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 <net/net_namespace.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ndisc.h>
53 #include <net/addrconf.h>
55 #include <linux/rtnetlink.h>
58 #include <net/netevent.h>
59 #include <net/netlink.h>
60 #include <net/nexthop.h>
62 #include <asm/uaccess.h>
65 #include <linux/sysctl.h>
69 RT6_NUD_FAIL_HARD = -3,
70 RT6_NUD_FAIL_PROBE = -2,
71 RT6_NUD_FAIL_DO_RR = -1,
75 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
76 const struct in6_addr *dest);
77 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
78 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
79 static unsigned int ip6_mtu(const struct dst_entry *dst);
80 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
81 static void ip6_dst_destroy(struct dst_entry *);
82 static void ip6_dst_ifdown(struct dst_entry *,
83 struct net_device *dev, int how);
84 static int ip6_dst_gc(struct dst_ops *ops);
86 static int ip6_pkt_discard(struct sk_buff *skb);
87 static int ip6_pkt_discard_out(struct sk_buff *skb);
88 static int ip6_pkt_prohibit(struct sk_buff *skb);
89 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
90 static void ip6_link_failure(struct sk_buff *skb);
91 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
92 struct sk_buff *skb, u32 mtu);
93 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
95 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
97 #ifdef CONFIG_IPV6_ROUTE_INFO
98 static struct rt6_info *rt6_add_route_info(struct net *net,
99 const struct in6_addr *prefix, int prefixlen,
100 const struct in6_addr *gwaddr, int ifindex,
102 static struct rt6_info *rt6_get_route_info(struct net *net,
103 const struct in6_addr *prefix, int prefixlen,
104 const struct in6_addr *gwaddr, int ifindex);
107 static void rt6_bind_peer(struct rt6_info *rt, int create)
109 struct inet_peer_base *base;
110 struct inet_peer *peer;
112 base = inetpeer_base_ptr(rt->_rt6i_peer);
116 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
118 if (!rt6_set_peer(rt, peer))
123 static struct inet_peer *__rt6_get_peer(struct rt6_info *rt, int create)
125 if (rt6_has_peer(rt))
126 return rt6_peer_ptr(rt);
128 rt6_bind_peer(rt, create);
129 return (rt6_has_peer(rt) ? rt6_peer_ptr(rt) : NULL);
132 static struct inet_peer *rt6_get_peer_create(struct rt6_info *rt)
134 return __rt6_get_peer(rt, 1);
137 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
139 struct rt6_info *rt = (struct rt6_info *) dst;
140 struct inet_peer *peer;
143 if (!(rt->dst.flags & DST_HOST))
146 peer = rt6_get_peer_create(rt);
148 u32 *old_p = __DST_METRICS_PTR(old);
149 unsigned long prev, new;
152 if (inet_metrics_new(peer))
153 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
155 new = (unsigned long) p;
156 prev = cmpxchg(&dst->_metrics, old, new);
159 p = __DST_METRICS_PTR(prev);
160 if (prev & DST_METRICS_READ_ONLY)
167 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
171 struct in6_addr *p = &rt->rt6i_gateway;
173 if (!ipv6_addr_any(p))
174 return (const void *) p;
176 return &ipv6_hdr(skb)->daddr;
180 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
184 struct rt6_info *rt = (struct rt6_info *) dst;
187 daddr = choose_neigh_daddr(rt, skb, daddr);
188 n = __ipv6_neigh_lookup(dst->dev, daddr);
191 return neigh_create(&nd_tbl, daddr, dst->dev);
194 static struct dst_ops ip6_dst_ops_template = {
196 .protocol = cpu_to_be16(ETH_P_IPV6),
199 .check = ip6_dst_check,
200 .default_advmss = ip6_default_advmss,
202 .cow_metrics = ipv6_cow_metrics,
203 .destroy = ip6_dst_destroy,
204 .ifdown = ip6_dst_ifdown,
205 .negative_advice = ip6_negative_advice,
206 .link_failure = ip6_link_failure,
207 .update_pmtu = ip6_rt_update_pmtu,
208 .redirect = rt6_do_redirect,
209 .local_out = __ip6_local_out,
210 .neigh_lookup = ip6_neigh_lookup,
213 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
215 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
217 return mtu ? : dst->dev->mtu;
220 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
221 struct sk_buff *skb, u32 mtu)
225 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
230 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
236 static struct dst_ops ip6_dst_blackhole_ops = {
238 .protocol = cpu_to_be16(ETH_P_IPV6),
239 .destroy = ip6_dst_destroy,
240 .check = ip6_dst_check,
241 .mtu = ip6_blackhole_mtu,
242 .default_advmss = ip6_default_advmss,
243 .update_pmtu = ip6_rt_blackhole_update_pmtu,
244 .redirect = ip6_rt_blackhole_redirect,
245 .cow_metrics = ip6_rt_blackhole_cow_metrics,
246 .neigh_lookup = ip6_neigh_lookup,
249 static const u32 ip6_template_metrics[RTAX_MAX] = {
250 [RTAX_HOPLIMIT - 1] = 0,
253 static const struct rt6_info ip6_null_entry_template = {
255 .__refcnt = ATOMIC_INIT(1),
257 .obsolete = DST_OBSOLETE_FORCE_CHK,
258 .error = -ENETUNREACH,
259 .input = ip6_pkt_discard,
260 .output = ip6_pkt_discard_out,
262 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
263 .rt6i_protocol = RTPROT_KERNEL,
264 .rt6i_metric = ~(u32) 0,
265 .rt6i_ref = ATOMIC_INIT(1),
268 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
270 static const struct rt6_info ip6_prohibit_entry_template = {
272 .__refcnt = ATOMIC_INIT(1),
274 .obsolete = DST_OBSOLETE_FORCE_CHK,
276 .input = ip6_pkt_prohibit,
277 .output = ip6_pkt_prohibit_out,
279 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
280 .rt6i_protocol = RTPROT_KERNEL,
281 .rt6i_metric = ~(u32) 0,
282 .rt6i_ref = ATOMIC_INIT(1),
285 static const struct rt6_info ip6_blk_hole_entry_template = {
287 .__refcnt = ATOMIC_INIT(1),
289 .obsolete = DST_OBSOLETE_FORCE_CHK,
291 .input = dst_discard,
292 .output = dst_discard,
294 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
295 .rt6i_protocol = RTPROT_KERNEL,
296 .rt6i_metric = ~(u32) 0,
297 .rt6i_ref = ATOMIC_INIT(1),
302 /* allocate dst with ip6_dst_ops */
303 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
304 struct net_device *dev,
306 struct fib6_table *table)
308 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
309 0, DST_OBSOLETE_FORCE_CHK, flags);
312 struct dst_entry *dst = &rt->dst;
314 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
315 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
316 rt->rt6i_genid = rt_genid_ipv6(net);
317 INIT_LIST_HEAD(&rt->rt6i_siblings);
322 static void ip6_dst_destroy(struct dst_entry *dst)
324 struct rt6_info *rt = (struct rt6_info *)dst;
325 struct inet6_dev *idev = rt->rt6i_idev;
326 struct dst_entry *from = dst->from;
328 if (!(rt->dst.flags & DST_HOST))
329 dst_destroy_metrics_generic(dst);
332 rt->rt6i_idev = NULL;
339 if (rt6_has_peer(rt)) {
340 struct inet_peer *peer = rt6_peer_ptr(rt);
345 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
348 struct rt6_info *rt = (struct rt6_info *)dst;
349 struct inet6_dev *idev = rt->rt6i_idev;
350 struct net_device *loopback_dev =
351 dev_net(dev)->loopback_dev;
353 if (dev != loopback_dev) {
354 if (idev && idev->dev == dev) {
355 struct inet6_dev *loopback_idev =
356 in6_dev_get(loopback_dev);
358 rt->rt6i_idev = loopback_idev;
365 static bool rt6_check_expired(const struct rt6_info *rt)
367 if (rt->rt6i_flags & RTF_EXPIRES) {
368 if (time_after(jiffies, rt->dst.expires))
370 } else if (rt->dst.from) {
371 return rt6_check_expired((struct rt6_info *) rt->dst.from);
376 static bool rt6_need_strict(const struct in6_addr *daddr)
378 return ipv6_addr_type(daddr) &
379 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
382 /* Multipath route selection:
383 * Hash based function using packet header and flowlabel.
384 * Adapted from fib_info_hashfn()
386 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
387 const struct flowi6 *fl6)
389 unsigned int val = fl6->flowi6_proto;
391 val ^= ipv6_addr_hash(&fl6->daddr);
392 val ^= ipv6_addr_hash(&fl6->saddr);
394 /* Work only if this not encapsulated */
395 switch (fl6->flowi6_proto) {
399 val ^= (__force u16)fl6->fl6_sport;
400 val ^= (__force u16)fl6->fl6_dport;
404 val ^= (__force u16)fl6->fl6_icmp_type;
405 val ^= (__force u16)fl6->fl6_icmp_code;
408 /* RFC6438 recommands to use flowlabel */
409 val ^= (__force u32)fl6->flowlabel;
411 /* Perhaps, we need to tune, this function? */
412 val = val ^ (val >> 7) ^ (val >> 12);
413 return val % candidate_count;
416 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
417 struct flowi6 *fl6, int oif,
420 struct rt6_info *sibling, *next_sibling;
423 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
424 /* Don't change the route, if route_choosen == 0
425 * (siblings does not include ourself)
428 list_for_each_entry_safe(sibling, next_sibling,
429 &match->rt6i_siblings, rt6i_siblings) {
431 if (route_choosen == 0) {
432 if (rt6_score_route(sibling, oif, strict) < 0)
442 * Route lookup. Any table->tb6_lock is implied.
445 static inline struct rt6_info *rt6_device_match(struct net *net,
447 const struct in6_addr *saddr,
451 struct rt6_info *local = NULL;
452 struct rt6_info *sprt;
454 if (!oif && ipv6_addr_any(saddr))
457 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
458 struct net_device *dev = sprt->dst.dev;
461 if (dev->ifindex == oif)
463 if (dev->flags & IFF_LOOPBACK) {
464 if (!sprt->rt6i_idev ||
465 sprt->rt6i_idev->dev->ifindex != oif) {
466 if (flags & RT6_LOOKUP_F_IFACE && oif)
468 if (local && (!oif ||
469 local->rt6i_idev->dev->ifindex == oif))
475 if (ipv6_chk_addr(net, saddr, dev,
476 flags & RT6_LOOKUP_F_IFACE))
485 if (flags & RT6_LOOKUP_F_IFACE)
486 return net->ipv6.ip6_null_entry;
492 #ifdef CONFIG_IPV6_ROUTER_PREF
493 struct __rt6_probe_work {
494 struct work_struct work;
495 struct in6_addr target;
496 struct net_device *dev;
499 static void rt6_probe_deferred(struct work_struct *w)
501 struct in6_addr mcaddr;
502 struct __rt6_probe_work *work =
503 container_of(w, struct __rt6_probe_work, work);
505 addrconf_addr_solict_mult(&work->target, &mcaddr);
506 ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
511 static void rt6_probe(struct rt6_info *rt)
513 struct neighbour *neigh;
515 * Okay, this does not seem to be appropriate
516 * for now, however, we need to check if it
517 * is really so; aka Router Reachability Probing.
519 * Router Reachability Probe MUST be rate-limited
520 * to no more than one per minute.
522 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
525 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
527 write_lock(&neigh->lock);
528 if (neigh->nud_state & NUD_VALID)
533 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
534 struct __rt6_probe_work *work;
536 work = kmalloc(sizeof(*work), GFP_ATOMIC);
539 __neigh_set_probe_once(neigh);
542 write_unlock(&neigh->lock);
545 INIT_WORK(&work->work, rt6_probe_deferred);
546 work->target = rt->rt6i_gateway;
547 dev_hold(rt->dst.dev);
548 work->dev = rt->dst.dev;
549 schedule_work(&work->work);
553 write_unlock(&neigh->lock);
555 rcu_read_unlock_bh();
558 static inline void rt6_probe(struct rt6_info *rt)
564 * Default Router Selection (RFC 2461 6.3.6)
566 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
568 struct net_device *dev = rt->dst.dev;
569 if (!oif || dev->ifindex == oif)
571 if ((dev->flags & IFF_LOOPBACK) &&
572 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
577 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
579 struct neighbour *neigh;
580 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
582 if (rt->rt6i_flags & RTF_NONEXTHOP ||
583 !(rt->rt6i_flags & RTF_GATEWAY))
584 return RT6_NUD_SUCCEED;
587 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
589 read_lock(&neigh->lock);
590 if (neigh->nud_state & NUD_VALID)
591 ret = RT6_NUD_SUCCEED;
592 #ifdef CONFIG_IPV6_ROUTER_PREF
593 else if (!(neigh->nud_state & NUD_FAILED))
594 ret = RT6_NUD_SUCCEED;
596 ret = RT6_NUD_FAIL_PROBE;
598 read_unlock(&neigh->lock);
600 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
601 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
603 rcu_read_unlock_bh();
608 static int rt6_score_route(struct rt6_info *rt, int oif,
613 m = rt6_check_dev(rt, oif);
614 if (!m && (strict & RT6_LOOKUP_F_IFACE))
615 return RT6_NUD_FAIL_HARD;
616 #ifdef CONFIG_IPV6_ROUTER_PREF
617 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
619 if (strict & RT6_LOOKUP_F_REACHABLE) {
620 int n = rt6_check_neigh(rt);
627 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
628 int *mpri, struct rt6_info *match,
632 bool match_do_rr = false;
634 if (rt6_check_expired(rt))
637 m = rt6_score_route(rt, oif, strict);
638 if (m == RT6_NUD_FAIL_DO_RR) {
640 m = 0; /* lowest valid score */
641 } else if (m == RT6_NUD_FAIL_HARD) {
645 if (strict & RT6_LOOKUP_F_REACHABLE)
648 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
650 *do_rr = match_do_rr;
658 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
659 struct rt6_info *rr_head,
660 u32 metric, int oif, int strict,
663 struct rt6_info *rt, *match;
667 for (rt = rr_head; rt && rt->rt6i_metric == metric;
668 rt = rt->dst.rt6_next)
669 match = find_match(rt, oif, strict, &mpri, match, do_rr);
670 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
671 rt = rt->dst.rt6_next)
672 match = find_match(rt, oif, strict, &mpri, match, do_rr);
677 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
679 struct rt6_info *match, *rt0;
685 fn->rr_ptr = rt0 = fn->leaf;
687 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
691 struct rt6_info *next = rt0->dst.rt6_next;
693 /* no entries matched; do round-robin */
694 if (!next || next->rt6i_metric != rt0->rt6i_metric)
701 net = dev_net(rt0->dst.dev);
702 return match ? match : net->ipv6.ip6_null_entry;
705 #ifdef CONFIG_IPV6_ROUTE_INFO
706 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
707 const struct in6_addr *gwaddr)
709 struct net *net = dev_net(dev);
710 struct route_info *rinfo = (struct route_info *) opt;
711 struct in6_addr prefix_buf, *prefix;
713 unsigned long lifetime;
716 if (len < sizeof(struct route_info)) {
720 /* Sanity check for prefix_len and length */
721 if (rinfo->length > 3) {
723 } else if (rinfo->prefix_len > 128) {
725 } else if (rinfo->prefix_len > 64) {
726 if (rinfo->length < 2) {
729 } else if (rinfo->prefix_len > 0) {
730 if (rinfo->length < 1) {
735 pref = rinfo->route_pref;
736 if (pref == ICMPV6_ROUTER_PREF_INVALID)
739 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
741 if (rinfo->length == 3)
742 prefix = (struct in6_addr *)rinfo->prefix;
744 /* this function is safe */
745 ipv6_addr_prefix(&prefix_buf,
746 (struct in6_addr *)rinfo->prefix,
748 prefix = &prefix_buf;
751 if (rinfo->prefix_len == 0)
752 rt = rt6_get_dflt_router(gwaddr, dev);
754 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
755 gwaddr, dev->ifindex);
757 if (rt && !lifetime) {
763 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
766 rt->rt6i_flags = RTF_ROUTEINFO |
767 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
770 if (!addrconf_finite_timeout(lifetime))
771 rt6_clean_expires(rt);
773 rt6_set_expires(rt, jiffies + HZ * lifetime);
781 #define BACKTRACK(__net, saddr) \
783 if (rt == __net->ipv6.ip6_null_entry) { \
784 struct fib6_node *pn; \
786 if (fn->fn_flags & RTN_TL_ROOT) \
789 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
790 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
793 if (fn->fn_flags & RTN_RTINFO) \
799 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
800 struct fib6_table *table,
801 struct flowi6 *fl6, int flags)
803 struct fib6_node *fn;
806 read_lock_bh(&table->tb6_lock);
807 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
810 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
811 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
812 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
813 BACKTRACK(net, &fl6->saddr);
815 dst_use(&rt->dst, jiffies);
816 read_unlock_bh(&table->tb6_lock);
821 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
824 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
826 EXPORT_SYMBOL_GPL(ip6_route_lookup);
828 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
829 const struct in6_addr *saddr, int oif, int strict)
831 struct flowi6 fl6 = {
835 struct dst_entry *dst;
836 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
839 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
840 flags |= RT6_LOOKUP_F_HAS_SADDR;
843 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
845 return (struct rt6_info *) dst;
852 EXPORT_SYMBOL(rt6_lookup);
854 /* ip6_ins_rt is called with FREE table->tb6_lock.
855 It takes new route entry, the addition fails by any reason the
856 route is freed. In any case, if caller does not hold it, it may
860 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
863 struct fib6_table *table;
865 table = rt->rt6i_table;
866 write_lock_bh(&table->tb6_lock);
867 err = fib6_add(&table->tb6_root, rt, info);
868 write_unlock_bh(&table->tb6_lock);
873 int ip6_ins_rt(struct rt6_info *rt)
875 struct nl_info info = {
876 .nl_net = dev_net(rt->dst.dev),
878 return __ip6_ins_rt(rt, &info);
881 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
882 const struct in6_addr *daddr,
883 const struct in6_addr *saddr)
891 rt = ip6_rt_copy(ort, daddr);
894 if (ort->rt6i_dst.plen != 128 &&
895 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
896 rt->rt6i_flags |= RTF_ANYCAST;
898 rt->rt6i_flags |= RTF_CACHE;
900 #ifdef CONFIG_IPV6_SUBTREES
901 if (rt->rt6i_src.plen && saddr) {
902 rt->rt6i_src.addr = *saddr;
903 rt->rt6i_src.plen = 128;
911 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
912 const struct in6_addr *daddr)
914 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
917 rt->rt6i_flags |= RTF_CACHE;
921 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
922 struct flowi6 *fl6, int flags)
924 struct fib6_node *fn;
925 struct rt6_info *rt, *nrt;
929 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
931 strict |= flags & RT6_LOOKUP_F_IFACE;
934 read_lock_bh(&table->tb6_lock);
937 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
940 rt = rt6_select(fn, oif, strict | reachable);
941 if (rt->rt6i_nsiblings)
942 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
943 BACKTRACK(net, &fl6->saddr);
944 if (rt == net->ipv6.ip6_null_entry ||
945 rt->rt6i_flags & RTF_CACHE)
949 read_unlock_bh(&table->tb6_lock);
951 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
952 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
953 else if (!(rt->dst.flags & DST_HOST))
954 nrt = rt6_alloc_clone(rt, &fl6->daddr);
959 rt = nrt ? : net->ipv6.ip6_null_entry;
963 err = ip6_ins_rt(nrt);
972 * Race condition! In the gap, when table->tb6_lock was
973 * released someone could insert this route. Relookup.
984 read_unlock_bh(&table->tb6_lock);
986 rt->dst.lastuse = jiffies;
992 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
993 struct flowi6 *fl6, int flags)
995 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
998 static struct dst_entry *ip6_route_input_lookup(struct net *net,
999 struct net_device *dev,
1000 struct flowi6 *fl6, int flags)
1002 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1003 flags |= RT6_LOOKUP_F_IFACE;
1005 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1008 void ip6_route_input(struct sk_buff *skb)
1010 const struct ipv6hdr *iph = ipv6_hdr(skb);
1011 struct net *net = dev_net(skb->dev);
1012 int flags = RT6_LOOKUP_F_HAS_SADDR;
1013 struct flowi6 fl6 = {
1014 .flowi6_iif = skb->dev->ifindex,
1015 .daddr = iph->daddr,
1016 .saddr = iph->saddr,
1017 .flowlabel = ip6_flowinfo(iph),
1018 .flowi6_mark = skb->mark,
1019 .flowi6_proto = iph->nexthdr,
1022 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1025 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1026 struct flowi6 *fl6, int flags)
1028 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1031 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1036 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1038 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1039 flags |= RT6_LOOKUP_F_IFACE;
1041 if (!ipv6_addr_any(&fl6->saddr))
1042 flags |= RT6_LOOKUP_F_HAS_SADDR;
1044 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1046 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1049 EXPORT_SYMBOL(ip6_route_output);
1051 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1053 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1054 struct dst_entry *new = NULL;
1056 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1060 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1061 rt6_init_peer(rt, net->ipv6.peers);
1064 new->input = dst_discard;
1065 new->output = dst_discard;
1067 if (dst_metrics_read_only(&ort->dst))
1068 new->_metrics = ort->dst._metrics;
1070 dst_copy_metrics(new, &ort->dst);
1071 rt->rt6i_idev = ort->rt6i_idev;
1073 in6_dev_hold(rt->rt6i_idev);
1075 rt->rt6i_gateway = ort->rt6i_gateway;
1076 rt->rt6i_flags = ort->rt6i_flags;
1077 rt->rt6i_metric = 0;
1079 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1080 #ifdef CONFIG_IPV6_SUBTREES
1081 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1087 dst_release(dst_orig);
1088 return new ? new : ERR_PTR(-ENOMEM);
1092 * Destination cache support functions
1095 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1097 struct rt6_info *rt;
1099 rt = (struct rt6_info *) dst;
1101 /* All IPV6 dsts are created with ->obsolete set to the value
1102 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1103 * into this function always.
1105 if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev)))
1108 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1111 if (rt6_check_expired(rt))
1117 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1119 struct rt6_info *rt = (struct rt6_info *) dst;
1122 if (rt->rt6i_flags & RTF_CACHE) {
1123 if (rt6_check_expired(rt)) {
1135 static void ip6_link_failure(struct sk_buff *skb)
1137 struct rt6_info *rt;
1139 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1141 rt = (struct rt6_info *) skb_dst(skb);
1143 if (rt->rt6i_flags & RTF_CACHE) {
1147 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1148 rt->rt6i_node->fn_sernum = -1;
1153 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1154 struct sk_buff *skb, u32 mtu)
1156 struct rt6_info *rt6 = (struct rt6_info*)dst;
1159 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1160 struct net *net = dev_net(dst->dev);
1162 rt6->rt6i_flags |= RTF_MODIFIED;
1163 if (mtu < IPV6_MIN_MTU) {
1164 u32 features = dst_metric(dst, RTAX_FEATURES);
1166 features |= RTAX_FEATURE_ALLFRAG;
1167 dst_metric_set(dst, RTAX_FEATURES, features);
1169 dst_metric_set(dst, RTAX_MTU, mtu);
1170 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1174 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1177 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1178 struct dst_entry *dst;
1181 memset(&fl6, 0, sizeof(fl6));
1182 fl6.flowi6_oif = oif;
1183 fl6.flowi6_mark = mark;
1184 fl6.daddr = iph->daddr;
1185 fl6.saddr = iph->saddr;
1186 fl6.flowlabel = ip6_flowinfo(iph);
1188 dst = ip6_route_output(net, NULL, &fl6);
1190 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1193 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1195 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1197 ip6_update_pmtu(skb, sock_net(sk), mtu,
1198 sk->sk_bound_dev_if, sk->sk_mark);
1200 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1202 /* Handle redirects */
1203 struct ip6rd_flowi {
1205 struct in6_addr gateway;
1208 static struct rt6_info *__ip6_route_redirect(struct net *net,
1209 struct fib6_table *table,
1213 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1214 struct rt6_info *rt;
1215 struct fib6_node *fn;
1217 /* Get the "current" route for this destination and
1218 * check if the redirect has come from approriate router.
1220 * RFC 4861 specifies that redirects should only be
1221 * accepted if they come from the nexthop to the target.
1222 * Due to the way the routes are chosen, this notion
1223 * is a bit fuzzy and one might need to check all possible
1227 read_lock_bh(&table->tb6_lock);
1228 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1230 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1231 if (rt6_check_expired(rt))
1235 if (!(rt->rt6i_flags & RTF_GATEWAY))
1237 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1239 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1245 rt = net->ipv6.ip6_null_entry;
1246 else if (rt->dst.error) {
1247 rt = net->ipv6.ip6_null_entry;
1250 BACKTRACK(net, &fl6->saddr);
1254 read_unlock_bh(&table->tb6_lock);
1259 static struct dst_entry *ip6_route_redirect(struct net *net,
1260 const struct flowi6 *fl6,
1261 const struct in6_addr *gateway)
1263 int flags = RT6_LOOKUP_F_HAS_SADDR;
1264 struct ip6rd_flowi rdfl;
1267 rdfl.gateway = *gateway;
1269 return fib6_rule_lookup(net, &rdfl.fl6,
1270 flags, __ip6_route_redirect);
1273 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1275 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1276 struct dst_entry *dst;
1279 memset(&fl6, 0, sizeof(fl6));
1280 fl6.flowi6_oif = oif;
1281 fl6.flowi6_mark = mark;
1282 fl6.daddr = iph->daddr;
1283 fl6.saddr = iph->saddr;
1284 fl6.flowlabel = ip6_flowinfo(iph);
1286 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1287 rt6_do_redirect(dst, NULL, skb);
1290 EXPORT_SYMBOL_GPL(ip6_redirect);
1292 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1295 const struct ipv6hdr *iph = ipv6_hdr(skb);
1296 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1297 struct dst_entry *dst;
1300 memset(&fl6, 0, sizeof(fl6));
1301 fl6.flowi6_oif = oif;
1302 fl6.flowi6_mark = mark;
1303 fl6.daddr = msg->dest;
1304 fl6.saddr = iph->daddr;
1306 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1307 rt6_do_redirect(dst, NULL, skb);
1311 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1313 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1315 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1317 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1319 struct net_device *dev = dst->dev;
1320 unsigned int mtu = dst_mtu(dst);
1321 struct net *net = dev_net(dev);
1323 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1325 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1326 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1329 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1330 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1331 * IPV6_MAXPLEN is also valid and means: "any MSS,
1332 * rely only on pmtu discovery"
1334 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1339 static unsigned int ip6_mtu(const struct dst_entry *dst)
1341 struct inet6_dev *idev;
1342 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1350 idev = __in6_dev_get(dst->dev);
1352 mtu = idev->cnf.mtu6;
1356 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1359 static struct dst_entry *icmp6_dst_gc_list;
1360 static DEFINE_SPINLOCK(icmp6_dst_lock);
1362 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1365 struct dst_entry *dst;
1366 struct rt6_info *rt;
1367 struct inet6_dev *idev = in6_dev_get(dev);
1368 struct net *net = dev_net(dev);
1370 if (unlikely(!idev))
1371 return ERR_PTR(-ENODEV);
1373 rt = ip6_dst_alloc(net, dev, 0, NULL);
1374 if (unlikely(!rt)) {
1376 dst = ERR_PTR(-ENOMEM);
1380 rt->dst.flags |= DST_HOST;
1381 rt->dst.output = ip6_output;
1382 atomic_set(&rt->dst.__refcnt, 1);
1383 rt->rt6i_gateway = fl6->daddr;
1384 rt->rt6i_dst.addr = fl6->daddr;
1385 rt->rt6i_dst.plen = 128;
1386 rt->rt6i_idev = idev;
1387 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1389 spin_lock_bh(&icmp6_dst_lock);
1390 rt->dst.next = icmp6_dst_gc_list;
1391 icmp6_dst_gc_list = &rt->dst;
1392 spin_unlock_bh(&icmp6_dst_lock);
1394 fib6_force_start_gc(net);
1396 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1402 int icmp6_dst_gc(void)
1404 struct dst_entry *dst, **pprev;
1407 spin_lock_bh(&icmp6_dst_lock);
1408 pprev = &icmp6_dst_gc_list;
1410 while ((dst = *pprev) != NULL) {
1411 if (!atomic_read(&dst->__refcnt)) {
1420 spin_unlock_bh(&icmp6_dst_lock);
1425 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1428 struct dst_entry *dst, **pprev;
1430 spin_lock_bh(&icmp6_dst_lock);
1431 pprev = &icmp6_dst_gc_list;
1432 while ((dst = *pprev) != NULL) {
1433 struct rt6_info *rt = (struct rt6_info *) dst;
1434 if (func(rt, arg)) {
1441 spin_unlock_bh(&icmp6_dst_lock);
1444 static int ip6_dst_gc(struct dst_ops *ops)
1446 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1447 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1448 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1449 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1450 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1451 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1454 entries = dst_entries_get_fast(ops);
1455 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1456 entries <= rt_max_size)
1459 net->ipv6.ip6_rt_gc_expire++;
1460 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1461 entries = dst_entries_get_slow(ops);
1462 if (entries < ops->gc_thresh)
1463 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1465 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1466 return entries > rt_max_size;
1473 int ip6_route_add(struct fib6_config *cfg)
1476 struct net *net = cfg->fc_nlinfo.nl_net;
1477 struct rt6_info *rt = NULL;
1478 struct net_device *dev = NULL;
1479 struct inet6_dev *idev = NULL;
1480 struct fib6_table *table;
1483 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1485 #ifndef CONFIG_IPV6_SUBTREES
1486 if (cfg->fc_src_len)
1489 if (cfg->fc_ifindex) {
1491 dev = dev_get_by_index(net, cfg->fc_ifindex);
1494 idev = in6_dev_get(dev);
1499 if (cfg->fc_metric == 0)
1500 cfg->fc_metric = IP6_RT_PRIO_USER;
1503 if (cfg->fc_nlinfo.nlh &&
1504 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1505 table = fib6_get_table(net, cfg->fc_table);
1507 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1508 table = fib6_new_table(net, cfg->fc_table);
1511 table = fib6_new_table(net, cfg->fc_table);
1517 rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1524 if (cfg->fc_flags & RTF_EXPIRES)
1525 rt6_set_expires(rt, jiffies +
1526 clock_t_to_jiffies(cfg->fc_expires));
1528 rt6_clean_expires(rt);
1530 if (cfg->fc_protocol == RTPROT_UNSPEC)
1531 cfg->fc_protocol = RTPROT_BOOT;
1532 rt->rt6i_protocol = cfg->fc_protocol;
1534 addr_type = ipv6_addr_type(&cfg->fc_dst);
1536 if (addr_type & IPV6_ADDR_MULTICAST)
1537 rt->dst.input = ip6_mc_input;
1538 else if (cfg->fc_flags & RTF_LOCAL)
1539 rt->dst.input = ip6_input;
1541 rt->dst.input = ip6_forward;
1543 rt->dst.output = ip6_output;
1545 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1546 rt->rt6i_dst.plen = cfg->fc_dst_len;
1547 if (rt->rt6i_dst.plen == 128)
1548 rt->dst.flags |= DST_HOST;
1550 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1551 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1556 dst_init_metrics(&rt->dst, metrics, 0);
1558 #ifdef CONFIG_IPV6_SUBTREES
1559 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1560 rt->rt6i_src.plen = cfg->fc_src_len;
1563 rt->rt6i_metric = cfg->fc_metric;
1565 /* We cannot add true routes via loopback here,
1566 they would result in kernel looping; promote them to reject routes
1568 if ((cfg->fc_flags & RTF_REJECT) ||
1569 (dev && (dev->flags & IFF_LOOPBACK) &&
1570 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1571 !(cfg->fc_flags & RTF_LOCAL))) {
1572 /* hold loopback dev/idev if we haven't done so. */
1573 if (dev != net->loopback_dev) {
1578 dev = net->loopback_dev;
1580 idev = in6_dev_get(dev);
1586 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1587 switch (cfg->fc_type) {
1589 rt->dst.error = -EINVAL;
1590 rt->dst.output = dst_discard;
1591 rt->dst.input = dst_discard;
1594 rt->dst.error = -EACCES;
1595 rt->dst.output = ip6_pkt_prohibit_out;
1596 rt->dst.input = ip6_pkt_prohibit;
1600 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1602 rt->dst.output = ip6_pkt_discard_out;
1603 rt->dst.input = ip6_pkt_discard;
1609 if (cfg->fc_flags & RTF_GATEWAY) {
1610 const struct in6_addr *gw_addr;
1613 gw_addr = &cfg->fc_gateway;
1614 rt->rt6i_gateway = *gw_addr;
1615 gwa_type = ipv6_addr_type(gw_addr);
1617 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1618 struct rt6_info *grt;
1620 /* IPv6 strictly inhibits using not link-local
1621 addresses as nexthop address.
1622 Otherwise, router will not able to send redirects.
1623 It is very good, but in some (rare!) circumstances
1624 (SIT, PtP, NBMA NOARP links) it is handy to allow
1625 some exceptions. --ANK
1628 if (!(gwa_type & IPV6_ADDR_UNICAST))
1631 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1633 err = -EHOSTUNREACH;
1637 if (dev != grt->dst.dev) {
1643 idev = grt->rt6i_idev;
1645 in6_dev_hold(grt->rt6i_idev);
1647 if (!(grt->rt6i_flags & RTF_GATEWAY))
1655 if (!dev || (dev->flags & IFF_LOOPBACK))
1663 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1664 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1668 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1669 rt->rt6i_prefsrc.plen = 128;
1671 rt->rt6i_prefsrc.plen = 0;
1673 rt->rt6i_flags = cfg->fc_flags;
1680 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1681 int type = nla_type(nla);
1684 if (type > RTAX_MAX) {
1689 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1695 rt->rt6i_idev = idev;
1696 rt->rt6i_table = table;
1698 cfg->fc_nlinfo.nl_net = dev_net(dev);
1700 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1712 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1715 struct fib6_table *table;
1716 struct net *net = dev_net(rt->dst.dev);
1718 if (rt == net->ipv6.ip6_null_entry) {
1723 table = rt->rt6i_table;
1724 write_lock_bh(&table->tb6_lock);
1725 err = fib6_del(rt, info);
1726 write_unlock_bh(&table->tb6_lock);
1733 int ip6_del_rt(struct rt6_info *rt)
1735 struct nl_info info = {
1736 .nl_net = dev_net(rt->dst.dev),
1738 return __ip6_del_rt(rt, &info);
1741 static int ip6_route_del(struct fib6_config *cfg)
1743 struct fib6_table *table;
1744 struct fib6_node *fn;
1745 struct rt6_info *rt;
1748 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1752 read_lock_bh(&table->tb6_lock);
1754 fn = fib6_locate(&table->tb6_root,
1755 &cfg->fc_dst, cfg->fc_dst_len,
1756 &cfg->fc_src, cfg->fc_src_len);
1759 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1760 if (cfg->fc_ifindex &&
1762 rt->dst.dev->ifindex != cfg->fc_ifindex))
1764 if (cfg->fc_flags & RTF_GATEWAY &&
1765 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1767 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1770 read_unlock_bh(&table->tb6_lock);
1772 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1775 read_unlock_bh(&table->tb6_lock);
1780 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1782 struct net *net = dev_net(skb->dev);
1783 struct netevent_redirect netevent;
1784 struct rt6_info *rt, *nrt = NULL;
1785 struct ndisc_options ndopts;
1786 struct inet6_dev *in6_dev;
1787 struct neighbour *neigh;
1789 int optlen, on_link;
1792 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1793 optlen -= sizeof(*msg);
1796 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1800 msg = (struct rd_msg *)icmp6_hdr(skb);
1802 if (ipv6_addr_is_multicast(&msg->dest)) {
1803 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1808 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1810 } else if (ipv6_addr_type(&msg->target) !=
1811 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1812 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1816 in6_dev = __in6_dev_get(skb->dev);
1819 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1823 * The IP source address of the Redirect MUST be the same as the current
1824 * first-hop router for the specified ICMP Destination Address.
1827 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1828 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1833 if (ndopts.nd_opts_tgt_lladdr) {
1834 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1837 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1842 rt = (struct rt6_info *) dst;
1843 if (rt == net->ipv6.ip6_null_entry) {
1844 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1848 /* Redirect received -> path was valid.
1849 * Look, redirects are sent only in response to data packets,
1850 * so that this nexthop apparently is reachable. --ANK
1852 dst_confirm(&rt->dst);
1854 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1859 * We have finally decided to accept it.
1862 neigh_update(neigh, lladdr, NUD_STALE,
1863 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1864 NEIGH_UPDATE_F_OVERRIDE|
1865 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1866 NEIGH_UPDATE_F_ISROUTER))
1869 nrt = ip6_rt_copy(rt, &msg->dest);
1873 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1875 nrt->rt6i_flags &= ~RTF_GATEWAY;
1877 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1879 if (ip6_ins_rt(nrt))
1882 netevent.old = &rt->dst;
1883 netevent.new = &nrt->dst;
1884 netevent.daddr = &msg->dest;
1885 netevent.neigh = neigh;
1886 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1888 if (rt->rt6i_flags & RTF_CACHE) {
1889 rt = (struct rt6_info *) dst_clone(&rt->dst);
1894 neigh_release(neigh);
1898 * Misc support functions
1901 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1902 const struct in6_addr *dest)
1904 struct net *net = dev_net(ort->dst.dev);
1905 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1909 rt->dst.input = ort->dst.input;
1910 rt->dst.output = ort->dst.output;
1911 rt->dst.flags |= DST_HOST;
1913 rt->rt6i_dst.addr = *dest;
1914 rt->rt6i_dst.plen = 128;
1915 dst_copy_metrics(&rt->dst, &ort->dst);
1916 rt->dst.error = ort->dst.error;
1917 rt->rt6i_idev = ort->rt6i_idev;
1919 in6_dev_hold(rt->rt6i_idev);
1920 rt->dst.lastuse = jiffies;
1922 if (ort->rt6i_flags & RTF_GATEWAY)
1923 rt->rt6i_gateway = ort->rt6i_gateway;
1925 rt->rt6i_gateway = *dest;
1926 rt->rt6i_flags = ort->rt6i_flags;
1927 rt6_set_from(rt, ort);
1928 rt->rt6i_metric = 0;
1930 #ifdef CONFIG_IPV6_SUBTREES
1931 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1933 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1934 rt->rt6i_table = ort->rt6i_table;
1939 #ifdef CONFIG_IPV6_ROUTE_INFO
1940 static struct rt6_info *rt6_get_route_info(struct net *net,
1941 const struct in6_addr *prefix, int prefixlen,
1942 const struct in6_addr *gwaddr, int ifindex)
1944 struct fib6_node *fn;
1945 struct rt6_info *rt = NULL;
1946 struct fib6_table *table;
1948 table = fib6_get_table(net, RT6_TABLE_INFO);
1952 read_lock_bh(&table->tb6_lock);
1953 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1957 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1958 if (rt->dst.dev->ifindex != ifindex)
1960 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1962 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1968 read_unlock_bh(&table->tb6_lock);
1972 static struct rt6_info *rt6_add_route_info(struct net *net,
1973 const struct in6_addr *prefix, int prefixlen,
1974 const struct in6_addr *gwaddr, int ifindex,
1977 struct fib6_config cfg = {
1978 .fc_table = RT6_TABLE_INFO,
1979 .fc_metric = IP6_RT_PRIO_USER,
1980 .fc_ifindex = ifindex,
1981 .fc_dst_len = prefixlen,
1982 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1983 RTF_UP | RTF_PREF(pref),
1984 .fc_nlinfo.portid = 0,
1985 .fc_nlinfo.nlh = NULL,
1986 .fc_nlinfo.nl_net = net,
1989 cfg.fc_dst = *prefix;
1990 cfg.fc_gateway = *gwaddr;
1992 /* We should treat it as a default route if prefix length is 0. */
1994 cfg.fc_flags |= RTF_DEFAULT;
1996 ip6_route_add(&cfg);
1998 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2002 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2004 struct rt6_info *rt;
2005 struct fib6_table *table;
2007 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2011 read_lock_bh(&table->tb6_lock);
2012 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
2013 if (dev == rt->dst.dev &&
2014 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2015 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2020 read_unlock_bh(&table->tb6_lock);
2024 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2025 struct net_device *dev,
2028 struct fib6_config cfg = {
2029 .fc_table = RT6_TABLE_DFLT,
2030 .fc_metric = IP6_RT_PRIO_USER,
2031 .fc_ifindex = dev->ifindex,
2032 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2033 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2034 .fc_nlinfo.portid = 0,
2035 .fc_nlinfo.nlh = NULL,
2036 .fc_nlinfo.nl_net = dev_net(dev),
2039 cfg.fc_gateway = *gwaddr;
2041 ip6_route_add(&cfg);
2043 return rt6_get_dflt_router(gwaddr, dev);
2046 void rt6_purge_dflt_routers(struct net *net)
2048 struct rt6_info *rt;
2049 struct fib6_table *table;
2051 /* NOTE: Keep consistent with rt6_get_dflt_router */
2052 table = fib6_get_table(net, RT6_TABLE_DFLT);
2057 read_lock_bh(&table->tb6_lock);
2058 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2059 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2060 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2062 read_unlock_bh(&table->tb6_lock);
2067 read_unlock_bh(&table->tb6_lock);
2070 static void rtmsg_to_fib6_config(struct net *net,
2071 struct in6_rtmsg *rtmsg,
2072 struct fib6_config *cfg)
2074 memset(cfg, 0, sizeof(*cfg));
2076 cfg->fc_table = RT6_TABLE_MAIN;
2077 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2078 cfg->fc_metric = rtmsg->rtmsg_metric;
2079 cfg->fc_expires = rtmsg->rtmsg_info;
2080 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2081 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2082 cfg->fc_flags = rtmsg->rtmsg_flags;
2084 cfg->fc_nlinfo.nl_net = net;
2086 cfg->fc_dst = rtmsg->rtmsg_dst;
2087 cfg->fc_src = rtmsg->rtmsg_src;
2088 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2091 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2093 struct fib6_config cfg;
2094 struct in6_rtmsg rtmsg;
2098 case SIOCADDRT: /* Add a route */
2099 case SIOCDELRT: /* Delete a route */
2100 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2102 err = copy_from_user(&rtmsg, arg,
2103 sizeof(struct in6_rtmsg));
2107 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2112 err = ip6_route_add(&cfg);
2115 err = ip6_route_del(&cfg);
2129 * Drop the packet on the floor
2132 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2135 struct dst_entry *dst = skb_dst(skb);
2136 switch (ipstats_mib_noroutes) {
2137 case IPSTATS_MIB_INNOROUTES:
2138 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2139 if (type == IPV6_ADDR_ANY) {
2140 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2141 IPSTATS_MIB_INADDRERRORS);
2145 case IPSTATS_MIB_OUTNOROUTES:
2146 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2147 ipstats_mib_noroutes);
2150 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2155 static int ip6_pkt_discard(struct sk_buff *skb)
2157 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2160 static int ip6_pkt_discard_out(struct sk_buff *skb)
2162 skb->dev = skb_dst(skb)->dev;
2163 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2166 static int ip6_pkt_prohibit(struct sk_buff *skb)
2168 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2171 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2173 skb->dev = skb_dst(skb)->dev;
2174 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2178 * Allocate a dst for local (unicast / anycast) address.
2181 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2182 const struct in6_addr *addr,
2185 struct net *net = dev_net(idev->dev);
2186 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2189 return ERR_PTR(-ENOMEM);
2193 rt->dst.flags |= DST_HOST;
2194 rt->dst.input = ip6_input;
2195 rt->dst.output = ip6_output;
2196 rt->rt6i_idev = idev;
2198 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2200 rt->rt6i_flags |= RTF_ANYCAST;
2202 rt->rt6i_flags |= RTF_LOCAL;
2204 rt->rt6i_gateway = *addr;
2205 rt->rt6i_dst.addr = *addr;
2206 rt->rt6i_dst.plen = 128;
2207 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2209 atomic_set(&rt->dst.__refcnt, 1);
2214 int ip6_route_get_saddr(struct net *net,
2215 struct rt6_info *rt,
2216 const struct in6_addr *daddr,
2218 struct in6_addr *saddr)
2220 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2222 if (rt->rt6i_prefsrc.plen)
2223 *saddr = rt->rt6i_prefsrc.addr;
2225 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2226 daddr, prefs, saddr);
2230 /* remove deleted ip from prefsrc entries */
2231 struct arg_dev_net_ip {
2232 struct net_device *dev;
2234 struct in6_addr *addr;
2237 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2239 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2240 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2241 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2243 if (((void *)rt->dst.dev == dev || !dev) &&
2244 rt != net->ipv6.ip6_null_entry &&
2245 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2246 /* remove prefsrc entry */
2247 rt->rt6i_prefsrc.plen = 0;
2252 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2254 struct net *net = dev_net(ifp->idev->dev);
2255 struct arg_dev_net_ip adni = {
2256 .dev = ifp->idev->dev,
2260 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2263 struct arg_dev_net {
2264 struct net_device *dev;
2268 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2270 const struct arg_dev_net *adn = arg;
2271 const struct net_device *dev = adn->dev;
2273 if ((rt->dst.dev == dev || !dev) &&
2274 rt != adn->net->ipv6.ip6_null_entry)
2280 void rt6_ifdown(struct net *net, struct net_device *dev)
2282 struct arg_dev_net adn = {
2287 fib6_clean_all(net, fib6_ifdown, &adn);
2288 icmp6_clean_all(fib6_ifdown, &adn);
2291 struct rt6_mtu_change_arg {
2292 struct net_device *dev;
2296 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2298 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2299 struct inet6_dev *idev;
2301 /* In IPv6 pmtu discovery is not optional,
2302 so that RTAX_MTU lock cannot disable it.
2303 We still use this lock to block changes
2304 caused by addrconf/ndisc.
2307 idev = __in6_dev_get(arg->dev);
2311 /* For administrative MTU increase, there is no way to discover
2312 IPv6 PMTU increase, so PMTU increase should be updated here.
2313 Since RFC 1981 doesn't include administrative MTU increase
2314 update PMTU increase is a MUST. (i.e. jumbo frame)
2317 If new MTU is less than route PMTU, this new MTU will be the
2318 lowest MTU in the path, update the route PMTU to reflect PMTU
2319 decreases; if new MTU is greater than route PMTU, and the
2320 old MTU is the lowest MTU in the path, update the route PMTU
2321 to reflect the increase. In this case if the other nodes' MTU
2322 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2325 if (rt->dst.dev == arg->dev &&
2326 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2327 (dst_mtu(&rt->dst) >= arg->mtu ||
2328 (dst_mtu(&rt->dst) < arg->mtu &&
2329 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2330 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2335 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2337 struct rt6_mtu_change_arg arg = {
2342 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2345 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2346 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2347 [RTA_OIF] = { .type = NLA_U32 },
2348 [RTA_IIF] = { .type = NLA_U32 },
2349 [RTA_PRIORITY] = { .type = NLA_U32 },
2350 [RTA_METRICS] = { .type = NLA_NESTED },
2351 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2354 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2355 struct fib6_config *cfg)
2358 struct nlattr *tb[RTA_MAX+1];
2361 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2366 rtm = nlmsg_data(nlh);
2367 memset(cfg, 0, sizeof(*cfg));
2369 cfg->fc_table = rtm->rtm_table;
2370 cfg->fc_dst_len = rtm->rtm_dst_len;
2371 cfg->fc_src_len = rtm->rtm_src_len;
2372 cfg->fc_flags = RTF_UP;
2373 cfg->fc_protocol = rtm->rtm_protocol;
2374 cfg->fc_type = rtm->rtm_type;
2376 if (rtm->rtm_type == RTN_UNREACHABLE ||
2377 rtm->rtm_type == RTN_BLACKHOLE ||
2378 rtm->rtm_type == RTN_PROHIBIT ||
2379 rtm->rtm_type == RTN_THROW)
2380 cfg->fc_flags |= RTF_REJECT;
2382 if (rtm->rtm_type == RTN_LOCAL)
2383 cfg->fc_flags |= RTF_LOCAL;
2385 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2386 cfg->fc_nlinfo.nlh = nlh;
2387 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2389 if (tb[RTA_GATEWAY]) {
2390 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2391 cfg->fc_flags |= RTF_GATEWAY;
2395 int plen = (rtm->rtm_dst_len + 7) >> 3;
2397 if (nla_len(tb[RTA_DST]) < plen)
2400 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2404 int plen = (rtm->rtm_src_len + 7) >> 3;
2406 if (nla_len(tb[RTA_SRC]) < plen)
2409 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2412 if (tb[RTA_PREFSRC])
2413 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2416 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2418 if (tb[RTA_PRIORITY])
2419 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2421 if (tb[RTA_METRICS]) {
2422 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2423 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2427 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2429 if (tb[RTA_MULTIPATH]) {
2430 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2431 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2439 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2441 struct fib6_config r_cfg;
2442 struct rtnexthop *rtnh;
2445 int err = 0, last_err = 0;
2448 rtnh = (struct rtnexthop *)cfg->fc_mp;
2449 remaining = cfg->fc_mp_len;
2451 /* Parse a Multipath Entry */
2452 while (rtnh_ok(rtnh, remaining)) {
2453 memcpy(&r_cfg, cfg, sizeof(*cfg));
2454 if (rtnh->rtnh_ifindex)
2455 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2457 attrlen = rtnh_attrlen(rtnh);
2459 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2461 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2463 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2464 r_cfg.fc_flags |= RTF_GATEWAY;
2467 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2470 /* If we are trying to remove a route, do not stop the
2471 * loop when ip6_route_del() fails (because next hop is
2472 * already gone), we should try to remove all next hops.
2475 /* If add fails, we should try to delete all
2476 * next hops that have been already added.
2482 /* Because each route is added like a single route we remove
2483 * this flag after the first nexthop (if there is a collision,
2484 * we have already fail to add the first nexthop:
2485 * fib6_add_rt2node() has reject it).
2487 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2488 rtnh = rtnh_next(rtnh, &remaining);
2494 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2496 struct fib6_config cfg;
2499 err = rtm_to_fib6_config(skb, nlh, &cfg);
2504 return ip6_route_multipath(&cfg, 0);
2506 return ip6_route_del(&cfg);
2509 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2511 struct fib6_config cfg;
2514 err = rtm_to_fib6_config(skb, nlh, &cfg);
2519 return ip6_route_multipath(&cfg, 1);
2521 return ip6_route_add(&cfg);
2524 static inline size_t rt6_nlmsg_size(void)
2526 return NLMSG_ALIGN(sizeof(struct rtmsg))
2527 + nla_total_size(16) /* RTA_SRC */
2528 + nla_total_size(16) /* RTA_DST */
2529 + nla_total_size(16) /* RTA_GATEWAY */
2530 + nla_total_size(16) /* RTA_PREFSRC */
2531 + nla_total_size(4) /* RTA_TABLE */
2532 + nla_total_size(4) /* RTA_IIF */
2533 + nla_total_size(4) /* RTA_OIF */
2534 + nla_total_size(4) /* RTA_PRIORITY */
2535 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2536 + nla_total_size(sizeof(struct rta_cacheinfo));
2539 static int rt6_fill_node(struct net *net,
2540 struct sk_buff *skb, struct rt6_info *rt,
2541 struct in6_addr *dst, struct in6_addr *src,
2542 int iif, int type, u32 portid, u32 seq,
2543 int prefix, int nowait, unsigned int flags)
2546 struct nlmsghdr *nlh;
2550 if (prefix) { /* user wants prefix routes only */
2551 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2552 /* success since this is not a prefix route */
2557 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2561 rtm = nlmsg_data(nlh);
2562 rtm->rtm_family = AF_INET6;
2563 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2564 rtm->rtm_src_len = rt->rt6i_src.plen;
2567 table = rt->rt6i_table->tb6_id;
2569 table = RT6_TABLE_UNSPEC;
2570 rtm->rtm_table = table;
2571 if (nla_put_u32(skb, RTA_TABLE, table))
2572 goto nla_put_failure;
2573 if (rt->rt6i_flags & RTF_REJECT) {
2574 switch (rt->dst.error) {
2576 rtm->rtm_type = RTN_BLACKHOLE;
2579 rtm->rtm_type = RTN_PROHIBIT;
2582 rtm->rtm_type = RTN_THROW;
2585 rtm->rtm_type = RTN_UNREACHABLE;
2589 else if (rt->rt6i_flags & RTF_LOCAL)
2590 rtm->rtm_type = RTN_LOCAL;
2591 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2592 rtm->rtm_type = RTN_LOCAL;
2594 rtm->rtm_type = RTN_UNICAST;
2596 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2597 rtm->rtm_protocol = rt->rt6i_protocol;
2598 if (rt->rt6i_flags & RTF_DYNAMIC)
2599 rtm->rtm_protocol = RTPROT_REDIRECT;
2600 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2601 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2602 rtm->rtm_protocol = RTPROT_RA;
2604 rtm->rtm_protocol = RTPROT_KERNEL;
2607 if (rt->rt6i_flags & RTF_CACHE)
2608 rtm->rtm_flags |= RTM_F_CLONED;
2611 if (nla_put(skb, RTA_DST, 16, dst))
2612 goto nla_put_failure;
2613 rtm->rtm_dst_len = 128;
2614 } else if (rtm->rtm_dst_len)
2615 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2616 goto nla_put_failure;
2617 #ifdef CONFIG_IPV6_SUBTREES
2619 if (nla_put(skb, RTA_SRC, 16, src))
2620 goto nla_put_failure;
2621 rtm->rtm_src_len = 128;
2622 } else if (rtm->rtm_src_len &&
2623 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2624 goto nla_put_failure;
2627 #ifdef CONFIG_IPV6_MROUTE
2628 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2629 int err = ip6mr_get_route(net, skb, rtm, nowait);
2634 goto nla_put_failure;
2636 if (err == -EMSGSIZE)
2637 goto nla_put_failure;
2642 if (nla_put_u32(skb, RTA_IIF, iif))
2643 goto nla_put_failure;
2645 struct in6_addr saddr_buf;
2646 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2647 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2648 goto nla_put_failure;
2651 if (rt->rt6i_prefsrc.plen) {
2652 struct in6_addr saddr_buf;
2653 saddr_buf = rt->rt6i_prefsrc.addr;
2654 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2655 goto nla_put_failure;
2658 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2659 goto nla_put_failure;
2661 if (rt->rt6i_flags & RTF_GATEWAY) {
2662 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2663 goto nla_put_failure;
2667 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2668 goto nla_put_failure;
2669 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2670 goto nla_put_failure;
2672 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2674 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2675 goto nla_put_failure;
2677 return nlmsg_end(skb, nlh);
2680 nlmsg_cancel(skb, nlh);
2684 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2686 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2689 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2690 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2691 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2695 return rt6_fill_node(arg->net,
2696 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2697 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2698 prefix, 0, NLM_F_MULTI);
2701 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2703 struct net *net = sock_net(in_skb->sk);
2704 struct nlattr *tb[RTA_MAX+1];
2705 struct rt6_info *rt;
2706 struct sk_buff *skb;
2709 int err, iif = 0, oif = 0;
2711 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2716 memset(&fl6, 0, sizeof(fl6));
2719 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2722 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2726 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2729 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2733 iif = nla_get_u32(tb[RTA_IIF]);
2736 oif = nla_get_u32(tb[RTA_OIF]);
2739 struct net_device *dev;
2742 dev = __dev_get_by_index(net, iif);
2748 fl6.flowi6_iif = iif;
2750 if (!ipv6_addr_any(&fl6.saddr))
2751 flags |= RT6_LOOKUP_F_HAS_SADDR;
2753 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2756 fl6.flowi6_oif = oif;
2758 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2761 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2768 /* Reserve room for dummy headers, this skb can pass
2769 through good chunk of routing engine.
2771 skb_reset_mac_header(skb);
2772 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2774 skb_dst_set(skb, &rt->dst);
2776 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2777 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2778 nlh->nlmsg_seq, 0, 0, 0);
2784 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2789 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2791 struct sk_buff *skb;
2792 struct net *net = info->nl_net;
2797 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2799 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2803 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2804 event, info->portid, seq, 0, 0, 0);
2806 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2807 WARN_ON(err == -EMSGSIZE);
2811 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2812 info->nlh, gfp_any());
2816 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2819 static int ip6_route_dev_notify(struct notifier_block *this,
2820 unsigned long event, void *ptr)
2822 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2823 struct net *net = dev_net(dev);
2825 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2826 net->ipv6.ip6_null_entry->dst.dev = dev;
2827 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2828 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2829 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2830 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2831 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2832 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2843 #ifdef CONFIG_PROC_FS
2845 static const struct file_operations ipv6_route_proc_fops = {
2846 .owner = THIS_MODULE,
2847 .open = ipv6_route_open,
2849 .llseek = seq_lseek,
2850 .release = seq_release_net,
2853 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2855 struct net *net = (struct net *)seq->private;
2856 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2857 net->ipv6.rt6_stats->fib_nodes,
2858 net->ipv6.rt6_stats->fib_route_nodes,
2859 net->ipv6.rt6_stats->fib_rt_alloc,
2860 net->ipv6.rt6_stats->fib_rt_entries,
2861 net->ipv6.rt6_stats->fib_rt_cache,
2862 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2863 net->ipv6.rt6_stats->fib_discarded_routes);
2868 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2870 return single_open_net(inode, file, rt6_stats_seq_show);
2873 static const struct file_operations rt6_stats_seq_fops = {
2874 .owner = THIS_MODULE,
2875 .open = rt6_stats_seq_open,
2877 .llseek = seq_lseek,
2878 .release = single_release_net,
2880 #endif /* CONFIG_PROC_FS */
2882 #ifdef CONFIG_SYSCTL
2885 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2886 void __user *buffer, size_t *lenp, loff_t *ppos)
2893 net = (struct net *)ctl->extra1;
2894 delay = net->ipv6.sysctl.flush_delay;
2895 proc_dointvec(ctl, write, buffer, lenp, ppos);
2896 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2900 struct ctl_table ipv6_route_table_template[] = {
2902 .procname = "flush",
2903 .data = &init_net.ipv6.sysctl.flush_delay,
2904 .maxlen = sizeof(int),
2906 .proc_handler = ipv6_sysctl_rtcache_flush
2909 .procname = "gc_thresh",
2910 .data = &ip6_dst_ops_template.gc_thresh,
2911 .maxlen = sizeof(int),
2913 .proc_handler = proc_dointvec,
2916 .procname = "max_size",
2917 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2918 .maxlen = sizeof(int),
2920 .proc_handler = proc_dointvec,
2923 .procname = "gc_min_interval",
2924 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2925 .maxlen = sizeof(int),
2927 .proc_handler = proc_dointvec_jiffies,
2930 .procname = "gc_timeout",
2931 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2932 .maxlen = sizeof(int),
2934 .proc_handler = proc_dointvec_jiffies,
2937 .procname = "gc_interval",
2938 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2939 .maxlen = sizeof(int),
2941 .proc_handler = proc_dointvec_jiffies,
2944 .procname = "gc_elasticity",
2945 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2946 .maxlen = sizeof(int),
2948 .proc_handler = proc_dointvec,
2951 .procname = "mtu_expires",
2952 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2953 .maxlen = sizeof(int),
2955 .proc_handler = proc_dointvec_jiffies,
2958 .procname = "min_adv_mss",
2959 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2960 .maxlen = sizeof(int),
2962 .proc_handler = proc_dointvec,
2965 .procname = "gc_min_interval_ms",
2966 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2967 .maxlen = sizeof(int),
2969 .proc_handler = proc_dointvec_ms_jiffies,
2974 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2976 struct ctl_table *table;
2978 table = kmemdup(ipv6_route_table_template,
2979 sizeof(ipv6_route_table_template),
2983 table[0].data = &net->ipv6.sysctl.flush_delay;
2984 table[0].extra1 = net;
2985 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2986 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2987 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2988 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2989 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2990 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2991 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2992 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2993 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2995 /* Don't export sysctls to unprivileged users */
2996 if (net->user_ns != &init_user_ns)
2997 table[0].procname = NULL;
3004 static int __net_init ip6_route_net_init(struct net *net)
3008 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3009 sizeof(net->ipv6.ip6_dst_ops));
3011 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3012 goto out_ip6_dst_ops;
3014 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3015 sizeof(*net->ipv6.ip6_null_entry),
3017 if (!net->ipv6.ip6_null_entry)
3018 goto out_ip6_dst_entries;
3019 net->ipv6.ip6_null_entry->dst.path =
3020 (struct dst_entry *)net->ipv6.ip6_null_entry;
3021 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3022 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3023 ip6_template_metrics, true);
3025 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3026 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3027 sizeof(*net->ipv6.ip6_prohibit_entry),
3029 if (!net->ipv6.ip6_prohibit_entry)
3030 goto out_ip6_null_entry;
3031 net->ipv6.ip6_prohibit_entry->dst.path =
3032 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3033 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3034 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3035 ip6_template_metrics, true);
3037 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3038 sizeof(*net->ipv6.ip6_blk_hole_entry),
3040 if (!net->ipv6.ip6_blk_hole_entry)
3041 goto out_ip6_prohibit_entry;
3042 net->ipv6.ip6_blk_hole_entry->dst.path =
3043 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3044 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3045 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3046 ip6_template_metrics, true);
3049 net->ipv6.sysctl.flush_delay = 0;
3050 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3051 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3052 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3053 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3054 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3055 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3056 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3058 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3064 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3065 out_ip6_prohibit_entry:
3066 kfree(net->ipv6.ip6_prohibit_entry);
3068 kfree(net->ipv6.ip6_null_entry);
3070 out_ip6_dst_entries:
3071 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3076 static void __net_exit ip6_route_net_exit(struct net *net)
3078 kfree(net->ipv6.ip6_null_entry);
3079 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3080 kfree(net->ipv6.ip6_prohibit_entry);
3081 kfree(net->ipv6.ip6_blk_hole_entry);
3083 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3086 static int __net_init ip6_route_net_init_late(struct net *net)
3088 #ifdef CONFIG_PROC_FS
3089 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3090 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3095 static void __net_exit ip6_route_net_exit_late(struct net *net)
3097 #ifdef CONFIG_PROC_FS
3098 remove_proc_entry("ipv6_route", net->proc_net);
3099 remove_proc_entry("rt6_stats", net->proc_net);
3103 static struct pernet_operations ip6_route_net_ops = {
3104 .init = ip6_route_net_init,
3105 .exit = ip6_route_net_exit,
3108 static int __net_init ipv6_inetpeer_init(struct net *net)
3110 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3114 inet_peer_base_init(bp);
3115 net->ipv6.peers = bp;
3119 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3121 struct inet_peer_base *bp = net->ipv6.peers;
3123 net->ipv6.peers = NULL;
3124 inetpeer_invalidate_tree(bp);
3128 static struct pernet_operations ipv6_inetpeer_ops = {
3129 .init = ipv6_inetpeer_init,
3130 .exit = ipv6_inetpeer_exit,
3133 static struct pernet_operations ip6_route_net_late_ops = {
3134 .init = ip6_route_net_init_late,
3135 .exit = ip6_route_net_exit_late,
3138 static struct notifier_block ip6_route_dev_notifier = {
3139 .notifier_call = ip6_route_dev_notify,
3143 int __init ip6_route_init(void)
3148 ip6_dst_ops_template.kmem_cachep =
3149 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3150 SLAB_HWCACHE_ALIGN, NULL);
3151 if (!ip6_dst_ops_template.kmem_cachep)
3154 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3156 goto out_kmem_cache;
3158 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3160 goto out_dst_entries;
3162 ret = register_pernet_subsys(&ip6_route_net_ops);
3164 goto out_register_inetpeer;
3166 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3168 /* Registering of the loopback is done before this portion of code,
3169 * the loopback reference in rt6_info will not be taken, do it
3170 * manually for init_net */
3171 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3172 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3173 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3174 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3175 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3176 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3177 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3181 goto out_register_subsys;
3187 ret = fib6_rules_init();
3191 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3193 goto fib6_rules_init;
3196 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3197 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3198 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3199 goto out_register_late_subsys;
3201 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3203 goto out_register_late_subsys;
3208 out_register_late_subsys:
3209 unregister_pernet_subsys(&ip6_route_net_late_ops);
3211 fib6_rules_cleanup();
3216 out_register_subsys:
3217 unregister_pernet_subsys(&ip6_route_net_ops);
3218 out_register_inetpeer:
3219 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3221 dst_entries_destroy(&ip6_dst_blackhole_ops);
3223 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3227 void ip6_route_cleanup(void)
3229 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3230 unregister_pernet_subsys(&ip6_route_net_late_ops);
3231 fib6_rules_cleanup();
3234 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3235 unregister_pernet_subsys(&ip6_route_net_ops);
3236 dst_entries_destroy(&ip6_dst_blackhole_ops);
3237 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);