2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
24 * Fixed routing subtrees.
27 #define pr_fmt(fmt) "IPv6: " fmt
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/export.h>
32 #include <linux/types.h>
33 #include <linux/times.h>
34 #include <linux/socket.h>
35 #include <linux/sockios.h>
36 #include <linux/net.h>
37 #include <linux/route.h>
38 #include <linux/netdevice.h>
39 #include <linux/in6.h>
40 #include <linux/mroute6.h>
41 #include <linux/init.h>
42 #include <linux/if_arp.h>
43 #include <linux/proc_fs.h>
44 #include <linux/seq_file.h>
45 #include <linux/nsproxy.h>
46 #include <linux/slab.h>
47 #include <linux/jhash.h>
48 #include <net/net_namespace.h>
51 #include <net/ip6_fib.h>
52 #include <net/ip6_route.h>
53 #include <net/ndisc.h>
54 #include <net/addrconf.h>
56 #include <linux/rtnetlink.h>
58 #include <net/dst_metadata.h>
60 #include <net/netevent.h>
61 #include <net/netlink.h>
62 #include <net/nexthop.h>
63 #include <net/lwtunnel.h>
64 #include <net/ip_tunnels.h>
65 #include <net/l3mdev.h>
66 #include <trace/events/fib6.h>
68 #include <linux/uaccess.h>
71 #include <linux/sysctl.h>
75 RT6_NUD_FAIL_HARD = -3,
76 RT6_NUD_FAIL_PROBE = -2,
77 RT6_NUD_FAIL_DO_RR = -1,
81 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
82 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
83 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
84 static unsigned int ip6_mtu(const struct dst_entry *dst);
85 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
86 static void ip6_dst_destroy(struct dst_entry *);
87 static void ip6_dst_ifdown(struct dst_entry *,
88 struct net_device *dev, int how);
89 static int ip6_dst_gc(struct dst_ops *ops);
91 static int ip6_pkt_discard(struct sk_buff *skb);
92 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
93 static int ip6_pkt_prohibit(struct sk_buff *skb);
94 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
95 static void ip6_link_failure(struct sk_buff *skb);
96 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
97 struct sk_buff *skb, u32 mtu);
98 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
100 static void rt6_dst_from_metrics_check(struct rt6_info *rt);
101 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
102 static size_t rt6_nlmsg_size(struct rt6_info *rt);
103 static int rt6_fill_node(struct net *net,
104 struct sk_buff *skb, struct rt6_info *rt,
105 struct in6_addr *dst, struct in6_addr *src,
106 int iif, int type, u32 portid, u32 seq,
108 static struct rt6_info *rt6_find_cached_rt(struct rt6_info *rt,
109 struct in6_addr *daddr,
110 struct in6_addr *saddr);
112 #ifdef CONFIG_IPV6_ROUTE_INFO
113 static struct rt6_info *rt6_add_route_info(struct net *net,
114 const struct in6_addr *prefix, int prefixlen,
115 const struct in6_addr *gwaddr,
116 struct net_device *dev,
118 static struct rt6_info *rt6_get_route_info(struct net *net,
119 const struct in6_addr *prefix, int prefixlen,
120 const struct in6_addr *gwaddr,
121 struct net_device *dev);
124 struct uncached_list {
126 struct list_head head;
129 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
131 static void rt6_uncached_list_add(struct rt6_info *rt)
133 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
135 rt->rt6i_uncached_list = ul;
137 spin_lock_bh(&ul->lock);
138 list_add_tail(&rt->rt6i_uncached, &ul->head);
139 spin_unlock_bh(&ul->lock);
142 static void rt6_uncached_list_del(struct rt6_info *rt)
144 if (!list_empty(&rt->rt6i_uncached)) {
145 struct uncached_list *ul = rt->rt6i_uncached_list;
147 spin_lock_bh(&ul->lock);
148 list_del(&rt->rt6i_uncached);
149 spin_unlock_bh(&ul->lock);
153 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
155 struct net_device *loopback_dev = net->loopback_dev;
158 if (dev == loopback_dev)
161 for_each_possible_cpu(cpu) {
162 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
165 spin_lock_bh(&ul->lock);
166 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
167 struct inet6_dev *rt_idev = rt->rt6i_idev;
168 struct net_device *rt_dev = rt->dst.dev;
170 if (rt_idev->dev == dev) {
171 rt->rt6i_idev = in6_dev_get(loopback_dev);
172 in6_dev_put(rt_idev);
176 rt->dst.dev = loopback_dev;
177 dev_hold(rt->dst.dev);
181 spin_unlock_bh(&ul->lock);
185 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
187 return dst_metrics_write_ptr(rt->dst.from);
190 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
192 struct rt6_info *rt = (struct rt6_info *)dst;
194 if (rt->rt6i_flags & RTF_PCPU)
195 return rt6_pcpu_cow_metrics(rt);
196 else if (rt->rt6i_flags & RTF_CACHE)
199 return dst_cow_metrics_generic(dst, old);
202 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
206 struct in6_addr *p = &rt->rt6i_gateway;
208 if (!ipv6_addr_any(p))
209 return (const void *) p;
211 return &ipv6_hdr(skb)->daddr;
215 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
219 struct rt6_info *rt = (struct rt6_info *) dst;
222 daddr = choose_neigh_daddr(rt, skb, daddr);
223 n = __ipv6_neigh_lookup(dst->dev, daddr);
226 return neigh_create(&nd_tbl, daddr, dst->dev);
229 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
231 struct net_device *dev = dst->dev;
232 struct rt6_info *rt = (struct rt6_info *)dst;
234 daddr = choose_neigh_daddr(rt, NULL, daddr);
237 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
239 if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
241 __ipv6_confirm_neigh(dev, daddr);
244 static struct dst_ops ip6_dst_ops_template = {
248 .check = ip6_dst_check,
249 .default_advmss = ip6_default_advmss,
251 .cow_metrics = ipv6_cow_metrics,
252 .destroy = ip6_dst_destroy,
253 .ifdown = ip6_dst_ifdown,
254 .negative_advice = ip6_negative_advice,
255 .link_failure = ip6_link_failure,
256 .update_pmtu = ip6_rt_update_pmtu,
257 .redirect = rt6_do_redirect,
258 .local_out = __ip6_local_out,
259 .neigh_lookup = ip6_neigh_lookup,
260 .confirm_neigh = ip6_confirm_neigh,
263 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
265 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
267 return mtu ? : dst->dev->mtu;
270 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
271 struct sk_buff *skb, u32 mtu)
275 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
280 static struct dst_ops ip6_dst_blackhole_ops = {
282 .destroy = ip6_dst_destroy,
283 .check = ip6_dst_check,
284 .mtu = ip6_blackhole_mtu,
285 .default_advmss = ip6_default_advmss,
286 .update_pmtu = ip6_rt_blackhole_update_pmtu,
287 .redirect = ip6_rt_blackhole_redirect,
288 .cow_metrics = dst_cow_metrics_generic,
289 .neigh_lookup = ip6_neigh_lookup,
292 static const u32 ip6_template_metrics[RTAX_MAX] = {
293 [RTAX_HOPLIMIT - 1] = 0,
296 static const struct rt6_info ip6_null_entry_template = {
298 .__refcnt = ATOMIC_INIT(1),
300 .obsolete = DST_OBSOLETE_FORCE_CHK,
301 .error = -ENETUNREACH,
302 .input = ip6_pkt_discard,
303 .output = ip6_pkt_discard_out,
305 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
306 .rt6i_protocol = RTPROT_KERNEL,
307 .rt6i_metric = ~(u32) 0,
308 .rt6i_ref = ATOMIC_INIT(1),
311 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
313 static const struct rt6_info ip6_prohibit_entry_template = {
315 .__refcnt = ATOMIC_INIT(1),
317 .obsolete = DST_OBSOLETE_FORCE_CHK,
319 .input = ip6_pkt_prohibit,
320 .output = ip6_pkt_prohibit_out,
322 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
323 .rt6i_protocol = RTPROT_KERNEL,
324 .rt6i_metric = ~(u32) 0,
325 .rt6i_ref = ATOMIC_INIT(1),
328 static const struct rt6_info ip6_blk_hole_entry_template = {
330 .__refcnt = ATOMIC_INIT(1),
332 .obsolete = DST_OBSOLETE_FORCE_CHK,
334 .input = dst_discard,
335 .output = dst_discard_out,
337 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
338 .rt6i_protocol = RTPROT_KERNEL,
339 .rt6i_metric = ~(u32) 0,
340 .rt6i_ref = ATOMIC_INIT(1),
345 static void rt6_info_init(struct rt6_info *rt)
347 struct dst_entry *dst = &rt->dst;
349 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
350 INIT_LIST_HEAD(&rt->rt6i_siblings);
351 INIT_LIST_HEAD(&rt->rt6i_uncached);
354 /* allocate dst with ip6_dst_ops */
355 static struct rt6_info *__ip6_dst_alloc(struct net *net,
356 struct net_device *dev,
359 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
360 1, DST_OBSOLETE_FORCE_CHK, flags);
368 struct rt6_info *ip6_dst_alloc(struct net *net,
369 struct net_device *dev,
372 struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
375 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
379 for_each_possible_cpu(cpu) {
382 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
383 /* no one shares rt */
387 dst_release_immediate(&rt->dst);
394 EXPORT_SYMBOL(ip6_dst_alloc);
396 static void ip6_dst_destroy(struct dst_entry *dst)
398 struct rt6_info *rt = (struct rt6_info *)dst;
399 struct rt6_exception_bucket *bucket;
400 struct dst_entry *from = dst->from;
401 struct inet6_dev *idev;
403 dst_destroy_metrics_generic(dst);
404 free_percpu(rt->rt6i_pcpu);
405 rt6_uncached_list_del(rt);
407 idev = rt->rt6i_idev;
409 rt->rt6i_idev = NULL;
412 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket, 1);
414 rt->rt6i_exception_bucket = NULL;
422 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
425 struct rt6_info *rt = (struct rt6_info *)dst;
426 struct inet6_dev *idev = rt->rt6i_idev;
427 struct net_device *loopback_dev =
428 dev_net(dev)->loopback_dev;
430 if (idev && idev->dev != loopback_dev) {
431 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
433 rt->rt6i_idev = loopback_idev;
439 static bool __rt6_check_expired(const struct rt6_info *rt)
441 if (rt->rt6i_flags & RTF_EXPIRES)
442 return time_after(jiffies, rt->dst.expires);
447 static bool rt6_check_expired(const struct rt6_info *rt)
449 if (rt->rt6i_flags & RTF_EXPIRES) {
450 if (time_after(jiffies, rt->dst.expires))
452 } else if (rt->dst.from) {
453 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
454 rt6_check_expired((struct rt6_info *)rt->dst.from);
459 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
460 struct flowi6 *fl6, int oif,
463 struct rt6_info *sibling, *next_sibling;
466 /* We might have already computed the hash for ICMPv6 errors. In such
467 * case it will always be non-zero. Otherwise now is the time to do it.
470 fl6->mp_hash = rt6_multipath_hash(fl6, NULL);
472 route_choosen = fl6->mp_hash % (match->rt6i_nsiblings + 1);
473 /* Don't change the route, if route_choosen == 0
474 * (siblings does not include ourself)
477 list_for_each_entry_safe(sibling, next_sibling,
478 &match->rt6i_siblings, rt6i_siblings) {
480 if (route_choosen == 0) {
481 if (rt6_score_route(sibling, oif, strict) < 0)
491 * Route lookup. rcu_read_lock() should be held.
494 static inline struct rt6_info *rt6_device_match(struct net *net,
496 const struct in6_addr *saddr,
500 struct rt6_info *local = NULL;
501 struct rt6_info *sprt;
503 if (!oif && ipv6_addr_any(saddr))
506 for (sprt = rt; sprt; sprt = rcu_dereference(sprt->dst.rt6_next)) {
507 struct net_device *dev = sprt->dst.dev;
510 if (dev->ifindex == oif)
512 if (dev->flags & IFF_LOOPBACK) {
513 if (!sprt->rt6i_idev ||
514 sprt->rt6i_idev->dev->ifindex != oif) {
515 if (flags & RT6_LOOKUP_F_IFACE)
518 local->rt6i_idev->dev->ifindex == oif)
524 if (ipv6_chk_addr(net, saddr, dev,
525 flags & RT6_LOOKUP_F_IFACE))
534 if (flags & RT6_LOOKUP_F_IFACE)
535 return net->ipv6.ip6_null_entry;
541 #ifdef CONFIG_IPV6_ROUTER_PREF
542 struct __rt6_probe_work {
543 struct work_struct work;
544 struct in6_addr target;
545 struct net_device *dev;
548 static void rt6_probe_deferred(struct work_struct *w)
550 struct in6_addr mcaddr;
551 struct __rt6_probe_work *work =
552 container_of(w, struct __rt6_probe_work, work);
554 addrconf_addr_solict_mult(&work->target, &mcaddr);
555 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
560 static void rt6_probe(struct rt6_info *rt)
562 struct __rt6_probe_work *work;
563 struct neighbour *neigh;
565 * Okay, this does not seem to be appropriate
566 * for now, however, we need to check if it
567 * is really so; aka Router Reachability Probing.
569 * Router Reachability Probe MUST be rate-limited
570 * to no more than one per minute.
572 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
575 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
577 if (neigh->nud_state & NUD_VALID)
581 write_lock(&neigh->lock);
582 if (!(neigh->nud_state & NUD_VALID) &&
585 rt->rt6i_idev->cnf.rtr_probe_interval)) {
586 work = kmalloc(sizeof(*work), GFP_ATOMIC);
588 __neigh_set_probe_once(neigh);
590 write_unlock(&neigh->lock);
592 work = kmalloc(sizeof(*work), GFP_ATOMIC);
596 INIT_WORK(&work->work, rt6_probe_deferred);
597 work->target = rt->rt6i_gateway;
598 dev_hold(rt->dst.dev);
599 work->dev = rt->dst.dev;
600 schedule_work(&work->work);
604 rcu_read_unlock_bh();
607 static inline void rt6_probe(struct rt6_info *rt)
613 * Default Router Selection (RFC 2461 6.3.6)
615 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
617 struct net_device *dev = rt->dst.dev;
618 if (!oif || dev->ifindex == oif)
620 if ((dev->flags & IFF_LOOPBACK) &&
621 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
626 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
628 struct neighbour *neigh;
629 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
631 if (rt->rt6i_flags & RTF_NONEXTHOP ||
632 !(rt->rt6i_flags & RTF_GATEWAY))
633 return RT6_NUD_SUCCEED;
636 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
638 read_lock(&neigh->lock);
639 if (neigh->nud_state & NUD_VALID)
640 ret = RT6_NUD_SUCCEED;
641 #ifdef CONFIG_IPV6_ROUTER_PREF
642 else if (!(neigh->nud_state & NUD_FAILED))
643 ret = RT6_NUD_SUCCEED;
645 ret = RT6_NUD_FAIL_PROBE;
647 read_unlock(&neigh->lock);
649 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
650 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
652 rcu_read_unlock_bh();
657 static int rt6_score_route(struct rt6_info *rt, int oif,
662 m = rt6_check_dev(rt, oif);
663 if (!m && (strict & RT6_LOOKUP_F_IFACE))
664 return RT6_NUD_FAIL_HARD;
665 #ifdef CONFIG_IPV6_ROUTER_PREF
666 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
668 if (strict & RT6_LOOKUP_F_REACHABLE) {
669 int n = rt6_check_neigh(rt);
676 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
677 int *mpri, struct rt6_info *match,
681 bool match_do_rr = false;
682 struct inet6_dev *idev = rt->rt6i_idev;
683 struct net_device *dev = rt->dst.dev;
685 if (dev && !netif_carrier_ok(dev) &&
686 idev->cnf.ignore_routes_with_linkdown &&
687 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
690 if (rt6_check_expired(rt))
693 m = rt6_score_route(rt, oif, strict);
694 if (m == RT6_NUD_FAIL_DO_RR) {
696 m = 0; /* lowest valid score */
697 } else if (m == RT6_NUD_FAIL_HARD) {
701 if (strict & RT6_LOOKUP_F_REACHABLE)
704 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
706 *do_rr = match_do_rr;
714 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
715 struct rt6_info *leaf,
716 struct rt6_info *rr_head,
717 u32 metric, int oif, int strict,
720 struct rt6_info *rt, *match, *cont;
725 for (rt = rr_head; rt; rt = rcu_dereference(rt->dst.rt6_next)) {
726 if (rt->rt6i_metric != metric) {
731 match = find_match(rt, oif, strict, &mpri, match, do_rr);
734 for (rt = leaf; rt && rt != rr_head;
735 rt = rcu_dereference(rt->dst.rt6_next)) {
736 if (rt->rt6i_metric != metric) {
741 match = find_match(rt, oif, strict, &mpri, match, do_rr);
747 for (rt = cont; rt; rt = rcu_dereference(rt->dst.rt6_next))
748 match = find_match(rt, oif, strict, &mpri, match, do_rr);
753 static struct rt6_info *rt6_select(struct net *net, struct fib6_node *fn,
756 struct rt6_info *leaf = rcu_dereference(fn->leaf);
757 struct rt6_info *match, *rt0;
762 return net->ipv6.ip6_null_entry;
764 rt0 = rcu_dereference(fn->rr_ptr);
768 /* Double check to make sure fn is not an intermediate node
769 * and fn->leaf does not points to its child's leaf
770 * (This might happen if all routes under fn are deleted from
771 * the tree and fib6_repair_tree() is called on the node.)
773 key_plen = rt0->rt6i_dst.plen;
774 #ifdef CONFIG_IPV6_SUBTREES
775 if (rt0->rt6i_src.plen)
776 key_plen = rt0->rt6i_src.plen;
778 if (fn->fn_bit != key_plen)
779 return net->ipv6.ip6_null_entry;
781 match = find_rr_leaf(fn, leaf, rt0, rt0->rt6i_metric, oif, strict,
785 struct rt6_info *next = rcu_dereference(rt0->dst.rt6_next);
787 /* no entries matched; do round-robin */
788 if (!next || next->rt6i_metric != rt0->rt6i_metric)
792 spin_lock_bh(&leaf->rt6i_table->tb6_lock);
793 /* make sure next is not being deleted from the tree */
795 rcu_assign_pointer(fn->rr_ptr, next);
796 spin_unlock_bh(&leaf->rt6i_table->tb6_lock);
800 return match ? match : net->ipv6.ip6_null_entry;
803 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
805 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
808 #ifdef CONFIG_IPV6_ROUTE_INFO
809 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
810 const struct in6_addr *gwaddr)
812 struct net *net = dev_net(dev);
813 struct route_info *rinfo = (struct route_info *) opt;
814 struct in6_addr prefix_buf, *prefix;
816 unsigned long lifetime;
819 if (len < sizeof(struct route_info)) {
823 /* Sanity check for prefix_len and length */
824 if (rinfo->length > 3) {
826 } else if (rinfo->prefix_len > 128) {
828 } else if (rinfo->prefix_len > 64) {
829 if (rinfo->length < 2) {
832 } else if (rinfo->prefix_len > 0) {
833 if (rinfo->length < 1) {
838 pref = rinfo->route_pref;
839 if (pref == ICMPV6_ROUTER_PREF_INVALID)
842 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
844 if (rinfo->length == 3)
845 prefix = (struct in6_addr *)rinfo->prefix;
847 /* this function is safe */
848 ipv6_addr_prefix(&prefix_buf,
849 (struct in6_addr *)rinfo->prefix,
851 prefix = &prefix_buf;
854 if (rinfo->prefix_len == 0)
855 rt = rt6_get_dflt_router(gwaddr, dev);
857 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
860 if (rt && !lifetime) {
866 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
869 rt->rt6i_flags = RTF_ROUTEINFO |
870 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
873 if (!addrconf_finite_timeout(lifetime))
874 rt6_clean_expires(rt);
876 rt6_set_expires(rt, jiffies + HZ * lifetime);
884 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
885 struct in6_addr *saddr)
887 struct fib6_node *pn, *sn;
889 if (fn->fn_flags & RTN_TL_ROOT)
891 pn = rcu_dereference(fn->parent);
892 sn = FIB6_SUBTREE(pn);
894 fn = fib6_lookup(sn, NULL, saddr);
897 if (fn->fn_flags & RTN_RTINFO)
902 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt,
905 struct rt6_info *rt = *prt;
907 if (dst_hold_safe(&rt->dst))
910 rt = net->ipv6.ip6_null_entry;
919 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
920 struct fib6_table *table,
921 struct flowi6 *fl6, int flags)
923 struct rt6_info *rt, *rt_cache;
924 struct fib6_node *fn;
927 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
929 rt = rcu_dereference(fn->leaf);
931 rt = net->ipv6.ip6_null_entry;
933 rt = rt6_device_match(net, rt, &fl6->saddr,
934 fl6->flowi6_oif, flags);
935 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
936 rt = rt6_multipath_select(rt, fl6,
937 fl6->flowi6_oif, flags);
939 if (rt == net->ipv6.ip6_null_entry) {
940 fn = fib6_backtrack(fn, &fl6->saddr);
944 /* Search through exception table */
945 rt_cache = rt6_find_cached_rt(rt, &fl6->daddr, &fl6->saddr);
949 if (ip6_hold_safe(net, &rt, true))
950 dst_use_noref(&rt->dst, jiffies);
954 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
960 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
963 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
965 EXPORT_SYMBOL_GPL(ip6_route_lookup);
967 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
968 const struct in6_addr *saddr, int oif, int strict)
970 struct flowi6 fl6 = {
974 struct dst_entry *dst;
975 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
978 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
979 flags |= RT6_LOOKUP_F_HAS_SADDR;
982 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
984 return (struct rt6_info *) dst;
990 EXPORT_SYMBOL(rt6_lookup);
992 /* ip6_ins_rt is called with FREE table->tb6_lock.
993 * It takes new route entry, the addition fails by any reason the
995 * Caller must hold dst before calling it.
998 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
999 struct mx6_config *mxc,
1000 struct netlink_ext_ack *extack)
1003 struct fib6_table *table;
1005 table = rt->rt6i_table;
1006 spin_lock_bh(&table->tb6_lock);
1007 err = fib6_add(&table->tb6_root, rt, info, mxc, extack);
1008 spin_unlock_bh(&table->tb6_lock);
1013 int ip6_ins_rt(struct rt6_info *rt)
1015 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
1016 struct mx6_config mxc = { .mx = NULL, };
1018 /* Hold dst to account for the reference from the fib6 tree */
1020 return __ip6_ins_rt(rt, &info, &mxc, NULL);
1023 /* called with rcu_lock held */
1024 static struct net_device *ip6_rt_get_dev_rcu(struct rt6_info *rt)
1026 struct net_device *dev = rt->dst.dev;
1028 if (rt->rt6i_flags & RTF_LOCAL) {
1029 /* for copies of local routes, dst->dev needs to be the
1030 * device if it is a master device, the master device if
1031 * device is enslaved, and the loopback as the default
1033 if (netif_is_l3_slave(dev) &&
1034 !rt6_need_strict(&rt->rt6i_dst.addr))
1035 dev = l3mdev_master_dev_rcu(dev);
1036 else if (!netif_is_l3_master(dev))
1037 dev = dev_net(dev)->loopback_dev;
1038 /* last case is netif_is_l3_master(dev) is true in which
1039 * case we want dev returned to be dev
1046 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
1047 const struct in6_addr *daddr,
1048 const struct in6_addr *saddr)
1050 struct net_device *dev;
1051 struct rt6_info *rt;
1057 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
1058 ort = (struct rt6_info *)ort->dst.from;
1061 dev = ip6_rt_get_dev_rcu(ort);
1062 rt = __ip6_dst_alloc(dev_net(dev), dev, 0);
1067 ip6_rt_copy_init(rt, ort);
1068 rt->rt6i_flags |= RTF_CACHE;
1069 rt->rt6i_metric = 0;
1070 rt->dst.flags |= DST_HOST;
1071 rt->rt6i_dst.addr = *daddr;
1072 rt->rt6i_dst.plen = 128;
1074 if (!rt6_is_gw_or_nonexthop(ort)) {
1075 if (ort->rt6i_dst.plen != 128 &&
1076 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
1077 rt->rt6i_flags |= RTF_ANYCAST;
1078 #ifdef CONFIG_IPV6_SUBTREES
1079 if (rt->rt6i_src.plen && saddr) {
1080 rt->rt6i_src.addr = *saddr;
1081 rt->rt6i_src.plen = 128;
1089 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
1091 struct net_device *dev;
1092 struct rt6_info *pcpu_rt;
1095 dev = ip6_rt_get_dev_rcu(rt);
1096 pcpu_rt = __ip6_dst_alloc(dev_net(dev), dev, rt->dst.flags);
1100 ip6_rt_copy_init(pcpu_rt, rt);
1101 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
1102 pcpu_rt->rt6i_flags |= RTF_PCPU;
1106 /* It should be called with rcu_read_lock() acquired */
1107 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
1109 struct rt6_info *pcpu_rt, **p;
1111 p = this_cpu_ptr(rt->rt6i_pcpu);
1114 if (pcpu_rt && ip6_hold_safe(NULL, &pcpu_rt, false))
1115 rt6_dst_from_metrics_check(pcpu_rt);
1120 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1122 struct rt6_info *pcpu_rt, *prev, **p;
1124 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1126 struct net *net = dev_net(rt->dst.dev);
1128 dst_hold(&net->ipv6.ip6_null_entry->dst);
1129 return net->ipv6.ip6_null_entry;
1132 dst_hold(&pcpu_rt->dst);
1133 p = this_cpu_ptr(rt->rt6i_pcpu);
1134 prev = cmpxchg(p, NULL, pcpu_rt);
1136 /* If someone did it before us, return prev instead */
1137 /* release refcnt taken by ip6_rt_pcpu_alloc() */
1138 dst_release_immediate(&pcpu_rt->dst);
1139 /* release refcnt taken by above dst_hold() */
1140 dst_release_immediate(&pcpu_rt->dst);
1141 dst_hold(&prev->dst);
1145 rt6_dst_from_metrics_check(pcpu_rt);
1149 /* exception hash table implementation
1151 static DEFINE_SPINLOCK(rt6_exception_lock);
1153 /* Remove rt6_ex from hash table and free the memory
1154 * Caller must hold rt6_exception_lock
1156 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1157 struct rt6_exception *rt6_ex)
1159 if (!bucket || !rt6_ex)
1161 rt6_ex->rt6i->rt6i_node = NULL;
1162 hlist_del_rcu(&rt6_ex->hlist);
1163 rt6_release(rt6_ex->rt6i);
1164 kfree_rcu(rt6_ex, rcu);
1165 WARN_ON_ONCE(!bucket->depth);
1169 /* Remove oldest rt6_ex in bucket and free the memory
1170 * Caller must hold rt6_exception_lock
1172 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1174 struct rt6_exception *rt6_ex, *oldest = NULL;
1179 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1180 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1183 rt6_remove_exception(bucket, oldest);
1186 static u32 rt6_exception_hash(const struct in6_addr *dst,
1187 const struct in6_addr *src)
1189 static u32 seed __read_mostly;
1192 net_get_random_once(&seed, sizeof(seed));
1193 val = jhash(dst, sizeof(*dst), seed);
1195 #ifdef CONFIG_IPV6_SUBTREES
1197 val = jhash(src, sizeof(*src), val);
1199 return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1202 /* Helper function to find the cached rt in the hash table
1203 * and update bucket pointer to point to the bucket for this
1204 * (daddr, saddr) pair
1205 * Caller must hold rt6_exception_lock
1207 static struct rt6_exception *
1208 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1209 const struct in6_addr *daddr,
1210 const struct in6_addr *saddr)
1212 struct rt6_exception *rt6_ex;
1215 if (!(*bucket) || !daddr)
1218 hval = rt6_exception_hash(daddr, saddr);
1221 hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1222 struct rt6_info *rt6 = rt6_ex->rt6i;
1223 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1225 #ifdef CONFIG_IPV6_SUBTREES
1226 if (matched && saddr)
1227 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1235 /* Helper function to find the cached rt in the hash table
1236 * and update bucket pointer to point to the bucket for this
1237 * (daddr, saddr) pair
1238 * Caller must hold rcu_read_lock()
1240 static struct rt6_exception *
1241 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1242 const struct in6_addr *daddr,
1243 const struct in6_addr *saddr)
1245 struct rt6_exception *rt6_ex;
1248 WARN_ON_ONCE(!rcu_read_lock_held());
1250 if (!(*bucket) || !daddr)
1253 hval = rt6_exception_hash(daddr, saddr);
1256 hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1257 struct rt6_info *rt6 = rt6_ex->rt6i;
1258 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1260 #ifdef CONFIG_IPV6_SUBTREES
1261 if (matched && saddr)
1262 matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1270 static int rt6_insert_exception(struct rt6_info *nrt,
1271 struct rt6_info *ort)
1273 struct rt6_exception_bucket *bucket;
1274 struct in6_addr *src_key = NULL;
1275 struct rt6_exception *rt6_ex;
1278 /* ort can't be a cache or pcpu route */
1279 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
1280 ort = (struct rt6_info *)ort->dst.from;
1281 WARN_ON_ONCE(ort->rt6i_flags & (RTF_CACHE | RTF_PCPU));
1283 spin_lock_bh(&rt6_exception_lock);
1285 if (ort->exception_bucket_flushed) {
1290 bucket = rcu_dereference_protected(ort->rt6i_exception_bucket,
1291 lockdep_is_held(&rt6_exception_lock));
1293 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1299 rcu_assign_pointer(ort->rt6i_exception_bucket, bucket);
1302 #ifdef CONFIG_IPV6_SUBTREES
1303 /* rt6i_src.plen != 0 indicates ort is in subtree
1304 * and exception table is indexed by a hash of
1305 * both rt6i_dst and rt6i_src.
1306 * Otherwise, the exception table is indexed by
1307 * a hash of only rt6i_dst.
1309 if (ort->rt6i_src.plen)
1310 src_key = &nrt->rt6i_src.addr;
1313 /* Update rt6i_prefsrc as it could be changed
1314 * in rt6_remove_prefsrc()
1316 nrt->rt6i_prefsrc = ort->rt6i_prefsrc;
1317 /* rt6_mtu_change() might lower mtu on ort.
1318 * Only insert this exception route if its mtu
1319 * is less than ort's mtu value.
1321 if (nrt->rt6i_pmtu >= dst_mtu(&ort->dst)) {
1326 rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1329 rt6_remove_exception(bucket, rt6_ex);
1331 rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1337 rt6_ex->stamp = jiffies;
1338 atomic_inc(&nrt->rt6i_ref);
1339 nrt->rt6i_node = ort->rt6i_node;
1340 hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1343 if (bucket->depth > FIB6_MAX_DEPTH)
1344 rt6_exception_remove_oldest(bucket);
1347 spin_unlock_bh(&rt6_exception_lock);
1349 /* Update fn->fn_sernum to invalidate all cached dst */
1351 fib6_update_sernum(ort);
1356 void rt6_flush_exceptions(struct rt6_info *rt)
1358 struct rt6_exception_bucket *bucket;
1359 struct rt6_exception *rt6_ex;
1360 struct hlist_node *tmp;
1363 spin_lock_bh(&rt6_exception_lock);
1364 /* Prevent rt6_insert_exception() to recreate the bucket list */
1365 rt->exception_bucket_flushed = 1;
1367 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1368 lockdep_is_held(&rt6_exception_lock));
1372 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1373 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist)
1374 rt6_remove_exception(bucket, rt6_ex);
1375 WARN_ON_ONCE(bucket->depth);
1380 spin_unlock_bh(&rt6_exception_lock);
1383 /* Find cached rt in the hash table inside passed in rt
1384 * Caller has to hold rcu_read_lock()
1386 static struct rt6_info *rt6_find_cached_rt(struct rt6_info *rt,
1387 struct in6_addr *daddr,
1388 struct in6_addr *saddr)
1390 struct rt6_exception_bucket *bucket;
1391 struct in6_addr *src_key = NULL;
1392 struct rt6_exception *rt6_ex;
1393 struct rt6_info *res = NULL;
1395 bucket = rcu_dereference(rt->rt6i_exception_bucket);
1397 #ifdef CONFIG_IPV6_SUBTREES
1398 /* rt6i_src.plen != 0 indicates rt is in subtree
1399 * and exception table is indexed by a hash of
1400 * both rt6i_dst and rt6i_src.
1401 * Otherwise, the exception table is indexed by
1402 * a hash of only rt6i_dst.
1404 if (rt->rt6i_src.plen)
1407 rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1409 if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1415 /* Remove the passed in cached rt from the hash table that contains it */
1416 int rt6_remove_exception_rt(struct rt6_info *rt)
1418 struct rt6_info *from = (struct rt6_info *)rt->dst.from;
1419 struct rt6_exception_bucket *bucket;
1420 struct in6_addr *src_key = NULL;
1421 struct rt6_exception *rt6_ex;
1425 !(rt->rt6i_flags | RTF_CACHE))
1428 if (!rcu_access_pointer(from->rt6i_exception_bucket))
1431 spin_lock_bh(&rt6_exception_lock);
1432 bucket = rcu_dereference_protected(from->rt6i_exception_bucket,
1433 lockdep_is_held(&rt6_exception_lock));
1434 #ifdef CONFIG_IPV6_SUBTREES
1435 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1436 * and exception table is indexed by a hash of
1437 * both rt6i_dst and rt6i_src.
1438 * Otherwise, the exception table is indexed by
1439 * a hash of only rt6i_dst.
1441 if (from->rt6i_src.plen)
1442 src_key = &rt->rt6i_src.addr;
1444 rt6_ex = __rt6_find_exception_spinlock(&bucket,
1448 rt6_remove_exception(bucket, rt6_ex);
1454 spin_unlock_bh(&rt6_exception_lock);
1458 /* Find rt6_ex which contains the passed in rt cache and
1461 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1463 struct rt6_info *from = (struct rt6_info *)rt->dst.from;
1464 struct rt6_exception_bucket *bucket;
1465 struct in6_addr *src_key = NULL;
1466 struct rt6_exception *rt6_ex;
1469 !(rt->rt6i_flags | RTF_CACHE))
1473 bucket = rcu_dereference(from->rt6i_exception_bucket);
1475 #ifdef CONFIG_IPV6_SUBTREES
1476 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1477 * and exception table is indexed by a hash of
1478 * both rt6i_dst and rt6i_src.
1479 * Otherwise, the exception table is indexed by
1480 * a hash of only rt6i_dst.
1482 if (from->rt6i_src.plen)
1483 src_key = &rt->rt6i_src.addr;
1485 rt6_ex = __rt6_find_exception_rcu(&bucket,
1489 rt6_ex->stamp = jiffies;
1494 static void rt6_exceptions_remove_prefsrc(struct rt6_info *rt)
1496 struct rt6_exception_bucket *bucket;
1497 struct rt6_exception *rt6_ex;
1500 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1501 lockdep_is_held(&rt6_exception_lock));
1504 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1505 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1506 rt6_ex->rt6i->rt6i_prefsrc.plen = 0;
1513 static void rt6_exceptions_update_pmtu(struct rt6_info *rt, int mtu)
1515 struct rt6_exception_bucket *bucket;
1516 struct rt6_exception *rt6_ex;
1519 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1520 lockdep_is_held(&rt6_exception_lock));
1523 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1524 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1525 struct rt6_info *entry = rt6_ex->rt6i;
1526 /* For RTF_CACHE with rt6i_pmtu == 0
1527 * (i.e. a redirected route),
1528 * the metrics of its rt->dst.from has already
1531 if (entry->rt6i_pmtu && entry->rt6i_pmtu > mtu)
1532 entry->rt6i_pmtu = mtu;
1539 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
1541 static void rt6_exceptions_clean_tohost(struct rt6_info *rt,
1542 struct in6_addr *gateway)
1544 struct rt6_exception_bucket *bucket;
1545 struct rt6_exception *rt6_ex;
1546 struct hlist_node *tmp;
1549 if (!rcu_access_pointer(rt->rt6i_exception_bucket))
1552 spin_lock_bh(&rt6_exception_lock);
1553 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1554 lockdep_is_held(&rt6_exception_lock));
1557 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1558 hlist_for_each_entry_safe(rt6_ex, tmp,
1559 &bucket->chain, hlist) {
1560 struct rt6_info *entry = rt6_ex->rt6i;
1562 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
1563 RTF_CACHE_GATEWAY &&
1564 ipv6_addr_equal(gateway,
1565 &entry->rt6i_gateway)) {
1566 rt6_remove_exception(bucket, rt6_ex);
1573 spin_unlock_bh(&rt6_exception_lock);
1576 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
1577 struct rt6_exception *rt6_ex,
1578 struct fib6_gc_args *gc_args,
1581 struct rt6_info *rt = rt6_ex->rt6i;
1583 if (atomic_read(&rt->dst.__refcnt) == 1 &&
1584 time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
1585 RT6_TRACE("aging clone %p\n", rt);
1586 rt6_remove_exception(bucket, rt6_ex);
1588 } else if (rt->rt6i_flags & RTF_GATEWAY) {
1589 struct neighbour *neigh;
1590 __u8 neigh_flags = 0;
1592 neigh = dst_neigh_lookup(&rt->dst, &rt->rt6i_gateway);
1594 neigh_flags = neigh->flags;
1595 neigh_release(neigh);
1597 if (!(neigh_flags & NTF_ROUTER)) {
1598 RT6_TRACE("purging route %p via non-router but gateway\n",
1600 rt6_remove_exception(bucket, rt6_ex);
1607 void rt6_age_exceptions(struct rt6_info *rt,
1608 struct fib6_gc_args *gc_args,
1611 struct rt6_exception_bucket *bucket;
1612 struct rt6_exception *rt6_ex;
1613 struct hlist_node *tmp;
1616 if (!rcu_access_pointer(rt->rt6i_exception_bucket))
1619 spin_lock_bh(&rt6_exception_lock);
1620 bucket = rcu_dereference_protected(rt->rt6i_exception_bucket,
1621 lockdep_is_held(&rt6_exception_lock));
1624 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1625 hlist_for_each_entry_safe(rt6_ex, tmp,
1626 &bucket->chain, hlist) {
1627 rt6_age_examine_exception(bucket, rt6_ex,
1633 spin_unlock_bh(&rt6_exception_lock);
1636 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1637 int oif, struct flowi6 *fl6, int flags)
1639 struct fib6_node *fn, *saved_fn;
1640 struct rt6_info *rt, *rt_cache;
1643 strict |= flags & RT6_LOOKUP_F_IFACE;
1644 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
1645 if (net->ipv6.devconf_all->forwarding == 0)
1646 strict |= RT6_LOOKUP_F_REACHABLE;
1650 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1653 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1657 rt = rt6_select(net, fn, oif, strict);
1658 if (rt->rt6i_nsiblings)
1659 rt = rt6_multipath_select(rt, fl6, oif, strict);
1660 if (rt == net->ipv6.ip6_null_entry) {
1661 fn = fib6_backtrack(fn, &fl6->saddr);
1663 goto redo_rt6_select;
1664 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1665 /* also consider unreachable route */
1666 strict &= ~RT6_LOOKUP_F_REACHABLE;
1668 goto redo_rt6_select;
1672 /*Search through exception table */
1673 rt_cache = rt6_find_cached_rt(rt, &fl6->daddr, &fl6->saddr);
1677 if (rt == net->ipv6.ip6_null_entry) {
1680 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1682 } else if (rt->rt6i_flags & RTF_CACHE) {
1683 if (ip6_hold_safe(net, &rt, true)) {
1684 dst_use_noref(&rt->dst, jiffies);
1685 rt6_dst_from_metrics_check(rt);
1688 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1690 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1691 !(rt->rt6i_flags & RTF_GATEWAY))) {
1692 /* Create a RTF_CACHE clone which will not be
1693 * owned by the fib6 tree. It is for the special case where
1694 * the daddr in the skb during the neighbor look-up is different
1695 * from the fl6->daddr used to look-up route here.
1698 struct rt6_info *uncached_rt;
1700 if (ip6_hold_safe(net, &rt, true)) {
1701 dst_use_noref(&rt->dst, jiffies);
1705 goto uncached_rt_out;
1709 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1710 dst_release(&rt->dst);
1713 /* Uncached_rt's refcnt is taken during ip6_rt_cache_alloc()
1714 * No need for another dst_hold()
1716 rt6_uncached_list_add(uncached_rt);
1718 uncached_rt = net->ipv6.ip6_null_entry;
1719 dst_hold(&uncached_rt->dst);
1723 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1727 /* Get a percpu copy */
1729 struct rt6_info *pcpu_rt;
1731 dst_use_noref(&rt->dst, jiffies);
1732 pcpu_rt = rt6_get_pcpu_route(rt);
1737 /* atomic_inc_not_zero() is needed when using rcu */
1738 if (atomic_inc_not_zero(&rt->rt6i_ref)) {
1739 /* We have to do the read_unlock first
1740 * because rt6_make_pcpu_route() may trigger
1741 * ip6_dst_gc() which will take the write_lock.
1743 * No dst_hold() on rt is needed because grabbing
1744 * rt->rt6i_ref makes sure rt can't be released.
1747 pcpu_rt = rt6_make_pcpu_route(rt);
1750 /* rt is already removed from tree */
1752 pcpu_rt = net->ipv6.ip6_null_entry;
1753 dst_hold(&pcpu_rt->dst);
1757 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1761 EXPORT_SYMBOL_GPL(ip6_pol_route);
1763 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1764 struct flowi6 *fl6, int flags)
1766 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1769 struct dst_entry *ip6_route_input_lookup(struct net *net,
1770 struct net_device *dev,
1771 struct flowi6 *fl6, int flags)
1773 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1774 flags |= RT6_LOOKUP_F_IFACE;
1776 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1778 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
1780 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
1781 struct flow_keys *keys)
1783 const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
1784 const struct ipv6hdr *key_iph = outer_iph;
1785 const struct ipv6hdr *inner_iph;
1786 const struct icmp6hdr *icmph;
1787 struct ipv6hdr _inner_iph;
1789 if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
1792 icmph = icmp6_hdr(skb);
1793 if (icmph->icmp6_type != ICMPV6_DEST_UNREACH &&
1794 icmph->icmp6_type != ICMPV6_PKT_TOOBIG &&
1795 icmph->icmp6_type != ICMPV6_TIME_EXCEED &&
1796 icmph->icmp6_type != ICMPV6_PARAMPROB)
1799 inner_iph = skb_header_pointer(skb,
1800 skb_transport_offset(skb) + sizeof(*icmph),
1801 sizeof(_inner_iph), &_inner_iph);
1805 key_iph = inner_iph;
1807 memset(keys, 0, sizeof(*keys));
1808 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1809 keys->addrs.v6addrs.src = key_iph->saddr;
1810 keys->addrs.v6addrs.dst = key_iph->daddr;
1811 keys->tags.flow_label = ip6_flowinfo(key_iph);
1812 keys->basic.ip_proto = key_iph->nexthdr;
1815 /* if skb is set it will be used and fl6 can be NULL */
1816 u32 rt6_multipath_hash(const struct flowi6 *fl6, const struct sk_buff *skb)
1818 struct flow_keys hash_keys;
1821 ip6_multipath_l3_keys(skb, &hash_keys);
1822 return flow_hash_from_keys(&hash_keys);
1825 return get_hash_from_flowi6(fl6);
1828 void ip6_route_input(struct sk_buff *skb)
1830 const struct ipv6hdr *iph = ipv6_hdr(skb);
1831 struct net *net = dev_net(skb->dev);
1832 int flags = RT6_LOOKUP_F_HAS_SADDR;
1833 struct ip_tunnel_info *tun_info;
1834 struct flowi6 fl6 = {
1835 .flowi6_iif = skb->dev->ifindex,
1836 .daddr = iph->daddr,
1837 .saddr = iph->saddr,
1838 .flowlabel = ip6_flowinfo(iph),
1839 .flowi6_mark = skb->mark,
1840 .flowi6_proto = iph->nexthdr,
1843 tun_info = skb_tunnel_info(skb);
1844 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1845 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1846 if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
1847 fl6.mp_hash = rt6_multipath_hash(&fl6, skb);
1849 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1852 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1853 struct flowi6 *fl6, int flags)
1855 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1858 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1859 struct flowi6 *fl6, int flags)
1863 if (rt6_need_strict(&fl6->daddr)) {
1864 struct dst_entry *dst;
1866 dst = l3mdev_link_scope_lookup(net, fl6);
1871 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1873 any_src = ipv6_addr_any(&fl6->saddr);
1874 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1875 (fl6->flowi6_oif && any_src))
1876 flags |= RT6_LOOKUP_F_IFACE;
1879 flags |= RT6_LOOKUP_F_HAS_SADDR;
1881 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1883 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1885 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1887 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1889 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1890 struct net_device *loopback_dev = net->loopback_dev;
1891 struct dst_entry *new = NULL;
1893 rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
1894 DST_OBSOLETE_NONE, 0);
1900 new->input = dst_discard;
1901 new->output = dst_discard_out;
1903 dst_copy_metrics(new, &ort->dst);
1905 rt->rt6i_idev = in6_dev_get(loopback_dev);
1906 rt->rt6i_gateway = ort->rt6i_gateway;
1907 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1908 rt->rt6i_metric = 0;
1910 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1911 #ifdef CONFIG_IPV6_SUBTREES
1912 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1916 dst_release(dst_orig);
1917 return new ? new : ERR_PTR(-ENOMEM);
1921 * Destination cache support functions
1924 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1927 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1928 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1931 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1935 if (!rt6_get_cookie_safe(rt, &rt_cookie) || rt_cookie != cookie)
1938 if (rt6_check_expired(rt))
1944 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1946 if (!__rt6_check_expired(rt) &&
1947 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1948 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1954 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1956 struct rt6_info *rt;
1958 rt = (struct rt6_info *) dst;
1960 /* All IPV6 dsts are created with ->obsolete set to the value
1961 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1962 * into this function always.
1965 rt6_dst_from_metrics_check(rt);
1967 if (rt->rt6i_flags & RTF_PCPU ||
1968 (unlikely(!list_empty(&rt->rt6i_uncached)) && rt->dst.from))
1969 return rt6_dst_from_check(rt, cookie);
1971 return rt6_check(rt, cookie);
1974 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1976 struct rt6_info *rt = (struct rt6_info *) dst;
1979 if (rt->rt6i_flags & RTF_CACHE) {
1980 if (rt6_check_expired(rt)) {
1992 static void ip6_link_failure(struct sk_buff *skb)
1994 struct rt6_info *rt;
1996 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1998 rt = (struct rt6_info *) skb_dst(skb);
2000 if (rt->rt6i_flags & RTF_CACHE) {
2001 if (dst_hold_safe(&rt->dst))
2004 struct fib6_node *fn;
2007 fn = rcu_dereference(rt->rt6i_node);
2008 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2015 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2017 struct net *net = dev_net(rt->dst.dev);
2019 rt->rt6i_flags |= RTF_MODIFIED;
2020 rt->rt6i_pmtu = mtu;
2021 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2024 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2026 return !(rt->rt6i_flags & RTF_CACHE) &&
2027 (rt->rt6i_flags & RTF_PCPU ||
2028 rcu_access_pointer(rt->rt6i_node));
2031 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2032 const struct ipv6hdr *iph, u32 mtu)
2034 const struct in6_addr *daddr, *saddr;
2035 struct rt6_info *rt6 = (struct rt6_info *)dst;
2037 if (rt6->rt6i_flags & RTF_LOCAL)
2040 if (dst_metric_locked(dst, RTAX_MTU))
2044 daddr = &iph->daddr;
2045 saddr = &iph->saddr;
2047 daddr = &sk->sk_v6_daddr;
2048 saddr = &inet6_sk(sk)->saddr;
2053 dst_confirm_neigh(dst, daddr);
2054 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2055 if (mtu >= dst_mtu(dst))
2058 if (!rt6_cache_allowed_for_pmtu(rt6)) {
2059 rt6_do_update_pmtu(rt6, mtu);
2060 /* update rt6_ex->stamp for cache */
2061 if (rt6->rt6i_flags & RTF_CACHE)
2062 rt6_update_exception_stamp_rt(rt6);
2064 struct rt6_info *nrt6;
2066 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
2068 rt6_do_update_pmtu(nrt6, mtu);
2069 if (rt6_insert_exception(nrt6, rt6))
2070 dst_release_immediate(&nrt6->dst);
2075 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2076 struct sk_buff *skb, u32 mtu)
2078 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
2081 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2082 int oif, u32 mark, kuid_t uid)
2084 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2085 struct dst_entry *dst;
2088 memset(&fl6, 0, sizeof(fl6));
2089 fl6.flowi6_oif = oif;
2090 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
2091 fl6.daddr = iph->daddr;
2092 fl6.saddr = iph->saddr;
2093 fl6.flowlabel = ip6_flowinfo(iph);
2094 fl6.flowi6_uid = uid;
2096 dst = ip6_route_output(net, NULL, &fl6);
2098 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
2101 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2103 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2105 struct dst_entry *dst;
2107 ip6_update_pmtu(skb, sock_net(sk), mtu,
2108 sk->sk_bound_dev_if, sk->sk_mark, sk->sk_uid);
2110 dst = __sk_dst_get(sk);
2111 if (!dst || !dst->obsolete ||
2112 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2116 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2117 ip6_datagram_dst_update(sk, false);
2120 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2122 /* Handle redirects */
2123 struct ip6rd_flowi {
2125 struct in6_addr gateway;
2128 static struct rt6_info *__ip6_route_redirect(struct net *net,
2129 struct fib6_table *table,
2133 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
2134 struct rt6_info *rt, *rt_cache;
2135 struct fib6_node *fn;
2137 /* Get the "current" route for this destination and
2138 * check if the redirect has come from appropriate router.
2140 * RFC 4861 specifies that redirects should only be
2141 * accepted if they come from the nexthop to the target.
2142 * Due to the way the routes are chosen, this notion
2143 * is a bit fuzzy and one might need to check all possible
2148 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2150 for_each_fib6_node_rt_rcu(fn) {
2151 if (rt6_check_expired(rt))
2155 if (!(rt->rt6i_flags & RTF_GATEWAY))
2157 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
2159 /* rt_cache's gateway might be different from its 'parent'
2160 * in the case of an ip redirect.
2161 * So we keep searching in the exception table if the gateway
2164 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway)) {
2165 rt_cache = rt6_find_cached_rt(rt,
2169 ipv6_addr_equal(&rdfl->gateway,
2170 &rt_cache->rt6i_gateway)) {
2180 rt = net->ipv6.ip6_null_entry;
2181 else if (rt->dst.error) {
2182 rt = net->ipv6.ip6_null_entry;
2186 if (rt == net->ipv6.ip6_null_entry) {
2187 fn = fib6_backtrack(fn, &fl6->saddr);
2193 ip6_hold_safe(net, &rt, true);
2197 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
2201 static struct dst_entry *ip6_route_redirect(struct net *net,
2202 const struct flowi6 *fl6,
2203 const struct in6_addr *gateway)
2205 int flags = RT6_LOOKUP_F_HAS_SADDR;
2206 struct ip6rd_flowi rdfl;
2209 rdfl.gateway = *gateway;
2211 return fib6_rule_lookup(net, &rdfl.fl6,
2212 flags, __ip6_route_redirect);
2215 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
2218 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2219 struct dst_entry *dst;
2222 memset(&fl6, 0, sizeof(fl6));
2223 fl6.flowi6_iif = LOOPBACK_IFINDEX;
2224 fl6.flowi6_oif = oif;
2225 fl6.flowi6_mark = mark;
2226 fl6.daddr = iph->daddr;
2227 fl6.saddr = iph->saddr;
2228 fl6.flowlabel = ip6_flowinfo(iph);
2229 fl6.flowi6_uid = uid;
2231 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
2232 rt6_do_redirect(dst, NULL, skb);
2235 EXPORT_SYMBOL_GPL(ip6_redirect);
2237 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
2240 const struct ipv6hdr *iph = ipv6_hdr(skb);
2241 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
2242 struct dst_entry *dst;
2245 memset(&fl6, 0, sizeof(fl6));
2246 fl6.flowi6_iif = LOOPBACK_IFINDEX;
2247 fl6.flowi6_oif = oif;
2248 fl6.flowi6_mark = mark;
2249 fl6.daddr = msg->dest;
2250 fl6.saddr = iph->daddr;
2251 fl6.flowi6_uid = sock_net_uid(net, NULL);
2253 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
2254 rt6_do_redirect(dst, NULL, skb);
2258 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
2260 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
2263 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
2265 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
2267 struct net_device *dev = dst->dev;
2268 unsigned int mtu = dst_mtu(dst);
2269 struct net *net = dev_net(dev);
2271 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
2273 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
2274 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
2277 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
2278 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
2279 * IPV6_MAXPLEN is also valid and means: "any MSS,
2280 * rely only on pmtu discovery"
2282 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
2287 static unsigned int ip6_mtu(const struct dst_entry *dst)
2289 const struct rt6_info *rt = (const struct rt6_info *)dst;
2290 unsigned int mtu = rt->rt6i_pmtu;
2291 struct inet6_dev *idev;
2296 mtu = dst_metric_raw(dst, RTAX_MTU);
2303 idev = __in6_dev_get(dst->dev);
2305 mtu = idev->cnf.mtu6;
2309 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
2311 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
2314 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
2317 struct dst_entry *dst;
2318 struct rt6_info *rt;
2319 struct inet6_dev *idev = in6_dev_get(dev);
2320 struct net *net = dev_net(dev);
2322 if (unlikely(!idev))
2323 return ERR_PTR(-ENODEV);
2325 rt = ip6_dst_alloc(net, dev, 0);
2326 if (unlikely(!rt)) {
2328 dst = ERR_PTR(-ENOMEM);
2332 rt->dst.flags |= DST_HOST;
2333 rt->dst.output = ip6_output;
2334 rt->rt6i_gateway = fl6->daddr;
2335 rt->rt6i_dst.addr = fl6->daddr;
2336 rt->rt6i_dst.plen = 128;
2337 rt->rt6i_idev = idev;
2338 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
2340 /* Add this dst into uncached_list so that rt6_ifdown() can
2341 * do proper release of the net_device
2343 rt6_uncached_list_add(rt);
2345 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
2351 static int ip6_dst_gc(struct dst_ops *ops)
2353 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
2354 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
2355 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
2356 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
2357 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
2358 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
2361 entries = dst_entries_get_fast(ops);
2362 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
2363 entries <= rt_max_size)
2366 net->ipv6.ip6_rt_gc_expire++;
2367 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
2368 entries = dst_entries_get_slow(ops);
2369 if (entries < ops->gc_thresh)
2370 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
2372 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
2373 return entries > rt_max_size;
2376 static int ip6_convert_metrics(struct mx6_config *mxc,
2377 const struct fib6_config *cfg)
2379 bool ecn_ca = false;
2387 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
2391 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
2392 int type = nla_type(nla);
2397 if (unlikely(type > RTAX_MAX))
2400 if (type == RTAX_CC_ALGO) {
2401 char tmp[TCP_CA_NAME_MAX];
2403 nla_strlcpy(tmp, nla, sizeof(tmp));
2404 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
2405 if (val == TCP_CA_UNSPEC)
2408 val = nla_get_u32(nla);
2410 if (type == RTAX_HOPLIMIT && val > 255)
2412 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
2416 __set_bit(type - 1, mxc->mx_valid);
2420 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
2421 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
2431 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
2432 struct fib6_config *cfg,
2433 const struct in6_addr *gw_addr)
2435 struct flowi6 fl6 = {
2436 .flowi6_oif = cfg->fc_ifindex,
2438 .saddr = cfg->fc_prefsrc,
2440 struct fib6_table *table;
2441 struct rt6_info *rt;
2442 int flags = RT6_LOOKUP_F_IFACE | RT6_LOOKUP_F_IGNORE_LINKSTATE;
2444 table = fib6_get_table(net, cfg->fc_table);
2448 if (!ipv6_addr_any(&cfg->fc_prefsrc))
2449 flags |= RT6_LOOKUP_F_HAS_SADDR;
2451 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
2453 /* if table lookup failed, fall back to full lookup */
2454 if (rt == net->ipv6.ip6_null_entry) {
2462 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg,
2463 struct netlink_ext_ack *extack)
2465 struct net *net = cfg->fc_nlinfo.nl_net;
2466 struct rt6_info *rt = NULL;
2467 struct net_device *dev = NULL;
2468 struct inet6_dev *idev = NULL;
2469 struct fib6_table *table;
2473 /* RTF_PCPU is an internal flag; can not be set by userspace */
2474 if (cfg->fc_flags & RTF_PCPU) {
2475 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
2479 if (cfg->fc_dst_len > 128) {
2480 NL_SET_ERR_MSG(extack, "Invalid prefix length");
2483 if (cfg->fc_src_len > 128) {
2484 NL_SET_ERR_MSG(extack, "Invalid source address length");
2487 #ifndef CONFIG_IPV6_SUBTREES
2488 if (cfg->fc_src_len) {
2489 NL_SET_ERR_MSG(extack,
2490 "Specifying source address requires IPV6_SUBTREES to be enabled");
2494 if (cfg->fc_ifindex) {
2496 dev = dev_get_by_index(net, cfg->fc_ifindex);
2499 idev = in6_dev_get(dev);
2504 if (cfg->fc_metric == 0)
2505 cfg->fc_metric = IP6_RT_PRIO_USER;
2508 if (cfg->fc_nlinfo.nlh &&
2509 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
2510 table = fib6_get_table(net, cfg->fc_table);
2512 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
2513 table = fib6_new_table(net, cfg->fc_table);
2516 table = fib6_new_table(net, cfg->fc_table);
2522 rt = ip6_dst_alloc(net, NULL,
2523 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
2530 if (cfg->fc_flags & RTF_EXPIRES)
2531 rt6_set_expires(rt, jiffies +
2532 clock_t_to_jiffies(cfg->fc_expires));
2534 rt6_clean_expires(rt);
2536 if (cfg->fc_protocol == RTPROT_UNSPEC)
2537 cfg->fc_protocol = RTPROT_BOOT;
2538 rt->rt6i_protocol = cfg->fc_protocol;
2540 addr_type = ipv6_addr_type(&cfg->fc_dst);
2542 if (addr_type & IPV6_ADDR_MULTICAST)
2543 rt->dst.input = ip6_mc_input;
2544 else if (cfg->fc_flags & RTF_LOCAL)
2545 rt->dst.input = ip6_input;
2547 rt->dst.input = ip6_forward;
2549 rt->dst.output = ip6_output;
2551 if (cfg->fc_encap) {
2552 struct lwtunnel_state *lwtstate;
2554 err = lwtunnel_build_state(cfg->fc_encap_type,
2555 cfg->fc_encap, AF_INET6, cfg,
2559 rt->dst.lwtstate = lwtstate_get(lwtstate);
2560 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
2561 rt->dst.lwtstate->orig_output = rt->dst.output;
2562 rt->dst.output = lwtunnel_output;
2564 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
2565 rt->dst.lwtstate->orig_input = rt->dst.input;
2566 rt->dst.input = lwtunnel_input;
2570 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
2571 rt->rt6i_dst.plen = cfg->fc_dst_len;
2572 if (rt->rt6i_dst.plen == 128)
2573 rt->dst.flags |= DST_HOST;
2575 #ifdef CONFIG_IPV6_SUBTREES
2576 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
2577 rt->rt6i_src.plen = cfg->fc_src_len;
2580 rt->rt6i_metric = cfg->fc_metric;
2582 /* We cannot add true routes via loopback here,
2583 they would result in kernel looping; promote them to reject routes
2585 if ((cfg->fc_flags & RTF_REJECT) ||
2586 (dev && (dev->flags & IFF_LOOPBACK) &&
2587 !(addr_type & IPV6_ADDR_LOOPBACK) &&
2588 !(cfg->fc_flags & RTF_LOCAL))) {
2589 /* hold loopback dev/idev if we haven't done so. */
2590 if (dev != net->loopback_dev) {
2595 dev = net->loopback_dev;
2597 idev = in6_dev_get(dev);
2603 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
2604 switch (cfg->fc_type) {
2606 rt->dst.error = -EINVAL;
2607 rt->dst.output = dst_discard_out;
2608 rt->dst.input = dst_discard;
2611 rt->dst.error = -EACCES;
2612 rt->dst.output = ip6_pkt_prohibit_out;
2613 rt->dst.input = ip6_pkt_prohibit;
2616 case RTN_UNREACHABLE:
2618 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
2619 : (cfg->fc_type == RTN_UNREACHABLE)
2620 ? -EHOSTUNREACH : -ENETUNREACH;
2621 rt->dst.output = ip6_pkt_discard_out;
2622 rt->dst.input = ip6_pkt_discard;
2628 if (cfg->fc_flags & RTF_GATEWAY) {
2629 const struct in6_addr *gw_addr;
2632 gw_addr = &cfg->fc_gateway;
2633 gwa_type = ipv6_addr_type(gw_addr);
2635 /* if gw_addr is local we will fail to detect this in case
2636 * address is still TENTATIVE (DAD in progress). rt6_lookup()
2637 * will return already-added prefix route via interface that
2638 * prefix route was assigned to, which might be non-loopback.
2641 if (ipv6_chk_addr_and_flags(net, gw_addr,
2642 gwa_type & IPV6_ADDR_LINKLOCAL ?
2643 dev : NULL, 0, 0)) {
2644 NL_SET_ERR_MSG(extack, "Invalid gateway address");
2647 rt->rt6i_gateway = *gw_addr;
2649 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
2650 struct rt6_info *grt = NULL;
2652 /* IPv6 strictly inhibits using not link-local
2653 addresses as nexthop address.
2654 Otherwise, router will not able to send redirects.
2655 It is very good, but in some (rare!) circumstances
2656 (SIT, PtP, NBMA NOARP links) it is handy to allow
2657 some exceptions. --ANK
2658 We allow IPv4-mapped nexthops to support RFC4798-type
2661 if (!(gwa_type & (IPV6_ADDR_UNICAST |
2662 IPV6_ADDR_MAPPED))) {
2663 NL_SET_ERR_MSG(extack,
2664 "Invalid gateway address");
2668 if (cfg->fc_table) {
2669 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
2672 if (grt->rt6i_flags & RTF_GATEWAY ||
2673 (dev && dev != grt->dst.dev)) {
2681 grt = rt6_lookup(net, gw_addr, NULL,
2682 cfg->fc_ifindex, 1);
2684 err = -EHOSTUNREACH;
2688 if (dev != grt->dst.dev) {
2694 idev = grt->rt6i_idev;
2696 in6_dev_hold(grt->rt6i_idev);
2698 if (!(grt->rt6i_flags & RTF_GATEWAY))
2707 NL_SET_ERR_MSG(extack, "Egress device not specified");
2709 } else if (dev->flags & IFF_LOOPBACK) {
2710 NL_SET_ERR_MSG(extack,
2711 "Egress device can not be loopback device for this route");
2720 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2721 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2722 NL_SET_ERR_MSG(extack, "Invalid source address");
2726 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2727 rt->rt6i_prefsrc.plen = 128;
2729 rt->rt6i_prefsrc.plen = 0;
2731 rt->rt6i_flags = cfg->fc_flags;
2735 rt->rt6i_idev = idev;
2736 rt->rt6i_table = table;
2738 cfg->fc_nlinfo.nl_net = dev_net(dev);
2747 dst_release_immediate(&rt->dst);
2749 return ERR_PTR(err);
2752 int ip6_route_add(struct fib6_config *cfg,
2753 struct netlink_ext_ack *extack)
2755 struct mx6_config mxc = { .mx = NULL, };
2756 struct rt6_info *rt;
2759 rt = ip6_route_info_create(cfg, extack);
2766 err = ip6_convert_metrics(&mxc, cfg);
2770 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc, extack);
2777 dst_release_immediate(&rt->dst);
2782 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2785 struct fib6_table *table;
2786 struct net *net = dev_net(rt->dst.dev);
2788 if (rt == net->ipv6.ip6_null_entry) {
2793 table = rt->rt6i_table;
2794 spin_lock_bh(&table->tb6_lock);
2795 err = fib6_del(rt, info);
2796 spin_unlock_bh(&table->tb6_lock);
2803 int ip6_del_rt(struct rt6_info *rt)
2805 struct nl_info info = {
2806 .nl_net = dev_net(rt->dst.dev),
2808 return __ip6_del_rt(rt, &info);
2811 static int __ip6_del_rt_siblings(struct rt6_info *rt, struct fib6_config *cfg)
2813 struct nl_info *info = &cfg->fc_nlinfo;
2814 struct net *net = info->nl_net;
2815 struct sk_buff *skb = NULL;
2816 struct fib6_table *table;
2819 if (rt == net->ipv6.ip6_null_entry)
2821 table = rt->rt6i_table;
2822 spin_lock_bh(&table->tb6_lock);
2824 if (rt->rt6i_nsiblings && cfg->fc_delete_all_nh) {
2825 struct rt6_info *sibling, *next_sibling;
2827 /* prefer to send a single notification with all hops */
2828 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
2830 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2832 if (rt6_fill_node(net, skb, rt,
2833 NULL, NULL, 0, RTM_DELROUTE,
2834 info->portid, seq, 0) < 0) {
2838 info->skip_notify = 1;
2841 list_for_each_entry_safe(sibling, next_sibling,
2844 err = fib6_del(sibling, info);
2850 err = fib6_del(rt, info);
2852 spin_unlock_bh(&table->tb6_lock);
2857 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2858 info->nlh, gfp_any());
2863 static int ip6_route_del(struct fib6_config *cfg,
2864 struct netlink_ext_ack *extack)
2866 struct rt6_info *rt, *rt_cache;
2867 struct fib6_table *table;
2868 struct fib6_node *fn;
2871 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2873 NL_SET_ERR_MSG(extack, "FIB table does not exist");
2879 fn = fib6_locate(&table->tb6_root,
2880 &cfg->fc_dst, cfg->fc_dst_len,
2881 &cfg->fc_src, cfg->fc_src_len,
2882 !(cfg->fc_flags & RTF_CACHE));
2885 for_each_fib6_node_rt_rcu(fn) {
2886 if (cfg->fc_flags & RTF_CACHE) {
2887 rt_cache = rt6_find_cached_rt(rt, &cfg->fc_dst,
2893 if (cfg->fc_ifindex &&
2895 rt->dst.dev->ifindex != cfg->fc_ifindex))
2897 if (cfg->fc_flags & RTF_GATEWAY &&
2898 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2900 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2902 if (cfg->fc_protocol && cfg->fc_protocol != rt->rt6i_protocol)
2904 if (!dst_hold_safe(&rt->dst))
2908 /* if gateway was specified only delete the one hop */
2909 if (cfg->fc_flags & RTF_GATEWAY)
2910 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2912 return __ip6_del_rt_siblings(rt, cfg);
2920 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2922 struct netevent_redirect netevent;
2923 struct rt6_info *rt, *nrt = NULL;
2924 struct ndisc_options ndopts;
2925 struct inet6_dev *in6_dev;
2926 struct neighbour *neigh;
2928 int optlen, on_link;
2931 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2932 optlen -= sizeof(*msg);
2935 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2939 msg = (struct rd_msg *)icmp6_hdr(skb);
2941 if (ipv6_addr_is_multicast(&msg->dest)) {
2942 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2947 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2949 } else if (ipv6_addr_type(&msg->target) !=
2950 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2951 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2955 in6_dev = __in6_dev_get(skb->dev);
2958 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2962 * The IP source address of the Redirect MUST be the same as the current
2963 * first-hop router for the specified ICMP Destination Address.
2966 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
2967 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2972 if (ndopts.nd_opts_tgt_lladdr) {
2973 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2976 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2981 rt = (struct rt6_info *) dst;
2982 if (rt->rt6i_flags & RTF_REJECT) {
2983 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2987 /* Redirect received -> path was valid.
2988 * Look, redirects are sent only in response to data packets,
2989 * so that this nexthop apparently is reachable. --ANK
2991 dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
2993 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2998 * We have finally decided to accept it.
3001 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
3002 NEIGH_UPDATE_F_WEAK_OVERRIDE|
3003 NEIGH_UPDATE_F_OVERRIDE|
3004 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
3005 NEIGH_UPDATE_F_ISROUTER)),
3006 NDISC_REDIRECT, &ndopts);
3008 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
3012 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
3014 nrt->rt6i_flags &= ~RTF_GATEWAY;
3016 nrt->rt6i_protocol = RTPROT_REDIRECT;
3017 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
3019 /* No need to remove rt from the exception table if rt is
3020 * a cached route because rt6_insert_exception() will
3023 if (rt6_insert_exception(nrt, rt)) {
3024 dst_release_immediate(&nrt->dst);
3028 netevent.old = &rt->dst;
3029 netevent.new = &nrt->dst;
3030 netevent.daddr = &msg->dest;
3031 netevent.neigh = neigh;
3032 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
3035 neigh_release(neigh);
3039 * Misc support functions
3042 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
3044 BUG_ON(from->dst.from);
3046 rt->rt6i_flags &= ~RTF_EXPIRES;
3047 dst_hold(&from->dst);
3048 rt->dst.from = &from->dst;
3049 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
3052 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
3054 rt->dst.input = ort->dst.input;
3055 rt->dst.output = ort->dst.output;
3056 rt->rt6i_dst = ort->rt6i_dst;
3057 rt->dst.error = ort->dst.error;
3058 rt->rt6i_idev = ort->rt6i_idev;
3060 in6_dev_hold(rt->rt6i_idev);
3061 rt->dst.lastuse = jiffies;
3062 rt->rt6i_gateway = ort->rt6i_gateway;
3063 rt->rt6i_flags = ort->rt6i_flags;
3064 rt6_set_from(rt, ort);
3065 rt->rt6i_metric = ort->rt6i_metric;
3066 #ifdef CONFIG_IPV6_SUBTREES
3067 rt->rt6i_src = ort->rt6i_src;
3069 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
3070 rt->rt6i_table = ort->rt6i_table;
3071 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
3074 #ifdef CONFIG_IPV6_ROUTE_INFO
3075 static struct rt6_info *rt6_get_route_info(struct net *net,
3076 const struct in6_addr *prefix, int prefixlen,
3077 const struct in6_addr *gwaddr,
3078 struct net_device *dev)
3080 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
3081 int ifindex = dev->ifindex;
3082 struct fib6_node *fn;
3083 struct rt6_info *rt = NULL;
3084 struct fib6_table *table;
3086 table = fib6_get_table(net, tb_id);
3091 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
3095 for_each_fib6_node_rt_rcu(fn) {
3096 if (rt->dst.dev->ifindex != ifindex)
3098 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
3100 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
3102 ip6_hold_safe(NULL, &rt, false);
3110 static struct rt6_info *rt6_add_route_info(struct net *net,
3111 const struct in6_addr *prefix, int prefixlen,
3112 const struct in6_addr *gwaddr,
3113 struct net_device *dev,
3116 struct fib6_config cfg = {
3117 .fc_metric = IP6_RT_PRIO_USER,
3118 .fc_ifindex = dev->ifindex,
3119 .fc_dst_len = prefixlen,
3120 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
3121 RTF_UP | RTF_PREF(pref),
3122 .fc_protocol = RTPROT_RA,
3123 .fc_nlinfo.portid = 0,
3124 .fc_nlinfo.nlh = NULL,
3125 .fc_nlinfo.nl_net = net,
3128 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
3129 cfg.fc_dst = *prefix;
3130 cfg.fc_gateway = *gwaddr;
3132 /* We should treat it as a default route if prefix length is 0. */
3134 cfg.fc_flags |= RTF_DEFAULT;
3136 ip6_route_add(&cfg, NULL);
3138 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
3142 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
3144 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
3145 struct rt6_info *rt;
3146 struct fib6_table *table;
3148 table = fib6_get_table(dev_net(dev), tb_id);
3153 for_each_fib6_node_rt_rcu(&table->tb6_root) {
3154 if (dev == rt->dst.dev &&
3155 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
3156 ipv6_addr_equal(&rt->rt6i_gateway, addr))
3160 ip6_hold_safe(NULL, &rt, false);
3165 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
3166 struct net_device *dev,
3169 struct fib6_config cfg = {
3170 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
3171 .fc_metric = IP6_RT_PRIO_USER,
3172 .fc_ifindex = dev->ifindex,
3173 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
3174 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
3175 .fc_protocol = RTPROT_RA,
3176 .fc_nlinfo.portid = 0,
3177 .fc_nlinfo.nlh = NULL,
3178 .fc_nlinfo.nl_net = dev_net(dev),
3181 cfg.fc_gateway = *gwaddr;
3183 if (!ip6_route_add(&cfg, NULL)) {
3184 struct fib6_table *table;
3186 table = fib6_get_table(dev_net(dev), cfg.fc_table);
3188 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
3191 return rt6_get_dflt_router(gwaddr, dev);
3194 static void __rt6_purge_dflt_routers(struct fib6_table *table)
3196 struct rt6_info *rt;
3200 for_each_fib6_node_rt_rcu(&table->tb6_root) {
3201 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
3202 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
3203 if (dst_hold_safe(&rt->dst)) {
3214 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
3217 void rt6_purge_dflt_routers(struct net *net)
3219 struct fib6_table *table;
3220 struct hlist_head *head;
3225 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
3226 head = &net->ipv6.fib_table_hash[h];
3227 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
3228 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
3229 __rt6_purge_dflt_routers(table);
3236 static void rtmsg_to_fib6_config(struct net *net,
3237 struct in6_rtmsg *rtmsg,
3238 struct fib6_config *cfg)
3240 memset(cfg, 0, sizeof(*cfg));
3242 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
3244 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
3245 cfg->fc_metric = rtmsg->rtmsg_metric;
3246 cfg->fc_expires = rtmsg->rtmsg_info;
3247 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
3248 cfg->fc_src_len = rtmsg->rtmsg_src_len;
3249 cfg->fc_flags = rtmsg->rtmsg_flags;
3251 cfg->fc_nlinfo.nl_net = net;
3253 cfg->fc_dst = rtmsg->rtmsg_dst;
3254 cfg->fc_src = rtmsg->rtmsg_src;
3255 cfg->fc_gateway = rtmsg->rtmsg_gateway;
3258 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
3260 struct fib6_config cfg;
3261 struct in6_rtmsg rtmsg;
3265 case SIOCADDRT: /* Add a route */
3266 case SIOCDELRT: /* Delete a route */
3267 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3269 err = copy_from_user(&rtmsg, arg,
3270 sizeof(struct in6_rtmsg));
3274 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
3279 err = ip6_route_add(&cfg, NULL);
3282 err = ip6_route_del(&cfg, NULL);
3296 * Drop the packet on the floor
3299 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
3302 struct dst_entry *dst = skb_dst(skb);
3303 switch (ipstats_mib_noroutes) {
3304 case IPSTATS_MIB_INNOROUTES:
3305 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
3306 if (type == IPV6_ADDR_ANY) {
3307 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
3308 IPSTATS_MIB_INADDRERRORS);
3312 case IPSTATS_MIB_OUTNOROUTES:
3313 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
3314 ipstats_mib_noroutes);
3317 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
3322 static int ip6_pkt_discard(struct sk_buff *skb)
3324 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
3327 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
3329 skb->dev = skb_dst(skb)->dev;
3330 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
3333 static int ip6_pkt_prohibit(struct sk_buff *skb)
3335 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
3338 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
3340 skb->dev = skb_dst(skb)->dev;
3341 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
3345 * Allocate a dst for local (unicast / anycast) address.
3348 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
3349 const struct in6_addr *addr,
3353 struct net *net = dev_net(idev->dev);
3354 struct net_device *dev = idev->dev;
3355 struct rt6_info *rt;
3357 rt = ip6_dst_alloc(net, dev, DST_NOCOUNT);
3359 return ERR_PTR(-ENOMEM);
3363 rt->dst.flags |= DST_HOST;
3364 rt->dst.input = ip6_input;
3365 rt->dst.output = ip6_output;
3366 rt->rt6i_idev = idev;
3368 rt->rt6i_protocol = RTPROT_KERNEL;
3369 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
3371 rt->rt6i_flags |= RTF_ANYCAST;
3373 rt->rt6i_flags |= RTF_LOCAL;
3375 rt->rt6i_gateway = *addr;
3376 rt->rt6i_dst.addr = *addr;
3377 rt->rt6i_dst.plen = 128;
3378 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
3379 rt->rt6i_table = fib6_get_table(net, tb_id);
3384 /* remove deleted ip from prefsrc entries */
3385 struct arg_dev_net_ip {
3386 struct net_device *dev;
3388 struct in6_addr *addr;
3391 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
3393 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
3394 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
3395 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
3397 if (((void *)rt->dst.dev == dev || !dev) &&
3398 rt != net->ipv6.ip6_null_entry &&
3399 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
3400 spin_lock_bh(&rt6_exception_lock);
3401 /* remove prefsrc entry */
3402 rt->rt6i_prefsrc.plen = 0;
3403 /* need to update cache as well */
3404 rt6_exceptions_remove_prefsrc(rt);
3405 spin_unlock_bh(&rt6_exception_lock);
3410 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
3412 struct net *net = dev_net(ifp->idev->dev);
3413 struct arg_dev_net_ip adni = {
3414 .dev = ifp->idev->dev,
3418 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
3421 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
3423 /* Remove routers and update dst entries when gateway turn into host. */
3424 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
3426 struct in6_addr *gateway = (struct in6_addr *)arg;
3428 if (((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
3429 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
3433 /* Further clean up cached routes in exception table.
3434 * This is needed because cached route may have a different
3435 * gateway than its 'parent' in the case of an ip redirect.
3437 rt6_exceptions_clean_tohost(rt, gateway);
3442 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
3444 fib6_clean_all(net, fib6_clean_tohost, gateway);
3447 struct arg_dev_net {
3448 struct net_device *dev;
3452 /* called with write lock held for table with rt */
3453 static int fib6_ifdown(struct rt6_info *rt, void *arg)
3455 const struct arg_dev_net *adn = arg;
3456 const struct net_device *dev = adn->dev;
3458 if ((rt->dst.dev == dev || !dev) &&
3459 rt != adn->net->ipv6.ip6_null_entry &&
3460 (rt->rt6i_nsiblings == 0 ||
3461 (dev && netdev_unregistering(dev)) ||
3462 !rt->rt6i_idev->cnf.ignore_routes_with_linkdown))
3468 void rt6_ifdown(struct net *net, struct net_device *dev)
3470 struct arg_dev_net adn = {
3475 fib6_clean_all(net, fib6_ifdown, &adn);
3477 rt6_uncached_list_flush_dev(net, dev);
3480 struct rt6_mtu_change_arg {
3481 struct net_device *dev;
3485 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
3487 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
3488 struct inet6_dev *idev;
3490 /* In IPv6 pmtu discovery is not optional,
3491 so that RTAX_MTU lock cannot disable it.
3492 We still use this lock to block changes
3493 caused by addrconf/ndisc.
3496 idev = __in6_dev_get(arg->dev);
3500 /* For administrative MTU increase, there is no way to discover
3501 IPv6 PMTU increase, so PMTU increase should be updated here.
3502 Since RFC 1981 doesn't include administrative MTU increase
3503 update PMTU increase is a MUST. (i.e. jumbo frame)
3506 If new MTU is less than route PMTU, this new MTU will be the
3507 lowest MTU in the path, update the route PMTU to reflect PMTU
3508 decreases; if new MTU is greater than route PMTU, and the
3509 old MTU is the lowest MTU in the path, update the route PMTU
3510 to reflect the increase. In this case if the other nodes' MTU
3511 also have the lowest MTU, TOO BIG MESSAGE will be lead to
3514 if (rt->dst.dev == arg->dev &&
3515 dst_metric_raw(&rt->dst, RTAX_MTU) &&
3516 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
3517 spin_lock_bh(&rt6_exception_lock);
3518 if (dst_mtu(&rt->dst) >= arg->mtu ||
3519 (dst_mtu(&rt->dst) < arg->mtu &&
3520 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
3521 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
3523 rt6_exceptions_update_pmtu(rt, arg->mtu);
3524 spin_unlock_bh(&rt6_exception_lock);
3529 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
3531 struct rt6_mtu_change_arg arg = {
3536 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
3539 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
3540 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
3541 [RTA_OIF] = { .type = NLA_U32 },
3542 [RTA_IIF] = { .type = NLA_U32 },
3543 [RTA_PRIORITY] = { .type = NLA_U32 },
3544 [RTA_METRICS] = { .type = NLA_NESTED },
3545 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
3546 [RTA_PREF] = { .type = NLA_U8 },
3547 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
3548 [RTA_ENCAP] = { .type = NLA_NESTED },
3549 [RTA_EXPIRES] = { .type = NLA_U32 },
3550 [RTA_UID] = { .type = NLA_U32 },
3551 [RTA_MARK] = { .type = NLA_U32 },
3554 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
3555 struct fib6_config *cfg,
3556 struct netlink_ext_ack *extack)
3559 struct nlattr *tb[RTA_MAX+1];
3563 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
3569 rtm = nlmsg_data(nlh);
3570 memset(cfg, 0, sizeof(*cfg));
3572 cfg->fc_table = rtm->rtm_table;
3573 cfg->fc_dst_len = rtm->rtm_dst_len;
3574 cfg->fc_src_len = rtm->rtm_src_len;
3575 cfg->fc_flags = RTF_UP;
3576 cfg->fc_protocol = rtm->rtm_protocol;
3577 cfg->fc_type = rtm->rtm_type;
3579 if (rtm->rtm_type == RTN_UNREACHABLE ||
3580 rtm->rtm_type == RTN_BLACKHOLE ||
3581 rtm->rtm_type == RTN_PROHIBIT ||
3582 rtm->rtm_type == RTN_THROW)
3583 cfg->fc_flags |= RTF_REJECT;
3585 if (rtm->rtm_type == RTN_LOCAL)
3586 cfg->fc_flags |= RTF_LOCAL;
3588 if (rtm->rtm_flags & RTM_F_CLONED)
3589 cfg->fc_flags |= RTF_CACHE;
3591 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
3592 cfg->fc_nlinfo.nlh = nlh;
3593 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
3595 if (tb[RTA_GATEWAY]) {
3596 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
3597 cfg->fc_flags |= RTF_GATEWAY;
3601 int plen = (rtm->rtm_dst_len + 7) >> 3;
3603 if (nla_len(tb[RTA_DST]) < plen)
3606 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
3610 int plen = (rtm->rtm_src_len + 7) >> 3;
3612 if (nla_len(tb[RTA_SRC]) < plen)
3615 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
3618 if (tb[RTA_PREFSRC])
3619 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
3622 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
3624 if (tb[RTA_PRIORITY])
3625 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
3627 if (tb[RTA_METRICS]) {
3628 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
3629 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
3633 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
3635 if (tb[RTA_MULTIPATH]) {
3636 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
3637 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
3639 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
3640 cfg->fc_mp_len, extack);
3646 pref = nla_get_u8(tb[RTA_PREF]);
3647 if (pref != ICMPV6_ROUTER_PREF_LOW &&
3648 pref != ICMPV6_ROUTER_PREF_HIGH)
3649 pref = ICMPV6_ROUTER_PREF_MEDIUM;
3650 cfg->fc_flags |= RTF_PREF(pref);
3654 cfg->fc_encap = tb[RTA_ENCAP];
3656 if (tb[RTA_ENCAP_TYPE]) {
3657 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
3659 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
3664 if (tb[RTA_EXPIRES]) {
3665 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
3667 if (addrconf_finite_timeout(timeout)) {
3668 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
3669 cfg->fc_flags |= RTF_EXPIRES;
3679 struct rt6_info *rt6_info;
3680 struct fib6_config r_cfg;
3681 struct mx6_config mxc;
3682 struct list_head next;
3685 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
3689 list_for_each_entry(nh, rt6_nh_list, next) {
3690 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6c nexthop %pI6c ifi %d\n",
3691 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
3692 nh->r_cfg.fc_ifindex);
3696 static int ip6_route_info_append(struct list_head *rt6_nh_list,
3697 struct rt6_info *rt, struct fib6_config *r_cfg)
3702 list_for_each_entry(nh, rt6_nh_list, next) {
3703 /* check if rt6_info already exists */
3704 if (rt6_duplicate_nexthop(nh->rt6_info, rt))
3708 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
3712 err = ip6_convert_metrics(&nh->mxc, r_cfg);
3717 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
3718 list_add_tail(&nh->next, rt6_nh_list);
3723 static void ip6_route_mpath_notify(struct rt6_info *rt,
3724 struct rt6_info *rt_last,
3725 struct nl_info *info,
3728 /* if this is an APPEND route, then rt points to the first route
3729 * inserted and rt_last points to last route inserted. Userspace
3730 * wants a consistent dump of the route which starts at the first
3731 * nexthop. Since sibling routes are always added at the end of
3732 * the list, find the first sibling of the last route appended
3734 if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->rt6i_nsiblings) {
3735 rt = list_first_entry(&rt_last->rt6i_siblings,
3741 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
3744 static int ip6_route_multipath_add(struct fib6_config *cfg,
3745 struct netlink_ext_ack *extack)
3747 struct rt6_info *rt_notif = NULL, *rt_last = NULL;
3748 struct nl_info *info = &cfg->fc_nlinfo;
3749 struct fib6_config r_cfg;
3750 struct rtnexthop *rtnh;
3751 struct rt6_info *rt;
3752 struct rt6_nh *err_nh;
3753 struct rt6_nh *nh, *nh_safe;
3759 int replace = (cfg->fc_nlinfo.nlh &&
3760 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
3761 LIST_HEAD(rt6_nh_list);
3763 nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
3764 if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
3765 nlflags |= NLM_F_APPEND;
3767 remaining = cfg->fc_mp_len;
3768 rtnh = (struct rtnexthop *)cfg->fc_mp;
3770 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
3771 * rt6_info structs per nexthop
3773 while (rtnh_ok(rtnh, remaining)) {
3774 memcpy(&r_cfg, cfg, sizeof(*cfg));
3775 if (rtnh->rtnh_ifindex)
3776 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3778 attrlen = rtnh_attrlen(rtnh);
3780 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3782 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3784 r_cfg.fc_gateway = nla_get_in6_addr(nla);
3785 r_cfg.fc_flags |= RTF_GATEWAY;
3787 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
3788 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
3790 r_cfg.fc_encap_type = nla_get_u16(nla);
3793 rt = ip6_route_info_create(&r_cfg, extack);
3800 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
3802 dst_release_immediate(&rt->dst);
3806 rtnh = rtnh_next(rtnh, &remaining);
3809 /* for add and replace send one notification with all nexthops.
3810 * Skip the notification in fib6_add_rt2node and send one with
3811 * the full route when done
3813 info->skip_notify = 1;
3816 list_for_each_entry(nh, &rt6_nh_list, next) {
3817 rt_last = nh->rt6_info;
3818 err = __ip6_ins_rt(nh->rt6_info, info, &nh->mxc, extack);
3819 /* save reference to first route for notification */
3820 if (!rt_notif && !err)
3821 rt_notif = nh->rt6_info;
3823 /* nh->rt6_info is used or freed at this point, reset to NULL*/
3824 nh->rt6_info = NULL;
3827 ip6_print_replace_route_err(&rt6_nh_list);
3832 /* Because each route is added like a single route we remove
3833 * these flags after the first nexthop: if there is a collision,
3834 * we have already failed to add the first nexthop:
3835 * fib6_add_rt2node() has rejected it; when replacing, old
3836 * nexthops have been replaced by first new, the rest should
3839 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
3844 /* success ... tell user about new route */
3845 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
3849 /* send notification for routes that were added so that
3850 * the delete notifications sent by ip6_route_del are
3854 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
3856 /* Delete routes that were already added */
3857 list_for_each_entry(nh, &rt6_nh_list, next) {
3860 ip6_route_del(&nh->r_cfg, extack);
3864 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3866 dst_release_immediate(&nh->rt6_info->dst);
3868 list_del(&nh->next);
3875 static int ip6_route_multipath_del(struct fib6_config *cfg,
3876 struct netlink_ext_ack *extack)
3878 struct fib6_config r_cfg;
3879 struct rtnexthop *rtnh;
3882 int err = 1, last_err = 0;
3884 remaining = cfg->fc_mp_len;
3885 rtnh = (struct rtnexthop *)cfg->fc_mp;
3887 /* Parse a Multipath Entry */
3888 while (rtnh_ok(rtnh, remaining)) {
3889 memcpy(&r_cfg, cfg, sizeof(*cfg));
3890 if (rtnh->rtnh_ifindex)
3891 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3893 attrlen = rtnh_attrlen(rtnh);
3895 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3897 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3899 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3900 r_cfg.fc_flags |= RTF_GATEWAY;
3903 err = ip6_route_del(&r_cfg, extack);
3907 rtnh = rtnh_next(rtnh, &remaining);
3913 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
3914 struct netlink_ext_ack *extack)
3916 struct fib6_config cfg;
3919 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
3924 return ip6_route_multipath_del(&cfg, extack);
3926 cfg.fc_delete_all_nh = 1;
3927 return ip6_route_del(&cfg, extack);
3931 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
3932 struct netlink_ext_ack *extack)
3934 struct fib6_config cfg;
3937 err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
3942 return ip6_route_multipath_add(&cfg, extack);
3944 return ip6_route_add(&cfg, extack);
3947 static size_t rt6_nlmsg_size(struct rt6_info *rt)
3949 int nexthop_len = 0;
3951 if (rt->rt6i_nsiblings) {
3952 nexthop_len = nla_total_size(0) /* RTA_MULTIPATH */
3953 + NLA_ALIGN(sizeof(struct rtnexthop))
3954 + nla_total_size(16) /* RTA_GATEWAY */
3955 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3957 nexthop_len *= rt->rt6i_nsiblings;
3960 return NLMSG_ALIGN(sizeof(struct rtmsg))
3961 + nla_total_size(16) /* RTA_SRC */
3962 + nla_total_size(16) /* RTA_DST */
3963 + nla_total_size(16) /* RTA_GATEWAY */
3964 + nla_total_size(16) /* RTA_PREFSRC */
3965 + nla_total_size(4) /* RTA_TABLE */
3966 + nla_total_size(4) /* RTA_IIF */
3967 + nla_total_size(4) /* RTA_OIF */
3968 + nla_total_size(4) /* RTA_PRIORITY */
3969 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3970 + nla_total_size(sizeof(struct rta_cacheinfo))
3971 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3972 + nla_total_size(1) /* RTA_PREF */
3973 + lwtunnel_get_encap_size(rt->dst.lwtstate)
3977 static int rt6_nexthop_info(struct sk_buff *skb, struct rt6_info *rt,
3978 unsigned int *flags, bool skip_oif)
3980 if (!netif_running(rt->dst.dev) || !netif_carrier_ok(rt->dst.dev)) {
3981 *flags |= RTNH_F_LINKDOWN;
3982 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3983 *flags |= RTNH_F_DEAD;
3986 if (rt->rt6i_flags & RTF_GATEWAY) {
3987 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3988 goto nla_put_failure;
3991 if (rt->rt6i_nh_flags & RTNH_F_OFFLOAD)
3992 *flags |= RTNH_F_OFFLOAD;
3994 /* not needed for multipath encoding b/c it has a rtnexthop struct */
3995 if (!skip_oif && rt->dst.dev &&
3996 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3997 goto nla_put_failure;
3999 if (rt->dst.lwtstate &&
4000 lwtunnel_fill_encap(skb, rt->dst.lwtstate) < 0)
4001 goto nla_put_failure;
4009 /* add multipath next hop */
4010 static int rt6_add_nexthop(struct sk_buff *skb, struct rt6_info *rt)
4012 struct rtnexthop *rtnh;
4013 unsigned int flags = 0;
4015 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
4017 goto nla_put_failure;
4019 rtnh->rtnh_hops = 0;
4020 rtnh->rtnh_ifindex = rt->dst.dev ? rt->dst.dev->ifindex : 0;
4022 if (rt6_nexthop_info(skb, rt, &flags, true) < 0)
4023 goto nla_put_failure;
4025 rtnh->rtnh_flags = flags;
4027 /* length of rtnetlink header + attributes */
4028 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
4036 static int rt6_fill_node(struct net *net,
4037 struct sk_buff *skb, struct rt6_info *rt,
4038 struct in6_addr *dst, struct in6_addr *src,
4039 int iif, int type, u32 portid, u32 seq,
4042 u32 metrics[RTAX_MAX];
4044 struct nlmsghdr *nlh;
4048 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
4052 rtm = nlmsg_data(nlh);
4053 rtm->rtm_family = AF_INET6;
4054 rtm->rtm_dst_len = rt->rt6i_dst.plen;
4055 rtm->rtm_src_len = rt->rt6i_src.plen;
4058 table = rt->rt6i_table->tb6_id;
4060 table = RT6_TABLE_UNSPEC;
4061 rtm->rtm_table = table;
4062 if (nla_put_u32(skb, RTA_TABLE, table))
4063 goto nla_put_failure;
4064 if (rt->rt6i_flags & RTF_REJECT) {
4065 switch (rt->dst.error) {
4067 rtm->rtm_type = RTN_BLACKHOLE;
4070 rtm->rtm_type = RTN_PROHIBIT;
4073 rtm->rtm_type = RTN_THROW;
4076 rtm->rtm_type = RTN_UNREACHABLE;
4080 else if (rt->rt6i_flags & RTF_LOCAL)
4081 rtm->rtm_type = RTN_LOCAL;
4082 else if (rt->rt6i_flags & RTF_ANYCAST)
4083 rtm->rtm_type = RTN_ANYCAST;
4084 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
4085 rtm->rtm_type = RTN_LOCAL;
4087 rtm->rtm_type = RTN_UNICAST;
4089 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
4090 rtm->rtm_protocol = rt->rt6i_protocol;
4092 if (rt->rt6i_flags & RTF_CACHE)
4093 rtm->rtm_flags |= RTM_F_CLONED;
4096 if (nla_put_in6_addr(skb, RTA_DST, dst))
4097 goto nla_put_failure;
4098 rtm->rtm_dst_len = 128;
4099 } else if (rtm->rtm_dst_len)
4100 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
4101 goto nla_put_failure;
4102 #ifdef CONFIG_IPV6_SUBTREES
4104 if (nla_put_in6_addr(skb, RTA_SRC, src))
4105 goto nla_put_failure;
4106 rtm->rtm_src_len = 128;
4107 } else if (rtm->rtm_src_len &&
4108 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
4109 goto nla_put_failure;
4112 #ifdef CONFIG_IPV6_MROUTE
4113 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
4114 int err = ip6mr_get_route(net, skb, rtm, portid);
4119 goto nla_put_failure;
4122 if (nla_put_u32(skb, RTA_IIF, iif))
4123 goto nla_put_failure;
4125 struct in6_addr saddr_buf;
4126 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
4127 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
4128 goto nla_put_failure;
4131 if (rt->rt6i_prefsrc.plen) {
4132 struct in6_addr saddr_buf;
4133 saddr_buf = rt->rt6i_prefsrc.addr;
4134 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
4135 goto nla_put_failure;
4138 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
4140 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
4141 if (rtnetlink_put_metrics(skb, metrics) < 0)
4142 goto nla_put_failure;
4144 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
4145 goto nla_put_failure;
4147 /* For multipath routes, walk the siblings list and add
4148 * each as a nexthop within RTA_MULTIPATH.
4150 if (rt->rt6i_nsiblings) {
4151 struct rt6_info *sibling, *next_sibling;
4154 mp = nla_nest_start(skb, RTA_MULTIPATH);
4156 goto nla_put_failure;
4158 if (rt6_add_nexthop(skb, rt) < 0)
4159 goto nla_put_failure;
4161 list_for_each_entry_safe(sibling, next_sibling,
4162 &rt->rt6i_siblings, rt6i_siblings) {
4163 if (rt6_add_nexthop(skb, sibling) < 0)
4164 goto nla_put_failure;
4167 nla_nest_end(skb, mp);
4169 if (rt6_nexthop_info(skb, rt, &rtm->rtm_flags, false) < 0)
4170 goto nla_put_failure;
4173 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
4175 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
4176 goto nla_put_failure;
4178 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
4179 goto nla_put_failure;
4182 nlmsg_end(skb, nlh);
4186 nlmsg_cancel(skb, nlh);
4190 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
4192 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
4193 struct net *net = arg->net;
4195 if (rt == net->ipv6.ip6_null_entry)
4198 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
4199 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
4201 /* user wants prefix routes only */
4202 if (rtm->rtm_flags & RTM_F_PREFIX &&
4203 !(rt->rt6i_flags & RTF_PREFIX_RT)) {
4204 /* success since this is not a prefix route */
4209 return rt6_fill_node(net,
4210 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
4211 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
4215 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4216 struct netlink_ext_ack *extack)
4218 struct net *net = sock_net(in_skb->sk);
4219 struct nlattr *tb[RTA_MAX+1];
4220 int err, iif = 0, oif = 0;
4221 struct dst_entry *dst;
4222 struct rt6_info *rt;
4223 struct sk_buff *skb;
4228 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
4234 memset(&fl6, 0, sizeof(fl6));
4235 rtm = nlmsg_data(nlh);
4236 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
4237 fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
4240 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
4243 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
4247 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
4250 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
4254 iif = nla_get_u32(tb[RTA_IIF]);
4257 oif = nla_get_u32(tb[RTA_OIF]);
4260 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
4263 fl6.flowi6_uid = make_kuid(current_user_ns(),
4264 nla_get_u32(tb[RTA_UID]));
4266 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
4269 struct net_device *dev;
4274 dev = dev_get_by_index_rcu(net, iif);
4281 fl6.flowi6_iif = iif;
4283 if (!ipv6_addr_any(&fl6.saddr))
4284 flags |= RT6_LOOKUP_F_HAS_SADDR;
4287 dst = ip6_route_input_lookup(net, dev, &fl6, flags);
4289 dst = ip6_route_lookup(net, &fl6, 0);
4293 fl6.flowi6_oif = oif;
4296 dst = ip6_route_output(net, NULL, &fl6);
4298 dst = ip6_route_lookup(net, &fl6, 0);
4302 rt = container_of(dst, struct rt6_info, dst);
4303 if (rt->dst.error) {
4304 err = rt->dst.error;
4309 if (rt == net->ipv6.ip6_null_entry) {
4310 err = rt->dst.error;
4315 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
4322 skb_dst_set(skb, &rt->dst);
4324 err = rt6_fill_node(net, skb, rt, NULL, NULL, iif,
4325 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
4328 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
4329 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
4336 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4341 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
4342 unsigned int nlm_flags)
4344 struct sk_buff *skb;
4345 struct net *net = info->nl_net;
4350 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
4352 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
4356 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
4357 event, info->portid, seq, nlm_flags);
4359 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
4360 WARN_ON(err == -EMSGSIZE);
4364 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
4365 info->nlh, gfp_any());
4369 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
4372 static int ip6_route_dev_notify(struct notifier_block *this,
4373 unsigned long event, void *ptr)
4375 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4376 struct net *net = dev_net(dev);
4378 if (!(dev->flags & IFF_LOOPBACK))
4381 if (event == NETDEV_REGISTER) {
4382 net->ipv6.ip6_null_entry->dst.dev = dev;
4383 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
4384 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4385 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
4386 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
4387 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
4388 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
4390 } else if (event == NETDEV_UNREGISTER &&
4391 dev->reg_state != NETREG_UNREGISTERED) {
4392 /* NETDEV_UNREGISTER could be fired for multiple times by
4393 * netdev_wait_allrefs(). Make sure we only call this once.
4395 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
4396 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4397 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
4398 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
4409 #ifdef CONFIG_PROC_FS
4411 static const struct file_operations ipv6_route_proc_fops = {
4412 .owner = THIS_MODULE,
4413 .open = ipv6_route_open,
4415 .llseek = seq_lseek,
4416 .release = seq_release_net,
4419 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
4421 struct net *net = (struct net *)seq->private;
4422 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
4423 net->ipv6.rt6_stats->fib_nodes,
4424 net->ipv6.rt6_stats->fib_route_nodes,
4425 net->ipv6.rt6_stats->fib_rt_alloc,
4426 net->ipv6.rt6_stats->fib_rt_entries,
4427 net->ipv6.rt6_stats->fib_rt_cache,
4428 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
4429 net->ipv6.rt6_stats->fib_discarded_routes);
4434 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
4436 return single_open_net(inode, file, rt6_stats_seq_show);
4439 static const struct file_operations rt6_stats_seq_fops = {
4440 .owner = THIS_MODULE,
4441 .open = rt6_stats_seq_open,
4443 .llseek = seq_lseek,
4444 .release = single_release_net,
4446 #endif /* CONFIG_PROC_FS */
4448 #ifdef CONFIG_SYSCTL
4451 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
4452 void __user *buffer, size_t *lenp, loff_t *ppos)
4459 net = (struct net *)ctl->extra1;
4460 delay = net->ipv6.sysctl.flush_delay;
4461 proc_dointvec(ctl, write, buffer, lenp, ppos);
4462 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
4466 struct ctl_table ipv6_route_table_template[] = {
4468 .procname = "flush",
4469 .data = &init_net.ipv6.sysctl.flush_delay,
4470 .maxlen = sizeof(int),
4472 .proc_handler = ipv6_sysctl_rtcache_flush
4475 .procname = "gc_thresh",
4476 .data = &ip6_dst_ops_template.gc_thresh,
4477 .maxlen = sizeof(int),
4479 .proc_handler = proc_dointvec,
4482 .procname = "max_size",
4483 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
4484 .maxlen = sizeof(int),
4486 .proc_handler = proc_dointvec,
4489 .procname = "gc_min_interval",
4490 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
4491 .maxlen = sizeof(int),
4493 .proc_handler = proc_dointvec_jiffies,
4496 .procname = "gc_timeout",
4497 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
4498 .maxlen = sizeof(int),
4500 .proc_handler = proc_dointvec_jiffies,
4503 .procname = "gc_interval",
4504 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
4505 .maxlen = sizeof(int),
4507 .proc_handler = proc_dointvec_jiffies,
4510 .procname = "gc_elasticity",
4511 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
4512 .maxlen = sizeof(int),
4514 .proc_handler = proc_dointvec,
4517 .procname = "mtu_expires",
4518 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
4519 .maxlen = sizeof(int),
4521 .proc_handler = proc_dointvec_jiffies,
4524 .procname = "min_adv_mss",
4525 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
4526 .maxlen = sizeof(int),
4528 .proc_handler = proc_dointvec,
4531 .procname = "gc_min_interval_ms",
4532 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
4533 .maxlen = sizeof(int),
4535 .proc_handler = proc_dointvec_ms_jiffies,
4540 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
4542 struct ctl_table *table;
4544 table = kmemdup(ipv6_route_table_template,
4545 sizeof(ipv6_route_table_template),
4549 table[0].data = &net->ipv6.sysctl.flush_delay;
4550 table[0].extra1 = net;
4551 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
4552 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
4553 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
4554 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
4555 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
4556 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
4557 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
4558 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
4559 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
4561 /* Don't export sysctls to unprivileged users */
4562 if (net->user_ns != &init_user_ns)
4563 table[0].procname = NULL;
4570 static int __net_init ip6_route_net_init(struct net *net)
4574 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
4575 sizeof(net->ipv6.ip6_dst_ops));
4577 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
4578 goto out_ip6_dst_ops;
4580 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
4581 sizeof(*net->ipv6.ip6_null_entry),
4583 if (!net->ipv6.ip6_null_entry)
4584 goto out_ip6_dst_entries;
4585 net->ipv6.ip6_null_entry->dst.path =
4586 (struct dst_entry *)net->ipv6.ip6_null_entry;
4587 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
4588 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
4589 ip6_template_metrics, true);
4591 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4592 net->ipv6.fib6_has_custom_rules = false;
4593 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
4594 sizeof(*net->ipv6.ip6_prohibit_entry),
4596 if (!net->ipv6.ip6_prohibit_entry)
4597 goto out_ip6_null_entry;
4598 net->ipv6.ip6_prohibit_entry->dst.path =
4599 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
4600 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
4601 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
4602 ip6_template_metrics, true);
4604 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
4605 sizeof(*net->ipv6.ip6_blk_hole_entry),
4607 if (!net->ipv6.ip6_blk_hole_entry)
4608 goto out_ip6_prohibit_entry;
4609 net->ipv6.ip6_blk_hole_entry->dst.path =
4610 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
4611 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
4612 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
4613 ip6_template_metrics, true);
4616 net->ipv6.sysctl.flush_delay = 0;
4617 net->ipv6.sysctl.ip6_rt_max_size = 4096;
4618 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
4619 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
4620 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
4621 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
4622 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
4623 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
4625 net->ipv6.ip6_rt_gc_expire = 30*HZ;
4631 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4632 out_ip6_prohibit_entry:
4633 kfree(net->ipv6.ip6_prohibit_entry);
4635 kfree(net->ipv6.ip6_null_entry);
4637 out_ip6_dst_entries:
4638 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
4643 static void __net_exit ip6_route_net_exit(struct net *net)
4645 kfree(net->ipv6.ip6_null_entry);
4646 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4647 kfree(net->ipv6.ip6_prohibit_entry);
4648 kfree(net->ipv6.ip6_blk_hole_entry);
4650 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
4653 static int __net_init ip6_route_net_init_late(struct net *net)
4655 #ifdef CONFIG_PROC_FS
4656 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
4657 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
4662 static void __net_exit ip6_route_net_exit_late(struct net *net)
4664 #ifdef CONFIG_PROC_FS
4665 remove_proc_entry("ipv6_route", net->proc_net);
4666 remove_proc_entry("rt6_stats", net->proc_net);
4670 static struct pernet_operations ip6_route_net_ops = {
4671 .init = ip6_route_net_init,
4672 .exit = ip6_route_net_exit,
4675 static int __net_init ipv6_inetpeer_init(struct net *net)
4677 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
4681 inet_peer_base_init(bp);
4682 net->ipv6.peers = bp;
4686 static void __net_exit ipv6_inetpeer_exit(struct net *net)
4688 struct inet_peer_base *bp = net->ipv6.peers;
4690 net->ipv6.peers = NULL;
4691 inetpeer_invalidate_tree(bp);
4695 static struct pernet_operations ipv6_inetpeer_ops = {
4696 .init = ipv6_inetpeer_init,
4697 .exit = ipv6_inetpeer_exit,
4700 static struct pernet_operations ip6_route_net_late_ops = {
4701 .init = ip6_route_net_init_late,
4702 .exit = ip6_route_net_exit_late,
4705 static struct notifier_block ip6_route_dev_notifier = {
4706 .notifier_call = ip6_route_dev_notify,
4707 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
4710 void __init ip6_route_init_special_entries(void)
4712 /* Registering of the loopback is done before this portion of code,
4713 * the loopback reference in rt6_info will not be taken, do it
4714 * manually for init_net */
4715 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
4716 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4717 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4718 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
4719 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4720 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
4721 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4725 int __init ip6_route_init(void)
4731 ip6_dst_ops_template.kmem_cachep =
4732 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
4733 SLAB_HWCACHE_ALIGN, NULL);
4734 if (!ip6_dst_ops_template.kmem_cachep)
4737 ret = dst_entries_init(&ip6_dst_blackhole_ops);
4739 goto out_kmem_cache;
4741 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
4743 goto out_dst_entries;
4745 ret = register_pernet_subsys(&ip6_route_net_ops);
4747 goto out_register_inetpeer;
4749 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
4753 goto out_register_subsys;
4759 ret = fib6_rules_init();
4763 ret = register_pernet_subsys(&ip6_route_net_late_ops);
4765 goto fib6_rules_init;
4768 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, 0) ||
4769 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, 0) ||
4770 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL,
4771 RTNL_FLAG_DOIT_UNLOCKED))
4772 goto out_register_late_subsys;
4774 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
4776 goto out_register_late_subsys;
4778 for_each_possible_cpu(cpu) {
4779 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
4781 INIT_LIST_HEAD(&ul->head);
4782 spin_lock_init(&ul->lock);
4788 out_register_late_subsys:
4789 unregister_pernet_subsys(&ip6_route_net_late_ops);
4791 fib6_rules_cleanup();
4796 out_register_subsys:
4797 unregister_pernet_subsys(&ip6_route_net_ops);
4798 out_register_inetpeer:
4799 unregister_pernet_subsys(&ipv6_inetpeer_ops);
4801 dst_entries_destroy(&ip6_dst_blackhole_ops);
4803 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
4807 void ip6_route_cleanup(void)
4809 unregister_netdevice_notifier(&ip6_route_dev_notifier);
4810 unregister_pernet_subsys(&ip6_route_net_late_ops);
4811 fib6_rules_cleanup();
4814 unregister_pernet_subsys(&ipv6_inetpeer_ops);
4815 unregister_pernet_subsys(&ip6_route_net_ops);
4816 dst_entries_destroy(&ip6_dst_blackhole_ops);
4817 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);