pktgen: add needed include file
[linux/fpc-iii.git] / net / ipv6 / route.c
blob74ab1f74abcddcb056d9dbe847a1364ab7bceb04
1 /*
2 * Linux INET6 implementation
3 * FIB front-end.
5 * Authors:
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.
14 /* Changes:
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.
23 * Ville Nuorvala
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>
48 #include <net/snmp.h>
49 #include <net/ipv6.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ndisc.h>
53 #include <net/addrconf.h>
54 #include <net/tcp.h>
55 #include <linux/rtnetlink.h>
56 #include <net/dst.h>
57 #include <net/xfrm.h>
58 #include <net/netevent.h>
59 #include <net/netlink.h>
60 #include <net/nexthop.h>
62 #include <asm/uaccess.h>
64 #ifdef CONFIG_SYSCTL
65 #include <linux/sysctl.h>
66 #endif
68 enum rt6_nud_state {
69 RT6_NUD_FAIL_HARD = -2,
70 RT6_NUD_FAIL_SOFT = -1,
71 RT6_NUD_SUCCEED = 1
74 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
75 const struct in6_addr *dest);
76 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
77 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
78 static unsigned int ip6_mtu(const struct dst_entry *dst);
79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
80 static void ip6_dst_destroy(struct dst_entry *);
81 static void ip6_dst_ifdown(struct dst_entry *,
82 struct net_device *dev, int how);
83 static int ip6_dst_gc(struct dst_ops *ops);
85 static int ip6_pkt_discard(struct sk_buff *skb);
86 static int ip6_pkt_discard_out(struct sk_buff *skb);
87 static void ip6_link_failure(struct sk_buff *skb);
88 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
89 struct sk_buff *skb, u32 mtu);
90 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
91 struct sk_buff *skb);
92 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
94 #ifdef CONFIG_IPV6_ROUTE_INFO
95 static struct rt6_info *rt6_add_route_info(struct net *net,
96 const struct in6_addr *prefix, int prefixlen,
97 const struct in6_addr *gwaddr, int ifindex,
98 unsigned int pref);
99 static struct rt6_info *rt6_get_route_info(struct net *net,
100 const struct in6_addr *prefix, int prefixlen,
101 const struct in6_addr *gwaddr, int ifindex);
102 #endif
104 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
106 struct rt6_info *rt = (struct rt6_info *) dst;
107 struct inet_peer *peer;
108 u32 *p = NULL;
110 if (!(rt->dst.flags & DST_HOST))
111 return NULL;
113 peer = rt6_get_peer_create(rt);
114 if (peer) {
115 u32 *old_p = __DST_METRICS_PTR(old);
116 unsigned long prev, new;
118 p = peer->metrics;
119 if (inet_metrics_new(peer))
120 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
122 new = (unsigned long) p;
123 prev = cmpxchg(&dst->_metrics, old, new);
125 if (prev != old) {
126 p = __DST_METRICS_PTR(prev);
127 if (prev & DST_METRICS_READ_ONLY)
128 p = NULL;
131 return p;
134 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
135 struct sk_buff *skb,
136 const void *daddr)
138 struct in6_addr *p = &rt->rt6i_gateway;
140 if (!ipv6_addr_any(p))
141 return (const void *) p;
142 else if (skb)
143 return &ipv6_hdr(skb)->daddr;
144 return daddr;
147 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
148 struct sk_buff *skb,
149 const void *daddr)
151 struct rt6_info *rt = (struct rt6_info *) dst;
152 struct neighbour *n;
154 daddr = choose_neigh_daddr(rt, skb, daddr);
155 n = __ipv6_neigh_lookup(dst->dev, daddr);
156 if (n)
157 return n;
158 return neigh_create(&nd_tbl, daddr, dst->dev);
161 static struct dst_ops ip6_dst_ops_template = {
162 .family = AF_INET6,
163 .protocol = cpu_to_be16(ETH_P_IPV6),
164 .gc = ip6_dst_gc,
165 .gc_thresh = 1024,
166 .check = ip6_dst_check,
167 .default_advmss = ip6_default_advmss,
168 .mtu = ip6_mtu,
169 .cow_metrics = ipv6_cow_metrics,
170 .destroy = ip6_dst_destroy,
171 .ifdown = ip6_dst_ifdown,
172 .negative_advice = ip6_negative_advice,
173 .link_failure = ip6_link_failure,
174 .update_pmtu = ip6_rt_update_pmtu,
175 .redirect = rt6_do_redirect,
176 .local_out = __ip6_local_out,
177 .neigh_lookup = ip6_neigh_lookup,
180 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
182 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
184 return mtu ? : dst->dev->mtu;
187 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
188 struct sk_buff *skb, u32 mtu)
192 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
193 struct sk_buff *skb)
197 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
198 unsigned long old)
200 return NULL;
203 static struct dst_ops ip6_dst_blackhole_ops = {
204 .family = AF_INET6,
205 .protocol = cpu_to_be16(ETH_P_IPV6),
206 .destroy = ip6_dst_destroy,
207 .check = ip6_dst_check,
208 .mtu = ip6_blackhole_mtu,
209 .default_advmss = ip6_default_advmss,
210 .update_pmtu = ip6_rt_blackhole_update_pmtu,
211 .redirect = ip6_rt_blackhole_redirect,
212 .cow_metrics = ip6_rt_blackhole_cow_metrics,
213 .neigh_lookup = ip6_neigh_lookup,
216 static const u32 ip6_template_metrics[RTAX_MAX] = {
217 [RTAX_HOPLIMIT - 1] = 0,
220 static const struct rt6_info ip6_null_entry_template = {
221 .dst = {
222 .__refcnt = ATOMIC_INIT(1),
223 .__use = 1,
224 .obsolete = DST_OBSOLETE_FORCE_CHK,
225 .error = -ENETUNREACH,
226 .input = ip6_pkt_discard,
227 .output = ip6_pkt_discard_out,
229 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
230 .rt6i_protocol = RTPROT_KERNEL,
231 .rt6i_metric = ~(u32) 0,
232 .rt6i_ref = ATOMIC_INIT(1),
235 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
237 static int ip6_pkt_prohibit(struct sk_buff *skb);
238 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
240 static const struct rt6_info ip6_prohibit_entry_template = {
241 .dst = {
242 .__refcnt = ATOMIC_INIT(1),
243 .__use = 1,
244 .obsolete = DST_OBSOLETE_FORCE_CHK,
245 .error = -EACCES,
246 .input = ip6_pkt_prohibit,
247 .output = ip6_pkt_prohibit_out,
249 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
250 .rt6i_protocol = RTPROT_KERNEL,
251 .rt6i_metric = ~(u32) 0,
252 .rt6i_ref = ATOMIC_INIT(1),
255 static const struct rt6_info ip6_blk_hole_entry_template = {
256 .dst = {
257 .__refcnt = ATOMIC_INIT(1),
258 .__use = 1,
259 .obsolete = DST_OBSOLETE_FORCE_CHK,
260 .error = -EINVAL,
261 .input = dst_discard,
262 .output = dst_discard,
264 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
265 .rt6i_protocol = RTPROT_KERNEL,
266 .rt6i_metric = ~(u32) 0,
267 .rt6i_ref = ATOMIC_INIT(1),
270 #endif
272 /* allocate dst with ip6_dst_ops */
273 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
274 struct net_device *dev,
275 int flags,
276 struct fib6_table *table)
278 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
279 0, DST_OBSOLETE_FORCE_CHK, flags);
281 if (rt) {
282 struct dst_entry *dst = &rt->dst;
284 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
285 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
286 rt->rt6i_genid = rt_genid(net);
287 INIT_LIST_HEAD(&rt->rt6i_siblings);
289 return rt;
292 static void ip6_dst_destroy(struct dst_entry *dst)
294 struct rt6_info *rt = (struct rt6_info *)dst;
295 struct inet6_dev *idev = rt->rt6i_idev;
296 struct dst_entry *from = dst->from;
298 if (!(rt->dst.flags & DST_HOST))
299 dst_destroy_metrics_generic(dst);
301 if (idev) {
302 rt->rt6i_idev = NULL;
303 in6_dev_put(idev);
306 dst->from = NULL;
307 dst_release(from);
309 if (rt6_has_peer(rt)) {
310 struct inet_peer *peer = rt6_peer_ptr(rt);
311 inet_putpeer(peer);
315 void rt6_bind_peer(struct rt6_info *rt, int create)
317 struct inet_peer_base *base;
318 struct inet_peer *peer;
320 base = inetpeer_base_ptr(rt->_rt6i_peer);
321 if (!base)
322 return;
324 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
325 if (peer) {
326 if (!rt6_set_peer(rt, peer))
327 inet_putpeer(peer);
331 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
332 int how)
334 struct rt6_info *rt = (struct rt6_info *)dst;
335 struct inet6_dev *idev = rt->rt6i_idev;
336 struct net_device *loopback_dev =
337 dev_net(dev)->loopback_dev;
339 if (dev != loopback_dev) {
340 if (idev && idev->dev == dev) {
341 struct inet6_dev *loopback_idev =
342 in6_dev_get(loopback_dev);
343 if (loopback_idev) {
344 rt->rt6i_idev = loopback_idev;
345 in6_dev_put(idev);
351 static bool rt6_check_expired(const struct rt6_info *rt)
353 if (rt->rt6i_flags & RTF_EXPIRES) {
354 if (time_after(jiffies, rt->dst.expires))
355 return true;
356 } else if (rt->dst.from) {
357 return rt6_check_expired((struct rt6_info *) rt->dst.from);
359 return false;
362 static bool rt6_need_strict(const struct in6_addr *daddr)
364 return ipv6_addr_type(daddr) &
365 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
368 /* Multipath route selection:
369 * Hash based function using packet header and flowlabel.
370 * Adapted from fib_info_hashfn()
372 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
373 const struct flowi6 *fl6)
375 unsigned int val = fl6->flowi6_proto;
377 val ^= ipv6_addr_hash(&fl6->daddr);
378 val ^= ipv6_addr_hash(&fl6->saddr);
380 /* Work only if this not encapsulated */
381 switch (fl6->flowi6_proto) {
382 case IPPROTO_UDP:
383 case IPPROTO_TCP:
384 case IPPROTO_SCTP:
385 val ^= (__force u16)fl6->fl6_sport;
386 val ^= (__force u16)fl6->fl6_dport;
387 break;
389 case IPPROTO_ICMPV6:
390 val ^= (__force u16)fl6->fl6_icmp_type;
391 val ^= (__force u16)fl6->fl6_icmp_code;
392 break;
394 /* RFC6438 recommands to use flowlabel */
395 val ^= (__force u32)fl6->flowlabel;
397 /* Perhaps, we need to tune, this function? */
398 val = val ^ (val >> 7) ^ (val >> 12);
399 return val % candidate_count;
402 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
403 struct flowi6 *fl6, int oif,
404 int strict)
406 struct rt6_info *sibling, *next_sibling;
407 int route_choosen;
409 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
410 /* Don't change the route, if route_choosen == 0
411 * (siblings does not include ourself)
413 if (route_choosen)
414 list_for_each_entry_safe(sibling, next_sibling,
415 &match->rt6i_siblings, rt6i_siblings) {
416 route_choosen--;
417 if (route_choosen == 0) {
418 if (rt6_score_route(sibling, oif, strict) < 0)
419 break;
420 match = sibling;
421 break;
424 return match;
428 * Route lookup. Any table->tb6_lock is implied.
431 static inline struct rt6_info *rt6_device_match(struct net *net,
432 struct rt6_info *rt,
433 const struct in6_addr *saddr,
434 int oif,
435 int flags)
437 struct rt6_info *local = NULL;
438 struct rt6_info *sprt;
440 if (!oif && ipv6_addr_any(saddr))
441 goto out;
443 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
444 struct net_device *dev = sprt->dst.dev;
446 if (oif) {
447 if (dev->ifindex == oif)
448 return sprt;
449 if (dev->flags & IFF_LOOPBACK) {
450 if (!sprt->rt6i_idev ||
451 sprt->rt6i_idev->dev->ifindex != oif) {
452 if (flags & RT6_LOOKUP_F_IFACE && oif)
453 continue;
454 if (local && (!oif ||
455 local->rt6i_idev->dev->ifindex == oif))
456 continue;
458 local = sprt;
460 } else {
461 if (ipv6_chk_addr(net, saddr, dev,
462 flags & RT6_LOOKUP_F_IFACE))
463 return sprt;
467 if (oif) {
468 if (local)
469 return local;
471 if (flags & RT6_LOOKUP_F_IFACE)
472 return net->ipv6.ip6_null_entry;
474 out:
475 return rt;
478 #ifdef CONFIG_IPV6_ROUTER_PREF
479 static void rt6_probe(struct rt6_info *rt)
481 struct neighbour *neigh;
483 * Okay, this does not seem to be appropriate
484 * for now, however, we need to check if it
485 * is really so; aka Router Reachability Probing.
487 * Router Reachability Probe MUST be rate-limited
488 * to no more than one per minute.
490 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
491 return;
492 rcu_read_lock_bh();
493 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
494 if (neigh) {
495 write_lock(&neigh->lock);
496 if (neigh->nud_state & NUD_VALID)
497 goto out;
500 if (!neigh ||
501 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
502 struct in6_addr mcaddr;
503 struct in6_addr *target;
505 if (neigh) {
506 neigh->updated = jiffies;
507 write_unlock(&neigh->lock);
510 target = (struct in6_addr *)&rt->rt6i_gateway;
511 addrconf_addr_solict_mult(target, &mcaddr);
512 ndisc_send_ns(rt->dst.dev, NULL, target, &mcaddr, NULL);
513 } else {
514 out:
515 write_unlock(&neigh->lock);
517 rcu_read_unlock_bh();
519 #else
520 static inline void rt6_probe(struct rt6_info *rt)
523 #endif
526 * Default Router Selection (RFC 2461 6.3.6)
528 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
530 struct net_device *dev = rt->dst.dev;
531 if (!oif || dev->ifindex == oif)
532 return 2;
533 if ((dev->flags & IFF_LOOPBACK) &&
534 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
535 return 1;
536 return 0;
539 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
541 struct neighbour *neigh;
542 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
544 if (rt->rt6i_flags & RTF_NONEXTHOP ||
545 !(rt->rt6i_flags & RTF_GATEWAY))
546 return RT6_NUD_SUCCEED;
548 rcu_read_lock_bh();
549 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
550 if (neigh) {
551 read_lock(&neigh->lock);
552 if (neigh->nud_state & NUD_VALID)
553 ret = RT6_NUD_SUCCEED;
554 #ifdef CONFIG_IPV6_ROUTER_PREF
555 else if (!(neigh->nud_state & NUD_FAILED))
556 ret = RT6_NUD_SUCCEED;
557 #endif
558 read_unlock(&neigh->lock);
559 } else {
560 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
561 RT6_NUD_SUCCEED : RT6_NUD_FAIL_SOFT;
563 rcu_read_unlock_bh();
565 return ret;
568 static int rt6_score_route(struct rt6_info *rt, int oif,
569 int strict)
571 int m;
573 m = rt6_check_dev(rt, oif);
574 if (!m && (strict & RT6_LOOKUP_F_IFACE))
575 return RT6_NUD_FAIL_HARD;
576 #ifdef CONFIG_IPV6_ROUTER_PREF
577 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
578 #endif
579 if (strict & RT6_LOOKUP_F_REACHABLE) {
580 int n = rt6_check_neigh(rt);
581 if (n < 0)
582 return n;
584 return m;
587 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
588 int *mpri, struct rt6_info *match,
589 bool *do_rr)
591 int m;
592 bool match_do_rr = false;
594 if (rt6_check_expired(rt))
595 goto out;
597 m = rt6_score_route(rt, oif, strict);
598 if (m == RT6_NUD_FAIL_SOFT && !IS_ENABLED(CONFIG_IPV6_ROUTER_PREF)) {
599 match_do_rr = true;
600 m = 0; /* lowest valid score */
601 } else if (m < 0) {
602 goto out;
605 if (strict & RT6_LOOKUP_F_REACHABLE)
606 rt6_probe(rt);
608 if (m > *mpri) {
609 *do_rr = match_do_rr;
610 *mpri = m;
611 match = rt;
613 out:
614 return match;
617 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
618 struct rt6_info *rr_head,
619 u32 metric, int oif, int strict,
620 bool *do_rr)
622 struct rt6_info *rt, *match;
623 int mpri = -1;
625 match = NULL;
626 for (rt = rr_head; rt && rt->rt6i_metric == metric;
627 rt = rt->dst.rt6_next)
628 match = find_match(rt, oif, strict, &mpri, match, do_rr);
629 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
630 rt = rt->dst.rt6_next)
631 match = find_match(rt, oif, strict, &mpri, match, do_rr);
633 return match;
636 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
638 struct rt6_info *match, *rt0;
639 struct net *net;
640 bool do_rr = false;
642 rt0 = fn->rr_ptr;
643 if (!rt0)
644 fn->rr_ptr = rt0 = fn->leaf;
646 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
647 &do_rr);
649 if (do_rr) {
650 struct rt6_info *next = rt0->dst.rt6_next;
652 /* no entries matched; do round-robin */
653 if (!next || next->rt6i_metric != rt0->rt6i_metric)
654 next = fn->leaf;
656 if (next != rt0)
657 fn->rr_ptr = next;
660 net = dev_net(rt0->dst.dev);
661 return match ? match : net->ipv6.ip6_null_entry;
664 #ifdef CONFIG_IPV6_ROUTE_INFO
665 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
666 const struct in6_addr *gwaddr)
668 struct net *net = dev_net(dev);
669 struct route_info *rinfo = (struct route_info *) opt;
670 struct in6_addr prefix_buf, *prefix;
671 unsigned int pref;
672 unsigned long lifetime;
673 struct rt6_info *rt;
675 if (len < sizeof(struct route_info)) {
676 return -EINVAL;
679 /* Sanity check for prefix_len and length */
680 if (rinfo->length > 3) {
681 return -EINVAL;
682 } else if (rinfo->prefix_len > 128) {
683 return -EINVAL;
684 } else if (rinfo->prefix_len > 64) {
685 if (rinfo->length < 2) {
686 return -EINVAL;
688 } else if (rinfo->prefix_len > 0) {
689 if (rinfo->length < 1) {
690 return -EINVAL;
694 pref = rinfo->route_pref;
695 if (pref == ICMPV6_ROUTER_PREF_INVALID)
696 return -EINVAL;
698 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
700 if (rinfo->length == 3)
701 prefix = (struct in6_addr *)rinfo->prefix;
702 else {
703 /* this function is safe */
704 ipv6_addr_prefix(&prefix_buf,
705 (struct in6_addr *)rinfo->prefix,
706 rinfo->prefix_len);
707 prefix = &prefix_buf;
710 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
711 dev->ifindex);
713 if (rt && !lifetime) {
714 ip6_del_rt(rt);
715 rt = NULL;
718 if (!rt && lifetime)
719 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
720 pref);
721 else if (rt)
722 rt->rt6i_flags = RTF_ROUTEINFO |
723 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
725 if (rt) {
726 if (!addrconf_finite_timeout(lifetime))
727 rt6_clean_expires(rt);
728 else
729 rt6_set_expires(rt, jiffies + HZ * lifetime);
731 ip6_rt_put(rt);
733 return 0;
735 #endif
737 #define BACKTRACK(__net, saddr) \
738 do { \
739 if (rt == __net->ipv6.ip6_null_entry) { \
740 struct fib6_node *pn; \
741 while (1) { \
742 if (fn->fn_flags & RTN_TL_ROOT) \
743 goto out; \
744 pn = fn->parent; \
745 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
746 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
747 else \
748 fn = pn; \
749 if (fn->fn_flags & RTN_RTINFO) \
750 goto restart; \
753 } while (0)
755 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
756 struct fib6_table *table,
757 struct flowi6 *fl6, int flags)
759 struct fib6_node *fn;
760 struct rt6_info *rt;
762 read_lock_bh(&table->tb6_lock);
763 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
764 restart:
765 rt = fn->leaf;
766 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
767 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
768 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
769 BACKTRACK(net, &fl6->saddr);
770 out:
771 dst_use(&rt->dst, jiffies);
772 read_unlock_bh(&table->tb6_lock);
773 return rt;
777 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
778 int flags)
780 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
782 EXPORT_SYMBOL_GPL(ip6_route_lookup);
784 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
785 const struct in6_addr *saddr, int oif, int strict)
787 struct flowi6 fl6 = {
788 .flowi6_oif = oif,
789 .daddr = *daddr,
791 struct dst_entry *dst;
792 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
794 if (saddr) {
795 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
796 flags |= RT6_LOOKUP_F_HAS_SADDR;
799 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
800 if (dst->error == 0)
801 return (struct rt6_info *) dst;
803 dst_release(dst);
805 return NULL;
808 EXPORT_SYMBOL(rt6_lookup);
810 /* ip6_ins_rt is called with FREE table->tb6_lock.
811 It takes new route entry, the addition fails by any reason the
812 route is freed. In any case, if caller does not hold it, it may
813 be destroyed.
816 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
818 int err;
819 struct fib6_table *table;
821 table = rt->rt6i_table;
822 write_lock_bh(&table->tb6_lock);
823 err = fib6_add(&table->tb6_root, rt, info);
824 write_unlock_bh(&table->tb6_lock);
826 return err;
829 int ip6_ins_rt(struct rt6_info *rt)
831 struct nl_info info = {
832 .nl_net = dev_net(rt->dst.dev),
834 return __ip6_ins_rt(rt, &info);
837 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
838 const struct in6_addr *daddr,
839 const struct in6_addr *saddr)
841 struct rt6_info *rt;
844 * Clone the route.
847 rt = ip6_rt_copy(ort, daddr);
849 if (rt) {
850 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
851 if (ort->rt6i_dst.plen != 128 &&
852 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
853 rt->rt6i_flags |= RTF_ANYCAST;
854 rt->rt6i_gateway = *daddr;
857 rt->rt6i_flags |= RTF_CACHE;
859 #ifdef CONFIG_IPV6_SUBTREES
860 if (rt->rt6i_src.plen && saddr) {
861 rt->rt6i_src.addr = *saddr;
862 rt->rt6i_src.plen = 128;
864 #endif
867 return rt;
870 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
871 const struct in6_addr *daddr)
873 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
875 if (rt)
876 rt->rt6i_flags |= RTF_CACHE;
877 return rt;
880 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
881 struct flowi6 *fl6, int flags)
883 struct fib6_node *fn;
884 struct rt6_info *rt, *nrt;
885 int strict = 0;
886 int attempts = 3;
887 int err;
888 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
890 strict |= flags & RT6_LOOKUP_F_IFACE;
892 relookup:
893 read_lock_bh(&table->tb6_lock);
895 restart_2:
896 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
898 restart:
899 rt = rt6_select(fn, oif, strict | reachable);
900 if (rt->rt6i_nsiblings)
901 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
902 BACKTRACK(net, &fl6->saddr);
903 if (rt == net->ipv6.ip6_null_entry ||
904 rt->rt6i_flags & RTF_CACHE)
905 goto out;
907 dst_hold(&rt->dst);
908 read_unlock_bh(&table->tb6_lock);
910 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
911 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
912 else if (!(rt->dst.flags & DST_HOST))
913 nrt = rt6_alloc_clone(rt, &fl6->daddr);
914 else
915 goto out2;
917 ip6_rt_put(rt);
918 rt = nrt ? : net->ipv6.ip6_null_entry;
920 dst_hold(&rt->dst);
921 if (nrt) {
922 err = ip6_ins_rt(nrt);
923 if (!err)
924 goto out2;
927 if (--attempts <= 0)
928 goto out2;
931 * Race condition! In the gap, when table->tb6_lock was
932 * released someone could insert this route. Relookup.
934 ip6_rt_put(rt);
935 goto relookup;
937 out:
938 if (reachable) {
939 reachable = 0;
940 goto restart_2;
942 dst_hold(&rt->dst);
943 read_unlock_bh(&table->tb6_lock);
944 out2:
945 rt->dst.lastuse = jiffies;
946 rt->dst.__use++;
948 return rt;
951 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
952 struct flowi6 *fl6, int flags)
954 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
957 static struct dst_entry *ip6_route_input_lookup(struct net *net,
958 struct net_device *dev,
959 struct flowi6 *fl6, int flags)
961 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
962 flags |= RT6_LOOKUP_F_IFACE;
964 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
967 void ip6_route_input(struct sk_buff *skb)
969 const struct ipv6hdr *iph = ipv6_hdr(skb);
970 struct net *net = dev_net(skb->dev);
971 int flags = RT6_LOOKUP_F_HAS_SADDR;
972 struct flowi6 fl6 = {
973 .flowi6_iif = skb->dev->ifindex,
974 .daddr = iph->daddr,
975 .saddr = iph->saddr,
976 .flowlabel = ip6_flowinfo(iph),
977 .flowi6_mark = skb->mark,
978 .flowi6_proto = iph->nexthdr,
981 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
984 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
985 struct flowi6 *fl6, int flags)
987 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
990 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
991 struct flowi6 *fl6)
993 int flags = 0;
995 fl6->flowi6_iif = LOOPBACK_IFINDEX;
997 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
998 flags |= RT6_LOOKUP_F_IFACE;
1000 if (!ipv6_addr_any(&fl6->saddr))
1001 flags |= RT6_LOOKUP_F_HAS_SADDR;
1002 else if (sk)
1003 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1005 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1008 EXPORT_SYMBOL(ip6_route_output);
1010 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1012 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1013 struct dst_entry *new = NULL;
1015 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1016 if (rt) {
1017 new = &rt->dst;
1019 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1020 rt6_init_peer(rt, net->ipv6.peers);
1022 new->__use = 1;
1023 new->input = dst_discard;
1024 new->output = dst_discard;
1026 if (dst_metrics_read_only(&ort->dst))
1027 new->_metrics = ort->dst._metrics;
1028 else
1029 dst_copy_metrics(new, &ort->dst);
1030 rt->rt6i_idev = ort->rt6i_idev;
1031 if (rt->rt6i_idev)
1032 in6_dev_hold(rt->rt6i_idev);
1034 rt->rt6i_gateway = ort->rt6i_gateway;
1035 rt->rt6i_flags = ort->rt6i_flags;
1036 rt->rt6i_metric = 0;
1038 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1039 #ifdef CONFIG_IPV6_SUBTREES
1040 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1041 #endif
1043 dst_free(new);
1046 dst_release(dst_orig);
1047 return new ? new : ERR_PTR(-ENOMEM);
1051 * Destination cache support functions
1054 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1056 struct rt6_info *rt;
1058 rt = (struct rt6_info *) dst;
1060 /* All IPV6 dsts are created with ->obsolete set to the value
1061 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1062 * into this function always.
1064 if (rt->rt6i_genid != rt_genid(dev_net(rt->dst.dev)))
1065 return NULL;
1067 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
1068 return dst;
1070 return NULL;
1073 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1075 struct rt6_info *rt = (struct rt6_info *) dst;
1077 if (rt) {
1078 if (rt->rt6i_flags & RTF_CACHE) {
1079 if (rt6_check_expired(rt)) {
1080 ip6_del_rt(rt);
1081 dst = NULL;
1083 } else {
1084 dst_release(dst);
1085 dst = NULL;
1088 return dst;
1091 static void ip6_link_failure(struct sk_buff *skb)
1093 struct rt6_info *rt;
1095 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1097 rt = (struct rt6_info *) skb_dst(skb);
1098 if (rt) {
1099 if (rt->rt6i_flags & RTF_CACHE) {
1100 dst_hold(&rt->dst);
1101 if (ip6_del_rt(rt))
1102 dst_free(&rt->dst);
1103 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1104 rt->rt6i_node->fn_sernum = -1;
1109 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1110 struct sk_buff *skb, u32 mtu)
1112 struct rt6_info *rt6 = (struct rt6_info*)dst;
1114 dst_confirm(dst);
1115 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1116 struct net *net = dev_net(dst->dev);
1118 rt6->rt6i_flags |= RTF_MODIFIED;
1119 if (mtu < IPV6_MIN_MTU) {
1120 u32 features = dst_metric(dst, RTAX_FEATURES);
1121 mtu = IPV6_MIN_MTU;
1122 features |= RTAX_FEATURE_ALLFRAG;
1123 dst_metric_set(dst, RTAX_FEATURES, features);
1125 dst_metric_set(dst, RTAX_MTU, mtu);
1126 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1130 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1131 int oif, u32 mark)
1133 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1134 struct dst_entry *dst;
1135 struct flowi6 fl6;
1137 memset(&fl6, 0, sizeof(fl6));
1138 fl6.flowi6_oif = oif;
1139 fl6.flowi6_mark = mark;
1140 fl6.flowi6_flags = 0;
1141 fl6.daddr = iph->daddr;
1142 fl6.saddr = iph->saddr;
1143 fl6.flowlabel = ip6_flowinfo(iph);
1145 dst = ip6_route_output(net, NULL, &fl6);
1146 if (!dst->error)
1147 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1148 dst_release(dst);
1150 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1152 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1154 ip6_update_pmtu(skb, sock_net(sk), mtu,
1155 sk->sk_bound_dev_if, sk->sk_mark);
1157 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1159 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1161 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1162 struct dst_entry *dst;
1163 struct flowi6 fl6;
1165 memset(&fl6, 0, sizeof(fl6));
1166 fl6.flowi6_oif = oif;
1167 fl6.flowi6_mark = mark;
1168 fl6.flowi6_flags = 0;
1169 fl6.daddr = iph->daddr;
1170 fl6.saddr = iph->saddr;
1171 fl6.flowlabel = ip6_flowinfo(iph);
1173 dst = ip6_route_output(net, NULL, &fl6);
1174 if (!dst->error)
1175 rt6_do_redirect(dst, NULL, skb);
1176 dst_release(dst);
1178 EXPORT_SYMBOL_GPL(ip6_redirect);
1180 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1182 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1184 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1186 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1188 struct net_device *dev = dst->dev;
1189 unsigned int mtu = dst_mtu(dst);
1190 struct net *net = dev_net(dev);
1192 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1194 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1195 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1198 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1199 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1200 * IPV6_MAXPLEN is also valid and means: "any MSS,
1201 * rely only on pmtu discovery"
1203 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1204 mtu = IPV6_MAXPLEN;
1205 return mtu;
1208 static unsigned int ip6_mtu(const struct dst_entry *dst)
1210 struct inet6_dev *idev;
1211 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1213 if (mtu)
1214 return mtu;
1216 mtu = IPV6_MIN_MTU;
1218 rcu_read_lock();
1219 idev = __in6_dev_get(dst->dev);
1220 if (idev)
1221 mtu = idev->cnf.mtu6;
1222 rcu_read_unlock();
1224 return mtu;
1227 static struct dst_entry *icmp6_dst_gc_list;
1228 static DEFINE_SPINLOCK(icmp6_dst_lock);
1230 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1231 struct flowi6 *fl6)
1233 struct dst_entry *dst;
1234 struct rt6_info *rt;
1235 struct inet6_dev *idev = in6_dev_get(dev);
1236 struct net *net = dev_net(dev);
1238 if (unlikely(!idev))
1239 return ERR_PTR(-ENODEV);
1241 rt = ip6_dst_alloc(net, dev, 0, NULL);
1242 if (unlikely(!rt)) {
1243 in6_dev_put(idev);
1244 dst = ERR_PTR(-ENOMEM);
1245 goto out;
1248 rt->dst.flags |= DST_HOST;
1249 rt->dst.output = ip6_output;
1250 atomic_set(&rt->dst.__refcnt, 1);
1251 rt->rt6i_dst.addr = fl6->daddr;
1252 rt->rt6i_dst.plen = 128;
1253 rt->rt6i_idev = idev;
1254 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1256 spin_lock_bh(&icmp6_dst_lock);
1257 rt->dst.next = icmp6_dst_gc_list;
1258 icmp6_dst_gc_list = &rt->dst;
1259 spin_unlock_bh(&icmp6_dst_lock);
1261 fib6_force_start_gc(net);
1263 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1265 out:
1266 return dst;
1269 int icmp6_dst_gc(void)
1271 struct dst_entry *dst, **pprev;
1272 int more = 0;
1274 spin_lock_bh(&icmp6_dst_lock);
1275 pprev = &icmp6_dst_gc_list;
1277 while ((dst = *pprev) != NULL) {
1278 if (!atomic_read(&dst->__refcnt)) {
1279 *pprev = dst->next;
1280 dst_free(dst);
1281 } else {
1282 pprev = &dst->next;
1283 ++more;
1287 spin_unlock_bh(&icmp6_dst_lock);
1289 return more;
1292 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1293 void *arg)
1295 struct dst_entry *dst, **pprev;
1297 spin_lock_bh(&icmp6_dst_lock);
1298 pprev = &icmp6_dst_gc_list;
1299 while ((dst = *pprev) != NULL) {
1300 struct rt6_info *rt = (struct rt6_info *) dst;
1301 if (func(rt, arg)) {
1302 *pprev = dst->next;
1303 dst_free(dst);
1304 } else {
1305 pprev = &dst->next;
1308 spin_unlock_bh(&icmp6_dst_lock);
1311 static int ip6_dst_gc(struct dst_ops *ops)
1313 unsigned long now = jiffies;
1314 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1315 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1316 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1317 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1318 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1319 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1320 int entries;
1322 entries = dst_entries_get_fast(ops);
1323 if (time_after(rt_last_gc + rt_min_interval, now) &&
1324 entries <= rt_max_size)
1325 goto out;
1327 net->ipv6.ip6_rt_gc_expire++;
1328 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1329 net->ipv6.ip6_rt_last_gc = now;
1330 entries = dst_entries_get_slow(ops);
1331 if (entries < ops->gc_thresh)
1332 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1333 out:
1334 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1335 return entries > rt_max_size;
1338 int ip6_dst_hoplimit(struct dst_entry *dst)
1340 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1341 if (hoplimit == 0) {
1342 struct net_device *dev = dst->dev;
1343 struct inet6_dev *idev;
1345 rcu_read_lock();
1346 idev = __in6_dev_get(dev);
1347 if (idev)
1348 hoplimit = idev->cnf.hop_limit;
1349 else
1350 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1351 rcu_read_unlock();
1353 return hoplimit;
1355 EXPORT_SYMBOL(ip6_dst_hoplimit);
1361 int ip6_route_add(struct fib6_config *cfg)
1363 int err;
1364 struct net *net = cfg->fc_nlinfo.nl_net;
1365 struct rt6_info *rt = NULL;
1366 struct net_device *dev = NULL;
1367 struct inet6_dev *idev = NULL;
1368 struct fib6_table *table;
1369 int addr_type;
1371 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1372 return -EINVAL;
1373 #ifndef CONFIG_IPV6_SUBTREES
1374 if (cfg->fc_src_len)
1375 return -EINVAL;
1376 #endif
1377 if (cfg->fc_ifindex) {
1378 err = -ENODEV;
1379 dev = dev_get_by_index(net, cfg->fc_ifindex);
1380 if (!dev)
1381 goto out;
1382 idev = in6_dev_get(dev);
1383 if (!idev)
1384 goto out;
1387 if (cfg->fc_metric == 0)
1388 cfg->fc_metric = IP6_RT_PRIO_USER;
1390 err = -ENOBUFS;
1391 if (cfg->fc_nlinfo.nlh &&
1392 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1393 table = fib6_get_table(net, cfg->fc_table);
1394 if (!table) {
1395 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1396 table = fib6_new_table(net, cfg->fc_table);
1398 } else {
1399 table = fib6_new_table(net, cfg->fc_table);
1402 if (!table)
1403 goto out;
1405 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1407 if (!rt) {
1408 err = -ENOMEM;
1409 goto out;
1412 if (cfg->fc_flags & RTF_EXPIRES)
1413 rt6_set_expires(rt, jiffies +
1414 clock_t_to_jiffies(cfg->fc_expires));
1415 else
1416 rt6_clean_expires(rt);
1418 if (cfg->fc_protocol == RTPROT_UNSPEC)
1419 cfg->fc_protocol = RTPROT_BOOT;
1420 rt->rt6i_protocol = cfg->fc_protocol;
1422 addr_type = ipv6_addr_type(&cfg->fc_dst);
1424 if (addr_type & IPV6_ADDR_MULTICAST)
1425 rt->dst.input = ip6_mc_input;
1426 else if (cfg->fc_flags & RTF_LOCAL)
1427 rt->dst.input = ip6_input;
1428 else
1429 rt->dst.input = ip6_forward;
1431 rt->dst.output = ip6_output;
1433 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1434 rt->rt6i_dst.plen = cfg->fc_dst_len;
1435 if (rt->rt6i_dst.plen == 128)
1436 rt->dst.flags |= DST_HOST;
1438 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1439 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1440 if (!metrics) {
1441 err = -ENOMEM;
1442 goto out;
1444 dst_init_metrics(&rt->dst, metrics, 0);
1446 #ifdef CONFIG_IPV6_SUBTREES
1447 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1448 rt->rt6i_src.plen = cfg->fc_src_len;
1449 #endif
1451 rt->rt6i_metric = cfg->fc_metric;
1453 /* We cannot add true routes via loopback here,
1454 they would result in kernel looping; promote them to reject routes
1456 if ((cfg->fc_flags & RTF_REJECT) ||
1457 (dev && (dev->flags & IFF_LOOPBACK) &&
1458 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1459 !(cfg->fc_flags & RTF_LOCAL))) {
1460 /* hold loopback dev/idev if we haven't done so. */
1461 if (dev != net->loopback_dev) {
1462 if (dev) {
1463 dev_put(dev);
1464 in6_dev_put(idev);
1466 dev = net->loopback_dev;
1467 dev_hold(dev);
1468 idev = in6_dev_get(dev);
1469 if (!idev) {
1470 err = -ENODEV;
1471 goto out;
1474 rt->dst.output = ip6_pkt_discard_out;
1475 rt->dst.input = ip6_pkt_discard;
1476 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1477 switch (cfg->fc_type) {
1478 case RTN_BLACKHOLE:
1479 rt->dst.error = -EINVAL;
1480 break;
1481 case RTN_PROHIBIT:
1482 rt->dst.error = -EACCES;
1483 break;
1484 case RTN_THROW:
1485 rt->dst.error = -EAGAIN;
1486 break;
1487 default:
1488 rt->dst.error = -ENETUNREACH;
1489 break;
1491 goto install_route;
1494 if (cfg->fc_flags & RTF_GATEWAY) {
1495 const struct in6_addr *gw_addr;
1496 int gwa_type;
1498 gw_addr = &cfg->fc_gateway;
1499 rt->rt6i_gateway = *gw_addr;
1500 gwa_type = ipv6_addr_type(gw_addr);
1502 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1503 struct rt6_info *grt;
1505 /* IPv6 strictly inhibits using not link-local
1506 addresses as nexthop address.
1507 Otherwise, router will not able to send redirects.
1508 It is very good, but in some (rare!) circumstances
1509 (SIT, PtP, NBMA NOARP links) it is handy to allow
1510 some exceptions. --ANK
1512 err = -EINVAL;
1513 if (!(gwa_type & IPV6_ADDR_UNICAST))
1514 goto out;
1516 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1518 err = -EHOSTUNREACH;
1519 if (!grt)
1520 goto out;
1521 if (dev) {
1522 if (dev != grt->dst.dev) {
1523 ip6_rt_put(grt);
1524 goto out;
1526 } else {
1527 dev = grt->dst.dev;
1528 idev = grt->rt6i_idev;
1529 dev_hold(dev);
1530 in6_dev_hold(grt->rt6i_idev);
1532 if (!(grt->rt6i_flags & RTF_GATEWAY))
1533 err = 0;
1534 ip6_rt_put(grt);
1536 if (err)
1537 goto out;
1539 err = -EINVAL;
1540 if (!dev || (dev->flags & IFF_LOOPBACK))
1541 goto out;
1544 err = -ENODEV;
1545 if (!dev)
1546 goto out;
1548 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1549 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1550 err = -EINVAL;
1551 goto out;
1553 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1554 rt->rt6i_prefsrc.plen = 128;
1555 } else
1556 rt->rt6i_prefsrc.plen = 0;
1558 rt->rt6i_flags = cfg->fc_flags;
1560 install_route:
1561 if (cfg->fc_mx) {
1562 struct nlattr *nla;
1563 int remaining;
1565 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1566 int type = nla_type(nla);
1568 if (type) {
1569 if (type > RTAX_MAX) {
1570 err = -EINVAL;
1571 goto out;
1574 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1579 rt->dst.dev = dev;
1580 rt->rt6i_idev = idev;
1581 rt->rt6i_table = table;
1583 cfg->fc_nlinfo.nl_net = dev_net(dev);
1585 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1587 out:
1588 if (dev)
1589 dev_put(dev);
1590 if (idev)
1591 in6_dev_put(idev);
1592 if (rt)
1593 dst_free(&rt->dst);
1594 return err;
1597 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1599 int err;
1600 struct fib6_table *table;
1601 struct net *net = dev_net(rt->dst.dev);
1603 if (rt == net->ipv6.ip6_null_entry) {
1604 err = -ENOENT;
1605 goto out;
1608 table = rt->rt6i_table;
1609 write_lock_bh(&table->tb6_lock);
1610 err = fib6_del(rt, info);
1611 write_unlock_bh(&table->tb6_lock);
1613 out:
1614 ip6_rt_put(rt);
1615 return err;
1618 int ip6_del_rt(struct rt6_info *rt)
1620 struct nl_info info = {
1621 .nl_net = dev_net(rt->dst.dev),
1623 return __ip6_del_rt(rt, &info);
1626 static int ip6_route_del(struct fib6_config *cfg)
1628 struct fib6_table *table;
1629 struct fib6_node *fn;
1630 struct rt6_info *rt;
1631 int err = -ESRCH;
1633 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1634 if (!table)
1635 return err;
1637 read_lock_bh(&table->tb6_lock);
1639 fn = fib6_locate(&table->tb6_root,
1640 &cfg->fc_dst, cfg->fc_dst_len,
1641 &cfg->fc_src, cfg->fc_src_len);
1643 if (fn) {
1644 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1645 if (cfg->fc_ifindex &&
1646 (!rt->dst.dev ||
1647 rt->dst.dev->ifindex != cfg->fc_ifindex))
1648 continue;
1649 if (cfg->fc_flags & RTF_GATEWAY &&
1650 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1651 continue;
1652 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1653 continue;
1654 dst_hold(&rt->dst);
1655 read_unlock_bh(&table->tb6_lock);
1657 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1660 read_unlock_bh(&table->tb6_lock);
1662 return err;
1665 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1667 struct net *net = dev_net(skb->dev);
1668 struct netevent_redirect netevent;
1669 struct rt6_info *rt, *nrt = NULL;
1670 struct ndisc_options ndopts;
1671 struct inet6_dev *in6_dev;
1672 struct neighbour *neigh;
1673 struct rd_msg *msg;
1674 int optlen, on_link;
1675 u8 *lladdr;
1677 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1678 optlen -= sizeof(*msg);
1680 if (optlen < 0) {
1681 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1682 return;
1685 msg = (struct rd_msg *)icmp6_hdr(skb);
1687 if (ipv6_addr_is_multicast(&msg->dest)) {
1688 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1689 return;
1692 on_link = 0;
1693 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1694 on_link = 1;
1695 } else if (ipv6_addr_type(&msg->target) !=
1696 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1697 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1698 return;
1701 in6_dev = __in6_dev_get(skb->dev);
1702 if (!in6_dev)
1703 return;
1704 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1705 return;
1707 /* RFC2461 8.1:
1708 * The IP source address of the Redirect MUST be the same as the current
1709 * first-hop router for the specified ICMP Destination Address.
1712 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1713 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1714 return;
1717 lladdr = NULL;
1718 if (ndopts.nd_opts_tgt_lladdr) {
1719 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1720 skb->dev);
1721 if (!lladdr) {
1722 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1723 return;
1727 rt = (struct rt6_info *) dst;
1728 if (rt == net->ipv6.ip6_null_entry) {
1729 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1730 return;
1733 /* Redirect received -> path was valid.
1734 * Look, redirects are sent only in response to data packets,
1735 * so that this nexthop apparently is reachable. --ANK
1737 dst_confirm(&rt->dst);
1739 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1740 if (!neigh)
1741 return;
1744 * We have finally decided to accept it.
1747 neigh_update(neigh, lladdr, NUD_STALE,
1748 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1749 NEIGH_UPDATE_F_OVERRIDE|
1750 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1751 NEIGH_UPDATE_F_ISROUTER))
1754 nrt = ip6_rt_copy(rt, &msg->dest);
1755 if (!nrt)
1756 goto out;
1758 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1759 if (on_link)
1760 nrt->rt6i_flags &= ~RTF_GATEWAY;
1762 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1764 if (ip6_ins_rt(nrt))
1765 goto out;
1767 netevent.old = &rt->dst;
1768 netevent.new = &nrt->dst;
1769 netevent.daddr = &msg->dest;
1770 netevent.neigh = neigh;
1771 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1773 if (rt->rt6i_flags & RTF_CACHE) {
1774 rt = (struct rt6_info *) dst_clone(&rt->dst);
1775 ip6_del_rt(rt);
1778 out:
1779 neigh_release(neigh);
1783 * Misc support functions
1786 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1787 const struct in6_addr *dest)
1789 struct net *net = dev_net(ort->dst.dev);
1790 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1791 ort->rt6i_table);
1793 if (rt) {
1794 rt->dst.input = ort->dst.input;
1795 rt->dst.output = ort->dst.output;
1796 rt->dst.flags |= DST_HOST;
1798 rt->rt6i_dst.addr = *dest;
1799 rt->rt6i_dst.plen = 128;
1800 dst_copy_metrics(&rt->dst, &ort->dst);
1801 rt->dst.error = ort->dst.error;
1802 rt->rt6i_idev = ort->rt6i_idev;
1803 if (rt->rt6i_idev)
1804 in6_dev_hold(rt->rt6i_idev);
1805 rt->dst.lastuse = jiffies;
1807 rt->rt6i_gateway = ort->rt6i_gateway;
1808 rt->rt6i_flags = ort->rt6i_flags;
1809 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1810 (RTF_DEFAULT | RTF_ADDRCONF))
1811 rt6_set_from(rt, ort);
1812 rt->rt6i_metric = 0;
1814 #ifdef CONFIG_IPV6_SUBTREES
1815 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1816 #endif
1817 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1818 rt->rt6i_table = ort->rt6i_table;
1820 return rt;
1823 #ifdef CONFIG_IPV6_ROUTE_INFO
1824 static struct rt6_info *rt6_get_route_info(struct net *net,
1825 const struct in6_addr *prefix, int prefixlen,
1826 const struct in6_addr *gwaddr, int ifindex)
1828 struct fib6_node *fn;
1829 struct rt6_info *rt = NULL;
1830 struct fib6_table *table;
1832 table = fib6_get_table(net, RT6_TABLE_INFO);
1833 if (!table)
1834 return NULL;
1836 read_lock_bh(&table->tb6_lock);
1837 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1838 if (!fn)
1839 goto out;
1841 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1842 if (rt->dst.dev->ifindex != ifindex)
1843 continue;
1844 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1845 continue;
1846 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1847 continue;
1848 dst_hold(&rt->dst);
1849 break;
1851 out:
1852 read_unlock_bh(&table->tb6_lock);
1853 return rt;
1856 static struct rt6_info *rt6_add_route_info(struct net *net,
1857 const struct in6_addr *prefix, int prefixlen,
1858 const struct in6_addr *gwaddr, int ifindex,
1859 unsigned int pref)
1861 struct fib6_config cfg = {
1862 .fc_table = RT6_TABLE_INFO,
1863 .fc_metric = IP6_RT_PRIO_USER,
1864 .fc_ifindex = ifindex,
1865 .fc_dst_len = prefixlen,
1866 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1867 RTF_UP | RTF_PREF(pref),
1868 .fc_nlinfo.portid = 0,
1869 .fc_nlinfo.nlh = NULL,
1870 .fc_nlinfo.nl_net = net,
1873 cfg.fc_dst = *prefix;
1874 cfg.fc_gateway = *gwaddr;
1876 /* We should treat it as a default route if prefix length is 0. */
1877 if (!prefixlen)
1878 cfg.fc_flags |= RTF_DEFAULT;
1880 ip6_route_add(&cfg);
1882 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1884 #endif
1886 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1888 struct rt6_info *rt;
1889 struct fib6_table *table;
1891 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1892 if (!table)
1893 return NULL;
1895 read_lock_bh(&table->tb6_lock);
1896 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1897 if (dev == rt->dst.dev &&
1898 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1899 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1900 break;
1902 if (rt)
1903 dst_hold(&rt->dst);
1904 read_unlock_bh(&table->tb6_lock);
1905 return rt;
1908 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1909 struct net_device *dev,
1910 unsigned int pref)
1912 struct fib6_config cfg = {
1913 .fc_table = RT6_TABLE_DFLT,
1914 .fc_metric = IP6_RT_PRIO_USER,
1915 .fc_ifindex = dev->ifindex,
1916 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1917 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1918 .fc_nlinfo.portid = 0,
1919 .fc_nlinfo.nlh = NULL,
1920 .fc_nlinfo.nl_net = dev_net(dev),
1923 cfg.fc_gateway = *gwaddr;
1925 ip6_route_add(&cfg);
1927 return rt6_get_dflt_router(gwaddr, dev);
1930 void rt6_purge_dflt_routers(struct net *net)
1932 struct rt6_info *rt;
1933 struct fib6_table *table;
1935 /* NOTE: Keep consistent with rt6_get_dflt_router */
1936 table = fib6_get_table(net, RT6_TABLE_DFLT);
1937 if (!table)
1938 return;
1940 restart:
1941 read_lock_bh(&table->tb6_lock);
1942 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1943 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
1944 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
1945 dst_hold(&rt->dst);
1946 read_unlock_bh(&table->tb6_lock);
1947 ip6_del_rt(rt);
1948 goto restart;
1951 read_unlock_bh(&table->tb6_lock);
1954 static void rtmsg_to_fib6_config(struct net *net,
1955 struct in6_rtmsg *rtmsg,
1956 struct fib6_config *cfg)
1958 memset(cfg, 0, sizeof(*cfg));
1960 cfg->fc_table = RT6_TABLE_MAIN;
1961 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1962 cfg->fc_metric = rtmsg->rtmsg_metric;
1963 cfg->fc_expires = rtmsg->rtmsg_info;
1964 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1965 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1966 cfg->fc_flags = rtmsg->rtmsg_flags;
1968 cfg->fc_nlinfo.nl_net = net;
1970 cfg->fc_dst = rtmsg->rtmsg_dst;
1971 cfg->fc_src = rtmsg->rtmsg_src;
1972 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1975 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1977 struct fib6_config cfg;
1978 struct in6_rtmsg rtmsg;
1979 int err;
1981 switch(cmd) {
1982 case SIOCADDRT: /* Add a route */
1983 case SIOCDELRT: /* Delete a route */
1984 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1985 return -EPERM;
1986 err = copy_from_user(&rtmsg, arg,
1987 sizeof(struct in6_rtmsg));
1988 if (err)
1989 return -EFAULT;
1991 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1993 rtnl_lock();
1994 switch (cmd) {
1995 case SIOCADDRT:
1996 err = ip6_route_add(&cfg);
1997 break;
1998 case SIOCDELRT:
1999 err = ip6_route_del(&cfg);
2000 break;
2001 default:
2002 err = -EINVAL;
2004 rtnl_unlock();
2006 return err;
2009 return -EINVAL;
2013 * Drop the packet on the floor
2016 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2018 int type;
2019 struct dst_entry *dst = skb_dst(skb);
2020 switch (ipstats_mib_noroutes) {
2021 case IPSTATS_MIB_INNOROUTES:
2022 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2023 if (type == IPV6_ADDR_ANY) {
2024 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2025 IPSTATS_MIB_INADDRERRORS);
2026 break;
2028 /* FALLTHROUGH */
2029 case IPSTATS_MIB_OUTNOROUTES:
2030 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2031 ipstats_mib_noroutes);
2032 break;
2034 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2035 kfree_skb(skb);
2036 return 0;
2039 static int ip6_pkt_discard(struct sk_buff *skb)
2041 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2044 static int ip6_pkt_discard_out(struct sk_buff *skb)
2046 skb->dev = skb_dst(skb)->dev;
2047 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2050 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2052 static int ip6_pkt_prohibit(struct sk_buff *skb)
2054 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2057 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2059 skb->dev = skb_dst(skb)->dev;
2060 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2063 #endif
2066 * Allocate a dst for local (unicast / anycast) address.
2069 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2070 const struct in6_addr *addr,
2071 bool anycast)
2073 struct net *net = dev_net(idev->dev);
2074 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2076 if (!rt) {
2077 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2078 return ERR_PTR(-ENOMEM);
2081 in6_dev_hold(idev);
2083 rt->dst.flags |= DST_HOST;
2084 rt->dst.input = ip6_input;
2085 rt->dst.output = ip6_output;
2086 rt->rt6i_idev = idev;
2088 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2089 if (anycast)
2090 rt->rt6i_flags |= RTF_ANYCAST;
2091 else
2092 rt->rt6i_flags |= RTF_LOCAL;
2094 rt->rt6i_dst.addr = *addr;
2095 rt->rt6i_dst.plen = 128;
2096 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2098 atomic_set(&rt->dst.__refcnt, 1);
2100 return rt;
2103 int ip6_route_get_saddr(struct net *net,
2104 struct rt6_info *rt,
2105 const struct in6_addr *daddr,
2106 unsigned int prefs,
2107 struct in6_addr *saddr)
2109 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2110 int err = 0;
2111 if (rt->rt6i_prefsrc.plen)
2112 *saddr = rt->rt6i_prefsrc.addr;
2113 else
2114 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2115 daddr, prefs, saddr);
2116 return err;
2119 /* remove deleted ip from prefsrc entries */
2120 struct arg_dev_net_ip {
2121 struct net_device *dev;
2122 struct net *net;
2123 struct in6_addr *addr;
2126 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2128 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2129 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2130 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2132 if (((void *)rt->dst.dev == dev || !dev) &&
2133 rt != net->ipv6.ip6_null_entry &&
2134 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2135 /* remove prefsrc entry */
2136 rt->rt6i_prefsrc.plen = 0;
2138 return 0;
2141 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2143 struct net *net = dev_net(ifp->idev->dev);
2144 struct arg_dev_net_ip adni = {
2145 .dev = ifp->idev->dev,
2146 .net = net,
2147 .addr = &ifp->addr,
2149 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2152 struct arg_dev_net {
2153 struct net_device *dev;
2154 struct net *net;
2157 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2159 const struct arg_dev_net *adn = arg;
2160 const struct net_device *dev = adn->dev;
2162 if ((rt->dst.dev == dev || !dev) &&
2163 rt != adn->net->ipv6.ip6_null_entry)
2164 return -1;
2166 return 0;
2169 void rt6_ifdown(struct net *net, struct net_device *dev)
2171 struct arg_dev_net adn = {
2172 .dev = dev,
2173 .net = net,
2176 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2177 icmp6_clean_all(fib6_ifdown, &adn);
2180 struct rt6_mtu_change_arg {
2181 struct net_device *dev;
2182 unsigned int mtu;
2185 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2187 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2188 struct inet6_dev *idev;
2190 /* In IPv6 pmtu discovery is not optional,
2191 so that RTAX_MTU lock cannot disable it.
2192 We still use this lock to block changes
2193 caused by addrconf/ndisc.
2196 idev = __in6_dev_get(arg->dev);
2197 if (!idev)
2198 return 0;
2200 /* For administrative MTU increase, there is no way to discover
2201 IPv6 PMTU increase, so PMTU increase should be updated here.
2202 Since RFC 1981 doesn't include administrative MTU increase
2203 update PMTU increase is a MUST. (i.e. jumbo frame)
2206 If new MTU is less than route PMTU, this new MTU will be the
2207 lowest MTU in the path, update the route PMTU to reflect PMTU
2208 decreases; if new MTU is greater than route PMTU, and the
2209 old MTU is the lowest MTU in the path, update the route PMTU
2210 to reflect the increase. In this case if the other nodes' MTU
2211 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2212 PMTU discouvery.
2214 if (rt->dst.dev == arg->dev &&
2215 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2216 (dst_mtu(&rt->dst) >= arg->mtu ||
2217 (dst_mtu(&rt->dst) < arg->mtu &&
2218 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2219 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2221 return 0;
2224 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2226 struct rt6_mtu_change_arg arg = {
2227 .dev = dev,
2228 .mtu = mtu,
2231 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2234 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2235 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2236 [RTA_OIF] = { .type = NLA_U32 },
2237 [RTA_IIF] = { .type = NLA_U32 },
2238 [RTA_PRIORITY] = { .type = NLA_U32 },
2239 [RTA_METRICS] = { .type = NLA_NESTED },
2240 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2243 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2244 struct fib6_config *cfg)
2246 struct rtmsg *rtm;
2247 struct nlattr *tb[RTA_MAX+1];
2248 int err;
2250 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2251 if (err < 0)
2252 goto errout;
2254 err = -EINVAL;
2255 rtm = nlmsg_data(nlh);
2256 memset(cfg, 0, sizeof(*cfg));
2258 cfg->fc_table = rtm->rtm_table;
2259 cfg->fc_dst_len = rtm->rtm_dst_len;
2260 cfg->fc_src_len = rtm->rtm_src_len;
2261 cfg->fc_flags = RTF_UP;
2262 cfg->fc_protocol = rtm->rtm_protocol;
2263 cfg->fc_type = rtm->rtm_type;
2265 if (rtm->rtm_type == RTN_UNREACHABLE ||
2266 rtm->rtm_type == RTN_BLACKHOLE ||
2267 rtm->rtm_type == RTN_PROHIBIT ||
2268 rtm->rtm_type == RTN_THROW)
2269 cfg->fc_flags |= RTF_REJECT;
2271 if (rtm->rtm_type == RTN_LOCAL)
2272 cfg->fc_flags |= RTF_LOCAL;
2274 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2275 cfg->fc_nlinfo.nlh = nlh;
2276 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2278 if (tb[RTA_GATEWAY]) {
2279 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2280 cfg->fc_flags |= RTF_GATEWAY;
2283 if (tb[RTA_DST]) {
2284 int plen = (rtm->rtm_dst_len + 7) >> 3;
2286 if (nla_len(tb[RTA_DST]) < plen)
2287 goto errout;
2289 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2292 if (tb[RTA_SRC]) {
2293 int plen = (rtm->rtm_src_len + 7) >> 3;
2295 if (nla_len(tb[RTA_SRC]) < plen)
2296 goto errout;
2298 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2301 if (tb[RTA_PREFSRC])
2302 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2304 if (tb[RTA_OIF])
2305 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2307 if (tb[RTA_PRIORITY])
2308 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2310 if (tb[RTA_METRICS]) {
2311 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2312 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2315 if (tb[RTA_TABLE])
2316 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2318 if (tb[RTA_MULTIPATH]) {
2319 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2320 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2323 err = 0;
2324 errout:
2325 return err;
2328 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2330 struct fib6_config r_cfg;
2331 struct rtnexthop *rtnh;
2332 int remaining;
2333 int attrlen;
2334 int err = 0, last_err = 0;
2336 beginning:
2337 rtnh = (struct rtnexthop *)cfg->fc_mp;
2338 remaining = cfg->fc_mp_len;
2340 /* Parse a Multipath Entry */
2341 while (rtnh_ok(rtnh, remaining)) {
2342 memcpy(&r_cfg, cfg, sizeof(*cfg));
2343 if (rtnh->rtnh_ifindex)
2344 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2346 attrlen = rtnh_attrlen(rtnh);
2347 if (attrlen > 0) {
2348 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2350 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2351 if (nla) {
2352 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2353 r_cfg.fc_flags |= RTF_GATEWAY;
2356 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2357 if (err) {
2358 last_err = err;
2359 /* If we are trying to remove a route, do not stop the
2360 * loop when ip6_route_del() fails (because next hop is
2361 * already gone), we should try to remove all next hops.
2363 if (add) {
2364 /* If add fails, we should try to delete all
2365 * next hops that have been already added.
2367 add = 0;
2368 goto beginning;
2371 /* Because each route is added like a single route we remove
2372 * this flag after the first nexthop (if there is a collision,
2373 * we have already fail to add the first nexthop:
2374 * fib6_add_rt2node() has reject it).
2376 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2377 rtnh = rtnh_next(rtnh, &remaining);
2380 return last_err;
2383 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2385 struct fib6_config cfg;
2386 int err;
2388 err = rtm_to_fib6_config(skb, nlh, &cfg);
2389 if (err < 0)
2390 return err;
2392 if (cfg.fc_mp)
2393 return ip6_route_multipath(&cfg, 0);
2394 else
2395 return ip6_route_del(&cfg);
2398 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2400 struct fib6_config cfg;
2401 int err;
2403 err = rtm_to_fib6_config(skb, nlh, &cfg);
2404 if (err < 0)
2405 return err;
2407 if (cfg.fc_mp)
2408 return ip6_route_multipath(&cfg, 1);
2409 else
2410 return ip6_route_add(&cfg);
2413 static inline size_t rt6_nlmsg_size(void)
2415 return NLMSG_ALIGN(sizeof(struct rtmsg))
2416 + nla_total_size(16) /* RTA_SRC */
2417 + nla_total_size(16) /* RTA_DST */
2418 + nla_total_size(16) /* RTA_GATEWAY */
2419 + nla_total_size(16) /* RTA_PREFSRC */
2420 + nla_total_size(4) /* RTA_TABLE */
2421 + nla_total_size(4) /* RTA_IIF */
2422 + nla_total_size(4) /* RTA_OIF */
2423 + nla_total_size(4) /* RTA_PRIORITY */
2424 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2425 + nla_total_size(sizeof(struct rta_cacheinfo));
2428 static int rt6_fill_node(struct net *net,
2429 struct sk_buff *skb, struct rt6_info *rt,
2430 struct in6_addr *dst, struct in6_addr *src,
2431 int iif, int type, u32 portid, u32 seq,
2432 int prefix, int nowait, unsigned int flags)
2434 struct rtmsg *rtm;
2435 struct nlmsghdr *nlh;
2436 long expires;
2437 u32 table;
2439 if (prefix) { /* user wants prefix routes only */
2440 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2441 /* success since this is not a prefix route */
2442 return 1;
2446 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2447 if (!nlh)
2448 return -EMSGSIZE;
2450 rtm = nlmsg_data(nlh);
2451 rtm->rtm_family = AF_INET6;
2452 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2453 rtm->rtm_src_len = rt->rt6i_src.plen;
2454 rtm->rtm_tos = 0;
2455 if (rt->rt6i_table)
2456 table = rt->rt6i_table->tb6_id;
2457 else
2458 table = RT6_TABLE_UNSPEC;
2459 rtm->rtm_table = table;
2460 if (nla_put_u32(skb, RTA_TABLE, table))
2461 goto nla_put_failure;
2462 if (rt->rt6i_flags & RTF_REJECT) {
2463 switch (rt->dst.error) {
2464 case -EINVAL:
2465 rtm->rtm_type = RTN_BLACKHOLE;
2466 break;
2467 case -EACCES:
2468 rtm->rtm_type = RTN_PROHIBIT;
2469 break;
2470 case -EAGAIN:
2471 rtm->rtm_type = RTN_THROW;
2472 break;
2473 default:
2474 rtm->rtm_type = RTN_UNREACHABLE;
2475 break;
2478 else if (rt->rt6i_flags & RTF_LOCAL)
2479 rtm->rtm_type = RTN_LOCAL;
2480 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2481 rtm->rtm_type = RTN_LOCAL;
2482 else
2483 rtm->rtm_type = RTN_UNICAST;
2484 rtm->rtm_flags = 0;
2485 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2486 rtm->rtm_protocol = rt->rt6i_protocol;
2487 if (rt->rt6i_flags & RTF_DYNAMIC)
2488 rtm->rtm_protocol = RTPROT_REDIRECT;
2489 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2490 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2491 rtm->rtm_protocol = RTPROT_RA;
2492 else
2493 rtm->rtm_protocol = RTPROT_KERNEL;
2496 if (rt->rt6i_flags & RTF_CACHE)
2497 rtm->rtm_flags |= RTM_F_CLONED;
2499 if (dst) {
2500 if (nla_put(skb, RTA_DST, 16, dst))
2501 goto nla_put_failure;
2502 rtm->rtm_dst_len = 128;
2503 } else if (rtm->rtm_dst_len)
2504 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2505 goto nla_put_failure;
2506 #ifdef CONFIG_IPV6_SUBTREES
2507 if (src) {
2508 if (nla_put(skb, RTA_SRC, 16, src))
2509 goto nla_put_failure;
2510 rtm->rtm_src_len = 128;
2511 } else if (rtm->rtm_src_len &&
2512 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2513 goto nla_put_failure;
2514 #endif
2515 if (iif) {
2516 #ifdef CONFIG_IPV6_MROUTE
2517 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2518 int err = ip6mr_get_route(net, skb, rtm, nowait);
2519 if (err <= 0) {
2520 if (!nowait) {
2521 if (err == 0)
2522 return 0;
2523 goto nla_put_failure;
2524 } else {
2525 if (err == -EMSGSIZE)
2526 goto nla_put_failure;
2529 } else
2530 #endif
2531 if (nla_put_u32(skb, RTA_IIF, iif))
2532 goto nla_put_failure;
2533 } else if (dst) {
2534 struct in6_addr saddr_buf;
2535 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2536 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2537 goto nla_put_failure;
2540 if (rt->rt6i_prefsrc.plen) {
2541 struct in6_addr saddr_buf;
2542 saddr_buf = rt->rt6i_prefsrc.addr;
2543 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2544 goto nla_put_failure;
2547 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2548 goto nla_put_failure;
2550 if (rt->rt6i_flags & RTF_GATEWAY) {
2551 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2552 goto nla_put_failure;
2555 if (rt->dst.dev &&
2556 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2557 goto nla_put_failure;
2558 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2559 goto nla_put_failure;
2561 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2563 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2564 goto nla_put_failure;
2566 return nlmsg_end(skb, nlh);
2568 nla_put_failure:
2569 nlmsg_cancel(skb, nlh);
2570 return -EMSGSIZE;
2573 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2575 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2576 int prefix;
2578 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2579 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2580 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2581 } else
2582 prefix = 0;
2584 return rt6_fill_node(arg->net,
2585 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2586 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2587 prefix, 0, NLM_F_MULTI);
2590 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2592 struct net *net = sock_net(in_skb->sk);
2593 struct nlattr *tb[RTA_MAX+1];
2594 struct rt6_info *rt;
2595 struct sk_buff *skb;
2596 struct rtmsg *rtm;
2597 struct flowi6 fl6;
2598 int err, iif = 0, oif = 0;
2600 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2601 if (err < 0)
2602 goto errout;
2604 err = -EINVAL;
2605 memset(&fl6, 0, sizeof(fl6));
2607 if (tb[RTA_SRC]) {
2608 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2609 goto errout;
2611 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2614 if (tb[RTA_DST]) {
2615 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2616 goto errout;
2618 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2621 if (tb[RTA_IIF])
2622 iif = nla_get_u32(tb[RTA_IIF]);
2624 if (tb[RTA_OIF])
2625 oif = nla_get_u32(tb[RTA_OIF]);
2627 if (iif) {
2628 struct net_device *dev;
2629 int flags = 0;
2631 dev = __dev_get_by_index(net, iif);
2632 if (!dev) {
2633 err = -ENODEV;
2634 goto errout;
2637 fl6.flowi6_iif = iif;
2639 if (!ipv6_addr_any(&fl6.saddr))
2640 flags |= RT6_LOOKUP_F_HAS_SADDR;
2642 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2643 flags);
2644 } else {
2645 fl6.flowi6_oif = oif;
2647 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2650 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2651 if (!skb) {
2652 ip6_rt_put(rt);
2653 err = -ENOBUFS;
2654 goto errout;
2657 /* Reserve room for dummy headers, this skb can pass
2658 through good chunk of routing engine.
2660 skb_reset_mac_header(skb);
2661 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2663 skb_dst_set(skb, &rt->dst);
2665 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2666 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2667 nlh->nlmsg_seq, 0, 0, 0);
2668 if (err < 0) {
2669 kfree_skb(skb);
2670 goto errout;
2673 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2674 errout:
2675 return err;
2678 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2680 struct sk_buff *skb;
2681 struct net *net = info->nl_net;
2682 u32 seq;
2683 int err;
2685 err = -ENOBUFS;
2686 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2688 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2689 if (!skb)
2690 goto errout;
2692 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2693 event, info->portid, seq, 0, 0, 0);
2694 if (err < 0) {
2695 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2696 WARN_ON(err == -EMSGSIZE);
2697 kfree_skb(skb);
2698 goto errout;
2700 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2701 info->nlh, gfp_any());
2702 return;
2703 errout:
2704 if (err < 0)
2705 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2708 static int ip6_route_dev_notify(struct notifier_block *this,
2709 unsigned long event, void *ptr)
2711 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2712 struct net *net = dev_net(dev);
2714 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2715 net->ipv6.ip6_null_entry->dst.dev = dev;
2716 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2717 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2718 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2719 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2720 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2721 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2722 #endif
2725 return NOTIFY_OK;
2729 * /proc
2732 #ifdef CONFIG_PROC_FS
2734 struct rt6_proc_arg
2736 char *buffer;
2737 int offset;
2738 int length;
2739 int skip;
2740 int len;
2743 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2745 struct seq_file *m = p_arg;
2747 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2749 #ifdef CONFIG_IPV6_SUBTREES
2750 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2751 #else
2752 seq_puts(m, "00000000000000000000000000000000 00 ");
2753 #endif
2754 if (rt->rt6i_flags & RTF_GATEWAY) {
2755 seq_printf(m, "%pi6", &rt->rt6i_gateway);
2756 } else {
2757 seq_puts(m, "00000000000000000000000000000000");
2759 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2760 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2761 rt->dst.__use, rt->rt6i_flags,
2762 rt->dst.dev ? rt->dst.dev->name : "");
2763 return 0;
2766 static int ipv6_route_show(struct seq_file *m, void *v)
2768 struct net *net = (struct net *)m->private;
2769 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2770 return 0;
2773 static int ipv6_route_open(struct inode *inode, struct file *file)
2775 return single_open_net(inode, file, ipv6_route_show);
2778 static const struct file_operations ipv6_route_proc_fops = {
2779 .owner = THIS_MODULE,
2780 .open = ipv6_route_open,
2781 .read = seq_read,
2782 .llseek = seq_lseek,
2783 .release = single_release_net,
2786 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2788 struct net *net = (struct net *)seq->private;
2789 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2790 net->ipv6.rt6_stats->fib_nodes,
2791 net->ipv6.rt6_stats->fib_route_nodes,
2792 net->ipv6.rt6_stats->fib_rt_alloc,
2793 net->ipv6.rt6_stats->fib_rt_entries,
2794 net->ipv6.rt6_stats->fib_rt_cache,
2795 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2796 net->ipv6.rt6_stats->fib_discarded_routes);
2798 return 0;
2801 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2803 return single_open_net(inode, file, rt6_stats_seq_show);
2806 static const struct file_operations rt6_stats_seq_fops = {
2807 .owner = THIS_MODULE,
2808 .open = rt6_stats_seq_open,
2809 .read = seq_read,
2810 .llseek = seq_lseek,
2811 .release = single_release_net,
2813 #endif /* CONFIG_PROC_FS */
2815 #ifdef CONFIG_SYSCTL
2817 static
2818 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2819 void __user *buffer, size_t *lenp, loff_t *ppos)
2821 struct net *net;
2822 int delay;
2823 if (!write)
2824 return -EINVAL;
2826 net = (struct net *)ctl->extra1;
2827 delay = net->ipv6.sysctl.flush_delay;
2828 proc_dointvec(ctl, write, buffer, lenp, ppos);
2829 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2830 return 0;
2833 struct ctl_table ipv6_route_table_template[] = {
2835 .procname = "flush",
2836 .data = &init_net.ipv6.sysctl.flush_delay,
2837 .maxlen = sizeof(int),
2838 .mode = 0200,
2839 .proc_handler = ipv6_sysctl_rtcache_flush
2842 .procname = "gc_thresh",
2843 .data = &ip6_dst_ops_template.gc_thresh,
2844 .maxlen = sizeof(int),
2845 .mode = 0644,
2846 .proc_handler = proc_dointvec,
2849 .procname = "max_size",
2850 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2851 .maxlen = sizeof(int),
2852 .mode = 0644,
2853 .proc_handler = proc_dointvec,
2856 .procname = "gc_min_interval",
2857 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2858 .maxlen = sizeof(int),
2859 .mode = 0644,
2860 .proc_handler = proc_dointvec_jiffies,
2863 .procname = "gc_timeout",
2864 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2865 .maxlen = sizeof(int),
2866 .mode = 0644,
2867 .proc_handler = proc_dointvec_jiffies,
2870 .procname = "gc_interval",
2871 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2872 .maxlen = sizeof(int),
2873 .mode = 0644,
2874 .proc_handler = proc_dointvec_jiffies,
2877 .procname = "gc_elasticity",
2878 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2879 .maxlen = sizeof(int),
2880 .mode = 0644,
2881 .proc_handler = proc_dointvec,
2884 .procname = "mtu_expires",
2885 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2886 .maxlen = sizeof(int),
2887 .mode = 0644,
2888 .proc_handler = proc_dointvec_jiffies,
2891 .procname = "min_adv_mss",
2892 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2893 .maxlen = sizeof(int),
2894 .mode = 0644,
2895 .proc_handler = proc_dointvec,
2898 .procname = "gc_min_interval_ms",
2899 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2900 .maxlen = sizeof(int),
2901 .mode = 0644,
2902 .proc_handler = proc_dointvec_ms_jiffies,
2907 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2909 struct ctl_table *table;
2911 table = kmemdup(ipv6_route_table_template,
2912 sizeof(ipv6_route_table_template),
2913 GFP_KERNEL);
2915 if (table) {
2916 table[0].data = &net->ipv6.sysctl.flush_delay;
2917 table[0].extra1 = net;
2918 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2919 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2920 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2921 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2922 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2923 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2924 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2925 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2926 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2928 /* Don't export sysctls to unprivileged users */
2929 if (net->user_ns != &init_user_ns)
2930 table[0].procname = NULL;
2933 return table;
2935 #endif
2937 static int __net_init ip6_route_net_init(struct net *net)
2939 int ret = -ENOMEM;
2941 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2942 sizeof(net->ipv6.ip6_dst_ops));
2944 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2945 goto out_ip6_dst_ops;
2947 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2948 sizeof(*net->ipv6.ip6_null_entry),
2949 GFP_KERNEL);
2950 if (!net->ipv6.ip6_null_entry)
2951 goto out_ip6_dst_entries;
2952 net->ipv6.ip6_null_entry->dst.path =
2953 (struct dst_entry *)net->ipv6.ip6_null_entry;
2954 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2955 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2956 ip6_template_metrics, true);
2958 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2959 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2960 sizeof(*net->ipv6.ip6_prohibit_entry),
2961 GFP_KERNEL);
2962 if (!net->ipv6.ip6_prohibit_entry)
2963 goto out_ip6_null_entry;
2964 net->ipv6.ip6_prohibit_entry->dst.path =
2965 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2966 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2967 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2968 ip6_template_metrics, true);
2970 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2971 sizeof(*net->ipv6.ip6_blk_hole_entry),
2972 GFP_KERNEL);
2973 if (!net->ipv6.ip6_blk_hole_entry)
2974 goto out_ip6_prohibit_entry;
2975 net->ipv6.ip6_blk_hole_entry->dst.path =
2976 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2977 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2978 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2979 ip6_template_metrics, true);
2980 #endif
2982 net->ipv6.sysctl.flush_delay = 0;
2983 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2984 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2985 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2986 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2987 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2988 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2989 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2991 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2993 ret = 0;
2994 out:
2995 return ret;
2997 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2998 out_ip6_prohibit_entry:
2999 kfree(net->ipv6.ip6_prohibit_entry);
3000 out_ip6_null_entry:
3001 kfree(net->ipv6.ip6_null_entry);
3002 #endif
3003 out_ip6_dst_entries:
3004 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3005 out_ip6_dst_ops:
3006 goto out;
3009 static void __net_exit ip6_route_net_exit(struct net *net)
3011 kfree(net->ipv6.ip6_null_entry);
3012 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3013 kfree(net->ipv6.ip6_prohibit_entry);
3014 kfree(net->ipv6.ip6_blk_hole_entry);
3015 #endif
3016 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3019 static int __net_init ip6_route_net_init_late(struct net *net)
3021 #ifdef CONFIG_PROC_FS
3022 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3023 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3024 #endif
3025 return 0;
3028 static void __net_exit ip6_route_net_exit_late(struct net *net)
3030 #ifdef CONFIG_PROC_FS
3031 remove_proc_entry("ipv6_route", net->proc_net);
3032 remove_proc_entry("rt6_stats", net->proc_net);
3033 #endif
3036 static struct pernet_operations ip6_route_net_ops = {
3037 .init = ip6_route_net_init,
3038 .exit = ip6_route_net_exit,
3041 static int __net_init ipv6_inetpeer_init(struct net *net)
3043 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3045 if (!bp)
3046 return -ENOMEM;
3047 inet_peer_base_init(bp);
3048 net->ipv6.peers = bp;
3049 return 0;
3052 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3054 struct inet_peer_base *bp = net->ipv6.peers;
3056 net->ipv6.peers = NULL;
3057 inetpeer_invalidate_tree(bp);
3058 kfree(bp);
3061 static struct pernet_operations ipv6_inetpeer_ops = {
3062 .init = ipv6_inetpeer_init,
3063 .exit = ipv6_inetpeer_exit,
3066 static struct pernet_operations ip6_route_net_late_ops = {
3067 .init = ip6_route_net_init_late,
3068 .exit = ip6_route_net_exit_late,
3071 static struct notifier_block ip6_route_dev_notifier = {
3072 .notifier_call = ip6_route_dev_notify,
3073 .priority = 0,
3076 int __init ip6_route_init(void)
3078 int ret;
3080 ret = -ENOMEM;
3081 ip6_dst_ops_template.kmem_cachep =
3082 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3083 SLAB_HWCACHE_ALIGN, NULL);
3084 if (!ip6_dst_ops_template.kmem_cachep)
3085 goto out;
3087 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3088 if (ret)
3089 goto out_kmem_cache;
3091 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3092 if (ret)
3093 goto out_dst_entries;
3095 ret = register_pernet_subsys(&ip6_route_net_ops);
3096 if (ret)
3097 goto out_register_inetpeer;
3099 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3101 /* Registering of the loopback is done before this portion of code,
3102 * the loopback reference in rt6_info will not be taken, do it
3103 * manually for init_net */
3104 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3105 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3106 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3107 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3108 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3109 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3110 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3111 #endif
3112 ret = fib6_init();
3113 if (ret)
3114 goto out_register_subsys;
3116 ret = xfrm6_init();
3117 if (ret)
3118 goto out_fib6_init;
3120 ret = fib6_rules_init();
3121 if (ret)
3122 goto xfrm6_init;
3124 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3125 if (ret)
3126 goto fib6_rules_init;
3128 ret = -ENOBUFS;
3129 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3130 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3131 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3132 goto out_register_late_subsys;
3134 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3135 if (ret)
3136 goto out_register_late_subsys;
3138 out:
3139 return ret;
3141 out_register_late_subsys:
3142 unregister_pernet_subsys(&ip6_route_net_late_ops);
3143 fib6_rules_init:
3144 fib6_rules_cleanup();
3145 xfrm6_init:
3146 xfrm6_fini();
3147 out_fib6_init:
3148 fib6_gc_cleanup();
3149 out_register_subsys:
3150 unregister_pernet_subsys(&ip6_route_net_ops);
3151 out_register_inetpeer:
3152 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3153 out_dst_entries:
3154 dst_entries_destroy(&ip6_dst_blackhole_ops);
3155 out_kmem_cache:
3156 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3157 goto out;
3160 void ip6_route_cleanup(void)
3162 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3163 unregister_pernet_subsys(&ip6_route_net_late_ops);
3164 fib6_rules_cleanup();
3165 xfrm6_fini();
3166 fib6_gc_cleanup();
3167 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3168 unregister_pernet_subsys(&ip6_route_net_ops);
3169 dst_entries_destroy(&ip6_dst_blackhole_ops);
3170 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);