projects / deliverable / linux.git / blob
? search:


c28cdda3f5dbc02a0da37b1c42596133c60c50cf
[deliverable/linux.git] / net / ipv6 / route.c
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
68 enum rt6_nud_state {
69 RT6_NUD_FAIL_HARD = -2,
70 RT6_NUD_FAIL_SOFT = -1,
71 RT6_NUD_SUCCEED = 1
72 };
73
74 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
75 const struct in6_addr *dest);
76 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
77 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
78 static unsigned int ip6_mtu(const struct dst_entry *dst);
79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
80 static void ip6_dst_destroy(struct dst_entry *);
81 static void ip6_dst_ifdown(struct dst_entry *,
82 struct net_device *dev, int how);
83 static int ip6_dst_gc(struct dst_ops *ops);
84
85 static int ip6_pkt_discard(struct sk_buff *skb);
86 static int ip6_pkt_discard_out(struct sk_buff *skb);
87 static void ip6_link_failure(struct sk_buff *skb);
88 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
89 struct sk_buff *skb, u32 mtu);
90 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
91 struct sk_buff *skb);
92 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
93
94 #ifdef CONFIG_IPV6_ROUTE_INFO
95 static struct rt6_info *rt6_add_route_info(struct net *net,
96 const struct in6_addr *prefix, int prefixlen,
97 const struct in6_addr *gwaddr, int ifindex,
98 unsigned int pref);
99 static struct rt6_info *rt6_get_route_info(struct net *net,
100 const struct in6_addr *prefix, int prefixlen,
101 const struct in6_addr *gwaddr, int ifindex);
102 #endif
103
104 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
105 {
106 struct rt6_info *rt = (struct rt6_info *) dst;
107 struct inet_peer *peer;
108 u32 *p = NULL;
109
110 if (!(rt->dst.flags & DST_HOST))
111 return NULL;
112
113 peer = rt6_get_peer_create(rt);
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
134 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
135 struct sk_buff *skb,
136 const void *daddr)
137 {
138 struct in6_addr *p = &rt->rt6i_gateway;
139
140 if (!ipv6_addr_any(p))
141 return (const void *) p;
142 else if (skb)
143 return &ipv6_hdr(skb)->daddr;
144 return daddr;
145 }
146
147 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
148 struct sk_buff *skb,
149 const void *daddr)
150 {
151 struct rt6_info *rt = (struct rt6_info *) dst;
152 struct neighbour *n;
153
154 daddr = choose_neigh_daddr(rt, skb, daddr);
155 n = __ipv6_neigh_lookup(dst->dev, daddr);
156 if (n)
157 return n;
158 return neigh_create(&nd_tbl, daddr, dst->dev);
159 }
160
161 static struct dst_ops ip6_dst_ops_template = {
162 .family = AF_INET6,
163 .protocol = cpu_to_be16(ETH_P_IPV6),
164 .gc = ip6_dst_gc,
165 .gc_thresh = 1024,
166 .check = ip6_dst_check,
167 .default_advmss = ip6_default_advmss,
168 .mtu = ip6_mtu,
169 .cow_metrics = ipv6_cow_metrics,
170 .destroy = ip6_dst_destroy,
171 .ifdown = ip6_dst_ifdown,
172 .negative_advice = ip6_negative_advice,
173 .link_failure = ip6_link_failure,
174 .update_pmtu = ip6_rt_update_pmtu,
175 .redirect = rt6_do_redirect,
176 .local_out = __ip6_local_out,
177 .neigh_lookup = ip6_neigh_lookup,
178 };
179
180 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
181 {
182 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
183
184 return mtu ? : dst->dev->mtu;
185 }
186
187 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
188 struct sk_buff *skb, u32 mtu)
189 {
190 }
191
192 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
193 struct sk_buff *skb)
194 {
195 }
196
197 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
198 unsigned long old)
199 {
200 return NULL;
201 }
202
203 static struct dst_ops ip6_dst_blackhole_ops = {
204 .family = AF_INET6,
205 .protocol = cpu_to_be16(ETH_P_IPV6),
206 .destroy = ip6_dst_destroy,
207 .check = ip6_dst_check,
208 .mtu = ip6_blackhole_mtu,
209 .default_advmss = ip6_default_advmss,
210 .update_pmtu = ip6_rt_blackhole_update_pmtu,
211 .redirect = ip6_rt_blackhole_redirect,
212 .cow_metrics = ip6_rt_blackhole_cow_metrics,
213 .neigh_lookup = ip6_neigh_lookup,
214 };
215
216 static const u32 ip6_template_metrics[RTAX_MAX] = {
217 [RTAX_HOPLIMIT - 1] = 0,
218 };
219
220 static const struct rt6_info ip6_null_entry_template = {
221 .dst = {
222 .__refcnt = ATOMIC_INIT(1),
223 .__use = 1,
224 .obsolete = DST_OBSOLETE_FORCE_CHK,
225 .error = -ENETUNREACH,
226 .input = ip6_pkt_discard,
227 .output = ip6_pkt_discard_out,
228 },
229 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
230 .rt6i_protocol = RTPROT_KERNEL,
231 .rt6i_metric = ~(u32) 0,
232 .rt6i_ref = ATOMIC_INIT(1),
233 };
234
235 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
236
237 static int ip6_pkt_prohibit(struct sk_buff *skb);
238 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
239
240 static const struct rt6_info ip6_prohibit_entry_template = {
241 .dst = {
242 .__refcnt = ATOMIC_INIT(1),
243 .__use = 1,
244 .obsolete = DST_OBSOLETE_FORCE_CHK,
245 .error = -EACCES,
246 .input = ip6_pkt_prohibit,
247 .output = ip6_pkt_prohibit_out,
248 },
249 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
250 .rt6i_protocol = RTPROT_KERNEL,
251 .rt6i_metric = ~(u32) 0,
252 .rt6i_ref = ATOMIC_INIT(1),
253 };
254
255 static const struct rt6_info ip6_blk_hole_entry_template = {
256 .dst = {
257 .__refcnt = ATOMIC_INIT(1),
258 .__use = 1,
259 .obsolete = DST_OBSOLETE_FORCE_CHK,
260 .error = -EINVAL,
261 .input = dst_discard,
262 .output = dst_discard,
263 },
264 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
265 .rt6i_protocol = RTPROT_KERNEL,
266 .rt6i_metric = ~(u32) 0,
267 .rt6i_ref = ATOMIC_INIT(1),
268 };
269
270 #endif
271
272 /* allocate dst with ip6_dst_ops */
273 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
274 struct net_device *dev,
275 int flags,
276 struct fib6_table *table)
277 {
278 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
279 0, DST_OBSOLETE_FORCE_CHK, flags);
280
281 if (rt) {
282 struct dst_entry *dst = &rt->dst;
283
284 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
285 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
286 rt->rt6i_genid = rt_genid_ipv6(net);
287 INIT_LIST_HEAD(&rt->rt6i_siblings);
288 }
289 return rt;
290 }
291
292 static void ip6_dst_destroy(struct dst_entry *dst)
293 {
294 struct rt6_info *rt = (struct rt6_info *)dst;
295 struct inet6_dev *idev = rt->rt6i_idev;
296 struct dst_entry *from = dst->from;
297
298 if (!(rt->dst.flags & DST_HOST))
299 dst_destroy_metrics_generic(dst);
300
301 if (idev) {
302 rt->rt6i_idev = NULL;
303 in6_dev_put(idev);
304 }
305
306 dst->from = NULL;
307 dst_release(from);
308
309 if (rt6_has_peer(rt)) {
310 struct inet_peer *peer = rt6_peer_ptr(rt);
311 inet_putpeer(peer);
312 }
313 }
314
315 void rt6_bind_peer(struct rt6_info *rt, int create)
316 {
317 struct inet_peer_base *base;
318 struct inet_peer *peer;
319
320 base = inetpeer_base_ptr(rt->_rt6i_peer);
321 if (!base)
322 return;
323
324 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
325 if (peer) {
326 if (!rt6_set_peer(rt, peer))
327 inet_putpeer(peer);
328 }
329 }
330
331 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
332 int how)
333 {
334 struct rt6_info *rt = (struct rt6_info *)dst;
335 struct inet6_dev *idev = rt->rt6i_idev;
336 struct net_device *loopback_dev =
337 dev_net(dev)->loopback_dev;
338
339 if (dev != loopback_dev) {
340 if (idev && idev->dev == dev) {
341 struct inet6_dev *loopback_idev =
342 in6_dev_get(loopback_dev);
343 if (loopback_idev) {
344 rt->rt6i_idev = loopback_idev;
345 in6_dev_put(idev);
346 }
347 }
348 }
349 }
350
351 static bool rt6_check_expired(const struct rt6_info *rt)
352 {
353 if (rt->rt6i_flags & RTF_EXPIRES) {
354 if (time_after(jiffies, rt->dst.expires))
355 return true;
356 } else if (rt->dst.from) {
357 return rt6_check_expired((struct rt6_info *) rt->dst.from);
358 }
359 return false;
360 }
361
362 static bool rt6_need_strict(const struct in6_addr *daddr)
363 {
364 return ipv6_addr_type(daddr) &
365 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
366 }
367
368 /* Multipath route selection:
369 * Hash based function using packet header and flowlabel.
370 * Adapted from fib_info_hashfn()
371 */
372 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
373 const struct flowi6 *fl6)
374 {
375 unsigned int val = fl6->flowi6_proto;
376
377 val ^= ipv6_addr_hash(&fl6->daddr);
378 val ^= ipv6_addr_hash(&fl6->saddr);
379
380 /* Work only if this not encapsulated */
381 switch (fl6->flowi6_proto) {
382 case IPPROTO_UDP:
383 case IPPROTO_TCP:
384 case IPPROTO_SCTP:
385 val ^= (__force u16)fl6->fl6_sport;
386 val ^= (__force u16)fl6->fl6_dport;
387 break;
388
389 case IPPROTO_ICMPV6:
390 val ^= (__force u16)fl6->fl6_icmp_type;
391 val ^= (__force u16)fl6->fl6_icmp_code;
392 break;
393 }
394 /* RFC6438 recommands to use flowlabel */
395 val ^= (__force u32)fl6->flowlabel;
396
397 /* Perhaps, we need to tune, this function? */
398 val = val ^ (val >> 7) ^ (val >> 12);
399 return val % candidate_count;
400 }
401
402 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
403 struct flowi6 *fl6, int oif,
404 int strict)
405 {
406 struct rt6_info *sibling, *next_sibling;
407 int route_choosen;
408
409 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
410 /* Don't change the route, if route_choosen == 0
411 * (siblings does not include ourself)
412 */
413 if (route_choosen)
414 list_for_each_entry_safe(sibling, next_sibling,
415 &match->rt6i_siblings, rt6i_siblings) {
416 route_choosen--;
417 if (route_choosen == 0) {
418 if (rt6_score_route(sibling, oif, strict) < 0)
419 break;
420 match = sibling;
421 break;
422 }
423 }
424 return match;
425 }
426
427 /*
428 * Route lookup. Any table->tb6_lock is implied.
429 */
430
431 static inline struct rt6_info *rt6_device_match(struct net *net,
432 struct rt6_info *rt,
433 const struct in6_addr *saddr,
434 int oif,
435 int flags)
436 {
437 struct rt6_info *local = NULL;
438 struct rt6_info *sprt;
439
440 if (!oif && ipv6_addr_any(saddr))
441 goto out;
442
443 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
444 struct net_device *dev = sprt->dst.dev;
445
446 if (oif) {
447 if (dev->ifindex == oif)
448 return sprt;
449 if (dev->flags & IFF_LOOPBACK) {
450 if (!sprt->rt6i_idev ||
451 sprt->rt6i_idev->dev->ifindex != oif) {
452 if (flags & RT6_LOOKUP_F_IFACE && oif)
453 continue;
454 if (local && (!oif ||
455 local->rt6i_idev->dev->ifindex == oif))
456 continue;
457 }
458 local = sprt;
459 }
460 } else {
461 if (ipv6_chk_addr(net, saddr, dev,
462 flags & RT6_LOOKUP_F_IFACE))
463 return sprt;
464 }
465 }
466
467 if (oif) {
468 if (local)
469 return local;
470
471 if (flags & RT6_LOOKUP_F_IFACE)
472 return net->ipv6.ip6_null_entry;
473 }
474 out:
475 return rt;
476 }
477
478 #ifdef CONFIG_IPV6_ROUTER_PREF
479 struct __rt6_probe_work {
480 struct work_struct work;
481 struct in6_addr target;
482 struct net_device *dev;
483 };
484
485 static void rt6_probe_deferred(struct work_struct *w)
486 {
487 struct in6_addr mcaddr;
488 struct __rt6_probe_work *work =
489 container_of(w, struct __rt6_probe_work, work);
490
491 addrconf_addr_solict_mult(&work->target, &mcaddr);
492 ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
493 dev_put(work->dev);
494 kfree(w);
495 }
496
497 static void rt6_probe(struct rt6_info *rt)
498 {
499 struct neighbour *neigh;
500 /*
501 * Okay, this does not seem to be appropriate
502 * for now, however, we need to check if it
503 * is really so; aka Router Reachability Probing.
504 *
505 * Router Reachability Probe MUST be rate-limited
506 * to no more than one per minute.
507 */
508 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
509 return;
510 rcu_read_lock_bh();
511 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
512 if (neigh) {
513 write_lock(&neigh->lock);
514 if (neigh->nud_state & NUD_VALID)
515 goto out;
516 }
517
518 if (!neigh ||
519 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
520 struct __rt6_probe_work *work;
521
522 work = kmalloc(sizeof(*work), GFP_ATOMIC);
523
524 if (neigh && work)
525 neigh->updated = jiffies;
526
527 if (neigh)
528 write_unlock(&neigh->lock);
529
530 if (work) {
531 INIT_WORK(&work->work, rt6_probe_deferred);
532 work->target = rt->rt6i_gateway;
533 dev_hold(rt->dst.dev);
534 work->dev = rt->dst.dev;
535 schedule_work(&work->work);
536 }
537 } else {
538 out:
539 write_unlock(&neigh->lock);
540 }
541 rcu_read_unlock_bh();
542 }
543 #else
544 static inline void rt6_probe(struct rt6_info *rt)
545 {
546 }
547 #endif
548
549 /*
550 * Default Router Selection (RFC 2461 6.3.6)
551 */
552 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
553 {
554 struct net_device *dev = rt->dst.dev;
555 if (!oif || dev->ifindex == oif)
556 return 2;
557 if ((dev->flags & IFF_LOOPBACK) &&
558 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
559 return 1;
560 return 0;
561 }
562
563 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
564 {
565 struct neighbour *neigh;
566 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
567
568 if (rt->rt6i_flags & RTF_NONEXTHOP ||
569 !(rt->rt6i_flags & RTF_GATEWAY))
570 return RT6_NUD_SUCCEED;
571
572 rcu_read_lock_bh();
573 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
574 if (neigh) {
575 read_lock(&neigh->lock);
576 if (neigh->nud_state & NUD_VALID)
577 ret = RT6_NUD_SUCCEED;
578 #ifdef CONFIG_IPV6_ROUTER_PREF
579 else if (!(neigh->nud_state & NUD_FAILED))
580 ret = RT6_NUD_SUCCEED;
581 #endif
582 read_unlock(&neigh->lock);
583 } else {
584 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
585 RT6_NUD_SUCCEED : RT6_NUD_FAIL_SOFT;
586 }
587 rcu_read_unlock_bh();
588
589 return ret;
590 }
591
592 static int rt6_score_route(struct rt6_info *rt, int oif,
593 int strict)
594 {
595 int m;
596
597 m = rt6_check_dev(rt, oif);
598 if (!m && (strict & RT6_LOOKUP_F_IFACE))
599 return RT6_NUD_FAIL_HARD;
600 #ifdef CONFIG_IPV6_ROUTER_PREF
601 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
602 #endif
603 if (strict & RT6_LOOKUP_F_REACHABLE) {
604 int n = rt6_check_neigh(rt);
605 if (n < 0)
606 return n;
607 }
608 return m;
609 }
610
611 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
612 int *mpri, struct rt6_info *match,
613 bool *do_rr)
614 {
615 int m;
616 bool match_do_rr = false;
617
618 if (rt6_check_expired(rt))
619 goto out;
620
621 m = rt6_score_route(rt, oif, strict);
622 if (m == RT6_NUD_FAIL_SOFT) {
623 match_do_rr = true;
624 m = 0; /* lowest valid score */
625 } else if (m < 0) {
626 goto out;
627 }
628
629 if (strict & RT6_LOOKUP_F_REACHABLE)
630 rt6_probe(rt);
631
632 if (m > *mpri) {
633 *do_rr = match_do_rr;
634 *mpri = m;
635 match = rt;
636 }
637 out:
638 return match;
639 }
640
641 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
642 struct rt6_info *rr_head,
643 u32 metric, int oif, int strict,
644 bool *do_rr)
645 {
646 struct rt6_info *rt, *match;
647 int mpri = -1;
648
649 match = NULL;
650 for (rt = rr_head; rt && rt->rt6i_metric == metric;
651 rt = rt->dst.rt6_next)
652 match = find_match(rt, oif, strict, &mpri, match, do_rr);
653 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
654 rt = rt->dst.rt6_next)
655 match = find_match(rt, oif, strict, &mpri, match, do_rr);
656
657 return match;
658 }
659
660 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
661 {
662 struct rt6_info *match, *rt0;
663 struct net *net;
664 bool do_rr = false;
665
666 rt0 = fn->rr_ptr;
667 if (!rt0)
668 fn->rr_ptr = rt0 = fn->leaf;
669
670 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
671 &do_rr);
672
673 if (do_rr) {
674 struct rt6_info *next = rt0->dst.rt6_next;
675
676 /* no entries matched; do round-robin */
677 if (!next || next->rt6i_metric != rt0->rt6i_metric)
678 next = fn->leaf;
679
680 if (next != rt0)
681 fn->rr_ptr = next;
682 }
683
684 net = dev_net(rt0->dst.dev);
685 return match ? match : net->ipv6.ip6_null_entry;
686 }
687
688 #ifdef CONFIG_IPV6_ROUTE_INFO
689 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
690 const struct in6_addr *gwaddr)
691 {
692 struct net *net = dev_net(dev);
693 struct route_info *rinfo = (struct route_info *) opt;
694 struct in6_addr prefix_buf, *prefix;
695 unsigned int pref;
696 unsigned long lifetime;
697 struct rt6_info *rt;
698
699 if (len < sizeof(struct route_info)) {
700 return -EINVAL;
701 }
702
703 /* Sanity check for prefix_len and length */
704 if (rinfo->length > 3) {
705 return -EINVAL;
706 } else if (rinfo->prefix_len > 128) {
707 return -EINVAL;
708 } else if (rinfo->prefix_len > 64) {
709 if (rinfo->length < 2) {
710 return -EINVAL;
711 }
712 } else if (rinfo->prefix_len > 0) {
713 if (rinfo->length < 1) {
714 return -EINVAL;
715 }
716 }
717
718 pref = rinfo->route_pref;
719 if (pref == ICMPV6_ROUTER_PREF_INVALID)
720 return -EINVAL;
721
722 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
723
724 if (rinfo->length == 3)
725 prefix = (struct in6_addr *)rinfo->prefix;
726 else {
727 /* this function is safe */
728 ipv6_addr_prefix(&prefix_buf,
729 (struct in6_addr *)rinfo->prefix,
730 rinfo->prefix_len);
731 prefix = &prefix_buf;
732 }
733
734 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
735 dev->ifindex);
736
737 if (rt && !lifetime) {
738 ip6_del_rt(rt);
739 rt = NULL;
740 }
741
742 if (!rt && lifetime)
743 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
744 pref);
745 else if (rt)
746 rt->rt6i_flags = RTF_ROUTEINFO |
747 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
748
749 if (rt) {
750 if (!addrconf_finite_timeout(lifetime))
751 rt6_clean_expires(rt);
752 else
753 rt6_set_expires(rt, jiffies + HZ * lifetime);
754
755 ip6_rt_put(rt);
756 }
757 return 0;
758 }
759 #endif
760
761 #define BACKTRACK(__net, saddr) \
762 do { \
763 if (rt == __net->ipv6.ip6_null_entry) { \
764 struct fib6_node *pn; \
765 while (1) { \
766 if (fn->fn_flags & RTN_TL_ROOT) \
767 goto out; \
768 pn = fn->parent; \
769 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
770 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
771 else \
772 fn = pn; \
773 if (fn->fn_flags & RTN_RTINFO) \
774 goto restart; \
775 } \
776 } \
777 } while (0)
778
779 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
780 struct fib6_table *table,
781 struct flowi6 *fl6, int flags)
782 {
783 struct fib6_node *fn;
784 struct rt6_info *rt;
785
786 read_lock_bh(&table->tb6_lock);
787 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
788 restart:
789 rt = fn->leaf;
790 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
791 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
792 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
793 BACKTRACK(net, &fl6->saddr);
794 out:
795 dst_use(&rt->dst, jiffies);
796 read_unlock_bh(&table->tb6_lock);
797 return rt;
798
799 }
800
801 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
802 int flags)
803 {
804 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
805 }
806 EXPORT_SYMBOL_GPL(ip6_route_lookup);
807
808 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
809 const struct in6_addr *saddr, int oif, int strict)
810 {
811 struct flowi6 fl6 = {
812 .flowi6_oif = oif,
813 .daddr = *daddr,
814 };
815 struct dst_entry *dst;
816 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
817
818 if (saddr) {
819 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
820 flags |= RT6_LOOKUP_F_HAS_SADDR;
821 }
822
823 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
824 if (dst->error == 0)
825 return (struct rt6_info *) dst;
826
827 dst_release(dst);
828
829 return NULL;
830 }
831
832 EXPORT_SYMBOL(rt6_lookup);
833
834 /* ip6_ins_rt is called with FREE table->tb6_lock.
835 It takes new route entry, the addition fails by any reason the
836 route is freed. In any case, if caller does not hold it, it may
837 be destroyed.
838 */
839
840 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
841 {
842 int err;
843 struct fib6_table *table;
844
845 table = rt->rt6i_table;
846 write_lock_bh(&table->tb6_lock);
847 err = fib6_add(&table->tb6_root, rt, info);
848 write_unlock_bh(&table->tb6_lock);
849
850 return err;
851 }
852
853 int ip6_ins_rt(struct rt6_info *rt)
854 {
855 struct nl_info info = {
856 .nl_net = dev_net(rt->dst.dev),
857 };
858 return __ip6_ins_rt(rt, &info);
859 }
860
861 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
862 const struct in6_addr *daddr,
863 const struct in6_addr *saddr)
864 {
865 struct rt6_info *rt;
866
867 /*
868 * Clone the route.
869 */
870
871 rt = ip6_rt_copy(ort, daddr);
872
873 if (rt) {
874 if (ort->rt6i_dst.plen != 128 &&
875 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
876 rt->rt6i_flags |= RTF_ANYCAST;
877
878 rt->rt6i_flags |= RTF_CACHE;
879
880 #ifdef CONFIG_IPV6_SUBTREES
881 if (rt->rt6i_src.plen && saddr) {
882 rt->rt6i_src.addr = *saddr;
883 rt->rt6i_src.plen = 128;
884 }
885 #endif
886 }
887
888 return rt;
889 }
890
891 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
892 const struct in6_addr *daddr)
893 {
894 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
895
896 if (rt)
897 rt->rt6i_flags |= RTF_CACHE;
898 return rt;
899 }
900
901 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
902 struct flowi6 *fl6, int flags)
903 {
904 struct fib6_node *fn;
905 struct rt6_info *rt, *nrt;
906 int strict = 0;
907 int attempts = 3;
908 int err;
909 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
910
911 strict |= flags & RT6_LOOKUP_F_IFACE;
912
913 relookup:
914 read_lock_bh(&table->tb6_lock);
915
916 restart_2:
917 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
918
919 restart:
920 rt = rt6_select(fn, oif, strict | reachable);
921 if (rt->rt6i_nsiblings)
922 rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
923 BACKTRACK(net, &fl6->saddr);
924 if (rt == net->ipv6.ip6_null_entry ||
925 rt->rt6i_flags & RTF_CACHE)
926 goto out;
927
928 dst_hold(&rt->dst);
929 read_unlock_bh(&table->tb6_lock);
930
931 if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
932 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
933 else if (!(rt->dst.flags & DST_HOST))
934 nrt = rt6_alloc_clone(rt, &fl6->daddr);
935 else
936 goto out2;
937
938 ip6_rt_put(rt);
939 rt = nrt ? : net->ipv6.ip6_null_entry;
940
941 dst_hold(&rt->dst);
942 if (nrt) {
943 err = ip6_ins_rt(nrt);
944 if (!err)
945 goto out2;
946 }
947
948 if (--attempts <= 0)
949 goto out2;
950
951 /*
952 * Race condition! In the gap, when table->tb6_lock was
953 * released someone could insert this route. Relookup.
954 */
955 ip6_rt_put(rt);
956 goto relookup;
957
958 out:
959 if (reachable) {
960 reachable = 0;
961 goto restart_2;
962 }
963 dst_hold(&rt->dst);
964 read_unlock_bh(&table->tb6_lock);
965 out2:
966 rt->dst.lastuse = jiffies;
967 rt->dst.__use++;
968
969 return rt;
970 }
971
972 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
973 struct flowi6 *fl6, int flags)
974 {
975 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
976 }
977
978 static struct dst_entry *ip6_route_input_lookup(struct net *net,
979 struct net_device *dev,
980 struct flowi6 *fl6, int flags)
981 {
982 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
983 flags |= RT6_LOOKUP_F_IFACE;
984
985 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
986 }
987
988 void ip6_route_input(struct sk_buff *skb)
989 {
990 const struct ipv6hdr *iph = ipv6_hdr(skb);
991 struct net *net = dev_net(skb->dev);
992 int flags = RT6_LOOKUP_F_HAS_SADDR;
993 struct flowi6 fl6 = {
994 .flowi6_iif = skb->dev->ifindex,
995 .daddr = iph->daddr,
996 .saddr = iph->saddr,
997 .flowlabel = ip6_flowinfo(iph),
998 .flowi6_mark = skb->mark,
999 .flowi6_proto = iph->nexthdr,
1000 };
1001
1002 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1003 }
1004
1005 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1006 struct flowi6 *fl6, int flags)
1007 {
1008 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1009 }
1010
1011 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1012 struct flowi6 *fl6)
1013 {
1014 int flags = 0;
1015
1016 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1017
1018 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1019 flags |= RT6_LOOKUP_F_IFACE;
1020
1021 if (!ipv6_addr_any(&fl6->saddr))
1022 flags |= RT6_LOOKUP_F_HAS_SADDR;
1023 else if (sk)
1024 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1025
1026 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1027 }
1028
1029 EXPORT_SYMBOL(ip6_route_output);
1030
1031 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1032 {
1033 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1034 struct dst_entry *new = NULL;
1035
1036 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1037 if (rt) {
1038 new = &rt->dst;
1039
1040 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1041 rt6_init_peer(rt, net->ipv6.peers);
1042
1043 new->__use = 1;
1044 new->input = dst_discard;
1045 new->output = dst_discard;
1046
1047 if (dst_metrics_read_only(&ort->dst))
1048 new->_metrics = ort->dst._metrics;
1049 else
1050 dst_copy_metrics(new, &ort->dst);
1051 rt->rt6i_idev = ort->rt6i_idev;
1052 if (rt->rt6i_idev)
1053 in6_dev_hold(rt->rt6i_idev);
1054
1055 rt->rt6i_gateway = ort->rt6i_gateway;
1056 rt->rt6i_flags = ort->rt6i_flags;
1057 rt->rt6i_metric = 0;
1058
1059 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1060 #ifdef CONFIG_IPV6_SUBTREES
1061 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1062 #endif
1063
1064 dst_free(new);
1065 }
1066
1067 dst_release(dst_orig);
1068 return new ? new : ERR_PTR(-ENOMEM);
1069 }
1070
1071 /*
1072 * Destination cache support functions
1073 */
1074
1075 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1076 {
1077 struct rt6_info *rt;
1078
1079 rt = (struct rt6_info *) dst;
1080
1081 /* All IPV6 dsts are created with ->obsolete set to the value
1082 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1083 * into this function always.
1084 */
1085 if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev)))
1086 return NULL;
1087
1088 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1089 return NULL;
1090
1091 if (rt6_check_expired(rt))
1092 return NULL;
1093
1094 return dst;
1095 }
1096
1097 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1098 {
1099 struct rt6_info *rt = (struct rt6_info *) dst;
1100
1101 if (rt) {
1102 if (rt->rt6i_flags & RTF_CACHE) {
1103 if (rt6_check_expired(rt)) {
1104 ip6_del_rt(rt);
1105 dst = NULL;
1106 }
1107 } else {
1108 dst_release(dst);
1109 dst = NULL;
1110 }
1111 }
1112 return dst;
1113 }
1114
1115 static void ip6_link_failure(struct sk_buff *skb)
1116 {
1117 struct rt6_info *rt;
1118
1119 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1120
1121 rt = (struct rt6_info *) skb_dst(skb);
1122 if (rt) {
1123 if (rt->rt6i_flags & RTF_CACHE) {
1124 dst_hold(&rt->dst);
1125 if (ip6_del_rt(rt))
1126 dst_free(&rt->dst);
1127 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1128 rt->rt6i_node->fn_sernum = -1;
1129 }
1130 }
1131 }
1132
1133 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1134 struct sk_buff *skb, u32 mtu)
1135 {
1136 struct rt6_info *rt6 = (struct rt6_info*)dst;
1137
1138 dst_confirm(dst);
1139 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1140 struct net *net = dev_net(dst->dev);
1141
1142 rt6->rt6i_flags |= RTF_MODIFIED;
1143 if (mtu < IPV6_MIN_MTU) {
1144 u32 features = dst_metric(dst, RTAX_FEATURES);
1145 mtu = IPV6_MIN_MTU;
1146 features |= RTAX_FEATURE_ALLFRAG;
1147 dst_metric_set(dst, RTAX_FEATURES, features);
1148 }
1149 dst_metric_set(dst, RTAX_MTU, mtu);
1150 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1151 }
1152 }
1153
1154 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1155 int oif, u32 mark)
1156 {
1157 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1158 struct dst_entry *dst;
1159 struct flowi6 fl6;
1160
1161 memset(&fl6, 0, sizeof(fl6));
1162 fl6.flowi6_oif = oif;
1163 fl6.flowi6_mark = mark;
1164 fl6.daddr = iph->daddr;
1165 fl6.saddr = iph->saddr;
1166 fl6.flowlabel = ip6_flowinfo(iph);
1167
1168 dst = ip6_route_output(net, NULL, &fl6);
1169 if (!dst->error)
1170 ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1171 dst_release(dst);
1172 }
1173 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1174
1175 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1176 {
1177 ip6_update_pmtu(skb, sock_net(sk), mtu,
1178 sk->sk_bound_dev_if, sk->sk_mark);
1179 }
1180 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1181
1182 /* Handle redirects */
1183 struct ip6rd_flowi {
1184 struct flowi6 fl6;
1185 struct in6_addr gateway;
1186 };
1187
1188 static struct rt6_info *__ip6_route_redirect(struct net *net,
1189 struct fib6_table *table,
1190 struct flowi6 *fl6,
1191 int flags)
1192 {
1193 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1194 struct rt6_info *rt;
1195 struct fib6_node *fn;
1196
1197 /* Get the "current" route for this destination and
1198 * check if the redirect has come from approriate router.
1199 *
1200 * RFC 4861 specifies that redirects should only be
1201 * accepted if they come from the nexthop to the target.
1202 * Due to the way the routes are chosen, this notion
1203 * is a bit fuzzy and one might need to check all possible
1204 * routes.
1205 */
1206
1207 read_lock_bh(&table->tb6_lock);
1208 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1209 restart:
1210 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1211 if (rt6_check_expired(rt))
1212 continue;
1213 if (rt->dst.error)
1214 break;
1215 if (!(rt->rt6i_flags & RTF_GATEWAY))
1216 continue;
1217 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1218 continue;
1219 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1220 continue;
1221 break;
1222 }
1223
1224 if (!rt)
1225 rt = net->ipv6.ip6_null_entry;
1226 else if (rt->dst.error) {
1227 rt = net->ipv6.ip6_null_entry;
1228 goto out;
1229 }
1230 BACKTRACK(net, &fl6->saddr);
1231 out:
1232 dst_hold(&rt->dst);
1233
1234 read_unlock_bh(&table->tb6_lock);
1235
1236 return rt;
1237 };
1238
1239 static struct dst_entry *ip6_route_redirect(struct net *net,
1240 const struct flowi6 *fl6,
1241 const struct in6_addr *gateway)
1242 {
1243 int flags = RT6_LOOKUP_F_HAS_SADDR;
1244 struct ip6rd_flowi rdfl;
1245
1246 rdfl.fl6 = *fl6;
1247 rdfl.gateway = *gateway;
1248
1249 return fib6_rule_lookup(net, &rdfl.fl6,
1250 flags, __ip6_route_redirect);
1251 }
1252
1253 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1254 {
1255 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1256 struct dst_entry *dst;
1257 struct flowi6 fl6;
1258
1259 memset(&fl6, 0, sizeof(fl6));
1260 fl6.flowi6_oif = oif;
1261 fl6.flowi6_mark = mark;
1262 fl6.daddr = iph->daddr;
1263 fl6.saddr = iph->saddr;
1264 fl6.flowlabel = ip6_flowinfo(iph);
1265
1266 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1267 rt6_do_redirect(dst, NULL, skb);
1268 dst_release(dst);
1269 }
1270 EXPORT_SYMBOL_GPL(ip6_redirect);
1271
1272 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1273 u32 mark)
1274 {
1275 const struct ipv6hdr *iph = ipv6_hdr(skb);
1276 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1277 struct dst_entry *dst;
1278 struct flowi6 fl6;
1279
1280 memset(&fl6, 0, sizeof(fl6));
1281 fl6.flowi6_oif = oif;
1282 fl6.flowi6_mark = mark;
1283 fl6.daddr = msg->dest;
1284 fl6.saddr = iph->daddr;
1285
1286 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1287 rt6_do_redirect(dst, NULL, skb);
1288 dst_release(dst);
1289 }
1290
1291 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1292 {
1293 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1294 }
1295 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1296
1297 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1298 {
1299 struct net_device *dev = dst->dev;
1300 unsigned int mtu = dst_mtu(dst);
1301 struct net *net = dev_net(dev);
1302
1303 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1304
1305 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1306 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1307
1308 /*
1309 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1310 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1311 * IPV6_MAXPLEN is also valid and means: "any MSS,
1312 * rely only on pmtu discovery"
1313 */
1314 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1315 mtu = IPV6_MAXPLEN;
1316 return mtu;
1317 }
1318
1319 static unsigned int ip6_mtu(const struct dst_entry *dst)
1320 {
1321 struct inet6_dev *idev;
1322 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1323
1324 if (mtu)
1325 return mtu;
1326
1327 mtu = IPV6_MIN_MTU;
1328
1329 rcu_read_lock();
1330 idev = __in6_dev_get(dst->dev);
1331 if (idev)
1332 mtu = idev->cnf.mtu6;
1333 rcu_read_unlock();
1334
1335 return mtu;
1336 }
1337
1338 static struct dst_entry *icmp6_dst_gc_list;
1339 static DEFINE_SPINLOCK(icmp6_dst_lock);
1340
1341 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1342 struct flowi6 *fl6)
1343 {
1344 struct dst_entry *dst;
1345 struct rt6_info *rt;
1346 struct inet6_dev *idev = in6_dev_get(dev);
1347 struct net *net = dev_net(dev);
1348
1349 if (unlikely(!idev))
1350 return ERR_PTR(-ENODEV);
1351
1352 rt = ip6_dst_alloc(net, dev, 0, NULL);
1353 if (unlikely(!rt)) {
1354 in6_dev_put(idev);
1355 dst = ERR_PTR(-ENOMEM);
1356 goto out;
1357 }
1358
1359 rt->dst.flags |= DST_HOST;
1360 rt->dst.output = ip6_output;
1361 atomic_set(&rt->dst.__refcnt, 1);
1362 rt->rt6i_gateway = fl6->daddr;
1363 rt->rt6i_dst.addr = fl6->daddr;
1364 rt->rt6i_dst.plen = 128;
1365 rt->rt6i_idev = idev;
1366 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1367
1368 spin_lock_bh(&icmp6_dst_lock);
1369 rt->dst.next = icmp6_dst_gc_list;
1370 icmp6_dst_gc_list = &rt->dst;
1371 spin_unlock_bh(&icmp6_dst_lock);
1372
1373 fib6_force_start_gc(net);
1374
1375 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1376
1377 out:
1378 return dst;
1379 }
1380
1381 int icmp6_dst_gc(void)
1382 {
1383 struct dst_entry *dst, **pprev;
1384 int more = 0;
1385
1386 spin_lock_bh(&icmp6_dst_lock);
1387 pprev = &icmp6_dst_gc_list;
1388
1389 while ((dst = *pprev) != NULL) {
1390 if (!atomic_read(&dst->__refcnt)) {
1391 *pprev = dst->next;
1392 dst_free(dst);
1393 } else {
1394 pprev = &dst->next;
1395 ++more;
1396 }
1397 }
1398
1399 spin_unlock_bh(&icmp6_dst_lock);
1400
1401 return more;
1402 }
1403
1404 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1405 void *arg)
1406 {
1407 struct dst_entry *dst, **pprev;
1408
1409 spin_lock_bh(&icmp6_dst_lock);
1410 pprev = &icmp6_dst_gc_list;
1411 while ((dst = *pprev) != NULL) {
1412 struct rt6_info *rt = (struct rt6_info *) dst;
1413 if (func(rt, arg)) {
1414 *pprev = dst->next;
1415 dst_free(dst);
1416 } else {
1417 pprev = &dst->next;
1418 }
1419 }
1420 spin_unlock_bh(&icmp6_dst_lock);
1421 }
1422
1423 static int ip6_dst_gc(struct dst_ops *ops)
1424 {
1425 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1426 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1427 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1428 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1429 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1430 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1431 int entries;
1432
1433 entries = dst_entries_get_fast(ops);
1434 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1435 entries <= rt_max_size)
1436 goto out;
1437
1438 net->ipv6.ip6_rt_gc_expire++;
1439 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1440 entries = dst_entries_get_slow(ops);
1441 if (entries < ops->gc_thresh)
1442 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1443 out:
1444 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1445 return entries > rt_max_size;
1446 }
1447
1448 /*
1449 *
1450 */
1451
1452 int ip6_route_add(struct fib6_config *cfg)
1453 {
1454 int err;
1455 struct net *net = cfg->fc_nlinfo.nl_net;
1456 struct rt6_info *rt = NULL;
1457 struct net_device *dev = NULL;
1458 struct inet6_dev *idev = NULL;
1459 struct fib6_table *table;
1460 int addr_type;
1461
1462 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1463 return -EINVAL;
1464 #ifndef CONFIG_IPV6_SUBTREES
1465 if (cfg->fc_src_len)
1466 return -EINVAL;
1467 #endif
1468 if (cfg->fc_ifindex) {
1469 err = -ENODEV;
1470 dev = dev_get_by_index(net, cfg->fc_ifindex);
1471 if (!dev)
1472 goto out;
1473 idev = in6_dev_get(dev);
1474 if (!idev)
1475 goto out;
1476 }
1477
1478 if (cfg->fc_metric == 0)
1479 cfg->fc_metric = IP6_RT_PRIO_USER;
1480
1481 err = -ENOBUFS;
1482 if (cfg->fc_nlinfo.nlh &&
1483 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1484 table = fib6_get_table(net, cfg->fc_table);
1485 if (!table) {
1486 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1487 table = fib6_new_table(net, cfg->fc_table);
1488 }
1489 } else {
1490 table = fib6_new_table(net, cfg->fc_table);
1491 }
1492
1493 if (!table)
1494 goto out;
1495
1496 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1497
1498 if (!rt) {
1499 err = -ENOMEM;
1500 goto out;
1501 }
1502
1503 if (cfg->fc_flags & RTF_EXPIRES)
1504 rt6_set_expires(rt, jiffies +
1505 clock_t_to_jiffies(cfg->fc_expires));
1506 else
1507 rt6_clean_expires(rt);
1508
1509 if (cfg->fc_protocol == RTPROT_UNSPEC)
1510 cfg->fc_protocol = RTPROT_BOOT;
1511 rt->rt6i_protocol = cfg->fc_protocol;
1512
1513 addr_type = ipv6_addr_type(&cfg->fc_dst);
1514
1515 if (addr_type & IPV6_ADDR_MULTICAST)
1516 rt->dst.input = ip6_mc_input;
1517 else if (cfg->fc_flags & RTF_LOCAL)
1518 rt->dst.input = ip6_input;
1519 else
1520 rt->dst.input = ip6_forward;
1521
1522 rt->dst.output = ip6_output;
1523
1524 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1525 rt->rt6i_dst.plen = cfg->fc_dst_len;
1526 if (rt->rt6i_dst.plen == 128)
1527 rt->dst.flags |= DST_HOST;
1528
1529 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1530 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1531 if (!metrics) {
1532 err = -ENOMEM;
1533 goto out;
1534 }
1535 dst_init_metrics(&rt->dst, metrics, 0);
1536 }
1537 #ifdef CONFIG_IPV6_SUBTREES
1538 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1539 rt->rt6i_src.plen = cfg->fc_src_len;
1540 #endif
1541
1542 rt->rt6i_metric = cfg->fc_metric;
1543
1544 /* We cannot add true routes via loopback here,
1545 they would result in kernel looping; promote them to reject routes
1546 */
1547 if ((cfg->fc_flags & RTF_REJECT) ||
1548 (dev && (dev->flags & IFF_LOOPBACK) &&
1549 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1550 !(cfg->fc_flags & RTF_LOCAL))) {
1551 /* hold loopback dev/idev if we haven't done so. */
1552 if (dev != net->loopback_dev) {
1553 if (dev) {
1554 dev_put(dev);
1555 in6_dev_put(idev);
1556 }
1557 dev = net->loopback_dev;
1558 dev_hold(dev);
1559 idev = in6_dev_get(dev);
1560 if (!idev) {
1561 err = -ENODEV;
1562 goto out;
1563 }
1564 }
1565 rt->dst.output = ip6_pkt_discard_out;
1566 rt->dst.input = ip6_pkt_discard;
1567 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1568 switch (cfg->fc_type) {
1569 case RTN_BLACKHOLE:
1570 rt->dst.error = -EINVAL;
1571 break;
1572 case RTN_PROHIBIT:
1573 rt->dst.error = -EACCES;
1574 break;
1575 case RTN_THROW:
1576 rt->dst.error = -EAGAIN;
1577 break;
1578 default:
1579 rt->dst.error = -ENETUNREACH;
1580 break;
1581 }
1582 goto install_route;
1583 }
1584
1585 if (cfg->fc_flags & RTF_GATEWAY) {
1586 const struct in6_addr *gw_addr;
1587 int gwa_type;
1588
1589 gw_addr = &cfg->fc_gateway;
1590 rt->rt6i_gateway = *gw_addr;
1591 gwa_type = ipv6_addr_type(gw_addr);
1592
1593 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1594 struct rt6_info *grt;
1595
1596 /* IPv6 strictly inhibits using not link-local
1597 addresses as nexthop address.
1598 Otherwise, router will not able to send redirects.
1599 It is very good, but in some (rare!) circumstances
1600 (SIT, PtP, NBMA NOARP links) it is handy to allow
1601 some exceptions. --ANK
1602 */
1603 err = -EINVAL;
1604 if (!(gwa_type & IPV6_ADDR_UNICAST))
1605 goto out;
1606
1607 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1608
1609 err = -EHOSTUNREACH;
1610 if (!grt)
1611 goto out;
1612 if (dev) {
1613 if (dev != grt->dst.dev) {
1614 ip6_rt_put(grt);
1615 goto out;
1616 }
1617 } else {
1618 dev = grt->dst.dev;
1619 idev = grt->rt6i_idev;
1620 dev_hold(dev);
1621 in6_dev_hold(grt->rt6i_idev);
1622 }
1623 if (!(grt->rt6i_flags & RTF_GATEWAY))
1624 err = 0;
1625 ip6_rt_put(grt);
1626
1627 if (err)
1628 goto out;
1629 }
1630 err = -EINVAL;
1631 if (!dev || (dev->flags & IFF_LOOPBACK))
1632 goto out;
1633 }
1634
1635 err = -ENODEV;
1636 if (!dev)
1637 goto out;
1638
1639 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1640 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1641 err = -EINVAL;
1642 goto out;
1643 }
1644 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1645 rt->rt6i_prefsrc.plen = 128;
1646 } else
1647 rt->rt6i_prefsrc.plen = 0;
1648
1649 rt->rt6i_flags = cfg->fc_flags;
1650
1651 install_route:
1652 if (cfg->fc_mx) {
1653 struct nlattr *nla;
1654 int remaining;
1655
1656 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1657 int type = nla_type(nla);
1658
1659 if (type) {
1660 if (type > RTAX_MAX) {
1661 err = -EINVAL;
1662 goto out;
1663 }
1664
1665 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1666 }
1667 }
1668 }
1669
1670 rt->dst.dev = dev;
1671 rt->rt6i_idev = idev;
1672 rt->rt6i_table = table;
1673
1674 cfg->fc_nlinfo.nl_net = dev_net(dev);
1675
1676 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1677
1678 out:
1679 if (dev)
1680 dev_put(dev);
1681 if (idev)
1682 in6_dev_put(idev);
1683 if (rt)
1684 dst_free(&rt->dst);
1685 return err;
1686 }
1687
1688 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1689 {
1690 int err;
1691 struct fib6_table *table;
1692 struct net *net = dev_net(rt->dst.dev);
1693
1694 if (rt == net->ipv6.ip6_null_entry) {
1695 err = -ENOENT;
1696 goto out;
1697 }
1698
1699 table = rt->rt6i_table;
1700 write_lock_bh(&table->tb6_lock);
1701 err = fib6_del(rt, info);
1702 write_unlock_bh(&table->tb6_lock);
1703
1704 out:
1705 ip6_rt_put(rt);
1706 return err;
1707 }
1708
1709 int ip6_del_rt(struct rt6_info *rt)
1710 {
1711 struct nl_info info = {
1712 .nl_net = dev_net(rt->dst.dev),
1713 };
1714 return __ip6_del_rt(rt, &info);
1715 }
1716
1717 static int ip6_route_del(struct fib6_config *cfg)
1718 {
1719 struct fib6_table *table;
1720 struct fib6_node *fn;
1721 struct rt6_info *rt;
1722 int err = -ESRCH;
1723
1724 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1725 if (!table)
1726 return err;
1727
1728 read_lock_bh(&table->tb6_lock);
1729
1730 fn = fib6_locate(&table->tb6_root,
1731 &cfg->fc_dst, cfg->fc_dst_len,
1732 &cfg->fc_src, cfg->fc_src_len);
1733
1734 if (fn) {
1735 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1736 if (cfg->fc_ifindex &&
1737 (!rt->dst.dev ||
1738 rt->dst.dev->ifindex != cfg->fc_ifindex))
1739 continue;
1740 if (cfg->fc_flags & RTF_GATEWAY &&
1741 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1742 continue;
1743 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1744 continue;
1745 dst_hold(&rt->dst);
1746 read_unlock_bh(&table->tb6_lock);
1747
1748 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1749 }
1750 }
1751 read_unlock_bh(&table->tb6_lock);
1752
1753 return err;
1754 }
1755
1756 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1757 {
1758 struct net *net = dev_net(skb->dev);
1759 struct netevent_redirect netevent;
1760 struct rt6_info *rt, *nrt = NULL;
1761 struct ndisc_options ndopts;
1762 struct inet6_dev *in6_dev;
1763 struct neighbour *neigh;
1764 struct rd_msg *msg;
1765 int optlen, on_link;
1766 u8 *lladdr;
1767
1768 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1769 optlen -= sizeof(*msg);
1770
1771 if (optlen < 0) {
1772 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1773 return;
1774 }
1775
1776 msg = (struct rd_msg *)icmp6_hdr(skb);
1777
1778 if (ipv6_addr_is_multicast(&msg->dest)) {
1779 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1780 return;
1781 }
1782
1783 on_link = 0;
1784 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1785 on_link = 1;
1786 } else if (ipv6_addr_type(&msg->target) !=
1787 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1788 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1789 return;
1790 }
1791
1792 in6_dev = __in6_dev_get(skb->dev);
1793 if (!in6_dev)
1794 return;
1795 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
1796 return;
1797
1798 /* RFC2461 8.1:
1799 * The IP source address of the Redirect MUST be the same as the current
1800 * first-hop router for the specified ICMP Destination Address.
1801 */
1802
1803 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1804 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1805 return;
1806 }
1807
1808 lladdr = NULL;
1809 if (ndopts.nd_opts_tgt_lladdr) {
1810 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1811 skb->dev);
1812 if (!lladdr) {
1813 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1814 return;
1815 }
1816 }
1817
1818 rt = (struct rt6_info *) dst;
1819 if (rt == net->ipv6.ip6_null_entry) {
1820 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1821 return;
1822 }
1823
1824 /* Redirect received -> path was valid.
1825 * Look, redirects are sent only in response to data packets,
1826 * so that this nexthop apparently is reachable. --ANK
1827 */
1828 dst_confirm(&rt->dst);
1829
1830 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1831 if (!neigh)
1832 return;
1833
1834 /*
1835 * We have finally decided to accept it.
1836 */
1837
1838 neigh_update(neigh, lladdr, NUD_STALE,
1839 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1840 NEIGH_UPDATE_F_OVERRIDE|
1841 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1842 NEIGH_UPDATE_F_ISROUTER))
1843 );
1844
1845 nrt = ip6_rt_copy(rt, &msg->dest);
1846 if (!nrt)
1847 goto out;
1848
1849 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1850 if (on_link)
1851 nrt->rt6i_flags &= ~RTF_GATEWAY;
1852
1853 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1854
1855 if (ip6_ins_rt(nrt))
1856 goto out;
1857
1858 netevent.old = &rt->dst;
1859 netevent.new = &nrt->dst;
1860 netevent.daddr = &msg->dest;
1861 netevent.neigh = neigh;
1862 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1863
1864 if (rt->rt6i_flags & RTF_CACHE) {
1865 rt = (struct rt6_info *) dst_clone(&rt->dst);
1866 ip6_del_rt(rt);
1867 }
1868
1869 out:
1870 neigh_release(neigh);
1871 }
1872
1873 /*
1874 * Misc support functions
1875 */
1876
1877 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1878 const struct in6_addr *dest)
1879 {
1880 struct net *net = dev_net(ort->dst.dev);
1881 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1882 ort->rt6i_table);
1883
1884 if (rt) {
1885 rt->dst.input = ort->dst.input;
1886 rt->dst.output = ort->dst.output;
1887 rt->dst.flags |= DST_HOST;
1888
1889 rt->rt6i_dst.addr = *dest;
1890 rt->rt6i_dst.plen = 128;
1891 dst_copy_metrics(&rt->dst, &ort->dst);
1892 rt->dst.error = ort->dst.error;
1893 rt->rt6i_idev = ort->rt6i_idev;
1894 if (rt->rt6i_idev)
1895 in6_dev_hold(rt->rt6i_idev);
1896 rt->dst.lastuse = jiffies;
1897
1898 if (ort->rt6i_flags & RTF_GATEWAY)
1899 rt->rt6i_gateway = ort->rt6i_gateway;
1900 else
1901 rt->rt6i_gateway = *dest;
1902 rt->rt6i_flags = ort->rt6i_flags;
1903 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1904 (RTF_DEFAULT | RTF_ADDRCONF))
1905 rt6_set_from(rt, ort);
1906 rt->rt6i_metric = 0;
1907
1908 #ifdef CONFIG_IPV6_SUBTREES
1909 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1910 #endif
1911 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1912 rt->rt6i_table = ort->rt6i_table;
1913 }
1914 return rt;
1915 }
1916
1917 #ifdef CONFIG_IPV6_ROUTE_INFO
1918 static struct rt6_info *rt6_get_route_info(struct net *net,
1919 const struct in6_addr *prefix, int prefixlen,
1920 const struct in6_addr *gwaddr, int ifindex)
1921 {
1922 struct fib6_node *fn;
1923 struct rt6_info *rt = NULL;
1924 struct fib6_table *table;
1925
1926 table = fib6_get_table(net, RT6_TABLE_INFO);
1927 if (!table)
1928 return NULL;
1929
1930 read_lock_bh(&table->tb6_lock);
1931 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1932 if (!fn)
1933 goto out;
1934
1935 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1936 if (rt->dst.dev->ifindex != ifindex)
1937 continue;
1938 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1939 continue;
1940 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1941 continue;
1942 dst_hold(&rt->dst);
1943 break;
1944 }
1945 out:
1946 read_unlock_bh(&table->tb6_lock);
1947 return rt;
1948 }
1949
1950 static struct rt6_info *rt6_add_route_info(struct net *net,
1951 const struct in6_addr *prefix, int prefixlen,
1952 const struct in6_addr *gwaddr, int ifindex,
1953 unsigned int pref)
1954 {
1955 struct fib6_config cfg = {
1956 .fc_table = RT6_TABLE_INFO,
1957 .fc_metric = IP6_RT_PRIO_USER,
1958 .fc_ifindex = ifindex,
1959 .fc_dst_len = prefixlen,
1960 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1961 RTF_UP | RTF_PREF(pref),
1962 .fc_nlinfo.portid = 0,
1963 .fc_nlinfo.nlh = NULL,
1964 .fc_nlinfo.nl_net = net,
1965 };
1966
1967 cfg.fc_dst = *prefix;
1968 cfg.fc_gateway = *gwaddr;
1969
1970 /* We should treat it as a default route if prefix length is 0. */
1971 if (!prefixlen)
1972 cfg.fc_flags |= RTF_DEFAULT;
1973
1974 ip6_route_add(&cfg);
1975
1976 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1977 }
1978 #endif
1979
1980 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1981 {
1982 struct rt6_info *rt;
1983 struct fib6_table *table;
1984
1985 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1986 if (!table)
1987 return NULL;
1988
1989 read_lock_bh(&table->tb6_lock);
1990 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1991 if (dev == rt->dst.dev &&
1992 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1993 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1994 break;
1995 }
1996 if (rt)
1997 dst_hold(&rt->dst);
1998 read_unlock_bh(&table->tb6_lock);
1999 return rt;
2000 }
2001
2002 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2003 struct net_device *dev,
2004 unsigned int pref)
2005 {
2006 struct fib6_config cfg = {
2007 .fc_table = RT6_TABLE_DFLT,
2008 .fc_metric = IP6_RT_PRIO_USER,
2009 .fc_ifindex = dev->ifindex,
2010 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2011 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2012 .fc_nlinfo.portid = 0,
2013 .fc_nlinfo.nlh = NULL,
2014 .fc_nlinfo.nl_net = dev_net(dev),
2015 };
2016
2017 cfg.fc_gateway = *gwaddr;
2018
2019 ip6_route_add(&cfg);
2020
2021 return rt6_get_dflt_router(gwaddr, dev);
2022 }
2023
2024 void rt6_purge_dflt_routers(struct net *net)
2025 {
2026 struct rt6_info *rt;
2027 struct fib6_table *table;
2028
2029 /* NOTE: Keep consistent with rt6_get_dflt_router */
2030 table = fib6_get_table(net, RT6_TABLE_DFLT);
2031 if (!table)
2032 return;
2033
2034 restart:
2035 read_lock_bh(&table->tb6_lock);
2036 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2037 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2038 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2039 dst_hold(&rt->dst);
2040 read_unlock_bh(&table->tb6_lock);
2041 ip6_del_rt(rt);
2042 goto restart;
2043 }
2044 }
2045 read_unlock_bh(&table->tb6_lock);
2046 }
2047
2048 static void rtmsg_to_fib6_config(struct net *net,
2049 struct in6_rtmsg *rtmsg,
2050 struct fib6_config *cfg)
2051 {
2052 memset(cfg, 0, sizeof(*cfg));
2053
2054 cfg->fc_table = RT6_TABLE_MAIN;
2055 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2056 cfg->fc_metric = rtmsg->rtmsg_metric;
2057 cfg->fc_expires = rtmsg->rtmsg_info;
2058 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2059 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2060 cfg->fc_flags = rtmsg->rtmsg_flags;
2061
2062 cfg->fc_nlinfo.nl_net = net;
2063
2064 cfg->fc_dst = rtmsg->rtmsg_dst;
2065 cfg->fc_src = rtmsg->rtmsg_src;
2066 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2067 }
2068
2069 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2070 {
2071 struct fib6_config cfg;
2072 struct in6_rtmsg rtmsg;
2073 int err;
2074
2075 switch(cmd) {
2076 case SIOCADDRT: /* Add a route */
2077 case SIOCDELRT: /* Delete a route */
2078 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2079 return -EPERM;
2080 err = copy_from_user(&rtmsg, arg,
2081 sizeof(struct in6_rtmsg));
2082 if (err)
2083 return -EFAULT;
2084
2085 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2086
2087 rtnl_lock();
2088 switch (cmd) {
2089 case SIOCADDRT:
2090 err = ip6_route_add(&cfg);
2091 break;
2092 case SIOCDELRT:
2093 err = ip6_route_del(&cfg);
2094 break;
2095 default:
2096 err = -EINVAL;
2097 }
2098 rtnl_unlock();
2099
2100 return err;
2101 }
2102
2103 return -EINVAL;
2104 }
2105
2106 /*
2107 * Drop the packet on the floor
2108 */
2109
2110 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2111 {
2112 int type;
2113 struct dst_entry *dst = skb_dst(skb);
2114 switch (ipstats_mib_noroutes) {
2115 case IPSTATS_MIB_INNOROUTES:
2116 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2117 if (type == IPV6_ADDR_ANY) {
2118 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2119 IPSTATS_MIB_INADDRERRORS);
2120 break;
2121 }
2122 /* FALLTHROUGH */
2123 case IPSTATS_MIB_OUTNOROUTES:
2124 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2125 ipstats_mib_noroutes);
2126 break;
2127 }
2128 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2129 kfree_skb(skb);
2130 return 0;
2131 }
2132
2133 static int ip6_pkt_discard(struct sk_buff *skb)
2134 {
2135 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2136 }
2137
2138 static int ip6_pkt_discard_out(struct sk_buff *skb)
2139 {
2140 skb->dev = skb_dst(skb)->dev;
2141 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2142 }
2143
2144 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2145
2146 static int ip6_pkt_prohibit(struct sk_buff *skb)
2147 {
2148 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2149 }
2150
2151 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2152 {
2153 skb->dev = skb_dst(skb)->dev;
2154 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2155 }
2156
2157 #endif
2158
2159 /*
2160 * Allocate a dst for local (unicast / anycast) address.
2161 */
2162
2163 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2164 const struct in6_addr *addr,
2165 bool anycast)
2166 {
2167 struct net *net = dev_net(idev->dev);
2168 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2169
2170 if (!rt) {
2171 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2172 return ERR_PTR(-ENOMEM);
2173 }
2174
2175 in6_dev_hold(idev);
2176
2177 rt->dst.flags |= DST_HOST;
2178 rt->dst.input = ip6_input;
2179 rt->dst.output = ip6_output;
2180 rt->rt6i_idev = idev;
2181
2182 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2183 if (anycast)
2184 rt->rt6i_flags |= RTF_ANYCAST;
2185 else
2186 rt->rt6i_flags |= RTF_LOCAL;
2187
2188 rt->rt6i_gateway = *addr;
2189 rt->rt6i_dst.addr = *addr;
2190 rt->rt6i_dst.plen = 128;
2191 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2192
2193 atomic_set(&rt->dst.__refcnt, 1);
2194
2195 return rt;
2196 }
2197
2198 int ip6_route_get_saddr(struct net *net,
2199 struct rt6_info *rt,
2200 const struct in6_addr *daddr,
2201 unsigned int prefs,
2202 struct in6_addr *saddr)
2203 {
2204 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2205 int err = 0;
2206 if (rt->rt6i_prefsrc.plen)
2207 *saddr = rt->rt6i_prefsrc.addr;
2208 else
2209 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2210 daddr, prefs, saddr);
2211 return err;
2212 }
2213
2214 /* remove deleted ip from prefsrc entries */
2215 struct arg_dev_net_ip {
2216 struct net_device *dev;
2217 struct net *net;
2218 struct in6_addr *addr;
2219 };
2220
2221 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2222 {
2223 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2224 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2225 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2226
2227 if (((void *)rt->dst.dev == dev || !dev) &&
2228 rt != net->ipv6.ip6_null_entry &&
2229 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2230 /* remove prefsrc entry */
2231 rt->rt6i_prefsrc.plen = 0;
2232 }
2233 return 0;
2234 }
2235
2236 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2237 {
2238 struct net *net = dev_net(ifp->idev->dev);
2239 struct arg_dev_net_ip adni = {
2240 .dev = ifp->idev->dev,
2241 .net = net,
2242 .addr = &ifp->addr,
2243 };
2244 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2245 }
2246
2247 struct arg_dev_net {
2248 struct net_device *dev;
2249 struct net *net;
2250 };
2251
2252 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2253 {
2254 const struct arg_dev_net *adn = arg;
2255 const struct net_device *dev = adn->dev;
2256
2257 if ((rt->dst.dev == dev || !dev) &&
2258 rt != adn->net->ipv6.ip6_null_entry)
2259 return -1;
2260
2261 return 0;
2262 }
2263
2264 void rt6_ifdown(struct net *net, struct net_device *dev)
2265 {
2266 struct arg_dev_net adn = {
2267 .dev = dev,
2268 .net = net,
2269 };
2270
2271 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2272 icmp6_clean_all(fib6_ifdown, &adn);
2273 }
2274
2275 struct rt6_mtu_change_arg {
2276 struct net_device *dev;
2277 unsigned int mtu;
2278 };
2279
2280 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2281 {
2282 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2283 struct inet6_dev *idev;
2284
2285 /* In IPv6 pmtu discovery is not optional,
2286 so that RTAX_MTU lock cannot disable it.
2287 We still use this lock to block changes
2288 caused by addrconf/ndisc.
2289 */
2290
2291 idev = __in6_dev_get(arg->dev);
2292 if (!idev)
2293 return 0;
2294
2295 /* For administrative MTU increase, there is no way to discover
2296 IPv6 PMTU increase, so PMTU increase should be updated here.
2297 Since RFC 1981 doesn't include administrative MTU increase
2298 update PMTU increase is a MUST. (i.e. jumbo frame)
2299 */
2300 /*
2301 If new MTU is less than route PMTU, this new MTU will be the
2302 lowest MTU in the path, update the route PMTU to reflect PMTU
2303 decreases; if new MTU is greater than route PMTU, and the
2304 old MTU is the lowest MTU in the path, update the route PMTU
2305 to reflect the increase. In this case if the other nodes' MTU
2306 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2307 PMTU discouvery.
2308 */
2309 if (rt->dst.dev == arg->dev &&
2310 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2311 (dst_mtu(&rt->dst) >= arg->mtu ||
2312 (dst_mtu(&rt->dst) < arg->mtu &&
2313 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2314 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2315 }
2316 return 0;
2317 }
2318
2319 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2320 {
2321 struct rt6_mtu_change_arg arg = {
2322 .dev = dev,
2323 .mtu = mtu,
2324 };
2325
2326 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2327 }
2328
2329 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2330 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2331 [RTA_OIF] = { .type = NLA_U32 },
2332 [RTA_IIF] = { .type = NLA_U32 },
2333 [RTA_PRIORITY] = { .type = NLA_U32 },
2334 [RTA_METRICS] = { .type = NLA_NESTED },
2335 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2336 };
2337
2338 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2339 struct fib6_config *cfg)
2340 {
2341 struct rtmsg *rtm;
2342 struct nlattr *tb[RTA_MAX+1];
2343 int err;
2344
2345 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2346 if (err < 0)
2347 goto errout;
2348
2349 err = -EINVAL;
2350 rtm = nlmsg_data(nlh);
2351 memset(cfg, 0, sizeof(*cfg));
2352
2353 cfg->fc_table = rtm->rtm_table;
2354 cfg->fc_dst_len = rtm->rtm_dst_len;
2355 cfg->fc_src_len = rtm->rtm_src_len;
2356 cfg->fc_flags = RTF_UP;
2357 cfg->fc_protocol = rtm->rtm_protocol;
2358 cfg->fc_type = rtm->rtm_type;
2359
2360 if (rtm->rtm_type == RTN_UNREACHABLE ||
2361 rtm->rtm_type == RTN_BLACKHOLE ||
2362 rtm->rtm_type == RTN_PROHIBIT ||
2363 rtm->rtm_type == RTN_THROW)
2364 cfg->fc_flags |= RTF_REJECT;
2365
2366 if (rtm->rtm_type == RTN_LOCAL)
2367 cfg->fc_flags |= RTF_LOCAL;
2368
2369 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2370 cfg->fc_nlinfo.nlh = nlh;
2371 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2372
2373 if (tb[RTA_GATEWAY]) {
2374 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2375 cfg->fc_flags |= RTF_GATEWAY;
2376 }
2377
2378 if (tb[RTA_DST]) {
2379 int plen = (rtm->rtm_dst_len + 7) >> 3;
2380
2381 if (nla_len(tb[RTA_DST]) < plen)
2382 goto errout;
2383
2384 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2385 }
2386
2387 if (tb[RTA_SRC]) {
2388 int plen = (rtm->rtm_src_len + 7) >> 3;
2389
2390 if (nla_len(tb[RTA_SRC]) < plen)
2391 goto errout;
2392
2393 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2394 }
2395
2396 if (tb[RTA_PREFSRC])
2397 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2398
2399 if (tb[RTA_OIF])
2400 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2401
2402 if (tb[RTA_PRIORITY])
2403 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2404
2405 if (tb[RTA_METRICS]) {
2406 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2407 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2408 }
2409
2410 if (tb[RTA_TABLE])
2411 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2412
2413 if (tb[RTA_MULTIPATH]) {
2414 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2415 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2416 }
2417
2418 err = 0;
2419 errout:
2420 return err;
2421 }
2422
2423 static int ip6_route_multipath(struct fib6_config *cfg, int add)
2424 {
2425 struct fib6_config r_cfg;
2426 struct rtnexthop *rtnh;
2427 int remaining;
2428 int attrlen;
2429 int err = 0, last_err = 0;
2430
2431 beginning:
2432 rtnh = (struct rtnexthop *)cfg->fc_mp;
2433 remaining = cfg->fc_mp_len;
2434
2435 /* Parse a Multipath Entry */
2436 while (rtnh_ok(rtnh, remaining)) {
2437 memcpy(&r_cfg, cfg, sizeof(*cfg));
2438 if (rtnh->rtnh_ifindex)
2439 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2440
2441 attrlen = rtnh_attrlen(rtnh);
2442 if (attrlen > 0) {
2443 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2444
2445 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2446 if (nla) {
2447 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2448 r_cfg.fc_flags |= RTF_GATEWAY;
2449 }
2450 }
2451 err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
2452 if (err) {
2453 last_err = err;
2454 /* If we are trying to remove a route, do not stop the
2455 * loop when ip6_route_del() fails (because next hop is
2456 * already gone), we should try to remove all next hops.
2457 */
2458 if (add) {
2459 /* If add fails, we should try to delete all
2460 * next hops that have been already added.
2461 */
2462 add = 0;
2463 goto beginning;
2464 }
2465 }
2466 /* Because each route is added like a single route we remove
2467 * this flag after the first nexthop (if there is a collision,
2468 * we have already fail to add the first nexthop:
2469 * fib6_add_rt2node() has reject it).
2470 */
2471 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
2472 rtnh = rtnh_next(rtnh, &remaining);
2473 }
2474
2475 return last_err;
2476 }
2477
2478 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2479 {
2480 struct fib6_config cfg;
2481 int err;
2482
2483 err = rtm_to_fib6_config(skb, nlh, &cfg);
2484 if (err < 0)
2485 return err;
2486
2487 if (cfg.fc_mp)
2488 return ip6_route_multipath(&cfg, 0);
2489 else
2490 return ip6_route_del(&cfg);
2491 }
2492
2493 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
2494 {
2495 struct fib6_config cfg;
2496 int err;
2497
2498 err = rtm_to_fib6_config(skb, nlh, &cfg);
2499 if (err < 0)
2500 return err;
2501
2502 if (cfg.fc_mp)
2503 return ip6_route_multipath(&cfg, 1);
2504 else
2505 return ip6_route_add(&cfg);
2506 }
2507
2508 static inline size_t rt6_nlmsg_size(void)
2509 {
2510 return NLMSG_ALIGN(sizeof(struct rtmsg))
2511 + nla_total_size(16) /* RTA_SRC */
2512 + nla_total_size(16) /* RTA_DST */
2513 + nla_total_size(16) /* RTA_GATEWAY */
2514 + nla_total_size(16) /* RTA_PREFSRC */
2515 + nla_total_size(4) /* RTA_TABLE */
2516 + nla_total_size(4) /* RTA_IIF */
2517 + nla_total_size(4) /* RTA_OIF */
2518 + nla_total_size(4) /* RTA_PRIORITY */
2519 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2520 + nla_total_size(sizeof(struct rta_cacheinfo));
2521 }
2522
2523 static int rt6_fill_node(struct net *net,
2524 struct sk_buff *skb, struct rt6_info *rt,
2525 struct in6_addr *dst, struct in6_addr *src,
2526 int iif, int type, u32 portid, u32 seq,
2527 int prefix, int nowait, unsigned int flags)
2528 {
2529 struct rtmsg *rtm;
2530 struct nlmsghdr *nlh;
2531 long expires;
2532 u32 table;
2533
2534 if (prefix) { /* user wants prefix routes only */
2535 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2536 /* success since this is not a prefix route */
2537 return 1;
2538 }
2539 }
2540
2541 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2542 if (!nlh)
2543 return -EMSGSIZE;
2544
2545 rtm = nlmsg_data(nlh);
2546 rtm->rtm_family = AF_INET6;
2547 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2548 rtm->rtm_src_len = rt->rt6i_src.plen;
2549 rtm->rtm_tos = 0;
2550 if (rt->rt6i_table)
2551 table = rt->rt6i_table->tb6_id;
2552 else
2553 table = RT6_TABLE_UNSPEC;
2554 rtm->rtm_table = table;
2555 if (nla_put_u32(skb, RTA_TABLE, table))
2556 goto nla_put_failure;
2557 if (rt->rt6i_flags & RTF_REJECT) {
2558 switch (rt->dst.error) {
2559 case -EINVAL:
2560 rtm->rtm_type = RTN_BLACKHOLE;
2561 break;
2562 case -EACCES:
2563 rtm->rtm_type = RTN_PROHIBIT;
2564 break;
2565 case -EAGAIN:
2566 rtm->rtm_type = RTN_THROW;
2567 break;
2568 default:
2569 rtm->rtm_type = RTN_UNREACHABLE;
2570 break;
2571 }
2572 }
2573 else if (rt->rt6i_flags & RTF_LOCAL)
2574 rtm->rtm_type = RTN_LOCAL;
2575 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2576 rtm->rtm_type = RTN_LOCAL;
2577 else
2578 rtm->rtm_type = RTN_UNICAST;
2579 rtm->rtm_flags = 0;
2580 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2581 rtm->rtm_protocol = rt->rt6i_protocol;
2582 if (rt->rt6i_flags & RTF_DYNAMIC)
2583 rtm->rtm_protocol = RTPROT_REDIRECT;
2584 else if (rt->rt6i_flags & RTF_ADDRCONF) {
2585 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2586 rtm->rtm_protocol = RTPROT_RA;
2587 else
2588 rtm->rtm_protocol = RTPROT_KERNEL;
2589 }
2590
2591 if (rt->rt6i_flags & RTF_CACHE)
2592 rtm->rtm_flags |= RTM_F_CLONED;
2593
2594 if (dst) {
2595 if (nla_put(skb, RTA_DST, 16, dst))
2596 goto nla_put_failure;
2597 rtm->rtm_dst_len = 128;
2598 } else if (rtm->rtm_dst_len)
2599 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2600 goto nla_put_failure;
2601 #ifdef CONFIG_IPV6_SUBTREES
2602 if (src) {
2603 if (nla_put(skb, RTA_SRC, 16, src))
2604 goto nla_put_failure;
2605 rtm->rtm_src_len = 128;
2606 } else if (rtm->rtm_src_len &&
2607 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2608 goto nla_put_failure;
2609 #endif
2610 if (iif) {
2611 #ifdef CONFIG_IPV6_MROUTE
2612 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2613 int err = ip6mr_get_route(net, skb, rtm, nowait);
2614 if (err <= 0) {
2615 if (!nowait) {
2616 if (err == 0)
2617 return 0;
2618 goto nla_put_failure;
2619 } else {
2620 if (err == -EMSGSIZE)
2621 goto nla_put_failure;
2622 }
2623 }
2624 } else
2625 #endif
2626 if (nla_put_u32(skb, RTA_IIF, iif))
2627 goto nla_put_failure;
2628 } else if (dst) {
2629 struct in6_addr saddr_buf;
2630 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2631 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2632 goto nla_put_failure;
2633 }
2634
2635 if (rt->rt6i_prefsrc.plen) {
2636 struct in6_addr saddr_buf;
2637 saddr_buf = rt->rt6i_prefsrc.addr;
2638 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2639 goto nla_put_failure;
2640 }
2641
2642 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2643 goto nla_put_failure;
2644
2645 if (rt->rt6i_flags & RTF_GATEWAY) {
2646 if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
2647 goto nla_put_failure;
2648 }
2649
2650 if (rt->dst.dev &&
2651 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2652 goto nla_put_failure;
2653 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2654 goto nla_put_failure;
2655
2656 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
2657
2658 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2659 goto nla_put_failure;
2660
2661 return nlmsg_end(skb, nlh);
2662
2663 nla_put_failure:
2664 nlmsg_cancel(skb, nlh);
2665 return -EMSGSIZE;
2666 }
2667
2668 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2669 {
2670 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2671 int prefix;
2672
2673 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2674 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2675 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2676 } else
2677 prefix = 0;
2678
2679 return rt6_fill_node(arg->net,
2680 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2681 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2682 prefix, 0, NLM_F_MULTI);
2683 }
2684
2685 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
2686 {
2687 struct net *net = sock_net(in_skb->sk);
2688 struct nlattr *tb[RTA_MAX+1];
2689 struct rt6_info *rt;
2690 struct sk_buff *skb;
2691 struct rtmsg *rtm;
2692 struct flowi6 fl6;
2693 int err, iif = 0, oif = 0;
2694
2695 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2696 if (err < 0)
2697 goto errout;
2698
2699 err = -EINVAL;
2700 memset(&fl6, 0, sizeof(fl6));
2701
2702 if (tb[RTA_SRC]) {
2703 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2704 goto errout;
2705
2706 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2707 }
2708
2709 if (tb[RTA_DST]) {
2710 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2711 goto errout;
2712
2713 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2714 }
2715
2716 if (tb[RTA_IIF])
2717 iif = nla_get_u32(tb[RTA_IIF]);
2718
2719 if (tb[RTA_OIF])
2720 oif = nla_get_u32(tb[RTA_OIF]);
2721
2722 if (iif) {
2723 struct net_device *dev;
2724 int flags = 0;
2725
2726 dev = __dev_get_by_index(net, iif);
2727 if (!dev) {
2728 err = -ENODEV;
2729 goto errout;
2730 }
2731
2732 fl6.flowi6_iif = iif;
2733
2734 if (!ipv6_addr_any(&fl6.saddr))
2735 flags |= RT6_LOOKUP_F_HAS_SADDR;
2736
2737 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2738 flags);
2739 } else {
2740 fl6.flowi6_oif = oif;
2741
2742 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2743 }
2744
2745 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2746 if (!skb) {
2747 ip6_rt_put(rt);
2748 err = -ENOBUFS;
2749 goto errout;
2750 }
2751
2752 /* Reserve room for dummy headers, this skb can pass
2753 through good chunk of routing engine.
2754 */
2755 skb_reset_mac_header(skb);
2756 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2757
2758 skb_dst_set(skb, &rt->dst);
2759
2760 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2761 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2762 nlh->nlmsg_seq, 0, 0, 0);
2763 if (err < 0) {
2764 kfree_skb(skb);
2765 goto errout;
2766 }
2767
2768 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2769 errout:
2770 return err;
2771 }
2772
2773 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2774 {
2775 struct sk_buff *skb;
2776 struct net *net = info->nl_net;
2777 u32 seq;
2778 int err;
2779
2780 err = -ENOBUFS;
2781 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2782
2783 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2784 if (!skb)
2785 goto errout;
2786
2787 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2788 event, info->portid, seq, 0, 0, 0);
2789 if (err < 0) {
2790 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2791 WARN_ON(err == -EMSGSIZE);
2792 kfree_skb(skb);
2793 goto errout;
2794 }
2795 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2796 info->nlh, gfp_any());
2797 return;
2798 errout:
2799 if (err < 0)
2800 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2801 }
2802
2803 static int ip6_route_dev_notify(struct notifier_block *this,
2804 unsigned long event, void *ptr)
2805 {
2806 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2807 struct net *net = dev_net(dev);
2808
2809 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2810 net->ipv6.ip6_null_entry->dst.dev = dev;
2811 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2812 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2813 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2814 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2815 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2816 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2817 #endif
2818 }
2819
2820 return NOTIFY_OK;
2821 }
2822
2823 /*
2824 * /proc
2825 */
2826
2827 #ifdef CONFIG_PROC_FS
2828
2829 static const struct file_operations ipv6_route_proc_fops = {
2830 .owner = THIS_MODULE,
2831 .open = ipv6_route_open,
2832 .read = seq_read,
2833 .llseek = seq_lseek,
2834 .release = seq_release_net,
2835 };
2836
2837 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2838 {
2839 struct net *net = (struct net *)seq->private;
2840 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2841 net->ipv6.rt6_stats->fib_nodes,
2842 net->ipv6.rt6_stats->fib_route_nodes,
2843 net->ipv6.rt6_stats->fib_rt_alloc,
2844 net->ipv6.rt6_stats->fib_rt_entries,
2845 net->ipv6.rt6_stats->fib_rt_cache,
2846 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2847 net->ipv6.rt6_stats->fib_discarded_routes);
2848
2849 return 0;
2850 }
2851
2852 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2853 {
2854 return single_open_net(inode, file, rt6_stats_seq_show);
2855 }
2856
2857 static const struct file_operations rt6_stats_seq_fops = {
2858 .owner = THIS_MODULE,
2859 .open = rt6_stats_seq_open,
2860 .read = seq_read,
2861 .llseek = seq_lseek,
2862 .release = single_release_net,
2863 };
2864 #endif /* CONFIG_PROC_FS */
2865
2866 #ifdef CONFIG_SYSCTL
2867
2868 static
2869 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2870 void __user *buffer, size_t *lenp, loff_t *ppos)
2871 {
2872 struct net *net;
2873 int delay;
2874 if (!write)
2875 return -EINVAL;
2876
2877 net = (struct net *)ctl->extra1;
2878 delay = net->ipv6.sysctl.flush_delay;
2879 proc_dointvec(ctl, write, buffer, lenp, ppos);
2880 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2881 return 0;
2882 }
2883
2884 struct ctl_table ipv6_route_table_template[] = {
2885 {
2886 .procname = "flush",
2887 .data = &init_net.ipv6.sysctl.flush_delay,
2888 .maxlen = sizeof(int),
2889 .mode = 0200,
2890 .proc_handler = ipv6_sysctl_rtcache_flush
2891 },
2892 {
2893 .procname = "gc_thresh",
2894 .data = &ip6_dst_ops_template.gc_thresh,
2895 .maxlen = sizeof(int),
2896 .mode = 0644,
2897 .proc_handler = proc_dointvec,
2898 },
2899 {
2900 .procname = "max_size",
2901 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2902 .maxlen = sizeof(int),
2903 .mode = 0644,
2904 .proc_handler = proc_dointvec,
2905 },
2906 {
2907 .procname = "gc_min_interval",
2908 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2909 .maxlen = sizeof(int),
2910 .mode = 0644,
2911 .proc_handler = proc_dointvec_jiffies,
2912 },
2913 {
2914 .procname = "gc_timeout",
2915 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2916 .maxlen = sizeof(int),
2917 .mode = 0644,
2918 .proc_handler = proc_dointvec_jiffies,
2919 },
2920 {
2921 .procname = "gc_interval",
2922 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2923 .maxlen = sizeof(int),
2924 .mode = 0644,
2925 .proc_handler = proc_dointvec_jiffies,
2926 },
2927 {
2928 .procname = "gc_elasticity",
2929 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2930 .maxlen = sizeof(int),
2931 .mode = 0644,
2932 .proc_handler = proc_dointvec,
2933 },
2934 {
2935 .procname = "mtu_expires",
2936 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2937 .maxlen = sizeof(int),
2938 .mode = 0644,
2939 .proc_handler = proc_dointvec_jiffies,
2940 },
2941 {
2942 .procname = "min_adv_mss",
2943 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2944 .maxlen = sizeof(int),
2945 .mode = 0644,
2946 .proc_handler = proc_dointvec,
2947 },
2948 {
2949 .procname = "gc_min_interval_ms",
2950 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2951 .maxlen = sizeof(int),
2952 .mode = 0644,
2953 .proc_handler = proc_dointvec_ms_jiffies,
2954 },
2955 { }
2956 };
2957
2958 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2959 {
2960 struct ctl_table *table;
2961
2962 table = kmemdup(ipv6_route_table_template,
2963 sizeof(ipv6_route_table_template),
2964 GFP_KERNEL);
2965
2966 if (table) {
2967 table[0].data = &net->ipv6.sysctl.flush_delay;
2968 table[0].extra1 = net;
2969 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2970 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2971 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2972 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2973 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2974 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2975 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2976 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2977 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2978
2979 /* Don't export sysctls to unprivileged users */
2980 if (net->user_ns != &init_user_ns)
2981 table[0].procname = NULL;
2982 }
2983
2984 return table;
2985 }
2986 #endif
2987
2988 static int __net_init ip6_route_net_init(struct net *net)
2989 {
2990 int ret = -ENOMEM;
2991
2992 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2993 sizeof(net->ipv6.ip6_dst_ops));
2994
2995 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2996 goto out_ip6_dst_ops;
2997
2998 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2999 sizeof(*net->ipv6.ip6_null_entry),
3000 GFP_KERNEL);
3001 if (!net->ipv6.ip6_null_entry)
3002 goto out_ip6_dst_entries;
3003 net->ipv6.ip6_null_entry->dst.path =
3004 (struct dst_entry *)net->ipv6.ip6_null_entry;
3005 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3006 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3007 ip6_template_metrics, true);
3008
3009 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3010 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3011 sizeof(*net->ipv6.ip6_prohibit_entry),
3012 GFP_KERNEL);
3013 if (!net->ipv6.ip6_prohibit_entry)
3014 goto out_ip6_null_entry;
3015 net->ipv6.ip6_prohibit_entry->dst.path =
3016 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3017 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3018 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3019 ip6_template_metrics, true);
3020
3021 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3022 sizeof(*net->ipv6.ip6_blk_hole_entry),
3023 GFP_KERNEL);
3024 if (!net->ipv6.ip6_blk_hole_entry)
3025 goto out_ip6_prohibit_entry;
3026 net->ipv6.ip6_blk_hole_entry->dst.path =
3027 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3028 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3029 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3030 ip6_template_metrics, true);
3031 #endif
3032
3033 net->ipv6.sysctl.flush_delay = 0;
3034 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3035 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3036 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3037 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3038 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3039 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3040 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3041
3042 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3043
3044 ret = 0;
3045 out:
3046 return ret;
3047
3048 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3049 out_ip6_prohibit_entry:
3050 kfree(net->ipv6.ip6_prohibit_entry);
3051 out_ip6_null_entry:
3052 kfree(net->ipv6.ip6_null_entry);
3053 #endif
3054 out_ip6_dst_entries:
3055 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3056 out_ip6_dst_ops:
3057 goto out;
3058 }
3059
3060 static void __net_exit ip6_route_net_exit(struct net *net)
3061 {
3062 kfree(net->ipv6.ip6_null_entry);
3063 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3064 kfree(net->ipv6.ip6_prohibit_entry);
3065 kfree(net->ipv6.ip6_blk_hole_entry);
3066 #endif
3067 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3068 }
3069
3070 static int __net_init ip6_route_net_init_late(struct net *net)
3071 {
3072 #ifdef CONFIG_PROC_FS
3073 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3074 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3075 #endif
3076 return 0;
3077 }
3078
3079 static void __net_exit ip6_route_net_exit_late(struct net *net)
3080 {
3081 #ifdef CONFIG_PROC_FS
3082 remove_proc_entry("ipv6_route", net->proc_net);
3083 remove_proc_entry("rt6_stats", net->proc_net);
3084 #endif
3085 }
3086
3087 static struct pernet_operations ip6_route_net_ops = {
3088 .init = ip6_route_net_init,
3089 .exit = ip6_route_net_exit,
3090 };
3091
3092 static int __net_init ipv6_inetpeer_init(struct net *net)
3093 {
3094 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3095
3096 if (!bp)
3097 return -ENOMEM;
3098 inet_peer_base_init(bp);
3099 net->ipv6.peers = bp;
3100 return 0;
3101 }
3102
3103 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3104 {
3105 struct inet_peer_base *bp = net->ipv6.peers;
3106
3107 net->ipv6.peers = NULL;
3108 inetpeer_invalidate_tree(bp);
3109 kfree(bp);
3110 }
3111
3112 static struct pernet_operations ipv6_inetpeer_ops = {
3113 .init = ipv6_inetpeer_init,
3114 .exit = ipv6_inetpeer_exit,
3115 };
3116
3117 static struct pernet_operations ip6_route_net_late_ops = {
3118 .init = ip6_route_net_init_late,
3119 .exit = ip6_route_net_exit_late,
3120 };
3121
3122 static struct notifier_block ip6_route_dev_notifier = {
3123 .notifier_call = ip6_route_dev_notify,
3124 .priority = 0,
3125 };
3126
3127 int __init ip6_route_init(void)
3128 {
3129 int ret;
3130
3131 ret = -ENOMEM;
3132 ip6_dst_ops_template.kmem_cachep =
3133 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3134 SLAB_HWCACHE_ALIGN, NULL);
3135 if (!ip6_dst_ops_template.kmem_cachep)
3136 goto out;
3137
3138 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3139 if (ret)
3140 goto out_kmem_cache;
3141
3142 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3143 if (ret)
3144 goto out_dst_entries;
3145
3146 ret = register_pernet_subsys(&ip6_route_net_ops);
3147 if (ret)
3148 goto out_register_inetpeer;
3149
3150 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3151
3152 /* Registering of the loopback is done before this portion of code,
3153 * the loopback reference in rt6_info will not be taken, do it
3154 * manually for init_net */
3155 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3156 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3157 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3158 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3159 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3160 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3161 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3162 #endif
3163 ret = fib6_init();
3164 if (ret)
3165 goto out_register_subsys;
3166
3167 ret = xfrm6_init();
3168 if (ret)
3169 goto out_fib6_init;
3170
3171 ret = fib6_rules_init();
3172 if (ret)
3173 goto xfrm6_init;
3174
3175 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3176 if (ret)
3177 goto fib6_rules_init;
3178
3179 ret = -ENOBUFS;
3180 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3181 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3182 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3183 goto out_register_late_subsys;
3184
3185 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3186 if (ret)
3187 goto out_register_late_subsys;
3188
3189 out:
3190 return ret;
3191
3192 out_register_late_subsys:
3193 unregister_pernet_subsys(&ip6_route_net_late_ops);
3194 fib6_rules_init:
3195 fib6_rules_cleanup();
3196 xfrm6_init:
3197 xfrm6_fini();
3198 out_fib6_init:
3199 fib6_gc_cleanup();
3200 out_register_subsys:
3201 unregister_pernet_subsys(&ip6_route_net_ops);
3202 out_register_inetpeer:
3203 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3204 out_dst_entries:
3205 dst_entries_destroy(&ip6_dst_blackhole_ops);
3206 out_kmem_cache:
3207 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3208 goto out;
3209 }
3210
3211 void ip6_route_cleanup(void)
3212 {
3213 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3214 unregister_pernet_subsys(&ip6_route_net_late_ops);
3215 fib6_rules_cleanup();
3216 xfrm6_fini();
3217 fib6_gc_cleanup();
3218 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3219 unregister_pernet_subsys(&ip6_route_net_ops);
3220 dst_entries_destroy(&ip6_dst_blackhole_ops);
3221 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3222 }
Linux kernel - Deliverables
This page took 0.135811 seconds and 4 git commands to generate.