Loading...
1/*
2 * Linux INET6 implementation
3 * FIB front-end.
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
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.
12 */
13
14/* Changes:
15 *
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.
25 */
26
27#include <linux/capability.h>
28#include <linux/errno.h>
29#include <linux/types.h>
30#include <linux/times.h>
31#include <linux/socket.h>
32#include <linux/sockios.h>
33#include <linux/net.h>
34#include <linux/route.h>
35#include <linux/netdevice.h>
36#include <linux/in6.h>
37#include <linux/mroute6.h>
38#include <linux/init.h>
39#include <linux/if_arp.h>
40#include <linux/proc_fs.h>
41#include <linux/seq_file.h>
42#include <linux/nsproxy.h>
43#include <linux/slab.h>
44#include <net/net_namespace.h>
45#include <net/snmp.h>
46#include <net/ipv6.h>
47#include <net/ip6_fib.h>
48#include <net/ip6_route.h>
49#include <net/ndisc.h>
50#include <net/addrconf.h>
51#include <net/tcp.h>
52#include <linux/rtnetlink.h>
53#include <net/dst.h>
54#include <net/xfrm.h>
55#include <net/netevent.h>
56#include <net/netlink.h>
57
58#include <asm/uaccess.h>
59
60#ifdef CONFIG_SYSCTL
61#include <linux/sysctl.h>
62#endif
63
64/* Set to 3 to get tracing. */
65#define RT6_DEBUG 2
66
67#if RT6_DEBUG >= 3
68#define RDBG(x) printk x
69#define RT6_TRACE(x...) printk(KERN_DEBUG x)
70#else
71#define RDBG(x)
72#define RT6_TRACE(x...) do { ; } while (0)
73#endif
74
75static struct rt6_info *ip6_rt_copy(const struct rt6_info *ort,
76 const struct in6_addr *dest);
77static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
78static unsigned int ip6_default_advmss(const struct dst_entry *dst);
79static unsigned int ip6_default_mtu(const struct dst_entry *dst);
80static struct dst_entry *ip6_negative_advice(struct dst_entry *);
81static void ip6_dst_destroy(struct dst_entry *);
82static void ip6_dst_ifdown(struct dst_entry *,
83 struct net_device *dev, int how);
84static int ip6_dst_gc(struct dst_ops *ops);
85
86static int ip6_pkt_discard(struct sk_buff *skb);
87static int ip6_pkt_discard_out(struct sk_buff *skb);
88static void ip6_link_failure(struct sk_buff *skb);
89static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
90
91#ifdef CONFIG_IPV6_ROUTE_INFO
92static struct rt6_info *rt6_add_route_info(struct net *net,
93 const struct in6_addr *prefix, int prefixlen,
94 const struct in6_addr *gwaddr, int ifindex,
95 unsigned pref);
96static struct rt6_info *rt6_get_route_info(struct net *net,
97 const struct in6_addr *prefix, int prefixlen,
98 const struct in6_addr *gwaddr, int ifindex);
99#endif
100
101static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
102{
103 struct rt6_info *rt = (struct rt6_info *) dst;
104 struct inet_peer *peer;
105 u32 *p = NULL;
106
107 if (!(rt->dst.flags & DST_HOST))
108 return NULL;
109
110 if (!rt->rt6i_peer)
111 rt6_bind_peer(rt, 1);
112
113 peer = rt->rt6i_peer;
114 if (peer) {
115 u32 *old_p = __DST_METRICS_PTR(old);
116 unsigned long prev, new;
117
118 p = peer->metrics;
119 if (inet_metrics_new(peer))
120 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
121
122 new = (unsigned long) p;
123 prev = cmpxchg(&dst->_metrics, old, new);
124
125 if (prev != old) {
126 p = __DST_METRICS_PTR(prev);
127 if (prev & DST_METRICS_READ_ONLY)
128 p = NULL;
129 }
130 }
131 return p;
132}
133
134static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst, const void *daddr)
135{
136 return __neigh_lookup_errno(&nd_tbl, daddr, dst->dev);
137}
138
139static struct dst_ops ip6_dst_ops_template = {
140 .family = AF_INET6,
141 .protocol = cpu_to_be16(ETH_P_IPV6),
142 .gc = ip6_dst_gc,
143 .gc_thresh = 1024,
144 .check = ip6_dst_check,
145 .default_advmss = ip6_default_advmss,
146 .default_mtu = ip6_default_mtu,
147 .cow_metrics = ipv6_cow_metrics,
148 .destroy = ip6_dst_destroy,
149 .ifdown = ip6_dst_ifdown,
150 .negative_advice = ip6_negative_advice,
151 .link_failure = ip6_link_failure,
152 .update_pmtu = ip6_rt_update_pmtu,
153 .local_out = __ip6_local_out,
154 .neigh_lookup = ip6_neigh_lookup,
155};
156
157static unsigned int ip6_blackhole_default_mtu(const struct dst_entry *dst)
158{
159 return 0;
160}
161
162static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
163{
164}
165
166static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
167 unsigned long old)
168{
169 return NULL;
170}
171
172static struct dst_ops ip6_dst_blackhole_ops = {
173 .family = AF_INET6,
174 .protocol = cpu_to_be16(ETH_P_IPV6),
175 .destroy = ip6_dst_destroy,
176 .check = ip6_dst_check,
177 .default_mtu = ip6_blackhole_default_mtu,
178 .default_advmss = ip6_default_advmss,
179 .update_pmtu = ip6_rt_blackhole_update_pmtu,
180 .cow_metrics = ip6_rt_blackhole_cow_metrics,
181 .neigh_lookup = ip6_neigh_lookup,
182};
183
184static const u32 ip6_template_metrics[RTAX_MAX] = {
185 [RTAX_HOPLIMIT - 1] = 255,
186};
187
188static struct rt6_info ip6_null_entry_template = {
189 .dst = {
190 .__refcnt = ATOMIC_INIT(1),
191 .__use = 1,
192 .obsolete = -1,
193 .error = -ENETUNREACH,
194 .input = ip6_pkt_discard,
195 .output = ip6_pkt_discard_out,
196 },
197 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
198 .rt6i_protocol = RTPROT_KERNEL,
199 .rt6i_metric = ~(u32) 0,
200 .rt6i_ref = ATOMIC_INIT(1),
201};
202
203#ifdef CONFIG_IPV6_MULTIPLE_TABLES
204
205static int ip6_pkt_prohibit(struct sk_buff *skb);
206static int ip6_pkt_prohibit_out(struct sk_buff *skb);
207
208static struct rt6_info ip6_prohibit_entry_template = {
209 .dst = {
210 .__refcnt = ATOMIC_INIT(1),
211 .__use = 1,
212 .obsolete = -1,
213 .error = -EACCES,
214 .input = ip6_pkt_prohibit,
215 .output = ip6_pkt_prohibit_out,
216 },
217 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
218 .rt6i_protocol = RTPROT_KERNEL,
219 .rt6i_metric = ~(u32) 0,
220 .rt6i_ref = ATOMIC_INIT(1),
221};
222
223static struct rt6_info ip6_blk_hole_entry_template = {
224 .dst = {
225 .__refcnt = ATOMIC_INIT(1),
226 .__use = 1,
227 .obsolete = -1,
228 .error = -EINVAL,
229 .input = dst_discard,
230 .output = dst_discard,
231 },
232 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
233 .rt6i_protocol = RTPROT_KERNEL,
234 .rt6i_metric = ~(u32) 0,
235 .rt6i_ref = ATOMIC_INIT(1),
236};
237
238#endif
239
240/* allocate dst with ip6_dst_ops */
241static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops,
242 struct net_device *dev,
243 int flags)
244{
245 struct rt6_info *rt = dst_alloc(ops, dev, 0, 0, flags);
246
247 if (rt != NULL)
248 memset(&rt->rt6i_table, 0,
249 sizeof(*rt) - sizeof(struct dst_entry));
250
251 return rt;
252}
253
254static void ip6_dst_destroy(struct dst_entry *dst)
255{
256 struct rt6_info *rt = (struct rt6_info *)dst;
257 struct inet6_dev *idev = rt->rt6i_idev;
258 struct inet_peer *peer = rt->rt6i_peer;
259
260 if (!(rt->dst.flags & DST_HOST))
261 dst_destroy_metrics_generic(dst);
262
263 if (idev != NULL) {
264 rt->rt6i_idev = NULL;
265 in6_dev_put(idev);
266 }
267 if (peer) {
268 rt->rt6i_peer = NULL;
269 inet_putpeer(peer);
270 }
271}
272
273static atomic_t __rt6_peer_genid = ATOMIC_INIT(0);
274
275static u32 rt6_peer_genid(void)
276{
277 return atomic_read(&__rt6_peer_genid);
278}
279
280void rt6_bind_peer(struct rt6_info *rt, int create)
281{
282 struct inet_peer *peer;
283
284 peer = inet_getpeer_v6(&rt->rt6i_dst.addr, create);
285 if (peer && cmpxchg(&rt->rt6i_peer, NULL, peer) != NULL)
286 inet_putpeer(peer);
287 else
288 rt->rt6i_peer_genid = rt6_peer_genid();
289}
290
291static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
292 int how)
293{
294 struct rt6_info *rt = (struct rt6_info *)dst;
295 struct inet6_dev *idev = rt->rt6i_idev;
296 struct net_device *loopback_dev =
297 dev_net(dev)->loopback_dev;
298
299 if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
300 struct inet6_dev *loopback_idev =
301 in6_dev_get(loopback_dev);
302 if (loopback_idev != NULL) {
303 rt->rt6i_idev = loopback_idev;
304 in6_dev_put(idev);
305 }
306 }
307}
308
309static __inline__ int rt6_check_expired(const struct rt6_info *rt)
310{
311 return (rt->rt6i_flags & RTF_EXPIRES) &&
312 time_after(jiffies, rt->rt6i_expires);
313}
314
315static inline int rt6_need_strict(const struct in6_addr *daddr)
316{
317 return ipv6_addr_type(daddr) &
318 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
319}
320
321/*
322 * Route lookup. Any table->tb6_lock is implied.
323 */
324
325static inline struct rt6_info *rt6_device_match(struct net *net,
326 struct rt6_info *rt,
327 const struct in6_addr *saddr,
328 int oif,
329 int flags)
330{
331 struct rt6_info *local = NULL;
332 struct rt6_info *sprt;
333
334 if (!oif && ipv6_addr_any(saddr))
335 goto out;
336
337 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
338 struct net_device *dev = sprt->rt6i_dev;
339
340 if (oif) {
341 if (dev->ifindex == oif)
342 return sprt;
343 if (dev->flags & IFF_LOOPBACK) {
344 if (sprt->rt6i_idev == NULL ||
345 sprt->rt6i_idev->dev->ifindex != oif) {
346 if (flags & RT6_LOOKUP_F_IFACE && oif)
347 continue;
348 if (local && (!oif ||
349 local->rt6i_idev->dev->ifindex == oif))
350 continue;
351 }
352 local = sprt;
353 }
354 } else {
355 if (ipv6_chk_addr(net, saddr, dev,
356 flags & RT6_LOOKUP_F_IFACE))
357 return sprt;
358 }
359 }
360
361 if (oif) {
362 if (local)
363 return local;
364
365 if (flags & RT6_LOOKUP_F_IFACE)
366 return net->ipv6.ip6_null_entry;
367 }
368out:
369 return rt;
370}
371
372#ifdef CONFIG_IPV6_ROUTER_PREF
373static void rt6_probe(struct rt6_info *rt)
374{
375 struct neighbour *neigh;
376 /*
377 * Okay, this does not seem to be appropriate
378 * for now, however, we need to check if it
379 * is really so; aka Router Reachability Probing.
380 *
381 * Router Reachability Probe MUST be rate-limited
382 * to no more than one per minute.
383 */
384 rcu_read_lock();
385 neigh = rt ? dst_get_neighbour(&rt->dst) : NULL;
386 if (!neigh || (neigh->nud_state & NUD_VALID))
387 goto out;
388 read_lock_bh(&neigh->lock);
389 if (!(neigh->nud_state & NUD_VALID) &&
390 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
391 struct in6_addr mcaddr;
392 struct in6_addr *target;
393
394 neigh->updated = jiffies;
395 read_unlock_bh(&neigh->lock);
396
397 target = (struct in6_addr *)&neigh->primary_key;
398 addrconf_addr_solict_mult(target, &mcaddr);
399 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
400 } else {
401 read_unlock_bh(&neigh->lock);
402 }
403out:
404 rcu_read_unlock();
405}
406#else
407static inline void rt6_probe(struct rt6_info *rt)
408{
409}
410#endif
411
412/*
413 * Default Router Selection (RFC 2461 6.3.6)
414 */
415static inline int rt6_check_dev(struct rt6_info *rt, int oif)
416{
417 struct net_device *dev = rt->rt6i_dev;
418 if (!oif || dev->ifindex == oif)
419 return 2;
420 if ((dev->flags & IFF_LOOPBACK) &&
421 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
422 return 1;
423 return 0;
424}
425
426static inline int rt6_check_neigh(struct rt6_info *rt)
427{
428 struct neighbour *neigh;
429 int m;
430
431 rcu_read_lock();
432 neigh = dst_get_neighbour(&rt->dst);
433 if (rt->rt6i_flags & RTF_NONEXTHOP ||
434 !(rt->rt6i_flags & RTF_GATEWAY))
435 m = 1;
436 else if (neigh) {
437 read_lock_bh(&neigh->lock);
438 if (neigh->nud_state & NUD_VALID)
439 m = 2;
440#ifdef CONFIG_IPV6_ROUTER_PREF
441 else if (neigh->nud_state & NUD_FAILED)
442 m = 0;
443#endif
444 else
445 m = 1;
446 read_unlock_bh(&neigh->lock);
447 } else
448 m = 0;
449 rcu_read_unlock();
450 return m;
451}
452
453static int rt6_score_route(struct rt6_info *rt, int oif,
454 int strict)
455{
456 int m, n;
457
458 m = rt6_check_dev(rt, oif);
459 if (!m && (strict & RT6_LOOKUP_F_IFACE))
460 return -1;
461#ifdef CONFIG_IPV6_ROUTER_PREF
462 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
463#endif
464 n = rt6_check_neigh(rt);
465 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
466 return -1;
467 return m;
468}
469
470static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
471 int *mpri, struct rt6_info *match)
472{
473 int m;
474
475 if (rt6_check_expired(rt))
476 goto out;
477
478 m = rt6_score_route(rt, oif, strict);
479 if (m < 0)
480 goto out;
481
482 if (m > *mpri) {
483 if (strict & RT6_LOOKUP_F_REACHABLE)
484 rt6_probe(match);
485 *mpri = m;
486 match = rt;
487 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
488 rt6_probe(rt);
489 }
490
491out:
492 return match;
493}
494
495static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
496 struct rt6_info *rr_head,
497 u32 metric, int oif, int strict)
498{
499 struct rt6_info *rt, *match;
500 int mpri = -1;
501
502 match = NULL;
503 for (rt = rr_head; rt && rt->rt6i_metric == metric;
504 rt = rt->dst.rt6_next)
505 match = find_match(rt, oif, strict, &mpri, match);
506 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
507 rt = rt->dst.rt6_next)
508 match = find_match(rt, oif, strict, &mpri, match);
509
510 return match;
511}
512
513static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
514{
515 struct rt6_info *match, *rt0;
516 struct net *net;
517
518 RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
519 __func__, fn->leaf, oif);
520
521 rt0 = fn->rr_ptr;
522 if (!rt0)
523 fn->rr_ptr = rt0 = fn->leaf;
524
525 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
526
527 if (!match &&
528 (strict & RT6_LOOKUP_F_REACHABLE)) {
529 struct rt6_info *next = rt0->dst.rt6_next;
530
531 /* no entries matched; do round-robin */
532 if (!next || next->rt6i_metric != rt0->rt6i_metric)
533 next = fn->leaf;
534
535 if (next != rt0)
536 fn->rr_ptr = next;
537 }
538
539 RT6_TRACE("%s() => %p\n",
540 __func__, match);
541
542 net = dev_net(rt0->rt6i_dev);
543 return match ? match : net->ipv6.ip6_null_entry;
544}
545
546#ifdef CONFIG_IPV6_ROUTE_INFO
547int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
548 const struct in6_addr *gwaddr)
549{
550 struct net *net = dev_net(dev);
551 struct route_info *rinfo = (struct route_info *) opt;
552 struct in6_addr prefix_buf, *prefix;
553 unsigned int pref;
554 unsigned long lifetime;
555 struct rt6_info *rt;
556
557 if (len < sizeof(struct route_info)) {
558 return -EINVAL;
559 }
560
561 /* Sanity check for prefix_len and length */
562 if (rinfo->length > 3) {
563 return -EINVAL;
564 } else if (rinfo->prefix_len > 128) {
565 return -EINVAL;
566 } else if (rinfo->prefix_len > 64) {
567 if (rinfo->length < 2) {
568 return -EINVAL;
569 }
570 } else if (rinfo->prefix_len > 0) {
571 if (rinfo->length < 1) {
572 return -EINVAL;
573 }
574 }
575
576 pref = rinfo->route_pref;
577 if (pref == ICMPV6_ROUTER_PREF_INVALID)
578 return -EINVAL;
579
580 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
581
582 if (rinfo->length == 3)
583 prefix = (struct in6_addr *)rinfo->prefix;
584 else {
585 /* this function is safe */
586 ipv6_addr_prefix(&prefix_buf,
587 (struct in6_addr *)rinfo->prefix,
588 rinfo->prefix_len);
589 prefix = &prefix_buf;
590 }
591
592 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
593 dev->ifindex);
594
595 if (rt && !lifetime) {
596 ip6_del_rt(rt);
597 rt = NULL;
598 }
599
600 if (!rt && lifetime)
601 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
602 pref);
603 else if (rt)
604 rt->rt6i_flags = RTF_ROUTEINFO |
605 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
606
607 if (rt) {
608 if (!addrconf_finite_timeout(lifetime)) {
609 rt->rt6i_flags &= ~RTF_EXPIRES;
610 } else {
611 rt->rt6i_expires = jiffies + HZ * lifetime;
612 rt->rt6i_flags |= RTF_EXPIRES;
613 }
614 dst_release(&rt->dst);
615 }
616 return 0;
617}
618#endif
619
620#define BACKTRACK(__net, saddr) \
621do { \
622 if (rt == __net->ipv6.ip6_null_entry) { \
623 struct fib6_node *pn; \
624 while (1) { \
625 if (fn->fn_flags & RTN_TL_ROOT) \
626 goto out; \
627 pn = fn->parent; \
628 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
629 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
630 else \
631 fn = pn; \
632 if (fn->fn_flags & RTN_RTINFO) \
633 goto restart; \
634 } \
635 } \
636} while(0)
637
638static struct rt6_info *ip6_pol_route_lookup(struct net *net,
639 struct fib6_table *table,
640 struct flowi6 *fl6, int flags)
641{
642 struct fib6_node *fn;
643 struct rt6_info *rt;
644
645 read_lock_bh(&table->tb6_lock);
646 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
647restart:
648 rt = fn->leaf;
649 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
650 BACKTRACK(net, &fl6->saddr);
651out:
652 dst_use(&rt->dst, jiffies);
653 read_unlock_bh(&table->tb6_lock);
654 return rt;
655
656}
657
658struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
659 const struct in6_addr *saddr, int oif, int strict)
660{
661 struct flowi6 fl6 = {
662 .flowi6_oif = oif,
663 .daddr = *daddr,
664 };
665 struct dst_entry *dst;
666 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
667
668 if (saddr) {
669 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
670 flags |= RT6_LOOKUP_F_HAS_SADDR;
671 }
672
673 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
674 if (dst->error == 0)
675 return (struct rt6_info *) dst;
676
677 dst_release(dst);
678
679 return NULL;
680}
681
682EXPORT_SYMBOL(rt6_lookup);
683
684/* ip6_ins_rt is called with FREE table->tb6_lock.
685 It takes new route entry, the addition fails by any reason the
686 route is freed. In any case, if caller does not hold it, it may
687 be destroyed.
688 */
689
690static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
691{
692 int err;
693 struct fib6_table *table;
694
695 table = rt->rt6i_table;
696 write_lock_bh(&table->tb6_lock);
697 err = fib6_add(&table->tb6_root, rt, info);
698 write_unlock_bh(&table->tb6_lock);
699
700 return err;
701}
702
703int ip6_ins_rt(struct rt6_info *rt)
704{
705 struct nl_info info = {
706 .nl_net = dev_net(rt->rt6i_dev),
707 };
708 return __ip6_ins_rt(rt, &info);
709}
710
711static struct rt6_info *rt6_alloc_cow(const struct rt6_info *ort,
712 const struct in6_addr *daddr,
713 const struct in6_addr *saddr)
714{
715 struct rt6_info *rt;
716
717 /*
718 * Clone the route.
719 */
720
721 rt = ip6_rt_copy(ort, daddr);
722
723 if (rt) {
724 struct neighbour *neigh;
725 int attempts = !in_softirq();
726
727 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
728 if (rt->rt6i_dst.plen != 128 &&
729 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
730 rt->rt6i_flags |= RTF_ANYCAST;
731 ipv6_addr_copy(&rt->rt6i_gateway, daddr);
732 }
733
734 rt->rt6i_flags |= RTF_CACHE;
735
736#ifdef CONFIG_IPV6_SUBTREES
737 if (rt->rt6i_src.plen && saddr) {
738 ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
739 rt->rt6i_src.plen = 128;
740 }
741#endif
742
743 retry:
744 neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
745 if (IS_ERR(neigh)) {
746 struct net *net = dev_net(rt->rt6i_dev);
747 int saved_rt_min_interval =
748 net->ipv6.sysctl.ip6_rt_gc_min_interval;
749 int saved_rt_elasticity =
750 net->ipv6.sysctl.ip6_rt_gc_elasticity;
751
752 if (attempts-- > 0) {
753 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
754 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
755
756 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
757
758 net->ipv6.sysctl.ip6_rt_gc_elasticity =
759 saved_rt_elasticity;
760 net->ipv6.sysctl.ip6_rt_gc_min_interval =
761 saved_rt_min_interval;
762 goto retry;
763 }
764
765 if (net_ratelimit())
766 printk(KERN_WARNING
767 "ipv6: Neighbour table overflow.\n");
768 dst_free(&rt->dst);
769 return NULL;
770 }
771 dst_set_neighbour(&rt->dst, neigh);
772
773 }
774
775 return rt;
776}
777
778static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
779 const struct in6_addr *daddr)
780{
781 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
782
783 if (rt) {
784 rt->rt6i_flags |= RTF_CACHE;
785 dst_set_neighbour(&rt->dst, neigh_clone(dst_get_neighbour_raw(&ort->dst)));
786 }
787 return rt;
788}
789
790static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
791 struct flowi6 *fl6, int flags)
792{
793 struct fib6_node *fn;
794 struct rt6_info *rt, *nrt;
795 int strict = 0;
796 int attempts = 3;
797 int err;
798 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
799
800 strict |= flags & RT6_LOOKUP_F_IFACE;
801
802relookup:
803 read_lock_bh(&table->tb6_lock);
804
805restart_2:
806 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
807
808restart:
809 rt = rt6_select(fn, oif, strict | reachable);
810
811 BACKTRACK(net, &fl6->saddr);
812 if (rt == net->ipv6.ip6_null_entry ||
813 rt->rt6i_flags & RTF_CACHE)
814 goto out;
815
816 dst_hold(&rt->dst);
817 read_unlock_bh(&table->tb6_lock);
818
819 if (!dst_get_neighbour_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
820 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
821 else if (!(rt->dst.flags & DST_HOST))
822 nrt = rt6_alloc_clone(rt, &fl6->daddr);
823 else
824 goto out2;
825
826 dst_release(&rt->dst);
827 rt = nrt ? : net->ipv6.ip6_null_entry;
828
829 dst_hold(&rt->dst);
830 if (nrt) {
831 err = ip6_ins_rt(nrt);
832 if (!err)
833 goto out2;
834 }
835
836 if (--attempts <= 0)
837 goto out2;
838
839 /*
840 * Race condition! In the gap, when table->tb6_lock was
841 * released someone could insert this route. Relookup.
842 */
843 dst_release(&rt->dst);
844 goto relookup;
845
846out:
847 if (reachable) {
848 reachable = 0;
849 goto restart_2;
850 }
851 dst_hold(&rt->dst);
852 read_unlock_bh(&table->tb6_lock);
853out2:
854 rt->dst.lastuse = jiffies;
855 rt->dst.__use++;
856
857 return rt;
858}
859
860static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
861 struct flowi6 *fl6, int flags)
862{
863 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
864}
865
866void ip6_route_input(struct sk_buff *skb)
867{
868 const struct ipv6hdr *iph = ipv6_hdr(skb);
869 struct net *net = dev_net(skb->dev);
870 int flags = RT6_LOOKUP_F_HAS_SADDR;
871 struct flowi6 fl6 = {
872 .flowi6_iif = skb->dev->ifindex,
873 .daddr = iph->daddr,
874 .saddr = iph->saddr,
875 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
876 .flowi6_mark = skb->mark,
877 .flowi6_proto = iph->nexthdr,
878 };
879
880 if (rt6_need_strict(&iph->daddr) && skb->dev->type != ARPHRD_PIMREG)
881 flags |= RT6_LOOKUP_F_IFACE;
882
883 skb_dst_set(skb, fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_input));
884}
885
886static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
887 struct flowi6 *fl6, int flags)
888{
889 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
890}
891
892struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
893 struct flowi6 *fl6)
894{
895 int flags = 0;
896
897 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
898 flags |= RT6_LOOKUP_F_IFACE;
899
900 if (!ipv6_addr_any(&fl6->saddr))
901 flags |= RT6_LOOKUP_F_HAS_SADDR;
902 else if (sk)
903 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
904
905 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
906}
907
908EXPORT_SYMBOL(ip6_route_output);
909
910struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
911{
912 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
913 struct dst_entry *new = NULL;
914
915 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, 0, 0);
916 if (rt) {
917 memset(&rt->rt6i_table, 0, sizeof(*rt) - sizeof(struct dst_entry));
918
919 new = &rt->dst;
920
921 new->__use = 1;
922 new->input = dst_discard;
923 new->output = dst_discard;
924
925 if (dst_metrics_read_only(&ort->dst))
926 new->_metrics = ort->dst._metrics;
927 else
928 dst_copy_metrics(new, &ort->dst);
929 rt->rt6i_idev = ort->rt6i_idev;
930 if (rt->rt6i_idev)
931 in6_dev_hold(rt->rt6i_idev);
932 rt->rt6i_expires = 0;
933
934 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
935 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
936 rt->rt6i_metric = 0;
937
938 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
939#ifdef CONFIG_IPV6_SUBTREES
940 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
941#endif
942
943 dst_free(new);
944 }
945
946 dst_release(dst_orig);
947 return new ? new : ERR_PTR(-ENOMEM);
948}
949
950/*
951 * Destination cache support functions
952 */
953
954static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
955{
956 struct rt6_info *rt;
957
958 rt = (struct rt6_info *) dst;
959
960 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie)) {
961 if (rt->rt6i_peer_genid != rt6_peer_genid()) {
962 if (!rt->rt6i_peer)
963 rt6_bind_peer(rt, 0);
964 rt->rt6i_peer_genid = rt6_peer_genid();
965 }
966 return dst;
967 }
968 return NULL;
969}
970
971static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
972{
973 struct rt6_info *rt = (struct rt6_info *) dst;
974
975 if (rt) {
976 if (rt->rt6i_flags & RTF_CACHE) {
977 if (rt6_check_expired(rt)) {
978 ip6_del_rt(rt);
979 dst = NULL;
980 }
981 } else {
982 dst_release(dst);
983 dst = NULL;
984 }
985 }
986 return dst;
987}
988
989static void ip6_link_failure(struct sk_buff *skb)
990{
991 struct rt6_info *rt;
992
993 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
994
995 rt = (struct rt6_info *) skb_dst(skb);
996 if (rt) {
997 if (rt->rt6i_flags&RTF_CACHE) {
998 dst_set_expires(&rt->dst, 0);
999 rt->rt6i_flags |= RTF_EXPIRES;
1000 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1001 rt->rt6i_node->fn_sernum = -1;
1002 }
1003}
1004
1005static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
1006{
1007 struct rt6_info *rt6 = (struct rt6_info*)dst;
1008
1009 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1010 rt6->rt6i_flags |= RTF_MODIFIED;
1011 if (mtu < IPV6_MIN_MTU) {
1012 u32 features = dst_metric(dst, RTAX_FEATURES);
1013 mtu = IPV6_MIN_MTU;
1014 features |= RTAX_FEATURE_ALLFRAG;
1015 dst_metric_set(dst, RTAX_FEATURES, features);
1016 }
1017 dst_metric_set(dst, RTAX_MTU, mtu);
1018 }
1019}
1020
1021static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1022{
1023 struct net_device *dev = dst->dev;
1024 unsigned int mtu = dst_mtu(dst);
1025 struct net *net = dev_net(dev);
1026
1027 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1028
1029 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1030 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1031
1032 /*
1033 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1034 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1035 * IPV6_MAXPLEN is also valid and means: "any MSS,
1036 * rely only on pmtu discovery"
1037 */
1038 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1039 mtu = IPV6_MAXPLEN;
1040 return mtu;
1041}
1042
1043static unsigned int ip6_default_mtu(const struct dst_entry *dst)
1044{
1045 unsigned int mtu = IPV6_MIN_MTU;
1046 struct inet6_dev *idev;
1047
1048 rcu_read_lock();
1049 idev = __in6_dev_get(dst->dev);
1050 if (idev)
1051 mtu = idev->cnf.mtu6;
1052 rcu_read_unlock();
1053
1054 return mtu;
1055}
1056
1057static struct dst_entry *icmp6_dst_gc_list;
1058static DEFINE_SPINLOCK(icmp6_dst_lock);
1059
1060struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1061 struct neighbour *neigh,
1062 const struct in6_addr *addr)
1063{
1064 struct rt6_info *rt;
1065 struct inet6_dev *idev = in6_dev_get(dev);
1066 struct net *net = dev_net(dev);
1067
1068 if (unlikely(idev == NULL))
1069 return NULL;
1070
1071 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops, dev, 0);
1072 if (unlikely(rt == NULL)) {
1073 in6_dev_put(idev);
1074 goto out;
1075 }
1076
1077 if (neigh)
1078 neigh_hold(neigh);
1079 else {
1080 neigh = ndisc_get_neigh(dev, addr);
1081 if (IS_ERR(neigh))
1082 neigh = NULL;
1083 }
1084
1085 rt->dst.flags |= DST_HOST;
1086 rt->dst.output = ip6_output;
1087 dst_set_neighbour(&rt->dst, neigh);
1088 atomic_set(&rt->dst.__refcnt, 1);
1089 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 255);
1090
1091 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1092 rt->rt6i_dst.plen = 128;
1093 rt->rt6i_idev = idev;
1094
1095 spin_lock_bh(&icmp6_dst_lock);
1096 rt->dst.next = icmp6_dst_gc_list;
1097 icmp6_dst_gc_list = &rt->dst;
1098 spin_unlock_bh(&icmp6_dst_lock);
1099
1100 fib6_force_start_gc(net);
1101
1102out:
1103 return &rt->dst;
1104}
1105
1106int icmp6_dst_gc(void)
1107{
1108 struct dst_entry *dst, **pprev;
1109 int more = 0;
1110
1111 spin_lock_bh(&icmp6_dst_lock);
1112 pprev = &icmp6_dst_gc_list;
1113
1114 while ((dst = *pprev) != NULL) {
1115 if (!atomic_read(&dst->__refcnt)) {
1116 *pprev = dst->next;
1117 dst_free(dst);
1118 } else {
1119 pprev = &dst->next;
1120 ++more;
1121 }
1122 }
1123
1124 spin_unlock_bh(&icmp6_dst_lock);
1125
1126 return more;
1127}
1128
1129static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1130 void *arg)
1131{
1132 struct dst_entry *dst, **pprev;
1133
1134 spin_lock_bh(&icmp6_dst_lock);
1135 pprev = &icmp6_dst_gc_list;
1136 while ((dst = *pprev) != NULL) {
1137 struct rt6_info *rt = (struct rt6_info *) dst;
1138 if (func(rt, arg)) {
1139 *pprev = dst->next;
1140 dst_free(dst);
1141 } else {
1142 pprev = &dst->next;
1143 }
1144 }
1145 spin_unlock_bh(&icmp6_dst_lock);
1146}
1147
1148static int ip6_dst_gc(struct dst_ops *ops)
1149{
1150 unsigned long now = jiffies;
1151 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1152 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1153 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1154 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1155 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1156 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1157 int entries;
1158
1159 entries = dst_entries_get_fast(ops);
1160 if (time_after(rt_last_gc + rt_min_interval, now) &&
1161 entries <= rt_max_size)
1162 goto out;
1163
1164 net->ipv6.ip6_rt_gc_expire++;
1165 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1166 net->ipv6.ip6_rt_last_gc = now;
1167 entries = dst_entries_get_slow(ops);
1168 if (entries < ops->gc_thresh)
1169 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1170out:
1171 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1172 return entries > rt_max_size;
1173}
1174
1175/* Clean host part of a prefix. Not necessary in radix tree,
1176 but results in cleaner routing tables.
1177
1178 Remove it only when all the things will work!
1179 */
1180
1181int ip6_dst_hoplimit(struct dst_entry *dst)
1182{
1183 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1184 if (hoplimit == 0) {
1185 struct net_device *dev = dst->dev;
1186 struct inet6_dev *idev;
1187
1188 rcu_read_lock();
1189 idev = __in6_dev_get(dev);
1190 if (idev)
1191 hoplimit = idev->cnf.hop_limit;
1192 else
1193 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1194 rcu_read_unlock();
1195 }
1196 return hoplimit;
1197}
1198EXPORT_SYMBOL(ip6_dst_hoplimit);
1199
1200/*
1201 *
1202 */
1203
1204int ip6_route_add(struct fib6_config *cfg)
1205{
1206 int err;
1207 struct net *net = cfg->fc_nlinfo.nl_net;
1208 struct rt6_info *rt = NULL;
1209 struct net_device *dev = NULL;
1210 struct inet6_dev *idev = NULL;
1211 struct fib6_table *table;
1212 int addr_type;
1213
1214 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1215 return -EINVAL;
1216#ifndef CONFIG_IPV6_SUBTREES
1217 if (cfg->fc_src_len)
1218 return -EINVAL;
1219#endif
1220 if (cfg->fc_ifindex) {
1221 err = -ENODEV;
1222 dev = dev_get_by_index(net, cfg->fc_ifindex);
1223 if (!dev)
1224 goto out;
1225 idev = in6_dev_get(dev);
1226 if (!idev)
1227 goto out;
1228 }
1229
1230 if (cfg->fc_metric == 0)
1231 cfg->fc_metric = IP6_RT_PRIO_USER;
1232
1233 table = fib6_new_table(net, cfg->fc_table);
1234 if (table == NULL) {
1235 err = -ENOBUFS;
1236 goto out;
1237 }
1238
1239 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops, NULL, DST_NOCOUNT);
1240
1241 if (rt == NULL) {
1242 err = -ENOMEM;
1243 goto out;
1244 }
1245
1246 rt->dst.obsolete = -1;
1247 rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1248 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1249 0;
1250
1251 if (cfg->fc_protocol == RTPROT_UNSPEC)
1252 cfg->fc_protocol = RTPROT_BOOT;
1253 rt->rt6i_protocol = cfg->fc_protocol;
1254
1255 addr_type = ipv6_addr_type(&cfg->fc_dst);
1256
1257 if (addr_type & IPV6_ADDR_MULTICAST)
1258 rt->dst.input = ip6_mc_input;
1259 else if (cfg->fc_flags & RTF_LOCAL)
1260 rt->dst.input = ip6_input;
1261 else
1262 rt->dst.input = ip6_forward;
1263
1264 rt->dst.output = ip6_output;
1265
1266 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1267 rt->rt6i_dst.plen = cfg->fc_dst_len;
1268 if (rt->rt6i_dst.plen == 128)
1269 rt->dst.flags |= DST_HOST;
1270
1271 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1272 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1273 if (!metrics) {
1274 err = -ENOMEM;
1275 goto out;
1276 }
1277 dst_init_metrics(&rt->dst, metrics, 0);
1278 }
1279#ifdef CONFIG_IPV6_SUBTREES
1280 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1281 rt->rt6i_src.plen = cfg->fc_src_len;
1282#endif
1283
1284 rt->rt6i_metric = cfg->fc_metric;
1285
1286 /* We cannot add true routes via loopback here,
1287 they would result in kernel looping; promote them to reject routes
1288 */
1289 if ((cfg->fc_flags & RTF_REJECT) ||
1290 (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK)
1291 && !(cfg->fc_flags&RTF_LOCAL))) {
1292 /* hold loopback dev/idev if we haven't done so. */
1293 if (dev != net->loopback_dev) {
1294 if (dev) {
1295 dev_put(dev);
1296 in6_dev_put(idev);
1297 }
1298 dev = net->loopback_dev;
1299 dev_hold(dev);
1300 idev = in6_dev_get(dev);
1301 if (!idev) {
1302 err = -ENODEV;
1303 goto out;
1304 }
1305 }
1306 rt->dst.output = ip6_pkt_discard_out;
1307 rt->dst.input = ip6_pkt_discard;
1308 rt->dst.error = -ENETUNREACH;
1309 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1310 goto install_route;
1311 }
1312
1313 if (cfg->fc_flags & RTF_GATEWAY) {
1314 const struct in6_addr *gw_addr;
1315 int gwa_type;
1316
1317 gw_addr = &cfg->fc_gateway;
1318 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1319 gwa_type = ipv6_addr_type(gw_addr);
1320
1321 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1322 struct rt6_info *grt;
1323
1324 /* IPv6 strictly inhibits using not link-local
1325 addresses as nexthop address.
1326 Otherwise, router will not able to send redirects.
1327 It is very good, but in some (rare!) circumstances
1328 (SIT, PtP, NBMA NOARP links) it is handy to allow
1329 some exceptions. --ANK
1330 */
1331 err = -EINVAL;
1332 if (!(gwa_type&IPV6_ADDR_UNICAST))
1333 goto out;
1334
1335 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1336
1337 err = -EHOSTUNREACH;
1338 if (grt == NULL)
1339 goto out;
1340 if (dev) {
1341 if (dev != grt->rt6i_dev) {
1342 dst_release(&grt->dst);
1343 goto out;
1344 }
1345 } else {
1346 dev = grt->rt6i_dev;
1347 idev = grt->rt6i_idev;
1348 dev_hold(dev);
1349 in6_dev_hold(grt->rt6i_idev);
1350 }
1351 if (!(grt->rt6i_flags&RTF_GATEWAY))
1352 err = 0;
1353 dst_release(&grt->dst);
1354
1355 if (err)
1356 goto out;
1357 }
1358 err = -EINVAL;
1359 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1360 goto out;
1361 }
1362
1363 err = -ENODEV;
1364 if (dev == NULL)
1365 goto out;
1366
1367 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1368 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1369 err = -EINVAL;
1370 goto out;
1371 }
1372 ipv6_addr_copy(&rt->rt6i_prefsrc.addr, &cfg->fc_prefsrc);
1373 rt->rt6i_prefsrc.plen = 128;
1374 } else
1375 rt->rt6i_prefsrc.plen = 0;
1376
1377 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1378 struct neighbour *n = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1379 if (IS_ERR(n)) {
1380 err = PTR_ERR(n);
1381 goto out;
1382 }
1383 dst_set_neighbour(&rt->dst, n);
1384 }
1385
1386 rt->rt6i_flags = cfg->fc_flags;
1387
1388install_route:
1389 if (cfg->fc_mx) {
1390 struct nlattr *nla;
1391 int remaining;
1392
1393 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1394 int type = nla_type(nla);
1395
1396 if (type) {
1397 if (type > RTAX_MAX) {
1398 err = -EINVAL;
1399 goto out;
1400 }
1401
1402 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1403 }
1404 }
1405 }
1406
1407 rt->dst.dev = dev;
1408 rt->rt6i_idev = idev;
1409 rt->rt6i_table = table;
1410
1411 cfg->fc_nlinfo.nl_net = dev_net(dev);
1412
1413 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1414
1415out:
1416 if (dev)
1417 dev_put(dev);
1418 if (idev)
1419 in6_dev_put(idev);
1420 if (rt)
1421 dst_free(&rt->dst);
1422 return err;
1423}
1424
1425static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1426{
1427 int err;
1428 struct fib6_table *table;
1429 struct net *net = dev_net(rt->rt6i_dev);
1430
1431 if (rt == net->ipv6.ip6_null_entry)
1432 return -ENOENT;
1433
1434 table = rt->rt6i_table;
1435 write_lock_bh(&table->tb6_lock);
1436
1437 err = fib6_del(rt, info);
1438 dst_release(&rt->dst);
1439
1440 write_unlock_bh(&table->tb6_lock);
1441
1442 return err;
1443}
1444
1445int ip6_del_rt(struct rt6_info *rt)
1446{
1447 struct nl_info info = {
1448 .nl_net = dev_net(rt->rt6i_dev),
1449 };
1450 return __ip6_del_rt(rt, &info);
1451}
1452
1453static int ip6_route_del(struct fib6_config *cfg)
1454{
1455 struct fib6_table *table;
1456 struct fib6_node *fn;
1457 struct rt6_info *rt;
1458 int err = -ESRCH;
1459
1460 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1461 if (table == NULL)
1462 return err;
1463
1464 read_lock_bh(&table->tb6_lock);
1465
1466 fn = fib6_locate(&table->tb6_root,
1467 &cfg->fc_dst, cfg->fc_dst_len,
1468 &cfg->fc_src, cfg->fc_src_len);
1469
1470 if (fn) {
1471 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1472 if (cfg->fc_ifindex &&
1473 (rt->rt6i_dev == NULL ||
1474 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1475 continue;
1476 if (cfg->fc_flags & RTF_GATEWAY &&
1477 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1478 continue;
1479 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1480 continue;
1481 dst_hold(&rt->dst);
1482 read_unlock_bh(&table->tb6_lock);
1483
1484 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1485 }
1486 }
1487 read_unlock_bh(&table->tb6_lock);
1488
1489 return err;
1490}
1491
1492/*
1493 * Handle redirects
1494 */
1495struct ip6rd_flowi {
1496 struct flowi6 fl6;
1497 struct in6_addr gateway;
1498};
1499
1500static struct rt6_info *__ip6_route_redirect(struct net *net,
1501 struct fib6_table *table,
1502 struct flowi6 *fl6,
1503 int flags)
1504{
1505 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1506 struct rt6_info *rt;
1507 struct fib6_node *fn;
1508
1509 /*
1510 * Get the "current" route for this destination and
1511 * check if the redirect has come from approriate router.
1512 *
1513 * RFC 2461 specifies that redirects should only be
1514 * accepted if they come from the nexthop to the target.
1515 * Due to the way the routes are chosen, this notion
1516 * is a bit fuzzy and one might need to check all possible
1517 * routes.
1518 */
1519
1520 read_lock_bh(&table->tb6_lock);
1521 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1522restart:
1523 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1524 /*
1525 * Current route is on-link; redirect is always invalid.
1526 *
1527 * Seems, previous statement is not true. It could
1528 * be node, which looks for us as on-link (f.e. proxy ndisc)
1529 * But then router serving it might decide, that we should
1530 * know truth 8)8) --ANK (980726).
1531 */
1532 if (rt6_check_expired(rt))
1533 continue;
1534 if (!(rt->rt6i_flags & RTF_GATEWAY))
1535 continue;
1536 if (fl6->flowi6_oif != rt->rt6i_dev->ifindex)
1537 continue;
1538 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1539 continue;
1540 break;
1541 }
1542
1543 if (!rt)
1544 rt = net->ipv6.ip6_null_entry;
1545 BACKTRACK(net, &fl6->saddr);
1546out:
1547 dst_hold(&rt->dst);
1548
1549 read_unlock_bh(&table->tb6_lock);
1550
1551 return rt;
1552};
1553
1554static struct rt6_info *ip6_route_redirect(const struct in6_addr *dest,
1555 const struct in6_addr *src,
1556 const struct in6_addr *gateway,
1557 struct net_device *dev)
1558{
1559 int flags = RT6_LOOKUP_F_HAS_SADDR;
1560 struct net *net = dev_net(dev);
1561 struct ip6rd_flowi rdfl = {
1562 .fl6 = {
1563 .flowi6_oif = dev->ifindex,
1564 .daddr = *dest,
1565 .saddr = *src,
1566 },
1567 };
1568
1569 ipv6_addr_copy(&rdfl.gateway, gateway);
1570
1571 if (rt6_need_strict(dest))
1572 flags |= RT6_LOOKUP_F_IFACE;
1573
1574 return (struct rt6_info *)fib6_rule_lookup(net, &rdfl.fl6,
1575 flags, __ip6_route_redirect);
1576}
1577
1578void rt6_redirect(const struct in6_addr *dest, const struct in6_addr *src,
1579 const struct in6_addr *saddr,
1580 struct neighbour *neigh, u8 *lladdr, int on_link)
1581{
1582 struct rt6_info *rt, *nrt = NULL;
1583 struct netevent_redirect netevent;
1584 struct net *net = dev_net(neigh->dev);
1585
1586 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1587
1588 if (rt == net->ipv6.ip6_null_entry) {
1589 if (net_ratelimit())
1590 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1591 "for redirect target\n");
1592 goto out;
1593 }
1594
1595 /*
1596 * We have finally decided to accept it.
1597 */
1598
1599 neigh_update(neigh, lladdr, NUD_STALE,
1600 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1601 NEIGH_UPDATE_F_OVERRIDE|
1602 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1603 NEIGH_UPDATE_F_ISROUTER))
1604 );
1605
1606 /*
1607 * Redirect received -> path was valid.
1608 * Look, redirects are sent only in response to data packets,
1609 * so that this nexthop apparently is reachable. --ANK
1610 */
1611 dst_confirm(&rt->dst);
1612
1613 /* Duplicate redirect: silently ignore. */
1614 if (neigh == dst_get_neighbour_raw(&rt->dst))
1615 goto out;
1616
1617 nrt = ip6_rt_copy(rt, dest);
1618 if (nrt == NULL)
1619 goto out;
1620
1621 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1622 if (on_link)
1623 nrt->rt6i_flags &= ~RTF_GATEWAY;
1624
1625 ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1626 dst_set_neighbour(&nrt->dst, neigh_clone(neigh));
1627
1628 if (ip6_ins_rt(nrt))
1629 goto out;
1630
1631 netevent.old = &rt->dst;
1632 netevent.new = &nrt->dst;
1633 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1634
1635 if (rt->rt6i_flags&RTF_CACHE) {
1636 ip6_del_rt(rt);
1637 return;
1638 }
1639
1640out:
1641 dst_release(&rt->dst);
1642}
1643
1644/*
1645 * Handle ICMP "packet too big" messages
1646 * i.e. Path MTU discovery
1647 */
1648
1649static void rt6_do_pmtu_disc(const struct in6_addr *daddr, const struct in6_addr *saddr,
1650 struct net *net, u32 pmtu, int ifindex)
1651{
1652 struct rt6_info *rt, *nrt;
1653 int allfrag = 0;
1654again:
1655 rt = rt6_lookup(net, daddr, saddr, ifindex, 0);
1656 if (rt == NULL)
1657 return;
1658
1659 if (rt6_check_expired(rt)) {
1660 ip6_del_rt(rt);
1661 goto again;
1662 }
1663
1664 if (pmtu >= dst_mtu(&rt->dst))
1665 goto out;
1666
1667 if (pmtu < IPV6_MIN_MTU) {
1668 /*
1669 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1670 * MTU (1280) and a fragment header should always be included
1671 * after a node receiving Too Big message reporting PMTU is
1672 * less than the IPv6 Minimum Link MTU.
1673 */
1674 pmtu = IPV6_MIN_MTU;
1675 allfrag = 1;
1676 }
1677
1678 /* New mtu received -> path was valid.
1679 They are sent only in response to data packets,
1680 so that this nexthop apparently is reachable. --ANK
1681 */
1682 dst_confirm(&rt->dst);
1683
1684 /* Host route. If it is static, it would be better
1685 not to override it, but add new one, so that
1686 when cache entry will expire old pmtu
1687 would return automatically.
1688 */
1689 if (rt->rt6i_flags & RTF_CACHE) {
1690 dst_metric_set(&rt->dst, RTAX_MTU, pmtu);
1691 if (allfrag) {
1692 u32 features = dst_metric(&rt->dst, RTAX_FEATURES);
1693 features |= RTAX_FEATURE_ALLFRAG;
1694 dst_metric_set(&rt->dst, RTAX_FEATURES, features);
1695 }
1696 dst_set_expires(&rt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1697 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1698 goto out;
1699 }
1700
1701 /* Network route.
1702 Two cases are possible:
1703 1. It is connected route. Action: COW
1704 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1705 */
1706 if (!dst_get_neighbour_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
1707 nrt = rt6_alloc_cow(rt, daddr, saddr);
1708 else
1709 nrt = rt6_alloc_clone(rt, daddr);
1710
1711 if (nrt) {
1712 dst_metric_set(&nrt->dst, RTAX_MTU, pmtu);
1713 if (allfrag) {
1714 u32 features = dst_metric(&nrt->dst, RTAX_FEATURES);
1715 features |= RTAX_FEATURE_ALLFRAG;
1716 dst_metric_set(&nrt->dst, RTAX_FEATURES, features);
1717 }
1718
1719 /* According to RFC 1981, detecting PMTU increase shouldn't be
1720 * happened within 5 mins, the recommended timer is 10 mins.
1721 * Here this route expiration time is set to ip6_rt_mtu_expires
1722 * which is 10 mins. After 10 mins the decreased pmtu is expired
1723 * and detecting PMTU increase will be automatically happened.
1724 */
1725 dst_set_expires(&nrt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1726 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1727
1728 ip6_ins_rt(nrt);
1729 }
1730out:
1731 dst_release(&rt->dst);
1732}
1733
1734void rt6_pmtu_discovery(const struct in6_addr *daddr, const struct in6_addr *saddr,
1735 struct net_device *dev, u32 pmtu)
1736{
1737 struct net *net = dev_net(dev);
1738
1739 /*
1740 * RFC 1981 states that a node "MUST reduce the size of the packets it
1741 * is sending along the path" that caused the Packet Too Big message.
1742 * Since it's not possible in the general case to determine which
1743 * interface was used to send the original packet, we update the MTU
1744 * on the interface that will be used to send future packets. We also
1745 * update the MTU on the interface that received the Packet Too Big in
1746 * case the original packet was forced out that interface with
1747 * SO_BINDTODEVICE or similar. This is the next best thing to the
1748 * correct behaviour, which would be to update the MTU on all
1749 * interfaces.
1750 */
1751 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, 0);
1752 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, dev->ifindex);
1753}
1754
1755/*
1756 * Misc support functions
1757 */
1758
1759static struct rt6_info *ip6_rt_copy(const struct rt6_info *ort,
1760 const struct in6_addr *dest)
1761{
1762 struct net *net = dev_net(ort->rt6i_dev);
1763 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops,
1764 ort->dst.dev, 0);
1765
1766 if (rt) {
1767 rt->dst.input = ort->dst.input;
1768 rt->dst.output = ort->dst.output;
1769 rt->dst.flags |= DST_HOST;
1770
1771 ipv6_addr_copy(&rt->rt6i_dst.addr, dest);
1772 rt->rt6i_dst.plen = 128;
1773 dst_copy_metrics(&rt->dst, &ort->dst);
1774 rt->dst.error = ort->dst.error;
1775 rt->rt6i_idev = ort->rt6i_idev;
1776 if (rt->rt6i_idev)
1777 in6_dev_hold(rt->rt6i_idev);
1778 rt->dst.lastuse = jiffies;
1779 rt->rt6i_expires = 0;
1780
1781 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1782 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1783 rt->rt6i_metric = 0;
1784
1785#ifdef CONFIG_IPV6_SUBTREES
1786 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1787#endif
1788 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1789 rt->rt6i_table = ort->rt6i_table;
1790 }
1791 return rt;
1792}
1793
1794#ifdef CONFIG_IPV6_ROUTE_INFO
1795static struct rt6_info *rt6_get_route_info(struct net *net,
1796 const struct in6_addr *prefix, int prefixlen,
1797 const struct in6_addr *gwaddr, int ifindex)
1798{
1799 struct fib6_node *fn;
1800 struct rt6_info *rt = NULL;
1801 struct fib6_table *table;
1802
1803 table = fib6_get_table(net, RT6_TABLE_INFO);
1804 if (table == NULL)
1805 return NULL;
1806
1807 write_lock_bh(&table->tb6_lock);
1808 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1809 if (!fn)
1810 goto out;
1811
1812 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1813 if (rt->rt6i_dev->ifindex != ifindex)
1814 continue;
1815 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1816 continue;
1817 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1818 continue;
1819 dst_hold(&rt->dst);
1820 break;
1821 }
1822out:
1823 write_unlock_bh(&table->tb6_lock);
1824 return rt;
1825}
1826
1827static struct rt6_info *rt6_add_route_info(struct net *net,
1828 const struct in6_addr *prefix, int prefixlen,
1829 const struct in6_addr *gwaddr, int ifindex,
1830 unsigned pref)
1831{
1832 struct fib6_config cfg = {
1833 .fc_table = RT6_TABLE_INFO,
1834 .fc_metric = IP6_RT_PRIO_USER,
1835 .fc_ifindex = ifindex,
1836 .fc_dst_len = prefixlen,
1837 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1838 RTF_UP | RTF_PREF(pref),
1839 .fc_nlinfo.pid = 0,
1840 .fc_nlinfo.nlh = NULL,
1841 .fc_nlinfo.nl_net = net,
1842 };
1843
1844 ipv6_addr_copy(&cfg.fc_dst, prefix);
1845 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1846
1847 /* We should treat it as a default route if prefix length is 0. */
1848 if (!prefixlen)
1849 cfg.fc_flags |= RTF_DEFAULT;
1850
1851 ip6_route_add(&cfg);
1852
1853 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1854}
1855#endif
1856
1857struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1858{
1859 struct rt6_info *rt;
1860 struct fib6_table *table;
1861
1862 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1863 if (table == NULL)
1864 return NULL;
1865
1866 write_lock_bh(&table->tb6_lock);
1867 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1868 if (dev == rt->rt6i_dev &&
1869 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1870 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1871 break;
1872 }
1873 if (rt)
1874 dst_hold(&rt->dst);
1875 write_unlock_bh(&table->tb6_lock);
1876 return rt;
1877}
1878
1879struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1880 struct net_device *dev,
1881 unsigned int pref)
1882{
1883 struct fib6_config cfg = {
1884 .fc_table = RT6_TABLE_DFLT,
1885 .fc_metric = IP6_RT_PRIO_USER,
1886 .fc_ifindex = dev->ifindex,
1887 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1888 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1889 .fc_nlinfo.pid = 0,
1890 .fc_nlinfo.nlh = NULL,
1891 .fc_nlinfo.nl_net = dev_net(dev),
1892 };
1893
1894 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1895
1896 ip6_route_add(&cfg);
1897
1898 return rt6_get_dflt_router(gwaddr, dev);
1899}
1900
1901void rt6_purge_dflt_routers(struct net *net)
1902{
1903 struct rt6_info *rt;
1904 struct fib6_table *table;
1905
1906 /* NOTE: Keep consistent with rt6_get_dflt_router */
1907 table = fib6_get_table(net, RT6_TABLE_DFLT);
1908 if (table == NULL)
1909 return;
1910
1911restart:
1912 read_lock_bh(&table->tb6_lock);
1913 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1914 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1915 dst_hold(&rt->dst);
1916 read_unlock_bh(&table->tb6_lock);
1917 ip6_del_rt(rt);
1918 goto restart;
1919 }
1920 }
1921 read_unlock_bh(&table->tb6_lock);
1922}
1923
1924static void rtmsg_to_fib6_config(struct net *net,
1925 struct in6_rtmsg *rtmsg,
1926 struct fib6_config *cfg)
1927{
1928 memset(cfg, 0, sizeof(*cfg));
1929
1930 cfg->fc_table = RT6_TABLE_MAIN;
1931 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1932 cfg->fc_metric = rtmsg->rtmsg_metric;
1933 cfg->fc_expires = rtmsg->rtmsg_info;
1934 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1935 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1936 cfg->fc_flags = rtmsg->rtmsg_flags;
1937
1938 cfg->fc_nlinfo.nl_net = net;
1939
1940 ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1941 ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1942 ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1943}
1944
1945int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1946{
1947 struct fib6_config cfg;
1948 struct in6_rtmsg rtmsg;
1949 int err;
1950
1951 switch(cmd) {
1952 case SIOCADDRT: /* Add a route */
1953 case SIOCDELRT: /* Delete a route */
1954 if (!capable(CAP_NET_ADMIN))
1955 return -EPERM;
1956 err = copy_from_user(&rtmsg, arg,
1957 sizeof(struct in6_rtmsg));
1958 if (err)
1959 return -EFAULT;
1960
1961 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1962
1963 rtnl_lock();
1964 switch (cmd) {
1965 case SIOCADDRT:
1966 err = ip6_route_add(&cfg);
1967 break;
1968 case SIOCDELRT:
1969 err = ip6_route_del(&cfg);
1970 break;
1971 default:
1972 err = -EINVAL;
1973 }
1974 rtnl_unlock();
1975
1976 return err;
1977 }
1978
1979 return -EINVAL;
1980}
1981
1982/*
1983 * Drop the packet on the floor
1984 */
1985
1986static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1987{
1988 int type;
1989 struct dst_entry *dst = skb_dst(skb);
1990 switch (ipstats_mib_noroutes) {
1991 case IPSTATS_MIB_INNOROUTES:
1992 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1993 if (type == IPV6_ADDR_ANY) {
1994 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1995 IPSTATS_MIB_INADDRERRORS);
1996 break;
1997 }
1998 /* FALLTHROUGH */
1999 case IPSTATS_MIB_OUTNOROUTES:
2000 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2001 ipstats_mib_noroutes);
2002 break;
2003 }
2004 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2005 kfree_skb(skb);
2006 return 0;
2007}
2008
2009static int ip6_pkt_discard(struct sk_buff *skb)
2010{
2011 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2012}
2013
2014static int ip6_pkt_discard_out(struct sk_buff *skb)
2015{
2016 skb->dev = skb_dst(skb)->dev;
2017 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2018}
2019
2020#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2021
2022static int ip6_pkt_prohibit(struct sk_buff *skb)
2023{
2024 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2025}
2026
2027static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2028{
2029 skb->dev = skb_dst(skb)->dev;
2030 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2031}
2032
2033#endif
2034
2035/*
2036 * Allocate a dst for local (unicast / anycast) address.
2037 */
2038
2039struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2040 const struct in6_addr *addr,
2041 int anycast)
2042{
2043 struct net *net = dev_net(idev->dev);
2044 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops,
2045 net->loopback_dev, 0);
2046 struct neighbour *neigh;
2047
2048 if (rt == NULL) {
2049 if (net_ratelimit())
2050 pr_warning("IPv6: Maximum number of routes reached,"
2051 " consider increasing route/max_size.\n");
2052 return ERR_PTR(-ENOMEM);
2053 }
2054
2055 in6_dev_hold(idev);
2056
2057 rt->dst.flags |= DST_HOST;
2058 rt->dst.input = ip6_input;
2059 rt->dst.output = ip6_output;
2060 rt->rt6i_idev = idev;
2061 rt->dst.obsolete = -1;
2062
2063 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2064 if (anycast)
2065 rt->rt6i_flags |= RTF_ANYCAST;
2066 else
2067 rt->rt6i_flags |= RTF_LOCAL;
2068 neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
2069 if (IS_ERR(neigh)) {
2070 dst_free(&rt->dst);
2071
2072 return ERR_CAST(neigh);
2073 }
2074 dst_set_neighbour(&rt->dst, neigh);
2075
2076 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
2077 rt->rt6i_dst.plen = 128;
2078 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2079
2080 atomic_set(&rt->dst.__refcnt, 1);
2081
2082 return rt;
2083}
2084
2085int ip6_route_get_saddr(struct net *net,
2086 struct rt6_info *rt,
2087 const struct in6_addr *daddr,
2088 unsigned int prefs,
2089 struct in6_addr *saddr)
2090{
2091 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2092 int err = 0;
2093 if (rt->rt6i_prefsrc.plen)
2094 ipv6_addr_copy(saddr, &rt->rt6i_prefsrc.addr);
2095 else
2096 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2097 daddr, prefs, saddr);
2098 return err;
2099}
2100
2101/* remove deleted ip from prefsrc entries */
2102struct arg_dev_net_ip {
2103 struct net_device *dev;
2104 struct net *net;
2105 struct in6_addr *addr;
2106};
2107
2108static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2109{
2110 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2111 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2112 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2113
2114 if (((void *)rt->rt6i_dev == dev || dev == NULL) &&
2115 rt != net->ipv6.ip6_null_entry &&
2116 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2117 /* remove prefsrc entry */
2118 rt->rt6i_prefsrc.plen = 0;
2119 }
2120 return 0;
2121}
2122
2123void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2124{
2125 struct net *net = dev_net(ifp->idev->dev);
2126 struct arg_dev_net_ip adni = {
2127 .dev = ifp->idev->dev,
2128 .net = net,
2129 .addr = &ifp->addr,
2130 };
2131 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2132}
2133
2134struct arg_dev_net {
2135 struct net_device *dev;
2136 struct net *net;
2137};
2138
2139static int fib6_ifdown(struct rt6_info *rt, void *arg)
2140{
2141 const struct arg_dev_net *adn = arg;
2142 const struct net_device *dev = adn->dev;
2143
2144 if ((rt->rt6i_dev == dev || dev == NULL) &&
2145 rt != adn->net->ipv6.ip6_null_entry) {
2146 RT6_TRACE("deleted by ifdown %p\n", rt);
2147 return -1;
2148 }
2149 return 0;
2150}
2151
2152void rt6_ifdown(struct net *net, struct net_device *dev)
2153{
2154 struct arg_dev_net adn = {
2155 .dev = dev,
2156 .net = net,
2157 };
2158
2159 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2160 icmp6_clean_all(fib6_ifdown, &adn);
2161}
2162
2163struct rt6_mtu_change_arg
2164{
2165 struct net_device *dev;
2166 unsigned mtu;
2167};
2168
2169static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2170{
2171 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2172 struct inet6_dev *idev;
2173
2174 /* In IPv6 pmtu discovery is not optional,
2175 so that RTAX_MTU lock cannot disable it.
2176 We still use this lock to block changes
2177 caused by addrconf/ndisc.
2178 */
2179
2180 idev = __in6_dev_get(arg->dev);
2181 if (idev == NULL)
2182 return 0;
2183
2184 /* For administrative MTU increase, there is no way to discover
2185 IPv6 PMTU increase, so PMTU increase should be updated here.
2186 Since RFC 1981 doesn't include administrative MTU increase
2187 update PMTU increase is a MUST. (i.e. jumbo frame)
2188 */
2189 /*
2190 If new MTU is less than route PMTU, this new MTU will be the
2191 lowest MTU in the path, update the route PMTU to reflect PMTU
2192 decreases; if new MTU is greater than route PMTU, and the
2193 old MTU is the lowest MTU in the path, update the route PMTU
2194 to reflect the increase. In this case if the other nodes' MTU
2195 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2196 PMTU discouvery.
2197 */
2198 if (rt->rt6i_dev == arg->dev &&
2199 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2200 (dst_mtu(&rt->dst) >= arg->mtu ||
2201 (dst_mtu(&rt->dst) < arg->mtu &&
2202 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2203 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2204 }
2205 return 0;
2206}
2207
2208void rt6_mtu_change(struct net_device *dev, unsigned mtu)
2209{
2210 struct rt6_mtu_change_arg arg = {
2211 .dev = dev,
2212 .mtu = mtu,
2213 };
2214
2215 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2216}
2217
2218static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2219 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2220 [RTA_OIF] = { .type = NLA_U32 },
2221 [RTA_IIF] = { .type = NLA_U32 },
2222 [RTA_PRIORITY] = { .type = NLA_U32 },
2223 [RTA_METRICS] = { .type = NLA_NESTED },
2224};
2225
2226static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2227 struct fib6_config *cfg)
2228{
2229 struct rtmsg *rtm;
2230 struct nlattr *tb[RTA_MAX+1];
2231 int err;
2232
2233 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2234 if (err < 0)
2235 goto errout;
2236
2237 err = -EINVAL;
2238 rtm = nlmsg_data(nlh);
2239 memset(cfg, 0, sizeof(*cfg));
2240
2241 cfg->fc_table = rtm->rtm_table;
2242 cfg->fc_dst_len = rtm->rtm_dst_len;
2243 cfg->fc_src_len = rtm->rtm_src_len;
2244 cfg->fc_flags = RTF_UP;
2245 cfg->fc_protocol = rtm->rtm_protocol;
2246
2247 if (rtm->rtm_type == RTN_UNREACHABLE)
2248 cfg->fc_flags |= RTF_REJECT;
2249
2250 if (rtm->rtm_type == RTN_LOCAL)
2251 cfg->fc_flags |= RTF_LOCAL;
2252
2253 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2254 cfg->fc_nlinfo.nlh = nlh;
2255 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2256
2257 if (tb[RTA_GATEWAY]) {
2258 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2259 cfg->fc_flags |= RTF_GATEWAY;
2260 }
2261
2262 if (tb[RTA_DST]) {
2263 int plen = (rtm->rtm_dst_len + 7) >> 3;
2264
2265 if (nla_len(tb[RTA_DST]) < plen)
2266 goto errout;
2267
2268 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2269 }
2270
2271 if (tb[RTA_SRC]) {
2272 int plen = (rtm->rtm_src_len + 7) >> 3;
2273
2274 if (nla_len(tb[RTA_SRC]) < plen)
2275 goto errout;
2276
2277 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2278 }
2279
2280 if (tb[RTA_PREFSRC])
2281 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2282
2283 if (tb[RTA_OIF])
2284 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2285
2286 if (tb[RTA_PRIORITY])
2287 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2288
2289 if (tb[RTA_METRICS]) {
2290 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2291 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2292 }
2293
2294 if (tb[RTA_TABLE])
2295 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2296
2297 err = 0;
2298errout:
2299 return err;
2300}
2301
2302static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2303{
2304 struct fib6_config cfg;
2305 int err;
2306
2307 err = rtm_to_fib6_config(skb, nlh, &cfg);
2308 if (err < 0)
2309 return err;
2310
2311 return ip6_route_del(&cfg);
2312}
2313
2314static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2315{
2316 struct fib6_config cfg;
2317 int err;
2318
2319 err = rtm_to_fib6_config(skb, nlh, &cfg);
2320 if (err < 0)
2321 return err;
2322
2323 return ip6_route_add(&cfg);
2324}
2325
2326static inline size_t rt6_nlmsg_size(void)
2327{
2328 return NLMSG_ALIGN(sizeof(struct rtmsg))
2329 + nla_total_size(16) /* RTA_SRC */
2330 + nla_total_size(16) /* RTA_DST */
2331 + nla_total_size(16) /* RTA_GATEWAY */
2332 + nla_total_size(16) /* RTA_PREFSRC */
2333 + nla_total_size(4) /* RTA_TABLE */
2334 + nla_total_size(4) /* RTA_IIF */
2335 + nla_total_size(4) /* RTA_OIF */
2336 + nla_total_size(4) /* RTA_PRIORITY */
2337 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2338 + nla_total_size(sizeof(struct rta_cacheinfo));
2339}
2340
2341static int rt6_fill_node(struct net *net,
2342 struct sk_buff *skb, struct rt6_info *rt,
2343 struct in6_addr *dst, struct in6_addr *src,
2344 int iif, int type, u32 pid, u32 seq,
2345 int prefix, int nowait, unsigned int flags)
2346{
2347 struct rtmsg *rtm;
2348 struct nlmsghdr *nlh;
2349 long expires;
2350 u32 table;
2351 struct neighbour *n;
2352
2353 if (prefix) { /* user wants prefix routes only */
2354 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2355 /* success since this is not a prefix route */
2356 return 1;
2357 }
2358 }
2359
2360 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2361 if (nlh == NULL)
2362 return -EMSGSIZE;
2363
2364 rtm = nlmsg_data(nlh);
2365 rtm->rtm_family = AF_INET6;
2366 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2367 rtm->rtm_src_len = rt->rt6i_src.plen;
2368 rtm->rtm_tos = 0;
2369 if (rt->rt6i_table)
2370 table = rt->rt6i_table->tb6_id;
2371 else
2372 table = RT6_TABLE_UNSPEC;
2373 rtm->rtm_table = table;
2374 NLA_PUT_U32(skb, RTA_TABLE, table);
2375 if (rt->rt6i_flags&RTF_REJECT)
2376 rtm->rtm_type = RTN_UNREACHABLE;
2377 else if (rt->rt6i_flags&RTF_LOCAL)
2378 rtm->rtm_type = RTN_LOCAL;
2379 else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2380 rtm->rtm_type = RTN_LOCAL;
2381 else
2382 rtm->rtm_type = RTN_UNICAST;
2383 rtm->rtm_flags = 0;
2384 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2385 rtm->rtm_protocol = rt->rt6i_protocol;
2386 if (rt->rt6i_flags&RTF_DYNAMIC)
2387 rtm->rtm_protocol = RTPROT_REDIRECT;
2388 else if (rt->rt6i_flags & RTF_ADDRCONF)
2389 rtm->rtm_protocol = RTPROT_KERNEL;
2390 else if (rt->rt6i_flags&RTF_DEFAULT)
2391 rtm->rtm_protocol = RTPROT_RA;
2392
2393 if (rt->rt6i_flags&RTF_CACHE)
2394 rtm->rtm_flags |= RTM_F_CLONED;
2395
2396 if (dst) {
2397 NLA_PUT(skb, RTA_DST, 16, dst);
2398 rtm->rtm_dst_len = 128;
2399 } else if (rtm->rtm_dst_len)
2400 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2401#ifdef CONFIG_IPV6_SUBTREES
2402 if (src) {
2403 NLA_PUT(skb, RTA_SRC, 16, src);
2404 rtm->rtm_src_len = 128;
2405 } else if (rtm->rtm_src_len)
2406 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2407#endif
2408 if (iif) {
2409#ifdef CONFIG_IPV6_MROUTE
2410 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2411 int err = ip6mr_get_route(net, skb, rtm, nowait);
2412 if (err <= 0) {
2413 if (!nowait) {
2414 if (err == 0)
2415 return 0;
2416 goto nla_put_failure;
2417 } else {
2418 if (err == -EMSGSIZE)
2419 goto nla_put_failure;
2420 }
2421 }
2422 } else
2423#endif
2424 NLA_PUT_U32(skb, RTA_IIF, iif);
2425 } else if (dst) {
2426 struct in6_addr saddr_buf;
2427 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0)
2428 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2429 }
2430
2431 if (rt->rt6i_prefsrc.plen) {
2432 struct in6_addr saddr_buf;
2433 ipv6_addr_copy(&saddr_buf, &rt->rt6i_prefsrc.addr);
2434 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2435 }
2436
2437 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2438 goto nla_put_failure;
2439
2440 rcu_read_lock();
2441 n = dst_get_neighbour(&rt->dst);
2442 if (n)
2443 NLA_PUT(skb, RTA_GATEWAY, 16, &n->primary_key);
2444 rcu_read_unlock();
2445
2446 if (rt->dst.dev)
2447 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2448
2449 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2450
2451 if (!(rt->rt6i_flags & RTF_EXPIRES))
2452 expires = 0;
2453 else if (rt->rt6i_expires - jiffies < INT_MAX)
2454 expires = rt->rt6i_expires - jiffies;
2455 else
2456 expires = INT_MAX;
2457
2458 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, 0, 0,
2459 expires, rt->dst.error) < 0)
2460 goto nla_put_failure;
2461
2462 return nlmsg_end(skb, nlh);
2463
2464nla_put_failure:
2465 nlmsg_cancel(skb, nlh);
2466 return -EMSGSIZE;
2467}
2468
2469int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2470{
2471 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2472 int prefix;
2473
2474 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2475 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2476 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2477 } else
2478 prefix = 0;
2479
2480 return rt6_fill_node(arg->net,
2481 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2482 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2483 prefix, 0, NLM_F_MULTI);
2484}
2485
2486static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2487{
2488 struct net *net = sock_net(in_skb->sk);
2489 struct nlattr *tb[RTA_MAX+1];
2490 struct rt6_info *rt;
2491 struct sk_buff *skb;
2492 struct rtmsg *rtm;
2493 struct flowi6 fl6;
2494 int err, iif = 0;
2495
2496 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2497 if (err < 0)
2498 goto errout;
2499
2500 err = -EINVAL;
2501 memset(&fl6, 0, sizeof(fl6));
2502
2503 if (tb[RTA_SRC]) {
2504 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2505 goto errout;
2506
2507 ipv6_addr_copy(&fl6.saddr, nla_data(tb[RTA_SRC]));
2508 }
2509
2510 if (tb[RTA_DST]) {
2511 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2512 goto errout;
2513
2514 ipv6_addr_copy(&fl6.daddr, nla_data(tb[RTA_DST]));
2515 }
2516
2517 if (tb[RTA_IIF])
2518 iif = nla_get_u32(tb[RTA_IIF]);
2519
2520 if (tb[RTA_OIF])
2521 fl6.flowi6_oif = nla_get_u32(tb[RTA_OIF]);
2522
2523 if (iif) {
2524 struct net_device *dev;
2525 dev = __dev_get_by_index(net, iif);
2526 if (!dev) {
2527 err = -ENODEV;
2528 goto errout;
2529 }
2530 }
2531
2532 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2533 if (skb == NULL) {
2534 err = -ENOBUFS;
2535 goto errout;
2536 }
2537
2538 /* Reserve room for dummy headers, this skb can pass
2539 through good chunk of routing engine.
2540 */
2541 skb_reset_mac_header(skb);
2542 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2543
2544 rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl6);
2545 skb_dst_set(skb, &rt->dst);
2546
2547 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2548 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2549 nlh->nlmsg_seq, 0, 0, 0);
2550 if (err < 0) {
2551 kfree_skb(skb);
2552 goto errout;
2553 }
2554
2555 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2556errout:
2557 return err;
2558}
2559
2560void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2561{
2562 struct sk_buff *skb;
2563 struct net *net = info->nl_net;
2564 u32 seq;
2565 int err;
2566
2567 err = -ENOBUFS;
2568 seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2569
2570 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2571 if (skb == NULL)
2572 goto errout;
2573
2574 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2575 event, info->pid, seq, 0, 0, 0);
2576 if (err < 0) {
2577 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2578 WARN_ON(err == -EMSGSIZE);
2579 kfree_skb(skb);
2580 goto errout;
2581 }
2582 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2583 info->nlh, gfp_any());
2584 return;
2585errout:
2586 if (err < 0)
2587 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2588}
2589
2590static int ip6_route_dev_notify(struct notifier_block *this,
2591 unsigned long event, void *data)
2592{
2593 struct net_device *dev = (struct net_device *)data;
2594 struct net *net = dev_net(dev);
2595
2596 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2597 net->ipv6.ip6_null_entry->dst.dev = dev;
2598 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2599#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2600 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2601 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2602 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2603 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2604#endif
2605 }
2606
2607 return NOTIFY_OK;
2608}
2609
2610/*
2611 * /proc
2612 */
2613
2614#ifdef CONFIG_PROC_FS
2615
2616struct rt6_proc_arg
2617{
2618 char *buffer;
2619 int offset;
2620 int length;
2621 int skip;
2622 int len;
2623};
2624
2625static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2626{
2627 struct seq_file *m = p_arg;
2628 struct neighbour *n;
2629
2630 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2631
2632#ifdef CONFIG_IPV6_SUBTREES
2633 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2634#else
2635 seq_puts(m, "00000000000000000000000000000000 00 ");
2636#endif
2637 rcu_read_lock();
2638 n = dst_get_neighbour(&rt->dst);
2639 if (n) {
2640 seq_printf(m, "%pi6", n->primary_key);
2641 } else {
2642 seq_puts(m, "00000000000000000000000000000000");
2643 }
2644 rcu_read_unlock();
2645 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2646 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2647 rt->dst.__use, rt->rt6i_flags,
2648 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2649 return 0;
2650}
2651
2652static int ipv6_route_show(struct seq_file *m, void *v)
2653{
2654 struct net *net = (struct net *)m->private;
2655 fib6_clean_all(net, rt6_info_route, 0, m);
2656 return 0;
2657}
2658
2659static int ipv6_route_open(struct inode *inode, struct file *file)
2660{
2661 return single_open_net(inode, file, ipv6_route_show);
2662}
2663
2664static const struct file_operations ipv6_route_proc_fops = {
2665 .owner = THIS_MODULE,
2666 .open = ipv6_route_open,
2667 .read = seq_read,
2668 .llseek = seq_lseek,
2669 .release = single_release_net,
2670};
2671
2672static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2673{
2674 struct net *net = (struct net *)seq->private;
2675 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2676 net->ipv6.rt6_stats->fib_nodes,
2677 net->ipv6.rt6_stats->fib_route_nodes,
2678 net->ipv6.rt6_stats->fib_rt_alloc,
2679 net->ipv6.rt6_stats->fib_rt_entries,
2680 net->ipv6.rt6_stats->fib_rt_cache,
2681 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2682 net->ipv6.rt6_stats->fib_discarded_routes);
2683
2684 return 0;
2685}
2686
2687static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2688{
2689 return single_open_net(inode, file, rt6_stats_seq_show);
2690}
2691
2692static const struct file_operations rt6_stats_seq_fops = {
2693 .owner = THIS_MODULE,
2694 .open = rt6_stats_seq_open,
2695 .read = seq_read,
2696 .llseek = seq_lseek,
2697 .release = single_release_net,
2698};
2699#endif /* CONFIG_PROC_FS */
2700
2701#ifdef CONFIG_SYSCTL
2702
2703static
2704int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2705 void __user *buffer, size_t *lenp, loff_t *ppos)
2706{
2707 struct net *net;
2708 int delay;
2709 if (!write)
2710 return -EINVAL;
2711
2712 net = (struct net *)ctl->extra1;
2713 delay = net->ipv6.sysctl.flush_delay;
2714 proc_dointvec(ctl, write, buffer, lenp, ppos);
2715 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2716 return 0;
2717}
2718
2719ctl_table ipv6_route_table_template[] = {
2720 {
2721 .procname = "flush",
2722 .data = &init_net.ipv6.sysctl.flush_delay,
2723 .maxlen = sizeof(int),
2724 .mode = 0200,
2725 .proc_handler = ipv6_sysctl_rtcache_flush
2726 },
2727 {
2728 .procname = "gc_thresh",
2729 .data = &ip6_dst_ops_template.gc_thresh,
2730 .maxlen = sizeof(int),
2731 .mode = 0644,
2732 .proc_handler = proc_dointvec,
2733 },
2734 {
2735 .procname = "max_size",
2736 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2737 .maxlen = sizeof(int),
2738 .mode = 0644,
2739 .proc_handler = proc_dointvec,
2740 },
2741 {
2742 .procname = "gc_min_interval",
2743 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2744 .maxlen = sizeof(int),
2745 .mode = 0644,
2746 .proc_handler = proc_dointvec_jiffies,
2747 },
2748 {
2749 .procname = "gc_timeout",
2750 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2751 .maxlen = sizeof(int),
2752 .mode = 0644,
2753 .proc_handler = proc_dointvec_jiffies,
2754 },
2755 {
2756 .procname = "gc_interval",
2757 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2758 .maxlen = sizeof(int),
2759 .mode = 0644,
2760 .proc_handler = proc_dointvec_jiffies,
2761 },
2762 {
2763 .procname = "gc_elasticity",
2764 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2765 .maxlen = sizeof(int),
2766 .mode = 0644,
2767 .proc_handler = proc_dointvec,
2768 },
2769 {
2770 .procname = "mtu_expires",
2771 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2772 .maxlen = sizeof(int),
2773 .mode = 0644,
2774 .proc_handler = proc_dointvec_jiffies,
2775 },
2776 {
2777 .procname = "min_adv_mss",
2778 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2779 .maxlen = sizeof(int),
2780 .mode = 0644,
2781 .proc_handler = proc_dointvec,
2782 },
2783 {
2784 .procname = "gc_min_interval_ms",
2785 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2786 .maxlen = sizeof(int),
2787 .mode = 0644,
2788 .proc_handler = proc_dointvec_ms_jiffies,
2789 },
2790 { }
2791};
2792
2793struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2794{
2795 struct ctl_table *table;
2796
2797 table = kmemdup(ipv6_route_table_template,
2798 sizeof(ipv6_route_table_template),
2799 GFP_KERNEL);
2800
2801 if (table) {
2802 table[0].data = &net->ipv6.sysctl.flush_delay;
2803 table[0].extra1 = net;
2804 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2805 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2806 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2807 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2808 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2809 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2810 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2811 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2812 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2813 }
2814
2815 return table;
2816}
2817#endif
2818
2819static int __net_init ip6_route_net_init(struct net *net)
2820{
2821 int ret = -ENOMEM;
2822
2823 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2824 sizeof(net->ipv6.ip6_dst_ops));
2825
2826 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2827 goto out_ip6_dst_ops;
2828
2829 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2830 sizeof(*net->ipv6.ip6_null_entry),
2831 GFP_KERNEL);
2832 if (!net->ipv6.ip6_null_entry)
2833 goto out_ip6_dst_entries;
2834 net->ipv6.ip6_null_entry->dst.path =
2835 (struct dst_entry *)net->ipv6.ip6_null_entry;
2836 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2837 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2838 ip6_template_metrics, true);
2839
2840#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2841 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2842 sizeof(*net->ipv6.ip6_prohibit_entry),
2843 GFP_KERNEL);
2844 if (!net->ipv6.ip6_prohibit_entry)
2845 goto out_ip6_null_entry;
2846 net->ipv6.ip6_prohibit_entry->dst.path =
2847 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2848 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2849 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2850 ip6_template_metrics, true);
2851
2852 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2853 sizeof(*net->ipv6.ip6_blk_hole_entry),
2854 GFP_KERNEL);
2855 if (!net->ipv6.ip6_blk_hole_entry)
2856 goto out_ip6_prohibit_entry;
2857 net->ipv6.ip6_blk_hole_entry->dst.path =
2858 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2859 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2860 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2861 ip6_template_metrics, true);
2862#endif
2863
2864 net->ipv6.sysctl.flush_delay = 0;
2865 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2866 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2867 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2868 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2869 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2870 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2871 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2872
2873#ifdef CONFIG_PROC_FS
2874 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2875 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2876#endif
2877 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2878
2879 ret = 0;
2880out:
2881 return ret;
2882
2883#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2884out_ip6_prohibit_entry:
2885 kfree(net->ipv6.ip6_prohibit_entry);
2886out_ip6_null_entry:
2887 kfree(net->ipv6.ip6_null_entry);
2888#endif
2889out_ip6_dst_entries:
2890 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2891out_ip6_dst_ops:
2892 goto out;
2893}
2894
2895static void __net_exit ip6_route_net_exit(struct net *net)
2896{
2897#ifdef CONFIG_PROC_FS
2898 proc_net_remove(net, "ipv6_route");
2899 proc_net_remove(net, "rt6_stats");
2900#endif
2901 kfree(net->ipv6.ip6_null_entry);
2902#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2903 kfree(net->ipv6.ip6_prohibit_entry);
2904 kfree(net->ipv6.ip6_blk_hole_entry);
2905#endif
2906 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2907}
2908
2909static struct pernet_operations ip6_route_net_ops = {
2910 .init = ip6_route_net_init,
2911 .exit = ip6_route_net_exit,
2912};
2913
2914static struct notifier_block ip6_route_dev_notifier = {
2915 .notifier_call = ip6_route_dev_notify,
2916 .priority = 0,
2917};
2918
2919int __init ip6_route_init(void)
2920{
2921 int ret;
2922
2923 ret = -ENOMEM;
2924 ip6_dst_ops_template.kmem_cachep =
2925 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2926 SLAB_HWCACHE_ALIGN, NULL);
2927 if (!ip6_dst_ops_template.kmem_cachep)
2928 goto out;
2929
2930 ret = dst_entries_init(&ip6_dst_blackhole_ops);
2931 if (ret)
2932 goto out_kmem_cache;
2933
2934 ret = register_pernet_subsys(&ip6_route_net_ops);
2935 if (ret)
2936 goto out_dst_entries;
2937
2938 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
2939
2940 /* Registering of the loopback is done before this portion of code,
2941 * the loopback reference in rt6_info will not be taken, do it
2942 * manually for init_net */
2943 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
2944 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2945 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2946 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
2947 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2948 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
2949 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2950 #endif
2951 ret = fib6_init();
2952 if (ret)
2953 goto out_register_subsys;
2954
2955 ret = xfrm6_init();
2956 if (ret)
2957 goto out_fib6_init;
2958
2959 ret = fib6_rules_init();
2960 if (ret)
2961 goto xfrm6_init;
2962
2963 ret = -ENOBUFS;
2964 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
2965 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
2966 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
2967 goto fib6_rules_init;
2968
2969 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2970 if (ret)
2971 goto fib6_rules_init;
2972
2973out:
2974 return ret;
2975
2976fib6_rules_init:
2977 fib6_rules_cleanup();
2978xfrm6_init:
2979 xfrm6_fini();
2980out_fib6_init:
2981 fib6_gc_cleanup();
2982out_register_subsys:
2983 unregister_pernet_subsys(&ip6_route_net_ops);
2984out_dst_entries:
2985 dst_entries_destroy(&ip6_dst_blackhole_ops);
2986out_kmem_cache:
2987 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2988 goto out;
2989}
2990
2991void ip6_route_cleanup(void)
2992{
2993 unregister_netdevice_notifier(&ip6_route_dev_notifier);
2994 fib6_rules_cleanup();
2995 xfrm6_fini();
2996 fib6_gc_cleanup();
2997 unregister_pernet_subsys(&ip6_route_net_ops);
2998 dst_entries_destroy(&ip6_dst_blackhole_ops);
2999 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3000}
1/*
2 * Linux INET6 implementation
3 * FIB front-end.
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
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.
12 */
13
14/* Changes:
15 *
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.
25 */
26
27#define pr_fmt(fmt) "IPv6: " fmt
28
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>
61
62#include <asm/uaccess.h>
63
64#ifdef CONFIG_SYSCTL
65#include <linux/sysctl.h>
66#endif
67
68enum rt6_nud_state {
69 RT6_NUD_FAIL_HARD = -3,
70 RT6_NUD_FAIL_PROBE = -2,
71 RT6_NUD_FAIL_DO_RR = -1,
72 RT6_NUD_SUCCEED = 1
73};
74
75static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
76 const struct in6_addr *dest);
77static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
78static unsigned int ip6_default_advmss(const struct dst_entry *dst);
79static unsigned int ip6_mtu(const struct dst_entry *dst);
80static struct dst_entry *ip6_negative_advice(struct dst_entry *);
81static void ip6_dst_destroy(struct dst_entry *);
82static void ip6_dst_ifdown(struct dst_entry *,
83 struct net_device *dev, int how);
84static int ip6_dst_gc(struct dst_ops *ops);
85
86static int ip6_pkt_discard(struct sk_buff *skb);
87static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb);
88static int ip6_pkt_prohibit(struct sk_buff *skb);
89static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb);
90static void ip6_link_failure(struct sk_buff *skb);
91static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
92 struct sk_buff *skb, u32 mtu);
93static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
94 struct sk_buff *skb);
95static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
96
97#ifdef CONFIG_IPV6_ROUTE_INFO
98static 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,
101 unsigned int pref);
102static 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);
105#endif
106
107static void rt6_bind_peer(struct rt6_info *rt, int create)
108{
109 struct inet_peer_base *base;
110 struct inet_peer *peer;
111
112 base = inetpeer_base_ptr(rt->_rt6i_peer);
113 if (!base)
114 return;
115
116 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
117 if (peer) {
118 if (!rt6_set_peer(rt, peer))
119 inet_putpeer(peer);
120 }
121}
122
123static struct inet_peer *__rt6_get_peer(struct rt6_info *rt, int create)
124{
125 if (rt6_has_peer(rt))
126 return rt6_peer_ptr(rt);
127
128 rt6_bind_peer(rt, create);
129 return (rt6_has_peer(rt) ? rt6_peer_ptr(rt) : NULL);
130}
131
132static struct inet_peer *rt6_get_peer_create(struct rt6_info *rt)
133{
134 return __rt6_get_peer(rt, 1);
135}
136
137static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
138{
139 struct rt6_info *rt = (struct rt6_info *) dst;
140 struct inet_peer *peer;
141 u32 *p = NULL;
142
143 if (!(rt->dst.flags & DST_HOST))
144 return NULL;
145
146 peer = rt6_get_peer_create(rt);
147 if (peer) {
148 u32 *old_p = __DST_METRICS_PTR(old);
149 unsigned long prev, new;
150
151 p = peer->metrics;
152 if (inet_metrics_new(peer) ||
153 (old & DST_METRICS_FORCE_OVERWRITE))
154 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
155
156 new = (unsigned long) p;
157 prev = cmpxchg(&dst->_metrics, old, new);
158
159 if (prev != old) {
160 p = __DST_METRICS_PTR(prev);
161 if (prev & DST_METRICS_READ_ONLY)
162 p = NULL;
163 }
164 }
165 return p;
166}
167
168static inline const void *choose_neigh_daddr(struct rt6_info *rt,
169 struct sk_buff *skb,
170 const void *daddr)
171{
172 struct in6_addr *p = &rt->rt6i_gateway;
173
174 if (!ipv6_addr_any(p))
175 return (const void *) p;
176 else if (skb)
177 return &ipv6_hdr(skb)->daddr;
178 return daddr;
179}
180
181static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
182 struct sk_buff *skb,
183 const void *daddr)
184{
185 struct rt6_info *rt = (struct rt6_info *) dst;
186 struct neighbour *n;
187
188 daddr = choose_neigh_daddr(rt, skb, daddr);
189 n = __ipv6_neigh_lookup(dst->dev, daddr);
190 if (n)
191 return n;
192 return neigh_create(&nd_tbl, daddr, dst->dev);
193}
194
195static struct dst_ops ip6_dst_ops_template = {
196 .family = AF_INET6,
197 .protocol = cpu_to_be16(ETH_P_IPV6),
198 .gc = ip6_dst_gc,
199 .gc_thresh = 1024,
200 .check = ip6_dst_check,
201 .default_advmss = ip6_default_advmss,
202 .mtu = ip6_mtu,
203 .cow_metrics = ipv6_cow_metrics,
204 .destroy = ip6_dst_destroy,
205 .ifdown = ip6_dst_ifdown,
206 .negative_advice = ip6_negative_advice,
207 .link_failure = ip6_link_failure,
208 .update_pmtu = ip6_rt_update_pmtu,
209 .redirect = rt6_do_redirect,
210 .local_out = __ip6_local_out,
211 .neigh_lookup = ip6_neigh_lookup,
212};
213
214static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
215{
216 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
217
218 return mtu ? : dst->dev->mtu;
219}
220
221static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
222 struct sk_buff *skb, u32 mtu)
223{
224}
225
226static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
227 struct sk_buff *skb)
228{
229}
230
231static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
232 unsigned long old)
233{
234 return NULL;
235}
236
237static struct dst_ops ip6_dst_blackhole_ops = {
238 .family = AF_INET6,
239 .protocol = cpu_to_be16(ETH_P_IPV6),
240 .destroy = ip6_dst_destroy,
241 .check = ip6_dst_check,
242 .mtu = ip6_blackhole_mtu,
243 .default_advmss = ip6_default_advmss,
244 .update_pmtu = ip6_rt_blackhole_update_pmtu,
245 .redirect = ip6_rt_blackhole_redirect,
246 .cow_metrics = ip6_rt_blackhole_cow_metrics,
247 .neigh_lookup = ip6_neigh_lookup,
248};
249
250static const u32 ip6_template_metrics[RTAX_MAX] = {
251 [RTAX_HOPLIMIT - 1] = 0,
252};
253
254static const struct rt6_info ip6_null_entry_template = {
255 .dst = {
256 .__refcnt = ATOMIC_INIT(1),
257 .__use = 1,
258 .obsolete = DST_OBSOLETE_FORCE_CHK,
259 .error = -ENETUNREACH,
260 .input = ip6_pkt_discard,
261 .output = ip6_pkt_discard_out,
262 },
263 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
264 .rt6i_protocol = RTPROT_KERNEL,
265 .rt6i_metric = ~(u32) 0,
266 .rt6i_ref = ATOMIC_INIT(1),
267};
268
269#ifdef CONFIG_IPV6_MULTIPLE_TABLES
270
271static const struct rt6_info ip6_prohibit_entry_template = {
272 .dst = {
273 .__refcnt = ATOMIC_INIT(1),
274 .__use = 1,
275 .obsolete = DST_OBSOLETE_FORCE_CHK,
276 .error = -EACCES,
277 .input = ip6_pkt_prohibit,
278 .output = ip6_pkt_prohibit_out,
279 },
280 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
281 .rt6i_protocol = RTPROT_KERNEL,
282 .rt6i_metric = ~(u32) 0,
283 .rt6i_ref = ATOMIC_INIT(1),
284};
285
286static const struct rt6_info ip6_blk_hole_entry_template = {
287 .dst = {
288 .__refcnt = ATOMIC_INIT(1),
289 .__use = 1,
290 .obsolete = DST_OBSOLETE_FORCE_CHK,
291 .error = -EINVAL,
292 .input = dst_discard,
293 .output = dst_discard_sk,
294 },
295 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
296 .rt6i_protocol = RTPROT_KERNEL,
297 .rt6i_metric = ~(u32) 0,
298 .rt6i_ref = ATOMIC_INIT(1),
299};
300
301#endif
302
303/* allocate dst with ip6_dst_ops */
304static inline struct rt6_info *ip6_dst_alloc(struct net *net,
305 struct net_device *dev,
306 int flags,
307 struct fib6_table *table)
308{
309 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
310 0, DST_OBSOLETE_FORCE_CHK, flags);
311
312 if (rt) {
313 struct dst_entry *dst = &rt->dst;
314
315 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
316 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
317 rt->rt6i_genid = rt_genid_ipv6(net);
318 INIT_LIST_HEAD(&rt->rt6i_siblings);
319 }
320 return rt;
321}
322
323static void ip6_dst_destroy(struct dst_entry *dst)
324{
325 struct rt6_info *rt = (struct rt6_info *)dst;
326 struct inet6_dev *idev = rt->rt6i_idev;
327 struct dst_entry *from = dst->from;
328
329 if (!(rt->dst.flags & DST_HOST))
330 dst_destroy_metrics_generic(dst);
331
332 if (idev) {
333 rt->rt6i_idev = NULL;
334 in6_dev_put(idev);
335 }
336
337 dst->from = NULL;
338 dst_release(from);
339
340 if (rt6_has_peer(rt)) {
341 struct inet_peer *peer = rt6_peer_ptr(rt);
342 inet_putpeer(peer);
343 }
344}
345
346static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
347 int how)
348{
349 struct rt6_info *rt = (struct rt6_info *)dst;
350 struct inet6_dev *idev = rt->rt6i_idev;
351 struct net_device *loopback_dev =
352 dev_net(dev)->loopback_dev;
353
354 if (dev != loopback_dev) {
355 if (idev && idev->dev == dev) {
356 struct inet6_dev *loopback_idev =
357 in6_dev_get(loopback_dev);
358 if (loopback_idev) {
359 rt->rt6i_idev = loopback_idev;
360 in6_dev_put(idev);
361 }
362 }
363 }
364}
365
366static bool rt6_check_expired(const struct rt6_info *rt)
367{
368 if (rt->rt6i_flags & RTF_EXPIRES) {
369 if (time_after(jiffies, rt->dst.expires))
370 return true;
371 } else if (rt->dst.from) {
372 return rt6_check_expired((struct rt6_info *) rt->dst.from);
373 }
374 return false;
375}
376
377/* Multipath route selection:
378 * Hash based function using packet header and flowlabel.
379 * Adapted from fib_info_hashfn()
380 */
381static int rt6_info_hash_nhsfn(unsigned int candidate_count,
382 const struct flowi6 *fl6)
383{
384 unsigned int val = fl6->flowi6_proto;
385
386 val ^= ipv6_addr_hash(&fl6->daddr);
387 val ^= ipv6_addr_hash(&fl6->saddr);
388
389 /* Work only if this not encapsulated */
390 switch (fl6->flowi6_proto) {
391 case IPPROTO_UDP:
392 case IPPROTO_TCP:
393 case IPPROTO_SCTP:
394 val ^= (__force u16)fl6->fl6_sport;
395 val ^= (__force u16)fl6->fl6_dport;
396 break;
397
398 case IPPROTO_ICMPV6:
399 val ^= (__force u16)fl6->fl6_icmp_type;
400 val ^= (__force u16)fl6->fl6_icmp_code;
401 break;
402 }
403 /* RFC6438 recommands to use flowlabel */
404 val ^= (__force u32)fl6->flowlabel;
405
406 /* Perhaps, we need to tune, this function? */
407 val = val ^ (val >> 7) ^ (val >> 12);
408 return val % candidate_count;
409}
410
411static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
412 struct flowi6 *fl6, int oif,
413 int strict)
414{
415 struct rt6_info *sibling, *next_sibling;
416 int route_choosen;
417
418 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
419 /* Don't change the route, if route_choosen == 0
420 * (siblings does not include ourself)
421 */
422 if (route_choosen)
423 list_for_each_entry_safe(sibling, next_sibling,
424 &match->rt6i_siblings, rt6i_siblings) {
425 route_choosen--;
426 if (route_choosen == 0) {
427 if (rt6_score_route(sibling, oif, strict) < 0)
428 break;
429 match = sibling;
430 break;
431 }
432 }
433 return match;
434}
435
436/*
437 * Route lookup. Any table->tb6_lock is implied.
438 */
439
440static inline struct rt6_info *rt6_device_match(struct net *net,
441 struct rt6_info *rt,
442 const struct in6_addr *saddr,
443 int oif,
444 int flags)
445{
446 struct rt6_info *local = NULL;
447 struct rt6_info *sprt;
448
449 if (!oif && ipv6_addr_any(saddr))
450 goto out;
451
452 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
453 struct net_device *dev = sprt->dst.dev;
454
455 if (oif) {
456 if (dev->ifindex == oif)
457 return sprt;
458 if (dev->flags & IFF_LOOPBACK) {
459 if (!sprt->rt6i_idev ||
460 sprt->rt6i_idev->dev->ifindex != oif) {
461 if (flags & RT6_LOOKUP_F_IFACE && oif)
462 continue;
463 if (local && (!oif ||
464 local->rt6i_idev->dev->ifindex == oif))
465 continue;
466 }
467 local = sprt;
468 }
469 } else {
470 if (ipv6_chk_addr(net, saddr, dev,
471 flags & RT6_LOOKUP_F_IFACE))
472 return sprt;
473 }
474 }
475
476 if (oif) {
477 if (local)
478 return local;
479
480 if (flags & RT6_LOOKUP_F_IFACE)
481 return net->ipv6.ip6_null_entry;
482 }
483out:
484 return rt;
485}
486
487#ifdef CONFIG_IPV6_ROUTER_PREF
488struct __rt6_probe_work {
489 struct work_struct work;
490 struct in6_addr target;
491 struct net_device *dev;
492};
493
494static void rt6_probe_deferred(struct work_struct *w)
495{
496 struct in6_addr mcaddr;
497 struct __rt6_probe_work *work =
498 container_of(w, struct __rt6_probe_work, work);
499
500 addrconf_addr_solict_mult(&work->target, &mcaddr);
501 ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
502 dev_put(work->dev);
503 kfree(w);
504}
505
506static void rt6_probe(struct rt6_info *rt)
507{
508 struct neighbour *neigh;
509 /*
510 * Okay, this does not seem to be appropriate
511 * for now, however, we need to check if it
512 * is really so; aka Router Reachability Probing.
513 *
514 * Router Reachability Probe MUST be rate-limited
515 * to no more than one per minute.
516 */
517 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
518 return;
519 rcu_read_lock_bh();
520 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
521 if (neigh) {
522 write_lock(&neigh->lock);
523 if (neigh->nud_state & NUD_VALID)
524 goto out;
525 }
526
527 if (!neigh ||
528 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
529 struct __rt6_probe_work *work;
530
531 work = kmalloc(sizeof(*work), GFP_ATOMIC);
532
533 if (neigh && work)
534 __neigh_set_probe_once(neigh);
535
536 if (neigh)
537 write_unlock(&neigh->lock);
538
539 if (work) {
540 INIT_WORK(&work->work, rt6_probe_deferred);
541 work->target = rt->rt6i_gateway;
542 dev_hold(rt->dst.dev);
543 work->dev = rt->dst.dev;
544 schedule_work(&work->work);
545 }
546 } else {
547out:
548 write_unlock(&neigh->lock);
549 }
550 rcu_read_unlock_bh();
551}
552#else
553static inline void rt6_probe(struct rt6_info *rt)
554{
555}
556#endif
557
558/*
559 * Default Router Selection (RFC 2461 6.3.6)
560 */
561static inline int rt6_check_dev(struct rt6_info *rt, int oif)
562{
563 struct net_device *dev = rt->dst.dev;
564 if (!oif || dev->ifindex == oif)
565 return 2;
566 if ((dev->flags & IFF_LOOPBACK) &&
567 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
568 return 1;
569 return 0;
570}
571
572static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
573{
574 struct neighbour *neigh;
575 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
576
577 if (rt->rt6i_flags & RTF_NONEXTHOP ||
578 !(rt->rt6i_flags & RTF_GATEWAY))
579 return RT6_NUD_SUCCEED;
580
581 rcu_read_lock_bh();
582 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
583 if (neigh) {
584 read_lock(&neigh->lock);
585 if (neigh->nud_state & NUD_VALID)
586 ret = RT6_NUD_SUCCEED;
587#ifdef CONFIG_IPV6_ROUTER_PREF
588 else if (!(neigh->nud_state & NUD_FAILED))
589 ret = RT6_NUD_SUCCEED;
590 else
591 ret = RT6_NUD_FAIL_PROBE;
592#endif
593 read_unlock(&neigh->lock);
594 } else {
595 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
596 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
597 }
598 rcu_read_unlock_bh();
599
600 return ret;
601}
602
603static int rt6_score_route(struct rt6_info *rt, int oif,
604 int strict)
605{
606 int m;
607
608 m = rt6_check_dev(rt, oif);
609 if (!m && (strict & RT6_LOOKUP_F_IFACE))
610 return RT6_NUD_FAIL_HARD;
611#ifdef CONFIG_IPV6_ROUTER_PREF
612 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
613#endif
614 if (strict & RT6_LOOKUP_F_REACHABLE) {
615 int n = rt6_check_neigh(rt);
616 if (n < 0)
617 return n;
618 }
619 return m;
620}
621
622static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
623 int *mpri, struct rt6_info *match,
624 bool *do_rr)
625{
626 int m;
627 bool match_do_rr = false;
628
629 if (rt6_check_expired(rt))
630 goto out;
631
632 m = rt6_score_route(rt, oif, strict);
633 if (m == RT6_NUD_FAIL_DO_RR) {
634 match_do_rr = true;
635 m = 0; /* lowest valid score */
636 } else if (m == RT6_NUD_FAIL_HARD) {
637 goto out;
638 }
639
640 if (strict & RT6_LOOKUP_F_REACHABLE)
641 rt6_probe(rt);
642
643 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
644 if (m > *mpri) {
645 *do_rr = match_do_rr;
646 *mpri = m;
647 match = rt;
648 }
649out:
650 return match;
651}
652
653static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
654 struct rt6_info *rr_head,
655 u32 metric, int oif, int strict,
656 bool *do_rr)
657{
658 struct rt6_info *rt, *match;
659 int mpri = -1;
660
661 match = NULL;
662 for (rt = rr_head; rt && rt->rt6i_metric == metric;
663 rt = rt->dst.rt6_next)
664 match = find_match(rt, oif, strict, &mpri, match, do_rr);
665 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
666 rt = rt->dst.rt6_next)
667 match = find_match(rt, oif, strict, &mpri, match, do_rr);
668
669 return match;
670}
671
672static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
673{
674 struct rt6_info *match, *rt0;
675 struct net *net;
676 bool do_rr = false;
677
678 rt0 = fn->rr_ptr;
679 if (!rt0)
680 fn->rr_ptr = rt0 = fn->leaf;
681
682 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
683 &do_rr);
684
685 if (do_rr) {
686 struct rt6_info *next = rt0->dst.rt6_next;
687
688 /* no entries matched; do round-robin */
689 if (!next || next->rt6i_metric != rt0->rt6i_metric)
690 next = fn->leaf;
691
692 if (next != rt0)
693 fn->rr_ptr = next;
694 }
695
696 net = dev_net(rt0->dst.dev);
697 return match ? match : net->ipv6.ip6_null_entry;
698}
699
700#ifdef CONFIG_IPV6_ROUTE_INFO
701int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
702 const struct in6_addr *gwaddr)
703{
704 struct net *net = dev_net(dev);
705 struct route_info *rinfo = (struct route_info *) opt;
706 struct in6_addr prefix_buf, *prefix;
707 unsigned int pref;
708 unsigned long lifetime;
709 struct rt6_info *rt;
710
711 if (len < sizeof(struct route_info)) {
712 return -EINVAL;
713 }
714
715 /* Sanity check for prefix_len and length */
716 if (rinfo->length > 3) {
717 return -EINVAL;
718 } else if (rinfo->prefix_len > 128) {
719 return -EINVAL;
720 } else if (rinfo->prefix_len > 64) {
721 if (rinfo->length < 2) {
722 return -EINVAL;
723 }
724 } else if (rinfo->prefix_len > 0) {
725 if (rinfo->length < 1) {
726 return -EINVAL;
727 }
728 }
729
730 pref = rinfo->route_pref;
731 if (pref == ICMPV6_ROUTER_PREF_INVALID)
732 return -EINVAL;
733
734 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
735
736 if (rinfo->length == 3)
737 prefix = (struct in6_addr *)rinfo->prefix;
738 else {
739 /* this function is safe */
740 ipv6_addr_prefix(&prefix_buf,
741 (struct in6_addr *)rinfo->prefix,
742 rinfo->prefix_len);
743 prefix = &prefix_buf;
744 }
745
746 if (rinfo->prefix_len == 0)
747 rt = rt6_get_dflt_router(gwaddr, dev);
748 else
749 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
750 gwaddr, dev->ifindex);
751
752 if (rt && !lifetime) {
753 ip6_del_rt(rt);
754 rt = NULL;
755 }
756
757 if (!rt && lifetime)
758 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
759 pref);
760 else if (rt)
761 rt->rt6i_flags = RTF_ROUTEINFO |
762 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
763
764 if (rt) {
765 if (!addrconf_finite_timeout(lifetime))
766 rt6_clean_expires(rt);
767 else
768 rt6_set_expires(rt, jiffies + HZ * lifetime);
769
770 ip6_rt_put(rt);
771 }
772 return 0;
773}
774#endif
775
776#define BACKTRACK(__net, saddr) \
777do { \
778 if (rt == __net->ipv6.ip6_null_entry) { \
779 struct fib6_node *pn; \
780 while (1) { \
781 if (fn->fn_flags & RTN_TL_ROOT) \
782 goto out; \
783 pn = fn->parent; \
784 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
785 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
786 else \
787 fn = pn; \
788 if (fn->fn_flags & RTN_RTINFO) \
789 goto restart; \
790 } \
791 } \
792} while (0)
793
794static struct rt6_info *ip6_pol_route_lookup(struct net *net,
795 struct fib6_table *table,
796 struct flowi6 *fl6, int flags)
797{
798 struct fib6_node *fn;
799 struct rt6_info *rt;
800
801 read_lock_bh(&table->tb6_lock);
802 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
803restart:
804 rt = fn->leaf;
805 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
806 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
807 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
808 BACKTRACK(net, &fl6->saddr);
809out:
810 dst_use(&rt->dst, jiffies);
811 read_unlock_bh(&table->tb6_lock);
812 return rt;
813
814}
815
816struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
817 int flags)
818{
819 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
820}
821EXPORT_SYMBOL_GPL(ip6_route_lookup);
822
823struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
824 const struct in6_addr *saddr, int oif, int strict)
825{
826 struct flowi6 fl6 = {
827 .flowi6_oif = oif,
828 .daddr = *daddr,
829 };
830 struct dst_entry *dst;
831 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
832
833 if (saddr) {
834 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
835 flags |= RT6_LOOKUP_F_HAS_SADDR;
836 }
837
838 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
839 if (dst->error == 0)
840 return (struct rt6_info *) dst;
841
842 dst_release(dst);
843
844 return NULL;
845}
846
847EXPORT_SYMBOL(rt6_lookup);
848
849/* ip6_ins_rt is called with FREE table->tb6_lock.
850 It takes new route entry, the addition fails by any reason the
851 route is freed. In any case, if caller does not hold it, it may
852 be destroyed.
853 */
854
855static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
856 struct nlattr *mx, int mx_len)
857{
858 int err;
859 struct fib6_table *table;
860
861 table = rt->rt6i_table;
862 write_lock_bh(&table->tb6_lock);
863 err = fib6_add(&table->tb6_root, rt, info, mx, mx_len);
864 write_unlock_bh(&table->tb6_lock);
865
866 return err;
867}
868
869int ip6_ins_rt(struct rt6_info *rt)
870{
871 struct nl_info info = {
872 .nl_net = dev_net(rt->dst.dev),
873 };
874 return __ip6_ins_rt(rt, &info, NULL, 0);
875}
876
877static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
878 const struct in6_addr *daddr,
879 const struct in6_addr *saddr)
880{
881 struct rt6_info *rt;
882
883 /*
884 * Clone the route.
885 */
886
887 rt = ip6_rt_copy(ort, daddr);
888
889 if (rt) {
890 if (ort->rt6i_dst.plen != 128 &&
891 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
892 rt->rt6i_flags |= RTF_ANYCAST;
893
894 rt->rt6i_flags |= RTF_CACHE;
895
896#ifdef CONFIG_IPV6_SUBTREES
897 if (rt->rt6i_src.plen && saddr) {
898 rt->rt6i_src.addr = *saddr;
899 rt->rt6i_src.plen = 128;
900 }
901#endif
902 }
903
904 return rt;
905}
906
907static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
908 const struct in6_addr *daddr)
909{
910 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
911
912 if (rt)
913 rt->rt6i_flags |= RTF_CACHE;
914 return rt;
915}
916
917static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
918 struct flowi6 *fl6, int flags)
919{
920 struct fib6_node *fn;
921 struct rt6_info *rt, *nrt;
922 int strict = 0;
923 int attempts = 3;
924 int err;
925 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
926
927 strict |= flags & RT6_LOOKUP_F_IFACE;
928
929relookup:
930 read_lock_bh(&table->tb6_lock);
931
932restart_2:
933 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
934
935restart:
936 rt = rt6_select(fn, oif, strict | reachable);
937 if (rt->rt6i_nsiblings)
938 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
939 BACKTRACK(net, &fl6->saddr);
940 if (rt == net->ipv6.ip6_null_entry ||
941 rt->rt6i_flags & RTF_CACHE)
942 goto out;
943
944 dst_hold(&rt->dst);
945 read_unlock_bh(&table->tb6_lock);
946
947 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
948 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
949 else if (!(rt->dst.flags & DST_HOST))
950 nrt = rt6_alloc_clone(rt, &fl6->daddr);
951 else
952 goto out2;
953
954 ip6_rt_put(rt);
955 rt = nrt ? : net->ipv6.ip6_null_entry;
956
957 dst_hold(&rt->dst);
958 if (nrt) {
959 err = ip6_ins_rt(nrt);
960 if (!err)
961 goto out2;
962 }
963
964 if (--attempts <= 0)
965 goto out2;
966
967 /*
968 * Race condition! In the gap, when table->tb6_lock was
969 * released someone could insert this route. Relookup.
970 */
971 ip6_rt_put(rt);
972 goto relookup;
973
974out:
975 if (reachable) {
976 reachable = 0;
977 goto restart_2;
978 }
979 dst_hold(&rt->dst);
980 read_unlock_bh(&table->tb6_lock);
981out2:
982 rt->dst.lastuse = jiffies;
983 rt->dst.__use++;
984
985 return rt;
986}
987
988static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
989 struct flowi6 *fl6, int flags)
990{
991 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
992}
993
994static struct dst_entry *ip6_route_input_lookup(struct net *net,
995 struct net_device *dev,
996 struct flowi6 *fl6, int flags)
997{
998 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
999 flags |= RT6_LOOKUP_F_IFACE;
1000
1001 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1002}
1003
1004void ip6_route_input(struct sk_buff *skb)
1005{
1006 const struct ipv6hdr *iph = ipv6_hdr(skb);
1007 struct net *net = dev_net(skb->dev);
1008 int flags = RT6_LOOKUP_F_HAS_SADDR;
1009 struct flowi6 fl6 = {
1010 .flowi6_iif = skb->dev->ifindex,
1011 .daddr = iph->daddr,
1012 .saddr = iph->saddr,
1013 .flowlabel = ip6_flowinfo(iph),
1014 .flowi6_mark = skb->mark,
1015 .flowi6_proto = iph->nexthdr,
1016 };
1017
1018 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1019}
1020
1021static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1022 struct flowi6 *fl6, int flags)
1023{
1024 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1025}
1026
1027struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1028 struct flowi6 *fl6)
1029{
1030 int flags = 0;
1031
1032 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1033
1034 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1035 flags |= RT6_LOOKUP_F_IFACE;
1036
1037 if (!ipv6_addr_any(&fl6->saddr))
1038 flags |= RT6_LOOKUP_F_HAS_SADDR;
1039 else if (sk)
1040 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1041
1042 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1043}
1044
1045EXPORT_SYMBOL(ip6_route_output);
1046
1047struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1048{
1049 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1050 struct dst_entry *new = NULL;
1051
1052 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1053 if (rt) {
1054 new = &rt->dst;
1055
1056 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1057 rt6_init_peer(rt, net->ipv6.peers);
1058
1059 new->__use = 1;
1060 new->input = dst_discard;
1061 new->output = dst_discard_sk;
1062
1063 if (dst_metrics_read_only(&ort->dst))
1064 new->_metrics = ort->dst._metrics;
1065 else
1066 dst_copy_metrics(new, &ort->dst);
1067 rt->rt6i_idev = ort->rt6i_idev;
1068 if (rt->rt6i_idev)
1069 in6_dev_hold(rt->rt6i_idev);
1070
1071 rt->rt6i_gateway = ort->rt6i_gateway;
1072 rt->rt6i_flags = ort->rt6i_flags;
1073 rt->rt6i_metric = 0;
1074
1075 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1076#ifdef CONFIG_IPV6_SUBTREES
1077 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1078#endif
1079
1080 dst_free(new);
1081 }
1082
1083 dst_release(dst_orig);
1084 return new ? new : ERR_PTR(-ENOMEM);
1085}
1086
1087/*
1088 * Destination cache support functions
1089 */
1090
1091static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1092{
1093 struct rt6_info *rt;
1094
1095 rt = (struct rt6_info *) dst;
1096
1097 /* All IPV6 dsts are created with ->obsolete set to the value
1098 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1099 * into this function always.
1100 */
1101 if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev)))
1102 return NULL;
1103
1104 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1105 return NULL;
1106
1107 if (rt6_check_expired(rt))
1108 return NULL;
1109
1110 return dst;
1111}
1112
1113static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1114{
1115 struct rt6_info *rt = (struct rt6_info *) dst;
1116
1117 if (rt) {
1118 if (rt->rt6i_flags & RTF_CACHE) {
1119 if (rt6_check_expired(rt)) {
1120 ip6_del_rt(rt);
1121 dst = NULL;
1122 }
1123 } else {
1124 dst_release(dst);
1125 dst = NULL;
1126 }
1127 }
1128 return dst;
1129}
1130
1131static void ip6_link_failure(struct sk_buff *skb)
1132{
1133 struct rt6_info *rt;
1134
1135 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1136
1137 rt = (struct rt6_info *) skb_dst(skb);
1138 if (rt) {
1139 if (rt->rt6i_flags & RTF_CACHE) {
1140 dst_hold(&rt->dst);
1141 if (ip6_del_rt(rt))
1142 dst_free(&rt->dst);
1143 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1144 rt->rt6i_node->fn_sernum = -1;
1145 }
1146 }
1147}
1148
1149static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1150 struct sk_buff *skb, u32 mtu)
1151{
1152 struct rt6_info *rt6 = (struct rt6_info*)dst;
1153
1154 dst_confirm(dst);
1155 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1156 struct net *net = dev_net(dst->dev);
1157
1158 rt6->rt6i_flags |= RTF_MODIFIED;
1159 if (mtu < IPV6_MIN_MTU) {
1160 u32 features = dst_metric(dst, RTAX_FEATURES);
1161 mtu = IPV6_MIN_MTU;
1162 features |= RTAX_FEATURE_ALLFRAG;
1163 dst_metric_set(dst, RTAX_FEATURES, features);
1164 }
1165 dst_metric_set(dst, RTAX_MTU, mtu);
1166 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1167 }
1168}
1169
1170void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1171 int oif, u32 mark)
1172{
1173 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1174 struct dst_entry *dst;
1175 struct flowi6 fl6;
1176
1177 memset(&fl6, 0, sizeof(fl6));
1178 fl6.flowi6_oif = oif;
1179 fl6.flowi6_mark = mark;
1180 fl6.daddr = iph->daddr;
1181 fl6.saddr = iph->saddr;
1182 fl6.flowlabel = ip6_flowinfo(iph);
1183
1184 dst = ip6_route_output(net, NULL, &fl6);
1185 if (!dst->error)
1186 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1187 dst_release(dst);
1188}
1189EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1190
1191void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1192{
1193 ip6_update_pmtu(skb, sock_net(sk), mtu,
1194 sk->sk_bound_dev_if, sk->sk_mark);
1195}
1196EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1197
1198/* Handle redirects */
1199struct ip6rd_flowi {
1200 struct flowi6 fl6;
1201 struct in6_addr gateway;
1202};
1203
1204static struct rt6_info *__ip6_route_redirect(struct net *net,
1205 struct fib6_table *table,
1206 struct flowi6 *fl6,
1207 int flags)
1208{
1209 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1210 struct rt6_info *rt;
1211 struct fib6_node *fn;
1212
1213 /* Get the "current" route for this destination and
1214 * check if the redirect has come from approriate router.
1215 *
1216 * RFC 4861 specifies that redirects should only be
1217 * accepted if they come from the nexthop to the target.
1218 * Due to the way the routes are chosen, this notion
1219 * is a bit fuzzy and one might need to check all possible
1220 * routes.
1221 */
1222
1223 read_lock_bh(&table->tb6_lock);
1224 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1225restart:
1226 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1227 if (rt6_check_expired(rt))
1228 continue;
1229 if (rt->dst.error)
1230 break;
1231 if (!(rt->rt6i_flags & RTF_GATEWAY))
1232 continue;
1233 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1234 continue;
1235 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1236 continue;
1237 break;
1238 }
1239
1240 if (!rt)
1241 rt = net->ipv6.ip6_null_entry;
1242 else if (rt->dst.error) {
1243 rt = net->ipv6.ip6_null_entry;
1244 goto out;
1245 }
1246 BACKTRACK(net, &fl6->saddr);
1247out:
1248 dst_hold(&rt->dst);
1249
1250 read_unlock_bh(&table->tb6_lock);
1251
1252 return rt;
1253};
1254
1255static struct dst_entry *ip6_route_redirect(struct net *net,
1256 const struct flowi6 *fl6,
1257 const struct in6_addr *gateway)
1258{
1259 int flags = RT6_LOOKUP_F_HAS_SADDR;
1260 struct ip6rd_flowi rdfl;
1261
1262 rdfl.fl6 = *fl6;
1263 rdfl.gateway = *gateway;
1264
1265 return fib6_rule_lookup(net, &rdfl.fl6,
1266 flags, __ip6_route_redirect);
1267}
1268
1269void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1270{
1271 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1272 struct dst_entry *dst;
1273 struct flowi6 fl6;
1274
1275 memset(&fl6, 0, sizeof(fl6));
1276 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1277 fl6.flowi6_oif = oif;
1278 fl6.flowi6_mark = mark;
1279 fl6.daddr = iph->daddr;
1280 fl6.saddr = iph->saddr;
1281 fl6.flowlabel = ip6_flowinfo(iph);
1282
1283 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1284 rt6_do_redirect(dst, NULL, skb);
1285 dst_release(dst);
1286}
1287EXPORT_SYMBOL_GPL(ip6_redirect);
1288
1289void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1290 u32 mark)
1291{
1292 const struct ipv6hdr *iph = ipv6_hdr(skb);
1293 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1294 struct dst_entry *dst;
1295 struct flowi6 fl6;
1296
1297 memset(&fl6, 0, sizeof(fl6));
1298 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1299 fl6.flowi6_oif = oif;
1300 fl6.flowi6_mark = mark;
1301 fl6.daddr = msg->dest;
1302 fl6.saddr = iph->daddr;
1303
1304 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1305 rt6_do_redirect(dst, NULL, skb);
1306 dst_release(dst);
1307}
1308
1309void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1310{
1311 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1312}
1313EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1314
1315static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1316{
1317 struct net_device *dev = dst->dev;
1318 unsigned int mtu = dst_mtu(dst);
1319 struct net *net = dev_net(dev);
1320
1321 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1322
1323 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1324 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1325
1326 /*
1327 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1328 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1329 * IPV6_MAXPLEN is also valid and means: "any MSS,
1330 * rely only on pmtu discovery"
1331 */
1332 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1333 mtu = IPV6_MAXPLEN;
1334 return mtu;
1335}
1336
1337static unsigned int ip6_mtu(const struct dst_entry *dst)
1338{
1339 struct inet6_dev *idev;
1340 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1341
1342 if (mtu)
1343 goto out;
1344
1345 mtu = IPV6_MIN_MTU;
1346
1347 rcu_read_lock();
1348 idev = __in6_dev_get(dst->dev);
1349 if (idev)
1350 mtu = idev->cnf.mtu6;
1351 rcu_read_unlock();
1352
1353out:
1354 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1355}
1356
1357static struct dst_entry *icmp6_dst_gc_list;
1358static DEFINE_SPINLOCK(icmp6_dst_lock);
1359
1360struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1361 struct flowi6 *fl6)
1362{
1363 struct dst_entry *dst;
1364 struct rt6_info *rt;
1365 struct inet6_dev *idev = in6_dev_get(dev);
1366 struct net *net = dev_net(dev);
1367
1368 if (unlikely(!idev))
1369 return ERR_PTR(-ENODEV);
1370
1371 rt = ip6_dst_alloc(net, dev, 0, NULL);
1372 if (unlikely(!rt)) {
1373 in6_dev_put(idev);
1374 dst = ERR_PTR(-ENOMEM);
1375 goto out;
1376 }
1377
1378 rt->dst.flags |= DST_HOST;
1379 rt->dst.output = ip6_output;
1380 atomic_set(&rt->dst.__refcnt, 1);
1381 rt->rt6i_gateway = fl6->daddr;
1382 rt->rt6i_dst.addr = fl6->daddr;
1383 rt->rt6i_dst.plen = 128;
1384 rt->rt6i_idev = idev;
1385 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1386
1387 spin_lock_bh(&icmp6_dst_lock);
1388 rt->dst.next = icmp6_dst_gc_list;
1389 icmp6_dst_gc_list = &rt->dst;
1390 spin_unlock_bh(&icmp6_dst_lock);
1391
1392 fib6_force_start_gc(net);
1393
1394 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1395
1396out:
1397 return dst;
1398}
1399
1400int icmp6_dst_gc(void)
1401{
1402 struct dst_entry *dst, **pprev;
1403 int more = 0;
1404
1405 spin_lock_bh(&icmp6_dst_lock);
1406 pprev = &icmp6_dst_gc_list;
1407
1408 while ((dst = *pprev) != NULL) {
1409 if (!atomic_read(&dst->__refcnt)) {
1410 *pprev = dst->next;
1411 dst_free(dst);
1412 } else {
1413 pprev = &dst->next;
1414 ++more;
1415 }
1416 }
1417
1418 spin_unlock_bh(&icmp6_dst_lock);
1419
1420 return more;
1421}
1422
1423static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1424 void *arg)
1425{
1426 struct dst_entry *dst, **pprev;
1427
1428 spin_lock_bh(&icmp6_dst_lock);
1429 pprev = &icmp6_dst_gc_list;
1430 while ((dst = *pprev) != NULL) {
1431 struct rt6_info *rt = (struct rt6_info *) dst;
1432 if (func(rt, arg)) {
1433 *pprev = dst->next;
1434 dst_free(dst);
1435 } else {
1436 pprev = &dst->next;
1437 }
1438 }
1439 spin_unlock_bh(&icmp6_dst_lock);
1440}
1441
1442static int ip6_dst_gc(struct dst_ops *ops)
1443{
1444 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1445 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1446 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1447 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1448 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1449 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1450 int entries;
1451
1452 entries = dst_entries_get_fast(ops);
1453 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1454 entries <= rt_max_size)
1455 goto out;
1456
1457 net->ipv6.ip6_rt_gc_expire++;
1458 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1459 entries = dst_entries_get_slow(ops);
1460 if (entries < ops->gc_thresh)
1461 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1462out:
1463 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1464 return entries > rt_max_size;
1465}
1466
1467/*
1468 *
1469 */
1470
1471int ip6_route_add(struct fib6_config *cfg)
1472{
1473 int err;
1474 struct net *net = cfg->fc_nlinfo.nl_net;
1475 struct rt6_info *rt = NULL;
1476 struct net_device *dev = NULL;
1477 struct inet6_dev *idev = NULL;
1478 struct fib6_table *table;
1479 int addr_type;
1480
1481 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1482 return -EINVAL;
1483#ifndef CONFIG_IPV6_SUBTREES
1484 if (cfg->fc_src_len)
1485 return -EINVAL;
1486#endif
1487 if (cfg->fc_ifindex) {
1488 err = -ENODEV;
1489 dev = dev_get_by_index(net, cfg->fc_ifindex);
1490 if (!dev)
1491 goto out;
1492 idev = in6_dev_get(dev);
1493 if (!idev)
1494 goto out;
1495 }
1496
1497 if (cfg->fc_metric == 0)
1498 cfg->fc_metric = IP6_RT_PRIO_USER;
1499
1500 err = -ENOBUFS;
1501 if (cfg->fc_nlinfo.nlh &&
1502 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1503 table = fib6_get_table(net, cfg->fc_table);
1504 if (!table) {
1505 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1506 table = fib6_new_table(net, cfg->fc_table);
1507 }
1508 } else {
1509 table = fib6_new_table(net, cfg->fc_table);
1510 }
1511
1512 if (!table)
1513 goto out;
1514
1515 rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1516
1517 if (!rt) {
1518 err = -ENOMEM;
1519 goto out;
1520 }
1521
1522 if (cfg->fc_flags & RTF_EXPIRES)
1523 rt6_set_expires(rt, jiffies +
1524 clock_t_to_jiffies(cfg->fc_expires));
1525 else
1526 rt6_clean_expires(rt);
1527
1528 if (cfg->fc_protocol == RTPROT_UNSPEC)
1529 cfg->fc_protocol = RTPROT_BOOT;
1530 rt->rt6i_protocol = cfg->fc_protocol;
1531
1532 addr_type = ipv6_addr_type(&cfg->fc_dst);
1533
1534 if (addr_type & IPV6_ADDR_MULTICAST)
1535 rt->dst.input = ip6_mc_input;
1536 else if (cfg->fc_flags & RTF_LOCAL)
1537 rt->dst.input = ip6_input;
1538 else
1539 rt->dst.input = ip6_forward;
1540
1541 rt->dst.output = ip6_output;
1542
1543 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1544 rt->rt6i_dst.plen = cfg->fc_dst_len;
1545 if (rt->rt6i_dst.plen == 128) {
1546 rt->dst.flags |= DST_HOST;
1547 dst_metrics_set_force_overwrite(&rt->dst);
1548 }
1549
1550#ifdef CONFIG_IPV6_SUBTREES
1551 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1552 rt->rt6i_src.plen = cfg->fc_src_len;
1553#endif
1554
1555 rt->rt6i_metric = cfg->fc_metric;
1556
1557 /* We cannot add true routes via loopback here,
1558 they would result in kernel looping; promote them to reject routes
1559 */
1560 if ((cfg->fc_flags & RTF_REJECT) ||
1561 (dev && (dev->flags & IFF_LOOPBACK) &&
1562 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1563 !(cfg->fc_flags & RTF_LOCAL))) {
1564 /* hold loopback dev/idev if we haven't done so. */
1565 if (dev != net->loopback_dev) {
1566 if (dev) {
1567 dev_put(dev);
1568 in6_dev_put(idev);
1569 }
1570 dev = net->loopback_dev;
1571 dev_hold(dev);
1572 idev = in6_dev_get(dev);
1573 if (!idev) {
1574 err = -ENODEV;
1575 goto out;
1576 }
1577 }
1578 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1579 switch (cfg->fc_type) {
1580 case RTN_BLACKHOLE:
1581 rt->dst.error = -EINVAL;
1582 rt->dst.output = dst_discard_sk;
1583 rt->dst.input = dst_discard;
1584 break;
1585 case RTN_PROHIBIT:
1586 rt->dst.error = -EACCES;
1587 rt->dst.output = ip6_pkt_prohibit_out;
1588 rt->dst.input = ip6_pkt_prohibit;
1589 break;
1590 case RTN_THROW:
1591 default:
1592 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1593 : -ENETUNREACH;
1594 rt->dst.output = ip6_pkt_discard_out;
1595 rt->dst.input = ip6_pkt_discard;
1596 break;
1597 }
1598 goto install_route;
1599 }
1600
1601 if (cfg->fc_flags & RTF_GATEWAY) {
1602 const struct in6_addr *gw_addr;
1603 int gwa_type;
1604
1605 gw_addr = &cfg->fc_gateway;
1606 rt->rt6i_gateway = *gw_addr;
1607 gwa_type = ipv6_addr_type(gw_addr);
1608
1609 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1610 struct rt6_info *grt;
1611
1612 /* IPv6 strictly inhibits using not link-local
1613 addresses as nexthop address.
1614 Otherwise, router will not able to send redirects.
1615 It is very good, but in some (rare!) circumstances
1616 (SIT, PtP, NBMA NOARP links) it is handy to allow
1617 some exceptions. --ANK
1618 */
1619 err = -EINVAL;
1620 if (!(gwa_type & IPV6_ADDR_UNICAST))
1621 goto out;
1622
1623 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1624
1625 err = -EHOSTUNREACH;
1626 if (!grt)
1627 goto out;
1628 if (dev) {
1629 if (dev != grt->dst.dev) {
1630 ip6_rt_put(grt);
1631 goto out;
1632 }
1633 } else {
1634 dev = grt->dst.dev;
1635 idev = grt->rt6i_idev;
1636 dev_hold(dev);
1637 in6_dev_hold(grt->rt6i_idev);
1638 }
1639 if (!(grt->rt6i_flags & RTF_GATEWAY))
1640 err = 0;
1641 ip6_rt_put(grt);
1642
1643 if (err)
1644 goto out;
1645 }
1646 err = -EINVAL;
1647 if (!dev || (dev->flags & IFF_LOOPBACK))
1648 goto out;
1649 }
1650
1651 err = -ENODEV;
1652 if (!dev)
1653 goto out;
1654
1655 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1656 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1657 err = -EINVAL;
1658 goto out;
1659 }
1660 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1661 rt->rt6i_prefsrc.plen = 128;
1662 } else
1663 rt->rt6i_prefsrc.plen = 0;
1664
1665 rt->rt6i_flags = cfg->fc_flags;
1666
1667install_route:
1668 rt->dst.dev = dev;
1669 rt->rt6i_idev = idev;
1670 rt->rt6i_table = table;
1671
1672 cfg->fc_nlinfo.nl_net = dev_net(dev);
1673
1674 return __ip6_ins_rt(rt, &cfg->fc_nlinfo, cfg->fc_mx, cfg->fc_mx_len);
1675
1676out:
1677 if (dev)
1678 dev_put(dev);
1679 if (idev)
1680 in6_dev_put(idev);
1681 if (rt)
1682 dst_free(&rt->dst);
1683 return err;
1684}
1685
1686static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1687{
1688 int err;
1689 struct fib6_table *table;
1690 struct net *net = dev_net(rt->dst.dev);
1691
1692 if (rt == net->ipv6.ip6_null_entry) {
1693 err = -ENOENT;
1694 goto out;
1695 }
1696
1697 table = rt->rt6i_table;
1698 write_lock_bh(&table->tb6_lock);
1699 err = fib6_del(rt, info);
1700 write_unlock_bh(&table->tb6_lock);
1701
1702out:
1703 ip6_rt_put(rt);
1704 return err;
1705}
1706
1707int ip6_del_rt(struct rt6_info *rt)
1708{
1709 struct nl_info info = {
1710 .nl_net = dev_net(rt->dst.dev),
1711 };
1712 return __ip6_del_rt(rt, &info);
1713}
1714
1715static int ip6_route_del(struct fib6_config *cfg)
1716{
1717 struct fib6_table *table;
1718 struct fib6_node *fn;
1719 struct rt6_info *rt;
1720 int err = -ESRCH;
1721
1722 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1723 if (!table)
1724 return err;
1725
1726 read_lock_bh(&table->tb6_lock);
1727
1728 fn = fib6_locate(&table->tb6_root,
1729 &cfg->fc_dst, cfg->fc_dst_len,
1730 &cfg->fc_src, cfg->fc_src_len);
1731
1732 if (fn) {
1733 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1734 if (cfg->fc_ifindex &&
1735 (!rt->dst.dev ||
1736 rt->dst.dev->ifindex != cfg->fc_ifindex))
1737 continue;
1738 if (cfg->fc_flags & RTF_GATEWAY &&
1739 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1740 continue;
1741 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1742 continue;
1743 dst_hold(&rt->dst);
1744 read_unlock_bh(&table->tb6_lock);
1745
1746 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1747 }
1748 }
1749 read_unlock_bh(&table->tb6_lock);
1750
1751 return err;
1752}
1753
1754static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1755{
1756 struct net *net = dev_net(skb->dev);
1757 struct netevent_redirect netevent;
1758 struct rt6_info *rt, *nrt = NULL;
1759 struct ndisc_options ndopts;
1760 struct inet6_dev *in6_dev;
1761 struct neighbour *neigh;
1762 struct rd_msg *msg;
1763 int optlen, on_link;
1764 u8 *lladdr;
1765
1766 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1767 optlen -= sizeof(*msg);
1768
1769 if (optlen < 0) {
1770 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1771 return;
1772 }
1773
1774 msg = (struct rd_msg *)icmp6_hdr(skb);
1775
1776 if (ipv6_addr_is_multicast(&msg->dest)) {
1777 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1778 return;
1779 }
1780
1781 on_link = 0;
1782 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1783 on_link = 1;
1784 } else if (ipv6_addr_type(&msg->target) !=
1785 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1786 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1787 return;
1788 }
1789
1790 in6_dev = __in6_dev_get(skb->dev);
1791 if (!in6_dev)
1792 return;
1793 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1794 return;
1795
1796 /* RFC2461 8.1:
1797 * The IP source address of the Redirect MUST be the same as the current
1798 * first-hop router for the specified ICMP Destination Address.
1799 */
1800
1801 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1802 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1803 return;
1804 }
1805
1806 lladdr = NULL;
1807 if (ndopts.nd_opts_tgt_lladdr) {
1808 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1809 skb->dev);
1810 if (!lladdr) {
1811 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1812 return;
1813 }
1814 }
1815
1816 rt = (struct rt6_info *) dst;
1817 if (rt == net->ipv6.ip6_null_entry) {
1818 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1819 return;
1820 }
1821
1822 /* Redirect received -> path was valid.
1823 * Look, redirects are sent only in response to data packets,
1824 * so that this nexthop apparently is reachable. --ANK
1825 */
1826 dst_confirm(&rt->dst);
1827
1828 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1829 if (!neigh)
1830 return;
1831
1832 /*
1833 * We have finally decided to accept it.
1834 */
1835
1836 neigh_update(neigh, lladdr, NUD_STALE,
1837 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1838 NEIGH_UPDATE_F_OVERRIDE|
1839 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1840 NEIGH_UPDATE_F_ISROUTER))
1841 );
1842
1843 nrt = ip6_rt_copy(rt, &msg->dest);
1844 if (!nrt)
1845 goto out;
1846
1847 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1848 if (on_link)
1849 nrt->rt6i_flags &= ~RTF_GATEWAY;
1850
1851 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1852
1853 if (ip6_ins_rt(nrt))
1854 goto out;
1855
1856 netevent.old = &rt->dst;
1857 netevent.new = &nrt->dst;
1858 netevent.daddr = &msg->dest;
1859 netevent.neigh = neigh;
1860 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1861
1862 if (rt->rt6i_flags & RTF_CACHE) {
1863 rt = (struct rt6_info *) dst_clone(&rt->dst);
1864 ip6_del_rt(rt);
1865 }
1866
1867out:
1868 neigh_release(neigh);
1869}
1870
1871/*
1872 * Misc support functions
1873 */
1874
1875static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1876 const struct in6_addr *dest)
1877{
1878 struct net *net = dev_net(ort->dst.dev);
1879 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1880 ort->rt6i_table);
1881
1882 if (rt) {
1883 rt->dst.input = ort->dst.input;
1884 rt->dst.output = ort->dst.output;
1885 rt->dst.flags |= DST_HOST;
1886
1887 rt->rt6i_dst.addr = *dest;
1888 rt->rt6i_dst.plen = 128;
1889 dst_copy_metrics(&rt->dst, &ort->dst);
1890 rt->dst.error = ort->dst.error;
1891 rt->rt6i_idev = ort->rt6i_idev;
1892 if (rt->rt6i_idev)
1893 in6_dev_hold(rt->rt6i_idev);
1894 rt->dst.lastuse = jiffies;
1895
1896 if (ort->rt6i_flags & RTF_GATEWAY)
1897 rt->rt6i_gateway = ort->rt6i_gateway;
1898 else
1899 rt->rt6i_gateway = *dest;
1900 rt->rt6i_flags = ort->rt6i_flags;
1901 rt6_set_from(rt, ort);
1902 rt->rt6i_metric = 0;
1903
1904#ifdef CONFIG_IPV6_SUBTREES
1905 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1906#endif
1907 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1908 rt->rt6i_table = ort->rt6i_table;
1909 }
1910 return rt;
1911}
1912
1913#ifdef CONFIG_IPV6_ROUTE_INFO
1914static struct rt6_info *rt6_get_route_info(struct net *net,
1915 const struct in6_addr *prefix, int prefixlen,
1916 const struct in6_addr *gwaddr, int ifindex)
1917{
1918 struct fib6_node *fn;
1919 struct rt6_info *rt = NULL;
1920 struct fib6_table *table;
1921
1922 table = fib6_get_table(net, RT6_TABLE_INFO);
1923 if (!table)
1924 return NULL;
1925
1926 read_lock_bh(&table->tb6_lock);
1927 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1928 if (!fn)
1929 goto out;
1930
1931 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1932 if (rt->dst.dev->ifindex != ifindex)
1933 continue;
1934 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1935 continue;
1936 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1937 continue;
1938 dst_hold(&rt->dst);
1939 break;
1940 }
1941out:
1942 read_unlock_bh(&table->tb6_lock);
1943 return rt;
1944}
1945
1946static struct rt6_info *rt6_add_route_info(struct net *net,
1947 const struct in6_addr *prefix, int prefixlen,
1948 const struct in6_addr *gwaddr, int ifindex,
1949 unsigned int pref)
1950{
1951 struct fib6_config cfg = {
1952 .fc_table = RT6_TABLE_INFO,
1953 .fc_metric = IP6_RT_PRIO_USER,
1954 .fc_ifindex = ifindex,
1955 .fc_dst_len = prefixlen,
1956 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1957 RTF_UP | RTF_PREF(pref),
1958 .fc_nlinfo.portid = 0,
1959 .fc_nlinfo.nlh = NULL,
1960 .fc_nlinfo.nl_net = net,
1961 };
1962
1963 cfg.fc_dst = *prefix;
1964 cfg.fc_gateway = *gwaddr;
1965
1966 /* We should treat it as a default route if prefix length is 0. */
1967 if (!prefixlen)
1968 cfg.fc_flags |= RTF_DEFAULT;
1969
1970 ip6_route_add(&cfg);
1971
1972 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1973}
1974#endif
1975
1976struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1977{
1978 struct rt6_info *rt;
1979 struct fib6_table *table;
1980
1981 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1982 if (!table)
1983 return NULL;
1984
1985 read_lock_bh(&table->tb6_lock);
1986 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1987 if (dev == rt->dst.dev &&
1988 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1989 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1990 break;
1991 }
1992 if (rt)
1993 dst_hold(&rt->dst);
1994 read_unlock_bh(&table->tb6_lock);
1995 return rt;
1996}
1997
1998struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1999 struct net_device *dev,
2000 unsigned int pref)
2001{
2002 struct fib6_config cfg = {
2003 .fc_table = RT6_TABLE_DFLT,
2004 .fc_metric = IP6_RT_PRIO_USER,
2005 .fc_ifindex = dev->ifindex,
2006 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2007 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2008 .fc_nlinfo.portid = 0,
2009 .fc_nlinfo.nlh = NULL,
2010 .fc_nlinfo.nl_net = dev_net(dev),
2011 };
2012
2013 cfg.fc_gateway = *gwaddr;
2014
2015 ip6_route_add(&cfg);
2016
2017 return rt6_get_dflt_router(gwaddr, dev);
2018}
2019
2020void rt6_purge_dflt_routers(struct net *net)
2021{
2022 struct rt6_info *rt;
2023 struct fib6_table *table;
2024
2025 /* NOTE: Keep consistent with rt6_get_dflt_router */
2026 table = fib6_get_table(net, RT6_TABLE_DFLT);
2027 if (!table)
2028 return;
2029
2030restart:
2031 read_lock_bh(&table->tb6_lock);
2032 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2033 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2034 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2035 dst_hold(&rt->dst);
2036 read_unlock_bh(&table->tb6_lock);
2037 ip6_del_rt(rt);
2038 goto restart;
2039 }
2040 }
2041 read_unlock_bh(&table->tb6_lock);
2042}
2043
2044static void rtmsg_to_fib6_config(struct net *net,
2045 struct in6_rtmsg *rtmsg,
2046 struct fib6_config *cfg)
2047{
2048 memset(cfg, 0, sizeof(*cfg));
2049
2050 cfg->fc_table = RT6_TABLE_MAIN;
2051 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2052 cfg->fc_metric = rtmsg->rtmsg_metric;
2053 cfg->fc_expires = rtmsg->rtmsg_info;
2054 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2055 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2056 cfg->fc_flags = rtmsg->rtmsg_flags;
2057
2058 cfg->fc_nlinfo.nl_net = net;
2059
2060 cfg->fc_dst = rtmsg->rtmsg_dst;
2061 cfg->fc_src = rtmsg->rtmsg_src;
2062 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2063}
2064
2065int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2066{
2067 struct fib6_config cfg;
2068 struct in6_rtmsg rtmsg;
2069 int err;
2070
2071 switch(cmd) {
2072 case SIOCADDRT: /* Add a route */
2073 case SIOCDELRT: /* Delete a route */
2074 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2075 return -EPERM;
2076 err = copy_from_user(&rtmsg, arg,
2077 sizeof(struct in6_rtmsg));
2078 if (err)
2079 return -EFAULT;
2080
2081 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2082
2083 rtnl_lock();
2084 switch (cmd) {
2085 case SIOCADDRT:
2086 err = ip6_route_add(&cfg);
2087 break;
2088 case SIOCDELRT:
2089 err = ip6_route_del(&cfg);
2090 break;
2091 default:
2092 err = -EINVAL;
2093 }
2094 rtnl_unlock();
2095
2096 return err;
2097 }
2098
2099 return -EINVAL;
2100}
2101
2102/*
2103 * Drop the packet on the floor
2104 */
2105
2106static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2107{
2108 int type;
2109 struct dst_entry *dst = skb_dst(skb);
2110 switch (ipstats_mib_noroutes) {
2111 case IPSTATS_MIB_INNOROUTES:
2112 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2113 if (type == IPV6_ADDR_ANY) {
2114 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2115 IPSTATS_MIB_INADDRERRORS);
2116 break;
2117 }
2118 /* FALLTHROUGH */
2119 case IPSTATS_MIB_OUTNOROUTES:
2120 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2121 ipstats_mib_noroutes);
2122 break;
2123 }
2124 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2125 kfree_skb(skb);
2126 return 0;
2127}
2128
2129static int ip6_pkt_discard(struct sk_buff *skb)
2130{
2131 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2132}
2133
2134static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2135{
2136 skb->dev = skb_dst(skb)->dev;
2137 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2138}
2139
2140static int ip6_pkt_prohibit(struct sk_buff *skb)
2141{
2142 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2143}
2144
2145static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2146{
2147 skb->dev = skb_dst(skb)->dev;
2148 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2149}
2150
2151/*
2152 * Allocate a dst for local (unicast / anycast) address.
2153 */
2154
2155struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2156 const struct in6_addr *addr,
2157 bool anycast)
2158{
2159 struct net *net = dev_net(idev->dev);
2160 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2161 DST_NOCOUNT, NULL);
2162 if (!rt)
2163 return ERR_PTR(-ENOMEM);
2164
2165 in6_dev_hold(idev);
2166
2167 rt->dst.flags |= DST_HOST;
2168 rt->dst.input = ip6_input;
2169 rt->dst.output = ip6_output;
2170 rt->rt6i_idev = idev;
2171
2172 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2173 if (anycast)
2174 rt->rt6i_flags |= RTF_ANYCAST;
2175 else
2176 rt->rt6i_flags |= RTF_LOCAL;
2177
2178 rt->rt6i_gateway = *addr;
2179 rt->rt6i_dst.addr = *addr;
2180 rt->rt6i_dst.plen = 128;
2181 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2182
2183 atomic_set(&rt->dst.__refcnt, 1);
2184
2185 return rt;
2186}
2187
2188int ip6_route_get_saddr(struct net *net,
2189 struct rt6_info *rt,
2190 const struct in6_addr *daddr,
2191 unsigned int prefs,
2192 struct in6_addr *saddr)
2193{
2194 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2195 int err = 0;
2196 if (rt->rt6i_prefsrc.plen)
2197 *saddr = rt->rt6i_prefsrc.addr;
2198 else
2199 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2200 daddr, prefs, saddr);
2201 return err;
2202}
2203
2204/* remove deleted ip from prefsrc entries */
2205struct arg_dev_net_ip {
2206 struct net_device *dev;
2207 struct net *net;
2208 struct in6_addr *addr;
2209};
2210
2211static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2212{
2213 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2214 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2215 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2216
2217 if (((void *)rt->dst.dev == dev || !dev) &&
2218 rt != net->ipv6.ip6_null_entry &&
2219 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2220 /* remove prefsrc entry */
2221 rt->rt6i_prefsrc.plen = 0;
2222 }
2223 return 0;
2224}
2225
2226void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2227{
2228 struct net *net = dev_net(ifp->idev->dev);
2229 struct arg_dev_net_ip adni = {
2230 .dev = ifp->idev->dev,
2231 .net = net,
2232 .addr = &ifp->addr,
2233 };
2234 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2235}
2236
2237#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2238#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2239
2240/* Remove routers and update dst entries when gateway turn into host. */
2241static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2242{
2243 struct in6_addr *gateway = (struct in6_addr *)arg;
2244
2245 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2246 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2247 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2248 return -1;
2249 }
2250 return 0;
2251}
2252
2253void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2254{
2255 fib6_clean_all(net, fib6_clean_tohost, gateway);
2256}
2257
2258struct arg_dev_net {
2259 struct net_device *dev;
2260 struct net *net;
2261};
2262
2263static int fib6_ifdown(struct rt6_info *rt, void *arg)
2264{
2265 const struct arg_dev_net *adn = arg;
2266 const struct net_device *dev = adn->dev;
2267
2268 if ((rt->dst.dev == dev || !dev) &&
2269 rt != adn->net->ipv6.ip6_null_entry)
2270 return -1;
2271
2272 return 0;
2273}
2274
2275void rt6_ifdown(struct net *net, struct net_device *dev)
2276{
2277 struct arg_dev_net adn = {
2278 .dev = dev,
2279 .net = net,
2280 };
2281
2282 fib6_clean_all(net, fib6_ifdown, &adn);
2283 icmp6_clean_all(fib6_ifdown, &adn);
2284}
2285
2286struct rt6_mtu_change_arg {
2287 struct net_device *dev;
2288 unsigned int mtu;
2289};
2290
2291static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2292{
2293 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2294 struct inet6_dev *idev;
2295
2296 /* In IPv6 pmtu discovery is not optional,
2297 so that RTAX_MTU lock cannot disable it.
2298 We still use this lock to block changes
2299 caused by addrconf/ndisc.
2300 */
2301
2302 idev = __in6_dev_get(arg->dev);
2303 if (!idev)
2304 return 0;
2305
2306 /* For administrative MTU increase, there is no way to discover
2307 IPv6 PMTU increase, so PMTU increase should be updated here.
2308 Since RFC 1981 doesn't include administrative MTU increase
2309 update PMTU increase is a MUST. (i.e. jumbo frame)
2310 */
2311 /*
2312 If new MTU is less than route PMTU, this new MTU will be the
2313 lowest MTU in the path, update the route PMTU to reflect PMTU
2314 decreases; if new MTU is greater than route PMTU, and the
2315 old MTU is the lowest MTU in the path, update the route PMTU
2316 to reflect the increase. In this case if the other nodes' MTU
2317 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2318 PMTU discouvery.
2319 */
2320 if (rt->dst.dev == arg->dev &&
2321 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2322 (dst_mtu(&rt->dst) >= arg->mtu ||
2323 (dst_mtu(&rt->dst) < arg->mtu &&
2324 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2325 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2326 }
2327 return 0;
2328}
2329
2330void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2331{
2332 struct rt6_mtu_change_arg arg = {
2333 .dev = dev,
2334 .mtu = mtu,
2335 };
2336
2337 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2338}
2339
2340static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2341 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2342 [RTA_OIF] = { .type = NLA_U32 },
2343 [RTA_IIF] = { .type = NLA_U32 },
2344 [RTA_PRIORITY] = { .type = NLA_U32 },
2345 [RTA_METRICS] = { .type = NLA_NESTED },
2346 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2347};
2348
2349static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2350 struct fib6_config *cfg)
2351{
2352 struct rtmsg *rtm;
2353 struct nlattr *tb[RTA_MAX+1];
2354 int err;
2355
2356 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2357 if (err < 0)
2358 goto errout;
2359
2360 err = -EINVAL;
2361 rtm = nlmsg_data(nlh);
2362 memset(cfg, 0, sizeof(*cfg));
2363
2364 cfg->fc_table = rtm->rtm_table;
2365 cfg->fc_dst_len = rtm->rtm_dst_len;
2366 cfg->fc_src_len = rtm->rtm_src_len;
2367 cfg->fc_flags = RTF_UP;
2368 cfg->fc_protocol = rtm->rtm_protocol;
2369 cfg->fc_type = rtm->rtm_type;
2370
2371 if (rtm->rtm_type == RTN_UNREACHABLE ||
2372 rtm->rtm_type == RTN_BLACKHOLE ||
2373 rtm->rtm_type == RTN_PROHIBIT ||
2374 rtm->rtm_type == RTN_THROW)
2375 cfg->fc_flags |= RTF_REJECT;
2376
2377 if (rtm->rtm_type == RTN_LOCAL)
2378 cfg->fc_flags |= RTF_LOCAL;
2379
2380 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2381 cfg->fc_nlinfo.nlh = nlh;
2382 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2383
2384 if (tb[RTA_GATEWAY]) {
2385 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2386 cfg->fc_flags |= RTF_GATEWAY;
2387 }
2388
2389 if (tb[RTA_DST]) {
2390 int plen = (rtm->rtm_dst_len + 7) >> 3;
2391
2392 if (nla_len(tb[RTA_DST]) < plen)
2393 goto errout;
2394
2395 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2396 }
2397
2398 if (tb[RTA_SRC]) {
2399 int plen = (rtm->rtm_src_len + 7) >> 3;
2400
2401 if (nla_len(tb[RTA_SRC]) < plen)
2402 goto errout;
2403
2404 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2405 }
2406
2407 if (tb[RTA_PREFSRC])
2408 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2409
2410 if (tb[RTA_OIF])
2411 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2412
2413 if (tb[RTA_PRIORITY])
2414 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2415
2416 if (tb[RTA_METRICS]) {
2417 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2418 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2419 }
2420
2421 if (tb[RTA_TABLE])
2422 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2423
2424 if (tb[RTA_MULTIPATH]) {
2425 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2426 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2427 }
2428
2429 err = 0;
2430errout:
2431 return err;
2432}
2433
2434static int ip6_route_multipath(struct fib6_config *cfg, int add)
2435{
2436 struct fib6_config r_cfg;
2437 struct rtnexthop *rtnh;
2438 int remaining;
2439 int attrlen;
2440 int err = 0, last_err = 0;
2441
2442beginning:
2443 rtnh = (struct rtnexthop *)cfg->fc_mp;
2444 remaining = cfg->fc_mp_len;
2445
2446 /* Parse a Multipath Entry */
2447 while (rtnh_ok(rtnh, remaining)) {
2448 memcpy(&r_cfg, cfg, sizeof(*cfg));
2449 if (rtnh->rtnh_ifindex)
2450 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2451
2452 attrlen = rtnh_attrlen(rtnh);
2453 if (attrlen > 0) {
2454 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2455
2456 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2457 if (nla) {
2458 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2459 r_cfg.fc_flags |= RTF_GATEWAY;
2460 }
2461 }
2462 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2463 if (err) {
2464 last_err = err;
2465 /* If we are trying to remove a route, do not stop the
2466 * loop when ip6_route_del() fails (because next hop is
2467 * already gone), we should try to remove all next hops.
2468 */
2469 if (add) {
2470 /* If add fails, we should try to delete all
2471 * next hops that have been already added.
2472 */
2473 add = 0;
2474 goto beginning;
2475 }
2476 }
2477 /* Because each route is added like a single route we remove
2478 * this flag after the first nexthop (if there is a collision,
2479 * we have already fail to add the first nexthop:
2480 * fib6_add_rt2node() has reject it).
2481 */
2482 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2483 rtnh = rtnh_next(rtnh, &remaining);
2484 }
2485
2486 return last_err;
2487}
2488
2489static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2490{
2491 struct fib6_config cfg;
2492 int err;
2493
2494 err = rtm_to_fib6_config(skb, nlh, &cfg);
2495 if (err < 0)
2496 return err;
2497
2498 if (cfg.fc_mp)
2499 return ip6_route_multipath(&cfg, 0);
2500 else
2501 return ip6_route_del(&cfg);
2502}
2503
2504static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2505{
2506 struct fib6_config cfg;
2507 int err;
2508
2509 err = rtm_to_fib6_config(skb, nlh, &cfg);
2510 if (err < 0)
2511 return err;
2512
2513 if (cfg.fc_mp)
2514 return ip6_route_multipath(&cfg, 1);
2515 else
2516 return ip6_route_add(&cfg);
2517}
2518
2519static inline size_t rt6_nlmsg_size(void)
2520{
2521 return NLMSG_ALIGN(sizeof(struct rtmsg))
2522 + nla_total_size(16) /* RTA_SRC */
2523 + nla_total_size(16) /* RTA_DST */
2524 + nla_total_size(16) /* RTA_GATEWAY */
2525 + nla_total_size(16) /* RTA_PREFSRC */
2526 + nla_total_size(4) /* RTA_TABLE */
2527 + nla_total_size(4) /* RTA_IIF */
2528 + nla_total_size(4) /* RTA_OIF */
2529 + nla_total_size(4) /* RTA_PRIORITY */
2530 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2531 + nla_total_size(sizeof(struct rta_cacheinfo));
2532}
2533
2534static int rt6_fill_node(struct net *net,
2535 struct sk_buff *skb, struct rt6_info *rt,
2536 struct in6_addr *dst, struct in6_addr *src,
2537 int iif, int type, u32 portid, u32 seq,
2538 int prefix, int nowait, unsigned int flags)
2539{
2540 struct rtmsg *rtm;
2541 struct nlmsghdr *nlh;
2542 long expires;
2543 u32 table;
2544
2545 if (prefix) { /* user wants prefix routes only */
2546 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2547 /* success since this is not a prefix route */
2548 return 1;
2549 }
2550 }
2551
2552 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2553 if (!nlh)
2554 return -EMSGSIZE;
2555
2556 rtm = nlmsg_data(nlh);
2557 rtm->rtm_family = AF_INET6;
2558 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2559 rtm->rtm_src_len = rt->rt6i_src.plen;
2560 rtm->rtm_tos = 0;
2561 if (rt->rt6i_table)
2562 table = rt->rt6i_table->tb6_id;
2563 else
2564 table = RT6_TABLE_UNSPEC;
2565 rtm->rtm_table = table;
2566 if (nla_put_u32(skb, RTA_TABLE, table))
2567 goto nla_put_failure;
2568 if (rt->rt6i_flags & RTF_REJECT) {
2569 switch (rt->dst.error) {
2570 case -EINVAL:
2571 rtm->rtm_type = RTN_BLACKHOLE;
2572 break;
2573 case -EACCES:
2574 rtm->rtm_type = RTN_PROHIBIT;
2575 break;
2576 case -EAGAIN:
2577 rtm->rtm_type = RTN_THROW;
2578 break;
2579 default:
2580 rtm->rtm_type = RTN_UNREACHABLE;
2581 break;
2582 }
2583 }
2584 else if (rt->rt6i_flags & RTF_LOCAL)
2585 rtm->rtm_type = RTN_LOCAL;
2586 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2587 rtm->rtm_type = RTN_LOCAL;
2588 else
2589 rtm->rtm_type = RTN_UNICAST;
2590 rtm->rtm_flags = 0;
2591 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2592 rtm->rtm_protocol = rt->rt6i_protocol;
2593 if (rt->rt6i_flags & RTF_DYNAMIC)
2594 rtm->rtm_protocol = RTPROT_REDIRECT;
2595 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2596 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2597 rtm->rtm_protocol = RTPROT_RA;
2598 else
2599 rtm->rtm_protocol = RTPROT_KERNEL;
2600 }
2601
2602 if (rt->rt6i_flags & RTF_CACHE)
2603 rtm->rtm_flags |= RTM_F_CLONED;
2604
2605 if (dst) {
2606 if (nla_put(skb, RTA_DST, 16, dst))
2607 goto nla_put_failure;
2608 rtm->rtm_dst_len = 128;
2609 } else if (rtm->rtm_dst_len)
2610 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2611 goto nla_put_failure;
2612#ifdef CONFIG_IPV6_SUBTREES
2613 if (src) {
2614 if (nla_put(skb, RTA_SRC, 16, src))
2615 goto nla_put_failure;
2616 rtm->rtm_src_len = 128;
2617 } else if (rtm->rtm_src_len &&
2618 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2619 goto nla_put_failure;
2620#endif
2621 if (iif) {
2622#ifdef CONFIG_IPV6_MROUTE
2623 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2624 int err = ip6mr_get_route(net, skb, rtm, nowait);
2625 if (err <= 0) {
2626 if (!nowait) {
2627 if (err == 0)
2628 return 0;
2629 goto nla_put_failure;
2630 } else {
2631 if (err == -EMSGSIZE)
2632 goto nla_put_failure;
2633 }
2634 }
2635 } else
2636#endif
2637 if (nla_put_u32(skb, RTA_IIF, iif))
2638 goto nla_put_failure;
2639 } else if (dst) {
2640 struct in6_addr saddr_buf;
2641 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2642 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2643 goto nla_put_failure;
2644 }
2645
2646 if (rt->rt6i_prefsrc.plen) {
2647 struct in6_addr saddr_buf;
2648 saddr_buf = rt->rt6i_prefsrc.addr;
2649 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2650 goto nla_put_failure;
2651 }
2652
2653 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2654 goto nla_put_failure;
2655
2656 if (rt->rt6i_flags & RTF_GATEWAY) {
2657 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2658 goto nla_put_failure;
2659 }
2660
2661 if (rt->dst.dev &&
2662 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2663 goto nla_put_failure;
2664 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2665 goto nla_put_failure;
2666
2667 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2668
2669 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2670 goto nla_put_failure;
2671
2672 return nlmsg_end(skb, nlh);
2673
2674nla_put_failure:
2675 nlmsg_cancel(skb, nlh);
2676 return -EMSGSIZE;
2677}
2678
2679int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2680{
2681 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2682 int prefix;
2683
2684 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2685 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2686 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2687 } else
2688 prefix = 0;
2689
2690 return rt6_fill_node(arg->net,
2691 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2692 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2693 prefix, 0, NLM_F_MULTI);
2694}
2695
2696static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2697{
2698 struct net *net = sock_net(in_skb->sk);
2699 struct nlattr *tb[RTA_MAX+1];
2700 struct rt6_info *rt;
2701 struct sk_buff *skb;
2702 struct rtmsg *rtm;
2703 struct flowi6 fl6;
2704 int err, iif = 0, oif = 0;
2705
2706 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2707 if (err < 0)
2708 goto errout;
2709
2710 err = -EINVAL;
2711 memset(&fl6, 0, sizeof(fl6));
2712
2713 if (tb[RTA_SRC]) {
2714 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2715 goto errout;
2716
2717 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2718 }
2719
2720 if (tb[RTA_DST]) {
2721 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2722 goto errout;
2723
2724 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2725 }
2726
2727 if (tb[RTA_IIF])
2728 iif = nla_get_u32(tb[RTA_IIF]);
2729
2730 if (tb[RTA_OIF])
2731 oif = nla_get_u32(tb[RTA_OIF]);
2732
2733 if (tb[RTA_MARK])
2734 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
2735
2736 if (iif) {
2737 struct net_device *dev;
2738 int flags = 0;
2739
2740 dev = __dev_get_by_index(net, iif);
2741 if (!dev) {
2742 err = -ENODEV;
2743 goto errout;
2744 }
2745
2746 fl6.flowi6_iif = iif;
2747
2748 if (!ipv6_addr_any(&fl6.saddr))
2749 flags |= RT6_LOOKUP_F_HAS_SADDR;
2750
2751 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2752 flags);
2753 } else {
2754 fl6.flowi6_oif = oif;
2755
2756 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2757 }
2758
2759 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2760 if (!skb) {
2761 ip6_rt_put(rt);
2762 err = -ENOBUFS;
2763 goto errout;
2764 }
2765
2766 /* Reserve room for dummy headers, this skb can pass
2767 through good chunk of routing engine.
2768 */
2769 skb_reset_mac_header(skb);
2770 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2771
2772 skb_dst_set(skb, &rt->dst);
2773
2774 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2775 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2776 nlh->nlmsg_seq, 0, 0, 0);
2777 if (err < 0) {
2778 kfree_skb(skb);
2779 goto errout;
2780 }
2781
2782 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2783errout:
2784 return err;
2785}
2786
2787void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2788{
2789 struct sk_buff *skb;
2790 struct net *net = info->nl_net;
2791 u32 seq;
2792 int err;
2793
2794 err = -ENOBUFS;
2795 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2796
2797 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2798 if (!skb)
2799 goto errout;
2800
2801 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2802 event, info->portid, seq, 0, 0, 0);
2803 if (err < 0) {
2804 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2805 WARN_ON(err == -EMSGSIZE);
2806 kfree_skb(skb);
2807 goto errout;
2808 }
2809 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2810 info->nlh, gfp_any());
2811 return;
2812errout:
2813 if (err < 0)
2814 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2815}
2816
2817static int ip6_route_dev_notify(struct notifier_block *this,
2818 unsigned long event, void *ptr)
2819{
2820 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2821 struct net *net = dev_net(dev);
2822
2823 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2824 net->ipv6.ip6_null_entry->dst.dev = dev;
2825 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2826#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2827 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2828 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2829 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2830 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2831#endif
2832 }
2833
2834 return NOTIFY_OK;
2835}
2836
2837/*
2838 * /proc
2839 */
2840
2841#ifdef CONFIG_PROC_FS
2842
2843static const struct file_operations ipv6_route_proc_fops = {
2844 .owner = THIS_MODULE,
2845 .open = ipv6_route_open,
2846 .read = seq_read,
2847 .llseek = seq_lseek,
2848 .release = seq_release_net,
2849};
2850
2851static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2852{
2853 struct net *net = (struct net *)seq->private;
2854 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2855 net->ipv6.rt6_stats->fib_nodes,
2856 net->ipv6.rt6_stats->fib_route_nodes,
2857 net->ipv6.rt6_stats->fib_rt_alloc,
2858 net->ipv6.rt6_stats->fib_rt_entries,
2859 net->ipv6.rt6_stats->fib_rt_cache,
2860 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2861 net->ipv6.rt6_stats->fib_discarded_routes);
2862
2863 return 0;
2864}
2865
2866static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2867{
2868 return single_open_net(inode, file, rt6_stats_seq_show);
2869}
2870
2871static const struct file_operations rt6_stats_seq_fops = {
2872 .owner = THIS_MODULE,
2873 .open = rt6_stats_seq_open,
2874 .read = seq_read,
2875 .llseek = seq_lseek,
2876 .release = single_release_net,
2877};
2878#endif /* CONFIG_PROC_FS */
2879
2880#ifdef CONFIG_SYSCTL
2881
2882static
2883int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2884 void __user *buffer, size_t *lenp, loff_t *ppos)
2885{
2886 struct net *net;
2887 int delay;
2888 if (!write)
2889 return -EINVAL;
2890
2891 net = (struct net *)ctl->extra1;
2892 delay = net->ipv6.sysctl.flush_delay;
2893 proc_dointvec(ctl, write, buffer, lenp, ppos);
2894 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2895 return 0;
2896}
2897
2898struct ctl_table ipv6_route_table_template[] = {
2899 {
2900 .procname = "flush",
2901 .data = &init_net.ipv6.sysctl.flush_delay,
2902 .maxlen = sizeof(int),
2903 .mode = 0200,
2904 .proc_handler = ipv6_sysctl_rtcache_flush
2905 },
2906 {
2907 .procname = "gc_thresh",
2908 .data = &ip6_dst_ops_template.gc_thresh,
2909 .maxlen = sizeof(int),
2910 .mode = 0644,
2911 .proc_handler = proc_dointvec,
2912 },
2913 {
2914 .procname = "max_size",
2915 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2916 .maxlen = sizeof(int),
2917 .mode = 0644,
2918 .proc_handler = proc_dointvec,
2919 },
2920 {
2921 .procname = "gc_min_interval",
2922 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2923 .maxlen = sizeof(int),
2924 .mode = 0644,
2925 .proc_handler = proc_dointvec_jiffies,
2926 },
2927 {
2928 .procname = "gc_timeout",
2929 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2930 .maxlen = sizeof(int),
2931 .mode = 0644,
2932 .proc_handler = proc_dointvec_jiffies,
2933 },
2934 {
2935 .procname = "gc_interval",
2936 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2937 .maxlen = sizeof(int),
2938 .mode = 0644,
2939 .proc_handler = proc_dointvec_jiffies,
2940 },
2941 {
2942 .procname = "gc_elasticity",
2943 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2944 .maxlen = sizeof(int),
2945 .mode = 0644,
2946 .proc_handler = proc_dointvec,
2947 },
2948 {
2949 .procname = "mtu_expires",
2950 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2951 .maxlen = sizeof(int),
2952 .mode = 0644,
2953 .proc_handler = proc_dointvec_jiffies,
2954 },
2955 {
2956 .procname = "min_adv_mss",
2957 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2958 .maxlen = sizeof(int),
2959 .mode = 0644,
2960 .proc_handler = proc_dointvec,
2961 },
2962 {
2963 .procname = "gc_min_interval_ms",
2964 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2965 .maxlen = sizeof(int),
2966 .mode = 0644,
2967 .proc_handler = proc_dointvec_ms_jiffies,
2968 },
2969 { }
2970};
2971
2972struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2973{
2974 struct ctl_table *table;
2975
2976 table = kmemdup(ipv6_route_table_template,
2977 sizeof(ipv6_route_table_template),
2978 GFP_KERNEL);
2979
2980 if (table) {
2981 table[0].data = &net->ipv6.sysctl.flush_delay;
2982 table[0].extra1 = net;
2983 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2984 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2985 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2986 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2987 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2988 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2989 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2990 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2991 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2992
2993 /* Don't export sysctls to unprivileged users */
2994 if (net->user_ns != &init_user_ns)
2995 table[0].procname = NULL;
2996 }
2997
2998 return table;
2999}
3000#endif
3001
3002static int __net_init ip6_route_net_init(struct net *net)
3003{
3004 int ret = -ENOMEM;
3005
3006 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3007 sizeof(net->ipv6.ip6_dst_ops));
3008
3009 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3010 goto out_ip6_dst_ops;
3011
3012 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3013 sizeof(*net->ipv6.ip6_null_entry),
3014 GFP_KERNEL);
3015 if (!net->ipv6.ip6_null_entry)
3016 goto out_ip6_dst_entries;
3017 net->ipv6.ip6_null_entry->dst.path =
3018 (struct dst_entry *)net->ipv6.ip6_null_entry;
3019 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3020 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3021 ip6_template_metrics, true);
3022
3023#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3024 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3025 sizeof(*net->ipv6.ip6_prohibit_entry),
3026 GFP_KERNEL);
3027 if (!net->ipv6.ip6_prohibit_entry)
3028 goto out_ip6_null_entry;
3029 net->ipv6.ip6_prohibit_entry->dst.path =
3030 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3031 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3032 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3033 ip6_template_metrics, true);
3034
3035 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3036 sizeof(*net->ipv6.ip6_blk_hole_entry),
3037 GFP_KERNEL);
3038 if (!net->ipv6.ip6_blk_hole_entry)
3039 goto out_ip6_prohibit_entry;
3040 net->ipv6.ip6_blk_hole_entry->dst.path =
3041 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3042 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3043 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3044 ip6_template_metrics, true);
3045#endif
3046
3047 net->ipv6.sysctl.flush_delay = 0;
3048 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3049 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3050 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3051 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3052 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3053 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3054 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3055
3056 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3057
3058 ret = 0;
3059out:
3060 return ret;
3061
3062#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3063out_ip6_prohibit_entry:
3064 kfree(net->ipv6.ip6_prohibit_entry);
3065out_ip6_null_entry:
3066 kfree(net->ipv6.ip6_null_entry);
3067#endif
3068out_ip6_dst_entries:
3069 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3070out_ip6_dst_ops:
3071 goto out;
3072}
3073
3074static void __net_exit ip6_route_net_exit(struct net *net)
3075{
3076 kfree(net->ipv6.ip6_null_entry);
3077#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3078 kfree(net->ipv6.ip6_prohibit_entry);
3079 kfree(net->ipv6.ip6_blk_hole_entry);
3080#endif
3081 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3082}
3083
3084static int __net_init ip6_route_net_init_late(struct net *net)
3085{
3086#ifdef CONFIG_PROC_FS
3087 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3088 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3089#endif
3090 return 0;
3091}
3092
3093static void __net_exit ip6_route_net_exit_late(struct net *net)
3094{
3095#ifdef CONFIG_PROC_FS
3096 remove_proc_entry("ipv6_route", net->proc_net);
3097 remove_proc_entry("rt6_stats", net->proc_net);
3098#endif
3099}
3100
3101static struct pernet_operations ip6_route_net_ops = {
3102 .init = ip6_route_net_init,
3103 .exit = ip6_route_net_exit,
3104};
3105
3106static int __net_init ipv6_inetpeer_init(struct net *net)
3107{
3108 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3109
3110 if (!bp)
3111 return -ENOMEM;
3112 inet_peer_base_init(bp);
3113 net->ipv6.peers = bp;
3114 return 0;
3115}
3116
3117static void __net_exit ipv6_inetpeer_exit(struct net *net)
3118{
3119 struct inet_peer_base *bp = net->ipv6.peers;
3120
3121 net->ipv6.peers = NULL;
3122 inetpeer_invalidate_tree(bp);
3123 kfree(bp);
3124}
3125
3126static struct pernet_operations ipv6_inetpeer_ops = {
3127 .init = ipv6_inetpeer_init,
3128 .exit = ipv6_inetpeer_exit,
3129};
3130
3131static struct pernet_operations ip6_route_net_late_ops = {
3132 .init = ip6_route_net_init_late,
3133 .exit = ip6_route_net_exit_late,
3134};
3135
3136static struct notifier_block ip6_route_dev_notifier = {
3137 .notifier_call = ip6_route_dev_notify,
3138 .priority = 0,
3139};
3140
3141int __init ip6_route_init(void)
3142{
3143 int ret;
3144
3145 ret = -ENOMEM;
3146 ip6_dst_ops_template.kmem_cachep =
3147 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3148 SLAB_HWCACHE_ALIGN, NULL);
3149 if (!ip6_dst_ops_template.kmem_cachep)
3150 goto out;
3151
3152 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3153 if (ret)
3154 goto out_kmem_cache;
3155
3156 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3157 if (ret)
3158 goto out_dst_entries;
3159
3160 ret = register_pernet_subsys(&ip6_route_net_ops);
3161 if (ret)
3162 goto out_register_inetpeer;
3163
3164 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3165
3166 /* Registering of the loopback is done before this portion of code,
3167 * the loopback reference in rt6_info will not be taken, do it
3168 * manually for init_net */
3169 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3170 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3171 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3172 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3173 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3174 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3175 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3176 #endif
3177 ret = fib6_init();
3178 if (ret)
3179 goto out_register_subsys;
3180
3181 ret = xfrm6_init();
3182 if (ret)
3183 goto out_fib6_init;
3184
3185 ret = fib6_rules_init();
3186 if (ret)
3187 goto xfrm6_init;
3188
3189 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3190 if (ret)
3191 goto fib6_rules_init;
3192
3193 ret = -ENOBUFS;
3194 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3195 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3196 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3197 goto out_register_late_subsys;
3198
3199 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3200 if (ret)
3201 goto out_register_late_subsys;
3202
3203out:
3204 return ret;
3205
3206out_register_late_subsys:
3207 unregister_pernet_subsys(&ip6_route_net_late_ops);
3208fib6_rules_init:
3209 fib6_rules_cleanup();
3210xfrm6_init:
3211 xfrm6_fini();
3212out_fib6_init:
3213 fib6_gc_cleanup();
3214out_register_subsys:
3215 unregister_pernet_subsys(&ip6_route_net_ops);
3216out_register_inetpeer:
3217 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3218out_dst_entries:
3219 dst_entries_destroy(&ip6_dst_blackhole_ops);
3220out_kmem_cache:
3221 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3222 goto out;
3223}
3224
3225void ip6_route_cleanup(void)
3226{
3227 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3228 unregister_pernet_subsys(&ip6_route_net_late_ops);
3229 fib6_rules_cleanup();
3230 xfrm6_fini();
3231 fib6_gc_cleanup();
3232 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3233 unregister_pernet_subsys(&ip6_route_net_ops);
3234 dst_entries_destroy(&ip6_dst_blackhole_ops);
3235 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3236}