ipv6: Use ERR_CAST in addrconf_dst_alloc.
[linux-2.6.git] / net / ipv6 / route.c
1 /*
2  *      Linux INET6 implementation
3  *      FIB front-end.
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *
8  *      This program is free software; you can redistribute it and/or
9  *      modify it under the terms of the GNU General Public License
10  *      as published by the Free Software Foundation; either version
11  *      2 of the License, or (at your option) any later version.
12  */
13
14 /*      Changes:
15  *
16  *      YOSHIFUJI Hideaki @USAGI
17  *              reworked default router selection.
18  *              - respect outgoing interface
19  *              - select from (probably) reachable routers (i.e.
20  *              routers in REACHABLE, STALE, DELAY or PROBE states).
21  *              - always select the same router if it is (probably)
22  *              reachable.  otherwise, round-robin the list.
23  *      Ville Nuorvala
24  *              Fixed routing subtrees.
25  */
26
27 #include <linux/capability.h>
28 #include <linux/errno.h>
29 #include <linux/types.h>
30 #include <linux/times.h>
31 #include <linux/socket.h>
32 #include <linux/sockios.h>
33 #include <linux/net.h>
34 #include <linux/route.h>
35 #include <linux/netdevice.h>
36 #include <linux/in6.h>
37 #include <linux/mroute6.h>
38 #include <linux/init.h>
39 #include <linux/if_arp.h>
40 #include <linux/proc_fs.h>
41 #include <linux/seq_file.h>
42 #include <linux/nsproxy.h>
43 #include <linux/slab.h>
44 #include <net/net_namespace.h>
45 #include <net/snmp.h>
46 #include <net/ipv6.h>
47 #include <net/ip6_fib.h>
48 #include <net/ip6_route.h>
49 #include <net/ndisc.h>
50 #include <net/addrconf.h>
51 #include <net/tcp.h>
52 #include <linux/rtnetlink.h>
53 #include <net/dst.h>
54 #include <net/xfrm.h>
55 #include <net/netevent.h>
56 #include <net/netlink.h>
57
58 #include <asm/uaccess.h>
59
60 #ifdef CONFIG_SYSCTL
61 #include <linux/sysctl.h>
62 #endif
63
64 /* Set to 3 to get tracing. */
65 #define RT6_DEBUG 2
66
67 #if RT6_DEBUG >= 3
68 #define RDBG(x) printk x
69 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
70 #else
71 #define RDBG(x)
72 #define RT6_TRACE(x...) do { ; } while (0)
73 #endif
74
75 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
76 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
77 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
78 static unsigned int      ip6_default_mtu(const struct dst_entry *dst);
79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
80 static void             ip6_dst_destroy(struct dst_entry *);
81 static void             ip6_dst_ifdown(struct dst_entry *,
82                                        struct net_device *dev, int how);
83 static int               ip6_dst_gc(struct dst_ops *ops);
84
85 static int              ip6_pkt_discard(struct sk_buff *skb);
86 static int              ip6_pkt_discard_out(struct sk_buff *skb);
87 static void             ip6_link_failure(struct sk_buff *skb);
88 static void             ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
89
90 #ifdef CONFIG_IPV6_ROUTE_INFO
91 static struct rt6_info *rt6_add_route_info(struct net *net,
92                                            struct in6_addr *prefix, int prefixlen,
93                                            struct in6_addr *gwaddr, int ifindex,
94                                            unsigned pref);
95 static struct rt6_info *rt6_get_route_info(struct net *net,
96                                            struct in6_addr *prefix, int prefixlen,
97                                            struct in6_addr *gwaddr, int ifindex);
98 #endif
99
100 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
101 {
102         struct rt6_info *rt = (struct rt6_info *) dst;
103         struct inet_peer *peer;
104         u32 *p = NULL;
105
106         if (!rt->rt6i_peer)
107                 rt6_bind_peer(rt, 1);
108
109         peer = rt->rt6i_peer;
110         if (peer) {
111                 u32 *old_p = __DST_METRICS_PTR(old);
112                 unsigned long prev, new;
113
114                 p = peer->metrics;
115                 if (inet_metrics_new(peer))
116                         memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
117
118                 new = (unsigned long) p;
119                 prev = cmpxchg(&dst->_metrics, old, new);
120
121                 if (prev != old) {
122                         p = __DST_METRICS_PTR(prev);
123                         if (prev & DST_METRICS_READ_ONLY)
124                                 p = NULL;
125                 }
126         }
127         return p;
128 }
129
130 static struct dst_ops ip6_dst_ops_template = {
131         .family                 =       AF_INET6,
132         .protocol               =       cpu_to_be16(ETH_P_IPV6),
133         .gc                     =       ip6_dst_gc,
134         .gc_thresh              =       1024,
135         .check                  =       ip6_dst_check,
136         .default_advmss         =       ip6_default_advmss,
137         .default_mtu            =       ip6_default_mtu,
138         .cow_metrics            =       ipv6_cow_metrics,
139         .destroy                =       ip6_dst_destroy,
140         .ifdown                 =       ip6_dst_ifdown,
141         .negative_advice        =       ip6_negative_advice,
142         .link_failure           =       ip6_link_failure,
143         .update_pmtu            =       ip6_rt_update_pmtu,
144         .local_out              =       __ip6_local_out,
145 };
146
147 static unsigned int ip6_blackhole_default_mtu(const struct dst_entry *dst)
148 {
149         return 0;
150 }
151
152 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
153 {
154 }
155
156 static struct dst_ops ip6_dst_blackhole_ops = {
157         .family                 =       AF_INET6,
158         .protocol               =       cpu_to_be16(ETH_P_IPV6),
159         .destroy                =       ip6_dst_destroy,
160         .check                  =       ip6_dst_check,
161         .default_mtu            =       ip6_blackhole_default_mtu,
162         .default_advmss         =       ip6_default_advmss,
163         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
164 };
165
166 static const u32 ip6_template_metrics[RTAX_MAX] = {
167         [RTAX_HOPLIMIT - 1] = 255,
168 };
169
170 static struct rt6_info ip6_null_entry_template = {
171         .dst = {
172                 .__refcnt       = ATOMIC_INIT(1),
173                 .__use          = 1,
174                 .obsolete       = -1,
175                 .error          = -ENETUNREACH,
176                 .input          = ip6_pkt_discard,
177                 .output         = ip6_pkt_discard_out,
178         },
179         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
180         .rt6i_protocol  = RTPROT_KERNEL,
181         .rt6i_metric    = ~(u32) 0,
182         .rt6i_ref       = ATOMIC_INIT(1),
183 };
184
185 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
186
187 static int ip6_pkt_prohibit(struct sk_buff *skb);
188 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
189
190 static struct rt6_info ip6_prohibit_entry_template = {
191         .dst = {
192                 .__refcnt       = ATOMIC_INIT(1),
193                 .__use          = 1,
194                 .obsolete       = -1,
195                 .error          = -EACCES,
196                 .input          = ip6_pkt_prohibit,
197                 .output         = ip6_pkt_prohibit_out,
198         },
199         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
200         .rt6i_protocol  = RTPROT_KERNEL,
201         .rt6i_metric    = ~(u32) 0,
202         .rt6i_ref       = ATOMIC_INIT(1),
203 };
204
205 static struct rt6_info ip6_blk_hole_entry_template = {
206         .dst = {
207                 .__refcnt       = ATOMIC_INIT(1),
208                 .__use          = 1,
209                 .obsolete       = -1,
210                 .error          = -EINVAL,
211                 .input          = dst_discard,
212                 .output         = dst_discard,
213         },
214         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
215         .rt6i_protocol  = RTPROT_KERNEL,
216         .rt6i_metric    = ~(u32) 0,
217         .rt6i_ref       = ATOMIC_INIT(1),
218 };
219
220 #endif
221
222 /* allocate dst with ip6_dst_ops */
223 static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops)
224 {
225         return (struct rt6_info *)dst_alloc(ops, 0);
226 }
227
228 static void ip6_dst_destroy(struct dst_entry *dst)
229 {
230         struct rt6_info *rt = (struct rt6_info *)dst;
231         struct inet6_dev *idev = rt->rt6i_idev;
232         struct inet_peer *peer = rt->rt6i_peer;
233
234         if (idev != NULL) {
235                 rt->rt6i_idev = NULL;
236                 in6_dev_put(idev);
237         }
238         if (peer) {
239                 rt->rt6i_peer = NULL;
240                 inet_putpeer(peer);
241         }
242 }
243
244 static atomic_t __rt6_peer_genid = ATOMIC_INIT(0);
245
246 static u32 rt6_peer_genid(void)
247 {
248         return atomic_read(&__rt6_peer_genid);
249 }
250
251 void rt6_bind_peer(struct rt6_info *rt, int create)
252 {
253         struct inet_peer *peer;
254
255         peer = inet_getpeer_v6(&rt->rt6i_dst.addr, create);
256         if (peer && cmpxchg(&rt->rt6i_peer, NULL, peer) != NULL)
257                 inet_putpeer(peer);
258         else
259                 rt->rt6i_peer_genid = rt6_peer_genid();
260 }
261
262 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
263                            int how)
264 {
265         struct rt6_info *rt = (struct rt6_info *)dst;
266         struct inet6_dev *idev = rt->rt6i_idev;
267         struct net_device *loopback_dev =
268                 dev_net(dev)->loopback_dev;
269
270         if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
271                 struct inet6_dev *loopback_idev =
272                         in6_dev_get(loopback_dev);
273                 if (loopback_idev != NULL) {
274                         rt->rt6i_idev = loopback_idev;
275                         in6_dev_put(idev);
276                 }
277         }
278 }
279
280 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
281 {
282         return (rt->rt6i_flags & RTF_EXPIRES) &&
283                 time_after(jiffies, rt->rt6i_expires);
284 }
285
286 static inline int rt6_need_strict(struct in6_addr *daddr)
287 {
288         return ipv6_addr_type(daddr) &
289                 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
290 }
291
292 /*
293  *      Route lookup. Any table->tb6_lock is implied.
294  */
295
296 static inline struct rt6_info *rt6_device_match(struct net *net,
297                                                     struct rt6_info *rt,
298                                                     struct in6_addr *saddr,
299                                                     int oif,
300                                                     int flags)
301 {
302         struct rt6_info *local = NULL;
303         struct rt6_info *sprt;
304
305         if (!oif && ipv6_addr_any(saddr))
306                 goto out;
307
308         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
309                 struct net_device *dev = sprt->rt6i_dev;
310
311                 if (oif) {
312                         if (dev->ifindex == oif)
313                                 return sprt;
314                         if (dev->flags & IFF_LOOPBACK) {
315                                 if (sprt->rt6i_idev == NULL ||
316                                     sprt->rt6i_idev->dev->ifindex != oif) {
317                                         if (flags & RT6_LOOKUP_F_IFACE && oif)
318                                                 continue;
319                                         if (local && (!oif ||
320                                                       local->rt6i_idev->dev->ifindex == oif))
321                                                 continue;
322                                 }
323                                 local = sprt;
324                         }
325                 } else {
326                         if (ipv6_chk_addr(net, saddr, dev,
327                                           flags & RT6_LOOKUP_F_IFACE))
328                                 return sprt;
329                 }
330         }
331
332         if (oif) {
333                 if (local)
334                         return local;
335
336                 if (flags & RT6_LOOKUP_F_IFACE)
337                         return net->ipv6.ip6_null_entry;
338         }
339 out:
340         return rt;
341 }
342
343 #ifdef CONFIG_IPV6_ROUTER_PREF
344 static void rt6_probe(struct rt6_info *rt)
345 {
346         struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
347         /*
348          * Okay, this does not seem to be appropriate
349          * for now, however, we need to check if it
350          * is really so; aka Router Reachability Probing.
351          *
352          * Router Reachability Probe MUST be rate-limited
353          * to no more than one per minute.
354          */
355         if (!neigh || (neigh->nud_state & NUD_VALID))
356                 return;
357         read_lock_bh(&neigh->lock);
358         if (!(neigh->nud_state & NUD_VALID) &&
359             time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
360                 struct in6_addr mcaddr;
361                 struct in6_addr *target;
362
363                 neigh->updated = jiffies;
364                 read_unlock_bh(&neigh->lock);
365
366                 target = (struct in6_addr *)&neigh->primary_key;
367                 addrconf_addr_solict_mult(target, &mcaddr);
368                 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
369         } else
370                 read_unlock_bh(&neigh->lock);
371 }
372 #else
373 static inline void rt6_probe(struct rt6_info *rt)
374 {
375 }
376 #endif
377
378 /*
379  * Default Router Selection (RFC 2461 6.3.6)
380  */
381 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
382 {
383         struct net_device *dev = rt->rt6i_dev;
384         if (!oif || dev->ifindex == oif)
385                 return 2;
386         if ((dev->flags & IFF_LOOPBACK) &&
387             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
388                 return 1;
389         return 0;
390 }
391
392 static inline int rt6_check_neigh(struct rt6_info *rt)
393 {
394         struct neighbour *neigh = rt->rt6i_nexthop;
395         int m;
396         if (rt->rt6i_flags & RTF_NONEXTHOP ||
397             !(rt->rt6i_flags & RTF_GATEWAY))
398                 m = 1;
399         else if (neigh) {
400                 read_lock_bh(&neigh->lock);
401                 if (neigh->nud_state & NUD_VALID)
402                         m = 2;
403 #ifdef CONFIG_IPV6_ROUTER_PREF
404                 else if (neigh->nud_state & NUD_FAILED)
405                         m = 0;
406 #endif
407                 else
408                         m = 1;
409                 read_unlock_bh(&neigh->lock);
410         } else
411                 m = 0;
412         return m;
413 }
414
415 static int rt6_score_route(struct rt6_info *rt, int oif,
416                            int strict)
417 {
418         int m, n;
419
420         m = rt6_check_dev(rt, oif);
421         if (!m && (strict & RT6_LOOKUP_F_IFACE))
422                 return -1;
423 #ifdef CONFIG_IPV6_ROUTER_PREF
424         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
425 #endif
426         n = rt6_check_neigh(rt);
427         if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
428                 return -1;
429         return m;
430 }
431
432 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
433                                    int *mpri, struct rt6_info *match)
434 {
435         int m;
436
437         if (rt6_check_expired(rt))
438                 goto out;
439
440         m = rt6_score_route(rt, oif, strict);
441         if (m < 0)
442                 goto out;
443
444         if (m > *mpri) {
445                 if (strict & RT6_LOOKUP_F_REACHABLE)
446                         rt6_probe(match);
447                 *mpri = m;
448                 match = rt;
449         } else if (strict & RT6_LOOKUP_F_REACHABLE) {
450                 rt6_probe(rt);
451         }
452
453 out:
454         return match;
455 }
456
457 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
458                                      struct rt6_info *rr_head,
459                                      u32 metric, int oif, int strict)
460 {
461         struct rt6_info *rt, *match;
462         int mpri = -1;
463
464         match = NULL;
465         for (rt = rr_head; rt && rt->rt6i_metric == metric;
466              rt = rt->dst.rt6_next)
467                 match = find_match(rt, oif, strict, &mpri, match);
468         for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
469              rt = rt->dst.rt6_next)
470                 match = find_match(rt, oif, strict, &mpri, match);
471
472         return match;
473 }
474
475 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
476 {
477         struct rt6_info *match, *rt0;
478         struct net *net;
479
480         RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
481                   __func__, fn->leaf, oif);
482
483         rt0 = fn->rr_ptr;
484         if (!rt0)
485                 fn->rr_ptr = rt0 = fn->leaf;
486
487         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
488
489         if (!match &&
490             (strict & RT6_LOOKUP_F_REACHABLE)) {
491                 struct rt6_info *next = rt0->dst.rt6_next;
492
493                 /* no entries matched; do round-robin */
494                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
495                         next = fn->leaf;
496
497                 if (next != rt0)
498                         fn->rr_ptr = next;
499         }
500
501         RT6_TRACE("%s() => %p\n",
502                   __func__, match);
503
504         net = dev_net(rt0->rt6i_dev);
505         return match ? match : net->ipv6.ip6_null_entry;
506 }
507
508 #ifdef CONFIG_IPV6_ROUTE_INFO
509 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
510                   struct in6_addr *gwaddr)
511 {
512         struct net *net = dev_net(dev);
513         struct route_info *rinfo = (struct route_info *) opt;
514         struct in6_addr prefix_buf, *prefix;
515         unsigned int pref;
516         unsigned long lifetime;
517         struct rt6_info *rt;
518
519         if (len < sizeof(struct route_info)) {
520                 return -EINVAL;
521         }
522
523         /* Sanity check for prefix_len and length */
524         if (rinfo->length > 3) {
525                 return -EINVAL;
526         } else if (rinfo->prefix_len > 128) {
527                 return -EINVAL;
528         } else if (rinfo->prefix_len > 64) {
529                 if (rinfo->length < 2) {
530                         return -EINVAL;
531                 }
532         } else if (rinfo->prefix_len > 0) {
533                 if (rinfo->length < 1) {
534                         return -EINVAL;
535                 }
536         }
537
538         pref = rinfo->route_pref;
539         if (pref == ICMPV6_ROUTER_PREF_INVALID)
540                 return -EINVAL;
541
542         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
543
544         if (rinfo->length == 3)
545                 prefix = (struct in6_addr *)rinfo->prefix;
546         else {
547                 /* this function is safe */
548                 ipv6_addr_prefix(&prefix_buf,
549                                  (struct in6_addr *)rinfo->prefix,
550                                  rinfo->prefix_len);
551                 prefix = &prefix_buf;
552         }
553
554         rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
555                                 dev->ifindex);
556
557         if (rt && !lifetime) {
558                 ip6_del_rt(rt);
559                 rt = NULL;
560         }
561
562         if (!rt && lifetime)
563                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
564                                         pref);
565         else if (rt)
566                 rt->rt6i_flags = RTF_ROUTEINFO |
567                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
568
569         if (rt) {
570                 if (!addrconf_finite_timeout(lifetime)) {
571                         rt->rt6i_flags &= ~RTF_EXPIRES;
572                 } else {
573                         rt->rt6i_expires = jiffies + HZ * lifetime;
574                         rt->rt6i_flags |= RTF_EXPIRES;
575                 }
576                 dst_release(&rt->dst);
577         }
578         return 0;
579 }
580 #endif
581
582 #define BACKTRACK(__net, saddr)                 \
583 do { \
584         if (rt == __net->ipv6.ip6_null_entry) { \
585                 struct fib6_node *pn; \
586                 while (1) { \
587                         if (fn->fn_flags & RTN_TL_ROOT) \
588                                 goto out; \
589                         pn = fn->parent; \
590                         if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
591                                 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
592                         else \
593                                 fn = pn; \
594                         if (fn->fn_flags & RTN_RTINFO) \
595                                 goto restart; \
596                 } \
597         } \
598 } while(0)
599
600 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
601                                              struct fib6_table *table,
602                                              struct flowi *fl, int flags)
603 {
604         struct fib6_node *fn;
605         struct rt6_info *rt;
606
607         read_lock_bh(&table->tb6_lock);
608         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
609 restart:
610         rt = fn->leaf;
611         rt = rt6_device_match(net, rt, &fl->fl6_src, fl->oif, flags);
612         BACKTRACK(net, &fl->fl6_src);
613 out:
614         dst_use(&rt->dst, jiffies);
615         read_unlock_bh(&table->tb6_lock);
616         return rt;
617
618 }
619
620 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
621                             const struct in6_addr *saddr, int oif, int strict)
622 {
623         struct flowi fl = {
624                 .oif = oif,
625                 .fl6_dst = *daddr,
626         };
627         struct dst_entry *dst;
628         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
629
630         if (saddr) {
631                 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
632                 flags |= RT6_LOOKUP_F_HAS_SADDR;
633         }
634
635         dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_lookup);
636         if (dst->error == 0)
637                 return (struct rt6_info *) dst;
638
639         dst_release(dst);
640
641         return NULL;
642 }
643
644 EXPORT_SYMBOL(rt6_lookup);
645
646 /* ip6_ins_rt is called with FREE table->tb6_lock.
647    It takes new route entry, the addition fails by any reason the
648    route is freed. In any case, if caller does not hold it, it may
649    be destroyed.
650  */
651
652 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
653 {
654         int err;
655         struct fib6_table *table;
656
657         table = rt->rt6i_table;
658         write_lock_bh(&table->tb6_lock);
659         err = fib6_add(&table->tb6_root, rt, info);
660         write_unlock_bh(&table->tb6_lock);
661
662         return err;
663 }
664
665 int ip6_ins_rt(struct rt6_info *rt)
666 {
667         struct nl_info info = {
668                 .nl_net = dev_net(rt->rt6i_dev),
669         };
670         return __ip6_ins_rt(rt, &info);
671 }
672
673 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
674                                       struct in6_addr *saddr)
675 {
676         struct rt6_info *rt;
677
678         /*
679          *      Clone the route.
680          */
681
682         rt = ip6_rt_copy(ort);
683
684         if (rt) {
685                 struct neighbour *neigh;
686                 int attempts = !in_softirq();
687
688                 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
689                         if (rt->rt6i_dst.plen != 128 &&
690                             ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
691                                 rt->rt6i_flags |= RTF_ANYCAST;
692                         ipv6_addr_copy(&rt->rt6i_gateway, daddr);
693                 }
694
695                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
696                 rt->rt6i_dst.plen = 128;
697                 rt->rt6i_flags |= RTF_CACHE;
698                 rt->dst.flags |= DST_HOST;
699
700 #ifdef CONFIG_IPV6_SUBTREES
701                 if (rt->rt6i_src.plen && saddr) {
702                         ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
703                         rt->rt6i_src.plen = 128;
704                 }
705 #endif
706
707         retry:
708                 neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
709                 if (IS_ERR(neigh)) {
710                         struct net *net = dev_net(rt->rt6i_dev);
711                         int saved_rt_min_interval =
712                                 net->ipv6.sysctl.ip6_rt_gc_min_interval;
713                         int saved_rt_elasticity =
714                                 net->ipv6.sysctl.ip6_rt_gc_elasticity;
715
716                         if (attempts-- > 0) {
717                                 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
718                                 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
719
720                                 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
721
722                                 net->ipv6.sysctl.ip6_rt_gc_elasticity =
723                                         saved_rt_elasticity;
724                                 net->ipv6.sysctl.ip6_rt_gc_min_interval =
725                                         saved_rt_min_interval;
726                                 goto retry;
727                         }
728
729                         if (net_ratelimit())
730                                 printk(KERN_WARNING
731                                        "ipv6: Neighbour table overflow.\n");
732                         dst_free(&rt->dst);
733                         return NULL;
734                 }
735                 rt->rt6i_nexthop = neigh;
736
737         }
738
739         return rt;
740 }
741
742 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
743 {
744         struct rt6_info *rt = ip6_rt_copy(ort);
745         if (rt) {
746                 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
747                 rt->rt6i_dst.plen = 128;
748                 rt->rt6i_flags |= RTF_CACHE;
749                 rt->dst.flags |= DST_HOST;
750                 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
751         }
752         return rt;
753 }
754
755 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
756                                       struct flowi *fl, int flags)
757 {
758         struct fib6_node *fn;
759         struct rt6_info *rt, *nrt;
760         int strict = 0;
761         int attempts = 3;
762         int err;
763         int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
764
765         strict |= flags & RT6_LOOKUP_F_IFACE;
766
767 relookup:
768         read_lock_bh(&table->tb6_lock);
769
770 restart_2:
771         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
772
773 restart:
774         rt = rt6_select(fn, oif, strict | reachable);
775
776         BACKTRACK(net, &fl->fl6_src);
777         if (rt == net->ipv6.ip6_null_entry ||
778             rt->rt6i_flags & RTF_CACHE)
779                 goto out;
780
781         dst_hold(&rt->dst);
782         read_unlock_bh(&table->tb6_lock);
783
784         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
785                 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
786         else
787                 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
788
789         dst_release(&rt->dst);
790         rt = nrt ? : net->ipv6.ip6_null_entry;
791
792         dst_hold(&rt->dst);
793         if (nrt) {
794                 err = ip6_ins_rt(nrt);
795                 if (!err)
796                         goto out2;
797         }
798
799         if (--attempts <= 0)
800                 goto out2;
801
802         /*
803          * Race condition! In the gap, when table->tb6_lock was
804          * released someone could insert this route.  Relookup.
805          */
806         dst_release(&rt->dst);
807         goto relookup;
808
809 out:
810         if (reachable) {
811                 reachable = 0;
812                 goto restart_2;
813         }
814         dst_hold(&rt->dst);
815         read_unlock_bh(&table->tb6_lock);
816 out2:
817         rt->dst.lastuse = jiffies;
818         rt->dst.__use++;
819
820         return rt;
821 }
822
823 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
824                                             struct flowi *fl, int flags)
825 {
826         return ip6_pol_route(net, table, fl->iif, fl, flags);
827 }
828
829 void ip6_route_input(struct sk_buff *skb)
830 {
831         struct ipv6hdr *iph = ipv6_hdr(skb);
832         struct net *net = dev_net(skb->dev);
833         int flags = RT6_LOOKUP_F_HAS_SADDR;
834         struct flowi fl = {
835                 .iif = skb->dev->ifindex,
836                 .fl6_dst = iph->daddr,
837                 .fl6_src = iph->saddr,
838                 .fl6_flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
839                 .mark = skb->mark,
840                 .proto = iph->nexthdr,
841         };
842
843         if (rt6_need_strict(&iph->daddr) && skb->dev->type != ARPHRD_PIMREG)
844                 flags |= RT6_LOOKUP_F_IFACE;
845
846         skb_dst_set(skb, fib6_rule_lookup(net, &fl, flags, ip6_pol_route_input));
847 }
848
849 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
850                                              struct flowi *fl, int flags)
851 {
852         return ip6_pol_route(net, table, fl->oif, fl, flags);
853 }
854
855 struct dst_entry * ip6_route_output(struct net *net, struct sock *sk,
856                                     struct flowi *fl)
857 {
858         int flags = 0;
859
860         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl->fl6_dst))
861                 flags |= RT6_LOOKUP_F_IFACE;
862
863         if (!ipv6_addr_any(&fl->fl6_src))
864                 flags |= RT6_LOOKUP_F_HAS_SADDR;
865         else if (sk)
866                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
867
868         return fib6_rule_lookup(net, fl, flags, ip6_pol_route_output);
869 }
870
871 EXPORT_SYMBOL(ip6_route_output);
872
873 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
874 {
875         struct rt6_info *rt = dst_alloc(&ip6_dst_blackhole_ops, 1);
876         struct rt6_info *ort = (struct rt6_info *) dst_orig;
877         struct dst_entry *new = NULL;
878
879         if (rt) {
880                 new = &rt->dst;
881
882                 new->__use = 1;
883                 new->input = dst_discard;
884                 new->output = dst_discard;
885
886                 dst_copy_metrics(new, &ort->dst);
887                 new->dev = ort->dst.dev;
888                 if (new->dev)
889                         dev_hold(new->dev);
890                 rt->rt6i_idev = ort->rt6i_idev;
891                 if (rt->rt6i_idev)
892                         in6_dev_hold(rt->rt6i_idev);
893                 rt->rt6i_expires = 0;
894
895                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
896                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
897                 rt->rt6i_metric = 0;
898
899                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
900 #ifdef CONFIG_IPV6_SUBTREES
901                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
902 #endif
903
904                 dst_free(new);
905         }
906
907         dst_release(dst_orig);
908         return new ? new : ERR_PTR(-ENOMEM);
909 }
910
911 /*
912  *      Destination cache support functions
913  */
914
915 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
916 {
917         struct rt6_info *rt;
918
919         rt = (struct rt6_info *) dst;
920
921         if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie)) {
922                 if (rt->rt6i_peer_genid != rt6_peer_genid()) {
923                         if (!rt->rt6i_peer)
924                                 rt6_bind_peer(rt, 0);
925                         rt->rt6i_peer_genid = rt6_peer_genid();
926                 }
927                 return dst;
928         }
929         return NULL;
930 }
931
932 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
933 {
934         struct rt6_info *rt = (struct rt6_info *) dst;
935
936         if (rt) {
937                 if (rt->rt6i_flags & RTF_CACHE) {
938                         if (rt6_check_expired(rt)) {
939                                 ip6_del_rt(rt);
940                                 dst = NULL;
941                         }
942                 } else {
943                         dst_release(dst);
944                         dst = NULL;
945                 }
946         }
947         return dst;
948 }
949
950 static void ip6_link_failure(struct sk_buff *skb)
951 {
952         struct rt6_info *rt;
953
954         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
955
956         rt = (struct rt6_info *) skb_dst(skb);
957         if (rt) {
958                 if (rt->rt6i_flags&RTF_CACHE) {
959                         dst_set_expires(&rt->dst, 0);
960                         rt->rt6i_flags |= RTF_EXPIRES;
961                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
962                         rt->rt6i_node->fn_sernum = -1;
963         }
964 }
965
966 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
967 {
968         struct rt6_info *rt6 = (struct rt6_info*)dst;
969
970         if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
971                 rt6->rt6i_flags |= RTF_MODIFIED;
972                 if (mtu < IPV6_MIN_MTU) {
973                         u32 features = dst_metric(dst, RTAX_FEATURES);
974                         mtu = IPV6_MIN_MTU;
975                         features |= RTAX_FEATURE_ALLFRAG;
976                         dst_metric_set(dst, RTAX_FEATURES, features);
977                 }
978                 dst_metric_set(dst, RTAX_MTU, mtu);
979         }
980 }
981
982 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
983 {
984         struct net_device *dev = dst->dev;
985         unsigned int mtu = dst_mtu(dst);
986         struct net *net = dev_net(dev);
987
988         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
989
990         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
991                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
992
993         /*
994          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
995          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
996          * IPV6_MAXPLEN is also valid and means: "any MSS,
997          * rely only on pmtu discovery"
998          */
999         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1000                 mtu = IPV6_MAXPLEN;
1001         return mtu;
1002 }
1003
1004 static unsigned int ip6_default_mtu(const struct dst_entry *dst)
1005 {
1006         unsigned int mtu = IPV6_MIN_MTU;
1007         struct inet6_dev *idev;
1008
1009         rcu_read_lock();
1010         idev = __in6_dev_get(dst->dev);
1011         if (idev)
1012                 mtu = idev->cnf.mtu6;
1013         rcu_read_unlock();
1014
1015         return mtu;
1016 }
1017
1018 static struct dst_entry *icmp6_dst_gc_list;
1019 static DEFINE_SPINLOCK(icmp6_dst_lock);
1020
1021 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1022                                   struct neighbour *neigh,
1023                                   const struct in6_addr *addr)
1024 {
1025         struct rt6_info *rt;
1026         struct inet6_dev *idev = in6_dev_get(dev);
1027         struct net *net = dev_net(dev);
1028
1029         if (unlikely(idev == NULL))
1030                 return NULL;
1031
1032         rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1033         if (unlikely(rt == NULL)) {
1034                 in6_dev_put(idev);
1035                 goto out;
1036         }
1037
1038         dev_hold(dev);
1039         if (neigh)
1040                 neigh_hold(neigh);
1041         else {
1042                 neigh = ndisc_get_neigh(dev, addr);
1043                 if (IS_ERR(neigh))
1044                         neigh = NULL;
1045         }
1046
1047         rt->rt6i_dev      = dev;
1048         rt->rt6i_idev     = idev;
1049         rt->rt6i_nexthop  = neigh;
1050         atomic_set(&rt->dst.__refcnt, 1);
1051         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 255);
1052         rt->dst.output  = ip6_output;
1053
1054 #if 0   /* there's no chance to use these for ndisc */
1055         rt->dst.flags   = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
1056                                 ? DST_HOST
1057                                 : 0;
1058         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1059         rt->rt6i_dst.plen = 128;
1060 #endif
1061
1062         spin_lock_bh(&icmp6_dst_lock);
1063         rt->dst.next = icmp6_dst_gc_list;
1064         icmp6_dst_gc_list = &rt->dst;
1065         spin_unlock_bh(&icmp6_dst_lock);
1066
1067         fib6_force_start_gc(net);
1068
1069 out:
1070         return &rt->dst;
1071 }
1072
1073 int icmp6_dst_gc(void)
1074 {
1075         struct dst_entry *dst, **pprev;
1076         int more = 0;
1077
1078         spin_lock_bh(&icmp6_dst_lock);
1079         pprev = &icmp6_dst_gc_list;
1080
1081         while ((dst = *pprev) != NULL) {
1082                 if (!atomic_read(&dst->__refcnt)) {
1083                         *pprev = dst->next;
1084                         dst_free(dst);
1085                 } else {
1086                         pprev = &dst->next;
1087                         ++more;
1088                 }
1089         }
1090
1091         spin_unlock_bh(&icmp6_dst_lock);
1092
1093         return more;
1094 }
1095
1096 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1097                             void *arg)
1098 {
1099         struct dst_entry *dst, **pprev;
1100
1101         spin_lock_bh(&icmp6_dst_lock);
1102         pprev = &icmp6_dst_gc_list;
1103         while ((dst = *pprev) != NULL) {
1104                 struct rt6_info *rt = (struct rt6_info *) dst;
1105                 if (func(rt, arg)) {
1106                         *pprev = dst->next;
1107                         dst_free(dst);
1108                 } else {
1109                         pprev = &dst->next;
1110                 }
1111         }
1112         spin_unlock_bh(&icmp6_dst_lock);
1113 }
1114
1115 static int ip6_dst_gc(struct dst_ops *ops)
1116 {
1117         unsigned long now = jiffies;
1118         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1119         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1120         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1121         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1122         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1123         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1124         int entries;
1125
1126         entries = dst_entries_get_fast(ops);
1127         if (time_after(rt_last_gc + rt_min_interval, now) &&
1128             entries <= rt_max_size)
1129                 goto out;
1130
1131         net->ipv6.ip6_rt_gc_expire++;
1132         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1133         net->ipv6.ip6_rt_last_gc = now;
1134         entries = dst_entries_get_slow(ops);
1135         if (entries < ops->gc_thresh)
1136                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1137 out:
1138         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1139         return entries > rt_max_size;
1140 }
1141
1142 /* Clean host part of a prefix. Not necessary in radix tree,
1143    but results in cleaner routing tables.
1144
1145    Remove it only when all the things will work!
1146  */
1147
1148 int ip6_dst_hoplimit(struct dst_entry *dst)
1149 {
1150         int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1151         if (hoplimit == 0) {
1152                 struct net_device *dev = dst->dev;
1153                 struct inet6_dev *idev;
1154
1155                 rcu_read_lock();
1156                 idev = __in6_dev_get(dev);
1157                 if (idev)
1158                         hoplimit = idev->cnf.hop_limit;
1159                 else
1160                         hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1161                 rcu_read_unlock();
1162         }
1163         return hoplimit;
1164 }
1165 EXPORT_SYMBOL(ip6_dst_hoplimit);
1166
1167 /*
1168  *
1169  */
1170
1171 int ip6_route_add(struct fib6_config *cfg)
1172 {
1173         int err;
1174         struct net *net = cfg->fc_nlinfo.nl_net;
1175         struct rt6_info *rt = NULL;
1176         struct net_device *dev = NULL;
1177         struct inet6_dev *idev = NULL;
1178         struct fib6_table *table;
1179         int addr_type;
1180
1181         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1182                 return -EINVAL;
1183 #ifndef CONFIG_IPV6_SUBTREES
1184         if (cfg->fc_src_len)
1185                 return -EINVAL;
1186 #endif
1187         if (cfg->fc_ifindex) {
1188                 err = -ENODEV;
1189                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1190                 if (!dev)
1191                         goto out;
1192                 idev = in6_dev_get(dev);
1193                 if (!idev)
1194                         goto out;
1195         }
1196
1197         if (cfg->fc_metric == 0)
1198                 cfg->fc_metric = IP6_RT_PRIO_USER;
1199
1200         table = fib6_new_table(net, cfg->fc_table);
1201         if (table == NULL) {
1202                 err = -ENOBUFS;
1203                 goto out;
1204         }
1205
1206         rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1207
1208         if (rt == NULL) {
1209                 err = -ENOMEM;
1210                 goto out;
1211         }
1212
1213         rt->dst.obsolete = -1;
1214         rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1215                                 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1216                                 0;
1217
1218         if (cfg->fc_protocol == RTPROT_UNSPEC)
1219                 cfg->fc_protocol = RTPROT_BOOT;
1220         rt->rt6i_protocol = cfg->fc_protocol;
1221
1222         addr_type = ipv6_addr_type(&cfg->fc_dst);
1223
1224         if (addr_type & IPV6_ADDR_MULTICAST)
1225                 rt->dst.input = ip6_mc_input;
1226         else if (cfg->fc_flags & RTF_LOCAL)
1227                 rt->dst.input = ip6_input;
1228         else
1229                 rt->dst.input = ip6_forward;
1230
1231         rt->dst.output = ip6_output;
1232
1233         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1234         rt->rt6i_dst.plen = cfg->fc_dst_len;
1235         if (rt->rt6i_dst.plen == 128)
1236                rt->dst.flags = DST_HOST;
1237
1238 #ifdef CONFIG_IPV6_SUBTREES
1239         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1240         rt->rt6i_src.plen = cfg->fc_src_len;
1241 #endif
1242
1243         rt->rt6i_metric = cfg->fc_metric;
1244
1245         /* We cannot add true routes via loopback here,
1246            they would result in kernel looping; promote them to reject routes
1247          */
1248         if ((cfg->fc_flags & RTF_REJECT) ||
1249             (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK)
1250                                               && !(cfg->fc_flags&RTF_LOCAL))) {
1251                 /* hold loopback dev/idev if we haven't done so. */
1252                 if (dev != net->loopback_dev) {
1253                         if (dev) {
1254                                 dev_put(dev);
1255                                 in6_dev_put(idev);
1256                         }
1257                         dev = net->loopback_dev;
1258                         dev_hold(dev);
1259                         idev = in6_dev_get(dev);
1260                         if (!idev) {
1261                                 err = -ENODEV;
1262                                 goto out;
1263                         }
1264                 }
1265                 rt->dst.output = ip6_pkt_discard_out;
1266                 rt->dst.input = ip6_pkt_discard;
1267                 rt->dst.error = -ENETUNREACH;
1268                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1269                 goto install_route;
1270         }
1271
1272         if (cfg->fc_flags & RTF_GATEWAY) {
1273                 struct in6_addr *gw_addr;
1274                 int gwa_type;
1275
1276                 gw_addr = &cfg->fc_gateway;
1277                 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1278                 gwa_type = ipv6_addr_type(gw_addr);
1279
1280                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1281                         struct rt6_info *grt;
1282
1283                         /* IPv6 strictly inhibits using not link-local
1284                            addresses as nexthop address.
1285                            Otherwise, router will not able to send redirects.
1286                            It is very good, but in some (rare!) circumstances
1287                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1288                            some exceptions. --ANK
1289                          */
1290                         err = -EINVAL;
1291                         if (!(gwa_type&IPV6_ADDR_UNICAST))
1292                                 goto out;
1293
1294                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1295
1296                         err = -EHOSTUNREACH;
1297                         if (grt == NULL)
1298                                 goto out;
1299                         if (dev) {
1300                                 if (dev != grt->rt6i_dev) {
1301                                         dst_release(&grt->dst);
1302                                         goto out;
1303                                 }
1304                         } else {
1305                                 dev = grt->rt6i_dev;
1306                                 idev = grt->rt6i_idev;
1307                                 dev_hold(dev);
1308                                 in6_dev_hold(grt->rt6i_idev);
1309                         }
1310                         if (!(grt->rt6i_flags&RTF_GATEWAY))
1311                                 err = 0;
1312                         dst_release(&grt->dst);
1313
1314                         if (err)
1315                                 goto out;
1316                 }
1317                 err = -EINVAL;
1318                 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1319                         goto out;
1320         }
1321
1322         err = -ENODEV;
1323         if (dev == NULL)
1324                 goto out;
1325
1326         if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1327                 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1328                 if (IS_ERR(rt->rt6i_nexthop)) {
1329                         err = PTR_ERR(rt->rt6i_nexthop);
1330                         rt->rt6i_nexthop = NULL;
1331                         goto out;
1332                 }
1333         }
1334
1335         rt->rt6i_flags = cfg->fc_flags;
1336
1337 install_route:
1338         if (cfg->fc_mx) {
1339                 struct nlattr *nla;
1340                 int remaining;
1341
1342                 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1343                         int type = nla_type(nla);
1344
1345                         if (type) {
1346                                 if (type > RTAX_MAX) {
1347                                         err = -EINVAL;
1348                                         goto out;
1349                                 }
1350
1351                                 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1352                         }
1353                 }
1354         }
1355
1356         rt->dst.dev = dev;
1357         rt->rt6i_idev = idev;
1358         rt->rt6i_table = table;
1359
1360         cfg->fc_nlinfo.nl_net = dev_net(dev);
1361
1362         return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1363
1364 out:
1365         if (dev)
1366                 dev_put(dev);
1367         if (idev)
1368                 in6_dev_put(idev);
1369         if (rt)
1370                 dst_free(&rt->dst);
1371         return err;
1372 }
1373
1374 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1375 {
1376         int err;
1377         struct fib6_table *table;
1378         struct net *net = dev_net(rt->rt6i_dev);
1379
1380         if (rt == net->ipv6.ip6_null_entry)
1381                 return -ENOENT;
1382
1383         table = rt->rt6i_table;
1384         write_lock_bh(&table->tb6_lock);
1385
1386         err = fib6_del(rt, info);
1387         dst_release(&rt->dst);
1388
1389         write_unlock_bh(&table->tb6_lock);
1390
1391         return err;
1392 }
1393
1394 int ip6_del_rt(struct rt6_info *rt)
1395 {
1396         struct nl_info info = {
1397                 .nl_net = dev_net(rt->rt6i_dev),
1398         };
1399         return __ip6_del_rt(rt, &info);
1400 }
1401
1402 static int ip6_route_del(struct fib6_config *cfg)
1403 {
1404         struct fib6_table *table;
1405         struct fib6_node *fn;
1406         struct rt6_info *rt;
1407         int err = -ESRCH;
1408
1409         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1410         if (table == NULL)
1411                 return err;
1412
1413         read_lock_bh(&table->tb6_lock);
1414
1415         fn = fib6_locate(&table->tb6_root,
1416                          &cfg->fc_dst, cfg->fc_dst_len,
1417                          &cfg->fc_src, cfg->fc_src_len);
1418
1419         if (fn) {
1420                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1421                         if (cfg->fc_ifindex &&
1422                             (rt->rt6i_dev == NULL ||
1423                              rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1424                                 continue;
1425                         if (cfg->fc_flags & RTF_GATEWAY &&
1426                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1427                                 continue;
1428                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1429                                 continue;
1430                         dst_hold(&rt->dst);
1431                         read_unlock_bh(&table->tb6_lock);
1432
1433                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1434                 }
1435         }
1436         read_unlock_bh(&table->tb6_lock);
1437
1438         return err;
1439 }
1440
1441 /*
1442  *      Handle redirects
1443  */
1444 struct ip6rd_flowi {
1445         struct flowi fl;
1446         struct in6_addr gateway;
1447 };
1448
1449 static struct rt6_info *__ip6_route_redirect(struct net *net,
1450                                              struct fib6_table *table,
1451                                              struct flowi *fl,
1452                                              int flags)
1453 {
1454         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1455         struct rt6_info *rt;
1456         struct fib6_node *fn;
1457
1458         /*
1459          * Get the "current" route for this destination and
1460          * check if the redirect has come from approriate router.
1461          *
1462          * RFC 2461 specifies that redirects should only be
1463          * accepted if they come from the nexthop to the target.
1464          * Due to the way the routes are chosen, this notion
1465          * is a bit fuzzy and one might need to check all possible
1466          * routes.
1467          */
1468
1469         read_lock_bh(&table->tb6_lock);
1470         fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1471 restart:
1472         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1473                 /*
1474                  * Current route is on-link; redirect is always invalid.
1475                  *
1476                  * Seems, previous statement is not true. It could
1477                  * be node, which looks for us as on-link (f.e. proxy ndisc)
1478                  * But then router serving it might decide, that we should
1479                  * know truth 8)8) --ANK (980726).
1480                  */
1481                 if (rt6_check_expired(rt))
1482                         continue;
1483                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1484                         continue;
1485                 if (fl->oif != rt->rt6i_dev->ifindex)
1486                         continue;
1487                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1488                         continue;
1489                 break;
1490         }
1491
1492         if (!rt)
1493                 rt = net->ipv6.ip6_null_entry;
1494         BACKTRACK(net, &fl->fl6_src);
1495 out:
1496         dst_hold(&rt->dst);
1497
1498         read_unlock_bh(&table->tb6_lock);
1499
1500         return rt;
1501 };
1502
1503 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1504                                            struct in6_addr *src,
1505                                            struct in6_addr *gateway,
1506                                            struct net_device *dev)
1507 {
1508         int flags = RT6_LOOKUP_F_HAS_SADDR;
1509         struct net *net = dev_net(dev);
1510         struct ip6rd_flowi rdfl = {
1511                 .fl = {
1512                         .oif = dev->ifindex,
1513                         .fl6_dst = *dest,
1514                         .fl6_src = *src,
1515                 },
1516         };
1517
1518         ipv6_addr_copy(&rdfl.gateway, gateway);
1519
1520         if (rt6_need_strict(dest))
1521                 flags |= RT6_LOOKUP_F_IFACE;
1522
1523         return (struct rt6_info *)fib6_rule_lookup(net, (struct flowi *)&rdfl,
1524                                                    flags, __ip6_route_redirect);
1525 }
1526
1527 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1528                   struct in6_addr *saddr,
1529                   struct neighbour *neigh, u8 *lladdr, int on_link)
1530 {
1531         struct rt6_info *rt, *nrt = NULL;
1532         struct netevent_redirect netevent;
1533         struct net *net = dev_net(neigh->dev);
1534
1535         rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1536
1537         if (rt == net->ipv6.ip6_null_entry) {
1538                 if (net_ratelimit())
1539                         printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1540                                "for redirect target\n");
1541                 goto out;
1542         }
1543
1544         /*
1545          *      We have finally decided to accept it.
1546          */
1547
1548         neigh_update(neigh, lladdr, NUD_STALE,
1549                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
1550                      NEIGH_UPDATE_F_OVERRIDE|
1551                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1552                                      NEIGH_UPDATE_F_ISROUTER))
1553                      );
1554
1555         /*
1556          * Redirect received -> path was valid.
1557          * Look, redirects are sent only in response to data packets,
1558          * so that this nexthop apparently is reachable. --ANK
1559          */
1560         dst_confirm(&rt->dst);
1561
1562         /* Duplicate redirect: silently ignore. */
1563         if (neigh == rt->dst.neighbour)
1564                 goto out;
1565
1566         nrt = ip6_rt_copy(rt);
1567         if (nrt == NULL)
1568                 goto out;
1569
1570         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1571         if (on_link)
1572                 nrt->rt6i_flags &= ~RTF_GATEWAY;
1573
1574         ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1575         nrt->rt6i_dst.plen = 128;
1576         nrt->dst.flags |= DST_HOST;
1577
1578         ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1579         nrt->rt6i_nexthop = neigh_clone(neigh);
1580
1581         if (ip6_ins_rt(nrt))
1582                 goto out;
1583
1584         netevent.old = &rt->dst;
1585         netevent.new = &nrt->dst;
1586         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1587
1588         if (rt->rt6i_flags&RTF_CACHE) {
1589                 ip6_del_rt(rt);
1590                 return;
1591         }
1592
1593 out:
1594         dst_release(&rt->dst);
1595 }
1596
1597 /*
1598  *      Handle ICMP "packet too big" messages
1599  *      i.e. Path MTU discovery
1600  */
1601
1602 static void rt6_do_pmtu_disc(struct in6_addr *daddr, struct in6_addr *saddr,
1603                              struct net *net, u32 pmtu, int ifindex)
1604 {
1605         struct rt6_info *rt, *nrt;
1606         int allfrag = 0;
1607 again:
1608         rt = rt6_lookup(net, daddr, saddr, ifindex, 0);
1609         if (rt == NULL)
1610                 return;
1611
1612         if (rt6_check_expired(rt)) {
1613                 ip6_del_rt(rt);
1614                 goto again;
1615         }
1616
1617         if (pmtu >= dst_mtu(&rt->dst))
1618                 goto out;
1619
1620         if (pmtu < IPV6_MIN_MTU) {
1621                 /*
1622                  * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1623                  * MTU (1280) and a fragment header should always be included
1624                  * after a node receiving Too Big message reporting PMTU is
1625                  * less than the IPv6 Minimum Link MTU.
1626                  */
1627                 pmtu = IPV6_MIN_MTU;
1628                 allfrag = 1;
1629         }
1630
1631         /* New mtu received -> path was valid.
1632            They are sent only in response to data packets,
1633            so that this nexthop apparently is reachable. --ANK
1634          */
1635         dst_confirm(&rt->dst);
1636
1637         /* Host route. If it is static, it would be better
1638            not to override it, but add new one, so that
1639            when cache entry will expire old pmtu
1640            would return automatically.
1641          */
1642         if (rt->rt6i_flags & RTF_CACHE) {
1643                 dst_metric_set(&rt->dst, RTAX_MTU, pmtu);
1644                 if (allfrag) {
1645                         u32 features = dst_metric(&rt->dst, RTAX_FEATURES);
1646                         features |= RTAX_FEATURE_ALLFRAG;
1647                         dst_metric_set(&rt->dst, RTAX_FEATURES, features);
1648                 }
1649                 dst_set_expires(&rt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1650                 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1651                 goto out;
1652         }
1653
1654         /* Network route.
1655            Two cases are possible:
1656            1. It is connected route. Action: COW
1657            2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1658          */
1659         if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1660                 nrt = rt6_alloc_cow(rt, daddr, saddr);
1661         else
1662                 nrt = rt6_alloc_clone(rt, daddr);
1663
1664         if (nrt) {
1665                 dst_metric_set(&nrt->dst, RTAX_MTU, pmtu);
1666                 if (allfrag) {
1667                         u32 features = dst_metric(&nrt->dst, RTAX_FEATURES);
1668                         features |= RTAX_FEATURE_ALLFRAG;
1669                         dst_metric_set(&nrt->dst, RTAX_FEATURES, features);
1670                 }
1671
1672                 /* According to RFC 1981, detecting PMTU increase shouldn't be
1673                  * happened within 5 mins, the recommended timer is 10 mins.
1674                  * Here this route expiration time is set to ip6_rt_mtu_expires
1675                  * which is 10 mins. After 10 mins the decreased pmtu is expired
1676                  * and detecting PMTU increase will be automatically happened.
1677                  */
1678                 dst_set_expires(&nrt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1679                 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1680
1681                 ip6_ins_rt(nrt);
1682         }
1683 out:
1684         dst_release(&rt->dst);
1685 }
1686
1687 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1688                         struct net_device *dev, u32 pmtu)
1689 {
1690         struct net *net = dev_net(dev);
1691
1692         /*
1693          * RFC 1981 states that a node "MUST reduce the size of the packets it
1694          * is sending along the path" that caused the Packet Too Big message.
1695          * Since it's not possible in the general case to determine which
1696          * interface was used to send the original packet, we update the MTU
1697          * on the interface that will be used to send future packets. We also
1698          * update the MTU on the interface that received the Packet Too Big in
1699          * case the original packet was forced out that interface with
1700          * SO_BINDTODEVICE or similar. This is the next best thing to the
1701          * correct behaviour, which would be to update the MTU on all
1702          * interfaces.
1703          */
1704         rt6_do_pmtu_disc(daddr, saddr, net, pmtu, 0);
1705         rt6_do_pmtu_disc(daddr, saddr, net, pmtu, dev->ifindex);
1706 }
1707
1708 /*
1709  *      Misc support functions
1710  */
1711
1712 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1713 {
1714         struct net *net = dev_net(ort->rt6i_dev);
1715         struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1716
1717         if (rt) {
1718                 rt->dst.input = ort->dst.input;
1719                 rt->dst.output = ort->dst.output;
1720
1721                 dst_copy_metrics(&rt->dst, &ort->dst);
1722                 rt->dst.error = ort->dst.error;
1723                 rt->dst.dev = ort->dst.dev;
1724                 if (rt->dst.dev)
1725                         dev_hold(rt->dst.dev);
1726                 rt->rt6i_idev = ort->rt6i_idev;
1727                 if (rt->rt6i_idev)
1728                         in6_dev_hold(rt->rt6i_idev);
1729                 rt->dst.lastuse = jiffies;
1730                 rt->rt6i_expires = 0;
1731
1732                 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1733                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1734                 rt->rt6i_metric = 0;
1735
1736                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1737 #ifdef CONFIG_IPV6_SUBTREES
1738                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1739 #endif
1740                 rt->rt6i_table = ort->rt6i_table;
1741         }
1742         return rt;
1743 }
1744
1745 #ifdef CONFIG_IPV6_ROUTE_INFO
1746 static struct rt6_info *rt6_get_route_info(struct net *net,
1747                                            struct in6_addr *prefix, int prefixlen,
1748                                            struct in6_addr *gwaddr, int ifindex)
1749 {
1750         struct fib6_node *fn;
1751         struct rt6_info *rt = NULL;
1752         struct fib6_table *table;
1753
1754         table = fib6_get_table(net, RT6_TABLE_INFO);
1755         if (table == NULL)
1756                 return NULL;
1757
1758         write_lock_bh(&table->tb6_lock);
1759         fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1760         if (!fn)
1761                 goto out;
1762
1763         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1764                 if (rt->rt6i_dev->ifindex != ifindex)
1765                         continue;
1766                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1767                         continue;
1768                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1769                         continue;
1770                 dst_hold(&rt->dst);
1771                 break;
1772         }
1773 out:
1774         write_unlock_bh(&table->tb6_lock);
1775         return rt;
1776 }
1777
1778 static struct rt6_info *rt6_add_route_info(struct net *net,
1779                                            struct in6_addr *prefix, int prefixlen,
1780                                            struct in6_addr *gwaddr, int ifindex,
1781                                            unsigned pref)
1782 {
1783         struct fib6_config cfg = {
1784                 .fc_table       = RT6_TABLE_INFO,
1785                 .fc_metric      = IP6_RT_PRIO_USER,
1786                 .fc_ifindex     = ifindex,
1787                 .fc_dst_len     = prefixlen,
1788                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1789                                   RTF_UP | RTF_PREF(pref),
1790                 .fc_nlinfo.pid = 0,
1791                 .fc_nlinfo.nlh = NULL,
1792                 .fc_nlinfo.nl_net = net,
1793         };
1794
1795         ipv6_addr_copy(&cfg.fc_dst, prefix);
1796         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1797
1798         /* We should treat it as a default route if prefix length is 0. */
1799         if (!prefixlen)
1800                 cfg.fc_flags |= RTF_DEFAULT;
1801
1802         ip6_route_add(&cfg);
1803
1804         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1805 }
1806 #endif
1807
1808 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1809 {
1810         struct rt6_info *rt;
1811         struct fib6_table *table;
1812
1813         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1814         if (table == NULL)
1815                 return NULL;
1816
1817         write_lock_bh(&table->tb6_lock);
1818         for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1819                 if (dev == rt->rt6i_dev &&
1820                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1821                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
1822                         break;
1823         }
1824         if (rt)
1825                 dst_hold(&rt->dst);
1826         write_unlock_bh(&table->tb6_lock);
1827         return rt;
1828 }
1829
1830 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1831                                      struct net_device *dev,
1832                                      unsigned int pref)
1833 {
1834         struct fib6_config cfg = {
1835                 .fc_table       = RT6_TABLE_DFLT,
1836                 .fc_metric      = IP6_RT_PRIO_USER,
1837                 .fc_ifindex     = dev->ifindex,
1838                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1839                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1840                 .fc_nlinfo.pid = 0,
1841                 .fc_nlinfo.nlh = NULL,
1842                 .fc_nlinfo.nl_net = dev_net(dev),
1843         };
1844
1845         ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1846
1847         ip6_route_add(&cfg);
1848
1849         return rt6_get_dflt_router(gwaddr, dev);
1850 }
1851
1852 void rt6_purge_dflt_routers(struct net *net)
1853 {
1854         struct rt6_info *rt;
1855         struct fib6_table *table;
1856
1857         /* NOTE: Keep consistent with rt6_get_dflt_router */
1858         table = fib6_get_table(net, RT6_TABLE_DFLT);
1859         if (table == NULL)
1860                 return;
1861
1862 restart:
1863         read_lock_bh(&table->tb6_lock);
1864         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1865                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1866                         dst_hold(&rt->dst);
1867                         read_unlock_bh(&table->tb6_lock);
1868                         ip6_del_rt(rt);
1869                         goto restart;
1870                 }
1871         }
1872         read_unlock_bh(&table->tb6_lock);
1873 }
1874
1875 static void rtmsg_to_fib6_config(struct net *net,
1876                                  struct in6_rtmsg *rtmsg,
1877                                  struct fib6_config *cfg)
1878 {
1879         memset(cfg, 0, sizeof(*cfg));
1880
1881         cfg->fc_table = RT6_TABLE_MAIN;
1882         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1883         cfg->fc_metric = rtmsg->rtmsg_metric;
1884         cfg->fc_expires = rtmsg->rtmsg_info;
1885         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1886         cfg->fc_src_len = rtmsg->rtmsg_src_len;
1887         cfg->fc_flags = rtmsg->rtmsg_flags;
1888
1889         cfg->fc_nlinfo.nl_net = net;
1890
1891         ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1892         ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1893         ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1894 }
1895
1896 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1897 {
1898         struct fib6_config cfg;
1899         struct in6_rtmsg rtmsg;
1900         int err;
1901
1902         switch(cmd) {
1903         case SIOCADDRT:         /* Add a route */
1904         case SIOCDELRT:         /* Delete a route */
1905                 if (!capable(CAP_NET_ADMIN))
1906                         return -EPERM;
1907                 err = copy_from_user(&rtmsg, arg,
1908                                      sizeof(struct in6_rtmsg));
1909                 if (err)
1910                         return -EFAULT;
1911
1912                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1913
1914                 rtnl_lock();
1915                 switch (cmd) {
1916                 case SIOCADDRT:
1917                         err = ip6_route_add(&cfg);
1918                         break;
1919                 case SIOCDELRT:
1920                         err = ip6_route_del(&cfg);
1921                         break;
1922                 default:
1923                         err = -EINVAL;
1924                 }
1925                 rtnl_unlock();
1926
1927                 return err;
1928         }
1929
1930         return -EINVAL;
1931 }
1932
1933 /*
1934  *      Drop the packet on the floor
1935  */
1936
1937 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1938 {
1939         int type;
1940         struct dst_entry *dst = skb_dst(skb);
1941         switch (ipstats_mib_noroutes) {
1942         case IPSTATS_MIB_INNOROUTES:
1943                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1944                 if (type == IPV6_ADDR_ANY) {
1945                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1946                                       IPSTATS_MIB_INADDRERRORS);
1947                         break;
1948                 }
1949                 /* FALLTHROUGH */
1950         case IPSTATS_MIB_OUTNOROUTES:
1951                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1952                               ipstats_mib_noroutes);
1953                 break;
1954         }
1955         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1956         kfree_skb(skb);
1957         return 0;
1958 }
1959
1960 static int ip6_pkt_discard(struct sk_buff *skb)
1961 {
1962         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1963 }
1964
1965 static int ip6_pkt_discard_out(struct sk_buff *skb)
1966 {
1967         skb->dev = skb_dst(skb)->dev;
1968         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1969 }
1970
1971 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1972
1973 static int ip6_pkt_prohibit(struct sk_buff *skb)
1974 {
1975         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1976 }
1977
1978 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1979 {
1980         skb->dev = skb_dst(skb)->dev;
1981         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1982 }
1983
1984 #endif
1985
1986 /*
1987  *      Allocate a dst for local (unicast / anycast) address.
1988  */
1989
1990 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1991                                     const struct in6_addr *addr,
1992                                     int anycast)
1993 {
1994         struct net *net = dev_net(idev->dev);
1995         struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1996         struct neighbour *neigh;
1997
1998         if (rt == NULL) {
1999                 if (net_ratelimit())
2000                         pr_warning("IPv6:  Maximum number of routes reached,"
2001                                    " consider increasing route/max_size.\n");
2002                 return ERR_PTR(-ENOMEM);
2003         }
2004
2005         dev_hold(net->loopback_dev);
2006         in6_dev_hold(idev);
2007
2008         rt->dst.flags = DST_HOST;
2009         rt->dst.input = ip6_input;
2010         rt->dst.output = ip6_output;
2011         rt->rt6i_dev = net->loopback_dev;
2012         rt->rt6i_idev = idev;
2013         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, -1);
2014         rt->dst.obsolete = -1;
2015
2016         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2017         if (anycast)
2018                 rt->rt6i_flags |= RTF_ANYCAST;
2019         else
2020                 rt->rt6i_flags |= RTF_LOCAL;
2021         neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
2022         if (IS_ERR(neigh)) {
2023                 dst_free(&rt->dst);
2024
2025                 return ERR_CAST(neigh);
2026         }
2027         rt->rt6i_nexthop = neigh;
2028
2029         ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
2030         rt->rt6i_dst.plen = 128;
2031         rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2032
2033         atomic_set(&rt->dst.__refcnt, 1);
2034
2035         return rt;
2036 }
2037
2038 struct arg_dev_net {
2039         struct net_device *dev;
2040         struct net *net;
2041 };
2042
2043 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2044 {
2045         const struct arg_dev_net *adn = arg;
2046         const struct net_device *dev = adn->dev;
2047
2048         if ((rt->rt6i_dev == dev || dev == NULL) &&
2049             rt != adn->net->ipv6.ip6_null_entry) {
2050                 RT6_TRACE("deleted by ifdown %p\n", rt);
2051                 return -1;
2052         }
2053         return 0;
2054 }
2055
2056 void rt6_ifdown(struct net *net, struct net_device *dev)
2057 {
2058         struct arg_dev_net adn = {
2059                 .dev = dev,
2060                 .net = net,
2061         };
2062
2063         fib6_clean_all(net, fib6_ifdown, 0, &adn);
2064         icmp6_clean_all(fib6_ifdown, &adn);
2065 }
2066
2067 struct rt6_mtu_change_arg
2068 {
2069         struct net_device *dev;
2070         unsigned mtu;
2071 };
2072
2073 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2074 {
2075         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2076         struct inet6_dev *idev;
2077
2078         /* In IPv6 pmtu discovery is not optional,
2079            so that RTAX_MTU lock cannot disable it.
2080            We still use this lock to block changes
2081            caused by addrconf/ndisc.
2082         */
2083
2084         idev = __in6_dev_get(arg->dev);
2085         if (idev == NULL)
2086                 return 0;
2087
2088         /* For administrative MTU increase, there is no way to discover
2089            IPv6 PMTU increase, so PMTU increase should be updated here.
2090            Since RFC 1981 doesn't include administrative MTU increase
2091            update PMTU increase is a MUST. (i.e. jumbo frame)
2092          */
2093         /*
2094            If new MTU is less than route PMTU, this new MTU will be the
2095            lowest MTU in the path, update the route PMTU to reflect PMTU
2096            decreases; if new MTU is greater than route PMTU, and the
2097            old MTU is the lowest MTU in the path, update the route PMTU
2098            to reflect the increase. In this case if the other nodes' MTU
2099            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2100            PMTU discouvery.
2101          */
2102         if (rt->rt6i_dev == arg->dev &&
2103             !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2104             (dst_mtu(&rt->dst) >= arg->mtu ||
2105              (dst_mtu(&rt->dst) < arg->mtu &&
2106               dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2107                 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2108         }
2109         return 0;
2110 }
2111
2112 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
2113 {
2114         struct rt6_mtu_change_arg arg = {
2115                 .dev = dev,
2116                 .mtu = mtu,
2117         };
2118
2119         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2120 }
2121
2122 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2123         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2124         [RTA_OIF]               = { .type = NLA_U32 },
2125         [RTA_IIF]               = { .type = NLA_U32 },
2126         [RTA_PRIORITY]          = { .type = NLA_U32 },
2127         [RTA_METRICS]           = { .type = NLA_NESTED },
2128 };
2129
2130 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2131                               struct fib6_config *cfg)
2132 {
2133         struct rtmsg *rtm;
2134         struct nlattr *tb[RTA_MAX+1];
2135         int err;
2136
2137         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2138         if (err < 0)
2139                 goto errout;
2140
2141         err = -EINVAL;
2142         rtm = nlmsg_data(nlh);
2143         memset(cfg, 0, sizeof(*cfg));
2144
2145         cfg->fc_table = rtm->rtm_table;
2146         cfg->fc_dst_len = rtm->rtm_dst_len;
2147         cfg->fc_src_len = rtm->rtm_src_len;
2148         cfg->fc_flags = RTF_UP;
2149         cfg->fc_protocol = rtm->rtm_protocol;
2150
2151         if (rtm->rtm_type == RTN_UNREACHABLE)
2152                 cfg->fc_flags |= RTF_REJECT;
2153
2154         if (rtm->rtm_type == RTN_LOCAL)
2155                 cfg->fc_flags |= RTF_LOCAL;
2156
2157         cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2158         cfg->fc_nlinfo.nlh = nlh;
2159         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2160
2161         if (tb[RTA_GATEWAY]) {
2162                 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2163                 cfg->fc_flags |= RTF_GATEWAY;
2164         }
2165
2166         if (tb[RTA_DST]) {
2167                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2168
2169                 if (nla_len(tb[RTA_DST]) < plen)
2170                         goto errout;
2171
2172                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2173         }
2174
2175         if (tb[RTA_SRC]) {
2176                 int plen = (rtm->rtm_src_len + 7) >> 3;
2177
2178                 if (nla_len(tb[RTA_SRC]) < plen)
2179                         goto errout;
2180
2181                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2182         }
2183
2184         if (tb[RTA_OIF])
2185                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2186
2187         if (tb[RTA_PRIORITY])
2188                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2189
2190         if (tb[RTA_METRICS]) {
2191                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2192                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2193         }
2194
2195         if (tb[RTA_TABLE])
2196                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2197
2198         err = 0;
2199 errout:
2200         return err;
2201 }
2202
2203 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2204 {
2205         struct fib6_config cfg;
2206         int err;
2207
2208         err = rtm_to_fib6_config(skb, nlh, &cfg);
2209         if (err < 0)
2210                 return err;
2211
2212         return ip6_route_del(&cfg);
2213 }
2214
2215 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2216 {
2217         struct fib6_config cfg;
2218         int err;
2219
2220         err = rtm_to_fib6_config(skb, nlh, &cfg);
2221         if (err < 0)
2222                 return err;
2223
2224         return ip6_route_add(&cfg);
2225 }
2226
2227 static inline size_t rt6_nlmsg_size(void)
2228 {
2229         return NLMSG_ALIGN(sizeof(struct rtmsg))
2230                + nla_total_size(16) /* RTA_SRC */
2231                + nla_total_size(16) /* RTA_DST */
2232                + nla_total_size(16) /* RTA_GATEWAY */
2233                + nla_total_size(16) /* RTA_PREFSRC */
2234                + nla_total_size(4) /* RTA_TABLE */
2235                + nla_total_size(4) /* RTA_IIF */
2236                + nla_total_size(4) /* RTA_OIF */
2237                + nla_total_size(4) /* RTA_PRIORITY */
2238                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2239                + nla_total_size(sizeof(struct rta_cacheinfo));
2240 }
2241
2242 static int rt6_fill_node(struct net *net,
2243                          struct sk_buff *skb, struct rt6_info *rt,
2244                          struct in6_addr *dst, struct in6_addr *src,
2245                          int iif, int type, u32 pid, u32 seq,
2246                          int prefix, int nowait, unsigned int flags)
2247 {
2248         struct rtmsg *rtm;
2249         struct nlmsghdr *nlh;
2250         long expires;
2251         u32 table;
2252
2253         if (prefix) {   /* user wants prefix routes only */
2254                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2255                         /* success since this is not a prefix route */
2256                         return 1;
2257                 }
2258         }
2259
2260         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2261         if (nlh == NULL)
2262                 return -EMSGSIZE;
2263
2264         rtm = nlmsg_data(nlh);
2265         rtm->rtm_family = AF_INET6;
2266         rtm->rtm_dst_len = rt->rt6i_dst.plen;
2267         rtm->rtm_src_len = rt->rt6i_src.plen;
2268         rtm->rtm_tos = 0;
2269         if (rt->rt6i_table)
2270                 table = rt->rt6i_table->tb6_id;
2271         else
2272                 table = RT6_TABLE_UNSPEC;
2273         rtm->rtm_table = table;
2274         NLA_PUT_U32(skb, RTA_TABLE, table);
2275         if (rt->rt6i_flags&RTF_REJECT)
2276                 rtm->rtm_type = RTN_UNREACHABLE;
2277         else if (rt->rt6i_flags&RTF_LOCAL)
2278                 rtm->rtm_type = RTN_LOCAL;
2279         else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2280                 rtm->rtm_type = RTN_LOCAL;
2281         else
2282                 rtm->rtm_type = RTN_UNICAST;
2283         rtm->rtm_flags = 0;
2284         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2285         rtm->rtm_protocol = rt->rt6i_protocol;
2286         if (rt->rt6i_flags&RTF_DYNAMIC)
2287                 rtm->rtm_protocol = RTPROT_REDIRECT;
2288         else if (rt->rt6i_flags & RTF_ADDRCONF)
2289                 rtm->rtm_protocol = RTPROT_KERNEL;
2290         else if (rt->rt6i_flags&RTF_DEFAULT)
2291                 rtm->rtm_protocol = RTPROT_RA;
2292
2293         if (rt->rt6i_flags&RTF_CACHE)
2294                 rtm->rtm_flags |= RTM_F_CLONED;
2295
2296         if (dst) {
2297                 NLA_PUT(skb, RTA_DST, 16, dst);
2298                 rtm->rtm_dst_len = 128;
2299         } else if (rtm->rtm_dst_len)
2300                 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2301 #ifdef CONFIG_IPV6_SUBTREES
2302         if (src) {
2303                 NLA_PUT(skb, RTA_SRC, 16, src);
2304                 rtm->rtm_src_len = 128;
2305         } else if (rtm->rtm_src_len)
2306                 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2307 #endif
2308         if (iif) {
2309 #ifdef CONFIG_IPV6_MROUTE
2310                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2311                         int err = ip6mr_get_route(net, skb, rtm, nowait);
2312                         if (err <= 0) {
2313                                 if (!nowait) {
2314                                         if (err == 0)
2315                                                 return 0;
2316                                         goto nla_put_failure;
2317                                 } else {
2318                                         if (err == -EMSGSIZE)
2319                                                 goto nla_put_failure;
2320                                 }
2321                         }
2322                 } else
2323 #endif
2324                         NLA_PUT_U32(skb, RTA_IIF, iif);
2325         } else if (dst) {
2326                 struct inet6_dev *idev = ip6_dst_idev(&rt->dst);
2327                 struct in6_addr saddr_buf;
2328                 if (ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2329                                        dst, 0, &saddr_buf) == 0)
2330                         NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2331         }
2332
2333         if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2334                 goto nla_put_failure;
2335
2336         if (rt->dst.neighbour)
2337                 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->dst.neighbour->primary_key);
2338
2339         if (rt->dst.dev)
2340                 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2341
2342         NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2343
2344         if (!(rt->rt6i_flags & RTF_EXPIRES))
2345                 expires = 0;
2346         else if (rt->rt6i_expires - jiffies < INT_MAX)
2347                 expires = rt->rt6i_expires - jiffies;
2348         else
2349                 expires = INT_MAX;
2350
2351         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, 0, 0,
2352                                expires, rt->dst.error) < 0)
2353                 goto nla_put_failure;
2354
2355         return nlmsg_end(skb, nlh);
2356
2357 nla_put_failure:
2358         nlmsg_cancel(skb, nlh);
2359         return -EMSGSIZE;
2360 }
2361
2362 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2363 {
2364         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2365         int prefix;
2366
2367         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2368                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2369                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2370         } else
2371                 prefix = 0;
2372
2373         return rt6_fill_node(arg->net,
2374                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2375                      NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2376                      prefix, 0, NLM_F_MULTI);
2377 }
2378
2379 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2380 {
2381         struct net *net = sock_net(in_skb->sk);
2382         struct nlattr *tb[RTA_MAX+1];
2383         struct rt6_info *rt;
2384         struct sk_buff *skb;
2385         struct rtmsg *rtm;
2386         struct flowi fl;
2387         int err, iif = 0;
2388
2389         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2390         if (err < 0)
2391                 goto errout;
2392
2393         err = -EINVAL;
2394         memset(&fl, 0, sizeof(fl));
2395
2396         if (tb[RTA_SRC]) {
2397                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2398                         goto errout;
2399
2400                 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2401         }
2402
2403         if (tb[RTA_DST]) {
2404                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2405                         goto errout;
2406
2407                 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2408         }
2409
2410         if (tb[RTA_IIF])
2411                 iif = nla_get_u32(tb[RTA_IIF]);
2412
2413         if (tb[RTA_OIF])
2414                 fl.oif = nla_get_u32(tb[RTA_OIF]);
2415
2416         if (iif) {
2417                 struct net_device *dev;
2418                 dev = __dev_get_by_index(net, iif);
2419                 if (!dev) {
2420                         err = -ENODEV;
2421                         goto errout;
2422                 }
2423         }
2424
2425         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2426         if (skb == NULL) {
2427                 err = -ENOBUFS;
2428                 goto errout;
2429         }
2430
2431         /* Reserve room for dummy headers, this skb can pass
2432            through good chunk of routing engine.
2433          */
2434         skb_reset_mac_header(skb);
2435         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2436
2437         rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl);
2438         skb_dst_set(skb, &rt->dst);
2439
2440         err = rt6_fill_node(net, skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2441                             RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2442                             nlh->nlmsg_seq, 0, 0, 0);
2443         if (err < 0) {
2444                 kfree_skb(skb);
2445                 goto errout;
2446         }
2447
2448         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2449 errout:
2450         return err;
2451 }
2452
2453 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2454 {
2455         struct sk_buff *skb;
2456         struct net *net = info->nl_net;
2457         u32 seq;
2458         int err;
2459
2460         err = -ENOBUFS;
2461         seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2462
2463         skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2464         if (skb == NULL)
2465                 goto errout;
2466
2467         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2468                                 event, info->pid, seq, 0, 0, 0);
2469         if (err < 0) {
2470                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2471                 WARN_ON(err == -EMSGSIZE);
2472                 kfree_skb(skb);
2473                 goto errout;
2474         }
2475         rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2476                     info->nlh, gfp_any());
2477         return;
2478 errout:
2479         if (err < 0)
2480                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2481 }
2482
2483 static int ip6_route_dev_notify(struct notifier_block *this,
2484                                 unsigned long event, void *data)
2485 {
2486         struct net_device *dev = (struct net_device *)data;
2487         struct net *net = dev_net(dev);
2488
2489         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2490                 net->ipv6.ip6_null_entry->dst.dev = dev;
2491                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2492 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2493                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2494                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2495                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2496                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2497 #endif
2498         }
2499
2500         return NOTIFY_OK;
2501 }
2502
2503 /*
2504  *      /proc
2505  */
2506
2507 #ifdef CONFIG_PROC_FS
2508
2509 struct rt6_proc_arg
2510 {
2511         char *buffer;
2512         int offset;
2513         int length;
2514         int skip;
2515         int len;
2516 };
2517
2518 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2519 {
2520         struct seq_file *m = p_arg;
2521
2522         seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2523
2524 #ifdef CONFIG_IPV6_SUBTREES
2525         seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2526 #else
2527         seq_puts(m, "00000000000000000000000000000000 00 ");
2528 #endif
2529
2530         if (rt->rt6i_nexthop) {
2531                 seq_printf(m, "%pi6", rt->rt6i_nexthop->primary_key);
2532         } else {
2533                 seq_puts(m, "00000000000000000000000000000000");
2534         }
2535         seq_printf(m, " %08x %08x %08x %08x %8s\n",
2536                    rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2537                    rt->dst.__use, rt->rt6i_flags,
2538                    rt->rt6i_dev ? rt->rt6i_dev->name : "");
2539         return 0;
2540 }
2541
2542 static int ipv6_route_show(struct seq_file *m, void *v)
2543 {
2544         struct net *net = (struct net *)m->private;
2545         fib6_clean_all(net, rt6_info_route, 0, m);
2546         return 0;
2547 }
2548
2549 static int ipv6_route_open(struct inode *inode, struct file *file)
2550 {
2551         return single_open_net(inode, file, ipv6_route_show);
2552 }
2553
2554 static const struct file_operations ipv6_route_proc_fops = {
2555         .owner          = THIS_MODULE,
2556         .open           = ipv6_route_open,
2557         .read           = seq_read,
2558         .llseek         = seq_lseek,
2559         .release        = single_release_net,
2560 };
2561
2562 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2563 {
2564         struct net *net = (struct net *)seq->private;
2565         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2566                    net->ipv6.rt6_stats->fib_nodes,
2567                    net->ipv6.rt6_stats->fib_route_nodes,
2568                    net->ipv6.rt6_stats->fib_rt_alloc,
2569                    net->ipv6.rt6_stats->fib_rt_entries,
2570                    net->ipv6.rt6_stats->fib_rt_cache,
2571                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2572                    net->ipv6.rt6_stats->fib_discarded_routes);
2573
2574         return 0;
2575 }
2576
2577 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2578 {
2579         return single_open_net(inode, file, rt6_stats_seq_show);
2580 }
2581
2582 static const struct file_operations rt6_stats_seq_fops = {
2583         .owner   = THIS_MODULE,
2584         .open    = rt6_stats_seq_open,
2585         .read    = seq_read,
2586         .llseek  = seq_lseek,
2587         .release = single_release_net,
2588 };
2589 #endif  /* CONFIG_PROC_FS */
2590
2591 #ifdef CONFIG_SYSCTL
2592
2593 static
2594 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2595                               void __user *buffer, size_t *lenp, loff_t *ppos)
2596 {
2597         struct net *net = current->nsproxy->net_ns;
2598         int delay = net->ipv6.sysctl.flush_delay;
2599         if (write) {
2600                 proc_dointvec(ctl, write, buffer, lenp, ppos);
2601                 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2602                 return 0;
2603         } else
2604                 return -EINVAL;
2605 }
2606
2607 ctl_table ipv6_route_table_template[] = {
2608         {
2609                 .procname       =       "flush",
2610                 .data           =       &init_net.ipv6.sysctl.flush_delay,
2611                 .maxlen         =       sizeof(int),
2612                 .mode           =       0200,
2613                 .proc_handler   =       ipv6_sysctl_rtcache_flush
2614         },
2615         {
2616                 .procname       =       "gc_thresh",
2617                 .data           =       &ip6_dst_ops_template.gc_thresh,
2618                 .maxlen         =       sizeof(int),
2619                 .mode           =       0644,
2620                 .proc_handler   =       proc_dointvec,
2621         },
2622         {
2623                 .procname       =       "max_size",
2624                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
2625                 .maxlen         =       sizeof(int),
2626                 .mode           =       0644,
2627                 .proc_handler   =       proc_dointvec,
2628         },
2629         {
2630                 .procname       =       "gc_min_interval",
2631                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2632                 .maxlen         =       sizeof(int),
2633                 .mode           =       0644,
2634                 .proc_handler   =       proc_dointvec_jiffies,
2635         },
2636         {
2637                 .procname       =       "gc_timeout",
2638                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2639                 .maxlen         =       sizeof(int),
2640                 .mode           =       0644,
2641                 .proc_handler   =       proc_dointvec_jiffies,
2642         },
2643         {
2644                 .procname       =       "gc_interval",
2645                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2646                 .maxlen         =       sizeof(int),
2647                 .mode           =       0644,
2648                 .proc_handler   =       proc_dointvec_jiffies,
2649         },
2650         {
2651                 .procname       =       "gc_elasticity",
2652                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2653                 .maxlen         =       sizeof(int),
2654                 .mode           =       0644,
2655                 .proc_handler   =       proc_dointvec,
2656         },
2657         {
2658                 .procname       =       "mtu_expires",
2659                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2660                 .maxlen         =       sizeof(int),
2661                 .mode           =       0644,
2662                 .proc_handler   =       proc_dointvec_jiffies,
2663         },
2664         {
2665                 .procname       =       "min_adv_mss",
2666                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2667                 .maxlen         =       sizeof(int),
2668                 .mode           =       0644,
2669                 .proc_handler   =       proc_dointvec,
2670         },
2671         {
2672                 .procname       =       "gc_min_interval_ms",
2673                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2674                 .maxlen         =       sizeof(int),
2675                 .mode           =       0644,
2676                 .proc_handler   =       proc_dointvec_ms_jiffies,
2677         },
2678         { }
2679 };
2680
2681 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2682 {
2683         struct ctl_table *table;
2684
2685         table = kmemdup(ipv6_route_table_template,
2686                         sizeof(ipv6_route_table_template),
2687                         GFP_KERNEL);
2688
2689         if (table) {
2690                 table[0].data = &net->ipv6.sysctl.flush_delay;
2691                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2692                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2693                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2694                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2695                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2696                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2697                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2698                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2699                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2700         }
2701
2702         return table;
2703 }
2704 #endif
2705
2706 static int __net_init ip6_route_net_init(struct net *net)
2707 {
2708         int ret = -ENOMEM;
2709
2710         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2711                sizeof(net->ipv6.ip6_dst_ops));
2712
2713         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2714                 goto out_ip6_dst_ops;
2715
2716         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2717                                            sizeof(*net->ipv6.ip6_null_entry),
2718                                            GFP_KERNEL);
2719         if (!net->ipv6.ip6_null_entry)
2720                 goto out_ip6_dst_entries;
2721         net->ipv6.ip6_null_entry->dst.path =
2722                 (struct dst_entry *)net->ipv6.ip6_null_entry;
2723         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2724         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2725                          ip6_template_metrics, true);
2726
2727 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2728         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2729                                                sizeof(*net->ipv6.ip6_prohibit_entry),
2730                                                GFP_KERNEL);
2731         if (!net->ipv6.ip6_prohibit_entry)
2732                 goto out_ip6_null_entry;
2733         net->ipv6.ip6_prohibit_entry->dst.path =
2734                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2735         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2736         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2737                          ip6_template_metrics, true);
2738
2739         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2740                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
2741                                                GFP_KERNEL);
2742         if (!net->ipv6.ip6_blk_hole_entry)
2743                 goto out_ip6_prohibit_entry;
2744         net->ipv6.ip6_blk_hole_entry->dst.path =
2745                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2746         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2747         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2748                          ip6_template_metrics, true);
2749 #endif
2750
2751         net->ipv6.sysctl.flush_delay = 0;
2752         net->ipv6.sysctl.ip6_rt_max_size = 4096;
2753         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2754         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2755         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2756         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2757         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2758         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2759
2760 #ifdef CONFIG_PROC_FS
2761         proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2762         proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2763 #endif
2764         net->ipv6.ip6_rt_gc_expire = 30*HZ;
2765
2766         ret = 0;
2767 out:
2768         return ret;
2769
2770 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2771 out_ip6_prohibit_entry:
2772         kfree(net->ipv6.ip6_prohibit_entry);
2773 out_ip6_null_entry:
2774         kfree(net->ipv6.ip6_null_entry);
2775 #endif
2776 out_ip6_dst_entries:
2777         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2778 out_ip6_dst_ops:
2779         goto out;
2780 }
2781
2782 static void __net_exit ip6_route_net_exit(struct net *net)
2783 {
2784 #ifdef CONFIG_PROC_FS
2785         proc_net_remove(net, "ipv6_route");
2786         proc_net_remove(net, "rt6_stats");
2787 #endif
2788         kfree(net->ipv6.ip6_null_entry);
2789 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2790         kfree(net->ipv6.ip6_prohibit_entry);
2791         kfree(net->ipv6.ip6_blk_hole_entry);
2792 #endif
2793         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2794 }
2795
2796 static struct pernet_operations ip6_route_net_ops = {
2797         .init = ip6_route_net_init,
2798         .exit = ip6_route_net_exit,
2799 };
2800
2801 static struct notifier_block ip6_route_dev_notifier = {
2802         .notifier_call = ip6_route_dev_notify,
2803         .priority = 0,
2804 };
2805
2806 int __init ip6_route_init(void)
2807 {
2808         int ret;
2809
2810         ret = -ENOMEM;
2811         ip6_dst_ops_template.kmem_cachep =
2812                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2813                                   SLAB_HWCACHE_ALIGN, NULL);
2814         if (!ip6_dst_ops_template.kmem_cachep)
2815                 goto out;
2816
2817         ret = dst_entries_init(&ip6_dst_blackhole_ops);
2818         if (ret)
2819                 goto out_kmem_cache;
2820
2821         ret = register_pernet_subsys(&ip6_route_net_ops);
2822         if (ret)
2823                 goto out_dst_entries;
2824
2825         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
2826
2827         /* Registering of the loopback is done before this portion of code,
2828          * the loopback reference in rt6_info will not be taken, do it
2829          * manually for init_net */
2830         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
2831         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2832   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2833         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
2834         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2835         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
2836         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2837   #endif
2838         ret = fib6_init();
2839         if (ret)
2840                 goto out_register_subsys;
2841
2842         ret = xfrm6_init();
2843         if (ret)
2844                 goto out_fib6_init;
2845
2846         ret = fib6_rules_init();
2847         if (ret)
2848                 goto xfrm6_init;
2849
2850         ret = -ENOBUFS;
2851         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2852             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2853             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2854                 goto fib6_rules_init;
2855
2856         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2857         if (ret)
2858                 goto fib6_rules_init;
2859
2860 out:
2861         return ret;
2862
2863 fib6_rules_init:
2864         fib6_rules_cleanup();
2865 xfrm6_init:
2866         xfrm6_fini();
2867 out_fib6_init:
2868         fib6_gc_cleanup();
2869 out_register_subsys:
2870         unregister_pernet_subsys(&ip6_route_net_ops);
2871 out_dst_entries:
2872         dst_entries_destroy(&ip6_dst_blackhole_ops);
2873 out_kmem_cache:
2874         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2875         goto out;
2876 }
2877
2878 void ip6_route_cleanup(void)
2879 {
2880         unregister_netdevice_notifier(&ip6_route_dev_notifier);
2881         fib6_rules_cleanup();
2882         xfrm6_fini();
2883         fib6_gc_cleanup();
2884         unregister_pernet_subsys(&ip6_route_net_ops);
2885         dst_entries_destroy(&ip6_dst_blackhole_ops);
2886         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2887 }