Loading...
1/*
2 * Linux INET6 implementation
3 * FIB front-end.
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14/* Changes:
15 *
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
23 * Ville Nuorvala
24 * Fixed routing subtrees.
25 */
26
27#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/dst_metadata.h>
58#include <net/xfrm.h>
59#include <net/netevent.h>
60#include <net/netlink.h>
61#include <net/nexthop.h>
62#include <net/lwtunnel.h>
63#include <net/ip_tunnels.h>
64#include <net/l3mdev.h>
65#include <trace/events/fib6.h>
66
67#include <asm/uaccess.h>
68
69#ifdef CONFIG_SYSCTL
70#include <linux/sysctl.h>
71#endif
72
73enum rt6_nud_state {
74 RT6_NUD_FAIL_HARD = -3,
75 RT6_NUD_FAIL_PROBE = -2,
76 RT6_NUD_FAIL_DO_RR = -1,
77 RT6_NUD_SUCCEED = 1
78};
79
80static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
81static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
82static unsigned int ip6_default_advmss(const struct dst_entry *dst);
83static unsigned int ip6_mtu(const struct dst_entry *dst);
84static struct dst_entry *ip6_negative_advice(struct dst_entry *);
85static void ip6_dst_destroy(struct dst_entry *);
86static void ip6_dst_ifdown(struct dst_entry *,
87 struct net_device *dev, int how);
88static int ip6_dst_gc(struct dst_ops *ops);
89
90static int ip6_pkt_discard(struct sk_buff *skb);
91static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
92static int ip6_pkt_prohibit(struct sk_buff *skb);
93static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
94static void ip6_link_failure(struct sk_buff *skb);
95static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
96 struct sk_buff *skb, u32 mtu);
97static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
98 struct sk_buff *skb);
99static void rt6_dst_from_metrics_check(struct rt6_info *rt);
100static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
101
102#ifdef CONFIG_IPV6_ROUTE_INFO
103static struct rt6_info *rt6_add_route_info(struct net *net,
104 const struct in6_addr *prefix, int prefixlen,
105 const struct in6_addr *gwaddr, int ifindex,
106 unsigned int pref);
107static struct rt6_info *rt6_get_route_info(struct net *net,
108 const struct in6_addr *prefix, int prefixlen,
109 const struct in6_addr *gwaddr, int ifindex);
110#endif
111
112struct uncached_list {
113 spinlock_t lock;
114 struct list_head head;
115};
116
117static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
118
119static void rt6_uncached_list_add(struct rt6_info *rt)
120{
121 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
122
123 rt->dst.flags |= DST_NOCACHE;
124 rt->rt6i_uncached_list = ul;
125
126 spin_lock_bh(&ul->lock);
127 list_add_tail(&rt->rt6i_uncached, &ul->head);
128 spin_unlock_bh(&ul->lock);
129}
130
131static void rt6_uncached_list_del(struct rt6_info *rt)
132{
133 if (!list_empty(&rt->rt6i_uncached)) {
134 struct uncached_list *ul = rt->rt6i_uncached_list;
135
136 spin_lock_bh(&ul->lock);
137 list_del(&rt->rt6i_uncached);
138 spin_unlock_bh(&ul->lock);
139 }
140}
141
142static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
143{
144 struct net_device *loopback_dev = net->loopback_dev;
145 int cpu;
146
147 if (dev == loopback_dev)
148 return;
149
150 for_each_possible_cpu(cpu) {
151 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
152 struct rt6_info *rt;
153
154 spin_lock_bh(&ul->lock);
155 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
156 struct inet6_dev *rt_idev = rt->rt6i_idev;
157 struct net_device *rt_dev = rt->dst.dev;
158
159 if (rt_idev->dev == dev) {
160 rt->rt6i_idev = in6_dev_get(loopback_dev);
161 in6_dev_put(rt_idev);
162 }
163
164 if (rt_dev == dev) {
165 rt->dst.dev = loopback_dev;
166 dev_hold(rt->dst.dev);
167 dev_put(rt_dev);
168 }
169 }
170 spin_unlock_bh(&ul->lock);
171 }
172}
173
174static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
175{
176 return dst_metrics_write_ptr(rt->dst.from);
177}
178
179static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
180{
181 struct rt6_info *rt = (struct rt6_info *)dst;
182
183 if (rt->rt6i_flags & RTF_PCPU)
184 return rt6_pcpu_cow_metrics(rt);
185 else if (rt->rt6i_flags & RTF_CACHE)
186 return NULL;
187 else
188 return dst_cow_metrics_generic(dst, old);
189}
190
191static inline const void *choose_neigh_daddr(struct rt6_info *rt,
192 struct sk_buff *skb,
193 const void *daddr)
194{
195 struct in6_addr *p = &rt->rt6i_gateway;
196
197 if (!ipv6_addr_any(p))
198 return (const void *) p;
199 else if (skb)
200 return &ipv6_hdr(skb)->daddr;
201 return daddr;
202}
203
204static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
205 struct sk_buff *skb,
206 const void *daddr)
207{
208 struct rt6_info *rt = (struct rt6_info *) dst;
209 struct neighbour *n;
210
211 daddr = choose_neigh_daddr(rt, skb, daddr);
212 n = __ipv6_neigh_lookup(dst->dev, daddr);
213 if (n)
214 return n;
215 return neigh_create(&nd_tbl, daddr, dst->dev);
216}
217
218static struct dst_ops ip6_dst_ops_template = {
219 .family = AF_INET6,
220 .gc = ip6_dst_gc,
221 .gc_thresh = 1024,
222 .check = ip6_dst_check,
223 .default_advmss = ip6_default_advmss,
224 .mtu = ip6_mtu,
225 .cow_metrics = ipv6_cow_metrics,
226 .destroy = ip6_dst_destroy,
227 .ifdown = ip6_dst_ifdown,
228 .negative_advice = ip6_negative_advice,
229 .link_failure = ip6_link_failure,
230 .update_pmtu = ip6_rt_update_pmtu,
231 .redirect = rt6_do_redirect,
232 .local_out = __ip6_local_out,
233 .neigh_lookup = ip6_neigh_lookup,
234};
235
236static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
237{
238 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
239
240 return mtu ? : dst->dev->mtu;
241}
242
243static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
244 struct sk_buff *skb, u32 mtu)
245{
246}
247
248static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
249 struct sk_buff *skb)
250{
251}
252
253static struct dst_ops ip6_dst_blackhole_ops = {
254 .family = AF_INET6,
255 .destroy = ip6_dst_destroy,
256 .check = ip6_dst_check,
257 .mtu = ip6_blackhole_mtu,
258 .default_advmss = ip6_default_advmss,
259 .update_pmtu = ip6_rt_blackhole_update_pmtu,
260 .redirect = ip6_rt_blackhole_redirect,
261 .cow_metrics = dst_cow_metrics_generic,
262 .neigh_lookup = ip6_neigh_lookup,
263};
264
265static const u32 ip6_template_metrics[RTAX_MAX] = {
266 [RTAX_HOPLIMIT - 1] = 0,
267};
268
269static const struct rt6_info ip6_null_entry_template = {
270 .dst = {
271 .__refcnt = ATOMIC_INIT(1),
272 .__use = 1,
273 .obsolete = DST_OBSOLETE_FORCE_CHK,
274 .error = -ENETUNREACH,
275 .input = ip6_pkt_discard,
276 .output = ip6_pkt_discard_out,
277 },
278 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
279 .rt6i_protocol = RTPROT_KERNEL,
280 .rt6i_metric = ~(u32) 0,
281 .rt6i_ref = ATOMIC_INIT(1),
282};
283
284#ifdef CONFIG_IPV6_MULTIPLE_TABLES
285
286static const struct rt6_info ip6_prohibit_entry_template = {
287 .dst = {
288 .__refcnt = ATOMIC_INIT(1),
289 .__use = 1,
290 .obsolete = DST_OBSOLETE_FORCE_CHK,
291 .error = -EACCES,
292 .input = ip6_pkt_prohibit,
293 .output = ip6_pkt_prohibit_out,
294 },
295 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
296 .rt6i_protocol = RTPROT_KERNEL,
297 .rt6i_metric = ~(u32) 0,
298 .rt6i_ref = ATOMIC_INIT(1),
299};
300
301static const struct rt6_info ip6_blk_hole_entry_template = {
302 .dst = {
303 .__refcnt = ATOMIC_INIT(1),
304 .__use = 1,
305 .obsolete = DST_OBSOLETE_FORCE_CHK,
306 .error = -EINVAL,
307 .input = dst_discard,
308 .output = dst_discard_out,
309 },
310 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
311 .rt6i_protocol = RTPROT_KERNEL,
312 .rt6i_metric = ~(u32) 0,
313 .rt6i_ref = ATOMIC_INIT(1),
314};
315
316#endif
317
318static void rt6_info_init(struct rt6_info *rt)
319{
320 struct dst_entry *dst = &rt->dst;
321
322 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
323 INIT_LIST_HEAD(&rt->rt6i_siblings);
324 INIT_LIST_HEAD(&rt->rt6i_uncached);
325}
326
327/* allocate dst with ip6_dst_ops */
328static struct rt6_info *__ip6_dst_alloc(struct net *net,
329 struct net_device *dev,
330 int flags)
331{
332 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
333 0, DST_OBSOLETE_FORCE_CHK, flags);
334
335 if (rt)
336 rt6_info_init(rt);
337
338 return rt;
339}
340
341struct rt6_info *ip6_dst_alloc(struct net *net,
342 struct net_device *dev,
343 int flags)
344{
345 struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
346
347 if (rt) {
348 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
349 if (rt->rt6i_pcpu) {
350 int cpu;
351
352 for_each_possible_cpu(cpu) {
353 struct rt6_info **p;
354
355 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
356 /* no one shares rt */
357 *p = NULL;
358 }
359 } else {
360 dst_destroy((struct dst_entry *)rt);
361 return NULL;
362 }
363 }
364
365 return rt;
366}
367EXPORT_SYMBOL(ip6_dst_alloc);
368
369static void ip6_dst_destroy(struct dst_entry *dst)
370{
371 struct rt6_info *rt = (struct rt6_info *)dst;
372 struct dst_entry *from = dst->from;
373 struct inet6_dev *idev;
374
375 dst_destroy_metrics_generic(dst);
376 free_percpu(rt->rt6i_pcpu);
377 rt6_uncached_list_del(rt);
378
379 idev = rt->rt6i_idev;
380 if (idev) {
381 rt->rt6i_idev = NULL;
382 in6_dev_put(idev);
383 }
384
385 dst->from = NULL;
386 dst_release(from);
387}
388
389static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
390 int how)
391{
392 struct rt6_info *rt = (struct rt6_info *)dst;
393 struct inet6_dev *idev = rt->rt6i_idev;
394 struct net_device *loopback_dev =
395 dev_net(dev)->loopback_dev;
396
397 if (dev != loopback_dev) {
398 if (idev && idev->dev == dev) {
399 struct inet6_dev *loopback_idev =
400 in6_dev_get(loopback_dev);
401 if (loopback_idev) {
402 rt->rt6i_idev = loopback_idev;
403 in6_dev_put(idev);
404 }
405 }
406 }
407}
408
409static bool __rt6_check_expired(const struct rt6_info *rt)
410{
411 if (rt->rt6i_flags & RTF_EXPIRES)
412 return time_after(jiffies, rt->dst.expires);
413 else
414 return false;
415}
416
417static bool rt6_check_expired(const struct rt6_info *rt)
418{
419 if (rt->rt6i_flags & RTF_EXPIRES) {
420 if (time_after(jiffies, rt->dst.expires))
421 return true;
422 } else if (rt->dst.from) {
423 return rt6_check_expired((struct rt6_info *) rt->dst.from);
424 }
425 return false;
426}
427
428/* Multipath route selection:
429 * Hash based function using packet header and flowlabel.
430 * Adapted from fib_info_hashfn()
431 */
432static int rt6_info_hash_nhsfn(unsigned int candidate_count,
433 const struct flowi6 *fl6)
434{
435 return get_hash_from_flowi6(fl6) % candidate_count;
436}
437
438static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
439 struct flowi6 *fl6, int oif,
440 int strict)
441{
442 struct rt6_info *sibling, *next_sibling;
443 int route_choosen;
444
445 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
446 /* Don't change the route, if route_choosen == 0
447 * (siblings does not include ourself)
448 */
449 if (route_choosen)
450 list_for_each_entry_safe(sibling, next_sibling,
451 &match->rt6i_siblings, rt6i_siblings) {
452 route_choosen--;
453 if (route_choosen == 0) {
454 if (rt6_score_route(sibling, oif, strict) < 0)
455 break;
456 match = sibling;
457 break;
458 }
459 }
460 return match;
461}
462
463/*
464 * Route lookup. Any table->tb6_lock is implied.
465 */
466
467static inline struct rt6_info *rt6_device_match(struct net *net,
468 struct rt6_info *rt,
469 const struct in6_addr *saddr,
470 int oif,
471 int flags)
472{
473 struct rt6_info *local = NULL;
474 struct rt6_info *sprt;
475
476 if (!oif && ipv6_addr_any(saddr))
477 goto out;
478
479 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
480 struct net_device *dev = sprt->dst.dev;
481
482 if (oif) {
483 if (dev->ifindex == oif)
484 return sprt;
485 if (dev->flags & IFF_LOOPBACK) {
486 if (!sprt->rt6i_idev ||
487 sprt->rt6i_idev->dev->ifindex != oif) {
488 if (flags & RT6_LOOKUP_F_IFACE)
489 continue;
490 if (local &&
491 local->rt6i_idev->dev->ifindex == oif)
492 continue;
493 }
494 local = sprt;
495 }
496 } else {
497 if (ipv6_chk_addr(net, saddr, dev,
498 flags & RT6_LOOKUP_F_IFACE))
499 return sprt;
500 }
501 }
502
503 if (oif) {
504 if (local)
505 return local;
506
507 if (flags & RT6_LOOKUP_F_IFACE)
508 return net->ipv6.ip6_null_entry;
509 }
510out:
511 return rt;
512}
513
514#ifdef CONFIG_IPV6_ROUTER_PREF
515struct __rt6_probe_work {
516 struct work_struct work;
517 struct in6_addr target;
518 struct net_device *dev;
519};
520
521static void rt6_probe_deferred(struct work_struct *w)
522{
523 struct in6_addr mcaddr;
524 struct __rt6_probe_work *work =
525 container_of(w, struct __rt6_probe_work, work);
526
527 addrconf_addr_solict_mult(&work->target, &mcaddr);
528 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL);
529 dev_put(work->dev);
530 kfree(work);
531}
532
533static void rt6_probe(struct rt6_info *rt)
534{
535 struct __rt6_probe_work *work;
536 struct neighbour *neigh;
537 /*
538 * Okay, this does not seem to be appropriate
539 * for now, however, we need to check if it
540 * is really so; aka Router Reachability Probing.
541 *
542 * Router Reachability Probe MUST be rate-limited
543 * to no more than one per minute.
544 */
545 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
546 return;
547 rcu_read_lock_bh();
548 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
549 if (neigh) {
550 if (neigh->nud_state & NUD_VALID)
551 goto out;
552
553 work = NULL;
554 write_lock(&neigh->lock);
555 if (!(neigh->nud_state & NUD_VALID) &&
556 time_after(jiffies,
557 neigh->updated +
558 rt->rt6i_idev->cnf.rtr_probe_interval)) {
559 work = kmalloc(sizeof(*work), GFP_ATOMIC);
560 if (work)
561 __neigh_set_probe_once(neigh);
562 }
563 write_unlock(&neigh->lock);
564 } else {
565 work = kmalloc(sizeof(*work), GFP_ATOMIC);
566 }
567
568 if (work) {
569 INIT_WORK(&work->work, rt6_probe_deferred);
570 work->target = rt->rt6i_gateway;
571 dev_hold(rt->dst.dev);
572 work->dev = rt->dst.dev;
573 schedule_work(&work->work);
574 }
575
576out:
577 rcu_read_unlock_bh();
578}
579#else
580static inline void rt6_probe(struct rt6_info *rt)
581{
582}
583#endif
584
585/*
586 * Default Router Selection (RFC 2461 6.3.6)
587 */
588static inline int rt6_check_dev(struct rt6_info *rt, int oif)
589{
590 struct net_device *dev = rt->dst.dev;
591 if (!oif || dev->ifindex == oif)
592 return 2;
593 if ((dev->flags & IFF_LOOPBACK) &&
594 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
595 return 1;
596 return 0;
597}
598
599static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
600{
601 struct neighbour *neigh;
602 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
603
604 if (rt->rt6i_flags & RTF_NONEXTHOP ||
605 !(rt->rt6i_flags & RTF_GATEWAY))
606 return RT6_NUD_SUCCEED;
607
608 rcu_read_lock_bh();
609 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
610 if (neigh) {
611 read_lock(&neigh->lock);
612 if (neigh->nud_state & NUD_VALID)
613 ret = RT6_NUD_SUCCEED;
614#ifdef CONFIG_IPV6_ROUTER_PREF
615 else if (!(neigh->nud_state & NUD_FAILED))
616 ret = RT6_NUD_SUCCEED;
617 else
618 ret = RT6_NUD_FAIL_PROBE;
619#endif
620 read_unlock(&neigh->lock);
621 } else {
622 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
623 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
624 }
625 rcu_read_unlock_bh();
626
627 return ret;
628}
629
630static int rt6_score_route(struct rt6_info *rt, int oif,
631 int strict)
632{
633 int m;
634
635 m = rt6_check_dev(rt, oif);
636 if (!m && (strict & RT6_LOOKUP_F_IFACE))
637 return RT6_NUD_FAIL_HARD;
638#ifdef CONFIG_IPV6_ROUTER_PREF
639 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
640#endif
641 if (strict & RT6_LOOKUP_F_REACHABLE) {
642 int n = rt6_check_neigh(rt);
643 if (n < 0)
644 return n;
645 }
646 return m;
647}
648
649static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
650 int *mpri, struct rt6_info *match,
651 bool *do_rr)
652{
653 int m;
654 bool match_do_rr = false;
655 struct inet6_dev *idev = rt->rt6i_idev;
656 struct net_device *dev = rt->dst.dev;
657
658 if (dev && !netif_carrier_ok(dev) &&
659 idev->cnf.ignore_routes_with_linkdown)
660 goto out;
661
662 if (rt6_check_expired(rt))
663 goto out;
664
665 m = rt6_score_route(rt, oif, strict);
666 if (m == RT6_NUD_FAIL_DO_RR) {
667 match_do_rr = true;
668 m = 0; /* lowest valid score */
669 } else if (m == RT6_NUD_FAIL_HARD) {
670 goto out;
671 }
672
673 if (strict & RT6_LOOKUP_F_REACHABLE)
674 rt6_probe(rt);
675
676 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
677 if (m > *mpri) {
678 *do_rr = match_do_rr;
679 *mpri = m;
680 match = rt;
681 }
682out:
683 return match;
684}
685
686static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
687 struct rt6_info *rr_head,
688 u32 metric, int oif, int strict,
689 bool *do_rr)
690{
691 struct rt6_info *rt, *match, *cont;
692 int mpri = -1;
693
694 match = NULL;
695 cont = NULL;
696 for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
697 if (rt->rt6i_metric != metric) {
698 cont = rt;
699 break;
700 }
701
702 match = find_match(rt, oif, strict, &mpri, match, do_rr);
703 }
704
705 for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
706 if (rt->rt6i_metric != metric) {
707 cont = rt;
708 break;
709 }
710
711 match = find_match(rt, oif, strict, &mpri, match, do_rr);
712 }
713
714 if (match || !cont)
715 return match;
716
717 for (rt = cont; rt; rt = rt->dst.rt6_next)
718 match = find_match(rt, oif, strict, &mpri, match, do_rr);
719
720 return match;
721}
722
723static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
724{
725 struct rt6_info *match, *rt0;
726 struct net *net;
727 bool do_rr = false;
728
729 rt0 = fn->rr_ptr;
730 if (!rt0)
731 fn->rr_ptr = rt0 = fn->leaf;
732
733 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
734 &do_rr);
735
736 if (do_rr) {
737 struct rt6_info *next = rt0->dst.rt6_next;
738
739 /* no entries matched; do round-robin */
740 if (!next || next->rt6i_metric != rt0->rt6i_metric)
741 next = fn->leaf;
742
743 if (next != rt0)
744 fn->rr_ptr = next;
745 }
746
747 net = dev_net(rt0->dst.dev);
748 return match ? match : net->ipv6.ip6_null_entry;
749}
750
751static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
752{
753 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
754}
755
756#ifdef CONFIG_IPV6_ROUTE_INFO
757int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
758 const struct in6_addr *gwaddr)
759{
760 struct net *net = dev_net(dev);
761 struct route_info *rinfo = (struct route_info *) opt;
762 struct in6_addr prefix_buf, *prefix;
763 unsigned int pref;
764 unsigned long lifetime;
765 struct rt6_info *rt;
766
767 if (len < sizeof(struct route_info)) {
768 return -EINVAL;
769 }
770
771 /* Sanity check for prefix_len and length */
772 if (rinfo->length > 3) {
773 return -EINVAL;
774 } else if (rinfo->prefix_len > 128) {
775 return -EINVAL;
776 } else if (rinfo->prefix_len > 64) {
777 if (rinfo->length < 2) {
778 return -EINVAL;
779 }
780 } else if (rinfo->prefix_len > 0) {
781 if (rinfo->length < 1) {
782 return -EINVAL;
783 }
784 }
785
786 pref = rinfo->route_pref;
787 if (pref == ICMPV6_ROUTER_PREF_INVALID)
788 return -EINVAL;
789
790 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
791
792 if (rinfo->length == 3)
793 prefix = (struct in6_addr *)rinfo->prefix;
794 else {
795 /* this function is safe */
796 ipv6_addr_prefix(&prefix_buf,
797 (struct in6_addr *)rinfo->prefix,
798 rinfo->prefix_len);
799 prefix = &prefix_buf;
800 }
801
802 if (rinfo->prefix_len == 0)
803 rt = rt6_get_dflt_router(gwaddr, dev);
804 else
805 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
806 gwaddr, dev->ifindex);
807
808 if (rt && !lifetime) {
809 ip6_del_rt(rt);
810 rt = NULL;
811 }
812
813 if (!rt && lifetime)
814 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
815 pref);
816 else if (rt)
817 rt->rt6i_flags = RTF_ROUTEINFO |
818 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
819
820 if (rt) {
821 if (!addrconf_finite_timeout(lifetime))
822 rt6_clean_expires(rt);
823 else
824 rt6_set_expires(rt, jiffies + HZ * lifetime);
825
826 ip6_rt_put(rt);
827 }
828 return 0;
829}
830#endif
831
832static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
833 struct in6_addr *saddr)
834{
835 struct fib6_node *pn;
836 while (1) {
837 if (fn->fn_flags & RTN_TL_ROOT)
838 return NULL;
839 pn = fn->parent;
840 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
841 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
842 else
843 fn = pn;
844 if (fn->fn_flags & RTN_RTINFO)
845 return fn;
846 }
847}
848
849static struct rt6_info *ip6_pol_route_lookup(struct net *net,
850 struct fib6_table *table,
851 struct flowi6 *fl6, int flags)
852{
853 struct fib6_node *fn;
854 struct rt6_info *rt;
855
856 read_lock_bh(&table->tb6_lock);
857 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
858restart:
859 rt = fn->leaf;
860 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
861 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
862 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
863 if (rt == net->ipv6.ip6_null_entry) {
864 fn = fib6_backtrack(fn, &fl6->saddr);
865 if (fn)
866 goto restart;
867 }
868 dst_use(&rt->dst, jiffies);
869 read_unlock_bh(&table->tb6_lock);
870
871 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
872
873 return rt;
874
875}
876
877struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
878 int flags)
879{
880 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
881}
882EXPORT_SYMBOL_GPL(ip6_route_lookup);
883
884struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
885 const struct in6_addr *saddr, int oif, int strict)
886{
887 struct flowi6 fl6 = {
888 .flowi6_oif = oif,
889 .daddr = *daddr,
890 };
891 struct dst_entry *dst;
892 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
893
894 if (saddr) {
895 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
896 flags |= RT6_LOOKUP_F_HAS_SADDR;
897 }
898
899 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
900 if (dst->error == 0)
901 return (struct rt6_info *) dst;
902
903 dst_release(dst);
904
905 return NULL;
906}
907EXPORT_SYMBOL(rt6_lookup);
908
909/* ip6_ins_rt is called with FREE table->tb6_lock.
910 It takes new route entry, the addition fails by any reason the
911 route is freed. In any case, if caller does not hold it, it may
912 be destroyed.
913 */
914
915static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
916 struct mx6_config *mxc)
917{
918 int err;
919 struct fib6_table *table;
920
921 table = rt->rt6i_table;
922 write_lock_bh(&table->tb6_lock);
923 err = fib6_add(&table->tb6_root, rt, info, mxc);
924 write_unlock_bh(&table->tb6_lock);
925
926 return err;
927}
928
929int ip6_ins_rt(struct rt6_info *rt)
930{
931 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
932 struct mx6_config mxc = { .mx = NULL, };
933
934 return __ip6_ins_rt(rt, &info, &mxc);
935}
936
937static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
938 const struct in6_addr *daddr,
939 const struct in6_addr *saddr)
940{
941 struct rt6_info *rt;
942
943 /*
944 * Clone the route.
945 */
946
947 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
948 ort = (struct rt6_info *)ort->dst.from;
949
950 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
951
952 if (!rt)
953 return NULL;
954
955 ip6_rt_copy_init(rt, ort);
956 rt->rt6i_flags |= RTF_CACHE;
957 rt->rt6i_metric = 0;
958 rt->dst.flags |= DST_HOST;
959 rt->rt6i_dst.addr = *daddr;
960 rt->rt6i_dst.plen = 128;
961
962 if (!rt6_is_gw_or_nonexthop(ort)) {
963 if (ort->rt6i_dst.plen != 128 &&
964 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
965 rt->rt6i_flags |= RTF_ANYCAST;
966#ifdef CONFIG_IPV6_SUBTREES
967 if (rt->rt6i_src.plen && saddr) {
968 rt->rt6i_src.addr = *saddr;
969 rt->rt6i_src.plen = 128;
970 }
971#endif
972 }
973
974 return rt;
975}
976
977static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
978{
979 struct rt6_info *pcpu_rt;
980
981 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
982 rt->dst.dev, rt->dst.flags);
983
984 if (!pcpu_rt)
985 return NULL;
986 ip6_rt_copy_init(pcpu_rt, rt);
987 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
988 pcpu_rt->rt6i_flags |= RTF_PCPU;
989 return pcpu_rt;
990}
991
992/* It should be called with read_lock_bh(&tb6_lock) acquired */
993static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
994{
995 struct rt6_info *pcpu_rt, **p;
996
997 p = this_cpu_ptr(rt->rt6i_pcpu);
998 pcpu_rt = *p;
999
1000 if (pcpu_rt) {
1001 dst_hold(&pcpu_rt->dst);
1002 rt6_dst_from_metrics_check(pcpu_rt);
1003 }
1004 return pcpu_rt;
1005}
1006
1007static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1008{
1009 struct fib6_table *table = rt->rt6i_table;
1010 struct rt6_info *pcpu_rt, *prev, **p;
1011
1012 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1013 if (!pcpu_rt) {
1014 struct net *net = dev_net(rt->dst.dev);
1015
1016 dst_hold(&net->ipv6.ip6_null_entry->dst);
1017 return net->ipv6.ip6_null_entry;
1018 }
1019
1020 read_lock_bh(&table->tb6_lock);
1021 if (rt->rt6i_pcpu) {
1022 p = this_cpu_ptr(rt->rt6i_pcpu);
1023 prev = cmpxchg(p, NULL, pcpu_rt);
1024 if (prev) {
1025 /* If someone did it before us, return prev instead */
1026 dst_destroy(&pcpu_rt->dst);
1027 pcpu_rt = prev;
1028 }
1029 } else {
1030 /* rt has been removed from the fib6 tree
1031 * before we have a chance to acquire the read_lock.
1032 * In this case, don't brother to create a pcpu rt
1033 * since rt is going away anyway. The next
1034 * dst_check() will trigger a re-lookup.
1035 */
1036 dst_destroy(&pcpu_rt->dst);
1037 pcpu_rt = rt;
1038 }
1039 dst_hold(&pcpu_rt->dst);
1040 rt6_dst_from_metrics_check(pcpu_rt);
1041 read_unlock_bh(&table->tb6_lock);
1042 return pcpu_rt;
1043}
1044
1045static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1046 struct flowi6 *fl6, int flags)
1047{
1048 struct fib6_node *fn, *saved_fn;
1049 struct rt6_info *rt;
1050 int strict = 0;
1051
1052 strict |= flags & RT6_LOOKUP_F_IFACE;
1053 if (net->ipv6.devconf_all->forwarding == 0)
1054 strict |= RT6_LOOKUP_F_REACHABLE;
1055
1056 read_lock_bh(&table->tb6_lock);
1057
1058 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1059 saved_fn = fn;
1060
1061 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1062 oif = 0;
1063
1064redo_rt6_select:
1065 rt = rt6_select(fn, oif, strict);
1066 if (rt->rt6i_nsiblings)
1067 rt = rt6_multipath_select(rt, fl6, oif, strict);
1068 if (rt == net->ipv6.ip6_null_entry) {
1069 fn = fib6_backtrack(fn, &fl6->saddr);
1070 if (fn)
1071 goto redo_rt6_select;
1072 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1073 /* also consider unreachable route */
1074 strict &= ~RT6_LOOKUP_F_REACHABLE;
1075 fn = saved_fn;
1076 goto redo_rt6_select;
1077 }
1078 }
1079
1080
1081 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1082 dst_use(&rt->dst, jiffies);
1083 read_unlock_bh(&table->tb6_lock);
1084
1085 rt6_dst_from_metrics_check(rt);
1086
1087 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1088 return rt;
1089 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1090 !(rt->rt6i_flags & RTF_GATEWAY))) {
1091 /* Create a RTF_CACHE clone which will not be
1092 * owned by the fib6 tree. It is for the special case where
1093 * the daddr in the skb during the neighbor look-up is different
1094 * from the fl6->daddr used to look-up route here.
1095 */
1096
1097 struct rt6_info *uncached_rt;
1098
1099 dst_use(&rt->dst, jiffies);
1100 read_unlock_bh(&table->tb6_lock);
1101
1102 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1103 dst_release(&rt->dst);
1104
1105 if (uncached_rt)
1106 rt6_uncached_list_add(uncached_rt);
1107 else
1108 uncached_rt = net->ipv6.ip6_null_entry;
1109
1110 dst_hold(&uncached_rt->dst);
1111
1112 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1113 return uncached_rt;
1114
1115 } else {
1116 /* Get a percpu copy */
1117
1118 struct rt6_info *pcpu_rt;
1119
1120 rt->dst.lastuse = jiffies;
1121 rt->dst.__use++;
1122 pcpu_rt = rt6_get_pcpu_route(rt);
1123
1124 if (pcpu_rt) {
1125 read_unlock_bh(&table->tb6_lock);
1126 } else {
1127 /* We have to do the read_unlock first
1128 * because rt6_make_pcpu_route() may trigger
1129 * ip6_dst_gc() which will take the write_lock.
1130 */
1131 dst_hold(&rt->dst);
1132 read_unlock_bh(&table->tb6_lock);
1133 pcpu_rt = rt6_make_pcpu_route(rt);
1134 dst_release(&rt->dst);
1135 }
1136
1137 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1138 return pcpu_rt;
1139
1140 }
1141}
1142
1143static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1144 struct flowi6 *fl6, int flags)
1145{
1146 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1147}
1148
1149static struct dst_entry *ip6_route_input_lookup(struct net *net,
1150 struct net_device *dev,
1151 struct flowi6 *fl6, int flags)
1152{
1153 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1154 flags |= RT6_LOOKUP_F_IFACE;
1155
1156 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1157}
1158
1159void ip6_route_input(struct sk_buff *skb)
1160{
1161 const struct ipv6hdr *iph = ipv6_hdr(skb);
1162 struct net *net = dev_net(skb->dev);
1163 int flags = RT6_LOOKUP_F_HAS_SADDR;
1164 struct ip_tunnel_info *tun_info;
1165 struct flowi6 fl6 = {
1166 .flowi6_iif = l3mdev_fib_oif(skb->dev),
1167 .daddr = iph->daddr,
1168 .saddr = iph->saddr,
1169 .flowlabel = ip6_flowinfo(iph),
1170 .flowi6_mark = skb->mark,
1171 .flowi6_proto = iph->nexthdr,
1172 };
1173
1174 tun_info = skb_tunnel_info(skb);
1175 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1176 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1177 skb_dst_drop(skb);
1178 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1179}
1180
1181static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1182 struct flowi6 *fl6, int flags)
1183{
1184 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1185}
1186
1187struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1188 struct flowi6 *fl6, int flags)
1189{
1190 struct dst_entry *dst;
1191 bool any_src;
1192
1193 dst = l3mdev_rt6_dst_by_oif(net, fl6);
1194 if (dst)
1195 return dst;
1196
1197 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1198
1199 any_src = ipv6_addr_any(&fl6->saddr);
1200 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1201 (fl6->flowi6_oif && any_src))
1202 flags |= RT6_LOOKUP_F_IFACE;
1203
1204 if (!any_src)
1205 flags |= RT6_LOOKUP_F_HAS_SADDR;
1206 else if (sk)
1207 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1208
1209 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1210}
1211EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1212
1213struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1214{
1215 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1216 struct dst_entry *new = NULL;
1217
1218 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1219 if (rt) {
1220 rt6_info_init(rt);
1221
1222 new = &rt->dst;
1223 new->__use = 1;
1224 new->input = dst_discard;
1225 new->output = dst_discard_out;
1226
1227 dst_copy_metrics(new, &ort->dst);
1228 rt->rt6i_idev = ort->rt6i_idev;
1229 if (rt->rt6i_idev)
1230 in6_dev_hold(rt->rt6i_idev);
1231
1232 rt->rt6i_gateway = ort->rt6i_gateway;
1233 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1234 rt->rt6i_metric = 0;
1235
1236 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1237#ifdef CONFIG_IPV6_SUBTREES
1238 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1239#endif
1240
1241 dst_free(new);
1242 }
1243
1244 dst_release(dst_orig);
1245 return new ? new : ERR_PTR(-ENOMEM);
1246}
1247
1248/*
1249 * Destination cache support functions
1250 */
1251
1252static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1253{
1254 if (rt->dst.from &&
1255 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1256 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1257}
1258
1259static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1260{
1261 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1262 return NULL;
1263
1264 if (rt6_check_expired(rt))
1265 return NULL;
1266
1267 return &rt->dst;
1268}
1269
1270static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1271{
1272 if (!__rt6_check_expired(rt) &&
1273 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1274 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1275 return &rt->dst;
1276 else
1277 return NULL;
1278}
1279
1280static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1281{
1282 struct rt6_info *rt;
1283
1284 rt = (struct rt6_info *) dst;
1285
1286 /* All IPV6 dsts are created with ->obsolete set to the value
1287 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1288 * into this function always.
1289 */
1290
1291 rt6_dst_from_metrics_check(rt);
1292
1293 if (rt->rt6i_flags & RTF_PCPU ||
1294 (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1295 return rt6_dst_from_check(rt, cookie);
1296 else
1297 return rt6_check(rt, cookie);
1298}
1299
1300static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1301{
1302 struct rt6_info *rt = (struct rt6_info *) dst;
1303
1304 if (rt) {
1305 if (rt->rt6i_flags & RTF_CACHE) {
1306 if (rt6_check_expired(rt)) {
1307 ip6_del_rt(rt);
1308 dst = NULL;
1309 }
1310 } else {
1311 dst_release(dst);
1312 dst = NULL;
1313 }
1314 }
1315 return dst;
1316}
1317
1318static void ip6_link_failure(struct sk_buff *skb)
1319{
1320 struct rt6_info *rt;
1321
1322 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1323
1324 rt = (struct rt6_info *) skb_dst(skb);
1325 if (rt) {
1326 if (rt->rt6i_flags & RTF_CACHE) {
1327 dst_hold(&rt->dst);
1328 ip6_del_rt(rt);
1329 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1330 rt->rt6i_node->fn_sernum = -1;
1331 }
1332 }
1333}
1334
1335static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1336{
1337 struct net *net = dev_net(rt->dst.dev);
1338
1339 rt->rt6i_flags |= RTF_MODIFIED;
1340 rt->rt6i_pmtu = mtu;
1341 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1342}
1343
1344static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1345{
1346 return !(rt->rt6i_flags & RTF_CACHE) &&
1347 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1348}
1349
1350static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1351 const struct ipv6hdr *iph, u32 mtu)
1352{
1353 struct rt6_info *rt6 = (struct rt6_info *)dst;
1354
1355 if (rt6->rt6i_flags & RTF_LOCAL)
1356 return;
1357
1358 dst_confirm(dst);
1359 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1360 if (mtu >= dst_mtu(dst))
1361 return;
1362
1363 if (!rt6_cache_allowed_for_pmtu(rt6)) {
1364 rt6_do_update_pmtu(rt6, mtu);
1365 } else {
1366 const struct in6_addr *daddr, *saddr;
1367 struct rt6_info *nrt6;
1368
1369 if (iph) {
1370 daddr = &iph->daddr;
1371 saddr = &iph->saddr;
1372 } else if (sk) {
1373 daddr = &sk->sk_v6_daddr;
1374 saddr = &inet6_sk(sk)->saddr;
1375 } else {
1376 return;
1377 }
1378 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1379 if (nrt6) {
1380 rt6_do_update_pmtu(nrt6, mtu);
1381
1382 /* ip6_ins_rt(nrt6) will bump the
1383 * rt6->rt6i_node->fn_sernum
1384 * which will fail the next rt6_check() and
1385 * invalidate the sk->sk_dst_cache.
1386 */
1387 ip6_ins_rt(nrt6);
1388 }
1389 }
1390}
1391
1392static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1393 struct sk_buff *skb, u32 mtu)
1394{
1395 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1396}
1397
1398void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1399 int oif, u32 mark)
1400{
1401 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1402 struct dst_entry *dst;
1403 struct flowi6 fl6;
1404
1405 memset(&fl6, 0, sizeof(fl6));
1406 fl6.flowi6_oif = oif;
1407 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1408 fl6.daddr = iph->daddr;
1409 fl6.saddr = iph->saddr;
1410 fl6.flowlabel = ip6_flowinfo(iph);
1411
1412 dst = ip6_route_output(net, NULL, &fl6);
1413 if (!dst->error)
1414 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1415 dst_release(dst);
1416}
1417EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1418
1419void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1420{
1421 struct dst_entry *dst;
1422
1423 ip6_update_pmtu(skb, sock_net(sk), mtu,
1424 sk->sk_bound_dev_if, sk->sk_mark);
1425
1426 dst = __sk_dst_get(sk);
1427 if (!dst || !dst->obsolete ||
1428 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1429 return;
1430
1431 bh_lock_sock(sk);
1432 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1433 ip6_datagram_dst_update(sk, false);
1434 bh_unlock_sock(sk);
1435}
1436EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1437
1438/* Handle redirects */
1439struct ip6rd_flowi {
1440 struct flowi6 fl6;
1441 struct in6_addr gateway;
1442};
1443
1444static struct rt6_info *__ip6_route_redirect(struct net *net,
1445 struct fib6_table *table,
1446 struct flowi6 *fl6,
1447 int flags)
1448{
1449 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1450 struct rt6_info *rt;
1451 struct fib6_node *fn;
1452
1453 /* Get the "current" route for this destination and
1454 * check if the redirect has come from approriate router.
1455 *
1456 * RFC 4861 specifies that redirects should only be
1457 * accepted if they come from the nexthop to the target.
1458 * Due to the way the routes are chosen, this notion
1459 * is a bit fuzzy and one might need to check all possible
1460 * routes.
1461 */
1462
1463 read_lock_bh(&table->tb6_lock);
1464 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1465restart:
1466 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1467 if (rt6_check_expired(rt))
1468 continue;
1469 if (rt->dst.error)
1470 break;
1471 if (!(rt->rt6i_flags & RTF_GATEWAY))
1472 continue;
1473 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1474 continue;
1475 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1476 continue;
1477 break;
1478 }
1479
1480 if (!rt)
1481 rt = net->ipv6.ip6_null_entry;
1482 else if (rt->dst.error) {
1483 rt = net->ipv6.ip6_null_entry;
1484 goto out;
1485 }
1486
1487 if (rt == net->ipv6.ip6_null_entry) {
1488 fn = fib6_backtrack(fn, &fl6->saddr);
1489 if (fn)
1490 goto restart;
1491 }
1492
1493out:
1494 dst_hold(&rt->dst);
1495
1496 read_unlock_bh(&table->tb6_lock);
1497
1498 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1499 return rt;
1500};
1501
1502static struct dst_entry *ip6_route_redirect(struct net *net,
1503 const struct flowi6 *fl6,
1504 const struct in6_addr *gateway)
1505{
1506 int flags = RT6_LOOKUP_F_HAS_SADDR;
1507 struct ip6rd_flowi rdfl;
1508
1509 rdfl.fl6 = *fl6;
1510 rdfl.gateway = *gateway;
1511
1512 return fib6_rule_lookup(net, &rdfl.fl6,
1513 flags, __ip6_route_redirect);
1514}
1515
1516void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1517{
1518 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1519 struct dst_entry *dst;
1520 struct flowi6 fl6;
1521
1522 memset(&fl6, 0, sizeof(fl6));
1523 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1524 fl6.flowi6_oif = oif;
1525 fl6.flowi6_mark = mark;
1526 fl6.daddr = iph->daddr;
1527 fl6.saddr = iph->saddr;
1528 fl6.flowlabel = ip6_flowinfo(iph);
1529
1530 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1531 rt6_do_redirect(dst, NULL, skb);
1532 dst_release(dst);
1533}
1534EXPORT_SYMBOL_GPL(ip6_redirect);
1535
1536void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1537 u32 mark)
1538{
1539 const struct ipv6hdr *iph = ipv6_hdr(skb);
1540 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1541 struct dst_entry *dst;
1542 struct flowi6 fl6;
1543
1544 memset(&fl6, 0, sizeof(fl6));
1545 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1546 fl6.flowi6_oif = oif;
1547 fl6.flowi6_mark = mark;
1548 fl6.daddr = msg->dest;
1549 fl6.saddr = iph->daddr;
1550
1551 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1552 rt6_do_redirect(dst, NULL, skb);
1553 dst_release(dst);
1554}
1555
1556void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1557{
1558 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1559}
1560EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1561
1562static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1563{
1564 struct net_device *dev = dst->dev;
1565 unsigned int mtu = dst_mtu(dst);
1566 struct net *net = dev_net(dev);
1567
1568 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1569
1570 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1571 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1572
1573 /*
1574 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1575 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1576 * IPV6_MAXPLEN is also valid and means: "any MSS,
1577 * rely only on pmtu discovery"
1578 */
1579 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1580 mtu = IPV6_MAXPLEN;
1581 return mtu;
1582}
1583
1584static unsigned int ip6_mtu(const struct dst_entry *dst)
1585{
1586 const struct rt6_info *rt = (const struct rt6_info *)dst;
1587 unsigned int mtu = rt->rt6i_pmtu;
1588 struct inet6_dev *idev;
1589
1590 if (mtu)
1591 goto out;
1592
1593 mtu = dst_metric_raw(dst, RTAX_MTU);
1594 if (mtu)
1595 goto out;
1596
1597 mtu = IPV6_MIN_MTU;
1598
1599 rcu_read_lock();
1600 idev = __in6_dev_get(dst->dev);
1601 if (idev)
1602 mtu = idev->cnf.mtu6;
1603 rcu_read_unlock();
1604
1605out:
1606 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1607}
1608
1609static struct dst_entry *icmp6_dst_gc_list;
1610static DEFINE_SPINLOCK(icmp6_dst_lock);
1611
1612struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1613 struct flowi6 *fl6)
1614{
1615 struct dst_entry *dst;
1616 struct rt6_info *rt;
1617 struct inet6_dev *idev = in6_dev_get(dev);
1618 struct net *net = dev_net(dev);
1619
1620 if (unlikely(!idev))
1621 return ERR_PTR(-ENODEV);
1622
1623 rt = ip6_dst_alloc(net, dev, 0);
1624 if (unlikely(!rt)) {
1625 in6_dev_put(idev);
1626 dst = ERR_PTR(-ENOMEM);
1627 goto out;
1628 }
1629
1630 rt->dst.flags |= DST_HOST;
1631 rt->dst.output = ip6_output;
1632 atomic_set(&rt->dst.__refcnt, 1);
1633 rt->rt6i_gateway = fl6->daddr;
1634 rt->rt6i_dst.addr = fl6->daddr;
1635 rt->rt6i_dst.plen = 128;
1636 rt->rt6i_idev = idev;
1637 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1638
1639 spin_lock_bh(&icmp6_dst_lock);
1640 rt->dst.next = icmp6_dst_gc_list;
1641 icmp6_dst_gc_list = &rt->dst;
1642 spin_unlock_bh(&icmp6_dst_lock);
1643
1644 fib6_force_start_gc(net);
1645
1646 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1647
1648out:
1649 return dst;
1650}
1651
1652int icmp6_dst_gc(void)
1653{
1654 struct dst_entry *dst, **pprev;
1655 int more = 0;
1656
1657 spin_lock_bh(&icmp6_dst_lock);
1658 pprev = &icmp6_dst_gc_list;
1659
1660 while ((dst = *pprev) != NULL) {
1661 if (!atomic_read(&dst->__refcnt)) {
1662 *pprev = dst->next;
1663 dst_free(dst);
1664 } else {
1665 pprev = &dst->next;
1666 ++more;
1667 }
1668 }
1669
1670 spin_unlock_bh(&icmp6_dst_lock);
1671
1672 return more;
1673}
1674
1675static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1676 void *arg)
1677{
1678 struct dst_entry *dst, **pprev;
1679
1680 spin_lock_bh(&icmp6_dst_lock);
1681 pprev = &icmp6_dst_gc_list;
1682 while ((dst = *pprev) != NULL) {
1683 struct rt6_info *rt = (struct rt6_info *) dst;
1684 if (func(rt, arg)) {
1685 *pprev = dst->next;
1686 dst_free(dst);
1687 } else {
1688 pprev = &dst->next;
1689 }
1690 }
1691 spin_unlock_bh(&icmp6_dst_lock);
1692}
1693
1694static int ip6_dst_gc(struct dst_ops *ops)
1695{
1696 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1697 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1698 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1699 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1700 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1701 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1702 int entries;
1703
1704 entries = dst_entries_get_fast(ops);
1705 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1706 entries <= rt_max_size)
1707 goto out;
1708
1709 net->ipv6.ip6_rt_gc_expire++;
1710 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1711 entries = dst_entries_get_slow(ops);
1712 if (entries < ops->gc_thresh)
1713 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1714out:
1715 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1716 return entries > rt_max_size;
1717}
1718
1719static int ip6_convert_metrics(struct mx6_config *mxc,
1720 const struct fib6_config *cfg)
1721{
1722 bool ecn_ca = false;
1723 struct nlattr *nla;
1724 int remaining;
1725 u32 *mp;
1726
1727 if (!cfg->fc_mx)
1728 return 0;
1729
1730 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1731 if (unlikely(!mp))
1732 return -ENOMEM;
1733
1734 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1735 int type = nla_type(nla);
1736 u32 val;
1737
1738 if (!type)
1739 continue;
1740 if (unlikely(type > RTAX_MAX))
1741 goto err;
1742
1743 if (type == RTAX_CC_ALGO) {
1744 char tmp[TCP_CA_NAME_MAX];
1745
1746 nla_strlcpy(tmp, nla, sizeof(tmp));
1747 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1748 if (val == TCP_CA_UNSPEC)
1749 goto err;
1750 } else {
1751 val = nla_get_u32(nla);
1752 }
1753 if (type == RTAX_HOPLIMIT && val > 255)
1754 val = 255;
1755 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1756 goto err;
1757
1758 mp[type - 1] = val;
1759 __set_bit(type - 1, mxc->mx_valid);
1760 }
1761
1762 if (ecn_ca) {
1763 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1764 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1765 }
1766
1767 mxc->mx = mp;
1768 return 0;
1769 err:
1770 kfree(mp);
1771 return -EINVAL;
1772}
1773
1774static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1775{
1776 struct net *net = cfg->fc_nlinfo.nl_net;
1777 struct rt6_info *rt = NULL;
1778 struct net_device *dev = NULL;
1779 struct inet6_dev *idev = NULL;
1780 struct fib6_table *table;
1781 int addr_type;
1782 int err = -EINVAL;
1783
1784 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1785 goto out;
1786#ifndef CONFIG_IPV6_SUBTREES
1787 if (cfg->fc_src_len)
1788 goto out;
1789#endif
1790 if (cfg->fc_ifindex) {
1791 err = -ENODEV;
1792 dev = dev_get_by_index(net, cfg->fc_ifindex);
1793 if (!dev)
1794 goto out;
1795 idev = in6_dev_get(dev);
1796 if (!idev)
1797 goto out;
1798 }
1799
1800 if (cfg->fc_metric == 0)
1801 cfg->fc_metric = IP6_RT_PRIO_USER;
1802
1803 err = -ENOBUFS;
1804 if (cfg->fc_nlinfo.nlh &&
1805 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1806 table = fib6_get_table(net, cfg->fc_table);
1807 if (!table) {
1808 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1809 table = fib6_new_table(net, cfg->fc_table);
1810 }
1811 } else {
1812 table = fib6_new_table(net, cfg->fc_table);
1813 }
1814
1815 if (!table)
1816 goto out;
1817
1818 rt = ip6_dst_alloc(net, NULL,
1819 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1820
1821 if (!rt) {
1822 err = -ENOMEM;
1823 goto out;
1824 }
1825
1826 if (cfg->fc_flags & RTF_EXPIRES)
1827 rt6_set_expires(rt, jiffies +
1828 clock_t_to_jiffies(cfg->fc_expires));
1829 else
1830 rt6_clean_expires(rt);
1831
1832 if (cfg->fc_protocol == RTPROT_UNSPEC)
1833 cfg->fc_protocol = RTPROT_BOOT;
1834 rt->rt6i_protocol = cfg->fc_protocol;
1835
1836 addr_type = ipv6_addr_type(&cfg->fc_dst);
1837
1838 if (addr_type & IPV6_ADDR_MULTICAST)
1839 rt->dst.input = ip6_mc_input;
1840 else if (cfg->fc_flags & RTF_LOCAL)
1841 rt->dst.input = ip6_input;
1842 else
1843 rt->dst.input = ip6_forward;
1844
1845 rt->dst.output = ip6_output;
1846
1847 if (cfg->fc_encap) {
1848 struct lwtunnel_state *lwtstate;
1849
1850 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1851 cfg->fc_encap, AF_INET6, cfg,
1852 &lwtstate);
1853 if (err)
1854 goto out;
1855 rt->dst.lwtstate = lwtstate_get(lwtstate);
1856 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1857 rt->dst.lwtstate->orig_output = rt->dst.output;
1858 rt->dst.output = lwtunnel_output;
1859 }
1860 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1861 rt->dst.lwtstate->orig_input = rt->dst.input;
1862 rt->dst.input = lwtunnel_input;
1863 }
1864 }
1865
1866 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1867 rt->rt6i_dst.plen = cfg->fc_dst_len;
1868 if (rt->rt6i_dst.plen == 128)
1869 rt->dst.flags |= DST_HOST;
1870
1871#ifdef CONFIG_IPV6_SUBTREES
1872 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1873 rt->rt6i_src.plen = cfg->fc_src_len;
1874#endif
1875
1876 rt->rt6i_metric = cfg->fc_metric;
1877
1878 /* We cannot add true routes via loopback here,
1879 they would result in kernel looping; promote them to reject routes
1880 */
1881 if ((cfg->fc_flags & RTF_REJECT) ||
1882 (dev && (dev->flags & IFF_LOOPBACK) &&
1883 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1884 !(cfg->fc_flags & RTF_LOCAL))) {
1885 /* hold loopback dev/idev if we haven't done so. */
1886 if (dev != net->loopback_dev) {
1887 if (dev) {
1888 dev_put(dev);
1889 in6_dev_put(idev);
1890 }
1891 dev = net->loopback_dev;
1892 dev_hold(dev);
1893 idev = in6_dev_get(dev);
1894 if (!idev) {
1895 err = -ENODEV;
1896 goto out;
1897 }
1898 }
1899 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1900 switch (cfg->fc_type) {
1901 case RTN_BLACKHOLE:
1902 rt->dst.error = -EINVAL;
1903 rt->dst.output = dst_discard_out;
1904 rt->dst.input = dst_discard;
1905 break;
1906 case RTN_PROHIBIT:
1907 rt->dst.error = -EACCES;
1908 rt->dst.output = ip6_pkt_prohibit_out;
1909 rt->dst.input = ip6_pkt_prohibit;
1910 break;
1911 case RTN_THROW:
1912 case RTN_UNREACHABLE:
1913 default:
1914 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1915 : (cfg->fc_type == RTN_UNREACHABLE)
1916 ? -EHOSTUNREACH : -ENETUNREACH;
1917 rt->dst.output = ip6_pkt_discard_out;
1918 rt->dst.input = ip6_pkt_discard;
1919 break;
1920 }
1921 goto install_route;
1922 }
1923
1924 if (cfg->fc_flags & RTF_GATEWAY) {
1925 const struct in6_addr *gw_addr;
1926 int gwa_type;
1927
1928 gw_addr = &cfg->fc_gateway;
1929 gwa_type = ipv6_addr_type(gw_addr);
1930
1931 /* if gw_addr is local we will fail to detect this in case
1932 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1933 * will return already-added prefix route via interface that
1934 * prefix route was assigned to, which might be non-loopback.
1935 */
1936 err = -EINVAL;
1937 if (ipv6_chk_addr_and_flags(net, gw_addr,
1938 gwa_type & IPV6_ADDR_LINKLOCAL ?
1939 dev : NULL, 0, 0))
1940 goto out;
1941
1942 rt->rt6i_gateway = *gw_addr;
1943
1944 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1945 struct rt6_info *grt;
1946
1947 /* IPv6 strictly inhibits using not link-local
1948 addresses as nexthop address.
1949 Otherwise, router will not able to send redirects.
1950 It is very good, but in some (rare!) circumstances
1951 (SIT, PtP, NBMA NOARP links) it is handy to allow
1952 some exceptions. --ANK
1953 */
1954 if (!(gwa_type & IPV6_ADDR_UNICAST))
1955 goto out;
1956
1957 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1958
1959 err = -EHOSTUNREACH;
1960 if (!grt)
1961 goto out;
1962 if (dev) {
1963 if (dev != grt->dst.dev) {
1964 ip6_rt_put(grt);
1965 goto out;
1966 }
1967 } else {
1968 dev = grt->dst.dev;
1969 idev = grt->rt6i_idev;
1970 dev_hold(dev);
1971 in6_dev_hold(grt->rt6i_idev);
1972 }
1973 if (!(grt->rt6i_flags & RTF_GATEWAY))
1974 err = 0;
1975 ip6_rt_put(grt);
1976
1977 if (err)
1978 goto out;
1979 }
1980 err = -EINVAL;
1981 if (!dev || (dev->flags & IFF_LOOPBACK))
1982 goto out;
1983 }
1984
1985 err = -ENODEV;
1986 if (!dev)
1987 goto out;
1988
1989 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1990 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1991 err = -EINVAL;
1992 goto out;
1993 }
1994 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1995 rt->rt6i_prefsrc.plen = 128;
1996 } else
1997 rt->rt6i_prefsrc.plen = 0;
1998
1999 rt->rt6i_flags = cfg->fc_flags;
2000
2001install_route:
2002 rt->dst.dev = dev;
2003 rt->rt6i_idev = idev;
2004 rt->rt6i_table = table;
2005
2006 cfg->fc_nlinfo.nl_net = dev_net(dev);
2007
2008 return rt;
2009out:
2010 if (dev)
2011 dev_put(dev);
2012 if (idev)
2013 in6_dev_put(idev);
2014 if (rt)
2015 dst_free(&rt->dst);
2016
2017 return ERR_PTR(err);
2018}
2019
2020int ip6_route_add(struct fib6_config *cfg)
2021{
2022 struct mx6_config mxc = { .mx = NULL, };
2023 struct rt6_info *rt;
2024 int err;
2025
2026 rt = ip6_route_info_create(cfg);
2027 if (IS_ERR(rt)) {
2028 err = PTR_ERR(rt);
2029 rt = NULL;
2030 goto out;
2031 }
2032
2033 err = ip6_convert_metrics(&mxc, cfg);
2034 if (err)
2035 goto out;
2036
2037 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2038
2039 kfree(mxc.mx);
2040
2041 return err;
2042out:
2043 if (rt)
2044 dst_free(&rt->dst);
2045
2046 return err;
2047}
2048
2049static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2050{
2051 int err;
2052 struct fib6_table *table;
2053 struct net *net = dev_net(rt->dst.dev);
2054
2055 if (rt == net->ipv6.ip6_null_entry ||
2056 rt->dst.flags & DST_NOCACHE) {
2057 err = -ENOENT;
2058 goto out;
2059 }
2060
2061 table = rt->rt6i_table;
2062 write_lock_bh(&table->tb6_lock);
2063 err = fib6_del(rt, info);
2064 write_unlock_bh(&table->tb6_lock);
2065
2066out:
2067 ip6_rt_put(rt);
2068 return err;
2069}
2070
2071int ip6_del_rt(struct rt6_info *rt)
2072{
2073 struct nl_info info = {
2074 .nl_net = dev_net(rt->dst.dev),
2075 };
2076 return __ip6_del_rt(rt, &info);
2077}
2078
2079static int ip6_route_del(struct fib6_config *cfg)
2080{
2081 struct fib6_table *table;
2082 struct fib6_node *fn;
2083 struct rt6_info *rt;
2084 int err = -ESRCH;
2085
2086 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2087 if (!table)
2088 return err;
2089
2090 read_lock_bh(&table->tb6_lock);
2091
2092 fn = fib6_locate(&table->tb6_root,
2093 &cfg->fc_dst, cfg->fc_dst_len,
2094 &cfg->fc_src, cfg->fc_src_len);
2095
2096 if (fn) {
2097 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2098 if ((rt->rt6i_flags & RTF_CACHE) &&
2099 !(cfg->fc_flags & RTF_CACHE))
2100 continue;
2101 if (cfg->fc_ifindex &&
2102 (!rt->dst.dev ||
2103 rt->dst.dev->ifindex != cfg->fc_ifindex))
2104 continue;
2105 if (cfg->fc_flags & RTF_GATEWAY &&
2106 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2107 continue;
2108 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2109 continue;
2110 dst_hold(&rt->dst);
2111 read_unlock_bh(&table->tb6_lock);
2112
2113 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2114 }
2115 }
2116 read_unlock_bh(&table->tb6_lock);
2117
2118 return err;
2119}
2120
2121static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2122{
2123 struct netevent_redirect netevent;
2124 struct rt6_info *rt, *nrt = NULL;
2125 struct ndisc_options ndopts;
2126 struct inet6_dev *in6_dev;
2127 struct neighbour *neigh;
2128 struct rd_msg *msg;
2129 int optlen, on_link;
2130 u8 *lladdr;
2131
2132 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2133 optlen -= sizeof(*msg);
2134
2135 if (optlen < 0) {
2136 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2137 return;
2138 }
2139
2140 msg = (struct rd_msg *)icmp6_hdr(skb);
2141
2142 if (ipv6_addr_is_multicast(&msg->dest)) {
2143 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2144 return;
2145 }
2146
2147 on_link = 0;
2148 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2149 on_link = 1;
2150 } else if (ipv6_addr_type(&msg->target) !=
2151 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2152 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2153 return;
2154 }
2155
2156 in6_dev = __in6_dev_get(skb->dev);
2157 if (!in6_dev)
2158 return;
2159 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2160 return;
2161
2162 /* RFC2461 8.1:
2163 * The IP source address of the Redirect MUST be the same as the current
2164 * first-hop router for the specified ICMP Destination Address.
2165 */
2166
2167 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2168 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2169 return;
2170 }
2171
2172 lladdr = NULL;
2173 if (ndopts.nd_opts_tgt_lladdr) {
2174 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2175 skb->dev);
2176 if (!lladdr) {
2177 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2178 return;
2179 }
2180 }
2181
2182 rt = (struct rt6_info *) dst;
2183 if (rt->rt6i_flags & RTF_REJECT) {
2184 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2185 return;
2186 }
2187
2188 /* Redirect received -> path was valid.
2189 * Look, redirects are sent only in response to data packets,
2190 * so that this nexthop apparently is reachable. --ANK
2191 */
2192 dst_confirm(&rt->dst);
2193
2194 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2195 if (!neigh)
2196 return;
2197
2198 /*
2199 * We have finally decided to accept it.
2200 */
2201
2202 neigh_update(neigh, lladdr, NUD_STALE,
2203 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2204 NEIGH_UPDATE_F_OVERRIDE|
2205 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2206 NEIGH_UPDATE_F_ISROUTER))
2207 );
2208
2209 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2210 if (!nrt)
2211 goto out;
2212
2213 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2214 if (on_link)
2215 nrt->rt6i_flags &= ~RTF_GATEWAY;
2216
2217 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2218
2219 if (ip6_ins_rt(nrt))
2220 goto out;
2221
2222 netevent.old = &rt->dst;
2223 netevent.new = &nrt->dst;
2224 netevent.daddr = &msg->dest;
2225 netevent.neigh = neigh;
2226 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2227
2228 if (rt->rt6i_flags & RTF_CACHE) {
2229 rt = (struct rt6_info *) dst_clone(&rt->dst);
2230 ip6_del_rt(rt);
2231 }
2232
2233out:
2234 neigh_release(neigh);
2235}
2236
2237/*
2238 * Misc support functions
2239 */
2240
2241static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2242{
2243 BUG_ON(from->dst.from);
2244
2245 rt->rt6i_flags &= ~RTF_EXPIRES;
2246 dst_hold(&from->dst);
2247 rt->dst.from = &from->dst;
2248 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2249}
2250
2251static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2252{
2253 rt->dst.input = ort->dst.input;
2254 rt->dst.output = ort->dst.output;
2255 rt->rt6i_dst = ort->rt6i_dst;
2256 rt->dst.error = ort->dst.error;
2257 rt->rt6i_idev = ort->rt6i_idev;
2258 if (rt->rt6i_idev)
2259 in6_dev_hold(rt->rt6i_idev);
2260 rt->dst.lastuse = jiffies;
2261 rt->rt6i_gateway = ort->rt6i_gateway;
2262 rt->rt6i_flags = ort->rt6i_flags;
2263 rt6_set_from(rt, ort);
2264 rt->rt6i_metric = ort->rt6i_metric;
2265#ifdef CONFIG_IPV6_SUBTREES
2266 rt->rt6i_src = ort->rt6i_src;
2267#endif
2268 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2269 rt->rt6i_table = ort->rt6i_table;
2270 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2271}
2272
2273#ifdef CONFIG_IPV6_ROUTE_INFO
2274static struct rt6_info *rt6_get_route_info(struct net *net,
2275 const struct in6_addr *prefix, int prefixlen,
2276 const struct in6_addr *gwaddr, int ifindex)
2277{
2278 struct fib6_node *fn;
2279 struct rt6_info *rt = NULL;
2280 struct fib6_table *table;
2281
2282 table = fib6_get_table(net, RT6_TABLE_INFO);
2283 if (!table)
2284 return NULL;
2285
2286 read_lock_bh(&table->tb6_lock);
2287 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2288 if (!fn)
2289 goto out;
2290
2291 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2292 if (rt->dst.dev->ifindex != ifindex)
2293 continue;
2294 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2295 continue;
2296 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2297 continue;
2298 dst_hold(&rt->dst);
2299 break;
2300 }
2301out:
2302 read_unlock_bh(&table->tb6_lock);
2303 return rt;
2304}
2305
2306static struct rt6_info *rt6_add_route_info(struct net *net,
2307 const struct in6_addr *prefix, int prefixlen,
2308 const struct in6_addr *gwaddr, int ifindex,
2309 unsigned int pref)
2310{
2311 struct fib6_config cfg = {
2312 .fc_metric = IP6_RT_PRIO_USER,
2313 .fc_ifindex = ifindex,
2314 .fc_dst_len = prefixlen,
2315 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2316 RTF_UP | RTF_PREF(pref),
2317 .fc_nlinfo.portid = 0,
2318 .fc_nlinfo.nlh = NULL,
2319 .fc_nlinfo.nl_net = net,
2320 };
2321
2322 cfg.fc_table = l3mdev_fib_table_by_index(net, ifindex) ? : RT6_TABLE_INFO;
2323 cfg.fc_dst = *prefix;
2324 cfg.fc_gateway = *gwaddr;
2325
2326 /* We should treat it as a default route if prefix length is 0. */
2327 if (!prefixlen)
2328 cfg.fc_flags |= RTF_DEFAULT;
2329
2330 ip6_route_add(&cfg);
2331
2332 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2333}
2334#endif
2335
2336struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2337{
2338 struct rt6_info *rt;
2339 struct fib6_table *table;
2340
2341 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2342 if (!table)
2343 return NULL;
2344
2345 read_lock_bh(&table->tb6_lock);
2346 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2347 if (dev == rt->dst.dev &&
2348 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2349 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2350 break;
2351 }
2352 if (rt)
2353 dst_hold(&rt->dst);
2354 read_unlock_bh(&table->tb6_lock);
2355 return rt;
2356}
2357
2358struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2359 struct net_device *dev,
2360 unsigned int pref)
2361{
2362 struct fib6_config cfg = {
2363 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2364 .fc_metric = IP6_RT_PRIO_USER,
2365 .fc_ifindex = dev->ifindex,
2366 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2367 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2368 .fc_nlinfo.portid = 0,
2369 .fc_nlinfo.nlh = NULL,
2370 .fc_nlinfo.nl_net = dev_net(dev),
2371 };
2372
2373 cfg.fc_gateway = *gwaddr;
2374
2375 ip6_route_add(&cfg);
2376
2377 return rt6_get_dflt_router(gwaddr, dev);
2378}
2379
2380void rt6_purge_dflt_routers(struct net *net)
2381{
2382 struct rt6_info *rt;
2383 struct fib6_table *table;
2384
2385 /* NOTE: Keep consistent with rt6_get_dflt_router */
2386 table = fib6_get_table(net, RT6_TABLE_DFLT);
2387 if (!table)
2388 return;
2389
2390restart:
2391 read_lock_bh(&table->tb6_lock);
2392 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2393 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2394 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2395 dst_hold(&rt->dst);
2396 read_unlock_bh(&table->tb6_lock);
2397 ip6_del_rt(rt);
2398 goto restart;
2399 }
2400 }
2401 read_unlock_bh(&table->tb6_lock);
2402}
2403
2404static void rtmsg_to_fib6_config(struct net *net,
2405 struct in6_rtmsg *rtmsg,
2406 struct fib6_config *cfg)
2407{
2408 memset(cfg, 0, sizeof(*cfg));
2409
2410 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2411 : RT6_TABLE_MAIN;
2412 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2413 cfg->fc_metric = rtmsg->rtmsg_metric;
2414 cfg->fc_expires = rtmsg->rtmsg_info;
2415 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2416 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2417 cfg->fc_flags = rtmsg->rtmsg_flags;
2418
2419 cfg->fc_nlinfo.nl_net = net;
2420
2421 cfg->fc_dst = rtmsg->rtmsg_dst;
2422 cfg->fc_src = rtmsg->rtmsg_src;
2423 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2424}
2425
2426int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2427{
2428 struct fib6_config cfg;
2429 struct in6_rtmsg rtmsg;
2430 int err;
2431
2432 switch (cmd) {
2433 case SIOCADDRT: /* Add a route */
2434 case SIOCDELRT: /* Delete a route */
2435 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2436 return -EPERM;
2437 err = copy_from_user(&rtmsg, arg,
2438 sizeof(struct in6_rtmsg));
2439 if (err)
2440 return -EFAULT;
2441
2442 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2443
2444 rtnl_lock();
2445 switch (cmd) {
2446 case SIOCADDRT:
2447 err = ip6_route_add(&cfg);
2448 break;
2449 case SIOCDELRT:
2450 err = ip6_route_del(&cfg);
2451 break;
2452 default:
2453 err = -EINVAL;
2454 }
2455 rtnl_unlock();
2456
2457 return err;
2458 }
2459
2460 return -EINVAL;
2461}
2462
2463/*
2464 * Drop the packet on the floor
2465 */
2466
2467static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2468{
2469 int type;
2470 struct dst_entry *dst = skb_dst(skb);
2471 switch (ipstats_mib_noroutes) {
2472 case IPSTATS_MIB_INNOROUTES:
2473 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2474 if (type == IPV6_ADDR_ANY) {
2475 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2476 IPSTATS_MIB_INADDRERRORS);
2477 break;
2478 }
2479 /* FALLTHROUGH */
2480 case IPSTATS_MIB_OUTNOROUTES:
2481 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2482 ipstats_mib_noroutes);
2483 break;
2484 }
2485 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2486 kfree_skb(skb);
2487 return 0;
2488}
2489
2490static int ip6_pkt_discard(struct sk_buff *skb)
2491{
2492 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2493}
2494
2495static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2496{
2497 skb->dev = skb_dst(skb)->dev;
2498 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2499}
2500
2501static int ip6_pkt_prohibit(struct sk_buff *skb)
2502{
2503 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2504}
2505
2506static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2507{
2508 skb->dev = skb_dst(skb)->dev;
2509 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2510}
2511
2512/*
2513 * Allocate a dst for local (unicast / anycast) address.
2514 */
2515
2516struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2517 const struct in6_addr *addr,
2518 bool anycast)
2519{
2520 u32 tb_id;
2521 struct net *net = dev_net(idev->dev);
2522 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2523 DST_NOCOUNT);
2524 if (!rt)
2525 return ERR_PTR(-ENOMEM);
2526
2527 in6_dev_hold(idev);
2528
2529 rt->dst.flags |= DST_HOST;
2530 rt->dst.input = ip6_input;
2531 rt->dst.output = ip6_output;
2532 rt->rt6i_idev = idev;
2533
2534 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2535 if (anycast)
2536 rt->rt6i_flags |= RTF_ANYCAST;
2537 else
2538 rt->rt6i_flags |= RTF_LOCAL;
2539
2540 rt->rt6i_gateway = *addr;
2541 rt->rt6i_dst.addr = *addr;
2542 rt->rt6i_dst.plen = 128;
2543 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2544 rt->rt6i_table = fib6_get_table(net, tb_id);
2545 rt->dst.flags |= DST_NOCACHE;
2546
2547 atomic_set(&rt->dst.__refcnt, 1);
2548
2549 return rt;
2550}
2551
2552int ip6_route_get_saddr(struct net *net,
2553 struct rt6_info *rt,
2554 const struct in6_addr *daddr,
2555 unsigned int prefs,
2556 struct in6_addr *saddr)
2557{
2558 struct inet6_dev *idev =
2559 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2560 int err = 0;
2561 if (rt && rt->rt6i_prefsrc.plen)
2562 *saddr = rt->rt6i_prefsrc.addr;
2563 else
2564 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2565 daddr, prefs, saddr);
2566 return err;
2567}
2568
2569/* remove deleted ip from prefsrc entries */
2570struct arg_dev_net_ip {
2571 struct net_device *dev;
2572 struct net *net;
2573 struct in6_addr *addr;
2574};
2575
2576static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2577{
2578 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2579 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2580 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2581
2582 if (((void *)rt->dst.dev == dev || !dev) &&
2583 rt != net->ipv6.ip6_null_entry &&
2584 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2585 /* remove prefsrc entry */
2586 rt->rt6i_prefsrc.plen = 0;
2587 }
2588 return 0;
2589}
2590
2591void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2592{
2593 struct net *net = dev_net(ifp->idev->dev);
2594 struct arg_dev_net_ip adni = {
2595 .dev = ifp->idev->dev,
2596 .net = net,
2597 .addr = &ifp->addr,
2598 };
2599 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2600}
2601
2602#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2603#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2604
2605/* Remove routers and update dst entries when gateway turn into host. */
2606static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2607{
2608 struct in6_addr *gateway = (struct in6_addr *)arg;
2609
2610 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2611 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2612 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2613 return -1;
2614 }
2615 return 0;
2616}
2617
2618void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2619{
2620 fib6_clean_all(net, fib6_clean_tohost, gateway);
2621}
2622
2623struct arg_dev_net {
2624 struct net_device *dev;
2625 struct net *net;
2626};
2627
2628static int fib6_ifdown(struct rt6_info *rt, void *arg)
2629{
2630 const struct arg_dev_net *adn = arg;
2631 const struct net_device *dev = adn->dev;
2632
2633 if ((rt->dst.dev == dev || !dev) &&
2634 rt != adn->net->ipv6.ip6_null_entry)
2635 return -1;
2636
2637 return 0;
2638}
2639
2640void rt6_ifdown(struct net *net, struct net_device *dev)
2641{
2642 struct arg_dev_net adn = {
2643 .dev = dev,
2644 .net = net,
2645 };
2646
2647 fib6_clean_all(net, fib6_ifdown, &adn);
2648 icmp6_clean_all(fib6_ifdown, &adn);
2649 if (dev)
2650 rt6_uncached_list_flush_dev(net, dev);
2651}
2652
2653struct rt6_mtu_change_arg {
2654 struct net_device *dev;
2655 unsigned int mtu;
2656};
2657
2658static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2659{
2660 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2661 struct inet6_dev *idev;
2662
2663 /* In IPv6 pmtu discovery is not optional,
2664 so that RTAX_MTU lock cannot disable it.
2665 We still use this lock to block changes
2666 caused by addrconf/ndisc.
2667 */
2668
2669 idev = __in6_dev_get(arg->dev);
2670 if (!idev)
2671 return 0;
2672
2673 /* For administrative MTU increase, there is no way to discover
2674 IPv6 PMTU increase, so PMTU increase should be updated here.
2675 Since RFC 1981 doesn't include administrative MTU increase
2676 update PMTU increase is a MUST. (i.e. jumbo frame)
2677 */
2678 /*
2679 If new MTU is less than route PMTU, this new MTU will be the
2680 lowest MTU in the path, update the route PMTU to reflect PMTU
2681 decreases; if new MTU is greater than route PMTU, and the
2682 old MTU is the lowest MTU in the path, update the route PMTU
2683 to reflect the increase. In this case if the other nodes' MTU
2684 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2685 PMTU discouvery.
2686 */
2687 if (rt->dst.dev == arg->dev &&
2688 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2689 if (rt->rt6i_flags & RTF_CACHE) {
2690 /* For RTF_CACHE with rt6i_pmtu == 0
2691 * (i.e. a redirected route),
2692 * the metrics of its rt->dst.from has already
2693 * been updated.
2694 */
2695 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2696 rt->rt6i_pmtu = arg->mtu;
2697 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2698 (dst_mtu(&rt->dst) < arg->mtu &&
2699 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2700 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2701 }
2702 }
2703 return 0;
2704}
2705
2706void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2707{
2708 struct rt6_mtu_change_arg arg = {
2709 .dev = dev,
2710 .mtu = mtu,
2711 };
2712
2713 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2714}
2715
2716static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2717 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2718 [RTA_OIF] = { .type = NLA_U32 },
2719 [RTA_IIF] = { .type = NLA_U32 },
2720 [RTA_PRIORITY] = { .type = NLA_U32 },
2721 [RTA_METRICS] = { .type = NLA_NESTED },
2722 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2723 [RTA_PREF] = { .type = NLA_U8 },
2724 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2725 [RTA_ENCAP] = { .type = NLA_NESTED },
2726 [RTA_EXPIRES] = { .type = NLA_U32 },
2727};
2728
2729static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2730 struct fib6_config *cfg)
2731{
2732 struct rtmsg *rtm;
2733 struct nlattr *tb[RTA_MAX+1];
2734 unsigned int pref;
2735 int err;
2736
2737 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2738 if (err < 0)
2739 goto errout;
2740
2741 err = -EINVAL;
2742 rtm = nlmsg_data(nlh);
2743 memset(cfg, 0, sizeof(*cfg));
2744
2745 cfg->fc_table = rtm->rtm_table;
2746 cfg->fc_dst_len = rtm->rtm_dst_len;
2747 cfg->fc_src_len = rtm->rtm_src_len;
2748 cfg->fc_flags = RTF_UP;
2749 cfg->fc_protocol = rtm->rtm_protocol;
2750 cfg->fc_type = rtm->rtm_type;
2751
2752 if (rtm->rtm_type == RTN_UNREACHABLE ||
2753 rtm->rtm_type == RTN_BLACKHOLE ||
2754 rtm->rtm_type == RTN_PROHIBIT ||
2755 rtm->rtm_type == RTN_THROW)
2756 cfg->fc_flags |= RTF_REJECT;
2757
2758 if (rtm->rtm_type == RTN_LOCAL)
2759 cfg->fc_flags |= RTF_LOCAL;
2760
2761 if (rtm->rtm_flags & RTM_F_CLONED)
2762 cfg->fc_flags |= RTF_CACHE;
2763
2764 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2765 cfg->fc_nlinfo.nlh = nlh;
2766 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2767
2768 if (tb[RTA_GATEWAY]) {
2769 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2770 cfg->fc_flags |= RTF_GATEWAY;
2771 }
2772
2773 if (tb[RTA_DST]) {
2774 int plen = (rtm->rtm_dst_len + 7) >> 3;
2775
2776 if (nla_len(tb[RTA_DST]) < plen)
2777 goto errout;
2778
2779 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2780 }
2781
2782 if (tb[RTA_SRC]) {
2783 int plen = (rtm->rtm_src_len + 7) >> 3;
2784
2785 if (nla_len(tb[RTA_SRC]) < plen)
2786 goto errout;
2787
2788 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2789 }
2790
2791 if (tb[RTA_PREFSRC])
2792 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2793
2794 if (tb[RTA_OIF])
2795 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2796
2797 if (tb[RTA_PRIORITY])
2798 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2799
2800 if (tb[RTA_METRICS]) {
2801 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2802 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2803 }
2804
2805 if (tb[RTA_TABLE])
2806 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2807
2808 if (tb[RTA_MULTIPATH]) {
2809 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2810 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2811 }
2812
2813 if (tb[RTA_PREF]) {
2814 pref = nla_get_u8(tb[RTA_PREF]);
2815 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2816 pref != ICMPV6_ROUTER_PREF_HIGH)
2817 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2818 cfg->fc_flags |= RTF_PREF(pref);
2819 }
2820
2821 if (tb[RTA_ENCAP])
2822 cfg->fc_encap = tb[RTA_ENCAP];
2823
2824 if (tb[RTA_ENCAP_TYPE])
2825 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2826
2827 if (tb[RTA_EXPIRES]) {
2828 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
2829
2830 if (addrconf_finite_timeout(timeout)) {
2831 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
2832 cfg->fc_flags |= RTF_EXPIRES;
2833 }
2834 }
2835
2836 err = 0;
2837errout:
2838 return err;
2839}
2840
2841struct rt6_nh {
2842 struct rt6_info *rt6_info;
2843 struct fib6_config r_cfg;
2844 struct mx6_config mxc;
2845 struct list_head next;
2846};
2847
2848static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2849{
2850 struct rt6_nh *nh;
2851
2852 list_for_each_entry(nh, rt6_nh_list, next) {
2853 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2854 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2855 nh->r_cfg.fc_ifindex);
2856 }
2857}
2858
2859static int ip6_route_info_append(struct list_head *rt6_nh_list,
2860 struct rt6_info *rt, struct fib6_config *r_cfg)
2861{
2862 struct rt6_nh *nh;
2863 struct rt6_info *rtnh;
2864 int err = -EEXIST;
2865
2866 list_for_each_entry(nh, rt6_nh_list, next) {
2867 /* check if rt6_info already exists */
2868 rtnh = nh->rt6_info;
2869
2870 if (rtnh->dst.dev == rt->dst.dev &&
2871 rtnh->rt6i_idev == rt->rt6i_idev &&
2872 ipv6_addr_equal(&rtnh->rt6i_gateway,
2873 &rt->rt6i_gateway))
2874 return err;
2875 }
2876
2877 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2878 if (!nh)
2879 return -ENOMEM;
2880 nh->rt6_info = rt;
2881 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2882 if (err) {
2883 kfree(nh);
2884 return err;
2885 }
2886 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2887 list_add_tail(&nh->next, rt6_nh_list);
2888
2889 return 0;
2890}
2891
2892static int ip6_route_multipath_add(struct fib6_config *cfg)
2893{
2894 struct fib6_config r_cfg;
2895 struct rtnexthop *rtnh;
2896 struct rt6_info *rt;
2897 struct rt6_nh *err_nh;
2898 struct rt6_nh *nh, *nh_safe;
2899 int remaining;
2900 int attrlen;
2901 int err = 1;
2902 int nhn = 0;
2903 int replace = (cfg->fc_nlinfo.nlh &&
2904 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2905 LIST_HEAD(rt6_nh_list);
2906
2907 remaining = cfg->fc_mp_len;
2908 rtnh = (struct rtnexthop *)cfg->fc_mp;
2909
2910 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2911 * rt6_info structs per nexthop
2912 */
2913 while (rtnh_ok(rtnh, remaining)) {
2914 memcpy(&r_cfg, cfg, sizeof(*cfg));
2915 if (rtnh->rtnh_ifindex)
2916 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2917
2918 attrlen = rtnh_attrlen(rtnh);
2919 if (attrlen > 0) {
2920 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2921
2922 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2923 if (nla) {
2924 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2925 r_cfg.fc_flags |= RTF_GATEWAY;
2926 }
2927 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2928 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2929 if (nla)
2930 r_cfg.fc_encap_type = nla_get_u16(nla);
2931 }
2932
2933 rt = ip6_route_info_create(&r_cfg);
2934 if (IS_ERR(rt)) {
2935 err = PTR_ERR(rt);
2936 rt = NULL;
2937 goto cleanup;
2938 }
2939
2940 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2941 if (err) {
2942 dst_free(&rt->dst);
2943 goto cleanup;
2944 }
2945
2946 rtnh = rtnh_next(rtnh, &remaining);
2947 }
2948
2949 err_nh = NULL;
2950 list_for_each_entry(nh, &rt6_nh_list, next) {
2951 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2952 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2953 nh->rt6_info = NULL;
2954 if (err) {
2955 if (replace && nhn)
2956 ip6_print_replace_route_err(&rt6_nh_list);
2957 err_nh = nh;
2958 goto add_errout;
2959 }
2960
2961 /* Because each route is added like a single route we remove
2962 * these flags after the first nexthop: if there is a collision,
2963 * we have already failed to add the first nexthop:
2964 * fib6_add_rt2node() has rejected it; when replacing, old
2965 * nexthops have been replaced by first new, the rest should
2966 * be added to it.
2967 */
2968 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2969 NLM_F_REPLACE);
2970 nhn++;
2971 }
2972
2973 goto cleanup;
2974
2975add_errout:
2976 /* Delete routes that were already added */
2977 list_for_each_entry(nh, &rt6_nh_list, next) {
2978 if (err_nh == nh)
2979 break;
2980 ip6_route_del(&nh->r_cfg);
2981 }
2982
2983cleanup:
2984 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2985 if (nh->rt6_info)
2986 dst_free(&nh->rt6_info->dst);
2987 kfree(nh->mxc.mx);
2988 list_del(&nh->next);
2989 kfree(nh);
2990 }
2991
2992 return err;
2993}
2994
2995static int ip6_route_multipath_del(struct fib6_config *cfg)
2996{
2997 struct fib6_config r_cfg;
2998 struct rtnexthop *rtnh;
2999 int remaining;
3000 int attrlen;
3001 int err = 1, last_err = 0;
3002
3003 remaining = cfg->fc_mp_len;
3004 rtnh = (struct rtnexthop *)cfg->fc_mp;
3005
3006 /* Parse a Multipath Entry */
3007 while (rtnh_ok(rtnh, remaining)) {
3008 memcpy(&r_cfg, cfg, sizeof(*cfg));
3009 if (rtnh->rtnh_ifindex)
3010 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3011
3012 attrlen = rtnh_attrlen(rtnh);
3013 if (attrlen > 0) {
3014 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3015
3016 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3017 if (nla) {
3018 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3019 r_cfg.fc_flags |= RTF_GATEWAY;
3020 }
3021 }
3022 err = ip6_route_del(&r_cfg);
3023 if (err)
3024 last_err = err;
3025
3026 rtnh = rtnh_next(rtnh, &remaining);
3027 }
3028
3029 return last_err;
3030}
3031
3032static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3033{
3034 struct fib6_config cfg;
3035 int err;
3036
3037 err = rtm_to_fib6_config(skb, nlh, &cfg);
3038 if (err < 0)
3039 return err;
3040
3041 if (cfg.fc_mp)
3042 return ip6_route_multipath_del(&cfg);
3043 else
3044 return ip6_route_del(&cfg);
3045}
3046
3047static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3048{
3049 struct fib6_config cfg;
3050 int err;
3051
3052 err = rtm_to_fib6_config(skb, nlh, &cfg);
3053 if (err < 0)
3054 return err;
3055
3056 if (cfg.fc_mp)
3057 return ip6_route_multipath_add(&cfg);
3058 else
3059 return ip6_route_add(&cfg);
3060}
3061
3062static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3063{
3064 return NLMSG_ALIGN(sizeof(struct rtmsg))
3065 + nla_total_size(16) /* RTA_SRC */
3066 + nla_total_size(16) /* RTA_DST */
3067 + nla_total_size(16) /* RTA_GATEWAY */
3068 + nla_total_size(16) /* RTA_PREFSRC */
3069 + nla_total_size(4) /* RTA_TABLE */
3070 + nla_total_size(4) /* RTA_IIF */
3071 + nla_total_size(4) /* RTA_OIF */
3072 + nla_total_size(4) /* RTA_PRIORITY */
3073 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3074 + nla_total_size(sizeof(struct rta_cacheinfo))
3075 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3076 + nla_total_size(1) /* RTA_PREF */
3077 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3078}
3079
3080static int rt6_fill_node(struct net *net,
3081 struct sk_buff *skb, struct rt6_info *rt,
3082 struct in6_addr *dst, struct in6_addr *src,
3083 int iif, int type, u32 portid, u32 seq,
3084 int prefix, int nowait, unsigned int flags)
3085{
3086 u32 metrics[RTAX_MAX];
3087 struct rtmsg *rtm;
3088 struct nlmsghdr *nlh;
3089 long expires;
3090 u32 table;
3091
3092 if (prefix) { /* user wants prefix routes only */
3093 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3094 /* success since this is not a prefix route */
3095 return 1;
3096 }
3097 }
3098
3099 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3100 if (!nlh)
3101 return -EMSGSIZE;
3102
3103 rtm = nlmsg_data(nlh);
3104 rtm->rtm_family = AF_INET6;
3105 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3106 rtm->rtm_src_len = rt->rt6i_src.plen;
3107 rtm->rtm_tos = 0;
3108 if (rt->rt6i_table)
3109 table = rt->rt6i_table->tb6_id;
3110 else
3111 table = RT6_TABLE_UNSPEC;
3112 rtm->rtm_table = table;
3113 if (nla_put_u32(skb, RTA_TABLE, table))
3114 goto nla_put_failure;
3115 if (rt->rt6i_flags & RTF_REJECT) {
3116 switch (rt->dst.error) {
3117 case -EINVAL:
3118 rtm->rtm_type = RTN_BLACKHOLE;
3119 break;
3120 case -EACCES:
3121 rtm->rtm_type = RTN_PROHIBIT;
3122 break;
3123 case -EAGAIN:
3124 rtm->rtm_type = RTN_THROW;
3125 break;
3126 default:
3127 rtm->rtm_type = RTN_UNREACHABLE;
3128 break;
3129 }
3130 }
3131 else if (rt->rt6i_flags & RTF_LOCAL)
3132 rtm->rtm_type = RTN_LOCAL;
3133 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3134 rtm->rtm_type = RTN_LOCAL;
3135 else
3136 rtm->rtm_type = RTN_UNICAST;
3137 rtm->rtm_flags = 0;
3138 if (!netif_carrier_ok(rt->dst.dev)) {
3139 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3140 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3141 rtm->rtm_flags |= RTNH_F_DEAD;
3142 }
3143 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3144 rtm->rtm_protocol = rt->rt6i_protocol;
3145 if (rt->rt6i_flags & RTF_DYNAMIC)
3146 rtm->rtm_protocol = RTPROT_REDIRECT;
3147 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3148 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3149 rtm->rtm_protocol = RTPROT_RA;
3150 else
3151 rtm->rtm_protocol = RTPROT_KERNEL;
3152 }
3153
3154 if (rt->rt6i_flags & RTF_CACHE)
3155 rtm->rtm_flags |= RTM_F_CLONED;
3156
3157 if (dst) {
3158 if (nla_put_in6_addr(skb, RTA_DST, dst))
3159 goto nla_put_failure;
3160 rtm->rtm_dst_len = 128;
3161 } else if (rtm->rtm_dst_len)
3162 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3163 goto nla_put_failure;
3164#ifdef CONFIG_IPV6_SUBTREES
3165 if (src) {
3166 if (nla_put_in6_addr(skb, RTA_SRC, src))
3167 goto nla_put_failure;
3168 rtm->rtm_src_len = 128;
3169 } else if (rtm->rtm_src_len &&
3170 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3171 goto nla_put_failure;
3172#endif
3173 if (iif) {
3174#ifdef CONFIG_IPV6_MROUTE
3175 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3176 int err = ip6mr_get_route(net, skb, rtm, nowait);
3177 if (err <= 0) {
3178 if (!nowait) {
3179 if (err == 0)
3180 return 0;
3181 goto nla_put_failure;
3182 } else {
3183 if (err == -EMSGSIZE)
3184 goto nla_put_failure;
3185 }
3186 }
3187 } else
3188#endif
3189 if (nla_put_u32(skb, RTA_IIF, iif))
3190 goto nla_put_failure;
3191 } else if (dst) {
3192 struct in6_addr saddr_buf;
3193 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3194 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3195 goto nla_put_failure;
3196 }
3197
3198 if (rt->rt6i_prefsrc.plen) {
3199 struct in6_addr saddr_buf;
3200 saddr_buf = rt->rt6i_prefsrc.addr;
3201 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3202 goto nla_put_failure;
3203 }
3204
3205 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3206 if (rt->rt6i_pmtu)
3207 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3208 if (rtnetlink_put_metrics(skb, metrics) < 0)
3209 goto nla_put_failure;
3210
3211 if (rt->rt6i_flags & RTF_GATEWAY) {
3212 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3213 goto nla_put_failure;
3214 }
3215
3216 if (rt->dst.dev &&
3217 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3218 goto nla_put_failure;
3219 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3220 goto nla_put_failure;
3221
3222 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3223
3224 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3225 goto nla_put_failure;
3226
3227 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3228 goto nla_put_failure;
3229
3230 lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3231
3232 nlmsg_end(skb, nlh);
3233 return 0;
3234
3235nla_put_failure:
3236 nlmsg_cancel(skb, nlh);
3237 return -EMSGSIZE;
3238}
3239
3240int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3241{
3242 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3243 int prefix;
3244
3245 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3246 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3247 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3248 } else
3249 prefix = 0;
3250
3251 return rt6_fill_node(arg->net,
3252 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3253 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3254 prefix, 0, NLM_F_MULTI);
3255}
3256
3257static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3258{
3259 struct net *net = sock_net(in_skb->sk);
3260 struct nlattr *tb[RTA_MAX+1];
3261 struct rt6_info *rt;
3262 struct sk_buff *skb;
3263 struct rtmsg *rtm;
3264 struct flowi6 fl6;
3265 int err, iif = 0, oif = 0;
3266
3267 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3268 if (err < 0)
3269 goto errout;
3270
3271 err = -EINVAL;
3272 memset(&fl6, 0, sizeof(fl6));
3273
3274 if (tb[RTA_SRC]) {
3275 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3276 goto errout;
3277
3278 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3279 }
3280
3281 if (tb[RTA_DST]) {
3282 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3283 goto errout;
3284
3285 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3286 }
3287
3288 if (tb[RTA_IIF])
3289 iif = nla_get_u32(tb[RTA_IIF]);
3290
3291 if (tb[RTA_OIF])
3292 oif = nla_get_u32(tb[RTA_OIF]);
3293
3294 if (tb[RTA_MARK])
3295 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3296
3297 if (iif) {
3298 struct net_device *dev;
3299 int flags = 0;
3300
3301 dev = __dev_get_by_index(net, iif);
3302 if (!dev) {
3303 err = -ENODEV;
3304 goto errout;
3305 }
3306
3307 fl6.flowi6_iif = iif;
3308
3309 if (!ipv6_addr_any(&fl6.saddr))
3310 flags |= RT6_LOOKUP_F_HAS_SADDR;
3311
3312 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3313 flags);
3314 } else {
3315 fl6.flowi6_oif = oif;
3316
3317 if (netif_index_is_l3_master(net, oif)) {
3318 fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3319 FLOWI_FLAG_SKIP_NH_OIF;
3320 }
3321
3322 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3323 }
3324
3325 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3326 if (!skb) {
3327 ip6_rt_put(rt);
3328 err = -ENOBUFS;
3329 goto errout;
3330 }
3331
3332 /* Reserve room for dummy headers, this skb can pass
3333 through good chunk of routing engine.
3334 */
3335 skb_reset_mac_header(skb);
3336 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3337
3338 skb_dst_set(skb, &rt->dst);
3339
3340 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3341 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3342 nlh->nlmsg_seq, 0, 0, 0);
3343 if (err < 0) {
3344 kfree_skb(skb);
3345 goto errout;
3346 }
3347
3348 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3349errout:
3350 return err;
3351}
3352
3353void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3354 unsigned int nlm_flags)
3355{
3356 struct sk_buff *skb;
3357 struct net *net = info->nl_net;
3358 u32 seq;
3359 int err;
3360
3361 err = -ENOBUFS;
3362 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3363
3364 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3365 if (!skb)
3366 goto errout;
3367
3368 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3369 event, info->portid, seq, 0, 0, nlm_flags);
3370 if (err < 0) {
3371 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3372 WARN_ON(err == -EMSGSIZE);
3373 kfree_skb(skb);
3374 goto errout;
3375 }
3376 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3377 info->nlh, gfp_any());
3378 return;
3379errout:
3380 if (err < 0)
3381 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3382}
3383
3384static int ip6_route_dev_notify(struct notifier_block *this,
3385 unsigned long event, void *ptr)
3386{
3387 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3388 struct net *net = dev_net(dev);
3389
3390 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3391 net->ipv6.ip6_null_entry->dst.dev = dev;
3392 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3393#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3394 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3395 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3396 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3397 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3398#endif
3399 }
3400
3401 return NOTIFY_OK;
3402}
3403
3404/*
3405 * /proc
3406 */
3407
3408#ifdef CONFIG_PROC_FS
3409
3410static const struct file_operations ipv6_route_proc_fops = {
3411 .owner = THIS_MODULE,
3412 .open = ipv6_route_open,
3413 .read = seq_read,
3414 .llseek = seq_lseek,
3415 .release = seq_release_net,
3416};
3417
3418static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3419{
3420 struct net *net = (struct net *)seq->private;
3421 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3422 net->ipv6.rt6_stats->fib_nodes,
3423 net->ipv6.rt6_stats->fib_route_nodes,
3424 net->ipv6.rt6_stats->fib_rt_alloc,
3425 net->ipv6.rt6_stats->fib_rt_entries,
3426 net->ipv6.rt6_stats->fib_rt_cache,
3427 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3428 net->ipv6.rt6_stats->fib_discarded_routes);
3429
3430 return 0;
3431}
3432
3433static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3434{
3435 return single_open_net(inode, file, rt6_stats_seq_show);
3436}
3437
3438static const struct file_operations rt6_stats_seq_fops = {
3439 .owner = THIS_MODULE,
3440 .open = rt6_stats_seq_open,
3441 .read = seq_read,
3442 .llseek = seq_lseek,
3443 .release = single_release_net,
3444};
3445#endif /* CONFIG_PROC_FS */
3446
3447#ifdef CONFIG_SYSCTL
3448
3449static
3450int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3451 void __user *buffer, size_t *lenp, loff_t *ppos)
3452{
3453 struct net *net;
3454 int delay;
3455 if (!write)
3456 return -EINVAL;
3457
3458 net = (struct net *)ctl->extra1;
3459 delay = net->ipv6.sysctl.flush_delay;
3460 proc_dointvec(ctl, write, buffer, lenp, ppos);
3461 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3462 return 0;
3463}
3464
3465struct ctl_table ipv6_route_table_template[] = {
3466 {
3467 .procname = "flush",
3468 .data = &init_net.ipv6.sysctl.flush_delay,
3469 .maxlen = sizeof(int),
3470 .mode = 0200,
3471 .proc_handler = ipv6_sysctl_rtcache_flush
3472 },
3473 {
3474 .procname = "gc_thresh",
3475 .data = &ip6_dst_ops_template.gc_thresh,
3476 .maxlen = sizeof(int),
3477 .mode = 0644,
3478 .proc_handler = proc_dointvec,
3479 },
3480 {
3481 .procname = "max_size",
3482 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3483 .maxlen = sizeof(int),
3484 .mode = 0644,
3485 .proc_handler = proc_dointvec,
3486 },
3487 {
3488 .procname = "gc_min_interval",
3489 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3490 .maxlen = sizeof(int),
3491 .mode = 0644,
3492 .proc_handler = proc_dointvec_jiffies,
3493 },
3494 {
3495 .procname = "gc_timeout",
3496 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3497 .maxlen = sizeof(int),
3498 .mode = 0644,
3499 .proc_handler = proc_dointvec_jiffies,
3500 },
3501 {
3502 .procname = "gc_interval",
3503 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3504 .maxlen = sizeof(int),
3505 .mode = 0644,
3506 .proc_handler = proc_dointvec_jiffies,
3507 },
3508 {
3509 .procname = "gc_elasticity",
3510 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3511 .maxlen = sizeof(int),
3512 .mode = 0644,
3513 .proc_handler = proc_dointvec,
3514 },
3515 {
3516 .procname = "mtu_expires",
3517 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3518 .maxlen = sizeof(int),
3519 .mode = 0644,
3520 .proc_handler = proc_dointvec_jiffies,
3521 },
3522 {
3523 .procname = "min_adv_mss",
3524 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3525 .maxlen = sizeof(int),
3526 .mode = 0644,
3527 .proc_handler = proc_dointvec,
3528 },
3529 {
3530 .procname = "gc_min_interval_ms",
3531 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3532 .maxlen = sizeof(int),
3533 .mode = 0644,
3534 .proc_handler = proc_dointvec_ms_jiffies,
3535 },
3536 { }
3537};
3538
3539struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3540{
3541 struct ctl_table *table;
3542
3543 table = kmemdup(ipv6_route_table_template,
3544 sizeof(ipv6_route_table_template),
3545 GFP_KERNEL);
3546
3547 if (table) {
3548 table[0].data = &net->ipv6.sysctl.flush_delay;
3549 table[0].extra1 = net;
3550 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3551 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3552 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3553 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3554 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3555 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3556 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3557 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3558 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3559
3560 /* Don't export sysctls to unprivileged users */
3561 if (net->user_ns != &init_user_ns)
3562 table[0].procname = NULL;
3563 }
3564
3565 return table;
3566}
3567#endif
3568
3569static int __net_init ip6_route_net_init(struct net *net)
3570{
3571 int ret = -ENOMEM;
3572
3573 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3574 sizeof(net->ipv6.ip6_dst_ops));
3575
3576 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3577 goto out_ip6_dst_ops;
3578
3579 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3580 sizeof(*net->ipv6.ip6_null_entry),
3581 GFP_KERNEL);
3582 if (!net->ipv6.ip6_null_entry)
3583 goto out_ip6_dst_entries;
3584 net->ipv6.ip6_null_entry->dst.path =
3585 (struct dst_entry *)net->ipv6.ip6_null_entry;
3586 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3587 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3588 ip6_template_metrics, true);
3589
3590#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3591 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3592 sizeof(*net->ipv6.ip6_prohibit_entry),
3593 GFP_KERNEL);
3594 if (!net->ipv6.ip6_prohibit_entry)
3595 goto out_ip6_null_entry;
3596 net->ipv6.ip6_prohibit_entry->dst.path =
3597 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3598 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3599 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3600 ip6_template_metrics, true);
3601
3602 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3603 sizeof(*net->ipv6.ip6_blk_hole_entry),
3604 GFP_KERNEL);
3605 if (!net->ipv6.ip6_blk_hole_entry)
3606 goto out_ip6_prohibit_entry;
3607 net->ipv6.ip6_blk_hole_entry->dst.path =
3608 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3609 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3610 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3611 ip6_template_metrics, true);
3612#endif
3613
3614 net->ipv6.sysctl.flush_delay = 0;
3615 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3616 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3617 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3618 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3619 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3620 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3621 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3622
3623 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3624
3625 ret = 0;
3626out:
3627 return ret;
3628
3629#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3630out_ip6_prohibit_entry:
3631 kfree(net->ipv6.ip6_prohibit_entry);
3632out_ip6_null_entry:
3633 kfree(net->ipv6.ip6_null_entry);
3634#endif
3635out_ip6_dst_entries:
3636 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3637out_ip6_dst_ops:
3638 goto out;
3639}
3640
3641static void __net_exit ip6_route_net_exit(struct net *net)
3642{
3643 kfree(net->ipv6.ip6_null_entry);
3644#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3645 kfree(net->ipv6.ip6_prohibit_entry);
3646 kfree(net->ipv6.ip6_blk_hole_entry);
3647#endif
3648 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3649}
3650
3651static int __net_init ip6_route_net_init_late(struct net *net)
3652{
3653#ifdef CONFIG_PROC_FS
3654 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3655 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3656#endif
3657 return 0;
3658}
3659
3660static void __net_exit ip6_route_net_exit_late(struct net *net)
3661{
3662#ifdef CONFIG_PROC_FS
3663 remove_proc_entry("ipv6_route", net->proc_net);
3664 remove_proc_entry("rt6_stats", net->proc_net);
3665#endif
3666}
3667
3668static struct pernet_operations ip6_route_net_ops = {
3669 .init = ip6_route_net_init,
3670 .exit = ip6_route_net_exit,
3671};
3672
3673static int __net_init ipv6_inetpeer_init(struct net *net)
3674{
3675 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3676
3677 if (!bp)
3678 return -ENOMEM;
3679 inet_peer_base_init(bp);
3680 net->ipv6.peers = bp;
3681 return 0;
3682}
3683
3684static void __net_exit ipv6_inetpeer_exit(struct net *net)
3685{
3686 struct inet_peer_base *bp = net->ipv6.peers;
3687
3688 net->ipv6.peers = NULL;
3689 inetpeer_invalidate_tree(bp);
3690 kfree(bp);
3691}
3692
3693static struct pernet_operations ipv6_inetpeer_ops = {
3694 .init = ipv6_inetpeer_init,
3695 .exit = ipv6_inetpeer_exit,
3696};
3697
3698static struct pernet_operations ip6_route_net_late_ops = {
3699 .init = ip6_route_net_init_late,
3700 .exit = ip6_route_net_exit_late,
3701};
3702
3703static struct notifier_block ip6_route_dev_notifier = {
3704 .notifier_call = ip6_route_dev_notify,
3705 .priority = 0,
3706};
3707
3708int __init ip6_route_init(void)
3709{
3710 int ret;
3711 int cpu;
3712
3713 ret = -ENOMEM;
3714 ip6_dst_ops_template.kmem_cachep =
3715 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3716 SLAB_HWCACHE_ALIGN, NULL);
3717 if (!ip6_dst_ops_template.kmem_cachep)
3718 goto out;
3719
3720 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3721 if (ret)
3722 goto out_kmem_cache;
3723
3724 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3725 if (ret)
3726 goto out_dst_entries;
3727
3728 ret = register_pernet_subsys(&ip6_route_net_ops);
3729 if (ret)
3730 goto out_register_inetpeer;
3731
3732 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3733
3734 /* Registering of the loopback is done before this portion of code,
3735 * the loopback reference in rt6_info will not be taken, do it
3736 * manually for init_net */
3737 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3738 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3739 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3740 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3741 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3742 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3743 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3744 #endif
3745 ret = fib6_init();
3746 if (ret)
3747 goto out_register_subsys;
3748
3749 ret = xfrm6_init();
3750 if (ret)
3751 goto out_fib6_init;
3752
3753 ret = fib6_rules_init();
3754 if (ret)
3755 goto xfrm6_init;
3756
3757 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3758 if (ret)
3759 goto fib6_rules_init;
3760
3761 ret = -ENOBUFS;
3762 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3763 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3764 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3765 goto out_register_late_subsys;
3766
3767 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3768 if (ret)
3769 goto out_register_late_subsys;
3770
3771 for_each_possible_cpu(cpu) {
3772 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3773
3774 INIT_LIST_HEAD(&ul->head);
3775 spin_lock_init(&ul->lock);
3776 }
3777
3778out:
3779 return ret;
3780
3781out_register_late_subsys:
3782 unregister_pernet_subsys(&ip6_route_net_late_ops);
3783fib6_rules_init:
3784 fib6_rules_cleanup();
3785xfrm6_init:
3786 xfrm6_fini();
3787out_fib6_init:
3788 fib6_gc_cleanup();
3789out_register_subsys:
3790 unregister_pernet_subsys(&ip6_route_net_ops);
3791out_register_inetpeer:
3792 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3793out_dst_entries:
3794 dst_entries_destroy(&ip6_dst_blackhole_ops);
3795out_kmem_cache:
3796 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3797 goto out;
3798}
3799
3800void ip6_route_cleanup(void)
3801{
3802 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3803 unregister_pernet_subsys(&ip6_route_net_late_ops);
3804 fib6_rules_cleanup();
3805 xfrm6_fini();
3806 fib6_gc_cleanup();
3807 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3808 unregister_pernet_subsys(&ip6_route_net_ops);
3809 dst_entries_destroy(&ip6_dst_blackhole_ops);
3810 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3811}
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
61#include <asm/uaccess.h>
62
63#ifdef CONFIG_SYSCTL
64#include <linux/sysctl.h>
65#endif
66
67static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
68 const struct in6_addr *dest);
69static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
70static unsigned int ip6_default_advmss(const struct dst_entry *dst);
71static unsigned int ip6_mtu(const struct dst_entry *dst);
72static struct dst_entry *ip6_negative_advice(struct dst_entry *);
73static void ip6_dst_destroy(struct dst_entry *);
74static void ip6_dst_ifdown(struct dst_entry *,
75 struct net_device *dev, int how);
76static int ip6_dst_gc(struct dst_ops *ops);
77
78static int ip6_pkt_discard(struct sk_buff *skb);
79static int ip6_pkt_discard_out(struct sk_buff *skb);
80static void ip6_link_failure(struct sk_buff *skb);
81static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
82
83#ifdef CONFIG_IPV6_ROUTE_INFO
84static struct rt6_info *rt6_add_route_info(struct net *net,
85 const struct in6_addr *prefix, int prefixlen,
86 const struct in6_addr *gwaddr, int ifindex,
87 unsigned int pref);
88static struct rt6_info *rt6_get_route_info(struct net *net,
89 const struct in6_addr *prefix, int prefixlen,
90 const struct in6_addr *gwaddr, int ifindex);
91#endif
92
93static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
94{
95 struct rt6_info *rt = (struct rt6_info *) dst;
96 struct inet_peer *peer;
97 u32 *p = NULL;
98
99 if (!(rt->dst.flags & DST_HOST))
100 return NULL;
101
102 if (!rt->rt6i_peer)
103 rt6_bind_peer(rt, 1);
104
105 peer = rt->rt6i_peer;
106 if (peer) {
107 u32 *old_p = __DST_METRICS_PTR(old);
108 unsigned long prev, new;
109
110 p = peer->metrics;
111 if (inet_metrics_new(peer))
112 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
113
114 new = (unsigned long) p;
115 prev = cmpxchg(&dst->_metrics, old, new);
116
117 if (prev != old) {
118 p = __DST_METRICS_PTR(prev);
119 if (prev & DST_METRICS_READ_ONLY)
120 p = NULL;
121 }
122 }
123 return p;
124}
125
126static inline const void *choose_neigh_daddr(struct rt6_info *rt, const void *daddr)
127{
128 struct in6_addr *p = &rt->rt6i_gateway;
129
130 if (!ipv6_addr_any(p))
131 return (const void *) p;
132 return daddr;
133}
134
135static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst, const void *daddr)
136{
137 struct rt6_info *rt = (struct rt6_info *) dst;
138 struct neighbour *n;
139
140 daddr = choose_neigh_daddr(rt, daddr);
141 n = __ipv6_neigh_lookup(&nd_tbl, dst->dev, daddr);
142 if (n)
143 return n;
144 return neigh_create(&nd_tbl, daddr, dst->dev);
145}
146
147static int rt6_bind_neighbour(struct rt6_info *rt, struct net_device *dev)
148{
149 struct neighbour *n = __ipv6_neigh_lookup(&nd_tbl, dev, &rt->rt6i_gateway);
150 if (!n) {
151 n = neigh_create(&nd_tbl, &rt->rt6i_gateway, dev);
152 if (IS_ERR(n))
153 return PTR_ERR(n);
154 }
155 dst_set_neighbour(&rt->dst, n);
156
157 return 0;
158}
159
160static struct dst_ops ip6_dst_ops_template = {
161 .family = AF_INET6,
162 .protocol = cpu_to_be16(ETH_P_IPV6),
163 .gc = ip6_dst_gc,
164 .gc_thresh = 1024,
165 .check = ip6_dst_check,
166 .default_advmss = ip6_default_advmss,
167 .mtu = ip6_mtu,
168 .cow_metrics = ipv6_cow_metrics,
169 .destroy = ip6_dst_destroy,
170 .ifdown = ip6_dst_ifdown,
171 .negative_advice = ip6_negative_advice,
172 .link_failure = ip6_link_failure,
173 .update_pmtu = ip6_rt_update_pmtu,
174 .local_out = __ip6_local_out,
175 .neigh_lookup = ip6_neigh_lookup,
176};
177
178static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
179{
180 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
181
182 return mtu ? : dst->dev->mtu;
183}
184
185static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
186{
187}
188
189static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
190 unsigned long old)
191{
192 return NULL;
193}
194
195static struct dst_ops ip6_dst_blackhole_ops = {
196 .family = AF_INET6,
197 .protocol = cpu_to_be16(ETH_P_IPV6),
198 .destroy = ip6_dst_destroy,
199 .check = ip6_dst_check,
200 .mtu = ip6_blackhole_mtu,
201 .default_advmss = ip6_default_advmss,
202 .update_pmtu = ip6_rt_blackhole_update_pmtu,
203 .cow_metrics = ip6_rt_blackhole_cow_metrics,
204 .neigh_lookup = ip6_neigh_lookup,
205};
206
207static const u32 ip6_template_metrics[RTAX_MAX] = {
208 [RTAX_HOPLIMIT - 1] = 255,
209};
210
211static struct rt6_info ip6_null_entry_template = {
212 .dst = {
213 .__refcnt = ATOMIC_INIT(1),
214 .__use = 1,
215 .obsolete = -1,
216 .error = -ENETUNREACH,
217 .input = ip6_pkt_discard,
218 .output = ip6_pkt_discard_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#ifdef CONFIG_IPV6_MULTIPLE_TABLES
227
228static int ip6_pkt_prohibit(struct sk_buff *skb);
229static int ip6_pkt_prohibit_out(struct sk_buff *skb);
230
231static struct rt6_info ip6_prohibit_entry_template = {
232 .dst = {
233 .__refcnt = ATOMIC_INIT(1),
234 .__use = 1,
235 .obsolete = -1,
236 .error = -EACCES,
237 .input = ip6_pkt_prohibit,
238 .output = ip6_pkt_prohibit_out,
239 },
240 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
241 .rt6i_protocol = RTPROT_KERNEL,
242 .rt6i_metric = ~(u32) 0,
243 .rt6i_ref = ATOMIC_INIT(1),
244};
245
246static struct rt6_info ip6_blk_hole_entry_template = {
247 .dst = {
248 .__refcnt = ATOMIC_INIT(1),
249 .__use = 1,
250 .obsolete = -1,
251 .error = -EINVAL,
252 .input = dst_discard,
253 .output = dst_discard,
254 },
255 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
256 .rt6i_protocol = RTPROT_KERNEL,
257 .rt6i_metric = ~(u32) 0,
258 .rt6i_ref = ATOMIC_INIT(1),
259};
260
261#endif
262
263/* allocate dst with ip6_dst_ops */
264static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops,
265 struct net_device *dev,
266 int flags)
267{
268 struct rt6_info *rt = dst_alloc(ops, dev, 0, 0, flags);
269
270 if (rt)
271 memset(&rt->rt6i_table, 0,
272 sizeof(*rt) - sizeof(struct dst_entry));
273
274 return rt;
275}
276
277static void ip6_dst_destroy(struct dst_entry *dst)
278{
279 struct rt6_info *rt = (struct rt6_info *)dst;
280 struct inet6_dev *idev = rt->rt6i_idev;
281 struct inet_peer *peer = rt->rt6i_peer;
282
283 if (!(rt->dst.flags & DST_HOST))
284 dst_destroy_metrics_generic(dst);
285
286 if (idev) {
287 rt->rt6i_idev = NULL;
288 in6_dev_put(idev);
289 }
290
291 if (!(rt->rt6i_flags & RTF_EXPIRES) && dst->from)
292 dst_release(dst->from);
293
294 if (peer) {
295 rt->rt6i_peer = NULL;
296 inet_putpeer(peer);
297 }
298}
299
300static atomic_t __rt6_peer_genid = ATOMIC_INIT(0);
301
302static u32 rt6_peer_genid(void)
303{
304 return atomic_read(&__rt6_peer_genid);
305}
306
307void rt6_bind_peer(struct rt6_info *rt, int create)
308{
309 struct inet_peer *peer;
310
311 peer = inet_getpeer_v6(&rt->rt6i_dst.addr, create);
312 if (peer && cmpxchg(&rt->rt6i_peer, NULL, peer) != NULL)
313 inet_putpeer(peer);
314 else
315 rt->rt6i_peer_genid = rt6_peer_genid();
316}
317
318static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
319 int how)
320{
321 struct rt6_info *rt = (struct rt6_info *)dst;
322 struct inet6_dev *idev = rt->rt6i_idev;
323 struct net_device *loopback_dev =
324 dev_net(dev)->loopback_dev;
325
326 if (dev != loopback_dev && idev && idev->dev == dev) {
327 struct inet6_dev *loopback_idev =
328 in6_dev_get(loopback_dev);
329 if (loopback_idev) {
330 rt->rt6i_idev = loopback_idev;
331 in6_dev_put(idev);
332 }
333 }
334}
335
336static bool rt6_check_expired(const struct rt6_info *rt)
337{
338 struct rt6_info *ort = NULL;
339
340 if (rt->rt6i_flags & RTF_EXPIRES) {
341 if (time_after(jiffies, rt->dst.expires))
342 return true;
343 } else if (rt->dst.from) {
344 ort = (struct rt6_info *) rt->dst.from;
345 return (ort->rt6i_flags & RTF_EXPIRES) &&
346 time_after(jiffies, ort->dst.expires);
347 }
348 return false;
349}
350
351static bool rt6_need_strict(const struct in6_addr *daddr)
352{
353 return ipv6_addr_type(daddr) &
354 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
355}
356
357/*
358 * Route lookup. Any table->tb6_lock is implied.
359 */
360
361static inline struct rt6_info *rt6_device_match(struct net *net,
362 struct rt6_info *rt,
363 const struct in6_addr *saddr,
364 int oif,
365 int flags)
366{
367 struct rt6_info *local = NULL;
368 struct rt6_info *sprt;
369
370 if (!oif && ipv6_addr_any(saddr))
371 goto out;
372
373 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
374 struct net_device *dev = sprt->dst.dev;
375
376 if (oif) {
377 if (dev->ifindex == oif)
378 return sprt;
379 if (dev->flags & IFF_LOOPBACK) {
380 if (!sprt->rt6i_idev ||
381 sprt->rt6i_idev->dev->ifindex != oif) {
382 if (flags & RT6_LOOKUP_F_IFACE && oif)
383 continue;
384 if (local && (!oif ||
385 local->rt6i_idev->dev->ifindex == oif))
386 continue;
387 }
388 local = sprt;
389 }
390 } else {
391 if (ipv6_chk_addr(net, saddr, dev,
392 flags & RT6_LOOKUP_F_IFACE))
393 return sprt;
394 }
395 }
396
397 if (oif) {
398 if (local)
399 return local;
400
401 if (flags & RT6_LOOKUP_F_IFACE)
402 return net->ipv6.ip6_null_entry;
403 }
404out:
405 return rt;
406}
407
408#ifdef CONFIG_IPV6_ROUTER_PREF
409static void rt6_probe(struct rt6_info *rt)
410{
411 struct neighbour *neigh;
412 /*
413 * Okay, this does not seem to be appropriate
414 * for now, however, we need to check if it
415 * is really so; aka Router Reachability Probing.
416 *
417 * Router Reachability Probe MUST be rate-limited
418 * to no more than one per minute.
419 */
420 rcu_read_lock();
421 neigh = rt ? dst_get_neighbour_noref(&rt->dst) : NULL;
422 if (!neigh || (neigh->nud_state & NUD_VALID))
423 goto out;
424 read_lock_bh(&neigh->lock);
425 if (!(neigh->nud_state & NUD_VALID) &&
426 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
427 struct in6_addr mcaddr;
428 struct in6_addr *target;
429
430 neigh->updated = jiffies;
431 read_unlock_bh(&neigh->lock);
432
433 target = (struct in6_addr *)&neigh->primary_key;
434 addrconf_addr_solict_mult(target, &mcaddr);
435 ndisc_send_ns(rt->dst.dev, NULL, target, &mcaddr, NULL);
436 } else {
437 read_unlock_bh(&neigh->lock);
438 }
439out:
440 rcu_read_unlock();
441}
442#else
443static inline void rt6_probe(struct rt6_info *rt)
444{
445}
446#endif
447
448/*
449 * Default Router Selection (RFC 2461 6.3.6)
450 */
451static inline int rt6_check_dev(struct rt6_info *rt, int oif)
452{
453 struct net_device *dev = rt->dst.dev;
454 if (!oif || dev->ifindex == oif)
455 return 2;
456 if ((dev->flags & IFF_LOOPBACK) &&
457 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
458 return 1;
459 return 0;
460}
461
462static inline int rt6_check_neigh(struct rt6_info *rt)
463{
464 struct neighbour *neigh;
465 int m;
466
467 rcu_read_lock();
468 neigh = dst_get_neighbour_noref(&rt->dst);
469 if (rt->rt6i_flags & RTF_NONEXTHOP ||
470 !(rt->rt6i_flags & RTF_GATEWAY))
471 m = 1;
472 else if (neigh) {
473 read_lock_bh(&neigh->lock);
474 if (neigh->nud_state & NUD_VALID)
475 m = 2;
476#ifdef CONFIG_IPV6_ROUTER_PREF
477 else if (neigh->nud_state & NUD_FAILED)
478 m = 0;
479#endif
480 else
481 m = 1;
482 read_unlock_bh(&neigh->lock);
483 } else
484 m = 0;
485 rcu_read_unlock();
486 return m;
487}
488
489static int rt6_score_route(struct rt6_info *rt, int oif,
490 int strict)
491{
492 int m, n;
493
494 m = rt6_check_dev(rt, oif);
495 if (!m && (strict & RT6_LOOKUP_F_IFACE))
496 return -1;
497#ifdef CONFIG_IPV6_ROUTER_PREF
498 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
499#endif
500 n = rt6_check_neigh(rt);
501 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
502 return -1;
503 return m;
504}
505
506static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
507 int *mpri, struct rt6_info *match)
508{
509 int m;
510
511 if (rt6_check_expired(rt))
512 goto out;
513
514 m = rt6_score_route(rt, oif, strict);
515 if (m < 0)
516 goto out;
517
518 if (m > *mpri) {
519 if (strict & RT6_LOOKUP_F_REACHABLE)
520 rt6_probe(match);
521 *mpri = m;
522 match = rt;
523 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
524 rt6_probe(rt);
525 }
526
527out:
528 return match;
529}
530
531static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
532 struct rt6_info *rr_head,
533 u32 metric, int oif, int strict)
534{
535 struct rt6_info *rt, *match;
536 int mpri = -1;
537
538 match = NULL;
539 for (rt = rr_head; rt && rt->rt6i_metric == metric;
540 rt = rt->dst.rt6_next)
541 match = find_match(rt, oif, strict, &mpri, match);
542 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
543 rt = rt->dst.rt6_next)
544 match = find_match(rt, oif, strict, &mpri, match);
545
546 return match;
547}
548
549static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
550{
551 struct rt6_info *match, *rt0;
552 struct net *net;
553
554 rt0 = fn->rr_ptr;
555 if (!rt0)
556 fn->rr_ptr = rt0 = fn->leaf;
557
558 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
559
560 if (!match &&
561 (strict & RT6_LOOKUP_F_REACHABLE)) {
562 struct rt6_info *next = rt0->dst.rt6_next;
563
564 /* no entries matched; do round-robin */
565 if (!next || next->rt6i_metric != rt0->rt6i_metric)
566 next = fn->leaf;
567
568 if (next != rt0)
569 fn->rr_ptr = next;
570 }
571
572 net = dev_net(rt0->dst.dev);
573 return match ? match : net->ipv6.ip6_null_entry;
574}
575
576#ifdef CONFIG_IPV6_ROUTE_INFO
577int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
578 const struct in6_addr *gwaddr)
579{
580 struct net *net = dev_net(dev);
581 struct route_info *rinfo = (struct route_info *) opt;
582 struct in6_addr prefix_buf, *prefix;
583 unsigned int pref;
584 unsigned long lifetime;
585 struct rt6_info *rt;
586
587 if (len < sizeof(struct route_info)) {
588 return -EINVAL;
589 }
590
591 /* Sanity check for prefix_len and length */
592 if (rinfo->length > 3) {
593 return -EINVAL;
594 } else if (rinfo->prefix_len > 128) {
595 return -EINVAL;
596 } else if (rinfo->prefix_len > 64) {
597 if (rinfo->length < 2) {
598 return -EINVAL;
599 }
600 } else if (rinfo->prefix_len > 0) {
601 if (rinfo->length < 1) {
602 return -EINVAL;
603 }
604 }
605
606 pref = rinfo->route_pref;
607 if (pref == ICMPV6_ROUTER_PREF_INVALID)
608 return -EINVAL;
609
610 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
611
612 if (rinfo->length == 3)
613 prefix = (struct in6_addr *)rinfo->prefix;
614 else {
615 /* this function is safe */
616 ipv6_addr_prefix(&prefix_buf,
617 (struct in6_addr *)rinfo->prefix,
618 rinfo->prefix_len);
619 prefix = &prefix_buf;
620 }
621
622 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
623 dev->ifindex);
624
625 if (rt && !lifetime) {
626 ip6_del_rt(rt);
627 rt = NULL;
628 }
629
630 if (!rt && lifetime)
631 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
632 pref);
633 else if (rt)
634 rt->rt6i_flags = RTF_ROUTEINFO |
635 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
636
637 if (rt) {
638 if (!addrconf_finite_timeout(lifetime))
639 rt6_clean_expires(rt);
640 else
641 rt6_set_expires(rt, jiffies + HZ * lifetime);
642
643 dst_release(&rt->dst);
644 }
645 return 0;
646}
647#endif
648
649#define BACKTRACK(__net, saddr) \
650do { \
651 if (rt == __net->ipv6.ip6_null_entry) { \
652 struct fib6_node *pn; \
653 while (1) { \
654 if (fn->fn_flags & RTN_TL_ROOT) \
655 goto out; \
656 pn = fn->parent; \
657 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
658 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
659 else \
660 fn = pn; \
661 if (fn->fn_flags & RTN_RTINFO) \
662 goto restart; \
663 } \
664 } \
665} while (0)
666
667static struct rt6_info *ip6_pol_route_lookup(struct net *net,
668 struct fib6_table *table,
669 struct flowi6 *fl6, int flags)
670{
671 struct fib6_node *fn;
672 struct rt6_info *rt;
673
674 read_lock_bh(&table->tb6_lock);
675 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
676restart:
677 rt = fn->leaf;
678 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
679 BACKTRACK(net, &fl6->saddr);
680out:
681 dst_use(&rt->dst, jiffies);
682 read_unlock_bh(&table->tb6_lock);
683 return rt;
684
685}
686
687struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
688 int flags)
689{
690 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
691}
692EXPORT_SYMBOL_GPL(ip6_route_lookup);
693
694struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
695 const struct in6_addr *saddr, int oif, int strict)
696{
697 struct flowi6 fl6 = {
698 .flowi6_oif = oif,
699 .daddr = *daddr,
700 };
701 struct dst_entry *dst;
702 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
703
704 if (saddr) {
705 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
706 flags |= RT6_LOOKUP_F_HAS_SADDR;
707 }
708
709 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
710 if (dst->error == 0)
711 return (struct rt6_info *) dst;
712
713 dst_release(dst);
714
715 return NULL;
716}
717
718EXPORT_SYMBOL(rt6_lookup);
719
720/* ip6_ins_rt is called with FREE table->tb6_lock.
721 It takes new route entry, the addition fails by any reason the
722 route is freed. In any case, if caller does not hold it, it may
723 be destroyed.
724 */
725
726static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
727{
728 int err;
729 struct fib6_table *table;
730
731 table = rt->rt6i_table;
732 write_lock_bh(&table->tb6_lock);
733 err = fib6_add(&table->tb6_root, rt, info);
734 write_unlock_bh(&table->tb6_lock);
735
736 return err;
737}
738
739int ip6_ins_rt(struct rt6_info *rt)
740{
741 struct nl_info info = {
742 .nl_net = dev_net(rt->dst.dev),
743 };
744 return __ip6_ins_rt(rt, &info);
745}
746
747static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
748 const struct in6_addr *daddr,
749 const struct in6_addr *saddr)
750{
751 struct rt6_info *rt;
752
753 /*
754 * Clone the route.
755 */
756
757 rt = ip6_rt_copy(ort, daddr);
758
759 if (rt) {
760 int attempts = !in_softirq();
761
762 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
763 if (ort->rt6i_dst.plen != 128 &&
764 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
765 rt->rt6i_flags |= RTF_ANYCAST;
766 rt->rt6i_gateway = *daddr;
767 }
768
769 rt->rt6i_flags |= RTF_CACHE;
770
771#ifdef CONFIG_IPV6_SUBTREES
772 if (rt->rt6i_src.plen && saddr) {
773 rt->rt6i_src.addr = *saddr;
774 rt->rt6i_src.plen = 128;
775 }
776#endif
777
778 retry:
779 if (rt6_bind_neighbour(rt, rt->dst.dev)) {
780 struct net *net = dev_net(rt->dst.dev);
781 int saved_rt_min_interval =
782 net->ipv6.sysctl.ip6_rt_gc_min_interval;
783 int saved_rt_elasticity =
784 net->ipv6.sysctl.ip6_rt_gc_elasticity;
785
786 if (attempts-- > 0) {
787 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
788 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
789
790 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
791
792 net->ipv6.sysctl.ip6_rt_gc_elasticity =
793 saved_rt_elasticity;
794 net->ipv6.sysctl.ip6_rt_gc_min_interval =
795 saved_rt_min_interval;
796 goto retry;
797 }
798
799 net_warn_ratelimited("Neighbour table overflow\n");
800 dst_free(&rt->dst);
801 return NULL;
802 }
803 }
804
805 return rt;
806}
807
808static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
809 const struct in6_addr *daddr)
810{
811 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
812
813 if (rt) {
814 rt->rt6i_flags |= RTF_CACHE;
815 dst_set_neighbour(&rt->dst, neigh_clone(dst_get_neighbour_noref_raw(&ort->dst)));
816 }
817 return rt;
818}
819
820static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
821 struct flowi6 *fl6, int flags)
822{
823 struct fib6_node *fn;
824 struct rt6_info *rt, *nrt;
825 int strict = 0;
826 int attempts = 3;
827 int err;
828 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
829
830 strict |= flags & RT6_LOOKUP_F_IFACE;
831
832relookup:
833 read_lock_bh(&table->tb6_lock);
834
835restart_2:
836 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
837
838restart:
839 rt = rt6_select(fn, oif, strict | reachable);
840
841 BACKTRACK(net, &fl6->saddr);
842 if (rt == net->ipv6.ip6_null_entry ||
843 rt->rt6i_flags & RTF_CACHE)
844 goto out;
845
846 dst_hold(&rt->dst);
847 read_unlock_bh(&table->tb6_lock);
848
849 if (!dst_get_neighbour_noref_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
850 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
851 else if (!(rt->dst.flags & DST_HOST))
852 nrt = rt6_alloc_clone(rt, &fl6->daddr);
853 else
854 goto out2;
855
856 dst_release(&rt->dst);
857 rt = nrt ? : net->ipv6.ip6_null_entry;
858
859 dst_hold(&rt->dst);
860 if (nrt) {
861 err = ip6_ins_rt(nrt);
862 if (!err)
863 goto out2;
864 }
865
866 if (--attempts <= 0)
867 goto out2;
868
869 /*
870 * Race condition! In the gap, when table->tb6_lock was
871 * released someone could insert this route. Relookup.
872 */
873 dst_release(&rt->dst);
874 goto relookup;
875
876out:
877 if (reachable) {
878 reachable = 0;
879 goto restart_2;
880 }
881 dst_hold(&rt->dst);
882 read_unlock_bh(&table->tb6_lock);
883out2:
884 rt->dst.lastuse = jiffies;
885 rt->dst.__use++;
886
887 return rt;
888}
889
890static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
891 struct flowi6 *fl6, int flags)
892{
893 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
894}
895
896static struct dst_entry *ip6_route_input_lookup(struct net *net,
897 struct net_device *dev,
898 struct flowi6 *fl6, int flags)
899{
900 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
901 flags |= RT6_LOOKUP_F_IFACE;
902
903 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
904}
905
906void ip6_route_input(struct sk_buff *skb)
907{
908 const struct ipv6hdr *iph = ipv6_hdr(skb);
909 struct net *net = dev_net(skb->dev);
910 int flags = RT6_LOOKUP_F_HAS_SADDR;
911 struct flowi6 fl6 = {
912 .flowi6_iif = skb->dev->ifindex,
913 .daddr = iph->daddr,
914 .saddr = iph->saddr,
915 .flowlabel = (* (__be32 *) iph) & IPV6_FLOWINFO_MASK,
916 .flowi6_mark = skb->mark,
917 .flowi6_proto = iph->nexthdr,
918 };
919
920 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
921}
922
923static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
924 struct flowi6 *fl6, int flags)
925{
926 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
927}
928
929struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
930 struct flowi6 *fl6)
931{
932 int flags = 0;
933
934 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
935 flags |= RT6_LOOKUP_F_IFACE;
936
937 if (!ipv6_addr_any(&fl6->saddr))
938 flags |= RT6_LOOKUP_F_HAS_SADDR;
939 else if (sk)
940 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
941
942 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
943}
944
945EXPORT_SYMBOL(ip6_route_output);
946
947struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
948{
949 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
950 struct dst_entry *new = NULL;
951
952 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, 0, 0);
953 if (rt) {
954 memset(&rt->rt6i_table, 0, sizeof(*rt) - sizeof(struct dst_entry));
955
956 new = &rt->dst;
957
958 new->__use = 1;
959 new->input = dst_discard;
960 new->output = dst_discard;
961
962 if (dst_metrics_read_only(&ort->dst))
963 new->_metrics = ort->dst._metrics;
964 else
965 dst_copy_metrics(new, &ort->dst);
966 rt->rt6i_idev = ort->rt6i_idev;
967 if (rt->rt6i_idev)
968 in6_dev_hold(rt->rt6i_idev);
969
970 rt->rt6i_gateway = ort->rt6i_gateway;
971 rt->rt6i_flags = ort->rt6i_flags;
972 rt6_clean_expires(rt);
973 rt->rt6i_metric = 0;
974
975 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
976#ifdef CONFIG_IPV6_SUBTREES
977 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
978#endif
979
980 dst_free(new);
981 }
982
983 dst_release(dst_orig);
984 return new ? new : ERR_PTR(-ENOMEM);
985}
986
987/*
988 * Destination cache support functions
989 */
990
991static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
992{
993 struct rt6_info *rt;
994
995 rt = (struct rt6_info *) dst;
996
997 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie)) {
998 if (rt->rt6i_peer_genid != rt6_peer_genid()) {
999 if (!rt->rt6i_peer)
1000 rt6_bind_peer(rt, 0);
1001 rt->rt6i_peer_genid = rt6_peer_genid();
1002 }
1003 return dst;
1004 }
1005 return NULL;
1006}
1007
1008static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1009{
1010 struct rt6_info *rt = (struct rt6_info *) dst;
1011
1012 if (rt) {
1013 if (rt->rt6i_flags & RTF_CACHE) {
1014 if (rt6_check_expired(rt)) {
1015 ip6_del_rt(rt);
1016 dst = NULL;
1017 }
1018 } else {
1019 dst_release(dst);
1020 dst = NULL;
1021 }
1022 }
1023 return dst;
1024}
1025
1026static void ip6_link_failure(struct sk_buff *skb)
1027{
1028 struct rt6_info *rt;
1029
1030 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1031
1032 rt = (struct rt6_info *) skb_dst(skb);
1033 if (rt) {
1034 if (rt->rt6i_flags & RTF_CACHE)
1035 rt6_update_expires(rt, 0);
1036 else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1037 rt->rt6i_node->fn_sernum = -1;
1038 }
1039}
1040
1041static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
1042{
1043 struct rt6_info *rt6 = (struct rt6_info*)dst;
1044
1045 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1046 rt6->rt6i_flags |= RTF_MODIFIED;
1047 if (mtu < IPV6_MIN_MTU) {
1048 u32 features = dst_metric(dst, RTAX_FEATURES);
1049 mtu = IPV6_MIN_MTU;
1050 features |= RTAX_FEATURE_ALLFRAG;
1051 dst_metric_set(dst, RTAX_FEATURES, features);
1052 }
1053 dst_metric_set(dst, RTAX_MTU, mtu);
1054 }
1055}
1056
1057static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1058{
1059 struct net_device *dev = dst->dev;
1060 unsigned int mtu = dst_mtu(dst);
1061 struct net *net = dev_net(dev);
1062
1063 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1064
1065 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1066 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1067
1068 /*
1069 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1070 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1071 * IPV6_MAXPLEN is also valid and means: "any MSS,
1072 * rely only on pmtu discovery"
1073 */
1074 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1075 mtu = IPV6_MAXPLEN;
1076 return mtu;
1077}
1078
1079static unsigned int ip6_mtu(const struct dst_entry *dst)
1080{
1081 struct inet6_dev *idev;
1082 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1083
1084 if (mtu)
1085 return mtu;
1086
1087 mtu = IPV6_MIN_MTU;
1088
1089 rcu_read_lock();
1090 idev = __in6_dev_get(dst->dev);
1091 if (idev)
1092 mtu = idev->cnf.mtu6;
1093 rcu_read_unlock();
1094
1095 return mtu;
1096}
1097
1098static struct dst_entry *icmp6_dst_gc_list;
1099static DEFINE_SPINLOCK(icmp6_dst_lock);
1100
1101struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1102 struct neighbour *neigh,
1103 struct flowi6 *fl6)
1104{
1105 struct dst_entry *dst;
1106 struct rt6_info *rt;
1107 struct inet6_dev *idev = in6_dev_get(dev);
1108 struct net *net = dev_net(dev);
1109
1110 if (unlikely(!idev))
1111 return ERR_PTR(-ENODEV);
1112
1113 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops, dev, 0);
1114 if (unlikely(!rt)) {
1115 in6_dev_put(idev);
1116 dst = ERR_PTR(-ENOMEM);
1117 goto out;
1118 }
1119
1120 if (neigh)
1121 neigh_hold(neigh);
1122 else {
1123 neigh = ip6_neigh_lookup(&rt->dst, &fl6->daddr);
1124 if (IS_ERR(neigh)) {
1125 in6_dev_put(idev);
1126 dst_free(&rt->dst);
1127 return ERR_CAST(neigh);
1128 }
1129 }
1130
1131 rt->dst.flags |= DST_HOST;
1132 rt->dst.output = ip6_output;
1133 dst_set_neighbour(&rt->dst, neigh);
1134 atomic_set(&rt->dst.__refcnt, 1);
1135 rt->rt6i_dst.addr = fl6->daddr;
1136 rt->rt6i_dst.plen = 128;
1137 rt->rt6i_idev = idev;
1138 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 255);
1139
1140 spin_lock_bh(&icmp6_dst_lock);
1141 rt->dst.next = icmp6_dst_gc_list;
1142 icmp6_dst_gc_list = &rt->dst;
1143 spin_unlock_bh(&icmp6_dst_lock);
1144
1145 fib6_force_start_gc(net);
1146
1147 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1148
1149out:
1150 return dst;
1151}
1152
1153int icmp6_dst_gc(void)
1154{
1155 struct dst_entry *dst, **pprev;
1156 int more = 0;
1157
1158 spin_lock_bh(&icmp6_dst_lock);
1159 pprev = &icmp6_dst_gc_list;
1160
1161 while ((dst = *pprev) != NULL) {
1162 if (!atomic_read(&dst->__refcnt)) {
1163 *pprev = dst->next;
1164 dst_free(dst);
1165 } else {
1166 pprev = &dst->next;
1167 ++more;
1168 }
1169 }
1170
1171 spin_unlock_bh(&icmp6_dst_lock);
1172
1173 return more;
1174}
1175
1176static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1177 void *arg)
1178{
1179 struct dst_entry *dst, **pprev;
1180
1181 spin_lock_bh(&icmp6_dst_lock);
1182 pprev = &icmp6_dst_gc_list;
1183 while ((dst = *pprev) != NULL) {
1184 struct rt6_info *rt = (struct rt6_info *) dst;
1185 if (func(rt, arg)) {
1186 *pprev = dst->next;
1187 dst_free(dst);
1188 } else {
1189 pprev = &dst->next;
1190 }
1191 }
1192 spin_unlock_bh(&icmp6_dst_lock);
1193}
1194
1195static int ip6_dst_gc(struct dst_ops *ops)
1196{
1197 unsigned long now = jiffies;
1198 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1199 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1200 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1201 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1202 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1203 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1204 int entries;
1205
1206 entries = dst_entries_get_fast(ops);
1207 if (time_after(rt_last_gc + rt_min_interval, now) &&
1208 entries <= rt_max_size)
1209 goto out;
1210
1211 net->ipv6.ip6_rt_gc_expire++;
1212 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1213 net->ipv6.ip6_rt_last_gc = now;
1214 entries = dst_entries_get_slow(ops);
1215 if (entries < ops->gc_thresh)
1216 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1217out:
1218 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1219 return entries > rt_max_size;
1220}
1221
1222/* Clean host part of a prefix. Not necessary in radix tree,
1223 but results in cleaner routing tables.
1224
1225 Remove it only when all the things will work!
1226 */
1227
1228int ip6_dst_hoplimit(struct dst_entry *dst)
1229{
1230 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1231 if (hoplimit == 0) {
1232 struct net_device *dev = dst->dev;
1233 struct inet6_dev *idev;
1234
1235 rcu_read_lock();
1236 idev = __in6_dev_get(dev);
1237 if (idev)
1238 hoplimit = idev->cnf.hop_limit;
1239 else
1240 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1241 rcu_read_unlock();
1242 }
1243 return hoplimit;
1244}
1245EXPORT_SYMBOL(ip6_dst_hoplimit);
1246
1247/*
1248 *
1249 */
1250
1251int ip6_route_add(struct fib6_config *cfg)
1252{
1253 int err;
1254 struct net *net = cfg->fc_nlinfo.nl_net;
1255 struct rt6_info *rt = NULL;
1256 struct net_device *dev = NULL;
1257 struct inet6_dev *idev = NULL;
1258 struct fib6_table *table;
1259 int addr_type;
1260
1261 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1262 return -EINVAL;
1263#ifndef CONFIG_IPV6_SUBTREES
1264 if (cfg->fc_src_len)
1265 return -EINVAL;
1266#endif
1267 if (cfg->fc_ifindex) {
1268 err = -ENODEV;
1269 dev = dev_get_by_index(net, cfg->fc_ifindex);
1270 if (!dev)
1271 goto out;
1272 idev = in6_dev_get(dev);
1273 if (!idev)
1274 goto out;
1275 }
1276
1277 if (cfg->fc_metric == 0)
1278 cfg->fc_metric = IP6_RT_PRIO_USER;
1279
1280 err = -ENOBUFS;
1281 if (cfg->fc_nlinfo.nlh &&
1282 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1283 table = fib6_get_table(net, cfg->fc_table);
1284 if (!table) {
1285 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1286 table = fib6_new_table(net, cfg->fc_table);
1287 }
1288 } else {
1289 table = fib6_new_table(net, cfg->fc_table);
1290 }
1291
1292 if (!table)
1293 goto out;
1294
1295 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops, NULL, DST_NOCOUNT);
1296
1297 if (!rt) {
1298 err = -ENOMEM;
1299 goto out;
1300 }
1301
1302 rt->dst.obsolete = -1;
1303
1304 if (cfg->fc_flags & RTF_EXPIRES)
1305 rt6_set_expires(rt, jiffies +
1306 clock_t_to_jiffies(cfg->fc_expires));
1307 else
1308 rt6_clean_expires(rt);
1309
1310 if (cfg->fc_protocol == RTPROT_UNSPEC)
1311 cfg->fc_protocol = RTPROT_BOOT;
1312 rt->rt6i_protocol = cfg->fc_protocol;
1313
1314 addr_type = ipv6_addr_type(&cfg->fc_dst);
1315
1316 if (addr_type & IPV6_ADDR_MULTICAST)
1317 rt->dst.input = ip6_mc_input;
1318 else if (cfg->fc_flags & RTF_LOCAL)
1319 rt->dst.input = ip6_input;
1320 else
1321 rt->dst.input = ip6_forward;
1322
1323 rt->dst.output = ip6_output;
1324
1325 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1326 rt->rt6i_dst.plen = cfg->fc_dst_len;
1327 if (rt->rt6i_dst.plen == 128)
1328 rt->dst.flags |= DST_HOST;
1329
1330 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1331 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1332 if (!metrics) {
1333 err = -ENOMEM;
1334 goto out;
1335 }
1336 dst_init_metrics(&rt->dst, metrics, 0);
1337 }
1338#ifdef CONFIG_IPV6_SUBTREES
1339 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1340 rt->rt6i_src.plen = cfg->fc_src_len;
1341#endif
1342
1343 rt->rt6i_metric = cfg->fc_metric;
1344
1345 /* We cannot add true routes via loopback here,
1346 they would result in kernel looping; promote them to reject routes
1347 */
1348 if ((cfg->fc_flags & RTF_REJECT) ||
1349 (dev && (dev->flags & IFF_LOOPBACK) &&
1350 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1351 !(cfg->fc_flags & RTF_LOCAL))) {
1352 /* hold loopback dev/idev if we haven't done so. */
1353 if (dev != net->loopback_dev) {
1354 if (dev) {
1355 dev_put(dev);
1356 in6_dev_put(idev);
1357 }
1358 dev = net->loopback_dev;
1359 dev_hold(dev);
1360 idev = in6_dev_get(dev);
1361 if (!idev) {
1362 err = -ENODEV;
1363 goto out;
1364 }
1365 }
1366 rt->dst.output = ip6_pkt_discard_out;
1367 rt->dst.input = ip6_pkt_discard;
1368 rt->dst.error = -ENETUNREACH;
1369 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1370 goto install_route;
1371 }
1372
1373 if (cfg->fc_flags & RTF_GATEWAY) {
1374 const struct in6_addr *gw_addr;
1375 int gwa_type;
1376
1377 gw_addr = &cfg->fc_gateway;
1378 rt->rt6i_gateway = *gw_addr;
1379 gwa_type = ipv6_addr_type(gw_addr);
1380
1381 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1382 struct rt6_info *grt;
1383
1384 /* IPv6 strictly inhibits using not link-local
1385 addresses as nexthop address.
1386 Otherwise, router will not able to send redirects.
1387 It is very good, but in some (rare!) circumstances
1388 (SIT, PtP, NBMA NOARP links) it is handy to allow
1389 some exceptions. --ANK
1390 */
1391 err = -EINVAL;
1392 if (!(gwa_type & IPV6_ADDR_UNICAST))
1393 goto out;
1394
1395 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1396
1397 err = -EHOSTUNREACH;
1398 if (!grt)
1399 goto out;
1400 if (dev) {
1401 if (dev != grt->dst.dev) {
1402 dst_release(&grt->dst);
1403 goto out;
1404 }
1405 } else {
1406 dev = grt->dst.dev;
1407 idev = grt->rt6i_idev;
1408 dev_hold(dev);
1409 in6_dev_hold(grt->rt6i_idev);
1410 }
1411 if (!(grt->rt6i_flags & RTF_GATEWAY))
1412 err = 0;
1413 dst_release(&grt->dst);
1414
1415 if (err)
1416 goto out;
1417 }
1418 err = -EINVAL;
1419 if (!dev || (dev->flags & IFF_LOOPBACK))
1420 goto out;
1421 }
1422
1423 err = -ENODEV;
1424 if (!dev)
1425 goto out;
1426
1427 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1428 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1429 err = -EINVAL;
1430 goto out;
1431 }
1432 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1433 rt->rt6i_prefsrc.plen = 128;
1434 } else
1435 rt->rt6i_prefsrc.plen = 0;
1436
1437 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1438 err = rt6_bind_neighbour(rt, dev);
1439 if (err)
1440 goto out;
1441 }
1442
1443 rt->rt6i_flags = cfg->fc_flags;
1444
1445install_route:
1446 if (cfg->fc_mx) {
1447 struct nlattr *nla;
1448 int remaining;
1449
1450 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1451 int type = nla_type(nla);
1452
1453 if (type) {
1454 if (type > RTAX_MAX) {
1455 err = -EINVAL;
1456 goto out;
1457 }
1458
1459 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1460 }
1461 }
1462 }
1463
1464 rt->dst.dev = dev;
1465 rt->rt6i_idev = idev;
1466 rt->rt6i_table = table;
1467
1468 cfg->fc_nlinfo.nl_net = dev_net(dev);
1469
1470 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1471
1472out:
1473 if (dev)
1474 dev_put(dev);
1475 if (idev)
1476 in6_dev_put(idev);
1477 if (rt)
1478 dst_free(&rt->dst);
1479 return err;
1480}
1481
1482static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1483{
1484 int err;
1485 struct fib6_table *table;
1486 struct net *net = dev_net(rt->dst.dev);
1487
1488 if (rt == net->ipv6.ip6_null_entry)
1489 return -ENOENT;
1490
1491 table = rt->rt6i_table;
1492 write_lock_bh(&table->tb6_lock);
1493
1494 err = fib6_del(rt, info);
1495 dst_release(&rt->dst);
1496
1497 write_unlock_bh(&table->tb6_lock);
1498
1499 return err;
1500}
1501
1502int ip6_del_rt(struct rt6_info *rt)
1503{
1504 struct nl_info info = {
1505 .nl_net = dev_net(rt->dst.dev),
1506 };
1507 return __ip6_del_rt(rt, &info);
1508}
1509
1510static int ip6_route_del(struct fib6_config *cfg)
1511{
1512 struct fib6_table *table;
1513 struct fib6_node *fn;
1514 struct rt6_info *rt;
1515 int err = -ESRCH;
1516
1517 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1518 if (!table)
1519 return err;
1520
1521 read_lock_bh(&table->tb6_lock);
1522
1523 fn = fib6_locate(&table->tb6_root,
1524 &cfg->fc_dst, cfg->fc_dst_len,
1525 &cfg->fc_src, cfg->fc_src_len);
1526
1527 if (fn) {
1528 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1529 if (cfg->fc_ifindex &&
1530 (!rt->dst.dev ||
1531 rt->dst.dev->ifindex != cfg->fc_ifindex))
1532 continue;
1533 if (cfg->fc_flags & RTF_GATEWAY &&
1534 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1535 continue;
1536 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1537 continue;
1538 dst_hold(&rt->dst);
1539 read_unlock_bh(&table->tb6_lock);
1540
1541 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1542 }
1543 }
1544 read_unlock_bh(&table->tb6_lock);
1545
1546 return err;
1547}
1548
1549/*
1550 * Handle redirects
1551 */
1552struct ip6rd_flowi {
1553 struct flowi6 fl6;
1554 struct in6_addr gateway;
1555};
1556
1557static struct rt6_info *__ip6_route_redirect(struct net *net,
1558 struct fib6_table *table,
1559 struct flowi6 *fl6,
1560 int flags)
1561{
1562 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1563 struct rt6_info *rt;
1564 struct fib6_node *fn;
1565
1566 /*
1567 * Get the "current" route for this destination and
1568 * check if the redirect has come from approriate router.
1569 *
1570 * RFC 2461 specifies that redirects should only be
1571 * accepted if they come from the nexthop to the target.
1572 * Due to the way the routes are chosen, this notion
1573 * is a bit fuzzy and one might need to check all possible
1574 * routes.
1575 */
1576
1577 read_lock_bh(&table->tb6_lock);
1578 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1579restart:
1580 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1581 /*
1582 * Current route is on-link; redirect is always invalid.
1583 *
1584 * Seems, previous statement is not true. It could
1585 * be node, which looks for us as on-link (f.e. proxy ndisc)
1586 * But then router serving it might decide, that we should
1587 * know truth 8)8) --ANK (980726).
1588 */
1589 if (rt6_check_expired(rt))
1590 continue;
1591 if (!(rt->rt6i_flags & RTF_GATEWAY))
1592 continue;
1593 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1594 continue;
1595 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1596 continue;
1597 break;
1598 }
1599
1600 if (!rt)
1601 rt = net->ipv6.ip6_null_entry;
1602 BACKTRACK(net, &fl6->saddr);
1603out:
1604 dst_hold(&rt->dst);
1605
1606 read_unlock_bh(&table->tb6_lock);
1607
1608 return rt;
1609};
1610
1611static struct rt6_info *ip6_route_redirect(const struct in6_addr *dest,
1612 const struct in6_addr *src,
1613 const struct in6_addr *gateway,
1614 struct net_device *dev)
1615{
1616 int flags = RT6_LOOKUP_F_HAS_SADDR;
1617 struct net *net = dev_net(dev);
1618 struct ip6rd_flowi rdfl = {
1619 .fl6 = {
1620 .flowi6_oif = dev->ifindex,
1621 .daddr = *dest,
1622 .saddr = *src,
1623 },
1624 };
1625
1626 rdfl.gateway = *gateway;
1627
1628 if (rt6_need_strict(dest))
1629 flags |= RT6_LOOKUP_F_IFACE;
1630
1631 return (struct rt6_info *)fib6_rule_lookup(net, &rdfl.fl6,
1632 flags, __ip6_route_redirect);
1633}
1634
1635void rt6_redirect(const struct in6_addr *dest, const struct in6_addr *src,
1636 const struct in6_addr *saddr,
1637 struct neighbour *neigh, u8 *lladdr, int on_link)
1638{
1639 struct rt6_info *rt, *nrt = NULL;
1640 struct netevent_redirect netevent;
1641 struct net *net = dev_net(neigh->dev);
1642
1643 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1644
1645 if (rt == net->ipv6.ip6_null_entry) {
1646 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1647 goto out;
1648 }
1649
1650 /*
1651 * We have finally decided to accept it.
1652 */
1653
1654 neigh_update(neigh, lladdr, NUD_STALE,
1655 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1656 NEIGH_UPDATE_F_OVERRIDE|
1657 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1658 NEIGH_UPDATE_F_ISROUTER))
1659 );
1660
1661 /*
1662 * Redirect received -> path was valid.
1663 * Look, redirects are sent only in response to data packets,
1664 * so that this nexthop apparently is reachable. --ANK
1665 */
1666 dst_confirm(&rt->dst);
1667
1668 /* Duplicate redirect: silently ignore. */
1669 if (neigh == dst_get_neighbour_noref_raw(&rt->dst))
1670 goto out;
1671
1672 nrt = ip6_rt_copy(rt, dest);
1673 if (!nrt)
1674 goto out;
1675
1676 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1677 if (on_link)
1678 nrt->rt6i_flags &= ~RTF_GATEWAY;
1679
1680 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1681 dst_set_neighbour(&nrt->dst, neigh_clone(neigh));
1682
1683 if (ip6_ins_rt(nrt))
1684 goto out;
1685
1686 netevent.old = &rt->dst;
1687 netevent.new = &nrt->dst;
1688 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1689
1690 if (rt->rt6i_flags & RTF_CACHE) {
1691 ip6_del_rt(rt);
1692 return;
1693 }
1694
1695out:
1696 dst_release(&rt->dst);
1697}
1698
1699/*
1700 * Handle ICMP "packet too big" messages
1701 * i.e. Path MTU discovery
1702 */
1703
1704static void rt6_do_pmtu_disc(const struct in6_addr *daddr, const struct in6_addr *saddr,
1705 struct net *net, u32 pmtu, int ifindex)
1706{
1707 struct rt6_info *rt, *nrt;
1708 int allfrag = 0;
1709again:
1710 rt = rt6_lookup(net, daddr, saddr, ifindex, 0);
1711 if (!rt)
1712 return;
1713
1714 if (rt6_check_expired(rt)) {
1715 ip6_del_rt(rt);
1716 goto again;
1717 }
1718
1719 if (pmtu >= dst_mtu(&rt->dst))
1720 goto out;
1721
1722 if (pmtu < IPV6_MIN_MTU) {
1723 /*
1724 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1725 * MTU (1280) and a fragment header should always be included
1726 * after a node receiving Too Big message reporting PMTU is
1727 * less than the IPv6 Minimum Link MTU.
1728 */
1729 pmtu = IPV6_MIN_MTU;
1730 allfrag = 1;
1731 }
1732
1733 /* New mtu received -> path was valid.
1734 They are sent only in response to data packets,
1735 so that this nexthop apparently is reachable. --ANK
1736 */
1737 dst_confirm(&rt->dst);
1738
1739 /* Host route. If it is static, it would be better
1740 not to override it, but add new one, so that
1741 when cache entry will expire old pmtu
1742 would return automatically.
1743 */
1744 if (rt->rt6i_flags & RTF_CACHE) {
1745 dst_metric_set(&rt->dst, RTAX_MTU, pmtu);
1746 if (allfrag) {
1747 u32 features = dst_metric(&rt->dst, RTAX_FEATURES);
1748 features |= RTAX_FEATURE_ALLFRAG;
1749 dst_metric_set(&rt->dst, RTAX_FEATURES, features);
1750 }
1751 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1752 rt->rt6i_flags |= RTF_MODIFIED;
1753 goto out;
1754 }
1755
1756 /* Network route.
1757 Two cases are possible:
1758 1. It is connected route. Action: COW
1759 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1760 */
1761 if (!dst_get_neighbour_noref_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
1762 nrt = rt6_alloc_cow(rt, daddr, saddr);
1763 else
1764 nrt = rt6_alloc_clone(rt, daddr);
1765
1766 if (nrt) {
1767 dst_metric_set(&nrt->dst, RTAX_MTU, pmtu);
1768 if (allfrag) {
1769 u32 features = dst_metric(&nrt->dst, RTAX_FEATURES);
1770 features |= RTAX_FEATURE_ALLFRAG;
1771 dst_metric_set(&nrt->dst, RTAX_FEATURES, features);
1772 }
1773
1774 /* According to RFC 1981, detecting PMTU increase shouldn't be
1775 * happened within 5 mins, the recommended timer is 10 mins.
1776 * Here this route expiration time is set to ip6_rt_mtu_expires
1777 * which is 10 mins. After 10 mins the decreased pmtu is expired
1778 * and detecting PMTU increase will be automatically happened.
1779 */
1780 rt6_update_expires(nrt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1781 nrt->rt6i_flags |= RTF_DYNAMIC;
1782 ip6_ins_rt(nrt);
1783 }
1784out:
1785 dst_release(&rt->dst);
1786}
1787
1788void rt6_pmtu_discovery(const struct in6_addr *daddr, const struct in6_addr *saddr,
1789 struct net_device *dev, u32 pmtu)
1790{
1791 struct net *net = dev_net(dev);
1792
1793 /*
1794 * RFC 1981 states that a node "MUST reduce the size of the packets it
1795 * is sending along the path" that caused the Packet Too Big message.
1796 * Since it's not possible in the general case to determine which
1797 * interface was used to send the original packet, we update the MTU
1798 * on the interface that will be used to send future packets. We also
1799 * update the MTU on the interface that received the Packet Too Big in
1800 * case the original packet was forced out that interface with
1801 * SO_BINDTODEVICE or similar. This is the next best thing to the
1802 * correct behaviour, which would be to update the MTU on all
1803 * interfaces.
1804 */
1805 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, 0);
1806 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, dev->ifindex);
1807}
1808
1809/*
1810 * Misc support functions
1811 */
1812
1813static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1814 const struct in6_addr *dest)
1815{
1816 struct net *net = dev_net(ort->dst.dev);
1817 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops,
1818 ort->dst.dev, 0);
1819
1820 if (rt) {
1821 rt->dst.input = ort->dst.input;
1822 rt->dst.output = ort->dst.output;
1823 rt->dst.flags |= DST_HOST;
1824
1825 rt->rt6i_dst.addr = *dest;
1826 rt->rt6i_dst.plen = 128;
1827 dst_copy_metrics(&rt->dst, &ort->dst);
1828 rt->dst.error = ort->dst.error;
1829 rt->rt6i_idev = ort->rt6i_idev;
1830 if (rt->rt6i_idev)
1831 in6_dev_hold(rt->rt6i_idev);
1832 rt->dst.lastuse = jiffies;
1833
1834 rt->rt6i_gateway = ort->rt6i_gateway;
1835 rt->rt6i_flags = ort->rt6i_flags;
1836 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1837 (RTF_DEFAULT | RTF_ADDRCONF))
1838 rt6_set_from(rt, ort);
1839 else
1840 rt6_clean_expires(rt);
1841 rt->rt6i_metric = 0;
1842
1843#ifdef CONFIG_IPV6_SUBTREES
1844 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1845#endif
1846 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1847 rt->rt6i_table = ort->rt6i_table;
1848 }
1849 return rt;
1850}
1851
1852#ifdef CONFIG_IPV6_ROUTE_INFO
1853static struct rt6_info *rt6_get_route_info(struct net *net,
1854 const struct in6_addr *prefix, int prefixlen,
1855 const struct in6_addr *gwaddr, int ifindex)
1856{
1857 struct fib6_node *fn;
1858 struct rt6_info *rt = NULL;
1859 struct fib6_table *table;
1860
1861 table = fib6_get_table(net, RT6_TABLE_INFO);
1862 if (!table)
1863 return NULL;
1864
1865 write_lock_bh(&table->tb6_lock);
1866 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1867 if (!fn)
1868 goto out;
1869
1870 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1871 if (rt->dst.dev->ifindex != ifindex)
1872 continue;
1873 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1874 continue;
1875 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1876 continue;
1877 dst_hold(&rt->dst);
1878 break;
1879 }
1880out:
1881 write_unlock_bh(&table->tb6_lock);
1882 return rt;
1883}
1884
1885static struct rt6_info *rt6_add_route_info(struct net *net,
1886 const struct in6_addr *prefix, int prefixlen,
1887 const struct in6_addr *gwaddr, int ifindex,
1888 unsigned int pref)
1889{
1890 struct fib6_config cfg = {
1891 .fc_table = RT6_TABLE_INFO,
1892 .fc_metric = IP6_RT_PRIO_USER,
1893 .fc_ifindex = ifindex,
1894 .fc_dst_len = prefixlen,
1895 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1896 RTF_UP | RTF_PREF(pref),
1897 .fc_nlinfo.pid = 0,
1898 .fc_nlinfo.nlh = NULL,
1899 .fc_nlinfo.nl_net = net,
1900 };
1901
1902 cfg.fc_dst = *prefix;
1903 cfg.fc_gateway = *gwaddr;
1904
1905 /* We should treat it as a default route if prefix length is 0. */
1906 if (!prefixlen)
1907 cfg.fc_flags |= RTF_DEFAULT;
1908
1909 ip6_route_add(&cfg);
1910
1911 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1912}
1913#endif
1914
1915struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1916{
1917 struct rt6_info *rt;
1918 struct fib6_table *table;
1919
1920 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1921 if (!table)
1922 return NULL;
1923
1924 write_lock_bh(&table->tb6_lock);
1925 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1926 if (dev == rt->dst.dev &&
1927 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1928 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1929 break;
1930 }
1931 if (rt)
1932 dst_hold(&rt->dst);
1933 write_unlock_bh(&table->tb6_lock);
1934 return rt;
1935}
1936
1937struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1938 struct net_device *dev,
1939 unsigned int pref)
1940{
1941 struct fib6_config cfg = {
1942 .fc_table = RT6_TABLE_DFLT,
1943 .fc_metric = IP6_RT_PRIO_USER,
1944 .fc_ifindex = dev->ifindex,
1945 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1946 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1947 .fc_nlinfo.pid = 0,
1948 .fc_nlinfo.nlh = NULL,
1949 .fc_nlinfo.nl_net = dev_net(dev),
1950 };
1951
1952 cfg.fc_gateway = *gwaddr;
1953
1954 ip6_route_add(&cfg);
1955
1956 return rt6_get_dflt_router(gwaddr, dev);
1957}
1958
1959void rt6_purge_dflt_routers(struct net *net)
1960{
1961 struct rt6_info *rt;
1962 struct fib6_table *table;
1963
1964 /* NOTE: Keep consistent with rt6_get_dflt_router */
1965 table = fib6_get_table(net, RT6_TABLE_DFLT);
1966 if (!table)
1967 return;
1968
1969restart:
1970 read_lock_bh(&table->tb6_lock);
1971 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1972 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1973 dst_hold(&rt->dst);
1974 read_unlock_bh(&table->tb6_lock);
1975 ip6_del_rt(rt);
1976 goto restart;
1977 }
1978 }
1979 read_unlock_bh(&table->tb6_lock);
1980}
1981
1982static void rtmsg_to_fib6_config(struct net *net,
1983 struct in6_rtmsg *rtmsg,
1984 struct fib6_config *cfg)
1985{
1986 memset(cfg, 0, sizeof(*cfg));
1987
1988 cfg->fc_table = RT6_TABLE_MAIN;
1989 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1990 cfg->fc_metric = rtmsg->rtmsg_metric;
1991 cfg->fc_expires = rtmsg->rtmsg_info;
1992 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1993 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1994 cfg->fc_flags = rtmsg->rtmsg_flags;
1995
1996 cfg->fc_nlinfo.nl_net = net;
1997
1998 cfg->fc_dst = rtmsg->rtmsg_dst;
1999 cfg->fc_src = rtmsg->rtmsg_src;
2000 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2001}
2002
2003int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2004{
2005 struct fib6_config cfg;
2006 struct in6_rtmsg rtmsg;
2007 int err;
2008
2009 switch(cmd) {
2010 case SIOCADDRT: /* Add a route */
2011 case SIOCDELRT: /* Delete a route */
2012 if (!capable(CAP_NET_ADMIN))
2013 return -EPERM;
2014 err = copy_from_user(&rtmsg, arg,
2015 sizeof(struct in6_rtmsg));
2016 if (err)
2017 return -EFAULT;
2018
2019 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2020
2021 rtnl_lock();
2022 switch (cmd) {
2023 case SIOCADDRT:
2024 err = ip6_route_add(&cfg);
2025 break;
2026 case SIOCDELRT:
2027 err = ip6_route_del(&cfg);
2028 break;
2029 default:
2030 err = -EINVAL;
2031 }
2032 rtnl_unlock();
2033
2034 return err;
2035 }
2036
2037 return -EINVAL;
2038}
2039
2040/*
2041 * Drop the packet on the floor
2042 */
2043
2044static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2045{
2046 int type;
2047 struct dst_entry *dst = skb_dst(skb);
2048 switch (ipstats_mib_noroutes) {
2049 case IPSTATS_MIB_INNOROUTES:
2050 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2051 if (type == IPV6_ADDR_ANY) {
2052 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2053 IPSTATS_MIB_INADDRERRORS);
2054 break;
2055 }
2056 /* FALLTHROUGH */
2057 case IPSTATS_MIB_OUTNOROUTES:
2058 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2059 ipstats_mib_noroutes);
2060 break;
2061 }
2062 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2063 kfree_skb(skb);
2064 return 0;
2065}
2066
2067static int ip6_pkt_discard(struct sk_buff *skb)
2068{
2069 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2070}
2071
2072static int ip6_pkt_discard_out(struct sk_buff *skb)
2073{
2074 skb->dev = skb_dst(skb)->dev;
2075 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2076}
2077
2078#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2079
2080static int ip6_pkt_prohibit(struct sk_buff *skb)
2081{
2082 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2083}
2084
2085static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2086{
2087 skb->dev = skb_dst(skb)->dev;
2088 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2089}
2090
2091#endif
2092
2093/*
2094 * Allocate a dst for local (unicast / anycast) address.
2095 */
2096
2097struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2098 const struct in6_addr *addr,
2099 bool anycast)
2100{
2101 struct net *net = dev_net(idev->dev);
2102 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops,
2103 net->loopback_dev, 0);
2104 int err;
2105
2106 if (!rt) {
2107 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2108 return ERR_PTR(-ENOMEM);
2109 }
2110
2111 in6_dev_hold(idev);
2112
2113 rt->dst.flags |= DST_HOST;
2114 rt->dst.input = ip6_input;
2115 rt->dst.output = ip6_output;
2116 rt->rt6i_idev = idev;
2117 rt->dst.obsolete = -1;
2118
2119 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2120 if (anycast)
2121 rt->rt6i_flags |= RTF_ANYCAST;
2122 else
2123 rt->rt6i_flags |= RTF_LOCAL;
2124 err = rt6_bind_neighbour(rt, rt->dst.dev);
2125 if (err) {
2126 dst_free(&rt->dst);
2127 return ERR_PTR(err);
2128 }
2129
2130 rt->rt6i_dst.addr = *addr;
2131 rt->rt6i_dst.plen = 128;
2132 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2133
2134 atomic_set(&rt->dst.__refcnt, 1);
2135
2136 return rt;
2137}
2138
2139int ip6_route_get_saddr(struct net *net,
2140 struct rt6_info *rt,
2141 const struct in6_addr *daddr,
2142 unsigned int prefs,
2143 struct in6_addr *saddr)
2144{
2145 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2146 int err = 0;
2147 if (rt->rt6i_prefsrc.plen)
2148 *saddr = rt->rt6i_prefsrc.addr;
2149 else
2150 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2151 daddr, prefs, saddr);
2152 return err;
2153}
2154
2155/* remove deleted ip from prefsrc entries */
2156struct arg_dev_net_ip {
2157 struct net_device *dev;
2158 struct net *net;
2159 struct in6_addr *addr;
2160};
2161
2162static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2163{
2164 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2165 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2166 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2167
2168 if (((void *)rt->dst.dev == dev || !dev) &&
2169 rt != net->ipv6.ip6_null_entry &&
2170 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2171 /* remove prefsrc entry */
2172 rt->rt6i_prefsrc.plen = 0;
2173 }
2174 return 0;
2175}
2176
2177void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2178{
2179 struct net *net = dev_net(ifp->idev->dev);
2180 struct arg_dev_net_ip adni = {
2181 .dev = ifp->idev->dev,
2182 .net = net,
2183 .addr = &ifp->addr,
2184 };
2185 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2186}
2187
2188struct arg_dev_net {
2189 struct net_device *dev;
2190 struct net *net;
2191};
2192
2193static int fib6_ifdown(struct rt6_info *rt, void *arg)
2194{
2195 const struct arg_dev_net *adn = arg;
2196 const struct net_device *dev = adn->dev;
2197
2198 if ((rt->dst.dev == dev || !dev) &&
2199 rt != adn->net->ipv6.ip6_null_entry)
2200 return -1;
2201
2202 return 0;
2203}
2204
2205void rt6_ifdown(struct net *net, struct net_device *dev)
2206{
2207 struct arg_dev_net adn = {
2208 .dev = dev,
2209 .net = net,
2210 };
2211
2212 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2213 icmp6_clean_all(fib6_ifdown, &adn);
2214}
2215
2216struct rt6_mtu_change_arg {
2217 struct net_device *dev;
2218 unsigned int mtu;
2219};
2220
2221static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2222{
2223 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2224 struct inet6_dev *idev;
2225
2226 /* In IPv6 pmtu discovery is not optional,
2227 so that RTAX_MTU lock cannot disable it.
2228 We still use this lock to block changes
2229 caused by addrconf/ndisc.
2230 */
2231
2232 idev = __in6_dev_get(arg->dev);
2233 if (!idev)
2234 return 0;
2235
2236 /* For administrative MTU increase, there is no way to discover
2237 IPv6 PMTU increase, so PMTU increase should be updated here.
2238 Since RFC 1981 doesn't include administrative MTU increase
2239 update PMTU increase is a MUST. (i.e. jumbo frame)
2240 */
2241 /*
2242 If new MTU is less than route PMTU, this new MTU will be the
2243 lowest MTU in the path, update the route PMTU to reflect PMTU
2244 decreases; if new MTU is greater than route PMTU, and the
2245 old MTU is the lowest MTU in the path, update the route PMTU
2246 to reflect the increase. In this case if the other nodes' MTU
2247 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2248 PMTU discouvery.
2249 */
2250 if (rt->dst.dev == arg->dev &&
2251 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2252 (dst_mtu(&rt->dst) >= arg->mtu ||
2253 (dst_mtu(&rt->dst) < arg->mtu &&
2254 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2255 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2256 }
2257 return 0;
2258}
2259
2260void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2261{
2262 struct rt6_mtu_change_arg arg = {
2263 .dev = dev,
2264 .mtu = mtu,
2265 };
2266
2267 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2268}
2269
2270static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2271 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2272 [RTA_OIF] = { .type = NLA_U32 },
2273 [RTA_IIF] = { .type = NLA_U32 },
2274 [RTA_PRIORITY] = { .type = NLA_U32 },
2275 [RTA_METRICS] = { .type = NLA_NESTED },
2276};
2277
2278static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2279 struct fib6_config *cfg)
2280{
2281 struct rtmsg *rtm;
2282 struct nlattr *tb[RTA_MAX+1];
2283 int err;
2284
2285 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2286 if (err < 0)
2287 goto errout;
2288
2289 err = -EINVAL;
2290 rtm = nlmsg_data(nlh);
2291 memset(cfg, 0, sizeof(*cfg));
2292
2293 cfg->fc_table = rtm->rtm_table;
2294 cfg->fc_dst_len = rtm->rtm_dst_len;
2295 cfg->fc_src_len = rtm->rtm_src_len;
2296 cfg->fc_flags = RTF_UP;
2297 cfg->fc_protocol = rtm->rtm_protocol;
2298
2299 if (rtm->rtm_type == RTN_UNREACHABLE)
2300 cfg->fc_flags |= RTF_REJECT;
2301
2302 if (rtm->rtm_type == RTN_LOCAL)
2303 cfg->fc_flags |= RTF_LOCAL;
2304
2305 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2306 cfg->fc_nlinfo.nlh = nlh;
2307 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2308
2309 if (tb[RTA_GATEWAY]) {
2310 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2311 cfg->fc_flags |= RTF_GATEWAY;
2312 }
2313
2314 if (tb[RTA_DST]) {
2315 int plen = (rtm->rtm_dst_len + 7) >> 3;
2316
2317 if (nla_len(tb[RTA_DST]) < plen)
2318 goto errout;
2319
2320 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2321 }
2322
2323 if (tb[RTA_SRC]) {
2324 int plen = (rtm->rtm_src_len + 7) >> 3;
2325
2326 if (nla_len(tb[RTA_SRC]) < plen)
2327 goto errout;
2328
2329 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2330 }
2331
2332 if (tb[RTA_PREFSRC])
2333 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2334
2335 if (tb[RTA_OIF])
2336 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2337
2338 if (tb[RTA_PRIORITY])
2339 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2340
2341 if (tb[RTA_METRICS]) {
2342 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2343 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2344 }
2345
2346 if (tb[RTA_TABLE])
2347 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2348
2349 err = 0;
2350errout:
2351 return err;
2352}
2353
2354static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2355{
2356 struct fib6_config cfg;
2357 int err;
2358
2359 err = rtm_to_fib6_config(skb, nlh, &cfg);
2360 if (err < 0)
2361 return err;
2362
2363 return ip6_route_del(&cfg);
2364}
2365
2366static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2367{
2368 struct fib6_config cfg;
2369 int err;
2370
2371 err = rtm_to_fib6_config(skb, nlh, &cfg);
2372 if (err < 0)
2373 return err;
2374
2375 return ip6_route_add(&cfg);
2376}
2377
2378static inline size_t rt6_nlmsg_size(void)
2379{
2380 return NLMSG_ALIGN(sizeof(struct rtmsg))
2381 + nla_total_size(16) /* RTA_SRC */
2382 + nla_total_size(16) /* RTA_DST */
2383 + nla_total_size(16) /* RTA_GATEWAY */
2384 + nla_total_size(16) /* RTA_PREFSRC */
2385 + nla_total_size(4) /* RTA_TABLE */
2386 + nla_total_size(4) /* RTA_IIF */
2387 + nla_total_size(4) /* RTA_OIF */
2388 + nla_total_size(4) /* RTA_PRIORITY */
2389 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2390 + nla_total_size(sizeof(struct rta_cacheinfo));
2391}
2392
2393static int rt6_fill_node(struct net *net,
2394 struct sk_buff *skb, struct rt6_info *rt,
2395 struct in6_addr *dst, struct in6_addr *src,
2396 int iif, int type, u32 pid, u32 seq,
2397 int prefix, int nowait, unsigned int flags)
2398{
2399 const struct inet_peer *peer;
2400 struct rtmsg *rtm;
2401 struct nlmsghdr *nlh;
2402 long expires;
2403 u32 table;
2404 struct neighbour *n;
2405 u32 ts, tsage;
2406
2407 if (prefix) { /* user wants prefix routes only */
2408 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2409 /* success since this is not a prefix route */
2410 return 1;
2411 }
2412 }
2413
2414 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2415 if (!nlh)
2416 return -EMSGSIZE;
2417
2418 rtm = nlmsg_data(nlh);
2419 rtm->rtm_family = AF_INET6;
2420 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2421 rtm->rtm_src_len = rt->rt6i_src.plen;
2422 rtm->rtm_tos = 0;
2423 if (rt->rt6i_table)
2424 table = rt->rt6i_table->tb6_id;
2425 else
2426 table = RT6_TABLE_UNSPEC;
2427 rtm->rtm_table = table;
2428 if (nla_put_u32(skb, RTA_TABLE, table))
2429 goto nla_put_failure;
2430 if (rt->rt6i_flags & RTF_REJECT)
2431 rtm->rtm_type = RTN_UNREACHABLE;
2432 else if (rt->rt6i_flags & RTF_LOCAL)
2433 rtm->rtm_type = RTN_LOCAL;
2434 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2435 rtm->rtm_type = RTN_LOCAL;
2436 else
2437 rtm->rtm_type = RTN_UNICAST;
2438 rtm->rtm_flags = 0;
2439 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2440 rtm->rtm_protocol = rt->rt6i_protocol;
2441 if (rt->rt6i_flags & RTF_DYNAMIC)
2442 rtm->rtm_protocol = RTPROT_REDIRECT;
2443 else if (rt->rt6i_flags & RTF_ADDRCONF)
2444 rtm->rtm_protocol = RTPROT_KERNEL;
2445 else if (rt->rt6i_flags & RTF_DEFAULT)
2446 rtm->rtm_protocol = RTPROT_RA;
2447
2448 if (rt->rt6i_flags & RTF_CACHE)
2449 rtm->rtm_flags |= RTM_F_CLONED;
2450
2451 if (dst) {
2452 if (nla_put(skb, RTA_DST, 16, dst))
2453 goto nla_put_failure;
2454 rtm->rtm_dst_len = 128;
2455 } else if (rtm->rtm_dst_len)
2456 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2457 goto nla_put_failure;
2458#ifdef CONFIG_IPV6_SUBTREES
2459 if (src) {
2460 if (nla_put(skb, RTA_SRC, 16, src))
2461 goto nla_put_failure;
2462 rtm->rtm_src_len = 128;
2463 } else if (rtm->rtm_src_len &&
2464 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2465 goto nla_put_failure;
2466#endif
2467 if (iif) {
2468#ifdef CONFIG_IPV6_MROUTE
2469 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2470 int err = ip6mr_get_route(net, skb, rtm, nowait);
2471 if (err <= 0) {
2472 if (!nowait) {
2473 if (err == 0)
2474 return 0;
2475 goto nla_put_failure;
2476 } else {
2477 if (err == -EMSGSIZE)
2478 goto nla_put_failure;
2479 }
2480 }
2481 } else
2482#endif
2483 if (nla_put_u32(skb, RTA_IIF, iif))
2484 goto nla_put_failure;
2485 } else if (dst) {
2486 struct in6_addr saddr_buf;
2487 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2488 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2489 goto nla_put_failure;
2490 }
2491
2492 if (rt->rt6i_prefsrc.plen) {
2493 struct in6_addr saddr_buf;
2494 saddr_buf = rt->rt6i_prefsrc.addr;
2495 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2496 goto nla_put_failure;
2497 }
2498
2499 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2500 goto nla_put_failure;
2501
2502 rcu_read_lock();
2503 n = dst_get_neighbour_noref(&rt->dst);
2504 if (n) {
2505 if (nla_put(skb, RTA_GATEWAY, 16, &n->primary_key) < 0) {
2506 rcu_read_unlock();
2507 goto nla_put_failure;
2508 }
2509 }
2510 rcu_read_unlock();
2511
2512 if (rt->dst.dev &&
2513 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2514 goto nla_put_failure;
2515 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2516 goto nla_put_failure;
2517 if (!(rt->rt6i_flags & RTF_EXPIRES))
2518 expires = 0;
2519 else if (rt->dst.expires - jiffies < INT_MAX)
2520 expires = rt->dst.expires - jiffies;
2521 else
2522 expires = INT_MAX;
2523
2524 peer = rt->rt6i_peer;
2525 ts = tsage = 0;
2526 if (peer && peer->tcp_ts_stamp) {
2527 ts = peer->tcp_ts;
2528 tsage = get_seconds() - peer->tcp_ts_stamp;
2529 }
2530
2531 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, ts, tsage,
2532 expires, rt->dst.error) < 0)
2533 goto nla_put_failure;
2534
2535 return nlmsg_end(skb, nlh);
2536
2537nla_put_failure:
2538 nlmsg_cancel(skb, nlh);
2539 return -EMSGSIZE;
2540}
2541
2542int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2543{
2544 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2545 int prefix;
2546
2547 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2548 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2549 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2550 } else
2551 prefix = 0;
2552
2553 return rt6_fill_node(arg->net,
2554 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2555 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2556 prefix, 0, NLM_F_MULTI);
2557}
2558
2559static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2560{
2561 struct net *net = sock_net(in_skb->sk);
2562 struct nlattr *tb[RTA_MAX+1];
2563 struct rt6_info *rt;
2564 struct sk_buff *skb;
2565 struct rtmsg *rtm;
2566 struct flowi6 fl6;
2567 int err, iif = 0, oif = 0;
2568
2569 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2570 if (err < 0)
2571 goto errout;
2572
2573 err = -EINVAL;
2574 memset(&fl6, 0, sizeof(fl6));
2575
2576 if (tb[RTA_SRC]) {
2577 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2578 goto errout;
2579
2580 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2581 }
2582
2583 if (tb[RTA_DST]) {
2584 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2585 goto errout;
2586
2587 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2588 }
2589
2590 if (tb[RTA_IIF])
2591 iif = nla_get_u32(tb[RTA_IIF]);
2592
2593 if (tb[RTA_OIF])
2594 oif = nla_get_u32(tb[RTA_OIF]);
2595
2596 if (iif) {
2597 struct net_device *dev;
2598 int flags = 0;
2599
2600 dev = __dev_get_by_index(net, iif);
2601 if (!dev) {
2602 err = -ENODEV;
2603 goto errout;
2604 }
2605
2606 fl6.flowi6_iif = iif;
2607
2608 if (!ipv6_addr_any(&fl6.saddr))
2609 flags |= RT6_LOOKUP_F_HAS_SADDR;
2610
2611 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2612 flags);
2613 } else {
2614 fl6.flowi6_oif = oif;
2615
2616 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2617 }
2618
2619 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2620 if (!skb) {
2621 dst_release(&rt->dst);
2622 err = -ENOBUFS;
2623 goto errout;
2624 }
2625
2626 /* Reserve room for dummy headers, this skb can pass
2627 through good chunk of routing engine.
2628 */
2629 skb_reset_mac_header(skb);
2630 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2631
2632 skb_dst_set(skb, &rt->dst);
2633
2634 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2635 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2636 nlh->nlmsg_seq, 0, 0, 0);
2637 if (err < 0) {
2638 kfree_skb(skb);
2639 goto errout;
2640 }
2641
2642 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2643errout:
2644 return err;
2645}
2646
2647void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2648{
2649 struct sk_buff *skb;
2650 struct net *net = info->nl_net;
2651 u32 seq;
2652 int err;
2653
2654 err = -ENOBUFS;
2655 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2656
2657 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2658 if (!skb)
2659 goto errout;
2660
2661 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2662 event, info->pid, seq, 0, 0, 0);
2663 if (err < 0) {
2664 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2665 WARN_ON(err == -EMSGSIZE);
2666 kfree_skb(skb);
2667 goto errout;
2668 }
2669 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2670 info->nlh, gfp_any());
2671 return;
2672errout:
2673 if (err < 0)
2674 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2675}
2676
2677static int ip6_route_dev_notify(struct notifier_block *this,
2678 unsigned long event, void *data)
2679{
2680 struct net_device *dev = (struct net_device *)data;
2681 struct net *net = dev_net(dev);
2682
2683 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2684 net->ipv6.ip6_null_entry->dst.dev = dev;
2685 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2686#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2687 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2688 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2689 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2690 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2691#endif
2692 }
2693
2694 return NOTIFY_OK;
2695}
2696
2697/*
2698 * /proc
2699 */
2700
2701#ifdef CONFIG_PROC_FS
2702
2703struct rt6_proc_arg
2704{
2705 char *buffer;
2706 int offset;
2707 int length;
2708 int skip;
2709 int len;
2710};
2711
2712static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2713{
2714 struct seq_file *m = p_arg;
2715 struct neighbour *n;
2716
2717 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2718
2719#ifdef CONFIG_IPV6_SUBTREES
2720 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2721#else
2722 seq_puts(m, "00000000000000000000000000000000 00 ");
2723#endif
2724 rcu_read_lock();
2725 n = dst_get_neighbour_noref(&rt->dst);
2726 if (n) {
2727 seq_printf(m, "%pi6", n->primary_key);
2728 } else {
2729 seq_puts(m, "00000000000000000000000000000000");
2730 }
2731 rcu_read_unlock();
2732 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2733 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2734 rt->dst.__use, rt->rt6i_flags,
2735 rt->dst.dev ? rt->dst.dev->name : "");
2736 return 0;
2737}
2738
2739static int ipv6_route_show(struct seq_file *m, void *v)
2740{
2741 struct net *net = (struct net *)m->private;
2742 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2743 return 0;
2744}
2745
2746static int ipv6_route_open(struct inode *inode, struct file *file)
2747{
2748 return single_open_net(inode, file, ipv6_route_show);
2749}
2750
2751static const struct file_operations ipv6_route_proc_fops = {
2752 .owner = THIS_MODULE,
2753 .open = ipv6_route_open,
2754 .read = seq_read,
2755 .llseek = seq_lseek,
2756 .release = single_release_net,
2757};
2758
2759static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2760{
2761 struct net *net = (struct net *)seq->private;
2762 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2763 net->ipv6.rt6_stats->fib_nodes,
2764 net->ipv6.rt6_stats->fib_route_nodes,
2765 net->ipv6.rt6_stats->fib_rt_alloc,
2766 net->ipv6.rt6_stats->fib_rt_entries,
2767 net->ipv6.rt6_stats->fib_rt_cache,
2768 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2769 net->ipv6.rt6_stats->fib_discarded_routes);
2770
2771 return 0;
2772}
2773
2774static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2775{
2776 return single_open_net(inode, file, rt6_stats_seq_show);
2777}
2778
2779static const struct file_operations rt6_stats_seq_fops = {
2780 .owner = THIS_MODULE,
2781 .open = rt6_stats_seq_open,
2782 .read = seq_read,
2783 .llseek = seq_lseek,
2784 .release = single_release_net,
2785};
2786#endif /* CONFIG_PROC_FS */
2787
2788#ifdef CONFIG_SYSCTL
2789
2790static
2791int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2792 void __user *buffer, size_t *lenp, loff_t *ppos)
2793{
2794 struct net *net;
2795 int delay;
2796 if (!write)
2797 return -EINVAL;
2798
2799 net = (struct net *)ctl->extra1;
2800 delay = net->ipv6.sysctl.flush_delay;
2801 proc_dointvec(ctl, write, buffer, lenp, ppos);
2802 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2803 return 0;
2804}
2805
2806ctl_table ipv6_route_table_template[] = {
2807 {
2808 .procname = "flush",
2809 .data = &init_net.ipv6.sysctl.flush_delay,
2810 .maxlen = sizeof(int),
2811 .mode = 0200,
2812 .proc_handler = ipv6_sysctl_rtcache_flush
2813 },
2814 {
2815 .procname = "gc_thresh",
2816 .data = &ip6_dst_ops_template.gc_thresh,
2817 .maxlen = sizeof(int),
2818 .mode = 0644,
2819 .proc_handler = proc_dointvec,
2820 },
2821 {
2822 .procname = "max_size",
2823 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2824 .maxlen = sizeof(int),
2825 .mode = 0644,
2826 .proc_handler = proc_dointvec,
2827 },
2828 {
2829 .procname = "gc_min_interval",
2830 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2831 .maxlen = sizeof(int),
2832 .mode = 0644,
2833 .proc_handler = proc_dointvec_jiffies,
2834 },
2835 {
2836 .procname = "gc_timeout",
2837 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2838 .maxlen = sizeof(int),
2839 .mode = 0644,
2840 .proc_handler = proc_dointvec_jiffies,
2841 },
2842 {
2843 .procname = "gc_interval",
2844 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2845 .maxlen = sizeof(int),
2846 .mode = 0644,
2847 .proc_handler = proc_dointvec_jiffies,
2848 },
2849 {
2850 .procname = "gc_elasticity",
2851 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2852 .maxlen = sizeof(int),
2853 .mode = 0644,
2854 .proc_handler = proc_dointvec,
2855 },
2856 {
2857 .procname = "mtu_expires",
2858 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2859 .maxlen = sizeof(int),
2860 .mode = 0644,
2861 .proc_handler = proc_dointvec_jiffies,
2862 },
2863 {
2864 .procname = "min_adv_mss",
2865 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2866 .maxlen = sizeof(int),
2867 .mode = 0644,
2868 .proc_handler = proc_dointvec,
2869 },
2870 {
2871 .procname = "gc_min_interval_ms",
2872 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2873 .maxlen = sizeof(int),
2874 .mode = 0644,
2875 .proc_handler = proc_dointvec_ms_jiffies,
2876 },
2877 { }
2878};
2879
2880struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2881{
2882 struct ctl_table *table;
2883
2884 table = kmemdup(ipv6_route_table_template,
2885 sizeof(ipv6_route_table_template),
2886 GFP_KERNEL);
2887
2888 if (table) {
2889 table[0].data = &net->ipv6.sysctl.flush_delay;
2890 table[0].extra1 = net;
2891 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2892 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2893 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2894 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2895 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2896 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2897 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2898 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2899 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2900 }
2901
2902 return table;
2903}
2904#endif
2905
2906static int __net_init ip6_route_net_init(struct net *net)
2907{
2908 int ret = -ENOMEM;
2909
2910 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2911 sizeof(net->ipv6.ip6_dst_ops));
2912
2913 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2914 goto out_ip6_dst_ops;
2915
2916 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2917 sizeof(*net->ipv6.ip6_null_entry),
2918 GFP_KERNEL);
2919 if (!net->ipv6.ip6_null_entry)
2920 goto out_ip6_dst_entries;
2921 net->ipv6.ip6_null_entry->dst.path =
2922 (struct dst_entry *)net->ipv6.ip6_null_entry;
2923 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2924 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2925 ip6_template_metrics, true);
2926
2927#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2928 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2929 sizeof(*net->ipv6.ip6_prohibit_entry),
2930 GFP_KERNEL);
2931 if (!net->ipv6.ip6_prohibit_entry)
2932 goto out_ip6_null_entry;
2933 net->ipv6.ip6_prohibit_entry->dst.path =
2934 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2935 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2936 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2937 ip6_template_metrics, true);
2938
2939 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2940 sizeof(*net->ipv6.ip6_blk_hole_entry),
2941 GFP_KERNEL);
2942 if (!net->ipv6.ip6_blk_hole_entry)
2943 goto out_ip6_prohibit_entry;
2944 net->ipv6.ip6_blk_hole_entry->dst.path =
2945 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2946 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2947 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2948 ip6_template_metrics, true);
2949#endif
2950
2951 net->ipv6.sysctl.flush_delay = 0;
2952 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2953 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2954 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2955 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2956 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2957 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2958 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2959
2960 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2961
2962 ret = 0;
2963out:
2964 return ret;
2965
2966#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2967out_ip6_prohibit_entry:
2968 kfree(net->ipv6.ip6_prohibit_entry);
2969out_ip6_null_entry:
2970 kfree(net->ipv6.ip6_null_entry);
2971#endif
2972out_ip6_dst_entries:
2973 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2974out_ip6_dst_ops:
2975 goto out;
2976}
2977
2978static void __net_exit ip6_route_net_exit(struct net *net)
2979{
2980 kfree(net->ipv6.ip6_null_entry);
2981#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2982 kfree(net->ipv6.ip6_prohibit_entry);
2983 kfree(net->ipv6.ip6_blk_hole_entry);
2984#endif
2985 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2986}
2987
2988static int __net_init ip6_route_net_init_late(struct net *net)
2989{
2990#ifdef CONFIG_PROC_FS
2991 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2992 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2993#endif
2994 return 0;
2995}
2996
2997static void __net_exit ip6_route_net_exit_late(struct net *net)
2998{
2999#ifdef CONFIG_PROC_FS
3000 proc_net_remove(net, "ipv6_route");
3001 proc_net_remove(net, "rt6_stats");
3002#endif
3003}
3004
3005static struct pernet_operations ip6_route_net_ops = {
3006 .init = ip6_route_net_init,
3007 .exit = ip6_route_net_exit,
3008};
3009
3010static struct pernet_operations ip6_route_net_late_ops = {
3011 .init = ip6_route_net_init_late,
3012 .exit = ip6_route_net_exit_late,
3013};
3014
3015static struct notifier_block ip6_route_dev_notifier = {
3016 .notifier_call = ip6_route_dev_notify,
3017 .priority = 0,
3018};
3019
3020int __init ip6_route_init(void)
3021{
3022 int ret;
3023
3024 ret = -ENOMEM;
3025 ip6_dst_ops_template.kmem_cachep =
3026 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3027 SLAB_HWCACHE_ALIGN, NULL);
3028 if (!ip6_dst_ops_template.kmem_cachep)
3029 goto out;
3030
3031 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3032 if (ret)
3033 goto out_kmem_cache;
3034
3035 ret = register_pernet_subsys(&ip6_route_net_ops);
3036 if (ret)
3037 goto out_dst_entries;
3038
3039 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3040
3041 /* Registering of the loopback is done before this portion of code,
3042 * the loopback reference in rt6_info will not be taken, do it
3043 * manually for init_net */
3044 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3045 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3046 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3047 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3048 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3049 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3050 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3051 #endif
3052 ret = fib6_init();
3053 if (ret)
3054 goto out_register_subsys;
3055
3056 ret = xfrm6_init();
3057 if (ret)
3058 goto out_fib6_init;
3059
3060 ret = fib6_rules_init();
3061 if (ret)
3062 goto xfrm6_init;
3063
3064 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3065 if (ret)
3066 goto fib6_rules_init;
3067
3068 ret = -ENOBUFS;
3069 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3070 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3071 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3072 goto out_register_late_subsys;
3073
3074 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3075 if (ret)
3076 goto out_register_late_subsys;
3077
3078out:
3079 return ret;
3080
3081out_register_late_subsys:
3082 unregister_pernet_subsys(&ip6_route_net_late_ops);
3083fib6_rules_init:
3084 fib6_rules_cleanup();
3085xfrm6_init:
3086 xfrm6_fini();
3087out_fib6_init:
3088 fib6_gc_cleanup();
3089out_register_subsys:
3090 unregister_pernet_subsys(&ip6_route_net_ops);
3091out_dst_entries:
3092 dst_entries_destroy(&ip6_dst_blackhole_ops);
3093out_kmem_cache:
3094 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3095 goto out;
3096}
3097
3098void ip6_route_cleanup(void)
3099{
3100 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3101 unregister_pernet_subsys(&ip6_route_net_late_ops);
3102 fib6_rules_cleanup();
3103 xfrm6_fini();
3104 fib6_gc_cleanup();
3105 unregister_pernet_subsys(&ip6_route_net_ops);
3106 dst_entries_destroy(&ip6_dst_blackhole_ops);
3107 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3108}