEWAY]) {
4846 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
4847 cfg->fc_flags |= RTF_GATEWAY;
4850 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
4855 int plen = (rtm->rtm_dst_len + 7) >> 3;
4857 if (nla_len(tb[RTA_DST]) < plen)
4860 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
4864 int plen = (rtm->rtm_src_len + 7) >> 3;
4866 if (nla_len(tb[RTA_SRC]) < plen)
4869 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
4872 if (tb[RTA_PREFSRC])
4873 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
4876 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4878 if (tb[RTA_PRIORITY])
4879 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4881 if (tb[RTA_METRICS]) {
4882 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4883 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
4887 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4889 if (tb[RTA_MULTIPATH]) {
4890 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4891 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
4893 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
4894 cfg->fc_mp_len, extack);
4900 pref = nla_get_u8(tb[RTA_PREF]);
4901 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4902 pref != ICMPV6_ROUTER_PREF_HIGH)
4903 pref = ICMPV6_ROUTER_PREF_MEDIUM;
4904 cfg->fc_flags |= RTF_PREF(pref);
4908 cfg->fc_encap = tb[RTA_ENCAP];
4910 if (tb[RTA_ENCAP_TYPE]) {
4911 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4913 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
4918 if (tb[RTA_EXPIRES]) {
4919 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
4921 if (addrconf_finite_timeout(timeout)) {
4922 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
4923 cfg->fc_flags |= RTF_EXPIRES;
4933 struct fib6_info *fib6_info;
4934 struct fib6_config r_cfg;
4935 struct list_head next;
4938 static int ip6_route_info_append(struct net *net,
4939 struct list_head *rt6_nh_list,
4940 struct fib6_info *rt,
4941 struct fib6_config *r_cfg)
4946 list_for_each_entry(nh, rt6_nh_list, next) {
4947 /* check if fib6_info already exists */
4948 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
4952 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
4956 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
4957 list_add_tail(&nh->next, rt6_nh_list);
4962 static void ip6_route_mpath_notify(struct fib6_info *rt,
4963 struct fib6_info *rt_last,
4964 struct nl_info *info,
4967 /* if this is an APPEND route, then rt points to the first route
4968 * inserted and rt_last points to last route inserted. Userspace
4969 * wants a consistent dump of the route which starts at the first
4970 * nexthop. Since sibling routes are always added at the end of
4971 * the list, find the first sibling of the last route appended
4973 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
4974 rt = list_first_entry(&rt_last->fib6_siblings,
4980 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
4983 static int ip6_route_multipath_add(struct fib6_config *cfg,
4984 struct netlink_ext_ack *extack)
4986 struct fib6_info *rt_notif = NULL, *rt_last = NULL;
4987 struct nl_info *info = &cfg->fc_nlinfo;
4988 enum fib_event_type event_type;
4989 struct fib6_config r_cfg;
4990 struct rtnexthop *rtnh;
4991 struct fib6_info *rt;
4992 struct rt6_nh *err_nh;
4993 struct rt6_nh *nh, *nh_safe;
4999 int replace = (cfg->fc_nlinfo.nlh &&
5000 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5001 LIST_HEAD(rt6_nh_list);
5003 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5004 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5005 nlflags |= NLM_F_APPEND;
5007 remaining = cfg->fc_mp_len;
5008 rtnh = (struct rtnexthop *)cfg->fc_mp;
5010 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
5011 * fib6_info structs per nexthop
5013 while (rtnh_ok(rtnh, remaining)) {
5014 memcpy(&r_cfg, cfg, sizeof(*cfg));
5015 if (rtnh->rtnh_ifindex)
5016 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5018 attrlen = rtnh_attrlen(rtnh);
5020 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5022 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5024 r_cfg.fc_gateway = nla_get_in6_addr(nla);
5025 r_cfg.fc_flags |= RTF_GATEWAY;
5027 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5028 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5030 r_cfg.fc_encap_type = nla_get_u16(nla);
5033 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5034 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5040 if (!rt6_qualify_for_ecmp(rt)) {
5042 NL_SET_ERR_MSG(extack,
5043 "Device only routes can not be added for IPv6 using the multipath API.");
5044 fib6_info_release(rt);
5048 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5050 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5053 fib6_info_release(rt);
5057 rtnh = rtnh_next(rtnh, &remaining);
5060 if (list_empty(&rt6_nh_list)) {
5061 NL_SET_ERR_MSG(extack,
5062 "Invalid nexthop configuration - no valid nexthops");
5066 /* for add and replace send one notification with all nexthops.
5067 * Skip the notification in fib6_add_rt2node and send one with
5068 * the full route when done
5070 info->skip_notify = 1;
5072 /* For add and replace, send one notification with all nexthops. For
5073 * append, send one notification with all appended nexthops.
5075 info->skip_notify_kernel = 1;
5078 list_for_each_entry(nh, &rt6_nh_list, next) {
5079 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5080 fib6_info_release(nh->fib6_info);
5083 /* save reference to last route successfully inserted */
5084 rt_last = nh->fib6_info;
5086 /* save reference to first route for notification */
5088 rt_notif = nh->fib6_info;
5091 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5092 nh->fib6_info = NULL;
5095 NL_SET_ERR_MSG_MOD(extack,
5096 "multipath route replace failed (check consistency of installed routes)");
5101 /* Because each route is added like a single route we remove
5102 * these flags after the first nexthop: if there is a collision,
5103 * we have already failed to add the first nexthop:
5104 * fib6_add_rt2node() has rejected it; when replacing, old
5105 * nexthops have been replaced by first new, the rest should
5108 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5113 event_type = replace ? FIB_EVENT_ENTRY_REPLACE : FIB_EVENT_ENTRY_ADD;
5114 err = call_fib6_multipath_entry_notifiers(info->nl_net, event_type,
5115 rt_notif, nhn - 1, extack);
5117 /* Delete all the siblings that were just added */
5122 /* success ... tell user about new route */
5123 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5127 /* send notification for routes that were added so that
5128 * the delete notifications sent by ip6_route_del are
5132 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5134 /* Delete routes that were already added */
5135 list_for_each_entry(nh, &rt6_nh_list, next) {
5138 ip6_route_del(&nh->r_cfg, extack);
5142 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5144 fib6_info_release(nh->fib6_info);
5145 list_del(&nh->next);
5152 static int ip6_route_multipath_del(struct fib6_config *cfg,
5153 struct netlink_ext_ack *extack)
5155 struct fib6_config r_cfg;
5156 struct rtnexthop *rtnh;
5159 int err = 1, last_err = 0;
5161 remaining = cfg->fc_mp_len;
5162 rtnh = (struct rtnexthop *)cfg->fc_mp;
5164 /* Parse a Multipath Entry */
5165 while (rtnh_ok(rtnh, remaining)) {
5166 memcpy(&r_cfg, cfg, sizeof(*cfg));
5167 if (rtnh->rtnh_ifindex)
5168 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5170 attrlen = rtnh_attrlen(rtnh);
5172 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5174 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5176 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5177 r_cfg.fc_flags |= RTF_GATEWAY;
5180 err = ip6_route_del(&r_cfg, extack);
5184 rtnh = rtnh_next(rtnh, &remaining);
5190 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5191 struct netlink_ext_ack *extack)
5193 struct fib6_config cfg;
5196 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5201 !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5202 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5207 return ip6_route_multipath_del(&cfg, extack);
5209 cfg.fc_delete_all_nh = 1;
5210 return ip6_route_del(&cfg, extack);
5214 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5215 struct netlink_ext_ack *extack)
5217 struct fib6_config cfg;
5220 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5224 if (cfg.fc_metric == 0)
5225 cfg.fc_metric = IP6_RT_PRIO_USER;
5228 return ip6_route_multipath_add(&cfg, extack);
5230 return ip6_route_add(&cfg, GFP_KERNEL, extack);
5233 /* add the overhead of this fib6_nh to nexthop_len */
5234 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5236 int *nexthop_len = arg;
5238 *nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */
5239 + NLA_ALIGN(sizeof(struct rtnexthop))
5240 + nla_total_size(16); /* RTA_GATEWAY */
5242 if (nh->fib_nh_lws) {
5243 /* RTA_ENCAP_TYPE */
5244 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5246 *nexthop_len += nla_total_size(2);
5252 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5257 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5258 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5261 struct fib6_nh *nh = f6i->fib6_nh;
5264 if (f6i->fib6_nsiblings) {
5265 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
5266 + NLA_ALIGN(sizeof(struct rtnexthop))
5267 + nla_total_size(16) /* RTA_GATEWAY */
5268 + lwtunnel_get_encap_size(nh->fib_nh_lws);
5270 nexthop_len *= f6i->fib6_nsiblings;
5272 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5275 return NLMSG_ALIGN(sizeof(struct rtmsg))
5276 + nla_total_size(16) /* RTA_SRC */
5277 + nla_total_size(16) /* RTA_DST */
5278 + nla_total_size(16) /* RTA_GATEWAY */
5279 + nla_total_size(16) /* RTA_PREFSRC */
5280 + nla_total_size(4) /* RTA_TABLE */
5281 + nla_total_size(4) /* RTA_IIF */
5282 + nla_total_size(4) /* RTA_OIF */
5283 + nla_total_size(4) /* RTA_PRIORITY */
5284 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5285 + nla_total_size(sizeof(struct rta_cacheinfo))
5286 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5287 + nla_total_size(1) /* RTA_PREF */
5291 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5292 unsigned char *flags)
5294 if (nexthop_is_multipath(nh)) {
5297 mp = nla_nest_start(skb, RTA_MULTIPATH);
5299 goto nla_put_failure;
5301 if (nexthop_mpath_fill_node(skb, nh))
5302 goto nla_put_failure;
5304 nla_nest_end(skb, mp);
5306 struct fib6_nh *fib6_nh;
5308 fib6_nh = nexthop_fib6_nh(nh);
5309 if (fib_nexthop_info(skb, &fib6_nh->nh_common,
5311 goto nla_put_failure;
5320 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5321 struct fib6_info *rt, struct dst_entry *dst,
5322 struct in6_addr *dest, struct in6_addr *src,
5323 int iif, int type, u32 portid, u32 seq,
5326 struct rt6_info *rt6 = (struct rt6_info *)dst;
5327 struct rt6key *rt6_dst, *rt6_src;
5328 u32 *pmetrics, table, rt6_flags;
5329 unsigned char nh_flags = 0;
5330 struct nlmsghdr *nlh;
5334 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5339 rt6_dst = &rt6->rt6i_dst;
5340 rt6_src = &rt6->rt6i_src;
5341 rt6_flags = rt6->rt6i_flags;
5343 rt6_dst = &rt->fib6_dst;
5344 rt6_src = &rt->fib6_src;
5345 rt6_flags = rt->fib6_flags;
5348 rtm = nlmsg_data(nlh);
5349 rtm->rtm_family = AF_INET6;
5350 rtm->rtm_dst_len = rt6_dst->plen;
5351 rtm->rtm_src_len = rt6_src->plen;
5354 table = rt->fib6_table->tb6_id;
5356 table = RT6_TABLE_UNSPEC;
5357 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5358 if (nla_put_u32(skb, RTA_TABLE, table))
5359 goto nla_put_failure;
5361 rtm->rtm_type = rt->fib6_type;
5363 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5364 rtm->rtm_protocol = rt->fib6_protocol;
5366 if (rt6_flags & RTF_CACHE)
5367 rtm->rtm_flags |= RTM_F_CLONED;
5370 if (nla_put_in6_addr(skb, RTA_DST, dest))
5371 goto nla_put_failure;
5372 rtm->rtm_dst_len = 128;
5373 } else if (rtm->rtm_dst_len)
5374 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5375 goto nla_put_failure;
5376 #ifdef CONFIG_IPV6_SUBTREES
5378 if (nla_put_in6_addr(skb, RTA_SRC, src))
5379 goto nla_put_failure;
5380 rtm->rtm_src_len = 128;
5381 } else if (rtm->rtm_src_len &&
5382 nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5383 goto nla_put_failure;
5386 #ifdef CONFIG_IPV6_MROUTE
5387 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5388 int err = ip6mr_get_route(net, skb, rtm, portid);
5393 goto nla_put_failure;
5396 if (nla_put_u32(skb, RTA_IIF, iif))
5397 goto nla_put_failure;
5399 struct in6_addr saddr_buf;
5400 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
5401 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5402 goto nla_put_failure;
5405 if (rt->fib6_prefsrc.plen) {
5406 struct in6_addr saddr_buf;
5407 saddr_buf = rt->fib6_prefsrc.addr;
5408 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5409 goto nla_put_failure;
5412 pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5413 if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5414 goto nla_put_failure;
5416 if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5417 goto nla_put_failure;
5419 /* For multipath routes, walk the siblings list and add
5420 * each as a nexthop within RTA_MULTIPATH.
5423 if (rt6_flags & RTF_GATEWAY &&
5424 nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5425 goto nla_put_failure;
5427 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5428 goto nla_put_failure;
5429 } else if (rt->fib6_nsiblings) {
5430 struct fib6_info *sibling, *next_sibling;
5433 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5435 goto nla_put_failure;
5437 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5438 rt->fib6_nh->fib_nh_weight) < 0)
5439 goto nla_put_failure;
5441 list_for_each_entry_safe(sibling, next_sibling,
5442 &rt->fib6_siblings, fib6_siblings) {
5443 if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5444 sibling->fib6_nh->fib_nh_weight) < 0)
5445 goto nla_put_failure;
5448 nla_nest_end(skb, mp);
5449 } else if (rt->nh) {
5450 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5451 goto nla_put_failure;
5453 if (nexthop_is_blackhole(rt->nh))
5454 rtm->rtm_type = RTN_BLACKHOLE;
5456 if (rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5457 goto nla_put_failure;
5459 rtm->rtm_flags |= nh_flags;
5461 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common,
5462 &nh_flags, false) < 0)
5463 goto nla_put_failure;
5465 rtm->rtm_flags |= nh_flags;
5468 if (rt6_flags & RTF_EXPIRES) {
5469 expires = dst ? dst->expires : rt->expires;
5473 if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5474 goto nla_put_failure;
5476 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5477 goto nla_put_failure;
5480 nlmsg_end(skb, nlh);
5484 nlmsg_cancel(skb, nlh);
5488 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5490 const struct net_device *dev = arg;
5492 if (nh->fib_nh_dev == dev)
5498 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5499 const struct net_device *dev)
5502 struct net_device *_dev = (struct net_device *)dev;
5504 return !!nexthop_for_each_fib6_nh(f6i->nh,
5505 fib6_info_nh_uses_dev,
5509 if (f6i->fib6_nh->fib_nh_dev == dev)
5512 if (f6i->fib6_nsiblings) {
5513 struct fib6_info *sibling, *next_sibling;
5515 list_for_each_entry_safe(sibling, next_sibling,
5516 &f6i->fib6_siblings, fib6_siblings) {
5517 if (sibling->fib6_nh->fib_nh_dev == dev)
5525 struct fib6_nh_exception_dump_walker {
5526 struct rt6_rtnl_dump_arg *dump;
5527 struct fib6_info *rt;
5533 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5535 struct fib6_nh_exception_dump_walker *w = arg;
5536 struct rt6_rtnl_dump_arg *dump = w->dump;
5537 struct rt6_exception_bucket *bucket;
5538 struct rt6_exception *rt6_ex;
5541 bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5545 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5546 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5552 /* Expiration of entries doesn't bump sernum, insertion
5553 * does. Removal is triggered by insertion, so we can
5554 * rely on the fact that if entries change between two
5555 * partial dumps, this node is scanned again completely,
5556 * see rt6_insert_exception() and fib6_dump_table().
5558 * Count expired entries we go through as handled
5559 * entries that we'll skip next time, in case of partial
5560 * node dump. Otherwise, if entries expire meanwhile,
5561 * we'll skip the wrong amount.
5563 if (rt6_check_expired(rt6_ex->rt6i)) {
5568 err = rt6_fill_node(dump->net, dump->skb, w->rt,
5569 &rt6_ex->rt6i->dst, NULL, NULL, 0,
5571 NETLINK_CB(dump->cb->skb).portid,
5572 dump->cb->nlh->nlmsg_seq, w->flags);
5584 /* Return -1 if done with node, number of handled routes on partial dump */
5585 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5587 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5588 struct fib_dump_filter *filter = &arg->filter;
5589 unsigned int flags = NLM_F_MULTI;
5590 struct net *net = arg->net;
5593 if (rt == net->ipv6.fib6_null_entry)
5596 if ((filter->flags & RTM_F_PREFIX) &&
5597 !(rt->fib6_flags & RTF_PREFIX_RT)) {
5598 /* success since this is not a prefix route */
5601 if (filter->filter_set &&
5602 ((filter->rt_type && rt->fib6_type != filter->rt_type) ||
5603 (filter->dev && !fib6_info_uses_dev(rt, filter->dev)) ||
5604 (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5608 if (filter->filter_set ||
5609 !filter->dump_routes || !filter->dump_exceptions) {
5610 flags |= NLM_F_DUMP_FILTERED;
5613 if (filter->dump_routes) {
5617 if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5619 NETLINK_CB(arg->cb->skb).portid,
5620 arg->cb->nlh->nlmsg_seq, flags)) {
5627 if (filter->dump_exceptions) {
5628 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5636 err = nexthop_for_each_fib6_nh(rt->nh,
5637 rt6_nh_dump_exceptions,
5640 err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5644 return count += w.count;
5650 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5651 const struct nlmsghdr *nlh,
5653 struct netlink_ext_ack *extack)
5658 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5659 NL_SET_ERR_MSG_MOD(extack,
5660 "Invalid header for get route request");
5664 if (!netlink_strict_get_check(skb))
5665 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5666 rtm_ipv6_policy, extack);
5668 rtm = nlmsg_data(nlh);
5669 if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5670 (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5671 rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5673 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5676 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5677 NL_SET_ERR_MSG_MOD(extack,
5678 "Invalid flags for get route request");
5682 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5683 rtm_ipv6_policy, extack);
5687 if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5688 (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5689 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5693 for (i = 0; i <= RTA_MAX; i++) {
5709 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5717 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5718 struct netlink_ext_ack *extack)
5720 struct net *net = sock_net(in_skb->sk);
5721 struct nlattr *tb[RTA_MAX+1];
5722 int err, iif = 0, oif = 0;
5723 struct fib6_info *from;
5724 struct dst_entry *dst;
5725 struct rt6_info *rt;
5726 struct sk_buff *skb;
5728 struct flowi6 fl6 = {};
5731 err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
5736 rtm = nlmsg_data(nlh);
5737 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
5738 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
5741 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
5744 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
5748 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
5751 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
5755 iif = nla_get_u32(tb[RTA_IIF]);
5758 oif = nla_get_u32(tb[RTA_OIF]);
5761 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
5764 fl6.flowi6_uid = make_kuid(current_user_ns(),
5765 nla_get_u32(tb[RTA_UID]));
5767 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
5770 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
5773 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
5775 if (tb[RTA_IP_PROTO]) {
5776 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
5777 &fl6.flowi6_proto, AF_INET6,
5784 struct net_device *dev;
5789 dev = dev_get_by_index_rcu(net, iif);
5796 fl6.flowi6_iif = iif;
5798 if (!ipv6_addr_any(&fl6.saddr))
5799 flags |= RT6_LOOKUP_F_HAS_SADDR;
5801 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
5805 fl6.flowi6_oif = oif;
5807 dst = ip6_route_output(net, NULL, &fl6);
5811 rt = container_of(dst, struct rt6_info, dst);
5812 if (rt->dst.error) {
5813 err = rt->dst.error;
5818 if (rt == net->ipv6.ip6_null_entry) {
5819 err = rt->dst.error;
5824 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
5831 skb_dst_set(skb, &rt->dst);
5834 from = rcu_dereference(rt->from);
5837 err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
5839 NETLINK_CB(in_skb).portid,
5842 err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
5843 &fl6.saddr, iif, RTM_NEWROUTE,
5844 NETLINK_CB(in_skb).portid,
5856 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5861 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
5862 unsigned int nlm_flags)
5864 struct sk_buff *skb;
5865 struct net *net = info->nl_net;
5870 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
5872 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
5876 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
5877 event, info->portid, seq, nlm_flags);
5879 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
5880 WARN_ON(err == -EMSGSIZE);
5884 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
5885 info->nlh, gfp_any());
5889 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
5892 void fib6_rt_update(struct net *net, struct fib6_info *rt,
5893 struct nl_info *info)
5895 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
5896 struct sk_buff *skb;
5899 /* call_fib6_entry_notifiers will be removed when in-kernel notifier
5900 * is implemented and supported for nexthop objects
5902 call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, rt, NULL);
5904 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
5908 err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
5909 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
5911 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
5912 WARN_ON(err == -EMSGSIZE);
5916 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
5917 info->nlh, gfp_any());
5921 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
5924 static int ip6_route_dev_notify(struct notifier_block *this,
5925 unsigned long event, void *ptr)
5927 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5928 struct net *net = dev_net(dev);
5930 if (!(dev->flags & IFF_LOOPBACK))
5933 if (event == NETDEV_REGISTER) {
5934 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
5935 net->ipv6.ip6_null_entry->dst.dev = dev;
5936 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
5937 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5938 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
5939 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
5940 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
5941 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
5943 } else if (event == NETDEV_UNREGISTER &&
5944 dev->reg_state != NETREG_UNREGISTERED) {
5945 /* NETDEV_UNREGISTER could be fired for multiple times by
5946 * netdev_wait_allrefs(). Make sure we only call this once.
5948 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
5949 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5950 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
5951 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
5962 #ifdef CONFIG_PROC_FS
5963 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
5965 struct net *net = (struct net *)seq->private;
5966 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
5967 net->ipv6.rt6_stats->fib_nodes,
5968 net->ipv6.rt6_stats->fib_route_nodes,
5969 atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
5970 net->ipv6.rt6_stats->fib_rt_entries,
5971 net->ipv6.rt6_stats->fib_rt_cache,
5972 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
5973 net->ipv6.rt6_stats->fib_discarded_routes);
5977 #endif /* CONFIG_PROC_FS */
5979 #ifdef CONFIG_SYSCTL
5982 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
5983 void __user *buffer, size_t *lenp, loff_t *ppos)
5991 net = (struct net *)ctl->extra1;
5992 delay = net->ipv6.sysctl.flush_delay;
5993 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
5997 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6004 static struct ctl_table ipv6_route_table_template[] = {
6006 .procname = "flush",
6007 .data = &init_net.ipv6.sysctl.flush_delay,
6008 .maxlen = sizeof(int),
6010 .proc_handler = ipv6_sysctl_rtcache_flush
6013 .procname = "gc_thresh",
6014 .data = &ip6_dst_ops_template.gc_thresh,
6015 .maxlen = sizeof(int),
6017 .proc_handler = proc_dointvec,
6020 .procname = "max_size",
6021 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
6022 .maxlen = sizeof(int),
6024 .proc_handler = proc_dointvec,
6027 .procname = "gc_min_interval",
6028 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6029 .maxlen = sizeof(int),
6031 .proc_handler = proc_dointvec_jiffies,
6034 .procname = "gc_timeout",
6035 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6036 .maxlen = sizeof(int),
6038 .proc_handler = proc_dointvec_jiffies,
6041 .procname = "gc_interval",
6042 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
6043 .maxlen = sizeof(int),
6045 .proc_handler = proc_dointvec_jiffies,
6048 .procname = "gc_elasticity",
6049 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6050 .maxlen = sizeof(int),
6052 .proc_handler = proc_dointvec,
6055 .procname = "mtu_expires",
6056 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6057 .maxlen = sizeof(int),
6059 .proc_handler = proc_dointvec_jiffies,
6062 .procname = "min_adv_mss",
6063 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
6064 .maxlen = sizeof(int),
6066 .proc_handler = proc_dointvec,
6069 .procname = "gc_min_interval_ms",
6070 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6071 .maxlen = sizeof(int),
6073 .proc_handler = proc_dointvec_ms_jiffies,
6076 .procname = "skip_notify_on_dev_down",
6077 .data = &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6078 .maxlen = sizeof(int),
6080 .proc_handler = proc_dointvec,
6087 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6089 struct ctl_table *table;
6091 table = kmemdup(ipv6_route_table_template,
6092 sizeof(ipv6_route_table_template),
6096 table[0].data = &net->ipv6.sysctl.flush_delay;
6097 table[0].extra1 = net;
6098 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6099 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
6100 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6101 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6102 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6103 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6104 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6105 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6106 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6107 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6109 /* Don't export sysctls to unprivileged users */
6110 if (net->user_ns != &init_user_ns)
6111 table[0].procname = NULL;
6118 static int __net_init ip6_route_net_init(struct net *net)
6122 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6123 sizeof(net->ipv6.ip6_dst_ops));
6125 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6126 goto out_ip6_dst_ops;
6128 net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6129 if (!net->ipv6.fib6_null_entry)
6130 goto out_ip6_dst_entries;
6131 memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6132 sizeof(*net->ipv6.fib6_null_entry));
6134 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6135 sizeof(*net->ipv6.ip6_null_entry),
6137 if (!net->ipv6.ip6_null_entry)
6138 goto out_fib6_null_entry;
6139 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6140 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6141 ip6_template_metrics, true);
6142 INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
6144 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6145 net->ipv6.fib6_has_custom_rules = false;
6146 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6147 sizeof(*net->ipv6.ip6_prohibit_entry),
6149 if (!net->ipv6.ip6_prohibit_entry)
6150 goto out_ip6_null_entry;
6151 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6152 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6153 ip6_template_metrics, true);
6154 INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
6156 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6157 sizeof(*net->ipv6.ip6_blk_hole_entry),
6159 if (!net->ipv6.ip6_blk_hole_entry)
6160 goto out_ip6_prohibit_entry;
6161 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6162 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6163 ip6_template_metrics, true);
6164 INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
6167 net->ipv6.sysctl.flush_delay = 0;
6168 net->ipv6.sysctl.ip6_rt_max_size = 4096;
6169 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6170 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6171 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6172 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6173 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6174 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6175 net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6177 net->ipv6.ip6_rt_gc_expire = 30*HZ;
6183 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6184 out_ip6_prohibit_entry:
6185 kfree(net->ipv6.ip6_prohibit_entry);
6187 kfree(net->ipv6.ip6_null_entry);
6189 out_fib6_null_entry:
6190 kfree(net->ipv6.fib6_null_entry);
6191 out_ip6_dst_entries:
6192 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6197 static void __net_exit ip6_route_net_exit(struct net *net)
6199 kfree(net->ipv6.fib6_null_entry);
6200 kfree(net->ipv6.ip6_null_entry);
6201 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6202 kfree(net->ipv6.ip6_prohibit_entry);
6203 kfree(net->ipv6.ip6_blk_hole_entry);
6205 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6208 static int __net_init ip6_route_net_init_late(struct net *net)
6210 #ifdef CONFIG_PROC_FS
6211 proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6212 sizeof(struct ipv6_route_iter));
6213 proc_create_net_single("rt6_stats", 0444, net->proc_net,
6214 rt6_stats_seq_show, NULL);
6219 static void __net_exit ip6_route_net_exit_late(struct net *net)
6221 #ifdef CONFIG_PROC_FS
6222 remove_proc_entry("ipv6_route", net->proc_net);
6223 remove_proc_entry("rt6_stats", net->proc_net);
6227 static struct pernet_operations ip6_route_net_ops = {
6228 .init = ip6_route_net_init,
6229 .exit = ip6_route_net_exit,
6232 static int __net_init ipv6_inetpeer_init(struct net *net)
6234 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6238 inet_peer_base_init(bp);
6239 net->ipv6.peers = bp;
6243 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6245 struct inet_peer_base *bp = net->ipv6.peers;
6247 net->ipv6.peers = NULL;
6248 inetpeer_invalidate_tree(bp);
6252 static struct pernet_operations ipv6_inetpeer_ops = {
6253 .init = ipv6_inetpeer_init,
6254 .exit = ipv6_inetpeer_exit,
6257 static struct pernet_operations ip6_route_net_late_ops = {
6258 .init = ip6_route_net_init_late,
6259 .exit = ip6_route_net_exit_late,
6262 static struct notifier_block ip6_route_dev_notifier = {
6263 .notifier_call = ip6_route_dev_notify,
6264 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6267 void __init ip6_route_init_special_entries(void)
6269 /* Registering of the loopback is done before this portion of code,
6270 * the loopback reference in rt6_info will not be taken, do it
6271 * manually for init_net */
6272 init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6273 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6274 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6275 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6276 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6277 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6278 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6279 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6283 int __init ip6_route_init(void)
6289 ip6_dst_ops_template.kmem_cachep =
6290 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6291 SLAB_HWCACHE_ALIGN, NULL);
6292 if (!ip6_dst_ops_template.kmem_cachep)
6295 ret = dst_entries_init(&ip6_dst_blackhole_ops);
6297 goto out_kmem_cache;
6299 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6301 goto out_dst_entries;
6303 ret = register_pernet_subsys(&ip6_route_net_ops);
6305 goto out_register_inetpeer;
6307 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6311 goto out_register_subsys;
6317 ret = fib6_rules_init();
6321 ret = register_pernet_subsys(&ip6_route_net_late_ops);
6323 goto fib6_rules_init;
6325 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6326 inet6_rtm_newroute, NULL, 0);
6328 goto out_register_late_subsys;
6330 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6331 inet6_rtm_delroute, NULL, 0);
6333 goto out_register_late_subsys;
6335 ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6336 inet6_rtm_getroute, NULL,
6337 RTNL_FLAG_DOIT_UNLOCKED);
6339 goto out_register_late_subsys;
6341 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6343 goto out_register_late_subsys;
6345 for_each_possible_cpu(cpu) {
6346 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6348 INIT_LIST_HEAD(&ul->head);
6349 spin_lock_init(&ul->lock);
6355 out_register_late_subsys:
6356 rtnl_unregister_all(PF_INET6);
6357 unregister_pernet_subsys(&ip6_route_net_late_ops);
6359 fib6_rules_cleanup();
6364 out_register_subsys:
6365 unregister_pernet_subsys(&ip6_route_net_ops);
6366 out_register_inetpeer:
6367 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6369 dst_entries_destroy(&ip6_dst_blackhole_ops);
6371 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6375 void ip6_route_cleanup(void)
6377 unregister_netdevice_notifier(&ip6_route_dev_notifier);
6378 unregister_pernet_subsys(&ip6_route_net_late_ops);
6379 fib6_rules_cleanup();
6382 unregister_pernet_subsys(&ipv6_inetpeer_ops);
6383 unregister_pernet_subsys(&ip6_route_net_ops);
6384 dst_entries_destroy(&ip6_dst_blackhole_ops);
6385 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);