?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
unix_diag: Write it into kbuild
[linux-2.6/libata-dev.git] / net / ipv6 / route.c
blobad438546d9159c7499a8a14b7bcf37e1583007b1
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/export.h>
30 #include <linux/types.h>
31 #include <linux/times.h>
32 #include <linux/socket.h>
33 #include <linux/sockios.h>
34 #include <linux/net.h>
35 #include <linux/route.h>
36 #include <linux/netdevice.h>
37 #include <linux/in6.h>
38 #include <linux/mroute6.h>
39 #include <linux/init.h>
40 #include <linux/if_arp.h>
41 #include <linux/proc_fs.h>
42 #include <linux/seq_file.h>
43 #include <linux/nsproxy.h>
44 #include <linux/slab.h>
45 #include <net/net_namespace.h>
46 #include <net/snmp.h>
47 #include <net/ipv6.h>
48 #include <net/ip6_fib.h>
49 #include <net/ip6_route.h>
50 #include <net/ndisc.h>
51 #include <net/addrconf.h>
52 #include <net/tcp.h>
53 #include <linux/rtnetlink.h>
54 #include <net/dst.h>
55 #include <net/xfrm.h>
56 #include <net/netevent.h>
57 #include <net/netlink.h>
58
59 #include <asm/uaccess.h>
60
61 #ifdef CONFIG_SYSCTL
62 #include <linux/sysctl.h>
63 #endif
64
65 /* Set to 3 to get tracing. */
66 #define RT6_DEBUG 2
67
68 #if RT6_DEBUG >= 3
69 #define RDBG(x) printk x
70 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
71 #else
72 #define RDBG(x)
73 #define RT6_TRACE(x...) do { ; } while (0)
74 #endif
75
76 static struct rt6_info *ip6_rt_copy(const struct rt6_info *ort,
77 const struct in6_addr *dest);
78 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
79 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
80 static unsigned int ip6_mtu(const struct dst_entry *dst);
81 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
82 static void ip6_dst_destroy(struct dst_entry *);
83 static void ip6_dst_ifdown(struct dst_entry *,
84 struct net_device *dev, int how);
85 static int ip6_dst_gc(struct dst_ops *ops);
86
87 static int ip6_pkt_discard(struct sk_buff *skb);
88 static int ip6_pkt_discard_out(struct sk_buff *skb);
89 static void ip6_link_failure(struct sk_buff *skb);
90 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
91
92 #ifdef CONFIG_IPV6_ROUTE_INFO
93 static struct rt6_info *rt6_add_route_info(struct net *net,
94 const struct in6_addr *prefix, int prefixlen,
95 const struct in6_addr *gwaddr, int ifindex,
96 unsigned pref);
97 static struct rt6_info *rt6_get_route_info(struct net *net,
98 const struct in6_addr *prefix, int prefixlen,
99 const struct in6_addr *gwaddr, int ifindex);
100 #endif
101
102 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
103 {
104 struct rt6_info *rt = (struct rt6_info *) dst;
105 struct inet_peer *peer;
106 u32 *p = NULL;
107
108 if (!(rt->dst.flags & DST_HOST))
109 return NULL;
110
111 if (!rt->rt6i_peer)
112 rt6_bind_peer(rt, 1);
113
114 peer = rt->rt6i_peer;
115 if (peer) {
116 u32 *old_p = __DST_METRICS_PTR(old);
117 unsigned long prev, new;
118
119 p = peer->metrics;
120 if (inet_metrics_new(peer))
121 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
122
123 new = (unsigned long) p;
124 prev = cmpxchg(&dst->_metrics, old, new);
125
126 if (prev != old) {
127 p = __DST_METRICS_PTR(prev);
128 if (prev & DST_METRICS_READ_ONLY)
129 p = NULL;
130 }
131 }
132 return p;
133 }
134
135 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst, const void *daddr)
136 {
137 return __neigh_lookup_errno(&nd_tbl, daddr, dst->dev);
138 }
139
140 static struct dst_ops ip6_dst_ops_template = {
141 .family = AF_INET6,
142 .protocol = cpu_to_be16(ETH_P_IPV6),
143 .gc = ip6_dst_gc,
144 .gc_thresh = 1024,
145 .check = ip6_dst_check,
146 .default_advmss = ip6_default_advmss,
147 .mtu = ip6_mtu,
148 .cow_metrics = ipv6_cow_metrics,
149 .destroy = ip6_dst_destroy,
150 .ifdown = ip6_dst_ifdown,
151 .negative_advice = ip6_negative_advice,
152 .link_failure = ip6_link_failure,
153 .update_pmtu = ip6_rt_update_pmtu,
154 .local_out = __ip6_local_out,
155 .neigh_lookup = ip6_neigh_lookup,
156 };
157
158 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
159 {
160 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
161
162 return mtu ? : dst->dev->mtu;
163 }
164
165 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
166 {
167 }
168
169 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
170 unsigned long old)
171 {
172 return NULL;
173 }
174
175 static struct dst_ops ip6_dst_blackhole_ops = {
176 .family = AF_INET6,
177 .protocol = cpu_to_be16(ETH_P_IPV6),
178 .destroy = ip6_dst_destroy,
179 .check = ip6_dst_check,
180 .mtu = ip6_blackhole_mtu,
181 .default_advmss = ip6_default_advmss,
182 .update_pmtu = ip6_rt_blackhole_update_pmtu,
183 .cow_metrics = ip6_rt_blackhole_cow_metrics,
184 .neigh_lookup = ip6_neigh_lookup,
185 };
186
187 static const u32 ip6_template_metrics[RTAX_MAX] = {
188 [RTAX_HOPLIMIT - 1] = 255,
189 };
190
191 static struct rt6_info ip6_null_entry_template = {
192 .dst = {
193 .__refcnt = ATOMIC_INIT(1),
194 .__use = 1,
195 .obsolete = -1,
196 .error = -ENETUNREACH,
197 .input = ip6_pkt_discard,
198 .output = ip6_pkt_discard_out,
199 },
200 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
201 .rt6i_protocol = RTPROT_KERNEL,
202 .rt6i_metric = ~(u32) 0,
203 .rt6i_ref = ATOMIC_INIT(1),
204 };
205
206 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
207
208 static int ip6_pkt_prohibit(struct sk_buff *skb);
209 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
210
211 static struct rt6_info ip6_prohibit_entry_template = {
212 .dst = {
213 .__refcnt = ATOMIC_INIT(1),
214 .__use = 1,
215 .obsolete = -1,
216 .error = -EACCES,
217 .input = ip6_pkt_prohibit,
218 .output = ip6_pkt_prohibit_out,
219 },
220 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
221 .rt6i_protocol = RTPROT_KERNEL,
222 .rt6i_metric = ~(u32) 0,
223 .rt6i_ref = ATOMIC_INIT(1),
224 };
225
226 static struct rt6_info ip6_blk_hole_entry_template = {
227 .dst = {
228 .__refcnt = ATOMIC_INIT(1),
229 .__use = 1,
230 .obsolete = -1,
231 .error = -EINVAL,
232 .input = dst_discard,
233 .output = dst_discard,
234 },
235 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
236 .rt6i_protocol = RTPROT_KERNEL,
237 .rt6i_metric = ~(u32) 0,
238 .rt6i_ref = ATOMIC_INIT(1),
239 };
240
241 #endif
242
243 /* allocate dst with ip6_dst_ops */
244 static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops,
245 struct net_device *dev,
246 int flags)
247 {
248 struct rt6_info *rt = dst_alloc(ops, dev, 0, 0, flags);
249
250 if (rt)
251 memset(&rt->rt6i_table, 0,
252 sizeof(*rt) - sizeof(struct dst_entry));
253
254 return rt;
255 }
256
257 static void ip6_dst_destroy(struct dst_entry *dst)
258 {
259 struct rt6_info *rt = (struct rt6_info *)dst;
260 struct inet6_dev *idev = rt->rt6i_idev;
261 struct inet_peer *peer = rt->rt6i_peer;
262
263 if (!(rt->dst.flags & DST_HOST))
264 dst_destroy_metrics_generic(dst);
265
266 if (idev) {
267 rt->rt6i_idev = NULL;
268 in6_dev_put(idev);
269 }
270 if (peer) {
271 rt->rt6i_peer = NULL;
272 inet_putpeer(peer);
273 }
274 }
275
276 static atomic_t __rt6_peer_genid = ATOMIC_INIT(0);
277
278 static u32 rt6_peer_genid(void)
279 {
280 return atomic_read(&__rt6_peer_genid);
281 }
282
283 void rt6_bind_peer(struct rt6_info *rt, int create)
284 {
285 struct inet_peer *peer;
286
287 peer = inet_getpeer_v6(&rt->rt6i_dst.addr, create);
288 if (peer && cmpxchg(&rt->rt6i_peer, NULL, peer) != NULL)
289 inet_putpeer(peer);
290 else
291 rt->rt6i_peer_genid = rt6_peer_genid();
292 }
293
294 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
295 int how)
296 {
297 struct rt6_info *rt = (struct rt6_info *)dst;
298 struct inet6_dev *idev = rt->rt6i_idev;
299 struct net_device *loopback_dev =
300 dev_net(dev)->loopback_dev;
301
302 if (dev != loopback_dev && idev && idev->dev == dev) {
303 struct inet6_dev *loopback_idev =
304 in6_dev_get(loopback_dev);
305 if (loopback_idev) {
306 rt->rt6i_idev = loopback_idev;
307 in6_dev_put(idev);
308 }
309 }
310 }
311
312 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
313 {
314 return (rt->rt6i_flags & RTF_EXPIRES) &&
315 time_after(jiffies, rt->rt6i_expires);
316 }
317
318 static inline int rt6_need_strict(const struct in6_addr *daddr)
319 {
320 return ipv6_addr_type(daddr) &
321 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
322 }
323
324 /*
325 * Route lookup. Any table->tb6_lock is implied.
326 */
327
328 static inline struct rt6_info *rt6_device_match(struct net *net,
329 struct rt6_info *rt,
330 const struct in6_addr *saddr,
331 int oif,
332 int flags)
333 {
334 struct rt6_info *local = NULL;
335 struct rt6_info *sprt;
336
337 if (!oif && ipv6_addr_any(saddr))
338 goto out;
339
340 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
341 struct net_device *dev = sprt->rt6i_dev;
342
343 if (oif) {
344 if (dev->ifindex == oif)
345 return sprt;
346 if (dev->flags & IFF_LOOPBACK) {
347 if (!sprt->rt6i_idev ||
348 sprt->rt6i_idev->dev->ifindex != oif) {
349 if (flags & RT6_LOOKUP_F_IFACE && oif)
350 continue;
351 if (local && (!oif ||
352 local->rt6i_idev->dev->ifindex == oif))
353 continue;
354 }
355 local = sprt;
356 }
357 } else {
358 if (ipv6_chk_addr(net, saddr, dev,
359 flags & RT6_LOOKUP_F_IFACE))
360 return sprt;
361 }
362 }
363
364 if (oif) {
365 if (local)
366 return local;
367
368 if (flags & RT6_LOOKUP_F_IFACE)
369 return net->ipv6.ip6_null_entry;
370 }
371 out:
372 return rt;
373 }
374
375 #ifdef CONFIG_IPV6_ROUTER_PREF
376 static void rt6_probe(struct rt6_info *rt)
377 {
378 struct neighbour *neigh;
379 /*
380 * Okay, this does not seem to be appropriate
381 * for now, however, we need to check if it
382 * is really so; aka Router Reachability Probing.
383 *
384 * Router Reachability Probe MUST be rate-limited
385 * to no more than one per minute.
386 */
387 rcu_read_lock();
388 neigh = rt ? dst_get_neighbour_noref(&rt->dst) : NULL;
389 if (!neigh || (neigh->nud_state & NUD_VALID))
390 goto out;
391 read_lock_bh(&neigh->lock);
392 if (!(neigh->nud_state & NUD_VALID) &&
393 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
394 struct in6_addr mcaddr;
395 struct in6_addr *target;
396
397 neigh->updated = jiffies;
398 read_unlock_bh(&neigh->lock);
399
400 target = (struct in6_addr *)&neigh->primary_key;
401 addrconf_addr_solict_mult(target, &mcaddr);
402 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
403 } else {
404 read_unlock_bh(&neigh->lock);
405 }
406 out:
407 rcu_read_unlock();
408 }
409 #else
410 static inline void rt6_probe(struct rt6_info *rt)
411 {
412 }
413 #endif
414
415 /*
416 * Default Router Selection (RFC 2461 6.3.6)
417 */
418 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
419 {
420 struct net_device *dev = rt->rt6i_dev;
421 if (!oif || dev->ifindex == oif)
422 return 2;
423 if ((dev->flags & IFF_LOOPBACK) &&
424 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
425 return 1;
426 return 0;
427 }
428
429 static inline int rt6_check_neigh(struct rt6_info *rt)
430 {
431 struct neighbour *neigh;
432 int m;
433
434 rcu_read_lock();
435 neigh = dst_get_neighbour_noref(&rt->dst);
436 if (rt->rt6i_flags & RTF_NONEXTHOP ||
437 !(rt->rt6i_flags & RTF_GATEWAY))
438 m = 1;
439 else if (neigh) {
440 read_lock_bh(&neigh->lock);
441 if (neigh->nud_state & NUD_VALID)
442 m = 2;
443 #ifdef CONFIG_IPV6_ROUTER_PREF
444 else if (neigh->nud_state & NUD_FAILED)
445 m = 0;
446 #endif
447 else
448 m = 1;
449 read_unlock_bh(&neigh->lock);
450 } else
451 m = 0;
452 rcu_read_unlock();
453 return m;
454 }
455
456 static int rt6_score_route(struct rt6_info *rt, int oif,
457 int strict)
458 {
459 int m, n;
460
461 m = rt6_check_dev(rt, oif);
462 if (!m && (strict & RT6_LOOKUP_F_IFACE))
463 return -1;
464 #ifdef CONFIG_IPV6_ROUTER_PREF
465 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
466 #endif
467 n = rt6_check_neigh(rt);
468 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
469 return -1;
470 return m;
471 }
472
473 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
474 int *mpri, struct rt6_info *match)
475 {
476 int m;
477
478 if (rt6_check_expired(rt))
479 goto out;
480
481 m = rt6_score_route(rt, oif, strict);
482 if (m < 0)
483 goto out;
484
485 if (m > *mpri) {
486 if (strict & RT6_LOOKUP_F_REACHABLE)
487 rt6_probe(match);
488 *mpri = m;
489 match = rt;
490 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
491 rt6_probe(rt);
492 }
493
494 out:
495 return match;
496 }
497
498 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
499 struct rt6_info *rr_head,
500 u32 metric, int oif, int strict)
501 {
502 struct rt6_info *rt, *match;
503 int mpri = -1;
504
505 match = NULL;
506 for (rt = rr_head; rt && rt->rt6i_metric == metric;
507 rt = rt->dst.rt6_next)
508 match = find_match(rt, oif, strict, &mpri, match);
509 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
510 rt = rt->dst.rt6_next)
511 match = find_match(rt, oif, strict, &mpri, match);
512
513 return match;
514 }
515
516 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
517 {
518 struct rt6_info *match, *rt0;
519 struct net *net;
520
521 RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
522 __func__, fn->leaf, oif);
523
524 rt0 = fn->rr_ptr;
525 if (!rt0)
526 fn->rr_ptr = rt0 = fn->leaf;
527
528 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
529
530 if (!match &&
531 (strict & RT6_LOOKUP_F_REACHABLE)) {
532 struct rt6_info *next = rt0->dst.rt6_next;
533
534 /* no entries matched; do round-robin */
535 if (!next || next->rt6i_metric != rt0->rt6i_metric)
536 next = fn->leaf;
537
538 if (next != rt0)
539 fn->rr_ptr = next;
540 }
541
542 RT6_TRACE("%s() => %p\n",
543 __func__, match);
544
545 net = dev_net(rt0->rt6i_dev);
546 return match ? match : net->ipv6.ip6_null_entry;
547 }
548
549 #ifdef CONFIG_IPV6_ROUTE_INFO
550 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
551 const struct in6_addr *gwaddr)
552 {
553 struct net *net = dev_net(dev);
554 struct route_info *rinfo = (struct route_info *) opt;
555 struct in6_addr prefix_buf, *prefix;
556 unsigned int pref;
557 unsigned long lifetime;
558 struct rt6_info *rt;
559
560 if (len < sizeof(struct route_info)) {
561 return -EINVAL;
562 }
563
564 /* Sanity check for prefix_len and length */
565 if (rinfo->length > 3) {
566 return -EINVAL;
567 } else if (rinfo->prefix_len > 128) {
568 return -EINVAL;
569 } else if (rinfo->prefix_len > 64) {
570 if (rinfo->length < 2) {
571 return -EINVAL;
572 }
573 } else if (rinfo->prefix_len > 0) {
574 if (rinfo->length < 1) {
575 return -EINVAL;
576 }
577 }
578
579 pref = rinfo->route_pref;
580 if (pref == ICMPV6_ROUTER_PREF_INVALID)
581 return -EINVAL;
582
583 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
584
585 if (rinfo->length == 3)
586 prefix = (struct in6_addr *)rinfo->prefix;
587 else {
588 /* this function is safe */
589 ipv6_addr_prefix(&prefix_buf,
590 (struct in6_addr *)rinfo->prefix,
591 rinfo->prefix_len);
592 prefix = &prefix_buf;
593 }
594
595 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
596 dev->ifindex);
597
598 if (rt && !lifetime) {
599 ip6_del_rt(rt);
600 rt = NULL;
601 }
602
603 if (!rt && lifetime)
604 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
605 pref);
606 else if (rt)
607 rt->rt6i_flags = RTF_ROUTEINFO |
608 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
609
610 if (rt) {
611 if (!addrconf_finite_timeout(lifetime)) {
612 rt->rt6i_flags &= ~RTF_EXPIRES;
613 } else {
614 rt->rt6i_expires = jiffies + HZ * lifetime;
615 rt->rt6i_flags |= RTF_EXPIRES;
616 }
617 dst_release(&rt->dst);
618 }
619 return 0;
620 }
621 #endif
622
623 #define BACKTRACK(__net, saddr) \
624 do { \
625 if (rt == __net->ipv6.ip6_null_entry) { \
626 struct fib6_node *pn; \
627 while (1) { \
628 if (fn->fn_flags & RTN_TL_ROOT) \
629 goto out; \
630 pn = fn->parent; \
631 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
632 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
633 else \
634 fn = pn; \
635 if (fn->fn_flags & RTN_RTINFO) \
636 goto restart; \
637 } \
638 } \
639 } while (0)
640
641 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
642 struct fib6_table *table,
643 struct flowi6 *fl6, int flags)
644 {
645 struct fib6_node *fn;
646 struct rt6_info *rt;
647
648 read_lock_bh(&table->tb6_lock);
649 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
650 restart:
651 rt = fn->leaf;
652 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
653 BACKTRACK(net, &fl6->saddr);
654 out:
655 dst_use(&rt->dst, jiffies);
656 read_unlock_bh(&table->tb6_lock);
657 return rt;
658
659 }
660
661 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
662 const struct in6_addr *saddr, int oif, int strict)
663 {
664 struct flowi6 fl6 = {
665 .flowi6_oif = oif,
666 .daddr = *daddr,
667 };
668 struct dst_entry *dst;
669 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
670
671 if (saddr) {
672 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
673 flags |= RT6_LOOKUP_F_HAS_SADDR;
674 }
675
676 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
677 if (dst->error == 0)
678 return (struct rt6_info *) dst;
679
680 dst_release(dst);
681
682 return NULL;
683 }
684
685 EXPORT_SYMBOL(rt6_lookup);
686
687 /* ip6_ins_rt is called with FREE table->tb6_lock.
688 It takes new route entry, the addition fails by any reason the
689 route is freed. In any case, if caller does not hold it, it may
690 be destroyed.
691 */
692
693 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
694 {
695 int err;
696 struct fib6_table *table;
697
698 table = rt->rt6i_table;
699 write_lock_bh(&table->tb6_lock);
700 err = fib6_add(&table->tb6_root, rt, info);
701 write_unlock_bh(&table->tb6_lock);
702
703 return err;
704 }
705
706 int ip6_ins_rt(struct rt6_info *rt)
707 {
708 struct nl_info info = {
709 .nl_net = dev_net(rt->rt6i_dev),
710 };
711 return __ip6_ins_rt(rt, &info);
712 }
713
714 static struct rt6_info *rt6_alloc_cow(const struct rt6_info *ort,
715 const struct in6_addr *daddr,
716 const struct in6_addr *saddr)
717 {
718 struct rt6_info *rt;
719
720 /*
721 * Clone the route.
722 */
723
724 rt = ip6_rt_copy(ort, daddr);
725
726 if (rt) {
727 struct neighbour *neigh;
728 int attempts = !in_softirq();
729
730 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
731 if (ort->rt6i_dst.plen != 128 &&
732 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
733 rt->rt6i_flags |= RTF_ANYCAST;
734 rt->rt6i_gateway = *daddr;
735 }
736
737 rt->rt6i_flags |= RTF_CACHE;
738
739 #ifdef CONFIG_IPV6_SUBTREES
740 if (rt->rt6i_src.plen && saddr) {
741 rt->rt6i_src.addr = *saddr;
742 rt->rt6i_src.plen = 128;
743 }
744 #endif
745
746 retry:
747 neigh = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway,
748 rt->rt6i_dev);
749 if (IS_ERR(neigh)) {
750 struct net *net = dev_net(rt->rt6i_dev);
751 int saved_rt_min_interval =
752 net->ipv6.sysctl.ip6_rt_gc_min_interval;
753 int saved_rt_elasticity =
754 net->ipv6.sysctl.ip6_rt_gc_elasticity;
755
756 if (attempts-- > 0) {
757 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
758 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
759
760 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
761
762 net->ipv6.sysctl.ip6_rt_gc_elasticity =
763 saved_rt_elasticity;
764 net->ipv6.sysctl.ip6_rt_gc_min_interval =
765 saved_rt_min_interval;
766 goto retry;
767 }
768
769 if (net_ratelimit())
770 printk(KERN_WARNING
771 "ipv6: Neighbour table overflow.\n");
772 dst_free(&rt->dst);
773 return NULL;
774 }
775 dst_set_neighbour(&rt->dst, neigh);
776
777 }
778
779 return rt;
780 }
781
782 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
783 const struct in6_addr *daddr)
784 {
785 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
786
787 if (rt) {
788 rt->rt6i_flags |= RTF_CACHE;
789 dst_set_neighbour(&rt->dst, neigh_clone(dst_get_neighbour_noref_raw(&ort->dst)));
790 }
791 return rt;
792 }
793
794 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
795 struct flowi6 *fl6, int flags)
796 {
797 struct fib6_node *fn;
798 struct rt6_info *rt, *nrt;
799 int strict = 0;
800 int attempts = 3;
801 int err;
802 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
803
804 strict |= flags & RT6_LOOKUP_F_IFACE;
805
806 relookup:
807 read_lock_bh(&table->tb6_lock);
808
809 restart_2:
810 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
811
812 restart:
813 rt = rt6_select(fn, oif, strict | reachable);
814
815 BACKTRACK(net, &fl6->saddr);
816 if (rt == net->ipv6.ip6_null_entry ||
817 rt->rt6i_flags & RTF_CACHE)
818 goto out;
819
820 dst_hold(&rt->dst);
821 read_unlock_bh(&table->tb6_lock);
822
823 if (!dst_get_neighbour_noref_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
824 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
825 else if (!(rt->dst.flags & DST_HOST))
826 nrt = rt6_alloc_clone(rt, &fl6->daddr);
827 else
828 goto out2;
829
830 dst_release(&rt->dst);
831 rt = nrt ? : net->ipv6.ip6_null_entry;
832
833 dst_hold(&rt->dst);
834 if (nrt) {
835 err = ip6_ins_rt(nrt);
836 if (!err)
837 goto out2;
838 }
839
840 if (--attempts <= 0)
841 goto out2;
842
843 /*
844 * Race condition! In the gap, when table->tb6_lock was
845 * released someone could insert this route. Relookup.
846 */
847 dst_release(&rt->dst);
848 goto relookup;
849
850 out:
851 if (reachable) {
852 reachable = 0;
853 goto restart_2;
854 }
855 dst_hold(&rt->dst);
856 read_unlock_bh(&table->tb6_lock);
857 out2:
858 rt->dst.lastuse = jiffies;
859 rt->dst.__use++;
860
861 return rt;
862 }
863
864 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
865 struct flowi6 *fl6, int flags)
866 {
867 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
868 }
869
870 void ip6_route_input(struct sk_buff *skb)
871 {
872 const struct ipv6hdr *iph = ipv6_hdr(skb);
873 struct net *net = dev_net(skb->dev);
874 int flags = RT6_LOOKUP_F_HAS_SADDR;
875 struct flowi6 fl6 = {
876 .flowi6_iif = skb->dev->ifindex,
877 .daddr = iph->daddr,
878 .saddr = iph->saddr,
879 .flowlabel = (* (__be32 *) iph) & IPV6_FLOWINFO_MASK,
880 .flowi6_mark = skb->mark,
881 .flowi6_proto = iph->nexthdr,
882 };
883
884 if (rt6_need_strict(&iph->daddr) && skb->dev->type != ARPHRD_PIMREG)
885 flags |= RT6_LOOKUP_F_IFACE;
886
887 skb_dst_set(skb, fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_input));
888 }
889
890 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
891 struct flowi6 *fl6, int flags)
892 {
893 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
894 }
895
896 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
897 struct flowi6 *fl6)
898 {
899 int flags = 0;
900
901 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
902 flags |= RT6_LOOKUP_F_IFACE;
903
904 if (!ipv6_addr_any(&fl6->saddr))
905 flags |= RT6_LOOKUP_F_HAS_SADDR;
906 else if (sk)
907 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
908
909 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
910 }
911
912 EXPORT_SYMBOL(ip6_route_output);
913
914 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
915 {
916 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
917 struct dst_entry *new = NULL;
918
919 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, 0, 0);
920 if (rt) {
921 memset(&rt->rt6i_table, 0, sizeof(*rt) - sizeof(struct dst_entry));
922
923 new = &rt->dst;
924
925 new->__use = 1;
926 new->input = dst_discard;
927 new->output = dst_discard;
928
929 if (dst_metrics_read_only(&ort->dst))
930 new->_metrics = ort->dst._metrics;
931 else
932 dst_copy_metrics(new, &ort->dst);
933 rt->rt6i_idev = ort->rt6i_idev;
934 if (rt->rt6i_idev)
935 in6_dev_hold(rt->rt6i_idev);
936 rt->rt6i_expires = 0;
937
938 rt->rt6i_gateway = ort->rt6i_gateway;
939 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
940 rt->rt6i_metric = 0;
941
942 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
943 #ifdef CONFIG_IPV6_SUBTREES
944 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
945 #endif
946
947 dst_free(new);
948 }
949
950 dst_release(dst_orig);
951 return new ? new : ERR_PTR(-ENOMEM);
952 }
953
954 /*
955 * Destination cache support functions
956 */
957
958 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
959 {
960 struct rt6_info *rt;
961
962 rt = (struct rt6_info *) dst;
963
964 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie)) {
965 if (rt->rt6i_peer_genid != rt6_peer_genid()) {
966 if (!rt->rt6i_peer)
967 rt6_bind_peer(rt, 0);
968 rt->rt6i_peer_genid = rt6_peer_genid();
969 }
970 return dst;
971 }
972 return NULL;
973 }
974
975 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
976 {
977 struct rt6_info *rt = (struct rt6_info *) dst;
978
979 if (rt) {
980 if (rt->rt6i_flags & RTF_CACHE) {
981 if (rt6_check_expired(rt)) {
982 ip6_del_rt(rt);
983 dst = NULL;
984 }
985 } else {
986 dst_release(dst);
987 dst = NULL;
988 }
989 }
990 return dst;
991 }
992
993 static void ip6_link_failure(struct sk_buff *skb)
994 {
995 struct rt6_info *rt;
996
997 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
998
999 rt = (struct rt6_info *) skb_dst(skb);
1000 if (rt) {
1001 if (rt->rt6i_flags & RTF_CACHE) {
1002 dst_set_expires(&rt->dst, 0);
1003 rt->rt6i_flags |= RTF_EXPIRES;
1004 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1005 rt->rt6i_node->fn_sernum = -1;
1006 }
1007 }
1008
1009 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
1010 {
1011 struct rt6_info *rt6 = (struct rt6_info*)dst;
1012
1013 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1014 rt6->rt6i_flags |= RTF_MODIFIED;
1015 if (mtu < IPV6_MIN_MTU) {
1016 u32 features = dst_metric(dst, RTAX_FEATURES);
1017 mtu = IPV6_MIN_MTU;
1018 features |= RTAX_FEATURE_ALLFRAG;
1019 dst_metric_set(dst, RTAX_FEATURES, features);
1020 }
1021 dst_metric_set(dst, RTAX_MTU, mtu);
1022 }
1023 }
1024
1025 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1026 {
1027 struct net_device *dev = dst->dev;
1028 unsigned int mtu = dst_mtu(dst);
1029 struct net *net = dev_net(dev);
1030
1031 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1032
1033 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1034 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1035
1036 /*
1037 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1038 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1039 * IPV6_MAXPLEN is also valid and means: "any MSS,
1040 * rely only on pmtu discovery"
1041 */
1042 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1043 mtu = IPV6_MAXPLEN;
1044 return mtu;
1045 }
1046
1047 static unsigned int ip6_mtu(const struct dst_entry *dst)
1048 {
1049 struct inet6_dev *idev;
1050 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1051
1052 if (mtu)
1053 return mtu;
1054
1055 mtu = IPV6_MIN_MTU;
1056
1057 rcu_read_lock();
1058 idev = __in6_dev_get(dst->dev);
1059 if (idev)
1060 mtu = idev->cnf.mtu6;
1061 rcu_read_unlock();
1062
1063 return mtu;
1064 }
1065
1066 static struct dst_entry *icmp6_dst_gc_list;
1067 static DEFINE_SPINLOCK(icmp6_dst_lock);
1068
1069 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1070 struct neighbour *neigh,
1071 struct flowi6 *fl6)
1072 {
1073 struct dst_entry *dst;
1074 struct rt6_info *rt;
1075 struct inet6_dev *idev = in6_dev_get(dev);
1076 struct net *net = dev_net(dev);
1077
1078 if (unlikely(!idev))
1079 return NULL;
1080
1081 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops, dev, 0);
1082 if (unlikely(!rt)) {
1083 in6_dev_put(idev);
1084 dst = ERR_PTR(-ENOMEM);
1085 goto out;
1086 }
1087
1088 if (neigh)
1089 neigh_hold(neigh);
1090 else {
1091 neigh = __neigh_lookup_errno(&nd_tbl, &fl6->daddr, dev);
1092 if (IS_ERR(neigh)) {
1093 dst_free(&rt->dst);
1094 return ERR_CAST(neigh);
1095 }
1096 }
1097
1098 rt->dst.flags |= DST_HOST;
1099 rt->dst.output = ip6_output;
1100 dst_set_neighbour(&rt->dst, neigh);
1101 atomic_set(&rt->dst.__refcnt, 1);
1102 rt->rt6i_dst.addr = fl6->daddr;
1103 rt->rt6i_dst.plen = 128;
1104 rt->rt6i_idev = idev;
1105 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 255);
1106
1107 spin_lock_bh(&icmp6_dst_lock);
1108 rt->dst.next = icmp6_dst_gc_list;
1109 icmp6_dst_gc_list = &rt->dst;
1110 spin_unlock_bh(&icmp6_dst_lock);
1111
1112 fib6_force_start_gc(net);
1113
1114 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1115
1116 out:
1117 return dst;
1118 }
1119
1120 int icmp6_dst_gc(void)
1121 {
1122 struct dst_entry *dst, **pprev;
1123 int more = 0;
1124
1125 spin_lock_bh(&icmp6_dst_lock);
1126 pprev = &icmp6_dst_gc_list;
1127
1128 while ((dst = *pprev) != NULL) {
1129 if (!atomic_read(&dst->__refcnt)) {
1130 *pprev = dst->next;
1131 dst_free(dst);
1132 } else {
1133 pprev = &dst->next;
1134 ++more;
1135 }
1136 }
1137
1138 spin_unlock_bh(&icmp6_dst_lock);
1139
1140 return more;
1141 }
1142
1143 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1144 void *arg)
1145 {
1146 struct dst_entry *dst, **pprev;
1147
1148 spin_lock_bh(&icmp6_dst_lock);
1149 pprev = &icmp6_dst_gc_list;
1150 while ((dst = *pprev) != NULL) {
1151 struct rt6_info *rt = (struct rt6_info *) dst;
1152 if (func(rt, arg)) {
1153 *pprev = dst->next;
1154 dst_free(dst);
1155 } else {
1156 pprev = &dst->next;
1157 }
1158 }
1159 spin_unlock_bh(&icmp6_dst_lock);
1160 }
1161
1162 static int ip6_dst_gc(struct dst_ops *ops)
1163 {
1164 unsigned long now = jiffies;
1165 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1166 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1167 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1168 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1169 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1170 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1171 int entries;
1172
1173 entries = dst_entries_get_fast(ops);
1174 if (time_after(rt_last_gc + rt_min_interval, now) &&
1175 entries <= rt_max_size)
1176 goto out;
1177
1178 net->ipv6.ip6_rt_gc_expire++;
1179 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1180 net->ipv6.ip6_rt_last_gc = now;
1181 entries = dst_entries_get_slow(ops);
1182 if (entries < ops->gc_thresh)
1183 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1184 out:
1185 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1186 return entries > rt_max_size;
1187 }
1188
1189 /* Clean host part of a prefix. Not necessary in radix tree,
1190 but results in cleaner routing tables.
1191
1192 Remove it only when all the things will work!
1193 */
1194
1195 int ip6_dst_hoplimit(struct dst_entry *dst)
1196 {
1197 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1198 if (hoplimit == 0) {
1199 struct net_device *dev = dst->dev;
1200 struct inet6_dev *idev;
1201
1202 rcu_read_lock();
1203 idev = __in6_dev_get(dev);
1204 if (idev)
1205 hoplimit = idev->cnf.hop_limit;
1206 else
1207 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1208 rcu_read_unlock();
1209 }
1210 return hoplimit;
1211 }
1212 EXPORT_SYMBOL(ip6_dst_hoplimit);
1213
1214 /*
1215 *
1216 */
1217
1218 int ip6_route_add(struct fib6_config *cfg)
1219 {
1220 int err;
1221 struct net *net = cfg->fc_nlinfo.nl_net;
1222 struct rt6_info *rt = NULL;
1223 struct net_device *dev = NULL;
1224 struct inet6_dev *idev = NULL;
1225 struct fib6_table *table;
1226 int addr_type;
1227
1228 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1229 return -EINVAL;
1230 #ifndef CONFIG_IPV6_SUBTREES
1231 if (cfg->fc_src_len)
1232 return -EINVAL;
1233 #endif
1234 if (cfg->fc_ifindex) {
1235 err = -ENODEV;
1236 dev = dev_get_by_index(net, cfg->fc_ifindex);
1237 if (!dev)
1238 goto out;
1239 idev = in6_dev_get(dev);
1240 if (!idev)
1241 goto out;
1242 }
1243
1244 if (cfg->fc_metric == 0)
1245 cfg->fc_metric = IP6_RT_PRIO_USER;
1246
1247 err = -ENOBUFS;
1248 if (cfg->fc_nlinfo.nlh &&
1249 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1250 table = fib6_get_table(net, cfg->fc_table);
1251 if (!table) {
1252 printk(KERN_WARNING "IPv6: NLM_F_CREATE should be specified when creating new route\n");
1253 table = fib6_new_table(net, cfg->fc_table);
1254 }
1255 } else {
1256 table = fib6_new_table(net, cfg->fc_table);
1257 }
1258
1259 if (!table)
1260 goto out;
1261
1262 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops, NULL, DST_NOCOUNT);
1263
1264 if (!rt) {
1265 err = -ENOMEM;
1266 goto out;
1267 }
1268
1269 rt->dst.obsolete = -1;
1270 rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1271 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1272 0;
1273
1274 if (cfg->fc_protocol == RTPROT_UNSPEC)
1275 cfg->fc_protocol = RTPROT_BOOT;
1276 rt->rt6i_protocol = cfg->fc_protocol;
1277
1278 addr_type = ipv6_addr_type(&cfg->fc_dst);
1279
1280 if (addr_type & IPV6_ADDR_MULTICAST)
1281 rt->dst.input = ip6_mc_input;
1282 else if (cfg->fc_flags & RTF_LOCAL)
1283 rt->dst.input = ip6_input;
1284 else
1285 rt->dst.input = ip6_forward;
1286
1287 rt->dst.output = ip6_output;
1288
1289 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1290 rt->rt6i_dst.plen = cfg->fc_dst_len;
1291 if (rt->rt6i_dst.plen == 128)
1292 rt->dst.flags |= DST_HOST;
1293
1294 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1295 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1296 if (!metrics) {
1297 err = -ENOMEM;
1298 goto out;
1299 }
1300 dst_init_metrics(&rt->dst, metrics, 0);
1301 }
1302 #ifdef CONFIG_IPV6_SUBTREES
1303 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1304 rt->rt6i_src.plen = cfg->fc_src_len;
1305 #endif
1306
1307 rt->rt6i_metric = cfg->fc_metric;
1308
1309 /* We cannot add true routes via loopback here,
1310 they would result in kernel looping; promote them to reject routes
1311 */
1312 if ((cfg->fc_flags & RTF_REJECT) ||
1313 (dev && (dev->flags & IFF_LOOPBACK) &&
1314 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1315 !(cfg->fc_flags & RTF_LOCAL))) {
1316 /* hold loopback dev/idev if we haven't done so. */
1317 if (dev != net->loopback_dev) {
1318 if (dev) {
1319 dev_put(dev);
1320 in6_dev_put(idev);
1321 }
1322 dev = net->loopback_dev;
1323 dev_hold(dev);
1324 idev = in6_dev_get(dev);
1325 if (!idev) {
1326 err = -ENODEV;
1327 goto out;
1328 }
1329 }
1330 rt->dst.output = ip6_pkt_discard_out;
1331 rt->dst.input = ip6_pkt_discard;
1332 rt->dst.error = -ENETUNREACH;
1333 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1334 goto install_route;
1335 }
1336
1337 if (cfg->fc_flags & RTF_GATEWAY) {
1338 const struct in6_addr *gw_addr;
1339 int gwa_type;
1340
1341 gw_addr = &cfg->fc_gateway;
1342 rt->rt6i_gateway = *gw_addr;
1343 gwa_type = ipv6_addr_type(gw_addr);
1344
1345 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1346 struct rt6_info *grt;
1347
1348 /* IPv6 strictly inhibits using not link-local
1349 addresses as nexthop address.
1350 Otherwise, router will not able to send redirects.
1351 It is very good, but in some (rare!) circumstances
1352 (SIT, PtP, NBMA NOARP links) it is handy to allow
1353 some exceptions. --ANK
1354 */
1355 err = -EINVAL;
1356 if (!(gwa_type & IPV6_ADDR_UNICAST))
1357 goto out;
1358
1359 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1360
1361 err = -EHOSTUNREACH;
1362 if (!grt)
1363 goto out;
1364 if (dev) {
1365 if (dev != grt->rt6i_dev) {
1366 dst_release(&grt->dst);
1367 goto out;
1368 }
1369 } else {
1370 dev = grt->rt6i_dev;
1371 idev = grt->rt6i_idev;
1372 dev_hold(dev);
1373 in6_dev_hold(grt->rt6i_idev);
1374 }
1375 if (!(grt->rt6i_flags & RTF_GATEWAY))
1376 err = 0;
1377 dst_release(&grt->dst);
1378
1379 if (err)
1380 goto out;
1381 }
1382 err = -EINVAL;
1383 if (!dev || (dev->flags & IFF_LOOPBACK))
1384 goto out;
1385 }
1386
1387 err = -ENODEV;
1388 if (!dev)
1389 goto out;
1390
1391 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1392 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1393 err = -EINVAL;
1394 goto out;
1395 }
1396 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1397 rt->rt6i_prefsrc.plen = 128;
1398 } else
1399 rt->rt6i_prefsrc.plen = 0;
1400
1401 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1402 struct neighbour *n = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1403 if (IS_ERR(n)) {
1404 err = PTR_ERR(n);
1405 goto out;
1406 }
1407 dst_set_neighbour(&rt->dst, n);
1408 }
1409
1410 rt->rt6i_flags = cfg->fc_flags;
1411
1412 install_route:
1413 if (cfg->fc_mx) {
1414 struct nlattr *nla;
1415 int remaining;
1416
1417 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1418 int type = nla_type(nla);
1419
1420 if (type) {
1421 if (type > RTAX_MAX) {
1422 err = -EINVAL;
1423 goto out;
1424 }
1425
1426 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1427 }
1428 }
1429 }
1430
1431 rt->dst.dev = dev;
1432 rt->rt6i_idev = idev;
1433 rt->rt6i_table = table;
1434
1435 cfg->fc_nlinfo.nl_net = dev_net(dev);
1436
1437 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1438
1439 out:
1440 if (dev)
1441 dev_put(dev);
1442 if (idev)
1443 in6_dev_put(idev);
1444 if (rt)
1445 dst_free(&rt->dst);
1446 return err;
1447 }
1448
1449 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1450 {
1451 int err;
1452 struct fib6_table *table;
1453 struct net *net = dev_net(rt->rt6i_dev);
1454
1455 if (rt == net->ipv6.ip6_null_entry)
1456 return -ENOENT;
1457
1458 table = rt->rt6i_table;
1459 write_lock_bh(&table->tb6_lock);
1460
1461 err = fib6_del(rt, info);
1462 dst_release(&rt->dst);
1463
1464 write_unlock_bh(&table->tb6_lock);
1465
1466 return err;
1467 }
1468
1469 int ip6_del_rt(struct rt6_info *rt)
1470 {
1471 struct nl_info info = {
1472 .nl_net = dev_net(rt->rt6i_dev),
1473 };
1474 return __ip6_del_rt(rt, &info);
1475 }
1476
1477 static int ip6_route_del(struct fib6_config *cfg)
1478 {
1479 struct fib6_table *table;
1480 struct fib6_node *fn;
1481 struct rt6_info *rt;
1482 int err = -ESRCH;
1483
1484 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1485 if (!table)
1486 return err;
1487
1488 read_lock_bh(&table->tb6_lock);
1489
1490 fn = fib6_locate(&table->tb6_root,
1491 &cfg->fc_dst, cfg->fc_dst_len,
1492 &cfg->fc_src, cfg->fc_src_len);
1493
1494 if (fn) {
1495 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1496 if (cfg->fc_ifindex &&
1497 (!rt->rt6i_dev ||
1498 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1499 continue;
1500 if (cfg->fc_flags & RTF_GATEWAY &&
1501 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1502 continue;
1503 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1504 continue;
1505 dst_hold(&rt->dst);
1506 read_unlock_bh(&table->tb6_lock);
1507
1508 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1509 }
1510 }
1511 read_unlock_bh(&table->tb6_lock);
1512
1513 return err;
1514 }
1515
1516 /*
1517 * Handle redirects
1518 */
1519 struct ip6rd_flowi {
1520 struct flowi6 fl6;
1521 struct in6_addr gateway;
1522 };
1523
1524 static struct rt6_info *__ip6_route_redirect(struct net *net,
1525 struct fib6_table *table,
1526 struct flowi6 *fl6,
1527 int flags)
1528 {
1529 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1530 struct rt6_info *rt;
1531 struct fib6_node *fn;
1532
1533 /*
1534 * Get the "current" route for this destination and
1535 * check if the redirect has come from approriate router.
1536 *
1537 * RFC 2461 specifies that redirects should only be
1538 * accepted if they come from the nexthop to the target.
1539 * Due to the way the routes are chosen, this notion
1540 * is a bit fuzzy and one might need to check all possible
1541 * routes.
1542 */
1543
1544 read_lock_bh(&table->tb6_lock);
1545 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1546 restart:
1547 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1548 /*
1549 * Current route is on-link; redirect is always invalid.
1550 *
1551 * Seems, previous statement is not true. It could
1552 * be node, which looks for us as on-link (f.e. proxy ndisc)
1553 * But then router serving it might decide, that we should
1554 * know truth 8)8) --ANK (980726).
1555 */
1556 if (rt6_check_expired(rt))
1557 continue;
1558 if (!(rt->rt6i_flags & RTF_GATEWAY))
1559 continue;
1560 if (fl6->flowi6_oif != rt->rt6i_dev->ifindex)
1561 continue;
1562 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1563 continue;
1564 break;
1565 }
1566
1567 if (!rt)
1568 rt = net->ipv6.ip6_null_entry;
1569 BACKTRACK(net, &fl6->saddr);
1570 out:
1571 dst_hold(&rt->dst);
1572
1573 read_unlock_bh(&table->tb6_lock);
1574
1575 return rt;
1576 };
1577
1578 static struct rt6_info *ip6_route_redirect(const struct in6_addr *dest,
1579 const struct in6_addr *src,
1580 const struct in6_addr *gateway,
1581 struct net_device *dev)
1582 {
1583 int flags = RT6_LOOKUP_F_HAS_SADDR;
1584 struct net *net = dev_net(dev);
1585 struct ip6rd_flowi rdfl = {
1586 .fl6 = {
1587 .flowi6_oif = dev->ifindex,
1588 .daddr = *dest,
1589 .saddr = *src,
1590 },
1591 };
1592
1593 rdfl.gateway = *gateway;
1594
1595 if (rt6_need_strict(dest))
1596 flags |= RT6_LOOKUP_F_IFACE;
1597
1598 return (struct rt6_info *)fib6_rule_lookup(net, &rdfl.fl6,
1599 flags, __ip6_route_redirect);
1600 }
1601
1602 void rt6_redirect(const struct in6_addr *dest, const struct in6_addr *src,
1603 const struct in6_addr *saddr,
1604 struct neighbour *neigh, u8 *lladdr, int on_link)
1605 {
1606 struct rt6_info *rt, *nrt = NULL;
1607 struct netevent_redirect netevent;
1608 struct net *net = dev_net(neigh->dev);
1609
1610 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1611
1612 if (rt == net->ipv6.ip6_null_entry) {
1613 if (net_ratelimit())
1614 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1615 "for redirect target\n");
1616 goto out;
1617 }
1618
1619 /*
1620 * We have finally decided to accept it.
1621 */
1622
1623 neigh_update(neigh, lladdr, NUD_STALE,
1624 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1625 NEIGH_UPDATE_F_OVERRIDE|
1626 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1627 NEIGH_UPDATE_F_ISROUTER))
1628 );
1629
1630 /*
1631 * Redirect received -> path was valid.
1632 * Look, redirects are sent only in response to data packets,
1633 * so that this nexthop apparently is reachable. --ANK
1634 */
1635 dst_confirm(&rt->dst);
1636
1637 /* Duplicate redirect: silently ignore. */
1638 if (neigh == dst_get_neighbour_noref_raw(&rt->dst))
1639 goto out;
1640
1641 nrt = ip6_rt_copy(rt, dest);
1642 if (!nrt)
1643 goto out;
1644
1645 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1646 if (on_link)
1647 nrt->rt6i_flags &= ~RTF_GATEWAY;
1648
1649 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1650 dst_set_neighbour(&nrt->dst, neigh_clone(neigh));
1651
1652 if (ip6_ins_rt(nrt))
1653 goto out;
1654
1655 netevent.old = &rt->dst;
1656 netevent.new = &nrt->dst;
1657 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1658
1659 if (rt->rt6i_flags & RTF_CACHE) {
1660 ip6_del_rt(rt);
1661 return;
1662 }
1663
1664 out:
1665 dst_release(&rt->dst);
1666 }
1667
1668 /*
1669 * Handle ICMP "packet too big" messages
1670 * i.e. Path MTU discovery
1671 */
1672
1673 static void rt6_do_pmtu_disc(const struct in6_addr *daddr, const struct in6_addr *saddr,
1674 struct net *net, u32 pmtu, int ifindex)
1675 {
1676 struct rt6_info *rt, *nrt;
1677 int allfrag = 0;
1678 again:
1679 rt = rt6_lookup(net, daddr, saddr, ifindex, 0);
1680 if (!rt)
1681 return;
1682
1683 if (rt6_check_expired(rt)) {
1684 ip6_del_rt(rt);
1685 goto again;
1686 }
1687
1688 if (pmtu >= dst_mtu(&rt->dst))
1689 goto out;
1690
1691 if (pmtu < IPV6_MIN_MTU) {
1692 /*
1693 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1694 * MTU (1280) and a fragment header should always be included
1695 * after a node receiving Too Big message reporting PMTU is
1696 * less than the IPv6 Minimum Link MTU.
1697 */
1698 pmtu = IPV6_MIN_MTU;
1699 allfrag = 1;
1700 }
1701
1702 /* New mtu received -> path was valid.
1703 They are sent only in response to data packets,
1704 so that this nexthop apparently is reachable. --ANK
1705 */
1706 dst_confirm(&rt->dst);
1707
1708 /* Host route. If it is static, it would be better
1709 not to override it, but add new one, so that
1710 when cache entry will expire old pmtu
1711 would return automatically.
1712 */
1713 if (rt->rt6i_flags & RTF_CACHE) {
1714 dst_metric_set(&rt->dst, RTAX_MTU, pmtu);
1715 if (allfrag) {
1716 u32 features = dst_metric(&rt->dst, RTAX_FEATURES);
1717 features |= RTAX_FEATURE_ALLFRAG;
1718 dst_metric_set(&rt->dst, RTAX_FEATURES, features);
1719 }
1720 dst_set_expires(&rt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1721 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1722 goto out;
1723 }
1724
1725 /* Network route.
1726 Two cases are possible:
1727 1. It is connected route. Action: COW
1728 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1729 */
1730 if (!dst_get_neighbour_noref_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
1731 nrt = rt6_alloc_cow(rt, daddr, saddr);
1732 else
1733 nrt = rt6_alloc_clone(rt, daddr);
1734
1735 if (nrt) {
1736 dst_metric_set(&nrt->dst, RTAX_MTU, pmtu);
1737 if (allfrag) {
1738 u32 features = dst_metric(&nrt->dst, RTAX_FEATURES);
1739 features |= RTAX_FEATURE_ALLFRAG;
1740 dst_metric_set(&nrt->dst, RTAX_FEATURES, features);
1741 }
1742
1743 /* According to RFC 1981, detecting PMTU increase shouldn't be
1744 * happened within 5 mins, the recommended timer is 10 mins.
1745 * Here this route expiration time is set to ip6_rt_mtu_expires
1746 * which is 10 mins. After 10 mins the decreased pmtu is expired
1747 * and detecting PMTU increase will be automatically happened.
1748 */
1749 dst_set_expires(&nrt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1750 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1751
1752 ip6_ins_rt(nrt);
1753 }
1754 out:
1755 dst_release(&rt->dst);
1756 }
1757
1758 void rt6_pmtu_discovery(const struct in6_addr *daddr, const struct in6_addr *saddr,
1759 struct net_device *dev, u32 pmtu)
1760 {
1761 struct net *net = dev_net(dev);
1762
1763 /*
1764 * RFC 1981 states that a node "MUST reduce the size of the packets it
1765 * is sending along the path" that caused the Packet Too Big message.
1766 * Since it's not possible in the general case to determine which
1767 * interface was used to send the original packet, we update the MTU
1768 * on the interface that will be used to send future packets. We also
1769 * update the MTU on the interface that received the Packet Too Big in
1770 * case the original packet was forced out that interface with
1771 * SO_BINDTODEVICE or similar. This is the next best thing to the
1772 * correct behaviour, which would be to update the MTU on all
1773 * interfaces.
1774 */
1775 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, 0);
1776 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, dev->ifindex);
1777 }
1778
1779 /*
1780 * Misc support functions
1781 */
1782
1783 static struct rt6_info *ip6_rt_copy(const struct rt6_info *ort,
1784 const struct in6_addr *dest)
1785 {
1786 struct net *net = dev_net(ort->rt6i_dev);
1787 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops,
1788 ort->dst.dev, 0);
1789
1790 if (rt) {
1791 rt->dst.input = ort->dst.input;
1792 rt->dst.output = ort->dst.output;
1793 rt->dst.flags |= DST_HOST;
1794
1795 rt->rt6i_dst.addr = *dest;
1796 rt->rt6i_dst.plen = 128;
1797 dst_copy_metrics(&rt->dst, &ort->dst);
1798 rt->dst.error = ort->dst.error;
1799 rt->rt6i_idev = ort->rt6i_idev;
1800 if (rt->rt6i_idev)
1801 in6_dev_hold(rt->rt6i_idev);
1802 rt->dst.lastuse = jiffies;
1803 rt->rt6i_expires = 0;
1804
1805 rt->rt6i_gateway = ort->rt6i_gateway;
1806 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1807 rt->rt6i_metric = 0;
1808
1809 #ifdef CONFIG_IPV6_SUBTREES
1810 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1811 #endif
1812 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1813 rt->rt6i_table = ort->rt6i_table;
1814 }
1815 return rt;
1816 }
1817
1818 #ifdef CONFIG_IPV6_ROUTE_INFO
1819 static struct rt6_info *rt6_get_route_info(struct net *net,
1820 const struct in6_addr *prefix, int prefixlen,
1821 const struct in6_addr *gwaddr, int ifindex)
1822 {
1823 struct fib6_node *fn;
1824 struct rt6_info *rt = NULL;
1825 struct fib6_table *table;
1826
1827 table = fib6_get_table(net, RT6_TABLE_INFO);
1828 if (!table)
1829 return NULL;
1830
1831 write_lock_bh(&table->tb6_lock);
1832 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1833 if (!fn)
1834 goto out;
1835
1836 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1837 if (rt->rt6i_dev->ifindex != ifindex)
1838 continue;
1839 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1840 continue;
1841 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1842 continue;
1843 dst_hold(&rt->dst);
1844 break;
1845 }
1846 out:
1847 write_unlock_bh(&table->tb6_lock);
1848 return rt;
1849 }
1850
1851 static struct rt6_info *rt6_add_route_info(struct net *net,
1852 const struct in6_addr *prefix, int prefixlen,
1853 const struct in6_addr *gwaddr, int ifindex,
1854 unsigned pref)
1855 {
1856 struct fib6_config cfg = {
1857 .fc_table = RT6_TABLE_INFO,
1858 .fc_metric = IP6_RT_PRIO_USER,
1859 .fc_ifindex = ifindex,
1860 .fc_dst_len = prefixlen,
1861 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1862 RTF_UP | RTF_PREF(pref),
1863 .fc_nlinfo.pid = 0,
1864 .fc_nlinfo.nlh = NULL,
1865 .fc_nlinfo.nl_net = net,
1866 };
1867
1868 cfg.fc_dst = *prefix;
1869 cfg.fc_gateway = *gwaddr;
1870
1871 /* We should treat it as a default route if prefix length is 0. */
1872 if (!prefixlen)
1873 cfg.fc_flags |= RTF_DEFAULT;
1874
1875 ip6_route_add(&cfg);
1876
1877 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1878 }
1879 #endif
1880
1881 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1882 {
1883 struct rt6_info *rt;
1884 struct fib6_table *table;
1885
1886 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1887 if (!table)
1888 return NULL;
1889
1890 write_lock_bh(&table->tb6_lock);
1891 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1892 if (dev == rt->rt6i_dev &&
1893 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1894 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1895 break;
1896 }
1897 if (rt)
1898 dst_hold(&rt->dst);
1899 write_unlock_bh(&table->tb6_lock);
1900 return rt;
1901 }
1902
1903 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1904 struct net_device *dev,
1905 unsigned int pref)
1906 {
1907 struct fib6_config cfg = {
1908 .fc_table = RT6_TABLE_DFLT,
1909 .fc_metric = IP6_RT_PRIO_USER,
1910 .fc_ifindex = dev->ifindex,
1911 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1912 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1913 .fc_nlinfo.pid = 0,
1914 .fc_nlinfo.nlh = NULL,
1915 .fc_nlinfo.nl_net = dev_net(dev),
1916 };
1917
1918 cfg.fc_gateway = *gwaddr;
1919
1920 ip6_route_add(&cfg);
1921
1922 return rt6_get_dflt_router(gwaddr, dev);
1923 }
1924
1925 void rt6_purge_dflt_routers(struct net *net)
1926 {
1927 struct rt6_info *rt;
1928 struct fib6_table *table;
1929
1930 /* NOTE: Keep consistent with rt6_get_dflt_router */
1931 table = fib6_get_table(net, RT6_TABLE_DFLT);
1932 if (!table)
1933 return;
1934
1935 restart:
1936 read_lock_bh(&table->tb6_lock);
1937 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1938 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1939 dst_hold(&rt->dst);
1940 read_unlock_bh(&table->tb6_lock);
1941 ip6_del_rt(rt);
1942 goto restart;
1943 }
1944 }
1945 read_unlock_bh(&table->tb6_lock);
1946 }
1947
1948 static void rtmsg_to_fib6_config(struct net *net,
1949 struct in6_rtmsg *rtmsg,
1950 struct fib6_config *cfg)
1951 {
1952 memset(cfg, 0, sizeof(*cfg));
1953
1954 cfg->fc_table = RT6_TABLE_MAIN;
1955 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1956 cfg->fc_metric = rtmsg->rtmsg_metric;
1957 cfg->fc_expires = rtmsg->rtmsg_info;
1958 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1959 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1960 cfg->fc_flags = rtmsg->rtmsg_flags;
1961
1962 cfg->fc_nlinfo.nl_net = net;
1963
1964 cfg->fc_dst = rtmsg->rtmsg_dst;
1965 cfg->fc_src = rtmsg->rtmsg_src;
1966 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1967 }
1968
1969 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1970 {
1971 struct fib6_config cfg;
1972 struct in6_rtmsg rtmsg;
1973 int err;
1974
1975 switch(cmd) {
1976 case SIOCADDRT: /* Add a route */
1977 case SIOCDELRT: /* Delete a route */
1978 if (!capable(CAP_NET_ADMIN))
1979 return -EPERM;
1980 err = copy_from_user(&rtmsg, arg,
1981 sizeof(struct in6_rtmsg));
1982 if (err)
1983 return -EFAULT;
1984
1985 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1986
1987 rtnl_lock();
1988 switch (cmd) {
1989 case SIOCADDRT:
1990 err = ip6_route_add(&cfg);
1991 break;
1992 case SIOCDELRT:
1993 err = ip6_route_del(&cfg);
1994 break;
1995 default:
1996 err = -EINVAL;
1997 }
1998 rtnl_unlock();
1999
2000 return err;
2001 }
2002
2003 return -EINVAL;
2004 }
2005
2006 /*
2007 * Drop the packet on the floor
2008 */
2009
2010 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2011 {
2012 int type;
2013 struct dst_entry *dst = skb_dst(skb);
2014 switch (ipstats_mib_noroutes) {
2015 case IPSTATS_MIB_INNOROUTES:
2016 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2017 if (type == IPV6_ADDR_ANY) {
2018 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2019 IPSTATS_MIB_INADDRERRORS);
2020 break;
2021 }
2022 /* FALLTHROUGH */
2023 case IPSTATS_MIB_OUTNOROUTES:
2024 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2025 ipstats_mib_noroutes);
2026 break;
2027 }
2028 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2029 kfree_skb(skb);
2030 return 0;
2031 }
2032
2033 static int ip6_pkt_discard(struct sk_buff *skb)
2034 {
2035 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2036 }
2037
2038 static int ip6_pkt_discard_out(struct sk_buff *skb)
2039 {
2040 skb->dev = skb_dst(skb)->dev;
2041 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2042 }
2043
2044 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2045
2046 static int ip6_pkt_prohibit(struct sk_buff *skb)
2047 {
2048 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2049 }
2050
2051 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2052 {
2053 skb->dev = skb_dst(skb)->dev;
2054 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2055 }
2056
2057 #endif
2058
2059 /*
2060 * Allocate a dst for local (unicast / anycast) address.
2061 */
2062
2063 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2064 const struct in6_addr *addr,
2065 bool anycast)
2066 {
2067 struct net *net = dev_net(idev->dev);
2068 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops,
2069 net->loopback_dev, 0);
2070 struct neighbour *neigh;
2071
2072 if (!rt) {
2073 if (net_ratelimit())
2074 pr_warning("IPv6: Maximum number of routes reached,"
2075 " consider increasing route/max_size.\n");
2076 return ERR_PTR(-ENOMEM);
2077 }
2078
2079 in6_dev_hold(idev);
2080
2081 rt->dst.flags |= DST_HOST;
2082 rt->dst.input = ip6_input;
2083 rt->dst.output = ip6_output;
2084 rt->rt6i_idev = idev;
2085 rt->dst.obsolete = -1;
2086
2087 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2088 if (anycast)
2089 rt->rt6i_flags |= RTF_ANYCAST;
2090 else
2091 rt->rt6i_flags |= RTF_LOCAL;
2092 neigh = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, rt->rt6i_dev);
2093 if (IS_ERR(neigh)) {
2094 dst_free(&rt->dst);
2095
2096 return ERR_CAST(neigh);
2097 }
2098 dst_set_neighbour(&rt->dst, neigh);
2099
2100 rt->rt6i_dst.addr = *addr;
2101 rt->rt6i_dst.plen = 128;
2102 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2103
2104 atomic_set(&rt->dst.__refcnt, 1);
2105
2106 return rt;
2107 }
2108
2109 int ip6_route_get_saddr(struct net *net,
2110 struct rt6_info *rt,
2111 const struct in6_addr *daddr,
2112 unsigned int prefs,
2113 struct in6_addr *saddr)
2114 {
2115 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2116 int err = 0;
2117 if (rt->rt6i_prefsrc.plen)
2118 *saddr = rt->rt6i_prefsrc.addr;
2119 else
2120 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2121 daddr, prefs, saddr);
2122 return err;
2123 }
2124
2125 /* remove deleted ip from prefsrc entries */
2126 struct arg_dev_net_ip {
2127 struct net_device *dev;
2128 struct net *net;
2129 struct in6_addr *addr;
2130 };
2131
2132 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2133 {
2134 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2135 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2136 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2137
2138 if (((void *)rt->rt6i_dev == dev || !dev) &&
2139 rt != net->ipv6.ip6_null_entry &&
2140 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2141 /* remove prefsrc entry */
2142 rt->rt6i_prefsrc.plen = 0;
2143 }
2144 return 0;
2145 }
2146
2147 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2148 {
2149 struct net *net = dev_net(ifp->idev->dev);
2150 struct arg_dev_net_ip adni = {
2151 .dev = ifp->idev->dev,
2152 .net = net,
2153 .addr = &ifp->addr,
2154 };
2155 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2156 }
2157
2158 struct arg_dev_net {
2159 struct net_device *dev;
2160 struct net *net;
2161 };
2162
2163 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2164 {
2165 const struct arg_dev_net *adn = arg;
2166 const struct net_device *dev = adn->dev;
2167
2168 if ((rt->rt6i_dev == dev || !dev) &&
2169 rt != adn->net->ipv6.ip6_null_entry) {
2170 RT6_TRACE("deleted by ifdown %p\n", rt);
2171 return -1;
2172 }
2173 return 0;
2174 }
2175
2176 void rt6_ifdown(struct net *net, struct net_device *dev)
2177 {
2178 struct arg_dev_net adn = {
2179 .dev = dev,
2180 .net = net,
2181 };
2182
2183 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2184 icmp6_clean_all(fib6_ifdown, &adn);
2185 }
2186
2187 struct rt6_mtu_change_arg
2188 {
2189 struct net_device *dev;
2190 unsigned mtu;
2191 };
2192
2193 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2194 {
2195 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2196 struct inet6_dev *idev;
2197
2198 /* In IPv6 pmtu discovery is not optional,
2199 so that RTAX_MTU lock cannot disable it.
2200 We still use this lock to block changes
2201 caused by addrconf/ndisc.
2202 */
2203
2204 idev = __in6_dev_get(arg->dev);
2205 if (!idev)
2206 return 0;
2207
2208 /* For administrative MTU increase, there is no way to discover
2209 IPv6 PMTU increase, so PMTU increase should be updated here.
2210 Since RFC 1981 doesn't include administrative MTU increase
2211 update PMTU increase is a MUST. (i.e. jumbo frame)
2212 */
2213 /*
2214 If new MTU is less than route PMTU, this new MTU will be the
2215 lowest MTU in the path, update the route PMTU to reflect PMTU
2216 decreases; if new MTU is greater than route PMTU, and the
2217 old MTU is the lowest MTU in the path, update the route PMTU
2218 to reflect the increase. In this case if the other nodes' MTU
2219 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2220 PMTU discouvery.
2221 */
2222 if (rt->rt6i_dev == arg->dev &&
2223 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2224 (dst_mtu(&rt->dst) >= arg->mtu ||
2225 (dst_mtu(&rt->dst) < arg->mtu &&
2226 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2227 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2228 }
2229 return 0;
2230 }
2231
2232 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
2233 {
2234 struct rt6_mtu_change_arg arg = {
2235 .dev = dev,
2236 .mtu = mtu,
2237 };
2238
2239 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2240 }
2241
2242 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2243 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2244 [RTA_OIF] = { .type = NLA_U32 },
2245 [RTA_IIF] = { .type = NLA_U32 },
2246 [RTA_PRIORITY] = { .type = NLA_U32 },
2247 [RTA_METRICS] = { .type = NLA_NESTED },
2248 };
2249
2250 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2251 struct fib6_config *cfg)
2252 {
2253 struct rtmsg *rtm;
2254 struct nlattr *tb[RTA_MAX+1];
2255 int err;
2256
2257 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2258 if (err < 0)
2259 goto errout;
2260
2261 err = -EINVAL;
2262 rtm = nlmsg_data(nlh);
2263 memset(cfg, 0, sizeof(*cfg));
2264
2265 cfg->fc_table = rtm->rtm_table;
2266 cfg->fc_dst_len = rtm->rtm_dst_len;
2267 cfg->fc_src_len = rtm->rtm_src_len;
2268 cfg->fc_flags = RTF_UP;
2269 cfg->fc_protocol = rtm->rtm_protocol;
2270
2271 if (rtm->rtm_type == RTN_UNREACHABLE)
2272 cfg->fc_flags |= RTF_REJECT;
2273
2274 if (rtm->rtm_type == RTN_LOCAL)
2275 cfg->fc_flags |= RTF_LOCAL;
2276
2277 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2278 cfg->fc_nlinfo.nlh = nlh;
2279 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2280
2281 if (tb[RTA_GATEWAY]) {
2282 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2283 cfg->fc_flags |= RTF_GATEWAY;
2284 }
2285
2286 if (tb[RTA_DST]) {
2287 int plen = (rtm->rtm_dst_len + 7) >> 3;
2288
2289 if (nla_len(tb[RTA_DST]) < plen)
2290 goto errout;
2291
2292 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2293 }
2294
2295 if (tb[RTA_SRC]) {
2296 int plen = (rtm->rtm_src_len + 7) >> 3;
2297
2298 if (nla_len(tb[RTA_SRC]) < plen)
2299 goto errout;
2300
2301 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2302 }
2303
2304 if (tb[RTA_PREFSRC])
2305 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2306
2307 if (tb[RTA_OIF])
2308 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2309
2310 if (tb[RTA_PRIORITY])
2311 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2312
2313 if (tb[RTA_METRICS]) {
2314 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2315 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2316 }
2317
2318 if (tb[RTA_TABLE])
2319 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2320
2321 err = 0;
2322 errout:
2323 return err;
2324 }
2325
2326 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2327 {
2328 struct fib6_config cfg;
2329 int err;
2330
2331 err = rtm_to_fib6_config(skb, nlh, &cfg);
2332 if (err < 0)
2333 return err;
2334
2335 return ip6_route_del(&cfg);
2336 }
2337
2338 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2339 {
2340 struct fib6_config cfg;
2341 int err;
2342
2343 err = rtm_to_fib6_config(skb, nlh, &cfg);
2344 if (err < 0)
2345 return err;
2346
2347 return ip6_route_add(&cfg);
2348 }
2349
2350 static inline size_t rt6_nlmsg_size(void)
2351 {
2352 return NLMSG_ALIGN(sizeof(struct rtmsg))
2353 + nla_total_size(16) /* RTA_SRC */
2354 + nla_total_size(16) /* RTA_DST */
2355 + nla_total_size(16) /* RTA_GATEWAY */
2356 + nla_total_size(16) /* RTA_PREFSRC */
2357 + nla_total_size(4) /* RTA_TABLE */
2358 + nla_total_size(4) /* RTA_IIF */
2359 + nla_total_size(4) /* RTA_OIF */
2360 + nla_total_size(4) /* RTA_PRIORITY */
2361 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2362 + nla_total_size(sizeof(struct rta_cacheinfo));
2363 }
2364
2365 static int rt6_fill_node(struct net *net,
2366 struct sk_buff *skb, struct rt6_info *rt,
2367 struct in6_addr *dst, struct in6_addr *src,
2368 int iif, int type, u32 pid, u32 seq,
2369 int prefix, int nowait, unsigned int flags)
2370 {
2371 struct rtmsg *rtm;
2372 struct nlmsghdr *nlh;
2373 long expires;
2374 u32 table;
2375 struct neighbour *n;
2376
2377 if (prefix) { /* user wants prefix routes only */
2378 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2379 /* success since this is not a prefix route */
2380 return 1;
2381 }
2382 }
2383
2384 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2385 if (!nlh)
2386 return -EMSGSIZE;
2387
2388 rtm = nlmsg_data(nlh);
2389 rtm->rtm_family = AF_INET6;
2390 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2391 rtm->rtm_src_len = rt->rt6i_src.plen;
2392 rtm->rtm_tos = 0;
2393 if (rt->rt6i_table)
2394 table = rt->rt6i_table->tb6_id;
2395 else
2396 table = RT6_TABLE_UNSPEC;
2397 rtm->rtm_table = table;
2398 NLA_PUT_U32(skb, RTA_TABLE, table);
2399 if (rt->rt6i_flags & RTF_REJECT)
2400 rtm->rtm_type = RTN_UNREACHABLE;
2401 else if (rt->rt6i_flags & RTF_LOCAL)
2402 rtm->rtm_type = RTN_LOCAL;
2403 else if (rt->rt6i_dev && (rt->rt6i_dev->flags & IFF_LOOPBACK))
2404 rtm->rtm_type = RTN_LOCAL;
2405 else
2406 rtm->rtm_type = RTN_UNICAST;
2407 rtm->rtm_flags = 0;
2408 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2409 rtm->rtm_protocol = rt->rt6i_protocol;
2410 if (rt->rt6i_flags & RTF_DYNAMIC)
2411 rtm->rtm_protocol = RTPROT_REDIRECT;
2412 else if (rt->rt6i_flags & RTF_ADDRCONF)
2413 rtm->rtm_protocol = RTPROT_KERNEL;
2414 else if (rt->rt6i_flags & RTF_DEFAULT)
2415 rtm->rtm_protocol = RTPROT_RA;
2416
2417 if (rt->rt6i_flags & RTF_CACHE)
2418 rtm->rtm_flags |= RTM_F_CLONED;
2419
2420 if (dst) {
2421 NLA_PUT(skb, RTA_DST, 16, dst);
2422 rtm->rtm_dst_len = 128;
2423 } else if (rtm->rtm_dst_len)
2424 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2425 #ifdef CONFIG_IPV6_SUBTREES
2426 if (src) {
2427 NLA_PUT(skb, RTA_SRC, 16, src);
2428 rtm->rtm_src_len = 128;
2429 } else if (rtm->rtm_src_len)
2430 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2431 #endif
2432 if (iif) {
2433 #ifdef CONFIG_IPV6_MROUTE
2434 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2435 int err = ip6mr_get_route(net, skb, rtm, nowait);
2436 if (err <= 0) {
2437 if (!nowait) {
2438 if (err == 0)
2439 return 0;
2440 goto nla_put_failure;
2441 } else {
2442 if (err == -EMSGSIZE)
2443 goto nla_put_failure;
2444 }
2445 }
2446 } else
2447 #endif
2448 NLA_PUT_U32(skb, RTA_IIF, iif);
2449 } else if (dst) {
2450 struct in6_addr saddr_buf;
2451 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0)
2452 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2453 }
2454
2455 if (rt->rt6i_prefsrc.plen) {
2456 struct in6_addr saddr_buf;
2457 saddr_buf = rt->rt6i_prefsrc.addr;
2458 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2459 }
2460
2461 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2462 goto nla_put_failure;
2463
2464 rcu_read_lock();
2465 n = dst_get_neighbour_noref(&rt->dst);
2466 if (n)
2467 NLA_PUT(skb, RTA_GATEWAY, 16, &n->primary_key);
2468 rcu_read_unlock();
2469
2470 if (rt->dst.dev)
2471 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2472
2473 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2474
2475 if (!(rt->rt6i_flags & RTF_EXPIRES))
2476 expires = 0;
2477 else if (rt->rt6i_expires - jiffies < INT_MAX)
2478 expires = rt->rt6i_expires - jiffies;
2479 else
2480 expires = INT_MAX;
2481
2482 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, 0, 0,
2483 expires, rt->dst.error) < 0)
2484 goto nla_put_failure;
2485
2486 return nlmsg_end(skb, nlh);
2487
2488 nla_put_failure:
2489 nlmsg_cancel(skb, nlh);
2490 return -EMSGSIZE;
2491 }
2492
2493 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2494 {
2495 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2496 int prefix;
2497
2498 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2499 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2500 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2501 } else
2502 prefix = 0;
2503
2504 return rt6_fill_node(arg->net,
2505 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2506 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2507 prefix, 0, NLM_F_MULTI);
2508 }
2509
2510 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2511 {
2512 struct net *net = sock_net(in_skb->sk);
2513 struct nlattr *tb[RTA_MAX+1];
2514 struct rt6_info *rt;
2515 struct sk_buff *skb;
2516 struct rtmsg *rtm;
2517 struct flowi6 fl6;
2518 int err, iif = 0;
2519
2520 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2521 if (err < 0)
2522 goto errout;
2523
2524 err = -EINVAL;
2525 memset(&fl6, 0, sizeof(fl6));
2526
2527 if (tb[RTA_SRC]) {
2528 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2529 goto errout;
2530
2531 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2532 }
2533
2534 if (tb[RTA_DST]) {
2535 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2536 goto errout;
2537
2538 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2539 }
2540
2541 if (tb[RTA_IIF])
2542 iif = nla_get_u32(tb[RTA_IIF]);
2543
2544 if (tb[RTA_OIF])
2545 fl6.flowi6_oif = nla_get_u32(tb[RTA_OIF]);
2546
2547 if (iif) {
2548 struct net_device *dev;
2549 dev = __dev_get_by_index(net, iif);
2550 if (!dev) {
2551 err = -ENODEV;
2552 goto errout;
2553 }
2554 }
2555
2556 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2557 if (!skb) {
2558 err = -ENOBUFS;
2559 goto errout;
2560 }
2561
2562 /* Reserve room for dummy headers, this skb can pass
2563 through good chunk of routing engine.
2564 */
2565 skb_reset_mac_header(skb);
2566 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2567
2568 rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl6);
2569 skb_dst_set(skb, &rt->dst);
2570
2571 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2572 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2573 nlh->nlmsg_seq, 0, 0, 0);
2574 if (err < 0) {
2575 kfree_skb(skb);
2576 goto errout;
2577 }
2578
2579 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2580 errout:
2581 return err;
2582 }
2583
2584 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2585 {
2586 struct sk_buff *skb;
2587 struct net *net = info->nl_net;
2588 u32 seq;
2589 int err;
2590
2591 err = -ENOBUFS;
2592 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2593
2594 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2595 if (!skb)
2596 goto errout;
2597
2598 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2599 event, info->pid, seq, 0, 0, 0);
2600 if (err < 0) {
2601 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2602 WARN_ON(err == -EMSGSIZE);
2603 kfree_skb(skb);
2604 goto errout;
2605 }
2606 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2607 info->nlh, gfp_any());
2608 return;
2609 errout:
2610 if (err < 0)
2611 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2612 }
2613
2614 static int ip6_route_dev_notify(struct notifier_block *this,
2615 unsigned long event, void *data)
2616 {
2617 struct net_device *dev = (struct net_device *)data;
2618 struct net *net = dev_net(dev);
2619
2620 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2621 net->ipv6.ip6_null_entry->dst.dev = dev;
2622 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2623 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2624 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2625 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2626 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2627 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2628 #endif
2629 }
2630
2631 return NOTIFY_OK;
2632 }
2633
2634 /*
2635 * /proc
2636 */
2637
2638 #ifdef CONFIG_PROC_FS
2639
2640 struct rt6_proc_arg
2641 {
2642 char *buffer;
2643 int offset;
2644 int length;
2645 int skip;
2646 int len;
2647 };
2648
2649 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2650 {
2651 struct seq_file *m = p_arg;
2652 struct neighbour *n;
2653
2654 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2655
2656 #ifdef CONFIG_IPV6_SUBTREES
2657 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2658 #else
2659 seq_puts(m, "00000000000000000000000000000000 00 ");
2660 #endif
2661 rcu_read_lock();
2662 n = dst_get_neighbour_noref(&rt->dst);
2663 if (n) {
2664 seq_printf(m, "%pi6", n->primary_key);
2665 } else {
2666 seq_puts(m, "00000000000000000000000000000000");
2667 }
2668 rcu_read_unlock();
2669 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2670 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2671 rt->dst.__use, rt->rt6i_flags,
2672 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2673 return 0;
2674 }
2675
2676 static int ipv6_route_show(struct seq_file *m, void *v)
2677 {
2678 struct net *net = (struct net *)m->private;
2679 fib6_clean_all(net, rt6_info_route, 0, m);
2680 return 0;
2681 }
2682
2683 static int ipv6_route_open(struct inode *inode, struct file *file)
2684 {
2685 return single_open_net(inode, file, ipv6_route_show);
2686 }
2687
2688 static const struct file_operations ipv6_route_proc_fops = {
2689 .owner = THIS_MODULE,
2690 .open = ipv6_route_open,
2691 .read = seq_read,
2692 .llseek = seq_lseek,
2693 .release = single_release_net,
2694 };
2695
2696 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2697 {
2698 struct net *net = (struct net *)seq->private;
2699 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2700 net->ipv6.rt6_stats->fib_nodes,
2701 net->ipv6.rt6_stats->fib_route_nodes,
2702 net->ipv6.rt6_stats->fib_rt_alloc,
2703 net->ipv6.rt6_stats->fib_rt_entries,
2704 net->ipv6.rt6_stats->fib_rt_cache,
2705 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2706 net->ipv6.rt6_stats->fib_discarded_routes);
2707
2708 return 0;
2709 }
2710
2711 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2712 {
2713 return single_open_net(inode, file, rt6_stats_seq_show);
2714 }
2715
2716 static const struct file_operations rt6_stats_seq_fops = {
2717 .owner = THIS_MODULE,
2718 .open = rt6_stats_seq_open,
2719 .read = seq_read,
2720 .llseek = seq_lseek,
2721 .release = single_release_net,
2722 };
2723 #endif /* CONFIG_PROC_FS */
2724
2725 #ifdef CONFIG_SYSCTL
2726
2727 static
2728 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2729 void __user *buffer, size_t *lenp, loff_t *ppos)
2730 {
2731 struct net *net;
2732 int delay;
2733 if (!write)
2734 return -EINVAL;
2735
2736 net = (struct net *)ctl->extra1;
2737 delay = net->ipv6.sysctl.flush_delay;
2738 proc_dointvec(ctl, write, buffer, lenp, ppos);
2739 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2740 return 0;
2741 }
2742
2743 ctl_table ipv6_route_table_template[] = {
2744 {
2745 .procname = "flush",
2746 .data = &init_net.ipv6.sysctl.flush_delay,
2747 .maxlen = sizeof(int),
2748 .mode = 0200,
2749 .proc_handler = ipv6_sysctl_rtcache_flush
2750 },
2751 {
2752 .procname = "gc_thresh",
2753 .data = &ip6_dst_ops_template.gc_thresh,
2754 .maxlen = sizeof(int),
2755 .mode = 0644,
2756 .proc_handler = proc_dointvec,
2757 },
2758 {
2759 .procname = "max_size",
2760 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2761 .maxlen = sizeof(int),
2762 .mode = 0644,
2763 .proc_handler = proc_dointvec,
2764 },
2765 {
2766 .procname = "gc_min_interval",
2767 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2768 .maxlen = sizeof(int),
2769 .mode = 0644,
2770 .proc_handler = proc_dointvec_jiffies,
2771 },
2772 {
2773 .procname = "gc_timeout",
2774 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2775 .maxlen = sizeof(int),
2776 .mode = 0644,
2777 .proc_handler = proc_dointvec_jiffies,
2778 },
2779 {
2780 .procname = "gc_interval",
2781 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2782 .maxlen = sizeof(int),
2783 .mode = 0644,
2784 .proc_handler = proc_dointvec_jiffies,
2785 },
2786 {
2787 .procname = "gc_elasticity",
2788 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2789 .maxlen = sizeof(int),
2790 .mode = 0644,
2791 .proc_handler = proc_dointvec,
2792 },
2793 {
2794 .procname = "mtu_expires",
2795 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2796 .maxlen = sizeof(int),
2797 .mode = 0644,
2798 .proc_handler = proc_dointvec_jiffies,
2799 },
2800 {
2801 .procname = "min_adv_mss",
2802 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2803 .maxlen = sizeof(int),
2804 .mode = 0644,
2805 .proc_handler = proc_dointvec,
2806 },
2807 {
2808 .procname = "gc_min_interval_ms",
2809 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2810 .maxlen = sizeof(int),
2811 .mode = 0644,
2812 .proc_handler = proc_dointvec_ms_jiffies,
2813 },
2814 { }
2815 };
2816
2817 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2818 {
2819 struct ctl_table *table;
2820
2821 table = kmemdup(ipv6_route_table_template,
2822 sizeof(ipv6_route_table_template),
2823 GFP_KERNEL);
2824
2825 if (table) {
2826 table[0].data = &net->ipv6.sysctl.flush_delay;
2827 table[0].extra1 = net;
2828 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2829 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2830 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2831 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2832 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2833 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2834 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2835 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2836 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2837 }
2838
2839 return table;
2840 }
2841 #endif
2842
2843 static int __net_init ip6_route_net_init(struct net *net)
2844 {
2845 int ret = -ENOMEM;
2846
2847 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2848 sizeof(net->ipv6.ip6_dst_ops));
2849
2850 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2851 goto out_ip6_dst_ops;
2852
2853 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2854 sizeof(*net->ipv6.ip6_null_entry),
2855 GFP_KERNEL);
2856 if (!net->ipv6.ip6_null_entry)
2857 goto out_ip6_dst_entries;
2858 net->ipv6.ip6_null_entry->dst.path =
2859 (struct dst_entry *)net->ipv6.ip6_null_entry;
2860 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2861 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2862 ip6_template_metrics, true);
2863
2864 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2865 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2866 sizeof(*net->ipv6.ip6_prohibit_entry),
2867 GFP_KERNEL);
2868 if (!net->ipv6.ip6_prohibit_entry)
2869 goto out_ip6_null_entry;
2870 net->ipv6.ip6_prohibit_entry->dst.path =
2871 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2872 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2873 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2874 ip6_template_metrics, true);
2875
2876 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2877 sizeof(*net->ipv6.ip6_blk_hole_entry),
2878 GFP_KERNEL);
2879 if (!net->ipv6.ip6_blk_hole_entry)
2880 goto out_ip6_prohibit_entry;
2881 net->ipv6.ip6_blk_hole_entry->dst.path =
2882 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2883 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2884 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2885 ip6_template_metrics, true);
2886 #endif
2887
2888 net->ipv6.sysctl.flush_delay = 0;
2889 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2890 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2891 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2892 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2893 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2894 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2895 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2896
2897 #ifdef CONFIG_PROC_FS
2898 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2899 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2900 #endif
2901 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2902
2903 ret = 0;
2904 out:
2905 return ret;
2906
2907 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2908 out_ip6_prohibit_entry:
2909 kfree(net->ipv6.ip6_prohibit_entry);
2910 out_ip6_null_entry:
2911 kfree(net->ipv6.ip6_null_entry);
2912 #endif
2913 out_ip6_dst_entries:
2914 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2915 out_ip6_dst_ops:
2916 goto out;
2917 }
2918
2919 static void __net_exit ip6_route_net_exit(struct net *net)
2920 {
2921 #ifdef CONFIG_PROC_FS
2922 proc_net_remove(net, "ipv6_route");
2923 proc_net_remove(net, "rt6_stats");
2924 #endif
2925 kfree(net->ipv6.ip6_null_entry);
2926 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2927 kfree(net->ipv6.ip6_prohibit_entry);
2928 kfree(net->ipv6.ip6_blk_hole_entry);
2929 #endif
2930 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2931 }
2932
2933 static struct pernet_operations ip6_route_net_ops = {
2934 .init = ip6_route_net_init,
2935 .exit = ip6_route_net_exit,
2936 };
2937
2938 static struct notifier_block ip6_route_dev_notifier = {
2939 .notifier_call = ip6_route_dev_notify,
2940 .priority = 0,
2941 };
2942
2943 int __init ip6_route_init(void)
2944 {
2945 int ret;
2946
2947 ret = -ENOMEM;
2948 ip6_dst_ops_template.kmem_cachep =
2949 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2950 SLAB_HWCACHE_ALIGN, NULL);
2951 if (!ip6_dst_ops_template.kmem_cachep)
2952 goto out;
2953
2954 ret = dst_entries_init(&ip6_dst_blackhole_ops);
2955 if (ret)
2956 goto out_kmem_cache;
2957
2958 ret = register_pernet_subsys(&ip6_route_net_ops);
2959 if (ret)
2960 goto out_dst_entries;
2961
2962 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
2963
2964 /* Registering of the loopback is done before this portion of code,
2965 * the loopback reference in rt6_info will not be taken, do it
2966 * manually for init_net */
2967 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
2968 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2969 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2970 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
2971 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2972 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
2973 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2974 #endif
2975 ret = fib6_init();
2976 if (ret)
2977 goto out_register_subsys;
2978
2979 ret = xfrm6_init();
2980 if (ret)
2981 goto out_fib6_init;
2982
2983 ret = fib6_rules_init();
2984 if (ret)
2985 goto xfrm6_init;
2986
2987 ret = -ENOBUFS;
2988 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
2989 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
2990 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
2991 goto fib6_rules_init;
2992
2993 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2994 if (ret)
2995 goto fib6_rules_init;
2996
2997 out:
2998 return ret;
2999
3000 fib6_rules_init:
3001 fib6_rules_cleanup();
3002 xfrm6_init:
3003 xfrm6_fini();
3004 out_fib6_init:
3005 fib6_gc_cleanup();
3006 out_register_subsys:
3007 unregister_pernet_subsys(&ip6_route_net_ops);
3008 out_dst_entries:
3009 dst_entries_destroy(&ip6_dst_blackhole_ops);
3010 out_kmem_cache:
3011 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3012 goto out;
3013 }
3014
3015 void ip6_route_cleanup(void)
3016 {
3017 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3018 fib6_rules_cleanup();
3019 xfrm6_fini();
3020 fib6_gc_cleanup();
3021 unregister_pernet_subsys(&ip6_route_net_ops);
3022 dst_entries_destroy(&ip6_dst_blackhole_ops);
3023 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3024 }
Linux 2.6 libata-dev (mirror)